Single-crystal – oriented-crystal – and epitaxy growth processes; – Forming from vapor or gaseous state – With a step of measuring – testing – or sensing
Reexamination Certificate
2001-02-22
2004-04-27
Kunemund, Robert (Department: 1765)
Single-crystal, oriented-crystal, and epitaxy growth processes;
Forming from vapor or gaseous state
With a step of measuring, testing, or sensing
C117S086000, C118S715000, C118S716000
Reexamination Certificate
active
06726767
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to layer processing and to a method and a system therefor, and more particularly (although not exclusively) to such a method and system for processing semiconductor material layers.
2. Discussion of Prior Art
In techniques for processing materials in thin layers inter alia for integrated circuits, there is a well-known problem of controlling linear dimensions (eg layer thickness or etch depth) and chemical composition in real time, ie during growth of a layer or while etching a surface. This problem is particularly relevant to growing layers of materials by low pressure vapour phase epitaxy (LPVPE). A degree of control over growth is available by controlling the relative proportions of partial pressures of constituent gases in a LPVPE source gas stream to be decomposed to produce a deposited layer. The prearranged sequence of gas mixtures, substrate temperatures and growth times is included in a growth recipe, the proportions of which should at least approximately be preserved in a layer grown from it. However, in practice there is drift in the calibration of gas flow control apparatus, which means growth diverges from the prearranged recipe, and the composition of a growing layer can alter unless there is some means for monitoring and controlling the layer composition and thickness in real time. Unfortunately it is not possible to make direct measurements of the parameters of layer thickness and chemical composition in real time during growth. To do so it is necessary to interrupt growth of a specimen and remove it from growth apparatus, which is most unattractive because it is time consuming, it interrupts the growth process and it may contaminate the layer being grown.
As a partial and indirect approach to solving the problem of monitoring chemical composition and linear dimensions in real time, it is known to use spectroscopic ellipsometry. Ellipsometry can be used for measurements on a growing specimen or a specimen being etched, but it does not give the necessary composition and thickness information directly. It uses reflection of light from a specimen surface to give optical parameters, but these involve a convolution of dimensions and refractive index which cannot be separated. This problem occurs in the growth of alloys such as silicon germanium alloy (Si
1−x
Ge
x
), where it is important to have accurate control over the alloy composition parameter x as well as layer thickness. It is particularly important in the growth of superlattices where the composition parameter x in a material system such as Si
1−x
Ge
x
alternates between successive layers in the region of eg 100 Angstroms thick.
However, it is possible to infer dimensions and composition from ellipsometric measurements combined with information from the chemical process taking place, eg a model of the process of a material being etched or of LPVPE layer growth derived from the gas mixture recipe. This again leads to further difficulties in practice because successive ellipsometric measurements taken at time intervals in the region of 1 or 2 seconds may be inaccurate or “noisy”, and do not necessarily give acceptable results for layer process control except under favourable circumstances.
In Appl. Phys. Lett. 57 (25), December 1990, Aspnes et al described optical control of growth of Al
x
Ga
1−x
As by organometallic molecular beam epitaxy. They disclosed a closed-loop control system for epitaxial growth of a homogeneous semiconductor crystal using monochromatic ellipsometry. The system related to homogeneous growth of a single layer where composition was controlled to remain constant. There was no disclosure of control of layer thickness.
In Thin Solid Films, 220, 1992, Urban reported development of artificial neural networks for real time in-situ ellipsometry data reduction They described monochromatic ellipsometry for a single homogeneous layer grown upon a substrate. The neural network was trained to provide seed values for an iterative model fitting routine which fit the layer composition and thickness parameters to the measured ellipsometric angles. There was no disclosure of using the resulting estimate to control growth.
In Thin Solid Films, 223, 1993, Johs et al describe using multi-wavelength ellipsometry for real-time monitoring and control during growth of CdTe by metal-organic vapour phase epitaxy (MOVPE). They disclosed making ellipsometry measurements at twelve wavelengths in less than three seconds. They introduced the virtual-interface method for determining the characteristics of the near-surface region of the growing crystal. They wished to estimate the rate of growth of homogeneous material (constant composition). The estimates of near-surface layer composition were obtained by fitting the parameters of the virtual-interface model (dielectric constants and layer depth) using an iterative model fitting algorithm. This relied upon using the algorithm to fit layer composition parameters to the most recent ellipsometry measurements.
In Applied Surface Science 63 (1993) pp 9-16, Duncan and Henck describe an etching process using a specimen with a known refractive index; ellipsometric measurements then gave thickness or etch depth directly. The specimen was SiO
2
2000 Angstroms thick upon an Si substrate. The etch depth measurements had uncertainties in the range 3 to 23 Angstroms. Measurements were made every 2 seconds approximately, and about 100 seconds were needed to etch through the specimen, so the incremental etch depth between measurements was 40 Angstroms. In consequence the uncertainty in the incremental etch depth varied between 15% and 57%, despite the SiO
2
/Si system being favourable for ellipsometric measurements; these materials have very different refractive indices of 1.4 and 3.9 respectively at 2 eV, and they are therefore easily discriminated by optical measurements.
In layer growth processing of compounds, eg alloys such as Si
1−x
Ge
x
, it is desirable to determine the thickness and composition of the layer contribution grown between successive pairs of ellipsometric measurements at intervals of 1-2 seconds. Si
1−x
Ge
x
, is grown at a rate of about 1 Angstrom per second, so layer contributions are 1-2 Angstroms thick. Since the composition of the layer contribution is unknown so also is the refractive index, and therefore the thickness cannot be determined directly. Si and Ge have similar refractive indices, eg 3.9 and 4 at 2 eV, and consequently the refractive index of Si
1−x
Ge
x
, is not very sensitive to changes in x and ellipsometric measurements give more inaccurate results than those for the SiO
2
/Si system.
In Diagnostic Techniques for Semiconductor Materials Processing II, Pang et. al. Eds, pp 87-94, Materials Research Soc., Pittsburgh, Pa. 1996, Vincent et. al. presented a method for in-situ estimation of etch rate using an extended Kalman filter based method for multi-wavelength reflectometry. Its possible application in real-time control was referred to but implementation details were not disclosed.
In the International Conference on Characterisation and Metrology for VLSI Technology, Gaithersburg, Md. USA, March 1998, Pickering et al disclose real-time process control using spectroscopic ellipsometry for Si/SiGe epitaxy. Si
1−x
Ge
x
was grown with x in the range 0 to 0.2—ie variable composition. They discussed the composition/growth rate correlation problem required for control of very thin near-surface layers. It was suggested that a principal component analysis algorithm might be used to obtain an estimate of growth rate which is independent of alloy composition. Moreover an analysis of composition based on an artificial neural network algorithm was given which was compared with SIMS data; after adjustment by scaling to allow for the lack of growth rate data given by this approach, a discrepancy of 0.02 for x in the range 0 to 0.2 was obtained, ie an error of at least 10% even when scaled.
SUMMARY OF THE INVENTION
It is an object of the
Dann Allister W. E.
Glasper John L
Marrs Alan D
Pickering Christopher
Robbins David J
Kunemund Robert
Nixon & Vanderhye P.C.
QinetiQ Limited
LandOfFree
Layer processing does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Layer processing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Layer processing will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3248402