C. Flink

Electrical and Recombination Properties of Copper‐Silicide Precipitates in Silicon

Copper-silicide precipitates in silicon obtained after copper diffusion and quench in different liquids were studied by transmission electron microscopy and capacitance spectroscopy techniques. A correlation between the quenching rate, geometric size, and deep level spectra of the copper-silicide precipitates was established. The unusually wide deep level spectra are shown to be due to a defect-related band in the bandgap. The parameters of the band are evaluated using numerical simulations. a positive charge of copper-silicide precipitates in p-type and moderately doped n-type Si is predicted by simulations and confirmed by minority carrier transient spectroscopy measurements. Strong recombination activity of the precipitates due to attraction of minority carriers by the electric field around the precipitates and their recombination via the defect band is predicted and confirmed by the experiments. The pairing of copper with boron is shown to be an important factor determining the precipitation kinetics of the interstitial copper at room temperature.

Diffusion, solubility and gettering of copper in silicon

Abstract The feasibility of quantitative predictive modeling of gettering of Cu in silicon, which requires quantitative understanding of its diffusivity and precipitation behavior, is discussed. Investigations of diffusion of Cu at low temperatures enabled us to determine the pairing constants of copper with boron, re-evaluate its diffusivity at room temperature in p-Si, and to predict its diffusivity in p + -Si substrates. We demonstrate that copper may either precipitate in the bulk of the wafer or diffuse to its surface, depending on the position of Fermi level in the sample. It is suggested that the Fermi level position determines the sign and magnitude of the electrostatic charge on the growing copper precipitates, and thus enhances or suppresses precipitation of interstitial copper ions. Modeling of p/p + segregation gettering of copper shows that while the copper can be gettered in p + layer during or after high-temperature anneals, it eventually will be released and will precipitate in the device region within the first few months of operation, unless more stable gettering (precipitation) sites for copper are utilized. An n-type layer is predicted to be an effective gettering site.

Gettering of iron by oxygen precipitates

In order to better understand and model internal gettering of iron in silicon, a quantitative investigation of iron precipitation in silicon containing different oxygen precipitate densities was performed. The number of iron precipitation sites was obtained from the iron precipitation kinetics using Ham’s Law. At low temperatures, the iron precipitate density corresponded to the oxygen precipitate density. A strong temperature dependence of the iron precipitate density was observed for the samples with larger oxygen precipitate densities. These data were used to simulate iron precipitation during a slow cool. From those simulations, optimal cooling rates were obtained for different silicon materials assuming various iron precipitation site densities in the epitaxial layer.

Interstitial copper-related center in n-type silicon

n-type silicon samples were measured by deep level transient spectroscopy (DLTS) immediately (within one hour of storage at room temperature, required for the preparation of Schottky-diodes) after copper diffusion and quench. A donor level at Ec-(0.15±0.01) eV with a concentration of up to 1013 cm−3 was detected. The amplitude of the DLTS peak decreased with the time of storage at room temperature, and stabilized at a concentration (4 to 7)×1011 cm−3 after 15–20 h. The activation energies and prefactors of the decay of the DLTS peak in n-type Si and the reactivation of copper-compensated boron in p-type Si concur. This correlation suggests that the deep level is interstitial copper itself or a complex of interstitial copper.

The dissociation energy and the charge state of a copper-pair center in silicon

Thermal dissociation of Cu pairs was studied in p-type silicon. The dissociation energy of the Cu pair was found to be 1.02±0.07 eV, twice as high as the binding energy of a Coulombically bound donor-acceptor pair placed on nearest neighbor 〈111〉 sites. This implies that the pair is either covalently bonded, or it consists of an ionically bonded doubly negatively charged acceptor and a singly charged donor. To distinguish between these two models, the dependence of the hole emission rate on the electric field in the depletion region was studied. The absence of the Pool-Frenkel emission enhancement ruled out the acceptor nature of the center and the purely ionic type of bonding. On the other hand, the polarization potential describing emission from a neutral impurity gave a satisfactory fit to the experimental data. It is concluded that the Cu pair is a donor with either covalent or mixed type of bonding.

Influence of interstitial copper on diffusion length and lifetime of minority carriers in p-type silicon

Though copper can be quenched interstitially in p-type silicon, it precipitates completely within 10–15 h at room temperature. The decay of concentration of interstitial copper in p-Si was monitored by capacitance–voltage characteristics (C–V) and surface photovoltage (SPV). It is shown that the time constant of change of minority carrier diffusion length, as measured by SPV, correlates well with the precipitation of interstitial copper. The capture cross section of interstitial copper is estimated to be in the range 10−15–10−17 cm2. It is shown that interstitial copper is far less deleterious than iron and can limit the diffusion length of commercial p-Si wafers only if its concentration is above 1013 cm−3. However, copper precipitates may be detrimental to lifetime even for much lower copper concentrations.

Electrical characterization of copper related defect reactions in silicon

Abstract Defect reactions involving interstitial copper impurities (Cu i ) in silicon are reviewed. The influence of the Coulomb interaction between positively charged copper and negatively charged defects, such as acceptor states of transition metals and lattice defects, on the complex formation rate is discussed in detail. The diffusivity of interstitial copper and the dissociation kinetics of copper–acceptor pairs are studied using the recently introduced transient ion drift (TID) method. TID results reveal that most interstitial copper impurities remain dissolved immediately after the quench and form pairs with shallow acceptors. It is shown that in moderately and heavily doped silicon the diffusivity of copper is trap limited, while in low B-doped silicon the interstitial copper–acceptor pairing is weak enough to allow the assessment of the copper intrinsic diffusion coefficient. The intrinsic diffusion barrier is estimated to be 0.18±0.01 eV. It is concluded that the Coulomb potential used in previous publications underestimated considerably the acceptor–copper interaction. In light of these results, a general discussion on Cu related defect reactions is given.

Analysis of Iron Precipitation in Silicon as a Basis for Gettering Simulations

In order to better understand and model internal gettering of iron in silicon, a quantitative investigation of iron precipitation in silicon containing different oxygen precipitate densities was performed. The number of iron precipitation sites was obtained from iron precipitation kinetics using Hams Law. At low temperatures, the iron precipitation site density corresponded to the oxygen precipitate density. A strong annealing temperature dependence of the iron precipitate density was observed for the samples with larger oxygen precipitate densities. These data were used to simulate iron precipitation during a slow cool. From these simulations, optimal cooling rates were obtained for different silicon materials assuming various iron precipitation site densities in the denuded zone.

Determination of parameters of deep level defects from numerical fit of deep level transient spectroscopy spectra: Analysis of accuracy and sensitivity to noise

The numerical fit of deep level transient spectroscopy (DLTS) spectra, used primarily to analyze complex DLTS spectra, is evaluated in terms of the accuracy of measuring deep levels and the sensitivity to noise. It is shown that by using numerical fit of DLTS spectra, the uncertainties in the emission activation energy and the capture cross section of deep level defects can be improved by three to four times over the standard Arrhenius plot method. Two modifications of the fitting procedure are tested: a fit of a DLTS spectrum using one rate window, and a simultaneous fit using five different rate windows. It is shown that simultaneous fit of spectra using different rate windows is significantly more accurate, has noticeably larger convergence radius for the initial values of parameters, and is less sensitive to noise. The advantages of the fitting routine are demonstrated on experimentally obtained noisy DLTS spectra.

Direct Correlation of Solar Cell Performance with Metal Impurity Distributions in Polycrystalline Silicon using Synchrotron-Based X-ray Analysis

Direct correlation of solar cell performance with metal impurity distributions in polycrystalline silicon using synchroton–based x–ray analysis