Jonas Lähnemann
home  |  physics  |  photo-gallery  |  kenya  |  israel  |  flevo-bike

Diplomarbeit (diploma thesis) on
Spectrally Resolved Current Losses in Cu(In,Ga)Se2 Thin-film Solar Cells (2007/2008)

Download pdf-version of my thesis.
Download pdf ready to print as A5 booklet on A4 (duplex, flip on short-edge)


1 Introduction

2 Cu(In,Ga)Se2 Solar Cells
    2.1 The Photovoltaic Principle
        2.1.1 p-n Junctions
        2.1.2 Semiconductor Solar Cells
    2.2 Cu(In,Ga)Se2 Thin-film Photovoltaics
        2.2.1 CIGS Material Properties
        2.2.2 Deposition
        2.2.3 The Window: Buffer Layer and TCO

3 Solar Cell Characterization
    3.1 Current-Voltage Characteristics
        3.1.1 Experimental Setup
    3.2 Quantum Efficiency
        3.2.1 Current Loss Mechanisms in Cu(In,Ga)Se2 Solar Cells
        3.2.2 Internal versus External Quantum Efficiency
        3.2.3 Recombination in Cu(In,Ga)Se2
        3.2.4 Voltage Bias
        3.2.5 Bias Illumination
        3.2.6 Analytical Modelling of Quantum Efficiencies
        3.2.7 Derivation of Parameters from QE Measurements

4 Quantum Efficiency Setup
    4.1 Measurement Principle
    4.2 Setup Details
        4.2.1 Optics
        4.2.2 Electronics
        4.2.3 System Summary
    4.3 Discussion of System Precision
        4.3.1 Systematic Error Sources
        4.3.2 Random Error
        4.3.3 Reproducibility
        4.3.4 Intercomparison
        4.3.5 Summary of QE Precision
    4.4 Continuous Illumination QE on Dye-sensitized Solar Cells

5 Effects of Varying Cu Supply in Three-stage Absorber Growth
    5.1 Growth Specifics
    5.2 Surface and Bulk Analysis
        5.2.1 Absorber Surface
        5.2.2 Absorber Bulk Composition and Growth Model
        5.2.3 Summary of Surface and Material Characteristics
    5.3 Photoelectric Characterization
        5.3.1 Current-Voltage Characteristics
    5.4 Quantum Efficiency
        5.4.1 Quantitative Analysis of Current Density Variations
        5.4.2 Series Resistance Deduced from QE
    5.5 Another View on Interface Formation and Recombination
    5.6 Conclusions

6 Summary

A Thin-film Material Characterization
    A.1 Scanning Electron Microscopy (SEM)
    A.2 Atomic Force Microscopy (AFM)
    A.3 Energy Dispersive X-ray Spectroscopy (EDX)
    A.4 X-ray Diffraction (XRD)
    A.5 Secondary Ion Mass Spectrometry (SIMS)


List of Figures

List of Tables


Among the most imminent challenges facing our modern society are the increasing global energy demand, dwindling of the conventional resources used for power generation and foremost the consequences coming along with an anthropogenic change in our earth's climate [1]. Nuclear power is not a solution, owing to the dangers from uranium mining, waste disposal and weapons proliferation, but also to limited resources. A key role in finding solutions to these problems, hence, comes to renewable energies and among them the use of solar energy [1, 2]. The sun delivers 162,000 terawatts (TW) in sunlight to the earth [3], about half of which reaches the earth's surface - in moderate climates we have about 1000W/m^2 irradiation on a clear summer day. This can be harnessed for thermal applications or converted to electricity by concentrating solar power [4] and photovoltaics (PV).

The photoelectric effect was first observed by Alexandre-Edmond Becquerel in 1839 and theoretically explained by Albert Einstein in 1905. With the development of semiconductor p-n-junctions during the 1940s and their theoretical description by William B. Shockley in 1950, the fundamentals of modern semiconductor electronics, among them transistors and photovoltaic cells, were given. This quickly led to the first successful silicon solar cells being built at the Bell laboratories. They were also the first to work on CuInSe_2 as material for solar cells in the early 1970s, while the application in thin-films was pioneered by Kazmerski et al. [5]. For a long time photovoltaics was driven mostly by space applications and remote powering needs, however, the last decade has seen an explosion of the market, with growth rates exceeding 30 %, driven by increased installation of grid integrated systems pioneered by Germany and Japan [6]. While silicon wafer technology has so far been the workhorse of the PV industry, thin-film solar cells based on heterojunctions of direct bandgap semiconductors have become commercially available in the last years and are projected to get increasing market shares in the coming years [7]. They are based on either Cu(In,Ga)Se_2 or CdTe absorbers, the former so far having shown superior record efficiencies up to 19.9 % [8] and module efficiencies up to 13.4 % [9]. In comparison to silicon technology, reduced cell thickness and the need for less pure, polycrystalline material, as well as direct fabrication of modules, are among the reasons to expect a price advantage for thin-film modules with increasing production [10]. Production of thin-film solar cells faces certain restrictions in terms of metal resources, esp. indium; nevertheless there is still a huge growth potential [11]. In view of the complex nature of polycrystalline compound semiconductors in heterojunctions, progress in the past was often based on empirical work and hence there is still need to deepen the understanding of device physics [10].

Quantum efficiency measurements are among the fundamental characterization techniques for solar cells; probably the most important after current-voltage analysis of the diode characteristics. Through spectrally resolving the current yield of a photovoltaic device, quantum efficiency gives a closer view on the short-circuit current - one of the basic parameters of a solar cell - and loss mechanisms restricting it. It helps understanding the physics of current generation, recombination and carrier diffusion mechanisms. Hence, quantum efficiency is a valuable tool for scientists in this field.

In the scope of this report a quantum efficiency system was drafted and set up at the Université de Nantes to complement the characterizational possibilities of the research groups in Cu(In,Ga)Se_2 thin-film photovoltaics and electrochemical dye-sensitized solar cells. This is presented and discussed in chapter 4. An application of the system in a study on the correlation of Cu(In,Ga)Se_2 absorber morphology and device performance resulting from varied duration of the Cu-rich interval during isothermal three-stage co-evaporation of the absorber is the content of chapter 5. Fundamentals on Cu(In,Ga)Se_2 solar cells are given in chapter 2. Aspects of their characterization are discussed in chapter 3 with a focus on quantum efficiency measurements. Appendix A introduces the techniques from material physics used in this work.

This report has given details on the quantum efficiency system developed and subsequently assembled for the photovoltaics research at the Université de Nantes. Such a system allows to spectrally resolve the photocurrent and analyze loss mechanisms limiting the current yield. The research in Nantes is focussed on thin-film solar cells with Cu(In,Ga)Se_2 absorber; in this framework, the QE system has been applied in a study on the effects of varying the duration and magnitude of temporary Cu-richness during a three-stage co-evaporation deposition. CIGS solar cells are a technology just entering the market, but there are still many pathways to improve their performance and since much of the past progress has been achieved empirically, a better understanding of the device physics is needed. The contribution of this work is twofold. On the one hand, the instrument built will facilitate the research of other scientists in this field and possibly even serve as model for other setups. On the other hand, an analysis and discussion on the impact of changing a specific deposition parameter was carried out with the help of the QE and other characterizational techniques.

Basics of the photovoltaic generation of electricity in a p-n junction have been presented in chapter 2; particularly, Cu(In,Ga)Se_2, a I-III-VI_2 compound semiconductor, as absorber material for heterojunction thin-film solar cells has been discussed. Through asymmetric doping and choice of bandgaps the junction partner ZnO:Al serves as window, while most absorption is in the CIGS. The device performance is improved through the introduction of buffer layers, giving a layered structure of glass/Mo/CIGS/CdS/i-ZnO/ZnO:Al for these solar cells. Absorbers for cells presented in this report were grown using co-evaporation (multi-source physical vapor deposition) following a three-stage In+Ga+Se/Cu+Se/In+Ga+Se procedure at constant substrate temperature.

As essential measurements towards understanding device physics and improving the performance of solar cells, the photoelectric characterization by recording current-voltage and quantum efficiency (QE) curves have been introduced in chapter 3. The focus of this report is on the latter, defined as spectrally resolved ratio of electrons collected from the device per incident photons. Technically, the photocurrent under monochromatic light is recorded during a spectral scan across the absorbed wavelengths. The knowledge of loss mechanisms limiting the QE allows conclusions on their influence on the device physics and the current yield. For CIGS thin-film cells, these losses include grid shading (in laboratory cells), front surface reflection, ZnO absorption above the bandgap and by free carriers, buffer absorption, incomplete absorption of long wavelengths in the CIGS, and finally incomplete collection of charge carriers from recombination of electrons and holes. Due to the large number of defects, Shockley-Read-Hall recombination in the absorber bulk or at the interface is dominant. A possibility to further investigate the collection is using QE with a bias voltage applied on the device. Reverse bias increases the space charge region and should improve collection if significant recombination occurs under unbiased conditions. Another variation is the superposition of continuous white light on the periodically modulated monochromatic light of lower intensity; this is termed light bias and allows probing under more realistic operating conditions. Both voltage and light bias have been discussed in view of device operation and particularly buffer photoconductivity. Some basics on the analytical description of the QE have been given; they allow the simulation of QE curves or the derivation of parameters from fits to the QE curve. Furthermore, procedures to derive the bandgap and the short-circuit current density from QE data have been described.

A quantum efficiency system has been designed to fit the requirements of the research groups in Nantes. Its setup has been described in chapter 4. The measurement procedure has been completely automated and is governed from a PC. The choice was to built a grating monochromator based dual-beam QE system, where all measurements are conducted in relation to a reference cell measured at the same time as the device under test or calibration cell. This corrects for fluctuations in the lamp intensity and reduces errors introduced by the spikes in the Xenon lamp spectrum. The quantum efficiency of a test device is deduced from that of a calibrated photovoltaic detector measured earlier in the same lightpath. For CIGS solar cells, the system operates in chopped light mode with lock-in detection. However, a steady-state mode (continuous illumination) using digital multimeters for detection is provided in order to allow QE measurements of dyesensitized solar cells. The use of a bifurcated fiber bundle as light-splitter allows great flexibility and easy use, but comes at the price of a significant (~ 80 %) loss in light intensity mostly resulting from the light coupling into the fiber. Nevertheless, the system's monochromatic illuminating power on the cells, that on average is 8 μW, has proven sufficient - unlinearities in all tested cells had saturated at the given illumination level. Straylight from the monochromator is negligible for the complete measurement range of 340-1360 nm. Options for light bias and voltage bias in chopped light mode were successfully implemented. Whilst the light bias is provided through an additional halogen lamp directly focussed on the cells, the supply for the voltage bias is included in the current to voltage preamplifier.

Systematic and random error sources for the instrument have been discussed. Among the systematic errors, that of the QE of the calibrated detector used in the calculation is the most dominant - stated by Hamamatsu at 5 %. The noise in the system was investigated and found to lie below 0.5 %. In repeated measurements on the same cells, the JSC calculated from the QE was found to vary by up to 2 % over a period of two months; this can be taken as an indicator for random errors. Repeatability within a measurement session showed much better agreement. Intercomparison with the National Renewable Energy Laboratory (NREL) in Boulder, Colorado, showed a deviation of only 1 %. Further comparisons were done to measurements from the Angstrom Solar Center in Uppsala and the Helmholtz Zentrum Berlin, falling within 3 % of the results from Nantes; this is an indicator for systematic errors. The shapes of the curves are also well reproduced between the laboratories. None of the systems seems to be particularly biased, as the deviation is almost equal to the reproducibility of measurements in Nantes. On the basis of the good comparability and reproducibility, the precision of the system is estimated at 5 %. The error for dye-sensitized solar cells lies somewhat above this, due to less spatial homogeneity of these cells. In chapter 5, Cu(In,Ga)Se_2 solar cells have been investigated that were grown from an isothermal three-stage co-evaporation process. The transitions between Cu-poor and Cu-rich stages has been monitored by end point detection (EPD), based on the sample emissivity detected via the heater output power necessary to keep the substrate temperature constant. The duration of the 2nd stage has been varied and, thus, also the maximum Cu/(In+Ga) ratio before passing back to an overall Cu-poor composition. From SEM and AFM micrographs this can be correlated to changes in surface morphology: Longer Cu deposition leads to greater surface roughness (root-mean-square roughness), larger crystal grains and deeper crevices between the grains. An increase of the final EPD signal, i.e. emissivity, for the most Cu-rich samples can be correlated to this increase in surface area. Variations seen in EDX and XRD measurements on the absorbers are not large enough to expect a strong influence on the device performance and don't point to a correlation with the change in Cu supply. From SIMS depth profiles and SEM data, together with growth models for three-stage co-evaporation, we can conclude a two section growth of the absorber. The first section, which is unaffected by the varied Cu/(In+Ga) ratio at the end of the 2nd stage, results from the interdiffusion of Cu, In and Ga between the Ga_xIn_ySe precursor from the 1st stage and Cu_xSe deposited in the 2nd stage. Such an interdiffusion leads to Ga gradients; these are visible in our SIMS profiles. After the film evolves Cu-rich, the excess Cu_xSe deposited is the precursor for the upper section of the CIGS formed through metal interdiffusion in the 3rd stage, when again In and Ga are evaporated in place of Cu. Such a conversion of Cu_xSe to CIGS is known to lead to the observed large grained morphology with deep crevices.

Analyzing the solar cells produced from these absorbers, high efficiencies are obtained as long as the film did not evolve too Cu-rich during the 2nd stage (max. Cu/(In+Ga) <= 1.5). The change in surface morphology appears to be the most important factor influencing the efficiency of resulting photovoltaic devices: an increase in surface roughness has a positive effect on the JSC, primarily by reducing interference - quantitatively a little more than 1 mA/cm^2. This increase is partially balanced by a slight decrease in open-circuit voltage and fill factor. For Cu/(In,Ga) > 1.5, all device parameters (FF, VOC, JSC) deteriorate strongly. The reduced VOC cannot be explained from the increased junction area, a relation which can be inferred from the diode model. The reduction of current density by up to 5 mA/cm^2 can clearly be attributed to collection losses when looking at the QE under reverse voltage bias. For the Cu-richest growth conditions, an additional reduction of the long-wavelength QE related to CdS photoconductivity is observed; measured as JSC, this amounts to 1 mA/cm^2. The collection losses are the result of a narrower space charge region and increased interface recombination, both possibly related to the crevices between crystal grains. This idea is confirmed by cross-sectional micrographs of the roughest sample; they show voids in the solar cells, i.e. the ZnO only covers but does not enter the crevices. It could be interesting to investigate if such a reduction in device performance is localized around the crevices, which the proper junction formation seen in the micrographs for the rest of the cell suggests. Additionally, other possible losses in the QE have been discussed and quantified. From measurements under forward bias, a value for the series resistance has been calculated; this disagrees with values from fitting to the current-voltage curve, which, however, is not surprising when considering the completely different illumination conditions.

The quantum efficiency system drafted and built during this diploma work to spectrally resolve current losses in thin-film solar cells has proven to give reliable results, both in various tests and in the application during the specific study on Cu(In,Ga)Se_2 solar cells presented in this report. This study has, furthermore, shown that when going too Cu-rich in the second stage of an isothermal threestage co-evaporation process, device performance is impaired by crevices formed between the absorber grains.

Univ. de Nantes
Institut des Materiaux Jean Rouxel Nantes (IMN)
Helmholtz Zentrum Berlin für Materialien und Energie

back to physics
Copyleft 2000-2017 by Jonas Lähnemann
Impressum / legal notice

Creative Commons License
unless stated otherwise