Optimization of Ga content gradient in Cu(In,Ga)Se<sub>2</sub> solar cells through machine learning and device simulation

Liu Wu; Zhu Cheng-Wan; Li Hao-Tian; Zhao Su-Ling; Qiao Bo; Xu Zheng; Song Dan-Dan

doi:10.7498/aps.70.20211234

Abstract

Cu(In,Ga)Se₂ (CIGS) solar cell is a kind of highly efficient thin film solar cell, for which Ga ratio (Ga/(Ga+In), GGI) gradient engineering is an efficient approach to achieving high open circuit voltage under no short circuit current loss. In this work, we firstly evaluate the room and the strategies for improving the device performance of the CIGS solar cells based on the comparison among their theoretical efficiency limits. Then we investigate the different schemes of “V” type GGI gradient and their effects on device performance through machine learning and device simulation. Based on these studies, we optimize the scheme of “V” type GGI gradient and obtain a high efficiency of 26% from device simulation. The carrier kinetics for the effect of modifying GGI gradient on device performance are analyzed. This work provides an approach to optimizing the GGI gradient to obtain highly efficient CIGS solar cells, which is referable for experimental optimization.

Keywords:

Authors and contacts

1.
Key Laboratory of Luminescence and Optical Information, Ministry of Education, Beijing Jiaotong University, Beijing 100044, China
2.
Institute of Optoelectronics Technology, Beijing Jiaotong University, Beijing 100044, China

Corresponding author: Song Dan-Dan, ddsong@bjtu.edu.cn

Funds: Project supported by the National Key R&D Program of China (Grant No. 2018YFB1500200).

FullText HTML : translate this paragraph

References

[1]	Nakamura M, Yamaguchi K, Kimoto Y, Yasaki Y, Kato T, Sugimoto H 2019 IEEE J. Photovoltaics 9 1863 Google Scholar
[2]	Belghachi A, Limam N 2017 Chin. J. Phys. 55 1127 Google Scholar
[3]	Peace B, Claypoole J, Sun N, Dwyer D, Eisaman M D, Haldar P, Efstathiadis H 2016 J. Alloys Compd. 657 873 Google Scholar
[4]	Lim D, Kim M Y, Song W 2016 Sci. Adv. Mater. 8 558 Google Scholar
[5]	Jackson P, Hariskos D, Wuerz R, et al. 2015 Phys. Status Solidi RRL 9 28 Google Scholar
[6]	Aissani H, Helmaoui A, Moughli H 2017 Int. J. Appl. Eng. Res. 12 227
[7]	Saadat M, Moradi M, Zahedifar M 2016 Superlattices. Microstruct. 92 303 Google Scholar
[8]	Majeed N, Saladina M, Krompiec M, Greedy S, Deibel C, MacKenzie R C I 2020 Adv. Funct. Mater. 30 1907259 Google Scholar
[9]	Buratti Y, Le Gia Q T, Dick J, Zhu Y, Hameiri Z 2020 NPJ Comput. Mater. 6 1 Google Scholar
[10]	Zhu C W, Liu W, Li Y Y, et al. 2021 Sol. Energy 228 45
[11]	Kato T, Wu J L, Hirai Y, Sugimoto H, Bermudez V 2018 IEEE J. Photovoltaics 9 325
[12]	Jackson P, Wuerz R, Hariskos D, Lotter E, Witte W, Powalla M 2016 Phys. Status Solidi RRL 10 583 Google Scholar
[13]	Tai K F, Kamada R, Yagioka T, Kato T, Sugimoto H 2017 Jpn. J. Appl. Phys. 56 08MC03 Google Scholar
[14]	Green M, Dunlop E, Hohl-Ebinger J, Yoshita M, Kopidakis N, Hao X 2021 Prog. Photovoltaics Res. Appl. 29 3 Google Scholar
[15]	Yoshikawa K, Kawasaki H, Yoshida W, Irie T, Konishi K, Nakano K, Uto T, Adachi D, Kanematsu M, Uzu H, Yamamoto K 2017 Nat. Energy 2 1
[16]	Haase F, Hollemann C, Schäfer S, Merkle A, Rienäcker M, Krügener J, Brendel R, Peibst R 2018 Sol. Energy Mater. Sol. Cells 186 184 Google Scholar
[17]	Yoo J J, Seo G, Chua M R, Park T G, Lu Y, Rotermund F, Kim Y K, Moon C S, Jeon N J, Correa-Baena J P, Bulović V, Shin S S, Bawendi M G, Seo J 2021 Nature 590 587 Google Scholar
[18]	Jeong M, Choi I W, Go E M, Cho Y, Kim M, Lee B, Jeong S, Jo Y, Choi H W, Lee J, Bae J H, Kwak S K, Kim D S, Yang C 2020 Science 369 1615 Google Scholar
[19]	Green M A, Dunlop E D, Levi D H, Hohl-Ebinger J, Yoshita M, Ho-Baillie A W Y 2019 Prog. Photovoltaics Res. Appl. 27 565 Google Scholar
[20]	Green M A, Hishikawa Y, Dunlop E D, Levi D H, Hohl-Ebinger J, Yoshita M, Ho-Baillie A W Y 2019 Prog. Photovoltaics Res. Appl. 27 3 Google Scholar
[21]	Polman A, Knight M, Garnett E C, Ehrler B, Sinke W C 2016 Science 352 aad4424 Google Scholar
[22]	Nayak P K, Cahen D 2014 Adv. Mater. 26 1622 Google Scholar
[23]	Jackson P, Hariskos D, Lotter E, Paetel S, Wuerz R, Menner R, Wischmann W, Powalla M 2011 Prog. Photovoltaics Res. Appl. 19 894 Google Scholar
[24]	Liu Y, Li B, Lin S, Liu W, Adam J, Madsen M, Rubahn H G, Sun Y 2018 J. Phys. Chem. Solids 120 190 Google Scholar
[25]	Saadat M, Moradi M, Zahedifar M 2016 J. Mater. Sci. -Mater. Electron. 27 1130 Google Scholar
[26]	Ji S, Hayakawa T, Suyama N, Nakada K, Yamada A 2020 Jpn. J. Appl. Phys. 59 041003 Google Scholar
[27]	刘芳芳, 孙云, 何青 2014 物理学报 63 047201 Google Scholar Liu F F, Sun Y, He Q 2014 Acta Phys. Sin. 63 047201 Google Scholar

Cited By

图 1 CIGS吸收层带隙梯度变化

Figure 1. Band gap gradient change of CIGS absorber layer.

DownLoad: Full-Size Img PowerPoint

图 2 实验与模拟得到的CIGS太阳能电池J-V特性

Figure 2. J-V characteristic of experimental data and simulation result for the CIGS solar cell.

DownLoad: Full-Size Img PowerPoint

图 3 (a) CIGS电池器件结构; (b) CIGS层中GGI梯度示意图

Figure 3. (a) Device structure of the CIGS solar cells; (b) schematic illustration of GGI gradient in CIGS.

DownLoad: Full-Size Img PowerPoint

图 4 RF算法预测的电池效率随GGI_F/GGI_M/GGI_B变化的曲线

Figure 4. Changes of predicted efficiency by RF algorithm with GGI_F/GGI_M/GGI_B.

DownLoad: Full-Size Img PowerPoint

图 5 RF算法预测的不同GGI值情况下的器件效率

Figure 5. Device efficiency predicted by RF algorithm under different GGI values.

DownLoad: Full-Size Img PowerPoint

图 6 RF算法预测的不同GGI梯度的器件性能. 其中, 0代表实验效率为23.35%的电池所采用的GGI梯度, 模型1—模型6为预测效率最高的6组器件的GGI梯度

Figure 6. Predicted device performance of the CIGS solar cells based on different GGI gradients by RF algorithm. 0 represents for the GGI gradient used in the device with 23.35% experimental efficiency, while 1–6 are the GGI gradients of the devices with the highest predicted efficiency.

DownLoad: Full-Size Img PowerPoint

图 7 器件模拟采用的14种“V”型双梯度模型　(a) GGI_F = 0.4; (b) GGI_F = 0.5; (c) GGI_F = 0.6

Figure 7. 14 types of “V”-shaped double GGI gradient schemes used in the device simulation: (a) GGI_F = 0.4; (b) GGI_F = 0.5; (c) GGI_F = 0.6.

DownLoad: Full-Size Img PowerPoint

图 8 器件模拟得到的不同GGI梯度模型的器件性能

Figure 8. Device performance of the CIGS solar cells based on different GGI gradient schemes obtained by device simulation.

DownLoad: Full-Size Img PowerPoint

图 9 器件模拟得到的CIGS电池不同位置载流子复合速率　(a) 前表面; (b) 电池内部; (c) 后表面. R为复合速率, 单位cm^–3·s^–1

Figure 9. Carrier recombination rate at different positions of CIGS solar cells obtained by device simulation: (a) Front surface; (b) interior; (c) rear surface. Here, R is the recombination rate, which unit is cm^–3·s^–1.

DownLoad: Full-Size Img PowerPoint

图 10 器件模拟得到的CIGS薄膜背电场强度

Figure 10. Back electric field intensity of CIGS film obtained by device simulation.

DownLoad: Full-Size Img PowerPoint

表 1 CIGS, c-Si, 钙钛矿三种电池的光伏特性参数、S-Q极限值及差值分析. E_g, PCE/V_OC/J_SC/FF, J_SQ, W_OC分别代表材料的带隙、电池的转换效率/开路电压/短路电流密度/填充因子、SQ极限计算的短路电流密度、V_OC损失

Table 1. Photovoltaic parameters, S-Q limit value and difference analysis of CIGS, c-Si, and perovskite solar cells. Here, E_g, PCE/V_OC/J_SC/FF, J_SQ, W_OC respectively represent the band gap of the absorption material, the photoelectric conversion efficiency/open circuit voltage/short-circuit current density/fill factor of the solar cells, the short-circuit current density calculated by the SQ limit, and the V_OC loss.

Cell type	E_th/eV	PCE/%	V_OC/mV	J_SC/(mA·cm^–2)	FF/%	J_SQ/(mA·cm^–2)	W_OC/mV	J_SC/J_SQ/%	机构	参考文献
CIGS	1.088	23.35	734.0	39.58	80.4	44.0	354.0	89.95	SF	[1]
CIGS	1.130	22.92	746.0	38.50	79.7	43.0	384.0	89.53	SF	[11]
CIGS	1.143	22.60	741.0	37.80	80.6	42.3	402.0	89.36	ZSW	[12]
CIGS	1.110	22.30	722.0	39.40	78.2	43.7	388.0	90.16	SF	[13]
c-Si	1.120	26.70	738.0	42.65	84.9	43.3	382.0	98.50	Kaneka	[14]
c-Si	1.120	26.60	740.3	42.50	84.7	43.3	379.7	98.15	Kaneka	[14]
c-Si	1.120	26.30	744.0	42.30	83.8	43.3	376.0	97.69	Kaneka	[15]
c-Si	1.120	26.10	726.6	42.62	84.3	43.3	393.4	98.43	ISFH	[16]
钙钛矿	1.550	25.20	1180.5	25.14	84.8	27.1	369.5	92.77	KRICT	[17]
钙钛矿	1.480	24.64	1181.4	26.18	79.6	29.5	298.6	83.53	UNIST	[18]
钙钛矿	1.574	24.20	1194.8	24.16	84.0	26.2	379.2	92.21	KRICT	[19]
钙钛矿	1.536	23.70	1169.7	25.40	79.8	27.6	366.3	92.03	ISCAS	[20]

DownLoad: CSV

表 2 器件模拟中使用的参数设置

Table 2. Parameter settings used in device simulation.

	B:ZnO	Zn(Mg, O)	Zn(O, S, OH)_x	CIGS
Thickness/nm	500	50	8	1000
Permittivity/1	9	10	9	13.6
Eg/eV	3.3	3.6	3.6	1.06—1.66
Affinity/eV	4.4	4.2	4.2	3.89—4.49
N_c/(10¹⁸ cm^–3)	2.2	2.2	2.2	2.2
N_v/(10¹⁹ cm^–3)	1.8	1.8	1.8	1.8
µ_n/(cm²·V^–1·s^–1)	100	100	100	100
µ_p/(cm²·V^–1·s^–1)	25	25	25	25
N_d/(10¹⁷ cm^–3)	10	10	1	0
N_a/(10¹⁶ cm^–3)	0	0	0	2

DownLoad: CSV

表 3 不同位置GGI与光伏参数之间的相关性

Table 3. Correlation between GGI value at different location and photovoltaic parameters.

	GGI_F	GGI_M	GGI_B
V_OC	0.38	0.35	0.34
J_SC	–0.26	–0.41	–0.12

DownLoad: CSV

表 4 wxAMPS器件模拟所用的GGI梯度模型

Table 4. Schemes of GGI gradient used for wxAMPS device simulation.

模型	1	2	3	4	5	6	7
GGI_F	0.40	0.40	0.40	0.40	0.40	0.40	0.50
GGI_M	0.10	0.15	0.20	0.25	0.30	0.35	0.10
GGI_B	0.84	0.77	0.71	0.65	0.58	0.52	0.80
模型	8	9	10	11	12	13	14
GGI_F	0.50	0.50	0.50	0.60	0.60	0.60	0.60
GGI_M	0.20	0.30	0.40	0.10	0.20	0.30	0.40
GGI_B	0.67	0.54	0.42	0.76	0.63	0.50	0.44

DownLoad: CSV

[1]	Nakamura M, Yamaguchi K, Kimoto Y, Yasaki Y, Kato T, Sugimoto H 2019 IEEE J. Photovoltaics 9 1863 Google Scholar
[2]	Belghachi A, Limam N 2017 Chin. J. Phys. 55 1127 Google Scholar
[3]	Peace B, Claypoole J, Sun N, Dwyer D, Eisaman M D, Haldar P, Efstathiadis H 2016 J. Alloys Compd. 657 873 Google Scholar
[4]	Lim D, Kim M Y, Song W 2016 Sci. Adv. Mater. 8 558 Google Scholar
[5]	Jackson P, Hariskos D, Wuerz R, et al. 2015 Phys. Status Solidi RRL 9 28 Google Scholar
[6]	Aissani H, Helmaoui A, Moughli H 2017 Int. J. Appl. Eng. Res. 12 227
[7]	Saadat M, Moradi M, Zahedifar M 2016 Superlattices. Microstruct. 92 303 Google Scholar
[8]	Majeed N, Saladina M, Krompiec M, Greedy S, Deibel C, MacKenzie R C I 2020 Adv. Funct. Mater. 30 1907259 Google Scholar
[9]	Buratti Y, Le Gia Q T, Dick J, Zhu Y, Hameiri Z 2020 NPJ Comput. Mater. 6 1 Google Scholar
[10]	Zhu C W, Liu W, Li Y Y, et al. 2021 Sol. Energy 228 45
[11]	Kato T, Wu J L, Hirai Y, Sugimoto H, Bermudez V 2018 IEEE J. Photovoltaics 9 325
[12]	Jackson P, Wuerz R, Hariskos D, Lotter E, Witte W, Powalla M 2016 Phys. Status Solidi RRL 10 583 Google Scholar
[13]	Tai K F, Kamada R, Yagioka T, Kato T, Sugimoto H 2017 Jpn. J. Appl. Phys. 56 08MC03 Google Scholar
[14]	Green M, Dunlop E, Hohl-Ebinger J, Yoshita M, Kopidakis N, Hao X 2021 Prog. Photovoltaics Res. Appl. 29 3 Google Scholar
[15]	Yoshikawa K, Kawasaki H, Yoshida W, Irie T, Konishi K, Nakano K, Uto T, Adachi D, Kanematsu M, Uzu H, Yamamoto K 2017 Nat. Energy 2 1
[16]	Haase F, Hollemann C, Schäfer S, Merkle A, Rienäcker M, Krügener J, Brendel R, Peibst R 2018 Sol. Energy Mater. Sol. Cells 186 184 Google Scholar
[17]	Yoo J J, Seo G, Chua M R, Park T G, Lu Y, Rotermund F, Kim Y K, Moon C S, Jeon N J, Correa-Baena J P, Bulović V, Shin S S, Bawendi M G, Seo J 2021 Nature 590 587 Google Scholar
[18]	Jeong M, Choi I W, Go E M, Cho Y, Kim M, Lee B, Jeong S, Jo Y, Choi H W, Lee J, Bae J H, Kwak S K, Kim D S, Yang C 2020 Science 369 1615 Google Scholar
[19]	Green M A, Dunlop E D, Levi D H, Hohl-Ebinger J, Yoshita M, Ho-Baillie A W Y 2019 Prog. Photovoltaics Res. Appl. 27 565 Google Scholar
[20]	Green M A, Hishikawa Y, Dunlop E D, Levi D H, Hohl-Ebinger J, Yoshita M, Ho-Baillie A W Y 2019 Prog. Photovoltaics Res. Appl. 27 3 Google Scholar
[21]	Polman A, Knight M, Garnett E C, Ehrler B, Sinke W C 2016 Science 352 aad4424 Google Scholar
[22]	Nayak P K, Cahen D 2014 Adv. Mater. 26 1622 Google Scholar
[23]	Jackson P, Hariskos D, Lotter E, Paetel S, Wuerz R, Menner R, Wischmann W, Powalla M 2011 Prog. Photovoltaics Res. Appl. 19 894 Google Scholar
[24]	Liu Y, Li B, Lin S, Liu W, Adam J, Madsen M, Rubahn H G, Sun Y 2018 J. Phys. Chem. Solids 120 190 Google Scholar
[25]	Saadat M, Moradi M, Zahedifar M 2016 J. Mater. Sci. -Mater. Electron. 27 1130 Google Scholar
[26]	Ji S, Hayakawa T, Suyama N, Nakada K, Yamada A 2020 Jpn. J. Appl. Phys. 59 041003 Google Scholar
[27]	刘芳芳, 孙云, 何青 2014 物理学报 63 047201 Google Scholar Liu F F, Sun Y, He Q 2014 Acta Phys. Sin. 63 047201 Google Scholar

[1]	Zhang Tong, Li Ri. Machine learning-based study of dynamic shrinkage behavior during solidification of castings. Acta Physica Sinica, 2025, 74(2): . doi: 10.7498/aps.74.20241581
[2]	Guo Yan, Lv Heng, Ding Chun-Ling, Yuan Chen-Zhi, Jin Rui-Bo. Fractional-order vortex beam diffraction process recognition using machine learning. Acta Physica Sinica, 2025, 74(1): 1-8. doi: 10.7498/aps.74.20241458
[3]	Song Rui, Liu Xue-Mei, Wang Hai-Bin, Lü Hao, Song Xiao-Yan. Hardness prediction of WC-Co cemented carbide based on machine learning model. Acta Physica Sinica, 2024, 73(12): 126201. doi: 10.7498/aps.73.20240284
[4]	Zhang Qiao, Tan Wei, Ning Yong-Qi, Nie Guo-Zheng, Cai Meng-Qiu, Wang Jun-Nian, Zhu Hui-Ping, Zhao Yu-Qing. Prediction of magnetic Janus materials based on machine learning and first-principles calculations. Acta Physica Sinica, 2024, 73(23): 230201. doi: 10.7498/aps.73.20241278
[5]	Zhang Xu, Ding Jin-Min, Hou Chen-Yang, Zhao Yi-Ming, Liu Hong-Wei, Liang Sheng. Machine learning based laser homogenization method. Acta Physica Sinica, 2024, 73(16): 164205. doi: 10.7498/aps.73.20240747
[6]	Zhang Jia-Hui. Machine learning for in silico protein research. Acta Physica Sinica, 2024, 73(6): 069301. doi: 10.7498/aps.73.20231618
[7]	Feng Nai-Xing, Wang Huan, Zhu Zi-Xian, Dong Chun-Zhi, Li Hong-Yang, Zhang Yu-Xian, Yang Li-Xia, Huang Zhi-Xiang. Intelligent algorithm of extreme gradient boosting based perfectly matched monolayer and its efficient absorption on airborne transient electromagnetics problems. Acta Physica Sinica, 2024, 73(6): 065201. doi: 10.7498/aps.73.20231724
[8]	Ouyang Xin-Jian, Zhang Yan-Xing, Wang Zhi-Long, Zhang Feng, Chen Wei-Jia, Zhuang Yuan, Jie Xiao, Liu Lai-Jun, Wang Da-Wei. Modeling ferroelectric phase transitions with graph convolutional neural networks. Acta Physica Sinica, 2024, 73(8): 086301. doi: 10.7498/aps.73.20240156
[9]	Deng Xiang-Wen, Wu Li-Yuan, Zhao Rui, Wang Jia-Ou, Zhao Li-Na. Application and prospect of machine learning in photoelectron spectroscopy. Acta Physica Sinica, 2024, 73(21): 210701. doi: 10.7498/aps.73.20240957
[10]	Guo Wei-Chen, Ai Bao-Quan, He Liang. Reveal flocking phase transition of self-propelled active particles by machine learning regression uncertainty. Acta Physica Sinica, 2023, 72(20): 200701. doi: 10.7498/aps.72.20230896
[11]	Liu Ye, Niu He-Ran, Li Bing-Bing, Ma Xin-Hua, Cui Shu-Wang. Application of machine learning in cosmic ray particle identification. Acta Physica Sinica, 2023, 72(14): 140202. doi: 10.7498/aps.72.20230334
[12]	Luo Qi-Rui, Shen Yi-Fan, Luo Meng-Bo. Computer simulation and machine learning of polymer collapse and critical adsorption phase transitions. Acta Physica Sinica, 2023, 72(24): 240502. doi: 10.7498/aps.72.20231058
[13]	Guan Xing-Yue, Huang Heng-Yan, Peng Hua-Qi, Liu Yan-Hang, Li Wen-Fei, Wang Wei. Machine learning in molecular simulations of biomolecules. Acta Physica Sinica, 2023, 72(24): 248708. doi: 10.7498/aps.72.20231624
[14]	Zhang Jia-Wei, Yao Hong-Bo, Zhang Yuan-Zheng, Jiang Wei-Bo, Wu Yong-Hui, Zhang Ya-Ju, Ao Tian-Yong, Zheng Hai-Wu. Self-powered sensing based on triboelectric nanogenerator through machine learning and its application. Acta Physica Sinica, 2022, 71(7): 078702. doi: 10.7498/aps.71.20211632
[15]	Ai Fei, Liu Zhi-Bing, Zhang Yuan-Tao. Numerical study of discharge characteristics of atmospheric dielectric barrier discharges by integrating machine learning. Acta Physica Sinica, 2022, 71(24): 245201. doi: 10.7498/aps.71.20221555
[16]	Lin Jian, Ye Meng, Zhu Jia-Wei, Li Xiao-Peng. Machine learning assisted quantum adiabatic algorithm design. Acta Physica Sinica, 2021, 70(14): 140306. doi: 10.7498/aps.70.20210831
[17]	Chen Jiang-Zhi, Yang Chen-Wen, Ren Jie. Machine learning based on wave and diffusion physical systems. Acta Physica Sinica, 2021, 70(14): 144204. doi: 10.7498/aps.70.20210879
[18]	Wang Wei, Jie Quan-Lin. Identifying phase transition point of J₁-J₂ antiferromagnetic Heisenberg spin chain by machine learning. Acta Physica Sinica, 2021, 70(23): 230701. doi: 10.7498/aps.70.20210711
[19]	Yang Zi-Xin, Gao Zhang-Ran, Sun Xiao-Fan, Cai Hong-Ling, Zhang Feng-Ming, Wu Xiao-Shan. High critical transition temperature of lead-based perovskite ferroelectric crystals: A machine learning study. Acta Physica Sinica, 2019, 68(21): 210502. doi: 10.7498/aps.68.20190942
[20]	Liu Fang-Fang, Sun Yun, He Qing. Influences of Ga gradient distribution on Cu(In, Ga)Se2 film. Acta Physica Sinica, 2014, 63(4): 047201. doi: 10.7498/aps.63.047201

Catalog

Vol. 70, Issue 23 - 2021-12-05

Metrics

Abstract views: 5151
PDF Downloads: 104
Cited By: 0

姓名
邮箱
手机号码
标题
留言内容
验证码

Search

Article

留言板

Optimization of Ga content gradient in Cu(In,Ga)Se₂ solar cells through machine learning and device simulation

Abstract

Authors and contacts

Corresponding author: Song Dan-Dan, ddsong@bjtu.edu.cn

FullText HTML : translate this paragraph

References

Cited By

Catalog

Metrics

Authors

Referees

About Journal

About

Search

Article

留言板

Optimization of Ga content gradient in Cu(In,Ga)Se2 solar cells through machine learning and device simulation

Abstract

Authors and contacts

Corresponding author: Song Dan-Dan, ddsong@bjtu.edu.cn

FullText HTML : translate this paragraph

References

Cited By

Catalog

Metrics

Publishing process

Authors

Referees

About Journal

About

Optimization of Ga content gradient in Cu(In,Ga)Se₂ solar cells through machine learning and device simulation