*Article* **Highly Multifunctional GNP**/**Epoxy Nanocomposites: From Strain-Sensing to Joule Heating Applications**

## **Xoan F. Sánchez-Romate \*, Alejandro Sans, Alberto Jiménez-Suárez \*, Mónica Campo, Alejandro Ureña and Silvia G. Prolongo**

Materials Science and Engineering Area, Escuela Superior de Ciencias Experimentales y Tecnología, Universidad Rey Juan Carlos, Calle Tulipán s/n, 28933 Móstoles, Madrid, Spain; a.sans@alumnos.urjc.es (A.S.); monica.campo@urjc.es (M.C.); alejandro.urena@urjc.es (A.U.); silvia.gonzalez@urjc.es (S.G.P.)

**\*** Correspondence: xoan.fernandez.sanchezromate@urjc.es (X.F.S.-R.); alberto.jimenez.suarez@urjc.es (A.J.-S.); Tel.: +34-914-884-771 (X.F.S.-R.); +34-914-887-141 (A.J.-S.)

Received: 16 November 2020; Accepted: 3 December 2020; Published: 5 December 2020

**Abstract:** A performance mapping of GNP/epoxy composites was developed according to their electromechanical and electrothermal properties for applications as strain sensors and Joule heaters. To achieve this purpose, a deep theoretical and experimental study of the thermal and electrical conductivity of nanocomposites has been carried out, determining the influence of both nanofiller content and sonication time. Concerning dispersion procedure, at lower contents, higher sonication times induce a decrease of thermal and electrical conductivity due to a more prevalent GNP breakage effect. However, at higher GNP contents, sonication time implies an enhancement of both electrical and thermal properties due to a prevalence of exfoliating mechanisms. Strain monitoring tests indicate that electrical sensitivity increases in an opposite way than electrical conductivity, due to a higher prevalence of tunneling mechanisms, with the 5 wt.% specimens being those with the best results. Moreover, Joule heating tests showed the dominant role of electrical mechanisms on the effectiveness of resistive heating, with the 8 wt.% GNP samples being those with the best capabilities. By taking the different functionalities into account, it can be concluded that 5 wt.% samples with 1 h sonication time are the most balanced for electrothermal applications, as shown in a radar chart.

**Keywords:** carbon nanotubes; thermal properties; electrical properties; strain sensing; joule heating

## **1. Introduction**

Nowadays, polymeric materials are gaining much attention. They present some interesting properties that make them suitable for use in a wide variety of applications, including as coatings for environmental and corrosion protection [1–3] or as a matrix in composite materials due to their high compatibility with most widely used reinforcements [4,5].

In this regard, the use of carbon nanoparticles such as graphene nanoplatelets (GNPs) or carbon nanotubes (CNTs) is now of interest. This can be explained by their excellent mechanical, thermal and electrical properties. In fact, they can reach values of Young's Modulus over 1 TPa, thermal conductivity around 5000 W/mK and electrical conductivities of 107 S/m [6–10]. These superior properties make them highly suitable for multiple applications. More specifically, they are commonly used as reinforcement for polymeric materials. Furthermore, when added to an insulator matrix, both electrical and thermal conductivity grow several orders of magnitude, becoming conductive materials, because of the creation of percolating networks inside the material [11–13]. These facts promote their use, for example, in Structural Health Monitoring (SHM), electromagnetic interference shield and as thermal interface materials (TIMs) [14–19].

This work is focused on the effect of GNP content and sonication time on several properties such as electrical and thermal conductivity, strain-sensing and Joule heating capabilities. Here, GNP nanocomposites have demonstrated good capabilities as strain sensors with gauge factors, that is, the correlation between the variation of the normalized electrical resistance divided by the applied strain is much superior to that of CNT-based ones, especially at higher strain levels [20–25]. This is explained by the 2D disposition of the GNPs within the material allowing a higher interparticle distance between adjacent nanoparticles, thus leading to a more prominent tunneling effect [26–28]. In addition, GNPs can be added into the resin in contents superior to those possible with CNTs without inducing a drastic degradation of mechanical properties. Therefore, these materials present much higher values of thermal conductivity, which mainly depends on nanofiller content [14,29,30].

However, the correlation among the different properties is sometimes not well understood, as there are multiple factors affecting the final properties of the nanocomposite, including content, dispersion and geometry of the nanofiller [11,24,31,32]. In this regard, several dispersion techniques are commonly used to ensure a proper homogenization of the nanofillers inside the polymer. Among others, three roll milling and sonication have proved to be the most effective techniques due to the higher shear or cavitation forces induced during the process, which lead to an adequate breakage of larger agglomerates along with some exfoliating mechanisms [33,34]. Moreover, the enhancement of one property can lead to the degradation of another, as observed in highly conductive nanocomposites that present low gauge factors, as the interparticle distance between adjacent nanoparticles is much lower [20]. For these reasons, this study aims to better understand the role of nanoparticle content and dispersion state in the final properties of the nanocomposites.

First, the electrical and thermal conductivity of GNP nanocomposites are determined for different combinations of GNP content and sonication time. Then, two examples of specific applications are also measured and deeply explored: their use as strain sensors by means of electrical measurements with applied strain and their capacity for Joule's effect resistive heating. Finally, a summary of the obtained results is shown by balancing the final properties of each material in order to select the optimum one as a function of the desired application.

## **2. Experimental Procedure**

## *2.1. Materials*

The nanocomposite is based on a GNP reinforced epoxy matrix. The resin is an *Araldite LY 556* from *Hunstman* supplied by *Antala* (Barcelona, Spain) with an amino hardener *XB 3473* in a stoichiometry proportion of 100:23 monomer to hardener from the same supplier.

GNPs are *M25* supplied by *XG Sciences* (Lansing, MI, USA) with a lateral size of 25 μm and a thickness of 6 to 10 nm.

## *2.2. Nanocomposite Manufacturing*

First, GNPs were manually dispersed into the epoxy resin. The mixture was then subjected to an ultrasonication process by using a horn sonicator *UP400S* supplied by *Hielscher* (Teltow, Germany) at an amplitude of 80% and a pulse period of 0.5 s. Sonication time and GNP content were varied in order to analyze their effects in the final properties of the nanocomposites. Nanofiller content and sonication time were varied accordingly to that shown in Table 1.

Once the dispersion was made, the mixture was degassed at 80 ◦C for 15 min in order to remove the possible entrapped air. Then, it was subjected to a curing cycle at 140 ◦C for 8 h.

Finally, the plates obtained were demolded and machined to the dimensions required by the different tests which are explained below.


**Table 1.** Nomenclature used for materials manufactured and tested.

## *2.3. Electrical, Thermal and Microstructural Characterization*

Four-probe DC volume conductivity tests were carried out for 4 different samples of 10 × 10 × 1 mm<sup>3</sup> dimensions for each condition (Figure 1a). Electrical resistance was determined as the slope of I-V curve, and electrical resistivity was determined accordingly to the geometry of the samples. The voltage range was set at 0–200 V for low conductive samples and 0–25 V for high conductive samples. The tests were performed in a *SMU, Keithley Instrument Inc. mod. 2410* (Cleveland, OH, USA).

**Figure 1.** Schematics of (**a**) electrode disposition on the electrical conductivity tests, (**b**) set-up of thermal conductivity measurements (where the red arrows indicate the sense of the heat flow), (**c**) Joule heating tests indicating the electrode's disposition in the sample and (**d**) an example of thermal image of a 5GNP-1 h sample (the dark shapes around the central region correspond to the silver paint coating of the electrodes).

Thermal conductivity was measured by estimating the heat flow through 50 mm diameter round samples by using a Heat Flow Meter (*FOX 50 Heat Flow Meter 190\_C VHS 220VAC*) from *TA Instruments* (New Castle, DE, USA) as shown in the schematic of Figure 1b. The thickness of the samples varied in the range of 4–5 mm. Two samples were tested for each condition and thermal conductivity was determined at 30, 90 and 180 ◦C.

The GNP distribution was determined by means of Scanning Electron Microscopy analysis. For this purpose, GNPs were filtrated after sonication process in an acetone bath using a 0.22 μm porous paper. The obtained powder was then analyzed by using a *Hitachi S 3400N* apparatus from *Hitachi Global* (Tokyo, Japan).

## *2.4. Strain Monitoring Tests*

Tensile tests were performed according to standard ASTM D638 at a test rate of 1 mm/min. Simultaneously, the electrical resistance was measured in order to characterize the strain monitoring capabilities of the manufactured materials by using an *Agilent* hardware *34410 A* (Agilent Technologies, Santa Clara, CA, USA).

To achieve this purpose, the electrical resistance was recorded between two electrodes made of copper wire and silver ink to ensure a good electrical contact with the substrate. Here, the sensitivity, also called, gauge factor (*GF*) of the materials has been determined.

*GF* is given by the ratio between the change of the normalized resistance divided by the applied strain:

$$GF = \frac{\frac{\Lambda R}{K\_0}}{\varepsilon} \tag{1}$$

where Δ*R*/*R*<sup>0</sup> denotes the change of the electrical resistance divided by the initial resistance of the specimen.

In the tests conducted, *GF* was determined at low strain levels where crack mechanisms are not supposed to taking place.

## *2.5. Joule E*ff*ect Heating Tests*

Electrothermal properties were determined by resistive heating. In this experiment, thermal conductivity samples were subjected to a varying applied voltage. The temperature of the samples was measured by using a FTIR thermal camera (FLIR E50) (FLIR Systems, Wilsonville, OR, USA) as shown in the schematics of Figure 1c,d. The electrodes were also made with copper wire and silver ink with a distance of 30 mm between them. The voltage was applied by steps of 50 V until a temperature of around 180 ◦C was reached in the sample as it is near the degradation of the epoxy matrix.

#### **3. Results and Discussion**

The electromechanical and thermal properties of GNP nanocomposites are discussed in this section. First, electrical conductivity measurements are shown, while thermal properties are also explored. Finally, the electromechanical characteristics of the proposed materials are given on the basis of strain monitoring tests.

#### *3.1. Electrical Properties of GNP*/*Epoxy Nanocomposites*

Figure 2 shows the values of the electrical conductivity for GNP nanocomposites. It can be observed that an increase of GNP content from 5 to 8 wt.% leads to a significant increase in the electrical conductivity, from values of 10−4–10−<sup>3</sup> S/m to values of 0.1–1 S/m. This is easily explained by the effect of the higher volume fraction of the nanofillers that induces the creation of a higher number of percolating networks inside the material as has been widely explored in other studies [35,36].

**Figure 2.** Electrical conductivity measurements for the different samples.

However, the effect of sonication time is quite more complex. Here, at lower GNP contents, it is observed that the increasing sonication time leads to a reduction of the electrical conductivity. This is explained by the effect that sonication has on the GNP mixture. On one hand, there is the prevalence of exfoliating mechanisms of graphene layers during the sonication process [37], leading to a reduction of GNP thickness and, thus, to an increase of the aspect ratio of the nanoparticles as well as to an enhancement of GNP dispersion [38]. However, it has been also widely investigated that very large sonication times lead to a significant breakage of the nanofillers due to the higher cavitation forces induced during the sonication process [35]. This breakage of GNPs leads to a reduction in the lateral size. At lower contents, the viscosity of the media is low, so the cavitation forces are more effective [39]. This means that the optimum sonication time to achieve the best electrical performance is lower and this fact explains that increasing this sonication time too much could result in a detriment of the electrical properties, because of a very aggressive rupture of GNPs that leads to a reduction of the effective aspect ratio.

On the other hand, when increasing the GNP content, the viscosity of the mixture is much higher, so that cavitation process is not so efficient and the optimum sonication time to achieve the desired properties is increased. For this reason, the highest electrical conductivity for 8 wt.% GNP nanocomposites is achieved at 3 h of sonication time.

In this regard, the SEM analysis of GNP powder after sonication process can confirm the previous statements. On one hand, when comparing 5GNP-1 h to 5GNP-3 h samples, an evident reduction of the lateral size can be pointed out (Figure 3a,b), which is more prevalent than the reduction of GNP stacking, which is qualitatively similar in both cases (Figure 3c,d) as, at low times, the sonication is effective at this GNP content. However, in the case of 8GNP samples, the increase of the sonication time from 1 h to 3 h promotes a very efficient reduction of the GNP stacking (Figure 3e,f) due to the previously commented higher efficiency of the sonication process at higher times explained by the higher viscosity of the mixture.

These statements are of high novelty, as sonication time can have a positive effect on GNP properties depending on the viscosity of the mixture, as observed for 8 wt.% samples whereas longer sonication times will have a negative effect at lower contents due to an initial higher efficiency of the process that trends to rapidly break the nanoplatelets.

**Figure 3.** SEM images of GNP powder after the sonication process for (**a**,**c**) 5GNP-1 h, (**b**,**d**) 5GNP-3 h, (**e**) 8GNP-1 h, and (**f**) 8GNP-3 h samples.

## *3.2. Thermal Properties of GNP*/*Epoxy Nanocomposites*

Figure 4 summarizes the values of the thermal conductivity for GNP nanocomposites. It is observed that an increase of GNP content induces an enhancement of the thermal conductivity, as expected, due to a higher presence of nanofiller. Here, the effect of sonication time is not so prevalent, and only induces slight differences in the thermal properties of the nanocomposites. This can be explained on the basis of the role of GNP geometry and distribution inside the material.

(**c**)

**Figure 4.** (**a**) Thermal conductivity measurements and theoretical estimations (dashed line), (**b**) variation of thermal conductivity for a 5 wt.% GNP sample accordingly to Hatta et al. model [40] as a function of GNP aspect ratio, and (**c**) estimation of aspect ratio accordingly to Hatta model for the different conditions.

Thermal conductivity can be estimated from Hatta et al. model [40] knowing the thermal conductivity of the epoxy and the GNPs:

$$\mathbf{S}\_{11} = \mathbf{S}\_{22} = \frac{\boldsymbol{\beta}}{2 \times \left(\boldsymbol{\beta}^2 - 1\right)^{\frac{3}{2}}} \times \left[\boldsymbol{\beta} \times \left(\boldsymbol{\beta}^2 - 1\right)^{\frac{1}{2}} - \cosh^{-1} \times \boldsymbol{\beta}\right] \mathbf{S}\_{33} = \left[1 - 2 \times \mathbf{S}\_{11}\right] \tag{2}$$

where β is the aspect ratio of GNPs, and *S*11, *S*<sup>22</sup> and *S*<sup>33</sup> are the thermal tensors in the principal axis. In the case of a 3D randomly distribution of nanofillers, the thermal conductivity of nanocomposite *k*<sup>c</sup> can be estimated from the thermal conductivity of matrix, *k*<sup>m</sup> (set as 0.22 W/mK) and from the nanoreinforcement *k*<sup>f</sup> (set as 100 W/mK), as well as from its volume fraction, φ:

$$\frac{k\_c}{k\_m} = 1 + \phi \times \left[ \left( k\_f - k\_m \right) \times \left( 2 \times \mathcal{S}\_{33} + \mathcal{S}\_{11} \right) + 3 \times k\_m \right] / I$$

$$J = 3 \times \left( 1 - \phi \right) \times \left( k\_f - k\_m \right) \times \mathcal{S}\_{11} \times \mathcal{S}\_{33} + k\_m \cdot \left[ 3 \times \left( \mathcal{S}\_{11} + \mathcal{S}\_{33} \right) - \phi \times \left( 2 \times \mathcal{S}\_{11} + \mathcal{S}\_{33} \right) \right] + \frac{3 \times k\_m^2}{\left( k\_f - k\_m \right)} \tag{3}$$

Therefore, the aspect ratio of the nanofillers also plays a significant role. In this context, the dashed green line in Figure 4a indicates the estimation of the thermal conductivity when supposing that the aspect ratio of GNPs is the same for every condition. It can be observed that at lower contents the model generally overestimates the value of the thermal conductivity while at higher contents the estimations are below the measured values. This can be attributed to the differences in the geometry between GNPs for each condition. In this regard, the influence of the aspect ratio on the thermal conductivity is analyzed in the graph of Figure 4b. Here, an increase of the aspect ratio leads to an increase of the thermal conductivity, which is more prevalent in a range of l/d from 10 to 1000.

Therefore, the increasing thermal conductivity with sonication time in the case of 8 wt.% GNP nanocomposites is explained by the increase of the aspect ratio due to a better correlation between the exfoliation induced by cavitation forces and the breakage of GNPs and, thus, a reduction on the lateral size. The opposite effect is observed at lower contents, as explained previously, as the sonication process is much more aggressive and, thus, at higher sonication time there is a more prevalent breakage of GNPs in comparison to exfoliating effect. In this context, the graph of Figure 4c shows the prediction of the aspect ratio of GNPs by adjusting the theoretical model to the experimental measurements. Here, the reduction of the aspect ratio due to a very aggressive breakage of GNPs can be stated when increasing the sonication time at lower contents, whereas the opposite effect is clearly seen at higher contents, validating the previous statements.

## *3.3. Analysis of Strain Monitoring Capabilities*

Figure 5 summarizes the measured gauge factor at a low strain level (ε ∼ 0.0025) for the different GNP nanocomposites. It can be noticed that GF values show the opposite trend when compared to the electrical conductivity measurements in Figure 2. This is explained by understanding the role of tunneling mechanisms inside the material. According to Simmons [41], the electrical resistance associated with tunneling mechanisms, *Rtunnel*, follows an exponential trend with the distance between adjacent nanoparticles, also called tunneling distance, *t*:

$$R\_{tunneling} = \frac{h^2 t}{A\epsilon^2 \sqrt{2m\rho}} \exp\left(\frac{4\pi t}{h} \sqrt{2m\Phi}\right) \tag{4}$$

where *h* is Planck's constant, *m* and *e* are the electron mass and charge, *A* the cross-sectional area of GNPs, and ϕ the height barrier of the matrix.

Therefore, the higher the tunneling distance, the more prevalent the exponential effect of tunneling resistance is. For this reason, lower values of conductivity, which imply higher values of tunneling resistance, usually lead to higher values of sensitivity, as seen in several studies [20,42].

Moreover, the electromechanical response of the samples shows a very prevalent exponential behavior, as can be seen in the graphs of Figure 5b. This is in good agreement with previous studies, where a prevalence of contact mechanisms takes places at a low strain level whereas the breakage of electrical pathways is dominant at higher strain levels, thus leading to a sharper increase of electrical resistance that is reflected in a higher GF at higher strain levels and, thus, to a very prevalent exponential response [24,42].

**Figure 5.** Electromechanic response of GNP nanocomposites showing (**a**) the measured GF and (**b**) an example of a strain-sensing curve.

In this case, the effect of sonication time can significantly affect the sensing properties of these materials. At lower contents and due to a higher efficiency of the dispersion method, there is a reduction of the aspect ratio, as commented, that leads to an increase of percolation threshold [43,44]. This higher percolation threshold implies a higher distance between adjacent nanoparticles and thus, a higher sensitivity when increasing this time. However, the opposite effect can be clearly seen at higher contents, where the highest sensitivities are observed at the lowest sonication time. Therefore, to achieve the best sensing response, the system with lower GNP content and higher sonication time will be selected.

### *3.4. Joule E*ff*ect Heating Analysis*

Figure 6 summarizes the results of the Joule effect resistive heating tests, where the applied voltage and its corresponding average temperature reached in the sample are correlated. Here, it can be observed that both GNP content and sonication time have a significant influence in the resistive heating capacities of the samples.

**Figure 6.** Average temperature reached as a function of the applied voltage for each tested condition (grey-colored area indicates the degradation zone of the epoxy resin).

On one side, by increasing the GNP content, the average temperature reached on the samples increases drastically. More specifically, the maximum allowable temperature, given by the degradation temperature of the epoxy resin (around 180–200 ◦C), is reached at 400–600 V under the 5GNP-1 h and 5GNP-2 h conditions. However, for 8 wt.% GNP samples, the applied voltage needed is around 150–200 V.

On the other hand, the sonication time also affects the Joule heating properties of the nanocomposites. For 5 wt.% GNP samples, an increase in the sonication time implies a drastic decrease of resistive heating capabilities. In fact, samples with a sonication time of 1 h present a limit voltage of 400 V while samples with a sonication time of 3 h do not reached the maximum allowable temperature at the range of the voltage tested. Nevertheless, sonication time does not have a prevalent effect for the samples with an 8 wt.% GNP content, where the maximum allowable temperature is reached at a similar applied voltage.

These results can be explained accordingly to Joule's Law:

$$Q = \dot{r}^2 \times \mathbb{R} \times t \tag{5}$$

where *Q* is the generated heat during the test, *i*, the current flow, *R*, the electrical resistance of the specimen and *t* the time that the specimen is subjected to resistive heating.

Therefore, the electrical properties of these materials play a crucial role in their resistive heating capabilities. Here, it can be concluded that the higher the electrical conductivity of the samples, the higher the current flow, *i*, and thus the heat generated during the Joule's effect tests. This is in good agreement with the previously mentioned electrical conductivity results shown in Figure 2. For 5 wt.% GNP samples, there is a significant variation of electrical conductivity with sonication time which is reflected in a poor heating capability for the samples at 3 h of sonication. Moreover, in the case of 8 wt.% GNP samples, their higher electrical conductivity thus leads to higher heating properties. Here, the differences observed among the different sonication times are less prevalent as the electrical

network formed inside the material is good enough to ensure proper electrical connections between adjacent nanoparticles.

Furthermore, the Joule heating tests show very good heating capabilities in comparison to other studies with similar reinforcements and equivalent geometries [45]. Here, the main difference is correlated with the dispersion technique which, in the case of sonication, tends to form a more homogeneous dispersion inside the material without seriously affect the electrical and thermal properties of the GNPs themselves than in three roll milling process, where there is a prevalent breakage of GNPs due to the high shear forces involved in the dispersion process, leading to lower values of electrical and thermal conductivity and, thus, lower resistive heating capabilities.

#### *3.5. Analysis of Optimum Conditions for Application*

In this section, the behavior of the different manufactured samples is analyzed depending on the property tested. The aim is to select the optimum conditions depending on the desired application. In this context, Figure 7 shows a head-to-head comparison between the thermal and electrical conductivity (Figure 7a) and between SHM and Joule heating capabilities of the different samples (Figure 7b).

**Figure 7.** Graphs representing a comparison (**a**) between electrical and thermal conductivity and (**b**) between SHM capabilities, given by the Gauge Factor and Joule heating properties, given by the applied voltage to reach the maximum allowable temperature.

In the first case, specimens with higher GNP contents and sonication times show the optimum combination of properties (hollow symbols of Figure 7a). This is explained by the prevalent role of the nanofiller content along with the selection of a higher sonication time that allows a more significant exfoliating effect without any substantial detriment on electrical and thermal properties of the GNPs themselves. However, when comparing the capability for SHM applications and Joule heating ones, the selection of an optimum condition is quite a bit more complex. This is explained by the opposite effect of Joule heating capabilities, which are mainly governed by the creation of a highly conductive electrical network inside the material and SHM ones, which are dominated by tunneling mechanisms that are more prevalent in less conductive networks. Here, 5GNP-1 h samples are very competitive (black solid symbol of Figure 7b), as they have a very high electrical sensitivity to strain due to a higher prevalence of tunneling mechanisms. In addition, their electrical conductivity is high enough to allow a relatively good Joule heating effect in comparison to 2 and 3 h samples because of a better GNP dispersion inside the material without affecting the intrinsic thermal and electrical properties, as previously explained.

The selection of an optimum condition will therefore depend on the desired functionality. For these reasons, a radar chart was constructed to obtain a complete overview.

In this chart, each measured property or functionality has been rescaled from 0 to 1, where 1 denotes the highest performance for this property. Therefore, the "best" material will have a factor of 1, whereas the rest of conditions were rescaled accordingly to their value of this property.

This re-scalation follows a linear trend for Joule Effect, Gauge Factor and thermal conductivity. However, due to the highest sensitivity to small variations of electrical conductivity, it has been rescaled following a logarithmic trend, where 1 denotes again the highest measured conductivity and 0, the value of conductivity at percolation threshold, fixed at 10−<sup>6</sup> S/m as observed in other studies [46,47].

Figure 8 shows the calculated values of the factors for each property and condition tested. Here, it can be observed that there is a high correspondence among electrical, thermal and Joule heating properties, whereas electrical sensitivity follows an opposite trend due to the previously commented factors. Accordingly, 5GNP-1 h seems to be a very promising solution for accomplishing all the analyzed functionalities, due to the good balance conferred by a good GNP dispersion without any detriment on nanoparticle intrinsic properties. More specifically, when compared to other works with similar nanoreinforcements, they show much higher Joule heating capabilities [45,48] and similar gauge factors at low strain levels [20,42], showing a high potential for diverse applications.

**Figure 8.** Radar chart of the different tested properties for (**a**) 5GNP and (**b**) 8GNP samples, scaled from 0 to 1.

## **4. Conclusions**

Thermo-electrical and strain-sensing capabilities of GNP nanocomposites were deeply studied by varying GNP content and sonication time.

It was observed that the strain-sensing capabilities and the electrical conductivity follow an opposite trend. Here, the highest strain-sensing gauge factors have been achieved for the samples with the lower GNP content and higher sonication time, which show the lowest electrical conductivity. Furthermore, the effect of the dispersion procedure by means of sonication time on electrical and electromechanical properties was also explored. At lower GNP contents, higher sonication times induced a higher breakage of GNPs, with this effect being more prevalent than the exfoliating effect. However, at higher GNP contents and due to the higher viscosity of the mixture, the exfoliating effect is more prevalent at higher sonication times, explaining the higher values of electrical conductivity reached for these samples.

Concerning the thermal and electrothermal properties, a similar trend to that of electrical conductivity is noticed. Here, the samples with higher GNP content show the highest thermal conductivities and Joule heating capabilities. However, dispersion procedure at higher contents does not play a crucial role, as there is a high enough percolating network to ensure good resistive heating responses.

Therefore, by comparing the measured properties, it is possible to select the optimum manufacturing conditions as a function of the desired application. In this regard, 5GNP-1 h samples show a good balance among properties, as their Joule heating capabilities are much higher than 2 and 3 h sonication samples and their sensitivity is also much higher than 8 wt.% GNP specimens. Furthermore, they are very competitive when compared to similar nanocomposites of the literature.

**Author Contributions:** X.F.S.-R. conceptualization, formal anaylisis, writing—original draft preparation, writing-review; A.S. methodology, formal analysis, A.J.-S. conceptualization, supervision, writing—review, funding acquisition; M.C. supervision, writing—review; A.U., funding acquisition, S.G.P. writing—review, supervision, funding acquisition; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministerio de Economía y Competitividad of Spanish Government [PROJECT PID2019-106703RB-I00], Comunidad de Madrid Regional Government [PROJECT ADITIMAT-CM (S2018/NMT-4411)] and Young Researchers R&D Project (Ref. M2183, SMART-MULTICOAT) financed by Universidad Rey Juan Carlos and Comunidad de Madrid.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Graphene-Based Scaffolds for Regenerative Medicine**

**Pietro Bellet 1,†, Matteo Gasparotto 1,†, Samuel Pressi 2,†, Anna Fortunato 2,†, Giorgia Scapin 3,\*, Miriam Mba 2,\*, Enzo Menna 2,\* and Francesco Filippini 1,\***

	- enzo.menna@unipd.it (E.M.); francesco.filippini@unipd.it (F.F.)

**Abstract:** Leading-edge regenerative medicine can take advantage of improved knowledge of key roles played, both in stem cell fate determination and in cell growth/differentiation, by mechanotransduction and other physicochemical stimuli from the tissue environment. This prompted advanced nanomaterials research to provide tissue engineers with next-generation scaffolds consisting of smart nanocomposites and/or hydrogels with nanofillers, where balanced combinations of specific matrices and nanomaterials can mediate and finely tune such stimuli and cues. In this review, we focus on graphene-based nanomaterials as, in addition to modulating nanotopography, elastic modulus and viscoelastic features of the scaffold, they can also regulate its conductivity. This feature is crucial to the determination and differentiation of some cell lineages and is of special interest to neural regenerative medicine. Hereafter we depict relevant properties of such nanofillers, illustrate how problems related to their eventual cytotoxicity are solved via enhanced synthesis, purification and derivatization protocols, and finally provide examples of successful applications in regenerative medicine on a number of tissues.

**Keywords:** graphene; graphene oxide; reduced graphene oxide; tissue regeneration; 2D-scaffolds; hydrogels; fibers; stem cell differentiation

## **1. Introduction**

Graphene consists of an atomic honeycomb lattice composed of carbon atoms that can be considered as an indefinite large polycyclic aromatic hydrocarbon with an infinite number of condensed benzene rings. Graphene family is constituted by several derivatives such as graphene oxide (GO), reduced graphene oxide (RGO), graphene quantum dots (GQDs), graphene nanosheets, monolayer graphene, and few-layer graphene [1]. A schematic representation of graphene-based materials (GBMs) taken into account in this review is shown in Figure 1. Although an accurate description of the state of the art in GBM synthesis is out of the scope of this review, a brief outline is provided in Section 2.1. It is vital to stress out that GBMs are highly heterogenous, especially when considering biological properties and applications. Therefore, careful choice of the synthetic method is required to obtain a material with the desired properties (i.e., dimensions, conductivity and eventual functional groups).

Due its high electrical conductivity, mechanical properties and aspect ratio, graphene has become attractive in many fields. In addition to being a rising star in scientific fields other than biology and medicine, graphene, GBMs and composites are widely used for important biotechnological and biomedical applications. Almost all graphene derivatives and composites are being used and tuned to develop special delivery carriers for theranostics [2], gene therapy and drug delivery, and a huge number of examples have been

**Citation:** Bellet, P.; Gasparotto, M.; Pressi, S.; Fortunato, A.; Scapin, G.; Mba, M.; Menna, E.; Filippini, F. Graphene-Based Scaffolds for Regenerative Medicine. *Nanomaterials* **2021**, *11*, 404. https://doi.org/ 10.3390/nano11020404

Academic Editor: José Miguel González-Domínguez Received: 16 December 2020 Accepted: 1 February 2021 Published: 5 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

reviewed in recent years [3–6]. Therefore, we can just list here a few examples of applications in biosensing and bioimaging, before moving to the focus of this review, which is regenerative medicine.

**Figure 1.** Structural overview of graphene-based materials.

Conductivity and high transporter capability of graphene allow for tuning biosensor surface features and outperforming many other biosensor types in terms of speed, accuracy, specificity, selectivity and sensitivity. In general, proteins (either catalysts or receptors/ligands) are associated to the graphene-based biosensor surface via electrostatic interaction, covalent bond or by polymer mediated capture. To avoid electrostatic interaction may alter the protein conformation, GBMs are combined with polymers in nanocomposites where mild electrostatics combines efficient binding and maintenance of the original conformation [7]. When instead the conformational dynamics of proteins has to be studied, multilayer graphene nanopore sensors can be used [8]. Several compounds can be detected electrochemically using graphene and GBMs as electrochemical sensor, e.g., cancer markers and cells, ATP, DNA, glucose, toxins, or even proteins [6]. Graphene-based field-effect transistor (FET) biosensors, which can be integrated with electronic chips, easing compatibility with industry standards, are especially applicable in detection of charged molecules such as DNA. Graphene-based fluorescence resonance energy transfer (FRET) biosensors are also widely used with small molecules, nucleic acids and proteins, as reviewed by Zhao et al. [9]. Some biosensors integrate graphene in the surface plasmon resonance (SPR) technology, showing improved sensitivity and detection range [10]. Graphene quantum dots (GQDs) are of special interest to bioimaging in vitro and in vivo because of their biocompatibility, tunable fluorescence with excellent photostability, ultra-small size and hydrophilicity [11]. Stable photoluminescence makes GQDs suitable for cancer bioimaging and has led to biofunctionalization for specific cancer cell imaging and real-time imaging in living cells [12]. GO and RGO are used in bioimaging as well, as their combination with different polymers (e.g. PGA), metal ions or bioimolecules can modulate emissions in three main fluorescence regions (blue, green and red), making (R)GO-derived platforms suitable for multiple tracing and bio-imaging purposes [13,14].

GBMs, and especially GO and RGO find plenty of applications in tissue engineering, where they are employed as scaffolds for tissue regeneration. Tissue engineering is an interdisciplinary technology that gains insights from material chemistry, engineering, cell biology, and immunology to develop biomaterials capable of restoring, maintaining, or

improving tissue function or a whole organ [15]. Scaffolds act as biological substitutes that enhance cellular interactions and are able to stimulate the differentiation of stem cells or precursor cells into the desired lineage. The extracellular environment provides biochemical, biophysical, and electrical signals, which all together define tissue-specific niches for proper tissue function and homeostasis. By recapitulating such features in biomimetic scaffolds, the goal of tissue engineering is to guide stem cell development and differentiation to resemble cell organization and behavior in the natural, tissue-specific environment. Such approach offers an interesting translational perspective for tissue repair and regeneration [16,17]. However, successfully reproducing a tissue is extremely challenging since a number of different aspects must be taken into account. In this scenario, nanocomposite materials have proven to be effective in mimicking the required characteristics.

Graphene-based scaffolds (GBSs) are a particular class of scaffold made from graphene, GO and/or RGO nanocomposites. Among the plethora of nanomaterials available, graphene and its derivatives are attractive candidates for developing tissue engineering scaffolds thanks to their tuneable electrical conductivity, excellent mechanical properties, biocompatibility, chemically modifiable surface, and nanoscale dimension matching cell surface receptors and extracellular matrix (ECM) nanoroughness/nanotopography. Morover, they display good capacity to adsorb proteins from the serum (e.g., fibronectin, laminin and albumin), favoring cell adhesion, proliferation and differentiation.

Graphene structural features and dimensions resemble many components of the extracellular environment such as proteins of the ECM (e.g. collagen), ion channels, signalling proteins and cytoskeletal elements [18]. Therefore, the introduction of graphene or its derivatives into polymeric scaffolds endows them with features that can be tailored to match the ones of the natural tissue of interest. For instance, each tissue has specific mechanical and electrical properties that should be matched by artificial scaffolds. Intuitively, scaffolds for bone regenerative medicine should be stiffer (E > 109 Pa), whereas nervous tissue requires much softer supports (E < 4·102 Pa) and muscles need substrate with intermediate stiffness (E > 10<sup>4</sup> Pa) [19].

Being one of the toughest and strongest nanomaterials discovered so far, graphene incorporation into polymeric scaffolds enhances their mechanical properties, toughness and tensile strength [20]. Therefore, graphene percentage within the scaffold can be modulated in order to better mimic the ECM mechanical properties of the tissue of interest. Moreover, graphene nanocomposite scaffolds are endowed with nanoroughness, which contributes to cell anchoring while modulating cell morphology [18]. This property is particularly important for the differentiation of neuronal cells as graphene establishes tight contact with the growth cone and guides the spreading of developing neurites [21]. Lastly, empirical evidence suggests that engineering the electrical conductivity of the scaffold plays a crucial role in producing a functional electroactive tissue. Since graphene is electrically conductive and its conductivity is stable in biological environments, its incorporation in polymeric scaffolds can reduce the polymer electrical resistance. As a result, graphene-based scaffolds can be used to mimic and regenerate the electroactive tissues like the cardiac and neural ones, but also to boost the repair of non-excitable cells that are subjected to electrical field after an injury, like during bone repair and wound healing [22]. However, Burnstine-Townley and co-workers pointed out that the actual role of scaffold conductivity in cell differentiation is not completely clear. Specifically, disentangling the effect of a single scaffold feature on cell fate can be challenging, as varying graphene content has effect on several properties, such as surface roughness, cellular adhesion and interaction with nutrients, growth factors and wastes [23].

At a glance, graphene ability to mimic the natural extracellular environment nanotopography, to retain signalling molecules, to be easily incorporated in both natural and synthetic polymers, and to modulate stiffness and conductivity of the scaffold make it the ideal nanomaterial to provide cues needed to guide cell behaviour and hence an invaluable tool for regenerative medicine applications. Nevertheless, the toxicology profile of graphene and its derivative has not been completely elucidated yet.

Several drawbacks of GBMs employment for regenerative medicine approaches have been reported which might include membrane damage, hydrophobic interaction, oxidative stress, genotoxicity, mitochondrial disorders and autophagy. However, safety risks should be evaluated case by case based on the intrinsic properties of GBMs, such as purity, surface functional groups, lateral size, stiffness, hydrophobicity and structural defects. Moreover, several reports showed that graphene cytotoxicity is influenced by multiple parameters such as cell population tested as well as graphene dispersibility and functionalization [24,25].

#### **2. Graphene-Based Scaffolds**

## *2.1. Methods for GBM Synthesis*

As thoroughly reviewed by Wu and co-workers graphene synthesis can be performed through a plethora of bottom-up or top-down approaches [26]. Among the most common, Chemical Vapor Deposition (CVD), Physical Vapor Deposition (PVD), spin coating, laser ablation and arch discharge needs to be mentioned. Although a systematic review of graphene synthetic methods is out of the scope of this review, it needs to be stressed that the final properties, complexity and cost of a nanomaterial are strictly related to its procedure of synthesis. Each protocol has its advantages and drawbacks, thus the choice should be done taking into consideration the final application of the product. For the sake of clarity, a brief overview of standard approaches is provided in this section.

CVD is often exploited to produce graphene for 2D composites and graphenic foams, described in the next sections. In CVD, gaseous precursors (typically hydrocarbons) are flowed at high temperatures over a metal surface, which acts as a catalyst for their decomposition and leads to the condensation of carbon atoms, forming a graphene sheet. In a typical process, graphene is grown onto a metal surface, supported with a polymer (e.g., poly(methyl methacrylate)—PMMA), and the catalyst is etched by acidic treatment. Subsequently, the graphene foil is transferred on a substrate and the supporting polymer is appropriately dissolved. The choice of metal or alloy for deposition changes process thermodynamics and kinetics, and allows to finely tune the number of graphene layers of the resulting material. The most common metal catalysts for CVD are nickel [27,28] and copper [29], with a preference for the latter due to its capability to produce single- and bi-layered graphene.

However, bulk production of graphene is more conveniently achieved starting from graphite and weakening the van der Waals forces between its stacked monoatomic carbon layers. Examples for such top-down approaches are liquid-phase exfoliation, surfactantassisted liquid-phase exfoliation and chemical functionalization. In the first two methods [30–34], exfoliation is achieved through different combinations of factors such as (i) the choice of a solvent with proper surface tension (e.g, γ = 40 mJ m-2) [35]; (ii) the use of surfactants, to minimize the interfacial tension between solvent and graphene; (iii) sonication or other external mechanical driving forces; (iv) centrifugation stages to remove thicker graphitic flakes. The principal shortcomings of these methods are the generation of defects and the reduced size attributed to sonication-induced cavitation [36,37].

Among the most common chemical top-down methods, there is the oxidation and subsequent exfoliation of graphite to GO, followed by either chemical reduction or thermal cleavage of oxidized groups to obtain RGO. In a typical procedure, graphite is mixed with sulfuric acid and oxidizing agents in an iterative and synergic action of intercalation and oxidation [38]. Subsequent exfoliation in water then easily yield GO, which can be further modified due its large amount of different oxygen functional groups (such as epoxy, hydroxyl, carbonyl and carboxyl groups). Even if the production of GO induces a large number of defects in the graphenic sp2 network, the enhanced hydrophilicity that results from the oxidation can be beneficial for its compatibility in different types of matrixes.

Oxidation is usually carried out at 40–50 ◦C. However, as demonstrated by Eigler and co-workers, working at lower temperatures could reduce damages to the basal plane. They demonstrated the possibility to synthesize a minimally damaged GO with an almost intact σ framework of C atoms [39] and superior thermal properties [40] while maintaining the oxidation temperature below 10 ◦C, and effectively controlling kinetics of process.

RGO is obtained from GO with different synthetic methods, yielding materials with different properties. Indeed, thermal treatment (often improperly called "thermal reduction") and chemical reduction of GO to RGO do not have the same effect on graphene structure, hence on the properties of the resulting materials. The disproportionation induced by thermal treatment of highly oxidized GO brings defective holes in the plane.

Chemical reduction, on the other hand, can be achieved with different reactants [41] leading to different results: as an example, hydrazine leads to *N*-doping in plane, while reduction with L-ascorbic acid leaves adsorbates on RGO that are not easily removed by washing procedures [42]. Since complete reduction of GO is not achieved, RGO differs from graphene due to the presence of residual functional groups; however, O/C ratio of RGO is much lower than that of GO. Even if the sp<sup>2</sup> network is partially restored, the performances are still lower than those of CVD graphene.

Stability and reactivity of GO are also affected by other parameters such as pH of the dispersion, which is often neglected or underestimated. Indeed, Hirsch and coworkers [43] found evidence that the carbon lattice is damaged by treatments with a base at 40 ◦C while at 10 ◦C the partial cleavage of epoxy groups is observed. According to the above-mentioned observations, assessing the O/C ratio, which is often the only parameter considered to assess the successful synthesis of RGO, is clearly not enough to describe the obtained material. It must also be emphasized that the choice of starting materials, different methods of synthesis, and purification procedures have a direct impact on the presence of impurities, that can have a biological effect and can lead to controversial results when materials are used for bio-applications.

In the next sections, graphene-based nanocomposites will be considered. In these materials, GBMs act as fillers while the matrix is typically an organic polymer (natural or synthetic FDA-approved polymer), though bioglasses and ceramics are also used. In the first part, two-dimensional (2D) scaffolds are discussed, whereas the second part is devoted to three-dimensional (3D) scaffolds (Figure 2). In particular, three types of 3D scaffolds are considered: porous foams, fibrous scaffolds, and hydrogels.

## *2.2. Two-Dimensional Scaffolds*

Two-dimensional scaffolds are relatively low cost and easy to fabricate, thus they are often used in preliminary studies to investigate the effect of a specific substrate on cell behavior.

The simplest example of a graphenic 2D-scaffold is represented by CVD-grown graphene (one or more layers) on a PMMA-supported metal catalyst, and then transferred onto a substrate after etching of the metal catalyst. Jangho et al. used this technique to transfer the monolayer graphene onto glass to study its effects on the reciprocal interactions between cells and substrate and to test the possible promotion of human mesenchymal stem cell (hMSC) neurogenesis and neurite outgrowth [44]. In a similar way Nayak et al. transferred CVD-grown graphene on different polymeric substrates to verify the effect of nanotopography induced by interactions between graphene and polymers. Differently from the glass control, their 2D scaffold exhibited nanoripples due to a weaker adhesion, and boosted hMSCs differentiation similarly to treatment with bone morphogenic protein BMP2 [45].

**Figure 2.** Examples of two-dimensional and three-dimensional scaffolds. (**a**) CVD graphene on Si/SiO2 chip and AFM images of graphene transferred on different polymeric substrates [45]; (**b**) graphene transferred on nanopatterned substrate and AFM image [46]; (**c**) PLLA-RGO film obtained as reported in [47]; (**d**) graphite oxide paper [48]; (**e**) GO foams and SEM images [49]; (**f**) PLLA-RGO electrospun fibers obtained as reported in [50]; (**g**) peptide–GO hybrid hydrogels and TEM images [51].

Another method to obtain a graphene-coated surface is based on the chemical modification of a substrate to enable specific interactions with graphene-based materials (GBMs). Ryoo et al. used (3-aminopropyl)triethoxylane (APTES) to decorate the surface of glass coverslips with aminic groups. As a result, they obtained a positively charged surface which could effectively interact with negatively charged GO. Similarly, they exploited (3-glycidyloxypropyl)trimethoxylane (GPTMS) to promote glass interaction with aminated carbon nanotubes (CNT). In vitro tests proved carbon nanomaterial-coated glass to be better at promoting the number and dimensionality of focal adhesions, suggesting good biocompatibility [52].

Two-dimensional graphene-based scaffolds can also be obtained by vacuum filtration of material suspensions. For instance, Jasin and co-workers fabricated graphene-based paper as a substrate for cell growth, air drying vacuum filtrated dispersions of three different starting materials: (i) graphite oxide and graphene oxide with (ii) small and (iii) large average lateral dimensions. Although they did not observe any significant difference on cell adhesion, morphology or proliferation, the smaller release of lactate dehydrogenase (LDH) enzyme compared to control samples, suggested that their scaffold can enhance cell viability [48].

A higher degree of versatility is achieved with hybrid or composite scaffolds, where graphene is used as a filler or coating for polymeric matrices. As an example, Pandele et al. prepared chitosan/GO composites by solution blending, obtaining films with a rough surface useful for cell adhesion. The homogeneous dispersion of GO in a polymeric matrix led to an enhancement of the mechanical properties due to the large aspect ratio of the nanomaterial and its interaction with the polymer chains [53]. Furthermore, Jin et al. tested the viability of a free-standing film composed of GO and bacterial cellulose (BC) obtained from *Gluconacetobacter intermedius*. GO was added to the growth media and *G. intermedius* bio-reduction capabilities were exploited to obtain BC-RGO composites. hMSCs seeded onto these materials showed higher proliferation compared to ones seeded onto films of RGO without the fibrous structure of cellulose [54]. Li et al. fabricated RGO-cellulose paper by drop-casting GO dispersions on cellulose paper, subsequently reducing it with L-ascorbic acid (Figure 3).

**Figure 3.** Assembly of RGO-cellulose hybrid paper through deposition of GO followed by in situ reduction [55].

These scaffolds showed low resistivity (∼300 Ω/sq), increased mechanical strength and a specific surface micro-topography induced by RGO, which led to improved stem cell adhesion and osteogenic induction. Furthermore, their 2D-scaffolds could be employed with pseudo-3D stacked multilayered constructs that can be configured by rolling or folding, allowing designing a large number of different setups [55].

To enhance their biological effects, two-dimensional scaffolds can be micro- or nanopatterned with specific topographical cues that can direct cell growth and differentiation. Different methods have been developed to this aim, and a pattern can be drawn with either the help of a positive photoresists spin-coated on graphene oxide surface [56], or by transferring CVD graphene on a polymeric nanopatterned substrate [46,57]. This latter approach was adopted by Jangho and co-workers. They transferred a graphene layer on a poly(urethane acrylate)-patterned surface featuring regular parallel nanogrooves, thus obtaining a chemically homogeneous but mechanically heterogeneous substrate. In fact, graphene has lower mechanical properties in regions where it is suspended between nanoridges. Indeed, alignment of hMSCs along the nanotopographical cues of the substrate was observed [46].

Among the plethora of chemical studies presenting new kinds of scaffolds, there is a modest number of works specifically focused on specific GBM functionalization strategies to improve biocompatibility or differentiation capabilities. As an example, Qi et al. functionalized GO with L-theanine, an amino acid that promotes neuronal differentiation. Its presence in a poly(lactic-co-glycolic acid (PLGA) film increased its hydrophilicity and enhanced neuronal differentiation of neuronal stem cells (NSCs) [58]. In our lab we [47,59] designed composite poly-L-lactic acid (PLLA) scaffolds with different carbon nanostructures (CNS) as filler—namely RGO, carbon nanohorns (CNH) and CNT—covalently functionalized with p-methoxyphenyl (PhOMe) groups in order to improve biocompatibility, and the electrical and mechanical properties of materials. RGO- and CNH-based scaffolds (RGO-PhOMe and CNH-PhOMe respectively) showed promising activity in enhancing the expression of myogenic markers during human circulating multipotent stem cell (hCMCs) differentiation. Moreover, electric percolation was found to take place within the considered range of RGO concentration, tough with lower performances compared to CNT-based samples. This difference is likely due to the influence of aspect ratios on electrical behavior.

Despite the aforementioned potentialities, 2D scaffolds have limitations. First of all, a two-dimensional environment is not suited to reproduce natural ECM. Then, nutrients are directly available to cells and wastes can diffuse to a limited extent. Lastly, altered cell–cell interactions may result in unpredictable cell responses. Therefore, in recent years the focus has shifted towards the study and design of 3D-scaffolds in order to overcome these limitations.

## *2.3. Three-Dimensional Scaffolds*

As already mentioned, 3D scaffolds recapitulate tissue biophysical features thus are better candidates for in vivo applications. Scaffolds with a three-dimensional architecture should be endowed with a highly interconnected porous network. Recently, Lutzweiler and co-workers reviewed the effects of porosity, pore size and shape, interconnectivity and curvature in scaffolds used for tissue regeneration: not only these properties directly influence migration of nutrients and wastes inside the scaffold, but also the permeation and communication between cells [60]. Recent evidence suggests that scaffolds with pore diameters between 100 and 750 μm are generally beneficial while larger pores make cells experience a planar pseudo-2D environment, which differs from their natural environment [61,62].

#### 2.3.1. Foams

The easiest method to fabricate porous scaffolds involve freeze-drying filtrates or suspensions. For example, Domínguez-Bajo et al. produced RGO foams by drying GO slurries, obtaining structures with 43% of porosity and 30 μm of pore size after thermal reduction. In addition, these scaffolds had a relatively low Young's modulus (~1.3 kPa) and made a good candidate for nervous tissue engineering. When their applicability on neural repair after spinal cord injury was tested in vivo, not only scaffolds were populated by nerve cells, but the authors also observed full vascularization [63]. In another instance, the same group exploited ice segregation-induced self-assembly, based on unidirectional freezing of dipped suspensions and lyophilization, to fabricate hierarchically channeled RGO scaffolds with controlled porosity and pore size (80% and 150 μm respectively) [64]. Liao et al. exploited a freeze-drying approach to produce a porous hybrid scaffold based on a copolymer composite of methacrylated chondroitin sulfate (CSMA) and poly(ethylene glycol) methyl ether-ε-caprolactone-acryloyl chloride (PECA) with GO, synthesized by heat initiated free radical polymerization. Not only scaffolds pore size could be tuned by CSMA:PECA ratio, but the compressive strength increased with PECA content, with values consistent with cartilage tissue. The plateau limit of conductivity (1.84 S/m) resulted at 3% GO content [65].

In a similar way, Hermenean et al. fabricated a porous chitosan/GO scaffold with improved mechanical performance—i.e., increased compressive strength and tunable Young's modulus while keeping scaffold flexibility—observing that the incorporation of 3% of GO significantly enhanced bone regeneration in vivo, compared to pure chitosan scaffolds, even in the absence of additional differentiating agents, confirming the active action of GO in facilitating cell infiltration and differentiation [66].

Graphene foams are the first porous structures composed of single layer graphene, applied in tissue engineering. Besides porosity, these scaffolds are endowed with a wrinkled topography induced by the synthetic process, which is beneficial for cell adhesion and proliferation since it better mimics the ECM [67,68]. Li et al. compared NSCs differentiation performance on 2D CVD graphene scaffolds and 3D graphene foam and observed improved proliferation and differentiation towards mature phenotypes on the latter substrate [69].

In the techniques described so far, pore size and interconnectivity depended on the Ni foam features. However, Xiao et al. recently managed to finely tailor these properties, fabricating an ordered architecture of Ni: they used photolithography to define a mask in which Ni was deposited by electroplating and aligned. Graphene was then grown through CVD on the resulting Ni template (Figure 4). Thanks to this procedure, they managed to design a scaffold with defined features by tuning pore and skeleton size (10–50 μm range), orientation angles (45◦ or 90◦), electrical conductivity (60–80 S cm−<sup>1</sup> range) and density (around 3–4 mg cm−3). Such a scaffold was able to direct neuronal growth and align neurons along a defined path to form a network [70].

**Figure 4.** (**I**) (**a**–**d**) Schematic illustration of the procedure used to fabricate a 3D-CG. (**II**) (**a**–**d**) SEM images of four layer freestanding 3D-CGs with different patterns, pores, and skeleton sizes [70].

Another method to fabricate porous structures has been employed by Rasch et al. [49]. Starting from tetrapod-shaped ZnO, pressed and annealed in a mold, they were able to synthesize templates with high porosity (50 to 98%). GBM deposition was obtained by infiltrating a GO suspension in the templates, followed by chemical etching with hydrochloric acid. Their protocol allowed easy, versatile and cost-effective deposition of nanomaterials. Moreover, biological evaluations of these scaffolds by Schmitt et al. showed they could be promising for nervous tissue engineering [71].

An alternative approach to induce porosity in scaffolds is supercritical foaming which allows to control scaffold morphology through a careful choice of experimental parameters, such as chamber pressure, temperature and decompression rate. Evlashin et al. exploited this process to manufacture RGO-reinforced polycaprolactone (PCL/RGO) and PCL/GO foams in a carbon dioxide atmosphere. Although the presence of RGO in the polymer matrix led to an increase of pore size, those foams showed poor cell adhesion properties. Conversely, they found PCL/GO scaffolds to enhance cell adhesion. However, both composites displayed lack of interconnected porosity, resulting in cells attaching only on scaffold surface [72]. Polymer-enriched hybrids can also be obtained starting from CVD graphene foams, by depositing the polymer from a solution through spin or dip coating. Resulting scaffolds show improved mechanical performances and cellular responses. In order to retain porosity, it is crucial to avoid pore saturation through fine optimization of dip coating time and by choosing a polymer with a favorable, near zero contact angle. Nieto et al. exploited this technique with a copolymer of polylactic acid (PLA) and poly-εcaprolactone (PCL) and achieved improved tensile strength due to filling of the pre-existing microcracks in pristine G foams. In vitro tests demonstrated these materials are able to support hMSCs viability and differentiation, making them suitable for musculoskeletal tissue engineering [73].

A layer-by-layer (LBL) technique was followed by Song and co-workers who deposited a positively charged polymer, poly(diallyl dimethylammonium) chloride (PDDA) on a negatively charged Ni template and subsequently placed negatively charged GO onto its surface, which was then thermally converted to RGO. Electrochemical deposition of polypyrrole (PPY) and hydroxyapatite (HA) on top, increased scaffold roughness and surface area, favoring cell adhesion and proliferation as confirmed by *in vitro* tests on the pre-osteoblast cell line MC3T3-E1 [74].

Besides metallic templates, polymeric organic foams are used for polymer replication technique, especially in the inorganic scaffold field. Deliormanlı et al. used polyurethane foam to fabricate HA scaffolds, eliminating the template and sintering HA by heat treatment. PCL/GO was added by dip coating, leading to a scaffold with improved mechanical performance and higher bioactivity [75].

The same procedure can be applied to bioactive glass, another important class of useful scaffolds in tissue engineering. As an example, Turk et al. incorporated 10% graphene directly in the glass matrix before sintering borate-based porous scaffolds, doubling the compressive strength and obtaining an electrical conductivity (0.060 S/cm) which could be exploited to electrically stimulate cell growth [76]. Moreover, Deliormanlı et al. fabricated more chemically stable and biocompatible silicate-based scaffolds coated with PCL/graphene with pore size between 100–500 μm, without detrimental effects of polymer coating on pore structure [77].

An alternative approach to porous structure design is 3D printing, which allows to accurately control scaffold geometrical features without the limitation of using a template. Jakus et al. exploited a PLGA-based ink where they incorporated graphene with the use of surfactants and plasticizers. The mechanical properties of composites are affected by graphenic particles, with an increase of elastic modulus to a value of 16 MPa at 20 vol% loading of graphene, but with detrimental effects at higher loadings (40–60 vol% of graphene). In addition, they observed an anisotropic alignment of graphene flakes, enhancing electrical conductivity due to shear forces produced during the 3D printing extrusion process, which increased with the decrease of the nozzle diameter [78].

Cabral et al. used extrusion 3D printing to produce multicomponent scaffolds, based on tricalcium phosphate chitosan and gelatin, which mimicked the inorganic and organic components of bones, respectively. GO was added to this blend and reduced to RGO in situ by L-ascorbic acid treatment. When comparing mechanical properties of scaffolds incorporating GO or RGO they found the latter to better mimic bone Young modulus, thus their scaffold might be useful as a temporary support for bone regeneration [79].

## 2.3.2. Electrospun Fibers

Fiber-based scaffolds are largely employed in tissue engineering because they intrinsically resemble the microstructure of natural tissues. Fiber diameter, porosity, and orientation are the main features that influence cell growth and tissue regeneration [80]. One of the most common techniques to produce continuous fibers is electrospinning. Electrospinning offers several advantages, including (i) ease of processing, (ii) possibility of large-scale production (iii) availability of advanced modes [81]. Moreover, it is highly versatile and electrospun fibers can be deposited in a random orientation or in an aligned fashion which enhances cell alignment and elongation along the contacted fiber direction [82]. Most thermoplastic materials can be electrospun by fine-tuning the properties of the polymeric solution and the electrospinning parameters such as voltage, electrodes distance and flow rate. The American Food and Drug Administration (FDA) approved several thermoplastic biomaterials for in vivo implantation. Nevertheless, their applications are restricted by the high hydrophobicity, low mechanical properties, lack of specific interactions with cells and sometimes relative slow in vivo degradation rate. Luckily, these limits can be easily overcome by introducing proper nanofillers, and several examples of electrospun thermoplastic materials reinforced with GBMs have been reported [83–85]. As highlighted by Song et al., the solubility of the filler strongly influences the mechanical properties of the final material: a poor dispersibility or a too-high loading leads to aggregation, which results in fractures and disconnections along the nanofibers. It has been observed that electrospinning graphene-based composites yields thinner fibers (Figure 5), but on the other hand, even a small amount of GO or RGO inside the electrospun fibers reinforces their structure and overcomes the detrimental effect of a reduced diameter on mechanical properties. Therefore, it is crucial to finely tune the CNS content in order to find the right balance between a uniform CNS dispersion, nanofiber diameter and reinforcement effect. Besides, it has been widely demonstrated that incorporation of CNS in fibrous scaffolds results in an improved biomimetic microenvironment that enhances cell adhesion and proliferation on different cell types [84].

**Figure 5.** Surface morphological and constructional images of PCL/GO composite nanofibers with different GO concentrations (wt%): (**A**) SEM images and (**B**) TEM images. The red ellipses in SEM images are the fractures [84].

Generally, the smaller thickness induced by graphene-based nanofillers on electrospun fibers allows mimicking the structure of ECM even better. It is believed that the effect of GBMs on fiber diameter is due to the electrical conductivity of the feeder solution, which is a key factor in determining the diameter and size distribution of the electrospun fibers. Moreover, it is reported that fiber diameter is highly correlated to the viscosity of the feeder solution. Scaffaro and colleagues pointed out that the decrease of viscosity of a PCL solution by addition of GO induces electrospinning of thinner fibers. On the other hand, they observed an opposite effect on viscosity (and fiber diameter) with GO-grafted-PEG (GO-g-PEG). Functionalization of the filler not only increased fiber diameter, but also improved dispersion of the filler and maximized the filler/matrix interfaced area, making GO-g-PEG more effective than GO in reinforcing composite fibers, in particular at low concentration [86].

In 2019, Basar and co-workers developed a PCL/GO composite scaffold [87] by functionalizing GO with either an RGD-peptide (GRGDSP), thiophene (Th) or both. Besides having the aforementioned effect on fiber diameter, GO functionalization yielded an enhanced electrical behavior to the scaffold, with conductivities reaching 15.06 μS cm−<sup>1</sup> in PCL/GO-GRGDSP-Th (2% of GO), a 15-fold increase compared to neat PCL (0.95 μS cm<sup>−</sup>1). However, while scaffolds with higher content of GO (2%) showed higher electrical performances, the elastic modulus and tensile strength of 0.5% GO-scaffolds were found to be higher. Once again, this result was associated with the uniform dispersion of GO in the polymer matrix. Interestingly, this modification resulted in an increment of both electric conductivity and mechanical stability due to the ability of sulfur moieties to enable the crosslink between GO and PCL [87].

Scaffold properties can also be altered by combining different organic or inorganic fillers. Lui et al. developed electrospun PLA scaffolds reinforced with GO (1–3 wt %) and/or nano-HA (15 wt %). Interestingly, addition of 15 wt % nano-HA improved both elastic modulus and tensile strength, whereas concentrations of GO above 2% diminished them due to filler aggregation. Nanofiller addition slightly increased scaffold glass transition temperature and modified the hydrophobicity of PLA, enhancing the polymer water uptake, which in turn assisted cell adhesion and proliferation [88].

Different strategies have been developed to obtain polymeric nanofiber scaffolds based on graphene and its derivatives. However, nanocomposites fail to provide a pure graphene interface. An alternative approach aims to immobilize nanostructures on the surface of polymeric nanofibers. The surface of aliphatic polyesters such as PCL and PLA can easily be functionalized with hydroxyl and amino groups by treating the polymeric scaffold with a diamine solution. In tissue engineering, aminolysis of polyesters improves their interactions with cells and allows them to form a stable graphenic coating [89]. Recently, Jalili-Firoozinezhad et al. reported an easy method to generate electrically conductive nanofibers by coating a PCL nanofibrous mat with GO liquid crystals, which were then

reduced to RGO to form PCL-templated graphene nanofibers [90]. Proper electrical conductivity and nanofibrous topography of these constructs make them an ideal platform for cell culture, tissue engineering, drug delivery, and biosensor applications. Preliminary in vitro analyses using hMSCs revealed no induced cytotoxicity and confirmed an enhanced cellular metabolism and proliferation rate compared to standard culture plates and PCL nanofibers.

Indeed, coated fibers can be obtained without any surface treatment. Wang et al. developed a conductive graphene-based fibrous scaffold by coating RGO via an in situ redox reaction of GO on the surface of silk fibroin/poly(L-lactic acid-co-caprolactone) (ApF/PLCL) composite nanofibers [91]. The authors highlighted that the coating did not affect the nanoscale topography of the scaffold and enhanced its mechanical properties, electroactivity and biocompatibility. They then investigated how these conductive scaffolds regulated in vitro and in vivo cell behavior and differentiation under electrical stimulation. RGO-coated ApF/PLCL scaffolds boosted cell migration, proliferation and myelin gene expression of Schwann cells (SCs), whereas pheochromocytoma-derived PC12 cells cultured on these scaffolds exhibited enhanced differentiation. In vivo implantation of the constructs promoted peripheral nerve regeneration in rats.

Polymer core-CNS shell fibers can be obtained by electrospinning the polymer into a solution of graphene or one of its derivatives. Subsequently, it is possible to further functionalize or reduce the shell. Jin et al. exploited this principle to develop an RGO core-shell nanofiber (RGO-CSNFM) [92]. The RGO core-shell structure displayed high mechanical, electrical conductivity (10.0 S cm<sup>−</sup>1) and a charge carrying capacity. This property is likely due to both RGO-CSNFM large surface areas and the extended π–π conjugated bond network generated over the surface of the RGO shell layer. Wu et al. developed an LBL method to coat electrospun nanofibers that mimic vascular ECM and enhance proliferation of endothelial cells. PLLA surface modification was achieved via electrostatic LBL selfassembly by alternately immersing PLLA fibers in a positively charged solution of 0.1 wt% chitosan and a negatively charged solution of 0.1 wt% heparin (PLLA-CS/Hep) or 0.1 wt% heparin/graphite oxide (PLLA-CS/Hep/GO). After the LBL coating, the hydrophilicity and mechanical properties of the modified PLLA nanofibers were greatly enhanced. Moreover, the CS/Hep/GO coating positively influenced cell attachment, viability, and proliferation of endothelial cells [93].

The versatility of electrospinning allows to obtain complex and ordered structures. A compelling example has been reported by Shao and co-workers [94] who used electrospinning to develop a 3D scaffold with multiple orthogonal aligned fibers. This peculiar architecture improved mechanical properties and decreased issues that may arise when working with parallel fibers or random networks. Moreover, a 3D structure better mimics the natural cellular environment. They developed an electrospun PLGA/silk fibroin/GO/hydroxyapatite (PLGA/TSF/GO/HA) 3D scaffold. hMSCs seeded onto these scaffolds showed enhanced proliferation and elongated morphology along the long axis of the nanofibers. Lastly, biological assays indicated that composite scaffolds enhanced osteogenesis and alkaline phosphatase activity.

In another work, Zhang and co-workers combined GO nanosheets and aligned aminolyzed PLLA nanofibers which favored nerve regeneration. The aminolysis of PLLA nanofibers allowed to form a stable GO coating. Schwann cells (SCs) cultured on these nanocomposite scaffolds displayed improved proliferation and elongation along the fiber direction compared to those grown on the aligned PLLA and aminolyzed-PLLA. The coated structure was also able to improve differentiation and neurite outgrowth of pheochromocytoma derived PC12 cell line. The authors suggest that these results may arise from the modification of surface chemistry and roughness induced by the GO coating [95].

#### 2.3.3. Hydrogels

Hydrogels are three-dimensional entangled networks able to retain large amounts of water. Despite being mostly liquid, they display a solid-like rheological behavior and recently they have been employed as scaffolds for tissue engineering [96,97]. Hydrogels can be categorized into two main classes based on the forces involved in building the network: (i) chemical and (ii) physical gels. The network of chemical hydrogels is obtained through covalent cross-linking of its components, which generates a permanent structure. On the other hand, the structure of physical hydrogel is characterized by reversible noncovalent interactions which make these gels suitable for cell encapsulation but highly susceptible to environmental conditions (i.e., ionic strength, pH, temperature), such that even minor changes can cause the network to collapse. Indeed, physical hydrogels exhibit lower mechanical properties than their chemical counterpart. However, even chemical gels generally cannot withstand high mechanical stress despite the covalent cross-links [98].

Graphene and graphene derivatives in hydrogels may play the role of (i) self-assembling gelator molecule or (ii) filler in order to prepare multi-functional nanocomposite hydrogels. Self-assembly has been recognized as one of the most effective "bottom-up" strategies for building structured networks. Driven by non-covalent π-π interactions that arise from their 2D structure, graphene and graphene derivatives spontaneously re-organize into a 3D structure. Self-assembled hydrogels can be prepared through a one-step hydrothermal method starting from a graphene-based solution [99]. For example, Yang and colleagues have demonstrated the jellification of GO at the solution–filter membrane interface, creating highly conductive and anisotropic films [100].

The employment of pure graphene and/or graphene derivatives hydrogels is quite restricted, thus they are mainly used as high-quality nanofillers for composite hydrogels [101]. Different synthetic and natural polymers able to form hydrogels are suitable for tissue engineering scaffolds. Among synthetic polymers we may mention polyethylene glycol (PEG), poly(acrylamide), poly(lactic acid) or synthetic peptides. Natural-derived polymers such as alginate, chitosan, collagen, silk or gelatin are also widely used to fabricate hydrogel scaffolds for tissue engineering. Polymeric scaffolds display good biocompability and biodegradability but lack, for example, the ability tolerate strong mechanical forces [102,103].

Alginate is a natural polysaccharide composed of β-D-mannuroic acid (M) and α-Lguluronic acid (G) typically obtained from brown seaweed. In the presence of various divalent cation (Ca2+, Mg2+), alginate polymers form gels via non-covalent cross-linking of the carboxylate groups of the G blocks on the polymer backbone. Even if the concentration of crosslinker, percentage of G content and jellification time allows to tune the properties of alginate hydrogels, other limitations cannot be overcome without the use of specific fillers. Particularly, alginate-based hydrogels do not permit good control over their internal architecture, they lack cell receptors adhesion sites and suffer from low protein adsorption capability [104]. As independently highlighted by Losic et al. and Chen et al. [105,106], the introduction of GO and RGO in an alginate matrix allows to modify and control the porosity of the gel (ca. 99%±0.3%), making the pores size uniform from surface to its inner core and fostering cellular activity. GO and RGO composite gels also allow to reach the optimal swelling index required for an efficient scaffold. Investigation of mechanical and electrical properties revealed an optimum GO content of 0.1 wt%. Above this concentration a detrimental effect was observed due to an imperfect dispersion of GO within the alginate matrix.

Chitosan, as well its derivatives, is a widely available natural polymer characterized by excellent biological properties (i.e., biocompatibility, coagulation activity, biodegradability). Agarose (AG), on the other hand, is a polysaccharide obtained from red algae, displays a thermo-sensitive behavior and exhibits mechanical properties similar to that of soft tissues. However, its employment is limited by the lack of cell recognition sites.

Sivashankari and Prabaharan used GO as a nanofiller for the fabrication of agarose/ chitosan (AG/CS)-based scaffold [107]. Through a freeze-drying method, they prepared 3D AG/CS/GO scaffolds with different concentrations of GO (0–1.5 wt %). GO introduced changes in the scaffold morphology, in their swelling behavior and in their water retention ability. In particular, AG/CS/GO scaffolds with 1 and 1.5 wt % of GO exhibited the highest

porosity (Figure 6), with an average pore size (237–274 μm) matching the demands for bone tissue regeneration [108,109]. Even with the increase in porosity, GO likewise enhanced the mechanical properties due to interactions established between fillers and polymer matrix and favored cell attachment and proliferation. Freeze-drying techniques are widely employed for scaffold generation and also allows to obtain anisotropic scaffolds. Liu et al. developed a highly oriented hydrogel through directional freezing of CS/GO suspension on a copper plate cooled with liquid nitrogen [110]. This method produced micro-sized ice rods within the suspension, which act as template for a honeycomb-like structure resembling a bone lamellae structure. The resulting hydrogel displayed anisotropic mechanical behavior improved by the incorporation of GO and were able to guide the growth of mouse osteoblastic MC3T3-E1 cells along the longitudinal direction of the honeycomb structure.

**Figure 6.** SEM images of (**A**) AG/CS (0% GO), (**B**) AG/CS/GO (0.5% GO), (**C**) AG/CS/GO (1% GO), and (**D**) AG/CS/GO (1.5% GO) composite scaffolds [107].

Self-assembling peptide-based hydrogels (SAPHs) have been widely employed as vehicle for drug delivery, but they can also be employed in tissue engineering due to their biocompatibility and non-immunogenic nature [104,111,112]. Ligorio et al. used GO as nanofiller in a peptide (FEFKFEFK) hydrogel for tissue engineering [51]. After conditioning with cell culture media (i.e., pH 7.4), all gels displayed an enhanced storage modulus. Bovine nucleus polposus (NP) cells were cultured on these hydrogels to assess cell viability and GO-hydrogels with shear modulus similar to the native NP showed higher viability and constant metabolic activity throughout the culture period.

Wang et al. prepared a silk fibroin scaffold incorporating exfoliated graphene [113]. An aligned silk fibroin hybrid hydrogel was obtained by application of an electric field. Even if aligned silk nanofiber gels were previously proven to be able to influence behavior such as cell orientation and migration [114], they failed to actively induce neural differentiation. The nanocomposite hydrogels displayed anisotropic mechanical properties, and the one with the highest content of graphene showed doubled parallel and orthogonal compressive moduli compared to graphene-free samples, making them suitable for nerve tissue engineering. After the addition of graphene, cell proliferation was further enhanced, indicating that graphene sheets effectively induced neurite differentiation.

In recent years, injectable hydrogels have drawn major attention since they need minimal invasive procedure to be administered and have reduced therapeutic costs. The hydrogel precursor should be injected as a controllable liquid (i.e., characterized by low viscosity) and must jellify into a robust hydrogel as quickly as possible *in situ* [115]. Finally, it is uttermost important that gelation occurs after injection and at physiological conditions (temperature and pH). Recently, Lee et al. developed an injectable GO-incorporated

glycolchitosan-oxidized hyaluronic acid (gCS/oHA) hydrogel [116]. Gelation of gCS/oHA was obtained through the cross-link between the aldehyde group on oHA and the amine groups of gCS (Schiff-base reaction). Frequency sweep experiments were used to investigate the mechanical properties in a plate–plate geometry. The results showed that when the GO content increased, the G' value gradually increased too, suggesting a more robust hydrogel formation. GO may enhance polymer cross-linking through hydrogen bonding interactions [117,118]. GO-incorporated hydrogels displayed lower cytotoxicity and higher osteogenic activity compared to control both in vitro and in vivo. High levels of COL1 expression observed in cultures hinted that these injectable gels could be suitable for treating bone injuries. Saravanan et al. explored chitosan-glycerophosphate-based injectable hydrogels for treatment of bone defects [119]. Due to newly introduced non-covalent interactions, GO (0.5% *w*/*v*) composite hydrogels significantly increased swelling, protein adsorption and cell interaction compared to their GO-free counterparts. Moreover, GO introduction reduced gelation times and controlled degradation rates.

Poor dispersion of GBMs within the polymer matrix causes aggregation [105], which may be detrimental for scaffold properties. To achieve homogeneous and stable dispersions of GBMs, covalent and/or non-covalent functionalization may be required. Díez-Pascual et al. [120] fabricated poly(propylene fumarate) (PPF)-based nanocomposites reinforced with GO, non-covalently functionalized with PEG (PEG-GO). PEG functionalization reduces the aggregation tendency and cytotoxicity of GO without impairing its unique features. The presence of PEG-GO leads to a threefold increase of Young's modulus at 3% loading of filler and improved cell adhesion and growth. The results have been ascribed to the roughness of the scaffold, the hydrogen-bonded network established between GO and the polymer and the good GO dispersion inside the matrix. Polymers may be also covalently bonded to GO, for example by esterification. Noh et al. designed a graphene oxide GO covalently functionalized with acrylated polyethylene glycol (PEGA-GO) through ester formation [121]. The PEGA-GO was photopolymerized with polyethylene glycol diacrylate (PEGDA) leading to gel formation. GO-doped hydrogels boosted cell adhesion and osteogenic differentiation, though no changes were observable in swelling and mechanical properties.

Wu et al. took advantage of the abundant functional groups on the surface of nanosized GO to link starch chains via esterification [122]. Starch is a widely available and cost-effective polysaccharide, which does not release degradation products that induce inflammations in vivo. Nanosized GO was synthesized from starch through microwaveassisted degradation and then covalently bonded to the polysaccharide itself to improve its mechanical features and bioactivity. In another example, Ruan et al. crosslinked carboxymethyl chitosan (CMC) to GO by amide bond formation [123]. The obtained GO-CMC scaffolds appeared rougher than their GO-free counterparts and showed better retention properties and slower degradation rates thanks to the higher cross-linking degree compared to the GO-free and CS/GO-CS samples. Water uptake and retention rates are important parameters, since the scaffold is the vessel for nutrients and metabolites for cell activity. The authors highlighted that GO introduction also deals with the poor mechanical strength typical of bare CMC [124].

## **3. Stem Cell Differentiation and Mechano-Transduction**

## *3.1. Tissue Engineering and Stem Cells*

Stem cells are non-specialized cells with self-renewal potential and the ability to differentiate into various cell types if directed with appropriate stimuli, making them a powerful tool for the regeneration of injured tissues [125]. Embryonic Stem Cells (ESC) are pluripotent stem cells able to originate all the cell types of the body [126]. Despite their ideal self-renewing capabilities and differentiation potential, they are not widely used for tissue engineering studies due to the ethical restrictions of human embryo use in research. As a valid alternative, tissue engineering switched the focus to adult stem cells, which are stem cells residing throughout the body whose role is to maintain and repair the tissue

in which there are found. Such cells have a limited differentiation potential compared to ESCs but offer the advantage of being isolated directly from the patient for autologous regenerative therapies. Good examples of adult stem cells are the Mesenchymal Stem Cells (MSC) and Hematopoietic Stem Cells (HSC). They both can be isolated from patient bone marrow and can regenerate bone, cartilage, and adipose tissue (MSCs), as well as the entire immune system (HSCs) [127,128]. However, some adult stem cells, like neuronal stem cells (NSCs), can be isolated only with very invasive procedures and in small quantities [129]. As of now, the most promising stem cell type for regenerative applications are the induced Pluripotent Stem Cells (iPSCs). iPSCs are generated from the "reprogramming" of somatic cells back to the pluripotent "embryonic" state [130]. Therefore, they show the same "unlimited" self-renewal and differentiation capabilities of the ESCs, with the advantage of being free from ethical issues as reprogrammed from patient or donor-matched somatic cells [131]. Challenges associated with the iPSC clinical use are (i) the difficulties in finding HLA-matched donors (especially for mixed-race patients) and (ii) the time and costs for the development of patient-derived iPSCs, particularly considering the extensive validation and stringent regulatory processes that would require each patient-derived cell line [132,133]. However, recent works proposed new strategies to engineer such iPSCs to make them "invisible" to the recipient immune system, showing that we are very close to the generation of off-the-shelf, universally compatible iPSCs for the allogenic treatment of a myriad of diseases [134–136]. The interaction between stem cells and the extracellular microenvironment is critical in controlling stem cell differentiation, as depicted Figure 7.

**Figure 7.** Schematic representation of biophysical and biochemical cues on cell differentiation.

#### *3.2. Cues Controlling Stem Cell Behavior*

Biochemical cues are provided by reciprocal interactions between the cell, soluble bioactive agents, and the ECM. Soluble molecules, such as growth factors, chemokines, and cytokines, diffuse to bind the cell surface receptors and have potent effects on cell growth, proliferation, and differentiation. Insoluble ECM macromolecules (e.g., collagens, elastin, and laminin), glycoproteins (e.g., fibronectin and vitronectin), and polysaccharides (e.g., heparan sulfate and hyaluronic acid) form a meshwork of fibers or fibrils with ECM glycoproteins incorporated into them. The resulting matrix is tissue-specific and functions as both a structural and signaling scaffold to cells [137].

Many works showed that some of the aforementioned molecules—if administrated both in vitro and in vivo—are able to elicit specific cell responses [138]; moreover, different strategies have been developed to link such proteins to biomaterial scaffolds in order to help delivery at the injured sites [139]. However, coating surfaces with recombinant proteins or native matrix macromolecules extracted from animal tissues encounters the problem of eliciting immune responses, in particular when using proteins from different species. Furthermore, their isolation and purification from native tissues or their production as

recombinant proteins at a larger scale for tissue engineering purposes is expensive and subject to batch-to-batch variability [140]. For these reasons, the production of specific motifs known to mediate regulatory signals as synthetic peptides presents significant advantages compared to using entire recombinant/native tissue proteins: (i) low immunogenic activity; (ii) increased stability; (iii) low production costs; and (iv) simplified preparation and immobilization onto substrates. Moreover, peptides can be: (v) presented to cells at surface densities significantly higher than those possibly achieved with entire proteins or domains; and (vi) tailored in composition for each tissue-specific application [141]. The biomimetic peptides most used for scaffold functionalization are the ones representing the ECM protein epitopes for integrin binding and therefore promoting cell adhesion [142]. In addition, tissue-specific peptides, resembling active motifs of growth factors and transmembrane proteins, have also been used to tune the cell differentiation [143,144].

Cells are capable of sensing and responding to biophysical cues, over a wide range of length scales. Many of these cues are provided by the ECM, which acts as a cellular scaffold and is the primary extracellular component in tissues. In vivo, the ECM, through its structure and molecular composition, presents a variety of geometrically defined, three-dimensional (3D) physical cues in the submicron to micron scale, referred to as topographies. Cell response to topographies is mediated by a phenomenon called contact guidance, which is known to affect cell adhesion, morphology, migration, and differentiation [145]. Another physical cue displayed by the ECM is mechanical stiffness through which, similar to topography, a diverse set of cellular functions can be modulated. Matrix sensing requires the ability of cells to pull against the matrix and cellular mechanotransducers to generate signals based on the force that the cell must generate to deform the matrix. Mechano-sensitive pathways subsequently convert these biophysical cues into biochemical signals that commit the cells to a specific lineage [145]. For example, MSC differentiation can be modulated by substrate stiffness [146], while developing neurons are able to transduce topographical stimuli through the interaction of the growth cone with the immediate environment. Such mechanical cues direct neurite extension, ensuring appropriate and regulated connectivity within the overall neural circuitry [147].

ECM mimicry can be achieved using either natural or synthetic polymers interconnected by physical and ionic interactions and even covalent linkages [148]. Electrospun polymer fibrous substrates with controlled fiber architectures and diameter provide topographical cues to cells by presenting geometries mimetic of the scale and 3D arrangement of the collagen and laminin fibrils of the ECM. Such polymer fibers present a high surface-tovolume ratio and porosity and are hence well-suited for promoting cell adhesion, growth, and differentiation and enable growth factor/drug loading; such properties are inherent to bioactive matrix microniches [50,149]. Recent advances in 3D bioprinting strongly improved our ability to imitate natural features of ECM. As an additive manufacture technology, the 3D bioprinting allows deposition of polymers, hydrogels, cells, growth factors, and peptide active motifs by using a layer-by-layer approach to build up arrangements favorable to tissue-like structure formation, which are endowed with superior differentiating properties compared to the conventional 2D culture vessels [148].

Endogenous electrical signals are present in many developing systems and influence crucial cellular behaviors—such as cell division, cell migration, and cell differentiation [150]. Some cell types, like osteoblasts, neurons, and cardiomyocytes, are especially sensitive to electrical signals as they activate membrane receptors and downstream intracellular signaling elements leading to specific cell responses [151]. Not only cells, but also extracellular matrix proteins, such as collagen, fibrin, and keratin, can generate electrical currents upon mechanical stress, a phenomenon known as piezoelectricity [152].

Electrically conductive scaffolds not only enhance path finding of growing axons [153], but also improve cell survival and functional integration after transplantation in vivo by providing structural support for transplanted cells and facilitating synaptogenesis with host cells by restoring the neuronal network activity [154,155].

Since electroactive myocytes are responsible for heart and muscle contraction, electrically conductive materials found applications in cardiac and muscle tissue engineering as they support and maintain cell electrophysiology [22].

Even though bone cells are non-excitable cells, stress-generated piezoelectricity has been shown to stimulate bone precursor cell proliferation and differentiation to restore the injured site, making electroactive materials and electrical stimulation a valid tool for bone regeneration strategies [156].

Stem cells are also sensitive to electrical cues and their differentiation can be modulated by electrical stimulation and culture on electroactive materials. NSC ability to undergo neuronal and glial differentiation is boosted by electroactive material and exogenous electrical field [157,158]. The use of electrically conductive material have also been shown to promote the neuronal differentiation of adult stem cells derived from non-neural tissue without the addition of neuron-specific growth factors and cytokines [159,160]. Cardiomyocyte differentiation of iPSCs, ESCs, and MSCs is possible with the use of chemically defined media, but it dramatically increases if coupled with electroactive materials, showing protein expression, cell morphology, and contractility of the natural tissue [161,162]. Similarly, MSC differentiation into osteoblasts can be achieved with specific osteogenic media; but it is further supported by the aid of electroactive scaffolds [163].

#### *3.3. The Importance of the "Nanoscale"*

Cells have micro and nanoscale sensitivity because the extracellular environment presents a variety of spatially defined cues in the sub-micron to micron scale (Figure 8).

**Figure 8.** The relative scale of biological molecules and structures.

At the nanometer level, the extracellular environment affects sub-cellular behaviors such as the organization of cell adhesion molecule receptors. At the micron level, the extracellular environment affects cellular and supracellular characteristics such as cell morphology and [163]. The nanoscale physical features of the scaffolds can affect cell behavior. Natural tissues have indeed a hierarchical structure ranging from the macroscale (>1 mm) to the microscale (1 μm–1 mm), and the nanoscale (<1 μm). As a result, individual cells (typically in the size range 10–50 μm) respond in different ways to structures at different length scales. It has been shown that integrin receptors possess characteristic dimensions on the order of 10 nm [164]. The basement membrane of organs consists of nanoscale fibers (line topography) and pores (holes) that range in diameter from a few nanometers to several hundred nanometers [165]. The tubular fibers of collagen also have nanoscale dimensions [166] and laminin shows a nanoscale texture as well [167]. Given that cell ECM is patterned down to the nanoscale, cell-biomaterial interactions in scaffolds can be optimized by incorporating features of nanoscale dimensions. Indeed, surfaces topographically structured at the submicron scale can affect a wide variety of growth parameters, such as cell adhesion, morphology, viability, genic regulation, apoptosis, motility, and differentiation [168]. Evidence from nanoscale topography analysis suggests that nanoscale features eliciting a cell response are in the same size range (50–70 nm) that is associated with integrin cluster formation [169]. Further studies showed that scaffold nanotopography can control cell fate by altering cell and nucleus shapes, hence activating

intracellular signal transduction and silent gene expression [125,170]. This is particularly true for neurons that, thanks to their growth cones, sense and actively respond to the surface nanotopography with a surprising sensitivity to variations of few nanometers [171].

Molecular deposition and lithographic techniques allow the patterning of tissuespecific molecules with nanometer resolutions. For example, the deposition of molecules that promote and support neuronal adhesion, growth, and differentiation on regenerating scaffolds enables the selective adhesion and growth of neural cells and a controlled neurite extension along the geometric pattern [172]. Apart from peptide/protein nanopatterns, nanomaterials and nanotechnology tools can also be used to develop special scaffolds able to recapitulate the architecture of structural proteins within ECM and the nanoscale features that model native ECM nanotopography [142]. Nanomaterials take advantage of their unique molecular features to induce, with high specificity, a number of desired physiological responses in target cells and tissues, while minimizing undesirable effects [173]. The peculiar mechanical and chemical properties of nanomaterials can be exploited for integration with native tissue in long-term implants; moreover, their nanoscale features have the potential to interact with the biological system at the molecular scale, while offering elevated levels of control [174].

Combinations of stimulatory cues may be used to incorporate nanoscale topographical, biochemical, and electrical cues in the same scaffold to provide an environment for tissue regeneration that is superior to inert scaffolds. This approach—able to precisely regulate cell differentiation, morphology, and polarization—is fundamental for the development of next-generation scaffolds suitable for clinical applications.

## *3.4. Role of the Biomolecular Corona*

Interactions between the surface chemistry of nanomaterials and surfaces of biological components (proteins, phospholipids, organelles, DNA etc.) are crucial to determine the effects on cells and tissues. As soon as a nanomaterial comes in contact with a biological fluid (i.e., cell culture media, blood or interstitial fluid) it is coated with ions and proteins and develop a new interface which is often referred to as the protein corona or biomolecular corona (BC). This layer at the nanobio interface defines the biological identity of the nanomaterial, determining cell interactions, uptake and clearance [175]. Protein adsorption by GBMs has been reported in numerous studies. Umadevi and Sastry [176] analyzed non-covalent interactions on the surface of carbon nanostructures and highlighted that the graphitic lattice of graphene allowed hydrophobic interactions and strong π-π stacking with aromatic amino acids (Phe, Tyr, Trp) with binding energies between 15 and 20 kcal mol<sup>−</sup>1. Surface chemistry is key in tuning the strength and type of interactions. Epoxide, hydroxyl and carboxyl groups on the surfaces favor hydrogen and electrostatic bonding with proteins, facilitating adsorption on GO compared to pristine graphene or RGO. GBMs have been shown to strongly bind to different serum proteins such as albumin, fibronectin (Fn), collagen, and laminin [177]. Therefore, when cells grow onto a graphene-based scaffold, they show an enhanced capacity to form focal adhesions by clustering integrin molecules and favoring cell adhesion [125]. In addition, GBMs capacity to adsorb proteins results in trapping growth factors produced by the cells during their differentiation. Such growth factors can progressively be released during cell maturation, allowing a continuous supply, which is suitable for long-term cell differentiation [177].

Not only graphene physical properties favor the adsorption of proteins, but they also offer tremendous opportunities for the covalent functionalization of protein active motifs and chemical groups [178]. Such approach allows the stable attachment of signalling molecules to the graphene structure to influence cell behaviour, but it also simplifies combination of graphene with both natural and synthetic polymers for the development of superior scaffolds combining multiple cues for cell growth and differentiation.

## **4. Nanotoxicology and Functionalization**

#### *4.1. In Vitro Cytotoxicity*

The use of graphene-based nanomaterials (GBMs) does not come without possible concerns about in vitro cytotoxicity and in vivo biocompatibility. As anticipated in Section 3.4, the biomolecular corona (BC) plays an important role in regulating the fate and toxicity of nanomaterials that interface with a biological environment. Due to the unique and distinct physico-chemical properties of graphene and its derivatives, there is an enormous variability at the nano-bio interface which leads to different intrinsic toxicological effects. Moreover, nanomaterials are often pre-bound to chemical moieties that originate from the manufacturing process, from stabilizers used in their preparation or from exposure to gasses or buffers, all of which might further impact biocompatibility. Therefore, any generalization would be inaccurate, possibly misleading and must be avoided [25,179].

Pristine GBMs have been shown to have a dose- and time-dependent in vitro toxicity in both procaryotic [180–183] and eucaryotic cells [184–187]. Graphene has a hydrophobic nature that often causes irreversible aggregation in cell culture media and it has been reported to agglomerate on cell membranes causing physical damage [188]. Conversely, oxidized derivatives of graphene, such as graphene oxide (GO) and reduced graphene oxide (RGO), are more hydrophilic and show little aggregation in biological buffers resulting in lesser cytotoxicity [179]. According to Chatterjee and co-workers [189], who performed a comprehensive study about biological interaction of oxidized graphene derivatives, GO and RGO had similar toxic responses with different dose-dependency and distinct molecular mechanisms which were attributed to their peculiar surface oxidation status. However, the presence of oxidative functional groups on the surface can lead to the generation of reactive oxygen species (ROS). In addition, if they are not correctly washed, graphene nanomaterials might retain residual chemicals applied to separate the graphitic layers or during the fabrication of oxidized derivatives.

To solve these problems, novel green approaches for nanoparticle synthesis and modification have been developed, involving the use of biocompatible surfactants and reducing agents. According to Askari et al. [190], graphene nanosheets can be synthetized in the presence of Herceptin, a natural antibody, using an ultrasonic-assisted exfoliation method. The toxicity of graphene was tested in 3D spheroid cultures of human breast adenocarcinoma cell line (SKBR-3) to better mimic the natural tissue micro-environment. The authors concluded that that presence of Herceptin and its residues on graphene nanoparticles created a biocompatible platform suitable for cell growth. In another study, Narayanan et al. [191] described a facile and green synthesis of reduced graphene oxide by the deoxygenation of GO under aqueous alkaline conditions in the presence of soluble starch as a reducing agent (SRGO). The cytotoxicity of SRGO on skin fibroblasts was evaluated using a Wst-1 assay and showed that SRGO showed a substantial increase in cell viability at high concentrations (200 μg mL−1) compared to non-reduced GO. The authors also investigated the hemocompatibility profiles of the nanomaterials and revealed that both caused a hemolysis effects compared to negative controls. However, SRGO did not exhibit a direct proportionality between hemolytic activity and concentration, with hemolysis staying as low as ~4.9% in maximum concentration samples.

#### *4.2. Hemocompatibility and Interaction with Immune System Cells*

Understanding interactions between nanomaterials and blood is key to determining in vivo biocompatibility due to the unavoidable contact between the two. Thanks to the protein corona effect, nanoparticles that touch blood or enter the bloodstream are coated by a milieu of proteins that may undergo conformational changes, exposing new epitopes and promoting phagocytosis or elimination from the circulation [192]. Nanomaterials can cause hemolysis and activate or interfere with clotting and coagulation cascades [193], seriously hindering the health of the organism.

The hemolytic property of nanoparticles is influenced by their distribution size, shape, surface charge and chemical composition [194]. Jaworski et al. [195] studied pristine

graphene, RGO and GO effects on chicken embryo red blood cells (RBCs) and reported altered RBC morphology with loss of biconcavity. All of the nanomaterials exhibited dose-dependent hemolytic activity towards RBCs, with highest hemolysis rates observed at 5 mg mL−1. Pristine graphene showed the highest hemolysis (73%), followed by RGO (42%) and GO (27%), correlating with the degree of surface oxidation. Lower hemolytic concentrations and activity have been reported by other groups [193]. However, according to Duan et al. [196] the hemolytic potential of GO can be largely reduced by pre-incubating it with BSA or FBS, exploiting their extremely high protein adsorption ability. In another work, Sasidharan et al. [197] provided evidence that pristine graphene and GO have excellent hemocompatibility showing no hemolysis, platelet activation or plasma coagulation up to a relatively high concentration (75 μg mL−1) and under in vitro conditions. The authors also highlighted that pristine graphene had the potential to upregulate the production under sterile conditions of pro-inflammatory cytokines, such as IL-6 and IL-8. Cytokines are soluble glycoproteins released during an inflammatory response that recruit immune cells in order to tackle foreign bodies that have entered the organism.

Understanding interactions of GBMs with the immune system is of considerable relevance both from a toxicological and biomedical perspective. The BC of carbon-based nanomaterials is abundant in complement proteins. These proteins play a central role in modulating the immune and inflammatory responses towards intruders and may be a key factor in generating chronic ailments (such as allergy and sterile inflammation) by recruiting neutrophils and macrophages [198,199]. In addition, complement activation can promote cell-mediated immunity by enhancing generation of antigen-specific immunoglobulins by B-cells, activation of T-cells and uptake by dendritic cells [200]. Neutrophils and macrophages are part of the reticuloendothelial system (RES) which is responsible for the uptake and clearance of foreign bodies that have entered the organisms: the former are normally the first to intervene in an inflammation reaction, whereas the latter arrive later and promote tissue healing. It has been reported that macrophages better uptake hydrophilic systems compared to hydrophobic graphene since it is poorly dispersible in water and remains blocked on the cell surface [197]. Similarly, neutrophils are involved in nanoparticle clearance and it has been shown that exposure to carbon-based nanomaterials may upregulate neutrophils infiltration in tissues [201]. Carbon nanomaterials are also known to trigger apoptosis and/or cell death in macrophages, causing significant impairment in the immune resistance of subjects if used in vivo. However, Lin et al. [202] reported in a recent study that macrophage viability and activation are found to be mainly unaffected by few-layered graphene (FLG) at doses up to 50 μg mL−<sup>1</sup> and therefore it is of little toxicity for M1 and M2 human macrophages, even though it triggers cell stress, ROS and inflammatory cytokines. Notably, neutrophils and macrophages are cleared from the circulation via the liver, spleen and kidneys and there is evidence that bone marrow may also play a major role in their clearance [203]. Therefore, nanomaterials carried by these cells can accumulate in those districts, causing unexpected issues and altering their fate.

#### *4.3. In Vivo Biocompatibility*

In vivo biodistribution and pharmacokinetics of GBMs have been studied in small and large animal models [204–209] in order to investigate the adsorption, distribution, metabolism and excretion (ADME). The fate of nanomaterials in organisms is influenced not only by their properties, but also from the pathway of exposure. Thanks to the wide range of potential applications of GBMs in biomedicine, exposure can occur in a number of ways including inhalation, intratracheal instillation, oral gauge, injection (intraperitoneal, intravenous or subcutaneous) or through debris generated from worn or biodegraded implants [204,210]. Once inside the organism, nanomaterials can make their way into the bloodstream even if not directly injected there and spread throughout the body. In addition, there is evidence that GBMs can diffuse across biological barriers such as the blood-air, blood–brain, blood-testis or blood-placental barrier, and accumulate in organs causing acute and chronic inflammation, tissue lesions and necrosis [211,212]. Krajnak and

coworkers [213] examined graphene nanoparticles of different sizes and different forms (carbon black, graphene, GO and RGO) to determine if pulmonary exposure resulted in changes in vascular function and expression of acute response markers in mice. It was observed that while graphene altered gene expression in cardiovascular system, no changes were produced in the peripheral vascular function. On the other hand, pulmonary exposure to the oxidized forms of graphene had a more acute effect on heart and kidneys and repeated exposure might lead to injury or dysfunctions. Another study reported that GO provokes severe and persistent injury in mice lungs including granulomas persisting for up to 90 days [214]. Biodistribution experiments on intratracheally instilled carbon-14 labeled FLG showed that even if it was mainly retained in lungs, it was also redistributed to the liver and spleen passing through the air-blood barrier. However, no detectable absorption of FLG was observed when administered orally [212]. Conversely, radioactive-labelled RGO given through an oral gauge was rapidly absorbed in the intestine, metabolized by the kidneys and then excreted via urine [215]. Intravenous injection of GO in mice elicited blood platelets aggregation and extensive pulmonary thromboembolism, while low uptake was observed in the RES [216]. Surprisingly, a recent study by Newman et al. [217] highlighted that GO sheets accumulate preferentially in the spleen and progressively biodegrade over nine months. They evaluated the potential consequences of this prolonged accumulation and found limited effects on spleen histopathology and splenic function. Cell-mediated immune response was measured by determining the populations of T lymphocytes, specifically CD4+ and CD8+ cells as the major immune component of the splenic white pulp. Moderate changes were seen in both cell populations in mice injected with GO (2.5, 5, and 10 mg/kg) at both 24 h and one month after administration and no significant differences in the levels of the proinflammatory cytokines IL-6 and TNF-α were detected at any time point compared to control. However, they registered a significant drop in anti-inflammatory cytokines expression at 24 h and at the one-month time point for all tested GO doses. The authors concluded that reduction in cytokine expression after GO treatment may indicate the involvement of the innate immune system in regulating the effects of GO.

#### *4.4. Minimizing GBM Toxicity*

Although the inherent toxicity of graphene and its derivatives is a major drawback for their biomedical applications, it is a well-known problem, and different strategies have been developed to overcome it. In an attempt to enhance their overall safety and minimize the risks for adverse reactions in humans from exposure, Bussy et al. [218] offered a set of rules for the development of graphene and its derivatives: (1) use small, individual graphene sheets which are more efficiently internalized by macrophages in the body and removed from the site of deposition; (2) minimize aggregation using hydrophilic, stable, colloidal dispersions of graphene sheets; (3) use excretable graphene material or chemically-modified graphene that can be degraded effectively. Biological responses to these nanomaterials depend on various properties and it has been reported that smaller particles and higher oxidation improve biocompatibility [219]. Most importantly, variations in surface chemistry play a major role determining their toxicity and pharmacokinetic profile [220]. Highly hydrophobic graphene tends to aggregate in aqueous solvents thanks to intermolecular attractive Van der Waals forces, π-π stacking, hydrogen bonds and electrostatic interactions [221]. This tendency makes it hard to manipulate and characterize their biocompatibility and it has been suggested that the high percentage of controversial results in toxicity statistics could be owing to the dissimilarities in GBMs solubility [222]. Therefore, improving the dispersion of graphene-based nanomaterials in various solvents is a prerequisite for their further applications. Recent strategies include sonication, stabilization with surfactant and surface functionalization.

Wojtoniszak et al. [223] showed that GO and RGO exhibit a surfactant-dependent toxicity by comparing the homogeneity of GO and RGO dispersions in phosphate buffered saline (PBS) and cell viability on mice fibroblasts L929 cells. Three different dispersants were used, namely PEG, poly(ethylene glycol)-*block*-poly(propylene glycol-*block*-poly(ethylene glycol) (Pluronic P123), and sodium deoxycholate (DOC). The authors concluded that both materials had relatively good cytocompatibility in the 3.125–12.5 μg mL−<sup>1</sup> range, with lowest toxicity detected in PEG-stabilized GO.

The BC can be tuned and modified by exploiting the ability of GBMs to adsorb moieties from the culture medium. As previously discussed, biological and bioactive species (DNA, carbohydrates and proteins) can be used as surfactants to stabilize graphene nanomaterials in aqueous solution, paving the way for different biomedical applications [191,224,225]. Pre-incubation in protein solutions was shown to form a thin coating on nanomaterials in suspension, minimizing cytotoxicity by limiting their direct interaction with cells. It has been reported that GO and RGO coated with BSA [226], FBS [227] or serum proteins [228] showed attenuate cytotoxicity and could improve biocompatibility. In addition, another interesting method of surface modification is functionalization through exposure to a specific enzyme, peptide or antibody [190,229]. Bussy et al. [230] exposed human lung carcinoma (A549) and bronchial epithelial (BEAS-2B) cell lines to GO and analyzed its effect in serum-free HEPES-buffered salt solution (BSS), Dulbecco's phosphate-buffered saline (PBS), and the normal media recommended for these cell lines (F12 for A549 and RPMI for BEAS-2B). Surprisingly, they reported more pronounced cellular responses in both BSS and PBS, but not in F12 or RPMI, and concluded that the interaction between GO and cells may differ depending on the concentration of salts and ions present in the aqueous environment. These charged moieties could influence nanofiller aggregation, bundling, stacking, or other colloidal properties of the negative surface-charged nanoparticles. In summary, the presence of proteins and other moieties in the cell culture medium influences the results on cytotoxicity and we could consider that GO and RGO might not be hemolytic in vivo where an abundant BC forms on their surfaces and protects the nanomaterial. These types of non-covalent surface functionalization are not stable in prolonged circulation and it is important to consider the dynamic changes of the BC as the nanoparticles translocate from one biological compartment to another or from the ECM intracellular locations [231].

Covalent functionalization is another strategy for enhancing solubility in different matrices and is frequently used to obtain nanocomposites, as previously discussed in Section 2. However, covalently bonding molecules to the surface leads to the disruption of the graphitic lattice changing its electronic and transport properties [232]. A number of surface modifications allow to obtain more hydrophilic GBMs with remarkably reduced toxicity. According to Kiew and coworkers [233], graphene-based nanomaterials with hydrophilic surfaces weaken the opsonin–protein interaction and could avoid being recognized by macrophage, thus an inflammatory response. Among the different strategies, the combination with polymeric materials represents a commonly used approach to overcome the limitations of graphene-based nanomaterials in biomedicine [234]. Several studies have reported that covalent modification with polyethilenglycole (PEGylation) can reduce cytotoxicity resulting in increased biocompatibility and stability in physiological buffers [235]. PEG is known to prolong particle circulation in the blood due to its ability to camouflage particle surfaces, sterically shielding against opsonization and uptake by the RES cells [192]. Other approaches involve the covalent attachment of conductive polymers (such as poly(pyrrole), poly(aniline), poly(allylamine)) or biodegradable synthetic polymers (e.g. poly(lactic) acid, poly(glycolic) acid, poly(lactide-co-glycolide)) or natural polysaccharides such as chitosan [236], alginate, hyaluronic acid and dextran (DEX). Like PEGylation, dextran coating reduces the adsorption of proteins the surface and improves biocompatibility. Compared with non-functionalized GO, GO–DEX conjugates showed improved stability in physiological solutions, accumulation in liver and spleen after intravenous injection, and most importantly clearance from body within a week without causing noticeable short-term toxicity [237].

Finally, graphene and its derivatives have been used in combination with biocompatible polymeric matrices to obtain conductive nanocomposites with enhanced cell adhesion, differentiation and biocompatibility [238]. These materials trigger reduced biological responses without the impairment of the GBMs capability to cross cell membranes and deliver therapeutic species [239]. As our lab has recently shown, organic-functionalized carbon nanofillers dispersed in a polymeric poly(L-lactic) acid exhibited enhanced cell viability (~90%) and supported cell growth [59], while having interest effects on the differentiation of neuronal precursors [50] and human circulating multipotent stem cells [47,159].

### **5. Examples of Tissue Regeneration**

## *5.1. Bone Regeneration*

Bones possess a remarkable regenerative capacity, as they maintain the ability to remodel themselves throughout adult life and they can repair fractures spontaneously [240]. After bone damage, soluble factors accumulate at the injury site and recruit mesenchymal stem cells, which, in turn, proliferate and differentiate toward osteoblasts. Subsequent calcification of the region results in a woven bone, which is finally remodeled by the renewing and resorptive actions of osteoblasts and osteoclasts [241]. Despite this regenerative process, there are instances where injuries may require clinical intervention to be completely healed. Autologous bone graft is a standard medical procedure for the treatment of bone-related diseases. Unfortunately, it is mostly limited by the availability of appropriate donor tissue [242]. Several scaffolds have been developed to enhance bone regeneration to overcome this issue and some representative examples are reviewed in this section.

For instance, graphene was used to coat three-dimensional hydroxyapatite scaffolds to support the growth and osteogenic differentiation of hMSCs. Scaffolds were found to be self-standing, as hMSCs differentiation did not require common differentiative molecules (i.e., dexamethasone or the bone morphogenetic protein 2) [160]. Moreover, graphene oxide was covalently linked to chitosan (CS), an animal-derived polymer already known to support cell adhesion and proliferation. The resulting polymer had better elastic modulus and hardness, which resulted in an increase in cell adhesion, spreading, proliferation, and formation of the extracellular matrix. Most importantly, cells grown onto GO-CS scaffolds showed an enhancement in calcium and phosphate deposition levels, a hallmark for osteoblastic differentiation [243].

Arnold and co-workers managed to enhance hMSCs osteogenic differentiation by directly functionalizing GO. Expressly, they set up an elegant universal synthetic procedure to covalently tether polyphosphates onto GO, generating a new phosphate-graphene material (CaPG) [244]. Their approach allowed them to obtain scaffolds with hydroxyapatite-like functionality at the interface, loaded with osteoinductive ions. They developed a 3D scaffold and assessed that its mechanical properties were comparable with bones (Young's modulus up to 1.8 GPa, compressive storage modulus up to 291 MPa, shear storage modulus up to 545 MPa, and ultimate compressive strengths up to 300 MPa). When hMSCs were seeded onto those scaffolds, a significant increase in the osteogenic marker alkaline phosphatase (ALP) and increased calcium deposits were observed, even when cells were cultured in growth medium (designed to maintain multipotency). Histological analyses of mice tissue after scaffold implantation showed no apparent damage, toxicological effects, or inflammation up to 8 weeks after treatment. More importantly, CaPG scaffolds enhanced donor cells' retention and provided differentiative signals favoring bone regeneration without using growth factors to direct osteogenesis.

Li and co-workers employed graphene oxide and lysozyme films to favor bone regeneration while minimizing the possibility of infection. Precisely by depositing overlapping layers of GO and lysozyme onto a chitosan base they obtained a construct not only able to support dental pulp stem cell growth and differentiation but also with improved antimicrobial activity. While GO is responsible for scaffolds stiffness and roughness, lysozyme improves the antimicrobial activity of GO by degrading the bacterial cell wall [245].

Among 3D scaffolds, Li and co-workers [246] provided an interesting proof of concept of the usage of 3D-printed alginate hydrogels as scaffolds for bone engineering. They used 3D-bioprinting to obtain gelatin-alginate scaffolds with defined porosity, then coated them with RGO. Although hydrogels are much less stiff than other composites, the authors observed a significant increase in adipose-derived stem cell (ADSC) differentiation toward the osteogenic lineage, as proven by the increase in ALP expression and calcification of the substrate.

Graphene oxide osteogenic potential was further investigated by Wu and co-workers, which grafted it with a peptide derived from the bone morphogenetic protein 2 [247]. GO-BMP2 was then bonded to silk-fibroin electrospun fibers to obtain biocompatible scaffolds that favor MSC adhesion and differentiation in vitro and in vivo and is are able to repair mice bone defects in less than 14 days.

## *5.2. Muscle Regeneration*

Skeletal muscles made up most of the mass of the human body and are essential for motion and support. They are composed of multinucleated myofibers, which developed from mononucleated stem cell precursors during embryonic development. Satellite cells are unipotent stem cells that remain associated with adult myofibers and are responsible for muscle growth and regeneration [248,249]. Because of them, muscle tissue is endowed with a remarkable regenerative capacity, and most injuries sustained during everyday life fully recover via well-characterized processes [250]. However, severe injuries such as volumetric muscle loss and neuromuscular degenerative diseases, or aging, can result in significant muscular impairment, severely dampening life quality. In recent years, the possibility to produce scaffold recapitulating features of adult muscle tissue to enhance regeneration has drawn much attention, and several features that can enhance muscle regeneration have been identified. Among those, Gilbert and co-workers found substrate elasticity to be pivotal for muscle regeneration, as substrates mimicking tissue elasticity (~12 kPa) were able to sustain muscle stem cell self-renewal in vitro and differentiation in vivo [251]. Moreover, it was found that electroconductive scaffolds can enhance myoblasts fusion into myotubes in vitro, possibly by mimicking neuromuscular activity [252,253]. Starting from the observation that scaffold elastic properties are pivotal to resist the dynamic condition of the muscle tissue environment, Jo and co-workers [254] developed polyurethane/graphene oxide nanocomposite fibrous scaffolds to form a flexible and myogenic stimulating matrix for tissue engineering. They found nanocomposite to have better tensile strength, hydrophilicity, and biocompatibility than pristine materials. When they seeded mouse skeletal muscle cells C2C12 (a standard model for muscle differentiation studies) onto their scaffolds, they found an enhancement in cell adhesion and spreading, as demonstrated by the increase in the expression of actin and vinculin. Scaffolds were also capable of inducing muscle differentiation, as immunocytochemistry against myosin heavy chain (MHC, a marker for mature muscle cells) and RT-PCR against MyoG, α-actinin, and MyoD (markers for differentiating muscle cells) showed an increase directly proportional to GO concentration. Most importantly, they also found that scaffolds were able to sustain dynamic tensional stimuli, which, in turn, further increased the expression of differentiative markers.

As muscle cells are aligned along the fiber axis, materials patterned with surface features resembling native extracellular environment can influence mechanotransduction and favor cell differentiation. Park and co-workers [255] employed femtosecond laser ablation (FLA) to produce GO and RGO-based micropatterned conductive PAAm-hydrogels, which can support muscle differentiation in vitro and proved to have good stability in vivo. All scaffolds resembled muscle tissue Young's modulus, but only rGO-based ones possessed enough conductivity to deliver signals to cells. FLA allowed them to pattern scaffolds with 20 μm wide, 10 μm deep canals, and only scaffolds with a pattern distance comparable to cell dimension (50–80 μm) proved to affect differentiation. Specifically, when fusion index (i.e., the ratio of nuclei inside myotubes to all nuclei) and nuclear shape (which becomes less rounded during differentiation) were considered, it was found they could be improved by 50 and 80 μm patterned scaffolds independently on their conductivity. Morphological analyses were confirmed by immunocytochemical and qRT-PCR analyses, which demonstrated an increase in the expression of differentiative and mature myoblast markers (i.e., MHC, MyoG, and MyoD). In spite of this, conductivity proved to be pivotal to enhance

cell aspect ratio, and electrical stimulation (2V, 10ms duration, 1 Hz) enhanced myotube formation with respect to untreated control. Hydrogels were also found to be suitable for implantation, as they remained intact for 4 weeks after subcutaneous implantation in mice, proving they can be a good platform for tissue implantation.

Besides skeletal muscle, cardiac muscle regeneration has drawn much attention because of the severity of heart diseases. Cardiomyocytes are specialized muscle cells which have a crucial role in the propagation of electric signal throughout the heart. Unlike skeletal muscle cells, cardiomyocytes have a reduced regenerative potential, and, after damage, are often replaced by scar tissue, which may lead to pathological heart failure. In an elegant comparative study, Lee and co-workers [256] compared the effects of gelatin methacrylic (GelMA) functionalized with either CNTs, GO, or RGO on the structural organization and functionality of rat primary cardiomyocytes. Even though all scaffolds resembled the elastic modulus of the heart, GO functionalized scaffolds exhibited low conductivity. Moreover, GO and RGO functionalized scaffolds displayed higher surface roughness compared to the GelMA and CNT-GelMA ones. Despite those differences, all scaffolds proved to support cell attachment and proliferation; however, they had different effects on cell differentiation. Specifically, when cells were stained against Cx43 (indicating electrical and metabolic coupling between cells), troponin-I and sarcomeric α-actin (both involved in muscle contraction), where enhanced only on RGO and CNT-GelMA but not in GO-GelMa. Moreover, even RGO failed to enhance the expression of troponin-I. Cells were further analyzed by patch-clamp to determine the extent and shape of the membrane action potential. Based on results, they found that CNT-GelMA led to the formation of ventricular like cardiomyocytes, whereas GO-GelMA resulted in an atrial-like phenotype. Instead, RGO-GelMA led to cells with a mixed phenotype. This finding suggests that different properties of the graphene derivative in the scaffold can be exploited to fine-tune cardiomyocyte phenotype.

In the context of injectable gels, Choe et al. developed an RGO-modified alginate gel and studied its antioxidant activity for cardiac tissue repair post myocardial infarction (MI) [257]. One of the hallmarks of MI is the high oxidative stress of heart tissues due to the formation of reactive oxygen species. Mesenchymal stem cell transplantation is a promising treatment for repairing heart tissues post MI, but after transplantation, their survival is compromise by the oxidative stress of the tissue. In their study, Choe et al. encapsulated hMSC in alginate microgels with a spherical shape (235 ± 11 μm diameter) suitable for easy injection. Nanocomposite microgels displayed higher cell viability than GO- and RGO-free beads. To further improve survival, hMSCs were first enclosed in GO/alginate hydrogels and then GO was reduced. Nanocomposite microgels showed greater scavenging activity in all assays, while the graphene-free counterpart had a negligible antioxidant activity. These injectable anti-oxidizing nanomaterial-embedded microgels were able to scavenge radicals and lower the oxidative stress post MI, support MSC viability and maturation, thus increasing therapeutic activities and regeneration of infarcted tissues.

#### *5.3. Nerve Regeneration*

The nervous system represents the most intricate and vulnerable system in the human body, as, despite its pivotal importance, it is substantially unable to regenerate itself after injury. Because of its vital role, its organization is extraordinarily complex. Briefly, the nervous system comprises two main classes of cells: the glial cells and neurons. Neurons act as functional units, as they are characterized by peculiar electrophysiological features which allow them to rapidly transmit information between each other. Connections are established during neuritogenesis by the sprouting of dendrites and axons from the cell body. Specifically, each growing axon is tipped by the growth cone, a complex molecular machinery that senses environmental stimuli to guide growth toward the proper target [258]. On the other hand, glial cells consist of various specialized cell types (including Schwann cells, oligodendrocytes, and astrocytes) that regulate homeostasis, form myelin sheets around axons, and provide support and protection for neurons by maintaining a proper microenvironment [259]. Anatomically, the nervous system has been divided into the central (CNS) and peripheral (PNS) nervous system. Besides their different physiological role, they also respond differently to damages. Central nervous system regeneration is made more challenging, mainly because adult CNS is naturally resilient to cell repair and differentiation. For instance, after axotomy, glial cells of the CNS secrete inhibitory cues and form a physical and chemical barrier, the glial scar, which prevents regenerating axons to cross the injury site and reach their new target. Moreover, the basal expression of anti-regenerative cues such as chondroitin sulfate proteoglycans, Nogo-A, and myelin-associated glycoproteins, semaphorin 4D, and ephrin, is upregulated, further suppressing the capacity of the axonal growth cone to elongate [260–262]. Conversely, PNS neurons are endowed with a higher regenerative capacity due to the lack of CNS inhibitory factors [263].

Because of the inability of central neurons to regenerate, traumatic brain injury, and spinal cord injury have profound adverse effects on life quality and are a significant cause of mortality [264]. Efforts from the scientific community to address this issue resulted in several pharmacological and surgical therapeutic strategies [265]. However, in recent years graphene and its derivatives emerged as intriguing tools to design biomaterials mimicking tissue properties, encapsulate biomolecules and favor stem cell differentiation or tissue regeneration [266–269]. Recently, Quian and co-workers [266] used 3D printing and layerby-layer casting methods to produce graphene and polycaprolactone scaffolds, which improved axonal regrowth and remyelination. Their technique allowed them to optimize quality control, mechanical strength, drug delivery distribution, and achieve the ideal electric conductivity for nerve growth. To increase scaffold biocompatibility, they coated it with polydopamine (PDA) and arginylglycylaspartic acid (RGD), which can encapsulate small molecules and favor cell adhesion, respectively. When tested with rat-immortalized Schwann cells, they found the optimal proliferation and viability rates on scaffolds at 1% graphene in PCL and that those scaffolds were able to support cell proliferation or up to seven days. Moreover, they found a higher expression of vinculin and *N*-cadherin on PDA/RGD-G/PCL scaffolds rather than on control scaffolds, indicating that graphene can have a role in promoting cell adhesion. Western blotting and qRT-PCR analyses indicated that not only scaffolds were able to induce expression of neural markers (such as glial fibrillary acidic protein, Class III ß-tubulin, and S100) but also they increase the expression of neurotrophic factors (NGF, BDNF, GDNF, and CNTF), which are vital to establishing a permissive environment for nerve regeneration. Moreover, when Schwann cell-loaded PDA/RGD-G/PCL scaffolds were grafted onto Sprague Dawley rats, histological and immunohistochemical observations 18 weeks after surgery suggested that regenerated nerves were well organized, lacked scar tissue and, most importantly, functional recovery was comparable to autograft implants.

As neurons require network formation to acquire proper function, tools to build 3D neuronal networks are required to enhance their function. An elegant method to encapsulate neurons onto a self-assembled micro-roll made of a bilayer of graphene and parylene-C [268], provided a proof of concept for designing a 3D neuronal network, which might also serve as a platform for modeling neurodegenerative diseases or producing cells suitable for transplantation. Their approach allowed them to create a support that allows neurons to interact with their surroundings without mixing with the external population, thus keeping a precisely controlled cell distribution. They exploited a sacrificial layer of calcium alginate to support a graphene layer, which was then coated with a parylene-C layer. Finally, the bilayer was patterned with an array of microscale pores to allow axons, but not cell bodies, to contact surrounding cells. Self-assembly into a tubular structure was induced by treating the sandwich with ethylenediaminetetraacetic acid (EDTA) to de-polymerize the alginate layer. Accessibility of reagents to the internal of the micro-roll was assessed by Ca2+ imaging in response to the addition of glutamate: encapsulated hippocampal neurons showed a coherent and coordinated response, and no delay with the response of external neurons was observed. Moreover, the formation of functional synapses between neurons was demonstrated both by immunocytochemistry against

synapsin I, which is expressed by neurons at the synapse puncta, and by monitoring the synchronization of spontaneous Ca2+ waves. Besides serving as support for cell growth, the authors claim the graphene in their scaffold might serve as an electrode for electrophysiological recording and neuronal activity stimulation.

In order to study the role of substrate conductivity in neuronal network formation and alignment, Wang and colleagues developed a 3D conductive GO-coated scaffold based on printed PLCL microfibers using a near-field electrostatic printing (NFEP) [270]. NFEP is a technique that combines electrostatic spinning and 3D printing that allows to obtain fiber sizes of a few micrometers and complex architectures [271]. By manipulating the motion of the collection surface along X-Y-Z axes, NEFP easily generates arbitrary patterns (2D or 3D). PLCL scaffolds with different fiber overlay angles, diameters, and spatial organization were coated with GO, which was then reduced to RGO in situ without damaging the architecture. Depending on the layer thickness, RGO coating improved electrical conductivity while increasing surface roughness. The scaffolds were then used to assess the correlation between electrical stimulation (ES) and neurite outgrowth of the pheochromocytoma-derived PC12 cell line and primary neurons from hippocampal tissue of embryonic mice. ES enhanced neurite outgrowth and alignment with respect to control without ES stimulation. Strikingly, while neurite outgrowth resulted in being strictly correlated with the strength of the electric field, its directionality did not seem to influence neurite alignment. However, it was found that neurite outgrowth tightly followed the orientation of the smaller microfiber pattern and a more dispersive distribution of neurites was observed on fibers with higher diameters, where neurites had a higher tendency to branch out and lose their directional orientation (Figure 9).

**Figure 9.** A diagram of the key procedures toward neuronal-like network formation with the guidance of conductive microfiber patterns under electrical stimulation [272].

Glial cells are as necessary as neurons to ensure proper nervous system functionality, therefore their regeneration after injury is as crucial as neuronal restoration [272]. Specifically, oligodendrocytes are responsible for myelination of central neurons and must be restored to ensure proper neuronal connectivity. The most common way to obtain oligodendrocytes is to differentiate multipotent NSCs or induced pluripotent stem cells (iPSCs). However, the process has proven to be challenging, as it requires long culture

periods (up to 150 days) and has a limited yield [273]. Shah and co-workers [267] developed a nanocomposite PCL-GO scaffold, which allowed for NSCs differentiation into oligodendrocytes in just 6 days of culture. They treated electrospun PCL nanofibers of 200–300 nm diameter with oxygen plasma to render their surface hydrophilic, then GO was deposited on their surface at either 0.1, 0.5, or 1 mg/mL. Finally, laminin, an ECM protein essential for adhesion, growth, and differentiation of NSCs, was used to coat scaffolds. Rat hippocampal NSCs displayed significant differences in cell morphology after just 6 days of culture. Moreover, concentrations of GO as low as 0.1 mg/mL were able to strongly enhance the expression of the myelin basic protein (MBP), a marker specific for oligodendrocyte differentiation. The absence of an effect on the expression of the neuronal marker Tubβ3 and the astrocytic marker GFAP further suggested those scaffolds were able to selectively direct differentiation toward the oligodendrocytic lineage. When they treated cells grown on PCL-GO with integrin signaling inhibitors, they observed a steep decrease in oligodendrocyte markers. This observation suggested that the GO-coating of the nanofiber scaffolds might promote differentiation through specific microenvironmental interactions that activate integrin-related intracellular signaling.

Besides rigid scaffolds, biocompatible conductive hydrogels have attracted much attention because of their ability to better reproduce the mechanical properties of host tissues. Javadi and co-workers [269] developed a biocompatible hydrogel, based on polyurethane (PU), poly(3,4-ethylenedioxythiophene) (PEDOT) doped with poly(4-styrenesulfonate) (PSS) and liquid crystal graphene oxide (LCGO). They obtained a formulation with excellent conductivity, tensile modulus, and yield strength to support neuronal stem cells differentiation toward neurons and glial cells (as proven by the increase of the neuronal marker Tubβ-3 and the astrocyte marker GFAP). The authors claim LCGO liquid crystal nature synergistically combined with the properties of PEDOT:PSS to increase hydrogel mechanical and electrical properties.

Starting from evidence that the cholinergic system is involved in several neuron protective processes, cortical plasticity, and functional recovery after brain injury, Pradhan and co-workers developed a choline-graphene oxide functionalized (CFGO) injectable hydrogel based on poly(acrylic acid). Not only their hydrogels were able to support neuronal cell growth and differentiation, but they also stabilized the actin cytoskeleton. As choline is involved in enhancing neural recovery in TBI treatment, they injected their hydrogels in mice with parietal cortex brain injuries. They found scaffolds were able to restore cortical loss in just 7 days of treatment [274].

#### *5.4. Wound Healing*

GBMs have also been employed as fillers for wound healing hydrogels. Rehman and co-workers [275] developed RGO-GelMA hydrogels which enhance migration of fibroblasts, keratinocytes and endothelial cells in vitro and favor angiogenesis, in vivo, in chicken embryos. The authors speculate that this property could be due to an increase in intracellular ROS levels caused by RGO. In another study, Li and co-workers [276] developed N-acetyl cysteine (NAC) loaded GO-collagen membranes. In this formulation, GO has been reported to enhance mechanical properties and water retention of the collagen scaffolds, whereas NAC is used to lower ROS levels in the damaged tissue. The membrane accelerated cell migration, maturation and angiogenesis, leading to rapid skin regeneration. Moreover, the expression of profibrotic factors was found to be downregulated, indicating those scaffolds could promote scarless wound healing. A common problem of wound healing hydrogel is their vulnerability to bacterial infection [277]. To solve this problem, Yan and co-workers [278] developed an Ag reduced GO sodium alginate film which not only is able to inhibit bacterial growth but also to stimulate rapid wound healing in vivo.

#### **6. Conclusions**

Graphene-based scaffolds have been proven to be versatile tools in mediating tissue regeneration, as highlighted by the examples of in vitro and in vivo applications that have been discussed in this review. However, much more effort is required from the scientific community to clarify and rationalize their mechanism of action. It is clear that different composites can be employed to obtain similar results and yet subtle changes in scaffold formulation may result in completely different results. Therefore, a systematic analysis of the effects of scaffold composition on differentiation is required in order to disentangle the role of each scaffold component on cell fate. This would allow for a direct comparison between different scaffolds and a finer rational design. Moreover, it needs to be stressed that most applications rely on qRT-PCR data to prove successful differentiation. However, this approach reveals only an average trend in cell differentiation, without detecting potentially significant and biologically important differences between cells seeded onto different spots of the scaffold. Even when immunocytochemical data are provided, most authors fail to mention whether or not those data are representative of the whole sample or are just isolate cases. Coupling scaffold engineering with single cell RNA sequencing would overcome those limitations, allowing both a better understanding of scaffold effects on cell physiology and comparisons between the in vitro culture and the in vivo reference.

To date, scaffold engineering has focused on regeneration of a single tissue. However, clinical applications often require grafting of whole organs. Obtaining a scaffold that is able to efficiently reproduce a whole organ, or even multiple tissues (e.g., innervate muscles), has proven challenging and further studies are required before any viable clinical usage. In our opinion, finding the rationale behind graphene regulation of cell fates will allow us to obtain scaffolds that can reliably support and differentiate cells in a number of pre-determined types at the same time.

Although a molecular rationale for graphene-mediated effects is still lacking, it is remarkable that graphene-based scaffolds are able to determine cell fate more rapidly and efficiently than any other differentiation protocol, even without the addition of exogenous pro-differentiation factors. Indeed, in recent years research on mechanotransduction has unveiled several details on how nanotopography and stiffness stimuli are perceived and transduced by cells, whereas further efforts are needed to elucidate the contribution of other chemical and physical stimuli from the tissue environment. In particular, conductivity is of special interest to neuronal cell fate and differentiation, as it is specific to the nervous system. In addition to the aforementioned nanotopographic features, graphene and GBMs are endowed with tunable conductivity. Therefore, graphene-based nanomaterials represent a useful and cost-effective tool to enhance neuronal differentiation and tissue repair.

**Author Contributions:** All authors contributed to writing manuscript parts, to the overall draft preparation, review and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding. The APC was funded by F.F.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Reduced Graphene Oxide Inserted into PEDOT:PSS Layer to Enhance the Electrical Behaviour of Light-Emitting Diodes**

**Fernando Rodríguez-Mas \*, Juan Carlos Ferrer, José Luis Alonso, Susana Fernández de Ávila and David Valiente**

Communications Engineering Department, Universidad Miguel Hernández, 03202 Elche, Spain; jc.ferrer@umh.es (J.C.F.); j.l.alonso@umh.es (J.L.A.); s.fdezavila@umh.es (S.F.d.Á.); dvaliente@umh.es (D.V.) **\*** Correspondence: fernando.rodriguezm@umh.es

**Abstract:** In this study, poly(9-vinylcarbazole) (PVK)-based LEDs doped with reduced graphene oxide (rGO) and cadmium sulphide (CdS) nanocrystals were fabricated by spin-coating. The hybrid LED structure was a layer sequence of glass/indium tin oxide (ITO)/PEDOT:PSS|rGO/PVK/Al. rGO was included in the poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS) layer due to its energy bands being close to PEDOT:PSS bands, and the possibility of using water for dispersing both polymer and flakes. Optical properties such as photoluminescence and UV-Vis absorption were not affected by the addition of rGO to the PEDOT:PSS solution. However, PVKbased LEDs with rGO showed increased current density compared to those without rGO in the hole transporting layer. Higher electroluminescence intensities were observed for rGO-enriched LEDs, although the shape of the spectrum was not modified. LEDs including CdS nanocrystals in the poly(9-vinylcarbazole) emissive layer did not show such dependence on the rGO presence. Though the addition of rGO to PEDOT:PSS still produces a slightly higher current density in CdS doped LEDs, this growth is no longer proportional to the rGO load.

**Keywords:** cadmium sulphide; PVK; hybrid light-emitting device; electroluminescence; nanocrystals; reduced graphene oxide

## **1. Introduction**

Organic-semiconductor devices have gained popularity in recent years due to several advantages they offer compared to their inorganic counterparts, namely mechanical flexibility, low cost [1], and a simple manufacturing processes [2,3]. Deposition of organic polymer layers can be achieved by means of simple techniques such as spin-coating. With this technique, uniform layers are applied onto a flexible or rigid substrate by means of centrifugal force. In these devices, specifically in light-emitting diodes (LEDs), the optical and electrical properties could be modified by the inclusion of different dopants, such as graphene and its derivatives, or semiconductor nanocrystals.

Since 2004 [4], graphene and its derivatives have generated high expectations in several areas, including as solar cells [5], field-effect transistors [6], electrocatalyst [7], transparent electrodes [8], graphene transistors [9], etc. Graphene has excellent properties, such as excellent thermal conductivity [10], electron mobility [11] and transparency [12], that make possible its involvement in the aforementioned areas. The simplest process for synthesizing graphene consists in the oxidation and exfoliation of graphite. This process transforms graphite into graphene oxide (GO). In a later step, GO is reduced, and either graphene or reduced graphene oxide (rGO) is synthesized depending on the reduction degree.

Not all the properties of graphene are extensible to graphene derivatives (GO or rGO). In the case of reduced graphene oxide, certain properties are maintained, but to a lesser extent, such as a conductivity of 200 S/m1. In this paper, we will study the possibility of manufacturing hybrid LEDs with graphene derivatives. Normally, in these devices, the

**Citation:** Rodríguez-Mas, F.; Ferrer, J.C.; Alonso, J.L.; Fernández de Ávila, S.; Valiente, D. Reduced Graphene Oxide Inserted into PEDOT:PSS Layer to Enhance the Electrical Behaviour of Light-Emitting Diodes. *Nanomaterials* **2021**, *11*, 645. https://doi.org/ 10.3390/nano11030645

Academic Editor: José Miguel González-Domínguez

Received: 12 February 2021 Accepted: 2 March 2021 Published: 5 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

hole transport layer is formed by Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS), but in our analysis, this layer will be doped with different proportions of rGO. Reduced graphene oxide was chosen because the value of its energy bands (−5.2~−4.9 eV) [13] is very close to PEDOT:PSS, (−5.1~−5 eV) [13]. Additionally, rGO is aqueous dispersible, like the PEDOT:PSS, eliminating the possible problems that a mixture of different solvents might cause in the spin-coating technique [14].

On the other hand, Cadmium sulphide (CdS) nanocrystals (NCs) have been studied in recent years [15] due to their ability to modify the optical and electrical properties of the organic devices [16]. Cadmium sulphide nanoparticles have been used to improve or modify the properties of different devices, such as organic solar cells [17], light-emitting devices [18], hybrid memory devices as a compound of polymer composites and CdS nanocrystals [19,20], and memory devices with other nanoparticles, such as graphene oxide nanocrystals [21]. CdS is a semiconductor with a direct bandgap energy of 2.4 eV. This bandgap energy makes it suitable for applications in the visible band of the electromagnetic spectrum. A simple way to synthesize CdS nanocrystals could be through the process of thiolate decomposition [22]. This route is based on a two-step process: (i) synthesis of a thiolate (compound formed by the future surface ligand and the precursor metal of the nanoparticle) and (ii) reaction with a sulphur source, originating the nanocrystals. It should not be overlooked that a critical factor of nanocrystal behaviour is their size [23,24], since the optical properties depend on this parameter. To control the size and solubility, the nanocrystals are coated with ligands [25]. In this route, the nanocrystals are coated with thiophenol. The presence of thiophenol makes them soluble in dimethyl sulfoxide (DMSO). However, since the control of the exact mass of nanoparticles in solution is imprecise with this method, they were dried following reference [14]. Moreover, thiophenol was chosen as a ligand because thiophenol is formed by an aromatic ring [26]. This aromatic ring is expected to improve the charge transfer between the nanocrystals and the surrounding polymer because of the ring resonance. The influence of rGO on devices with their emission layer doped with CdS was also studied.

The aim of this paper is to analyse the influence of rGO and CdS nanoparticles embedded in PVK polymer on the electrical and optical properties of hybrid LEDs with active layers based on these materials.

## **2. Materials and Methods**

## *2.1. Materials*

Cadmium nitrate-tetrahydrate (Cd(NO3)2·4H2O, 99.99%), thiophenol (99%), sulphur powder (99.98%), poly(9-vinylcarbazole) (PVK, 98%), poly(3,4-ethylenedi-oxythiophene): poly(styrene sulfonate) (PEDOT:PSS, 1.3% water solution), toluene, methanol and dimethyl sulfoxide (DMSO) were purchased from Sigma-Aldrich (Darmstadt, Germany) and used without further purification. Reduced graphene oxide (rGO) was obtained from Graphenano "nanotechnologies" (Yecla, Spain) and used without further purification.

## *2.2. Characterization*

The measurements of optical absorption were carried out with a T92+ UV/VIS spectrophotometer from PG instruments Ltd. (Lutterworth, UK), and the measurements of photoluminescence (PL) were performed with a Modular Spectrofluorometer Fluorolog-3 from Horiba Scientific (Madrid, Spain). In all the photoluminescence measurements, the excitation wavelength was fixed at λexc = 365 nm.

Transmission electron microscopy (TEM) analysis was performed using a Jeol 2010 (Tokyo, Japan) operating at 200 kV. High-resolution TEM (HRTEM) images were obtained by phase contrast at Scherzer defocus in order to obtain easily interpretable images.

Current density vs. voltage (J–V) curves of the LEDs were measured using Keithley 2400 Sourcemeter equipment (Bracknell, UK).

Electroluminescence (EL) characterization was performed with a Triax 190 monochromator (Madrid, Spain) and a multichannel thermoelectrically cooled CCD Symphony detector by Horiba Jobin Yvon (Madrid, Spain).

#### *2.3. Synthetic Pathway of PEDOT:PSS|rGo Dispersions*

A rGO solution was prepared and mixed with PEDOT:PSS. A reduced graphene oxide flake solution was prepared with distilled water at 4 wt%. When the solution was completely dispersed, it was mixed with the PEDOT:PSS, 1.3% water solution, in accordance with the proportions in volume shown in Table 1. The rGO was redispersed in distilled water because PEDOT:PSS was dissolved in water, too. In this sense, possible problems resulting from the phase separation of different solvents were avoided. In the spin-coating process, the use of solution with different solvents could cause craters and agglomerations in the deposited layer, as we observed in our previous study [14].


**Table 1.** Summary of the quantities used in the PEDOT:PSS layer doped with rGO.

## *2.4. Synthesis of CdS Nanocrystals Powder*

The route used for the synthesis of CdS NCs was an extension of the thiolate decomposition method [14]. The main advantage of this method is that nanocrystals end up in a powder, making it simpler to embed them in the organic polymer. The first step in this process was the synthesis of cadmium thiolate, Cd(C6H5S)2. For this, 1.54 g of Cd(NO3)2·4H2O was dissolved in 25 mL of distilled water and 25 mL of methanol and the solution was stirred for 30 min. In another flask, 1.03 mL of thiophenol was dissolved in 50 mL of distilled water. After that, when both solutions were well dissolved, the solutions were mixed and the blend was stirred. When Reaction (1) evolved, the mixture turned a whitish colour and the cadmium thiolate precipitated. The white powder was filtered and dried obtaining the cadmium thiolate.

$$\text{Cd(NO}\_3\text{)}\_2\text{-}4\text{H}\_2\text{O} + \text{C}\_6\text{H}\_5\text{-SH} \rightarrow \text{Cd(C}\_6\text{H}\_5\text{S})\_2\downarrow \tag{1}$$

In a second step, two more solutions were prepared, 0.08 g of Cd(C6H5S)2 was dissolved in 2 mL of DMSO and, in another vial, 0.17 g of sulphur was dissolved in 20 mL of toluene. Both solutions were stirred for thirty minutes and when the thiolate was dissolved, 0.4 mL of sulphur solution was added. The mixture changed and turned yellowish, indicating that the CdS NCs had been synthesized correctly. The nanocrystals ended up dissolved in DMSO and it was necessary to eliminate the solvent in order to avoid phase separation when mixed with the polymer solvent. To achieve this, the yellowish solution was heated to 200 ◦C for one hour. Once the solvent was evaporated, CdS NCs were obtained.

## *2.5. Hybrid LEDs Fabrication*

Hybrid light-emitting diodes with a layer sequence of ITO/PEDOT:PSS|rGO/PVK: CdS/Al were fabricated by spin-coating. Commercial glass substrates covered with a semitransparent ITO layer were routinely cleaned by sequential sonication in 1,2,4 trichlorobenzene, acetone and isopropyl alcohol, and then dried with N2.

Aqueous PEDOT-PSS dispersion with rGO was spin-coated onto the clean ITO surface and then annealed at 100 ◦C for 60 min. Then, the active layers were spin-coated and dried at 80 ◦C for 60 min.

Finally, the metallization of the cathodes was performed by evaporating aluminium in a high-vacuum chamber (10−<sup>6</sup> mbar) until a thickness of 200 nm was achieved.

#### **3. Results**

## *3.1. CdS NCs Characterization*

Absorbance and photoluminescence measurements were performed. The PL spectra are shown in Figure 1A. Therein, a narrow peak is observed at low wavelengths, the maximum of the narrow peak had a wavelength at 411 nm. This peak corresponded with the maximum photoluminescence peak of toluene (blue line). The solvent peaks are not usually observed in PL graphs, because the emission intensity of nanoparticles normally hides the emission intensity of solvents. In this case, toluene was present in the CdS curve because the concentration of measured solution was very low. This low concentration made the emission intensity of nanoparticles lower. The low concentration can also explain the noise that was observed in the CdS photoluminescence. In this PL spectrum, the main peak shows its maximum at 576 nm.

**Figure 1.** Photoluminescence spectra (**A**) of CdS nanocrystals (black line) and toluene (blue line). Optical absorption spectrum (**B**) of CdS NCs (black line).

To correctly observe the absorption edge and calculate the bandgap energy, the Tauc relation [27] was applied. In Figure 1B, (αhν) <sup>2</sup> versus hν are plotted, showing an excitonic shoulder, pointing to the absorption edge (dotted line). The resulting value of the band gap is 3.07 eV (403.7 nm). According to this, the wavelength corresponding to the maximum intensity at the CdS NCs photoluminescence spectrum (Figure 1A) should be at lower wavelengths, close to the absorption edge. This shift towards higher wavelengths is due to defects in the nanocrystal surface. The defects originate deep surface states, producing trap emission.

To characterize the size of the nanoparticles, two different methods were used, a theoretical method, whose calculations are based on the band gap energy, and the direct measurement from TEM images. As a theoretical method, the equation suggested by Brus was employed [28].

$$E\_n = E\_b + (\hbar^2 \pi^2 / 2R^2) \times (1/m\_e \text{\*} + 1/m\_h \text{\*}) - 1.8e^2 / (4\pi\varepsilon\_0 \varepsilon R),\tag{2}$$

where *En* is the nanoparticles band gap, *Eb* is the energy gap of the bulk material, *me*\* and *mh*\* are the effective masses of electrons and holes, and ε is the dielectric constant. Solving the Brus equation (Equation (2)), these nanoparticles had an average size of 3.11 nm.

To prepare samples for TEM analysis, a drop of CdS NCs solution in toluene was deposited on a carbon grid and then dried at room temperature. A high-resolution image of the CdS nanocrystals is shown in Figure 2, as well as the histogram of the size distribution of more than 50 particles. Diameters range from 2.2 nm to 4.4 nm, and the average size is 3.20 ± 0.06 nm. When comparing the size calculated from the Brus equation (Table 2) to the image measurement, a minimal difference can be observed (0.09 nm) that can be neglected. It was verified that the Brus equation performs a good approximation of the real size.

**Figure 2.** TEM images corresponding to CdS NCs and histogram of the size distribution for the nanocrystals.

**Table 2.** Summary of the CdS NCs characterization. Absorption edge, PL peak and size.


When measuring the interplanar distances of the CdS crystals in the HRTEM images, two different types of planes arise. The planes match the distances of the cubic zinc-blendetype structure and the hexagonal wurzite-type structure. In Figure 2, the {100} family planes of the cubic structure and the {011} family planes of the hexagonal phase are identified in two particles. The presence of both crystalline structures can justify the dispersion in the histogram of Figure 2. In addition, in Figure 1B, a second shoulder can be located at lower energy values. The dispersion in the histogram and the presence of the two shoulders in UV-Vis absorption, Figure 1B, confirm the two different structures [14]. The heat used to evaporate the solvent changes the NC structure. The energy could be enough to increase the size and transform the nanoparticles with cubic zinc blend structures into hexagonal wurzite-type structures [14,29,30].

## *3.2. Synthesis and Characterization of the Hybrid Solution*

As discussed in the introduction, the active layer of the devices was doped with CdS NCs. The active layer was prepared from a solution of PVK mixed with CdS nanocrystals. CdS nanoparticle powder was weighed and added to a solution formed by PVK dissolved in toluene. This solution had a PVK:CdS mass ratio of 8:1 at 3 wt%. Moreover, another solution with pristine PVK was also prepared with a concentration of 3 wt%. Optical absorption and photoluminescence measurements of these solutions were performed with the results shown in Figure 3.

**Figure 3.** Photoluminescence (**A**) and absorption spectra (**B**) from pristine PVK and PVK doped with CdS NCs and photoluminescence (**C**) and absorption spectra (**D**) from PEDOT:PSS doped with different proportions of reduced graphene oxide.

The PL curves of pristine PVK and PVK with CdS NCs are plotted in Figure 3A. Both spectra exhibited a main peak corresponding to the PVK polymer located around 385 nm not shown in the figure. The influence of the nanoparticles was clearly visible in the blue line. The maximum of the secondary peak of the PVK solution with CdS NCs was localized at 577 nm. The same wavelength value that was presented in Figure 1A, where the CdS nanocrystals were measured without PVK. The photoluminescence of hybrid PVK shows a normal decrease in the PVK emission because the content of PVK polymer was reduced in the hybrid solution. Additionally, PVK photoluminescence quenching was observed [31]. In the hybrid solution, the excited electrons of the highest molecular orbital (HOMO) of the PVK polymer can either drift to the polymer's lowest unoccupied molecular orbital (LUMO) or go to the conduction band of CdS. This migration produces a charge transfer, lowering the PVK intensity [13]. As for the optical absorption, a constant background absorption was observed in the total range of wavelengths (Figure 3B) due to the high concentration of CdS NCs.

To study the influence of reduced graphene oxide on the devices, rGO was included in several hybrid LEDs with different proportions. For this purpose, the hole transport layer, represented by the PEDOT:PSS polymer, was doped with different rGO masses.

We performed optical absorption and photoluminescence measurements to check the influence of rGO in our devices. To observe if the addition of reduced graphene oxide had any influence on the optical properties of the hole transport layer solution, absorbance and PL measurements were performed on the samples described in Table 1. The excitation wavelength for the photoluminescence measurements was λexc = 365 nm. To perform UV-Vis absorption measurements, PEDOT:PSS and PEDOT:PSS doped with rGO solutions were diluted to prevent re-absorption. Diluted solutions were measured.

No difference was observed between the PL curves of PEDOT:PSS doped with rGO and those of PEDOT:PSS, as shown in Figure 3C. There was no significant difference in optical absorption measurements, either (Figure 3D). Since the introduction of reduced graphene oxide on PEDOT:PSS did not influence the optical characteristics of PEDOT:PSS, we assume that it should not influence the optical characteristics of hybrid LEDs.

## *3.3. Hybrid LEDs with rGO*

Once the CdS nanocrystals and the different solutions were synthesized and characterized, hybrid light-emitting diodes were manufactured to check the influence of rGO on the devices. The hybrid LEDs consisted of a structure formed by the following stacked layers based on the solutions presented in the previous sections: glass/ITO/PEDOT: PSS(|rGO)/PVK(:CdS)/Al; Figure 4. Different devices were manufactured, employing PE-DOT:PSS as a hole transport layer and PVK as the active layer. Excluding the changes in the concentrations of graphene and CdS nanoparticles, the same conditions were maintained for the fabricating process of LEDs.

**Figure 4.** Structure of the different devices fabricated. ITO/PEDOT:PSS/PVK/Al (**A**), ITO/PEDOT:PSS|rGO/PVK/Al (**B**), ITO/PEDOT:PSS/PVK:CdS/Al (**C**), and ITO/PEDOT:PSS|rGO/PVK:CdS/Al (**D**).

Polymer and hybrid layers were deposited using the spin-coating technique. The hole transport layers were spin-coated at 2000 rpm on the substrate. When the hole transport layers were dried, the emission layers were deposited at 4000 rpm by spin-coating.

Two types of hybrid LEDs were produced, without or with CdS NCs embedded in the PVK active layer. In the first one, the solutions indicated in Table 1 were employed to spin cast the hole transport layers, and pristine PVK was used as the emissive layer. An LED without reduced graphene oxide was manufactured as a reference. The emission layer was composed of PVK at 3% wt. When the devices were fabricated, we collected the measurements of current density vs. voltage (J-V). The curves are shown in Figure 5A.

Hybrid LEDs doped with rGO exhibited changes in electrical properties. The current density increases with the inclusion of rGO. The maximum improvement was found for the hybrid LED with [PEDOT:PSS|rGO] ≡ [5:1], and this decreases with the ratio of rGO. The current density values at the threshold voltage are shown in Figure 5C. This current increases with rGO concentration. On the other hand, the threshold voltage is reduced by the presence of rGO in the PEDOT:PSS layer.

The variation of the electrical behaviour in the hybrid LEDs indicated that the rGO inclusion in the PEDOT:PSS layer improved the hole transport [32]. rGO is a two-dimensional structure with graphene distributions. In graphene distributions, the charge carriers tend to move. The reduction by means of which graphene transforms into reduced graphene oxide eliminates part of the graphene conducting regions present in rGO [33]. In these distributions, the π-bond (C=C) and σ-bond (C-H) alternate, generating optimal paths for

the hole transport. With the increase of reduced graphene oxide, the number of graphene distributions increases in the layer, augmenting the optimal paths for transport.

**Figure 5.** J-V curves (**A**), threshold voltage (**B**), and current density (**C**) at threshold voltage for the hybrid LEDs with rGO in the PEDOT:PSS layer.

Next, we studied the electroluminescence of the hybrid LEDs doped with rGO. Regarding the emission intensity, LEDs doped with reduced graphene oxide showed higher emission intensity than reference PVK-LEDs, as indicated in [34,35].

In Figure 6, the spectra were normalized to the maximum peak emission. All EL curves showed a similar morphology, with a narrow uppermost peak and two shoulders at longer wavelengths. To study these features, Gaussian deconvolutions were carried out for all spectra, indicated as non-solid lines in Figure 6 for PEDOT: PSS + PVK. Gaussian deconvolution is a statistical process where electroluminescence is decomposed by Gaussian curves, where the mean of each Gaussian curve coincides with the wavelength of the maximum emission peak, and the standard deviation is the difference of the wavelength of the maximum emission peak (mean) and the wavelength at which a - <sup>1</sup> <sup>−</sup> 1/√*<sup>e</sup>* % decrease from the maximum emission peak is produced. We allowed three Gaussian curves to fit each spectrum. The wavelengths corresponding to the maximum of each Gaussian emission peak are detailed in Table 3.

**Table 3.** Positions of emission peaks for the hybrid LEDs with rGO and PVK LED.


**Figure 6.** Normalized electroluminescence of PVK LED and hybrid LEDs using rGO in their hole transport layer. PEDOT:PSS + PVK (black), PEDOT:PSS|rGO [30:1] + PVK (red), PEDOT:PSS|rGO [15:1] + PVK (blue) and PEDOT:PSS|rGO [5:1] + PVK (green). Gaussian deconvolution of PEDOT:PSS + PVK (non-solid lines).

The EL spectra are different from the PL curves (Figure 3A). Like any polymer, PVK is a chain of monomers (a small molecule) bonded together. It is also known that photoluminescence of PVK has two peaks, one around 390 nm, produced by a small superposition of monomers, called p-PVK, because it is related to the phosphorescence of PVK. Another contribution is located around 410 nm, and is produced by the total superposition of the monomers, known as f-PVK, due to its relation to fluorescence of PVK [36]. In the EL curve of the PVK-reference, the peaks are not in the same positions as those observed for photoluminescence. The PVK electroluminescence presents a shift towards higher wavelengths. According to Ye et al. [36], with the increase of temperature, the emission intensity of p-PVK decreases and the emission intensity of the f-PVK increases and, at low temperatures, the PVK photoluminescence exhibits a broad peak at high wavelengths (~550 nm). This peak corresponds to the radiative transition of triplet states. In addition, Ye affirms that the PVK is a polar polymer, and that an electric current polarizes PVK. The polarization of PVK enhances the effect of f-PVK, since it reduces the intermolecular distance, and hence, the energy. Therefore, the peak of f-PVK is shifted towards higher wavelengths.

In PVK-reference, the maximum emission occurs for the narrowest peak, located at 428 nm, and the two shoulders that showed the EL curve were at 495 and 595 nm. The peak at 428 nm corresponds to the peak f-PVK enhanced by the electric field at room temperature. The two shoulders did not have any correspondence with the photoluminescence. Ye [36] stated that, for EL processes, the proportion of triplet states is increased compared to PL processes; this enables the observation of phosphorescent emissions at room temperature. This is the origin of the peak located at 495 nm. Finally, the lower shoulder observed in the EL spectrum close to a wavelength of 600 nm is attributed to the electromer of PVK. This emission is only visible in EL, not in PL, because photoexcitation does not normally generate the free charge carriers needed to form this complex.

For LEDs doped with rGO, the position of the f-PVK around 425 nm was not significantly modified. Additionally, the two shoulders that showed the EL curves of hybrid LEDs did not exhibit variations in wavelength with respect to the PVK-reference. Therefore, the rGO modified the emission intensity of the EL spectra, but this did not change the shape of the EL curves.

To verify that rGO does not modify the electroluminescence spectra, we compared the CIE 1931 chromatic coordinates of the manufactured LEDs. The colour coordinates for PVK-LED were (0.28, 0.28). The use of rGO did not qualitatively modify the chromatic coordinates. These were (0.28, 0.29) for PEDOT:PSS|rGO [30:1] + PVK; (0.27, 0.27) for PEDOT:PSS|rGO [15:1] + PVK and (0.28, 0.27) for PEDOT:PSS|rGO [5:1] + PVK.

## *3.4. Hybrid LEDs with rGO and CdS NCs*

Having determined the electrical behaviour due to the inclusion of rGO, we studied the rGO influence on LEDs doped with CdS nanocrystals. In Figure 7A, the J-V curves are plotted. In these curves, it can be observed that at the same voltage values, the current density is higher in the doped LEDs (with any dopant, rGO or CdS NCs) than in the pristine PVK LED. In devices without rGO, the CdS inclusion considerably increased the electrical conduction. Since in PVK LED, the PVK lowest unoccupied molecular orbital (LUMO) is −2.3 eV, and for aluminium it is −4.3 eV, these values cause a potential barrier, the value of which is 2 eV. The inclusion of CdS nanocrystals in the active layer generates distributions where the potential barrier decreases, favouring the electron transport. The conduction band of CdS NCs is closer to that of aluminium than that of PVK LUMO, (the conduction band of CdS NCs is around −4 eV). Therefore, some electrons will occupy the CdS conduction band, producing improvements in electrical behaviour [14]. Additionally, LEDs doped with rGO and their active layer composed of PVK with CdS nanocrystals also increased the electrical conduction of the pristine PVK LED. With the same voltage, the CdS NCs augmented the current density of the LEDs manufactured previously. This modification of electrical behaviour can be justified by changing the hole transport. However, in this case, the improvement of the hole transport is not the only factor present. To the enhancement by rGO is added the influence of the nanoparticles. CdS nanoparticles reduce the potential barrier for electron injection.

**Figure 7.** J-V curves (**A**), threshold voltage (**B**), and current density (**C**) at threshold voltage for the hybrid LEDs with rGO in the PEDOT:PSS layer and CdS NCs in the active layer.

In LEDs doped with reduced graphene oxide and CdS nanocrystals, a change was observed. These LEDs had a higher current density at threshold voltage than the reference PVK LED, Figure 7C. Additionally, the doped devices had a lower threshold voltage than the reference, Figure 7B. However, in this case, the direct relation between the rGO load and

improvement of electronic transport was eliminated by the presence of the nanoparticles, and was not observed, as indicated in Figure 7B,C. Nevertheless, hybrid LEDs with rGO in the hole transport layer show higher currents and lower threshold voltages than LEDs without rGO. We expected that the electrical conduction would modify with the rGO increase, because more areas of rGO are present, but as shown in Figure 7A, this relation was not produced. The inclusion of CdS nanocrystals produced a higher electrical increase than the inclusion of rGO. In addition, the electrical conduction was higher in LEDs with two dopants than in the devices with a single dopant, but in these devices, the relation "electrical evolution–rGO" was not perceived.

A possible hypothesis to explain this behaviour is related to the morphology of the layers, mainly the interface between the hole transport layer (HTL) and the emissive layer.

Spin-coated layers of PEDOT:PSS are quite smooth, but their roughness increases when rGO is embedded, proportionally to the ratio. If the layer deposited on top of this HTL is pristine polymer, the roughness is corrected to some grade because of the ability of this soft material to coat the underlying surface. Thus, devices with a PVK emissive layer show an influence on the electrical characteristics correlated with the rGO load. On the other hand, if the layer deposited on top of the PEDOT:PSS|rGO layer contains inorganic nanoparticles that are rigid and non-deformable, the interface between the HTL and the emissive layer will have an abrupt profile with several imperfections acting, most probably, as charge carrier traps and recombination centres.

This might be the reason for CdS NCs enriching LEDs with higher rGO load in the HTL; the electrical characteristics do not show the same trend observed for LEDs without CdS nanocrystals.

The electroluminescence of hybrid LEDs doped with rGO and CdS nanocrystals was measured. To observe more clearly the influence of nanoparticles in the EL spectra, the proportion of PVK versus CdS NCs increased to the ratio [PVK:CdS] = [2:1] by weight. Thanks to this increase, the influence of CdS was more evident.

As in devices doped exclusively with rGO, all EL curves of hybrid LED doped with rGO and CdS NCs were normalized, and the average was calculated. All the results are plotted in Figure 8. Gaussian deconvolution was also performed to locate the peak positions of the spectra.

**Figure 8.** Normalized electroluminescence of PVK LED and hybrid LEDs using rGO in their hole transport layer and doped with CdS NCs. PEDOT:PSS + PVK:CdS [2:1] (black), PEDOT:PSS|rGO [30:1] + PVK:CdS [2:1] (red), PEDOT:PSS|rGO [15:1] + PVK:CdS [2:1] (blue), and PEDOT:PSS|rGO [5:1] + PVK:CdS [2:1] (green). Gaussian deconvolution of PEDOT:PSS + PVK:CdS [2:1] (non-solid lines).

The electroluminescence of hybrid PVK LEDs presented a narrow peak at shorter wavelengths and a broad emission for longer wavelengths. We carried out Gaussian analysis (Table 4) to determine the origin of the broad peak and the composition of the emission spectra.

**Table 4.** Positions of emission peaks for the hybrid LEDs with rGO and PVK LED doped with CdS NCs. As reference, pristine PVK LED is also listed.


The narrow peak visible at lower wavelengths (Table 4) is due to the phosphorescence emission of PVK as in PVK-LED reference (428 nm) [14]. The broad peak of hybrid LEDs with CdS NCs was decomposed into two peaks. In Section 3.1, we indicated that two types of nanocrystals were present in the devices: cubic and hexagonal nanocrystals. These different structures are related to different sizes of CdS NCs, so the hybrid LEDs exhibit both contributions in their electroluminescence [14].

The addition of different amounts of rGO to hybrid LEDs with CdS NCs did not significantly shift the wavelength of the emission peaks' maximum intensity. In Figure 8, the highest peak is identified with CdS nanocrystals. In these devices, the charge carriers from the HOMO of PVK can migrate to the LUMO of PVK or to CdS. Due to this fact, the luminescence of the polymer and the nanoparticles are present in the electroluminescence. In our previous study [14], we demonstrated that with an increasing quantity of nanoparticles, the CdS luminescence is extended, because more charge carriers are able to migrate to CdS. At this point, the quantity of CdS nanoparticles is sufficient for it to account for the highest peak. As in Section 3.3, we studied the CIE 1931 colour coordinates, too. In our previous research, a shift towards the white colour was produced by the inclusion of CdS NCs in pristine PVK-LEDs [14]. The nanocrystals shifted the CIE coordinates to CdS NCs light emission [14]. Thus, we verified the relation of rGO and CdS NCs with PVK-LED, in their chromatic coordinates. The colour coordinates of PEDOT:PSS + PVK:CdS [2:1] were (0.37, 0.34). The nanocrystals produced the abovementioned shift towards the coordinates of CdS light emission (0.45, 0.50). As in the case of the previous study, the presence of rGO did not introduce qualitative variations to the chromatic coordinates of PVK-LEDs doped with CdS NCs. The CIE coordinates of all the manufactured LEDs are collected in Table 5. Apparently, hybrid LEDs with rGO [15:1] and CdS NCs are the closest to D65 white light emission.

**Table 5.** CIE 1931 colour coordinates of emission peaks for manufactured LEDs.


## **4. Conclusions**

Organic light-emitting diodes with the transport hole layer doped with reduced graphene oxide, and the active layer doped with CdS nanoparticles, were successfully fabricated.

Devices without nanoparticles showed an evolution in their electrical behaviour. This change was proportional to the amount of rGO. The rGO caused a decrease in the threshold voltage and an increase in the current density of the threshold voltage. Devices doped with rGO and CdS modified their electrical behaviour, too. However, the inclusion of CdS nanocrystals eliminated the dependence of the evolution on the amount of rGO.

Electroluminescence measurements were performed. As a result, it was demonstrated that rGO inclusion did not modify the position of the observed emissions in the spectra.

We did not find a clear influence of rGO on the electroluminescent emission in PVKbased LEDs, with or without CdS NCs. The CIE coordinates for LEDs doped with CdS NCs are quite close to white light sources; in particular, the LEDs including an intermediate load of rGO in the PEDOT:PSS layer emitted the light that was closest to the average midday light in Europe, known as D65.

**Author Contributions:** F.R.-M. designed the experiments; F.R.-M. performed the experiments; F.R.-M., J.C.F., J.L.A. and S.F.d.Á. analysed the data; F.R.-M. wrote the original paper draft; S.F.d.Á. supervised and performed critical revisions of the paper; J.C.F., J.L.A. and D.V. revised the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors gratefully acknowledge support from Graphenano "nanotechnology".

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Communication* **Drying-Time Study in Graphene Oxide**

**Talia Tene 1, Marco Guevara 2, Andrea Valarezo 3, Orlando Salguero 3, Fabian Arias Arias 4, Melvin Arias 3,5, Andrea Scarcello 3,6,7, Lorenzo S. Caputi 3,6 and Cristian Vacacela Gomez 2,3,\***


**Abstract:** Graphene oxide (GO) exhibits different properties from those found in free-standing graphene, which mainly depend on the type of defects induced by the preparation method and post-processing. Although defects in graphene oxide are widely studied, we report the effect of drying time in GO and how this modifies the presence or absence of edge-, basal-, and sp3-type defects. The effect of drying time is evaluated by Raman spectroscopy, UV-visible spectroscopy, and transmission electron microscopy (TEM). The traditional D, G, and 2D peaks are observed together with other less intense peaks called the D', D\*, D\*\*, D+G, and G+D. Remarkably, the D\* peak is activated/deactivated as a direct consequence of drying time. Furthermore, the broad region of the 2D peak is discussed as a function of its deconvoluted 2D1A, 2D2A, and D+G bands. The main peak in UV-visible absorption spectra undergoes a redshift as drying time increases. Finally, TEM measurements demonstrate the stacking of exfoliated GO sheets as the intercalated (water) molecules are removed.

**Keywords:** graphite; few-layer graphene; graphene oxide; Raman; TEM; UV-vis; Lorentzian fitting

## **1. Introduction**

Graphene, a two-dimensional carbon nanomaterial arranged in hexagonal symmetry, has already demonstrated excellent electronic, mechanical, electric, magnetic, and thermal properties [1–3] which guarantee exciting applications, including composites [4], energy storage [5], catalysis [6], field-effect transistors [7], and plasmonics [8–10]. Several strategies are currently used for preparing graphene, for instance, chemical vapor deposition (CVD) [11], epitaxial growth [12], liquid-phase exfoliation [13], shear-exfoliation [14], zeolite-shear exfoliation [15], and chemical exfoliation [16].

Among them, the liquid-phase exfoliation and chemical exfoliation strategies are the most practical methods for preparing graphene in large quantities. In particular, the chemical exfoliation uses strong oxidizing agents to produce graphite oxide [17,18], which under sonication yields graphene oxide (GO) [19]. After the oxidation process, GO is covered by different functional groups (hydroxyl, epoxide, carboxyl, and carbonyl groups [20,21]) which increase the interlayer spacing up to 0.87 nm [22]. Moreover, GO is a large-band gap material [23], limiting its use for electronic applications, but opening a

**Citation:** Tene, T.; Guevara, M.; Valarezo, A.; Salguero, O.; Arias Arias, F.; Arias, M.; Scarcello, A.; Caputi, L.S.; Vacacela Gomez, C. Drying-Time Study in Graphene Oxide. *Nanomaterials* **2021**, *11*, 1035. https://doi.org/10.3390/ nano11041035

Academic Editor: José Miguel González-Domínguez

Received: 26 March 2021 Accepted: 12 April 2021 Published: 19 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

multitude of other applications as the removal of heavy metals [24] or dyes [25] as well as GO-based hydrogels [26].

Depending on the oxidation process, e.g., Hummers [27], Marcano [28], and Chen [29], GO could present different structural defects, for example, edge-, vacancy-, and sp3-type defects [30,31]. The study and control of such defects is of vital importance before choosing any of its widespread intended applications. In this respect, Raman spectroscopy is the most used characterization technique to scrutinize the quality of the as-made GO. The Raman spectra of graphite, single-layer graphene, few-layer graphene (FLG), and GO are widely reported in the literature, characterized by three prominent peaks, namely the D band, the G band, and the 2D band [32]. Less intense peaks called the D', D\*, D\*\*, D+G, and G+D can also be observed [33].

The study of these bands is fundamental in characterizing graphene-derived materials. In this context, Kaniyoor et al. showed a study of the Raman spectrum of GO, considering seven different preparation strategies [33]. It is worth noting that the Raman spectrum of GO is significantly different from that of single-layer graphene (SLG) and must be carefully analyzed. Consequently, there is a lot of discrepancy in the literature over the Raman spectra of oxidized graphenes. Some reports show similarly intense D and G peaks with a highly broadened and low intense 2D band region [34–36]. In fact, in many published papers, the 2D band is neither shown nor discussed.

The different strategies for preparing GO present environmental issues during the preparation process. As an example, Hummers et al. [27] reported the strategy most used: The oxidation process is carried out by treating natural graphite with KMnO4 and NaNO3 in concentrated H2SO4. This procedure involves the generation of toxic gases, such as NO2 and N2O4, limiting the large-scale production of GO. Recently, we have demonstrated a scalable eco-friendly protocol by excluding NaNO3 from chemical reaction [29,37] and subjecting the resulting graphite oxide to simple purification steps to obtain GO. Therefore, an investigation of the Raman spectrum of the as-made GO is not reported yet.

In this paper, such a study is presented. Instead of giving a comparison between GO preparation methods (oxidation or reduction strategies) as reported in [33], we report the effect of drying time to obtain GO powder, by using Raman spectroscopy. This study is complemented by transmission electron microscopy (TEM) and UV-visible (UV-vis) absorption measurements. For comparison purpose, the Raman spectra of natural graphite and FLG prepared in ethanol, are also reported.

## **2. Materials and Method**

All chemicals were used as received, without further purification. Graphite powder (<150 μm, 99.99%), sulfuric acid (H2SO4, ACS reagent, 95.0–98.0%), potassium permanganate (KMnO4, ACS reagent, ≥99.0%), hydrochloric acid (HCl, ACS reagent, 37%), hydrogen peroxide (H2O2, 30%, Merk, Kenilworth, NJ, USA), and ethanol (purity ≥ 98.0%, CAS: 64–17–5) were obtained from Sigma-Aldrich (St. Louis, MO, USA).

## *2.1. Synthesis of GO*

GO was prepared as reported in our earlier paper [37]. Briefly, 3.0 g of graphite powder was added to 70 mL H2SO4 while stirring in an ice-water bath. Then, 9.0 g KMnO4 were added. The resulting mixture was transferred to an oil bath and agitated for about 0.5 h. After that, 150 mL distilled water was added, and the solution was stirred for 20 min. Additionally, 500 mL distilled water was added, followed by 15 mL H2O2 and stirred up to see a yellowish solution. The resulting graphite oxide suspension was washed with 1:10 HCl solution and distilled water eight times through centrifugation. The precipitated material was re-dispersed in water by sonication using an ultrasonic bath (Branson 2510 Ultrasonic Cleaner). The suspension was centrifugated at 1000 rpm for 0.5 h, and then, dried in a drying stove (2005142, 60 Hz, 1600 W, J.P. Selecta, Barcelona, Spain) at 80 ◦C for 0.5 h, 1 h, 3 h, 5 h, 24 h, taking 15 mL samples from the suspension. The obtained GO powder was used in subsequent characterization.

## *2.2. Synthesis of Few-Layer Graphene*

To perform the sonication process as simple as possible, 100 mg of graphite was added into 100 mL of ethanol using closed tube containers. The resulting mixture was sonicated employing an ultrasonic bath (Branson 2510 Ultrasonic Cleaner, 40 kHz, 130 W) in continuous operation. The sonication time was set to 7 h, and the resulting dispersion was centrifuged for 10 min at 1000 rpm to remove non-exfoliated graphite particles. Solvent evaporation is avoided because the sonication is made in sealed containers, and the temperature of the bath is controlled by fluxing fresh water every 0.5 h.

## *2.3. Characterization*

Raman spectra of graphite, FLG, and oxidized graphenes were obtained using a Jasco NRS-500 spectrometer with a 532 nm laser wavelength (0.3 mW, 100× objective). The surface morphologies of the samples were taken out on a transmission electron microscope (TEM, JEM 1400 Plus) operating at 80 kV. For TEM and Raman characterization, GO and FLG samples was prepared by drop casting onto formvar-coated copper grids, and glass substrates, respectively. Similarly, the GO samples subject to drying-time experiment were directly deposited on the corresponding substrates. The UV-visible measurements were recorded using a UV-vis spectrometer (Thermo Scientific, Evolution 220, Waltham, MA, USA), re-dispersing the treated GO by mid-sonication for 5 min.

## **3. Results and Discussion**

Let us stress again, the main goal of the present work is the Raman study of the effect of the drying time on graphene oxide samples. As commented, [33] showed a detailed study of the Raman spectra of GO considering different preparations processes, and the effect of the temperature was widely understood [38]. The dependency of graphene oxide layers on drying methods was also briefly reported [39]. Very recently, the study of the effect of the Raman excitation laser on GO was reported [40].

## *3.1. Raman Spectrum of Graphite*

Natural graphite has a crystal structure made up of flat graphene layers. These layers are stacked in a hexagonal honeycomb-like network, usually following the AB Bernal stacking or AA' stacking. The interatomic in-plane distance is 1.42 Å, while the out-ofplane distance between graphene layers (due to van der Waals interactions) can have values ranging from is 3.35 Å up to 3.70 Å.

The Raman spectrum of natural graphite is shown in Figure 1. The main feature is the first order spectrum displaying the E2g in-plane optical mode (commonly called G peak) at 1577 cm−<sup>1</sup> (Figure 1a). This narrow G peak appears due to the bond stretching of all pairs of sp<sup>2</sup> hybridized carbon atoms in both rings and chains [41]. The G\* peak found at 2447 cm−<sup>1</sup> is characteristic of graphitic materials. The 2D peak appears at 2720 cm−1, characterized by two bands (Figure 1b), the intense 2D2A band at 2720 cm−1, and a less intense 2D1A band at 2677 cm−1. These bands are originated due to the splitting of *π* electrons as an effect of the interaction between stacked graphene layers.

The position and shape of the 2D peak depends, mainly, on the number of layers. Indeed, the 2D peak in SLG is fitted by a single Lorentzian function, say, the 2D1A band which is located around 2680 cm−<sup>1</sup> [42]. Moreover, the intensity ratio I2D/IG in SLG is >11 while our starting graphite shows a I2D2A/IG ≈ 0.45, typical of any natural graphite.

Hence, the G peak is related to the C−C stretching mode in sp<sup>2</sup> carbon bonds and the 2D peak is a fingerprint to evaluate "qualitatively" the number of layers in the obtained graphene or graphene-derived material. The absent (or negligible intensity) of the D peak evidences a defect-free pristine graphite due to the D peak is ascribed to the basal/edge structural imperfections, corresponding with an increase in the amount of disorderly carbon and a decrease in the graphite crystal size [43].

**Figure 1.** Raman spectrum of graphite (**a**) from 1000 to 2000 cm−<sup>1</sup> and (**b**) from 2300 to 3000 cm−<sup>1</sup> recorded using 532 excitation laser. The intensity was normalized by the most intense peak and the fitting of the peaks using Lorentzian functions.

#### *3.2. Raman Spectrum of Few-Layer Graphene*

The Raman spectrum of FLG prepared in ethanol by sonication was discussed in detail in Refs. [44,45]. The three significant peaks of FLG are depicted in Figure 2, the D peak at 1339 cm−1, the G peak at 1578 cm−1, and the 2D peak at 2719 cm−1. Additionally, other less intense peaks are detected, the D\*\* peak at 1385 cm−1, the D' peak at 1615 cm−1, the D+G at 2901 cm<sup>−</sup>1, and the G+D' at 2961 cm−1.

**Figure 2.** Raman spectrum of FLG (**a**) from 1000 to 2000 cm−<sup>1</sup> and (**b**) from 2300 to 3000 cm−<sup>1</sup> recorded using 532 excitation laser. The intensity was normalized by the most intense peak and the fitting of the peaks using Lorentzian functions.

Here, the D and D' peaks are activated due to (i) the induced structural damage after the sonication process and (ii) the reduction of the lateral size of the graphene sheets. However, the D peak is narrow and significantly less intense than the G peak, suggesting edge defects (folded edge samples) rather than basal defects (vacancies or impurities). The intensity ratio ID/IG was found to be <0.5. The type of defects can be deduced analyzing the D' peak and the intensity ratio ID/ID'. In particular, the value of ID/ID' ∼3.7 is associated to edge-type defects, but basal defects cannot be ruled out completely since the D\*\* and D+G peaks are also present. Particularly, these peaks (with an appreciable intensity) have been found in highly disordered carbon samples.

Therefore, the origin of these low-intensity peaks (D\*\*, and D+G) in FLG appear as an effect of the structural damage due to the sonication time, solvent type, and sonication power, crucial parameters to be controlled to attain the scalable production of graphene dispersions. On the other hand, defects do not activate the G+D' peak because of momentum conservation restrictions [46].

Compared to graphite, the 2D peak in FLG is also characterized by two bands, the 2D1A at 2685 cm−<sup>1</sup> and the 2D2A at 2720 cm<sup>−</sup>1. Three main characteristics are observed: (i) The complete shape of the 2D peak changes, (ii) the intensity of the 2D1A band increases close to that of the 2D2A band, and (iii) the intensity ratio of I2D2A/I2D1A is 1.12 (in graphite, I2D2A/I2D1A = 2.78). All these evidences support the transformation of graphite into FLG, likely no more than 10 layers.

#### *3.3. Raman Spectrum of Non-Dried Graphene Oxide*

We now move to the focus of this communication. The Raman spectrum of GO (obviously different from that of graphite and FLG) is shown in Figure 3, and the corresponding peak position and full-width at half maximum (FWHM) are reported in Tables 1 and 2. Notice that the G+D' peak is not detected in the wavenumber window analyzed from 1000 cm−<sup>1</sup> to 3000 cm−<sup>1</sup> (Figure 3b). This peak is assumed to be shifted at higher wavenumber values (>3100 cm<sup>−</sup>1) because of the spectral weight of the D+G peak.

**Figure 3.** Raman spectra of GO (**a**) from 1000 to 2000 cm−<sup>1</sup> and (**b**) from 2300 to 3000 cm−<sup>1</sup> recorded using 532 excitation laser. The intensity was normalized by the most intense peak and the fitting of the peaks using Lorentzian functions.

**Table 1.** Peak position and full-width at half maximum (FWHM) of GO at different drying times in the region from 1000 to 2000 cm<sup>−</sup>1. The FWHM was obtained using Lorentzian fitting.



**Table 2.** Peak position and full-width at half maximum (FWHM) of GO at different drying times in the region from 2300 to 3000 cm<sup>−</sup>1. The FWHM was obtained using Lorentzian fitting.

It is well known that in GO, the D and D' peaks are related to the presence of defects such as folded edges, vacancies, impurities (functional groups or remaining metal species), and the change from sp<sup>2</sup> to sp<sup>3</sup> hybridization [47–50]. In particular, a decrease in the D' peak intensity can be considered as straight evidence of GO reduction. On the other hand, the intensity ratio ID/ID' in our as-made GO is ∼ 2.2, erroneously suggesting only the predominance of edge defects in its structure, even a lower value than that observed in FLG.

The D\*\* band is due to contributions from the phonon density of states in finite size graphitic crystals, C−H vibrations in hydrogenated carbon, and hopping-like defects [51]. Depending on the preparation process, GO also displays a D\* band related to the sp<sup>3</sup> diamond line on disordered amorphous carbons, but the broad region between ∼1400 cm−<sup>1</sup> and ∼1650 cm−<sup>1</sup> is not attributed to diamond carbon phases [31,52,53]. Although the D\* band is not perceptible in non-dried GO, this band appears after the drying-time experiment (discussed below).

Interestingly enough, the 2D band region (Figure 3b) is characterized by intense 2D1A and D+G bands and a less intense 2D2A band in contrast to those observed in natural graphite (Figure 1b) and FLG (Figure 2b). The intensity ratio I2D2A/I2D1A = 0.22 in GO decreases about 12 times and 5 times compared to graphite and FLG, respectively. These outcomes are probably due to a good chemical exfoliation and reduction of the number of layers, but the resulting material has sufficient defects to activate the D, D', D\*\*, and D+G bands.

## *3.4. Raman Spectrum of Dried Graphene Oxide*

Figure 4 shows the Raman spectra of GO dried at different times from 0.5 h to 24 h, keeping the temperature fixed (80 ◦C). The D, G, and D' peaks are observed which are not substantially affected in position, FWHM, or intensity in the drying-time testing (Table 1, Figure 5). In fact, a thermal reduction of the material obtained is not expected, as confirmed by FT-IR results (no shown here). Interestingly, the D\* band is not as intense at 0.5 h and 1 h, and can be seen as the drying time increases from 3 h to 24 h.

Two intervals are observed: (i) From 0.5 h to 3 h, the D\* peak is shifted at 1135 cm−<sup>1</sup> while the FWHM value remains relatively unchanged, and (ii) from 5 h to 24 h, the peak position is shifted to lower wavenumber values while the FWHM value notably increases, giving the highest value at 5 h (FWHM = 174) and 24 h (FWHM = 89). In contrast, the peak position and FWHM value of the D\*\* peak are not clearly affected by the drying time, but its intensity increases along with the intensity of the D\* band.

As seen in Figure 5, the relative intensity of the D' (blue line) and G (orange line) bands shows a constant trend, while the D\* (green line) and D\*\* (magenta line) display a square-root-like dispersion. From the curve fitting, it can be seen that the most significant effect of drying time is ≤10 h. These outcomes show that it is not possible to transform GO into reduce graphene oxide (rGO) through drying time at 80 ◦C. Nevertheless, if it is possible to increase the density of sp3-type and hopping-like defects associated with the presence of the D\* and D\*\* band, respectively.

**Figure 4.** Raman spectra of GO from 1000 to 2000 cm−<sup>1</sup> recorded using 532 excitation laser, subject to 80 ◦C, and considering different drying times: (**a**) 0.5 h, (**b**) 1 h, (**c**) 3 h, (**d**) 5 h, and (**e**) 24 h. The intensity was normalized by the most intense peak and the fitting of the peaks using Lorentzian functions.

**Figure 5.** Intensity of the D\*, D\*\*, G, and D' peaks as function of the drying time (0.5 h, 1 h, 3 h, 5 h, 24 h). The maximum intensity was obtained from the fitting using Lorentzian functions.

Therefore, it cannot be said that there is only one type of defects in GO. Instead of doing an analysis based on the intensity ratio ID/ID', we hypothesize that the study of defects in GO must be accompanied through the ratio of intensities, i.e., ID/ID' for edge-, ID/ID\* for sp3-, and ID/ID\*\* for hopping-like defects. In the present work, the corresponding intensity ratio values as function of the drying time are: 23.95 < ID/ID\* < 6.05, 6.14 < ID/ID\*\* < 3.31, and 1.70 < ID/ID' < 1.76. This hypothesis and intensity ratio values should be corroborated future studies. Most importantly, it can be noted that the longer the drying

time, the lower the intensity ratio of ID/ID\* and ID/ID\*\* while the intensity ratio ID/ID' slightly increases. The ID/ID' values reasonably agree with those reported in [54].

The presence of defects in GO also affects the 2D band region which is deconvoluted in the 2D1A, 2D2A, and D+G bands (Figure 6). In the literature, these bands are not discussed in detail when talking about conventionally prepared GO because the very low intensity and only noticeable when rGO is obtained, a sign of the recovery of the graphene structure. An important feature of GO is the presence of the 2D band region, suggesting non-critical damage to graphene structure after oxidation process. Thus, non-aggressive, environmentally friendly, and highly efficient reducing agents could be used, for instance, citric acid, ascorbic acid, or citrus hystrix [55].

**Figure 6.** Raman spectra of GO from 2300 to 3000 cm−<sup>1</sup> recorded using 532 excitation laser, subject to 80 ◦C, and considering different drying times: (**a**) 0.5 h, (**b**) 1 h, (**c**) 3 h, (**d**) 5 h, and (**e**) 24 h. The intensity was normalized by the most intense peak and the fitting of the peaks using Lorentzian functions.

The Raman spectra, peak position, FWHM, and relative intensity of the deconvoluted 2D band region as a function of drying time, are reported in Figure 6, Table 2 and Figure 7, respectively. The 2D1A and D+G bands are clearly observed, while the 2D2A band begins to appear as an effect of the drying time after 3 h, reaching an intensity comparable to that of the 2D1A and D+G bands at 5 h (Figure 6d) and 24 h (Figure 6e).

The decrease of the intensity of the 2D2A band at 0.5 h (Figure 6a) and 1 h (Figure 6b), can be interpreted as due to GO with few layers and wrinkled structure, as can be concluded from the prevalence of the 2D1A and D+G bands, respectively. The presence of the 2D2A band after 3 h of drying is attributed to the evaporation of water molecules between the GO layers, causing the stacking of the chemically exfoliated sheets. Trying to explain this fact, we have carried out TEM measurements (discussed below).

Although the FWHM values of the 2D1A band does not critically change after 1 h of drying, its position is shifted at 2685 cm−1. After that, the value of the FWHM decreases (e.g., FWHM = 188 at 24 h) and the peak position moves to lower wavenumber values (2653 cm−<sup>1</sup> at 5 h and 24 h), even lower than that observed in graphite (2677 cm<sup>−</sup>1) or FLG (2685 cm<sup>−</sup>1).

**Figure 7.** Intensity of the 2D1A, 2D2A, D+G peaks as function of the drying time (0.5 h, 1 h, 3 h, 5 h, 24 h). The maximum intensity was obtained from the fitting using Lorentzian functions.

Compared to the peak position of the 2D2A band in graphite or FLG (2720 cm<sup>−</sup>1), this band in GO is found at ∼2750 cm−<sup>1</sup> and is not affected by drying time, but the FWHM value decreases down to 46 cm<sup>−</sup>1, where the intensity of the 2D2A band is barely perceptible (0.5 h and 1 h). On the other hand, the peak position and FWHM value of the D+G band decrease (e.g., 2922 cm−<sup>1</sup> and FWHM = 131 cm−<sup>1</sup> at 24 h) as the drying time increases. The peak position found at 5 h and 24 h of drying are close to that of FLG (2901 cm−1). The latter corroborates our statement that the D+G band moves to higher wavenumber values as an effect of the oxidation process, also causing the displacement of the G+D' band. Therefore, it is demonstrated the D+G band has a high dependence on drying time, and its effect must be carefully considered when characterizing GO samples.

The intensity ratio I2D2A/I2D1A increases from 0.22 to 0.89, a value close to that observed in FLG (I2D2A/I2D1A = 1.12) while the intensity ratio ID+G/I2D2A decreases from 1.21 to 1.14. The latter suggests a slight reduction of defects. Most importantly, Figure 7 shows the relative intensity of the 2D1A (black line), 2D2A (blue line), and D+G (red line) bands. The 2D1A and D+G bands are characterized by an exponentially decreasing behavior while the 2D2A band shows a sigmoid growth trend. From the curve fitting, it can be seen that the most significant effect is obtained for drying times ≤5 h. This crucial result shows that for long drying times, the water molecules and other possible oxygen-containing molecules are removed, allowing the exfoliated GO sheets to pile up, probably, to re-form graphite oxide.

With all this in mind, the Raman results showed that a GO not very disordered and with few layers can be obtained with a drying time of 1 to 3 h. After that, a highly disordered graphitic-like structure is expected, suggesting that long drying times being unnecessary in a practical large-scale GO production. Furthermore, long drying times may not facilitate the recovery of the graphene structure after even aggressive reduction processes, for example, using hydrazine. This statement motivates more extended works.

The UV-vis absorption spectra of GO subject to different drying times are depicted in Figure 8. At 0 h of drying, GO exhibits an absorption peak at ∼233 nm and a shoulder at ∼304 nm, which are attributed to the *π* − *π*<sup>∗</sup> transition in C−C bonds and the *n* − *π*<sup>∗</sup> transition in C=O bonds, respectively. As drying time increases, the peak and the shoulder gradually redshift at ∼250 nm and at ∼325 nm, respectively, suggesting that the electronic conjugation within graphene structure starts to be restored. Although it is not possible to affirm the transformation of GO into rGO (as evidenced by Raman measurements), this result offers the possibility to adapt the optical and electrical properties of GO.

**Figure 8.** UV-vis spectra of GO subject to different drying times.

In addition to the results obtained by the drying time experiment, we also briefly describe the effect of rehydration in Figures S1 and S2 (Raman spectra), and Figure S3 (UVvis measurements). The dried GO samples (at 1 h, 5 h, and 24 h) were newly redispersed in continuous stirring for 30 min. Particularly, the D\* band is not perceptible in all rehydrated samples, and the intensity of the 2D2A band substantially decreases. UV-vis results evidence a blueshift of the absorption *π* − *π*<sup>∗</sup> peak. These outcomes can be attributed a well redispersion of GO due to its hydrophilic properties.

## *3.5. TEM Analysis of Graphene Oxide*

TEM micrographs of non-dried GO and GO subjected to different drying times from 0.5 h to 24 h are shown in Figure 9. Non-dried GO sample appears as a transparent and thin nanosheet with some wrinkles and folds on the surface and edges, but not with a very disordered surface morphology as GO prepared by conventional methods. Instead of point defects, these wrinkles are associated with surface defects formed due to the folding or twisting of the exfoliated GO sheets, causing deviation from the sp<sup>2</sup> to sp3 character. This wrinkled structure is a characteristic of graphene-like materials, whereas they are not present in other carbon nanostructures, e.g., amorphous carbons.

At a first approximation, the GO sample subject to 0.5 h of drying (Figure 9b) appears to be very similar to non-dried GO. However, a clear difference is observed beyond the wrinkles and folds, i.e., opaque regions are detected, which are attributed to the stacking of exfoliated GO sheets because the intercalated water molecules between the layers begin to be removed. This effect is more clearly observed after 1 h and 3 h of drying. At 5 h and 24 h, the GO samples look dark, suggesting a high removal of the intercalated molecules between layers, which causes a large stacking of the exfoliated GO sheets. This outcome supports the presence of the 2D2A band observed in the Raman study. In particular, Figure 9f demonstrates the stacked layers after 24 h of drying, and a GO sample with a lateral size larger than the previous ones, which also allows to observe that the GO flakes are reassembling. The inset in Figure 9f shows that the flake with lateral size in the micrometers range is made of a superposition of restacked GO layers.

**Figure 9.** TEM analyses of GO subject to 80 ◦C and considering different drying times: (**a**) 0.0 h, (**b**) 0.5 h, (**c**) 1.0 h, (**d**) 3.0 h, (**e**) 5.0 h, and (**f**) 24 h.

#### **4. Conclusions**

In summary, we have reported the effect of drying time (from 0.0 h to 24 h) on GO by a systematic Raman study. For comparison, the Raman spectrum of graphite and FLG were also discussed. The work is complemented using UV-vis and TEM measurements. The D, G, and D' peaks were not affected by increasing the drying time, but the D\*, D\*\*, and 2D2A peaks seemed to be very sensitive. The relative intensity of the different Raman bands and corresponding FWHM's were discussed. UV-vis results evidenced a redshift from ∼233 nm (0.0 h) to ∼250 (24 h) nm. TEM results showed the stacking and reassembly of GO sheets as a direct consequence of the drying time.

The study was carried out with a process as simple as possible, subjecting each sample to 80 ◦C in an oven, and characterizing them as soon as possible. For this reason, in future studies, a controlled environment should be considered as well as the time the samples remain at room temperature before being characterized. The water molecules would continue to evaporate for long periods. Additionally, we suggest to consider the type of graphite since here only graphite powder has been used. In fact, large lateral graphite or expanded graphite could present different results compared to those reported here.

Our findings are intended to contribute to the control of the technical parameters involved in the synthesis process to achieve the large-scale production of GO powder.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/nano11041035/s1, Figure S1: Raman spectra of rehydrated GO samples (1 h, 5 h, 24 h) from 1000 to 2000 cm−<sup>1</sup> recorded using 532 excitation laser, Figure S2: Raman spectra of rehydrated GO samples (1 h, 5 h, 24 h) from 2300 to 3000 cm−<sup>1</sup> recorded using 532 excitation laser, Figure S3: UV-vis spectra of rehydrated GO samples (1 h, 5 h, 24 h).

**Author Contributions:** Conceptualization, supervision, C.V.G., L.S.C. and T.T.; methodology, A.V., F.A.A. and O.S.; validation, T.T., M.G. and L.S.C.; formal analysis, C.V.G., M.A., A.S. and L.S.C.; investigation, C.V.G.; resources, T.T.; data curation, T.T. and M.G.; writing—original draft preparation, C.V.G.; writing—review and editing, C.V.G.; visualization, M.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Universidad Técnica Particular de Loja (UTPL-Ecuador), project "222-Radon adsorption on activated carbon and graphene filters" (code: PROY\_INNOV\_QUI\_ 2021\_3019). Part of this work has been also supported by the FONDOCyT from Ministry of Higher Education Science and Technology of the Dominican Republic (grant no. 2018-2019-3A9-139).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that supports the findings of this study are available within the article. Any additional data relevant to this study are available from the author upon reasonable request.

**Acknowledgments:** T.T., M.G. and C.V.G. wish to thank Escuela Superior Politécnica de Chimborazo for hospitality during the completion of this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Waterborne Graphene- and Nanocellulose-Based Inks for Functional Conductive Films and 3D Structures**

**Jose M. González-Domínguez 1,\*, Alejandro Baigorri 1, Miguel Á. Álvarez-Sánchez 1, Eduardo Colom 1, Belén Villacampa 2, Alejandro Ansón-Casaos 1, Enrique García-Bordejé 1, Ana M. Benito <sup>1</sup> and Wolfgang K. Maser <sup>1</sup>**


**Abstract:** In the vast field of conductive inks, graphene-based nanomaterials, including chemical derivatives such as graphene oxide as well as carbon nanotubes, offer important advantages as per their excellent physical properties. However, inks filled with carbon nanostructures are usually based on toxic and contaminating organic solvents or surfactants, posing serious health and environmental risks. Water is the most desirable medium for any envisioned application, thus, in this context, nanocellulose, an emerging nanomaterial, enables the dispersion of carbon nanomaterials in aqueous media within a sustainable and environmentally friendly scenario. In this work, we present the development of water-based inks made of a ternary system (graphene oxide, carbon nanotubes and nanocellulose) employing an autoclave method. Upon controlling the experimental variables, low-viscosity inks, high-viscosity pastes or self-standing hydrogels can be obtained in a tailored way. The resulting inks and pastes are further processed by spray- or rod-coating technologies into conductive films, and the hydrogels can be turned into aerogels by freeze-drying. The film properties, with respect to electrical surface resistance, surface morphology and robustness, present favorable opportunities as metal-free conductive layers in liquid-phase processed electronic device structures.

**Keywords:** graphene; carbon nanotubes; nanocellulose; conductive inks; liquid-phase processing; film fabrication; sustainability; metal-free electrodes

## **1. Introduction**

Liquid-phase processing (LPP) is currently the preferred route for building up advanced layered film devices such as environmental, physical or biological sensors, logic circuits, radiofrequency transmitters or screens [1]. The cornerstone of this field is to integrate high-performance electronic materials into functional systems in a low-cost configuration, with high performance and ease of manufacturing [2]. Therefore, the wide diversity of deposition methods in the liquid phase are very suitable for this purpose, and able to provide high-resolution patterns through liquid inks in a fully scalable manner [3,4]. Inks are complex mixtures of certain components and additives within a liquid medium (generally an organic and high boiling point solvent) whose concentrations and chemical nature determine their physical properties, namely rheology, surface tension and conductivity. The development of inks based on electrically conductive species is therefore critical to progress in this field of work [4,5].

Conductive nanomaterials have attracted considerable interest in this regard, as they offer excellent electronic properties and high compatibility with LPP [5]. Among the most studied nanostructures, those based upon carbon, such as carbon nanotubes (CNTs) and graphene derivatives, stand out. They possess extraordinary electrical, thermal and mechanical properties [5], as well as being extremely light and prone to biocompatibility.

**Citation:** González-Domínguez, J.M.; Baigorri, A.; Álvarez-Sánchez, M.Á.; Colom, E.; Villacampa, B.; Ansón-Casaos, A.; García-Bordejé, E.; Benito, A.M.; Maser, W.K. Waterborne Graphene- and Nanocellulose-Based Inks for Functional Conductive Films and 3D Structures. *Nanomaterials* **2021**, *11*, 1435. https://doi.org/ 10.3390/nano11061435

Academic Editor: Jin-Suk Chung

Received: 4 May 2021 Accepted: 27 May 2021 Published: 29 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Recent efforts of the scientific community in the area of LPP of carbon nanostructures have focused on the development of inks, showing enormous potential in technological applications and some advantages like the abundance of the source material (graphite) [6]. Beyond the graphene materials obtained by direct graphite exfoliation [7], the scientific community has also successfully employed its related chemical derivatives, such as graphene oxide (GO) or reduced GO.

Traditionally, noble metals have been incorporated into conductive inks, at concentrations of ~60% of conductive metal to reach acceptable conductivity values. In addition to the high cost involved, their high concentration generates problems of chemical stability and reaction with other neighboring species such as air and solvents. Graphene-based nanostructures can solve this drawback due to their high chemical stability [3,8]. Moreover, conductive aqueous inks with metals can cause toxicity [9], while inks based on aqueous solutions of graphene have been tested in human skin cells, resulting in neither toxicity nor morphological changes at a cellular level [10]. Therefore, these inks may be safe if the rest of the additives are harmless. In this sense, GO, as a hydrophilic derivative of graphene, has shown excellent performance when processed from water-based dispersions into graphene-based conductive films [11]. Thus, to date, graphene material-based inks have demonstrated their superiority in a wide range of functionalities, such as flexible interconnections, electrodes, transparent conductors and supercapacitors [12]. Surface electrical resistance values in the range of ~100 Ω/ are required for acting as electrode materials in organic or perovskite solar cell devices [13,14]. Thus, in order to establish versatility in the application of a certain conductive material by use of LPP, it is necessary to control the resistivity of the deposited layer. Furthermore, its roughness and compaction mainly depend on the deposition method, influenced in turn by the viscosity of the ink. The preparation of inks made of graphene-based nanomaterials is hence a challenging task, since several rheological properties of the ink (namely density, surface tension and viscosity) have a great impact [15].

As a matter of fact, high-viscosity inks (hereafter called pastes), with viscosities higher than 500 cP, would be suitable to form thicker layers (for instance, by screen printing techniques), while those with less viscosity would be more suitable for spray or inkjet deposition, allowing for a finer patterning and, in particular, for the case of spray coating technologies, the coverage of large areas. In general, these values can be adjusted by experimental parameters such as the concentration of the conductive additive or the dispersant, and also by the solvent choice. As mentioned earlier, typical liquid media in inks are non-volatile organic solvents. The most used ones for the production graphenebased liquid suspensions are *N*-methyl-2-pyrrolidone (NMP), *N*,*N*-dimethylformamide (DMF) and dimethylsulfoxide (DMSO), which are the ones that behave best as liquid-phase exfoliants [16]. These are very polluting solvents, with a boiling point of >170 ◦C, posing a serious risk of toxicity in humans. However, the use of water does not automatically solve these problems either, since it requires the incorporation of high concentrations of surfactants and other additives [9,17,18], indefinitely remaining in the ink, possibly leading to environmental and toxicity problems. Thus, new methods compatible with lower boiling point solvents (such as water or alcohols), together with non-toxic dispersants, are in demand in order to attain a truly environmentally friendly LPP, without raising the manufacturing cost or jeopardizing the overall electric/optoelectronic device performance once the deposition method takes place. Graphene-based nanostructures have promising potential, amongst currently known nanomaterials, to fulfill this purpose [19]. In particular, the GO derivatives have advantageous features for their use in conductive inks, such as affordable commercial availability, and the facile LPP due to the huge content of oxygen groups, in turn responsible for the higher interlayer spacing between planes. As stated by many authors, the GO LPP in a myriad of solvents and media does not need the addition of stabilizers, meaning a processing benefit [19]. For this reason, we have chosen GO as a target of interest for the preparation of aqueous inks. The only drawback is that GO inks require a subsequent reduction step to turn those inks into conductive material, be it by chemical, thermal or electrochemical means, usually entailing acceptable but poorer conductive properties than pristine graphene due to a larger number of structural defects [19]. Therefore, in the present work, the joint action of GO and CNTs was pursued to optimize the eventual conductive properties.

However, if one wishes to process carbon nanostructures from the liquid phase with greener approaches, avoiding toxic solvents or surfactants, a game-changing strategy is needed. In such a scenario, nanocrystalline cellulose (NCC) in particular acquires great relevance because of its sustainability. This nanomaterial is obtained from natural cellulose sources, by selective hydrolysis of the non-crystalline domains [20], resulting in fibrillar or needle-like nanostructures with widths and thicknesses around 3–20 nm and lengths of a few hundred nanometers. Due to its intrinsic chemical nature and the sustainability of the source material, nanocellulose may be considered an environmentally friendly nanomaterial. Despite the scarcity of scientific studies, there are already some examples showing the enormous potential that nanocellulose has as an aqueous dispersing agent of reduced GO and carbon nanotubes [21,22]. The structural diversity of NCC is defined by its crystalline allomorphs, among which types I and II stand out [22]. While type I NCC (exhibiting cellulose chains parallel to each other) is dominant in nature, type II is artificially synthesized, presenting polymer chains in antiparallel arrangements and typically requiring extreme caustic conditions or recrystallization processes for its synthesis. However, we have recently implemented a method to synthesize both NCC allomorphs by one-pot acid hydrolysis with sulfuric acid without any post-treatment step [22]. We also demonstrated the feasibility of dispersing CNTs in NCC, without any previous chemical modification on the NCC, leading to very stable aqueous colloids with proven bioactivity towards colon cancer (Caco-2) cells. In fact, the combination of carbon nanostructures with nanocellulose is an emerging trend, leading to useful hybrid nanomaterials with potential applications in biomedicine [23]. Beyond their biological response, nanocellulose paves the way towards new conductive inks based on carbon nanostructures, both dispersible in water and obeying green principles.

In this work, we disclose the development of carbon nanostructure-based low-viscosity inks and high-viscosity pastes able to be processed into films, by taking advantage of the impressive properties of NCC, standing out as a sustainable and green dispersing agent in water. We herein present a parametric study of ternary inks or pastes, or even self-standing hydrogels, by combining GO, CNTs and NCC and using a hydrothermal method in an autoclave. Further, conductive films were fabricated by different deposition techniques, and also hydrogel-derived porous materials, with the potential to become reference components as LPP electrodes and interfaces in electric/optoelectronic layered film device structures, such as batteries, supercapacitors, sensors and solar cells, among others.

## **2. Materials and Methods**

#### *2.1. Starting Materials and Reagents*

The source of cellulose used in this work was microcrystalline cellulose (MCC) powder from cotton linters with an average particle size of 20 microns (Sigma-Aldrich, San Luis, MO, USA, ref 310697). Sulfuric acid (98%) for acid hydrolysis was purchased from Labkem (Barcelona, Spain). Ultrapure water with a conductivity of 0.055 μS/cm was obtained from a Siemens Ultraclear device (München, Germany) and used in every step of the present work. Multi-walled CNTs (MWCNTs) were acquired from NANOCYL® (NC7000TM variety), produced by a catalytic chemical vapor deposition method. Before use, MWCNTs underwent mild oxidation in liquid phase with HNO3 (1.5 M, 2 h under reflux) to render a more hydrophilic surface without compromising their structure [24]. GO came from a commercial aqueous dispersion (0.4 wt%, 4 mg/mL), purchased from Graphenea® (San Sebastián, Spain). The characterization of mildly oxidized MWCNTs is presented in the Supplementary Materials (Figures S1 and S2), while the characterization of GO is provided by the manufacturer [25].

## *2.2. Synthesis of NCC (Types I and II)*

As discussed in the Introduction, the synthesis of NCC was carried out according to a methodology developed in our laboratory [22]. In a typical experiment, 10 g of commercial MCC were added to 45 mL of ultrapure water inside a round-bottom flask, and dispersed with the aid of an ultrasonic bath for 10 min. The flask was then placed in an ice bath at 0 ◦C and 45 mL of 98% sulfuric acid were added dropwise, under constant magnetic stirring. Once added, the final concentration in the flask was 64% and the mixture was quite viscous, thus requiring vigorous and constant stirring to avoid locally high concentrations which could burn the cellulose. Right after the last droplet of acid was added, the flask was removed from the ice bath and transferred to a heating plate with magnetic stirring. At this point, the procedure varied according to the type of NCC to be obtained. For NCC type I, the reaction medium was heated for 10 min at 70 ◦C, while for type II it was heated for 5 h at 27 ◦C. Once the heating step was finished, the reaction medium was poured into ultrapure cold water, least ~10 times of the initial volume (90 mL) in a 1 L beaker and left overnight at 4 ◦C, in order to favor sedimentation. After decanting the supernatant liquid, a dispersion was left with a very acidic pH that was neutralized by dialysis. For this, the aqueous dispersion was inserted into a dialysis membrane (SpectraPor®, Spectrum Labs, regenerated cellulose, 6–8 kDa cutoff molecular weight) immersed in 5 L of ultrapure water. The dialysis water was changed periodically until neutral pH in the washing waters was achieved. Then, the dialyzed medium was centrifuged at ~9300 rcf, and the supernatant liquid was kept and subsequently freeze-dried, in order to use NCC as a fine and light powder. Full characterization of NCC can be found in reference [22].

## *2.3. Hydrothermal Treatment to Obtain Inks, Pastes and Hydrogels*

Obtaining aqueous formulations able to be processed into films by different deposition techniques requires a certain control over their viscosity. This has been successfully attained through hydrothermal processes carried out in an autoclave. By heating a mixture of nanomaterials in water at a constant temperature (180 ◦C) in pressurized containers, with controlled times and pH, the chemical crosslinking of nanomaterials is favored [26]. In our case, this led to inks and pastes of different viscosities or even self-standing hydrogels. In general, what is herein termed as ink was of the order of 60 cP, while viscous pastes exhibited values of around 1500 cP.

We have undertaken a parametric study of different hydrothermal experiments, mixing variable amounts of NCC and CNTs and a fixed amount of GO, in order to ascertain the precise conditions under which inks, pastes or hydrogels form. All samples were brought to a final volume of 10 mL, and all samples were prepared at natural pH or alkalinized by the addition of aqueous ammonia (NH4OH), since basic pH plays a critical role by favoring the graphenic nanomaterials' aggregation during hydrothermal treatment [26]. The experimental procedure followed was: in a flat-bottom quartz vial, a specific amount of NCC (yielding final concentrations of 2.5 mg/mL or 5 mg/mL) was weighed. Unless stated otherwise, NCC refers to its type I polymorph. Then, a certain amount of CNTs were added to the vial (to attain final concentrations of 0.1 mg/mL, 0.2 mg/mL, 0.5 mg/mL, 1 mg/mL or 2 mg/mL), enabling the analysis of the influence of the amount of CNTs on the conductivity and viscosity of the resulting composites. Further, a variable volume of ultrapure water (4.8 mL or 5.0 mL) was added depending on whether or not NH4OH (200 μL) was added, respectively. At this point, the vial was subjected to an ultrasound bath (45 kHz) for 3 min. After that, a semi-homogeneous dispersion could be observed inside the vial. Finally, a fixed volume (5.0 mL) of the aqueous dispersion of GO was added, resulting in a final GO concentration of 2 mg/mL. The resulting medium was again bath-sonicated for another 2 min, thus obtaining a homogeneous dispersion with a final volume of 10 mL.

The hydrothermal treatment consisted of placing the vials in a Teflon vessel, inserting it into a metallic autoclave, tightly closed, and then heating it in an oven. The autoclaves were placed in the oven after pre-heating it at 180 ◦C, with a heating ramp of 20 ◦C, and the treatment lasted for a specific duration. The studied times were mainly 60 min, 30 min and 15 min. After one of these times had elapsed, the autoclave was removed from the oven and allowed to cool down in ambient conditions to room temperature before taking the sample out of the autoclave. The time needed for the sample inside the autoclave to reach 180 ◦C was estimated to be compensated by the time needed for cooling down to a safe handling temperature, so the overall treatment times were consistent in all cases. Figure 1 shows a visual scheme of the process.

**Figure 1.** General scheme for the preparation of inks and pastes following the autoclave method.

#### *2.4. Film Preparation and Characterization*

Once the inks (low-viscosity formulations) and pastes (high-viscosity formulations) were obtained by the procedures described above, the different possible deposition techniques were studied, in order to obtain the optimum conductive films. Spray-coating with an airbrush onto glass substrates was the chosen approach for the deposition of inks [27]. Films obtained from the viscous pastes were deposited by means of a rod-coating method using an agate rod, also over glass substrates. In both cases, the substrate was placed on a heating plate at 60 ◦C. For every sample, a total of 10 mL of the ink or paste was deposited in subsequent passages on the glass substrate, leading to films prepared under comparable conditions.

In order to measure the resistivity of the films, a Keithley 4200 unit was used, working in the range from −100 to 100 mV. An in-line 4-point probe configuration with equidistant probe separations of 2.24 mm was utilized, with controlled and homogeneous pressure over the conductive film. For the resistivity measurement, the geometry of the film and the distance between the electrodes were taken into account [18].

Morphology and microstructure of the deposited films were assessed through both optical and scanning electron (SEM) microscopies. Optical images were taken with a Zeiss AXIO microscope (Jena, Germany) with 20× and 50× objective lenses (N.A. = 0.4 and 0.7, respectively). SEM images were taken using Hitachi S3400N equipment (Tokyo, Japan), working in the secondary electron mode at a voltage of 15 kV and a distance of 5 mm.

Sample thickness as well as the roughness of the surface were evaluated by using a contact DektakXT Stylus Profiler (from Bruker, Billerica, MA, USA). The radius of the stylus used in the measurements was 2.5 μm. The height of the step of each deposited layer was measured in different areas along the layer edge. The depths of grooves made in the central part of the sample (reaching the substrate) were also determined. The layer thickness was obtained as the mean value of such measurements. The profile roughness was analyzed using Dektak analytical software (from Bruker, Jena, Germany).

#### *2.5. Aerogel Preparation and Characterization*

In order to obtain aerogels from hydrogels, a unidirectional freezing was applied before lyophilization. For that purpose, hydrogels were placed in an empty quartz vessel with a bottom platform made of metal which was immersed in liquid nitrogen. Once the hydrogel was fully frozen starting from the metal base, it was placed in a Telstar Cryodos freeze-drier, working at −49 ◦C and 0.3 mbar [28].

In order to measure the resistivity of the aerogels, the aforementioned Keithley unit was employed in a two-probe configuration. Aerogels were held between spring-loaded copper foils. The tungsten needle probes were brought into contact with the copper sheets, avoiding damage to the aerogels. The resistivities of the aerogels were calculated, taking into account the geometry, given by the distance between the electrodes and the specimen diameter, thus assuming a perfectly cylindrical shape.

## **3. Results and Discussion**

## *3.1. Hydrothermal Development of Inks, Pastes and Hydrogels: Unraveling Critical Parameters*

As stated earlier, different experiments were carried out at different concentrations of NCC and CNTs, whereas the GO concentration remained fixed. Different hydrothermal treatment times were applied (always at 180 ◦C), and pH was also varied by adding NH4OH or not to the medium. The systematic combination of such variables provided a roadmap for this ternary system (CNTs, GO and NCC type I) in water upon hydrothermal treatment in an autoclave. The two main possibilities pursued in the present work were liquid inks and viscous pastes, but it is worth recalling that the hydrothermal treatment of aqueous GO suspensions can also provide self-standing hydrogels, even for short times (generally beyond 30–45 min at 180 ◦C), based on the crosslinking of the GO sheets [28]. The viscosity of the medium increases progressively with the treatment time, also aided by the presence of basic pH. Thus, by controlling the thermal treatment time, homogeneous liquid dispersions, pastes or hydrogels can be obtained (Figures 2 and 3).

**Figure 2.** Three-dimensional scatter plot representing the conditions to obtain inks (blue), viscous pastes (orange) and self-standing hydrogels (gray). The effect of adding aqueous ammonia is herein represented, at different amounts of CNTs.

**Figure 3.** Three-dimensional scatter plot representing the conditions to obtain liquid inks (blue), viscous pastes (orange) and self-standing hydrogels (gray). The effect of CNTs is herein represented. All samples here had basic pH by adding aqueous ammonia.

As observed in Figures 2 and 3, hydrogels are always obtained for hydrothermal treatment times longer than 30 min, regardless of the other parameters. This suggests that treatment time is the dominant variable. This is consistent with our previous studies on the hydrothermal treatment of GO suspensions, with or without basic medium, in the absence of any other additive [28]. However, at 30 min of treatment time, the determining variable is the basicity of the medium. Samples containing NH4OH result in pastes or hydrogels, and those without NH4OH result in inks or pastes depending on the presence of CNTs. In fact, the joint presence of CNTs and NCC while adding NH4OH seems to be responsible for the rise in viscosity, leading to pastes.

For the experiments carried out for 15 min, the importance of the pH is observed again. Samples containing NH4OH give rise to pastes or hydrogels, and those without NH4OH give rise to inks. Again, the presence of NCC is decisive, since (at a given time of hydrothermal treatment with NH4OH) it leads to a paste, but in the absence of NCC the resulting outcome is a hydrogel. It can be reasonably postulated that the interactions between NCC and GO during hydrothermal treatment partly prevent the self-crosslinking of the GO and lead to viscous pastes instead of fully crosslinked hydrogels. Finally, it was observed that the concentration of NCC in the system was not a critical variable for the kind of aqueous formulations obtained. Therefore, in subsequent tests, this variable was set at 5 mg/mL.

In order to unravel the effect of crystalline polymorphism of NCC in the hydrothermal process, a series of experiments was also carried out with type II NCC. In this case, all experiments were performed for 15 min (since in the abovementioned results, there were no significant differences between 15 and 30 min). When NCC type II is incorporated into the system instead of type I, surprisingly, all formulations obtained in each experiment are low-viscosity inks (Table 1). This may be related to the aforesaid hindrance of GO crosslinking, as type II NCC could experience stronger interactions with the functional groups on the surface of GO, impeding its aggregation and thus leading to low-viscosity inks. This effect is stronger than that observed for type I NCC, which is consistent with the higher content of ester sulfate groups in type II NCC [22], responsible for its negative surface charge. As in the case for type I NCC, above 30 min of treatment time, in the presence of NH4OH, hydrogels are always obtained with type II NCC.

**Table 1.** The outcome of the hydrothermal treatment and composition of the GO/CNT/NCC (type II) ternary system in water at 180 ◦C for 15 min.


## *3.2. Morphology of Films Derived from Inks and Pastes*

Low-viscosity inks and high-viscosity pastes require the use of different deposition methods to attain conductive films over glass substrates. Inks were deposited by means of spray coating, while pastes were processed using the rod-coating method. The general appearance of such films is depicted in Figure 4. Films made from inks had a very compact aspect, because spray coating allowed for a tight coverage. Viscous pastes could not be deposited by spraying, with the rod coating method being the one showing the highest effectiveness, leading to films with an apparently more porous aspect. Deeper insights into the surface morphology of these films were obtained by both optical and scanning electron microscopies (Figure 5).

**Figure 4.** Photographs of films made from low-viscosity inks (**left**) and high-viscosity pastes (**right**) on glass substrates.

As confirmed by microscopy images, films derived from liquid inks and viscous pastes have a very distinct morphology both at the millimeter and the micrometer scale. According to the millimeter scale, films derived from inks exhibit no major irregularities, and a negligible presence of pores. In contrast, those coming from rod-coated pastes show the presence of large pores, together with more irregularities. These pores may be important for subsequent processing of such films, such as liquid-phase infiltration or interfacing with other species. As for the film surface microstructure, unraveled by

SEM, the observation from optical microscopy is corroborated. Films derived from inks present higher homogeneity and compaction, while those coming from pastes are more topographically irregular. In essence, both kinds of films display a very unique topography.

**Figure 5.** Optical photographs (**a**–**d**) and SEM images (**e**–**h**) of films obtained with low-viscosity inks (**a**,**b**,**e**,**f**) and highviscosity pastes (**c**,**d**,**g**,**h**). Scale bars (in white) = 1 mm (**a**,**b**), 400 μm (**c**,**d**), 100 μm (**e**,**g**) and 50 μm (**f**,**h**). Each image (either optical or from SEM) corresponds to a random point of each sample, none is the direct magnification of another.

Additional results were obtained through the profilometer, by which thickness and roughness were quantified. The mean values for the films studied in the present work are shown in Figure 6. Inks fabricated with type I NCC are able to generate films of a low thickness (~10 μm on average) up to a certain CNT content. From 0.5 mg/mL CNTs and up, the sprayed films in identical conditions drastically increase in thickness, reaching values in the range of ~30–45 μm. Conversely, inks containing type II NCC present the opposite trend; at a null content of CNTs, the sprayed films display an average thickness of ~25 μm and progressively decrease with increasing content of CNTs. Films made from pastes showed a steady value regardless of the CNT content, with an average of ~10–15 μm thickness. Film roughness shows an identical trend as thickness for inks with type I NCC; low roughness (~1–2 μm) until reaching 0.5 mg/mL CNTs, after which this value becomes 3- or 4-fold larger. Inks from type II NCC and pastes lead to films of comparable roughness, in the range of ~2–3 μm. All of these data may be related to the different interaction that NCC has with carbon nanostructures. In a previous work [24], we reported that type I NCC hybrids with CNTs presented a discrete distribution of cellulose nanocrystals adsorbed on the sidewalls, while type II NCC led to a heavily wrapped nanohybrid. In addition, type I cellulose nanocrystals are much longer and needle-like whereas the type II ones are much shorter and slightly thicker [22]. The kind of interaction among nanostructures in the reaction mixture during the hydrothermal treatment could determine their packing upon film deposition. Spraying liquid inks with type I NCC may provide a less efficient packing of nanostructures when the CNT content is high enough, causing the thickness and roughness to increase. In contrast, type II NCC, could have led to efficient packing regardless of the CNT content, given its very distinct structure. As for pastes, the explanation could lie in the nature of the deposition technique, as rod-coating of viscous formulations may not provide a tight packing nor attain highly thick or rough surfaces. Additionally, the rod-coating method also exhibited a visible thickening of the film edges. These observations can be better understood by observing some representative profiles of each case (Figure 7). Films coming from both inks show a more regular aspect with periodic spikes in the presence of CNTs, which determine the higher or lower mean roughness, while pastes present a more irregular profile, with some visible pores in the millimeter scale, as observed by optical microscopy. The profilometer also served as a means to perform a scratching test on selected film samples (see Supplementary Materials, Video S1). The conclusions drawn from these tests is that the films have a high adhesion and mechanical resistance, since they are not damaged at all when scratched with high local pressures (3.75 MPa).

A final surface characterization of these films was conducted through Raman spectroscopy (Supplementary Materials, Figure S4). In the observed spectra, it is visible how the D-band (~1349 cm−1) and the G-band (~1590 cm−1) correspond to both multi-walled CNTs and reduced GO indistinguishably, but are different from the starting GO due to the better definition of the band at 2700 cm−<sup>1</sup> (2D) in the films, and the rougher profile of GO. The D/G intensity ratio is lower than 1 for GO and films without CNTs, while in the presence of the latter, this D/G ratio may reach values of ≥1.

### *3.3. Characterization of Electrical Properties*

The four-probe electrical measurements performed on the prepared films allowed us to obtain their surface resistivity (Figure 8). It is observed that the surface resistivity values are lower as the concentration of CNTs in the liquid formulation (ink or paste) increases. Indeed, if the presence of GO is the basis for the hydrothermal aggregation, and the incorporation of NCC exerts control over the viscosity of the medium, the control over the conductive behavior of the derived films can be ascribed to CNTs. Those films without CNTs present very similar resistivity values amongst them (~1.5·10<sup>3</sup> <sup>Ω</sup>/-), regardless of whether they come from more or less viscous formulations before deposition, most likely owing to the lone effect of hydrothermally reduced GO. In fact, NCC does not generate any kind of char in the working conditions (Figure S3, Supplementary Materials). Regarding

the inks, there are no significant differences in surface resistivity between those films coming from type I or II NCC. Both kinds of inks provide films with surface resistivities in the range of ~103–102 Ω/ up to 1 mg/mL CNTs, reaching further values below 100 Ω/- at 2 mg/mL CNTs. All these values decrease almost linearly on a logarithmic scale, typical for a behavior beyond the percolation threshold in the studied conditions. Films made from pastes exhibit a steeper decrease in the surface resistivity with an increasing amount of CNTs, attaining ~100 Ω/ at 1 mg/mL CNTs. This seems to be a 'plateau' value as resistivity does not change at 2 mg/mL CNTs. It is worth pointing out that these films could become perfect candidates for metal-free electrode components or current collectors in liquid-phase processed electric/optoelectronic layered film device structures. In order to account for the variation trends in film thickness with increasing CNT content (Figure 6), the bulk conductivity of the samples could also be calculated (see Figure S5, Supplementary Materials). These results also show a generally direct proportionality between both parameters.

**Figure 6.** Film thicknesses (**a**) and roughnesses (**b**) corresponding to samples coming from sprayed inks or rod-coated pastes.

**Figure 7.** Topographic profiles across 6 mm distance for films made from inks with type I NCC (**a**), from inks with type II NCC (**b**) and from pastes with type I NCC (**c**), at the indicated CNT ratio in the initial formulation. All samples came from formulations with 2 mg/mL GO and 5 mg/mL NCC.

**Figure 8.** Surface resistivity measurements for different films prepared from inks and pastes. The concentration of CNTs refers to the one in the liquid formulation before deposition. All samples on this graph came from formulations with 2 mg/mL GO and 5 mg/mL NCC.

## *3.4. Post-Synthesis Processing Versatility*

Up to this point, we have presented a green approach to water-based inks with carbon nanostructures. This has led to formulations of varied viscosity able to be deposited by different means, leading to conductive films with good surface resistivity, unique surface morphologies and a potential feasibility to be applied as electrodes. Nonetheless, the properties of such films may be still improved, in terms of electrical conductivity and tolerance to organics, upon heat treatment. As a proof of concept, we chose a candidate film presenting one of the lowest values of surface resistivity (97.4 Ω/-, coming from a rod-coated paste with 2 mg/mL GO, 5 mg/mL NCC and 2 mg/mL CNTs), and subjected it to heat treatment and organic solvent exposure.

In one experiment, the selected film was inserted into an oven and heated at 400 ◦C for 4h under air atmosphere, and the outcome was a film with intact integrity and lower surface resistivity (Figure 8). The production of three different replicas provided values in the range of ~25–35 Ω/-, being of special interest for the construction of conductive carbon layers in layered film optoelectronic devices [29]. The morphology of the films before and after heat treatment was assessed again by SEM (Figure 9). It becomes clear that the effect of heat treatment is to smoothen the surface topography, probably together with the burning of NCC. In the heat-treated films, the presence of CNTs is better discerned, very probably entailing an improved contact between CNTs and reduced GO, thus leading to lower interparticle contact resistance. These results are very encouraging for the replacement of commercial carbon pastes with many toxic additives and binders, as these pastes have been reported to peel off and deform when heated beyond 250–300 ◦C [30].

**Figure 9.** SEM images from films obtained with high-viscosity pastes, before (**a**,**b**) and after (**c**,**d**) heat treatment at 400 ◦C. Scale bars (in white) = 30 μm (**a**,**c**) and 2 μm (**b**,**d**).

An additional advantage is the stability over time of the conductive properties of the films. Along the course of this research, we corroborated the stability of the conductive properties of our films in a time frame of many hours or a few days, meaning that the measurement of the surface resistivity of a freshly prepared film is generally coincident with the measurement after a short–medium time frame. Additionally, these observations are also valid for long-term periods (months). All of this shows that the preservation of the conductive properties of these films is possible across lengthy time periods with regular shelf storage.

In another experiment, the films were subjected to a treatment with typical organic solvents in the LPP of carbon nanostructures (NMP, DMSO), in order to study the tolerance of the films to such solvents. For this, the films were heated at 60 ◦C for 5 min. Then, several droplets of one of these organic solvents in a volume of 45 μL were randomly scattered across different areas of the films, in order to evaluate their effect towards its integrity. The film was maintained at 60 ◦C for 10–15 min to ensure infiltration through the pores and placed in the oven at 60 ◦C for 4 h. The film perfectly resisted this treatment, as no evident damage was observed by eye nor the optical microscope, and no peeling off occurred either. In addition, its surface resistivity changed from 97.4 to 94.4 Ω/-, which could be considered negligible and within experimental error.

In essence, these tests reassert the robustness and versatility of our conductive films, as they can endure (and be improved in terms of surface resistivity by) thermal treatments at high temperatures and long durations, as well as withstand the exposure to high boiling point organic solvents in aggressive conditions without any morphological harm nor damage to the electrical properties. This latter fact is of special relevance for their future performance as electrode components, since the whole integrity and surface resistivity of the films are retained after aggressive treatments, and depending on the application in mind, the resistance of the films towards certain critical solvents could be an important advantage.

## *3.5. Properties of Hydrogel-Derived Aerogels*

Some of the working conditions led to porous and light graphene-based aerogels [26,28]. We have taken advantage of the possibilities granted by unidirectional freezing, followed by lyophilization, which are able to create an anisotropic internal microstructure composed of parallel channels, with critical applications in energy and environmental remediation [28,31,32]. In the present case, hydrogels were prepared under specific conditions and subjected to hydrothermal treatment with the ternary GO/NCC/CNT system in water (see Section 3.1 and Figure 10), and when subjected to unidirectional freezing prior to lyophilization, they also presented an anisotropic porous microstructure.

**Figure 10.** Real images of a hydrogel derived from hydrothermal treatment in water (**left**) and preparation scheme of aerogels by unidirectional freezing followed by lyophilization (**right**).

As inferred from the SEM images (Figure 11), aerogels resulting from the unidirectional freeze-drying of hydrogels presented an anisotropic microstructure, with continuous straight pores parallel to the aerogel's longitudinal axis. This demonstrates that not only can such a microstructure be obtained from the reduction of GO in hydrothermal conditions [28], but also in the presence of NCC and CNTs. According to measurements of weight and dimensions, we elucidated that these aerogels presented an average density of 0.028 ± 0.006 g/cm3, and axial resistivities (measured in a two-probe configuration) in the range of 10–100 Ω·m, for CNT concentrations in the range of 0.1 to 0.5 mg/mL. Aerogels obtained in similar conditions but without CNTs or NCC presented lower densities (~0.005 g/cm3), but higher electrical resistivity (~1000 <sup>Ω</sup>·m). Parallel to the case of inks and pastes, the presence of CNTs governs the electrical properties of freeze-dried hydrogels, but in this case, NCC has a significant influence. When only GO and NCC are present in the hydrothermal medium, the density rises by one order of magnitude (~0.036 g/cm3) with respect to GO-only aerogels, but the axial resistivity is lowered by one order of magnitude. Therefore, all aerogels with GO, CNTs and NCC display electrical resistivities from ~102 down to ~10<sup>0</sup> in the studied range (from 0.1 to 0.5 mg/mL), revealing the importance of NCC in the process, which seems to play a role in the GO reduction during the hydrothermal treatment. NCC seems to boost the GO hydrothermal reduction and the number of interparticle contacts, hence favoring the decrease in electrical resistance at the expense of aerogel density. In summary, it is possible to produce light and porous aerogels with a GO/CNT/NCC ternary system, with an anisotropic microstructure and fairly low electrical resistivity in the absence of any post-treatment, comparable to the ones with only GO [32].

**Figure 11.** SEM images of aerogels derived from the unidirectional freezing followed by lyophilization of hydrothermally prepared hydrogels. Longitudinal (**a**,**b**) and transversal (**c**,**d**) cuts. Scale bars (in white): 400 μm (**a**), 200 μm (**b**), 100 μm (**c**) and 20 μm (**d**).

#### **4. Conclusions**

Nanocrystalline cellulose has been demonstrated to be a green and sustainable means to generate conductive water-based ink formulations of tailored viscosity, with carbon nanotubes and graphene oxide, as well as three-dimensional structures. Control of the processing parameters allows the preparation of inks, pastes or hydrogels using the same approach. The liquid inks and pastes have been used to fabricate films exhibiting diverse surface morphologies (depending on the deposition method and composition), with low resistivity values (<100 Ω/-). These conductive films also exhibit great robustness, being able to avoid disintegration upon aggressive treatments with organic solvents, as well as being electrically improvable by high-temperature treatments. Our studies show that formulations of this kind have great potential as metal-free electrodes in liquid-phase processed layered films, devices and structures. The green processes herein described, in addition to the capability of easily tuning the viscosity of liquid formulations, can be a great starting point for industrial development.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/nano11061435/s1, Figure S1: Thermogravimetric analysis in N2 atmosphere, Figure S2: TEM images of mildly oxidized MWCNTs, Figure S3: Photograph of NCC aqueous colloids subjected to hydrothermal treatment. Figure S4. Raman spectra (measured with a 532nm laser, Horiba Jobin Yvon equipment) for different selected conductive films. Each spectrum is the average of at least 5 random points across the film surface. Figure S5. Electrical conductivity values for different films coming from inks or pastes with different carbon nanotube contents. Videos S1: Recording of the stylus-based scratching tests, by applying a constant pressure of 3.75 MPa with the diamond tip (in the center of the crosshead). The distance traveled was 3 mm in each case. After scratching the films, no grooves or trails were observed, and the final profile seen corresponds to the neat sample topography. This means that these films are not damaged at all when scratched at high local pressures.

**Author Contributions:** Conceptualization, J.M.G.-D.; experimental work, J.M.G.-D., A.B., B.V., M.Á.Á.-S.; data processing, J.M.G.-D., A.B., E.C., A.A.-C.; resources, J.M.G.-D., E.G.-B., A.M.B., W.K.M.; writing—original draft preparation, J.M.G.-D.; writing—review and editing, J.M.G.-D., E.C., B.V., E.G.-B., A.A.-C., A.M.B., W.K.M. All authors have substantially contributed to, read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Spanish MINEICO/MICINN, under project references ENE2016-79282-C5-1-R (AEI/UE/FEDER) and PID2019-104272RB-C51/AEI/10.13039/501100011033, and the Gobierno de Aragón (Grupo Reconocido DGA-T03\_20R). J.M.G.-D. greatly acknowledges Spanish MINEICO for his 'Juan de la Cierva Incorporation' contract and associated research funds (ref. IJCI-2016-27789). The APC predoctoral contracts of E.C. and M.Á.Á.-S. were funded by the Spanish MINEICO (BES2017-080020, including EU Social Funds) and the Gobierno de Aragón, respectively. The work performed at the University of Zaragoza was funded by the MICINN (PID2019-104307GB-I00/AEI/10.13039/501100011033) and Gobierno de Aragón (E47\_20R).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Acknowledgments:** The authors acknowledge Carlos Martínez Barón for his support.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Review* **Developments in Synthesis and Potential Electronic and Magnetic Applications of Pristine and Doped Graphynes**

**Gisya Abdi 1,2, Abdolhamid Alizadeh 3, Wojciech Grochala <sup>1</sup> and Andrzej Szczurek 1,\***


**Abstract:** Doping and its consequences on the electronic features, optoelectronic features, and magnetism of graphynes (GYs) are reviewed in this work. First, synthetic strategies that consider numerous chemically and dimensionally different structures are discussed. Simultaneous or subsequent doping with heteroatoms, controlling dimensions, applying strain, and applying external electric fields can serve as effective ways to modulate the band structure of these new sp2/sp allotropes of carbon. The fundamental band gap is crucially dependent on morphology, with low dimensional GYs displaying a broader band gap than their bulk counterparts. Accurately chosen precursors and synthesis conditions ensure complete control of the morphological, electronic, and physicochemical properties of resulting GY sheets as well as the distribution of dopants deposited on GY surfaces. The uniform and quantitative inclusion of non-metallic (B, Cl, N, O, or P) and metallic (Fe, Co, or Ni) elements into graphyne derivatives were theoretically and experimentally studied, which improved their electronic and magnetic properties as row systems or in heterojunction. The effect of heteroatoms associated with metallic impurities on the magnetic properties of GYs was investigated. Finally, the flexibility of doped GYs' electronic and magnetic features recommends them for new electronic and optoelectronic applications.

**Keywords:** graphyne-like materials; synthesis and doping; electronic and magnetic properties; electronic transport; photodetectors

## **1. Introduction**

Among all chemical elements, carbon exhibits the greatest flexibility of its first coordination sphere, which is usually presented in textbooks as sp, sp2, and sp3 hybridizations. This plasticity leads to three available types of bonds (single-, double-, and triple-bonded C atoms) that may occur in diverse, practically unlimited, connectivities. Altogether, this markedly influences the allotropy of carbon, which is the richest among all chemical elements. Familiar allotropic forms of carbon include graphite, rhombohedral graphite, diamond, lonsdaleite, amorphous carbon (soot with variable sp2/sp3 carbon atom contents), and a huge variety of human-made high-specific surface area carbons, carbon aerogels [1,2], carbon foams [3,4], glassy carbon [5,6], polyynes [7], diverse fullerenes from C12 to as large as C960 [8,9], a multitude of single wall nanotubes [10,11], nano-onions [12], and more. Last but not least, they include graphene, which has unique physical properties [12]. Other exotic forms such as ultra-high pressure BC8 [13,14], Po-32 [15], and ferromagnetic carbon [16] have previously been theorized [17] (C18 was even reported [18]), but some are still disputed. Nevertheless, this structural diversity and versatility of chemical bonding brings an enormous pool of physicochemical properties, reactivities, and so on; crystal structures of over 500 periodic allotropes, known and hypothesized, have been collected

**Citation:** Abdi, G.; Alizadeh, A.; Grochala, W.; Szczurek, A. Developments in Synthesis and Potential Electronic and Magnetic Applications of Pristine and Doped Graphynes. *Nanomaterials* **2021**, *11*, 2268. https://doi.org/10.3390/ nano11092268

Academic Editor: José Miguel González-Domínguez

Received: 23 July 2021 Accepted: 30 August 2021 Published: 31 August 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in a unique Sacada database (https://www.sacada.info/) [19]. Despite the long-lasting research of carbon-based materials, some fundamental issues related to the shape of the phase diagram and mutual stability of polymorphs, or even their existence, remain unresolved to this day [20–22]; e.g., it has been recently claimed that lonsdaleite is not a genuine allotropic form but a twin of cubic crystals, which raised controversy [23,24]. One illustration of the intensity of the research field of carbon materials can be provided by an inspection of the Web of Science database; this resource lists approximately 114,000 papers using the keyword 'diamond', approximately 147,000 papers discussing 'graphite', and approximately 240,000 papers featuring 'graphene'. It can be safely estimated that—even given some overlap—well over half a million of over 90 million indexed scientific works (1900–2021) have been devoted to carbon allotropes. The discovery of graphene triggered the production of diverse 2D carbonaceous materials including graphone, graphane, and graphene oxide [25–27]. Graphynes (GYs) are layered two-dimensional structures built from sp- and sp2-hybridized carbon atoms (Figure 1A).

**Figure 1.** The structure of graphyne with indicated distribution of sp (red rings) and sp<sup>2</sup> (blue circles) hybridized carbons (**A**); The representation of graphdiyne structure (**B**).

Wide tunability in structural, mechanical, physical, and chemical properties make GYs fascinating candidates for use in energy storage, solar cells, electronic and spintronic devices, UV light detectors, as well as adsorbents in the separation of gases [28]. A GY can be a useful catalyst in water purification [28]. GYs have shown improved electronic properties and charge carriers in the optics and electronics industries. Graphone and graphane are hydrogenated forms of graphene—with fundamental band gaps of 2.45 and 5.4 eV, respectively, and interesting magnetic properties—that have shown potential for nanoelectronics and spintronics [29–31]. Due to structural similarities with graphene, these materials are considered to be excellent candidates for carbonaceous electronic devices, and they surely will be the subject of advanced studies for multiple and versatile applications. Scientists have been theoretically studying graphynes since the 1980s, which initially attracted attention after the discovery of fullerenes [32]. Though the structure of GY was first proposed in 1987 by Baughman et al. [33], the demanding synthesis of GYs hindered their dynamic development for more than 20 years [34]. Among all the GYs, the risingstar γ-graphdiyne (γ-GDY), seen in Figure 1B, was the first GDY member experimentally synthesized and reported by Li et al. in 2010 [35]. Because of the promising physical, optical, and mechanical features of GYs, a tremendous amount of research effort has been dedicated by theoretical, applied, and synthetic chemists. It is believed that GYs might pose competition for more common sp2-hybridized carbon systems, particularly graphene, and

meet the increasing demand for an alternative candidate to carbonaceous materials. Recent years have brought a sharp increase in research interest on the synthesis and theoretical prediction of GYs' properties in different dimensions, e.g., one-dimensional nanowires (such as nanotube arrays and ordered stripe arrays), and two-dimensional nanowalls (2D) and nanosheets (2D), and 3D frameworks [36–38]. More and more works are concerning pristine GYs' structures enriched with diverse heteroatoms (B, F, N, O, etc.) [37]. Due to the broad spectrum of intense scientific activities related to the development of new forms of GYs, this work emphasizes recent developments in this field, especially dealing with newly obtained heteroatom-doped structures and investigating the synergic effects of heteroatoms, metal oxides, and metal ions on the electronic and optoelectronic properties of doped GYs. In this work, we also focus our attention on theoretical and experimental research into the magnetism of pristine and doped GYs.

## **2. From Atomic Structure Suggestion to Experimental Appearance**

In 1987, Baughman et al. [33] proposed a novel two-dimensional allotrope of carbon assembled by aromatic centers (sp2) linked to each other by acetylenic bridges (sp); these were named graphynes (GYs). In the beginning, GYs (a general name given to this type of material) were designated according to the number of carbons included in the various rings forming a given network (shaped pores in the structure). For example, A, B, and C networks were called 18,18,18-graphyne (Figure 2A), 12,12,12-graphyne (Figure 2B), and 6,6,6-graphyne (Figure 2C), respectively. The latter theoretically present the lowest energy of carbon phases consisting of flat molecular sheets obtained by inserting acetylene linkers between the aromatic rings into the pristine honeycomb graphene structure. As energetically favorable, the 6,6,6-graphyne has a special place in the GY family, and the term "graphyne" without specifying the number of carbons is reserved for this phase [39].

**Figure 2.** Representation of different graphyne and graphdiyne–like structures. α-graphyne (**A**); β-graphyne (**B**); γ-grapyne (**C**); α-graphdiyne (**D**); β-graphdiyne (**E**); γ-graphdiyne (**F**). The red dots represent acetylenic (graphynes) and diacetylenic (graphdiynes) bridges (Csp) and yellow ones—sp2 hybridized carbons.

In 1997, Haley et al. proposed new forms of GYs called graphdiyne (GDY), a nanostructure formed from two adjacent acetylenic linkers between aromatic carbon atoms in contrast to graphyne, which is made of carbons connected by one triple bond [40]. There-

fore, to investigate new structures, different GY networks designations were developed and used in the literature [39,41]. Figure 2 shows several types of GY structures named in accordance with the widespread terminology including α-graphyne (Figure 2A), β-graphyne (Figure 2B), γ-graphyne (Figure 2C), α-graphdiyne (Figure 2D), β-graphdiyne (Figure 2E), and γ-graphdiyne (Figure 2F). The last nomenclature method named γ-graphynes with various numbers of acetylenic linkers in a structure, such as graph-n-yne (*n* = 1, 2, 3 ... , where *n* is the number of triple bonds). The mentioned nomenclature methods properly operate when naming the most conventional GYs and GDYs. However, the graphyne naming system causes considerable problems and should be standardized, especially for structures containing heteroatoms. In this paper, GYs, GDYs, and their derivatives are all briefly designated as graphyne- or graphdiyne-like structures with the abbreviation "GYs" and "GDYs", respectively.

In order to prepare extended structures of GYs, chemists developed versatile synthetic methods grouped into two subsections: top-down and bottom-up [42]. The top-down methods have the potential to obtain two dimensional structures from their bulk precursors through processes such as mechanochemical synthesis and vapor-liquid-solid (VLS) growth. The bottom-up formulations result in the formation of thin films or a few layers of GYs through coupling reactions of desired substrates in carefully chosen and strictly controlled conditions. "On-surface" coupling synthesis conducted in an ultra-high vacuum (UHVS), chemical vapor deposition (CVD), thermal treatments, and wet-synthetic approaches are practical subdivisions of the bottom-up method.

By summarizing the most representative and effective synthetic methods, we hope they can create an idea and solid foundation for preparing GYs with given applications in various research fields. The wet-synthetic strategy is the most successful and high-demand method for the preparation of a wide range of applicable GYs, in both pristine and doped natures. In order to facilitate understanding and comparison, the wet-synthetic methods are classified into two categories based on the different phases where the coupling reaction of the reactants occurs amid the synthesis process: one-phase (homogenous reaction) and twophase (interfacial reaction). Two-phase methods have three sub-divisions: liquid/liquid, solid/liquid, and gas/liquid.

We intend to summarize the developments in preparation methods resulting in different GY morphologies (nanowires, nanotubes, nanowalls, nanoribbons, nanosheets, etc.), as well as different chemical compositions. The broad-spectrum synthesis protocols employed in cross-coupling or homocoupling reactions of graphynes' subunits and derivatives were named after their inventors as Glaser [43], Eglinton [44], Hay [45], Negishi [46], Hiyama [47], and Sonogashira [48] reactions. All these methods, briefly presented in Table 1, were proven to be applicable to the preparation of valuable and scientifically important heteroatom-doped GYs.


**Table 1.** Synthetic methods employed in cross-coupling or homocoupling reactions for the preparation of graphyne derivatives; year when a given method was invented is indicated.


**Table 1.** *Cont.*

Theoretically proposed graphynes (GYs) induced tremendous effort to find applicable routes, resulting in the fabrication of these materials [39]. The controlled oligo-trimerization of cyclo-carbons such as cyclo-C18 (cycle made of nine acetylenic groups) and cyclo-C12 (cycle made of six acetylenic groups) and the oxidative polymerization of monomeric acetylenic precursors or synthetic macrocyclic compounds under Glaser–Hay coupling conditions are potential methods for the synthesis of γ-GDY, γ-GY, β-GDY, and other derivatives [49–52]. As an alternative strategy to investigate the potential properties of GYs, Haley et al. proposed a method for the synthetic preparation of γ-GY and γ-GDY substructures [40,53]. These subunits can be next used as the first block in the construction of extended GY structures. This breakthrough in synthetic approaches to alkynyl carbon materials revealed possibilities for the synthesis of low-dimensional carbonaceous nanomaterials involving acetylenic scaffolds [40,54–57]. Since the first report of γ-GDY preparation in 2010, exciting progress has been made in the experimental preparation of GYs. In the next sections, we describe different, well-developed synthetic protocols leading to graphyne-like structures composed of C, as well as structures chemically modified by heteroatoms (e.g., N, F, Cl, H, S, and B).

#### *2.1. Mechanochemical Synthesis*

A mechanochemical synthesis route is used in one of the most effective methods. Its strength lies in its simplicity, rapidity, and repeatability. As a result, solid extended GYs or their subunits with reproducible features may be produced. Under mechanical impact and at an elevated local temperature resulting from particle collisions, selected bonds of substrates are broken and new compounds may form in the solid-state, thus overcoming the problems associated with solution-based chemistry processes.

The reaction of hexachlorobenzene (HCB; known as a persistent organic pollutant) and calcium carbide (CaC2; known as an efficient and safe co-milling reagent) in a planetary ball mill at room temperature within 20 min of milling at a mass ratio of CaC2/HCB = 0.9 and a rotation speed of 300 rpm, as proposed by Li et al. [58] in 2017, caught the attention of materials chemists in the preparation of GYs. After this development, a mature mechanochemical approach was applied by Li et al. in 2017 for the one-step high-yield synthesis of GY monomers [59]. One year later, Cui et al. synthesized hydrogen-substituted graphyne (H-GY) and γ-graphyne (GY) via the ball-milling-driven mechanochemical cross-coupling of 1,3,5-tribromobenzene (PhBr3), hexabromobenzene (PhBr6), and CaC2 as precursors under a vacuum. Finally, the impurities were removed with diluted nitric acid and benzene (Figure 3) [60,61].

**Figure 3.** Schematic representation of mechanochemical synthesis for preparation of GY and its derivatives. Redrawn [60,61].

#### *2.2. The Vapor–Liquid–Solid Growth*

Synthesis reactions carried out with vapor-liquid-solid (VLS) growth allow one to control the production of different types of nanomaterials. In general, GY sheets are harvested on carefully prepared surfaces of monocrystalline silicon coated by metallic nanoparticles (Au, Fe, Zn, and Ni). The prepared substrate-catalyst support provides favorable physical features (surface energy, stability, and crystal structure) for the effective growth of nanostructural materials. The VLS method was applied by Li et al. for graphdiyne (GDY) film synthesis [62]. In their approach, powdered and vaporized GDY was deposited on the surface of ZnO nanorods grown on a silicon wafer (Figure 4a). Liquid nanodroplets of melted zinc (~419 ◦C) formed on one of the ends of the ZnO nanorods (ZnO NRs) (Figure 4b) served as energetically favorable adsorption sites of incoming vapors of GDY molecules. Due to the mixing of vapors with melted zinc, a solution of graphdiyne and Zn was formed (Figure 4c). The continuous influx of fresh portions of GDY molecules resulted in the formation of a supersaturated solution (GDY-Zn), as well as the fusion of drops and an increase in their size of droplets, thereby facilitating lateral growth due to the small edge energy of 2D materials.

#### *2.3. Thermal Treatment*

The heating of hexaethynylbenzene (HEB), N-rich precursors (2,4,6-triethynyl-1,3,5 triazine, TET), and pentaethynylpyridine (PEP) was applied by Zuo et al. to force a homocoupling reaction, resulting in GDY nanostructures with different nitrogen percentages and morphologies (nanoribbon, nanochain, and 3D-networks) [63,64]. Notably, this reaction could be carried out without using any metal catalyst. The powder of N-rich precursors were slowly delivered to the preheated conical flask (120 ◦C), leading to an explosive reaction whereby black GDY was obtained (Figure 5a,b). A gradual heating process (10 ◦C/min) to 120 ◦C in nitrogen converted the light-yellow HEB into black nanoribbon-like morphologies without volume change (Figure 5c,i). On the other hand, the implementation of this treatment in an air atmosphere resulted in GDY nanochains uniformly grown on the 3D network with a remarkable volume increase of 6 times (Figure 5c,ii). This means that the oxygen accelerated the dehydrogenation for the coupling reaction. However, the addition of HEB into a preheated air environment (120 ◦C) rapidly caused a more violent reaction, and an ultrafine nanochain with a 48-fold volume increase was obtained (Figure 5c,iii).

**Figure 4.** The illustration of the increase in a GDY layer in the VLS process on the head of a single ZnO NR (**a**) ZnO NRs in the presence of GDY vapors, (**b**) Zn droplet on the head of the ZnO NRs, and (**c**) Zn droplets scattered in GDY thin film, produced by connected several neighboring flakes of GDY. Redrawn [62].

The doping of carbonaceous materials with N atoms can be realized via the three following routes: chemical vapor deposition (CVD), the pyrolysis (annealing) of N-containing precursors, and heating with N-rich chemicals. For instance, N-doped GDY, B-doped GDY, F-doped GDY, and S-doped GDY have been prepared using an annealing strategy [65–70]. In such a case, GDY-based nanomaterials have been thermally treated with relevant chemicals, such as ammonia (N), B2O3 (B), NH4F (F), and thiourea (S), at a chosen temperature [70]. The considered methods, however, suffer from drawbacks, such as the randomness of the doping sites and uncontrolled dopant's percentage. Therefore, there is still a need to develop a synthetic strategy that provides 2D carbon materials with a homogeneous distribution of atoms of the desired type and at specific desired locations.

#### *2.4. "On-Surface" Synthesis under an Ultra-High Vacuum*

The so-called "on-surface synthesis" approach is carried out under an ultra-high vacuum (UHV) and presents considerable potential in building new types of nanomaterials. In this method, the starting building blocks are deposited onto the surface of a metallic substrate (Ag and Au) [71–85]. Then, the coupling of precursors occurs, resulting in singleatom-thick 1D and 2D materials. In contrast to conventional "wet" reactions, it helps here to eliminate possible undesired influences from surroundings. The chemical character of the organic precursors and the nature of the applied substrate are crucial factors that determine the final properties of GDY nanostructures [41]. Features such as the dimensionality of the organic precursors, the reactivity of their functionalities, the geometry of the surface, and interactions occurring between the organic building blocks and substrate have significant effects on the reaction and molecular surface patterns. The reactivity and mass of the used molecules have significant impacts on the success of synthesis. High reactivity or

weight may prevent them from sublimating on the surface of the substrate. On the other hand, molecules that are too small will escape from the reaction chamber. The metallic substrates play a double role as templates and catalysts of the coupling reaction. In the following sections, we discuss the most effective method for the synthesis of graphyne analog (sub-structures or infinite nanostructures) based on "on-surface" coupling reactions.

**Figure 5.** Thermal treatments of (**a**) TET, (**b**) PEP, and (**c**) HEB: (i) gradually heated to 120 ◦C/N2; (ii) rapidly heated to 120 ◦C/Air; and (iii) gradually heated to 120 ◦C/Air. Redrawn [63,64].

The production of acetylenic frameworks at interfaces involving the formation of self-assembled monolayers (SAMs) followed by a cross-linking step to form linked monolayers is a direct way to create carbonaceous materials such as carbyne, graphyne, and graphdiyne [71,72]. "On-surface" homocoupling reactions requiring the detachments of halogens or hydrogen from precursors functionalized with alkynyl groups have been reported as effective fabrication methods of low-dimensional carbon-based nanostructures. An overview of the production of 1D carbonaceous nanomaterials via "on-surface" approaches was previously described in the literature [73].

Rubben et al. reported a surface-assisted dehydrogenative homocoupling reaction of terminal alkynes (Csp–H), such as triethynylbenzene (TEB) and (1,3,5-tris-(4-ethynylphenyl) benzene (Ext-TEB), that was conducted on a Ag(111) surface, wherein the hydrogen was the only by-product of the reaction. The authors stated that it was an ideal method for the synthesis of individual chemicals or polymeric structures containing a conjugated backbone (after annealing at 400 K) [74]. In 2015, Wu et al. investigated the reaction of 2,5-diethynyl-1,4-bis (phenylethynyl)-benzene (DEBPB) taking place on the surface of silver with different facets including (111), (110), and (100). The reaction was carried out with the aid of scanning tunneling microscopy (STM). The Glaser synthesis conducted on Ag(111) was dominant and yielded one-dimensional, covalently bonded wires. On the contrary, reactions conducted on Ag(110) and Ag(100) surfaces resulted in one-dimensional organometallic frameworks built on terminal alkynes and metal atoms. (Figure 6) [75].

Klappenberger et al. employed a Ag(877) support to obtain one-dimensional conjugated molecular threads as components of extended GYs with lengths reaching 30 nm [76]. Thermal dehydrogenative reactions carried out on a flat Ag(111) plate were found to be associated with several undesirable side reactions that resulted in the formation of branched, irregular nanostructures. Liu et al. induced a dehydrogenative reaction of 2,5-diethynyl-1,4-bis(4-bromophenylethynyl)benzene and noticed that bromine adatoms affected the activation of C–H groups in terminal alkynes occurring at 298 K on a Ag(111) surface [77]. The STM studies disclosed the formation of organometallic species followed by their partial conversion to covalently bonded nanostructures after annealing at 420 K.

**Figure 6.** Schematic representation of the DEBPB coupling (Glaser reaction) carried out on silver surfaces. Redrawn [75].

In 2018, Xu et al. applied a dehalogenative homocoupling reaction to tribromoethylbenzene (TBP), 1,3-bis(tribromomethyl)benzene (bTBP), and 1,3,5-tris(tribromomethyl) benzene (tTBP), and they converted tribromomethyl functional groups (Csp3) to form C–C triple bonds (Csp) as structural motifs of dimeric structures, such as wires or 2D networks of GYs grown on a Au(111) surface [78].

In 2020, Xu et al. created on-surface graphyne nanowires through dehalogenative homocoupling reactions via the stepwise activation of two different types of C–Br bonds (involving Csp3–Br and Csp2–Br) in a 1-bromo-4-(tribromomethyl)benzene (BTBMB) compound on both Au(111) and Ag(110) surfaces [79]. Sun et al. also reported the successful formation of dimer structures with acetylenic linkers (wires and networks) via the on-surface C–Br activation of alkenyl carbon atoms [80]. Two-dimensional networks with acetylenic linkages were obtained after the homocoupling reaction of 1,3,5-tris(bromoethynyl)benzene (tBEP). In the first step, the precursor was deposited on a Au(111) surface at room temperature and was slightly heated up to 320 K. As a result, organometallic networks were obtained. The further increase in temperature up to ∼450 K led to the release of gold atoms and the formation of the final product. Figure 7 shows STM studies of this reaction. Considering the published results, the Ag(111) surface seems to be the most effective substrate for Glaser coupling because it causes a smaller number of side reactions [81] than Au(111) plates, for which the cyclization of the terminal alkyne to the benzene ring is common [82–84]. When used as a substrate, a Cu(111) surface presented low activity towards "on-surface" Glaser coupling. This stands in contrast to wet coupling reactions for which Cu ions show very high catalytic activity, and Cu is regarded as the most effective catalyst for such reactions [85].

**Figure 7.** Schematic representation of the growth of the organometallic framework after the deposition of tribromophenyl on Au(111) surface (from RT to 320 K) and the formation of GDY after heat treatment at 450 K (**a**). Scanning tunneling microscopy micrograph revealing the creation of GDY before and after heat treatment at 450 K (**b**,**c**). Detailed STM pictures of the C–Au–C framework (**d**), the mixture of C–Au–C networks and GDY fragments (**e**), and the GDY layer (**f**). The modeled structures of all considered networks are superimposed on the STM micrographs. Reproduced with permission [80]. Copyright 2016, American Chemical Society.

## *2.5. "On-Surface" Synthesis by Chemical Vapor Deposition*

Another synthesis strategy based on the covalent coupling of organic monomers occurring at the metal surface, so-called "on-surface" synthesis, is chemical vapor deposition (CVD).

Furthermore, the CVD method is recognized as one of the most promising routes for the creation of novel 2D materials. This approach relies on the transfer of vapors of monomers to reaction chambers and their embedding and coupling on the surface of a preheated metallic substrate. However, this method has severe limitations. As the reaction is conducted on a metal substrate without any additional catalysts, the reaction stops when the surface is fully covered with GY monolayer films (Figure 8) [86]. It has been statistically shown that silver seems to be the most efficient substrate for carrying out such reactions. In contrast to other investigated metallic substrates (such as Au and Cu), silver was found to ensure the lowest proportion of side reactions [86].

**Figure 8.** The formation of the GDY sheets on the silver surface with the aid of the CVD technique. Redrawn [86].

#### *2.6. Wet Chemical Synthesis*

Since the first successful fabrication of γ-GDY through a liquid/solid interfacial reaction on Cu foil as a catalytic substrate for an acceleration coupling reaction in an organic solvent, directed efforts have been carried out to prepare graphyne derivatives. Homo-coupling and cross-coupling reactions of hexaethynylbenzene (HEB) as an efficient precursor have been realized as operational pathways for the preparation of graphyne analogs (fragments, oligomers, or infinite structures) and have recently attracted immense attention. These methodologies employ metal salts (copper salts: Cu(I), Cu(II), or Pd(II)) in a homogenous conditions or on a surface template, e.g., Cu foil or other arbitrary surfaces such as graphene, Au, and 3D foam (which may have catalytic properties) to assist in a heterogeneous reaction [41].

#### 2.6.1. Developments of Coupling Reactions (All Reactants in Solution Phase)

The coupling reaction for the synthesis of graphyne analogs proposed by Moroni et al. combined dibromoaryls (I) and 1,4-diethynylaryls (II) in the presence of PdCl2, Cu(OAc)2, and triphenylphosphine (PPh)3 in a triethylamine/THF mixture (Figure 9A), where R1, R2, R3, and R4 could be the same or different (H, NO2, alkyl ether, alkyl thioether, or alkyl ester). Homopolymers or copolymers with phenyl, thienyl, anthryl, or stilbene groups as aryl units were synthesized [87]. Despite this progress in synthetic methods, extended structures of GYs are still unachievable.

The combined Negishi and Sonogashira cross-coupling reactions for the formulations of various kinds of substituted hexaethynylbenzenes from chloroiodobenzenes put researchers on a fast track towards the fabrication of GYs [88,89]. In 2007, Jiang et al. obtained poly(aryleneethynylene) networks with highly developed porous structures. To do so, they applied a Pd-supported Sonogashira-Hagihara reaction, which had previously been employed to synthesize different polymeric compounds such as polymers and ligands for coordination-polymer synthesis, wires, and shape-persistent macrocycles [90]. In 2010, Dowson et al. showed that porous properties (BET surface area and pore volume) are strictly controlled by the kind of solvent used as the environment of the reaction [91]. Toluene, tetrahydrofuran (THF), N-dimethylformamide (DMF), and 1,4-dioxane were tested for these reactions. Authors showed that DMF is the most proper solvent, as its received nanostructures are characterized by the highest BET surface areas (up to 1260 m2/g) [91,92]. In homogenous coupling reactions, materials chemists prepared oligomers and macromolecules, but infinite structures are still elusive.

After 2010, remarkable progress in GY preparation was achieved. The synthesis of poly(aryleneethynylene)s (PAEs) using Pd and Mo/W was thoroughly investigated by Bunz in 2010 [93]. Wu et al. utilized commercially available tris(*t*-butoxy)(2,2-dimeth ylpropylidyne)-tungsten(VI) as the catalyst in the synthesis of hydrogen-substituted graphyne (H-GY), as shown in Figure 9B [94]. Ding et al. prepared γ-graphyne in a homogenous ultrasound-driven reaction of hexabromobenzene (PhBr6) and calcium carbide (CaC2) in an inert atmosphere without a metal catalyst [95]. Wen et al. prepared new N-doped graphyne analogs (Figure 9C) in the reaction of nucleophilic substitution (SNAr) of cyanuric chloride and para-dilithium aromatic reagents. The process was carried out under mild conditions in a diglyme or bis(2-methoxyethyl) ether (solvents with high boiling points) solution. That designed reaction allowed them to obtain N-GYs on a gram scale [96]. The development of GDY synthesis was a breakthrough in the preparation of different morphologies such as films, nanowires, nanotube arrays, nanoribbons, nanosheets, and nanowalls of GYs with versatile properties, as well as reductions in the dimensionality [97].

Graphdiyne nanoribbons (GDYNRs) comprise a class of 1D GDY materials that stresses well-defined edges and nanometer size [97]. There have already been numerous theoretical efforts regarding GDYNRs seeking connections between their structures and properties. The results of this research are discussed in Section 3.2.3. A bottom-up chemical formulation could provide structurally uniform and well-defined nanostructures of GDYNRs. It is, however, necessary to perform the selective stepwise coupling of ethynyl groups during the synthesis procedure. Zhou et al. proposed a two-step method of intermolecular polymerization followed by the intramolecular cross-coupling of acetylenic moieties, as seen in Figure 9D (red fragment). First, the polymerization of the ethynyl units in the central part of the monomer ensures one-dimensional growth (Figure 9D, red fragment). Secondly, the intramolecular reaction of ethynyl groups on the established facing side chains and the bulky groups (such as the 3,5-di-tert-butylbenzyl group) on the outer side occurs. The latter works to sterically hinder intermolecular coupling. This strategy was applied to build GDYNRs nanostructures (Figure 9D, blue fragment) made of rhomboid units with benzene as junctions and butadiyne as linkers for the first time. The structures showed a well-defined width of ~4 nm and a length of hundreds of nanometers [97].

### 2.6.2. Two-Phase Methods (Interfacial Synthesis Utilizing Two Immiscible Liquids)

Atomic, ionic, or molecular compounds may be successfully applied as starting materials to the direct, bottom–up synthesis of ultrathin GDY nanostructures with in-plane periodicity. In 2017, Sakamoto et al. strived to create graphdiyne at the interface between two immiscible fluids (Figure 10A,B) [98].

The upper aqueous phase held copper (II) acetate and pyridine, which catalyzed ethynyl homocoupling (Eglinton coupling). The lower dichloromethane phase contained the HEB monomer. The continuous catalytic reaction for 24 h under an inert atmosphere at room temperature resulted in the development of a layered GDY (thickness: 24 nm; domain size: >25 μm). In 2019, Song et al. described the liquid/liquid interfacial formulation as a comprehensive way to obtain GYs via a reaction between terminal ethynyl groups and an aryl halide. The reactions catalyzed by PdCl2(PPh3)2 and CuI resulted in various forms of GY nanostructures, including hydrogen-substituted graphyne (H-GY), methylsubstituted graphyne (Me-GY), and fluorinated graphyne (F-GY) [99]. H-GY is a framework consisting of duplicated sections of benzene rings joined by ethynyl linkers at all meta sites. Likewise, the repeating units of Me-GY (or F-GY) are 1,3,5-trimethylbenzene (or 1,3,5-trifluorobenzene) rings attached to benzene rings within acetylene bridges. Twodimensional N-graphdiyne sheets were recently prepared via reactions conducted at the interfaces (Figure 10D,E) [100]. Nitrogen heterocycles (triazine and pyrazine) bearing terminal ethynyl groups were polymerized through Glaser coupling reactions at interfaces. This procedure was expanded to the synthesis of S-doped graphdiyne (TTF-GDY) structures comprising tetrathiafulvalene fractions (Figure 10C) and has this potential to be applied as a robust route for the synthesis of a wide range of heteroatom-rich graphyne-like structures in the future [101].

**Figure 9.** Homogenous coupling reactions in the preparation of graphyne derivatives (**A**). Synthetic scheme of the graphyne (**B**) and N-doped graphyne-like nanostructures (**C**). Redrawn [87,94]. (**D**) Intermolecular Glaser-Hay cross-coupling reactions of red fragments for GDY nanoribbons by copper(I) chloride, TMEDA, and acetone/tetrahydrofuran at RT (i); tetra-n-butylammonium fluoride and THF at RT (ii); intramolecular coupling reaction of blue fragments by Cu(OAc)2 and pyridine, H2O, and CH2Cl2 at RT (iii); R = triisopropylsilyl; TMEDA: *N*,*N*,*N* ,*N* -tetramethylethane-1,2-diamine; RT: room temperature. Redrawn [63,64].

**Figure 10.** Homocoupling reaction in interlayer between two phases: (**A**) liquid/liquid and gas/liquid (**B**) in the preparation of sulfur-rich graphdiyne (**C**) and N-graphdiyne (**D**,**E**). Redrawn [100,101].

2.6.3. Developments of Heterogeneous Coupling Reactions at Liquid/Solid Interfaces on Diverse Substrates

In liquid/solid interface reactions, substrates such as copper foil, plate, foam, and walls have been applied to bring reactants together and speed up the reaction procedure [102,103]. The first γ-GDY was prepared through the heterogeneous homo-coupling reaction of hexaethynylbenzene on copper foil, playing double roles of catalyst and substrate (Figure 11A) [28]. It was reported that both the oligomer evaporation process and the kinetics of the coupling reaction were strictly controlled by temperature [104,105]. Furthermore, the factors determining the structural properties of the obtained nanostructures were found to be catalyst distribution and monomer concentration. The formation of γ-GDY nanowalls via the Glaser-Hay reaction was successfully carried out in the presence of *N*,*N*,*N*',*N*'-tetramethylethylenediamine (TMEDA) due to its ease in complexing copper ions. The copper envelope catalysis strategy was employed for the synthesis of γ-GDY nanowalls on various substrates, including one-dimensional Si nanowires; twodimensional Au, Ni, and W foils; quartz; 3D stainless steel mesh; and 3D graphene foam (GY), as seen in Figure 11B [106]. For this purpose, these were the chosen substrates used in this method, and the target substrates were wrapped in a Cu-based envelope.

**Figure 11.** Proposed reaction process of GDY nanowall formation on Cu foil (**A**); envelope strategy for preparation GDY nanowalls on arbitrary substrates (**B**). Redrawn [35,106]; Glaser homocoupling reaction on Cu foil (liquid/solid method) (**C**). Redrawn [107–112]; boron-graphdiyne (B-GDY) preparation through homocoupling reaction on Cu foil (solid/liquid method) (**D**). Redrawn [113].

Well-defined films of triazine-based graphdiyne (TA-GDY), aminated-graphdiyne (NH2-GDY), β-GDY, H-GY, Cl-GDY, and F-GDY were prepared through Glaser homocoupling reactions on Cu foil (Figure 11C) [107–112]. 2,4,6-triethynyl-1,3,5-triazine, 2,4,6 triethynylaniline, tetraethynylethene (TEE), 1,3,5-triethynylbenzene, 1,3,5-trichloro-2,4,6 triethynylbenzene, and 1,3,5-triethynyl-2,4,6-trifluorobenzene were the respective starting compounds (prepared from the deprotection of trimethylsilyl group by tetra-butyl ammonium fluoride (TBAF) from corresponding silylated substrates) and were applied in homocoupling reactions to prepare extended GYs on Cu foil. The resulting films were peeled off from the Cu foil by a FeCl3-saturated solution and then rinsed with H2O, acetone, DMF, and ethanol. Wang et al. prepared boron-doped graphdiyne (B-GDY) through the aforementioned synthetic procedure, which was recognized to be an effective technique to prepare 2D carbonaceous nanostructures with strictly controlled and well-organized chemical structures (Figure 11D) [113]. In contrast to the copper foil, the copper nanowires (CuNWs) worked as templates and delivered more reactive sites for developing γ-GDY structures [114,115]. As a result, high-quality nanostructures with well-developed surface areas were obtained. For instance, the thin films of graphdiyne (average thickness: approximately 1.9 nm) were obtained on CuNWs (100 nm in diameter) [114]. In the end, polymeric films were isolated by washing a crude product in a mixture of hydrochloric acid and FeCl3.

### **3. Electronic Properties**

#### *3.1. Dirac Cone*

A Dirac cone is a distinctive feature in an electronic arrangement in which the energy levels of the valence and conduction bands meet at one specific point in the first Brillouin zone (named Dirac points), hence setting the Fermi level; the band structure in its vicinity resembles a double cone with linear dispersion [116]. The presence of a Dirac cone renders a given material "the zero-gap semiconductor" rather than metal and results in several unusual features such as ballistic electronic transport and enormous thermal conductivity. The occurrence of Dirac cones in graphene, predicted by Wallace in 1947 and experimentally demonstrated by Novoselov et al. in 2005, has sparked unceasing research in recent years [117,118]. Some efforts have been directed towards exploring the possibility of the occurrence of the Dirac cones in GYs. The presence of Dirac cones in *α*-graphyne, *β*-graphyne, and γ-graphyne-n with hexagonal symmetry structures was demonstrated. Recently, Vines et al. discovered that 6,6,12-GY, with rectangular symmetry, has two self-doped non-equivalent and distorted Dirac cones [119,120]. These results shed new light on the electronic properties of GY-like materials and suggest that rigorous hexagonal symmetry is not a feature that determines the appearance of cones in these materials.

## *3.2. Electronic Band Structure*

The electronic band structures of diverse GYs and GDYs have been investigated using theoretical methods, particularly density functional theory (DFT). Local density approximation (LDA) and the generalized gradient approximation (GGA) have been widely applied to study the structural, mechanical, electronic, and magnetic properties of GYs. Nevertheless, the underestimation of the band gap levels still is one of the critical problems of LDA and GGA. Hybrid Heyd-Scuseria-Ernzerhof-type functionals (HSE) have improved total energy evaluation by admixing the nonlocal Hartree-Fock exchange. They also lead to more realistic band gaps than LDA or GGA functionals. However, those conducted calculations generate significantly higher computational costs. To describe the van der Waals force in vdW-optPBE layered compounds, some extent of correction leads to improved results. Next, we review theoretical studies on GYs and compare diverse calculated parameters with the aforementioned theoretical approaches.

## 3.2.1. The Electronic Band Structure of GYs

The optimized geometry and electronic structures of diverse GY materials (graphyne, graphdiyne, graphyne-3, and graphyne-4) were computed by applying the full-potential linear combination of atomic orbitals (LCAO) approach by Narita et al. in 1998 [121]. The unit cell of all considered GY derivatives was similar to graphyne and is shown as a parallelogram in Figure 12. This unit cell was found to contain 12 carbons, the a and b lattice vectors were found to be equal (a = b), and the angle between them was found to be γ = 120◦. The Brillouin zone of the investigated material was an equilateral hexagon. As a result of geometry optimization, all bond angles are either 120◦ or 180◦ in graphyne-n. The lattice parameters of graphyne-n structures (*n* = 1, 2, 3, and 4) are 6.86, 9.44, 12.02, and 14.6 Å, respectively, and binding energies are 7.95, 7.78, 7.70, and 7.66 eV, respectively. It turned out that the reported hexagon presents a bond length almost equal to those found for graphite. Moreover, it is a bit longer than the bond that extends outside a hexagon. The bridges between hexagons are not formed by cumulenic linkers =C=C=; rather, they are formed by ethynyl ones (–C≡C–). The presence of conjugated multiple bonds is a typical feature of graphyne and its derivatives.

The indirect band gap is the separation between the conduction band minimum (CBmin) and the valence band maximum (VBmax) within an electronic band structure, whereas the direct band gap denotes the smallest of gaps at one particular point in the Brillouin zone [122]. Chen et al. studied the band structures of optimized α-, β- and γ-GY structures (Figure 12A) [123]. They showed that all investigated materials were direct semiconductors. In the case of α-graphyne, the conduction band minimum meets the valence band maximum at K-point in the Brillouin zone (Figure 12Ba) and symmetric Dirac cones are formed. In the case of β- and γ-GY, the energy bands coincide at M-points in the Brillouin zone (Figure 12Bb,c) and show quasi-Dirac cone structures, albeit with a slightly open band gap. The values of band gaps were found to be equal to 0 eV (α-GY), 0.028 eV (β-GY), and 0.447 eV (γ-GY) when applying the generalized gradient approximation of the Perdew–Burke–Ernzerhof (GGA-PBE) method (Figure 12Ba–c). Moreover, *γ*-GDY is a direct semiconductor with a Dirac point at the zone center (Γ point of the first Brillouin zone).

The estimated band gaps of graphyne-like families (monolayer, bilayer, multilayer, nanotubes, and nanoribbons) with and without strain by different functionals are compiled in Table 2. A wide range of band gap levels were found for *γ*-GY [121–129] and *γ*-GDY [42,106,121,125–127,130–134] (0.447–2.23 eV for *γ*-GY and 0.44–1.21 eV for *γ*-GDY), which is strictly related to the applied calculation functionals, as reviewed in Table 2.

The manipulation of the band structure of GYs has been attempted through strategies including strain tuning, structural engineering (the fabrication of GYs with different dimensions), doping, and the application of external electric fields. The obtained results suggest that GYs are valuable materials for nanoelectronic and optoelectronic devices or sensors. For example, strained graphyne-n was obtained by applying different types of strain, including homogeneous biaxial (H-strain in both the x and y directions) and uniaxial strains (x-direction A-strain and y-direction Z-strain), as was reported by Li et al. in 2013 (Figure 13a) [127]. For the H-strain case, hexagonal symmetry was preserved after the strain (Figure 13b). The A-strain deformation was found to be in the direction of the propagation of the ethynyl linkers, while the Z-strain described deformation perpendicular to the acetylenic linkers with bent hexagonal symmetry (Figure 13c,d). Unlike graphene, which exhibits a band gap insensitivity to applied strain, GYs have shown band gap modulations under different straining approaches. Comprehensive studies have proven that homogeneous tensile stress expands the band gap of GYs, whereas uniaxial tensile and compressive strains lead to band gap decreases (Figure 13e). Direct band gaps at either the M or S point of the Brillouin zone have been observed for both graphyne and graphyne-3 subjected to different tensile strains. In contrast, graphyne-4 and graphdiyne have been shown to display a direct band gap established at the Γ point, whatever the nature of the implemented strain [122,127]. Subsequent studies by Qui et al. confirmed the effect of

biaxial tensile stress in increasing the gap of γ-GDY within the range of 0.47–1.39 eV, while uniaxial tensile strain was found to decrease the band gap to approximately zero at the PW91 level [134].

**Figure 12.** Illustration of graphyne-n structures. The red parallelogram shows the unit cell of GYs. The Brillouin zone is shown in the upper-right, and various acetylenic linkers are shown in the bottom-right (**A**); (**B**) band gaps (Eg) and density of states (DOS) of α-graphyne (**a**), β-graphyne (**b**), and γ-graphyne (**c**). s and p describe the partial densities of states of the s and p orbitals in carbon atoms, respectively. The sum of both elements gives a value of the total density of states. Reproduced with permission [123]. Copyright 2018, MDPI.


**Table 2.** Calculated band gaps of GYs based on different summarized methods.

<sup>a</sup> Lattice constant of a = 9.44 Å and b = 6.90 Å; <sup>b</sup> lattice constants of a = 12.02 Å; <sup>c</sup> lattice constants of a = 14.6 Å; <sup>d</sup> AAA configuration presents metallic band structure; <sup>e</sup> AA stacked structure shows also metallic behavior.

**Figure 13.** The structure of GY film. The yellow parallelogram indicates the unit cell (**a**). Brillouin zone with high-symmetry points marked beneath H-strain (**b**), A-strain (**c**), and Z-strain (**d**). The band gap shift under different applied strains determined from GGA-PBE (**e**). Reproduced with permission [127]. Copyright 2013, American Chemical Society.

## 3.2.2. The Electronic Structure of GDYs

Heteroatom doping is an effective method to alter the band structure of graphynelike structures. The effect of the functionalization of sp-carbon atoms in γ-GDY on band structure was thoroughly investigated by Koo et al. in 2014 [130]. The hydrogenation of acetylenic linkers increased the band gap from 0.49 to a maximum of 5.11 eV, while fluorination increased the band gap up to 4.5 eV. Figure 14A–C illustrate different configurations for monolayered N-graphdiyne holding different numbers of C and N atoms varying from 24 to 42 entitled C18N6, C24N4, and C36N6, respectively, as reported by Singh in 2019 [144]. A hexagonal lattice was observed for C18N6 (a = 16.04 Å) and C36N6 (a = 18.66 Å), while C24N4 presented a rectangular unit cell (a = 15.97 Å; b = 9.67 Å). All considered crystals displayed semiconducting performance, with band gaps of 2.20 eV (C18N6), 0.50 eV (C24N4), and 1.10 eV (C36N6) [144].

The high-temperature treatment of carbonaceous materials is a conventional method of heteroatom doping. Until now, nitrogen-, sulfur-, and phosphorous-doped GDY derivatives have been prepared by this technique [65–69,145,146]. Chen et al. constructed X-doped graphdiynes by replacing a carbon atom (named C1, C2, and C3) (Figure 14Da) in GDY with heteroatom X, where X = B, N, P, and S. As a consequence, five models for N- (N1, N2, N3, pdN, and NH2), three models for each B- (B1, B2, and B3), P- (as P1, P2, and P3), and S- (S1, S2, and S3) doped GDY were constructed. The pdN and NH2 models represent the GDY structures doped with pyridinic nitrogen (Figure 14Db) and amino-derived functionalities (Figure 14Dc), respectively. The carbon atoms were divided into three classes that were labeled C1, C2, and C3 according to the arrangement of the structure. Additionally, nine adsorption sites, as indicated by red dots, were investigated in discussed studies.

The doping of pristine GDY with diverse elements (B, N, P, or S) has been found to result in a decrease in the band gap for all analyzed models when using spin-polarized DFT

computations and PBE functionals [146]. When the concentration of the dopant was about 1.4 at%, N- and P-doped GDYs were found to be metallic, while doping with boron and sulfur was found to reduce the band gap of starting GDYs from 0.46 to 0.16 and 0.28 eV, respectively. The further decrease in the band gaps was induced by an increase in dopant concentration. B-doped GDYs were found to become metallic after adding 5.6 at% of boron, whereas the band gap of S-doped GDYs decreased to 0.09 eV (Model S3). The planar structure of GDY after doping with B or N atoms was found to be preserved. Due to larger atomic radii of P and S atoms, the out-of-plane distortion of planarity, except for P3 and S3 models, was observed. In terms of the cohesive energies, N, P, and S elements were found to prefer the X2 position, while boron favored the X1 position.

In 2019, Yang et al. theoretically studied the changes in electronic properties caused by the adsorption of H2 and O2 atoms on GDY [147]. As Figure 14E shows, there were nine possible sites for their adsorption: three top sites (T1, T2, and T3), two hollow sites (H1 and H2), and four bridge sites. It was shown that the most stable adsorption positions are those located at T2 and H2. The adsorption of atoms on the GDY surface is related to the generation of distinguished high adsorption energies (GDY/H = 3.73 eV; GDY/O = 7.53). The latter meant that strong chemical interactions occurred between the adsorbed atoms and the surface. When H atoms were adsorbed, an inadequate 0.1 eV reduction in the band gap was observed. On the contrary, the adsorption of oxygen led to a higher band gap (an increase of 0.2 eV). Furthermore, the introduction of oxygen species to the structure resulted in weak magnetism because of the broken spin degeneration. With the advancement of research with single-layer GYs, researchers have turned their attention to other geometric shapes such as one-dimensional graphyne family members (e.g., nanoribbons and nanotubes) and three-dimensional ones represented by few-layer systems with different stacking arrangements.

**Figure 14.** N-graphdiyne nanosheets unit cells of C18N6 (**A**), C24N4 (**B**), and C36N6 (**C**). Reproduced with permission [144]. Copyright 2018, American Chemical Society. (**D**) The available position to be occupied by dopants (Da), pyridinic Ndoped GDY (pdN) (Db), and amino-group-doped GDY (NH2) (Dc). (**E**) Optimized H2 and O2 adsorption sites on pristine γ-graphdiyne. Redrawn [146,147].

3.2.3. The Electronic Band Structure of GY Nanoribbons

A nanoribbon (NR), as a 1D derivative of infinite GY sheets, is an example of tuning a band gap by changing the geometry of 2D monolayer GYs. Armchair (AGYNRs) and zig-zag (ZGYNRs) graphyne nanoribbons (differing in widths) were obtained by cutting through a graphyne (or graphdiyne) film along the x and y directions terminated to a benzene ring or acetylene group [140,142]. Gao et al. investigated nanoribbons with benzene rings at their edges. Figure 15 shows AGYNR and ZGYNR arrangements, where n is the number of recurring segments. In contrast to AGYNRs, the n number in ZGYNRs can vary by a half-integer (n + 1/2).

**Figure 15.** Armchair and zig-zag GY nanoribbons were achieved through the scissoring of the layer along the x and y directions. Redrawn [140].

The band gap of the nanoribbon decreased as the width increased because the ribbon tended to revert to a two-dimensional structure; in such a case, the expected band gap of the ribbon was 0.5 eV. The results confirmed that all considered nanoribbons showed semiconductive features, where band gaps were controlled by the widths and the nature of the edge. With the DFT-LDA method, the calculated energy gaps of AGYNR graphynebased nanoribbons were in the range of 0.59–1.25 eV, while ZGYNRs showed band gaps ranging from 0.75 to 1.32 eV. In the case of graphdiyne nanoribbons (GDYNRs), the energy gaps were in ranges of 0.54–0.97 eV and 0.73–1.65 eV for armchair (AGDYNRs) and zig-zag (ZGYDNRs) configurations, respectively. The effects of the edge arrangement (AGDYNTs and/or ZGYDNRs) of graphdiyne nanotubes (GDYNTs) on band structure were investigated in detail by Shohany et al. [139]. All GDYNTs under investigation exhibited semiconducting behavior, with a fundamental band gap ranging from 0.65 to 0.5 for AGDYNTs and from 0.95 to 0.55 for ZGDYNTs, as summarized in Table 2.

Kang et al. reported that an intersecting electric field could provoke the giant Stark effect in one-dimensional nanostructures, leading to a diminished or even disappeared band gap [148]. In comparison with other band gap modification approaches, the experimental control of an electric field seems to be a much easier way to play with GYs' band gaps. The effect of an electric field on the band structures of GDYNRs and the band gap decreased as

the electric field strength increased due to the strong localization of band-edge states. The band gap decreasing rate was found to be linearly dependent on the ribbon width.

#### 3.2.4. The Electronic Band Structure of Bulky GYs and GDYs

It was shown that 3D graphynes, thanks to the different stack arrangements, could be considered to be both as semiconductors and metals. In 2000, Narita et al. applied the first-principle calculations using a full-potential linear combination of the atomic orbitals method and local-spin-density approximation (LSDA) to optimize the geometry and investigate the electronic properties of one AAA stacking arrangement structure represented as α and three ABA configurations denoted as β1, β2, and β3 of three-dimensional graphyne (Figure 16a) [136].

**Figure 16.** ABA stacking arrangements of 3D graphyne. The A and B sheets are represented by grey and pink structures (**a**). Redrawn [136]. (**b**) Optimized arrangements of double-layer GDY labeled AB(β1) and AB(β2) (red top layer and grey bottom layer, respectively); (**c**) three potential forms of the trilayer GDYs: ABA(γ1), ABC(γ2), and ABC(γ3) arrangements (blue top layer, red middle layer, and grey bottom layer). Redrawn [138]. (**d**) AB-1, AB-2, and AB-3 represent structures of bulk GDY. Red arrows show the directions of the in-plane shift of two sheets in the cell. Redrawn [108].

In Table 3, it was shown that interactions occurring between orbitals 2pπ in α and β3 were greater than those appearing in β1 and β2. As a result, a large split of the π orbitals could be observed, consequently leading to the overlapping of conduction and valence bands in α and β3. The optimized plane lattice constant (a) and bond lengths were nearly the same as in 2D graphyne. The calculated interlayer distance (d) of graphyne (~3.3 Å) was longer than that of graphite (3.17 Å). However, the greater core/core repulsion between neighboring sheets in α and β3 resulted in lower binding energies compared with those found for β1 and β2. Furthermore, the obtained values were approximately 90% of the binding energy of graphite (8.867 eV/atom). This may suggest that graphyne is metastable while it is formulated. In contrast, the more stable nanostructures of β1 and β2 exhibited

semiconducting properties. The band gap of β1 was equal to 0.19 eV at the M and L points, whereas the energy gap of β2 was 0.50 eV at the L point. Thus, three-dimensional graphyne was assumed to have layered β1 and β2 organizations and semiconductive features with appropriate energy gaps.

**Table 3.** Optimized lattice constant, interlayer distance, binding energies, and band gap for 2D and 3D graphyne and graphite a.


<sup>a</sup> Using the projector-augmented wave method and the PBE functional.

In 2012, Zheng et al. conducted PW91 calculations and showed that the most stable bilayer and trilayer GDYs had their hexagonal rings arranged in the AB (direct Eg = 0.35 eV) and ABA (indirect Eg = 0.33 eV) configurations, respectively [138]. The decline in energy gaps, compared with monolayer species (0.46 eV), was caused by the occurrence of interlayer interactions. The application of an external electric field has been shown to be an effective technique to control the electronic and optical properties of few-layer graphyne-like materials. The two most stable arrangements for bilayer GDY (AB(β1) and AB(β2), owning band gaps of 0.35 and 0.14 eV, respectively) are depicted in Figure 16b. The stable trilayer configurations labeled ABA(γ1), ABC(γ3), and ABC(γ2) were found to present band gaps as high as 0.32, 0.33, and 0.18 eV, respectively (Figure 16c). The less stable AA and AAA configurations of GDY exhibited metallic properties.

In 2013, Nagase et al. employed a vdW-optPBE functional to investigate the relationships between the optical and electronic features of bulky GDY nanomaterials and their configurations. The obtained results were next compared with the results obtained using Heyd-Scuseria-Ernzerhof (HSE06) and LDA functionals [137]. They found that the AA configuration presented the lowest structural stability, accompanied by three AB configurations with energy gaps equal to 0.05, 0.74, and 0.35 eV, respectively. The investigations of Leenaerts et al. on the two-layer α-GY revealed that its band structure was qualitatively different from its single-layer derivative and was affected by the stacking modes of the two layers [149]. The AB staking arrangements exhibited a zero-gap feature similar to the AB configuration of bilayer graphene. It was shown again that electronic properties may be controlled by an applied electric field. The fluorinated GDY exhibited direct semiconductive behavior, with band gaps equal to 2.17 (AB stack-1) and 2.30 eV (AB stack-2), which were more than that of pristine GDY (~0.46 eV) (Figure 16d) [108]. Furthermore, the theoretically estimated band gap of the AB stack-1 configuration was consistent with the experimental value. The experimentally determined band gap of randomly fluorinated triazine-based graphyne by XeF2, reported by Szczurek et al., ranged from 3.12 to 3.34 eV and grew with increasing fluorine concentration [150]. An increase in the band gap upon the fluorination of acetylenic bridges is consistent with the decoupling of benzenic chromophores. Multilayer boron-graphdiyne (B-GDY) was comprehensively studied by Li et al. [113]. The calculated band gap energy of the monolayer showed that B-GDY was a direct band gap semiconductor at the Γ point with a value of 1.2 eV, which corresponded well to the experimentally derived band gap (1.1 eV) of the synthetic compound.

## *3.3. Electronic Transport*

The carrier mobility of different forms of GYs and GDys was theoretically predicted by Chen et al. In their experiments, the Boltzmann transport equation coupled with the deformation potential theory was applied to a-GY, b-GY, 6,6,12-GY, and GDY, and graphene was used as a reference. The obtained results revealed that almost all GYs and GDYs showed charge mobility values lower by one order of magnitude than graphene. 6,6,12-GY was an exception and presented a higher charge mobility (in the a direction) of around 25% for holes and 37% for electrons than the charge mobility found for graphene. Furthermore 6,6,12-GY presented a tremendous anisotropy of charge mobility along the *a* and *b* axes. The high carrier mobility of 6,6,12-graphyne might be explained by weaker electron–phonon coupling energy and longer relaxation times. Moreover, the rectangular arrangement of the 6,6,12-graphyne framework might cause the anisotropy of charge mobility in the structure [135].

In 2018, Nasri and Fotoohi theoretically investigated the electronic transport characteristics of a device built on N-doped (right electrode) and boron-doped (left electrode) α-armchair graphyne nanoribbons [151]. Four different devices differing in N and B substitution sites (sp or sp2) were proposed (Figure 17A). The current–voltage characteristics of the considered systems revealed effective non-linear behavior that led to the generation of a p–n junctions, which, in turn, resulted in rectifying behavior. The rectification properties, however, heavily depended on the deposition site of dopant atoms. The devices with doping atoms substituted on sp<sup>2</sup> sites (sp2–sp and sp2–sp2) showed a rectification ratio of around ten times lower than those having doping atoms attached to sp sites (sp–sp and sp–sp2) measured in the same bias region (Figure 17B). The rectifying behaviors of the described devices may be associated with asymmetric electrode arrangements. The N and B dopants caused crucial variations in electronic structures of α-graphynes, as they generated new sub-bands in valence (VB) and conductive (CB) bands. Those sub-bands seemed to ensure effective conduction and rectification in the described devices. This finding was supported by measurements conducted on pristine α-graphyne nanoribbons. The device built on the latter showed no rectification effect due to the symmetry of both electrodes. The suitability of graphyne and its h-BN (hexagonal boron nitride) derivatives (h-BNynes) to work as a field-effect transistor (FET) was theoretically investigated by Jhon et al. in 2014 [152]. With the aid of non-equilibrium Green's function combined with the density functional theory (NEGF–DFT) method, they examined the electronic transport of graphene–graphyne–graphene devices by varying graphyne size and carbon chain length.

**Figure 17.** *Cont*.

**Figure 17.** The scheme of devices built on an armchair graphyne nanoribbon (AGyNR) doped with boron and nitrogen atoms (**A**). The I-V curves of doped devices confronted with a pristine system; the inset figure shows the rectification ratio, RR (V), of all doped systems (**B**). Reproduced with permission [151]. Copyright 2018, Elsevier. The schemes of STM devices designed for the investigation SMC of carbobenzene, hexabenzocoronene, and carbo-n-butadiene (DBA). The molecules were attached via NH2 linkers to gold STM electrodes; TIPS: triisopropylsilyl (**C**). The single molecule conductance (SMC) and gate potential (VG) relationship found for the carbo-benzene sample. In this experiment trihexyl tetradecyl-phosphonium-bis(2,4,4-trimethylphenyl)phosphinate) was applied as the electrochemical gating electrolyte and the bias voltage was constant (0.1 V); SCE: saturated calomel electrode (**D**). Redrawn [153].

The concept of such constructed transistors was based on the outstanding electronic mobility of graphene and non-zero band gap graphyne, along with structural/compositional similarities between graphene and graphyne, being robustly connected between graphene electrodes. DFT calculations revealed that both graphyne and h-BNyne-based thin-film transistors (TFTs) showed good on/off ratios on the order of 102–103. Noteworthily, the size of such transistors might be reduced to below 1 nm while maintaining good switching features. Electronic orbital analysis disclosed that, contrary to h-BNynes, electrons in the conduction and valence orbitals were considerably delocalized in graphyne TFTs. The latter finding may suggest that graphyne TFTs could offer more facilitated electron mobilities than h-BNynes.

Single molecule conductance (SMC) was experimentally and theoretically investigated by Li et al. [153]. The measurements were performed using a scanning tunneling microscope, where the α-graphyne unit was chemically attached to the Au electrode and the STM tip. Amine (NH2) functionalities were used as an anchoring agent due to their ability to create molecular joints with negligible conductance aberrations (Figure 17C). The measurements conducted on α-graphyne and hexabenzocoronene representing the graphene molecular unit revealed that the α-graphyne units showed much higher conductance (106 nS, 1.9 nm) than the shorter units (1.4 nm), thus potentially being more conductive than hexabenzocoronene (14 nS). This unusual behavior originated from the electronic structure of both compounds. It turned out that the α-graphyne units had smaller HOMO-LUMO gaps than the graphene molecular units, so the transmission through the α-graphyne core was higher. The high molecular conductance of α-graphyne molecules also came from their rigid and planar structure. The measurements revealed that carbobutadiene wires showed a single molecular conductance that was 40 times smaller than that found for α-graphyne. The reason for this feature lies in the flexibility of the n-p-conjugated DBA framework, capable of adopting diverse geometries. The transport properties of different conformations of DBA molecules carried out with the NEGF-DFT model were strongly influenced by the measure of the rotation angle around the –C–C=C–C– sites. The increase in twist angle resulted in a higher HOMO-LUMO gap and thus a lower transmission between them. Finally, it was shown that the α-graphyne-based device showed excellent gating properties, understood as an increment of the SMC with increasing negativity of the gating potential (Figure 17D). Furthermore, the on/off ratio found for α-graphyne had an order of magnitude of ~15.

## *3.4. Optoelectronic Properties*

The optical features of N- and B-doped graphynes were theoretically studied by Bhattacharya and Sarkar [154]. The investigated structures presented, similar to pristine graphyne, optical anisotropy independent to the direction of applied electric field (Figure 17A, first column). The authors found that below an energy of 0.4 eV, the optical response was governed by the intra-band shift coming from free charges (Figure 18A, second column). The analysis of the static dielectric tensors revealed that doped graphynes showed a better electric conductivity and higher mobility of charges than pristine GY, creating the opportunities for their application in optoelectronics. The spin-polarized optoelectronic properties of α-graphyne were theoretically investigated by Yang et al. using NEGF–DFT [155]. They showed that photocurrents were generated by irradiating investigated devices with different light wavelengths (from UV to IR), and the polarization of the formed photocurrents strictly depended on the type of contact applied. Both M1 and M2 devices (Figure 18B) produced spin-down photocurrents, whereas the M2 device could also generate spin-up ones (Figure 18C). It was revealed that photocurrents might also be guided by reversing the electrodes' magnetization. Generally speaking, two spatially separated spin photocurrents appeared on the antiparallel polarized electrodes. Functionalization is another approach to manage the optoelectronic characteristics of GY and GDY nanostructures. Theoptoelectronic application of Gdy:ZnO nanocomposites was experimentally investigated by Jin et al. [156] Their experiment involved measuring I–V characteristics in the dark and under the influence of UV radiation. When the samples were illuminated, an increase in current due to light absorption was observed. Chronoamperometry was employed to record the rise/decay time of the current measured without and with light irradiation. The obtained results revealed that the current changes strongly depended on the photochemical character of investigated devices. The junction created between the GDY and ZnO nanoparticles strongly enhanced the charge transfer between these components and, consequently, the photoresponse. The GDY:ZnO/ZnO system manifested an excellent optical response of 1260 A W−1, with a rise/decay time as short as 6.1/2.1 s. The device built only on zinc oxide nanoparticles held a responsivity of 174 A W−<sup>1</sup> and a much more extended rise/decay time (32.1/28.7 s). Li et al. proposed deep UV photodetectors (Figure 18D) based on TiO2:GDY nanocomposites [157]. The published results proved that the considered composites were suitable as UV photodetectors, disclosing an outstanding optical response of 76 mAW−<sup>1</sup> and a rise/decay time of 3.5/2.7 s. The cited examples show the promising and efficient properties of graphyne derivatives. Their high charge mobility, on/off ratio, and photoresponse makes them prospective materials for constructing nanoscale electronics and optoelectronics.

It is noteworthy that metal oxide dopants in GDY nanocomposites, the optoelectronic amplifiers, do not seem to be mandatory additives. Zhang et al. constructed metalfree, flexible photodetectors built on GDY:PET composites. Those built devices were characterized by excellent mechanical, electronic, and optoelectronic properties. The experimentally investigated responsivity and photocurrent reached outstanding values of 1086.96 μA W−<sup>1</sup> and 5.98 μA cm−2, respectively. Moreover, the considered composites showed good photoresponses, even after undergoing 1000 bending and twisting cycles. The recorded loss in photocurrent was 25.6% (bending force) and 35% (twisting force) [158].

**Figure 18.** The 1B (**Ai**) and 1N (**Aii**) doped GYs' relationships between the imaginary part of the dielectric function and photon energies taken for different directions of electric fields. The reflectivities of inter- and intraband transitions found for 1B (**Aiii**) and 1N (**Aiv**) doped GYs. Reproduced with permission [154]. Copyright 2016, American Chemical Society. Scheme of the designed devices (**B**). The parallel and antiparallel spin-polarized photoresponse of M1 (**Ca–d**) and M2 (**Ce-h**) devices. Reproduced with permission [155]. Copyright 2017, IOP Publishing Ltd. I–V sweeps of the considered photodetectors working without (**Di**) and (**Dii**) with UV light (254 nm) irradiation. The on/off features of the TiO2:GDY/MZO bilayer PD (**Diii**) and the TiO2/MZO bilayer PD (**Div**). Reproduced with permission [157]. Copyright 2020, Elsevier.

## **4. Magnetism of Pure and Doped Graphyne-Like Materials**

## *4.1. Theoretically Investigated Magnetic Properties*

Local defects and substituents in carbonaceous materials might provoke superconducting or ferromagnetic features that can even appear at ambient temperature. Until now, magnetism in carbon has been experimentally revealed for (i) interacting radicals, (ii) carbons with a mixed hybridization (sp<sup>2</sup> + sp3), (iii) amorphous carbonaceous materials doped with trivalent atoms (P, N, or B), (iv) diverse nanostructures (graphite, diamond, and foams), and (v) fullerenes [159–162]. Apart from that, vicinity to metallic ferromagnets or the contamination of transition metals (Fe, Cr, or V) has been shown to generate spin polarization in carbon structures [159,160,162]. Localized magnetism and the zig-zag rib of graphene nanoribbons or neighboring vacancies and dangling bonds in pure carbon structure are representative of this concept. The theoretical considerations and fruitful synthesis of monolayer graphyne have led to considerable interest in research on the magnetic properties of pure or/and doped graphyne structures.

#### 4.1.1. Metal-Doped GYs

The magnetic features found in materials with electrons occupying only *s* and *p* orbitals, rather than the traditional *d* or *f* ones, might be remarkably appealing to spintronic

applications. However, the source of magnetism in pure carbonaceous materials is not fully understood. In this section, we report works addressing the magnetic properties of GYs. The effect of dopant distribution and functional groups on adjusting the electronic or magnetic properties of GYs might lead to new promising electronic, optoelectronic, and spintronic devices. The calculations revealed that unmodified γ-graphyne is a nonmagnetic semiconductor [128], while the adsorption of transition metal (TM) atoms might drastically change the electronic structure of and add ferromagnetic features to GY nanostructures [163,164]. In 2012, He et al. theoretically investigated (DFT + U) the electronic structure and magnetism of GDY and GY doped with single 3*d* transition metals (V, Cr, Mn, Fe, Co, and Ni) [164] The adsorption of metal atoms on the GDY and GY surfaces generated charge transfer between metallic adatoms and polymeric sheets. Moreover, the adsorption might generate electron redistribution in the *s*, *p*, and *d* orbitals of transition metal atoms. Except for vanadium, the mentioned factors caused decreases in the magnetic moments of adsorbed metals (Cr, Mn, Fe, and Co). Additionally, they were ranked as follows: Cr > Mn > V > Fe > Co for TM-GDY and Cr > V > Mn > Fe > Co for TM-GY; see Figure 19A. The energy of spin polarization (ΔEspin), taken as the difference between the nonmagnetic and magnetic states, was higher than 1.1 eV, suggesting the substantial stability of the spin-polarized states of transition for metal-doped GDY and GY nanostructures.

In 2017, Lee et al. applied DFT calculations to investigate the doping efficiency of the transition metals of 3*d*, 4*d*, and 5*d* groups deposited on a γ-GY surface [165]. For this purpose, different high-symmetry adsorption sites such as top (T), bridge (B), and hollow (H) were chosen, as depicted in Figure 19B. It was shown that adatoms typically occupy the H1 sites over ethynyl rings of γ-graphyne (taking Fe (μB: ~2.08) as an example). The lower atomic radii of Co (μB: ~1), Ni (μB: ~0), and Cu (μB: ~0), as well as Re (μB: ~1), caused those dopant atoms to be preferably placed at the H3 sites of the acetylenic rings. Magnetic moments of metals of the 4*f* group were found to be higher in comparison with 3*d* transition metals, and lanthanides with sufficiently large atomic radii are good candidates to be introduced in GY rings. Ren et al. used comprehensive first-principle calculations to study the magnetic properties of *β*-GY doped with different rare-earth (RE) atoms (La, Ce, Pr, Nd, Pm, Sm, and Eu) [166]. The β-GY was found to undergo a transition from semiconductor to metal. The introduction of external atoms such as neodymium, promethium, samarium, and europium (local magnetic moments in the range of 4.1–7.3) were able to translate into higher values of magnetic moments for metal-graphyne complexes (>4.1 μB). As expected, the carbon atoms neighboring dopant atoms were found to have a modest contribution to generated magnetic moments. In 2014, Alaei et al. studied two zig-zag graphyne nanotubes (ZGYNTs) and two armchair graphyne nanotubes (AGYNTs) doped with iron, cobalt, and nickel [167]. It was shown that a 12-membered ring (12-C), a hollow site surrounded by acetylenic linkers in GY, was the most preferable (the most stable) site for the deposition of those metals. The adatoms nested in the plane of the ethynyl rings and formed bonds with adjacent carbons. Complexes of Fe (μB: ~2.06) and Co (μB: ~1) with different GYNTs were magnetic and showed many features typical of metals, semimetals, half semimetals, and half-semiconductors [168]. Ni complexes (μB: ~0.01) were found to be nonmagnetic semiconductors exhibiting energy gaps narrower than those found for starting nanotubes (Figure 19A).

## 4.1.2. Non-Metal Doped GYs and GDYs

The electronic properties and magnetism of GYs can be also driven by applying diverse non-metal doping agents. In 2014, Drogar et al. stated that widening the band gap (~2 eV) and provoking a magnetic moment (~1 μB) of α-GY could be realized via the simple hydrogenation reaction of the latter due to the cleavage of π-bonds and the creation of unpaired electrons [169]. Subsequently, in 2018, Wang et al. studied the tuning of monoand bilayer GY features after the hydrogenation of different carbon atoms in GY [170]. Unlike the distribution of μB in TM-GY, in which the magnetism derives from the d- or f-electrons, the unpaired 2p electrons essentially provide the magnetic moments at the

non-hydrogenated carbons of GY. As indicated in Figure 19C, there are two sites (C1 (T) and C2 (B)) on which hydrogenation can occur. Twelve hydrogen atoms are located at Ti sites in the aromatic ring and the B1 sites in acetylenic linkers of GY nanosheets. It was observed that the magnetic moment of the monolayer reached a maximal value (1.59 μB) for three hydrogen atoms, whereas the magnetic moments progressively declined to 0 μB when the number of hydrogen atoms varied from 4 to 6. These results suggest that the hydrogenation of half C atoms (sp- or/and sp2-hybridized) may result in a maximal magnetic moment, similarly to what was found for graphene [171]. Theoretical calculations of the total magnetic moments of bilayers with different stacking arrangements yielded values larger than 1.52 μB. DFT is the most frequently utilized computation method to investigate the electronic properties and magnetism of single-layer graphdiyne (GDY) doped with nonmetallic elements, such as boron, nitrogen, oxygen, phosphorous, or sulfur [172]. The position of dopant atoms has a profound impact on the relevant characteristics of GDYs. Three possible places for doping are aromatic carbons (X-b) and two carbons in ethynyl linkers (X-1 and X-2), depicted in Figure 19D. Considering the cohesion energies found for the studied structures, one can conclude that boron and sulfur prefer to exchange the sp<sup>2</sup> aromatic carbons. In turn, the N, O, and P atoms favor substituting the carbons in the ethynyl bridges. It was shown that nitrogen atoms deposited on both GDYs and GY monolayers did not modify their magnetic properties, and all of them remained nonmagnetic [154]. In contrast, GDY structures doped with B, O, P, or S (structures X-1 and X-2) deposited at acetylenic chains offered spin polarization and were magnetic (μB varying from 0.31 to 1.26 depending on the type of dopant and applied functionals). The deposition of these elements at the aromatic site did not alter magnetic features, and doped GDYs were nonmagnetic.

**Figure 19.** The magnetic moment (μB) and ΔEspin (**A**). Reproduced with permission [164] Copyright 2012, American Chemical Society. The adsorption sites (green spheres) of transition metal atoms on a γ-graphyne film seen from the top (upper image); the in-plane case of H3 and its relation to the H1 site seen from a side (bottom image), blue points represent carbons (**B**). Redrawn [165]. Possible sites on GY for hydrogenation (**C**). Redrawn based on [171]. Possible site for doping by X = B, N, O, P, and S atoms (**D**). Reproduced with permission [172]. Copyright 2020, Elsevier.

## *4.2. Experimentally Investigated Magnetic Properties*

Following comprehensive theoretical studies on the magnetic properties of modified GYs, efforts have been directed to experimentally explore and measure magnetism in GYs [173]. In 2017, Huang et al. investigated the impact of N doping on the paramagnetic properties of GDYs and demonstrated the crucial role of nitrogen atoms deposited on the benzene ring in building the local magnetic moment [174]. Based on the M-H curves (Figure 20A,B) measured for GDY and N-GDY at 2 K, the authors showed that N-GDY and GDY presented low magnetization values of 0.96 and 0.51 emu/g, respectively. The magnetization obtained for N-GDY containing 5.29% of N atoms was on the same level as the value found for fluorinated RGO [175]. These results contradicted recent theoretical findings, which claimed that N in a chain or ring is non-paramagnetic. As such, more research is required to find the origination of paramagnetism [154,172].

**Figure 20.** M-H curves obtained at 4.2 K (**A**) and 2 K (**B**) for GDY and N-GDY. Obtained results were fitted to the Brillouin function with J = 1/2 (solid line) and J = 1 (dashed line) (**B**). Spin-resolved DOS of N-GDY films with nitrogen atoms deposited on the benzene ring in GDY. The upper panel (**C**,**D**) shows the paramagnetic S = <sup>1</sup> <sup>2</sup> system, while the bottom one the antiferromagnetic one (**E**,**F**). Reproduced with permission [174]. Copyright 2017, Springer Nature.

The spin-polarized calculations proved that a distinct local magnetic moment (μ<sup>B</sup> = 0.98) originates from nanostructures with deposited asymmetric pyridinic nitrogen (Py-1N), seen in Figure 20C,D. The structures bearing symmetric pyridinic nitrogen substitution (Py-2N) or N atoms attached to ethynyl linkers appeared to be nonmagnetic (Figure 20E,F). Finally, the investigated systems did not show any ordered ferromagnetic or ferrimagnetic properties.

Further research aiming to discover the relationship between doped heteroatoms and magnetization in GY structures continued. In 2017, Zheng et al. [176] reported that, contrary to pyridinic-N, the vacancy, carbonyl, and hydroxyl functionalities of GDY contribute to the magnetic properties of thermally treated GDYs. Furthermore, DFT calculations indicated that the OH groups at the chain of the GDY layer are a considerable source of unpaired electrons and may favor antiferromagnetism in annealed GDY. Moreover, the annealing of GDY at 600 ◦C (GDY-600) may lead to complex magnetic properties, depending on the applied measurement temperatures. Paramagnetic characteristics could be noticed below 50 K, and a hump seen in the range of 50–200 K demonstrated that both paramagnetic and antiferromagnetic phases coexist in the considered materials.

In 2019, Huang et al. investigated the effect of sulfurization on the induction of ferromagnetic characteristics into GDYs. They found that S-doped GDY presented strong residual magnetization (>0.047 emu/g) at ambient temperatures. The investigated systems were characterized by the transition temperature being close to 460 K [177]. The local magnetic moment and electron interactions occurring between C and S atoms were found to be responsible for the appearance of the ordered internal ferromagnetism in the investigated materials. The ferromagnetic behavior was also confirmed by magnetic hysteresis (M-H) loop measurements with different temperatures. In order to investigate the influence of sulfur doping on magnetic properties of GDY, temperature-dependent magnetic susceptibility χ-T curves (the applied magnetic field H = 500 Oe) and magnetization M-H curves for GDY350 and S-GDY by VSM were measured (Figure 21a–d) [177]. These findings are very promising in terms of S-doped GDY suitability in magnetic storage devices.

**Figure 21.** The χ-T relationship examined in a temperature range of 2–300 K (**a**) and an M-H curve taken at 2 K for GDY 350 (**b**). Temperature–magnetization relationship, where FC is field cooled mode and ZFC–zero field cooled mode (**c**) and hysteresis loops recorded for S-doped GDY (**d**). Reproduced with permission [177]. Copyright 2019, American Chemical Society.

#### **5. Mechanical Properties of Graphynes**

Graphene with an intrinsic tensile strength of 130 GPa, and Young's modulus equal to 1TPa is recognized as the strongest material ever tested [178]. Zhang et al. theoretically investigated mechanical properties of different forms of GYs (*α*-, *β*-, *γ*-, and 6,6,12-GYs), as well as graphene used as a reference [179]. The tensile stress (~125 GPa) and Young's modulus (0.99 TPa) obtained by molecular dynamic calculations confirmed the extraordinary mechanical resistance of graphene. In contrast, all considered graphynes presented 50–70% lower mechanical strength than graphene, whereas the decline of tensile stress and Young's modulus of graphynes was strictly related to an increasing amount of acetylenic linkage. A similar dependence of elastic properties on the proportion of acetylenic bridges in *α*-, *β*-, *γ*-graphynes was also observed by Hou et al. in 2014 [123]. Noteworthy, a deviation from the hexagonal structure of graphene and graphynes, as it happens for 6,6,12-graphyne, resulted in the appearance of anisotropy of mechanical properties measured along with x (zig-zag) and y (armchair) directions [179,180]. The anisotropy in mechanical and electronic properties makes 6,6,12-graphyne useful in diverse potential applications.5.

## **6. Conclusions and Outlook**

Synthetic methods for the preparation of graphyne-like structures doped with heteroatoms and controllable size and dimensions were collected and summarized. Doping (i.e., the replacement of carbon atoms or covalent bond formation with foreign atoms) methods such as heating and annealing techniques have crucial drawbacks in the preparation of carbon materials; uncertainty in the values and position of heteroatoms, along with the destruction of the intrinsic properties of pristine materials, are inevitable disadvantages. To avoid the abovementioned flaws, engineering the surface of GYs to include desired heteroatom is necessary. Different methodologies have been successfully applied with high reproducibility to prepare a wide range of GYs with ordered structures in which heteroatom occupy predictable positions on the surface. The adjustment of morphology (such as 1D, 2D, and 3D) and composition (doping with N, P, F, S, Cl, H, O, B, etc.) are influential approaches to modulate the band structures and (subsequently) electric, optical, and magnetic properties of graphyne-like structures. The size- and composition-dependent band structure, electronic transport, spin-polarized optoelectronic properties, or photosensitivity, and appearing magnetic moment of these allotropes of carbon material have been investigated. There is an understanding that research into the magnetic properties of carbon materials is still in its infancy stage, and there is a substantial gap between calculations and experimental findings. Though this topic needs many more firm, particularly experimental discoveries, it is already clear that this type of material could be exploited in spintronics technology. Benefiting from the high specific capacity, remarkable cycle performance, and inflated adsorption capacity of ions and gases, GYs could also have a promising future for applications in electrochemical energy storage such as batteries, capacitors, and hydrogen fuel cells. Theoretical research has proved GY's usefulness for not only understanding but also predicting the structures and properties of new modified systems as well as for preselecting those that most merit experimental study.

**Author Contributions:** G.A. writing—original draft preparation; A.A. writing—original draft; W.G. writing—review and editing, and supervision; A.S. writing—review and editing, project administration, and funding acquisition. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Science Centre (NCN), Sonata Bis project (2016/22/E/ST5/00529).

**Acknowledgments:** *In memoriam* to Andrzej Maria Huczko (1949–2021).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Cationic Pollutant Removal from Aqueous Solution Using Reduced Graphene Oxide**

**Talia Tene 1, Stefano Bellucci 2, Marco Guevara 3,4, Edwin Viteri 5, Malvin Arias Polanco 6,7, Orlando Salguero 7, Eder Vera-Guzmán 7, Sebastián Valladares 7, Andrea Scarcello 7,8,9, Francesca Alessandro 7,8, Lorenzo S. Caputi 7,8 and Cristian Vacacela Gomez 3,7,\***


**Abstract:** Reduced graphene oxide (rGO) is one of the most well-known graphene derivatives, which, due to its outstanding physical and chemical properties as well as its oxygen content, has been used for wastewater treatment technologies. Particularly, extra functionalized rGO is widely preferred for treating wastewater containing dyes or heavy metals. Nevertheless, the use of non-extra functionalized (pristine) rGO for the removal of cationic pollutants is not explored in detail or is ambiguous. Herein, pristine rGO—prepared by an eco-friendly protocol—is used for the removal of cationic pollutants from water, i.e., methylene blue (MB) and mercury-(II) (Hg-(II)). This work includes the eco-friendly synthesis process and related spectroscopical and morphological characterization. Most importantly, the investigated rGO shows an adsorption capacity of 121.95 mg g−<sup>1</sup> for MB and 109.49 mg g−<sup>1</sup> for Hg (II) at 298 K. A record adsorption time of 30 min was found for MB and 20 min for Hg (II) with an efficiency of about 89% and 73%, respectively. The capture of tested cationic pollutants on rGO exhibits a mixed physisorption–chemisorption process. The present work, therefore, presents new findings for cationic pollutant adsorbent materials based on oxidized graphenes, providing a new perspective for removing MB molecules and Hg(II) ions.

**Keywords:** graphene oxide; reduce graphene oxide; dyes; heavy metals; pollutant removal

## **1. Introduction**

Water pollution is one of the world's most serious problems due to a large amount of wastewater being produced and poured into the water bodies every year [1]. Among different types of wastewaters, water contaminated with dyes and heavy metals deserves significant attention. With the continuous development of industrialization processes, large amounts of dyes or heavy metals are released into the environment, making water unsafe for human use and disrupting aquatic ecosystems [2–4]. A large amount of dye

**Citation:** Tene, T.; Bellucci, S.; Guevara, M.; Viteri, E.; Arias Polanco, M.; Salguero, O.; Vera-Guzmán, E.; Valladares, S.; Scarcello, A.; Alessandro, F.; et al. Cationic Pollutant Removal from Aqueous Solution Using Reduced Graphene Oxide. *Nanomaterials* **2022**, *12*, 309. https://doi.org/10.3390/ nano12030309

Academic Editor: José Miguel González-Domínguez

Received: 27 December 2021 Accepted: 13 January 2022 Published: 18 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

pollutants in the wastewater brings a huge risk to both aquatic organisms and humans because they can reduce sunlight transmission and normally contain toxic substances such as heavy metals [5], causing mutagenicity, carcinogenicity as well as the dysfunction of the kidney, liver, brain, reproductive system, and central nervous system [6]. On the other hand, agricultural processes or mining activities have increased the concentration of heavy metals in water around the world, which has led to immediate legislation by various governments [7]. When these toxic metal ions enter the food chain and then the human body, they accumulate in an organ above the allowed limits, originating serious health-related diseases, for instance: skin irritation, vomiting, stomach cramps, and cancer of the lungs and kidney [8]. Additionally, infants may have delays in physical or mental development, children may have deficits in attention span and learning activities, and adults may have high blood pressure [9].

To remove dyes or heavy metals, different physicochemical and electrochemical methods have been proposed. Physicochemical processes include membrane filtration [10], ion-exchange [11], and adsorption [12]. Electrochemical processes include electrocoagulation [13], electroflotation [14], and electrodeposition [15]. Among all these possible methods, including on-site sensing ones [16,17], those cost-effective, environmentally friendly and no further pollutant features are required. Therefore, adsorption is one of the most prominent approaches to water and wastewater decontamination. Due to this fact, many adsorbents with different structural conformations and compositions have been prepared or modified, for instance, clays/zeolites [18], biosorbents [19], agricultural solid wastes [20], and industrial by-products [21]. In practical operations, activated carbon is one of the most used adsorbents, thanks to its excellent adsorption performance [22]. However, activated carbon has been limited by the high cost and the complicated regeneration processes. To tackle the aforementioned limitations, carbon-based nanomaterials, such as carbon nanotubes [23], graphene oxide (GO) [24], and reduced graphene oxide (rGO) [25], have been proposed.

Recently, rGO has attracted increased interest as an effective adsorbent for dyes or heavy metal ions [25]. While the properties of rGO are quite different from those of pristine graphene or beyond-graphene materials (i.e., unique electronic, optical, mechanical, plasmonic, and thermal properties [26–29] as well as promising applications in hybrid capacitors [30]), rGO is characterized by interesting hydrophilic and semiconducting properties [31]. Nowadays, the synthesis of rGO follows an eco-friendly and cost-effective preparation method that can be used for large-scale water treatment technologies [32,33]. Moreover, the presence of oxygen functional groups (mainly hydroxyl and epoxide groups [34]) on the rGO surface allows covalent modifications with strong chelating groups, which, in turn, present a high affinity to metal ions or more complex organic/inorganic molecules.

There is extensive literature on the use of extra functionalized rGO or GO (e.g., GONR [35], S-GO [36], GO-TSC [37], S-doped g-C3N4/LGO [38], GSH-NiFe2O4/GO [39], HT-rGO-N [40]) for treating water and wastewater; however, the use of pristine rGO is scarce (and sometimes unclear) when removing cationic heavy metals and cationic dyes from aqueous media. In this work, such a comparative study is presented, considering methyl blue (MB) as a cationic dye and mercury (II) (Hg(II)) as cationic heavy metal. These cationic pollutants have been selected, in particular, because MB can cause various poisoning problems and methemoglobinemia [41] and Hg (II) can cause substantial neurodevelopmental risk in fetuses, newborns, and children [42]. Various kinetics, isotherms, and thermodynamic studies are carried out to demonstrate the adsorption of MB and Hg(II) on as-made rGO. Additionally, the present work includes the synthesis of adsorbent material and the corresponding morphological and spectroscopical characterizations of raw graphite, GO, and rGO.

## **2. Materials and Method**

Although there are several methods for synthesizing graphene and its derivatives (e.g., liquid exfoliation [43,44], zeolite-assisted exfoliation [45], hydrothermal exfoliation [46], and microwave-assisted exfoliation [47–53]), the present study focuses on the eco-friendly oxidation-reduction protocol [33] to transform as-made GO into rGO.

#### *2.1. Materials*

All chemicals were used as received, without further purification:


#### *2.2. Preparation of Oxidized Graphenes*

A round-bottom flask was charged with graphite (1.5 g), H2SO4 (35 mL), maintaining a uniform and moderate circular agitation. The mixture was located in an ice-water bath, and then KMnO4 (4.5 g) was slowly added. The resulting mixture was agitated on a stirring plate while adding 75 mL of distilled water, being careful not to exceed 363 K.

Additionally, 250 mL of distilled water was added, followed by 7.5 mL of H2O2. The resulting solution was distributed to be washed by centrifugation with HCl solution and distilled water several times to adjust the pH~6 [33] and then dried (drying stove, 60 Hz, 1600 W) at 353 K for 2 h to obtain graphite oxide flakes.

After the oxidation process, 50 mg of graphite oxide flakes were dispersed in 500 mL of distilled water by sonication for 0.5 h [25]. The resulting solution was centrifuged to separate GO from non-exfoliated graphite oxide particles [32]. Under agitation, 1.0 g citric acid (CA) was added to the centrifuged suspension, setting the reduction temperature at 368 K. The precipitated rGO was collected, washed with distilled water by centrifugation, and dried at 353 K for 2 h to obtain rGO powder [33]. The resulting rGO was used for the cationic pollutant removal.

#### *2.3. Characterization of GO and rGO*

The surface morphology of raw graphite, GO, and rGO were carried out on a transmission electron microscope (TEM, JEM 1400 Plus, Akishima, Tokyo, Japan) operating at 80 kV, and a scanning electron microscope (SEM, JSM-IT100 InTouchScope, Akishima, Tokyo, Japan) equipped with a JEOL dispersive X-ray spectrometer (EDS) (Billerica, MA, USA), with the accelerating voltage of 15 kV. Raman spectra of graphite and oxidized graphene were obtained using a Jasco NRS-500 spectrometer (Oklahoma City, OK, USA), with a 532 nm laser wavelength (0.3 mW, 100X objective). Infrared spectra were collected using a Fourier transform infrared spectrometer (Jasco FT/IR 4000, Oklahoma City, OK, USA). UV-visible measurements were recorded using a UV–vis spectroscopy (Thermo Scientific, Evolution 220, Waltham, MA, USA). X-ray diffraction measurements were performed using an X-ray diffractometer (PANalytical Pro X-ray, Malvern, UK) in the diffraction angle (2θ) window of 5–90◦ using Cu Kα irradiation under the acceleration voltage of 60 kV and a current of 55 mA. The thermal stability of GO and rGO was examined using thermogravimetric analysis (TGA, PerkinElmer simultaneous thermal analyzer, STA 6000, Waltham, MA, USA).

SEM samples were mounted on aluminum substrates with adhesive, coated with 40–60 nm of metal such as Gold/Palladium and then observed in the microscope. TEM samples were arranged by drop-casting onto formvar-coated copper grids once the samples were cut into very thin cross-sections, allowing electrons to pass directly through the sample. Similarly, Raman samples were deposited directly by drop-casting on glass substrates and dried for a few seconds with the incident beam. To record the UV-visible spectra in the

window range 190–1000 nm, the samples were redispersed in distilled water by midsonication for 5 min.

## *2.4. Preparation of MB Solutions and Experimental Set-Up of MB Adsorption on rGO*

MB was dissolved in ultra-pure water to obtain a stock solution of 1000 mg L−1, and the working solutions were used in the test through serial dilutions. The pH of the solutions was adjusted using HCL and NaOH and controlled by a pH meter (HI221 Hanna Instruments).

The adsorption experiments were carried out in triplicate. In total, 500 mg of rGO was added into 250 mL of the MB solution with a concentration of 100 mg L−<sup>1</sup> to evaluate the adsorption kinetics and contact time effect (batch test) [25]. The resulting mixture was agitated up to 60 min at 298 K. Adsorption isotherms were obtained from batch experiments by adding 200 mg rGO in 50 mL of MB solutions, considering different concentrations in the range of 10–100 mg L−<sup>1</sup> and three different temperatures (298, 313, and 333 K). pH studies were carried out by adding 200 mg rGO in 100 mL of MB solution with a concentration of 100 mg L−1. Various aliquots were extracted from the solution to be evaluated by UV–vis spectroscopy. In all adsorption experiments (except in the study of the effect of pH), the pH was fixed to 6.02 ± 0.07. To investigate the pH effect, the pH of MB solutions was adjusted by HCl (0.1 M) and NaOH (0.1 M), and immediately, rGO was added.

#### *2.5. Preparation of Hg(II) Solutions and Experimental Set-Up of Hg(II) Adsorption on rGO*

rGO was placed in a dilute aqueous solution of HgO. The adsorption kinetics studies were performed by adding 300 mL aqueous HgO (pH = 6.41 ± 0.05) to a falcon tube. Then, 200 mg rGO was added to form a slurry. The mixture was stirred at 298 K for 60 min. After that, the mixture was filtered at intervals through a 0.45 mm membrane, and then the filtrated samples were analyzed by using an AAS-cold vapor to determinate the remaining Hg(II) content (standard methods 3112-B; 3111-B.4b) [13]. Adsorption isotherms were obtained by adding 2.5 mg rGO to each falcon tube containing 50 mL of HgO solution with different concentrations from 10 to 100 mg L−<sup>1</sup> and considering three different temperatures (298, 313, and 333 K). The resulting mixtures were stirred at room temperature for 30 min and then filtered separately through a 0.45-mm membrane filter. The filtrates were analyzed by using AAS-cold vapor to determine the remaining Hg(II) content (standard methods 3112-B; 3111-B.4b) [13]. The pH effect was investigated by adjusting the pH of HgO solutions with HCl (0.1 M) and NaOH (0.1 M) at room temperature, and immediately, rGO was added.

## **3. Results and Discussion**

#### *3.1. Characterization of Graphite*

Graphite has a crystal structure made up of stacked graphene layers in which the separation distance of the layers is 3.35 A˚ , whereas the separation of atoms within a layer is 1.42 A˚ . At the microscale, the starting (powder) graphite shows an irregular bulk structure with a lateral size ranging from 2 μm to 50 μm (Figure 1a). The XRD pattern of raw graphite is shown in Figure 1b. The most intense peak at 2θ = 26.73◦ corresponds to the graphite stacking crystallinity (002) [43]. The less intense peak at 2θ = 55.82◦ displays the long-range order of stacked graphene layers (004) [43].

Figure 1c,d show the Raman spectrum of raw graphite. The main features observed are: (i) the absence of the D peak demonstrating a defect-free starting graphite, (ii) the G peak at 1577 cm−<sup>1</sup> is ascribed to the C-C strC stretching mode in sp2 carbon bonds, and (iii) the 2D peak at 2720 cm−<sup>1</sup> is characterized by two bands, the intense 2D2A band at 2720 cm−<sup>1</sup> and a shoulder 2D1A band at 2677 cm<sup>−</sup>1. In particular, these bands originate as the effect of the splitting of *π* electron bands due to the interaction between stacked graphene layers. The G\* peak found at 2447 cm−<sup>1</sup> is characteristic of carbon-based materials with a graphitic-like structure.

**Figure 1.** Characterization of starting graphite source: (**a**) SEM morphology, (**b**) XRD measurement, and (**c**,**d**) Raman spectrum from 1000 to 3000 cm−<sup>1</sup> recorded using 532 excitation laser. The intensity was normalized by the most intense peak. The Raman spectrum was fitted using Lorentzian functions.

Compared to graphite, the Raman spectrum of oxidized graphenes shows a highly broadened and very-low intense 2D peak. With this in mind, the 2D band region of GO and rGO is not analyzed here, and instead, we focus on the region from 1000 to 2000 cm−<sup>1</sup> to scrutinize the crystallinity and, most importantly, the basal/edge defects of the obtained materials after the oxidation-reduction process (discussed below).

#### *3.2. Characterization of GO and rGO*

While, in the present work, rGO is used for the adsorption of cationic pollutants, it is extremely important to discuss its transformation from GO. SEM micrographs of GO and rGO are shown in Figure 2a,c, respectively. The surface morphology of GO indicates a face-to-face stacking of flakes as well as randomly aggregated flakes with wrinkles and folds on the surface (Figure 2a). Instead, rGO shows a surface morphology with mesopores and micropores and randomly organized flakes (Figure 2c). The highly distorted porous surface of rGO can avoid the face-to-face stacking of flakes, as observed in GO.

EDS measurements were carried out to determine the elemental composition of GO and rGO, considering a bombarded region large enough. Then, the carbon and oxygen content were C: 49.7% and O: 50.3% for GO (after oxidation process) and C: 62.9% and O: 37.1% for rGO (after de reduction process). The oxygen content decreased by 26.2% using CA as an alternative green-reducing agent, confirming the (partial) removal of oxygen functional groups.

**Figure 2.** SEM morphology of (**a**) GO and (**c**) rGO. EDS measurements of (**b**) GO and (**d**) rGO. TEM images of (**e**) GO and (**f**) rGO.

Representative TEM graphs of GO and rGO are shown in Figure 2e,f, respectively. GO looks like a semi-transparent thin nanosheet with various wrinkles and folds on the surface and edges (Figure 2e). The observed wrinkled/folded structure is attributed to surface defects because of the deviation from sp<sup>2</sup> to sp<sup>3</sup> hybridization as the effect of a high density of oxygen-containing functional groups [54]. After the reduction process with CA, well-defined and impurity-free nanosheets with slightly wrinkled regions are observed in rGO, suggesting the recovery of sp<sup>2</sup> hybridization by the removal of functional groups. The observed regular surface allows concluding that rGO did not undergo severe in-plane disruption compared to GO.

Raman analyses were performed to further corroborate the transformation of GO into rGO (Figure 3a,b, respectively). As is typical for oxidized graphenes, two characteristic peaks are observed in GO and rGO: (i) the D peak at ∼1349 cm−<sup>1</sup> is attributed to the breathing mode of aromatic carbon rings, which is Raman active by structural defects [32], and (ii) the G peak at ∼1588 cm−<sup>1</sup> is due to the C-C stretching mode in the sp2 hybridized carbon structure [46]. A detailed analysis using Lorentz functions shows the existence of three prominent bands: the D band (yellow line), the G band (green line), and the D' band (blue line). In particular, the D' band confirms the presence of basal/edge defects, and a decrease in the D' band intensity is a direct indication of GO reduction, which is observed in the Raman spectrum of rGO (Figure 3b). On the other hand, the ID/IG intensity ratio can be used as an indicator of the density of structural defects in the obtained oxidized graphenes [43]. It was found that the intensity ratio of GO (2.2) is larger than that of rGO (1.65), indicating that the size of the graphene-like domains increases after the reduction process.

The absorbance spectra of GO and rGO are shown in Figure 3c,d, respectively. Using the Lorentzian function, GO has the main absorbance band at 230 nm (darker green line) and a shoulder band at 329 nm (yellow line), which are related to the *π* − *π*<sup>∗</sup> transitions of C-C bonds and *n* − *π*<sup>∗</sup> transitions of C=O bonds, respectively. To confirm the transformation of GO into rGO, two characteristics are needed: (i) a redshift of the main absorbance band and (ii) the loss of the shoulder band. After the reduction process, rGO only meets the first point when the main absorbance band shifts to 261 nm, but the second one is observed at 324 nm, suggesting a close content of oxygen-containing functional groups, particularly hydroxyl and epoxide groups.

**Figure 3.** Raman spectra of (**a**) GO and (**b**) rGO from 1000 to 2000 cm<sup>−</sup>1. The intensity was normalized by the most intense peak. The Raman spectrum was fitted using Lorentzian functions. UV–Vis spectra of (**c**) GO and (**d**) rGO and the absorbance spectra were fitted by Lorentzian functions.

The presence and type of oxygen functional groups are confirmed by the FTIR analysis (Figure 4a). It is widely accepted that the hydroxyl and epoxide groups are attached to the basal in-plane of the graphene, whereas the carboxyl and carbonyl groups are located at the edges. The FTIR spectrum of GO shows the following bands: C-O-C at 1044 cm<sup>−</sup>1, C-O at 1222 cm<sup>−</sup>1, C=C at 1644 cm<sup>−</sup>1, and C=O at 1729 cm<sup>−</sup>1. The broadband observed at ∼<sup>3426</sup> cm−<sup>1</sup> is due to the presence of the hydroxyl groups (C-H) as well as adsorbed water molecules between GO flakes. The latter provides a hydrophilic characteristic in GO to be highly dispersible in water. It is worth noting that a higher hydrophilic property could interfere with the removal of pollutants from aqueous media, giving a poor adsorption process. After the reduction, these bands are significantly attenuated and weakened in the rGO spectrum, evidencing the removal of oxygen-containing functional groups [33].

To determine the thermal stability of as-made oxidized graphenes and the effect on the oxygen-containing functional groups, we carried out TGA analyses on GO and rGO (Figure 4b). In GO, the weight loss below 100 ◦C is ascribed to the loss of water molecules [33]. The significant weight loss in the region of 200–300 ◦C is attributed to the pyrolysis of unstable molecules (such as CO, CO2, and H2O) [33]. In the region of 300–600 ◦C, the weight loss is due to the removal of stable oxygen functional groups [33]. Instead, rGO shows relative thermal stability, but the observed TGA curve follows a similar trend as GO, confirming a reduced density of oxygen functional groups.

**Figure 4.** (**a**) Infrared spectra, (**b**) thermogravimetric study, and (**c**) XRD patterns of GO (black) and rGO (red), respectively.

Finally, the crystallinity changes from GO to rGO were revealed by XRD analysis (Figure 4c). As mentioned, graphite is characterized by an intense crystalline peak at 2θ = 26.73◦ related to a lattice spacing of 0.334 nm, which corresponds to the (002) interplane distance [43] (Figure 1b). In GO, this peak is found at 2θ = 10.93◦ with a lattice spacing of 0.81 nm, indicating the oxidation of graphite. The increased interlayer spacing appears as an effect of the intercalation of water molecules and oxygen functional groups. Additionally, the very low width of this peak demonstrates an ordered stacking along the out-of-plane axis. After the reduction process, the peak becomes broader due to the partial breakdown of the long-range order, and it shifts towards higher angles, 2θ = 22◦, showing a decrease in the lattice spacing (~0.39 nm) [43].

All these facts and pieces of evidence demonstrate the transformation of GO into rGO, which will be used for the removal of cationic pollutants from aqueous media, i.e., MB and Hg(II).

#### *3.3. Adsorption Kinetics*

We begin analyzing the effectiveness of rGO for removing MB and Hg(II) from water by using the following expression:

$$q\_t = \frac{(\mathbb{C}\_0 - \mathbb{C}\_t)V}{W} \tag{1}$$

where *C*<sup>0</sup> and *Ct* are the initial pollutant concentration (mg L<sup>−</sup>1) and the pollutant concentration at time *t*, respectively. *W* is the adsorbent mass (g), and *V* represents the volume of the aqueous solution (L). At the equilibrium, the equilibrium concentration is *Ce* = *Ct*, and the equilibrium adsorption capacity is *qe* = *qt*.

The removal efficiency (*RE*%) of the as-made rGO material can be defined by the following simple equation:

$$RE\% = \frac{\left(\mathcal{C}\_0 - \mathcal{C}\_\varepsilon\right)}{\mathcal{C}\_0} \times 100\tag{2}$$

Figure 5 shows the adsorption kinetics of MB and Hg (II) onto rGO at 298 K considering a contact time of up to 60 min. It can be seen that rGO rapidly captures MB molecules after 30 min (Figure 5a), while for Hg (II), the equilibrium time of adsorption is 20 min (Figure 5b). These results highlight the effectiveness of rGO for removing cationic pollutants from aqueous solutions compared with conventional benchmark sorbents [35–40]. In particular, the effectiveness of rGO can be attributed to the recovered surface area after the reduction process as well as the presence of oxygen functional groups.

**Figure 5.** Adsorption kinetics of (**a**) MB on rGO and (**b**) Hg(II) on rGO as a function of contact time (60 min) at 298 K.

The parameters of the adsorption kinetic process were determined by the pseudo-firstorder model and pseudo-second-order model. Specifically, Tene et al. stated that the first model assumes that the rate of change of the adsorption capacity is proportional to the concentration of available active sites per unit mass of adsorbent material [25], whereas Arias et al. stated that the second model assumes that the rate of change of the concentration of occupied active sites per unit mass of the adsorbent material is proportional to the square of the concentration of free active sites per unit mass of sorbent [13].

The pseudo-first-order model (red line) and pseudo-second-order model (blue line) can be described as follows:

$$
\log\left(q\_t - q\_t\right) = \log q\_t - \frac{k\_1}{2.303}t \tag{3}
$$

and

$$\frac{t}{q\_t} = \frac{1}{k\_2 q\_c^2} + \frac{1}{q\_c} t \tag{4}$$

where *k*<sup>1</sup> and *k*<sup>2</sup> are the pseudo-first-order and pseudo-second-order rate constants, respectively. The estimated parameters of the adsorption kinetics are summarized in Table 1.


**Table 1.** Estimated parameters at 298 K of the pseudo-first-order model and the pseudo-secondorder model.

A close picture of the pseudo-first-order and pseudo-second-order parameters shows that, in the case of MB, the calculated values of the equilibrium adsorption capacity (*qe*(*cal*) = 69.82 mg g−<sup>1</sup> and *qe*(*cal*) = 70.72 mg g−1, respectively) are very close to the experimental value (*qe*(*exp*) = 68.21 mg g−1). In the case of Hg(II), the calculated adsorption capacity (*qe*(*cal*) = 143.71 mg g−1) from the pseudo-first-model is close enough to the experimental value (*qe*(*exp*) = 142.26 mg g−1). However, the pseudo-second-order model overestimates the equilibrium adsorption capacity (*qe*(*cal*) = 151.32 mg g<sup>−</sup>1). By the comparison of SSE and R2 metrics, the adsorption kinetics of MB onto rGO are more in line with the pseudo-second-order model (SSE = 5.24, R2 = 0.999), whereas the adsorption kinetics of Hg(II) onto rGO are more in line with the pseudo-first-order model (SSE = 1826, R2 = 0.949).

#### *3.4. Intraparticle Diffusion Study*

The diffusion process of any pollutant into porous solid materials, such as our rGO (Figure 2c), mostly involves several steps characterized by different rates. This fact can be calculated by the intraparticle diffusion (IPD) model [13,25], which is given by the following expression:

$$q\_t = k\_p t^{0.5} + \mathbb{C} \tag{5}$$

where *kp* is the intraparticle diffusion rate constant (g mg−<sup>1</sup> min−1), and the intercept *C* reflects the boundary layer or surface adsorption. The respective plot and estimated parameters of the IPD model are shown in Figure 6 and Table 2.

**Figure 6.** Intraparticle diffusion (IPD) study of (**a**) rGO+Mb and (**b**) rGO+Hg(II) at 298 K, showing different regions of linearity (MB concentration 100 mg L−<sup>1</sup> and Hg(II) concentration 150 mg L<sup>−</sup>1).


**Table 2.** Estimated parameters of the intraparticle diffusion (IPD) model at 298 K.

As Ofomaja et al. [55] stated, the larger the intercept value, the greater the contribution of the surface in the adsorption process. Indeed, the values observed in MB (*C* = 59.17) and Hg(II) (*C* = 44.28) indicate that a greater amount of surface adsorption occurred, leading to a decrease in the rate of diffusion of MB molecules and Hg(II) ions from the adsorbent external surface to the adsorbent internal structure. From the linearized plot of the IPD model, different regions are observed: (i) the initial region (faster stage) is related to the movement of the pollutant from the solution to the rGO surface, (ii) the second region (intermediate stage) is related to the gradual diffusion of the pollutant into the large pores of the rGO structure, and (iii) the final region (lower stage) involves a very slow diffusion of the pollutant from larger pores to smaller ones.

Interestingly, the adsorption mechanism of MB on rGO is characterized by only two regions, regions I and II (Figure 6a), while all three regions are observed when Hg(II) becomes adsorbed onto rGO. In light of understanding this fact, we hypothesize that the size of pollutants plays a significant role in the diffusion procedure, i.e., as the MB molecules exhibit larger sizes compared to Hg(II) ions, MB cannot reach region III, particularly from larger to smaller pores.

The initial adsorption factor (*Ri*) can be estimated to further understand the abovementioned regions (Table 2) as follows:

$$R\_i = \frac{q\_{ref} - \mathcal{C}}{q\_{ref}} \tag{6}$$

where *qref* is the final adsorption amount at the longest time. In the MB-rGO system, the estimated *Ri* value is much less than 0.5, which confirms that most of the adsorption of MB occurs on the surface of rGO. In contrast, for the Hg(II)-rGO system, the value of *Ri* ∼0.49 indicates a limit between the strong initial adsorption (related to region I) and intermediate initial adsorption (related to region II), which means that the adsorption process of Hg(II) ions could occur at almost the same time in both regions.

## *3.5. Adsorption Isotherms*

Adsorption isotherms were carried out to analyze the interaction between MB molecules or Hg(II) ions and rGO considering a contact time of 30 min for MB and 20 min for Hg(II). The experimental data can be fitted using the Langmuir model and Freundlich model using the following equations, respectively:

$$\eta\_{\varepsilon} = \frac{q\_{m} \,\, \mathcal{K}\_{L} \,\, \mathcal{C}\_{\varepsilon}}{1 + \mathcal{K}\_{L} \,\, \mathcal{C}\_{\varepsilon}} \tag{7}$$

and

$$q\_{\mathcal{E}} = K \rho \mathbf{C}\_{\mathcal{E}}^{1/n} \tag{8}$$

where *qm* represents the maximum adsorption capacity (mg g−1), *KL* is the Langmuir constant (L g−1), *KF* is Freundlich constant (mg L−1), and *n* is the surface heterogeneity of adsorbent material. The corresponding results and estimated parameters at different temperatures (298, 313, 333 K) are shown in Figures 7 and 8 and Tables 3 and 4.

**Figure 7.** Adsorption isotherms of MB on rGO considering three different temperatures (289–333 K). (**a**) Langmuir model and (**b**) Freundlich model.

**Figure 8.** Adsorption isotherm of Hg(II) on rGO considering three different temperatures (289–333 K). (**a**) Langmuir model and (**b**) Freundlich model.

**Table 3.** Parameters of the Langmuir and Freundlich models for the adsorption isotherms of MB onto rGO considering three different temperatures.


**Table 4.** Parameters of the Langmuir and Freundlich models for the adsorption isotherms of Hg(II) onto rGO considering three different temperatures.


Taking the high correlation R<sup>2</sup> values (Tables 3 and 4), it can be seen that the measured points are more in line with the Langmuir model. Although the temperature does not dramatically modify the chemical composition of rGO at temperatures below 100 ◦C (Figure 4b), it seems to be an important parameter in the adsorption process because when the temperature increases, in the case of MB on rGO, a slight decrease in the maximum adsorption capacity is observed from 121.95 to 107.53 mg g<sup>−</sup>1. In contrast, in the case of Hg (II) on rGO, a significant increase in the maximum adsorption capacity is detected from 109.49 to 255.04 mg g<sup>−</sup>1. The temperature is a key point to be considered if rGO is used to treat water or wastewater at an industrial scale.

From the Freundlich model, the estimated values of *n* for MB-rGO (Table 3) or Hg(II) rGO (Table 4) systems indicate that the adsorbent heterogeneity tends to be homogeneous as the temperature rises. Indeed, values of *n* close to zero (<0.1) indicate strong surface heterogeneity. The affinity of the tested cationic pollutants for rGO can also be determined by the *KL* parameter, where the estimated values were found to be much less than 0.1, suggesting a good affinity of rGO to capture cationic pollutants, i.e., MB molecules and Hg(II) ions. However, this statement motivates more extended work for testing more cationic and non-cationic pollutants.

#### *3.6. Effect of pH and Initial Concentration*

To scrutinize the effect of the pH on the process of cationic pollutant removal, the experiments were carried out at different pH values ranging from 2 to 12 and setting the temperature at 298 K.

For MB on rGO (Figure 9a), the adsorption increases, starting from a removal percentage of about 76% at pH = 3 up to 92% at pH = 6. The removal percentage remains relatively constant from pH = 6 to pH = 8. The removal percentage decreases down to 83% for pH ≥ 10. To understand this fact, the effect of pH can be divided into three different regions: (i) from pH = 2 to pH = 4, the acid region is rich in cations which are captured together with the cationic MB molecules; (ii) the (relatively) neutral region from pH = 6 to pH = 8 is free from cations in the medium, and therefore, only the cationic dye molecules are captured by rGO; and (iii) the basis region (pH ≥ 10) is characterized by an excess of OH− ions that interact with the cationic dye molecules, remaining suspended in the aqueous media.

**Figure 9.** Removal percentage as a function of the pH (from 2 to 12) at 298 K of (**a**) MB on rGO and (**b**) Hg(II) on rGO (MB concentration 100 mg L−<sup>1</sup> and Hg(II) concentration 100 mg L<sup>−</sup>1).

For Hg(II) on rGO (Figure 9b), a removal percentage of about 39% at pH = 2 is observed, and the maximum removal percentage is found at pH = 10 (∼82%). The removal percentage remains relatively constant for pH ≥ 6, with an average value of 76.58%. To understand these results, a similar description can be given: (i) for pH ≤ 4, the cations in the acidic medium fight with the mercury cations for the active sites of rGO; (ii) in the neutral region, mercury cations easily reach the active sites of rGO; and, interestingly, (iii) for pH ≥ 8, mercury cations sometimes prefer to interact with the active sites of rGO rather than OH− ions due to the variation of the removal percentage when the pH increases.

The adsorption capacity (*qe*) of rGO increases quite linearly with the initial concentration of MB in solution (*C*0), almost in the range from 10 to 80 mg L−1; however, at high concentrations (≥90 mg L<sup>−</sup>1), a deviation from linearity does occur (Figure 10a). Similarly, the *qe* values of rGO increase linearly with the initial concentration of Hg(II) in the solution, from 10 to 50 mg L−1, and at concentrations ≥50 mg L−1, a deviation from linearity is also observed (Figure 10b). These results suggest that rGO has a finite amount of active adsorbent sites, which is fixed by its quality and the experimental conditions, such as temperature, pH, and solution volume/adsorbent mass ratio. To further emphasize, at the beginning of the adsorption process, rGO has a vast number of active sites, increasing the *qe* value as long as free active sites are available on rGO. Then, if all the active sites are involved, the saturation, and therefore the maximum adsorbent capacity (*qm*), is attained [13,25].

**Figure 10.** Effect of the initial concentration on the adsorption process at 298 K of (**a**) MB or rGO and (**b**) Hg(II) on rGO. Adsorption capacity (black markers) and removal percentage (red markers).

The adsorption effectiveness of rGO—defined as the percentage of cationic pollutant removal from water—is almost independent of *C*<sup>0</sup> in the adsorption of MB onto rGO, assuming an average value of 89.21%. Interestingly, a clear dependence on *C*<sup>0</sup> is observed for the adsorption of Hg(II) onto rGO, i.e., an abrupt drop from 92.89% (*C*<sup>0</sup> = 30 mg L−1) to 48.85% (*C*<sup>0</sup> = 100 mg L<sup>−</sup>1), giving an average value of mercury removal of 72.93%.

## *3.7. Adsorption Thermodynamics*

To acquire information about the energy changes due to the involved adsorption process [13], the Gibbs free energy (Δ*G*0), enthalpy change (Δ*H*0), and entropy change (Δ*S*0) were calculated by the following expressions:

$$K\_d = \frac{q\_c}{C\_0} \tag{9}$$

$$
\ln K\_d = \frac{\Delta S^0}{R} - \frac{\Delta H^0}{R \ T} \tag{10}
$$

$$
\Delta G^0 = -R \ T \ln \mathcal{K}\_d \tag{11}
$$

where *Kd* represent the distribution coefficient [13]. Δ*H*<sup>0</sup> and Δ*S*<sup>0</sup> were calculated from the slope and intercept of Van't Hoff plot of ln *Kd* as a function of *T*−<sup>1</sup> [25]. The Van't Hoff plot and estimated parameters are shown in Figure 11 and Table 5, respectively.

**Figure 11.** Van't Hoff study for the adsorption of (**a**) MB on rGO and (**b**) Hg(II) on rGO.


**Table 5.** Thermodynamics parameters for Mb and Hg(II) adsorption on rGO at three different temperatures.

The negative Δ*G*<sup>0</sup> values observed at different temperatures indicate spontaneous adsorption of MB molecules and Hg(II) ions onto the rGO surface. It is worth noting that, for <sup>Δ</sup>*G*<sup>0</sup> values in the range from 0 to −20 kJ mol−1, the adsorption process is assigned to physisorption or multilayer adsorption [25], while in the range from −80 to −400 kJ mol<sup>−</sup>1, the adsorption is assigned to chemisorption or monolayer adsorption [25]. The region from −20 to −80 kJ mol−<sup>1</sup> remains unclear, and a combined adsorption process can be assumed. With this in mind, the estimated <sup>Δ</sup>*G*<sup>0</sup> values for MB onto rGO (−22.75, −23.81, and −25.16 kJ mol−1) and Hg(II) onto rGO (−39.84, 31.55, 32.97 kJ mol−1) suggest that the adsorption process of tested cationic pollutants on rGO is governed by a mixed physisorption–chemisorption process. Interestingly, for MB on rGO, the Δ*G*<sup>0</sup> value increases by 5% at 313 K and by 10% at 333 K. In contrast, an inversely proportional relationship is observed for Hg(II) on rGO; say, the Δ*G*<sup>0</sup> value decreases by 21% at 313 k and by 17% at 333 K.

The negative Δ*H*<sup>0</sup> values indicate the exothermic nature of the adsorption process, i.e., a negative enthalpy implies that the temperature increase had a negative impact, particularly on the adsorption of MB (Δ*H*<sup>0</sup> = −2.20 kJ mol<sup>−</sup>1). However, in the adsorption of Hg(II) on rGO, the value observed (Δ*H*<sup>0</sup> = −0.14 kJ mol−1) is very small and could be considered negligible since increasing the temperature significantly increases the maximum adsorption capacity of rGO, as evidenced by the Langmuir model (Figure 8a and Table 4). The positive values of <sup>Δ</sup>*S*<sup>0</sup> = 0.069 kJ mol−<sup>1</sup> · <sup>K</sup>−<sup>1</sup> and <sup>Δ</sup>*S*<sup>0</sup> = 0.079 kJ mol−<sup>1</sup> · <sup>K</sup>−<sup>1</sup> corroborate the affinity of MB molecules and Hg(II) ions toward the rGO surface.

## *3.8. Final Remarks*

Table 6 shows the estimated *qm* values for MB (*qm* = 121.95 mg g−1) and Hg(II) (*qm* = 109.49 mg g<sup>−</sup>1) at 298 K, which are compared to those of recent studies.


**Table 6.** Comparative adsorption capacity of several adsorbents for the removal of MB and Hg(II).

The estimated *qm* value of the dye adsorption is higher than those previously reported and only surpassed by GO reduced by Citrus hystrix (*qm* = 276.06 mg g<sup>−</sup>1), suggesting that as-made (non-extra functionalized) rGO are excellent platforms to replace conventional adsorbent materials. In the case of heavy metal adsorption, the estimated *qm* value is higher than some functionalized/decorated GO and rGO. However, S-GO seems to be more profitable to be used for the removal of mercury (*qm* = 3490 mg g<sup>−</sup>1), but this is due to the fact that, obviously, the presence of sulfur improves the affinity and specificity for Hg (II) ions in any adsorbent material.

Although, in the present work, the regeneration of rGO was not studied, which motivates more extended work, we propose the following well-known techniques or processes: (i) the adsorbed rGO-pollutant system can be separated from aqueous media by filtration using filters with a pore size less than 1 μm since rGO is within the order of few micrometers, (ii) pollutant can be released from rGO by applying the concept of ionic force, i.e., by applying buffer solutions, and (iii) the isolated MB molecules or Hg(II) ions can be extracted by sulfide precipitation.

## **4. Conclusions**

In summary, we have demonstrated the effective and efficient removal of cationic pollutants (i.e., MB molecules and Hg(II) ions) from aqueous solutions using an eco-friendly and as-made rGO. The adsorbent material shows fast adsorption with a saturation capacity of 121.95 mg g−<sup>1</sup> and 109.49 mg g−<sup>1</sup> at 298 K, suggesting a good affinity for MB molecules and Hg(II) ions. These results are superior to those recently reported for other graphenebased benchmark materials [35–40]. By means of several chemical physics analyses, we have also shown that rGO keeps a good efficiency over a wide range of initial cationic pollutant concentrations and a broad range of pH values. Specifically, the maximum removal percentage as a function of pH was found in the range of 6 to 8 for MB and 6 to 10 for Hg(II). Our results allowed us to conclude that the MB-rGO and Hg(II)-rGO adsorption interaction follows a combined physisorption–chemisorption process due to the fact that the Gibbs free energy was found from −22.75 to −25.16 kJ mol−<sup>1</sup> for MB and from −39.84 to −32.97 kJ mol−<sup>1</sup> for Hg(II). The present study proposes non-extra functionalized rGO as a potential green adsorbent for wastewater decontamination.

**Author Contributions:** Conceptualization, S.B.; Data curation, T.T., M.G. and C.V.G.; Formal analysis, L.S.C. and C.V.G.; Funding acquisition, T.T.; Investigation, T.T., L.S.C. and C.V.G.; Methodology, E.V., M.A.P., O.S., E.V.-G., S.V., S.B., F.A., A.S. and L.S.C.; Validation, M.G.; Visualization, M.G.; Writing—original draft, S.B. and C.V.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Universidad Técnica Particular de Loja (UTPL-Ecuador), project "222-Radon adsorption on activated carbon and graphene filters" (code: PROY\_INNOV\_QUI- \_2021\_3019), and by the FONDOCyT from the Ministry of Higher Education Science and Technology of the Dominican Republic (grant no. 2018-2019-3A9-139).

**Data Availability Statement:** Not applicable.

**Acknowledgments:** T.T., M.G. and C.V.G. wish to thank Escuela Superior Politécnica de Chimborazo for its hospitality during the completion of this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Nanomaterials* Editorial Office E-mail: nanomaterials@mdpi.com www.mdpi.com/journal/nanomaterials

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-4784-8