Mechanisms for long carrier lifetime in Cd(Se)Te double heterostructures

CdTe-based semiconductors are used in detectors, imaging, and other applications. In the past decade, CdTe photovoltaics costs have declined significantly, making this technology one of the most economical approaches to add new electricity generation to the grid. In fact, CdTe photovoltaics supplied ∼40% of the 2019 U.S. utility photovoltaics market, and the levelized cost of electricity is generally less than traditional power generating resources like nuclear, coal, and natural gas.^1,2 Due to direct bandgaps and high absorption, most thin-film photovoltaic materials only require a few micrometers of semiconductor material. This enables implementing durable and inexpensive substrates like ultrathin glass, metal, and plastic for diverse lightweight and flexible applications.^3–8 CdTe solar panels formed on conventional glass already are competitive in cost and performance with Si technology for conventional applications,^2,9,10 and relative to silicon, have significant headroom available for improvement toward theoretical maximum efficiency and even lower LCOE. The latter is enabled by low-cost substrates and high throughput deposition, which also creates challenges. The CdSeTe/CdTe absorber layer is deposited on lattice-mismatched amorphous or nanocrystalline emitter layers,¹¹ which affect critical absorber layer properties including grain boundary (GB), interfacial, and bulk recombination. A CdCl₂, MgCl₂, or similar treatment is necessary to significantly reduce recombination in thin polycrystalline films,^12–14 but this process can also limit hole density due to carrier compensation, which in turn affects device performance.^15–18 Overcoming these challenges can open pathways to improve CdTe technology.^19–21 

Incorporating selenium and expanding grain size have increased bulk lifetimes of CdTe-based devices,^22,23 but typical CdTe device lifetimes still range from several ns to tens of ns, while CuIn_xGa_(1-x)Se₂ (CIGS), perovskites, GaAs, and Si show lifetimes of hundreds of ns and greater.^24–27 Reaching lifetimes of hundreds of ns is important for CdTe solar technology because resulting high diffusion lengths enable more minority carriers to reach the back of the device, which can make back surface fields and passivation efforts more effective in improving performance. Recent computational analysis indicates that when the bulk lifetime is less than 10 ns, or similarly there is significant front interface recombination, there is no significant advantage to passivating the back surface or forming electron reflectors. On the other hand, for low front interface recombination velocity and bulk lifetimes exceeding 100 ns, efficiencies as high as 28% are possible provided other device properties are also optimal.²¹ Longer lifetimes coupled with passivated back contacts can also enable CdTe bifacial and interdigitated back contact (IBC) technologies.^10,28

To realize long carrier lifetime in practice, it is important to understand recombination in test structures that simplify analysis by reducing charge separation, diffusion, and other effects caused by the p–n junction in regular devices.²⁹ While this approach has been used often by the III–V community,²⁵ effective double heterostructures (DHs) with long lifetimes and diffusion lengths have only recently been established for polycrystalline CdSeTe absorbers.^30–32 Here, DHs made with CdTe or CdSeTe placed between alumina (Al₂O₃) passivation layers were fabricated to examine the material properties that contribute to long lifetime. The CdSeTe DHs demonstrate bulk lifetimes reaching values exceeding 1 μs, yet for CdTe DHs, relatively modest lifetimes of tens of ns are observed. Detailed characterization indicates that interface, grain boundary, and intragrain attributes contribute to these differences.

A 100-nm alumina layer was deposited by electron-beam evaporation onto clean 7059 glass. A 5-μm thick CdSeTe or CdTe layer was then deposited by thermal evaporation. These films were subsequently annealed in a close-spaced sublimation (CSS) arrangement in which the film was suspended over a graphite pocket containing a powder of CdSe_0.4Te_0.6 for 15 min in 100 Torr He. Films were nominally heated to 550 °C while the powders were heated to 545 °C. It was observed that this anneal reduced CdCl₂-related discolorations at the front interface (high-temperature anneals can lead to considerable penetration of CdCl₂ to the front interface which can be observed as a hazy discoloration through the glass) and allowed the use of higher temperatures (>450 °C) during subsequent CdCl₂ treatments. Higher temperature CdCl₂ anneals are preferred since these generally increase intra-grain and GB passivation in CdTe-based pX films. The net-result is that higher CdCl₂ anneal temperatures without delamination should result in higher carrier lifetimes. It was observed, however, that the Se anneal (CdSe_0.4Te_0.6 powder anneal) made lifetimes worse for the CdTe DHs relative to those that were not annealed, and therefore, CdTe DHs were fabricated both with and without the Se anneal. Note that the results pertaining to CdTe DHs in this work have not had the Se anneal for optimum lifetimes.

All films were then subjected to a high-temperature vapor CdCl₂ treatment at 500 °C in a similar CSS chamber for 10 min in a He ambient at 400 Torr. The substrate temperature was held 5 °C above the CdCl₂ source temperature to minimize CdCl₂ supersaturation and possible condensation. Any CdCl₂ residue was subsequently removed using a 15-s methanol rinse prior to depositing the second alumina layer (20 nm) to complete the double heterostructures as illustrated in Fig. 1.

Initially, different Al₂O₃ growth approaches were evaluated including reactive sputtering and evaporation; however, we did not observe a significant effect on the DH performance. Results shown in this paper used electron-beam evaporated Al₂O₃ deposited at a rate of 1-nm/s.

Two-photon excitation time-resolved photoluminescence (2PE-TRPL) was measured with pulsed laser excitation using a wavelength of 1120 nm incident through the glass. The laser light was focused within the CdSeTe or CdTe layer to create two-photon excitation of electron–hole pairs in the absorber with an approximate beam diameter of 60 μm. The repetition rate and pulse length were 1.1 MHz and 0.3 ps, respectively. The luminescence was collected through a 44-nm bandpass filter centered at 819 nm. Photoluminescence decay curves were generated using time-correlated single photon counting, and the lifetimes were determined from single-exponential fits to TRPL decay curves.³³ 

Cathodoluminescence (CL) spectrum imaging was measured at room temperature with a JEOL 7600F scanning electron microscope equipped with a Horiba H-cathodoluminescence universal extension system. The electron-beam conditions were 7.5-kV accelerating voltage and about 3-nA current. Luminescence spectra were recorded at each image pixel using an iHR320 spectrometer with a Syncerity charge-coupled device (CCD) detector and a grating with 300 grooves mm⁻¹, blazed to 600 nm. Surface preparation for CL included ion milling with an Ar⁺ ion beam at a glancing angle using a JEOL Cross-Section Polisher operating at 3 kV. No alumina remained after ion milling.

Scanning electron microscopy (SEM) images were obtained by a FEI Nova 630 NanoSEM. To avoid shading effects due to the roughness of the surface and cross section, the samples were ion milled in a JEOL cross section polisher.

X-ray photoelectron spectroscopy (XPS) measurements were conducted in a custom designed cluster tool described in Ref. 34. Using an epoxy (H2OE EPO-TEK) compatible with ultra-high vacuum (UHV), devices were affixed to clean pieces of aluminum foil and allowed to cure for 16 h at 65 °C. In an argon-filled glovebox attached to a cluster tool, epoxied samples were immersed in a bath of liquid nitrogen where they spontaneously cleaved at the CdTe/alumina interface closest to the glass substrate. For each sample, both pieces were then withdrawn from the liquid nitrogen bath into a stream of dry argon until they reached room temperature. Samples were loaded into the UHV transport system and moved to the XPS system such that air exposure of the cleaved surfaces was avoided. Data were processed with PHI MultiPak v9.6 using standard elemental sensitivity factors. High-resolution XPS spectra were taken at a normal takeoff angle using monochromatic Al Kα radiation with a 11.75-eV pass energy and in angle-integrating lens mode (±7°). An electron beam was used for charge neutralization on samples in which the alumina prevented good electrical contact to the spectrometer. Surface potential differences resulting from these experimental differences required post-acquisition alignment of peaks to the Te²⁻ 3d_5/2 peak, which in conducting samples is at 572.4 eV.

Figure 2 shows 2PE-TRPL curves for CdTe and CdSe_xTe_1−x (x = 0.1, 0.2, and 0.3) DH structures passivated with alumina (as shown in Fig. 1). The inset on the top right corner shows the measured lifetimes for these different Se compositions, which agrees well with data from DHs made by different methods in Ref. 30 (red diamonds). Figure 2 illustrates that the 2PE lifetime increases monotonically with Se percentage, reaching 800 ns and 1200 ns for x = 0.2 and 0.3, respectively. The CdTe lifetime is significantly less despite similar passivation.³⁰ The high lifetimes observed in DHs containing Se translate to diffusion lengths several times greater than the DH thickness [e.g., L ∼14 μm (Ref. 32) for x = 0.2] implying low recombination velocities at the alumina/film interfaces. Indeed, the alumina/CdSeTe interface recombination velocity for the lifetimes reported can be less than 100 cm/s.^14,31

CL and XPS measurements were performed to understand the recombination differences. Figure 3 shows room-temperature CL integrated intensity images taken on the back surface of the CdTe and CdSeTe DHs. Large grains on the order of 5–10 μm are seen after the high-temperature CdCl₂ treatment, mixed with scattered smaller grains on the order of 1–2 μm for both CdTe and CdSeTe. The CL images illustrate that the CdTe intragrain regions appear more adversely affected by a range of extended defects. The CdSeTe intragrain region luminescence is generally more homogenous; however, some defect clusters are still clear. Figure 4 correlates the CL and backscatter electron images. The dark defective CdSeTe regions correspond to voids or precipitates formed during CdSeTe deposition and/or during the high-temperature CdCl₂ treatments. Outside these regions, the overall CdSeTe intragrain quality is superior to that of CdTe.

To obtain the average CL profiles for CdTe and CdSeTe DHs shown in Fig. 5, the grain-boundary (GB) pixels were located using a watershed segmentation algorithm, and the CL intensity was averaged for all GB pixels. Then, the number of data points included around the GB were successively dilated, and the intensity was averaged (see the supplementary material). The profile distance is approximate and corresponds to the pixel length times the number of dilation steps. As seen in Fig. 5, the relative decrease in the luminescence intensity at the CdSeTe grain boundaries, 10%–12%, is significantly less than at the CdTe grain boundaries, 30%–35%.

If GB recombination were identical in both films, one would expect the CdSeTe films to have higher GB contrast because the intragrain recombination is less in CdSeTe. Here, the converse is the case, indicating that the CdSeTe intragrain and grain-boundary recombination are both significantly less than CdTe samples.

Thermomechanical cleaving (described in the methods section) was applied to examine interfacial chemistry at the scale of nanometers by XPS. Samples spontaneously cleaved at the alumina/CdSeTe or alumina/CdTe front interface (closer to the glass) rather than the outer CdTe/alumina interface. The surfaces were not exposed to air, polishing, ion milling, or other damaging sample preparation techniques that could potentially alter the surface chemistry of the front interface. XPS on cleaved CdSeTe DHs shows high fractions of Te⁴⁺ species at the front (alumina side) interface [Fig. 6(a)]. Te⁴⁺ is associated with oxidized Te sites and the data are consistent with an oxide layer such as CdTeO₃. Overall, merging signals from both the alumina and CdSeTe sides, it can be seen that the [Te⁴⁺]/[Te²⁻ + Te⁴⁺] ratio for CdSeTe is 0.12 for sample 1 and 0.07 for sample 2. In contrast, as shown in Fig. 6(b), there is no detectable amount of oxidized Te species at the glass/alumina interface in selenium-free CdTe DHs. Thus, similar to the transparent conducting oxide (TCO) interface of actual devices, the presence of Te⁴⁺ is correlated with better passivation and less interface recombination.³⁵ One possible explanation for the passivating effect of oxygen is that a Cd–O cation dangling bond is more ionic than a Cd–Te dangling bond closer to the band edge, and therefore, not an active recombination center. A similar effect occurs with zinc substitution of Cd.³⁶ 

CdTe has long suffered from poor lifetimes relative to other photovoltaic materials like GaAs, perovskites, and Si. In this work, we observe lifetimes of 800 ns in CdSeTe DHs, whereas the lifetimes in CdTe DHs remain limited to 20 ns. The long lifetimes observed in CdSeTe can be attributed to a combination of several mechanisms. While both CdTe and CdSeTe have large grains of about 5–10 μm after similar CdCl₂ treatments, the intragrain quality of CdSeTe is significantly better than CdTe. The CdTe has less CL intensity and greater defect concentration throughout, while CdSeTe shows generally greater homogeneity and intragrain quality albeit with some scattered defect clusters that are potentially associated with film voids. CdSeTe also shows extremely low GB contrast ranging from 10% to 12%, which is superior to 30%–35% observed in CdTe. Additionally, XPS shows CdSeTe interfaces form Te⁴⁺, whereas CdTe interfaces do not. Hence, a combination of intragrain, GB, and interface attributes contribute to CdSeTe lifetimes approaching 1 μs. The results help illustrate the characteristics necessary to achieve similar lifetimes in solar cells. While the increased lifetime in and of itself can greatly increase performance, front interface and bulk passivation can also enhance performance gains from back surface reflectors and passivation, providing pathways to high efficiency and new designs like interdigitated back contacts and bifacial CdTe solar technology.

See the supplementary material which illustrates how the GB contrast was estimated by dilating the number of data points included around the GB.

Techniques for fabricating double-heterostructures discussed in this work were developed through funding by the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy (EERE) under the Solar Energy Technologies Office (SETO) Contract No. DE-EE0008552, Agreement Nos. 34345 and 34353. This work was authored in part by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government.