Laser annealing in thin-film chalcogenide photovoltaics
Thin-film photovoltaic (TFPV) solar cells have become a viable alternative to established crystalline silicon technologies in terms of both cost and efficiency. However, to compete with conventional electricity generation, manufacturing costs need to be further reduced. One of the bottlenecks in cost is thermal annealing of the semiconductor light-absorbing layer. The two leading chalcogenide TFPV technologies are based on polycrystalline copper indium gallium diselenide (CIGSe) or cadmium telluride (CdTe) absorber layers, 2–4μm thick, with appropriate front and back contact layers deposited on a glass substrate. When furnace annealing is used to improve the structural properties of the film, it takes a huge thermal budget to heat both the glass substrate (a poor thermal conductor) and the absorber layer. This makes the process both time- and cost-intensive.
In recent years, our group has investigated the use of laser annealing in both CdTe and CIGSe. Laser annealing has the potential to increase throughput and thus reduce manufacturing costs compared with standard thermal processes, and may also unlock unique property regimes because of transient and non-equilibrium kinetics. Choosing the right wavelengths and dwell times of the incident laser makes it possible to selectively heat the TFPV device stack. This allows processing in a transient regime, which in turn enables heating to higher temperatures and thus shorter processing times than with conventional furnace annealing. In this article, we review some general principles and results from our collaborative work.1–9
For timescales longer than ∼1ns, heat will flow through the entire thin-film stack, but it may take ∼100s or more to establish a steady-state 1D temperature profile through a glass substrate. At <1ns timescales, light absorption and thus annealing can be targeted within a thin-film stack. This further reduces the thermal budget, which potentially can reduce cost and increase manufacturing throughput. Figure 1 shows the bandgaps (Egap) of common TFPV materials and lasers suitable for annealing the full >1m2 TFPV panel area. Process tuning can result in faster annealing kinetics at higher temperatures without overheating the substrate.
High-energy pulses from pulsed lasers can localize annealing to a near-surface region of the sample without affecting the bulk of the active layer. In fact, UV excimer annealing of transparent conductive oxides is commonplace in the thin-film display industry at >1m2 sizes: the majority of smartphone displays have been laser annealed. We exploited this capability of pulsed-laser annealing to demonstrate a dry back-contact treatment for CdTe solar cells.4, 6,10,11 CdTe has a high work function (5.9eV) and typically forms Schottky contacts to most metals. Devices treated with 100 pulses of a 25ns excimer laser (248nm) at 50mJ/cm2 at the back contact did not show a rollover effect (reduced current density on increasing voltage) in the current density vs. voltage curve as generally observed in devices with a non-ohmic back contact. Moreover, the specific contact resistivity decreased monotonically with fluence even after one pulse.4
Post-deposition thermal processing of CdTe/CdS (cadmium sulfide) TFPV stacks in the presence of chlorine is critical for high efficiency. But it takes many minutes, thus limiting throughput. We have been working to demonstrate ways to accelerate manufacturing, such as low-quality CdTe deposition followed by rapid laser annealing at higher-than-typical temperatures. Using a scanned continuous-wave (CW) Nd:YAG laser, we have demonstrated improvements in the structural quality of sputtered CdTe.12 Interestingly, 1064nm Nd:YAG is sub-bandgap in perfect CdTe, which thus should be weakly absorbing or transparent to it. However, defects and disorder in films can be sufficient to result in very strong optical absorption only slightly weaker than bandgap absorption. This allows for the possibility of self-limiting sub-bandgap annealing: as defects are removed by the annealing, the optical absorption of the laser light becomes weaker.12 We also have used the transmission through the cell stack to monitor the changes during annealing in situ.12 Lastly, even for this long-wavelength CW laser annealing, we found that the recombination velocity and Fermi level pinning at the CdTe surface were reduced by the removal of defect states at the surface.2 This led to back-surface passivation of the CdTe layer by laser treatment.
Another unique application is recrystallization and annealing of CIGSe precursor layers by CW laser annealing. We first attempted liquid-phase processing, but molten CIGSe de-wets on microsecond timescales.13, 14 Consequently, we moved to CW solid-phase annealing and demonstrated significant structural improvements for electro-deposited copper indium diselenide films with dwell times of only 1s. Optoelectronic improvements, including the first laser-annealed working solar cells, were made by maintaining high selenium overpressure.3, 5,7,8
While the work is far from completed, our results have begun to illuminate the possibilities and limitations in terms of process windows and phenomena for area annealing of CIGSe and CdTe TFPVs. The rise of multi-kilowatt-class direct-diode and fiber-coupled lasers will undoubtedly usher in new opportunities for annealing large areas with good lateral uniformity and short process times. Our published work demonstrates the potential of laser annealing to be a much faster process than conventional furnace annealing (seconds compared to minutes). This can lead to faster throughput and thereby reduce manufacturing costs. Our current efforts are aimed at further optimizing laser annealing, exploitation of fast quenching, and investigating nonthermal effects.
We thank our past and present collaborators, especially B. J. Simonds, C. Ferekides, H. J. Meadows, P. Dale, V. Kheraj, B. Van Devener, A. Bhatia, and V. Palekis.
Mike Scarpulla is an associate professor in the Materials Science department and the Engineering and Electrical and Computer Engineering department. He is an IEEE senior member and editor of the IEEE Journal of Photovoltaics. His lab specializes in light-based processing and characterization of compound semiconductors, especially for TFPVs.
Sudhajit Misra is a fourth-year PhD candidate in Mike Scarpulla's lab, specializing in laser processing of CdTe and related materials.