Friday Afternoon Sessions (June 28) TMS Logo

About the 1996 Electronic Materials Conference: Friday Afternoon Sessions (June 28)



June 26-28, 1996 · 38TH ELECTRONIC MATERIALS CONFERENCE · Santa Barbara, California

Session DD: Real Time Monitoring and Control

Session Chairman: Henry P. Lee, 2231 Engineering Gateway, Department of ECE, University of California, Irvine, CA 92715. Co-Chairman: J.M. Kuo, AT&T Bell Laboratories, Crawfords Corner Road, Murray Hill, NJ 07733-1988

1:30PM, DD1

"Substrate Temperature Measurement During MBE Growth of Narrow Bandgap Semiconductors on Freely-Mounted Substances:" T.J. DE LYON, J.A. Roth, D.H. Chow, Hughes Research Laboratories, 3011 Malibu Canyon Road, Malibu, CA 90265

Measurement and control of substrate temperature is of critical importance when In-free substrate mounting is employed during MBE growth. An especially challenging problem arises when a film with bandgap narrower than the substrate is grown. In this case, the substrate temperature can increase significantly due to increased absorption of the substrate heater radiation by the epilayer/substrate structure [1]. Measurement of substrate temperature by pyrometry in this situation. i.e., during the interval when the epilayer is making a transition from semi-transparency to opacity, is frustrated by the changing optical properties (e.g. emissivity) of the epilayer/substrate combination. In addition. many narrow bandgap semiconductors such as GalnSb and HgCdTe, which are used for infrared emitters and detectors, are grown at temperatures sufficiently low that pyrometry is impractical.

Absorption edge spectroscopy (ARES) has been investigated as an alternative means of remote sensing of substrate temperature during MBE [2]. This technique is capable of measuring substrate temperatures much lower than can be determined with conventional pyrometry and so is ideally suited, in principle. to measurement of substrate temperature during low-temperature (180-450deg.C) MBE growth of narrow-bandage semiconductors. This method relies on sensing of shifts in the step in the substrate reflectivity spectrum at the substrate bandedge. While ABES works well to determined the temperature of a bare substrate, deposition of an absorbing overlayer severely attenuates the step in substrate reflectivity and eventually renders the step undetectable as the absorbing film thickness increases.

To enable ABES to successfully measure substrate temperature throughout the entire course of low-temperature MBE growth of narrow bandgap epilayers, we have developed a technique involving insertion of an optically reflecting layer, such as a quarter-wave stack, between the substrate and narrow bandgap epilayer. When ARES is implemented in reflection mode from the substrate backside, this reflecting layer prevents analysis radiation from penetrating into the growing narrow bandgap layer, thus ensuring a strong reflected signal that carries the absorption edge signature of the substrate. We will present specific examples of substrate temperature measurement, using this reflecting interlayer technique, for the cases of optically opaque films of GaSb on GaAs and HgCdTe on CdTe. Substrate temperatures from 200-360deg.C have been successfully measured using this approach.

[1] B.V. Shanabrook, J.R Waterman, J.L. Davis. and R.J. Wagner, Appl. Phys. Lett. 61, 2338 (1992).
[2] J.A. Roth. T.I. de Lyon. and M.E. Adel, Mat. Res. Soc. Symp. Proc. 324. 353 (1994).

1:50PM, DD2

"Measurement and Control of the Substrate Temperature During MBE of Lattice-Matched In.53Ga.47As on InP:" J.A. ROTH, D.H. Chow, J.-J. Dubray, 3011 Malibu Canyon Road, Hughes Research Laboratories, Malibu, CA 90265

Currently there is a great deal of interest in devices, such as heterojunction bipolar transistors (HBTs) and resonant tunneling diodes (RTDs), which employ lattice-watched InGaAs layers grown on InP substrates, for application to high-speed nanoelectrodes. Integrated circuits which utilize stacked HBT/RTD structures are also currently being fabricated. The material requirements for these devices include the growth of thick epitaxial layers of heavily-doped In.53Ga.47As, and in stacked structures, the total InGaAs thickness can exceed 2um. The measurement and control of the true substrate temperature daring growth of these layers is of critical importance to ensure a high degree of lattice matching with optimum material quality. The problem of temperature control is particularly challenging since substrate temperature rises by 100deg.C under constant heater power-conditions due to the increased absorption of heater radiation by the epitaxial film, and this rise must be measured and compensated in real time during the growth.

We presently experimental results showing that absorption-edge spectroscopy (ABES), implemented in the transmission mode, can he used to sense and control the true substrate temperature during growth of InGaAs, and the ABES signal can be used in a feedback algorithm to maintain constant substrate temperature. Measurements were performed using a Fisons V8OH MBE chamber equipped with an "in-line" substrate Manipulator modified to provide optical access to the back side of the substrate. Data will be presented comparing the substrate temperature history during growth of InGaAs using constant power, constant thermocouple temperature, and with constant substrate temperature controlled by feedback from the ABES sensor. Apparent temperature shifts due to distortion of the InP absorption-edge spectrum by the presence of an InGaAs overlayer have been determined and appropriate corrections made in the temperature control algorithm. The observed rise in substrate temperature when InGaAs growth is conducted at constant heater power or constant thermocouple temperature is explained using a steady-state model of the radiant energy transfer between the heater and the substrate which accounts for the change in the optical properties of the substrate overlayer system that occurs during InGaAs deposition.

1:50PM, DD2

"Measurement and Control of the Substrate Temperature During MBE of Lattice-Matched In.53Ga.47As on InP:" J.A. ROTH, D.H. Chow, J.-J. Dubray, 3011 Malibu Canyon Road, Hughes Research Laboratories, Malibu, CA 90265

Currently there is a great deal of interest in devices, such as heterojunction bipolar transistors (HBTs) and resonant tunneling diodes (RTDs), which employ lattice-watched InGaAs layers grown on InP substrates, for application to high-speed nanoelectrodes. Integrated circuits which utilize stacked HBT/RTD structures are also currently being fabricated. The material requirements for these devices include the growth of thick epitaxial layers of heavily-doped In.53Ga.47As, and in stacked structures, the total InGaAs thickness can exceed 2um. The measurement and control of the true substrate temperature daring growth of these layers is of critical importance to ensure a high degree of lattice matching with optimum material quality. The problem of temperature control is particularly challenging since substrate temperature rises by 100deg.C under constant heater power-conditions due to the increased absorption of heater radiation by the epitaxial film, and this rise must be measured and compensated in real time during the growth.

We presently experimental results showing that absorption-edge spectroscopy (ABES), implemented in the transmission mode, can he used to sense and control the true substrate temperature during growth of InGaAs, and the ABES signal can be used in a feedback algorithm to maintain constant substrate temperature. Measurements were performed using a Fisons V8OH MBE chamber equipped with an "in-line" substrate Manipulator modified to provide optical access to the back side of the substrate. Data will be presented comparing the substrate temperature history during growth of InGaAs using constant power, constant thermocouple temperature, and with constant substrate temperature controlled by feedback from the ABES sensor. Apparent temperature shifts due to distortion of the InP absorption-edge spectrum by the presence of an InGaAs overlayer have been determined and appropriate corrections made in the temperature control algorithm. The observed rise in substrate temperature when InGaAs growth is conducted at constant heater power or constant thermocouple temperature is explained using a steady-state model of the radiant energy transfer between the heater and the substrate which accounts for the change in the optical properties of the substrate overlayer system that occurs during InGaAs deposition.

2:10PM, DD3+

"Real-Time, Non-Invasive Temperature Control of Wafer Processing Based on Diffuse Reflectance Spectroscopy:" ZHONGZE WANG, Sui L. Kwan, T.P. Pearsall, J.L. Booth, B.T. Beard, S.R. Johnson, Department of Electrical Engineering, University of Washington, Seattle, WA 98195; Thermionics Northwest Inc., Port Townsend, WA 98368; Department of Physics, University of British Columbia, Vancouver, B.C. V6T 1Z1

We demonstrate a new real-time temperature control system for semiconductor wafer processing based on direct, non-invasive, in-situ measurement of wafer temperature by diffuse reflectance spectroscopy. We tested the new control system over a temperature range from 25deg.C to 600deg.C for semi-insulating GaAs wafers. The stability and precision of the controlled wafer temperature were maintained within +/-1deg.C with an update time of 2 seconds. Compared to the traditional temperature control technique where the wafer heater temperature is controlled, the direct control of wafer temperature using diffuse reflectance spectroscopy is shown to be a significant improvement in accuracy, stability, repeatability and response rate, for dynamic control of wafer temperature in the range of 25deg.C to 600deg.C. Our technology can also be applied to Si and other semiconductor wafers following proper calibration. We also discuss the large discrepancy between the behavior of the heater temperature and the wafer temperature during the heating process. Our results show that the wafer temperature response rate to the heater temperature is very sensitive to wafer temperature. This implies that during the direct control of wafer temperature, the control parameters, such as the PID settings, should be adjusted to maintain the best response rate at different wafer temperatures.

2:30PM, DD4+

"A Temperature-Dependent Model for the Complex Dielectric Function of GaAs for Growth Control:" LEONARD I. KAMLET, Fred L. Terry, Jr., Department of Electrical Engineering & Computer Science, 111 DTM Bldg., The University of Michigan, 2360 Bonisteel Blvd., Ann Arbor, MI 48109-2108; George N. Maracas, Motorola Phoenix Corporate Research Laboratories, 2100 E. Elliot Road, MD EL508, Tempe, AZ 85284

Nondestructive optical techniques such as spectroscopic ellipsometry or reflectometry provide information about a material's thickness and optical properties. These methods do not interfere with most semiconductor processes, and are currently being used as in situ monitors in many growth, deposition, and etch chambers. With appropriate feedback algorithms, they can also be used for real-time control of semiconductor processing. To be used effectively in these applications, appropriate models must be created for the complex dielectric function versus wavelength and material properties.

One property often overlooked in the modeling of a semiconductor's dielectric function is the dependence on temperature of the sample. For ex situ measurements, from which most models are constructed, this is not usually considered to be an issue. For in situ growth applications, it is crucial that models correctly account for the temperature in order to obtain accurate thickness information.

In this work, we discuss a temperature-dependent model for the complex dielectric function for GaAs valid for the temperature range 31deg.C <=T <= 634deg.C. We describe our model, which is an extension of the critical point parabolic band method and which has been previously demonstrated for composition-dependent models of the dielectric function for lattice-matched materials systems. One advantage of this model is that it is based on the physical processes occurring in the semiconductor while providing an excellent fit to the data. We demonstrate the quality of the model in fitting optical data for individual temperatures, and compare our results to other established models. The data used for each fitting ranges from 1.25 to 4.5 eV.

Using results obtained from the individual fits, we generate a temperature-dependent model that is valid for the range of temperatures given above. Also, we show how this model can be used to accurately determine the temperature (+/-5deg.C) of a material whose dielectric response has been obtained but was not included when generating the model. Extension of this work to simultaneously solve for the composition and temperature of AlGaAs will be discussed.

*This material is based on work supported by an ARPA program on smart ECR etching, and with support from National, Science Foundation Graduate Fellowship.

2:50PM, DD5

"Real-Time Monitoring of SiNx Sidewall Technology:" W. PROST, C. van den Berg, S. van Waasen, F.J. Tegude, Gerhard-Mercator-University Duisburg, Sonderforschungsbereich SFB 254, Solid-State Electronics Department, Kommandantenstr. 60, D-47057 Duisburg, Germany

High performance III/V-semiconductor devices with non-planar topologies require optimized metallization techniques with insulating sidewall layers in the nm-thickness regime [e.g. 1]. In this contribution we report on fast process development and reliable manufacture of insulating thin sidewall layers based on real-time in-situ monitoring using ellipsometry.

We use a PE-CVD system with ECR source allowing both deposition [2] and subsequent etching of SiNx. Both processes are carried out at room-temperature with excellent quality and may be combined with photo-resist patterning. N2, SiH4 and a CF4/O2 mixture are used for deposition and etching, respectively. Sidewall layers are formed due to an high isotropy difference between conformal deposition and vertical etching. At a pressure of 5 mTorr a conformal and nearly isotropic (1:2) deposition of vertical device mesa's is assured. Anisotropic ([[congruent]]1:20) etching is obtained in a flow-controlled process (p <2 mTorr). It is important to note, that we have established both processes in one run in a remote plasma system by means of process pressure handling, only. In comparison to the conventional approach [e.g. 1] no bias is applied and hence the plasma induced damage is minimized. On the other hand this process mode is more sensitive to random process fluctuations and an in-situ real-time monitoring system is highly required. We have installed a single-wavelength rotating-analyzer ellipsometer for accurate high speed on-line process monitoring. A brief description of the system including data calibration and important details for reliable measurements will be given. The contribution of this system to process development and control is manifold. We will report on multiple deposition and etch rate determination during one run. Random and systematic process failures are identified on-line and corrected. E.g. the influence of etch rate enhancement due to inert gases is analyzed in detail. With respect to the sidewall process a very high accuracy for the end-point detection is of pronounced importance We will show that using our system an accuracy better than 3 nm is easily attained which results in a reliable as well as a damage-free process. Finally, the application of the realized sidewall layers for vertical devices like HBT and RTD will be outlined.

______________________________________

[1]Randall J.N., Newell B.L. "Fifteen namometer features by sidewall processing and pattern transfer", J. Vac. Sci. Tech. Bl2(6).3631-3634, Nov./Dec. 1994.

[2]Wiersch A., Heedt C, Schneiders Tilders R.. Kuebart W., Frost W,. Tegude F.J. "Room-Temperature Deposition of SiNx, using ECR-PECVD for III/V Semiconductor Microelectronics in Lift-Off Technique", Journal on Non-Crystalline Solids. 187(7), 334-339(1995).

3:30PM, DD6

"Continuous In Situ Growth Rate Extraction Using Pyrometric Interferometry and Laser Reflectance Measurements During Molecular Beam Epitaxy:" J.J. ZHOU, P. Thompson, H.P. Lee, Department of Electrical and Computer Engineering, University of California, Irvine, CA 92715; Y.C. Kao, F.G. Celii, Corporate Research & Development, Texas Instrument, Dallas, TX 75265

Precise and reproducible control of epitaxial layer thickness and composition is crucial to the manufacturability of many heterostructure-based electronic and photonic devices. It is highly desirable to develop real-time in situ monitoring methods and data analysis algorithms that yield information on the cumulative thickness and instantaneous growth rate of the deposited materials. In situ narrow-band pyrometric interferometry and laser reflectance are two such monitoring methods that have been used for such applications. Until recently, growth rates were determined by measuring the period (T) of the quasi-sinusoidal monitoring signal from a layered sample using Gr =[[lambda]]m/(2nTcos[[theta]]) where lm is the monitoring wavelength, [[theta]] is propagating angle of light through the materials, and n is the refractive index of the growing layer at the growth temperature. To reduce statistical variance, a relatively thick calibration layer covering a number of oscillation periods is generally required. From the data processing point of view, this method is inefficient since only data at the peak and valley of the monitored signal are being sampled for growth rate measurement. Recently, we have developed a least square estimation method in which the quasi-sinusoidal monitored waveform is treated as an amplitude and phase modulated signal [1,2]. The progression of-889-7 the layer thickness is expressed in terms of a continuous variable, namely the cumulative "signal phase" (in the unit of radian) which can be extracted from the monitored signal in real-time. The least square phase estimation (LSPE) method has shown to be particular effective for analyzing quasi-sinusoidal signals subject to amplitude modulation (arising from temperature variation in the PI measurement, and intensity drift in the reflectance measurement) and had demonstrated excellent growth reproducibility when used as end point detection for 990 nm AlAs/GaAs DBR structures (peak reflectance deviation < 2nm)[2]. To implement the LSPE scheme for growth rate extraction, a reference `signal phase' is first obtained either from calculated monitoring signals, or empirical monitoring data from prior measurements in which the layer thickness has been verified by post growth measurement. The instantaneous growth rate, treated as a continuous variable is extracted in real-time based on the optimum fit between the experimental signal phase and the reference phase. Our results on AlAs/GaAs layered structures have shown that accurate growth rate measurements can be obtained with considerably shorter calibration layer thickness compared to the existing method. Using a thinner calibration layer not only reduces the time and cost for growth rates calibration, but on many cases may permit the incorporation of the calibration structure within the device structure itself.

____________________________________

1 X.Liu, E.Ranalli, D.L.Sato, Y.Li, and H.P.Lee, J. Vac. Sci.Tech. B13, 742. (1995).
2. H. P. Lee, Y. Li, D. Sato and J.J. Zhou, to appear in J. Vac. Sci and Technol. B. (May/June) 1996.

3:50PM, DD7

"Real-Time Composition Analysis During MBE Growth of Group IV and III-V Systems Using Reflection Electron Energy Loss Spectrometry:" CHANNING AHN, Gang He, Hiroshi Yoshino, Shouleh Nikzad*, Harry Atwater, Thomas J. Watson Laboratory of Applied Physics, 128-95 Caltech, Pasadena, CA 91125; *Center for Space Microelectronics Technology, Jet Propulsion Laboratory, Pasadena, CA 91109

Inelastically scattered electrons accompany and dominate the RHEED pattern which is typically used for surface structural analysis during MBE growth. Spectral analysis of this inelastic distribution of electrons enables us to determine surface composition during film growth by probing core loss ionization intensities of the incident RHEED electrons such as Ga and As L2,3 at 1115, and 1323 eV respectively, Si K at 1840 eV. For higher Z elements like Sn shallower M4,5 ionizations can be used. We have successfully analyzed GeSi and SnSi alloy compositions, as well as surface contaminant levels during MBE growth using a magnetic prism assembly, in which the dispersion of loss electrons are ramped across an energy selecting slit and detected with a scintillator photomultiplier tube.

We presently employ two different configurations for spectroscopic analysis. One system which is presently used for a III-V growth chamber is a serial electron energy analyzer which consists of a modified, differentially-pumped conventional prism mounted to the growth chamber with tilt and transitional capability. While the sample height and port geometries are fixed, variation of the incident angle at a particular sample position is accomplished by mechanical manipulators and beam steering is achieved by electromagnetic dipoles.

The second system which is used on our Group IV chamber is a parallel data acquisition system which measures the whole of the dispersion of loss electrons simultaneously by using a high dynamic range, CCD with which deep core losses can be measured and evaluated in times consistent with feedback typical for epitaxial growth. Matching of the dispersion to the diode-to-diode spacing of the CCD is accomplished through the use of post prism quadrupoles and the dispersion is projected onto a phosphor screen which is optically coupled to the detector. This configuration has a detective quantum efficiency of 0.4, enabling quantifiable data to be obtained at ~l hz.

4:10PM, DD8

"Closed-Loop MBE Control of Ternary Quantum Well Photoluminescence Energy Using an Atomic Absorption Flux Monitor:" PAUL PINSUKANJANA, Andrew Jackson, Jan Tofte, Kevin Maranowski, Larry Coldren, Arthur Gossard, ECE Department, Materials Department, University of California, Santa Barbara, CA 93106

Using a dual pass multi-channel atomic absorption Optical-based flux Monitor (OFM), we have demonstrated feedback controlled growth of InGaAs quantum wells. The OFM simultaneously monitors Ga and In fluxes, so growth rate and composition can be determined in real time. To obtain the desired photoluminescence (PL) energy, the OFM compensates for changes in composition during the quantum well growth by adjusting the quantum well total thickness.

The atomic absorption is measured by passing the atomic emission radiation from Ga and In hollow cathode lamps through the growth chamber. The OFM absorption signal is calibrated using Reflection High Energy Electron Diffraction (RHEED) oscillations. By simultaneously measuring the absorption signal and RHEED oscillations, the growth rate can be correlated to the absorption. The data were fined to a curve of the form: absorption = 1 - exp[-([[alpha]] + [[beta]]r)r], where r is the growth rate, and [[alpha]] and [[beta]] are the two fit parameters.

To generate an accurate model of the PL energy of an lnGaAs quantum well, E(x,w), as a function of composition (x) and thickness (w), we grew several quantum wells and measured the PL spectra. These data were fitted to a function of thickness and composition of the form: E(x,w)=A+Bx+(C+Dx)/ w2where A, B. C, and D are adjustable fit parameters. This model was then used to control the PL energy of InGaAs quantum wells during intentional ramping of the In cell temperature.

We have demonstrated the control of ternary quantum wells by the growth of two samples. The first sample consists of three InGaAs quantum wells grown with no feedback control (each well was grown for the same length of time). The composition was intentionally varied during growth by changing the indium cell temperature. The photoluminescence spectrum of the first sample shows three clearly separated peaks corresponding to the three quantum wells. Using the same temperature profile for the indium cell, a second sample was grown. The second sample also consists of three quantum wells, but feedback control of the shutters was used to maintain a constant PL energy. Instead of three separate peaks, the PL spectrum of the second sample shows a single strong peak corresponding to a superposition of the luminescence spectra from all three quantum wells. This demonstrates the ability of the technique to simultaneously monitor quantum well thickness and composition during real-time growth, resulting in structures with closely controlled optical properties.

4:30PM, DD9+

"Modeling and Algorithm Development for Automatic Optical Endpointing of an HBT Emitter Etch:" C.K. HANISH, L.I. Kamlet, S. Thomas III, J.W. Grizzle, S.W. Pang, F.L. Terry Jr., Department of Electrical Engineering and Computer Science, The University of Michigan, Ann Arbor, MI 48109-2122

This paper discusses the development of a high-accuracy, self-tuning endpointing algorithm for the emitter etch of a heterojunction bipolar transistor (HBT). Fabrication of high-performance HBT's using self-aligned base-emitter processes requires etching through the emitter layer and stopping with very high accuracy on the base layer. The lack of selectivity in dry etching coupled with the high etch rates possible in high density plasmas render the use of a standard timed overetch impractical, especially as device layers continue to became thinner. The etch process under study requires the complete removal of an AlInAs emitter while etching no more than 2 nm of the underlying GaInAs base layer.

Etch products can be monitored using either optical emission spectroscopy (OES) or mass spectrometry (MS) to determine etch endpoint. The process under study relies on the intensity of the 417.2 nm Ga emission line. The detection of the Ga line indicates that the etch has reached the GaInAs base layer. However, the presence of a time varying Ga baseline signal before endpoint and significant noise in the OES signal necessitate more than a simple threshold scheme for critical endpoint detection. The OES signal can he modeled as a line of unknown slope with independent Gaussian white noise with zero mean during the AlInAs etch when the etch reaches completion, the emission rises from the baseline level due to the presence of Ga in the plasma and can be modeled as a line with unknown positive slope plus independent Gaussian white noise with zero mean. An endpoint algorithm detects the resulting breakpoint in the signal.

The breakpoint can be detected by testing the hypothesis that the measured signal minus the unknown slope, which is estimated in real time, is independent Gaussian white noise with zero mean. Briefly, when k of m samples exceed a threshold, where k, m, and the threshold are chosen appropriately, the hypothesis is falsified and the etch is stopped. The threshold depends on the variance of the noise in the current run. which is also estimated in real time. This algorithm is robust to variance in the optical gains of the measurement equipment and is applicable to other etch processes. Other detection methods are also under consideration. Experimental tests of automated endpointing using this algorithm will be presented along with pre- and post-etch ex situ film thickness measurements.

This work was supported by an ARPA program on smart ECR etching, and with support from National Science Foundation Graduate Fellowships.

4:50PM, DD10

LATE NEWS


The information on this page is maintained by TMS Customer Service Center (csc@tms.org).

Search TMS Specialty Meetings Page TMS Meetings Page About TMS TMS OnLine