Renyu Chen

Phosphorus vacancy cluster model for phosphorus diffusion gettering of metals in Si

In this work, we develop models for the gettering of metals in silicon by high phosphorus concentration. We first performed ab initio calculations to determine favorable configurations of complexes involving phosphorus and transition metals (Fe, Cu, Cr, Ni, Ti, Mo, and W). Our ab initio calculations found that the P4V cluster, a vacancy surrounded by 4 nearest-neighbor phosphorus atoms, which is the most favorable inactive P species in heavily doped Si, strongly binds metals such as Cu, Cr, Ni, and Fe. Based on the calculated binding energies, we build continuum models to describe the P deactivation and Fe gettering processes with model parameters calibrated against experimental data. In contrast to previous models assuming metal-P1V or metal-P2V as the gettered species, the binding of metals to P4V satisfactorily explains the experimentally observed strong gettering behavior at high phosphorus concentrations.

A model for phosphosilicate glass deposition via POCl3 for control of phosphorus dose in Si

Effective control of the dose of diffused phosphorus emitter profiles is crucial for optimization of crystalline silicon solar cells, but it requires detailed understanding of the POCl3 doping process. We measure concentration profiles within the deposited phosphosilicate glass (PSG) layer for a range of POCl3 doping conditions and find that (i) its composition is nearly independent of process conditions and (ii) it is separated from Si by a thin SiO2 layer. We also find strong accumulation of P at the SiO2-Si interface. As common linear-parabolic models cannot fully explain the observed kinetics of PSG thickness and phosphorus dose in Si, we present an improved model including oxygen depletion and dose saturation, giving a better explanation of the experimental data. In contrast to previous models that adjust the peak phosphorus concentration at the Si surface to match the measured profiles, our models accurately predict the time-dependent dose behavior under different experimental conditions. We further...

Improving the predictive power of modeling the emitter diffusion by fully including the phosphsilicate glass (PSG) layer

Presently, the PV industry is switching to the selective emitter design, where the phosphorus density is significantly reduced between the front metal fingers. Current diffusion models simply adjust the peak phosphorus density at the Si surface to match the measured profiles, and therefore they are unable to predict the necessary gas flows and the temperatures during predeposition and drive-in to realize an optimum emitter profile. In order to achieve better prediction capabilities, we implement a model for the phosphosilicate glass (PSG) layer and for its coupling to silicon. We combine this model with coupled dopant/defect diffusion models in Si to calculate the resulting dopant profiles. With our improvements, we reproduce the profile measurements for a range of POCl3 flows at temperatures typically chosen in industrial fabrication.

Analyzing emitter dopant inhomogeneities at textured Si surfaces by using 3D process and device simulations in combination with SEM imaging

The lowering of the phosphorus dopant density in the emitter of Si solar cells is a current topic in the photovoltaic industry. In lowly-doped emitters, diffusion inhomogeneities between the tops of the pyramids and the valleys affect the saturation current density J0. We quantify diffusion inhomogeneities by means of 3D process simulations, and we evaluate J0 by means of 3D device simulations. Finally, we compare the simulated diffusion results with a 2D dopant-contrast analysis obtained with a scanning electron microscope (SEM). Both methods show a deeper in-diffusion at the top of the pyramid and a shallower in-diffusion in the valley regions. Within the pyramidal faces, the diffusion is between both of these extreme points and comparable with in-diffusion at planar structures. The results of the device simulations indicate that the increase of J0 from planar to textured surfaces depends on the dopant profile and the surface passivation, but that a factor of about 5 is observed in our example, as is observed experimentally.

Kinetic lattice Monte Carlo simulations of interdiffusion in strained silicon germanium alloys

Point-defect-mediated diffusion processes are investigated in strained SiGe alloys using kinetic lattice Monte Carlo (KLMC) simulation technique. The KLMC simulator incorporates an augmented lattice domain and includes defect structures, atomistic hopping mechanisms, and the stress dependence of transition rates obtained from density functional theory calculation results. Vacancy-mediated interdiffusion in strained SiGe alloys is analyzed, and the stress effect caused by the induced strain of germanium is quantified separately from that due to germanium-vacancy binding. The results indicate that both effects have substantial impact on interdiffusion.

Coupled modeling of the competitive gettering of transition metals and impact on performance of lifetime sensitive devices

This work models competitive gettering of metals (Cu, Ni, Fe, Mo, and W) by boron, phosphorus, and dislocation loops, and connects those results directly to device performance. Density functional theory calculations were first performed to determine the binding energies of metals to the gettering sites, and based on that, continuum models were developed to model the redistribution and trapping of the metals. Our models found that Fe is most strongly trapped by the dislocation loops while Cu and Ni are most strongly trapped by the P4V clusters formed in high phosphorus concentrations. In addition, it is found that none of the mentioned gettering sites are effective in gettering Mo and W. The calculated metal redistribution along with the associated capture cross sections and trap energy levels are passed to device simulation via the recombination models to calculate carrier lifetime and the resulting device performance. Thereby, a comprehensive and predictive TCAD framework is developed to optimize the pro...

Coupled modeling of evolution of impurity/defect distribution and cell performance

We demonstrate the use of end-to-end predictive modeling to optimize silicon solar cell fabrication processes. Coupled continuum models for dopants, metals, and light elements (e.g., O) are used to predict the distribution of electronically active defects. The models include point defect-mediated diffusion of boron and phosphorus from solid sources at both the emitter and back surface field. Interactions between metals and boron, boron and oxygen, dopants and point defects, and metals and dopant-defect complexes are modeled. The resulting impurity and defect distributions along with associated trap levels and capture cross-sections are passed to device simulation. The modeling results suggest strategies to optimize device performance in the presence of contamination.

Understanding coupled oxide growth and phosphorus diffusion in POCl 3 deposition for control of phosphorus emitter diffusion

Effective control of diffused phosphorus profiles in crystalline silicon requires detailed understanding of the doping process. We analyze concentration profiles within the deposited phosphosilicate glass (PSG) for a range of POCl3 conditions and develop a model to account for the experimentally observed time dependence of PSG thickness and dose of phosphorus in Si. A simple linear-parabolic model cannot fully explain the kinetics of thickness and dose; while an improved growth model including oxygen dependence and dose saturation gives better fits to the experiments. We further couple the growth model with phosphorus diffusion and deactivation models in silicon and provide full modeling of the POCl3 doping process.

End-to-end predictive modeling of silicon solar cell performance: From process recipe to device simulation

We demonstrate the use of end-to-end predictive modeling to optimize silicon solar cell fabrication processes. Coupled continuum models for boron, phosphorus, iron, and oxygen are used to predict the distribution of electronically active defects. The models include point defect-mediated diffusion of boron and phosphorus from solid sources (e.g., POCl3) at both the emitter and back surface field. Interactions between iron and boron, boron and oxygen, and phosphorus and vacancies are modeled. From the resulting dopant and defect distributions, carrier lifetimes are computed and, along with the solar cell structure, are passed to a device simulation. We find a competing effect between lifetime-enhancing iron segregation to the boron-rich region and lifetime-degrading BO2 complex formation. Guided by modeling results, we suggest strategies to restore device performance in the presence of iron contamination and moderate oxygen concentrations, demonstrating the utility of an “end-to-end” modeling framework.