**Sustainable Environmental Solutions**

Editors

**Sergio Ferro Marco Vocciante**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Editors* Sergio Ferro Ecas4 Australia Pty Ltd Mile End South South Australia Australia

Marco Vocciante Department of Chemistry and Industrial Chemistry University of Genoa Genova Italy

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Applied Sciences* (ISSN 2076-3417) (available at: www.mdpi.com/journal/applsci/special issues/ Sustainable Environmental Solutions).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Volume Number*, Page Range.

**ISBN 978-3-0365-1812-1 (Hbk) ISBN 978-3-0365-1811-4 (PDF)**

© 2021 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**


Reprinted from: *Applied Sciences* **2020**, *10*, 893, doi:10.3390/app10030893 . . . . . . . . . . . . . . **127**


## **About the Editors**

#### **Sergio Ferro**

Sergio Ferro was born in Italy in 1972 and obtained his Ph.D. in Chemical Sciences in 2002. His interests include electrochemical reactivity, material chemistry, surface science, and environmental chemistry, with particular attention to application aspects, such as the development of new electrode materials for industrial electrochemistry, the use of electrochemical methods for water disinfection, and the remediation of water and soils. In 2015, he left the University of Ferrara and accepted the position of technical manager at Ecas4 Australia, where he deals with the technical aspects related to the use of electrochemically activated solutions. Having worked with companies such as Industrie De Nora, Henkel, ENI, Saipem, HERA, and Electrolux, he has learned to combine commercial interests (and related intellectual property) with a decent scientific production. He authored about a hundred publications, including scientific articles and book chapters, and about ten patent applications.

#### **Marco Vocciante**

Marco Vocciante was born in Italy in 1987 and obtained his Ph.D. in Chemical Engineering in 2016; he is currently an Assistant Professor at the Department of Chemistry and Industrial Chemistry (DCCI), University of Genoa, Italy. His interests include process intensification and energy optimization, equipment fluid dynamics investigation and modeling, remediation technologies, raw materials recovery from urban and industrial waste, circular economy, management and automation of metallurgical processes, and nanoparticles synthesis by innovative and sustainable methods.

He is author of about 60 publications, including scientific articles in peer-reviewed ISI journals and book chapters. He is currently a member of the research team of the ReTeST project on the technical, economic, sustainability and risk analysis of innovative environmental technologies, in collaboration with ENI SpA and IRET-CNR, and the Erasmus+ coordinator for the interchange between DCCI and the Paderborn University, Germany.

## *Editorial* **Sustainable Environmental Solutions**

**Sergio Ferro 1,\* and Marco Vocciante <sup>2</sup>**


**Abstract:** In recent decades, increasing attention has been paid to the sustainability of products and processes, including activities aimed at environmental protection, site reclamation or treatment of contaminated effluents, as well as the valorization of waste through the recovery of resources. Although implemented with 'noble intentions', these processes are often highly invasive, unsustainable and socially unacceptable, as they involve significant use of chemical products or energy. This Special Issue is aimed at collecting research activities focused on the development of new processes to replace the above-cited obsolete practices. Taking inspiration from real problems and the need to face real cases of contamination or prevent potentially harmful situations, the development and optimization of 'smart' solutions, i.e., sustainable not only from an environmental point of view but also economically, are discussed in order to encourage as much as possible their actual implementation.

**Keywords:** environmental pollution and remediation; hazardous waste management; circular economy; soil and water reclamation; nanomaterials; sustainable processes

#### **1. Introduction**

The term 'sustainability' is generally used today when discussing possible improvements to problems such as excessive exploitation of natural resources, excessive use of energy, or the release of polluting by-products during manufacturing operations. Starting from the assumption (not always assured) that ecosystems will continue to operate and maintain the conditions that make it possible not to diminish the quality of life of today's modern societies, if a process or action causes little, less or no damage to the natural world, this process or action is considered 'sustainable'. The goal of a sustainable process/action is to maintain a balance between the exploitation of resources and the improvement of the quality of life of our modern societies and to increase the current and future potential to satisfy human needs and aspirations.

This Special Issue proposes 'sustainable environmental solutions' in relation to various activities, which have been divided into the following categories: sustainable remediation, sustainable development, and sustainable production. We hope that readers will be able to find some interesting answers or an incentive to contribute to Volume 2 of this editorial work.

#### **2. Sustainable Remediation**

Sustainable remediation is a modality of intervention in which the effects of the implementation of environmental restoration are taken into account through actions that minimize the environmental footprint, i.e., the demands in terms of energy (through the use of renewable energy), the use of materials and the production of waste (through the reuse and recycling of materials and waste), the use of water and/or the impact on water resources, the emission of air pollutants and greenhouse gas and the use of land and the impact on ecosystems.

**Citation:** Ferro, S.; Vocciante, M. Sustainable Environmental Solutions. *Appl. Sci.* **2021**, *11*, 6868. https:// doi.org/10.3390/app11156868

Received: 14 July 2021 Accepted: 23 July 2021 Published: 26 July 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/).

The use of Luminescent Solar Concentrators (LSC) in combination with phytoremediation is an example of how energy savings can be exploited to achieve self-sufficiency of greenhouses [1].

The efficiency improvement of simple and economical processes such as solid–liquid separation (a key operation in wastewater treatment) can be promoted through the optimization of the hydrodynamic behavior of suspended particles and the rheology of sludge [2].

With regard to wastewater treatment, Pietrelli et al. [3] evaluated the potential application of chitosan, a low-cost and environmentally friendly adsorbent, in the treatment of compounds with highly toxic and carcinogenic effects on biological systems such as chromium ions, while Kim and Han [4] investigated on the use of hydrogen nanobubbles to improve the electrokinetic remediation of copper-contaminated soils. Heidarrezaei et al. [5] managed to isolate and characterize a new bacterium (*Lysinibacillus boronitolerans*) capable of breaking down trichloroacetic acid, a member of the class of halogenated organic compounds widely used as solvents, herbicides and pesticides, but unfortunately carcinogenic to humans and animals, and Kulikova et al. [6] proposed a cost-effective approach to the synthesis of a magnesium potassium phosphate matrix, which is promising for the solidification of radioactive waste on an industrial scale.

On the other hand, a sustainable approach to remediation can also start from the sustainable detection of the contaminants to be addressed. An example of this is the development of a cork-modified carbon paste electrode for the determination of Pb (II), which has proved to be a sensitive electrochemical sensor capable of meeting stringent environmental control requirements while being economical, simple and highly selective [7]. In some cases, it may also be useful to conduct an economic evaluation of the intervention through the contingent evaluation (CV) to verify if the reduction is also socially advantageous or if the willingness to pay (WTP) for the reduction is greater than the costs involved in the reduction [8].

#### **3. Sustainable Development**

Speaking of sustainable development means referring to practices that make it possible to satisfy the needs of the present without compromising the ability of future generations to satisfy theirs. It means thinking about the future by balancing environmental, social and economic considerations to pursue a better quality of life. The logical or necessary consequences for society, culture, economy and the environment are interconnected and must be considered as such.

In the line of pursuing better energy efficiency in human activities, which would result in more sustainable use of resources, techniques aimed at improving the energy performances of buildings are of paramount importance, with the construction sector responsible for almost 40% of both energy consumption and the release of pollutants into the atmosphere. Among these, green roofs are becoming increasingly popular due to their ability to reduce the (electrical) energy requirements for the (summer) climatization of buildings, thus also positively influencing the internal comfort levels for the occupants [9]. The transition towards a low-carbon path should also involve agritourism buildings through the issuing of community directives, laws in member states and technical rules, including evaluation tools to assess the environmental improvements resulting from energy efficiency interventions in buildings [10].

The main obstacles to sustainable consumption are the lack of adequate infrastructure and a lack of knowledge. Infrastructure barriers in some situations make sustainable consumer behavior impossible or inconvenient (who therefore prefer other types of consumption), or in some cases require additional expenditure of time and money, thus leading to a reduction in the practice of sustainable consumer behavior [11].

To achieve the goal of ecological sustainability, influencing factors that could reduce ecological consumption need to be explored, to provide guidance for evidence-based policymaking on reducing ecological consumption [12].

Finally, the concept of virtual water, as a new approach to addressing water shortage and safety issues, can help support sustainable development in water-scarce regions [13].

#### **4. Sustainable Production**

The production of products is always linked to the extraction and consumption of natural raw materials and the use of the land. During the production process, pollutants are released into the soil, air and water and along the entire supply chain. The goal of sustainable production is to guarantee the conservation of resources and the ability of the environment to regenerate; this can be achieved by relying on processes and systems that are: non-polluting; able to limit the consumption of energy and natural resources; economically sustainable; safe and healthy for workers, communities and consumers and socially and creatively rewarding for all workers.

The growing pressure to comply with legislation and to adopt environmental strategies due to environmental concerns has led to the development of new sustainable supply chains, where a new area for a production and reconditioning system has emerged. In this complex scenario, optimal decisions can no longer avoid simultaneously considering strategies on carbon emissions, carbon tax and compulsory emissions [14].

This also applies to the food industry, with the constant search for sustainable strategies to maximize the effectiveness of the approaches and minimize the number of processes required for the production of safe food [15], the amount of drinking water required [16] and the chemicals involved, which must be as environmentally friendly and cost-effective as possible [17].

To minimize the environmental impact of industrial production, the treatment of waste linked to production is decisive, which at least in the past has always suffered from scarce attention as it is perceived as not aimed at generating value and therefore profit. In this regard, it is also appropriate to evaluate investment projects for waste treatment and try to understand their impact on the development of environmental policies [18].

#### **5. Future Advances in Sustainable Environmental Solutions**

It is widely believed that reconciling economic and environmental interests is not possible because they are conflicting interests. Many wonder whether it is actually possible to meet people's needs for food, water and energy by doing more to protect nature. We think the answer is yes, but we need a path to get there, and we need to make it urgently. Changing course over the next 10 years requires global collaboration at levels likely to be comparable to those seen after the Second World War. Protecting nature and providing water, food and energy to a growing world do not necessarily need to be mutually exclusive interests; on the contrary, energy, water, air, health and ecosystem initiatives are needed that intelligently balance economic growth and resource conservation needs. Achieving a sustainable future will depend on our ability to ensure both thriving human communities and abundant and healthy natural ecosystems.

We look forward to reporting on further advances in Volume 2 of the Special Issue 'Sustainable Environmental Solutions', which will soon be available to receive contributions from authors from around the world.

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

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

#### **References**


## *Article* **New Light on Phytoremediation: The Use of Luminescent Solar Concentrators**

**Francesca Pedron <sup>1</sup> , Martina Grifoni <sup>1</sup> , Meri Barbafieri <sup>1</sup> , Gianniantonio Petruzzelli <sup>1</sup> , Elisabetta Franchi <sup>2</sup> , Carmen Samà 3 , Liliana Gila <sup>3</sup> , Stefano Zanardi <sup>3</sup> , Stefano Palmery <sup>3</sup> , Antonio Proto <sup>3</sup> and Marco Vocciante 4,\***


**Featured Application: Under Luminescent Solar Concentrators (LSC), plants used in a phytoremediation feasibility test appear to grow better than plants grown in conventional greenhouse. This result and the energy savings characteristics of LSC highlight the prospective of LSC to further contribute in developing green remediation strategies.**

**Abstract:** The latest developments in photovoltaic studies focus on the best use of the solar spectrum through Luminescent Solar Concentrators (LSC). Due to their structural characteristics, LSC panels allow considerable energy savings. This significant saving can also be of great interest in the remediation of contaminated sites, which nowadays requires green interventions characterized by high environmental sustainability. This study reported the evaluation of LSC panels in phytoremediation feasibility tests. Three plant species were used at a microcosm scale on soil contaminated by arsenic and lead. The experiments were conducted by comparing plants grown under LSC panels doped with Lumogen Red F305 (BASF) with plants grown under polycarbonate panels used for greenhouse construction. The results showed a higher production of biomass by the plants grown under the LSC panels. The uptake of the two contaminants by plants was the same in both the growing conditions, thus resulting in an increased total accumulation (defined as metal concentration times produced biomass) in plants grown under LSC panels, indicating an overall higher phytoextraction efficiency. This seems to confirm the potential that LSCs have to be building-integrated on greenhouse roofs, canopies, and shelters to produce electricity while increasing plants productivity, thus reducing environmental pollution, and increasing sustainability.

**Keywords:** soil remediation; soil contamination; greenhouse; phytoextraction; mobilizing agents; photosynthetic efficiency; photovoltaics; luminescent dyes; energy savings; sustainability

#### **1. Introduction**

In recent years, research for photovoltaics development has been oriented towards the search for lower costs and higher conversion efficiencies. One of the fields of investigation concerns the optimal use of the solar spectrum by means of Luminescent Solar Concentrator (LSC). A typical LSC consists of a sheet of transparent material (generally polymeric matrix as PolyMethylMethAcrylate (PMMA)) where luminescent particles are homogeneously dispersed. The luminescent particles can selectively absorb the solar radiation and re-emit the energy absorbed at longer wavelengths, where PhotoVoltaic (PV) cells exhibit the highest efficiency. Therefore, the light is captured by the larger, planar surface of the slab,

**Citation:** Pedron, F.; Grifoni, M.; Barbafieri, M.; Petruzzelli, G.; Franchi, E.; Samà, C.; Gila, L.; Zanardi, S.; Palmery, S.; Proto, A.; et al. New Light on Phytoremediation: The Use of Luminescent Solar Concentrators. *Appl. Sci.* **2021**, *11*, 1923. https:// doi.org/10.3390/app11041923

Academic Editor: Luisa F. Cabeza

Received: 19 January 2021 Accepted: 19 February 2021 Published: 22 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/).

and the main part (~75%) of the converted radiation is waveguided in the slab's plane (thanks to the total internal reflection) and concentrated on the smaller PV cell area on the edges to produce electricity (Figure 1). This reduces the amount of Silicon cells needed to generate a particular amount of energy and the overall cost of the panel, since the waveguiding material is inexpensive.

**Figure 1.** Diagram of incident photons and photons emitted by a dye molecule inside the Luminescent Solar Concentrator (LSC).

The hypothesis of converting the incident solar spectrum into monochromatic light in the LSCs was already proposed at the end of the 1970s, when LSCs based on organic dyes were first introduced as a low-cost technology to enhance the power conversion efficiency (PCE) of PV solar cells through improving their spectral response [1]. They can collect both diffuse and direct solar radiation, making them a suitable technology to be used in countries where diffuse solar radiation is dominant (more than 50%) such as in northern European countries [2]. Unlike traditional PV panels, LSCs are transparent and can be made with a wide variety of colors and shapes. Thanks to these characteristics, LSCs can be seen as potential structural energy components in the design of PV windows, skylights, and colored PV panels in building facades, but also of noise barriers, advertisement signs, shelters, agricultural covers, and so on, reducing environmental pollution caused by fossil fuels.

The considerable energy savings achievable by using LSC panels could also have an interesting application in the field of remediation of contaminated sites. Several different physicochemical and biological approaches have been suggested for the remediation of contaminated water [3–5] and soils [6,7], so that selecting a suitable technology is often a difficult yet crucial step for the successful reclamation of a contaminated site [8,9]. However, activities aimed at the remediation of contaminated sites or the treatment of effluents also have an environmental impact, since they make use of chemical products or processes, with consequent consumption of raw materials and energy. In many cases these aspects are not negligible, and could compromise the sustainability of the approach or even invalidate its beneficial aspects. This is the case with the Electro Kinetic Remediation Technology (EKRT) [10,11], which has proved particularly interesting and efficient in dealing with various types of contaminations allowing in situ interventions, and demonstrates a more ecological character, compared to other approaches [12], but suffers from some critical issues including high energy consumption [13] that require new solutions to confirm the technology as operationally valid.

In light of a growing demand for remediation interventions characterized by high environmental sustainability, there is a considerable request to promote those technologies with a reduced impact on the environment, among which a primary role is attributed to phytoremediation. This term includes a series of technologies based on the use of plants to remediate organic and inorganic contaminants in soil and other environmental matrices (sediments, water). The interest in these phytotechnologies has increased over time, given some significant advantages in their use compared to traditional remediation technologies: low cost, simplicity of operation, environmental benefits. Recent contributions based on Life Cycle Assessment (LCA) comparison of different technologies clearly show the major advantage of phytoremediation in environmental impact and ecological footprint with respect to consolidated technologies or excavation and landfill disposal [14].

Phytotechnologies fall entirely within the green remediation category [15], and given their minimal environmental impact, have been proposed as an effective nature-based solution (NBS), as they ensure environmental remediation in a sustainable and economically efficient way [16,17].

Phytoremediation includes several decontamination processes; the most used are:


Among the various phytotechnologies in contaminated sites, phytoextraction is the most important for the removal of heavy metals. In the last twenty years, several innovative strategies have been developed to maximize the efficiency of the approach, which depends both on the biomass produced and the quantity of metal absorbed by the plants [18]. These strategies are aimed both to improve the performance of plants by using species with high biomass production and by increasing the bioavailability of contaminants through appropriate soil treatments with additives capable of increasing the concentration of metals in the liquid phase of the soil [19,20]. Among these innovative strategies, the possibility of relying on plant growth-promoting bacteria (PGPB) to improve the effectiveness of phytoextraction processes is of particular interest. These beneficial plant bacteria living in close association with roots have several positive effects on plant growth and develop PGPB can significantly contribute to increase plants metal uptake and, consequently, the efficiency and the rate of phytoextraction [20,21]. The increase in uptake is often further enhanced by simultaneous addition of metals mobilizing agents such as chelating compounds [22], to be selected and dosed appropriately in order not to penalize the sustainability of the approach, and possibly monitor with non-invasive approaches to avoid uncontrolled diffusion in the environment [23].

However, since phytoremediation is a highly site-specific approach, each innovation requires preliminary tests on an increasing scale to verify its effectiveness before full-scale applicability, including the setting up of dedicated greenhouses as a suitable environment for feasibility tests to optimally prepare full-scale interventions [24]. These increasing-scale trials mostly take place in greenhouses that are subject to considerable costs when they require the maintenance of optimal temperature conditions.

In a controlled environment, it is possible to better study the use of additives that increase the bioavailability of contaminants. The evaluation of the responses to plant stress and all those measures that can have significant positive effects on the efficiency of phytoremediation, particularly root-microorganism interactions, is often crucial to the success of the technology. Proper management of greenhouse conditions (light, temperature, humidity, and irrigation) can considerably improve phytoremediation feasibility tests' efficiency and speed. In semi-controlled and protected conditions, some obstacles to plant growth can be studied and overcome, such as the reduced biological activity of plants due to seasonality and the bioavailability dynamics of pollutants.

In particular, aim of this study was to provide the information necessary for evaluating the possibility of coupling Luminescent Solar Concentrators with phytoremediation in remediation procedures. Indeed, as the wave-guiding material is semi-transparent and wavelength selective, LSCs could also find promising applications in greenhouse roof panels: by selecting only the light that plants do not use for photosynthesis [25], it should be possible to produce electricity without penalizing plants growth, or possibly even increase agricultural productivity.

For this purpose, phytoextraction tests were carried out with three plant species (*Brassica juncea, Helianthus annuus* and *Lupinus albus*) to assess the effect of LSC panels on plant growth and the absorption capacity of contaminants by plants on a soil polluted by arsenic and lead. Experiments were carried out by comparing plants grown under LSC panels versus plants grown under polycarbonate panels. LSC panels doped with Lumogen Red F305 (BASF) as luminescent dye were used to be evaluated as optical filter and not as a PV device for electricity production. The parameters examined were those essential for a phytoremediation feasibility test:


This is a highly innovative perspective since, to the best of our knowledge, there are no consolidated studies and results on the use of LSC in the remediation of contaminated sites using phytoremediation technologies, when, due to soil contamination, plants should grow under significant stress conditions.

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

#### *2.1. Soil*

The soil considered was collected from a former industrial site in Tuscany (Italy) contaminated by lead (Pb) and arsenic (As) arising from manufacturing activities of various chemicals. Soil samples were collected from the 0 to 20 cm layer, air-dried, and sieved through a 2 mm sieve before laboratory analysis. In Table 1, soil pH was determined in a soil/water ratio of 1:2.5 [26], cation exchange capacity (CEC) using barium chloride (pH = 8.1) [27], and texture by the pipette method [28]. Total nitrogen (N) was determined by the Kjeldahl method [29], available phosphorus (P) by extraction with sodium bicarbonate [30], and organic matter by wet combustion [31].

**Table 1.** Physical–chemical properties of As and Pb contaminated soil. Values represent the mean (*n* = 3) ± standard deviation.


#### *2.2. Chelating Agents*

Plant uptake is mainly influenced by the bioavailable fractions rather than the total amount in the soil. For the highest efficiency in soil phytoextraction, an increased availability of soluble forms of the contaminants is required. Bioavailability depends on the soil characteristics that determine the release of Pb and As in the soil solution and plants ability to uptake and transfer the metals to their tissues.

A high cation exchange capacity (CEC) and alkaline pH reduce Pb mobility and bioavailability. Consequently, in soils contaminated by Pb, phytoextraction has many limitations, deriving from the behavior of the element in the soil environment.

To increase bioavailability, the uptake and translocation of metals, the addition of chelating agents has been extensively used in phytoextraction, with organic acids being particularly effective in increasing the solubility of metals [32,33].

For many years, chelating agents have been used to increase plants uptake of micronutrients from the soil. Chelating agents' action is mainly based on the release of metals from the soil–solid surfaces and the formation of stable metal complexes in soil solution available for plant uptake. In this experimental campaign, Ethylene Diaminete Traacetic Acid (EDTA) was selected being one of the most used chelant, which increases the uptake of several metals, Pb in particular [34]. Indeed, it was preferred to opt for a well-known and commonly used solution, with positive results in assisted phytoextraction processes, even very recent ones [35,36], in order to reduce uncertainty about this factor and focus more attention on the effect of using LSC panels.

#### *2.3. Photosynthetic Process and Selection of Luminescent Dye*

In the process of photosynthesis, plants absorb the solar radiation in the range 400–700 nanometers (this range is called Photosynthetically Active Radiation, PAR), primarily using the pigment Chlorophyll, the most abundant in the plant. In addition to chlorophyll, plants also use other pigments belonging to the carotenoid group.

The chlorophyll exists in two forms: chlorophyll *a* and chlorophyll *b***.** Chlorophyll *a* shows a strong light absorption in the blue zone of the solar spectrum as well in the red zone; chlorophyll *b* absorbs mostly blue and orange light. In contrast, both absorb poorly the green and near-green light, which reflects the typical green color of the leaves. In the near-green zone, the absorption of carotenoids takes place (Figure 2).

**Figure 2.** Absorption spectrum for chlorophylls and carotenoids (based on data from [37]).

The photosynthetic efficiency (i.e., the fraction of light energy converted into chemical energy during photosynthesis) depends on the wavelength of light. The red light (600–700 nm) is the most efficient; the efficiency increases when coupled to an equal far-red light (700–800 nm). The violet-blue light, even if less efficient (efficiency = 0.65–0.75) [38], is

necessary for photosynthesis because it promotes the development of chloroplasts, where photosynthesis takes place [39]. A low percent of violet-blue light (~7%) is enough to ensure plants good health [40]. Outside the visible spectrum, the UV light (200–400 nm) damages the chloroplasts hindering the photosynthetic process [41].

Considering this, luminescent dye Lumogen F305 (perylene-based molecule by BASF) was chosen for its spectroscopic features, high quantum yield (100%) and good photostability (5% of degraded dye after 4600 h accelerated ageing). Indeed, the absorption spectrum of Lumogen is characterized by a strong band in the range 500–600 nm with maximum absorption at 576 nm and a weak band between 400 and 500 nm; the photoluminescence emission takes place between 600 and 750 nm with a maximum fluorescence peak at 615 nm (Figure 3).

**Figure 3.** Absorption (red curve) and photoluminescent emission (blue curve) spectra of Lumogen F305.

Green light of solar spectrum, which is largely captured by Lumogen, is not absorbed by chlorophylls and carotenoids. Therefore, a LSC panel doped with Lumogen does not interfere with the photosynthetic process.

High intensity and correct distribution of transmitted sunlight are required to LSC plates for greenhouse application; these conditions allow high growth and good development of plants. Increased conversion of sunlight absorbed in electricity is also needed for making self-supporting greenhouses. In this respect, the amount of luminescent dye in the plate plays a key role.

Different dye concentrations were evaluated to determine the best one in terms of quantity and quality of transmitted light and electricity production. Transmitted light and its distribution in PAR range were measured from the transmittance UV–Vis spectra by using a Perkin Elmer UV–Vis–NIR Lambda 950 spectrometer. Every spectrum was firstly weighted for the solar spectrum by the AM 1.5 Reference Solar Spectrum. Then, the integrated area of the solar weighted transmission spectrum (in the range 400–700 nm) of the sample was perceptualized using the AM 1.5 as reference.

PAR attenuation (%) is the attenuation of the transmitted light (ITL) with respect to the incident light (IIL), as reported in Equation (1):

$$(\mathrm{I\_{TL}} - \mathrm{I\_{IL}})/\mathrm{I\_{IL}} \text{ \* 100} \tag{1}$$

#### *2.4. LSC Panel Design*

A typical greenhouse hard cover is made of double-wall PolyCarbonate (PC) panels with sizes that range from 1.2 × 0.6 m to 3 × 2 m. In contrast with this, large LSC modules rarely exceed a short-side length of 50–60 cm. In fact, the performances of larger devices are limited by self-absorption of the fluorescent dye (multiple absorption-emission events that reduce the light transport efficiency, due to the overlap between the absorption and emission spectra, Figure 3).

A reasonable trade-off is a module with a fixed short side of 50–60 cm and a variable length of 0.5 to 2 m; this size permits to have modules with a large area but where the optical path inside the slab is not so long to limit the performance due to self-absorption [42].

In addition, proper structural characteristics as rigidity and thermal insulation are required in practical use as roofing. Hence, a suitable LSC device for greenhouses requires a dedicated design with the integration of additional materials to fulfill these specifications.

In the present work, LSC panels with dimensions of 50 × 50 × 0.6 cm and 100 × 100 × 0.6 cm were used, fabricated by Altuglas (Arkema Group) using an industrial method of "cell-casting polymerization" [43]. In particular, PMMA ShieldUp® (impact resistant) was used as transparent material. LSCs with ShieldUp® are a patented technology [44] resulting from the collaboration between Arkema Group and Eni S.p.A, specially developed to provide a polymer composition that is highly resistant to shocks, remains transparent regardless of temperature, and possesses greater flexibility, all while it absorbs and re-emits light. These characteristics make it suitable for greenhouse roofing.

Measures of electrical efficiency of a complete LSC device were also performed. Eight silicon cells IXYS SLMD142H01LE (dimensions 24.7 × 0.6 cm each and an active surface of 14.8 cm<sup>2</sup> ) were glued on the four edges of the slab, wired in series, and connected to a Keithley 2602A (3A DC, 10A Pulse) digital multimeter to record the power response. A 50 × 50 × 0.6 cm device was exposed directly to the sun and the current-voltage (I-V) curves were collected.

The corresponding power conversion efficiency (PCE) was estimated though the following formula:

$$\frac{V\_{\text{OC}} \cdot f\_{\text{SC}} \cdot \text{FF}}{P\_{\text{in}}} \tag{2}$$

where in *Voc* is the open-circuit voltage, *Jsc* is the short-circuit current density, *Pin* is the intensity of the light incident on the device (Global Normal Irradiance GNI = 1000 W/m<sup>2</sup> ), and *FF* (Fill Factor) is defined by the following ratio:

$$FF = \frac{V\_{MPP} \cdot f\_{MPP}}{V\_{OC} \cdot f\_{SC}} \tag{3}$$

with *VMPP* and *JMPP* defined as the voltage and current density, respectively, corresponding to the maximum power point.

Considering Equations (2) and (3), it results *PCE* = *Pmax*/*Pin* = *VMPP* ∗ *JMPP*/*Pin*.

The power production from the panel estimated in this way represents the peak value, obtained with a naked LSC slab. However, the effective power conversion efficiency is influenced by the final configuration of the device, so measures on the building-integrated panel are deemed necessary.

#### *2.5. Experimental Design*

The microcosm tests were carried out in two different conditions: inside a greenhouse (first phase) and outdoor (second phase). In the first phase (Figure 4), the plants were grown on a microcosm scale under a red LSC panel of 100 × 100 cm supplied by Eni, positioned at the height of about 50 cm to allow adequate plant growth. As a comparison, the same tests were set up outside the LSC panel. The first phase aimed to evaluate if the red LSC panel could hinder or reduce the biomass production of the selected species.

**Figure 4.** Picture of the experiments inside the greenhouse.

In the second phase (Figure 5), outside the greenhouse, tests were set up using small boxes made of polycarbonate and LSC panels, built and supplied by Eni with a size of 50 × 50 cm. Microcosms were placed inside the red LSC box, so that the plants were totally subject to the action of the LSC panels. The same number of microcosms for each species were placed in the polycarbonate box, in order to have a comparison at the same conditions. In this second phase, metals bioavailability has been increased by the addition of EDTA for some samples.

**Figure 5.** Picture of outdoor experiments. The boxes are made of transparent polycarbonate and red LSC panels. Note: pots filled with white stones at the base of the box structures have the only function of anchoring the panels to avoid possible adverse effects in the event of a mighty wind.

In both phases, the growth of plant species commonly used for phytoremediation, *Brassica juncea (B), Lupinus albus (L),* and *Helianthus annuus (H)* [20,22] was considered to evaluate the biomass yield and the accumulation of the target metals (As and Pb).

#### 2.5.1. Indoor Microcosm Tests

The phytoextraction test was carried out in 400 mL microcosms, i.e. pots filled with soil in which the selected species are grown. Pots were filled with 300 g of As and Pb contaminated soil.

Ten microcosms per species were prepared, for a total number of microcosms of 30, 15 grown under the LSC panel and 15 outside the panel. Sowing was carried out using 0.5 g of *Brassica juncea*, 6 seeds of *Helianthus annuus,* and 5 seeds of *Lupinus albus*. The experiment was organized in a randomized complete block design.

Microcosms were watered daily (at least twice a day) according to the needs of the plants. *B. juncea* was the plant species that needed the least water. On average 20 mL for *B. juncea* and 25 mL for *L. albus* and *H. annuus*, twice a day.

The whole experiment lasted 30 days. Plants were separated into roots and shoots. Vegetal samples were accurately washed with deionized water, and roots were further sonicated for 5 min with a Branson Sonifier 250 ultrasonic processor (Branson Ultrasonic Corporation, Danbury, CT, USA) to remove the soil particles possibly still present, and then rinsed with deionized water. Vegetal samples were dried up to constant weight in a ventilated oven at 40 ◦C and each dry weight (DW) was recorded.

#### 2.5.2. Outdoor Microcosm Tests

The microcosm test was conducted outdoor under two boxes measuring 50 × 50 × 50 cm, one in transparent polycarbonate and the other consisting of red LSC panels. The amount of soil per microcosm was 300 g. The selected plant species were the same as the first phase: *B. juncea* (B), *L. albus* (L) and *H. annuus* (H). Sowing was carried out using 0.5 g of brassica seeds, 6 sunflower seeds, and 5 lupine seeds. For each species, tests were conducted in triplicate both without and with the addition of EDTA. The trial lasted about 30 days. Irrigation was carried out based on the daily need of the plants. About 20 days after sowing, treatment with the 2 mM EDTA solution started. A total dose of 10 mL of EDTA solution was added for each treated microcosm, divided into 5 days, diluting the daily dose with water. At the end of both tests, the plants were harvested by separating the leaves and stems from the roots. The fresh weight of the developed biomass was measured and after careful washing, the plant samples were placed in an oven at about 45 ◦C to dry. The dry weight was then determined, and the samples were prepared to be analyzed and evaluate the amount of As and Pb accumulated in vegetable tissues.

#### *2.6. Lead and Arsenic Analysis in Soil and Plants*

Each plant sample (roots and shoots) was ground into fine particles (<1 mm) and digested according to US-EPA 3052 [45]. Total As (using a method for the generation of hydrides) and Pb concentration in soil and plants (aerial part and roots) were analyzed by ICP-OES (Varian AX Liberty. Varian Inc., Palo Alto, CA, USA).

#### *2.7. Quality Assurance and Quality Control*

QA/QC were performed by testing two standard solutions (0.5 and 2 mg L−<sup>1</sup> ) every 10 samples. Certified reference materials, CRM ERM e CC141 for soil and CRM ERM - CD281 for plants, were used. The limit of quantification (LOQ) for Pb and As were of 5 and 50 mg L−<sup>1</sup> , respectively. The recovery of spiked samples ranged from 95% to 101% with a Relative Standard Deviation (RSD) of 1.88 of the mean for Pb and from 94 to 101% with a RSD of 1.91 of the mean for As.

#### *2.8. Statistical Analysis*

Statistical analysis was performed using STATISTICA version 6.0 (Statsoft, Inc., Tulsa, OK, USA). Effects of treatments were analyzed using one-way analysis of variance (ANOVA). Differences between means were compared and a post-hoc analysis of variance was performed using the Tukey's honestly significant difference test (*P* < 0.05).

#### **3. Results**

#### *3.1. LSC Panels Properties and Performance*

Tests with various concentrations of luminescent dye were performed in the range 40–160 ppm, in order to identify the right composition to both maximize the photosynthesis process and the energy production. LSC plate doped with 160 ppm of Lumogen F305 has been identified to have the right characteristics to be used in microcosm experiments. As reported in Table 2, it transmits a sufficient PAR light (about 30%), a low UV light (0.6%), and a high red and far-red light (90%); the blue light is also satisfactory (7.8%) [42].

**Table 2.** Photosynthetically Active Radiation (PAR) attenuation and distribution (%) of transmitted radiation at different wavelengths through PolyMethylMethAcrylate (PMMA) ShieldUp® doped with 160 ppm of Lumogen F305 luminescent dye.


A power conversion efficiency (PCE) of 1.5% (corresponding to 15 W/m<sup>2</sup> ) was obtained for the naked LSC slab (0.5 <sup>×</sup> 0.5 m<sup>2</sup> ) using Equations (1) and (2). Starting from 1.5% that is the maximum value obtained, the PMMA ShieldUp® implementation in the final configuration (to achieve the necessary structural characteristics) determines a drop, albeit modest, of the effective power conversion efficiency.

The power produced is reasonably sufficient to fulfill all or part of greenhouse needs, as airflow and water pumping irrigation, based on the characteristics of the specific greenhouse. Indeed, greenhouse electrical consumption depends on several parameters (as the location, the season of the year, technological innovation level and so on). However, the large surfaces of the greenhouse roofs can be potentially entirely covered by LSC ShieldUp® devices to maximize energy production.

#### *3.2. First phase: Indoor LSC Panel*

The first phase aimed to evaluate if the red LSC panel could hinder or reduce the biomass production of the selected species. The obtained results for the fresh weight for the aerial part and roots are reported in Figure 6. In the following, B LSC, H LSC, and L LSC indicate the brassica, sunflower and lupine plants grown under the LSC panel. B, H, and L refer to species grown outside the LSC panel, also indicated with the label CT (control).

From data showed, it can be noticed a trend towards a higher production of fresh biomass for the aerial part of the plants grown under the LSC panel. On average, considering the mean values, this increase ranges from around 24% to 31%. In Figure 6, the fresh weight values obtained for the roots have also been reported, even if these data are not particularly significant on a microcosm scale due to the difficulty of their harvesting and the modest production. Additionally, in this case, a general trend towards higher (root) biomass production under the LSC panel can be seen, although the differences between the plants grown outside or below the LSC panel are not very large.

To compare the overall biomass production, it is possible to consider the whole set of microcosms and calculate the sum of the biomass produced of the five microcosms of each species. The results are reported in Figure 7.

5.76 5.83

**Figure 6.** Fresh biomass yield (g) for aerial part and roots. Values are reported as mean ± standard deviation.

7.66

3.66

4.56

Roots Aerial

27 Fresh Biomass Yield (g)**Figure 7.** The yield of the biomass (g) of the five microcosms for each species. Data refer to the fresh weight of shoots and roots. Values are reported as mean ± standard deviation.

28.78 29.13 38.29 18 These results from fresh weights are also confirmed by the trend of the dry weight of the biomass, as shown in Table 3.

22.82

6.31 6.41 5.45 4.71 5.73 7.35 18.31 9 **Table 3.** Dry weight (mg) of the biomass of the aerial part and of the roots of plants. Values are reported as mean ± standard deviation (SD).


Note: Diff% is the increased percentage of the mean values of biomass grown under the LSC panel compared to the plants grown outside the panel.

23.20

4.64

2

4

Fresh Biomass Yield (g)

6

8

10

In this case, there are significant differences between the aerial parts of plants grown under the red panel and those grown outside the panel, with an increase of about 25%, 27%, and 28% of the mean values for *B. Juncea*, *H. Annuus,* and *L. Albus,* respectively.

Considering that energy saving is one of the strengths of "green remediation", the usefulness of LSC panels can be demonstrated if the plants cultivated grow and develop like those grown in traditional greenhouses. The results obtained seem to support this thesis. By absorbing mainly green light, red LSC panels maintain the blue spectral range necessary to activate photosynthesis. At the same time, the quantity and quality of light transmitted by the luminescent dye incorporated in these panels can improve the spectrum red fraction where the photosynthetic activity is highest [46–48]. Indeed, it should also be considered that these results were obtained with plants grown in contaminated soil, therefore under stress conditions.

#### *3.3. Arsenic and Lead Uptake by Plants*

An essential parameter to evaluate the feasibility of a phytoremediation intervention is the plant ability to absorb contaminants. The concentration of As absorbed by the plants is reported in Table 4.

**Table 4.** Concentration values (mg kg−<sup>1</sup> ) of As absorbed by the plants. Values are reported as mean ± standard deviation (SD).


No difference was found between the amount absorbed by plants grown under the LSC panel and those grown outside the panel. The values obtained, which are the average over five replicates, are not significantly different from each other; thus, it appears that the LSC panel did not have a negative influence on As uptake by plants both in the aerial part and in the roots of plants.

The average values of Pb concentration were shown in Table 5. Further, in the case of Pb, the LSC panel did not affect the absorption of the metal. The mean Pb concentration values are not significantly different in plants grown under or outside the panel.

**Table 5.** Concentration values (mg kg−<sup>1</sup> ) of Pb absorbed by the plants. Values are reported as mean ± standard deviation (SD).


#### *3.4. Total Accumulation*

The "total accumulation" (i.e., the total metal amount extracted by plants) was evaluated as product of metal concentration and aerial biomass [49]. This parameter provides an

estimation of phytoextraction efficiency, since it includes both metal uptake and vegetal biomass production [20].

Data of total accumulation are reported in Figure 8.

−

**Figure 8.** Total accumulation of As and Pb in the aerial part of plants grown under and outside the red panel. Values are reported as mean ± standard deviation.

The tests conducted in the greenhouse show positive effects on plants grown under the red LSC panel. On the contrary, since there are no differences in the absorption of contaminants, the increase in plant biomass grown under the red LSC panel also shows a beneficial effect on the total uptake values. On balance, the LSC panel could improve plant growth and development, with a consequent increase in the amount of metals removed from the contaminated soil.

#### *3.5. Second Phase: Outdoor Comparison between LSC and Polycarbonate Boxes*

As described above, the microcosm test was conducted outdoor using two boxes, one in transparent polycarbonate and the other consisting of red LSC panels.

In this case, the lighting conditions of the plants in the microcosms are significantly different from those of the first phase of the experiments. Indeed, there is no longer the shielding due to the greenhouse under which the first phase tests were conducted: the two boxes were placed outdoor, directly under the sunlight. In addition, for the LSC box all sides are made of red LSC panels, condition that should simulate on a small scale the effect of a hypothetical greenhouse made up exclusively of red LSC panels (in the comparison box all sides are made of polycarbonate).

In general, the plants grew well, even after adding EDTA. Additionally, in this case, the results obtained show that the production of fresh biomass, especially in the aerial part, is higher for plants grown under the red box (Figure 9).

**Figure 9.** Fresh biomass production of shoots and roots. LSC the plants grown under the red panels and T the plants that have been treated with Ethylene Diamine Tetraacetic Acid (EDTA). Values are reported as mean ± standard deviation.

The trend of dry weight of the biomass of the plants is similar to that of the fresh weight; the data has been reported in Table 6.


**Table 6.** Dry weight (mg) of the biomass of the aerial part and of the roots of plants. Values are reported as mean ± standard deviation (SD).

Note: Diff% is the increased percentage of the mean values of biomass grown under the LSC panel compared to the plants grown outside the panel.

A picture of plants grown under the two different boxes before EDTA addition is reported in Figure 10. As an example, for each of the three plant species, two microcosms grown in the LSC box (on the left of the viewer) are compared with two microcosms grown in the polycarbonate box (on the right).

**Figure 10.** A picture of plants grown under the two different boxes, (**a**) *L. albus*, (**b**) *H. annuus,* and (**c**) *B. juncea*. On the left of the beholder, the plants grown under the LSC box.

> It must be emphasized that the addition of EDTA (not reported in Figure 10) did not show a negative effect on biomass production. This result can be attributed to both the fractional addition of the chelating agent and the short growth period of the plants in the microcosm tests.

> The concentration values of As and Pb in the aerial part and in the roots of the plants are shown in Table 7.

> From the results obtained, it can be seen that the concentration of Pb and As is in general very similar for the plants grown under the red box and under the polycarbonate one. In some cases, the plants under the red box even showed a slightly improvement in the absorption of the contaminants. In general, the present experiment results showed that in all plants, EDTA addition increased Pb concentrations in shoots compared with the control.

> The addition of EDTA increased the Pb content in the brassica plants in the aerial part by more than four times, without notable differences between the plants grown under the red box and those under the polycarbonate box.

> − − − − A similar increase was also found in the aerial part of the sunflower, with an increase in the concentration of Pb of about 4.5 times in plants grown under the polycarbonate box and about 6 times in those grown under the red box. The most relevant effects of the addition of EDTA were found in the lupine plants with increases in Pb concentration of the aerial part of about 10 times the value found in the plants not treated with the chelating agent. The results were the same for both boxes.

> − The effect of EDTA was instead not very evident in the root system, where the difference in the concentration of Pb between treated and untreated microcosms was always minimal.

> Lead is not easily transferred to above-ground plant biomass, since it is mainly stored in root cells [50,51]. In this experiment, the addition of EDTA has proved to be particularly

effective for phytoextraction because it seems to have favored the translocation of the metal in the aerial part. It can be supposed that EDTA chelates Pb in the soil liquid phase then the soluble Pb–EDTA complex enters the roots and Pb is transported through the plant and accumulated in the aerial part [52].

**Table 7.** Mean concentration values (mg kg−<sup>1</sup> ) of As and Pb absorbed by the plants. Values are reported as mean ± standard deviation (SD).


The results did not show any adverse effect of EDTA on arsenic uptake. This can be ascribed to the action that EDTA carries out on iron oxides, partially solubilized by the complexing agent [53]. The mobility of arsenic in soil is greatly influenced by the presence of Fe-oxides where significant amounts of As are adsorbed [54]. Arsenate forms outer-sphere complexes by electrostatic coulombic interactions on all variable charge minerals [54]. It can be hypothesized that the disruptive effects of the oxides by EDTA release the arsenic that goes into soil solution becoming available for plants. As a matter of fact, in the specific contaminated soil, the plants can uptake both the contaminants even if not essential elements. There were no differences in the concentration of the two metals between the plants grown in the polycarbonate box and the LSC red box. After all, the EDTA metal complexes are poorly photodegradable in the soil, mostly when plants have grown and in the alkaline conditions of soil pH [55].

The total accumulation of the two metals for the three investigated plants is reported in Figure 11 for As and Pb.

This parameter, which, as previously stated, is essential in evaluating the efficiency of phytoremediation, shows a positive effect of the LSC panels, as it is generally higher for tests carried out in the red box. The results showed that plants development under the red box also appears visually higher than that under the polycarbonate box. Thus, the positive effects are more explicit in outdoor tests rather than in greenhouses, as in general the light conditions in the greenhouse tend to decrease due to the shielding effect of the walls of the structure.

μ **Figure 11.** Total accumulation (µg) of Pb and As in the aerial part of the plants. Values are reported as mean ± standard deviation.

#### **4. Conclusions**

In a context of growing international attention linked to the need to resort to renewable energy sources, the possibility of using LSC panels for the growth of vegetables is becoming a path pursued with great interest. LSCs can collect both diffuse and direct solar radiation, making them a suitable technology to be used in countries where diffuse solar radiation is dominant such as in northern European countries (with more than 50% diffuse light). Research is still at an early stage regarding the beneficial effects of LSCs on the plant growth, which will require a better understanding of the potential impacts of this technology on growth across the huge diversity of vegetable species.

Nevertheless, the results of the present experimentation show that LSC panels in PMMA ShieldUp doped with 160 ppm Lumogen Red F305 have the right characteristics to be used for greenhouse application, since they do not penalize the growth of plants, but rather they contribute to enhance the photosynthetic efficiency: the fluorescent dye transmits a sufficient PAR light (about 30%), a low UV light (0.6%) and a high red and far-red light (90%), while maintaining a satisfactory blue light (7.8%), necessary to improve the photosynthetic efficiency.

Despite the preliminary nature of the conducted tests, which to the best of the authors knowledge are absolutely innovative as there are no similar experiences even at an international level, the results seem absolutely promising. Indeed, despite stressful conditions due to a high concentration of contaminants in the soil, the plants under the LSC panel grew well showing a higher uptake capacity with respect to plants grown in the traditional greenhouse in polycarbonate. At the same time, plants grown in the LSC greenhouse showed an interesting increase in fresh biomass production on microcosm scale. These results offer some ideas for the possible use of these materials in the field of remediation through phytoremediation, especially on a larger scale where both LSCs and phytoremediation techniques deemed to be further proved.

An on-site greenhouse capable of being energy self-sufficient allows feasibility tests to be carried out in optimal times, regardless of the climatic conditions (protection from high temperatures in summer and low temperatures in winter) even with the addition of chemical additives and biological (PGPR) to choose the best strategy for full-scale remediation activities. In this sense, the power produced by relying on LSC panels could be utilized to fulfill all or part of the greenhouse needs, as airflow and water pumping irrigation. Indeed, the large surfaces of the greenhouse roofs can be potentially entirely building-integrated with by LSC devices to maximize energy production.

About this, measures of electrical efficiency of a complete LSC device (50 <sup>×</sup> <sup>50</sup> <sup>×</sup> 0.6 cm<sup>3</sup> ) with 160 ppm Lumogen Red F305 concentration showed a power conversion efficiency (PCE) of 1.5%, equal to a maximum power production of 15 W/m<sup>2</sup> .

Depending on the greenhouse size, location, year season and technological innovation degree, a use could also be envisaged for ex situ phytoremediation activities. For example, by digging the contaminated soil at a certain depth it may be possible to place it inside the greenhouse and program the growth of plants on the soil to be reclaimed in any climatic season, considerably reducing the remediation times especially in the presence of contamination by organic compounds.

On a full scale, LSC panels can also be used as canopies, placed at a certain height on the area to be reclaimed, using appropriate anchoring systems, allowing the LSC properties to be exploited. The height of the panels should be established based on the potentially used agricultural machinery, and sowing should also be arranged in such a way as to allow access and use of these means.

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

**Funding:** This research was supported by Eni S.p.A, Research & Technological Innovation Department, San Donato Milanese (Italy) and fully funded by Syndial S.p.A. (now Eni Rewind S.p.A.), agreement number 3500047416.

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data is contained within the present article.

**Acknowledgments:** The Authors thank Irene Rosellini, IRET-CNR, for technical assistance in Phytoremediation experiments.

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

#### **References**


## *Article* **Sedimentation of Fractal Aggregates in Shear-Thinning Fluids**

#### **Marco Trofa † and Gaetano D'Avino \*,†**

Dipartimento di Ingegneria Chimica, dei Materiali e della Produzione Industriale, Università di Napoli Federico II, Piazza Giorgio Ascarelli 80, 80125 Napoli, Italy; marco.trofa@unina.it

**\*** Correspondence: gadavino@unina.it; Tel.: +39-081-7682241

† These authors contributed equally to this work.

Received: 10 April 2020; Accepted: 2 May 2020; Published: 8 May 2020

**Abstract:** Solid–liquid separation is a key unit operation in the wastewater treatment, generally consisting of coagulation and flocculation steps to promote aggregation and increase the particle size, followed by sedimentation, where the particles settle due to the effect of gravity. The sedimentation efficiency is related to the hydrodynamic behavior of the suspended particles that, in turn, depends on the aggregate morphology. In addition, the non-Newtonian rheology of sludges strongly affects the drag coefficient of the suspended particles, leading to deviations from the known settling behavior in Newtonian fluids. In this work, we use direct numerical simulations to study the hydrodynamic drag of fractal-shaped particles suspended in a shear-thinning fluid modeled by the power-law constitutive equation. The fluid dynamics governing equations are solved for an applied force with different orientations uniformly distributed over the unit sphere. The resulting particle velocities are interpolated to compute the aggregate dynamics and the drag correction coefficient. A remarkable effect of the detailed microstructure of the aggregate on the sedimentation process is observed. The orientational dynamics shows a rich behavior characterized by steady-state, bistable, and periodic regimes. In qualitative agreement with spherical particles, shear-thinning increases the drag correction coefficient. Elongated aggregates sediment more slowly than sphere-like particles, with a lower terminal velocity as the aspect ratio increases.

**Keywords:** sedimentation; drag; fractal aggregates; shear-thinning; non-Newtonian fluids; suspensions; numerical simulations

#### **1. Introduction**

Separation of solid particles suspended in liquids is a fundamental operation in the treatment of wastewater. This process generally consists of a sequence of steps, namely, coagulation and flocculation, followed by sedimentation [1]. The first operation aims at destabilizing the suspension through the addition of coagulants that neutralize the negative charges on fine solids. The small destabilized particles are able to come into contact and form larger particles called microflocs. In the flocculation step, the microflocs collide and stick together forming larger and larger particles (macroflocs). Once the suspended aggregates have reached a desired dimension, the suspension undergoes the sedimentation step, i.e., the particles are separated from the liquid through gravity or centrifugal force [2].

The knowledge of the hydrodynamic drag force, which counterbalances the sedimentation force, is crucial to the design and optimization of wastewater treatment plants. Such a force depends on the size and shape of the suspended particles, and on the rheological properties of the liquid. During the flocculation stage, the aggregates assume fractal shapes [3–6] described by the following equation [7],

$$N\_{\rm P} = k\_{\rm f} \left(\frac{R\_{\rm g}}{a}\right)^{D\_{\rm f}} \tag{1}$$

where *N*<sup>p</sup> is the number of primary particles forming the aggregate, *R*<sup>g</sup> is the radius of gyration, *D*<sup>f</sup> is the fractal dimension, *k*<sup>f</sup> is the fractal pre-factor, and *a* is the radius of the primary particles. Aggregates with a shape satisfying Equation (1) are referred to as fractal-like or quasi-fractal aggregates, as the scaling law relation is independent of whether the particle has a real scale-invariant (self-similar) morphology [8]. The fractal dimension provides a scaling law between the number of primary particles composing the aggregate and a characteristic cluster size (e.g., the gyration radius). It assumes values between 1 and 3 corresponding to rod-like and spherical-like particles, respectively. The fractal prefactor is a descriptor of the aggregate local structure and is related to the packing factor. Finally, the radius of gyration is a geometric measure of the spatial mass distribution about the aggregate center of mass.

The particle morphology has a relevant influence on the settling velocity, and thus it must be accounted for when dealing with the sedimentation process of flocculated particles [9,10]. Therefore, it is not surprising that the hydrodynamic drag of particles with complex shapes has been thoroughly studied in the literature. Several methodologies have been proposed to compute the hydrodynamic drag of a set of spherical particles in contact, based on expansions of analytical solutions for Stokes flows [11–13] or direct numerical simulations [14–16]. Due to the linearity of the creeping flow equations, the dynamics of a particle with arbitrary shape can be determined by the mobility tensor that univocally relates translational and angular velocities to the forces and torques acting on the particle [17]. The knowledge of the mobility tensor allows to completely predict the translational and orientational motion of the aggregate. In this regards, it is well known that the settling velocity depends on the aggregate orientation. The average velocity over all possible orientations can be linearly related to the applied force, with a proportionality constant given by the arithmetic mean of the three eigenvalues of the translational mobility tensor divided by the fluid viscosity [13,18,19]. A hydrodynamic radius can be defined as the radius of a sphere that gives the same drag force acting on the aggregate in a uniform flow [13]. It can be readily seen that the hydrodynamic radius is inversely proportional to the aforementioned mean of the eigenvalues of the translational mobility tensor. The knowledge of the hydrodynamic radius for an aggregate with arbitrary shape is, then, sufficient to characterize its average settling velocity. The ratio of the hydrodynamic radius and the gyration radius has been found to be a function of the parameters of the fractal Equation (1) [13,19,20]. Specifically, such a ratio is an increasing function of *D*<sup>f</sup> , assuming a value of ~1 for *D*<sup>f</sup> = 2, up to a limiting value of about 1.29 for *D*<sup>f</sup> = 3 [13,15,20]. The number of primary particles strongly affects the ratio for low fractal dimensions (rod-like particles), whereas it has a weak influence for more spherical aggregates. Finally, increasing the fractal pre-factor moves the ratio to higher values without altering the dependence on *D*<sup>f</sup> and *N*<sup>p</sup> [13].

All the aforementioned studies consider a Newtonian suspending liquid. In addition, the developed methodologies used to compute the hydrodynamic drag are only applicable to Newtonian fluids. Active sludges, however, show a non-Newtonian rheology [21,22], which has a strong influence on the particle settling dynamics. For a spherical particle in an unbounded shear-thinning inelastic fluid, modeled with the power-law constitutive equation, several works are available showing that the drag force deviates from the Stokes' law [23–26]. A drag correction coefficient has been defined as the ratio between the applied force and the Stokes' drag law, where the viscosity is replaced by the power-law constitutive equation with a characteristic shear rate given by the terminal settling velocity divided by the particle diameter. The coefficient is 1 for a Newtonian fluid and increases as the flow index decreases (i.e., fluid shear-thinning increases). For non-spherical particles, Tripathi et al. [27] carried out finite element simulations to study the flow of a power-law fluid over prolate and oblate spheroidal particles aligned with the flow direction. In this particular orientation, the dependence of the total drag coefficient on the flow index was found to be qualitatively similar to that observed for spherical particles. As the aspect ratio of the prolate spheroid increases, the drag becomes relatively insensitive to the degree of shear-thinning. Oblate spheroids with high aspect ratio experience a lower drag as compared to spheres.

In summary, many works exist dealing with drag calculation of particles with non-spherical shape in Newtonian liquids. Concerning power-law fluids, the only results available are for spheres [25] and spheroids [27], the latter only considering particles oriented with a principal axis along the force direction. To the best of our knowledge, a study on the combined effect of non-Newtonian rheology and complex particle shape on the hydrodynamic drag is missing.

In this work, we investigate the hydrodynamic drag experienced by fractal aggregates suspended in a non-Newtonian fluid by numerical simulations. We assume that the aggregates are sufficiently large to neglect Brownian motion and that their concentration is low enough (less than 5% in volume) to avoid hydrodynamic interactions. This allows us to consider a single-particle problem. The suspending fluid is assumed to be inelastic and shear-thinning, and is modeled by the power-law constitutive equation. A map of particle velocities is precomputed by running finite element simulations for orientations of the applied force uniformly distributed over the unit sphere. Such velocities are then interpolated and used to reconstruct the aggregate dynamics by integrating the evolution equation of the particle position and orientation. The drag correction coefficient at long times is averaged over several initial orientations and particle shapes with the same fractal parameters. The effect of the fractal dimension, the number of primary particles forming the aggregate, and the flow index is investigated.

#### **2. Mathematical Model and Numerical Method**

#### *2.1. Governing Equations*

A rigid non-Brownian aggregate is suspended in a fluid and subjected to a constant force *F*. The fluid is at rest far from the aggregate. The computational domain, shown in Figure 1e, is a sphere with radius much larger than the maximum size of the particle. The aggregate is placed at the center of the sphere. A Cartesian reference frame is selected with *x* denoting the direction of the applied force *F*, i.e., *F* = (*F*, 0, 0). The fluid velocity is set to zero on the external spherical surface whereas a rigid-body motion is imposed on the particle boundary. We denote by *x*<sup>p</sup> and *θ*<sup>p</sup> the position of the particle center of volume and the rotation angle, and by *U*<sup>p</sup> and *ω*<sup>p</sup> the translational and angular particle velocities, respectively. All the symbols used in this work are reported in Table 1.

We model the aggregate shape by a set of primary spherical particles with radius *a* arranged to satisfy the fractal Equation (1). The construction of aggregate shapes satisfying such equation can be done in several ways, for instance by iteratively adding spherical particles (particle–cluster methods) or by directly connecting clusters of particles (cluster–cluster methods) [28–32]. In this work, we adopt the particle–cluster aggregation method proposed in [33,34]. All the available algorithms are based on the generation of pseudorandom numbers. Therefore, infinite shapes for the same fractal parameters can be obtained by changing the seed of the random number generator. We report in Figure 1 two examples of structures generated with *N*<sup>p</sup> = 20, *k*<sup>f</sup> = 1.3, and *D*<sup>f</sup> = 1.5 (Figure 1a) or *D*<sup>f</sup> = 2.5 (Figure 1b).

Assuming negligible fluid and particle inertia, the fluid dynamics of the investigated system is governed by the following mass and momentum balance equations,

$$\nabla \cdot \mathbf{u} = \mathbf{0} \tag{2}$$

$$\nabla \cdot \sigma = \mathbf{0} \tag{3}$$

$$
\sigma = -p\mathbf{I} + 2\eta(\dot{\gamma})\mathbf{D} \tag{4}
$$

where *<sup>u</sup>*, *<sup>σ</sup>*, *<sup>p</sup>*, *<sup>I</sup>*, *<sup>η</sup>*, and *<sup>D</sup>* are the velocity vector, the stress tensor, the pressure, the 3 <sup>×</sup> 3 unity tensor, the fluid viscosity, and the rate-of-deformation tensor *D* = (∇*u* + (∇*u*) T )/2, respectively. We model the suspending fluid by the power-law constitutive equation:

$$
\eta(\dot{\gamma}) = m\dot{\gamma}^{n-1} \tag{5}
$$

where *m* is the consistency index, *n* is the flow index, and *γ*˙ = √ 2*D* : *D* is the effective deformation rate. This model predicts shear-thinning for *n* < 1. For *n* = 1, a Newtonian fluid with (constant) viscosity *m* is recovered.

The fluid is at rest far from the aggregate and rigid-body motion is applied at the particle boundary, resulting in the following boundary conditions at the external spherical surface,

$$
\mathfrak{u} = \mathbf{0} \tag{6}
$$

and at the surface of the aggregate,

$$
\boldsymbol{\mu} = \mathbf{U}\_{\mathrm{P}} + \boldsymbol{\omega}\_{\mathrm{P}} \times (\mathbf{x} - \mathbf{x}\_{\mathrm{P}}) \tag{7}
$$

with *x* a point of the particle boundary.


**Table 1.** List of symbols.

**Figure 1.** Examples of aggregate shapes obtained from the particle–cluster method for *N*<sup>p</sup> = 20, *k*<sup>f</sup> = 1.3, and *D*<sup>f</sup> = 1.5 (**a**), and *D*<sup>f</sup> = 2.5 (**b**). To avoid numerical issues due to the tangent point, the centers of the spheres in contact are connected with a set of cylinders with radius 0.732*a*. In panels (**c**,**d**), the final geometry of the aggregates and the surface mesh are shown. The computational domain and the mesh on the external spherical surface is displayed in panel (**e**).

Finally, we need to specify the hydrodynamic total force and torque acting on the aggregate. Under the assumption of inertialess particle, the following equations hold,

$$\mathbf{F} = \int\_{S} \boldsymbol{\sigma} \cdot \mathbf{n} \, dS \tag{8}$$

$$\mathbf{T} = \int\_{S} (\mathbf{x} - \mathbf{x}\_{\mathbb{P}}) \times (\boldsymbol{\sigma} \cdot \mathbf{n}) \, dS = \mathbf{0} \tag{9}$$

where *T* is the total torque on the particle surface *S* and *n* is the unit vector normal to the particle surface pointing from the fluid to the boundary. Notice that the aggregate is torque-free, whereas the only external force is the applied force *F*.

The solution of the governing equations gives the fluid velocity and pressure fields, along with the particle translational and angular velocities. The translational and orientational dynamics can be computed by integrating the following equations,

$$\frac{d\mathbf{x\_p}}{dt} = \mathbf{U\_p} \tag{10}$$

$$\frac{d\theta\_\mathrm{p}}{dt} = \omega\_\mathrm{p} \tag{11}$$

with initial conditions *<sup>x</sup>*p|*t*=<sup>0</sup> <sup>=</sup> *<sup>x</sup>*p,0 and *<sup>θ</sup>*p|*t*=<sup>0</sup> <sup>=</sup> *<sup>θ</sup>*p,0. Notice that, for the problem under investigation (settling dynamics of a particle in an unbounded fluid), the evolution of the aggregate center of volume does not affect the orientational dynamics. Therefore, Equation (10) can be removed from the set of equations to be solved.

The governing equations can be made dimensionless by choosing appropriate characteristic quantities for length, time, and stress. As characteristic length, we choose the effective radius of the aggregate, defined as the radius of a sphere with the same volume *V* of the aggregate, *R*eff = ( <sup>3</sup>*<sup>V</sup>* 4*π* ) 1/3 . The characteristic time is chosen as the inverse of a characteristic shear rate (*U*/*R*eff) −1 , where *U* is the velocity along the direction of the applied force. The characteristic stress is *m*(*U*/*R*eff) *n* . By using such characteristic quantities, we can recast the governing equations and boundary conditions in their dimensionless form. In these equations, only the flow index *n* appears as dimensionless parameter. Therefore, the investigated system is fully determined by specifying *n* and the geometry of the aggregate defined by the parameters in Equation (1). In this regard, the radius of the primary particles *a* is related to *R*eff through *N*p. Moreover, the radius of gyration *R*<sup>g</sup> is determined once the other three parameters in Equation (1) are chosen. Therefore, the geometrical parameters that need to be specified are the number of particles *N*p, the fractal dimension *D*<sup>f</sup> , and the fractal pre-factor *k*<sup>f</sup> .

To quantify the hydrodynamic resistance of the particle to the applied force, we introduce the drag correction coefficient [25]:

$$X = \frac{F}{6\pi m \left(\frac{U}{2R\_{\text{eff}}}\right)^{n-1} U R\_{\text{eff}}} \tag{12}$$

defined as the applied force divided by the modified Stokes drag coefficient where the Newtonian viscosity is replaced by the power-law model. Of course, *X* = 1 for a sphere in a Newtonian fluid. The value of *X* depends on the orientation of the aggregate and, as such, changes in time since the particle varies its orientation while sedimenting. As it will be discussed below, the aggregate orientation can achieve various regimes, leading to different regime drag correction coefficients *X*R. As *X*<sup>R</sup> is affected by the initial particle orientation, we average over *C* initial configurations:

$$
\langle X\_{\mathbb{R}} \rangle = \frac{1}{\mathbb{C}} \sum\_{\mathbb{C}} X\_{\mathbb{R}} \tag{13}
$$

Finally, to make the results independent of the seed used in the algorithm to generate the aggregate, for each set of fractal parameters in Equation (1), the simulation is repeated with different seeds. The ensemble-average drag correction coefficient is computed as

$$
\langle \mathbf{X\_{R}} \rangle\_{\mathbf{m}} = \frac{1}{N\_{\text{seed}}} \sum\_{\mathbf{N\_{seed}}} \langle \mathbf{X\_{R}} \rangle \tag{14}
$$

with *N*seed the number of seeds.

In this work, the fractal pre-factor is fixed to *k*<sup>f</sup> = 1.3, which is a value commonly used in the literature to describe realistic aggregate shapes [8]. The sedimentation dynamics is studied by varying the flow index, the fractal dimension, and the number of primary spheres forming the particle.

#### *2.2. Numerical Method*

The calculation of the regime drag correction coefficient requires the knowledge of the orbit followed by the aggregate while pulled by the force. This would need, at each time step, the calculation of the particle angular velocity from the solution of the set of equations presented in the previous section (as discussed above, the translational dynamics is irrelevant). Instead of directly solving the governing equations to compute the orbit, we can greatly speed-up the calculations by precomputing a database of particle translational and angular velocities for different orientations of the aggregate [35,36]. The velocities needed at the right-hand side of Equations (10) and (11) are, then, obtained by interpolating the data of the database. In this way, the computational effort is only due to the construction of the database that can be used to compute the orbit for any initial orientation. Of course, the database must be re-computed for every particle morphology.

Due to the irregular particle shape, two orthogonal vectors fixed with the particle are needed to track the evolution of its orientation. However, it should be noted that any configuration obtained by rotating the aggregate around the applied force is equivalent in terms of particle dynamics and drag coefficient. Consequently, one orientation vector is sufficient to fully describe all the possible configurations of the aggregate with respect to the applied force. This can be readily seen if we consider the particle fixed in the laboratory frame and the applied force is rotated. All the possible configurations are, indeed, obtained by rotating the applied force around the unit sphere (so just considering one orientation vector). This is, in fact, the procedure we adopted to build the database, i.e., we fix the aggregate orientation as generated by the particle–cluster algorithm and solve the fluid dynamics problem for several orientations of the applied force uniformly distributed over the unit sphere. More specifically, we divide the unit sphere in a triangular mesh with icosahedral symmetry. The orientations of the force are taken as the directions connecting the center of the unit sphere and the vertices of the icosahedral mesh. We select an icosahedral subdivision with 162 vertices, verifying that this subdivision is sufficient to assure a good accuracy of the interpolation. Indeed, by computing the interpolating functions on an icosahedral grid with 42 vertices, the maximum relative error is ~3%.

Once the database of particle velocities (and the corresponding drag correction coefficients) has been computed, the rotational dynamics of the aggregate can be calculated by integrating Equation (11). It is, however, more convenient to compute the particle dynamics in the body reference frame, i.e., by fixing the orientation of the particle and rotating the applied force. The adopted procedure is summarized in Algorithm 1. We have denoted with *f* the (time-dependent) applied force vector acting on the aggregate, and with *f* 0 its initial value. In step 3, the search procedure described in [37] is used. A linear interpolation over the spherical triangle mesh is done in step 4 [37,38]. The update of the force *f* is carried out through quaternions [39]. Specifically, in step 5, a third-order Adams–Bashforth scheme is used to integrate in time the quaternions through the angular velocity computed at the previous step. A rotation matrix in terms of quaternions is constructed and used to update *f*. The procedure just described gives the time-evolution of the applied force *f* and of the drag correction coefficient *X*(*t*) that can be used to calculate the regime drag correction coefficient *X*R. Finally, the aggregate orbit can

be reconstructed by transforming the force from the body to the lab reference frame, i.e., by rotating the aggregate according to the rotation matrix that, at every instant, transform *f* in *F* = (*F*, 0, 0) (that is the applied force in the lab frame). The procedure summarized in Algorithm 1 has been adopted to simulate the rheology of a dilute suspension of fractal aggregates in shear-thinning fluids and validated by reproducing the dynamics of spheroids in shear flow [36].

**Algorithm 1** Procedure used to update the applied force in the body reference frame


The solution of the governing equations to build the database is done by the finite element method. The particle translational and angular velocities are treated as additional unknowns, and are included in the weak form of momentum equation. Lagrange multipliers in each node of the particle surface are employed to enforce the conditions in Equations (8) and (9) [40,41]. The fluid domain is discretized by tetrahedral elements. Mesh generation issues arise due to the contact points between the spheres generated by the particle–cluster algorithm. To overcome this problem, we perform a Boolean union operation of the spheres with a set of cylinders connecting the centers of the spheres in contact. The radius of the cylinders is 0.732*a*. We checked that lower values of the cylinder radius do not significantly alter the results. The Boolean union, smoothing, and meshing of the aggregate surface is done by the library PyMesh [42]. Examples of the surface meshes for the aggregates in Figure 1a,b are shown in Figure 1c,d. The tetrahedral volume mesh is generated by Gmsh [43].

The mesh and geometrical parameters used in the simulations are reported in Table 2 for the three values of *N*<sup>p</sup> considered in this work. The symbols ∆*x*, ∆*x*out, and *R*out denote the size of the elements on the aggregate surface and on the external domain (made dimensionless by the primary particle radius *a*), and the radius of the external sphere (see Figures 1e). The number of tetrahedral elements *N*elem is reported in the last column of Table 2. Notice that, to neglect the effect of the boundary condition far from the particle, very large external domains are needed. Furthermore, as expected, bigger aggregates (i.e., higher values of *N*p) require larger external domains. We verified mesh and domain size convergence by reducing both ∆*x* and ∆*x*out, and by further increasing *R*out. The (dimensionless) time-step size depends on the flow index *n*, ranging from about 0.01 for *n* = 1 to 0.005 for *n* = 0.6.

**Table 2.** Mesh and geometrical parameters.


The accuracy of the finite element solution is checked by comparing the results for a spherical particle in a power-law fluid with those reported in Dazhi and Tanner [25]. In Figure 2, the drag correction coefficient is reported as a function of the flow index *n*. The black circles are the simulation results by Dazhi and Tanner [25] and the triangles are obtained by our simulations for different mesh resolutions and size of the external domain (the radius of the spherical particle is 1, meshes M1 and M2 have approximately 50,000 and 70,000 elements). First of all, the superposition of the triangles denotes that the results are independent of the mesh and domain size used. A fair agreement between triangles and circles is observed for values of the flow index between 0.8 and 1. For lower *n*-values, deviations between the two sets of data are observed. We believe this is due to the coarser mesh used in Dazhi and Tanner that is particularly problematic for low values of the flow index due to large gradients of the velocity field around the particle. We have further examined this point by solving the same problem in a 2D axisymmetric geometry allowing for a much more refined mesh. The results show that the data for an extremely fine mesh overlap the triangles (deviations are lower than 1%). Furthermore, by progressively coarsening the mesh, the value of *X* moves towards the black circles. It should be noted, however, that the maximum deviation between the triangles and the circles (at *n* = 0.4) is ~4%, which is a relatively low value.

**Figure 2.** Drag correction coefficient *X* as a function of the flow index *n* for a spherical particle in an unbounded power-law fluid. The black circles are the simulation results by Dazhi and Tanner [25], and the triangles are obtained by our simulations for different mesh resolutions and size of the external domain.

#### **3. Results**

We investigate the aggregate dynamics and the resulting drag correction coefficient by varying the fractal dimension, the flow index, and the number of primary particles forming the aggregate. The values selected for the three parameters are *D*<sup>f</sup> = [1.5, 2.0, 2.5], *n* = [1.0, 0.8, 0.6], and *N*<sup>p</sup> = [20, 50, 100]. As discussed in Section 2.2, for each set of these parameters, we first run single-step simulations for different orientations of the applied force uniformly distributed over the unit sphere. Figure 3a–c reports the drag correction coefficient *X* as a function of the polar and azimuthal spherical coordinates (0 ≤ *θ* ≤ *π*, −*π* ≤ *φ* ≤ *π*), identifying the orientation of the applied force for *n* = 1 (i.e., the Newtonian case), *n* = 0.8, and *n* = 0.6, respectively. The aggregate is the one shown in Figure 1a, i.e., with *N*<sup>p</sup> = 20 and *D*<sup>f</sup> = 1.5. It can be readily observed that (i) the drag correction coefficient depends on the orientation of the force; (ii) the distributions are symmetric since *X* is the same for a specific force orientation (*θ*, *φ*) and its opposite (*π* − *θ*, *φ* ± *π*); (iii) in agreement with the spherical particle case [25], the drag correction coefficient increases as the flow index decreases (see the bar legends on the right of the panels); and (iv) the distributions are not affected by the flow index (for instance, the maxima and minima are observed for the same orientations of the force). Previous results have evidenced a trend between the drag force experienced by a fractal aggregate and its area projected along the direction of the applied force [14], although this geometrical quantity is not sufficient to accurately predict the drag. The dimensionless area of the aggregate projected along the force direction is reported in Figure 3d. Specifically, we take the directions identified by the 162 vertices of the unity sphere and, for each of them, we compute the area of the aggregate projected on a plane orthogonal to that direction (identified by the spherical coordinates *θ* and *φ*). The comparison with panels (a)–(c) shows some similarities between the distributions, e.g., the position of the maxima and minima is approximately the same, although a strict correlation is not observed.

**Figure 3.** (**a**–**c**) Drag correction coefficient for the aggregate shown in Figure 1a (*N*<sup>p</sup> = 20, *D*<sup>f</sup> = 1.5) as a function of the force direction identified by the spherical coordinates (*θ*, *φ*) for the Newtonian fluid (a), the power-law fluid with *n* = 0.8 (b), and *n* = 0.6 (c). (**d**) Dimensionless area projected along the direction of the applied force for the same aggregate as in panels (a–c). The symbols denote the direction of the force attained at long times for the Newtonian fluid (circle), power-law fluid with *n* = 0.8 (square) and *n* = 0.6 (triangle).

The same quantities are reported in Figure 4 for the more sphere-like aggregate in Figure 1b (*N*<sup>p</sup> = 20, *D*<sup>f</sup> = 2.5). As for the previous case, similar distributions are observed as the flow index is varied, with higher values of *X* for more shear-thinning fluids. The projected area also shows a trend similar to the drag correction coefficient. Due to the higher value of the fractal dimension leading to an aggregate with a more spherical shape, the range of variation of both the projected area and the drag correction coefficient is narrower than in the case at *D*<sup>f</sup> = 1.5, i.e., the influence of the force orientation is weaker.

**Figure 4.** (**a**–**c**) Drag correction coefficient for the aggregate shown in Figure 1b (*N*<sup>p</sup> = 20, *D*<sup>f</sup> = 2.5) as a function of the force direction identified by the spherical coordinates (*θ*, *φ*) for the Newtonian fluid (a), the power-law fluid with *n* = 0.8, (b) and *n* = 0.6 (c). (**d**) Dimensionless area projected along the direction of the applied force for the same aggregate as in panels (a–c). The symbols denote the direction of the force attained at long times for the Newtonian fluid (circle), power-law fluid with *n* = 0.8 (square) and *n* = 0.6 (triangle).

The data presented so far refer to the instantaneous drag correction factor, i.e., the one obtained by solving the fluid dynamics equations for a fixed orientation of the force (or, equivalently, of the aggregate for a fixed force). The applied force, however, generates a rotation of the aggregate (and, of course, a translation) leading to a change of the orientation and, in turn, of the drag correction coefficient. The knowledge of the orientational dynamics of the aggregate is then crucial to determine the time evolution of the drag correction coefficient and the regime attained by the particle. By using the procedure described in the previous section, we compute the orientational dynamics of the applied force for different initial orientations. Figure 5 shows the orbits for the aggregates reported in Figure 1a (top row) and Figure 1b (bottom row), and for the Newtonian (left column) and power-law fluid with *n* = 0.6 (right column). Twelve initial orientations uniformly distributed over the unit sphere are considered (blue circles). For these sets of parameters, the orbits converge towards a unique equilibrium point (green circle) regardless of the initial orientation. In the fixed reference frame, this means that the aggregate achieves a stable orientation. Specifically, our simulations show that, once the regime is achieved, the particle still rotates around the applied force, although with a very small rotation rate (the resulting linear velocity, obtained as the angular velocity around the applied force times the effective radius, is 2–3 orders of magnitude smaller than the sedimentation velocity). It is important to point out, however, that this rotation does not influence the drag as any configuration

around the force is equivalent (i.e., the force direction is a symmetry axis). The orientations of the force corresponding to the equilibrium points (green circles in Figure 5) are shown as symbols in the previous Figures 3 and 4. It can be readily observed that shear-thinning slightly affects the equilibrium orientation only for rod-like particles, whereas it has no influence for higher values of the fractal dimension (the symbols in Figure 4 overlap). Moreover, in both cases, the equilibrium orientation does not correspond to any special value of the projected area (for instance the minimum). Therefore, this quantity is not representative of the final orientation achieved by the aggregate and, as such, it is not helpful to estimate the drag correction factor at long times. On the contrary, the detailed microstructure of the aggregate needs to be considered to correctly predict the sedimentation dynamics.

**Figure 5.** Orbits described by the orientation of the applied force for 12 initial orientations (blue circles) uniformly distributed over the unit sphere for: the aggregate shown in Figure 1a (*N*<sup>p</sup> = 20, *D*<sup>f</sup> = 1.5) in a Newtonian (**a**) or power-law fluid with *n* = 0.6 (**b**), the aggregate shown in Figure 1b (*N*<sup>p</sup> = 20, *D*<sup>f</sup> = 2.5) in a Newtonian (**c**) or power-law fluid with *n* = 0.6 (**d**). The equilibrium points are denoted by green circles.

To further investigate on the effect of aggregate morphology, we have repeated the calculations by varying the seed of the random number generator. We recall that, although the fractal parameters in Equation (1) are fixed, the morphologies obtained by varying the seed are different. In the leftmost panels of Figure 6, the regime drag correction coefficient is shown as a function of the seed for *N*<sup>p</sup> = 20 and for different values of the fractal dimension and the flow index. If the force reaches an equilibrium point, regardless of the initial orientation like the orbits shown in Figure 5, *X*<sup>R</sup> is taken as the steady-state value. These points are represented as solid circles in Figure 6. The data show that, for *N*<sup>p</sup> = 20, the specific morphology (seed) has a relatively weak effect on *X*R, with a maximum relative deviations of 7% from the average value. Furthermore, in all the investigated cases, a single equilibrium orientation is achieved, except in one case (*D*<sup>f</sup> = 2.5, seed = 1, and *n* = 0.6) that will be discussed later.

**Figure 6.** Regime drag correction coefficient *X*<sup>R</sup> and its average over the initial orientations h*X*Ri for different number of primary particles (columns), fractal dimension (rows), random seed for the aggregate generation (bands), and flow index (orange *n* = 1, red *n* = 0.8, and green *n* = 0.6). Solid circles and open squares denote steady-state and periodic regimes. The black dashes represent h*X*Ri.

A relevant quantity for the sedimentation process is the time *t*<sup>R</sup> needed to achieve the final regime. For instance, with reference to Figure 5, this is the time needed to travel along the orbits from the initial orientation to the green circle. Of course, the time strongly depends on the initial orientation. Thus, we compute the orbits followed by the aggregate with orientation starting from the 162 vertices of the spherical triangle mesh discussed in the previous section and, for each orbit, we estimate the time needed to achieve the regime within a certain tolerance. In case a single steady-state regime exists, we calculate the time the force requires to align with the equilibrium orientation within an angle tolerance of 5◦ . The results for *N*<sup>p</sup> = 20 and different values of the seed, fractal dimension, and flow index are shown as box plots in Figure 7. The lower and higher limits of each box plot represent the first and third quartile over the different initial orientations, whereas the black dash is the median. In general, the (dimensionless) time decreases as the flow index decreases, whereas it is rather unaffected by the fractal dimension, ranging between 10 and 100 for almost all the examined cases. There are, however, some exceptions leading to remarkably longer times.

**Figure 7.** Box plot of the times needed for an aggregate to reach a stable regime as a function of particle random seed, flow index, and fractal dimension. The number of primary particles is *N*p = 20. The black dash within each box represents the median of the distribution.

To investigate these particular cases more in detail , we show in Figure 8 the orbits described by the orientation of the force for (i) *n* = 1, *D*<sup>f</sup> = 1.5, seed = 9 (Figure 8a corresponding to the highest box plot in the top panel of Figure 7); (ii) *n* = 0.8, *D*<sup>f</sup> = 2.5, seed = 1 (Figure 8b corresponding to the leftmost red box plot in the bottom panel of Figure 7); and (iii) *n* = 0.6, *D*<sup>f</sup> = 2.5, seed = 1 (Figure 8c corresponding to the leftmost green box plot in the bottom panel of Figure 7). In the first case, we still observe a dynamics similar to what reported in Figure 5 with all the orbits converging to a single equilibrium point. However, at variance with the previous cases where the orbits independently moved towards the equilibrium point, now each trajectory converges first towards a common orbit and then, very slowly, to the equilibrium orientation, resulting in a drastic increase of the time needed to reach the steady-state regime. A similar dynamics is also observed for the same seed and for *n* = 0.8 (red box plot) and *n* = 0.6 (green box plot), as well as for *D*<sup>f</sup> = 2.5 and seed = 3. A different scenario is observed for the second case (Figure 8b), where the orientation of the force follows spiraling trajectories before reaching the equilibrium point, also resulting in a longer transient dynamics. In the third case reported in Figure 8c, the regime becomes periodic with the presence of a limit cycle. Therefore, while settling, the aggregate continuously changes its orientation around the applied force, coming back to the same configuration after a certain period. Notice that the cases in Figure 8b,c correspond to the same aggregate (the fractal parameters and the seed are the same) and differ for the flow index. Further, the same aggregate in a Newtonian fluid (*n* = 1) gives orbits like the ones shown in Figure 5. Thus, we conclude that, for this aggregate shape, a decrease of the flow index leads to the appearance of a bifurcation (specifically a Hopf bifurcation [44]) with a qualitative change in the regime attained by the aggregate. In the case of a periodic regime, the time reported in Figure 7 is evaluated as the

time needed to reach the limit cycle within a tolerance of 5% on *X*. Notice that the appearance of the bifurcation inverts the trend of *t*<sup>R</sup> with the flow index (the values of the box plots corresponding to seed = 1 in Figure 7c increase with decreasing *n*). As the number of primary particles of the aggregate increases, another possible scenario, depicted in Figure 8d for *N*<sup>p</sup> = 50, *D*<sup>f</sup> = 1.5, *n* = 0.6, appears. Two equilibrium regimes are observed, identified by the blue and red orbits. Therefore, depending on the initial configuration, the aggregate can orient along one of the two stable orientations.

**Figure 8.** Orbits described by the orientation of the applied force for 12 initial orientations (blue circles) uniformly distributed over the unit sphere. The parameters are (**a**) *n* = 1, *N*<sup>p</sup> = 20, *D*<sup>f</sup> = 1.5, seed = 9; (**b**) *n* = 0.8, *N*<sup>p</sup> = 20, *D*<sup>f</sup> = 2.5, seed = 1; (**c**) *n* = 0.6, *N*<sup>p</sup> = 20, *D*<sup>f</sup> = 2.5, seed = 1; (**d**) *n* = 0.6, *N*<sup>p</sup> = 50, *D*<sup>f</sup> = 1.5, seed = 9.

By increasing the complexity of the shape, i.e., by increasing *N*<sup>p</sup> and decreasing *D*<sup>f</sup> , spiraling orbits, periodic, and bistable regimes are more frequent, leading to a substantial increase of the time needed to reach the final orientation, and, more importantly, to a significant effect of the detailed morphology (i.e., the seed used to generate the aggregate) on the settling dynamics. This is illustrated in the middle and right panels of Figure 6 where the regime drag correction coefficient is shown as a function of the seed for *N*<sup>p</sup> = 50 and *N*<sup>p</sup> = 100. The periodic regime is denoted by two open squares identifying the maximum and minimum values of the oscillation, with a corresponding *X*<sup>R</sup> calculated by averaging *X* over a period. The bistability is indicated by two closed circles that represent the values of *X*<sup>R</sup> for the two equilibrium points. In Figure 6, the average of the regime drag correction coefficient over all the initial orientations (h*X*Ri in Equation (13)) is also reported as a black dash. In case of a single equilibrium orientation, the unique solid circle coincides, in fact, with the dash. When multiple regimes coexist, the black dash is closer to the one that attracts more orbits. Notice that, in some cases (see, e.g., *N*<sup>p</sup> = 50, *D*<sup>f</sup> = 1.5, seed = 9, and *n* = 0.6), the values of *X*<sup>R</sup> for the two equilibrium points are remarkably different, resulting in a relevant quantitative effect of the initial orientation on the terminal velocity. At variance with the case at *N*<sup>p</sup> = 20, the effect of the microstructure (seed) is much more relevant, leading to deviations up to 25% from the value of the drag correction coefficient averaged over the seeds. In particular, maximum deviations are found for more elongated aggregates rather than sphere-like shapes. (Indeed, in the limiting case of a spherical aggregate, different seeds would produce the same shape.)

By averaging the data in Figure 6 over the seeds, we obtain the ensemble-average drag correction coefficient h*X*Ri<sup>m</sup> reported in Figure 9. The data are shown as a function of the fractal dimension, where each panel refers to a fixed number of primary particles and the curves are parametric in the flow index. For the Newtonian case (orange symbols), h*X*Ri<sup>m</sup> can be used to calculate the hydrodynamic radius, which is found to be quantitatively consistent with the one reported in [33]. In the investigated range of *D*f , the drag correction factor is a linear decreasing function of the fractal dimension, i.e., an aggregate with a more spherical shape sediments faster than one with a rod-like morphology. In agreement with previous results for spheres [25], shear-thinning increases the drag correction coefficient. As already noted in Figure 6, higher values of h*X*Ri<sup>m</sup> are observed as *N*<sup>p</sup> increases. This effect is more evident for low fractal dimensions where a variation of the number of primary particles affects the aspect ratio of the aggregate, in turn altering the drag experienced by the particle. On the contrary, as previously remarked, for aggregates with a sphere-like shape (high *D*<sup>f</sup> ), the number of primary particles mainly affects the "resolution" of the microstructure, without substantially changing the main geometrical features. For the same reason, the error bars are larger for low *D*<sup>f</sup> and high *N*p. As a final note, we recall that, especially for low fractal dimension, a variation of the number of particles and flow index may lead to different dynamics followed by the aggregate while sedimenting. In some cases, our simulations evidenced a qualitative change of the regime attained by the particle (e.g., a bifurcation) as one of these parameters is varied, with obvious consequences on the drag correction factor and on the time needed to achieve such regime. These observations prevent us to derive a simple scaling of h*X*Ri<sup>m</sup> with *N*<sup>p</sup> and *n*. In this regard, also the averaging of h*X*Ri over different seeds is, in some sense, misleading as it combines drag correction coefficients of aggregates that can experience very different dynamics. Therefore, we point out once again the importance of considering the detailed microstructure of the aggregate to correctly predict its sedimentation dynamics.

40

**Figure 9.** Ensemble-average drag correction coefficient as a function of fractal dimension, parametric in the number of aggregate primary particles and for different flow indexes. The data standard deviation and trend line are also reported.

#### **4. Conclusions**

In this work, the hydrodynamic drag experienced by a fractal aggregate suspended in a non-Newtonian fluid is studied by numerical simulations. The aggregate shape is generated through a particle–cluster method combining equally-sized spherical particles. The power-law constitutive equation is used to model the suspending fluid. Finite element simulations are employed to solve the fluid dynamics governing equations, for orientations of the applied force uniformly distributed over the unit sphere. These velocities are interpolated and used to reconstruct the translational and orientational aggregate dynamics. The drag correction coefficient at long times is averaged over several initial orientations and particle shapes with the same fractal parameters.

The results show a relevant effect of the aggregate morphology and shear-thinning on the sedimentation dynamics. Depending on the fractal dimension, the number of primary particles forming the aggregate, and the flow index, the aggregate can undergo a variety of rotational dynamics while settling. These can lead to a stable orientation, periodic oscillations around the force direction, or coexistence of multiple equilibrium orientations, with relevant implications on the terminal velocity and the time needed to achieve the long-time regime.

The ensemble-average drag correction coefficient linearly decreases by increasing the fractal dimension in the investigated range, i.e., rod-like aggregates sediment more slowly than particles with an isotropic shape. Shear-thinning further reduces the settling velocity. At low fractal dimensions, the number of primary spheres has a relevant influence on the drag correction coefficient as it affects the aggregate aspect ratio. On the contrary, a weak effect is observed for aggregates with a sphere-like shape as an increase of the number of spheres does not produce a relevant change of the overall morphology.

The results reported in the present work highlight that the detailed particle shape needs to be considered to properly predict the sedimentation dynamics. As a matter of fact, a variation of the morphology, even with the same fractal characteristics, may lead to different transients and final regimes. Therefore, to properly understand the settling phenomenon, the connection between the shape of the aggregate and the resulting translational and rotational dynamics needs to be investigated. This will be the subject of future works.

Finally, we would like to point out that the present results, although discussed in the context of the sedimentation process, apply to any system in which a particle of fractal shape moves in a shear-thinning liquid in a uniform flow field. Indeed, regardless of the nature of the applied force, the particle experiences an hydrodynamic resistance that can be predicted from the results reported in this work. In this regard, neglecting the details of the specific morphology, our calculation can be exploited to derive a drag correlation model to be included in a computational fluid dynamic solver for simulating particle laden flows [45].

**Author Contributions:** Conceptualization, M.T. and G.D.; methodology, M.T. and G.D.; software, M.T. and G.D.; validation, M.T. and G.D.; formal analysis, M.T. and G.D.; writing—original draft preparation, M.T. and G.D.; writing—review and editing, M.T. and G.D.; visualization, M.T. and G.D.; funding acquisition, G.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was carried out in the context of the VIMMP project (www.vimmp.eu). The VIMMP project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 760907.

**Conflicts of Interest:** The authors declare no conflicts 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* **Chromium(III) Removal from Wastewater by Chitosan Flakes**

**Loris Pietrelli 1,2,\*, Iolanda Francolini <sup>2</sup> , Antonella Piozzi <sup>2</sup> , Maria Sighicelli <sup>1</sup> , Ilaria Silvestro <sup>2</sup> and Marco Vocciante <sup>3</sup>**


Received: 30 January 2020; Accepted: 7 March 2020; Published: 11 March 2020

**Featured Application: The ability of chitosan as a low-cost and environmentally friendly Cr(III) adsorbent was studied to evaluate its potential application in the field of tannery wastewater treatment, in terms of the removal of chromium ions avoiding their conversion into Cr(VI), the compounds of which exert highly toxic and carcinogenic e**ff**ects on biological systems.**

**Abstract:** Chitosan is very effective in removing metal ions through their adsorption. A preliminary investigation of the adsorption of chromium(III) by chitosan was carried out by means of batch tests as a function of contact time, pH, ion competition, and initial chromium(III) concentration. The rate of adsorption was rather rapid (t1/<sup>2</sup> < 18 min) and influenced by the presence of other metal ions. The obtained data were tested using the Langmuir and Freundlich isotherm models and, based on R<sup>2</sup> values, the former appeared better applicable than the latter. Chitosan was found to have an excellent loading capacity for chromium(III), namely 138.0 mg Cr per g of chitosan at pH = 3.8, but metal ions adsorption was strongly influenced by the pH. About 76% of the recovered chromium was then removed simply by washing the used chitosan with 0.1 M EDTA (Ethylenediaminetetraacetic acid) solution. This study demonstrates that chitosan has the potential to become an effective and low-cost agent for wastewater treatment (e.g., tannery waste) and in situ environmental remediation.

**Keywords:** chitosan; chromium; heavy metals; adsorption; kinetics; low-cost adsorbent; tannery; ion exchange

#### **1. Introduction**

The removal of heavy metal ions from aqueous solutions, both for pollution control and for the recovery of raw materials, has assumed increasing importance in recent years. Among the many metals potentially harmful to the environment and human health, chromium pollution is of considerable concern, as the metal is widely used in many industrial activities such as electroplating, leather tanning, nuclear power plants, and textile industries [1,2].

To address this problem, numerous separation techniques are currently available (e.g., adsorption, ion exchange, selective precipitation, nanofiltration, etc.), the selection of which, however, is far from trivial and deserves extensive attention to avoid a suboptimal choice or the failure of the reclamation activity [3]. In general, adsorption-based technologies have proven to be among the most viable alternatives proposed for the treatment of industrial wastewater contaminated by a wide variety of pollutants, both organic [4] and inorganic [5,6], due to the low processing and

instrumentation costs, the simplicity of operation, and the availability of different types of low-cost and environmentally friendly adsorbents. A wide range of materials, including activated carbon [7], metal oxides, carbon nanotubes, polymers, agricultural residues [8], and natural and modified clays [9], have been used successfully to adsorb heavy metals from aqueous solutions. This is even more interesting when it is possible to exploit low-cost adsorbent materials from industrial waste [10], with a double advantage for the environment, in line with operational guidelines such as the Circular Economy and the "near-zero discharge" of hazardous waste [11] established by the most recent European laws.

In this context, a very promising and inexpensive material is chitosan (poly-β-(1→4)-2-amino-2-deoxy-D-glucose), a nitrogenous polysaccharide prepared from chitin by partially deacetylating its acetoamine groups using strong alkaline solutions at about 70 ◦C. Chitosan has a high potential for the adsorption of metal ions, since it has both amino and hydroxyl groups that can act as chelation sites for metal ions. One of the most interesting advantages of chitosan is its versatility, since the material can be easily physically modified to obtain different forms of polymers such as beads [12], membranes [13], or sponges [14] for different applications. Chitosan can also be easily chemically modified to increase its applications [15]. Recently, several critical reviews have been published on the many applications of chitosan as an environmentally friendly biomaterial [16], ranging from the medical field [17] to food technology [18] and environmental protection [19].

Chromium can be found in the environment in the forms Cr(III) and Cr(VI), as its other oxidation states are not stable in aerated aqueous media [20]. The trivalent state is the most stable form in reducing conditions and is present as a cationic species (Cr(OH)2+, Cr(OH)<sup>2</sup> <sup>+</sup>), with the first or second hydrolysis products dominating at pH values from 4 to 8. The low solubility of Cr(OH)<sup>3</sup> (log k = −16.19) considerably limits the concentration of Cr(III) for pH values above about 5.

Given its high danger to biological systems, many studies have focused on the removal of Cr(VI), while very few articles deal with the adsorption of Cr(III) by chitosan. Maruca et al. [21] reported the uptake of Cr(III) ions by chitosan flakes and the effect of PO<sup>4</sup> <sup>3</sup><sup>−</sup> on the adsorption mechanism. Chui and collaborators [22] studied the removal of Cr(III) using a packed column filled with crab chitosan. More recently, Singh & Nagendran [23] reported comparative studies on the sorption of Cr(III) on chitin and chitosan in terms of a comparison between Langmuir and Freundlich isotherms. Overall, the papers often concern the application of chitosan membranes or beads, and not of chitosan flakes.

Considering that equilibrium analysis is the most important fundamental study required to evaluate the affinity of a sorbent, the ability of chitosan to remove chromium(III) by adsorption was studied in the present work to evaluate its potential application in the field of tannery wastewater treatment [1]. Numerous adsorption tests of chromium(III) on chitosan flakes were conducted to investigate the effects of contact time, pH, initial Cr(III) concentration, and, using real wastewater, ion competition. The thermodynamic behavior was assessed using the well-known Langmuir and Freundlich isotherm models.

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

#### *2.1. Material and Reagents*

Chitosan (molecular weight: 400 k, 66.9% <40 mesh, degree of deacetylation: 84–86%) was provided by Merck KGaA (Darmstadt, Germany) and used without further purification. Its surface area, estimated with the nitrogen adsorption method (BET, Quantachrome Nova 2200, Quantachrome Instruments, Boynton Beach, FL, U.S.), was equal to 1.578 m<sup>2</sup> g −1 . The water content of this commercial chitosan, determined by thermogravimetric analysis (TGA1, Mettler Toledo, OH, U.S.), was 12.7%, while its decomposition temperature was 292.12 ◦C (Figure 1). In order to study the effect of particle size on the adsorption of metal ions, two granulometric fractions (<0.42 mm and >0.42 mm) were obtained from the starting material by using a sieve shaker. Analytical grade chemicals were supplied by Merck Co. (Kenilworth, NJ, U.S.); aqueous solutions were prepared dissolving Cr(NO3)<sup>3</sup> at different

concentrations in deionized water (Millipore Milli-Q, Merck KGaA, Darmstadt, Germany), and the initial pH was adjusted by adding a few drops of HNO<sup>3</sup> and NaOH solutions.

**Figure 1.** Thermogravimetric analysis (TGA) of the chitosan sample.

#### *2.2. Adsorption Tests*

−

− − −

μ

Batch experiments (in triplicate) were carried out using 100 and 500 mg of adsorbent each time; chitosan was added to 50 mL of Cr solution in a conical flask. The stirring rate was set at 120 rpm for all adsorption/desorption tests using a temperature-controlled magnetic stirrer. Analysis of metal ions was carried out using an ICP-optical emission spectrometer (Inductively Coupled Plasma, Optima 2000 DV, Perkin Elmer, Waltham, MA, U.S.).

− − For experiments on the pH effect, the solutions were initially adjusted with aqueous solutions of acid or base (0.01 M HNO<sup>3</sup> and/or 0.01 M NaOH) to reach pH values between 0.5 and 5, and thus avoid the precipitation of Cr(OH)3. Isotherms were recorded during the execution of adsorption experiments with various initial metal concentrations (C<sup>0</sup> = 50–2000 mg L−<sup>1</sup> ) at 20 ◦C. Kinetic tests were performed using 100 mg of chitosan flakes, 100 and 500 mg L−<sup>1</sup> as the initial metal concentration at 20 ◦C, and pH = 3.8 for fixed time intervals during adsorption (t = 0–24 h). The effect of the granulometry of the flakes on the adsorption capacity was investigated using the two granulometric fractions of chitosan (<0.42 mm and >0.42 mm) under the same conditions. For all adsorption batch tests, a contact time of 120 min was set.

The equilibrium amount of metal in the solid phase, expressed as Q<sup>e</sup> (mg g−<sup>1</sup> ), was determined with reference to the mass balance equation: Q<sup>e</sup> <sup>=</sup> (C<sup>0</sup> <sup>−</sup> <sup>C</sup>e)×(V/m), where C<sup>0</sup> and C<sup>e</sup> (mg L−<sup>1</sup> ) are the initial and equilibrium metal concentrations, respectively, V (L) is the volume of the aqueous solutions, and m (g) is the mass of the adsorbent.

− − − Dynamic tests were performed using a glass column with an internal diameter of 0.9 cm and a bed high of 40 cm, filled with 2 g of chitosan. Tests were performed using both a real wastewater solution and a 500 mg L−<sup>1</sup> chromium(III) solution at pH = 3.5, imposing a flow rate of 23.6 mL h−<sup>1</sup> (1 BV h−<sup>1</sup> ) by using a peristaltic pump.

The real wastewater, a tannery washing solution, had the following composition: pH = 3.2, Chemical Oxygen Demand, COD = 9.1 g L−<sup>1</sup> , Total Suspended Solids, TSS < 1 g L−<sup>1</sup> , Cr3<sup>+</sup> = 635 mg L−<sup>1</sup> , Na<sup>+</sup> = 1050 mg L−<sup>1</sup> , Mg2<sup>+</sup> = 760 mg L−<sup>1</sup> , Ca2<sup>+</sup> = 300 mg L−<sup>1</sup> , Zn2<sup>+</sup> = 115 mg L−<sup>1</sup> , Cd2<sup>+</sup> = 87 mg L−<sup>1</sup> , SO<sup>4</sup> <sup>2</sup><sup>−</sup> = 1.820 mg L−<sup>1</sup> , and Cl<sup>−</sup> = 818 mg L−<sup>1</sup> .

− − − − − Desorption experiments were performed in batch mode (T = 20 ◦C, t = 24 h). In particular, after the end of the adsorption phase, the adsorbent material was separated from the supernatant using filtration membranes (0.22 µm). Then, desorption tests were performed using 50 mL solutions of H2SO<sup>4</sup> and EDTA as desorption reagents at a concentration of 0.1 M and 0.05 M, respectively.

− − − − − −

The quantitative evaluation of desorption was carried out using desorption percentages calculated from the difference between the amount of metal loaded on the adsorbent after adsorption and the amount of metal in solution after desorption. To investigate the reuse capacity of the adsorbents, the above procedure was repeated 5 times under the same conditions (first adsorption and then desorption).

#### **3. Results**

#### *3.1. Adsorption Dynamics*

The uptake of metal ions from the solution involves several steps, necessary for the transfer of the solute from the liquid phase to the specific sites within the chitosan particles (e.g., external diffusion and intraparticle diffusion).

In the case of chitosan, its chains have a large number of the –NH<sup>2</sup> and –OH groups distributed throughout the structure, making the kinetic or mass transfer representation likely to be global. The –NH<sup>2</sup> groups are the most important binding sites for metal ions [24], yet the hydroxyl groups can also contribute as coordinator groups, especially those in C-3 position [15,22]. To examine the adsorption mechanism of the metal ion of interest, two kinetic models were tested:

i. the pseudo-first-order equation described by Lagergren [25], which can be rearranged to obtain a linear form as shown by Equation (1):

$$\mathcal{L}\log(q\_{\varepsilon} - q\_{t}) = \mathcal{L}q(q\_{\varepsilon}) - (k\_{1}/2.303)t \tag{1}$$

ii. a pseudo-second-order equation based on the equilibrium adsorption capacity, which can be expressed as in Equation (2):

$$\mathbf{t}/\boldsymbol{q}\_{\mathrm{f}} = \left(\mathbf{1}/k\_{2}\boldsymbol{q}\_{\mathrm{e}}^{2}\right) + \left(\mathbf{1}/q\_{\mathrm{e}}\right)\mathbf{t} \tag{2}$$

In the above equations, *q<sup>e</sup>* (mg g−<sup>1</sup> ) represents the quantity of Cr(III) adsorbed when the system is at equilibrium, *q<sup>t</sup>* (mg g−<sup>1</sup> ) is the quantity of Cr(III) adsorbed at time *t*, and *k*<sup>1</sup> (min−<sup>1</sup> ) and *k*<sup>2</sup> (g mg−<sup>1</sup> min−<sup>1</sup> ) are the rate constants of the pseudo-first and pseudo-second order kinetic models, respectively.

Given that Equations (1) and (2) are not able to provide information on the adsorption mechanism, the simplified intraparticle diffusion model [26] was also tested, being *k<sup>i</sup>* (g mg−<sup>1</sup> min−<sup>1</sup> ) the rate constant of the model:

$$q\_t = k\_i \, t^{1/2} \tag{3}$$

The validity of these models was assessed by analyzing the slopes and intercepts of *Log*(*q<sup>e</sup>* − *qt*) vs. *t*, *t*/*q<sup>t</sup>* vs. *t*, and *q<sup>t</sup>* vs. *t* 1/2 for each of the linearized equations.

The results obtained with different concentrations of chromium(III) are shown in Table 1 in terms of correlation coefficients (R<sup>2</sup> ) as well as calculated and experimental adsorption capacity values.

**Table 1.** Values of the adsorption kinetic constants at T = 20 ◦C, pH = 3.8, C<sup>0</sup> = 0.5 g L−<sup>1</sup> .


The correlation coefficient R<sup>2</sup> for the pseudo-second-order adsorption model was the highest and, in fact, its estimate of the equilibrium adsorption capacity *qe, cal* was quite close to the experimental *q<sup>t</sup>* values (23–28.5 mg g−<sup>1</sup> as shown in Figure 2 for the two grain-sizes). These results suggest that a second-order mechanism is predominant and that the overall Cr(III) adsorption rate is controlled by a chemisorption process.

**Figure 2.** Effect of the grain size of the flakes on the adsorption of Cr(III) on chitosan (100 mg), at pH = 3.8, T = 20 ◦C, and C<sup>0</sup> = 100 and 500 mg L−<sup>1</sup> of Cr(III).

−

#### *3.2. Grain Size E*ff*ect*

Figure 2 shows the effect of the grain size of the flakes on the adsorption capacity at pH = 3.8 and T = 20 ◦C. It can be observed that the metal uptake was higher on particles with a small size (<0.42 mm). This is likely due to the higher surface area exposed by these particles, which favors the removal of Cr(III) from the solution in the initial stages of the adsorption process. This phenomenon, previously reported for the adsorption on chitin [21], chitosan [21,27,28], and Neem sawdust [29], was further improved by the ability of metal ions to penetrate into the internal structure of chitosan.

Figure 2 also confirms that the adsorption process was rather rapid, with t1/<sup>2</sup> < 18 min and the maximum adsorption obtained in about 120 min. The reference time for the subsequent equilibrium tests was thus set at 120 min, an adequate compromise between accuracy and speed in the execution of experimental tests.

#### *3.3. E*ff*ect of pH*

≈

Figure 3 shows the effect of pH on Cr(III) adsorption on chitosan. Notably, the pH of the solution strongly affected the adsorption of metal ions, with the latter increasing with the pH of the solution. Under acidic conditions, the amino groups (R–NH<sup>3</sup> <sup>+</sup>) and the hydroxyl groups (R–OH<sup>2</sup> <sup>+</sup>) are protonated and the molecule is a sort of polycation, with a reduced number of binding sites available for the adsorption of Cr(III); according to [30], the pKa of the amine groups is 6.3. In addition, the positive surface charge may hinder the adsorption of metal ions. On the contrary, a high pH will favor their adsorption since the nitrogen free electron doublet is responsible for cations coordination. Considering that Kps <sup>=</sup> [Cr3+] <sup>×</sup> [OH]<sup>3</sup> <sup>=</sup> 6.7 <sup>×</sup> <sup>10</sup>−31, chromium(III) hydroxide begins to precipitate at pH <sup>≈</sup> 6.5; for pH values higher than 3.8, there is a significant reduction of the Cr(III) fraction, with formation of Cr(OH)2<sup>+</sup> and Cr(OH)<sup>2</sup> <sup>+</sup> hydrolyzed complex species [20]. The result is an increase in chromium adsorption due mainly to hydrolyzed forms. Therefore, the adsorption of metal ions is mainly due to the electrostatic interactions between counter ions.

− As previously reported [21], the final pH values of the equilibrated solutions were higher as the Cr(III) concentration became smaller (Table 2). This is probably due to the fact that Cr ions are Lewis acids; therefore, the lower the concentration, the higher the pH (Figure 3). Moreover, the chromium adsorption capacity increases by increasing the metal ions concentration (see Section 3.4), which causes a greater competition with H<sup>+</sup> protons.

**Figure 3.** Effect of pH (initial value) on chromium(III) adsorption on chitosan (C<sup>0</sup> = 500 mg L−<sup>1</sup> ).

−



As already stated, it is accepted that chitosan amino groups are the main reactive sites for metal ions and that hydroxyl groups (in particular in C-3 position) may contribute to sorption. Metal sorption may involve different mechanisms (chelating, electrostatic attraction) depending on the pH, the solution, and the metal (concentration, speciation, etc.). Protonation of the amino groups at acidic pH increases the adsorption of anionic species, while cations interactions increase with the pH due to deprotonation of amino/hydroxyl groups. Moreover, increasing the ions competition, due to increase in metal concentration, the pH decreases. The fraction of free (accessible) amine groups is the key parameter.

#### *3.4. Adsorption Isotherms*

−

To determine the maximum adsorption capacity of Cr(III) on chitosan, a study was carried out on the adsorption isotherm by comparing the most common models; in particular, data were analyzed using the Langmuir and Freundlich equations:

$$q\_{\varepsilon} = \bigotimes^{\circ} \mathcal{k}\_{\mathsf{L}} \mathsf{C}\_{\mathsf{eq}} / \left(1 + \mathcal{k}\_{\mathsf{L}} \mathsf{C}\_{\mathsf{eq}}\right) \quad \text{or, linearized} \quad 1/q\_{\varepsilon} = 1/\left(\bigotimes^{\circ} \mathcal{k}\_{\mathsf{L}}\right) \left(1/\mathcal{C}\_{\mathsf{eq}}\right) + 1/\mathcal{Q} \tag{4}$$

− −

$$q\_{\varepsilon} = k\_{\text{F}} \mathbb{C}\_{\text{eq}}^{1/n} \quad \text{or, linearized} \quad \text{Log } (q\_{\varepsilon}) = \text{Log } (k\_{\text{F}}) + 1/n \, \text{Log } (\mathbb{C}\_{\text{eq}}) \tag{5}$$

 = °/(1 + ) 1/ = 1/(°) (1/) + 1/° = ி ଵ/ () = (ி) + 1/ () where *q<sup>e</sup>* (mg g−<sup>1</sup> ) is the amount of Cr(III) on the solid phase at equilibrium, and *Ceq* (mg L−<sup>1</sup> ) is the equilibrium concentration of Cr(III) in the aqueous phase. According to Langmuir's equation, *Q*◦ (mg g−<sup>1</sup> ) is the amount of Cr(III) required for a complete coverage of available adsorption sites, while *k<sup>L</sup>* is an empirical coefficient related to the affinity of adsorption sites for the adsorbed species. With reference to Freundlich's equation, *k<sup>F</sup>* and *n* are empirical constants representing the adsorption capacity and adsorption intensity, respectively; all parameters can be estimated through the intercepts and slopes of the linearized forms of isotherm equations.

= 1/ሾ1 + (

)ሿ

The essential characteristics of the Langmuir equation can be expressed in terms of a dimensionless separation factor *RL*, which has been defined in [27] as:

$$R\_L = 1/\left[1 + \left(k\_L \mathbb{C}\_o\right)\right] \tag{6}$$

−

where *C<sup>o</sup>* is the highest initial Cr(III) ion concentration (mg L−<sup>1</sup> ). *R<sup>L</sup>* is related to the shape of the isotherm: the adsorption is unfavorable if *R<sup>L</sup>* > 1, favorable if 0 < *R<sup>L</sup>* < 1, irreversible if *R<sup>L</sup>* = 0, and linear if *R<sup>L</sup>* = 1. All the estimated isotherm parameters are reported in Table 3. The *R<sup>L</sup>* value confirms the affinity between chitosan and chromium ions, and the adsorption equilibrium data correlate well with the Langmuir isotherm equation, with a maximum adsorption capacity estimated at 138.04 mg g−<sup>1</sup> . This implies a monolayer interaction of chromium on the adsorbent [31]. Ngah and colleagues [23] found that *Q*◦ was 30.03 mg g−<sup>1</sup> using Cr(III) in the range 4–14 mg L−<sup>1</sup> and identified the Langmuir isotherm as the best model for the adsorption on cross-linked chitosan. However, the chromium concentrations considered were much lower than those investigated in the present work, and the number of amino groups available for ions coordination in cross-linked chitosan is limited, which obviously results in a lower adsorption capacity. Eiden and colleagues [32] found that *Q*◦ was 62 mg g−<sup>1</sup> for chitosan flakes at pH = 4, but no information regarding chitosan characteristics was provided (especially regarding its degree of deacetylation). − − −

**Table 3.** Langmuir's and Freundlich's isotherm parameters at 20 ◦C, C<sup>0</sup> = 0.5 g L−<sup>1</sup> , and pH = 3.8.


Indeed, the basis of the high adsorption capacity found in our study lies precisely in the fact that the investigated chromium concentrations are much higher than those reported in other works (see Figure 2, which shows that adsorption on chitosan increases with the concentration of the target species). Although the different experimental conditions make it difficult to compare the results obtained, it can be observed that at lower concentrations, the adsorption is in line with the values found in the literature (Figures 2 and 4).

− **Figure 4.** Adsorption isotherm at 20 ◦C, C<sup>0</sup> = 50–2000 mg L−<sup>1</sup> , and pH = 3.8.

#### *3.5. Desorption*

− After Cr(III) adsorption, the chitosan flakes were washed thoroughly with deionized water and treated with the desorption agents. Desorption tests were performed using H2SO<sup>4</sup> and EDTA as

−

desorption reagents; the chromium desorption efficiency of chitosan flakes for different washing solutions is reported in Table 4.


**Table 4.** Desorption efficiency (%) for each adsorption cycle.

As reported in [23], EDTA is an efficient desorption agent: being an hexadentate chelating agent, it is capable to form a strong complex with Cr(III) ions. Considering the pH effect on the chromium adsorption, sulfuric acid (such as other acidic media) was also considered as a reagent potentially able to remove Cr(III) ions from the chitosan polymer.

Despite being efficient in terms of chitosan regeneration, EDTA persists in municipal wastewater treatments, making its use in technical applications potentially unwelcome. Although future research may investigate other chelating agents that are biodegradable and more suitable for wastewater treatment plants, the current approach has considered the possibility of treating the spent solution to recover EDTA by precipitation in acidic media as a precautionary measure.

#### *3.6. Dynamic Tests*

−

The effect of competing ions can be observed from the breakthrough curves shown in Figure 5. The inhibition effect of anions such as chlorides and sulfates has been reported for Cr(VI) [33]; in addition, it has been reported that chitosan forms complexes with transition metal ions, but not with complexes with alkali and alkali earth metal ions due to the absence of *d* and *f* unsaturated orbitals [34,35]. Therefore, the presence of Na, Mg, and Ca does not reduce the chromium adsorption. In contrast, the removal of chromium is influenced by the presence of zinc and cadmium due to their affinity with chitosan [36]. When real wastewater was considered, chitosan was able to absorb 104.7 mg g−<sup>1</sup> of chromium, while 127.5 mg g−<sup>1</sup> were adsorbed using a chromium solution. − −

− − **Figure 5.** Breakthrough tests performed with real wastewater solution (pH = 3.2) and 500 mg L−<sup>1</sup> chromium(III) solution (at pH = 3.5), using a flow rate of 1 BV h−<sup>1</sup> (BV = 23.6 mL).

The breakthrough curve should be symmetrical and sigmoid in shape; however, a deviation from a symmetrical S-shape and not very steep curves can be observed in Figure 5. The packed column was probably not very homogeneous (chitosan flakes seem to be inappropriate to be used in a packed column) and the flow-rate was too fast, causing physical non-equilibrium processes and high mass transfer zone [37].

#### **4. Conclusions**

The present study indicated the suitability of chitosan in applications aimed at removing chromium(III) ions from aqueous solutions. Chitosan is a low-cost reagent and its utilization is novel, non-toxic, and environmentally compatible. Although its effectiveness has been proven in the adsorption of several metal ions, such as Cu, Ag, Cd, Zn, Mo, V, Pb, and Cr(VI), Cr(III) has been poorly investigated. Moreover, among the usable forms of chitosan, flakes are the least studied. In this work, the adsorption of Cr(III) on chitosan flakes was investigated considering concentrations (up to 2000 mg L−<sup>1</sup> ) that are much higher than those reported in the few studies present in literature. This made it possible to highlight a much higher adsorption capacity of chitosan (up to 138.45 mg g−<sup>1</sup> ), well beyond the values observed so far.

The pH significantly influences the adsorption capacity of the biopolymer, the latter increasing with the pH of the solution. The adsorption process is rather rapid: it was found that 2 h of contact time are sufficient to reach about 95% of the adsorption equilibrium; the Langmuir equation provided the best fit over the entire concentration range, thus suggesting a monolayer interaction of chromium on the adsorbent.

The adsorption capacity for Cr(III) in dynamic tests was found to be penalized by the competition between ions, and particularly influenced by the presence of zinc and cadmium due to their strong affinity with chitosan. The recovery of Cr(III) ions (and the consequent regeneration of chitosan) can be obtained by desorption with a 0.1 M EDTA solution, with the latter being recoverable by precipitation in an acidic medium.

Since chitosan is basically a low-cost and environmentally friendly material, its use as it is produced (i.e., flakes) is desirable, especially considering the greater theoretical adsorbent capacity due to a major availability of both amino and hydroxyl groups, but involves some critical issues. In particular, since the shape and size characteristics of commercial chitosan flakes introduce hydrodynamic limitations such as column clogging, a batch reactor is probably more appropriate as a experimental setup. Further investigations are therefore underway to quantify the effectiveness of alternative operating solutions.

**Author Contributions:** Conceptualization, L.P.; methodology, L.P. and A.P.; validation, L.P. and I.F.; investigation, I.S. and M.S.; data curation, M.V.; writing—original draft preparation, L.P. and M.V.; writing—review and editing, L.P. and M.V.; project administration, L.P.; funding acquisition, L.P. All authors have read and agree to the published version of the manuscript.

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

**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* **Remediation of Copper Contaminated Soils Using Water Containing Hydrogen Nanobubbles**

### **Dongchan Kim <sup>1</sup> and Junggeun Han 2,\***


Received: 5 February 2020; Accepted: 18 March 2020; Published: 23 March 2020

**Featured Application: Nanobubbles were manufactured by using pressurized hydrogen gas and were applied to soils contaminated with heavy metals, to confirm their function as enhancers. The manufactured nanobubbles remained for an extended period of time. A batch-test experiment revealed that the nanobubbles improved desorption of heavy metals. In addition, when the nanobubbles were used as enhancers in electrokinetic experiments, the nanobubbles had better remediation e**ff**ects than distilled water (DW). The remediation of heavy metals is expected to have a significant impact when using the nanobubbles.**

**Abstract:** This basic research study was undertaken to use ecofriendly nanobubbles that can improve the electrokinetic remediation of copper-contaminated soil, as well as to determine that remediation efficiency. The nanobubbles were generated by using pressurized hydrogen gas, and the quantity of hydrogen gas bubble that remained over 14 days was measured. The generated nanobubbles were used as an enhancer to remove a heavy metal on contaminated soil, and their applicability was confirmed. A batch test was used to compare the remediation effects of nanobubbles and distilled water on copper-contaminated soil. The results proved that the nanobubbles are a proper desorption agent for copper-contaminated sand and clay specimens. The solid–liquid ratio and the contact time for desorption of the sand and clay were then respectively determined. A large amount of effluent was obtained from electrokinetic remediation of the sand sample after applying the nanobubbles as an enhancer. The remediation efficiency demonstrated with sand proved to be higher than that for clay. This greater efficiency was attributed to a wider specific surface area, demonstrating the potential use of the nanobubbles as an enhancer for soil contaminated by copper with a large amount of effluent outflow. It was also assumed to be affected by the moving capability of the nanobubbles in the soil layer. Thus, the nanobubbled water can be used to improve the removal of heavy metals from contaminated soils. An ecofriendly enhancer for electrokinetic remediation with a relatively large void ratio and fast flowrate was confirmed by the nanobubbles.

**Keywords:** nanobubbles; contaminated soil; electrokinetic; in situ; remediation

#### **1. Introduction**

The use of chemical substances is increasing, and the resulting outflows from factories, gas stations, and industrial complexes are also increasing. Soil contamination, an obstacle to development, needs to be addressed, and contaminated land needs to be reclaimed. While the buffering capacity of land is comparatively large compared to the volume of contaminants, much depends on the soil characteristics and environmental factors, and so the buffering capacity of the land varies [1]. In particular, soil

contamination sourced from chemical facility effluents has characteristics that can be difficult to effectively treat as compared to water or air pollution.

Soil remediation methods can be categorized into in situ or ex situ, depending on where treatment takes place. Offsite processing, the ex situ method of restoration, is more effective and faster, but it is also more expensive. By contrast, onsite processing, the in situ method, has a lower cost and uses technologies pertinent to the characteristics of the site [2]. While some in situ remediation methods employ ecofriendly enhancers, chemical solvents are also employed, due to the inherent difficulty in desorbing heavy metals from fine-grained soil. Toxic heavy metals exist in water media, and effective removal represents a continuing challenge for the scientific community as a whole.

To remedy this, numerous studies have been published discussing the use of adsorbent materials for aqueous media decontamination from heavy metal ions [3–6]. However, the improper extraction of enhancers employed for remediation may result in enhancers remaining in the soil at the conclusion of treatment, causing secondary contamination [7].

The diameter of the nanobubbles ranges from 1 to 999 nm [8]. Bubbles with a diameter greater than 50 µm are classified as microbubbles, while bubbles with a diameter size between 10 and 50 µm are classified as microbubbles, and bubbles with a diameter under 200 nm are nanobubbles [9]. The microbubbles have a fast rate of rise due to their large buoyant force. It does not last long in the water and rapidly rises to the surface, and it bursts when it reaches the surface. On the other hand, the microbubbles become smaller over time as they stay in water due to their slow rate of increase. If the diameter of the microbubble is 10 µm, the internal pressure of the microbubble is about 0.3 atm. As the diameter of the microbubble decreases to 1 µm and 100 nm, the internal pressure of the microbubbles rises up to 3 and 30 atm, respectively.

During this process, the internal gas spreads according to Henry's Law, so all gases in the microbubble spread into the surrounding water until the microbubble disappears [10]. On the other hand, the nanobubbles exist in water for a long time once they are created. This is known to be due to the strong hydrogen bonds found in ice and gas that hydrate the surface of the nanobubbles. These bonds prevent the spread of internal gases and maintain an adequate mechanical balance against high internal pressures, as observed via Attenuated Total Reflectance Infrared (ATR-IR) [11]. These various characteristics of the nanobubbles have been observed by many researchers in different areas, including water purification, drag reduction, purification of surface contamination, removal of tumors using shock waves caused by the collapse of bubbles, growth promotion of animals and plants, and contrast medium.

The characteristics of the nanobubbles are that they have a wide specific area, they generate radicals on their boundary surfaces, and they permit high internal pressure in the liquids. Thus, studies looking into how to exploit these beneficial characteristics of the nanobubbles are being actively pursued in various fields, including surface cleaning, remediation of poor water quality, and as a supersonic contrast agent [9]. Kyzas [12] investigated the effect of the nanobubbles (NBs) on dissolved heavy metal adsorption with activated carbon and confirmed that the inherent property of the NBs to accept charged particles onto their interface assists in the diffusion and penetration of lead ions into the activated carbon pores. Choi [13] sought to determine the removal efficiency and degradation rate for total petroleum hydrocarbons (TPH). Thus, the objective of this study was to perform batch tests to evaluate the efficacy of applying nanobubbles as enhancers. The solid–liquid ratio and contact time required for desorption of the sand and clay were then calculated. Furthermore, electrokinetic in situ remediation using nanobubbles as an enhancer for remediation of heavy metals was performed for both copper-contaminated sand soils and copper-contaminated clay soils, in an attempt to determine the optimal remediation method.

#### **2. Experimental Methods and Condition**

#### *2.1. Soils*

To simulate soils contaminated with heavy metals, standard Jumunjin sand and clay specimens were used. These clay specimens are comprised of a mixture of seaside sand and cohesive shore sediment soils. Their physical characteristics are summarized in Table 1. Copper was selected as a contaminant, and the soil specimens were contaminated, using copper nitrate (Cu(NO3)2). The soil properties were measured, using the Korea Standard, and the pH measurements followed, using KSI ISO 10390.


**Table 1.** Material properties of soils.

※ SW: well-graded sand, CL: low plastic clay, NP: non plastic.

#### *2.2. Nanobubbles*

The hydrogen nanobubbles were created by using the equipment shown in Figure 1. Nanobubbles were created with hydrogen gas that was pressurized and filtered through a ceramic filter, so as to create tiny bubbles. These tiny bubbles were transferred to the surface of the water and crushed, thereby creating other nanobubble nuclei [14]. The nanobubbles were created for a period of 24 h. Upon completion, their zeta-potential was measured so as to appraise their stability. The parameters for the zeta-potential measurements and the measured value are presented in Table 2. The prepared nanobubbles were measured by using zeta-potential analysis equipment (ZetaPALS, Brookhaven Corp., USA), to measure the zeta potential of the nanobubbles (Figure 2c). The size and quantity of the generated nanobubbles were measured, using a laser within the Nanoparticles Tracking Analysis (NATA) equipment, as shown in Figure 1. Nanobubbles with an average diameter of 171 ± 6.11 nm, a mode of 130 <sup>±</sup> 9.85 nm, and a concentration of 1.5 <sup>±</sup> 0.03 <sup>×</sup> <sup>10</sup><sup>8</sup> particles/mL were thus obtained. The pH measured after the manufacture of the nanobubbles was confirmed to be neutral 7.

**Table 2.** Parameters for ζ-potential measurement and value of zeta-potential of nanobubble.


The stability of the nanobubbles was confirmed over two months. For the experiment, stable 14-day-old nanobubbles were used [15]. Figure 3 shows how long nanobubbles are maintained. Figure 3a confirms the persistence by making nanobubbles in gasoline, and the results confirm that the nanobubbles are maintained for 120 h. Figure 3b shows a comparison of the changes in the concentration of the nanobubbles over 14 days, and it is the same as the nanobubbles used in this

experiment. After day 14, approximately 15% of the nanobubbles were lost. Thus, the long-term existence of the nanobubbles was confirmed.

**Figure 1.** Schematic view of a nanobubble Generator [16]. (**a**) Pressure gauge (air pressure in water tank), (**b**) pressure gauge (air pressure in air tank), (**c**) water tank, (**d**) air, (**e**) gas tank, (**f**) water, (**g**) pressure gauge (outflow), (**h**) bubble, (**i**) filter, and (**j**) pressure gauge (inflow).

**Figure 2.** Nanobubble (**a**) NTA Instrument (Malvern Nanosight LM10, Malvern Panalytical Ltd, Unites Kingdom), (**b**) image of nanobubbles [17], and (**c**) Zeta Potential Analyzer (ZetaPALS).

**Figure 3.** Average particle concentration of nanobubbles: (**a**) concentration of nanobubbles in gasoline [15]; (**b**) concentration of nanobubbles [16].

#### *2.3. Batch Test*

– – Batch testing was carried out in order to identify desorption characteristics of the nanobubbles as an enhancer. The experimental conditions set for the batch test are summarized in Table 3. The solid–liquid ratio, contact time, and desorption of heavy metals by the nanobubbles used as an enhancer were tested. The copper-contaminated solution was prepared, using copper nitrate powder (Cu(NO3)2) in distilled water. The contamination concentration of the copper was referred to as the Korea Standard. An agitator was used for the batch test, while a centrifuge was used to collect the supernatant solution for analysis by means of inductively coupled plasma (ICP), following the Korea Standard Test Method. Different soil–liquid ratios and contact times for contaminant desorption were determined respectively to attain the optimal conditions of desorption.


**Table 3.** Batch-test conditions.

#### *2.4. Electrokinetic Test*

A laboratory test was carried out to apply an electrokinetic remediation process to soils contaminated with copper. The same soil specimens from the batch test, sand and clay, were used to construct contaminated and uncontaminated soil. The copper-contaminated solution was prepared, using copper nitrate powder (Cu(NO3)2) in DW. Then, 500 ppm of copper was contaminated to soil, at a water content of 40%. Clay was tested immediately after contamination. Sand was used as a simulation of the ground condition, and it was contaminated with 500 ppm of copper solution and dried in the shade. Dry sand contaminated with copper was installed with the maximum consolidation and tested. The contamination concentration of copper was referred to as the Korea Standard. To identify the remediation characteristics, different conditions of soils with Jumunjin standard sand and clay were used. In addition, nanobubbles identified as an appropriate enhancer from the batch test were placed into a Mariotte bottle and injected constantly. The experimental device designed for electrokinetic remediation is shown in Figure 4. The cell and Mariotte bottle are made of acrylic, and the electrode plate used graphite. The nanobubbles were injected into the cell, using a hydraulic gradient. An electrode water tank was installed at each end of the cell, to provide unrestricted outflow of electro-osmosis during the experiments. The testers were designed to conduct five simultaneous tests. The experimental conditions are summarized in Table 4.

**Figure 4.** (**a**) Side view. (**b**) Floor plan. Diagram of the Electrokinetic system.


**Table 4.** Electrokinetic test conditions.

#### **3. Results**

#### *3.1. Batch Test*

Batch testing was carried out by varying the solid-liquid ratio and the contact time for different conditions of copper-contaminated soils, so as to allow samples to be purified, using nanobubbles as an enhancer. The results obtained from the batch test are as shown below.

Figure 5 shows the incremental desorption percentage of copper according to changes in the solid-liquid ratios of soils contaminated with copper. After agitation for 24 h, the level of desorption of the heavy metal by the nanobubbles was found to be greater in sand than in clay. The increase in solid-liquid ratios in both sand and clay also increased the efficiency with which heavy metals were removed than compared to the use of distilled water alone. However, contrary to the proportional increase of the removal efficiency of the copper to solid-liquid ratio in sand, a solid-liquid ratio greater

clay's

clay's

than 1:20 in clay rendered no significant difference in the efficiency of the copper removal by DW and NBW. Taking both efficiency and economy into account, the optimal solid–liquid ratio, while using nanobubbles as an enhancer in sand, was determined to be 1:20. This finding was attributed to the comparatively larger specific surface area and zeta-potential of the nanobubbles, which is known to be advantageous for desorption and transfer of copper from the surface of particles [15,18]. –

**Figure 5.** Solid-liquid ration test.

The results of the batch test, according to variations in the reaction time with a constant solid-liquid ratio of 1:20, are presented in Figure 6. The nanobubbles demonstrated a greater efficiency for removing copper than distilled water. The efficiency of the copper removal varied according to the reaction time. Both distilled water and the nanobubbles initially increased desorption of copper. In the case of sand with 6 h of contact time, the highest removal efficiency of copper, by both distilled water and the nanobubbles, was 15.61% and 18.48% respectively. Thereafter, the removal efficiency of both distilled water and the nanobubbles tended to converge. In the case of clay, the highest removal efficiency (desorption of copper) of both distilled water and the nanobubbles was observed within the initial first hour. The appropriate interval for contact time by the nanobubbles was found to be 6 h for sand and 1 h for clay. The explanation for this variation could be that the characteristics of re-adsorption of copper from the surface of clay are greatly influenced by the clay's surface charge.

**Figure 6.** Contact time test (S:L = 1:20).

#### *3.2. Electrokinetic Test*

An electrokinetic remediation test was carried out, with the nanobubbles acting as an enhancer for copper-contaminated soil. The current density and effluent changes, according to the soil type, were then determined. The copper and pH in soil after completion of the experiment were also determined.

Figure 7 illustrates changes in the current density according to different types of soil. Clay (DW) corresponds to a peak current density of 9.21 (mA/cm<sup>2</sup> ) at 5 h from the start of the experiment. Clay (NBW) corresponds to a peak current density of 9.81 (mA/cm<sup>2</sup> ) at 8 h from the start of the experiment. Sand (DW) corresponds to a peak current density of 0.88 (mA/cm<sup>2</sup> ) at 24 h from the start of the experiment, and Sand (NBW) corresponds to a peak current density of 0.74 (mA/cm<sup>2</sup> ) at 5 h from the start of the experiment. Both sand and clay exhibited an initial increase, followed by a decrease in the current density. The result of all the tests tended to converge as time went by. The migration of heavy metal ions increased initially and then decreased. Han et al. [19] postulate that this might be attributable to the migration and sedimentation of the heavy metal ions.

**Figure 7.** Current density.

Figure 8 illustrates the amount of effluent according to the remediation time for sand and clay as part of the remediation experiment, where nanobubbles were used as an enhancer. Effluents of Clay (DW), Clay (NBW), Sand (DW), and Sand (NBW) were 729, 662, 5855, and 1198 mL, respectively. The amount of effluent from the one treated with distilled water was found to be greater than the outflow observed with the nanobubbles treatment. This was attributed to bubbles of hydrogen gas being generated at the cathode by electrokinetic remediation with nanobubbles that disturbed the discharge of the effluent. The amount of effluent from sand appeared greater than that outflowing from clay. The removal of heavy metal from sand with greater effluent discharge was found to be highly efficient. This could be attributed to the comparatively larger specific surface area and zeta-potential of the nanobubbles that resulted in remediation being more efficient.

Figure 9 illustrates the concentrations of residual heavy metals and pH level measured at each position in the cell upon completion of the experiment. In the case of sand and clay, all exhibited an increased pH due to the collision of acid and base at the point of 0.75 x/L by electrokinetic remediation.

Remediation of copper contamination occurred, as compared to the initial state of the specimen. The migration and accumulation of heavy metal by electrokinetic remediation were identified at a point of 0.75 (x/L) [20]. The efficiency of the removal of heavy metals from the specimens contaminated by heavy metal increased by the contact of the nanobubbles and by the high ratio of the void in sand, which increased the flow of effluent through the copper-contaminated soil under basic conditions [21,22]. The nanobubbles were made of hydrogen gas in distilled water and have an initial neutral pH 7. In the case of sand, the initial pH was influenced by the initial pH due to the inflow of nanobubbles in water, and it can be seen that the pH change at the point of 0.1 to 0.7 (x/L) is small. Sand has a smaller void than clay, and heavy metals adhere to the surface, so the sand is considered to be desorbed when passing through the void of the nanobubbles.

**Figure 9.** Final Copper Profile and pH at the end of Electrokinetic Test.

#### **4. Conclusions**

The present study carried out tests to determine the effect of applying nanobubbles to copper-contaminated sand and clay specimens for heavy metal remediation. An enhancer can be employed to solve the problems associated with sedimentation of heavy metals due to the basicity of soils in electrokinetic remediation. The enhancer is mainly comprised of surfactants and solutions with acidity or basicity. However, any enhancer that remains in the soil can produce secondary contamination. Thus, the objective of the present study was to use ecofriendly hydrogen gas, which would not lead to

secondary contamination, as an enhancer for in situ remediation. The results of the experiment carried out in the present study are as follows.

Batch testing was performed to determine the optimal solid–liquid ratio and contact time to remediate copper-contaminated sand and clay, using nanobubbles. The optimal solid-liquid ratio was found to be 1:20 for both sand and clay, while the optimal contact time was 6 h for sand and 1 h for clay. The shorter contact time for clay was attributed to the re-adsorption of desorbed copper according to the varied surface charge of clay. The nanobubbles exhibited a higher copper removal efficiency than distilled water, regardless of the soil conditions.

The electrokinetic remediation tests for sand and clay specimens were carried out, using nanobubbles. The current density was found to increase initially, after which it converged at a certain point. The amount of effluent from sand was found to exceed that of clay due to its higher permeability. The remediation efficiency for heavy metals in sand was also found to be higher than that for clay, due to more frequent contact with the nanobubbles. Additionally, the pH in soils at the cathode was found to be low, suggesting less influence by the sedimentation of the heavy metals.

Based on the results of the present study, the use of nanobubbles is expected to be applicable for electrokinetic remediation of the heavy-metal-contaminated sand and clay. By considering the engineering characteristics of soils and contaminants corresponding to diverse soil conditions, nanobubbles are a promising approach applicable for the in situ remediation of land contaminated by heavy metals. Based on this study, a remediation experiment using nanobubbles is needed for application to field soil.

**Author Contributions:** All authors contributed to this study; J.H., conceptualization, funding acquisition, project administration, supervision, writing of review, and editing; D.K., investigation, formal analysis, and writing of the original draft. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Korea Agency for Infrastructure Technology Advancement under the Ministry of Land, Infrastructure and Transport of the Korean government. (Project Number: 19SCIP-B108153-05) and by a grant from the National Research Foundation (NRF) of Korea, funded by the Korea government (MSIP) (NRF-2019R1A2C2088962).This research was funded by the Human Resources Development (No. 20204030200090) of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korea government Ministry of Trade, Industry and Energy by X-mind Corps program of National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (2017H1D8A1030599). And The APC was funded by a grant from the National Research Foundation (NRF) of Korea, funded by the Korea government (MSIP) (NRF-2019R1A2C2088962).

**Acknowledgments:** This research was supported by the Korea Agency for Infrastructure Technology Advancement under the Ministry of Land, Infrastructure and Transport of the Korean government. (Project Number: 19SCIP-B108153-05) and by a grant from the National Research Foundation (NRF) of Korea, funded by the Korea government (MSIP) (NRF-2019R1A2C2088962).

This research was supported by the Human Resources Development (No. 20204030200090) of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korea government Ministry of Trade, Industry and Energy by X-mind Corps program of National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (2017H1D8A1030599).

We are grateful to Sohee Jeong and Heewon Chae for help in publishing our paper.

**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*
