by Katherine Derbyshire, Contributing Editor, Solid State Technology

For system users, photovoltaic efficiency is a single number showing what fraction of the photons impinging on the array emerges as usable current at the attached load. For scientists trying to build better devices, efficiency is a more complicated proposition. Absorbing light and generating free carriers isn’t sufficient. To perform useful work, the carriers must reach the cell electrodes. If the electron-hole pair recombines too quickly, then it can’t contribute to the photocurrent.

In crystalline silicon solar cells, getting carriers from the junction where they are generated to the cell electrodes is straightforward. At the junction, electron-hole pairs split, which each carrier traveling through the n-type or p-type material, as appropriate. That’s the way CIGS (copper indium gallium diselenide) solar cells are supposed to work, too. Though most carriers are excited in the CIGS layer, a CdS “window” layer actually defines the junction between n-type and p-type material.

Yet measurements on real cells reveal a more complex picture. On one hand, “good” cells occur over a broad range of CIGS compositions. On the other hand, most of the companies attempting to produce CIGS cells in commercial quantities have had trouble producing “good” cells consistently. Cells processed under seemingly identical conditions turn out to have widely varying properties.

Some researchers have attributed this variability to the complexity of the system, suggesting that the solution lies in more precise control of process conditions and film composition. In contrast, B.J. Stanbery, founder of HelioVolt Corp., argues that the problem, and its solution, derive from the basic structure of the material ^[1].

Though much of the work in this field has focused on the simpler CIS (gallium-free) material, gallium freely substitutes for indium, and the same conclusions apply. In the CIS system, the best cells lie in a two-phase region of the quaternary phase diagram. In this region, a CuInSe₂ α-phase and In-rich CuIn₃Se₅ β-phase coexist. The fraction of each phase varies with the overall film composition, but compositions of the individual phases do not.

Stanbery argues that the α-phase is a p-type conductor, the β-phase is n-type, and that the cell performance depends on the intra-absorber junction (IAJ) between the two. In this model, the ideal cell structure is one in which the two phases form two continuous interpenetrating networks. Carriers generated too far away from an intra-absorber junction will recombine before they can reach a transport layer. Without this interpenetrating structure, cell efficiency is likely limited to 14% or less, suggests John Langdon, VP of marketing at HelioVolt. (The US National Renewable Energy Laboratory has set a target of 15.2% efficiency by 2010.)

The IAJ model appears to explain several puzzling aspects of CIGS cell behavior. As long as the minority phase forms a continuous path for carriers, the relative fractions of the α- and β-phase, and thus the composition of the film, are relatively unimportant. On the other hand, the behavior of the film depends on nanoscale composition fluctuations — which drive phase segregation — and not on structural features such as grain boundaries. The α- and β-phases are crystallographically coherent, creating no defects at the interface between the two regions. Thus, a film with completely uniform grain structure can still fail to have the desired phase structure, and vice-versa.

In CIGS, as opposed to CIS, the presence of gallium can help drive creation of the desired nanostructure. Gallium tends to segregate to the α-phase, while indium segregates to the β-phase. However, the bandgap of each phase depends on its composition. If gallium concentration in the film as a whole or in a local area exceeds 35%, the band difference between the α- and β-phase regions will become too large for a useful junction.

Actually creating the nanoscale phase structure in CIGS films turns out to be complicated. Most processes, both vacuum-based (PVD, evaporation) and non-vacuum (powder methods), use a final thermal annealing step to react the metal (or oxide) precursors with selenium. Yet the high temperature required for this step allows diffusion and phase segregation. The film structure after annealing will not necessarily be the same as the structure created by the metal precursors. Given that the initial metal structure can be difficult to control (see Composition control the key to CIGS efficiency“), the widely varying results seen in practice are not surprising.

HelioVolt’s process coats two precursor “inks” onto two different substrates. Bringing them together and flash heating the sandwich forms a dense CIGS film with the desired nanostructure. Rapid heating minimizes diffusion, while separate preparation of two precursor films gives control over the structure. The company claims that its method gives superior cell efficiency, while the low thermal budget makes the process compatible with a wide range of substrates. — K.D.

[1] B.J. Stanbery, “The Intra-absorber Junction (IAJ) Model for the Device Physics of Copper Indium Selenide-Based Photovoltaics,” IEEE Photovoltaic Specialists Conference, pp 355 – 358 (2005).