Anthony Teal

An Improved Method to Measure the Point Spread Function of Cameras Used for Electro- and Photoluminescence Imaging of Silicon Solar Cells

When silicon solar cells are investigated by electro- or photoluminescence (PL) imaging using a silicon-based camera, photon scattering in the detector chip leads to a certain degree of blurring of the images, which can be removed by image deconvolution. The necessary point spread function (PSF) was originally measured directly by evaluating fine light spots, but this procedure needed the evaluation of several images. It was shown recently that evaluating an image with a sharp contrast edge in the middle may lead to a PSF spreading over longer distances by evaluating only one image. Here, we show that the previous backwards substitution method for obtaining this PSF does still lead to residual errors. An alternative method to derive the PSF from a measured contrast edge image is introduced here. It uses an iterative method, which leads to most precise results. In addition to the local deconvolution, nonlocal (homogeneous) light scattering is corrected. The correct reconstruction of the dark part of the image used for obtaining the PSF is a proof for the accuracy of this PSF.

Correcting the inherent distortion in luminescence images of silicon solar cells

Luminescence imaging of Silicon solar cells is typically performed with a silicon CCD, which is a poor absorber of silicon luminescence (900-1300 nm). This leads to a phenomenon referred to as photon smearing in the CCD, where a photon incident on one pixel may be absorbed in another. This makes the image blurry and quantitative analysis of this data inaccurate. Also resolution, contrast, and sharpness of the image are reduced at features such as grain boundaries, and sample edges. An already established method of recovering the original luminescence signal incident on the CCD is to deconvolve a Point Spread Function (PSF) with the resultant image. This paper focuses on a novel method for determining the PSF from a measurement of the Edge Spread Function, which greatly increases the Signal to Noise ratio over methods where the PSF is measured directly. The determined PSF and its application to luminescence images, is compared and contrasted with previously published PSF determination methods.

Lifetime analysis of laser crystallized silicon films on glass

Only recently, the quality of liquid phase crystallized silicon directly on glass substrates made a huge leap towards the quality of multi-crystalline wafers with open circuit voltages well above 600 mV. In this paper, we investigate the material quality in order to identify the factors limiting further performance improvements. We employ photoluminescence imaging on a state of the art test structure with lifetime calibration by transient photoluminescence. The resulting lifetime map is converted into an effective diffusion length map and the origin of regions with short lifetimes is investigated with electron backscattering and transmission electron microscopy. High local dislocation densities in areas with dissociated coincidence site lattice boundaries were found to be responsible for the localised quenching of the photoluminescence signal.

Photoluminescence Imaging Using Silicon Line-Scanning Cameras

We experimentally demonstrate photoluminescence imaging using a silicon line-scanning camera. The narrow rectangular geometry of the sensor effectively lowers the image-blur, while keeping the sensitivity relatively high. Line scanning is demonstrated to provide sharp and high contrast images. To maintain fast measurement speeds, higher injections are used to compensate for the lower sensitivity of a line-scan measurement, where localized parts of the image are acquired sequentially. However, low-injection measurements remain possible at the expense of slower scan speeds if cooling is applied to the sensor. Using a detailed point-spread analysis and deconvolution, we quantify and correct for the remaining light spreading effects. Experimental results suggest an effective suppression of the cumulative nonlocal point spread commonly observed using area sensors, confirming silicon (Si) line sensors to be only susceptible to local point spreading effects.

Evaluating the accuracy of point spread function deconvolutions applied to luminescence images

Luminescence imaging is a widely used characterization technique for silicon photovoltaics. However, the tools used to acquire images typically utilize a silicon CCD array for detection, which is a poor absorber at silicon luminescence wavelengths. This leads to a smearing effect in the measured image which can be characterized by a point spread function (PSF). If the true PSF is known then the measured image can be restored through deconvolution. Several methods exist for determining a PSF for a particular imaging system and different extraction techniques can lead to variations in the PSF result, yet no studies have provided comprehensive analysis of PSF deconvolution accuracy when applied to luminescence imaging. In this work, several new techniques have been designed and investigated in order to test PSF deconvolution results, with a view to quantifying improvement or errors generated and potentially leading towards improved image restoration.

Dual-beam laser thermal processing of silicon photovoltaic materials

We have developed an all-laser processing technique by means of two industrially-relevant continuous-wave fiber lasers operating at 1070 nm. This approach is capable of both substrate heating with a large defocused beam and material processing with a second scanned beam, and is suitable for a variety of photovoltaic applications. We have demonstrated this technique for rapid crystallization of thin film (~10 μm) silicon on glass, which is a low cost alternative to wafer-based solar cells. We have also applied this technique to wafer silicon to control dopant diffusion at the surface region where the focused line beam rapidly melts the substrate that then regrows epitaxially. Finite element simulations have been used to model the melt depth as a function of preheat temperature and line beam power. This process is carried out in tens of seconds for an area approximately 10 cm2 using only about 1 kW of total optical power and is readily scalable. In this paper, we will discuss our results with both c-Si wafers and thin-film silicon.

Improved spatial resolution of luminescence images acquired with a silicon line scanning camera

Luminescence imaging is currently being used to provide spatially resolved defect in high volume silicon solar cell production. One option to obtain the high throughput required for on the fly detection is the use a silicon line scan cameras. However, when using a silicon based camera, the spatial resolution is reduced as a result of the weakly absorbed light scattering within the cameras chip. This paper address this issue by applying deconvolution from a measured point spread function. This paper extends the methods for determining the point spread function of a silicon area camera to a line scan camera with charge transfer. The improvement in resolution is quantified in the Fourier domain and in spatial domain on an image of a multicrystalline silicon brick. It is found that light spreading beyond the active sensor area is significant in line scan sensors, but can be corrected for through normalization of the point spread function. The application of this method improves the raw data, allowing effecti...