Review—Electrochemistry for Sustainable Solar Photovoltaics

We can estimate the amount of zinc required for long-term storage of solar electricity. In another paper in this Focus Issue, ¹⁰ we argue that by 2050, the scale of installed solar panels must reach about 100 TWp in order to reduce our carbon emission by about 25% from the 2020 level. With a 100 TWp solar electricity capacity by 2050, a significant portion would require storage. Let us assume that 90% of the 100 TWp capacity would be consumed in real time or stored for short to medium terms. That is, only 10% of the 100 TWp capacity requires long-term storage, which is 10 TWp. Due to the intermittency of solar energy and the power losses in photovoltaic systems, the time-averaged output of a solar panel is about 15% of its peak wattage. Therefore, the amount of solar electricity for long-term storage on a daily basis is 100 TWp × 10% × 24 h/d × 15% = 3.6 × 10¹⁰ kWh d⁻¹.

The Gibbs free energy of ZnO is –320.5 kJ mol⁻¹ or –0.089 kWh mol⁻¹ at 25 °C. ⁹ To store 3.6 × 10¹⁰ kWh of solar electricity on a daily basis, we need at least 3.6 × 10¹⁰ ÷ 0.089 = 4.0 × 10¹¹ moles of zinc a day or about 26 million tons (Mt) of zinc a day. If the turn-around time between shipping zinc rods to users and getting spent anodes back for regeneration is 180 days, the total amount of zinc required is 180 days × 26 Mt/d = 4.8 billion tons (Bt). The known reserve of zinc on our planet is 1.9 Bt, ¹¹ may or may not enough for the proposed storage loop. This highlights the importance of Earth-abundant materials as the media for solar electricity storage.

The estimation above is rough and many of the numbers used are arguable. The point of the estimation is that whatever storage medium we choose, the amount of the material required is likely in the range of billion tons.

There are about a dozen metals which have a reserve of at least 1 Bt according to Ref. 11. Most of them are listed in Table II. The storage medium should come from this table, and the question now is which metal should be our choice?

Table II. Figure of merit for Earth-abundant metal↔metal oxide loops and their standard reduction potentials. ^1,7

We can estimate the theoretical performance of different metal↔metal oxide loops for storage of solar electricity. A figure of merit (FoM) is defined as the ratio between the maximum possible amount of energy to be released when a metal oxidizes and the minimum amount of charge needed to reduce its oxide to metal:

$$\begin{eqnarray}&&FoM=\displaystyle \frac{Gibbs\,Free\,Energy}{Minimum\,Reduction\,Charge}\end{eqnarray} \tag{ 1 }$$

The unit for figure of merit is joule/coulomb (J/C). The minimum amount of charge to reduce 1 mole of a metal oxide is calculated from the valence of the metal in the oxide assuming a 100% coulombic efficiency.

Table II lists the figures of merit for a dozen Earth-abundant metals. They are listed in the order of decreasing figure of merit, except for water and carbon dioxide (CO₂). Burning fossil fuels takes the C↔CO₂ loop. The H₂↔H₂O loop has been proposed for storage of solar electricity. They are not as good as the Zn↔ZnO loop in terms of theoretical performance. This contradicts Table I in which hydrogen is slighter better than zinc in terms of energy capacity per kilowatthour of solar electricity. The discrepancy is attributed to the different energy efficiencies of the two electroreduction processes due to factors like over-potential. ⁶ A high over-potential increases the throughput but reduces the energy efficiency.

There are eight metals which have better figures of merit than Zn↔ZnO in Table II, but they are excluded due to process practicality or technical readiness.

Electroreduction of a metal oxide to metal is performed either in an aqueous solution or molten salt. A major difference between the two electrolytic processes is the temperature. Aqueous electrolysis is carried out at or near room temperature, but molten-salt electrolysis requires a high temperature around 1000 °C. The difficulty comes when we try to drive molten salt electrolysis by solar energy. That is, how do we keep the salt at 1000 °C for more than 15 h a day from late afternoon to early morning? We could use fossil fuels or grid electricity to maintain the temperature, but they undo the purpose of solar electricity storage. Even during daytime, concentrated solar power is needed to heat the salt to 1000 °C. Therefore, solar-powered molten-salt electrolysis requires a far more complicated setup than solar-powered aqueous electrolysis. In addition, concentrated solar power curtails the geographic applicability of molten-salt electrolysis as it requires direct-beam sunlight.

The last column in Table II lists the standard reduction potentials of Earth-abundant metal redox pairs. ⁹ The standard reduction potential of water is –0.8277 V vs the standard hydrogen electrode. Oxides of calcium, magnesium, aluminum, barium, and titanium (CaO, MgO, Al₂O₃, BaO, and Ti₃O₅) all have more negative reduction potentials than water, so they require molten-salt electroreduction which is unsuitable for solar electricity storage. If one tries electroreduction of these oxides in an aqueous solution, the product is not the metal but hydrogen.

The two key technologies needed to practice the proposed metal↔metal oxide loop are:

• An electroreduction process of a metal oxide in an aqueous solution, and
• A mechanically-recharged metal/air battery for the metal.

In Table II, there is no established electroreduction process for oxides of silicon and boron (SiO₂ and B₂O₃). For chromium oxide (Cr₂O₃), we have not found any report on a chromium/air battery.

Now we are down to Zn↔ZnO in Table II. Zinc is Earth abundant. The figure of merit for Zn↔ZnO is 1.66 J C⁻¹, better than H₂↔H₂O. Commercially zinc is produced electrolytically from zinc sulfate (ZnSO₄) by dissolving ZnO in sulfuric acid (H₂SO₄). ¹² Solar electroreduction of zinc oxide uses the same commercial process. Zinc/air batteries for solar electricity storage are being commercialized as they have a lower cost than other battery technologies for storage, ^4,13 although the research focus has been on electrically-recharged batteries instead of mechanically-recharged ones. ¹³ Therefore, the two technologies required to practice the Zn↔ZnO loop are commercially either available or almost ready.

The annual global production of zinc is about 13 Mt today, ¹¹ but the Zn↔ZnO loop requires 26 Mt of zinc a day. Even with an Earth-abundant metal, the challenges for long-term storage of solar electricity are still significant as the zinc production needs to be scaled up by about 740 times. The technologies yet to be developed for the Zn↔ZnO loop include:

• A low-cost high-efficiency system for solar electrolysis, and
• An optimized zinc/air battery specifically for mechanical recharging.

It is noticed that almost all the prior efforts on silicon electrorefining utilized the two-electrode electrolyzer as shown in Fig. 3, with an anode and a cathode. This two-electrode configuration is widely employed in the chemical and mining industries to electrolytically produce various raw materials, although the specifics for the electrolyzer are different depending on the material produced. The highest purity it has achieved is about 3 N's, but those electrolytically-produced materials do not require an ultrahigh purity. Electrorefining relies on the different redox potentials of various impurities from silicon. Therefore, controlling the potential applied for oxidation or reduction is critical to achieve an ultrahigh purity. In the two-electrode electrolyzer, the potential difference between the anode and cathode is controlled, but there is no way to independently control the oxidation potential on the anode or the reduction potential on the cathode. This is why ultrapure materials are difficult in the two-electrode electrolyzer. ²² The only paper we have found on three-electrode silicon electrorefining did not investigate the effect of anode or cathode potential on the purity of the produced silicon. ²³

Zoom In Zoom Out Reset image size

We are taking a new approach to silicon electrorefining in an attempt to achieve ultrapure silicon of 6 N's–9 N's. ²² It involves three-electrode electrolysis in two steps in a molten salt (Fig. 4). In the first step, metallurgical-grade silicon is the anode and the anode potential is controlled vs a reference electrode to keep all the more noble impurities than silicon in the anode (e.g., boron, phosphorous, iron, and copper). In the second step, the cathode from the first step is used as the anode, and the potential on the new cathode is controlled to keep the less noble impurities than silicon in the salt (e.g., aluminum and calcium). The cathode from the second step is expected to be solar-grade silicon with at least a 6 N purity.

Zoom In Zoom Out Reset image size

Our approach is analogous to the fractional distillation process for SiHCl₃. There are two steps in fractional distillation. In the first step, the vaporization temperature is controlled vs a reference temperature (e.g., the freezing point of water, 0 °C) to keep all the chlorides with boiling points higher than SiHCl₃ in the liquid phase. In the second step, the vapor from the first step is condensed and the condensation temperature is controlled to keep all the chlorides with boiling points lower than SiHCl₃ in the gas phase. That is, each step can remove only one type of impurities with a common characteristic.

We have extended the Nernst equation for impurity segregation in electrolysis. ²² The segregation coefficient, $k,$ is defined as the concentration ratio of an impurity in an electrode (either anode or cathode) and in the molten salt:

$$\begin{eqnarray}&&k=\displaystyle \frac{Impurity\,concentration\,in\,an\,electrode}{Impurity\,concentration\,in\,salt}\end{eqnarray} \tag{ 2 }$$

The applied reduction potential ${E}_{red}$ and the segregation coefficient follow the revised Nernst equation ²² :

$$\begin{eqnarray}&&{E}_{red}-{E}_{red}^{o}=-\displaystyle \frac{RT}{zF}\,\mathrm{ln}\,k\end{eqnarray} \tag{ 3 }$$

where ${E}_{red}^{o}$ is the standard reduction potential, $R$ the gas constant, $T$ the absolute temperature, $F$ the Faraday constant, and $z$ the number of electrons transferred in the redox reaction. The purpose of electrorefining is to maximize the segregation coefficient for the anode, ${k}_{anode},$ in the first step of Fig. 4 so impurities stay in the anode and minimize it for the cathode, ${k}_{cathode},$ in the second step so impurities remain in the salt.

Figure 5 plots Eq. 3, i.e., segregation coefficient vs under-potential ${E}_{red}-{E}_{red}^{o},$ at 850 °C for different $z$ values. Obviously the segregation coefficient in electrolysis is controlled by the applied potential. By switching from an under-potential to an over-potential, the segregation coefficient changes from smaller than unity to larger than unity. That is, we can keep an impurity either in the salt (by an under-potential) or in the anode (by an over-potential). By increasing the applied potential, we can also push the segregation coefficient far away from unity. Therefore, electrorefining is more versatile and effective than purification methods based on impurity segregation between two phases (e.g., between silicon melt and silicon solid in float-zone growth), for which the segregation coefficient of an impurity has a fixed value at the melting point of silicon.

Zoom In Zoom Out Reset image size

Equation 3 also provides the required potentials to achieve solar-grade silicon if the electrorefining process is run under equilibrium. ²² Table III shows the impurity concentrations in the salt at 850 °C when 2.064 V vs the O₂/2O^2– reference electrode is applied to the working electrode serving as the anode. All the standard reduction potentials at 850 °C in a molten salt are from Ref. 24. For silicon, it is –1.99 V vs O₂/2O^2–. With the typical impurity concentrations in metallurgical-grade silicon ²⁵ (column 2 in Table III), the concentrations of all the impurities in the salt (last column) are below the targets for solar-grade silicon (column 3).

Table III. Impurity segregation between anode and salt at 2.064 V vs O₂/2O^2– and 850 °C. ²²

Impurity	MG-Si (ppm)	SG-Si (ppm)	Redox pair	${E}_{red}^{o}$	${E}_{red}-{E}_{red}^{o}$	$k$ achieved	${C}_{salt}$ (ppm)
B	40	<1	B3+/B	–1.63 V	–0.434 V	8.00 × 105	5.00 × 10–5
P	20	<5	P2+/P	–1.27 V	–0.794 V	1.59 × 107	1.26 × 10–6
O	3000	<10	O2/2O2–	0.00 V	–2.064 V	1.23 × 1036	∼0
C	600	<10	C4+/C	–1.07 V	–0.994 V	4.84 × 1016	∼0
Fe	2000	<10	Fe3+/Fe	–0.81 V	–1.254 V	1.12 × 1016	∼0
Ti	200	<1	Ti2+/Ti	–1.81 V	–0.254 V	201	0.995
Cu	?	?	Cu2+/Cu	–0.22 V	–1.844 V	1.12 × 1016	~0

Controlling the anode potential removes only those more noble impurities than silicon. A second electrorefining step removes those less noble impurities than silicon. Table IV lists the impurity concentrations in the cathode at 850 °C when –2.083 V vs O₂/2O^2– is applied to the working electrode as the cathode. All the impurity concentrations in the cathode are within the targets for solar-grade silicon. That is, a two-step three-electrode process has the potential to achieve solar-grade silicon from metallurgical-grade silicon.

Table IV. Impurity segregation between salt and cathode at –2.083 V vs O₂/2O^2– and 850°C. ²²

Impurity	MG-Si (ppm)	SG-Si (ppm)	Redox pair	${E}_{red}^{o}$	${E}_{red}-{E}_{red}^{o}$	k achieved	Ccathode (ppm)
Al	200	<2	Al3+/Al	–2.23 V	0.24 V	1.00 × 10–2	2.00
Ca	600	<2	Ca2+/Ca	–2.56 V	0.57 V	4.73 × 10–5	2.84 × 10–2

The section above describes idealistic, simplistic, and equilibrium electrorefining, but the actual molten-salt silicon electrorefining process can be much more complicated. Kinetics will no doubt play a significant role. It is also possible that the system involves irreversible reactions and/or formation of complexes, which may prevent effective electrorefining.

There are more practical challenges. Molten-salt electrolysis requires about 1000 °C. At this temperature, silicon is oxidized in air and the electrically-insulating SiO₂ formed on the surface of silicon electrodes prevents further electrorefining. ²⁴ Therefore, silicon electrorefining must be carried out in high vacuum of about 10^–6 Torr. On the other hand, the salts used for silicon electrorefining are usually chlorides and fluorides. They have vapor pressures in the millitorr range at 1000 °C. Performing a high-temperature process in high vacuum with a highly corrosive vapor is by no means an easy task. Some of the difficulties include:

• Metallic contamination: Vacuum chambers are typically made of stainless steel which contains iron. Electrical wires are typically made of copper. Both are detrimental to solar-grade silicon even in the smallest amount. The apparatus can be designed in such a way that the iron or copper-containing components are not in direct contact with silicon, but the presence of fluorine and chlorine at high temperatures could carry iron and copper around through the gas phase causing contamination.
• Corrosion of apparatus: Not many materials can withstand fluorine and chlorine at high temperatures. For example, we have noticed that a molybdenum component rusted like a piece of iron in seawater after several months in the apparatus. We have so far identified several materials which are stable under these conditions including graphite, glassy carbon, alumina (Al₂O₃), and aluminum. There are also design tricks to minimize the detrimental impact of fluorine and chlorine. All these challenges must be overcome before electrorefined solar-grade silicon becomes a reality.

There are three scenarios for decommissioned solar panels: panel reuse, component reuse, or material reuse. ²⁸ If a decommissioned panel is still functional, it could be reused as a lower-quality product. If the panel is not reusable due to damage or low efficiency, the components in the panel could be recovered for reuse including the silicon cells and the glass pane. If the components are not reusable, the materials in the components could be extracted for reuse including silver, lead, tin, copper, and silicon. Table V provides the metallic contents of a typical silicon panel. ²⁹

Component reuse is unlikely to be practiced in the near future. This is because of the many different types of silicon cells in commercial panels today. ²⁸ Different cell structures produce different powers and efficiencies. Even for cells of the same structure, there are efficiency variations of about 2% absolute. For example, the efficiency of commercial aluminum back-surface field cells varies between 17% and 19%. Cells of different power, efficiency, voltage, or current can not be packaged into the same panel due to mismatch losses. With the dozen or so different cell structures and efficiency variations for the same cell structure, collecting cells of the same structure and matched efficiency is a significant challenge for the recyclers. They would have to have hundreds of bins, each for a particular cell structure with a particular efficiency. They would have to process a large number of waste panels before enough cells of the same structure and same efficiency could be accumulated for a new panel. This would significantly increase the cost of reused components.

Several companies are in the panel reuse business today. It generates a relatively-high revenue, about $20/panel, with few processing steps. The uncertainty for panel reuse is whether there will be a large and sustained market on the order of several hundreds of gigawatts peak a year for reused panels? ²⁸ Material reuse generates a lower revenue. Table VI shows the potential revenue if all the materials were extracted in their pure, high-value forms from silicon panels, $8.45/panel. ²⁸ The valuable materials include silver, aluminum, glass, silicon, and copper. Aluminum and glass come from the panel, and Fig. 6 shows the structure of the most-common commercial silicon panels without the junction box. Silver and silicon come from the cells (Fig. 7). Lead comes from the solder in the cells, which must be removed from the recycling sludge as it is toxic. Although material reuse generates a lower revenue, it is the ultimate solution as reused panels will eventually fail and require material extraction.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 8 compares the current recycling process with the two processes we propose, one for panel reuse and one for material reuse. ²⁸ The current process recovers only aluminum, glass, and copper for about $3/panel. Panel reuse generates the highest revenue ($20/panel) with the fewest processing steps, while material reuse produces a reasonable revenue ($8/panel) with the most processing steps. For this reason, all the decommissioned panels should begin with the same processing steps to check their reusability. If panel reuse is not possible, more processing steps are added for material extraction. Therefore, the first two steps are identical for the proposed processes in Fig. 8.

Zoom In Zoom Out Reset image size

Panel reuse is being practiced today. For material extraction, there are commercial tools to strip the junction box off for copper and then strip the aluminum frame off. After removal of junction box and aluminum frame, most recyclers today shred the remaining panels for glass, so the silicon cells with silver and lead are shredded with the glass. Our proposed process for material reuse requires a technology to cleanly separate the cells from the glass so the materials in the cells can be extracted. A Japanese company, NPC Incorporated, developed a hot-knife method in which a steel blade heated to about 300 °C slices through the ethylene vinyl acetate (EVA) layer to separate the cells from the glass. ³⁰ After hot knife, the polymers including the ethylene vinyl acetate encapsulant and the fluoropolymer backsheet must be removed from the glass and the cells by pyrolysis. ²⁸ The glass cullet after pyrolysis becomes high-transparency solar glass cullet and the cells without polymers enable material extraction.

The metals worth recovery from silicon cells include silver from the electrodes, lead and tin from the solder, copper from the interconnection, and silicon (Fig. 7). The silicon nitride (SiN_x) antireflection layer and the aluminum back electrode are difficult to recover and they have little value. ²⁸

The approaches in the literature for metal extraction from silicon cells all involve chemicals. ^28,31 The most common chemical to leach metals out of silicon cells is nitric acid (HNO₃). ³² Silver, lead, and copper dissolve in it. Tin precipitates out as white powder of tin oxide (SnO₂) in HNO₃, which requires filtration or sedimentation to recover. Although the metallic contents in silicon panels are low (Table V), the nitric leachate can process many batches of silicon cells so it contains high concentrations of the metals. Now the task is to extract the multiple metals from the leachate one by one in their pure forms. There are two methods in the literature which deal with multiple metal recovery from the leachate. Jung et al. ³³ used more than half a dozen different chemicals in combination with chemical extraction, electrowinning, and filtration to recover the three metals.

A simpler method for multiple metal recovery from the leachate is sequential electrowinning. ³² The nitric leachate contains three metals: silver, lead, and copper. As shown in Table VII, they have different reduction potentials vs the standard hydrogen electrode. We can apply an appropriate reduction potential on the cathode to extract silver first. Once silver is gone, a more negative potential is applied to recover copper. Finally an even more negative potential would recover lead.

Table VII. Standard reduction potentials for Ag, Cu, and Pb.

Redox Pair	${E}_{red}^{o}$ (V)
Ag+/Ag	0.7996
Cu2+/Cu	0.3419
Pb2+/Pb	–0.1262

The method has been demonstrated and reported in Ref. 32. A three-electrode electrolyzer was used for the experiment, with a titanium working electrode as the cathode, a platinum counter electrode as the anode, and a silver/silver chloride (Ag/AgCl) reference electrode. Figure 9 is the schematic of the electrolyzer. Figure 10 shows the voltammetric scans of the leachate before and after silver recovery. Before silver recovery, the silver reduction peak occurs around 0.35 V vs the Ag/AgCl reference electrode and copper reduction around 0.1 V. The lead reduction peak is overshadowed by the copper peak as the concentration of copper is about 10 times higher than that of lead in the leachate. For silver recovery, a potential of 0.3 V vs Ag/AgCl was applied for 5.5 h. After the silver was extracted, the silver peak disappeared in the voltammetric scan (Fig. 10). A potential of 0.1 V vs Ag/AgCl was then applied for 24 h causing copper to deposit on the cathode. It was also noticed that when copper was depositing on the cathode, the lead in the leachate was further oxidized and deposited on the anode. That is, instead of a three-step electrowinning process, the three metals in the nitric leachate can be recovered in their pure forms by two-step sequential electrowinning.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 11 shows an energy-dispersive X-ray analysis of the electrowon silver. Only two elements are observed: silver and titanium from the cathode. With a detection limit better than 1%, the electrowon silver is at least 99% pure. Figure 12 shows the energy-dispersive X-ray analyses of the deposits on the cathode and the anode during copper recovery. Since there is more copper than silver in the leachate, the thick copper deposit covers the titanium cathode and prevents it from being detected. On the anode, both the platinum anode and lead are detected.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

One of the major issues with sequential electrowinning is the costly platinum anode, which must be replaced with a low-cost anode. Another significant issue is the low silver recovery rate. With a nitric leachate, the silver recovery rate is only about 70%, copper 80%, but lead has a good recovery rate of 99%. ³² As shown in Table VI, silver is the most valuable material from silicon panels. Recovering only 70% of the silver means we would lose about $1.10/panel in potential revenue. The reason for the low silver and copper recovery rates is the thin, dendritic silver and copper deposits on the cathode. They often detach from the cathode and then redissolve in the nitric leachate. Innovations in leaching chemistry and/or electrowining method are needed to improve the silver and copper recovery rates to 100%.