Bandgap Engineering of CIGS: Active Control of Composition Gradient

Abstract

The tunable bandgap of CIGSe has established bandgap engineering as a pivotal research direction for advancing the efficiency frontiers of solar cells. In particular, the proposal of the V-shaped bandgap gradient has motivated extensive efforts to achieve precise control over elemental composition and spatial distribution within the absorber layer. Against this backdrop, this review systematically classifies active control strategies—such as surface sulfurization, Ga grading, and Ag alloying—according to their doping mechanisms and the resulting bandgap profiles. It further evaluates emerging profiles, including the “hockey-stick” distribution, against the conventional V-shaped benchmark, and explores future pathways for bandgap engineering in next-generation, high-efficiency photovoltaic devices. Further improvements in photovoltaic efficiency can effectively boost power generation and lower solar power costs, providing a practical solution to future energy and environmental challenges.

1. Introduction

Against the backdrop of growing global energy demand, the energy supply chain faces increasing strain from resource depletion and environmental pollution caused by heavy reliance on fossil fuels [1]. As a viable alternative, photovoltaic power generation stands out among clean energy sources for its cost-effectiveness, ease of installation, and minimal environmental impact. Over the years, continuous breakthroughs have been achieved in the power conversion efficiency of various photovoltaic technologies, including crystalline silicon [2,3], copper indium gallium selenide (CIGSe) [4,5], and perovskite solar cells [6,7,8], etc. Among them, CIGSe—a p-type direct-bandgap semiconductor—has emerged as a highly promising absorber material, particularly for lightweight and flexible applications. Its advantages are manifold: (1) a high absorption coefficient that enables efficient light capture with only micrometer-thick layers; (2) a fully inorganic composition that ensures excellent long-term stability and device lifetime; and (3) a tunable bandgap that can be optimized via elemental ratio engineering to better match the solar spectrum [9,10,11].

The two most recent world-record efficiencies for CIGSe solar cells were achieved via entirely distinct fabrication routes: Solar Frontier employed a sequential sulfurization process after selenization of metallic precursors, whereas Uppsala University utilized a three-stage co-evaporation method with Ag alloying. These approaches also differ in material composition, resulting in distinct bandgap gradient profiles. A comparative analysis of these differences is crucial for identifying new pathways to higher efficiency. The key distinctions include: (1) Alkali treatment: RbF (Uppsala University) vs. Cs (Solar Frontier); (2) Ag content: AAC ~0.19 (Uppsala University) vs. <0.04 (Solar Frontier); and (3) Bandgap profile: a ‘hockey-stick’-like GGI (Uppsala University) vs. a V-shaped gradient (Solar Frontier). Remarkably, despite these profound differences, both devices achieved similarly high V_oc and FF. This compelling parallel necessitates a re-examination of the impact of bandgap engineering on solar cell efficiency.

Bandgap engineering in CIGSe is broadly categorized into forward gradient, reverse gradient, and double gradient (V-shaped) bandgap [12,13,14]. Forward gradient bandgap typically refers to a bandgap configuration that increases linearly from the heterojunction to the Mo back contact, while reverse gradient bandgap is the opposite. Generally, due to Ga accumulation near the back contact in CIGSe, forward gradient bandgap readily forms and helps reduce reverse conduction of minority carrier, thereby suppressing carrier recombination in the quasi-neutral region. However, the surface exhibits a low bandgap due to its lower Ga content. Without introducing surface dopants such as S to enhance the bandgap, recombination intensifies within the space charge region (SCR). Consequently, the V_oc of the corresponding device is generally low, comparable to CuInSe₂ cells at around 0.4 V [15,16]. For a single reverse gradient bandgap, it can increase the quasi-Fermi level splitting at the heterojunction. but often results in an excessively high surface bandgap. This hinders carrier generation and transport at the heterojunction and severely degrades the J_sc of the devices. The double-gradient bandgap combines the advantages of both structures. It features a V-shaped structure with an increasing bandgap from the center toward both ends, referred to as the front-gradient and back-gradient, respectively. Therefore, it is widely employed to enhance photovoltaic efficiency. Two benefits have been demonstrated: (1) the wide bandgap near the two interfaces reduces carrier recombination and enhances the splitting of the quasi-Fermi level, thereby increasing V_oc; (2) the narrow bandgap in the center does not affect the absorption of low-energy photons. Meanwhile, the gradually variation in bandgap distribution does not impair carrier collection, thereby favoring increases in J_sc and FF.

Theoretically, the upper limit of open-circuit voltage (V_oc) increases with absorber bandgap due to enhanced quasi-Fermi level splitting [17,18]. However, in practice, this trend is sublinear due to non-radiative recombination at grain boundaries and trap states [19,20]. The V_oc deficit (V_oc,def) is defined as the difference between the theoretical maximum open-circuit voltage and the actual measured open-circuit voltage, expressed by the formula E_g/q − V_oc. V_oc,def typically reflects the consumption of photogenerated carriers due to non-radiative recombination processes within the device. And it typically worsens with a larger bandgap. This is especially true for wide-bandgap CIGSe, where V_oc saturates or even declines. Thus, effective bandgap engineering for V_oc enhancement must navigate the trade-off between a higher bandgap and the resultant recombination losses. Critically, while bandgap and recombination are not explicitly coupled, they are co-influenced by the same doping strategies.

As a I-III-VI semiconductor, the complex elemental composition of CIGSe facilitates diverse doping strategies. From the perspective of energy band structure, the valence band maximum (VBM) in CISe is primarily formed by the antibonding states of Cu 3d and Se 4p orbitals, while the conduction band minimum (CBM) consists of the antibonding states of In 5s and Se 4p orbitals [21,22]. This electronic structure dictates that substitutional doping at the Cu site predominantly affects the VBM, whereas doping at the In site modifies the CBM. Doping at the Se site, in contrast, can influence both band edges. These diverse doping mechanisms enable extensive tuning of the elemental distribution in CIGSe, thereby allowing for broad bandgap tailoring. However, this compositional complexity also introduces intricate elemental interactions during synthesis and promotes the formation of associated defects, which are major sources of non-radiative recombination. Therefore, a comprehensive analysis of how different dopants affect both the bandgap and carrier recombination is essential for identifying key technologies to further enhance the photovoltaic conversion efficiency of CIGSe devices.

This review systematically examines various elemental doping strategies for bandgap engineering in CIGSe solar cells, such as surface sulfurization, Ga gradient control, and Ag alloying. Particularly, we put a focus on their distinct impacts on V_oc and non-radiative recombination pathways. Furthermore, we compare the influence of conventional V-shaped bandgap profiles with the emerging ‘hockey-stick’-like gradient. Building on this analysis, we conclude by proposing strategic directions for further performance enhancement in CIGSe photovoltaics.

2. Surface Sulfurization

The intentional substitution of Se with S is an established strategy for increasing the surface bandgap of CIGSe absorbers. This alloying effect electronically manifests as a downshift of the VBM and an upshift of the CBM, resulting in a widened bandgap [23,24]. This modification creates a hole transport barrier at the buffer/absorber interface, which repels holes back into the absorber bulk and effectively suppresses interface recombination. Furthermore, sulfur’s propensity to diffuse rapidly along grain boundaries leads to its segregation at these sites and at defects. This creates a beneficial band bending at grain boundaries that suppresses majority carrier (hole) accumulation while enhancing minority carrier (electron) collection efficiency [25]. An additional benefit is the passivation of deep-level trap states by S, which further suppresses non-radiative recombination channels [26]. Nevertheless, the sulfurization process requires precise control. Inappropriate dosing can reduce the fill factor (FF), for instance, through the formation of an overly thick, pure sulfide surface phase. This detrimental phase can introduce an unfavorable spike-like conduction band alignment, acting as a barrier for photogenerated current collection [27]. Alternatively, S may etch away In, promoting the formation of a conductive Cu-(Se,S) ternary phase that creates shunting paths and diminishes the device’s parallel resistance.

The extent of S incorporation is conventionally represented by the ratio y = S/(S + Se). An empirical formula that describes the correlation between the chemical composition of the Cu(In_1-xGa_x)(Se_1-yS_y) phase and its corresponding bandgap is given below [28]:

E_g = 1.00 + 0.13x² + 0.08x²y + 0.13xy + 0.55x+ 0.54y

(1)

The bandgap profile across the absorber is governed by the depth-dependent distributions of the Ga/(In + Ga) ratio (x) and the S/(S + Se) ratio (y). Therefore, effective bandgap engineering necessitates precise control over these elemental gradients along the depth direction. Surface sulfurization, a technique more prevalent in sequential fabrication processes, is a key method for this purpose. Leading research institutions, including Solar Frontier, have extensively developed sulfurization techniques, successfully demonstrating them in high-efficiency solar cells. These sulfurization strategies can be categorized into two primary routes based on process timing: (1) post-sulfurization treatment [29,30,31], and (2) sulfurization after selenization (SAS) [32,33,34]. The choice of approach directly dictates the resulting S distribution, thereby enabling the formation of distinct bandgap profiles.

Post-sulfurization treatment involves exposing the fully selenized CIGSe absorber to a S-containing atmosphere. Once the absorber approaches stoichiometry after complete selenization, the diffusion rate of S is generally suppressed. Under fixed sulfurization conditions, the extent of S incorporation exhibits a clear dependence on the overall Cu/(Ga + In) (CGI) ratio, as illustrated in Figure 1a [29]. In Cu-poor absorbers, S diffusion is facilitated by a higher density of copper vacancies (V_Cu) and grain boundaries. A lower CGI ratio is often associated with inferior crystallinity, which manifests as a higher density of V_Cu and a greater abundance of grain boundaries. In turn, these microstructural features facilitate S diffusion, resulting in its deeper penetration and a higher overall S content within the absorber [35]. This phase acts as a sulfur reservoir, promoting subsequent S incorporation into the chalcopyrite lattice. However, analysis of V_oc trends reveals that deeper sulfurization does not linearly translate into greater V_oc gains, as shown in Figure 1b. When the surface S/(S + Se) ratio is comparable, a shallower sulfurization depth within the absorber is found to be more beneficial for achieving higher V_oc. This indicates that confining sulfurization to a narrow region near the surface is a more effective strategy for V_oc enhancement. The optimal bandgap grading (CGI = 0.94) markedly enhanced V_oc from 0.42 V in the control group to 0.58 V, while the FF increased from 63% to 71%. Since this graded structure did not change the overall minimum bandgap of the absorber layer, J_sc remained relatively stable without significant decline. However, the elevated front-gradient bandgap increased the activation energy for carrier recombination, thereby diminishing non-radiative recombination losses. This resulted in a greater quasi-Fermi level splitting, which contributed to substantial improvements in both V_oc and FF.

The predominant method for post-sulfurization treatment employs H₂S and S vapor at a high temperature (>500 °C). Due to the elevated temperature and abundant sulfur source, S atoms exhibit high reactivity to substitute Se atoms on the surface of CIGSe. However, this high reactivity readily leads to excessive sulfurization. Consequently, researchers have investigated alternative approaches to surface sulfurization using S-based solutions at a lower temperature. One such method involves immersing CIGSSe films in a solution of Group III chlorides and thioacetamide (TA) at 80 °C. This forms a thin CuInS₂ wide-bandgap layer on the surface, significantly enhancing device efficiency [36,37]. Another approach utilizes ammonium sulfide ((NH₄)₂S) vapor as the sulfur source, relying on the spontaneous decomposition of small amounts of H₂S at room temperature for sulfurization [38,39]. These low-temperature sulfurization processes induce a reduction in V_Se density, which is a deep-level trap defect on the surface. However, excessively low temperatures result in reduced sulfur activity, making it difficult to achieve sufficient sulfurization and limiting the passivation effect. Yuan et al. [25] developed a sulfurization process utilizing thioacetamide (TA) at 300 °C. At this temperature, TA undergoes thermal decomposition:

CH₃CSNH₂(s) → CH₃CN(g) + H₂S(g)

The released H₂S exhibits higher reactivity at 300 °C, promoting the inward diffusion of S to passivate V_Se defects. By optimizing the concentration of TA solution, the S/(S + Se) ratio on the CIGSSe surface increased from 3.9% to 9.4%, achieving locally optimal passivation. Regarding device performance, V_oc increased from 0.60 V to 0.65 V, and FF improved from 65.9% to 72.2%. J_sc slightly decreased from 33.16 to 32.53 mA/cm² due to the increased minimum bandgap. Notably, although the aforementioned methods employed different sulfur sources, they essentially relied on H₂S decomposition to supply S atoms for surface sulfurization. The variation in sulfurization temperatures resulted in differences in the reactivity of S atoms.

In addition to tailoring the bandgap, sulfurization plays a crucial role in surface passivation. The passivation effects manifest in two aspects: (1) it passivates deep-level trap defects (V_Se); (2) it modifies the electronic properties at grain boundaries. Multiple studies indicate that regardless of the sulfur source employed, this post-treatment allows S atoms to effectively occupy V_Se. A significant reduction in the signal intensity of V_Se defects was observed after sulfurization treatment in low-temperature PL emission spectra [40]. This directly indicates a decrease of V_Se density on the surface. However, large grains had already grown sufficiently on the surface during the selenization stage, resulting in a reduction in grain boundaries density. The diffusion of S atoms is confined to the surface, limiting the occupation of V_Se. Furthermore, studies indicate that sulfurization also modifies electronic properties at grain boundaries [25]. S atoms tend to aggregate at grain boundaries, thus the VBM bend downward relative to the grain interior. This downward band bending attracts electrons while repelling holes at grain boundaries, thereby enhancing carrier collection and suppressing hole–electron recombination.

Sulfurization after selenization (SAS) denotes a process where sulfurization is conducted immediately following the selenization step. This method, employed by Solar Frontier to achieve multiple world-record efficiencies, proceeds through reaction pathways established by time-freeze experiments [32] (see Figure 2a). A thin CuIn(S,Se)₂ formed at the absorber surface is the key feature of this route. And the challenge is how to confine the sulfurization to a shallow surface region and create the intended double-gradient bandgap profile. Because sulfurization follows selenization without interruption, the surface remains highly reactive. This makes the fill factor (FF) particularly sensitive to the intensity of the surface reaction. Overly aggressive sulfurization can generate highly conductive secondary phases that severely degrade device performance. To improve control over the absorber formation, SF subsequently integrated a rapid thermal process (RTP) into the standard SAS recipe [33]. This modification significantly altered the crystal growth pathway and intermediate phases, resulting in a reduced V_oc,def (Figure 2b). From a band structure perspective, the RTP-modified process flattens the overall bandgap gradient across the absorber. Specifically, the minimum bandgap (E_g) shifts closer to the heterojunction interface while the difference between the minimum E_g and the surface bandgap is narrowed [41] (Figure 2c), thereby further mitigating the performance losses associated with excessive sulfurization. The narrower surface bandgap gradient increased V_oc from 0.72 to 0.75 V while simultaneously raising FF from 78.2% to 79.7%. This improvement stems from the enhanced suppression of non-radiative recombination due to the further increase in surface bandgap. However, the minimum bandgap expansion led to a reduction in J_sc from 39.4 to 38.5 mA/cm².

For SAS processes, S diffusion is strongly influenced by the overall stoichiometry and Ga content of the absorber after selenization. The S diffusion coefficient in Cu-rich absorber is approximately two orders of magnitude lower than that in the Cu-poor absorber, indicating that Cu-poor films facilitate more extensive S diffusion. During sequential processing, Ga accumulation near the Mo back contact promotes the formation of fine-grained regions with low CGI, where S also tends to accumulate [42,43,44]. Huang et al. [45] investigated the elemental distributions of S and Ga under different sulfurization durations. They observed that the initially formed CIGSSe (CuIn_1-xGa_xS_ySe_1-y) phase showed Ga- and S-depleted regions near the absorber surface. With prolonged high-temperature treatment, both Ga and S diffused toward the front surface, leading to a more uniform redistribution. The similar distribution profiles of Ga and S suggest the possible formation of a Ga–S binary compound in the back region, which may further promote co-diffusion driven by the S atmosphere at the surface (as shown in Figure 3b,c). In contrast, Başol et al. [35] argued that the co-accumulation of Ga and S in the back region may not originate from a preferential chemical reaction between them, since the S/Ga molar ratio is not constant. Instead, they proposed that Ga accumulation leads to a fine-grained, low-CGI layer, whose high grain boundary density favors S accumulation. The resulting high local S concentration can affect band alignment between the CIGSSe absorber and the Mo back contact: S doping lowers the VBM, creating a barrier that hinders hole transport. When carrier transport is blocked by this barrier, a characteristic inversion behavior appears in the J–V curve. Siew et al. [46] demonstrated that heat soaking treatment can modify the S distribution near the Mo interface, reducing the width of S accumulation and thus improving back-contact properties (as shown in Figure 3d). But the heat soaking treatment has a negligible effect on the doping concentration at the front interface of the absorber layer, thus exerting little influence on the device’s V_oc and J_sc. However, the improvement of the bandgap gradient at the Mo back contact effectively enhances the FF of the cells from 61.2% to 65.9%, thereby significantly boosting the overall cell efficiency.

It can be seen that the front interface bandgap gradient significantly influences V_oc and J_sc. Combined with the analysis of surface sulfurization treatment discussed earlier, this effect manifests in two aspects: (1) The increase in bandgap near the heterojunction interface effectively reduces non-radiative recombination in the SCR, thereby enhancing V_oc performance. (2) The value and distribution width of the minimum bandgap within the bandgap gradient determine the overall spectral response range of the absorber layer, consequently affecting variations in J_sc. The FF is influenced by both the front and back bandgap gradients. For sulfurization treatment, the S substitution doping simultaneously shifts the CBM upshift and the VBM downshift, which will change the band alignment at the interfaces. Band alignment is typically evaluated by the VBO and CBO value of materials on both sides of the interface. The VBO is the difference in their VBM level (VBM_absorber − VBM_buffer) while CBO is the difference in their CBM level (CBM_absorber − CBM_buffer). The ideal CBO value should fall within the range of 0 to 0.4 eV while VBO need to be >0 [47]. For CBO < 0, carriers are more prone to non-radiative recombination. For CBO > 0.4 eV or VBO < 0, the collection of photogenerated carriers is hindered.

Therefore, the S/(S + Se) ratio at both ends should not be excessively high. If the ratio near the back contact is too high, the large valence band offset (VBO) between CIGSSe/Mo will significantly impair the collection efficiency of hole carriers at the Mo back contact. Conversely, excessive S concentration at the front interface may result in an excessive conduction band offset (CBO) between CIGSSe/CdS, which significantly hinders the mobility of photogenerated electron within the heterojunction region. Gi Soon Park et al. [48] obtained a S-rich surface in CIGSeS via H₂S sulfurization, leading to a maximum S/(S + Se) ratio of 0.65 at the surface. However, this S-rich surface creates an unfavorable cliff-like band alignment in the conventional CdS/CIGSeS heterojunction structure. This phenomenon occurs due to the substantial upshift of the CBM at the absorber surface caused by high S doping, as shown in Figure 4a. The incorporation of a (Cd,Zn)S buffer layer with a higher CBM level reinstated a spike-like band alignment, thereby improving electron transport at the heterojunction. This improvement is reflected in the enhancement of signal intensity within the long-wavelength range of the EQE spectrum, as depicted in Figure 4c. Ultimately, the optimized band structure of the heterojunction increased the V_oc from 0.55 V to 0.58 V, boosted the J_sc from 32.6 mA/cm² to 34.7 mA/cm², and elevated the FF from 70.9% to 71.0%.

3. Ga Gradient

The involvement of Ga marked a landmark advance in the development of CIGSe solar cells. By adjusting the In/Ga ratio, the bandgap profile of the absorber can be precisely tuned [49]. While a higher GGI ratio widens the bandgaps and thus promotes a higher V_oc, the benefit is counteracted once GGI exceeds 0.4, as the increase in V_oc,def outweighs the voltage gain from bandgap widening [50,51]. Research indicates that this trade-off arises from a severe reduction in minority carrier lifetime, which drops from approximately 400 ns at E_g = 1.13 eV to about 0.2 ns at E_g = 1.56 eV [52]. Consequently, optimizing the spatial distribution of Ga has emerged as a more effective strategy than simply raising the overall GGI ratio [53].

However, the resulting Ga distribution profile is highly dependent on the film deposition process, making precise control essential for achieving the designed bandgap structure. In sequential processes, the different activation energies of In-Se and Ga-Se reaction lead to preferential Ga accumulation near the back contact during selenization [54,55]. M. Marudachalam et al. [56] compared the selenization behaviors of CuGa and CuIn precursors. They revealed that complete selenization of CuGa requires a substrate temperature of at least 450 °C, whereas CuIn can be fully selenized at 400 °C with shorter processing times. This phenomenon highlights the thermodynamic disparity between the two reactions. Therefore, CuInSe₂ grows faster on the surface and gradually penetrates inward as reaction time increases. During this process, most In diffuses into the near-surface region, naturally displacing Ga toward the Mo back contact. For a narrow-bandgap CIGSe (GGI ≈ 0.2) with a total thickness of approximately 2 μm, the CuGaSe₂ layer measures about 0.4 μm. This implies Ga must diffuse 1.6 μm while In requires only 0.4 μm to achieve uniformity. According to Fick’s law (t ≈ L²/(4D)), even with identical diffusion coefficients for In and Ga, Ga requires a longer diffusion time to achieve homogenization. This determines a forward bandgap gradient, which refers to a gradually increasing GGI distribution toward the back contact.

However, research indicates that the diffusion coefficient of Ga is smaller than In, meaning that Ga homogenization requires a longer diffusion time. Witte et al. [57] further elucidated the kinetic aspect by calculating the V_Cu-mediated diffusion barriers, showing that the barrier for Ga diffusion in CuInSe₂ (1.41 eV) is substantially higher than that for In diffusion in CuGaSe₂ (0.87 eV). Therefore, according to the formula D~exp(-E_a/kT), the D_Ga in CuInSe₂ is smaller than D_In. Experimental results by Schroeder et al. [58] corroborate this observation: at 1000 K, D_Ga is approximately 10⁻¹² cm²/s for the Ga diffusivity into CuInSe₂ single-crystal epitaxial layers approaching stoichiometry, which is about three orders of magnitude smaller than the In self-diffusion rate in CuInSe₂ single crystals at the same temperature. Consequently, once ternary compounds form separately at high temperatures, In can diffuse readily into Ga-rich regions, whereas Ga diffusion into the In-rich front surface is strongly hindered. These thermodynamic and kinetic constraints make it particularly challenging to establish a Ga-front gradient in sequential processes.

The three-stage co-evaporation process offers active control over the evaporation timing and rates of individual elements, thereby modifying the CIGSe growth pathway and mitigating issues related to Ga accumulation. By dynamically adjusting the evaporation rates of Ga and In during different stages, the In/Ga ratio can be spatially tailored, enabling customization of the position and magnitude of the V-shaped bandgap gradient [59,60]. Xiao et al. [61] systematically optimized the V-shaped bandgap profile by precisely regulating the In and Ga evaporation rates in the first and third stages (Figure 5a). Through correlation with device performance, they identified key characteristics of the optimal bandgap structure (Figure 5b), which can be summarized as follows:

(1) A pronounced back gradient helps establish a stronger back-surface field, promoting the drift of photogenerated electrons from the quasi-neutral region into the space-charge region for efficient collection.

(2) A relatively wide notch width is beneficial, as it enhances the absorption of low-energy infrared photons—whose absorption coefficient is inherently low—by providing sufficient optical path length.

(3) A steep bandgap rise should be maintained near the heterojunction interface, where a narrow yet rapidly increasing bandgap helps improve V_oc without excessively elevating the CBM, thus avoiding unfavorable band alignment with the buffer layer.

These design principles not only apply to Ga grading but also offer general guidance for bandgap engineering using other dopants.

Specifically, Xiao et al.’s adjustment to the GGI bandgap profile ultimately yielded two major beneficial effects on device performance. First, the gradual increase in the front bandgap gradient elevated V_oc from 0.69 V to 0.73 V, achieved by suppressing non-radiative recombination losses and enhancing quasi-Fermi level splitting. Second, a significant reduction in the minimum bandgap broadened the spectral response range of the absorber layer, substantially increasing J_sc from 29.8 to 34.1 mA/cm². Devices implementing the optimized profile achieved a maximum conversion efficiency of 20.3%. However, this adjustment also had a detrimental effect, causing FF to decrease from 76.5% to 74.7%. Although the paper does not provide a detailed analysis of this phenomenon, we speculate that excessive GGI on the surface may have negatively impacted the bandgap alignment of the heterojunction.

Although the sequential process inherently induces Ga accumulation near the Mo back contact, several strategies have been reported to enhance Ga diffusion toward the front side of the absorber. These include regulating alkali metal content [63,64], tuning selenization temperature [65,66], and engineering precursor structure [67]. Nevertheless, such approaches offer only limited modification of the Ga distribution profile.

Beyond these methods, optimizing the Se supply during selenization provides an effective route to tailor the Ga profile. For instance, Kim et al. demonstrated that both the three-step H₂Se/Ar/H₂S and H₂Se/Ar/H₂Se processes significantly improve Ga uniformity [62]. A key stage in these processes is the Ar annealing step, during which rapid grain growth occurs, and Ga undergoes substantial redistribution (Figure 5c). The homogenization of Ga can be attributed to two main factors. First, after the initial selenization, a considerable amount of Ga remains near the Mo back contact in the form of Cu₉Ga₄ intermetallic compounds, while the absorber maintains a fine-grained structure. This configuration preserves high interface and/or surface energy, facilitating grain boundary migration and Ga homogenization during Ar annealing. Second, in the Se-free atmosphere of this stage, residual Cu-Ga intermetallics react with pre-formed CIGSe at the back side of the absorber. The Cu₉Ga₄ phase is considered to provide a fast diffusion pathway for Ga ions, as it coexists in liquid and solid states above 485 °C [68,69]. Consistent with this, Schmidt et al. [70] also showed that reducing the Se supply during intermediate annealing accelerates In/Ga inter-diffusion, thereby alleviating pronounced Ga accumulation.

Facilitating Ga diffusion through the ordered vacancy compound (OVC) phase represents another approach to controlling Ga gradients [71,72]. The OVC phase often refers to the Cu(In,Ga)₃Se₅ phase. Different from the stoichiometric ratio of Cu(In,Ga)Se₂, the OVC phase can provide a large amount of V_Cu to accelerate Ga diffusion. Tu et al. [73] demonstrated that under relatively low-CGI conditions, high Se vapor pressure promotes the formation of an OVC phase in the early stages of selenization. Although Ga diffusion in CIGSe is generally hindered by a high energy barrier, the abundance of V_Cu in the OVC structure significantly enhances Ga mobility toward the front region of the absorber. As illustrated in Figure 6a, this mechanism enables an optimized Ga distribution and improves overall cell efficiency. The results indicate that as the P_Se increases, the GGI on the surface rises from 0 to 0.30, significantly expanding the surface bandgap. This change corresponds to an increase in V_oc from 0.42 V to 0.59 V, and the FF increases from 38.4% to 72.0%. J_sc exhibits minimal fluctuation with changes in P_Se, probably due to a balance between the effects of bandgap widening and grain size enlargement.

The formation of the OVC phase is strongly influenced by both CGI and Se vapor pressure. At high CGI (=0.92), OVC formation is suppressed, preserving a steep Ga gradient. In contrast, under low CGI (=0.77) combined with high Se vapor pressure, Ga accumulation is substantially reduced, leading to a more uniform Ga distribution, as shown in Figure 6b. It should be noted, however, that high Se vapor pressure often promotes the excessive growth of a MoSe₂ layer, which may necessitate the use of a diffusion barrier to optimize back-contact properties.

The use of single-quaternary CIGSe ceramic targets in sputtering yields a uniform in-depth element distribution, as In and Ga are pre-alloyed with Se, thus requiring external layers to introduce bandgap grading. While a Ga-front gradient has not yet been realized via this route, a few studies have aimed at establishing a Ga-back gradient near the Mo interface. Employing Ga₂Se₃ as a back-side Ga source is challenged by its thermodynamic instability: under high-power sputtering, it reacts with residual oxygen to form Ga₂O₃ phase. This detrimental phase is an n-type wide-gap phase that impairs device performance if incorporated at the back contact. The use of CuGaSe₂ (CGSe) layers presents another option [74]. Since Ga loss during deposition induces Cu-rich secondary phases like Cu_2-xSe, precise Cu composition control is essential. Co-sputtering CIGSe and CGSe targets allows flexible design of Ga gradients and has been associated with improved efficiency, as depicted in Figure 6c,d. Wang et al. [74] increased GGI near the back contact from 0.26 to 0.53 through this approach. This modification redirected minority carriers away from recombination in the back region, thereby prolonging carrier lifetime. For corresponding device performance, V_oc significantly increased from 0.54 V to 0.61 V, and FF rose from 71.8% to 75.2%. Though the minimum bandgap remained unchanged, the back bandgap gradient enhanced carrier collection efficiency in the long-wavelength spectrum, boosting J_sc from 30.5 to 33.1 mA/cm².

4. Ag Alloying

Ag alloying has been demonstrated to offer several advantages in CIGSe solar cells, including: (1) suppression of lattice disorder induced by Ga incorporation [75,76], (2) formation of a low-melting-point liquid phase that enhances crystallization [77,78], and (3) bandgap tuning through substitution for Cu in the lattice [79,80].

Theoretically, since the VBM in CISe is primarily derived from Cu 3d antibonding orbitals, Ag substitution was expected to mainly influence the VBM. However, detailed studies reveal that Ag doping affects both the Cu 4s and 3d orbitals, thereby modifying both the CBM and VBM. As shown in Figure 7a [81], under a fixed GGI ratio, increasing the Ag/(Ag + Cu) (AAC) content simultaneously lowers both CBM and VBM energy levels, with a more pronounced shift in the VBM due to its strong Cu 3d character. Consequently, Ag alloying effectively increases the optical bandgap.

Expressing AAC as w, the bandgap E_g can be empirical described by the following relation [82]:

E_g = 0.24w² + 0.03w + 0.08x² + 0.61x − 0.11xw + 1.01

(2)

This demonstrates that tailoring the spatial distribution of Ag can similarly be used to establish a gradient bandgap within the absorber layer. Nevertheless, research on utilizing Ag to construct specific bandgap profiles remains limited, which may be attributed to the relatively low AAC doping levels (typically between 0.01 and 0.2) commonly employed in high-efficiency ACIGSe (Ag_wCu_1-wIn_1-xGa_xSe₂) solar cells. As illustrated in Figure 7a, within this low doping range, Ag has a negligible influence on the absorber bandgap. A significant increase in bandgap is only observed when AAC exceeds 0.5.

As mentioned above, S and Ga doping both raise the CBM to widen the bandgap but risk forming a cliff-like band alignment at the buffer/absorber interface under excessive doping. However, Ag doping reduces the CBM, avoiding this band mismatch risk. As shown in Figure 7b, this CBM downshift facilitates a favorable spike-like band alignment between ACIGSe and the conventional CdS buffer layer. Such an interface helps mitigate unfavorable carrier transport in high-GGI solar cells caused by an excessive conduction band offset (CBO < 0). However, excessively high Ag doping concentrations can also cause an excessive downshift in CBM. In turn, this leads to an inverse excess CBO (CBO > 0.4 eV), forming a barrier that hinders the transport of photogenerated electrons.

In addition, high concentrations of Ag doping also introduce several other challenges. For one thing, similar to the formation of Cu_2-xSe secondary phases in Cu-rich absorbers, Ag-rich ACIGSe layers are prone to form highly conductive Ag_2-xSe secondary phases. These can create shunt paths that severely degrade solar cell performance. For another thing, the introduction of large amounts of Ag significantly occupies V_Cu, leading to a decrease in V_Cu density within the absorber layer. However, as shallow-level defects in CIGSe, V_Cu provides free carriers and serves as the source of the p-type characteristics in CIGSe semiconductors. Therefore, a substantial reduction in V_Cu would negatively affect the carrier density in the absorber layer, which is highly detrimental to device performance.

Tu et al. [80] achieved a notable cell efficiency of 19.7% by employing a CuGa/In/AgGa triple-layer precursor structure as shown in Figure 7c. This design resulted in a front-gradient Ag distribution with a high concentration of Ag doping (AAC = 0.5). To compensate for the carrier density reduction induced by heavy Ag doping, K doping was intentionally incorporated [83]. Moreover, despite a high [I]/[III] ratio of 0.94, a benign OVC phase formed on the surface instead of the detrimental Ag_2-xSe phase. Previous studies suggest that Ag_2-xSe typically forms during the second stage of co-evaporation, as (In,Ga)₂Se₃ reacts more readily with Cu than with Ag. In contrast, the AgGa/AgIn alloy can effectively prevent Ag from directly reacting with Se. As a result, regardless of whether the Ag-containing layer is placed above or below the precursors, no Ag_2-xSe is detected in the film.

The observed Ag front-gradient appears to be closely related to the stacked precursor architecture. Although Ag generally exhibits a high diffusion rate and tends to distribute uniformly, placing the Ag-containing layer on top of the precursors (T-ACIGSe) helps maintain relatively high Ag content near the front surface after selenization. This phenomenon can be attributed to distinct phase formation pathways during film growth (as illustrated in Figure 7c), ultimately leading to a slight Ag enrichment at the front. Furthermore, the introduction of high Ag concentrations promotes the formation of OVC phases at the surface. First-principles calculations reveal that OVC phases possess lower surface energy than ACIGSe phases, making it thermodynamically favorable for the ACIGSe surface to decompose into OVC phases as a passivation mechanism—a finding consistent with experimental observations. These surface OVC phases increase the valence band offset (VBO), thereby effectively suppressing interfacial recombination. The reduction in non-radiative recombination losses enhances the splitting of the quasi-Fermi level within the space charge region, causing a significant rise in V_oc from 0.47 V to 0.70 V. Meanwhile, the FF also significantly increases from 63.1% to 72.6%. Although the distribution of the minimum bandgap remains unaffected, J_sc slightly increases from 32.1 to 32.3 mA/cm², probably due to the enhanced crystallinity introduced by Ag.

Another approach for bandgap tailoring via Ag doping involves the Ga distribution profile by promoting atomic diffusion. Kim et al. [84] investigated the effect of Ag precursor thicknesses at 450 °C on the V-shaped Ga gradient formed by a three-stages co-evaporation method. They observed that as the Ag doping content increases, the steep V-shaped bandgap notch becomes progressively flatter (Figure 8g). It has been demonstrated that an excessively deep bandgap gradient can create an electron collection barrier and impede electron transport. The flattening of the bandgap notch facilitates more efficient transport of photogenerated electrons and reduces electron–hole recombination, thereby enhancing photoelectron collection and consequently improving the V_oc and FF of the solar cells. As the Ag content increases, both the front and back bandgap gradients become smoother, thereby enhancing carrier collection efficiency. Consequently, the V_oc increases from 0.63 V to 0.68 V, and the FF significantly rises from 44.3% to 70.6%. Although the minimum bandgap notably increases compared to the control group, the enhanced collection of photogenerated carriers counteracts the effects of bandgap changes, resulting in negligible variation in J_sc.

The mechanism by which Ag doping flattens the bandgap notch can be attributed to two primary factors. First, the Ag-Se system forms a pure liquid phase at a significantly lower temperature (221 °C) compared to the Cu-Se system, which begins to form a liquid phase only above the eutectic temperature of 523 °C. This substantial difference in melting behavior enables Ag to promote the formation of the ACIGSe phases at lower temperature and enhance grain growth, as shown in Figure 8a–f. Second, the dissociation energy of the Ag-Se bond (201 kJ/mol) is lower than that of the Cu-Se bond (255 kJ/mol). This facilitates the segregation of Ag to grain boundary and the formation of a Ag-Se liquid phase under high Se flux. However, this property also favors the preferential reaction of Cu with (In,Ga)₂Se₃, resulting in the formation of the Ag_2-xSe secondary phases at a high Ag doping concentration.

Prathapani et al. [85] systematically compared the effects of different AAC levels on the distribution of Ag and Ga in absorbers with different CIG ratios and temperatures. When the total [I]/[III] ratio is between 0.99~1.03, Ag consistently segregates to the surface and becomes depleted in the bulk. In absorbers with CGI = 0.86, low AAC levels lead to a relatively uniform Ag distribution, whereas higher doping (AAC = 0.06) still results in Ag accumulation at the surface. Further reducing the CGI ratio can alleviate this surface accumulation (Figure 8h–m). It should be noted that Ag accumulation on the surface does not constitute the ideal front-gradient profile [86]. Instead, it predominantly leads to the formation of Ag_2-xSe secondary phase, which creates shunt paths and degrades device performance. Therefore, an appropriate Ag doping range combined with a low CGI is essential to achieve an optimal Ag gradient and enhance solar cell efficiency.

Furthermore, although a lower CGI promotes more Ag incorporation into the absorbers, studies indicate that as Ag doping increases, excessive Cu-poor ACIGSe (CGI < 0.84) results in a higher V_oc,def, which limits further efficiency improvement.

5. Analysis of V-Shaped vs. ‘Hockey Stick’-like Bandgap Distributions

The V-shaped bandgap gradient strategy has been extensively studied and validated through both theoretical simulations and experiments, with its efficacy in enhancing solar cell efficiency well-documented [87,88,89]. For instance, the 22.6% world-record efficiency by ZSW in 2016 utilized a three-stage co-evaporation method to construct a V-shaped GGI profile, which effectively suppressed bulk recombination and improved carrier collection [90]. Similarly, the 23.35% champion device from Solar Frontier in 2019 employed a sequential process that leveraged a natural back gradient from Ga accumulation and a front gradient from sulfurization to form the V-shaped bandgap. In contrast, the latest state-of-the-art device (23.64%, Uppsala University, 2023) deviates from this established approach. It utilizes a ‘hockey-stick’-like GGI profile, yet achieves comparably high V_oc and FF, as shown in

Table 1. PV parameters of state-of-the-art (A)CIGS solar cells from the literature [4].

	Voc (mV)	Jsc (mA/cm2)	FF (%)	η (%)
ZSW 2016	741	37.8	80.6	22.6
SF 2019	734	39.6	80.4	23.35
UU 2023	767	38.3	80.5	23.64

. This departure from the conventional V-shaped design compels a re-examination of the characteristics of both gradient architectures to identify the key factors limiting further efficiency breakthroughs in CIGSe devices.

The ‘hockey-stick’-like bandgap reported by Uppsala University is characterized by a compositionally uniform front surface and a Ga-gradient distributed back surface, as illustrated in Figure 9a. This design, combined with a relatively high Ag doping concentration (AAC = 0.19) and a standard RbF post-deposition treatment, effectively suppresses lateral and in-depth bandgap fluctuations within the absorber. The resulting champion device achieved an ideality factor of n = 1.30, which is significantly lower than that of previously reported high-efficiency cells. This lower value indicates a substantial reduction in the overall recombination rate within the space-charge region (SCR), an effect likely attributable to the synergistic roll of Rb and Ag. Notably, despite the high Ag content (AAC ~0.19), the overall [I]/[III] ratio in the absorber remains low (0.83~0.84). This specific composition prevents Ag surface accumulation while promoting a uniform distribution of other elements near the front surface. Although the elemental distribution is relatively uniform in the bulk of the front absorber, the grain boundary exhibits a distinct chemical character. As shown in Figure 9d–j, grain boundary is depleted in Cu but enriched in Ag and Rb. It has been reported that such Cu depletion and Rb segregation at grain boundary can induce a beneficial upward band bending, which helps to suppress carrier recombination [91,92,93].

In summary, suppressing non-radiative recombination in the space-charge region presents a key advantage of the ‘hockey-stick’-like bandgap profile. This design enables high V_oc and FF even in the absence of a conventional front bandgap gradient—an outcome that can be rationalized theoretically. It is well-explained that a front-gradient bandgap serves two primary roles: first, it enhances the built-in electric field near the surface without sacrificing photon absorption, thanks to the lower bandgap in the bulk; second, it increases the activation energy for carrier recombination in the space-charge region, thereby reducing the non-radiative recombination. In the record Uppsala University cell, the incorporation of Rb and Ag appears to mimic the second function, effectively lowering the overall recombination. While the first function—strengthening the built-in electric field—may remain irreplaceable under certain conditions, the comparable FF and V_oc achieved without it indicate that the built-in electric field is not the primary factor limiting further efficiency improvement in high-performance devices. Consequently, future research efforts are likely to focus increasingly on understanding and mitigating non-radiative recombination within the absorber.

Homogeneous surfaces generally exhibit fewer strain-related defects and reduced bandgap fluctuations, which in turn suppress non-radiative recombination, increase V_oc and lower V_oc,def [94]. Furthermore, expanding the thickness of the region with a small and constant minimum E_g enhances the absorption of low-energy photons, thereby improving J_sc. Indeed, the V-shaped bandgap gradient strategy also allows for thickening the E_g,min region—that is, flattening the V-shaped notch. In practice, the three-stage co-evaporation method achieves this by carefully regulating the evaporation rates of In and Ga to position the notch closer to the surface. Solar Frontier’s sequential process employs RTP to optimize the GGI profile for a flatter bandgap gradient. Both approaches confirm that broadening the notch area contributes to higher solar cells efficiency.

In fact, a wider notch area within the space-charge region functionally resembles a homogeneous surface. Therefore, the two bandgap gradient strategies are not fundamentally conflicting. The conventional V-shaped profile can be optimized by increasing the notch thickness and shifting it closer to the heterojunction surface. However, achieving precise control over the notch profile during actual fabrication remains a significant challenge.

6. Conclusions and Perspectives

In summary, this review has analyzed substitutional doping approaches in CIGSe semiconductors, linking different fabrication methods to their resulting bandgap profiles. Controlling the bandgap distribution has been identified as a central theme. A key insight is that while sophisticated bandgap engineering can improve all photovoltaic parameters, it introduces a critical challenge: the pursuit of optimal electronic properties necessitates unprecedented precision in controlling elemental distribution throughout the manufacturing process.

In sequential fabrication processes, surface sulfurization is commonly employed to establish a V-shaped bandgap gradient, leveraging the inherent back-gradient accumulation of Ga. However, this approach demands precise control over both the degree of surface sulfurization and the resulting diffusion depth. Among various doping strategies, Ga doping continues to be the central focus and a key challenge in tailoring the bandgap distribution of CIGSe absorbers. For instance, Solar Frontier further improved device efficiency by optimizing the GGI profile via RTP, underscoring the critical role of Ga-mediated bandgap engineering. Although several methods have been proposed to regulate Ga diffusion in sequential processes, this area still requires deeper investigation. In contrast, reports on constructing Ag alloying gradients remain relatively scarce. While early studies mainly utilized low Ag doping concentrations, recent years have seen a growing number of reports demonstrating high-efficiency devices achieved with high-concentration Ag alloying [95,96,97]. Notably, the latest state-of-the-art solar cell employed a relatively high Ag concentration (AAC ≈ 0.19). The homogeneous surface composition and unique bandgap profile reported by Uppsala University invite a renewed evaluation of current bandgap engineering paradigms. Additionally, alkali metal post-deposition treatment (PDT) is often combined with Ag alloying to achieve superior electronic properties in the absorber layer [86,98,99].

This review further posits that a uniform front bandgap combined with Ag/Rb doping represents a refinement, not a rejection, of the V-shaped bandgap concept, emphasizing the critical role of non-radiative recombination suppression. From this perspective, the integration of the two bandgap engineering approaches may yield unexpected synergistic effects. (1) We can first employ high-concentration Ag alloying to achieve compositional homogenization, thereby reducing non-radiative recombination loss caused by compositional and bandgap fluctuations. (2) Heavy alkali metal PDT can compensate for the decrease in carrier concentration induced by Ag alloying while further passivating grain boundaries. (3) Employing an appropriate post-sulfurization process can enhance the surface bandgap and establishes a rational, narrow front-gradient bandgap. This can further strengthen the splitting of the quasi-Fermi level at the heterojunction. Each of these measures has been, respectively, validated to suppress non-radiative recombination losses in CIGSe, and they do not present any insurmountable technical conflicts or difficulties in process integration. However, whether all elements can indeed distribute themselves rationally as anticipated and fulfill their respective roles requires further experimental verification.