20%-efficient polycrystalline Cd(Se,Te) thin-film solar cells with compositional gradient near the front junction

The study addresses how to reduce nonradiative recombination at the front junction of Cd(Se,Te) thin-film solar cells while using a commercial SnO₂ buffer, in order to improve open-circuit voltage (Voc) and overall efficiency. CdTe photovoltaics are cost-competitive, and incorporating Se to form Cd(Se,Te) reduces bandgap to <1.4 eV, improving Jsc but historically compromising Voc. Prior gains in Cd(Se,Te) devices often relied on a ZMO buffer, which creates a beneficial conduction-band spike and downward band bending near the front junction to suppress recombination, enabling Voc improvement. However, ZMO suffers from low electron conductivity (leading to S-kinks and low FF) and moisture sensitivity, harming reproducibility and stability. Using SnO₂ is attractive due to stability and reproducibility, but the SnO₂/Cd(Se,Te) interface exhibits higher nonradiative recombination. A bandgap gradient within the absorber is a proven route in CIGS and CZTSe devices to suppress front-interface recombination without introducing a detrimental heterointerface. In Cd(Se,Te), simply adding CdS tends to form a photo-inactive wurtzite Cd(S,Se) layer and an additional heterointerface, increasing recombination. The research hypothesis is that introducing a photoactive, wider bandgap region within the same crystal structure near the front junction—by incorporating oxygenated CdS and CdSe that intermix with CdTe—can create a Cd(O,S,Se,Te) region, produce favorable band alignment and carrier densities, suppress nonradiative recombination at the front junction, and increase Voc and device efficiency while retaining the practical advantages of SnO₂ buffers.

- Bandgap grading improves Voc in Cu(In,Ga)Se₂ and Cu(Zn,Sn)Se₂ by suppressing front-interface recombination; typically achieved via homovalent alloying (e.g., varying Ga/In). Recent work also used Ag incorporation to create grading regions. - In Cd(Se,Te), the ZMO buffer provides a higher CBM than Cd(Se,Te), forming a CBM spike at the ZMO/Cd(Se,Te) interface and inducing beneficial downward band bending, reducing recombination and enabling higher Voc. However, ZMO has low electron conductivity (even with doping), causing S-kinks and low FF, and is moisture-sensitive, raising stability concerns. - SnO₂ buffers are stable, reproducible, and have desirable electron conductivity and have been used in CdTe manufacturing for decades, but the SnO₂/Cd(Se,Te) interface lacks the beneficial band alignment, resulting in higher nonradiative recombination and lower Voc. - A straightforward attempt to form a graded region by stacking CdS/CdSe/CdTe and interdiffusing during CdCl₂ treatment tends to form a separate, photo-inactive wurtzite Cd(S,Se) layer and a heterointerface to Cd(Se,Te), which increases recombination and degrades Voc, Jsc, and PCE in prior reports.

Device concept and fabrication: - Baseline stack uses commercial SnO₂ as n-type buffer/emitter. Conventional fabrication deposits CdTe onto CdSe; after CdCl₂ treatment, interdiffusion forms a Cd(Se,Te) absorber with Se decreasing and Te increasing from front to back. - Proposed approach introduces oxygenated CdS and CdSe layers: deposit Cd(O,S) and Cd(O,Se) prior to CdTe. During CdCl₂ treatment, these amorphous, oxygen-containing layers fully intermix with CdTe to form a photoactive penternary Cd(O,S,Se,Te) region near the front junction with the same zinc blende structure as the absorber, avoiding a detrimental heterointerface. - Control devices: SnO₂/Cd(Se,Te)/CdTe without S incorporation (no Cd(O,S,Se,Te) region). SCAPS-1D simulation: - Absorber modeled as three layers without internal composition gradients: Cd(S,Se,Te) (front), Cd(Se,Te) (middle), CdTe (back). All assumed zinc blende crystal structure and no detrimental heterointerfaces. - Example parameters: S concentration ~10% giving Eg = 1.5 eV for the front 100 nm Cd(S,Se,Te) layer; free hole density p = 2×10^16 cm^-3 in Cd(S,Se,Te) and 5×10^16 cm^-3 in Cd(Se,Te); conduction-band spike ΔEc = +0.05 eV and valence-band offset ΔEv = −0.07 eV at Cd(S,Se,Te)/Cd(Se,Te); interface trap density at SnO₂/absorber fixed at 6×10^13 cm^-3. - Outputs: band diagrams, recombination current components (including interface recombination J_interf), and J–V characteristics. Additional parameter sweeps varied Cd(S,Se,Te) bandgap and front-junction trap density. Structural and compositional characterization: - Cross-sectional STEM-HAADF with EDS mapping and line profiles compared stacks made with Cd(O,S)/CdSe versus Cd(O,S)/Cd(O,Se). The former showed a distinct Cd(S,Se) layer and interface, while the latter showed continuous mixing with no heterointerface and a graded Cd(O,S,Se,Te) region. - Time-of-flight SIMS profiling of peeled absorber, starting from the Cd(O,S,Se,Te) side, quantified O, S, Se, Te near the front. Initial concentrations: O 2.75×10^21 cm^-3, S 9.88×10^20 cm^-3, Se 3.24×10^21 cm^-3, Te 5.42×10^24 cm^-3, corresponding to CdO0.222S0.08Se0.261Te0.437 (or approximated CdS0.102Se0.336Te0.562 neglecting O). Oxygen concentration is high at the interface and decays rapidly within ~0.5 μm. - Bandgap estimation: Co-evaporated Cd(S,Se,Te) films with varying S content (without O due to deposition tool limits) were measured to derive Eg versus S; Eg for CdS0.102Se0.336Te0.562 estimated ≈1.5 eV. With O incorporation, the Cd(O,S,Se,Te) bandgap is inferred to be >1.5 eV. Across the absorber, Eg trends from >1.5 eV (front) to ~1.38 eV (Cd(Se,Te) mid) to ~1.54 eV (CdTe back). Optoelectronic characterization: - Steady-state photoluminescence (PL) measured from the glass side with excitation at 405 nm and 633 nm to probe depth-dependent recombination near the front interface. - Time-resolved PL (TRPL) measured with 633 nm excitation from the glass side; decay fits used a tri-exponential model to extract τ1, τ2, τ3 and average lifetime. - Photoluminescence quantum yield (PLQY) used to estimate ideal Voc (iVoc) and ΔiVoc via ΔiVoc = (kT/q)·ln(PLQY_target/PLQY_control). Device performance: - J–V and EQE measured for statistically significant sets (n=15 per group). Cu doping applied to both control and target devices. Champion devices reported with Voc, Jsc, FF, and PCE.

- Simulation indicates beneficial bandgap grading: Introducing a 100 nm Cd(S,Se,Te) front region (Eg ~1.5 eV, ΔEc ~+0.05 eV, p ~2×10^16 cm^-3) produces downward band bending in Cd(Se,Te), reduces front-interface recombination current density J_interf (especially for forward bias >0.7 V), and increases Voc from 0.832 V to 0.882 V. Simulated PCE increases from 18.9% to 20.3% with the graded region, at fixed interface trap density (6×10^13 cm^-3). Higher front-region bandgap and lower front-junction trap density further increase Voc. - Structural evidence: STEM-HAADF/EDS shows that using oxygenated CdS and CdSe avoids the formation of a separate wurtzite Cd(S,Se) layer and any detrimental heterointerface; instead, a continuous, photoactive Cd(O,S,Se,Te) region with compositional gradient forms near the front junction. - Composition and bandgap: TOF-SIMS indicates a front-region composition around CdO0.222S0.08Se0.261Te0.437 (approx. CdS0.102Se0.336Te0.562 neglecting O). Eg for the analogous O-free composition is ~1.5 eV; with O present, Eg is inferred to be >1.5 eV. - Recombination suppression evidenced by PL/TRPL: • Control (SnO₂/Cd(Se,Te)/CdTe): PL intensity decreases by ~60% when excitation changes from 633 nm to 405 nm, indicating stronger nonradiative recombination near the front interface. • Target (SnO₂/Cd(O,S,Se,Te)/Cd(Se,Te)/CdTe): PL intensity at 405 nm is ~200% of that at 633 nm, consistent with reduced front-interface recombination. • TRPL lifetimes (target vs control): τ1 1.5 ns vs 0.36 ns; τ2 94 ns vs 4.8 ns; τ3 518 ns vs 323 ns; average lifetime 450 ns vs 198 ns, showing marked reduction in fast (interface-related) recombination and improved bulk carrier survival. • PLQY-derived iVoc: 893 ± 12.2 mV (target) vs 858 ± 13.5 mV (control), ΔiVoc ≈ 35 mV. - Device performance (statistics, n=15): Target devices show clear improvements in Voc, FF, and PCE compared to controls, with enhanced reproducibility relative to ZMO-buffered devices owing to stable, conductive SnO₂. - Champion devices: Target PCE 20.03% with Voc 0.863 V, Jsc 29.2 mA cm^-2, FF 79.5%; Control PCE 18.3% with Voc 0.834 V, Jsc 28.9 mA cm^-2, FF 76.1%. EQE shows slight enhancement at short wavelengths for the target, consistent with reduced front-interface losses.

The findings validate the hypothesis that a photoactive bandgap gradient near the front junction can suppress nonradiative recombination and boost Voc in Cd(Se,Te) solar cells while using a commercial SnO₂ buffer. By incorporating oxygenated CdS and CdSe layers that fully intermix with CdTe during CdCl₂ treatment, a penternary Cd(O,S,Se,Te) region is formed within the same crystal structure, eliminating the detrimental heterointerface associated with conventional CdS/CdSe layering. The resulting type-I alignment and small conduction-band spike located inside the absorber, combined with a lower hole density in the graded region, induce downward band bending that promotes electron-hole separation and reduces front-interface recombination. Simulation and experiments coherently indicate lowered J_interf, longer carrier lifetimes (especially for fast, interface-related components), higher PLQY-derived iVoc, and improved device metrics. The approach achieves Voc values comparable to those obtained with ZMO buffers but avoids the drawbacks of low conductivity and moisture sensitivity, thereby enabling higher reproducibility with SnO₂. Collectively, the work demonstrates a practical path to 20%-class Cd(Se,Te) devices through bandgap engineering without sacrificing manufacturability or stability.

This work demonstrates a viable route to implement a beneficial bandgap gradient in Cd(Se,Te) thin-film solar cells by forming a photoactive Cd(O,S,Se,Te) region near the front junction via oxygenated CdS and CdSe precursor layers. The approach avoids forming a detrimental heterointerface, reduces front-interface nonradiative recombination, and yields substantial gains in Voc and overall device efficiency. A champion device achieves 20.03% efficiency with Voc 0.863 V, Jsc 29.2 mA cm^-2, and FF 79.5%, comparable to the best ZMO-buffered Cd(Se,Te) cells, while retaining the reproducibility and stability advantages of a commercial SnO₂ buffer. Potential future research directions include: optimizing the spatial profile and composition (O, S, Se, Te) of the graded region to tune the band offsets and spike magnitude; refining control over oxygen incorporation and intermixing dynamics during CdCl₂ treatment; investigating long-term operational and environmental stability; exploring dopant strategies and defect passivation within the graded region; and scaling to module-level processes while maintaining uniformity and performance.

- The SCAPS simulations assume identical crystal structure across graded layers and the absence of any detrimental heterointerfaces; real devices may deviate from idealized assumptions, potentially affecting recombination and band alignment. - Band offsets and doping levels used in simulations (e.g., ΔEc ~0.05 eV, hole densities 2×10^16 vs 5×10^16 cm^-3, fixed interface trap density 6×10^13 cm^-3) are model inputs that may vary in actual devices. - Experimental estimation of the Cd(O,S,Se,Te) bandgap relies on measurements of O-free Cd(S,Se,Te) alloys due to deposition tool limitations; oxygen’s effect on Eg is inferred rather than directly measured for the exact penternary composition. - The oxygen concentration profile is high near the front and decays over ~0.5 μm; precise control and reproducibility of this profile across large-area manufacturing were not addressed in detail.