Investigation of defects in crystalline silicon solar cells by confocal Raman spectroscopy

Introduction

This study is born out of the need to provide experimental contribution to the vibrational behaviour as well as the structural property of polycrystalline silicon solar cell (poly-Si). Unlike the theoretical investigation of poly-Si vibrational properties that has created great interest among researchers, over the past 40 years, ¹ little has been done on the experimental investigation. Majority of the theoretical studies that have been done were only concerned with the qualitative comparisons of the vibrational density of states of the structural models. ¹ In terms of quantitative investigation, there are a handful of papers that relate the structural to vibrational properties, but most of these studies are theoretical studies.^2–5 Having a good knowledge of the quantitative relationship between the structural and vibrational properties is very important. This is because it is useful for relating the structural properties and vibrational properties to the spectrum width (linewidth) of the spectrum transverse optic Raman peak, as well as the extent of disorderliness.^6–9 This is an effective approach in characterising the defect level of the sample when using confocal Raman microscope. The width of the band is influenced to a large extent by the coordination defects localised in such spot.^10,11 In this present work, Raman spectra were used to qualitatively and quantitatively diagnose poly-Si. The qualitative evaluation of the spectra linewidth of the transverse optic Raman peak was used to analyse the degree of defect and the stress. The basic images were used to investigate the structural properties of both the affected and the non-affected regions. Raman spectroscopy is a well-suitable method for photovoltaic (PV) analysis because of the nature of silicon–silicon bond. Silicon–silicon bonds are known to be symmetrical and produce strong Raman scattering which makes it easy to study any variations that might occur. In addition, crystalline silicon has highly ordered structure, bond angle and bond length; at a wavelength of 521 cm⁻¹, it exhibits a sharp Raman peak. The relevance of this study lies in the use of both qualitative and quantitative analysis to study the effect of degradation, to reveal the effect of degradation and its associative defects on the acquired Raman spectra. The technique gives a direct approach in addressing the issues of long-term degradation issues resulting from defect.

Research methodology

Three poly-crystalline solar cells of 156 mm × 156 mm dimensions with 2 busbar and 200 μm thickness were reverse biased by passing a 10 V power source for 30 min. At the end of 30 min, one of the cells was found to have degraded. Infrared (IR) images were taken and from them the hotspot regions were identified and marked out for further analysis.^9, Measurements were done by WITEC 300 R+ using 50× and 100× objectives lens with 785 nm excitation laser. The use of this high excitation laser influences the amount of penetration depth which the light can go inside the sample. The samples of interest were carefully prepared and cleaned with compressed air to eliminate the presence of any moisture and dust particles. Thereafter they were placed on the sample holder and illuminated with laser light. The WITEC Alpha 300 R+ Raman spectroscopy is equipped with WITEC KTS-NOOA 1504003 software, which helps with the analysis of data mapping. While a Zeiss EC Epiplan of focal magnification power of 20×/0.4 as well as EC Epiplan-Neofluar 100×/0.9 Dic focal power were used to focus the illuminating laser light on the sample. Two different light sources of wavelengths 785 nm and 532 nm were used to determine the trench hold wavelength for these samples. Therefore, the 532 nm wavelength which is the second light wavelength functions as a countercheck. From the two Raman spectra, the 785 nm wavelength was adjudged the best because of the resulting single spectra of distinctive intensities. Hence, only results from 785 nm light source are reported in this study. The 785 nm laser excitation gave penetration depth of about 100 nm to 1 μm . ¹² Due to the high thermal sensitivity of the Si material, the illuminating laser power voltage was kept constant at 0.5 mV throughout, and this helped to prevent heating and subsequent recrystallization of the analysed sample. The geometry of the grain and defects of the sample were analysed by scanning through the surface while spectral factors like the full width at half-maximum (FWHM) and the Raman peak frequency were evaluated by averaging the entire resulting images to give the basic analysis. The spectrometer of the device has an estimated frequency resolution of 2.3 cm⁻¹.

Results and discussion

Defect analysis

The large area scan presented in this section was done on the area indicated by the red rectangular box in Figure 1(a). The level of defects in the sample can be estimated from their Raman intensity of the Si peak. A high degree of defect decreases the Si peak in the case of deep impurities. The degree of defect in the samples has been studied from single Raman spectra acquired from combined images of both the affected and non-affected regions. The appearance of other peaks in the spectrum is an indication of the defect caused by elemental impurities. Under the same measurement condition, the single spectra acquired from pure samples are expected to have uniform intensities, but the presence of defect affects the intensity of a particular chemical bond, in this case Si–Si tetrahedral bond as shown in Figure 2.

OPEN IN VIEWER

Figure 1.

Large area scan of reverse biased poly-Si cell showing the: (a) stitching image. (b–e) Drawn images representing different phase distribution of chemical impurities. (f) Combined image of different distributions indicating the existence of different phases.

OPEN IN VIEWER

Figure 2.

Different spectra indicating the existence of five different phases at the surface of the sample; this gives the corresponding five colour codes.

The stitching image was used to identify the location of the defects, Figure 1(a), and the light red rectangular box region was further scanned so as to give a detailed information about the defect. Four different drawn images resulted from the scanned area and each of them is displayed in Figure 1(b) to (e). The difference in each image is displayed in their colour codes, which helps to identity the different elemental impurities, present in the defective area. The combination of the various colour codes resulted to the combined image presented in Figure 1(f). The data acquired from the various drawn images in Figure 1 provide the Raman spectra of the analysed sample; this is presented in Figure 2.

The data extracted from the combined image are analysed in Figure 2. Each of the colour code represents a particular spectrum presented in Figure 1(f). In Figure 2, it is clear that only the red spectrum contains Si peak, while the other spectrum contains a different element which differs from Si, and hence they are acting as impurities. It is important to note that only the sample with defective cluster is analysed in Figure 2, hence the low quantity of Si.

As stated earlier, the spectra acquired from the combined basic images are given in Figure 2. The concentration of each measured spectrum is defined by the intensity of the Raman peak, which is defined by equation (1)^13–15

I=CNI 0 σ Ω LF ( T )

(1)

where I is the Raman signal intensity, C is arbitrary constant, N is the number density of the element, I 0 is the laser intensity which is the incident light beam from the confocal Raman, σ Ω is the Raman cross-sectional area, σ Ω L is the scattering angle, L is the laser beam segment length and F(T) is the temperature-dependent factor.

The temperature-dependent factor is a function of the spectral width, system resolution detector and mechanical properties of the analysed sample. In the defect-free region of the sample, the silicon bond vibration corresponds to the red spectrum which peaks at 525 cm⁻¹ as represented in Figure 2. However, Si Raman peak is theoretically given as 520 cm⁻¹ but in this study, the three known Si phonon Raman peaks degenerated at 525 cm⁻¹ due to the Si reference used as observed during the calibration process. Hence, 525 cm⁻¹ is used for Si in this study. Generally, Si has vibration modes at spectra line of 300 cm⁻¹, 520 cm⁻¹ (narrow line) and overtone at 940 cm⁻¹.^13,14 The higher and broader peaks in Figure 2 correspond to the forbidden transverse optical peaks, which are not the characteristics of any of the known three basic Si phonons; hence, they represent impurities.

In addition, from Figure 2, the first-order Raman spectrum of the poly-Si reveals extra phonon frequencies more than the theoretically recognised ones for crystalline solar cells. Apart from Si Raman peak that is normally present between 518 cm⁻¹ and 525 cm⁻¹, the other observed bands, 78 cm⁻¹, 227 cm⁻¹ and 1300 cm⁻¹, are attributed to impurities due to defects at the region where total junction breakdown occurred. The presence of unusual vibrational mode also appears at 288 cm⁻¹ due to co-implantation of oxygen (O) in the defected region and this is possible because of the high O content associated with hotspot centre.^15,16

FHWM analysis of different regions

The origins of Raman shift emanated from shift in the bond-to-bond vibrational frequency induced due to strain or stress in the analysed sample. The interatomic vibrational alteration is best identified via Raman spectroscopy to show the distribution of stress across the material.^{17,12–14,18} In analysing the degree of shift experienced in this study, Gaussian line was used to analyse the spectra in order to have the best analysis. This theory is based on experimental test observed during the analytical part of this work and this agrees with the work of Anastassakis. ¹⁸ Although a combination of Gaussian–Lorentzian shape could also yield a good analysis, the complexity surrounding the method due to additional fitting parameters makes this method less attractive.^19–21 Figure 3 indicates the Raman shift experienced by the spectra from the affected and the non-affected regions. Note that the affected spectrum is presented with a red line, while the region outside the affected region is presented in black. The inserted spectra in Figure 3 are the actual spectra before applying fitting; thus, the inserted spectra are the main spectra of the analysed sample.

OPEN IN VIEWER

Figure 3.

Qualitative analysis of stress/strain in and outside the defective region using FWHM, showing a shift in the Raman peak from 525 cm⁻¹ to 518 cm⁻¹ and a decrease in the spectrum width as a sign of low quality.

The most significant observation is the shift in the Raman peak from high to low frequency due to strain and decrease in the spectrum width. The decrease in the bandwidth is a sign of low material quality in the defective region; this can be attributed to the high concentration of impurities observed.

Deep level analysis

From single Raman spectra, it is clear that the defect resulted from elemental impurities, but the depth of impurities could not be determined; hence, there is a need for depth profiling. The section of the sample analysed in Figure 1 was further subjected to deep level scanning to ascertain the depth of the defect cluster. The depth analysis shows that the impurities in the defected centre as seen in the Raman spectra are not surface contaminations, rather a deep level impurity defects. The influence of tensile stress in the material active layer resulted to the precipitation of the moonlike defect, which has been addressed as the defective cluster in this study. The deep level scan of the defect cluster is presented in Figure 4. The blue line in Figure 4(a) indicates the spot where the deep level scan was performed, while Figure 4(b) to (d) show the different drawn images; the combined image is shown in Figure 4(e).

OPEN IN VIEWER

Figure 4.

Large area scan of a reverse biased poly-Si cell showing the depth profiling of the defective cluster region. (a) The optical image. (b–d) The drawn images. (e) The combined drawn image.

As stated before, the aim of the depth scan measurement was to investigate how deep the defect is in the sample substrate. Figure 4 shows an in-depth profile of the defected cluster which is acting at contaminants. This profiling was made possible because the sample matrix is transparent to the incoming laser. The position identified was manually focused and the resulted images show the distribution and origin of the defect. The profile reveals that the contaminants have high concentration at the middle as seen from the distribution. The depth profiling gave detail information about the defect; this technique work on the principle of optical sectioning by improving the aperture lateral and axial spatial resolution of the device; hence, depth profile data are acquired in an incremental order deep into the material. This helps eliminate the possibility of sample contamination, since no extra preparation is needed.

The depth profile image shows that there are three distinct elemental impurities that constitute the defective cluster. The fourth impurity observed from the spectra presented in Figure 2 is a metal substance, while the fifth is purely silicon, which is the parent element of the solar cell. The depth scan also confirms the various concentrations that resulted to the difference in the intensity signal resulting from Raman spectra.

Conclusion

The application of confocal Raman spectrometer in characterising defects in poly-Si solar cell has been demonstrated. This paper revealed the effect of impurity on the defect level of poly-Si solar cell. Raman spectra show that the observed moonlike defect clusters originated from high concentration of chemical impurities. The precipitation of this defect cluster is attributed to hotspot formation which was enhanced during the reverse biasing process. The changes in the Raman peaks and the presence of other peaks show the presence of impurities in the analysed sample. In this study, FWHM analysis was used to quantify the stress level in the affected region, and it was observed that the interatomic bonding of the defected region experiences strain because the Raman shift moves towards the strain level compared to the stress position. The changes in the single spectra widths and variation in the intensities made it possible to probe the defect level and the nature of the stress.