We have made an extensive study of Si 0.82 Ge 0.18 film relaxation on silicon on insulator (SOI) substrates having a top Si layer 40, 70, 330nm, and 10[.proportional]m thick. SiGe films were deposited with a thickness up to 1.2[.proportional]m in an ultrahigh vacuum chemical vapor deposition system at 630°C. Following growth, films were characterized by X-ray diffraction and a dislocation revealing etch. The same level of relaxation is reached for each thickness of SiGe film independent of the substrate structure. Accompanying the film relaxation is the development of a tetragonal tensile strain in the thin Si layer of the SOI substrates. This strain reached 0.22% for the 1.2[.proportional]m film on the 40nm SOI and decreases with SOI thickness. The Si thickness of the SOI substrate also effected the threading dislocation density. For 85% relaxed films the density fell from 7×10 6 pits/cm 2 on bulk Si to 10 3 pits/cm 2 for the 40, 70, and 330nm SOI substrates. The buried amorphous layer of the SOI substrate alters the dislocation dynamics by allowing dislocation core spreading or dislocation dissociation. The reduced strain field of these dislocations reduces dislocation interactions and the pinning that results. Without the dislocation pinning, the misfit dislocations can extend longer distances yielding a greatly reduced threading dislocation density.

We demonstrate the use of low-energy electron microscopy (LEEM) as a tool for studying dis-location formation in low-Ge-content SiGe films on Si(001) and silicon-on-insulator. Compared to TEM, sample preparation for LEEM consists only of conventional surface cleaning. Yet, because of its sensitivity to local variations in surface strain on Si(001), LEEM can detect dislocations at the earliest stages of strain relaxation. In identically prepared SiGe films, the typical dislocation extends over the entire viewable region of several hundred microns in SiGe/Si, but is less than 100 microns in SiGe/SOI. In addition, dislocation cross-slip and threading segments are common in SiGe/SOI, but virtually non-existent in SiGe/Si. We have also observed dislocation formation in real-time during high temperature annealing. Preliminary results appear to demonstrate dislocation multiplication and blocking at a perpendicular glide plane. The applicability of LEEM to strain relaxation in other Si-based systems will be discussed.

InGaP/GaAs HBTs have been achieved via base doping and base compositional grades focused adjacent to the collector. This has been done while avoiding unacceptable increases in base sheet resistances which are typically seen with plain linear grades. Additionally, semi-ballistic launching of carriers into the base layer is shown to further enhance base electron velocity and device gain. These base grading and launching techniques enable enhanced dc gain which, among other benefits, can offset the gain killing effects of very heavily doped contact layers for advanced wireless devices.

Heterojunction Bipolar Transistor (HBT) epi wafer manufacturing at thousands of wafers per week requires careful monitoring and control to ensure excellent quality and high yields. All product epi wafers undergo non-destructive testing which gives data on epi wafer sheet resistance, particle density, and surface roughness. These non-destructive measurements give only a limited view of wafer quality, therefore periodic destructive testing of selected epi wafers (HBT device fabrication and measurement) is required. It is highly desirable to directly measure key epi layer parameters on each product wafer in order to continue pushing toward the goals of perfect quality and yields. Therefore we have been working with the Laytec EpiTT in situ optical reflectance monitoring system. These data allow for immediate detection of shifts in the epi layers. Therefore problems can be detected and corrective actions taken before failed wafers are grown.

Photoreflectance (PR) is an optical measurement technique used to study material properties and device structures. It is uniquely able to non-destructively measure electric fields within devices. Some device structures such as HBTs have a rough highly doped InGaAs surface layer. In some cases the surface roughness is sufficient to produce light scattering that impedes optical measurements such as PR. We have developed a new measurement configuration separating the pump laser, which causes the light scattering, from the probe laser. Rapid, non-destructive measurements of the emitter and collector depletion regions are possible with this configuration on product wafers with rough surface layers INTRODUCTION The production of heterojunction bipolar transistors (HBT) is a high volume process involving multiple production tools operating 24 hours a day 7 days a week. This production achieves tight specifications and high levels of quality assurance. Quality assurance relies on destructive device fabrication and measurement of individual wafers to indicate the overall quality of production. It is important to reduce the number of product wafers consumed for this testing yet maintain the high levels of quality. This has driven the industry to develop non-destructive measurements of product wafers. Measurements of sheet resistance, particles, and surface haze are commonly performed. We have previously demonstrated the utility of using a fully automated X-ray diffraction tool to monitor the base doping of GaAs HBTs [1]. This is complemented by a new configuration for a photoreflectance (PR) measurement that is able to measure product HBT wafers, which previously produced too much light scattering to be measured. PR has been used for decades to study material and device properties [2]. The measurement is routinely used to measure band-gap, ordering of InGaP alloys, and electric field. Its ability to measure electric fields distinguishes it from other optical measurement techniques such as ellipsometry or photoluminescence. This ability allows key information about the collector and emitter depletion regions to be analyzed. Production HBTs utilize a rough InGaAs contact layer to reduce emitter contact resistance. This rough layer strongly scatters light, which interferes with the PR measurement. EXPERIMENTAL We have found a new measurement configuration that allows a rough sample to be measured and nullifies the problem of the light scattering. The typical PR configuration is drawn in Figure 1. The measurement involves two light sources. Here a red diode laser is used as the pump laser, which is electronically chopped. The probe beam is generated from a halogen lamp with the wavelength selected by a monochromator. The electrons and holes generated by the chopped pump beam modify the electric

We have experimentally studied the effect of two new base doping profiles on the base transit time of a GaAs npn heterojunction bipolar transistor. The doping in a region close to the collector is reduced either by a doping grade or a stepwise reduction. Quasi-electric fields resulting from these doping gradients increase the minority carrier velocity and the beta of large area transistors. By focusing these doping changes adjacent to the collector, the amount of low-doped base material and the resulting increase in base sheet resistance can be minimized. For both a step change in doping or graded doping change a 10% decrease in base transit time is achieved while only causing a 4 % increase in base sheet resistance. The impact on base transit time is confirmed with f/sub T/ data on small area devices.

We experimentally investigate the details of the minority electron transport across the base of a GaAs heterojunction bipolar transistor (HBT). A stepwise doping change in the base layer is employed to locally accelerate the electrons and enhance the dc current gain. The position of the step change is varied to probe the electron velocity across the base layer in multiple sample sets with different base thicknesses. Positioning the doping change within 200 /spl Aring/ of the collector provides the largest velocity increase (/spl sim/15%), independent of base thickness. This indicates that the electrons are moving slower as they approach the collector, an observation which sharply contrasts with the conventional quasi-ballistic drift-diffusion models of base transport that predict the highest electron velocities are reached near the collector.

We have developed an InGaP/GaInAsN/GaAs double heterojunction bipolar transistor technology that substantially improves upon existing GaAs-based HBTs. Band-gap engineering with dilute nitride GaInAsN alloys is utilized to enhance a variety of key device characteristics, including lower operating voltages, improved temperature stability and increased RF performance. Furthermore, GaInAsN-based HBTs are fully compatible with existing high-volume MOVPE and IC fabrication processes. While poor lifetimes have limited the applicability of dilute nitride materials in photovoltaic applications, we achieve minority carrier characteristics that approach those of conventional GaAs HBTs. We have found that a combination of growth algorithm optimization and compositional grading are critical for improving minority carrier properties in GaInAsN. In this work, we characterize the impact of both carbon and nitrogen doping on minority carrier lifetimes in GaInAsN base layers. Minority carrier lifetimes are extracted from direct measurements on bipolar transistor device structures. Specifically, lifetime is derived from the DC current gain, or β, taken in the bias regime dominated by neutral base recombination. Lifetimes extracted using this technique are observed to be inversely proportional to both carbon and nitrogen doping. As with conventional C-doped GaAs HBTs, current soaking (i.e. burn-in) is found to have a significant impact on GaInAsN HBTs. While we can replicate poor as-grown lifetimes consistent with those reported in photovoltaic dilute nitride materials, our best material to date exhibits nearly 30 × higher lifetime after current soaking.

High volume HBT (Heterojunction Bipolar Transistor) production benefits from non-destructive measurement techniques to verify the quality of every product wafer. X-ray diffraction is non-destructive and sensitive to layer thicknesses and compositions. The thin layers typical of HBT designs yield weak diffraction intensity, which are overwhelmed by the signal from the substrate in the conventional (004) orientation. We have found that by diffracting from the (117) planes, much more information can be obtained. As the HBT layer peaks move away from the substrate signal this enables the in-line measurement of HBT product wafer characteristics, such as base doping. In this paper, we develop a new curve fitting methodology, and demonstrate the precise measurement of the base doping of HBT wafers in the production line for the first time. This technique is demonstrated to measure base doping changes in production as small as 1.8%.

The strain relaxation behavior of Si0.82Ge0.18 films on silicon-on-insulator (SOI) substrates was investigated for films grown beyond the critical thickness and strain-relaxed during growth and metastable films, grown beyond the critical thickness, which relaxed during subsequent thermal annealing. The thickness of the top silicon layer of the SOI substrate was varied over a range from 40 nm to 10 mum. In all cases, the SiGe film relaxation occurred via the nucleation and propagation of dislocations with the same onset of film relaxation and same relaxation rate for both SOI and bulk Si substrates. The SOI substrate does not serve as a compliant substrate but does alter the dislocation structure and motion. The buried amorphous oxide layer in the SOI substrate leads to the relaxation of the dislocation strain field through the removal of the dislocation line tension. This removal of the dislocation line tension drives dislocation motion and leads to the development of strain in the thin Si layer of the SOI substrate. Models of this dislocation behavior for SiGe growth on the SOI substrate are presented and calculation of the equilibrium strain of the thin Si substrate layer closely fits the measured strain of several SOI substrates. The article addresses the implications of the modified dislocation structure and kinetics for film relaxation on SOI substrates. (C) 2003 American Institute of Physics.

Data presented herein demonstrate that the DC current gain of GaInP/GaInAsN/GaAs DHBTs is almost independent of temperature over the range 225[ T[ 475 K and is relatively insensitive to current density for Jc / 1 A/cm 2 . Direct comparisons are made between DHBTs with different GaInAsN base alloy compositions and grading schemes and a high-performance GaInP/GaAs HBT.