Impact of Band‐gap Gradient in Semi‐Transparent and Bifacial Ultra‐Thin Cu(In,Ga)Se2 Solar Cells

Ultra‐thin Cu(In,Ga)Se2 (CIGSe) solar cells on transparent conductive oxide back contact reduce the material consumption of rare indium and gallium and simultaneously exhibit great potential for semi‐transparent bifacial application. For highly efficient CIGSe solar cells, a steep back Ga grading and Na treatment are expected. However, Na will promote the formation of highly resistive GaOx at the rear interface owing to Ga accumulation. In this work, the three‐stage co‐evaporation process is renewed and the effect of the deposition sequence in the first stage on the Ga distribution as well as the cross‐correlated influence of Na is explored. In particular, the standard deposition sequence of Ga+In is altered to start with In. When a thin In layer is pre‐deposited on the back contact, the fill factor and efficiency increase. The deposition of In+Ga+In in the first stage of CIGSe growth leads to efficiencies 28% (on average) higher than for the standard deposition sequence of Ga+In. Additionally, 2.74% efficiency is reached under rear and 9.32% under simultaneous front and rear illumination. Therefore, adapting the deposition sequence in the first stage of CIGSe growth is identified as a key to improving the device performance on transparent back contact.


Introduction
Photovoltaics has taken the lead in the transition to power generation by renewables.Compared to the most established crystalline silicon technology, absorbers with direct band gaps get along with thicknesses in the μm range and thus allow lower material costs and flexible devices.Amongst these thin-film technologies, Cu(In,Ga)Se 2 or short CIGSe is a highly stable material that has achieved efficiency records up to 23.6%. [1]The standard thickness for the CIGSe absorber (2-2.5 μm) requires high consumption of scarce indium, and one strategy for lowering DOI: 10.1002/admi.202400085 the manufacturing costs and enhancing the large-scale deployment is thinning down the absorber layer thickness to ultra-thin (500 nm). [2]Additionally, the fabrication of an ultra-thin CIGSe absorber can also shorten the deposition duration, which can reduce energy consumption and operation costs.So far, the champion efficiency of ultra-thin CIGSe solar cells (with a 490 nm absorber) is 15.2%. [3]2b] The standard back contact of CIGSe solar cells is molybdenum (Mo), since a self-forming MoSe 2 layer evolves during the absorber deposition, resulting in an electrically beneficial quasi-Ohmic contact between the CIGSe absorber and the Mo back contact. [5]7b,8] In other words, CIGSe on TCO back contact has the potential to improve the usage of unabsorbed light by reflecting it back into the absorber via a mirror on the rear side of the glass, by transmitting it to another application (e.g., bottom solar cell in a tandem) or by exploiting bifacial illumination (i.e., from both front and rear side of the solar cell).
In a CIGSe absorber, the band gap can be adjusted by changing the Ga/(Ga+In) (GGI) ratio. [9]9a,10] The stronger back gradient results in a wider band gap near the rear interface of the absorber compared to the bulk and still compared to the front surface, leading to a slightly higher valence band maximum (VBM) and a significantly higher conduction band minimum (CBM) there. [11]9b] Additionally, a slight increase in the GGI ratio at the front surface promotes the formation of a spike-like conduction band alignment at the CdS/CIGSe interface, which allows the electrons to move more easily to the CdS layer and thus reduces recombination. [12]However, an excessive Ga gradient can lead to a deterioration of the morphology and hence increase recombination. [13]Therefore, the optimization of the Ga-gradient is crucial.
2b,c,14] However, a highly resistive GaO x layer can form at the CIGSe/ITO rear interface, especially during high-temperature deposition, which leads to blocking the hole extraction from the CIGSe absorber. [5]2c] An alternative strategy to suppress the GaO x formation is to pre-deposit an interface modification layer between the CIGSe absorber and the ITO back contact as it has been shown for the solution-based growth of CISSe. [15]9b] A further approach for performance enhancement of CIGSe solar cells is the treatment with alkali metals such as Na, K, and Rb, which can improve the open circuit voltage (V oc ). [16]Different strategies for alkali supply are researched for CIGSe: before, or after the absorber deposition.When the CIGSe absorber is deposited at high temperatures, Na will promote the formation of a highly resistive GaO x layer at the CIGSe/TCO interface owing to Ga accumulation. [5]For CIGSe absorbers grown by coevaporation, the most commonly applied configuration is alkali post-deposition treatment (PDT), that is, the supply of NaF in a fourth stage and in presence of selenium. [16,17]2c,16b] The Na atoms can passivate the CIGSe grain boundaries, and simultaneously occupy the Cu vacancies of the CIGSe surface. [18]Consequently, Na Cu defects are formed, leading to a high Na concentration at the front surface.However, an excessively high Na concentration near the front surface of the absorber hinders the formation of the Cd Cu donor defect during the CdS CBD deposition, resulting in a degraded pn-junction. [18,19]Therefore, the NaF content for alkali treatment should be controlled.2c] NaF pre-deposition (Pre-DT) is an alternative process for alkali treatment. [20]The Na 2 Se x , which is mainly located at the grain boundaries, acts as a Se reservoir during CIGS grain growth, passivating defects and filling Cu vacancies, thus increasing the p-type carrier concentration and V oc . [21]n this research, we alter the standard deposition sequence during the first stage of the CIGSe co-evaporation process to achieve a slightly In-rich rear interface and a band-gap gradient supporting carrier extraction: Whereas the first stage generally starts with Ga for the purpose of back Ga gradient for-

Results
To tune the band-gap gradient, which is affecting the electron and hole transport, we designed three different deposition sequences in the first stage of the CIGSe absorber growth.For the reference CIGSe absorber, first Ga then In was deposited during the first stage (Ga+In CIGSe, top line of Figure 1).To introduce a thin Ga-poor layer directly at the back contact, we altered the deposition sequence in the first stage and initially supplied In followed by Ga (In+Ga CIGSe, second line in Figure 1).With the aim of keeping the Ga-poor area minimal, we further altered the first stage to provide shortly In, then Ga, then In again (In+Ga+In CIGSe, third line in Figure 1).NaF PDT was utilized for Na incorporation into the CIGSe absorbers fabricated by these three recipes.In addition, the NaF Pre-DT was tested for the In+Ga+In strategy, described as Na+In+Ga+In CIGSe (bottom line in Figure 1).
The GD-OES results for the CIGSe absorbers deposited on Mo back contact with various deposition strategies in the first stage are shown in Figure 2. Corresponding composition depth profiles and a comparison of the individual element profiles of these absorbers are displayed in Figures S1 and S2 (Supporting Information), respectively.Generally, the CIGSe absorbers exhibit a Cu-poor front surface, which is beneficial to form a highquality pn-junction with CdS and the ZnO layers (Figure 2a).The In+Ga CIGSe, however, shows an increase in CGI ratio at the immediate front surface, which is related to a lower-quality pnjunction because of hindered Cd 2+ diffusion during the CdS CBD process. [18,22]Furthermore, the CGI across the entire absorber is higher for In+Ga CIGSe compared to the other samples.A Ushaped GGI ratio with a minimum in the first third from the front interface and a maximum in the second half is observed for all CIGSe absorbers (Figure 2b).In more detail, it can also be regarded as a W-shape since there are additional minima at the very front and rear positions.The maximum GGI value of Ga+In CIGSe is located at the rear side of the absorber.No matter, if NaF PDT or NaF Pre-DT is applied, for (Na+)In+Ga+In CIGSe, the maximum GGI values are found close to 80% absorber depth.On the contrary, the maximum GGI value of In+Ga CIGSe shifts to 50% absorber depth.The Ga+In and In+Ga+In CIGSe with NaF PDT display-as expected-a higher Na concentration near the front surface (0%-20% depth) than the NaF Pre-DT sample (Na+In+Ga+In) (Figure 2c).This observation is in line with reports that Na atoms can occupy Cu vacancies to form Na Cu defects. [18,19]For the In+Ga CIGSe with NaF PDT, however, a high Cu concentration was found at the front surface (Figure S2, Supporting Information) resulting in a significantly lower Na concentration than for Ga+In and In+Ga+In CIGSe.Interestingly, a very pronounced rise of the Na content toward the rear interface is observed for Ga+In CIGSe (rather than for Na+In+Ga+In).2c,13,23] Band gap grading profile of these CIGSe absorbers can be calculated from the elemental distribution based on the following equation: [24] E g (x, y) = ( 1.00 + 0.13x 2 + 0.08x 2 y + 0.13xy + 0.55x + 0.54y ) eV (1)   where x and y are Ga/(Ga+In) and S/(S+Se) ratio, respectively.In our experiment, sulfur is absent and y = S/(S+Se) = 0. Therefore, the band gap of these CIGSe absorbers depends on the GGI ratios only and a close correlation is found (compare Figure 2b,d).
11a] The conduction band minimum (CBM), however, shifts toward the vacuum level by increasing the GGI ratio, indicating that the CBM of CuGaSe 2 is closer to the vacuum level than that of CuInSe 2 . [11]Therefore, the band gap of the CIGSe absorber is widened with increasing GGI ratio.In other words, the band gap profile in a CIGSe absorber is determined by the depth-dependent CBM resulting from the GGI distribution.The maximum GGI of Ga+In CIGSe is closer to the rear interface than for the other samples, implying that the Ga+In CIGSe has a steep band-gap grading and a significantly higher CBM toward the rear interface.11b] In contrast, the GGI result of In+Ga CIGSe shows a peak-like band gap (or CBM) profile, in which the maximum GGI value is observed at ≈ 50% absorber depth.The photo-generated electrons in the bottom absorber region flow to the rear interface direction, whereas the photo-generated electrons in the front region flow toward the front surface. [25]This peak-like band gap profile thus hinders the electron transport and exhibits a severe shunting behavior, [25] since the photo-generated electrons in the bottom region of the In+Ga CIGSe absorber cannot be extracted effectively.
Figure 3 presents the current density-voltage (J-V) curves of the best CIGSe solar cells on Mo back contact from various deposition strategies in the first stage of the absorbers growth.The best Ga+In CIGSe solar cell of this series exhibits a 7.25% efficiency, with a V oc of 549.5 mV, a J sc of 22.72 mA cm −2 , and a FF of 58.0%.When In+Ga is deposited in the first stage, all the PV parameters of the best device significantly decrease as compared to the best Ga+In CIGSe solar cell (Figure 3 and Table 1).The best efficiency of In+Ga is significantly reduced to 3.80%.The degraded PV performance can be linked to the band gap profile, see the explanations related to Figure 2d.The In+Ga+In samples show a reduction of V oc and J sc but an improvement of FF and efficiency compared to Ga+In CIGSe.Maximum efficiencies of 6.95% and 7.57% are reached for In+Ga+In with NaF PDT and NaF Pre-DT, respectively.2c,19a] Figure 3b,c shows the statistical distributions of the PV parameters for the CIGSe solar cells on Mo (10 CIGSe sub-cells) with absorbers from various deposition strategies (summarized also in Table 1).These PV parameters reveal the same trends as the corresponding highest-efficient CIGSe solar cells.It is indicated that the In+Ga+In and Na+In+Ga+In CIGSe can obtain similar efficiency as the Ga+In sample.The In+Ga+In samples, both with NaF Pre-DT and PDT, show the potential to achieve highly efficient CIGSe solar cells on Mo back contact with the maximum efficiency reached for Na+In+Ga+In.External quantum efficiency (EQE) spectra of these bestperformaning CIGSe devices are depicted in Figure S3a (Supporting Information).They show a similar behavior in the short wavelength range (300-390 nm).The Ga+In CIGSe exhibits a significantly higher spectral response from 540 to 1200 nm wavelength than CIGSe from the other strategies, leading to a higher J sc of the Ga+In CIGSe device.The band gap (E g ) values of these CIGSe absorbers are extracted from the plot of [E×ln(1−EQE)] 2 versus E (Figure S3b, Supporting Information). [15]The Ga+In CIGSe presents a smaller E g value (1.03 eV) than the other three types of CIGSe in agreement with the slightly higher J sc of the Ga+In CIGSe solar cell (Figure 3a).This E g extracted from EQE is linked to the minimum value of the E g profile (Figure 2d), [24b] because the semiconductor materials can only absorb light starting from the band gap energy.The Urbach energies of the CIGSe absorbers can be extracted by fitting the slope of the ln(EQE) spectrum at the long-wavelength edge. [26]It is reported that E U impacts the carrier mobility and lifetime of the absorbers as well as the open-circuit voltage deficit of thin film solar cells. [26]26a] The In+Ga CIGSe solar cell owns a larger E U value (22.71 meV) than those from the other three deposition strategies (Figure S3c, Supporting Information), which can explain the low efficiency of the In+Ga device (Figure 3a and Table 1).However, the Na+In+Ga+In CIGSe solar cell exhibits the best efficiency (Figure 3a and Table 1) and smallest E U value (16.03 meV, Figure S3c, Supporting Information), indicating that the rear interface modification is a potential strategy for achieving high-performance CIGSe solar cells.
For the semi-transparent solar cells on ITO back contact, the GD-OES results of the CIGSe absorbers with various deposition strategies in the first stage are shown in Figure 4. Corresponding composition depth profiles and a comparison of the individual element profiles of these absorbers are displayed in Figures S4 and S5 (Supporting Information), respectively.All these CIGSe absorbers exhibit a Cu-poor composition (CGI < 1, Figure 4a).11b]  The In+Ga CIGSe shows the highest CGI, also at the front surface, just as observed on the Mo substrate coated in the same process (Figure 2a).This observation implies that the In+Ga absorber has fewer Cu vacancies than Ga+In, In+Ga+In, and Na+In+Ga+In CIGSe.
The GGI ratios of the various CIGSe absorbers present a Wshaped Ga gradient. [10]In other words, starting from the front surface, the GGI ratio first shows an increase followed by a slight decrease, and then increases further toward the rear interface where it drops again (Figure 4b).For the Ga+In CIGSe absorber, the maximum GGI ratio is reached near the rear interface, resulting from the Ga accumulation. [13]Compared to Ga+In CIGSe, a slight shift toward the front surface of the maximum GGI ratio is observed for both In+Ga+In and Na+In+Ga+In CIGSe, owing to the initial deposition of In in the first stage.With increasing time of initial In deposition as conducted for In+Ga CIGSe, the maximum GGI ratio is found in the middle of the absorber.
The Na concentrations of these CIGSe absorbers are shown in Figure 4c.Compared to In+Ga and Na+In+Ga+In CIGSe, higher Na concentrations are observed for the Ga+In and the In+Ga+In samples near the front surface (0%-18% of absorber thickness).It is implied that a larger number of Cu vacancies are occupied by Na to form Na Cu defects at the surface of the Ga+In and In+Ga+In CIGSe absorbers.According to literature, the Na atoms of Na Cu defects can be replaced by Cd 2+ to format Cd Cu donor defects during the CdS CBD process, leading to a better quality of the pn-junction. [18,19]19b] The Na concentrations increase near the rear interface, especially in the Ga+In and Na+In+Ga+In absorbers.19b,21b] The maximum band gap value of Ga+In CIGSe is close to the rear interface (90% absorber depth, Figure 4d), implying a steep Ga grading toward the back contact.When the deposition sequence in the first stage is In+Ga, the maximum band gap value and hence the highest CBM are observed at 50% absorber depth.9b] The In+Ga+In and Na+In+Ga+In CIGSe absorbers display a similar band gap distribution, yet, with the maximum values at 75% absorber depth and thus closer to the rear interface.
The PV parameters of the CIGSe solar cells on ITO back contact reveal maximum efficiencies of 5.71%, 2.84%, 6.18%, and 5.15% for Ga+In, In+Ga, In+Ga+In, and Na+In+Ga+In CIGSe devices, respectively (Figure 5a and Table 2).The best efficiency is thus achieved for the In+Ga+In sample confirming that a slight addition of In at the rear interface is beneficial for achieving highquality CIGSe solar cells.Comparing the best CIGSe devices, the V oc , J sc , and efficiency decrease for In+Ga compared to the Ga+In CIGSe device.The reduction of the PV parameters can be attributed to the maximum GGI being located in the bulk of the absorber rather than near the rear interface, leading to a barrier for the electron transport (Figure 4b).Even though the J sc of the In+Ga+In CIGSe device is smaller than for Ga+In CIGSe, V oc , FF, and efficiency are higher.A lower PV performance is observed for In+Ga+In with Pre-DT compared to NaF PDT because the initial addition of Na catalyzes the GaO x interlayer formation on ITO.The GaO x leads to a larger series resistance (1.23 Ω cm 2 for Na+In+Ga+In and 0.97 Ω cm 2 for In+Ga+In, see Figure S6, Supporting Information), which hinders the separation and transport of photo-generated carriers. [5,25]Thus, for ITO substrates, PDT is superior to Pre-DT for Na incorporation into ultra-thin CIGSe absorbers.
The statistical distributions of the PV parameters for the CIGSe solar cells on ITO (10 CIGSe sub-cells) with the absorbers obtained from various deposition strategies in the first stage are visualized in Figure 5b,c (and summarized in Table 2).These PV parameters show the same trends as the highest-efficient CIGSe solar cells with the maximum efficiency for the In+Ga+In sample related to the significantly increased V oc and FF.Additionally, all the PV parameters of the In+Ga+In CIGSe solar cells present smaller error bars as compared to Ga+In CIGSe (Figure 5b,c and Table 2), indicating that this proposed novel deposition sequence in the first stage cannot only lead to improved efficiency but also to higher reproducibility.
EQE spectra of the semi-transparent CIGSe solar cells are displayed in Figure S7a (Supporting Information).These EQE spectra reveal a comparable behavior to those of CIGSe on Mo back contact.In detail, these semi-transparent CIGSe solar cells have a similar EQE spectral response from 300 to 390 nm wavelength.Furthermore, the Ga+In CIGSe exhibits a significantly higher spectral response from 540 to 1200 nm wavelength than CIGSe of the other strategies, which can explain a higher J sc of the Ga+In CIGSe device.The In+Ga+In and Na+In+Ga+In CIGSe present an overlapping EQE spectrum, leading to similar J sc values (Table 2).Compared to CIGSe solar cells on Mo back contact, these semi-transparent CIGSe exhibit an overall larger E g (Figure S7b, Supporting Information).Simultaneously, the Ga+In CIGSe presents a smaller E g value (1.09 eV) than the other three types of CIGSe in correlation with a slightly higher J sc of the Ga+In CIGSe solar cells.The semi-transparent CIGSe solar cells exhibit larger E U values than the Mo-based ones (Figure S7c, Supporting Information, compared to Figure S3c, Supporting Information).16a] Finally, also the CIGSe growth on semi-transparent back contacts may impact the absorber and the Schottky contact the carrier transport.Therefore, the relationship between Urbach energies and photovoltaic performance of semi-transparent CIGSe solar cells is still to be explored in the future.The bifacial J-V measurements are carried out three months after the standard J-V measurements.The then measured J-V curves under front and rear illumination of the formerly best CIGSe solar cells deposited on ITO back contact are shown in Figure 6a.From the front illumination measurement, we observe that all CIGSe solar cells, which were kept in ambient condition for three months, present an increase in J sc (Table 3).This may be attributed to the formation of a thin In 2 S 3 layer at the CdS/CIGSe interface with time, owing to the small standard molar enthalpy of formation of In 2 S 3 or In 2 (O,S) 3 . [37]However, the CIGSe solar cells with NaF PDT exhibit a FF reduction (Table 3), in contrast to the stable FF for CIGSe with NaF Pre-DT.Compared to the fresh measurement, the re-measured Ga+In and In+Ga+In CIGSe solar cells present reductions in V oc and efficiency (Table 3).On the contrary, an improvement of these PV parameters is observed for the In+Ga and Na+In+Ga+In CIGSe solar cells.
When comparing the highest efficient CIGSe devices obtained from different deposition strategies in the first stage of absorber growth under rear and front illumination, similar trends are  [28] Chu et al., [29] Nishimura et al., [30] Mavlonov et al., [31] Nakada et al., [32] Mavlonov et al., [33] Shin et al, [34] Moon et al., [35] Shin et al, [7b] Yang et al., [9b] Li et al., [36] and Li et al. [8a] found (Figure 6a).However, when the CIGSe devices are measured under the rear illumination, the PV parameters are significantly lower than for front illumination.7b,8] The highest rearilluminated efficiency (3.47%) is achieved by the Ga+In CIGSe solar cell.The In+Ga+In devices, both with NaF PDT and Pre-DT, present lower efficiencies (2.74% and 2.43%, respectively) due to the remarkably lower J sc (11.19 and 12.06 mA cm −2 for In+Ga+In and Na+In+Ga+In compared to 19.43 mA cm −2 for Ga+In).The rear illumination efficiency of the In+Ga CIGSe device is negligible (0.34%), resulting from severe drops in all PV parameters.Interestingly, for the In+Ga+In CIGSe device, the FF improves under rear illumination, underlining the high quality rear-contact formation.When the CIGSe solar cells are measured under simultaneous front and rear illumination, V oc , J sc , and efficiency experience gains compared to front illumination only (Figure 6b and Table 3).The J sc and efficiency under bifacial illumination are close to the sum of the individual front and rear illumination measurements.The FF values for the devices with NaF PDT reveal a slight increase under bifacial compared to front illumination only.Just the Na+In+Ga+In CIGSe device shows a drop in FF, which may be linked to the high series resistance (Figure S6, Supporting Information).The best Ga+In CIGSe solar cell achieves 10.71% efficiency under 1 sun front + 1 sun rear illumination, which is higher than for the best In+Ga+In CIGSe device (9.32%).The main reason is the lower efficiency of In+Ga+In compared to Ga+In CIGSe under rear illumination, so the higher front illumination efficiency cannot prevail.When the In+Ga+In CIGSe device is measured under rear illumination, the short wavelength light is absorbed in the absorber's bottom region.However, a barrier at 75% depth of the absorber will hinder the electron transport toward the front surface, resulting in a decrease in charge collection.Thus, despite a minimum in In content at the rear interface-as successfully achieved by the In+Ga+In deposition sequence-being beneficial, it is important to restrict it to the immediate vicinity of the back contact.
In Figure 6c, we plot the efficiencies of semi-transparent CIGSe solar cells with various absorber thicknesses (under front illumination) for state-of-the-art devices and our work.The ratios of front-illuminated efficiency/rear-illuminated efficiency (PCE Rear /PCE Front ), that is, the bifaciality factors with respect to efficiency, of these CIGSe solar cells are presented in Figure 6d.It is observed that our strategy of first-stage adaptation leaves room for further efficiency enhancement.Regarding the bifaciality, this is still highest for the conventional case of Ga+In (66%).While moving the Ga-peak away from the immediate rear interface is beneficial for the front-illuminated efficiency, the bifaciality factor lags behind with 51% for In+Ga+In and 44% for Na+In+Ga+In.Across the literature, the thin CIGSe solar cells (with absorber thickness below 1 μm) generally reveal higher ratios of PCE Rear /PCE Front than the thick CIGSe solar cells (with absorber thickness above 1 μm).These results indicate that the ultra-thin CIGSe solar cells show potential in agrivoltaics, building-integration, or bifacial operation rather than the standard-thick CIGSe solar cells (with 2-3 μm absorber).

Discussion
It is reported that a steep Ga grading toward the rear reduces recombination of the charges at the back contact.Compared to the band gap of CuInSe 2 , CIGSe shows an increasing band gap with rising Ga content. [11]9a] When the CIGSe absorbers are deposited on both Mo and ITO back contact by the same deposition strategy, they present a similar GGI ratio and band gap gradient.In addition, these CIGSe devices show a similar average visible transmittance (AVT) in the spectral range of 380-770 nm (Figure S8, Supporting Information).Therefore, the semi-transparent CIGSe absorbers (on ITO back contact) are utilized for studying the mechanism of the charge carrier separation and transport.
The schematic diagram of charge carrier transport in these graded CIGSe solar cells is shown in Figure 7.When the CIGSe absorber is fabricated with the standard deposition sequence of Ga+In in the first stage, a strong back Ga gradient with a maximum GGI and thus E g around 90% absorber depth is observed (Figure 4d) and the charge carrier separation explained in Figure 7a: The electrons are repelled from the back interface due to the higher CBM and migrate toward the front surface of the CIGSe absorber (CdS side). [9,10]Simultaneously, the holes can be extracted by the ITO back contact.9a,10,13] When In+Ga is deposited in the first stage of the coevaporation process, the maximum GGI value is located at ≈ 50% of the CIGSe absorber depth (Figure 4b).Therefore, the widest band gap (the highest CBM) is also obtained in this depth of the absorber (Figure 4d and Figure 7b), leading to the formation of a central electron transport barrier. [25]The photo-generated electrons at the bottom region cannot go through that barrier and therefore, serious recombination occurs at the back side of the In+Ga CIGSe absorber.Since the change in VBM with GGI and hence E g variation is much smaller, the holes are less influenced but tend to be trapped in the region of high GGI and thus are not extracted efficiently either.In other words, the rear interface region of the In+Ga CIGSe solar cell cannot contribute with photogenerated charge carriers to the current generation.
For In+Ga+In and Na+In+Ga+In CIGSe the maximum E g values are observed at ≈ 75% of the absorber depth (Figure 7c,d).
Resulting are increased recombination losses compared to Ga+In CIGSe, following the arguments for In+Ga CIGSe even if the effects are not as pronounced for the later CBM maximum.Both Ga+In and In+Ga+In CIGSe exhibit a narrower band gap value than Na+In+Ga+In CIGSe at the front surface, which is beneficial to form a spike-like band alignment with the CdS buffer layer. [12]Therefore, higher efficiencies are achieved for Ga+In (5.71%) and In+Ga+In (6.18%) CIGSe rather than for Na+In+Ga+In (5.15%), while In+Ga CIGSe leads to significantly lower numbers (Table 2).The highest efficiency is observed for the In+Ga+In CIGSe device owing to the In-rich rear interface and a simultaneous occurrence of the band gap maximum in the second half of the absorber.
To achieve highly efficient semi-transparent CIGSe solar cells, besides the formation of a high back Ga gradient, approaches to avoid GaO x formation and foster the growth of large grains (as reported for high In content [13] ) are essential.The here presented strategy shows that we can to a certain extent move the Ga maximum further inside the absorber without deteriorating the carrier collection.Our results reveal that this approach can be further optimized by minimizing the time of In at the beginning of the first stage while establishing the start with In as a promising strategy for further enhancement of semi-transparent CIGSe solar cells.Therefore, the first stage deposition sequence is crucial for the fabrication of semi-transparent CIGSe solar cell.9b,38]

Conclusion
We have investigated the PV performance of CIGSe solar cells deposited on ITO and for comparison on Mo back contact by various deposition strategies in the first stage of the co-evaporated absorbers.In particular, we altered the standard deposition sequence of Ga+In to start with In, that is, In+Ga and alternatively In+Ga+In.Additionally, we disclosed the different effects of NaF pre-and post-deposition treatment for the In+Ga+In configuration.An In-rich absorber at the rear interface is desired to enable large grain growth and to avoid GaO x formation on ITO back contact.At the same time, we revealed that the Inrich phase should be kept to the close vicinity of the back contact to maintain efficient charge carrier transport and extraction.When the CIGSe is deposited on Mo (with Na diffusion barrier), the highest efficiency of 7.57% is observed for CIGSe with Na+In+Ga+In sequence in the first stage.The optimum performance of semi-transparent CIGSe solar cells on ITO is achieved by the deposition of In+Ga+In in the first stage combined with NaF PDT (6.18% efficiency, i.e., 28% on average relatively higher than for the standard deposition sequence Ga+In).Furthermore, this CIGSe solar cell presents an efficiency of 2.74% under rear illumination and 9.32% under simultaneous front and rear irradiance.All in all, starting the first stage of the CIGSe co-evaporation with a short phase of In supply is highly beneficial for enhancing the solar cell performance on Mo and in particular on ITO back contact.This newly introduced modification of the three-stage process of CIGSe growth will open up the doors for efficiency enhancement of semi-transparent devices with multiple applica-tions, for example, in agrivoltaics, building-integration, or bifacial operation.

Experimental Section
Materials: Barium-borosilicate glass 7059 (alkali content below 0.3%, purchased from CORNING) is selected to serve as alkali-free glass substrate.400 nm thick ITO was fabricated in the laboratory onto the glass substrates by DC-sputtering at a rate of 2.5 Å s −1 in Ar atmosphere.800 nm thick Mo with a Si 3 N 4 barrier was purchased from PVcomB-Kompetenzzentrum Photovoltaik, Berlin.
Fabrication of CIGSe Absorbers: The sub-500 nm CIGSe absorbers were fabricated on top of Mo or ITO by co-evaporation of the elements (PVD tool Dreva 450).Following the typical three-stage process for CIGSe growth, Se was evaporated continuously during the whole duration of the process, while the fluxes of Cu, In, and Ga were controlled by shutters.Ga and In were provided in the first stage at 410°C substrate temperature to form an In-Ga-Se precursor.After that, the substrate temperature was increased to 450 °C, and Cu was evaporated during the second stage.By the end of the second stage, a Cu-rich thin film [Cu/(Ga+In) = 1.06] had grown owing to the formation of a Cu x Se phase.This Cu/(Ga+In) ratio was adjusted to 0.85-0.90by simultaneous evaporation of In and Ga during the third stage.In case of NaF PDT, the Na addition was carried out directly after CIGSe deposition without breaking the vacuum and in a continuous supply of Se.In the experiment, both NaF PDT and NaF Pre-DT were utilized for Na incorporation.For the NaF Pre-DT, NaF plus Se were deposited before the first stage of the absorber deposition.The substrate temperature was kept at 360 °C during the Pre-DT and PDT process.
Fabrication of CIGSe Photovoltaic Devices: On top of the finished absorbers, an 80 nm CdS buffer layer was coated by chemical bath deposition (CBD).Following, 80 nm intrinsic zinc oxide (i-ZnO) and 300 nm AZO were coated by radio-frequency sputtering.10 nm Ni and 2000 nm Al contact grids were deposited by thermal evaporation before the samples were mechanically scribed into eight 1 cm × 0.5 cm cells.
Characterization: X-ray fluorescence (XRF, type XEPOS from Spectro) was applied to characterize the absorber thickness (480-495 nm), and average Cu/(Ga+In) (CGI = 0.85-0.90)and Ga/(Ga+In) (GGI = 0.26-0.29)ratio.Glow discharge optical emission spectroscopy (GD-OES, instrument GDA 650HR from Spectruma) measurements were performed for depth-profiling of the CIGSe absorber s, which had been subject to etching in 10% HCl solution for 2 min to remove excessive Na.The current density-voltage (J-V) curves were measured for performance evaluation under a standard AM 1.5 G sun simulator at room temperature (type WA-COM WXS-10-Super).During the measurement, the samples were kept at a constant temperature of 25 °C via a cooling chuck.The external quantum efficiency (EQE) was measured by a home-built system applying calibrated Si and Ge diodes as references.
Statistical Analysis: As-obtained J-V results are imported to an Excel table.Some sub-cells are destroyed during the solar cell fabrication process.Their fill factor values are below 25%, and they are considered outliers.Therefore, the PV performance of these sub-cells is not included in the statistical analysis.The PV performance is presented with (average ± standard deviation) calculated from 10 sub-cells for each experimental recipe.

Figure 1 .
Figure 1.CIGSe absorber deposition by a three-stage co-evaporation process with various strategies of In and Ga supply in the first stage.

Figure 2 .
Figure 2. a) Cu/(Ga+In) ratios, b) Ga/(Ga+In) ratios, c) Na molar concentration, and d) band gap of CIGSe absorbers deposited on Mo back contact with various deposition strategies in the first stage as measured by or derived from GD-OES.The x-axis is normalized to 100% absorber thickness.

Figure 3 .
Figure 3. a) J-V curves of the best CIGSe solar cells on Mo back contact fabricated by various deposition strategies in the first stage of the absorber growth.Statistical distribution of b) open circuit voltage V oc and short circuit current density J sc , and c) fill factor FF and power conversion efficiency PCE derived from 10 devices for each strategy.

Figure 4 .
Figure 4. a) Cu/(Ga+In) ratios, b) Ga/(Ga+In) ratios, c) Na concentration, and d) band gap of CIGSe absorbers on ITO with various deposition strategies in the first stage measured by or derived from GD-OES.The x-axis is normalized to 100% absorber thickness.

Figure 5 .
Figure 5. a) J-V curves of the best CIGSe solar cells on ITO back contact by various deposition strategies in the first stage for the absorber.Statistical distribution of b) open circuit voltage V oc and short circuit current density J sc , and c) fill factor FF and power conversion efficiency PCE derived from 10 devices for each strategy.

Figure 7 .
Figure 7. Schematic diagram of charge transport in bandgraded CIGSe absorbers.x-axis is the absorber thickness and y-axis is the band gap at each depth of CIGSe absorber.

Table 1 .
Electrical parameters of CIGSe on Mo back contact with various deposition strategies in the first stage.

Table 2 .
PV performance parameters (front illumination) of CIGSe on ITO back contact with various deposition strategies in the first stage.

Table 3 .
Best PV performance parameters of CIGSe on ITO back contact under front or rear illumination, and bifacial illumination.