WE HEREBY RECOMMEND THAT THE DISSERTATION PREPARED UNDER OUR SUPERVISION BY MARKUS GLOECKLER ENTITLED DEVICE PHYSICS OF Cu(In,Ga)Se2 THIN-FILM SOLAR CELLS BE ACCEPTED AS FULFILLING IN PART REQUIREMENTS FOR THE DEGREE OF DOCTOR

OF DISSERTATION DEVICE PHYSICS OF Cu(In,Ga)Se2 THIN-FILM SOLAR CELLS Thin-film solar cells have the potential to be an important contributor to the global energy demand by the mid-21st-century. Cu(In,Ga)Se2 (CIGS) solar cells, which have achieved laboratory efficiencies close to 20%, are highly attractive, because their band gap is near the optimal value, their polycrystallinity is not significantly detrimental to their performance, and the broad choice of heterojunction partners available allows additional degrees of freedom for optimizing their performance. Although steady progress has been made for CIGS solar cells, this progress has largely been driven by empirical optimization rather than by in-depth understanding of appropriate physical models. This thesis is intended to fill some of the gaps that exist between state-of-the-art experimental solar cells and their device physics. The level of complexity involved is largely prohibitive to analytical treatment and, hence, numerical approaches are primarily utilized. The dominant topics for CIGS solar cells covered in this dissertation are (1) variation in the Ga/(Ga+In) stoichiometry (“grading”), (2) the formation of “good” heterojunctions, (3) photoconductive effects in window or buffer materials, (4) the apparently benign or even beneficial presence of grain-boundaries (GBs), including a discussion of charged GBs and the effects of Cu-depletion near GBs. This work establishes a baseline model for CIGS solar cells and, from this starting point, the device physics relating to these questions is discussed and principles are identified that apply to a broad range of devices. CIGS grading is shown to have only small potential to improve device performance. This conclusion conflicts with earlier studies, and it is shown that the difference arises in the evaluation of the grading benefit, in particular, the proper choice of the reference performance. Band-gap increases toward the front of the device are most likely detrimental, while band-gap increases toward the back can be modestly beneficial. The popular “double grading” approach achieves only very small additional gains over the simple back grading approach. Very thin-absorber cells can benefit substantially from back grading, because in this case, grading can mitigate detrimental back-contact surface recombination. iii The window/absorber interface is studied and, in good agreement with experiments, a limitation of the open-circuit voltage is observed for wide-band-gap CdS/CIGS solar cells. This limitation can be circumvented by a change in window or buffer material to establish a positive conduction band-offset, a down-shift of the valence-band at the interface by Cudepletion, or appropriate pinning of the Fermi level at the interface. All these mechanisms deplete the interface of hole carriers and lead to lower interface recombination. Photoconductivity in CdS buffer layers has been thoroughly studied in the past, but approaches were mostly phenomenological and a quantified description of the transport mechanisms involved has been missing. This topic is revisited, a model that reproduces experimental results is established, and the details of carrier transport are discussed. Grain boundaries (GBs) in CIGS solar cells have not previously been investigated with respect to their influence on device performance. In particular, charging of GBs and a possible depletion of Cu at grain surfaces may be critically important for high performance polycrystalline CIGS cells. Based on first-principles calculations and experiments, Cu depletion shifts the valence-band edge downward, enlarging the band-gap energy. A two-dimensional model allows the study of recombination at GBs, charged defects at GBs, and valence-band offsets at GBs. It is found that (1) the GB recombination velocity would need to be less than 104 cm/s to allow for the observed record-level efficiencies, (2) small electron-attractive GB potentials lower both photocurrent and photovoltage, (3) large electron-attractive GB potentials increase photocurrent, but lower photovoltage with a net decrease in device efficiency, (4) small hole-attractive GB potentials are somewhat beneficial, (5) valence-band offsets reduce GB recombination and result in performances comparable to GB-free cells. Analytical approaches are used to solidify the understanding of the obtained results. It is concluded that GBs in CIGS solar cells are unlikely to be strongly charged and that record-level efficiencies require the inherent absence of GB recombination centers or an effective passivation such as that established by a valence-band offset at GBs. Limits and approximations of the GB models are discussed. Markus Gloeckler Physics Department Colorado State University Fort Collins, CO 80523 Summer 2005 iv