Cost Analysis of Perovskite Tandem Photovoltaics.

Cost Analysis of Perovskite Tandem Photovoltaics


 
Joule_Cost Analysis of Perovskite Tandem Photovoltaics
Joule_Cost Analysis of Perovskite Tandem Photovoltaics


Highlights

• Low LCOE is achieved due to extreme low cost of perovskites in tandem PVs

• Further improving tandem module lifetime and efficiency reduces LCOE

• LCOE decrease rates are used to measure the research efforts

Context & Scale

Solar energy is the most abundant resource to power this planet. Photovoltaics (PV) harvest solar energy in a clean manner, wherein the relevant technologies are mostly based on crystalline silicon. Featuring skyrocketing efficiency and extreme low cost, hybrid halide perovskite solar cells have emerged as the most promising next-generation PV technology. Moreover, they can be coupled with a complimentary absorber to form tandem solar cells, which may face fewer obstacles for market penetration by capitalizing on the established PV industry. It is thus important to establish the technoeconomic competitiveness of final electricity production. We estimated the levelized cost of electricity (LCOE) using a sensitivity analysis by varying the materials, module efficiency, and lifetime. We found that perovskite tandem PVs are potentially competitive, and further efforts are required to simultaneously improve the efficiency and lifetime of perovskite PVs to stand over the entire energy sector.

Summary

Recently, metal halide perovskite photovoltaics (PV) are marching toward commercialization. The possibility for perovskite absorbers to be incorporated into multi-junction solar cells is also being discussed, which suggests alternative market entry. Although intensive investigations are being made on their technical feasibility, serious analysis on the cost of perovskite-based tandem modules is lacking. The levelized cost of electricity (LCOE) of solar modules is often used to evaluate technoeconomic competitiveness. Here, we performed a detailed cost analysis on two perovskite-based tandem modules (the perovskite/c-silicon and the perovskite/perovskite tandem module) compared with standard multi-crystalline silicon and single-junction perovskite solar cells. We found that perovskite PVs (both single junction and multi-junction) are competitive in the context of LCOE if the module lifetime is comparable with that of c-silicon solar cells. This encourages further efforts to push perovskite tandem modules onto the market in the future.

Introduction

The world's energy demands to power society keep on increasing with the evolution of human civilization. Global electricity consumption reached 21,190 TWh in 2016, which was a significant portion of the world's total energy comsumption.1 Photovoltaics (PV) provide electricity in a clean and renewable manner, and the PV market has grown dramatically in the past decades. In 2017, newly installed PV capacity increased by more than 75 GW, while the cumulatively installed capacity exceeded 300 GW worldwide.2, 3 With the fast development of PV technology, the final price of electricity generated by PV has exhibited competitiveness among the whole energy sector. For example, Abu Dhabi Water and Electric Authority (ADWEA) reported price bidding for PV as low as 2.42 US cents/kWh,4 even lower than that of fossil energy.

With the rapid progress in PV technology, metal halide perovskite is emerging as the most promising candidate for the next-generation PV materials. In less than 8 years, perovskite solar cells have attained power conversion efficiency (PCE) of 22.7%,5 and thus have attracted considerable attention in both academia and industry. The community has already advanced significantly to validate the technical feasibility of single-junction perovskite solar cells. Very recently, perovskite-based multi-junction PV devices have received intensive attention as an exciting research topic,6, 7 thanks to their tunable band gap, superior optoelectronic properties, and ease of fabrication. To date, perovskites have been incorporated with c-silicon8, 9, 10, 11, 12, 13 or CIGS13, 14, 15, 16, 17, 18 to construct hybrid tandem solar cells with respectable PCEs. By carefully tuning the band gap of the perovskite absorber, the theoretical PCEs for perovskite/silicon solar cells and perovskite/perovskite solar cells are predicted to be 39% and 34%, respectively.19 In addition, all-perovskite tandem solar cells were also successfully demonstrated.20, 21, 22 Similar to that of perovskite single-junction modules, the development of tandem devices is facing the same major technical problems, such as stability, upscaling, and reproducibility.23 In view of the encouraging theoretical PCE limit and feasible pragmatic routes, perovskite-based tandem solar cells are expected to penetrate the market swiftly by capitalizing on the existing c-silicon industry.6

However, manufacturing cost, as one essential factor governing the success of PV techniques, has received limited attention. Recently, Cai et al.24 analyzed two representative perovskite solar modules and calculated the corresponding levelized cost of electricity (LCOE). They concluded that the LCOE of perovskite PV was estimated to be 3.5–4.9 US cents/kWh with 15 years lifetime, which was even lower than that of traditional fossil energy. They suggested that more attention should be given to improving device efficiency and lifetime instead of low-cost material replacement. Recently, the LCOE of a single-junction module was also analyzed by Chang et al.25 They made conservative assumptions on limited materials availability and estimated the module cost to be US $107/m2, which led to a significantly higher LCOE than in Cai's report. During the preparation of this manuscript, Song et al.26 proposed a feasible manufacturing flow for single-junction perovskite PV fabrication. They estimated the LCOE to be 5.82 US cents/kWh in Wichita, Kansas, if a system lifetime of 30 years can be achieved. The apparent discrepancy in LCOE estimation encourages further discussion and careful justification to project the manufacturing costs.

Given the roadmap of the development of perovskite PV, tandem modules serve as one important sector.6, 7 Nevertheless, they have received limited consideration from a technoeconomic perspective, which may influence future research efforts in this field. Here, we adopted a bottom-up cost model to estimate the module cost and further LCOE, and carried out a sensitivity analysis of LCOE varying the efficiency and lifetime of relevant PV modules. We carefully compared four modules: mc-silicon (the passivated emitter and rear cell [PERC]), perovskite single junction, perovskite/c-silicon (heterojunction with intrinsic thin layer [HIT]) tandem, and perovskite/perovskite tandem. The latter three perovskite-based modules are shown to have potential for commercialization within the assumptions of this study. We focused this investigation on the technoeconomic competitiveness among the four PV techniques in the given boundary conditions. Some financing and policy assumptions are reasonably simplified, which does not affect the effectiveness of the conclusions within the relative comparison. We found that perovskite PVs exhibit low materials cost, which reduces the LCOE substantially in both the single-junction devices and the tandem devices. Still, module efficiency and lifetime are the dominant parameters that affect the LCOE significantly. In addition, the idea of LCOE decrease rate is introduced for the first time as a clear measure to guide the direction of research and development for perovskite-based PVs and tandem solar cells.

【Results and Discussion

◆Modules Configurations and Fabrication

Four representative configurations of solar cells are listed in parallel for comparison, as shown in Figure 1. Module A uses the conventional PERC device based on p-type multi-crystalline silicon. PCE of ∼21.0% in line production is claimed, which is predominant in the PV market.27 This represents one of the major competitors that perovskite PVs face. The cell in module B is based on the planar structure of a perovskite solar device, which can be fabricated using low-temperature processing techniques such as screen printing and dip coating.28 The best certified cell PCE is 22.7% for ∼0.1 cm2 devices5, and 20.1% PCE is reported for 1 cm2 devices with a similar architecture.29 The planar perovskite device has already been adopted for commercialization trials. We assume in-line production will achieve cell PCE of 19% in the 2020s as a result of the fast development of perovskite solar cells.5 In module C, perovskite/c-silicon tandem cells are used with HIT sub-cells. HIT cells have achieved the best PCE of 26.6%,5 which is one of the most promising next-generation silicon-based techniques ready for mass production. Currently, successful demonstration of perovskite/c-silicon tandem cells showed a PCE of 23.6%.30 By capitalizing on accumulative efforts in silicon solar cells, we assume these HIT/perovskite tandem cells can attain a PCE of 25% in line production in this study. Module D proposes a perovskite/perovskite tandem device configuration adopting two different absorbers with wide/narrow band gaps, respectively. The feasibility of fabrication has recently been demonstrated in practice and has revealed solar cell efficiency of 25.9% by modeling.19 We investigated its technoeconomic competitiveness by assuming cell efficiency of 22% for mass production in the coming years, which might serve as a reference point for future research and design of all-perovskite tandem cells.

Schematic Diagram of Modules

Figure 1: Schematic Diagram of Modules

(A) Module A is composed of traditional silicon cells.

(B) Module B is composed of planar perovskite cells.

(C) Module C is composed of silicon/perovskite tandem cells.

(D) Module D is composed of perovskite/perovskite tandem cells.

Spin coating, which is often used in the lab, is replaced by screen printing for large-scale manufacturing in our simulation.24, 25, 26 The materials utilization ratio for most techniques in this paper is 80%, whereas that of evaporation is assumed to be 60%.25, 26 The material costs of some key components are listed in Table 1 (see Supplemental Information for details). The fabrication process flows for each module are summarized in Supplemental Information and Tables S1–S4.


▼Table 1: The Costs of Key Components in LCOE Simulation

Component Raw Material Price ($/kg) Weight (g/m2) Material Cost ($/m2)
ETM SnO2
100
0.22
0.024
Perovskite layer PbI2
700 (400–1,400)
1.40
0.980
Perovskite layer CH3NH3I
450 (240–920)
0.14
0.066
Perovskite layer CH3NH3Br
400
0.11
0.044
Electrode Cu
9.0
1.10
0.010
HTM NiO
200 (150–400)
0.667
0.143
Recombination layer C60
160,000
0.026
4.100
Recombination layer LiF
8,000
0.003
0.027
Recombination layer PEDOT:PSS
2,700
0.10
0.280

The LCOE represents the unit cost (per kilowatt hour) of electricity over the lifetime of certain generating entity, which is often used to measure the overall competitiveness of different technologies. In the context of physical chemistry, the chemical potential eventually determines the occurrence of a reaction. Similarly, the LCOE provides a cost estimation, which governs the commercialization process of a generating technology. Based on the above-mentioned assumptions, we have thus conducted LCOE analysis on the four representative modules, which is summarized in Figure 2. Here, we retain an accuracy of 0.01 cent/kWh in the final value to comply with the conventional LCOE calculations; it is not easy to achieve such accuracy considering the uncertainty in the technique and economy development. Modules B, C, and D exhibit LCOEs of 4.34, 5.22, and 4.22 US cents kWh−1, respectively. As a comparison, the LCOE for module A is 5.50 US cents kWh−1. Specifically, the LCOE of the single-junction perovskite solar cell (module B) is in line with the previous report,24 which is 21% lower than that of a traditional silicon solar cell (module A). This shows the great commercialization potential of perovskite solar cells if the final products can reach those assumptions during manufacturing. By incorporating perovskites into silicon solar cells (module C), the LCOE of the tandem device can be reduced by 5% from that of conventional silicon solar cells. Moreover, we calculated the most aggressive tandem device with two perovskite sub-cells (module D), with the largest LCOE reduction compared with that of module A. Interestingly, materials costs for modules B and D contribute much less to the LCOE compared with module A and C. This implies that perovskite-based PVs are less sensitive to the materials cost than the currently predominant multi-crystalline silicon PVs. These LCOE values are estimated based on a few favorable assumptions, such as good module efficiency and acceptable module lifetime (see Experimental Procedures). Some of these assumptions remain challenging (e.g., the lifetime), and some are expected to be achieved soon during mass production (e.g., the efficiency). Given the same operation conditions, however, it is inferred that the tandem solar cell design has a positive impact on reducing the LCOE when perovskite is incorporated. This can likely be attributed to the low module cost as discussed later.

Module cost affects the LCOE estimation significantly, which differs among different PV technologies. The estimated manufacturing cost breakdown of the different modules proposed above was carefully investigated. The module manufacturing cost is mainly composed of materials cost, depreciation and maintenance, labor, utilities and others (e.g., weighted average cost of capital [WACC], inverter price, balance of system [BOS], operation cost, etc).26 We performed the estimation based on the internal data provided by existing manufacturers, literature reports, and reasonable assumptions (see Experimental Procedures for details). As shown in Figure 2A, the manufacturing costs are estimated to be 89.59, 32.69, 121.18 and 45.23 US$ m−2 for modules A, B, C, and D, respectively. Taking the PCE into account, the module costs are 0.43, 0.17, 0.48, and 0.21 US $/W for the four modules. The material cost is a significant part of each module cost: 88.0%, 78.2%, 81.3%, and 81.2% for modules A, B, C, and D, respectively. It dominates the overall cost and thus the final LCOE of the corresponding technology. Interestingly, the cost for the perovskite absorber layer only contributes 4.1%, 1.1%, and 3.0% to the total material cost in modules B, C, and D, respectively. It clearly shows the cost advantage of perovskite absorber, which is in accordance with the previous analysis (Figure 2). In short, the perovskite PV is approaching the sweet zone to balance the module efficiency and cost, which leads to the attractive LCOE in the PV market.

The Estimated LCOE and Breakdown of Each Module

Figure 2: The Estimated LCOE and Breakdown of Each Module

(A) The comparision of LCOE. (B–E) The cost breakdown of each module. (B) Module A, composed of traditional silicon cells; (C) Module B, composed of planar perovskite solar cells; (D) Module C, composed of silicon/perovskite tandem cells; and (E) Module D, composed of perovskite/perovskite tandem cells.

A variety of materials have been reported to couple with the perovskite absorber to construct the heterojunction. In our estimation, we assume that economically favored materials are used for module fabrication. To reveal how the materials selection affects the LCOE, we conducted the sensitivity analysis for modules B and C (Figure 3B). In respect of carrier transport materials, [6,6]-phenyl-C61-butyric acid methyl ester(PCBM), 2,2′,7,7′-Tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9′-spirobifluorene(Spiro-OMeTAD), and Poly[bis(4-phenyl)(2,4,6-trimethylphenyl)aminePoly(triarylamine)](PTAA) are used as replacements for NiO or SnO2. They lead to increases in manufacturing costs of 801%, 348%, and 622% in module B, and 299%, 172%, and 252% in module C, respectively. Consequently, the LCOE increases to 286%, 166%, and 238% in module B, and 218%, 143%, and 190% in module C. When copper is replaced by gold in the electrodes, we found significant increment of module cost to 493% and 215% in modules B and C, respectively. The corresponding LCOE increased to 204% and 168% in modules B and C. This is consistent with previous estimations of manufacturing cost,25, 26 indicating usage of expensive material probably prevents any market penetration of perovskite PVs. Interestingly, when silver is used to replace copper, we observes negligible increment of LCOE in both modules B and C. Development of cheaper materials to couple with perovskite absorber is required for the commercialization of perovskite PVs. However, it remains challenging to obtain highly efficient and stable perovskite solar cells with low-cost materials. Moreover, it is critical to rationally design and construct tunneling junctions for tandem devices based on cheap building blocks, which is still less investigated.

The Estimation of the Module Cost and the Effect of Material on Module and LCOE

Figure 3: The Estimation of the Module Cost and the Effect of Material on Module and LCOE.


(A) Breakdown of the estimated cost for each module in US$ m−2.
(B and C) Variation of estimated (B) module cost and (C) LCOE with the use of alternative materials in modules B and C.


Furthermore, we attempted to correlate the LCOE with different components (e.g., materials, facility, etc.) in different modules. Among all the components that contribute to LCOE, material cost plays a significant role. It consists of all the materials involved during the manufacturing process, including the main materials, solvent, additives, etc. Figure 4A shows the linear relationship between material cost and LCOE, wherein the marker on each line represents the assumptions used in the current study. The slopes of the four curves are almost the same, which indicates that the dependence of LCOE on materials cost is similar for all four PV modules. It is interesting to note that the curves overlap for modules A and D, indicating the module D, if realized, follows a similar LCOE function for materials cost as currently predominant PVs. This means that the LCOE is the same for modules A and D if their module costs are the same. This is likely due to the similar module efficiency and the same module lifetime for the two modules, which further indicates costs other rather than the module cost (e.g., facility cost) have a relatively small impact on LCOE. Moreover, it is clearly revealed that the extremely low LCOE of modules B and D correlates with their low material cost. To reach the same LCOE as that of module B (if realized), the material cost of module A needs to be further reduced by 45%. It thus suggests reduction of materials cost as one effective strategy to reduce the final LCOE for silicon PVs and future perovskite/c-silicon tandem PVs. Comparing modules A and C in Figure 4A, we can further see the advantage of applying a tandem design to achieve lower LCOE even with higher materials cost.

On the other hand, the LCOE is found to be relatively insensitive to the facility cost. In our simulation, facility cost describes the cost of the main equipment to perform module production, such as building, equipment, etc. Figure 4B shows the dependence of LCOE of each module on the facility, wherein the marker on each line represents the assumptions used in the current study. As expected, the LCOE increases linearly with the increase in the facility cost in all four modules. We found that the dependence of LCOE on facility cost is similar for all four PV modules. Even if the facility investment for module B doubles, the LCOE increases by just 3%, which is significantly different than the impact of module cost on LCOE (Figure 4A). It clearly suggests that the initial equipment cost affects the LCOE to a minor extent, given 20 years constant throughput. It also indicates that facility investment should not be a concern if upgrading the c-silicon module fabrication line to a c-silicon/perovskite tandem module. Perovskite-based solar cells have a significant advantage of low materials cost to reduce the LCOE if the prerequisites of commercialization are satisfied.

The Sensitivity Analysis Regarding Material Cost and Facility Cost


Figure 4: The Sensitivity Analysis Regarding Material Cost and Facility Cost.
(A and B) The linear dependence of LCOE to (A) the materials cost, and (B) the facility cost in each module. The marker in each line represents the assumptions used in the current estimation



Stability has been a serious concern since the beginning of the perovskite cell development.31 However, tremendous progress has been achieved in the last 5 years in device reliability over thousands of hours with various degradation protocols.32, 33 Perovskite cells are required to work for 20 years with over 80% of their initial power output, which is in line with the typical field warranty for PV panels. In the LCOE simulation, the lifetime describes how long the PV system (project) would operate, which generally correlates to the module lifetime. During the project operation life cycle, the power degradation rate for each module would remain at 1% per year. We found the lifetime significantly affects the LCOE, wherein the lifetime obeys an inverse proportion relationship with the LCOE (Figure 5A). We found that lifetime plays a great role in minimizing the LCOE value; that is, a longer lifetime can result in a significant decrease of the LCOE. Taking the perovskite single-junction module (module B) as an example, the LCOE decreases substantially from 7.52 to 4.34 US cents/kWh if the lifetime is improved from 10 to 20 years. According to the current market requirement that commercial c-silicon modules (module A) have a guaranteed lifetime of 20 years, we find that the perovskite single-junction module (module B) requires only 14.7 years to achieve the same LCOE. It means the requirement for lifetime is not as strict as that of conventional c-silicon PVs, which might suggest a different business model for the commercialization of perovskite PVs. Nevertheless, it is still desirable to obtain long-term operational stability of perovskite-based PVs to compete against other energy sectors.

The Sensitivity Analysis Regarding Module Lifetime and Efficiency


Figure 5: The Sensitivity Analysis Regarding Module Lifetime and Efficiency
(A and B) LCOE variation as functions of (A) the module lifetime and (B) the module efficiency in each module, with the other parameters related to LCOE are fixed. The markers indicate the assumed conditions for each module in the LCOE calculation.


Similar to the module lifetime, module efficiency also plays an essential role to determine the LCOE value. In general, an increment in module efficiency leads to a significant decrease of the LCOE. Taking module B as an example, the LCOE decreases from 4.9 to 3.9 US $/kWh if the module efficiency increases from 17% to 22%. This applies to all four PV modules investigated here. It suggests research should focus on improving PCE as an effective approach to a competitive LCOE in perovskite PVs especially at the current stage, wherein their module efficiency during mass production is far from the theoretic limit. Although challenging, the predicted efficiency requirements are based on some assumptions wherein the dynamic development of the PV community is mostly overlooked. The calculations show theoretically that the perovskite PV is capable of achieving the same LCOE at a relatively lower efficiency than c-silicon PVs in either single-junction or tandem devices. This economic advantage is mostly ascribed to the low manufacturing cost originating from cheap materials and a low energy-consuming process. However, perovskite PVs are far from being commercially available. A fully developed industrial manufacturing process for perovskite solar modules and practical implementation of these modules are not yet available. The estimation predicts their potential technoeconomic competitiveness over c-silicon PVs, and improving the module lifetime and efficiency are feasible strategies to lower their LCOE.

The LCOE decreases with the increase of the module lifetime and/or efficiency. With different module lifetimes and/or efficiency however, the LCOE of different modules changes at different rates. A clear measure is required to analyze the effectiveness of different approaches to reducing LCOE. Therefore, the LCOE decrease rate is introduced to investigate the relationship between LCOE and module lifetime and efficiency. The LCOE decrease rate is defined as the first derivative of LCOE for either the lifetime or module efficiency (Figure 6). With increasing lifetime, the LCOE decrease rate gradually approaches zero (Figure 6A). It means the positive impact of reducing LCOE diminishes by prolonging the module lifetime. For discretion, we assume negligible effects on the LCOE when the LCOE decrease rate achieves −0.1 US cents/kWh/year. And we obtain effective module lifetimes of 30.6, 27.0, 29.6, and 25.8 years for modules A, B, C, and D, respectively. Comparing with silicon solar cells (module A), the current lifetimes of perovskite modules are far below the effective module lifetime as calculated, motivating efforts to extend of the lifetime of perovskite PVs with quantitative justifications.

Similarly, LCOE decrease rate over module efficiency was obtained for each module (Figure 6B). Compared with that over module lifetime (Figure 6A), the decrease rates over module efficiency of each module at their current module efficiency are relatively higher. This implies that the LCOE can be reduced considerably by improving module efficiency at the current stage. By setting the value of LCOE decrease rate at −0.1 US cents/kWh/%abs, we calculated the effective module efficiency to be 33.6%, 28.4%, 35.3%, and 30.0% for module A, B, C, and D, respectively. Thus, it is concluded that improving the module efficiency is a universal approach to the reduction of LCOE for all four PV technologies investigated here. It is found that modules with lower materials cost show lower effective module efficiency, which means the LCOE reduction for low-cost modules is less dependent on module efficiency. The selection of −0.1 cent/kWh/year or −0.1 cent/kWh/%abs is arbitrary and is shown as an example in this study. The choice of effective numbers was made to guide research efforts. These two numbers do not necessarily apply to guide mass production. It is found that effective module lifetime and effective module efficiency serve as yardsticks to measure the effectiveness of different techniques for LCOE reduction, which sheds light on judicious R&D for perovskite PV technology.

The LCOE Decrease Rate of Module Lifetime and Efficiency


Figure 6: The LCOE Decrease Rate of Module Lifetime and Efficiency
(A and B) The LCOE decrease rate changing with (A) the lifetime and (B) the module efficiency. The solid line represents the current value given by previous research and the dotted line represents the estimated value when the module efficiency and the lifetime are improved in the future. The LCOE decrease rate around −0.10 US cents kWh−1 is magnified in the insets.



Conclusions

In conclusion, we investigated four solar modules (e.g., silicon solar cells, perovskite solar cells, perovskite/silicon tandem solar cells and perovskite/perovskite tandem solar cells) with major emphasis on their LCOE and relevant contributors. The proposed tandem structures show great commercial potential if their performance (PCE and lifetime) on lab scale can be achieved after scaling up with low-cost materials with a satisfactory material utilization ratio. We adopted the bottom-up cost model to estimate the manufacturing cost and subsequently the LCOE of each module based on current progress and assumptions. We conclude (1) single-junction perovskite solar cells are promising to achieve a competitive LCOE in the PV market; (2) tandem PVs by incorporation of perovskite absorbers are capable of lowering the LCOE effectively due to the extremely low manufacturing cost and high efficiency of perovskite sub-cells; (3) a significant decrease in the LCOE can be achieved by improving the lifetime and the efficiency of solar modules; and (4) LCOE decrease rates are defined and investigated for the first time, and demonstrated to serve as a yardstick to measure the effectiveness regarding the research direction for PV development. In our analysis, the single-junction perovskite PV shows a lower LCOE than that of the c-silicon/perovskite tandem device. However, the latter may face less barriers when penetrating the market by potentially leveraging the established PV industry. Although the current analysis indicates that perovskite PVs have great potential from a technoeconomic perspective, they are still far from practical use; the assumed device performance has not been fully achieved yet. It calls for further efforts to simultaneously improve the efficiency and lifetime of perovskite modules, which will ultimately improve the competitiveness of perovskite PVs over the entire energy sector.

Acknowledgments

Q.C. and H.Z. acknowledge funding support from National Natural Science Foundation of China (51673025, 51672008), Q.C. thanks National Key Research and Development Program of China (grant no. 2016YFB0700700) and Young Talent Thousand Program. The authors sincerely appreciate the valuable discussions with Dr. Q. Wang of JA Solar, Mr. S. Wang of Macro Economic Research Institute of the State Development and Reform Commission of the People's Republic of China, and Ms. I. Ke of Enlitech.

Author Contributions

Q.C. and H.Z. conceived and supervised this work, Z.L., Y.Z., X.W., Y.L., and Q.C. contributed to the discussion and writing of the manuscript, Y.S., Z.Z., Y.L., X.W., and H.Z. contributed to data collection. Z.Q., X.W., H.Z., and Q.C. participated in analysis of the results.

★Cite Source: https://www.cell.com/joule/fulltext/S2542-4351(18)30191-0

2018-06-03