Promotor / Supervisor
Promoter: Prof. dr. Bart Vermang
Co-promoter: Prof. dr. Jessica de Wild
Samenvatting van het onderzoek / Summary of Research
In times in which renewable energy is more important than ever to tackle the global climate change crisis, solar energy is the most commonly available source of renewable energy. Among the readily available photovoltaic technologies, Cu(In,Ga)Se2 (CIGS) is one of the most advanced to help produce this renewable solar energy that we need. In order for the technology to stay competitive in the future it is necessary to either find new uses in niche areas, or to find ways to further reduce its production costs.
In both cases, the main way to make CIGS more appealing for large scale production is to reduce the use of critical raw materials, such as In and Ga. This can be done by either drastically reducing the thickness of the absorber material to use less material overall, or by simplifying its architecture and processing, by avoiding the use of a complex Ga grading. These adaptations promise reductions in production costs through lower material costs as well as a decreased thermal budget. However, these steps come with a series of significant drawbacks that need to be addressed. Reducing the thickness leads to absorption losses, and reducing the Ga grading severely affects the quality of the CIGS interfaces. Since efficiency has to be sufficiently high to warrant mass production, it is necessary to come up with ways to counteract these drawbacks. In addition, it is necessary for the solar cells to be produced using the updated process to satisfy common industry requirements in terms of reproducibility and reliability.
While various ways to solve each problem individually exist, a high performing solar cell ultimately relies on a sufficiently high quality absorber material to limit losses to the bare tolerable minimum, both in the bulk absorber and at the interfaces, and over the lifetime of the cell. Due to this, in the present contribution, we focus on different ways to improve the CIGS absorber layer, using different deposition methods and processing steps, and discuss the possibility of designing an ultra-thin CIGS absorber materials that would be both reproducible and reliable over extended periods of time in the field.
First, we performed a series of experiments on the two main deposition methods for CIGS, themal (co)- evaporation and sequential processing, to determine whether they can be efficiently used to create an ultra-thin absorber material. In combination with the investigation into the deposition methods, we tried various modifications to the standard processes in order to identify potential areas for improvements to the absorber quality, and increases in overall performance. Second, we performed accelerated lifetime tests on the solar cells to investigate whether the solar cells produced using the updated architecture can be made to align with industry requirements concerning field reliability. Finally, we developed and tested a novel measurement and analysis methodology, called bias dependent admittance spectroscopy (CVf ) mapping, using a combination of commonly used capacitive characterization techniques and electrical simulations, to help with the identification of defects and failure modes in our solar cell structure, and potentially prevent them in the future.
The work performed on the development of a baseline recipe for the (co)-evaporation of CIGS absorbers showed that it is possible to create recipes that lead to the growth of very reproducible absorber materials with predictable elemental gradients and phase formations. However, using a straightforward one-stage coevaporation process to do so inevitably leads to the formation of very small crystallites and an absorber material with a very high concentration of grain boundaries. To tackle this problem, we developed another two-stage coevaporation baseline recipe by adding an initial Cu-rich deposition stage. Similar to the one-stage process Ga and In crucible temperatures are kept constant during the whole deposition process to ensure the growth of a CIGS material with a flat band profile. The addition of a Cu-rich processing stage improved crystallite growth significantly. However the solar cell performance remained mostly unaffected when compared to the reference one-stage process. For the two-stage solar cells, this could be corrected by performing in-situ alkali post deposition treatments in an Se poor atmosphere. However, for the one-stage grown absorbers, these treatments failed to show any significant improvements, and mostly showed slight drops in performance. The predominant theory is that the Se-poor atmosphere during NaF evaporation leads to a loss of Se in the CIGS absorber which, in turn, leads to a higher defectivity of the interface region, and a poorer performance. This being said, the NaF treatment showed slight increases in performance in the case of the two-stage absorbers. This might be due to an improved interface with CdS, the advantages of which outweigh the disadvantages from the Se loss. Over the course of this project, it became apparent that, while it is possible to increase the performance of CIGS solar cells processed with a flat band gap grading, these solar cells will inevitably be limited by their high concentration of grain boundaries. Strategies to improve the solar cell performance of such material must, therefore, address the passivation of said grain boundaries. Development work on the sequential processing recipes for the annealing of metallic precursors lead to the observation that, while it is a promising route toward higher performance, it does not allow for the growth of an absorber layer without Ga-grading. In addition, it appeared that the absorbers produced through this method suffer from significant reproducibility issues, and their exact composition varies wildly from one experimental run to the next. Follow up experiments showed that this inhomogeneity and variability is likely due to an uneven diffusion of the different metallic precursors during annealing, leading to significant differences and gradients in the elemental profiles, both laterally and in depth. Investigations into the substrate temperature dependence of this type of processes highlighted a set of specific temperatures as the ideal candidates for higher performance. Interestingly, we were also able to highlight a dependency of said annealing temperature on the substrate orientation. Samples annealed with the precursor side facing towards the heating source experienced an, on average, 40 C higher temperature than when the precursors were annealed facing away from the heat source. This observation led us to conclude that, rather than depending on the annealing temperature directly, the performance of the solar cells depends on the elemental intermixing state within the absorber. This intermixing state is, in turn, dependent on the temperature experienced by the elements during the processing. While sequential processing is certainly a possibility for the growth of ultra-thin CIGS, this possibility is, however, strictly linked to the necessity to have a flawless process control to prevent excessive compositional variation and insure satisfactory reproducibility.
Using atom probe tomography, we studied the solar cell prototypes that underwent accelerated lifetime testing (ALT), and discovered that during extended exposure to humidity and heat, the absorber showed a pronounced presence of water related ions in the areas identified as grain boundaries. This indicated that water seeps into the absorber layer, using the grain boundaries between crystallites as a primary pathway. The performance results of related solar cells did not show significant changes to either current density or open circuit voltage. Rather, the performance of the ageing solar cells appeared to degrade mainly due to changes in the series and shunt resistances. This led us to the conclusion that the water seeping into the absorber material is not the main reason for the solar cell performance degradation, but that it is due to pronounced ageing in either the buffer or the transparent conductive layers, upon exposure to humidity. Follow-up degradation experiments excluding either of these layers from degradation indicated that while their degradation seems primordial to the degradation of the solar cell as a whole, it is also of uttermost importance to protect the various interfaces from the impact of humidity and related oxidation. This indicates that the TCO and CdS are the critical factors when it comes to solar cell performance losses upon moisture exposure. This might open up interesting possibilities when it comes to the recycling of aged CIGS solar cells and panels. Further research in this direction might unequivocally prove that the degradation of the CIGS with exposure to moisture is indeed negligible, and, therefore, that aged absorber layers could be reused upon restoration with fresh CdS and/or TCO. By performing extra analysis of our study samples with our newly developed CVf mapping technique, we determined that it is possible to identify specific defect types in the solar cell structure using their respective capacitive signature on the CVf maps. This allowed us to show that a commonly observed defective response on the CVf maps can be related to an apparent bulk defect. By comparing samples treated with KF PDT and ALT, we succeeded in identifying the grain boundaries as the probable main origin of this defect response. In addition we were able to highlight the importance of the sodium supply during absorber growth. It became apparent that an absorber layer grown in sodium deficient or excess conditions showed this significant bulk defect related response, while layers grown with a medium NaF supply did not. Moreover, by increasing the sodium supply, it becomes evident that the frontmost layers of the CIGS absorber are depleting of Cu and In, which leads to changes in the band alignment at the interface with CdS. This change in band alignment could in turn be observed on the CVf maps in the shape of an additional response domain at high frequency and forward bias. This result will become increasingly significant in years to come, as CIGS processing is slowly moving away from the standard Na rich substrates, and towards Na-free flexible solutions. A full transition towards a flexible substrate will require careful calibration of the Na amount provided to the CIGS material during its growth. In addition, it is highly likely that the exact amount required for an ideal performance will differ depending on the overall thickness of the deposited absorber.
Overall, the results obtained in this thesis highlight the potential power of the CVf mapping technique, and its wide field of potential applications. In combination with other commonly used characterization techniques, this methodology can highlight some critical knowledge and help with the root cause detection and prevention for poor solar cell performance. While our results in this work are limited to CIGS solar cells, nothing stands in the way of applying this approach to other types of materials and using the insight obtained to improve deposition processes for all types of solar cells.