James P. Lavine

The effect of rapid thermal annealing on the precipitation of oxygen in silicon

We report observations of the effect of heat pulses on the precipitation of oxygen in single‐crystal silicon. In our experiments, heat pulses are applied for various durations at a fixed temperature of 1200 °C prior to a two‐step precipitation sequence. When no pulse is applied, precipitation is at a maximum and is close to that expected from considerations of solid solubility, provided account is taken of the loss of oxygen due to evaporation. For very short pulses (e.g., 2 s), precipitation is suppressed by several fold. As the length of the pulse is increased to roughly 100 s, the precipitation recovers to its initial value. For very long pulses (greater than 10 000 s), precipitation decreases to nearly zero. There are no further changes in the precipitation characteristics for pulse times up to 50 000 s. We interpret these results in terms of a model in which the thermal pulses modify an initial distribution of heterogeneous nucleation sites. The fraction of heterogeneous sites that survive the two‐st...

Steady-state photocarrier collection in silicon imaging devices

Solid-state imagers lose resolution when photocarriers generated in one imaging site diffuse to a nearby site where they are collected. These processes are modeled by solving the steady-state diffusion equation for minority carriers. A source term represents the absorption of photons and the generation of photocarriers, and a linear term represents the loss of photocarriers by recombination. This is equivalent to studying the Helmholtz equation with an inhomogeneous term. The problem is simplified when the light source has symmetry. A line source or a cylindrically symmetric source leads to a two-dimensional problem. The approach of Seib, Crowell, and Labuda allows a solution by quadrature if the further assumption of a smooth top boundary is made. We calculate the integrated normal flux over each imaging site to see how many carriers diffuse from under the illuminated site to another site. We compare our predicted line- and point-spread functions to those measured on imagers and find reasonable agreement. This allows us to extract minority-carrier diffusion lengths. Further calculations show how the diffusion of carriers depends on the photon wavelength and the pixel size. We generalize Seibs approach and apply it to a solid-state imager covered with color filters. This allows us to see the extent of color mixing due to carrier diffusion. We also discuss a finite-difference solution of the diffusion equation that employs the method of conjugate gradients. This approach is useful for problems where the top boundary is not smooth.

A model for charge transfer in buried-channel charge-coupled devices at low temperature

Charge transfer in buried-channel charge-coupled devices (CCDs) is explored with a one-dimensional numerical model which describes the capture and emission of electrons from a shallow donor level in silicon through the use of the Shockley-Read-Hall generation-recombination theory. Incorporated in the model are the three-dimensional Poole-Frenkel barrier lowering theory of A. K. Jonscher (1967) and J. L. Hartke (1968) and the low-temperature form of Poissons equation. Reasonable agreement of the model with experimental data taken from the buried-channel CCDs of a PtSi Schottky barrier infrared image sensor is found. Moreover, the value for the capture cross section of electrons to the shallow phosphorus level in silicon inferred from the model follows the cascade theory for capture by M. Lax (1959) and agrees roughly with determinations made by other experimenters. >

An analytical, aperture, and two-layer carrier diffusion MTF and quantum efficiency model for solid-state image sensors

A two-dimensional analytical model is formulated for calculating the pixel response, modulation transfer function (MTF), and quantum efficiency of front-side illuminated, solid-state image sensors. Included in this unified model are the effects of lateral diffusion of charge carriers within a two-layer substrate and less than full pixel sampling apertures. The results of this model are compared to those of a numerical, three-dimensional Monte Carlo algorithm and to the analytical results reported by Blouke and Robinson. We find good agreement between the quantum efficiency and MTF calculated by the present model and by the three-dimensional Monte Carlo method. However, we find higher quantum efficiency and lower MTF than the previously reported analytical two-layer model. The unified aspect of the present model correctly combines the effects of sampling aperture and lateral diffusion. >

Latch-up and image crosstalk suppression by internal gettering

Internal gettering can be used to reduce crosstalk in imagers and latch-up susceptibility in CMOS circuits. The internal gettering process forms defects in the bulk of the silicon wafers that are effective recombination sites for minority carriers in the substrate. Experimental and theoretical results are presented for the crosstalk reduction obtained in an area imager. Also, the current gain β of the parasitic lateral n-p-n transistors formed in the substrate in CMOS circuits was considerably lower for the internally gettered wafers. The trigger current needed to initiate latch-up in the n-p-n-p structures increased as 1/β, in accordance with the theory. A Monte Carlo method was developed to calculate the expected lateral transistor current gain. The calculated βs are in excellent agreement with the measured values.

Probing Metal Defects in CCD Image Sensors

We have investigated the role of heavy metals in causing visible pixel defects in Charge Coupled Device (CCD) image sensors. Using a technique we call dark current spectroscopy, we can probe for deep-level traps in the active areas of completed image sensors with a sensitivity of 1 × 10 9 traps/cm 3 or better. Analysis of histograms of dark current images from these sensors shows that the presence of traps causes quantization in the dark current. Different metal traps have characteristic dark current generation rates that can identify the contaminant trap. By examining the temperature dependence of the dark current generation, we have calculated the energy levels and generation cross sections for gold, iron, nickel, and cobalt. Our results show the relationship of these traps to the “white spot” defects reported for image sensors.

Phosphorus diffusion in polycrystalline silicon

The diffusion of phosphorus in crystallized amorphous Si layers was studied with secondary‐ion mass spectroscopy. A two‐dimensional diffusion model is used to find effective grain (Dg) and grain‐boundary (Dgb) diffusion coefficients. This simplified model leads to Dgb ≤ 10Dg, which is significantly lower than what has been deduced from conventional, larger grained polysilicon. Our result is consistent with specific‐gravity measurements, which found a significantly lower ‘‘mass defect’’ for layers deposited amorphous and subsequently crystallized as compared to initially polycrystalline layers.

Monte Carlo simulation of the photoelectron crosstalk in silicon imaging devices

The Monte Carlo method is used to evaluate the extent of the crosstalk in solid-state imagers. The calculations are performed in three dimensions and are in excellent agreement with experiment. The Monte Carlo method is used because it handles adjacent regions that either collect or reflect minority carriers.

A 1/3" format image sensor with refractory metal light shield for color video applications

The authors report results obtained on a full-color interline transfer CCD (charged-coupled device) image sensor with pixel dimensions of 8.6 mu m(H)*6.8 mu m(V) using 1.2- mu m design rules and a two-phase, single-polysilicon-per-phase technology. In order to reduce image smear and to provide suitable topography for integral color filters, a refractory light shield with a flowed glass overlayer was incorporated. The basic sensor and pixel architecture is shown. Image smear as a percent of full well, measured with 10% vertical illumination at saturated intensity, is shown as a function of wavelength. Smear is lowest at short wavelengths but is at an acceptable level for applications with controlled illumination.<<ETX>>

Calculated moments for the implantation of boron into silicides

The Monte Carlo computer program trim is used to calculate the projected range, the standard deviation, the skewness, and the kurtosis of implanted boron ion distributions in silicon, silicon dioxide, molybdenum disilicide, tantalum disilicide, titanium disilicide, and tungsten disilicide. The boron implantation energy ranges from 10 to 500 keV.