**Experimental Study on the Lubrication and Cooling E**ff**ect of Graphene in Base Oil for Si3N4**/**Si3N<sup>4</sup> Sliding Pairs**

## **Lixiu Zhang 1,2, Xiaoyi Wei 1,2,\*, Junhai Wang 1,2, Yuhou Wu 1,3, Dong An 1,2 and Dongyang Xi <sup>4</sup>**


Received: 18 December 2019; Accepted: 22 January 2020; Published: 3 February 2020

**Abstract:** Recently, the engineering structural ceramics as friction and wear components in manufacturing technology and devices have attracted much attention due to their high strength and corrosion resistance. In this study, the tribological properties of Si3N4/Si3N<sup>4</sup> sliding pairs were investigated by adding few-layer graphene to base lubricating oil on the lubrication and cooling under different experimental conditions. Test results showed that lubrication and cooling performance was obviously improved with the addition of graphene at high rotational speeds and low loads. For oil containing 0.1 wt% graphene at a rotational speed of 3000 r·min−<sup>1</sup> and 40 N loads, the average friction coefficient was reduced by 76.33%. The cooling effect on Si3N4/Si3N<sup>4</sup> sliding pairs, however, was optimal at low rotational speeds and high loads. For oil containing 0.05 wt% graphene at a lower rotational speed of 500 r·min−<sup>1</sup> and a higher load of 140 N, the temperature rise was reduced by 19.76%. In addition, the wear mark depth would decrease when adding appropriate graphene. The mechanism behind the reduction in friction and anti-wear properties was related to the formation of a lubricating protective film.

**Keywords:** Si3N4; graphene; lubrication; friction; temperature rise

## **1. Introduction**

With the development of advanced manufacturing technology, the devices are often operated under conditions such as high temperature, high pressure, or less lubrication etc. However, traditional metal materials and metal tribo-pairs were not available due to its vulnerability to rupture and corrosion damage. Engineering ceramic tribo-pairs, for example Si3N4/Si3N4, Al2O3/Al2O3, etc., have excellent properties such as low density, significant thermal stability, and high hardness [1]. These properties are suitable for a wide variety of tribological applications [2,3]. Therefore, the research and practical application of Si3N4/Si3N<sup>4</sup> as friction and wear components has become one of the hot spots of devices and material science.

Lubrication is an important measure to reduce wear, save energy, and improve industrial efficiency and reliability. In addition, lubricating oil additives are also important for improving the performance of lubricating oil. Nanomaterials as a kind of lubricant additive play a good role in anti-wear and anti-friction. Among them, as the component part of graphite used as the traditional solid lubricating

material, graphene has attracted much attention in the tribological field due to its unique friction and wear properties [4,5]. Graphene possesses an extremely thin laminated structure, high load-bearing capacity, high chemical stability, and low surface energy, and thus, it can offer lower shear stress and prevent direct contact between metal interfaces [6]. Liu et al. [7] prepared novel composite coatings of diamond-like carbon/ionic liquid/graphene. They found that 0.075 mg·mL−<sup>1</sup> graphene in the composite coatings exhibited the lowest friction coefficient, and the highest bearing capacity in a simulated space environment. Huang et al. [8] investigated the tribological behavior of the graphite nanosheets as an additive in paraffin oil. Their result showed that the load-carrying capacity and anti-wear ability of the lubricating oil were improved. Lin et al. [9], likewise, carried out a new kind of lubricating oil containing modified graphene platelets. The results indicated that it could clearly improve the wear resistance and load-carrying capacity of the machine.

Although many studies have shown that graphene as a lubricant additive can improve the tribological properties of sliding trio-pairs, most of the studies have focused on steel/steel tribo-pairs. In our previous study, we investigated the tribological behaviors of Si3N4/GCr15 sliding pairs lubricated with graphene oxide [10]. However, there is less attention on the lubrication of Si3N4/Si3N<sup>4</sup> tribo-pairs, as the structure and performance characteristics of engineering ceramic device material are very different from those of metal materials. In addition, there are also few reports on the cooling effect of graphene as a lubricant additive. The purpose of this work is to explore the lubrication and cooling effect for Si3N4/Si3N<sup>4</sup> tribo-pairs by adding graphene to base lubricating oil. Factors influencing friction coefficient include rotational speed, load, and graphene concentration. We have investigated the tribological behavior of Si3N4/Si3N<sup>4</sup> sliding pairs under three different experimental conditions by using a Rtec MFT 5000 Tribometer with the ball-on-disk mode. Finally, the effect on lubrication and cooling and the lubrication mechanism is discussed.

#### **2. Materials and Methods**

#### *2.1. Materials*

Few-layer graphene (FLG) was obtained from Detong Nanotechnology Co. Ltd. (Qingdao, China). Mobile DTE oil light (Mdol) were selected as the base lubricating oil and purchased from Huijie Development Co. Ltd. (Changsha, China). FLG was obtained by physical methods and used directly without further purification. The kinematic viscosity of Mdol was 5.34 mm<sup>2</sup> ·s −1 , when the temperature was 100 ◦C and 29.77 mm<sup>2</sup> ·s <sup>−</sup><sup>1</sup> with a temperature of 40 ◦C. Mdol containing different concentrations of FLG (0.025, 0.05, 0.075, and 0.1 wt%) was prepared and followed by ultrasonication for about 30 min to make sure that FLG was evenly dispersed in the base lubricating oil.

#### *2.2. Tribological Test*

Si3N<sup>4</sup> ceramic balls and disks were purchased from Zhihai Bearing Co., Ltd. (Shanghai, China). The ball was a commercial product with a diameter of 9.525 mm and a surface roughness of no more than 140 nm. The thickness of the disk was 5 mm and its surface roughness did not exceed 250 nm. The tribological behaviors of Si3N4/Si3N<sup>4</sup> tribo-pairs lubricated without and with FLG were examined using a Rtec MFT5000 Tribometer (Rtec, San Jose, CA, USA) with the ball-on-disk mode, with the cylindrical upper tribo-pairs as a Si3N<sup>4</sup> ball and the lower tribo-pairs as a Si3N<sup>4</sup> disk. The temperature of the Si3N4/Si3N<sup>4</sup> sliding pairs was recorded by a temperature-rise detection device. The temperature-rise detection device included a data acquisition device, a standard rod, and the spindle error analyzer software. The temperature sensor used in data acquisition was a magnetic thermistor attached to the cylindrical pin above the Si3N<sup>4</sup> ceramic ball. It recorded temperature changes in real time during the experiment.

In a typical test, Mdol containing 0.025, 0.05, 0.075, and 0.1 wt% FLG were prepared and ultrasonic for about 30 min to ensure uniform dispersion of the graphene in the base lubricating oil. In order to avoid interference from other factors, the graphene additive lubricating oil did not use a surfactant. Since factors influencing the tribological properties was carried out, including FLG concentration (0.025, 0.05, 0.075, and 0.1 wt%), load (40, 80, 140 N), and rotational speed (500, 3000 r·min−<sup>1</sup> ). All tribological tests were repeated at least three times. The friction coefficient was automatically recorded by the experimental device and the real-time temperature was recorded by the temperature-rise detection system. 3000 r·min−1). All tribological tests were repeated at least three times. The friction coefficient was automatically recorded by the experimental device and the real-time temperature was recorded by the temperature-rise detection system. *2.3. Characterization* 

*Micromachines* **2020**, *11*, x 3 of 9

#### *2.3. Characterization* The structure of FLG was imaged with Raman spectroscope (HR800, Horiba, Paris, France) with

The structure of FLG was imaged with Raman spectroscope (HR800, Horiba, Paris, France) with a confocal Raman microscope mode and a laser wavelength of 532 nm. Nanoparticle analyzer is an instrument that uses a physical method to test the size and distribution of particles. The particle diameter distribution of FLG was measured by the NanoPlus-3 nanoparticle analyzer (Micromeritics, New York, NY, USA). The wear mark depth of the Si3N<sup>4</sup> disk was characterized by the OLS4100 3D laser measuring microscope (Olympus, Tokyo, Japan). S-4800 scanning electron microscope (SEM, Hitachi, Tokyo, Japan) equipped with an energy-dispersive X-ray spectroscope (EDS, Hitachi, Tokyo, Japan) and Raman spectroscope (Horiba, Paris, France) were used to observe the wear mark of the Si3N<sup>4</sup> disk. a confocal Raman microscope mode and a laser wavelength of 532 nm. Nanoparticle analyzer is an instrument that uses a physical method to test the size and distribution of particles. The particle diameter distribution of FLG was measured by the NanoPlus-3 nanoparticle analyzer (Micromeritics, New York, NY, USA). The wear mark depth of the Si3N4 disk was characterized by the OLS4100 3D laser measuring microscope (Olympus, Tokyo, Japan). S-4800 scanning electron microscope (SEM, Hitachi, Tokyo, Japan) equipped with an energy-dispersive X-ray spectroscope (EDS, Hitachi, Tokyo, Japan) and Raman spectroscope (Horiba, Paris, France) were used to observe the wear mark of the Si3N4 disk. **3. Results and Discussion** 

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

#### *3.1. Materials Characterization 3.1. Materials Characterization*  Figure 1 illustrates the characterization of FLG. Figure 1a is the SEM image of FLG, showing

Figure 1 illustrates the characterization of FLG. Figure 1a is the SEM image of FLG, showing that FLG retains its original laminated structure, which is transparent with folding at the edges, suggesting very few layers. Figure 1b portrays the Raman spectrum of FLG. Three typical features of FLG, the D-band (1342 cm−<sup>1</sup> ), G-band (1575 cm−<sup>1</sup> ), and 2D-band (2701 cm−<sup>1</sup> ), are observed. As shown in Figure 1b, the peak shape of the 2D-peak is widest, the intensity of the G-peak is very strong, and the intensity ratio of the G to 2D band (*I*G/*I*2D) is more than 1, demonstrating that the graphene samples exhibit a few-layered structure [11,12], as is proved in SEM (Figure 1a). However, the presence of the D-band (1342 cm−<sup>1</sup> ) illustrates the occurrence of disorder and defects in the graphene samples [13,14]. that FLG retains its original laminated structure, which is transparent with folding at the edges, suggesting very few layers. Figure 1b portrays the Raman spectrum of FLG. Three typical features of FLG, the D-band (1342 cm−1), G-band (1575 cm−1), and 2D-band (2701 cm−1), are observed. As shown in Figure 1b, the peak shape of the 2D-peak is widest, the intensity of the G-peak is very strong, and the intensity ratio of the G to 2D band (*I*G/*I*2D) is more than 1, demonstrating that the graphene samples exhibit a few-layered structure [11,12], as is proved in SEM (Figure 1a). However, the presence of the D-band (1342 cm−1) illustrates the occurrence of disorder and defects in the graphene samples [13,14].

**Figure 1.** Structure images of few-layer graphene (FLG). (**a**) Scanning electron microscope (SEM) **Figure 1.** Structure images of few-layer graphene (FLG). (**a**) Scanning electron microscope (SEM) image; (**b**) Raman spectrum.

#### image; (**b**) Raman spectrum. *3.2. Tribological Properties*

*3.2. Tribological Properties*  The average friction coefficient (COF) values of the Si3N4/Si3N4 sliding pairs lubricated by Mdol with and without FLG at different experimental conditions are shown in Figure 2. Figure 2a shows the friction coefficient curves of the Si3N4/Si3N4 lubricated by Mdol with 0.1 wt% FLG at 500 r·min−<sup>1</sup> speed under 40 N loads. When FLG is added to the base oil, the friction coefficient of the Si3N4/Si3N4 sliding pairs is drastically reduced. Figure 2b shows the COF of the Si3N4/Si3N4 sliding pairs lubricated with and without different contents of FLG in three different working conditions. The average friction coefficient (COF) values of the Si3N4/Si3N<sup>4</sup> sliding pairs lubricated by Mdol with and without FLG at different experimental conditions are shown in Figure 2. Figure 2a shows the friction coefficient curves of the Si3N4/Si3N<sup>4</sup> lubricated by Mdol with 0.1 wt% FLG at 500 r·min−<sup>1</sup> speed under 40 N loads. When FLG is added to the base oil, the friction coefficient of the Si3N4/Si3N<sup>4</sup> sliding pairs is drastically reduced. Figure 2b shows the COF of the Si3N4/Si3N<sup>4</sup> sliding pairs lubricated with and without different contents of FLG in three different working conditions. Figure 2b clearly shows that different amounts of graphene added to the lubricating oil have less influence on the lubrication

<sup>e</sup>ffect at low speeds (such as 500 r·min−<sup>1</sup> ), independent of load. Moreover, the lubricating effect is better at high speed and low load. At low speed and high load, the excessive load causes the protective film to rupture. Higher rotational speeds produce a lower COF. When the FLG content is 0.05 wt%, the COF is reduced the least, since the FLG is quickly removed from the wear track due to centrifugal force. As FLG content in base oil is increased, although the FLG is removed by centrifugal force at the same rate, the additional FLG on the wear track improves the lubricating effect depicted by a low friction coefficient. When FLG content is 0.1 wt%, the COF is reduced the most, decreasing by 76.33% (from 0.169 to 0.04). the lubricating effect is better at high speed and low load. At low speed and high load, the excessive load causes the protective film to rupture. Higher rotational speeds produce a lower COF. When the FLG content is 0.05 wt%, the COF is reduced the least, since the FLG is quickly removed from the wear track due to centrifugal force. As FLG content in base oil is increased, although the FLG is removed by centrifugal force at the same rate, the additional FLG on the wear track improves the lubricating effect depicted by a low friction coefficient. When FLG content is 0.1 wt%, the COF is reduced the most, decreasing by 76.33% (from 0.169 to 0.04). the lubricating effect is better at high speed and low load. At low speed and high load, the excessive load causes the protective film to rupture. Higher rotational speeds produce a lower COF. When the FLG content is 0.05 wt%, the COF is reduced the least, since the FLG is quickly removed from the wear track due to centrifugal force. As FLG content in base oil is increased, although the FLG is removed by centrifugal force at the same rate, the additional FLG on the wear track improves the lubricating effect depicted by a low friction coefficient. When FLG content is 0.1 wt%, the COF is reduced the most, decreasing by 76.33% (from 0.169 to 0.04).

*Micromachines* **2020**, *11*, x 4 of 9

**Figure 2.** Friction coefficient (COF) results for Si3N4/Si3N4 sliding pairs under three working conditions. (**a**) COF with a 40 N load and 3000 r·min<sup>−</sup>1 rotating speed with and without 0.1 wt% FLG; (**b**) comparison of COF results under three different working conditions. **Figure 2.** Friction coefficient (COF) results for Si3N<sup>4</sup> /Si3N<sup>4</sup> sliding pairs under three working conditions. (**a**) COF with a 40 N load and 3000 r·min−<sup>1</sup> rotating speed with and without 0.1 wt% FLG; (**b**) comparison of COF results under three different working conditions. conditions. (**a**) COF with a 40 N load and 3000 r·min<sup>−</sup>1 rotating speed with and without 0.1 wt% FLG; (**b**) comparison of COF results under three different working conditions.

The wear mark depth (WMD) profile curves of the Si3N4 disk lubricated by different FLG concentrations under a load of 40 N, at a rotational speed of 3000 r·min−1 are shown in Figure 3. It can be seen that the FLG concentration has a certain influence on the WMD of the Si3N4 disk. The WMD can be reduced when the Si3N4 disk is lubricated by a small amount of FLG. However, when the FLG concentration is too much, it will lead to aggregation and to the increased wear. The wear mark depth (WMD) profile curves of the Si3N<sup>4</sup> disk lubricated by different FLG concentrations under a load of 40 N, at a rotational speed of 3000 r·min−<sup>1</sup> are shown in Figure 3. It can be seen that the FLG concentration has a certain influence on the WMD of the Si3N<sup>4</sup> disk. The WMD can be reduced when the Si3N<sup>4</sup> disk is lubricated by a small amount of FLG. However, when the FLG concentration is too much, it will lead to aggregation and to the increased wear. The wear mark depth (WMD) profile curves of the Si3N4 disk lubricated by different FLG concentrations under a load of 40 N, at a rotational speed of 3000 r·min−1 are shown in Figure 3. It can be seen that the FLG concentration has a certain influence on the WMD of the Si3N4 disk. The WMD can be reduced when the Si3N4 disk is lubricated by a small amount of FLG. However, when the FLG concentration is too much, it will lead to aggregation and to the increased wear.

**Figure 3.** Wear mark depth (WMD) profile curve of the Si3N<sup>4</sup> disk lubricated by different FLG contents.

**Figure 3.** Wear mark depth (WMD) profile curve of the Si3N4 disk lubricated by different FLG contents. Figure 4 shows SEM images of wear scars on the Si3N4 disks lubricated by pure Mdol (Figure 4a) and Mdol containing 0.1 wt% FLG (Figure 4b). Many pores and deep scratches can be observed on the rubbing surface lubricated by pure Mdol, and EDS analysis reveals a surface carbon content of only 10.3 wt%. On the contrary, the tribo-surface lubricated by Mdol containing 0.1 wt% FLG becomes significantly smoother with less scratches and fewer pores compared to that lubricated **Figure 3.** Wear mark depth (WMD) profile curve of the Si3N4 disk lubricated by different FLG contents. Figure 4 shows SEM images of wear scars on the Si3N4 disks lubricated by pure Mdol (Figure 4a) and Mdol containing 0.1 wt% FLG (Figure 4b). Many pores and deep scratches can be observed on the rubbing surface lubricated by pure Mdol, and EDS analysis reveals a surface carbon content of only 10.3 wt%. On the contrary, the tribo-surface lubricated by Mdol containing 0.1 wt% FLG becomes significantly smoother with less scratches and fewer pores compared to that lubricated with pure Mdol. In addition, carbon content has increased to more than 25%. This is because FLG Figure 4 shows SEM images of wear scars on the Si3N<sup>4</sup> disks lubricated by pure Mdol (Figure 4a) and Mdol containing 0.1 wt% FLG (Figure 4b). Many pores and deep scratches can be observed on the rubbing surface lubricated by pure Mdol, and EDS analysis reveals a surface carbon content of only 10.3 wt%. On the contrary, the tribo-surface lubricated by Mdol containing 0.1 wt% FLG becomes significantly smoother with less scratches and fewer pores compared to that lubricated with pure Mdol. In addition, carbon content has increased to more than 25%. This is because FLG enters the interface of tribo-pairs and forms a lubricating protective film, preventing direct contact between the sliding pairs [15].

with pure Mdol. In addition, carbon content has increased to more than 25%. This is because FLG

between the sliding pairs [15].

between the sliding pairs [15].

enters the interface of tribo-pairs and forms a lubricating protective film, preventing direct contact

*Micromachines* **2020**, *11*, x 5 of 9

enters the interface of tribo-pairs and forms a lubricating protective film, preventing direct contact

**Figure 4.** SEM image of worn surfaces lubricated by (**a**) pure Mdol and (**b**), Mdol containing 0.1 wt% FLG; (**c**,**d**) Energy-dispersive X-ray spectroscope (EDS) maps of the areas pointed out in (**a**,**b**). **Figure 4.** SEM image of worn surfaces lubricated by (**a**) pure Mdol and (**b**), Mdol containing 0.1 wt% FLG; (**c**,**d**) Energy-dispersive X-ray spectroscope (EDS) maps of the areas pointed out in (**a**,**b**). FLG; (**c**,**d**) Energy-dispersive X-ray spectroscope (EDS) maps of the areas pointed out in (**a**,**b**).

**Figure 4.** SEM image of worn surfaces lubricated by (**a**) pure Mdol and (**b**), Mdol containing 0.1 wt%

#### *3.3. Cooling Properties 3.3. Cooling Properties 3.3. Cooling Properties*

Friction causes heat and temperature fluctuations during friction and wear testing. Temperature change with time for the Si3N4/Si3N4 sliding pairs lubricated by Mdol with various contents of FLG under different conditions are shown in Figure 5. Friction causes heat and temperature fluctuations during friction and wear testing. Temperature change with time for the Si3N4/Si3N<sup>4</sup> sliding pairs lubricated by Mdol with various contents of FLG under different conditions are shown in Figure 5. Friction causes heat and temperature fluctuations during friction and wear testing. Temperature change with time for the Si3N4/Si3N4 sliding pairs lubricated by Mdol with various contents of FLG under different conditions are shown in Figure 5.

**Figure 5.** Temperature rise for Si3N4/Si3N4 sliding pairs under different conditions. (**a**) 80 N load and 500 r·min<sup>−</sup>1 rotational speed, (**b**) 140 N load and 500 r·min<sup>−</sup>1 rotational speed, (**c**) 40 N load and 3000 r·min<sup>−</sup>1 rotational speed, and (**d**) comparison of temperature rise results of the above. **Figure 5.** Temperature rise for Si3N<sup>4</sup> /Si3N<sup>4</sup> sliding pairs under different conditions. (**a**) 80 N load and 500 r·min−<sup>1</sup> rotational speed, (**b**) 140 N load and 500 r·min−<sup>1</sup> rotational speed, (**c**) 40 N load and 3000 r·min−<sup>1</sup> rotational speed, and (**d**) comparison of temperature rise results of the above.

cooling effect of a small amount of graphene.

has formed a lubricating protective film on the wear surface.

surface area) [16]. When a small amount of FLG is added to Mdol, the overall thermal conductivity of the mixed liquid is increased [17]. During the sliding process, the COF was small, and the lubricant oil flow will take away some of the heat, furthermore, FLG particles distributed in the Mdol will promote heat transfer [18]. Therefore, the heat transfer performance is improved, resulting in a smaller temperature rise. When FLG content increases, a lot of graphene will agglomerate, leading to slower lubricant flow and reduced heat transfer. Secondly, for low rotational speeds, at the same content of FLG, the temperature rise at high loads is higher. This is because the friction coefficient is higher at heavy loads, resulting in more heat generation and therefore a rise in temperature rise. Interestingly, lubricating oils with different graphene concentrations have less influence on the temperature rise at high speed and low load, which is because the centrifugal force has a large influence at high rotation speed, causing some graphene to be scooped out of the wear track. In addition, the friction generates more heat at high speed, which cannot be offset by the

By testing different contents of FLG to the base oil, we observe that a low content of FLG is

Since Raman spectroscopy is a superior, sensitive, and widely used non-destructive characterization technique and is also an effective tool for assessing the quality of carbon materials. In this work, we use this technique to analyze the rubbing surfaces. Figure 6 shows the Raman spectra of the wear surfaces of the Si3N4 disks lubricated by Mdol containing different contents of FLG, as well as that of original FLG powder and in the case of dry friction. In contrast to dry friction (Figure 6a), it can be inferred that the band at 2194 cm−1 is the typical signal of Mdol (Figure 6b). When Si3N4 disks are lubricated by Mdol containing 0.1 wt% FLG (Figure 6c), Raman spectra of the wear surfaces not only show the lubricating oil signal, but also the FLG signal at the bands 1343 and 1595 cm−1. This indicates that FLG is adhered to the wear surfaces and forms a lubricating protective film. Compared to the Raman signal of FLG powder (Figure 6e), for all FLG (Figure 6d) and Mdol with 0.1% FLG, the intensity ratios of the D and G bands (*I*D/*I*G) are increased, which illustrates that the adhered graphene has become severely disordered [19], and as a result, the peak shape of 2D becomes weak and wide [20]. Raman spectroscopy was carried out to further confirm that the improved tribological behavior of Si3N4/Si3N4 sliding pairs is indeed attributable to the presence of FLG on the rubbing surface. These results validate our assumption that FLG is indeed present and

By testing different contents of FLG to the base oil, we observe that a low content of FLG is better in deterring temperature rise at low-speed (500 r·min−<sup>1</sup> rotating speed). First of all, this can be attributed to the excellent physical properties of FLG (high thermal conductivity and large specific surface area) [16]. When a small amount of FLG is added to Mdol, the overall thermal conductivity of the mixed liquid is increased [17]. During the sliding process, the COF was small, and the lubricant oil flow will take away some of the heat, furthermore, FLG particles distributed in the Mdol will promote heat transfer [18]. Therefore, the heat transfer performance is improved, resulting in a smaller temperature rise. When FLG content increases, a lot of graphene will agglomerate, leading to slower lubricant flow and reduced heat transfer. Secondly, for low rotational speeds, at the same content of FLG, the temperature rise at high loads is higher. This is because the friction coefficient is higher at heavy loads, resulting in more heat generation and therefore a rise in temperature rise. Interestingly, lubricating oils with different graphene concentrations have less influence on the temperature rise at high speed and low load, which is because the centrifugal force has a large influence at high rotation speed, causing some graphene to be scooped out of the wear track. In addition, the friction generates more heat at high speed, which cannot be offset by the cooling effect of a small amount of graphene.

Since Raman spectroscopy is a superior, sensitive, and widely used non-destructive characterization technique and is also an effective tool for assessing the quality of carbon materials. In this work, we use this technique to analyze the rubbing surfaces. Figure 6 shows the Raman spectra of the wear surfaces of the Si3N<sup>4</sup> disks lubricated by Mdol containing different contents of FLG, as well as that of original FLG powder and in the case of dry friction. In contrast to dry friction (Figure 6a), it can be inferred that the band at 2194 cm−<sup>1</sup> is the typical signal of Mdol (Figure 6b). When Si3N<sup>4</sup> disks are lubricated by Mdol containing 0.1 wt% FLG (Figure 6c), Raman spectra of the wear surfaces not only show the lubricating oil signal, but also the FLG signal at the bands 1343 and 1595 cm−<sup>1</sup> . This indicates that FLG is adhered to the wear surfaces and forms a lubricating protective film. Compared to the Raman signal of FLG powder (Figure 6e), for all FLG (Figure 6d) and Mdol with 0.1% FLG, the intensity ratios of the D and G bands (*I*D/*I*G) are increased, which illustrates that the adhered graphene has become severely disordered [19], and as a result, the peak shape of 2D becomes weak and wide [20]. Raman spectroscopy was carried out to further confirm that the improved tribological behavior of Si3N4/Si3N<sup>4</sup> sliding pairs is indeed attributable to the presence of FLG on the rubbing surface. These results validate our assumption that FLG is indeed present and has formed a lubricating protective film on the wear surface. *Micromachines* **2020**, *11*, x 7 of 9

**Figure 6.** Raman spectra of the rubbing surfaces of the Si3N4 disks. (**a**) Without any lubrication; (**b**–**d**) lubricated by Mdol containing different contents of FLG; (**e**) FLG powder. **Figure 6.** Raman spectra of the rubbing surfaces of the Si3N<sup>4</sup> disks. (**a**) Without any lubrication; (**b**–**d**) lubricated by Mdol containing different contents of FLG; (**e**) FLG powder.

Based on the above analysis, a schematic illustration of the tribological mechanism using FLG as a lubricant additive is shown in Figure 7. A schematic diagram of the friction experiment is shown Based on the above analysis, a schematic illustration of the tribological mechanism using FLG as a lubricant additive is shown in Figure 7. A schematic diagram of the friction experiment is shown in

rubbing interface and is sheared owing to its two-dimensional structure [21], as shown in Figure 7c. The addition of FLG into Mdol prevents direct contact between the Si3N4/Si3N4 sliding pairs because of the formation of a lubricating protective film. In addition, since FLG is easily torn under high Hertz contact pressure, disordered FLG with a much smaller particle size is formed [22]. FLG then wraps the abrasions to prevent direct contact between the sliding pairs. The reduction of the COF and anti-wear properties may be related to these mechanisms. A lubricating protective film was also confirmed in this paper, as measured by EDS analysis and Raman spectroscopy. The lubricating protective film effectively avoids direct contact between sliding pairs, thereby reducing the wear of

**Figure 7.** Schematic mechanism for FLG in Mdol. (**a**) Schematic diagram of the friction experiment, (**b**) lubrication diagram of pure lubricating oil, and (**c**) lubrication diagram of FLG suspended in base

The effects of FLG as a potential lubricating additive have been studied in detail in different experimental conditions. The results clearly show that FLG has a good effect on Mdol in improving the tribological properties of Si3N4/Si3N4 sliding pairs for high speeds and low loads, as well as decreasing temperature rise for low speeds regardless of the load level. The FLG content in Mdol has

the surface and decreasing the average friction coefficient.

oil.

**4. Conclusions** 

Figure 7a. As shown in Figure 7b, direct contact between the Si3N<sup>4</sup> ceramic ball and disk results in high Hertz contact stress. Therefore, the oil film is quickly broken, and then, the contact surface of Si3N4/Si3N<sup>4</sup> sliding pairs is seriously destroyed. When FLG is added to the base oil, it easily enters the rubbing interface and is sheared owing to its two-dimensional structure [21], as shown in Figure 7c. The addition of FLG into Mdol prevents direct contact between the Si3N4/Si3N<sup>4</sup> sliding pairs because of the formation of a lubricating protective film. In addition, since FLG is easily torn under high Hertz contact pressure, disordered FLG with a much smaller particle size is formed [22]. FLG then wraps the abrasions to prevent direct contact between the sliding pairs. The reduction of the COF and anti-wear properties may be related to these mechanisms. A lubricating protective film was also confirmed in this paper, as measured by EDS analysis and Raman spectroscopy. The lubricating protective film effectively avoids direct contact between sliding pairs, thereby reducing the wear of the surface and decreasing the average friction coefficient. in Figure 7a. As shown in Figure 7b, direct contact between the Si3N4 ceramic ball and disk results in high Hertz contact stress. Therefore, the oil film is quickly broken, and then, the contact surface of Si3N4/Si3N4 sliding pairs is seriously destroyed. When FLG is added to the base oil, it easily enters the rubbing interface and is sheared owing to its two-dimensional structure [21], as shown in Figure 7c. The addition of FLG into Mdol prevents direct contact between the Si3N4/Si3N4 sliding pairs because of the formation of a lubricating protective film. In addition, since FLG is easily torn under high Hertz contact pressure, disordered FLG with a much smaller particle size is formed [22]. FLG then wraps the abrasions to prevent direct contact between the sliding pairs. The reduction of the COF and anti-wear properties may be related to these mechanisms. A lubricating protective film was also confirmed in this paper, as measured by EDS analysis and Raman spectroscopy. The lubricating protective film effectively avoids direct contact between sliding pairs, thereby reducing the wear of the surface and decreasing the average friction coefficient.

**Figure 6.** Raman spectra of the rubbing surfaces of the Si3N4 disks. (**a**) Without any lubrication; (**b**–**d**)

as a lubricant additive is shown in Figure 7. A schematic diagram of the friction experiment is shown

lubricated by Mdol containing different contents of FLG; (**e**) FLG powder.

*Micromachines* **2020**, *11*, x 7 of 9

**Figure 7.** Schematic mechanism for FLG in Mdol. (**a**) Schematic diagram of the friction experiment, (**b**) lubrication diagram of pure lubricating oil, and (**c**) lubrication diagram of FLG suspended in base **Figure 7.** Schematic mechanism for FLG in Mdol. (**a**) Schematic diagram of the friction experiment, (**b**) lubrication diagram of pure lubricating oil, and (**c**) lubrication diagram of FLG suspended in base oil.

#### oil. **4. Conclusions**

**4. Conclusions**  The effects of FLG as a potential lubricating additive have been studied in detail in different experimental conditions. The results clearly show that FLG has a good effect on Mdol in improving the tribological properties of Si3N4/Si3N4 sliding pairs for high speeds and low loads, as well as decreasing temperature rise for low speeds regardless of the load level. The FLG content in Mdol has The effects of FLG as a potential lubricating additive have been studied in detail in different experimental conditions. The results clearly show that FLG has a good effect on Mdol in improving the tribological properties of Si3N4/Si3N<sup>4</sup> sliding pairs for high speeds and low loads, as well as decreasing temperature rise for low speeds regardless of the load level. The FLG content in Mdol has little effect on the width of the disks in the same conditions. On the contrary, it can reduce the WMD of the Si3N<sup>4</sup> disks. This paper presented a lubrication mechanism to explain the test results. The presence of lubricating protective film can effectively avoid the direct contact between the sliding pairs. In addition, abrasive particles are coated with FLG. As a result, it reduced damage to the surface of the wear scar. FLG as an additive in lubricating oil can provide a good cooling effect for the Si3N4/Si3N<sup>4</sup> sliding pairs, which can be attributed to the high thermal conductivity of FLG.

**Author Contributions:** Characterization, J.W. and D.X.; Formal analysis, X.W.; Writing—original draft preparation, L.Z.; Writing—review and editing, X.W.; Visualization, D.A.; Project administration, Y.W.; Funding acquisition, L.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Science Foundation of China (Nos. 51675353; 51805336), Natural Science Foundation of Liaoning Province (Nos. 20180550002 and 2019-ZD-0687).

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

### **References**


© 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/).

## *Article* **Lift-O**ff **Assisted Patterning of Few Layers Graphene**

#### **Alessio Verna 1,\* , Simone Luigi Marasso 1,2 , Paola Rivolo <sup>1</sup> , Matteo Parmeggiani 1,3 , Marco Laurenti <sup>1</sup> and Matteo Cocuzza 1,2**


Received: 29 May 2019; Accepted: 21 June 2019; Published: 25 June 2019

**Abstract:** Graphene and 2D materials have been exploited in a growing number of applications and the quality of the deposited layer has been found to be a critical issue for the functionality of the developed devices. Particularly, Chemical Vapor Deposition (CVD) of high quality graphene should be preserved without defects also in the subsequent processes of transferring and patterning. In this work, a lift-off assisted patterning process of Few Layer Graphene (FLG) has been developed to obtain a significant simplification of the whole transferring method and a conformal growth on micrometre size features. The process is based on the lift-off of the catalyst seed layer prior to the FLG deposition. Starting from a SiO<sup>2</sup> finished Silicon substrate, a photolithographic step has been carried out to define the micro patterns, then an evaporation of Pt thin film on Al2O<sup>3</sup> adhesion layer has been performed. Subsequently, the Pt/Al2O<sup>3</sup> lift-off step has been attained using a dimethyl sulfoxide (DMSO) bath. The FLG was grown directly on the patterned Pt seed layer by Chemical Vapor Deposition (CVD). Raman spectroscopy was applied on the patterned area in order to investigate the quality of the obtained graphene. Following the novel lift-off assisted patterning technique a minimization of the de-wetting phenomenon for temperatures up to 1000 ◦C was achieved and micropatterns, down to 10 µm, were easily covered with a high quality FLG.

**Keywords:** graphene; patterning; Pt; 2D materials; chemical vapor deposition (CVD)

### **1. Introduction**

In the last years, Single Layer or Few Layer Graphene (SLG-FLG) have been widely exploited to obtain novel high performance devices for different types of applications from electronics [1] to energy storage [2], sensors [3], biomedical implants [4] and others [5,6]. The quality of the SLG or FLG has been found to be a critical issue for the functionality of the devices and hence it is fundamental to improve the production and synthesis steps to avoid defects. To develop an SLG and FLG based device, the typical technological processes involved are: Chemical Vapor Deposition (CVD) on metal catalyst seed layers such as Cu or Ni foils [6]; the transferring step on polymer,—that is, poly(methyl methacrylate) (PMMA)—by spin coating and wet etching; the deposition on active regions, that is, metallic electrodes on SiO<sup>2</sup> finished Si substrates; and finally, the patterning step by plasma etching, laser ablation or other techniques [7]. The transfer of graphene onto arbitrary substrates is generally accomplished by polymer-assisted procedures. The transfer process consists of removing graphene from the growing substrate with the aid of a sacrificial polymer layer. For this purpose, several polymers like poly(methyl methacrylate) and polyvinylidene fluoride (PVDF) are widely employed as sacrificial supports [7,8]. Removing of the growing catalyst substrate is accomplished by chemical

etching [9] or by electrochemical delamination (ED) [10]. Finally, the graphene/polymer self-standing membrane is attached on the target substrate and the polymer is then removed with a suitable solvent [8]. Despite being widely used, this approach may lead to the formation of undesirable defects and damages within the graphene layer. Hence, a major challenge is to minimize such defects by contact transfer processes at wafer scale [11]. As an alternative, the direct growth of graphene on insulating substrates using Cu vapor [12] or ultrathin Cu [13,14] and Ni [15] films as catalyst has been attempted; however, these methods do not allow for a precise control of the defects. Another interesting approach is to pattern the graphene directly on Cu substrates before transfer [16] ensuring a precise reproduction of the pattern but involving time consuming polymer transfer. Due to the difficulties to grow graphene directly on Cu or Ni thin films and due to the incompatibility of these metals in most biological/electrochemical applications, usually graphene is transferred on Au or Pt electrodes exploiting the previously described methods.

Several applications require the use of Pt electrodes coated with graphene. These include atomic force microscope (AFM) tips for nanoscale electrical characterization [17], friction reduction in Micro Electro Mechanical System (MEMS) [18], counter electrodes for dye-sensitized solar cells [19], microelectrodes for neurostimulation [20], amperometric sensors in electrophoresis devices [21], electrodes for the electrochemical adsorption of dyes by cyclic voltammetry [22] and electrodes decorated with Pt nanoparticles for electrochemical applications [23,24]. Furthermore, graphene patterning on this type of electrodes is crucial to define the active area of graphene on devices like sensors [3,25,26]. This step introduces further defects and damages on the edges [27] or needs for additional technological processes [28].

Due to the high melting point (T<sup>m</sup> <sup>=</sup> <sup>1768</sup> ◦C) and low vapor pressure (1.3 <sup>×</sup> <sup>10</sup>−<sup>14</sup> mmHg), Pt is a well-known catalyst for growing graphene by CVD [29]. In fact, it has been demonstrated that sputtered Pt thin films do not suffer from de-wetting issues, typical of Cu and Ni, during monolayer or few layers graphene growth [30]. The stability of Pt thin films against de-wetting [31] phenomena at high temperature on Si/SiO<sup>2</sup> wafers can be also tuned and improved by increasing the thickness of the film [30]. In addition, the use of adhesion layers may lead to similar results with thinner films as well as e-beam evaporated films. Usual adhesion layers for Pt are metals like Ti, Ta or Cr. However, their exposure to high temperatures causes inter-diffusion or oxide formation and their subsequent degradation [32]. Alumina (Al2O3) is preferred as Pt adhesion layer in high temperature applications [33] due to its thermal and chemical stability. Moreover, it can be deposited with common techniques such as sputtering or e-beam evaporation [34] and then easily integrated in Pt deposition process. Finally, Pt is also an optimal choice for electrodes in chemical/biological sensors [35] as well as for high temperature micro-hotplates sensors [36,37] due to its high chemical and temperature stability. Therefore, the direct growth of high quality graphene on patterned Pt thin films may represent an advantage and simplification of the entire device fabrication process.

Here, the direct growth of CVD FLG on Pt thin film was obtained by a lift-off assisted patterning. Al2O<sup>3</sup> was used as adhesion layer to avoid de-wetting of Pt film. FLG was grown on patterned Al2O3/Pt substrates with features down to 10 µm. The graphene quality on patterned areas was evaluated and compared to the graphene grown on un-patterned film by Raman analysis.

#### **2. Materials and Methods**

#### *2.1. Lift-O*ff *Assisted Patterning*

The proposed novel method is based on the lift-off of the catalyst seed layer prior to the FLG deposition (Figure 1). Single side polished, P type, (100) silicon wafers (resistivity 1–10 Ω·cm) finished with 1 µm thermal oxide (supplied by Si-Mat, Kaufering, Germany) were employed for the patterning process. Two cm × two cm squares samples were used in order to fit into the graphene deposition system, which was the NANOCVD-8G system from Moorfield Nanotechnology Ltd. (Cheshire, UK). Samples were cleaned in an acetone bath, rinsed with isopropyl alcohol and then patterned using

image reversal photoresist (Microchemicals AZ 5214E, Ulm, Germany) and standard UV (ultraviolet) lithography, through a photomask. The next step was the deposition of 30 nm of Al2O<sup>3</sup> (purity 99.99%) followed by 60 nm of Pt (purity 99.99%) by electron beam evaporation (ULVAC EBX-14D, Chigasaki, Japan) with a deposition rate of 2–3 Å/s, both for Al2O<sup>3</sup> and Pt, in high vacuum (<10−<sup>5</sup> mTorr) and heating the samples at 150 ◦C during the deposition process. patterned using image reversal photoresist (Microchemicals AZ 5214E, Ulm, Germany) and standard UV (ultraviolet) lithography, through a photomask. The next step was the deposition of 30 nm of Al2O3 (purity 99.99%) followed by 60 nm of Pt (purity 99.99%) by electron beam evaporation (ULVAC EBX-14D, Chigasaki, Japan) with a deposition rate of 2–3 Å/s, both for Al2O3 and Pt, in high vacuum (<10−5 mTorr) and heating the samples at 150 °C during the deposition process.

After the deposition, the photoresist was stripped with dimethyl sulfoxide (DMSO) at 50 ◦C and the samples were rinsed with deionized water (DI) and dried with nitrogen. After the deposition, the photoresist was stripped with dimethyl sulfoxide (DMSO) at 50 °C and the samples were rinsed with deionized water (DI) and dried with nitrogen.

**Figure 1.** Process flow: (**a**) starting substrate, (**b**) photolithography, (**c**) Al2O3/Pt deposition and **Figure 1.** Process flow: (**a**) starting substrate, (**b**) photolithography, (**c**) Al2O<sup>3</sup> /Pt deposition and lift-off, (**d**) graphene growth.

#### lift-off, (**d**) graphene growth. *2.2. Graphene Deposition on Pt Film*

*2.2. Graphene Deposition on Pt Film*  Graphene was grown on Si/SiO2/Al2O3/Pt substrates by a cold-wall chemical vapor deposition (CVD) reactor operating at low pressure. To remove contaminants from the surface, a two-step annealing of the substrates was performed: 2 min at 900 °C under reducing flow of Ar at 190 sccm and H2 at 10 sccm (flow control regime) and then 30 s at 1000 °C in an atmosphere composed of Ar 90% and H2 10% at 10 torr (pressure control regime). The growth of graphene was then carried out at 1000 °C for 300 s, in a mixed atmosphere of Ar (80%), H2 (10%) and CH4 (10%) at 10 torr. The samples were finally cooled down to 200 °C under a reducing flow of Ar + H2 (190 sccm and 10 sccm) and Graphene was grown on Si/SiO2/Al2O3/Pt substrates by a cold-wall chemical vapor deposition (CVD) reactor operating at low pressure. To remove contaminants from the surface, a two-step annealing of the substrates was performed: 2 min at 900 ◦C under reducing flow of Ar at 190 sccm and H<sup>2</sup> at 10 sccm (flow control regime) and then 30 s at 1000 ◦C in an atmosphere composed of Ar 90% and H<sup>2</sup> 10% at 10 torr (pressure control regime). The growth of graphene was then carried out at 1000 ◦C for 300 s, in a mixed atmosphere of Ar (80%), H<sup>2</sup> (10%) and CH<sup>4</sup> (10%) at 10 torr. The samples were finally cooled down to 200 ◦C under a reducing flow of Ar + H<sup>2</sup> (190 sccm and 10 sccm) and then to room temperature in Ar atmosphere.

#### then to room temperature in Ar atmosphere. *2.3. Characterization*

*2.3. Characterization*  Un-patterned Al2O3/Pt thin films were characterized with field emission scanning electron microscopy (FESEM) after an annealing treatment at 900 °C, 1000 °C and 1050 °C to investigate the high temperature effect related to the CVD graphene growth process. Images were obtained with FESEM ZEISS Supra 40 (Oberkochen, Germany). For this purpose, the annealing was performed in the same atmosphere and time duration previously described for the growth of graphene but Un-patterned Al2O3/Pt thin films were characterized with field emission scanning electron microscopy (FESEM) after an annealing treatment at 900 ◦C, 1000 ◦C and 1050 ◦C to investigate the high temperature effect related to the CVD graphene growth process. Images were obtained with FESEM ZEISS Supra 40 (Oberkochen, Germany). For this purpose, the annealing was performed in the same atmosphere and time duration previously described for the growth of graphene but excluding CH<sup>4</sup> in the gas mixture.

excluding CH4 in the gas mixture. X-Ray Diffraction (XRD) was performed on un-patterned Si/SiO2/Al2O3/Pt substrates with the twofold aim of analysing the corresponding crystal structure and orientation and evaluating the effect of the thermal annealing at 1000 °C. XRD patterns were collected using a Panalytical X'Pert Diffractometer (PANalytical, Almelo, The Netherlands) in Bragg-Brentano configuration, equipped X-Ray Diffraction (XRD) was performed on un-patterned Si/SiO2/Al2O3/Pt substrates with the twofold aim of analysing the corresponding crystal structure and orientation and evaluating the effect of the thermal annealing at 1000 ◦C. XRD patterns were collected using a Panalytical X'Pert Diffractometer (PANalytical, Almelo, The Netherlands) in Bragg-Brentano configuration, equipped with a Cu Kα radiation as X-ray source (λ = 1.54059 Å).

with a Cu Kα radiation as X-ray source (λ = 1.54059 Å). Pt/FLG substrates were characterized by means of a Renishaw InVia Reflex micro-Raman spectrometer (Renishaw plc, Wottonunder-Edge, UK), equipped with a cooled CCD camera. The Raman source was a laser diode (λ = 514.5 nm) and samples inspection occurred in backscattering Pt/FLG substrates were characterized by means of a Renishaw InVia Reflex micro-Raman spectrometer (Renishaw plc, Wottonunder-Edge, UK), equipped with a cooled CCD camera. The Raman source was a laser diode (λ = 514.5 nm) and samples inspection occurred in backscattering light collection through a 50× microscope objective for all the single spectra acquisition. The spectra of the

light collection through a 50× microscope objective for all the single spectra acquisition. The spectra

patterned structures were obtained by focusing the laser spot on their centre, while a Raman map of the 10 µm-wide circle was collected by scanning, by means of a long working distance 100× objective, a 16 µm × 16 µm area, with a 0.5 µm step. The spectral map analysis was performed by means of the Renishaw WiRE 3.2 software. To collect both the single spectra and the map, 50 mW laser power, 60 s of exposure time and 4 accumulations were employed. of the patterned structures were obtained by focusing the laser spot on their centre, while a Raman map of the 10 µm-wide circle was collected by scanning, by means of a long working distance 100× objective, a 16 µm × 16 µm area, with a 0.5 µm step. The spectral map analysis was performed by means of the Renishaw WiRE 3.2 software. To collect both the single spectra and the map, 50 mW laser power, 60 s of exposure time and 4 accumulations were employed. *Micromachines* **2019**, *10*, x 4 of 11 of the patterned structures were obtained by focusing the laser spot on their centre, while a Raman map of the 10 µm-wide circle was collected by scanning, by means of a long working distance 100× objective, a 16 µm × 16 µm area, with a 0.5 µm step. The spectral map analysis was performed by

*Micromachines* **2019**, *10*, x 4 of 11

Optical images were acquired with a Nikon Eclipse ME600 microscope (Nikon, Tokyo, Japan). Optical images were acquired with a Nikon Eclipse ME600 microscope (Nikon, Tokyo, Japan). means of the Renishaw WiRE 3.2 software. To collect both the single spectra and the map, 50 mW laser power, 60 s of exposure time and 4 accumulations were employed.

#### **3. Results 3. Results**  Optical images were acquired with a Nikon Eclipse ME600 microscope (Nikon, Tokyo, Japan).

Lift-off assisted patterning of FLG has been successfully obtained (Figure 2) on Al2O3/Pt catalyst film. Lift-off assisted patterning of FLG has been successfully obtained (Figure 2) on Al2O3/Pt catalyst film. **3. Results** Lift-off assisted patterning of FLG has been successfully obtained (Figure 2) on Al2O3/Pt catalyst

**Figure 2.** Lift-off assisted patterning of few layers graphene (FLG) on Pt/Al2O3 catalyst: optical images of the patterned catalyst with different sizes features. **Figure 2.** Lift-off assisted patterning of few layers graphene (FLG) on Pt/Al2O<sup>3</sup> catalyst: optical images of the patterned catalyst with different sizes features. **Figure 2.** Lift-off assisted patterning of few layers graphene (FLG) on Pt/Al2O3 catalyst: optical

*3.1. Morphologica Characterization of De-Wetting Dynamic 3.1. Morphologica Characterization of De-Wetting Dynamic* images of the patterned catalyst with different sizes features.

The temperature effects were evaluated by thermal annealing tests on Al2O3/Pt layer at 900 °C, 1000 °C and 1050 °C (Figure 3). The temperature effects were evaluated by thermal annealing tests on Al2O3/Pt layer at 900 ◦C, 1000 ◦C and 1050 ◦C (Figure 3). *3.1. Morphologica Characterization of De-Wetting Dynamic*  The temperature effects we

annealed at 900 °C (**a**), 1000 °C (**b**) and 1050 °C (**c**). The film becomes highly discontinuous at 1050 °C although de-wetting process starts below 900 °C. **Figure 3.** Comparison of field emission scanning electron microscope (FESEM) images of Pt/Al2O3 annealed at 900 °C (**a**), 1000 °C (**b**) and 1050 °C (**c**). The film becomes highly discontinuous at 1050 °C although de-wetting process starts below 900 °C. **Figure 3.** Comparison of field emission scanning electron microscope (FESEM) images of Pt/Al2O<sup>3</sup> annealed at 900 ◦C (**a**), 1000 ◦C (**b**) and 1050 ◦C (**c**). The film becomes highly discontinuous at 1050 ◦C although de-wetting process starts below 900 ◦C.

FESEM images demonstrate that the Al2O<sup>3</sup> adhesion layer allows for controlling the de-wetting process (see supplementary information) to achieve a FLG growth temperature up to 1000 ◦C. It can be noticed that a detrimental effect appears at 1050 ◦C, where a discontinuous film is formed, making it impractical to use for most technological applications. process (see supplementary information) to achieve a FLG growth temperature up to 1000 °C. It can be noticed that a detrimental effect appears at 1050 °C, where a discontinuous film is formed, making it impractical to use for most technological applications.

FESEM images demonstrate that the Al2O3 adhesion layer allows for controlling the de-wetting

#### *3.2. XRD Characterization* The diffraction spectrum obtained by XRD investigation (Figure 4) shows a comparison

*3.2. XRD Characterization* 

The diffraction spectrum obtained by XRD investigation (Figure 4) shows a comparison between as-grown Al2O3/Pt samples and the 1000 ◦C annealed one. Apart from the contribution coming from the Si substrate (2θ–69.2◦ ), a single diffraction peak is detected at 40.1◦ in both cases and ascribed to the family of Pt(111) crystal planes (JCPDS Card 04-0802). After annealing, the crystal quality of the Pt layers turns out to be improved, as demonstrated by the higher Pt(111) peak intensity. Moreover, the (111) crystal orientation is also highly desirable for promoting graphene growth [23]. This characterization validates Al2O<sup>3</sup> as adhesion layer for this kind of application; indeed, with respect to previous work on a similar process [31], Al2O<sup>3</sup> prevents a premature de-wetting for e-beam evaporated Pt film. between as-grown Al2O3/Pt samples and the 1000 °C annealed one. Apart from the contribution coming from the Si substrate (2θ–69.2°), a single diffraction peak is detected at 40.1° in both cases and ascribed to the family of Pt(111) crystal planes (JCPDS Card 04-0802). After annealing, the crystal quality of the Pt layers turns out to be improved, as demonstrated by the higher Pt(111) peak intensity. Moreover, the (111) crystal orientation is also highly desirable for promoting graphene growth [23]. This characterization validates Al2O3 as adhesion layer for this kind of application; indeed, with respect to previous work on a similar process [31], Al2O3 prevents a premature de-wetting for e-beam evaporated Pt film.

**Figure 4.** X-ray diffraction (XRD) patterns of Al2O3/Pt samples, before and after annealing at 1000 °C. Annealed sample shows the amplification of Pt(111) phase which is suitable for graphene growth. **Figure 4.** X-ray diffraction (XRD) patterns of Al2O<sup>3</sup> /Pt samples, before and after annealing at 1000 ◦C. Annealed sample shows the amplification of Pt(111) phase which is suitable for graphene growth. The inset shows a magnification of the Pt(111) phase.

#### *3.3. Raman Characterization of Patterned Pt*

The inset shows a magnification of the Pt(111) phase.

*3.3. Raman Characterization of Patterned Pt*  The patterned Al2O3/Pt was characterized by Raman spectroscopy to evaluate the quality of the The patterned Al2O3/Pt was characterized by Raman spectroscopy to evaluate the quality of the grown graphene, which according to Wang et al. was about 2–3 layers [38].

grown graphene, which according to Wang et al. was about 2–3 layers [38]. The analysis was performed with the aim to evaluate the selective growth of FLG on Pt patterns with respect to SiO2 and the correlation between the FLG defectivity and the patterns sizes. Furthermore, growing temperature effect was investigated. Raman spectra of FLG on both un-patterned (red curve) and patterned Pt samples ranging from 5 to 100 µm wide strips were reported (Figure 5). For the patterned Pt samples, the Raman spectra were collected on a 1–2 µm wide area, far from the edges. The intensity (I), position and shape of D, G and 2D peaks (centred at ~1350, 1580 and 2700 cm−1 respectively) are similar for the 100 µm patterned and un-patterned areas but the intensity of the D peak differs on the 5 µm pattern. The presence of an ubiquitous peak at The analysis was performed with the aim to evaluate the selective growth of FLG on Pt patterns with respect to SiO<sup>2</sup> and the correlation between the FLG defectivity and the patterns sizes. Furthermore, growing temperature effect was investigated. Raman spectra of FLG on both un-patterned (red curve) and patterned Pt samples ranging from 5 to 100 µm wide strips were reported (Figure 5). For the patterned Pt samples, the Raman spectra were collected on a 1–2 µm wide area, far from the edges. The intensity (I), position and shape of D, G and 2D peaks (centred at ~1350, 1580 and 2700 cm−<sup>1</sup> respectively) are similar for the 100 µm patterned and un-patterned areas but the intensity of the D peak differs on the 5 µm pattern. The presence of an ubiquitous peak at ~2324 cm−<sup>1</sup> can be related to atmospheric N<sup>2</sup> gas fundamental vibration-rotation as previously reported [39].

~2324 cm−1 can be related to atmospheric N2 gas fundamental vibration-rotation as previously reported [39]. For the FLG growth on the un-patterned Pt sample, the G peak only differs from the patterned ones in shape and intensity with respect to the D band: a narrower and more symmetric band and an ID/I<sup>G</sup>

relevant detrimental effect.

ratio of ~0.20 are observable. Moreover, the calculated ID/I<sup>G</sup> is ~0.31 on the 100 µm pattern and ~0.73 on the 5 µm pattern suggesting that the presence of the microstructures could induce a more disordered superposition of the graphene few layers. As pointed out in the review by Ferrari and Basko [40], other factors confirm the disorder induced in the case of patterned Pt/FLG with respect to plain film; these include: the increased dispersion of the G peak; the small elbow at the right of the G peak which could be associated with a small D' peak and the noise at the left of 2D which could be associated with the D" peak. On the other hand, the shape and position of the 2D band for both samples (un-patterned and patterned) indicate that the number and the quality of the deposited graphene sheets are quite comparable, as the peak, though symmetric, cannot be fitted by one Lorentzian and it has a FWHM of ~77 cm−<sup>1</sup> , 64 cm−<sup>1</sup> and 84 cm−<sup>1</sup> respectively for un-patterned, 100 µm and 5 µm patterns, which are compatible with the characteristics of FLG grown on a nickel-coated SiO2/Si substrate, previously reported by Park et al. [41]. ~0.73 on the 5 µm pattern suggesting that the presence of the microstructures could induce a more disordered superposition of the graphene few layers. As pointed out in the review by Ferrari and Basko [40], other factors confirm the disorder induced in the case of patterned Pt/FLG with respect to plain film; these include: the increased dispersion of the G peak; the small elbow at the right of the G peak which could be associated with a small D' peak and the noise at the left of 2D which could be associated with the D'' peak. On the other hand, the shape and position of the 2D band for both samples (un-patterned and patterned) indicate that the number and the quality of the deposited graphene sheets are quite comparable, as the peak, though symmetric, cannot be fitted by one Lorentzian and it has a FWHM of ~77 cm−1, 64 cm−1 and 84 cm−1 respectively for un-patterned, 100 µm and 5 µm patterns, which are compatible with the characteristics of FLG grown on a nickel-coated SiO2/Si substrate, previously reported by Park et al. [41].

*Micromachines* **2019**, *10*, x 6 of 11

ID/IG ratio of ~0.20 are observable. Moreover, the calculated ID/IG is ~0.31 on the 100 µm pattern and

**Figure 5.** Comparative Raman spectra of un-patterned Pt/FLG (**a**), 100 µm strip pattern (**b**), 5 µm strip pattern (**c**) and blank silicon collected between two 100 µm Pt strips (**d**). The main peaks labels **Figure 5.** Comparative Raman spectra of un-patterned Pt/FLG (**a**), 100 µm strip pattern (**b**), 5 µm strip pattern (**c**) and blank silicon collected between two 100 µm Pt strips (**d**). The main peaks labels are shown.

are shown. Patterned Pt/FLG grown at 900 °C, 1000 °C and 1050 °C was further characterized by Raman spectroscopy (Figure 6). It can be noticed that at 1050 °C graphene quality improves although Pt thin film undergoes de-wetting effect and, with the increasing temperature, the ratio between 2D and G peaks also increases indicating a reduction in the number of layers. Moreover, both D peak intensity reduction and G peak sharpness indicate a minimization of the graphene defects. But from the comparison with morphological analysis (Figure 3), at 1050 °C the de-wetting of the Pt film has Patterned Pt/FLG grown at 900 ◦C, 1000 ◦C and 1050 ◦C was further characterized by Raman spectroscopy (Figure 6). It can be noticed that at 1050 ◦C graphene quality improves although Pt thin film undergoes de-wetting effect and, with the increasing temperature, the ratio between 2D and G peaks also increases indicating a reduction in the number of layers. Moreover, both D peak intensity reduction and G peak sharpness indicate a minimization of the graphene defects. But from the comparison with morphological analysis (Figure 3), at 1050 ◦C the de-wetting of the Pt film has relevant detrimental effect.

more disordered growth.

**Figure 6.** Comparative Raman spectra of 100 µm patterns of Pt/FLG grown at 900 °C (**a**), at 1000 °C (**b**) and at 1050 °C (**c**). The ratio between 2D and G peaks intensities increases indicating a reduction **Figure 6.** Comparative Raman spectra of 100 µm patterns of Pt/FLG grown at 900 ◦C (**a**), at 1000 ◦C (**b**) and at 1050 ◦C (**c**). The ratio between 2D and G peaks intensities increases indicating a reduction in the number of layers with temperature. The main peaks labels are shown.

in the number of layers with temperature. The main peaks labels are shown. Figure 7 shows FLG Raman spectra from the centre of a 5 µm strip to the border of the same pattern and then in a region 2.5 µm far from the edge. A transition from graphene to graphitic carbon residual is observed as the developing of D and G peaks indicate the presence of sp2 carbon Figure 7 shows FLG Raman spectra from the centre of a 5 µm strip to the border of the same pattern and then in a region 2.5 µm far from the edge. A transition from graphene to graphitic carbon residual is observed as the developing of D and G peaks indicate the presence of sp<sup>2</sup> carbon with a consistent number of defects as previously reported [42].

with a consistent number of defects as previously reported [42]. In order to verify the homogeneity of the FLG distribution and possible physical boundary effects, the scanning of a 16 × 16 µm2 area, including circle-shaped patterns (10 µm in diameter), was performed. The collected spectral Raman map (Figure 8) highlights that, in the inner region of the microstructure, the intensities of the D, G and 2D bands are quite constant in distribution and mutual ratio. Then, all the peak intensities increase by approaching the edge of the micro-circle, suggesting an accumulation of more defective and lower quality graphene sheets within such regions. Beyond the microstructure boundaries, no Raman features related to FLG are present, in accordance with blank spectrum (black curve) of Figure 6. This demonstrates the high selectivity of the growing process. Regarding defects accumulation on the edges, it is possible to assume that the In order to verify the homogeneity of the FLG distribution and possible physical boundary <sup>e</sup>ffects, the scanning of a 16 <sup>×</sup> <sup>16</sup> <sup>µ</sup>m<sup>2</sup> area, including circle-shaped patterns (10 <sup>µ</sup>m in diameter), was performed. The collected spectral Raman map (Figure 8) highlights that, in the inner region of the microstructure, the intensities of the D, G and 2D bands are quite constant in distribution and mutual ratio. Then, all the peak intensities increase by approaching the edge of the micro-circle, suggesting an accumulation of more defective and lower quality graphene sheets within such regions. Beyond the microstructure boundaries, no Raman features related to FLG are present, in accordance with blank spectrum (black curve) of Figure 6. This demonstrates the high selectivity of the growing process. Regarding defects accumulation on the edges, it is possible to assume that the discontinuities on the catalyst can affect the formation of graphene crystal domains thus leading to a more disordered growth.

discontinuities on the catalyst can affect the formation of graphene crystal domains thus leading to a

*Micromachines* **2019**, *10*, x 8 of 11

**Figure 7.** Raman spectra on 5 µm wide strips at different positions at the centre of the strip (**a**), on the border (**b**) and between two strips (**c**). The main peaks labels are shown. The inset shows an optical image of the sample. A graphitic carbon residual is observed in (**c**) while in un-patterned areas (Figure 5d) no carbon residuals are present. This shows that amorphous carbon deposition is **Figure 7.** Raman spectra on 5 µm wide strips at different positions at the centre of the strip (**a**), on the border (**b**) and between two strips (**c**). The main peaks labels are shown. The inset shows an optical image of the sample. A graphitic carbon residual is observed in (**c**) while in un-patterned areas (Figure 5d) no carbon residuals are present. This shows that amorphous carbon deposition is catalysed by the presence of platinum also in a halo of 1–2 µm outside the pattern. **Figure 7.** Raman spectra on 5 µm wide strips at different positions at the centre of the strip (**a**), on the border (**b**) and between two strips (**c**). The main peaks labels are shown. The inset shows an optical image of the sample. A graphitic carbon residual is observed in (**c**) while in un-patterned areas (Figure 5d) no carbon residuals are present. This shows that amorphous carbon deposition is catalysed by the presence of platinum also in a halo of 1–2 µm outside the pattern.

**Figure 8.** Raman map of a 10 µm wide circle. The map is superimposed to the optical image of 4 circles pattern for comparison. Map shows intensities distribution at 1350 cm<sup>−</sup>1 (**a**), 1580 cm<sup>−</sup>1 (**b**), 2700 cm<sup>−</sup>1 (**c**), **Figure 8.** Raman map of a 10 µm wide circle. The map is superimposed to the optical image of 4 circles pattern for comparison. Map shows intensities distribution at 1350 cm<sup>−</sup>1 (**a**), 1580 cm<sup>−</sup>1 (**b**), 2700 cm<sup>−</sup>1 (**c**), overlap of all the selected Raman shift intensity distributions (**d**). **Figure 8.** Raman map of a 10 µm wide circle. The map is superimposed to the optical image of 4 circles pattern for comparison. Map shows intensities distribution at 1350 cm−<sup>1</sup> (**a**), 1580 cm−<sup>1</sup> (**b**), 2700 cm−<sup>1</sup> (**c**), overlap of all the selected Raman shift intensity distributions (**d**).

#### overlap of all the selected Raman shift intensity distributions (**d**). **4. Discussion 4. Discussion**

**4. Discussion**  The reported analysis demonstrates that the lift-off assisted patterning is a valid method to obtain good quality FLG on Pt layer. A significant time reduction with respect to traditional process was achieved since the typical transferring steps were completely skipped. In addition, this method The reported analysis demonstrates that the lift-off assisted patterning is a valid method to obtain good quality FLG on Pt layer. A significant time reduction with respect to traditional process was achieved since the typical transferring steps were completely skipped. In addition, this method is not affected by the contamination of supporting polymers as PMMA. The obtained optimal repeatability on micrometric patterns allows for covering Pt film with every layouts and, more The reported analysis demonstrates that the lift-off assisted patterning is a valid method to obtain good quality FLG on Pt layer. A significant time reduction with respect to traditional process was achieved since the typical transferring steps were completely skipped. In addition, this method is not affected by the contamination of supporting polymers as PMMA. The obtained optimal repeatability on micrometric patterns allows for covering Pt film with every layouts and, more important, Pt can

is not affected by the contamination of supporting polymers as PMMA. The obtained optimal

be deposited with common techniques such as sputtering or e-beam evaporation and then easily integrated in a full device fabrication process. Pt represents an optimal metal selection for electrodes in chemical/biological sensors [5] as well as for high temperature micro-hotplates in Micro Electro Mechanical System (MEMS) [6], due to its high chemical and temperature stability and hence the implementation of this method is of high relevance for a wide range of applications from biosensing to neuronal stimulation.

### **5. Conclusions**

This FLG was grown directly on the patterned Pt seed layer by Chemical Vapor Deposition (CVD). The use of a proper adhesion layer, Al2O3, for the Pt film allows for raising the FLG growth temperature up to 1000 ◦C. The lift-off process of the catalyst, obtained by a standard photolithographic step, leads to a significant time reduction and consequent costs, of the graphene patterning since the typical transferring and etching steps were completely skipped, moreover an optimal repeatability on micrometric patterns can be easily obtained. The Raman characterization shows that the micropatterning was effective, and an accumulation of defects was mostly observed on the edges due to the discontinuity of the patterns. Since Pt is one of the most used materials for electrochemical or gas sensors due to its high thermal and chemical stability, the presented patterning approach has a potential high impact on the fabrication of graphene-based devices, when high quality graphene is required on noble metal electrodes. Moreover, the presented process can be applied to fabricate microelectrodes directly decorated with graphene on a whole wafer of any size avoiding the constraints correlated to polymer-assisted graphene transfer and etching.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-666X/10/6/426/s1, Figure S1: FESEM images at different magnifications of graphene growth on Pt at 900 ◦C. Figure S2: FESEM images at different magnifications of graphene growth on Pt at 1000 ◦C. Figure S3: FESEM images at different magnifications of graphene growth on Pt at 1050 ◦C. Table S1: Percentage of Pt coverage to evaluate de-wetting.

**Author Contributions:** Conceptualization, A.V.; validation, P.R., M.L.; formal analysis, P.R. and M.P.; investigation, A.V.; writing—original draft preparation, A.V.; writing—review and editing, S.L.M.; supervision, M.C.

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

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

### **References**


© 2019 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/).

*Article*
