The CW CO2 laser annealing is an effective method for removing deep level defects induced by ion-implantation. The available laser density range is 350˜600 W/cm2, which is consistent with that needed to fully activate implanted impurities. At high power density (>600W/cm2 ) slip dislocations can be induced. The energy spectra of dislocations were measured with DLTS technique. One peak at Ev + 0.33eV in p-Si and two at peaks at Ec −0.37eV and Ec −0.50eV in n-Si have been found. These peaks are stable during thermal annealing at high temperature. But they can be passivated with hydrogen plasma annealing and reactivated by subsequent thermal annealing in vacuum. The dislocation core is reconstructed.

Five different RTP units have been examined. These are the Varian RTP-800, the Eaton ROA-400, the A.G. Associates 2101/2106, the Tamarack 180A-C and the Varian IA-200. Each system is a cassette-to-cassette serial processor designed for use in a high volume semiconductor fabrication line. These units will heat a wafer from room temperature to 400–1400C in times on the order of a few seconds. Each unit uses radiation to heat and cool the wafer. The different radiation sources, wafer handling systems, temperature measurements and control computers are discussed. The control of slip and nonuniform temperatures is presented as well as information regarding the types of ambients that can be used in each system.

Rapid Thermal Processing (RTP) of silicon or other semiconductor materials requires accurate measurement and control of temperature. In a typical RTP cycle, heating of the wafer takes place in seconds, making accurate control of the wafer temperature very critical. Non contact wafer temperature sensing is achieved using an optical pyrometer. Precise temperature control from 400° C to 1350° C is maintained with a closed loop control system consisting of an optical pyrometer and a computer. Sources of errors due to variations in emissivity as a function of wafer temperature, surface conditions and background radiation are discussed. Calibration of the system is achieved by using a thermocouple instrumented wafer.

When high temperature processes are being carried out in a diffusion furnace, temperature control is accomplished in a straight forward manner. By using thermocouples and a proportional controller, sufficient accuracy and control are available to the process engineer.

But with the current shift toward the more rapid annealing schemes, better control is a necessity. In order to achieve this accuracy, especially when the heat source is a 100 kilowatt arc lamp, more sophisticated control methods must be employed. By using an optical pyrometer tuned to operate at a wavelength where external interferences are minimized, wafer temperature may be monitored to a high degree of accuracy and with the fast response times needed.

Coupling the output of the pyrometer to a microprocessor based controller allows the engineer to create and use multi stepped time/temperature profiles which are highly repeatable.

This paper will describe such a system and how it is used in a production oriented rapid optical annealing system.

Pyrometers are used to measure wafer temperature in rapid thermal processors. This paper focuses on the effect of thin dielectric films on the accuracy of pyrometric temperature measurement. The thin films affect the emissivity of the wafer, which in turn affects the radiation emitted by the wafer. Wafers with different oxide thicknesses were monitored by both pyrometer and thermocouple. Temperature errors of the order of 10°C to 50°C were found for wafers of differing oxide thicknesses. The chamber reflectivity was found to be an important factor in determining the ‘effective’ emissivity of the wafer. Pyrometer recalibration is necessary to accurately measure wafer temperature.

Short time anneal (STA) experiments were performed in a specially designed system which uses either a bank of Tungsten-Halogen lamps or a moving Mercury arc lamp as the light source. Both STA results were compared with conventional furnace anneal. The Mercury arc lamp anneal is found to be a critical process but optimization for wafer damage free conditions can be achieved. The anneals result in general in a better profile control which is most outspoken for shallow Boron implants while the electrical activation is comparable for all elements to that obtained with Tungsten-Halogen or furnace anneal.

The rapid thermal annealing (RTA) of p+n and n+p diodes, fabricated by the LOCOS process, and its subsequent effects on junction leakage current, junction depth and dopant activation were investigated. The reverse bias diode leakage currents of implanted Si <100> samples (As+: 60 KeY, 5×1014 5×1015 cm−2, B+: 25 KeV, l×1014, l×1015 cm−2 and BF2+: 45 KeV, 1×1015cm−2 ) were measured as functions of annealing temperature, and dwell time. The annealing was performed using an Eaton RTA system (Nova ROA-400) at temperatures ranging from 950 °C to 1150 °C. Annealing times ranged from 0.2 sec. to 10 sec. The results from the diode leakage current analysis are correlated with those from Secondary Ion Mass Spectroscopy (SIMS) and differential Hall measurements. The reverse-biased leakage currents from the RTA-treated samples are compared with those from furnace-annealed samples.

This study describes the use of a steam environment to reflow phosphosilicate glass (PSG) samples using a HEATPULSE® rapid thermal annealer. The samples comprised PSG over poly steps and of open contacts in PSG. It was observed that reflow occurs 50°C lower in steam than in dry O2. An acceptable flow cycle for 8 w/o P in PSG glass is 1050°C for 10 seconds in steam, while for 6 w/o P PSG it is 1100°C for 10 seconds. Steam is found to be an effective amibient for densification of the PSG film. The thermal oxide grown in the contact during opening reflow was determined to be near 140 A. The operating regime for a junction depth <0.4 um and a reflow angle < 75° is presented for 8 w/o P.

Double-diffused shallow junctions have been formed by ion implantation of both phosphorus and arsenic ions into silicon substrates and rapid thermal annealing. Experimental results on defect removal, impurity activation and redistribution, effects of Si preamorphization, and electrical characteristics of Ti-silicided junctions are presented.

A study on the rapid isothermal fusion (RIF) properties of borophosphosilicate glass (BPSG) films is presented. Conventional furnace fusion (FF) results are also reported for comparison purposes. The effects of varyinq the temperature (950°-1150°C for RIF, and 800°-1000°C for FF) and the cycle duration (5–60 s for RIF, and 7–60 min for FF) were investigated in a N2 ambient. Two deposition methods were used for the BPSG films, namely atmos- Dheric pressure CVD (APCVD) and low pressure CVD (LPCVD), both with the low temperature oxidation of SiH4, PH3 and B2H6 as a deposition reaction mechanism. The boron and phosphorus concentrations were maintained at 4 w/o for both species. Step coveraae fusion test structures were used to characterize the qlass flow. It was found that for both RIF and FF techniques, a considerable amount of fusion resulted from heating cycles which are compatible with the shallow Junctions required for VLSIC's. With BPSG, the choice between RIF and FF can thus be based on criteria other than olass fusion, such as shallow junction annealing. Also, more fusion was obtained for LPCVD films than for APCVD films.

In the first part of this paper the role of rapid thermal annealing in sulicide processing is viewed from a theoretical point of view with respect to what is known about metal-silicon reactions. The second part is a brief survey of the quickly expanding literature on the rapid thermal annealing of silicides. Whereas RTA does not appear to solve all, or perhaps even any, of the problems encountered in the use of silicides, it is concluded that RTA presents a number of definite advantages.

Rapid thermal annealing is used to form cobalt silicide directly on unimplanted as well as B, As, and P implanted wafers. The films are characterized by sheet resistance, X-ray diffraction, SEM, SIMS, and contact resistance measurements. The direct silicidation of cobalt on Si by rapid thermal annealing yields smooth, low resistivity films with minimal dopant redistribution.

Sputter deposited Ti films on Si and SiO2 substrates have been rapid thermal annealed in N2 at temperatures of 400–1100°C, and the reaction followed using AES, TEM, electron diffraction and sheet resistance measurements. The Ti initially becomes contaminated with oxygen before being nitrided at the surface and silicided at the interface. The oxygen is expelled from the siuicide and a TiNxO1−x/TiSix bilayer eventually results. With Si substrates, TiNxO1−x is much the thinner layer whereas the reverse is true on SiO2. Extended annealing in N2 completely converts the TiSi2 layer on Si into TiN.

Tungsten silicide samples were formed by sputter depositing 80 nm W metal onto (100) oriented, 5 ohm-cm Si wafers. After deposition, the samples were fast radiatively processed in an RTA system using quartz-halogen tungsten lamps as radiation sources for time intervals ranging from 20 to 60s under high vacuum. Films processed at 22–25 W/cm2 radiation with the film side of the samples oriented away from the lamps result in films which are metallic or cloudy in color, and have mixed composition as evidenced by x-ray diffraction (W, W5 Si3 and WSi2). Films processed with the film side oriented toward the lamps show the occurrence of a phase transforF.ation clearly nucleated at the film edge. The new phase is grey and mostly composed of WSi2, with film resistivities 55–60 μΩ-cm. AES measurements show a uniformly distributed Si diffusion profile into the metal film, while SEM indicates a nucleation and growth of the phase once diffusion occurs.

W-Ti bilayers on Si have been processed by fast thermal annealing in air, in N2/H2 (5%), and vacuum (1 E-3 and 1 E-6 Torr). The results indicate that the W layer on top of Ti acts as an effective protection barrier against oxidation of Ti in all atmospheres except air. Titanium-rich silicides are formed after 500 °C, while TiSi2 is completed at about 600 °C. At 700 °C a bilayer of WSi2/TiSi2 is found. Silicon is found to be the diffusing specie during the formation of WSi2. Finally, a ternary silicide, Ti0.6W0.4Si2, starts to form after 780 °C. The final ternary silicide phase has'an Mlectrical resistivity of about 60 μΩ-cm.

Limited reaction processing has been used to grow thin, high quality layers of epitaxial silicon, silicon dioxide, and polycrystalline silicon. Multiple layers of these materials were deposited in – situ to demonstrate control of layer thicknesses, interface transition regions, and interface cleanliness. MOS and p-n junction structures were fabricated which exhibit electrical characteristics comparable to those of devices produced by conventional techniques.

Single crystal silicon wafers have been oxidised by exposure to an oxygen ambient at atmospheric pressure during RTP using an A G Associates Heatpulse 2101 Rapid Thermal Annealer. Wafers of the standard orientations used in silicon device processing <100>, <111> and <110> were studied. Oxidation was carried out in the temperature range 900 to 1250°C for times of RTP from 4 to 25U seconds producing oxides up to 60nm in thickness. Oxidation rates and their orientation and temperature dependence were derived from ellipsometric measurements of oxide thickness. X-ray photoelectron spectroscopy (XPS) and infra-red absorption spectrophotometry were also employed in the oxide characterisation. Preliminary electrical characterisation of the oxides, investigated using MOS C-V analyses, showed that the interface state densities were comparable to those in conventional furnace grown oxides.

The uses of Rapid Thermal Annealing or Rapid Thermal Processing (RTP) have been expanding beyond the original post implant annealing. RTP has been used to reflow low temperature oxides (PSG or BPSG), anneal silicides and to sinter contacts. One application of RTP which is beginning to receive attention is the growth of oxides or nitrides of silicon.

This paper will examine the use of a commercial rapid thermal processing system based on a very high power water-wall DC arc lamp to grow oxides on silicon wafers. The work includes a study of the growth rates of oxides at different temperatures. Direct feedback control of wafer temperature and high ramp-up and cool-down rates are used to minimize the effects of temperature errors or “tails” in the temperature/time profiles. Ellipsometry is used as the primary measurement tool to characterize the oxide films.

In addition to using a pure, dry oxygen atmosphere, several oxygen-argon mixtures are used. The effects of atmosphere on the growth rate of the oxide film are reported.

In order to become a practical application of RTP, oxide growth must be accomplished uniformly and reproducibly. These characteristics are machine-dependent. The uniformity of films grown in this system are discussed. The growth of oxide films and the uniformity measurements are used as an indirect technique to characterize the uniformity of the system. The reproducibility of film thickness is also examined.

This paper reports the results of a study of thin SiO2 growth and annealing in a lamp heater (Heatpulse). Good quality 5–10 mn SiO2 films on silicon have been obtained. For good oxides the average destructive breakdown- field was 13 MV/cm and very few or no low-field breakdowns were found. The breakdown statistics were obtained with Al gate capacitors and by a current staircase-ramp technique which is described in the paper. Processing with a single oxidation step, oxidationannealing, and oxidation-annealing-oxidation are discussed.

Rapid thermal processing is a promising tool for studying the kinetics of interface state annealing and other process phenomena on short time scales. We have studied the decay of interface states with a variety of ambients, temperatures, and oxide thicknesses. Annealing kinetics appear to be controlled by a surface reaction process, and not hydrogen diffusion through the oxide. The annealing behavior depends strongly on temperature but less so on other process parameters. Our experimental methodology for temporal process modeling is discussed.