Classifications

• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L22/00—Testing or measuring during manufacture or treatment; Reliability measurements, i.e. testing of parts without further processing to modify the parts as such; Structural arrangements therefor
  • H01L22/20—Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
• • C—CHEMISTRY; METALLURGY
  • C07—ORGANIC CHEMISTRY
  • C07K—PEPTIDES
  • C07K14/00—Peptides having more than 20 amino acids; Gastrins; Somatostatins; Melanotropins; Derivatives thereof
  • C07K14/195—Peptides having more than 20 amino acids; Gastrins; Somatostatins; Melanotropins; Derivatives thereof from bacteria
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
  • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
  • G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
  • G01N21/71—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light thermally excited
  • G01N21/73—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light thermally excited using plasma burners or torches
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01J—ELECTRIC DISCHARGE TUBES OR DISCHARGE LAMPS
  • H01J37/00—Discharge tubes with provision for introducing objects or material to be exposed to the discharge, e.g. for the purpose of examination or processing thereof
  • H01J37/32—Gas-filled discharge tubes
  • H01J37/32917—Plasma diagnostics
  • H01J37/32935—Monitoring and controlling tubes by information coming from the object and/or discharge
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01J—ELECTRIC DISCHARGE TUBES OR DISCHARGE LAMPS
  • H01J37/00—Discharge tubes with provision for introducing objects or material to be exposed to the discharge, e.g. for the purpose of examination or processing thereof
  • H01J37/32—Gas-filled discharge tubes
  • H01J37/32917—Plasma diagnostics
  • H01J37/32935—Monitoring and controlling tubes by information coming from the object and/or discharge
  • H01J37/32963—End-point detection
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01J—ELECTRIC DISCHARGE TUBES OR DISCHARGE LAMPS
  • H01J37/00—Discharge tubes with provision for introducing objects or material to be exposed to the discharge, e.g. for the purpose of examination or processing thereof
  • H01J37/32—Gas-filled discharge tubes
  • H01J37/32917—Plasma diagnostics
  • H01J37/32935—Monitoring and controlling tubes by information coming from the object and/or discharge
  • H01J37/32972—Spectral analysis
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61K—PREPARATIONS FOR MEDICAL, DENTAL OR TOILETRY PURPOSES
  • A61K39/00—Medicinal preparations containing antigens or antibodies
• • G—PHYSICS
  • G05—CONTROLLING; REGULATING
  • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
  • G05B2219/00—Program-control systems
  • G05B2219/30—Nc systems
  • G05B2219/32—Operator till task planning
  • G05B2219/32187—Correlation between controlling parameters for influence on quality parameters
• • G—PHYSICS
  • G05—CONTROLLING; REGULATING
  • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
  • G05B2219/00—Program-control systems
  • G05B2219/30—Nc systems
  • G05B2219/45—Nc applications
  • G05B2219/45031—Manufacturing semiconductor wafers
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L21/00—Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices or of parts thereof
  • H01L21/67—Apparatus specially adapted for handling semiconductor or electric solid state devices during manufacture or treatment thereof; Apparatus specially adapted for handling wafers during manufacture or treatment of semiconductor or electric solid state devices or components ; Apparatus not specifically provided for elsewhere
  • H01L21/67005—Apparatus not specifically provided for elsewhere
  • H01L21/67011—Apparatus for manufacture or treatment
  • H01L21/67017—Apparatus for fluid treatment
  • H01L21/67063—Apparatus for fluid treatment for etching
  • H01L21/67069—Apparatus for fluid treatment for etching for drying etching
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L21/00—Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices or of parts thereof
  • H01L21/67—Apparatus specially adapted for handling semiconductor or electric solid state devices during manufacture or treatment thereof; Apparatus specially adapted for handling wafers during manufacture or treatment of semiconductor or electric solid state devices or components ; Apparatus not specifically provided for elsewhere
  • H01L21/67005—Apparatus not specifically provided for elsewhere
  • H01L21/67242—Apparatus for monitoring, sorting or marking
  • H01L21/67253—Process monitoring, e.g. flow or thickness monitoring
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L22/00—Testing or measuring during manufacture or treatment; Reliability measurements, i.e. testing of parts without further processing to modify the parts as such; Structural arrangements therefor
  • H01L22/20—Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
  • H01L22/26—Acting in response to an ongoing measurement without interruption of processing, e.g. endpoint detection, in-situ thickness measurement

Definitions

• This invention relates generally to semiconductor fabrication technologv, and more particularly to monitoring etching processes during semiconductor fabrication using optical emission spectroscopy
• an etching process such as a reactive ion etch (RI E) process is empioved for etelling tine line patterns in a silicon wafer RIE involves positioning a masked wafer in a chamber l i that contains a plasma The plasma contains etchant gases that are dissociated m a radio trequencv (RF) field so that reactive ions contained in the etchant gases are vertically accelerated toward the wafer surface I he accelerated reactive ions combine chemicallv with unmasked material on the water surface As a result volatile etch products are produced During such etching single or multiple lavers ot material or films mav be removed Such material includes tor example, silicon dioxide ( SIOT ). polysilicon ( poly) and silicon nitride (S ⁇ -N,) fndpoint
• determination or detection refers to control of an etch step and is useful in etching processes in general and in RIE processes in particular
• the volatile etch products are incorporated into the plasma As the RIE process approaches the interlace or end ot the layer being etched the amount of volatile etch product found in the plasma decreases since the amount ot unmasked material being etched is reduced due to the etching
• the amount of volatile etch product in the plasma mav be tracked to determine the endpoint ot the RIE process
• the depletion or reduction in the amount ot volatile etch product in the plasma dur ⁇ n « the RIL process tvpicaliv can be used as an indicatoi tor the end of the etching process
• plasma discharge materials such as etchant neutral and
• optical emission spectrometer diffracts this light into its component wavelengths Since each species emits light at a wavelength characteristic only ot that species, it is possible to associate a certain wavelength with a particular species, and to use this information to detect an etch endpoint
• CO carbon monoxide
• Patent No 5,658,423, to Angell et al entitled "Monitoring and Controlling Plasma Processes via Optical Emission Using Principal Component Analysis '
• These conventional techniques ty pically still entail singling out one wavelength to be used for signaling an etch endpoint, however
• a conventional technique tor effecting process control bv statistical analysis of the optical spectrum ot a product produced in a chemical process is described, for example, in U S Patent No 5,862,060 to Murray, Jr , entitled “Maintenance ot process control by statistical analysis of product optical spectrum” (the '060 patent)
• the '060 patent describes measuring the optical spectrum of each member of a calibration sample set of selected products, determining bv Principal Component Analysis
• PCA Partial Least Squares PLS
• the '060 patent describes, tor example, that a very small number ot Principal Components, usually no more than 4. suffice to define accurately that sample spectrum space for the purpose ot process control and that in some cases onlv 2 or 3 Principal Components need to be used However, that still leaves an undesirable amount of uncertainty in whether to use 2. 3 or 4 Principal Components furthermore, this uncertainty can lead these conventional techniques to be cumbersome and slow and difficult to implement on the fly" du ⁇ ng real-time etching processes, for example
• I he present invention is directed to overcoming, or at least reducing the ettects of, one or more of the problems set forth above
• a method is provided tor determining an etch endpoint The method includes collecting intensity data representative of optical emission spectral wavelengths during a plasma etch process I he method further includes calculating Scores from at least a portion ot the collected intensity data using at most tirst, second, third and fourth Principal Components derived from a model The method also includes determining the etch endpoint using Scores corresponding to at least one ot the first second, third and fourth Principal C mponents as an indicator for the etch endpoint.
• a computer-readable, program storage device is provided, encoded w ith instructions that when executed bv a computer, perform a method the method including collecting intensity data representative of optical emission spectral wavelengths du ⁇ ng a plasma etch process The method further includes calculating Scores from at least a portion of the collected intensity data using at most first, second, third and fourth Principal Components derived from a model The method also includes determining the etch
• a computer programmed to perform a method including collecting intensity data representative of optical emission spectral wavelengths during a plasma etch process The method further includes calculating Scores from at least a portion of the collected intensity data using at most first, second, third and fourth Principal Components de ⁇ ved from a model The method also includes determining the etch endpoint using Scores corresponding to at least one of the first, second, third and fourth Principal Components as an indicator for the etch endpoint
• Figures 1-7 schematically illustrate a flow diagram for various embodiments of a method according to the present invention.
• Figures 8- 14 schematically illustrate a flow diagram tor various alternative embodiments of a method according to the present invention
• Figures 15-21 schematically illustrate a flow diagram for yet other various embodiments of a method according to the present invention
• Figures 22 and 23 schematically illustrate first and second Principal Components tor respective data sets.
• Figure 24 schematically illustrates OES spectrometer counts plotted against wavelengths.
• Figure 25 schematically illustrates a time trace of OES spectrometer counts at a particular wavelength,
• Figure 26 schematically illustrates representative mean-scaled spectrometer counts tor OES traces of a contact hole etch plotted against wavelengths and time.
• Figure 27 schematically illustrates a time trace of Scores for the second Principal Component used to determine an etch endpoint
• Figures 28 and 29 schematically illustrate geometrically Principal Components Analysis for respective data sets.
• Figure 30 schematically illustrates a time trace of OES spectrometer counts at a particular wavelength and a reconstructed time trace of the OES spectrometer counts at the particular wavelength.
• Figure 3 1 schematically illustrates a method for fabricating a semiconductor device practiced in accordance with the present invention
• Figure 32 schematically illustrates workpieces being processed using a high-density plasma (HDP) etch processing tool, using a plurality of control input signals, in accordance with the present invention
• HDP high-density plasma
• OES optical emission spectroscopy
• PCA Principal Components Analysis
• the second Principal Component contains a very robust, high signal-to-noise indicator for etch endpoint determination
• the first four Principal Components similarly may be useful as indicators tor etch endpoint determination as well as for data compression of OES data
• PCA mav be applied to the OES data, either the whole spectrum or at least a portion ot the whole spectrum If the engineer and/or controller knows that onlv a portion of the OES data contains useful information. PCA mav be applied onlv to that portion, for example
• archived data sets of OES w avelengths (or frequencies), trom wafers that had previously been plasma etched, mav be processed and Loadings for the first through fourth Principal Components determined from the archived OES data sets mav be used as model Loadings to calculate approximate Scores corresponding to newlv acquired OES data
• These approximate Scores, along with the mean values for each wavelength, may then be stored as compressed OES data
• archived data sets of OES wavelengths (or frequencies), from wafers that had previously been plasma etched may be processed and Loadings for the first through fourth Principal Components determined from the archived OES data sets may be used as model Loadings to calculate approximate Scores corresponding to newly acquired OES data 1 hese approximate Scores may be used as an etch endpoint indicator to determine an endpoint for an etch process
• FIG. 15-21 archived data sets of OES wavelengths (or frequencies), from wafers that had previously been plasma etched, may be processed and Loadings tor the first through fourth Principal Components, determined from the archived OES data sets, may be used as model Loadings to calculate approximate Scores corresponding to newly acquired OES data These approximate Scores, along with model mean values also determined trom the archived OES data sets, may then be stored as more compressed OES data These approximate Scores may also be used as an etch endpoint indicator to determine an endpoint for an etch process
• any ot these three embodiments may be applied in real-time etch processing
• either ot the last two illustrative embodiments may be used as an identification technique when using batch etch processing, with archived data being applied statistically, to determine an etch endpoint tor the batch
• Various embodiments of the present invention are applicable to any plasma etching process affording a characteristic data set whose quality may be said to define the ' success ' of the plasma etching process and whose identity may be monitored by suitable spectroscopic techniques such as OES
• OES spectroscopic techniques
• the nature ot the plasma etching process itself is not critical, nor is the specific identity ot the workpieces (such as semiconducting silicon waters) whose spectra are being obtained
• an 'ideal or target characteristic data set should be able to be defined, having a known and/or determinable OES spectrum, and v ariations in the plasma etching process awav trom the target characteristic data set should be able to be
• test stream mav the spectrum ot the stream hereafter referred to as the test stream mav be obtained continuously (or in a near-continuous manner) on-line, and compared to the spectrum ot the target ' stream, hereafter referred to as the standard stream
• the ditterence in the two spectra may then be used to ad
• the complex spectral data will be reduced and/or compressed to no more than 4 numerical values that define the coordinates ot the spectrum in the Principal Component or T actor space of the process sub
• ustments will be made so that the test stream approaches the standard stream while minimizing oscillation particularly oscillations that will tend to lose process control rather than exercise control Adjustments may be made according to one or more suitable algorithms based on process modeling, process experience, and/or artificial intelligence feedback,
• more than one process variable may be sub
• one test stream and one process variable under control one mav analogize the foregoing to the use of a thermocouple in a reaction chamber to generate a heating voltage based on the sensed temperature, and to use the difference between the generated heating voltage and a set point heating voltage to send power to the reaction chamber in proportion to the difference between the actual temperature and the desired temperature which, presumably, has been predetermined to be the optimum temperature Since the result of a given change in a process variable can be determined quickly, this new approach opens up the possibility of controlling the process by an automated "trial and error" feedback system, since unfavorable changes can be detected quickly
• Illustrative embodiments of the present invention may operate as null-detectors with feedback from
• OES spectra are taken of characteristic data sets of plasma etching processes of various grades and quality, spanning the maximum range of values typical of the particular plasma etching processes Such spectra then are representative of the entire range of plasma etching processes and are often referred to as calibration samples Note that because the characteristic data sets are representative of those formed in the plasma etching processes, the characteristic data sets constitute a subset of those that define the boundaries of representative processes It will be recognized that there is no subset that is unique, that many different subsets may be used to define the boundaries, and that the specific samples selected are not critical
• the spectra of the calibration samples are subjected to the well-known statistical technique of Principal Component Analysis (PCA) to afford a small number of Principal Components (or Factors) that largely determine the spectrum of any sample
• PCA Principal Component Analysis
• PLS Partial Least Squares
• any new sample may be assigned various contributions ot these Principal Components that would reproduce the spectrum of the new sample
• the amount of each Principal Component required is called a Score, and time traces of these Scores, tracking how various of the Scores are changing with time, are used to detect deviations from the "target spectrum
• a set of m time samples of an OES for a workpiece (such as a semiconductor wafer having various process layers formed thereon) taken at n channels or wavelengths or frequencies may be arranged as a rectangular nxm matrix X
• the rectangular nxm matrix X may be comprised of 1 to n rows (each row corresponding to a separate OES channel or wavelength or frequency time sample) and I to m columns (each column corresponding to a separate OES spectrum time sample)
• the values of the rectangular nxm matrix X may be counts representing the intensity of the OES spectrum, or ratios of spectral intensities (normalized to a reference intensity), or logarithms of such ratios, for example
• the rectangular nxm matrix X may have rank r, where r ⁇ m ⁇ n ⁇ m,n ⁇ is the maximum number of independent variables in the matrix X
• PCA for example, generates a set of Principal Components P (whose "Loadings," or components, represent
• M is a rectangular nxm matrix of the mean values of the columns of X (the m columns of M are each the column mean vector j nx
• any sample mav be expressed as a 2-d ⁇ mens ⁇ onal representation of the intensitv of mission at a particular w avelength vs the wavelength That is one axis represents intensity the other wavelength
• the toregoing characterization ot a spectrum is intended to incorporate various transformations that are mathematically cova ⁇ ant c ⁇ instead ot emission one might use absorption and/or transmission, and either mav be expressed as a percentage or logarithmically Whatever the details each spectrum may be viewed as a vector
• the group of spectra arising trom a group ot samples similarly corresponds to a group ot vectors If the number ot samples is N there are at most N distinct spectra If in fact none ot the spectra can be expressed as a linear combination of the other spectra then the set of spectra define an /V-d ⁇ mens ⁇ onal spectrum space However, in the cases of interest here w here a particular stream in an invariant plasma etching process is being sample
• NPALS nonlinear iterative partial least squares
• each of the first two methods, EIG and SVD simultaneously calculates all possible Principal Components
• the NIPALS method allows for calculation of one Principal Component at a time
• the power method described more fully below, is an iterative approach to finding eigenvalues and eigenvectors, and also allows for calculation of one Principal Component at a time
• the power method may efficiently use computing time
• EIG reveals that the eigenvalues ⁇ of the matrix product A T A are 3 and 2
• a trial eigenvector rj ( 1 , 1 , 1 ) may be used
• EIG reveals that the eigenvalues of the matrix product B T B are 4, 2 and 2
• a trial eigenvector 2 T ( 1 , 1 , 1 , 1 ) may be used
• a Gram-Schmidt orthonormalization procedure may be used, for example
• the matrices A and B discussed above have been used for the sake of simplifying the presentation of PCA and the power method and are much smaller than the data matrices encountered in illustrative embodiments of the present invention
• 18 wafers may be run and corresponding OES data collected
• a scatte ⁇ lot 2200 ot data points 2210 mav be plotted in an n-dimensional variable space (n-3 in T ⁇ gure 22)
• the mean vector 2220 mav be determined bv taking the average of the columns
• the Principal Component ellipsoid 2230 mav have a first Principal Component 2240 (maior axis in Figure 22) with a length equal to the largest eigenvalue of the mean-scaled OES data matrix X-M and a second Principal Component 2250 (minor axis in Figure 22) with a length equal to the next largest eigenvalue of the mean-scaled OES data matrix X-M
• the 3x4 matrix B r given above mav be taken as the overall OES data matrix X (again for the sake of simplicity) corresponding to 4 scans taken at 3 wavelengths ⁇ s shown in Figure 23 a scatte ⁇ lot 2300 of data points 23 10 mav be plotted in a 3-d ⁇ mens ⁇ onal variable space 1 h mean vector 2320 u mav lie at the center of a 2-d ⁇ mens ⁇ onal Principal Component ellipsoid 2330 (reallv a circle a degenerate ellipsoid)
• the mean vector 2320 u mav be determined bv taking the average ot the columns ot the overall OES 3x4 data matrix B r
• the Principal Component ellipsoid 2330 mav have a first Principal Component 2340 ( maior axis in Figure 23) and a second Principal Component 2350 ( minor axis in Figure 23)
• each wafer mav be taken on wavelengths between about 240-1 100 nm at a high sample rate of about one per second
• 555 1 sampling points/spectrum/second corresponding to 1 scan per wafer per second taken at 5551 wavelengths or 7 scans per water per second taken at 793 wavelengths or 13 scans per water per second taken at 427 wavelengths or 61 scans per wafer per second taken at 91 wavelengths
• 555 1 sampling points/spectrum/second may be collected in real time, during etching ot a contact hole using an Applied Materials AMAT 5300 Centura etching chamber, to produce high resolution and broad band OES spectra
• a representative OES trace 2500 ot a contact hole etch is illustrated Time, measured in seconds (sec) is plotted along the horizontal axis against spectrometer counts plotted along the vertical axis As shown in Figure 25, by about 40 seconds into the etching process, as indicated by dashed line 2510, the
• FIG. 26 representative OES traces 2600 ot a contact hole etch are illustrated Wavelengths, measured in nanometers (nm) are plotted along a first axis, time, measured in seconds (sec) is plotted along a second axis, and mean-scaled OES spectrometer counts, for example are plotted along a third (vertical) axis As shown in Figure 26, over the course ot about 150 seconds of etching three clusters of wavelengths 2610, 2620 and
• any one of the three clusters of wavelengths 2610 2620 and 2630 mav be used, either taken alone or taken in any combination with any one (or both) ot the others, as an indicator variable signaling an etch endpoint
• spectrometer count value (for example, about 200, as shown in Figure 26) may be used, either taken alone or taken together, as an indicator variable signaling an etch endpoint
• only one cluster ot wavelengths 2630 having an absolute value ot mean-scaled OES spectrometer counts that exceeds a preselected threshold absolute mean-scaled OES spectrometer count value (for example, about 300. as shown in Figure 26) mav be used as an indicator variable signaling an etch endpoint ⁇
• PCA Principal Components Analysis
• a scatte ⁇ lot 2800 of OES data points 2810 and 2820. with coordinates ( 1 1.1) and (- 1,0,1 ). respectively, may be plotted in a 3-d ⁇ mens ⁇ onal variable space where the variables are respective spectrometer counts for each of the 3 wavelengths
• the mean vector 2830 u may be determined by taking the average of the columns of the overall OES 3x2 matrix C
• the Principal Component ellipsoid 2840 mav have a first Principal Component 2850 (the "major” axis in Figure 28. with length V5, lying along a first Principal Component axis 2860) and no second or third Principal Component lying along second or third Principal Component axes 2870 and 2880, respectively
• two of the eigenvalues of the mean-scaled OES data matrix C-M are equal to zero, so the lengths of the "minor" axes in Figure 28 are both equal to zero.
• the mean vector 2830 u_ is given by
• PCA is nothing more than a. principal axis rotation of the original variable axes (here, the OES spectrometer counts for 3 wavelengths) about the endpoint of the mean vector 2830 ⁇ _, with coordinates (0,1/2, 1 ) with respect to the original coordinate axes and coordinates [0,0,0] with respect to the new Principal Component axes 2860, 2870 and 2880
• the Loadings are merely the direction cosines of the new Principal Component axes 2860, 2870 and 2880 with respect to the original variable axes
• the Scores are simply the coordinates of the OES data points 2810 and 2820, [5 ⁇ 3 /2 0.0] and [-5 0 : 72,0,0], respectively, referred to the new Principal Component axes 2860, 2870 and 2880
• the mean-scaled 3x2 OES data matrix C-M, its transpose, the 2x3 matrix (C-M) 1 , their 2x2 matrix product (C-M) T (C-M), and their 3x3 matrix product (C-M) (C-M) ⁇ are given by
• EIG reveals that the eigenvalues ⁇ of the matrix product (C-M) T (C-M) are 5/2 and 0, for example b
• C-M PT r
• P the Principal Component matrix (whose columns are orthonormalized eigenvectors proportional to pi, £ 2 an d 2 3 , an ⁇ whose elements are the Loadings, the direction cosines of the new Principal Component axes 2860, 2870 and 2880 related to the original va ⁇ able axes)
• T the Scores matrix (whose rows are the coordinates of the OES data points 2810 and 2820, referred to the new Principal Component axes 2860, 2870 and 2880)
• the transpose of the Scores matrix (T ⁇ ) is given by the product of the matrix ot eigenvalues of C-M with a matrix whose rows are orthonormalized eigenvectors proportional to t, and f>
• the direction cosine (Loading) of the first Principal Component axis 2860 with respect to the wavelength I counts axis is given by COS ⁇
• the direction cosine (Loading) of the first Principal Component axis 2860 with respect to the wavelength 3 counts axis is given by C ⁇ s ⁇ 3
• 0
• the mean vector 2920 u mav e at the center of a 2-d ⁇ mens ⁇ onal Principal Component ellipsoid 2930 (really a circle, a somewhat degenerate ellipsoid)
• the mean vector 2920 u may be determined by taking the average of the columns of the overall OES 3 ⁇ 4 matrix D
• the Principal Component ellipsoid 2930 may have a first Principal Component 2940 (the ma
• PCA is nothing more than a principal axis rotation of the original va ⁇ able axes (here, the OES spectrometer counts for 3 wavelengths) about the endpoint of the mean vector 2920 u_, with coordinates ( 1 ,0 0) w ith respect to the original coordinate axes and coordinates [0,0,0] with respect to the new Principal Component axes 2950, 2970 and 2980
• the Loadings are merely the direction cosines of the new Principal Component axes 2950.
• the 3x3 matrix product (D-M)(D-M) T is given by
• EIG reveals that the eigenvalues of the matrix product (D-M)(D-M) T are 0, 2 and 2
• the columns of the transpose of the Scores matrix T ⁇ are. indeed, the coordinates of the OES data points. [ 1 ,0,0], [0.1 0], [0,- 1 ,0] and [- 1 ,0,0], respectively, referred to the new Principal Component axes 2950, 2970 and 2980
• the overall mean-scaled OES rectangular nxm data matrix X llm -M nm may be decomposed into a portion corresponding to the first and second Principal Components and respective Loadings and Scores, and a residual portion
• Both the SPE test and the Hotellmg T test can be used to monitor the etching process for example
• the overall mean-scaled OES rectangular nxm data matrix X tract m -M nm may be decomposed into a portion corresponding to the first through fourth Principal Components and respective Loadings and Scores, and a residual portion-
• T PC - is an m ⁇ 4 Scores matrix for the first through fourth Principal Components
• Tp is a 4 ⁇ m Scores matrix transpose
• Pp r T PC T Xpc is an nxm matrix
• P res is an n ⁇ (m-4) matrix whose columns are the residual Principal Components
• T rcs is an m ⁇ (m-4) Scores matrix for the residual Principal Components
• T rL is a (m-4) ⁇ m Scores matrix transpose
• P res T rCi r X res ' s an nxm matrix
• n -Ppc P ⁇ H is the projection ot x k into the residual subspace orthogonal to the PCS spanned by the first through fourth Principal Components.
• (Xpc ) k ' H Ac >s the projection ot x k ' into the Principal Component subspace (PCS) spanned by the first through fourth Principal Components.
• the overall mean-scaled OES rectangular nxm data matrix X nm -M ⁇ m of rank r w here r ⁇ m ⁇ n ⁇ m,n ⁇ may be decomposed into a portion corresponding to the first through rth Principal Components and respective Scores, and a residual portion
• P PC is an nxr matrix whose columns are the first through rth Principal Components.
• T ⁇ is an mxr Scores matrix for the first through rth Principal Components.
• T PC is an rxm Scores matrix transpose
• P PC T PC T X PC is an nxm matrix.
• T 1 x i P r ⁇ ⁇ '2 p ⁇ ⁇ "
• archived data sets (Y nxm ) °f OES wavelengths (or frequencies), trom waters that had previously been plasma etched, may be processed and the weighted linear combination ot the intensity data, representative of the archived OCS w avelengths (or frequencies ) collected over time during the plasma etch defined by the first through pth Principal Components, mav be used to compress newly acquired OES data
• the rectangular nxm matrix Y (Y wrench,, m ) may have rank r, where r ⁇ m ⁇ n ⁇ m,n ⁇ is the maximum number of independent variables in the matrix Y
• the workpiece 100 is sent from the etching preprocessing step j 105 to an etching step j+ 1 1 10 In the etching step j+ 1 1 10.
• the workpiece 100 may be sent trom the etching step j+ 1 1 10 and delivered to a postetchmg processing step j+2 1 15 for further postetch processing, and then sent on from the postetching processing step j +2 1 15
• the etching step j+ 1 1 10 may be the final step in the processing of the workpiece 100
• OES spectra are measured in situ bv an OES spectrometer (not shown), producing raw OES data 120 (X nxm ) indicative of the state of the workpiece 100 during the etching
• about 5500 samples ot each wafer mav be taken on wavelengths between about 240- 1 100 nm at a high sample rate ot about one per second
• 5551 sampling points/spectrum/second corresponding to 1 scan per water per second taken at 555 1 wavelengths
• the raw OES data 120 (X grip ⁇ m ) is sent trom the etching step j+ 1 1 10 and delivered to a mean-scaling step 125, producing a means matrix (M n m ), whose m columns are each the column mean vector (u_ nx ⁇ ) of the raw OES data 120 (X n x m ), and mean-scaled OES data (X- ⁇ - - , whatsoever)
• the mean values are treated as part of a model built from the archived data sets (Y nxm ) of OCS wavelengths (or frequencies), from wafers that had previouslv been plasma etched
• a means matrix (N nxm ) previously determined from the archived data sets (Y groove m ) of OES wavelengths (or frequencies), trom waters that had previously been plasma etched, is used to generate alternative mean-scaled OES data ( X
• the means matrix (M nxm ) and the mean-scaled OES data (X n ⁇ m -M ⁇ xrn ) 130 are sent from the mean scaling step 125 to a Scores calculating step 135, producing approximate Scores (T mxp )
• the mean-scaled OES data (X naturally ⁇ m -M ⁇ m ) are multiplied on the left by the transpose of the Principal Component (Loadings) matrix Qphon speaking p , with columns rj , o ⁇ , c
• the columns of the transpose of the Scores matrix T or, equivalents,
• the Loadings (Q nxp ), previously determined from the archived mean-scaled data sets (Y nxm -N ⁇ ⁇ m ) of OES wavelengths (or frequencies), from wafers that had previously been plasma etched, are used to generate the approximate Scores (T mxp ) corresponding to the mean-scaled OES data (X penetrate xm -M n m ) de ⁇ ved from the raw OES data 120 (X nxm )
• the Loadings are defined by the first through pth Principal Components
• p is in a range of 1-4
• p 2
• the first through pth Principal Components may be determmed off-line from the archived data sets (Y groove xm ) of OES wavelengths (or frequencies), for example, by any of the techniques discussed above
• the values of the rectangular nxm matrix Y (Y nxm ) f° r the archived data sets may be counts representing the intensity of the archived OES spectrum or ratios of spectral intensities (normalized to a reference intensity), or logarithms of such ratios, for example
• the rectangular nxm matrix Y (Y nxm ) for the archived data sets may have rank r where r ⁇ m ⁇ n ⁇ m,n ⁇ is the maximum number of independent variables in the matrix Y
• PCA tor example
• a feedback control signal 140 may be sent from the Scores calculating step 135 to the etching step j+1 1 10 to adjust the processing performed in the etching step j+ 1 1 10
• the feedback control signal 140 may be used to signal the etch endpoint
• the means matrix (M ⁇ xm ) and the approximate Scores (T, nxp ) 145 are sent from the Scores calculating step 135 and delivered to a save compressed PCA data step 150
• the means matrix (M nxm ) and the approximate Scores (T mxp ) 145 are saved and/or stored to be used in reconstructing X , ⁇ m , the decompressed approximation to the raw OES data 120 (X grip xm )
• the Loadings (Q 5551 X 4 ) are determined off-line from archived data sets of OES wavelengths (or frequencies), for example, by any of the techniques discussed above, and need not be separately stored with each water OES data set, so the storage volume of 5551 x4 for the Loadings (Q ⁇ X4 ) does not have to be taken into account in determining an effective compression ratio for the OES wafer data
• 0 o x ) only require
• the Loadings (Q, 35 , ⁇ ) are determined off-line from archived data sets (Ysssi x ioo) of OES wavelengths (or frequencies), tor example, by any of the techniques discussed above, and need not be separately stored with each water OES data set, so the storage volume of 5551 x4 for the
• the effective compression ratio tor the OES water data in this illustrative embodiment is about (555 l ⁇ l 00)/(793 ⁇ l ) or about 700 to 1 ( 100 1 ) More precisely, the compression ratio in this illustrative embodiment is about (5551 ⁇ l 00)/(793 ⁇ l + 100 ⁇ 4) or about 465 to 1 (465 1 )
• the effective compression ratio for the OES wafer data in this illustrative embodiment is about (5551 ⁇ i00)/(5551 ⁇ 5 ) or about 20 to 1 (20 1 ) More precisely, the compression ratio in this illustrative embodiment is about (5551 ⁇ l 00)/(555 1 x5+ 100x4) or about 19 7 to I ( 19 7 1 )
• the Loadings (Q 555 i x ) are determined off-line from archived data sets of OES wavelengths (or frequencies), for example, by any of the techniques discussed above and need not be separately stored with each wafer OES data set, so the storage volume of 55
• FIG. 30 a representative OES trace 3000 of a contact hole etch is illustrated Time, measured in seconds (sec) is plotted along the horizontal axis against spectrometer counts plotted along the vertical axis As shown in Figure 30, by about 40 seconds into the etching process, as indicated by dashed line 3010, the
• OES trace 3000 of spectrometer counts "settles down to a range of values less than or about 300. tor example
• a representative reconstructed OES trace 3020 (corresponding to X die m ), for times to the right of the dashed line 3010 (greater than or equal to about 40 seconds, for example), is schematically illustrated and compared with the corresponding noisv raw OES trace 3030 (corresponding to X n ⁇ m ), also for times to the right of the dashed line 3010
• the reconstructed OES trace 3020 (corresponding to X n%m ) is much smoother and less noisy than the raw OES trace 3030 (corresponding to X nxr ⁇ )
• a feedback control signal 155 may be sent trom the save compressed PCA data step 150 to the etching step j + 1 1 10 to adjust the processing performed in the etching step j ⁇ l 1 10
• the feedback control signal 155 may be used to signal the etch endpoint
• the workpiece 800 is sent from the etching preprocessing step j 805 to an etching step j+ 1 810 In the etching step j+ 1 810.
• the workpiece 800 may be sent from the etching step j+ 1 810 and delivered to a postetching processing step j+2 815 for further postetch processing, and then sent on from the postetching processing step j+2 815
• the etching step j+ 1 810 may be the final step in the processing of the workpiece 800 In the etching step j+ 1 810.
• OES spectra are measured m situ by an OES
• about 5500 samples ot each wafer may be taken on wavelengths between about 240- 1 100 nm at a high sample rate of about one per second
• 5551 sampling points/spectrum/second corresponding to 1 scan per wafer per second taken at 5551 wavelengths
• the raw OES data 820 (X ⁇ xm ) is sent from the etching step j-H 810 and delivered to a mean-scaling step 825, producing a means matrix (M nxm ), whose m columns are each the column mean vector ( ⁇ nx i) of the raw OES data 820 (X mm ), and mean-scaled OES data (X n ⁇ m -M nxr ⁇ ) In the mean-scaling step 825.
• M nxm means matrix
• the mean values are treated as part ot a model built from the archived data sets (Y nxm ) of OES wavelengths (or frequencies), from waters that had previously been plasma etched
• a means matrix (N Native xm ) previously determined from the archived data sets (Y nxm ) of OES wavelengths (or Irequencies), trom waters that had previously been plasma etched is used to generate alternative mean-scaled OES data (X Wayne xm -N n m )
• the mean values for each wafer and/or mean value for each wavelength for example, are determined as discussed above, and are used to generate the mean-scaled OES data ( ⁇ din m -M Channel m )
• the means matrix (M nxm ) and the mean-scaled OES data (X grip m -M practic , whatsoever) 830 are sent trom the mean scaling step 825 to a Scores calculating step 835. producing approximate Scores ( T r ⁇ l p )
• the mean-scaled OES data ( X , m -M - ⁇ l ) are multiplied on the left bv the transpose of the Principal Component (Loadings) matrix Q personallyx P , with columns U , rjj, ( jp , that are the first p orthonormalized eigenvectors of the matrix product ( Y-N)(Y-N) T
• the approximate Scores (T mxp ) are calculated using the Loadings (Q nxp ) derived from the model built trom the archived mean-scaled data sets (Y n ⁇ m -N n , reconsider) of OES wavelengths (or frequencies), from waters that had previously been plasma etched
• the Loadings (Q n p ) previously determined from the archived mean-scaled data sets (Y groove m -N m ) of OES wavelengths (or frequencies) trom wafers that had previously been plasma etched are used to generate the approximate Scores (T mxp ) corresponding to the mean-scaled OES data (X penetrate xm -M nxm ) derived from the raw OES data 820 (X nxm )
• the Loadings (O n P ) are defined by the first through pth Principal Components
• p ⁇ r in various illustrative embodiments p is in a range of 1 -4 in various alternative illustrative embodiments p 2
• the first through pth Principal Components may be determined off-line from the archived data sets (Y perennial m ) of OES wavelengths (or frequencies) for example, by any of the techniques discussed above
• the values ot the rectangular nxm matrix Y (Y chord xm ) tor the archived data sets mav be counts representing the intensity of the archived OES spectrum, or ratios ot spectral intensities (normalized to a reference intensity), or logarithms ot such ratios, for example
• the rectangular nxm matrix Y (Y nx ) for the archived data sets may have rank r.
• 1 ,2, ,m, and the rectangular mxn matrix Y ⁇ .
• a feedback control signal 840 may be sent from the Scores calculating step 835 to the etching step j+ 1 810 to adjust the processing performed in the etching step j+ 1 810 For example, based on the determination of the approximate Scores (T mxp ) calculated using the Loadings (Q nxP ) derived trom the model built from the archived mean-scaled data sets (Y nxm -N n ⁇ m ) of OES wavelengths (or frequencies), from wafers that had previously been plasma etched, the feedback control signal 840 mav be used to signal the etch endpoint
• the approximate Scores (T mxp ) 845 are sent trom the Scores calculating step 835 and delivered to a use Scores as etch indicator step 850
• the approximate Scores (T mxp ) 845 are used as an etch indicator
• a representative Scores time trace 2700 corresponding to the second Principal Component during a contact hole etch is illustrated Time, measured in seconds (sec) is plotted along the horizontal axis against Scores (in arbitrary units) plotted along the vertical axis
• the Scores time trace 2700 corresponding to the second Principal Component during a contact hole etch may start at a relatively high value initially, decrease with time, pass through a minimum value, and then begin increasing before leveling off
• the inflection point indicated by dashed line 2710, and approximately where the second derivative of the Scores time trace 2700 with respect to time vanishes
• a teedback control signal 855 mav be sent trom the use Scores as etch indicator step 850 to the etching step j + 1 810 to adjust the processing performed in the etching step j+ 1 810
• the feedback control signal 855 may be used to signal the etch endpoint
• These approximate Scores (T mx ) with or without the mean values for each wavelength (N nxm ), effectively the column mean vector ( ⁇ nX ⁇ ) of the archived OES data (Y nxm ), may then be stored as compressed OES data
• These approximate Scores (T mx4 ) may also be used as an etch endpoint indicator to determine an endpoint for an etch process
• a workpiece 1500 such as a semiconducting substrate or wafer, having one or more process layers and/or semiconductor devices such as an MOS transistor disposed thereon, tor example, is delivered to an etching preprocessing step j 1505.
• the workpiece 1500 is sent from the etching preprocessing step j 1505 to an etching step j+ 1 15 10
• the workpiece 1500 may be sent trom the etching step j+ 1 1510 and delivered to a postetching processing step j+2 1515 for further postetch processing, and then sent on from the postetching processing step j+2 15 15
• the etching step j+ 1 15 10 may be the final step in the processing of the workpiece 1500 In the etching step j+ 1 1510 OES spectra are measured in situ by an OES spectrome
• each water mav be taken on wavelengths between about 240- 1 100 nm at a high sample rate ot about one per second
• 555 1 sampling pomts/spectrum/second corresponding to 1 scan per wafer per second taken at 555 1 wavelengths
• mav be collected in real time, during etching ot a contact hole using an Applied Materials AMAT 5300 Centura etching chamber to produce high resolution and broad band OES spectra
• the raw OES data 1520 (X Guide x , n ) is sent trom the etching step j+ 1 15 10 and delivered to a Scores calculating step 1525, where a means matrix (N Primatician,) whose m columns are each the column mean vector (u_ practice x
• the mean values are treated as part ot a model built from the archived data sets (Y n ⁇ ) of OES wavelengths (or frequencies) from wafers that had previously been plasma etched
• a means matrix (N réelle m ) previously determined from the archived data sets (Y nxm ) of OES w avelengths (or frequencies) trom waters that had previously been plasma tched is used to generate the
• the alternative mean-scaled OES data ( ⁇ n familiar,-N 1Mm ) are used to produce alternative approximate Scores (T m x P )
• the columns of the transpose of the Scores matrix T ⁇ or, equivalently the rows ot the alternative approximate Scores matrix (T
• the alternative approximate Scores (T mxp ) are calculated using the Loadings (Q ⁇ xp ) derived from the model built from the archived mean-scaled data sets (Y ⁇ ⁇ -N n ⁇ m ) of OES wavelengths (or frequencies), from wafers that had previously been plasma etched
• the Loadings (Q nxp ) previously determined from the archived mean-scaled data sets (Y nxm -N nxm ) of
• OES wavelengths (or frequencies) trom wafers that had previously been plasma etched are used to generate the alternative approximate Scores (T m p ) corresponding to the mean-scaled OES data (X nxm - nxm ) derived from the raw OES data 1520 (X nxrn )
• the Loadings (Q tine xp ) are defined by the first through pth Principal Components
• the first through pth Principal Components may be determined off-line from the archived data sets (Y nxm ) of OES wavelengths (or frequencies), for example, by any of the techniques discussed above
• the values of the rectangular nxm matrix Y (Y nxm ) for the archived data sets may be counts representing the intensity of the archived OES spectrum, or ratios of spect
• a feedback control signal 1530 may be sent from the Scores calculating step 1535 to the etching step j+1 1510 to adjust the processing performed in the etching step j+ 1 1510
• the feedback control signal 1530 may be used to signal the etch endpoint
• the alternative approximate Scores (T mxp ) 1535 are sent from the Scores calculating step 1525 and delivered to a save compressed PCA data step 1540 In the save compressed PCA data step 1540, the alternative approximate Scores (T mxp ) 1535 are saved and/or stored to be used in reconstructing
• the means matrix (N 555 i x l oo) is determined off-line from archived data sets of OES wavelengths (or frequencies), and need not be separately stored with each wafer OES data set
• the storage volume of 5551 x 1 for the means matrix (N 5551X 100 ) where all the 100 columns of the means matrix (N 555 i x ⁇ oo) are identical (each of the 5551 rows of each column being the mean value for that wavelength or channel over the 100 scans), does not have to be taken into account in determining an effective compression ratio for the OES water data
• x4 ) are also determined off-line from archived data sets of OES wavelengths (or frequencies), tor example, by any of the techniques discussed above, and also
• a representative OES trace 3000 of a contact hole etch is illustrated Time, measured in seconds (sec) is plotted along the horizontal axis against spectrometer counts plotted along the vertical axis As shown in Figure 30, by about 40 seconds into the etching process, as indicated by dashed line 3010, the OES trace 3000 of spectrometer counts ' settles down to a range of values less than or about 300, for example
• a representative reconstructed OES trace 3020 (corresponding to X nxm ), for times to the right of the dashed line 3010 (greater than or equal to about 40 seconds for example), is schematically illustrated and compared with the corresponding noisv raw OES trace 3030 (corresponding to X tract ⁇ m ), also for times to the right ot the dashed line 3010
• the reconstructed OES trace 3020 (corresponding to X route ⁇ resort, ) is much smoother and less noisv than the raw OES trace 3030 (corresponding to X tract , context )
• the alternative approximate Scores (T mxp ) 1545 are sent trom the save compressed PCA data step 1540 and delivered to a use Scores as etch indicator step 1550 In the use Scores as etch indicator step 1550.
• the alternative approximate Scores ( r m p ) 1545 are used as an etch indicator
• a representative Scores time trace 2700 corresponding to the second Principal Component during a contact hole etch is illustrated Time, measured in seconds (sec) is plotted along the horizontal axis against Scores (in arbitrary units) plotted along the vertical axis As shown in Figure 27.
• the Scores time trace 2700 corresponding to the second Principal Component during a contact hole etch mav start at a relatively high value initially, decrease with time, pass through a minimum value and then begin increasing before leveling otf
• the inflection point indicated bv dashed line 2710 and approximately where the second derivative ot the Scores time trace 2700 with respect to time vanishes
• a feedback control signal 1555 mav be sent trom the use Scores as etch indicator step 1550 to the etching step j+ 1 1510 to adjust the processing performed in the etching step j+ 1 1510
• the feedback control signal 1 55 may be used to signal the etch endpoint
• Figure 3 1 illustrates one particular embodiment of a method 3100 practiced in accordance with the present invention
• Figure 32 illustrates one particular apparatus 3200 with which the method 3100 may be practiced
• the method 3100 shall be disclosed in the context of the apparatus 3200
• the invention is not so limited and admits wide variation, as is discussed further below
• the etch processing tool 3210 may be any etch processing tool known to the art, such as Applied Materials AMAT 5300 Centura etching chamber, provided it includes the requisite control capabilities
• the etch processing tool 3210 includes an etch processing tool controller 3215 for this pu ⁇ ose
• the nature and function of the etch processing tool controller 3215 will be implementation specific
• the etch processing tool controller 3215 may control etch control input parameters such as etch recipe control input parameters and etch endpoint control parameters, and the like
• Four workpieces 3205 are shown in Figure 32, but the lot ot workpieces or wafers, / , the ' wafer lot,' may be any practicable number ot waters from one to any finite number
• the method 3 100 begins, as set forth in box 3 120, by measuring parameters such as OES spectral data characteristic of the etch processing performed on the workpiece 3205 in the etch processing tool 3210
• parameters such as OES spectral data characteristic of the etch processing performed on the workpiece 3205 in the etch processing tool 3210
• OES spectral data characteristic of the etch processing performed on the workpiece 3205 in the etch processing tool 3210
• capabilities tor monitoring process parameters vary, to some degree, trom tool to tool Greater sensing capabilities may permit wider latitude in the characteristic parameters that are identified and measured and the manner in which this is done Conversely, lesser sensing capabilities may restrict this latitude
• the etch process characteristic parameters are measured and/or monitored by tool sensors (not shown)
• the outputs of these tool sensors are transmitted to a computer svstem 3230 over a line 3220
• the computer system 3230 analyzes these sensor outputs to ldentifv the characteristic parameters
• the method 3 100 proceeds by modeling the measured and identified characteristic parameter using PCA as set forth in box 3 130
• the computer system 3230 in Figure 32 is, in this particular embodiment, programmed to model the characteristic parameter using PCA
• the manner in which this PCA modeling occurs will be implementation specific
• a database 3235 stores a plurality of PCA models and/or archived PCA data sets that might potentially be applied, depending upon which characteristic parameter is identified This particular embodiment, therefore, requires some a priori knowledge of the characteristic parameters that might be measured
• the computer system 3230 then extracts an appropriate model from the database 3235 ot potential models to apply to the identified characteristic parameters If the database 3235 does not include an appropriate model, then the characteristic parameter may be ignored, or the computer system 3230 may attempt to develop one, if so programmed
• the database 3235 mav be stored on anv kind ot computer-readable, program storage medium, such as an optical disk 3240.
• the database 3235 mav also be stored on a separate computer system (not shown) that interfaces with the computer system 3230
• Modeling of the identified characteristic parameter may be implemented differently in alternative embodiments
• the computer svstem 3230 may be programmed using some form of artificial intelligence to analvze the sensor outputs and controller inputs to develop a PCA model on-the-fly in a real-time PCA implementation This approach might be a useful adjunct to the embodiment illustrated in Figure 32, and discussed above where characteristic parameters are measured and identified for which the database 3235 has no appropriate model
• the method 3 100 of Figure 31 then proceeds by applying the PCA model to compress the OES data and/or determine an etch endpoint, as set forth in box 3 140
• the OES data compressed using PCA may be stored on any kind of computer-readable, program storage medium, such as an optical disk 3240. a floppy disk 3245, or a hard disk drive (not shown) of the computer svstem 3230. and/or together with the database 3235
• the OES data compressed using PCA according to any ot the various illustrative embodiments of the present invention may also be stored on a separate computer system (not shown) that interfaces with the computer system 3230
• applying the PCA model may yield either a new value for the etch endpoint control parameter or a correction and/or update to the existing etch endpoint control parameter
• the new etch endpoint control parameter is then formulated from the value yielded by the PCA model and is transmitted to the etch processing tool controller 3215 over the line 3220
• the etch processing tool controller 3215 then controls subsequent etch process operations in accordance with the new etch control inputs
• Some alternative embodiments may employ a form ot feedback to improve the PCA modeling of characteristic parameters
• the implementation ot this feedback is dependent on several disparate facts, including the tool s sensing capabilities and economics
• One technique for doing this would be to monitor at least one effect of the PCA model s implementation and update the PCA model based on the effect(s) monitored
• the update may also depend on the PCA model For instance, a linear model mav require a different update than would a non-linear model, all other factors being the same
• some features ot the present invention are implemented in software for instance the acts set forth in the boxes 3120-3 140 in F ⁇ gure 31 are in the illustrated embodiment, sottware-implemented in whole or in part
• some features ot the present invention are implemented as instructions encoded on a computer-readable, program storage medium
• the program storage medium may be ot any type suitable to the particular implementation
• the program storage medium w ill typically be magnetic, such as the floppy disk 3245 or the computer 3230 hard disk drive (not shown ) or optical, such as the optical disk 3240
• thev perform the disclosed functions
• the computer may be a desktop computer, such as the computer 3230
• the computer might alternatively be a processor embedded in the etch processing tool 3210
• the computer might also be a laptop, a workstation, or a mainframe in various other embodiments
• the scope of the invention is not limited bv the tvpe or nature of the program storage medium or computer with which embodiments ot the
• a process engineer may be provided with advanced process data monitoring capabilities, such as the ability to provide historical parametric data in a user-friendly format, as well as event logging, real-time graphical display of both current processing parameters and the processing parameters of the entire run, and remote, / e , local site and worldwide, monitoring
• advanced process data monitoring capabilities such as the ability to provide historical parametric data in a user-friendly format, as well as event logging, real-time graphical display of both current processing parameters and the processing parameters of the entire run, and remote, / e , local site and worldwide, monitoring
• critical processing parameters such as throughput accuracy, stability and repeatability, processing temperatures, mechanical tool parameters, and the like
• This more optimal control of critical processing parameters reduces this variability
• This reduction in variability manifests itself as fewer within-run disparities, fewer run-to-run disparities and fewer tool-to- tool disparities
• This reduction in the number of these disparities that can propagate means fewer deviations in product quality and performance
• a monitoring and diagnostics system may be
• An etch endpoint determination signal as in any of the embodiments disclosed above may have a high signal-to-noise ratio and may be reproducible over the variations of the incoming waters and the state of the processing chamber, for example, whether or not the internal hardware in the processing chamber is worn or new, or whether or not the processing chamber is in a 'clean ' or a dirty condition
• an etch endpoint determination signal as in any of the embodiments disclosed above mav have a high enough signal-to-noise ratio to overcome the inherently very low signal-to-noise ratio that may arise simply by virtue of the small percentage ( 1 % or so) of surface area being etched
• the etch endpoint signal becomes very stable, and may throughput may be improved by reducing the ma etch time from approximately 145 seconds, for example to approximately 90-100 seconds, depending on the thickness of the oxide
• a longer etch time is conventionally needed to insure that all the material to be etched awav has been adequately etched away, even in vias and contact holes with high aspect ratios
• the presence of a robust etch endpoint determination signal as in anv of the embodiments disclosed above thus allows for a shorter etch time, and. hence, increased throughput, compared to conventional etching processes
• embodiments of the present invention fill a need in present day and future technology for optimizing selection ot wavelengths to monitor for endpoint determination or detection during etching
• embodiments of the present invention fill a need in present day and future technology for being able to determine an etch endpoint expeditiously, robustly, rehablv and reproduciblv under a variety of different conditions, even in real-time processing
• RIE reactive ion etching
• ICP inductively coupled plasma
• ECR electron cyclotron resonance
• Data compression of OES spectra as in any of the embodiments disclosed above may solve the set of problems is posed by the sheer number of OES frequencies or wavelengths available to monitor
• the monitoring typically generates a large amount of data
• a data file for each wafer monitored may be as large as 2-3 megabytes (MB), and each etcher can typically process about 500-700 wafers per day
• Conventional storage methods would require over a gigabytes (GB) of storage space per etcher per day and over 365 GB per etcher per year
• the raw OES data generated in such monitoring is typically "noisy" and unenhanced Compression ratios of 100 1, as in various of the illustrative embodiments disclosed above, would only require tens of MB of storage per etcher per day and only about 4 GB per etcher per year
• data compression and reconstruction of OES spectra as in any of the embodiments disclosed above may smooth and enhance the otherwise noisy and unenhanced raw OES data generated in etch monitoring
• Data compression of OES spectra as in any of the embodiments disclosed above may feature high compression ratios for the raw OES data ranging from about 20 1 to about 400 1, provide for efficient noise removal from the raw OES data and preserve relevant features of raw OES data
• Data compression of OES spectra as in any of the embodiments disclosed above may also have fast computation characteristics, such as fast compression and fast reconstruction, so as to be suitable for robust on-line, real-time implementation
• Data compression of OES spectra as in any of the embodiments disclosed above may further be compatible with real-time, PCA-based fault detection and classification (FDC), improve the accuracy and efficiency of etch processing, simplify manufacturing, lower overall costs and increase throughput
• FDC fault detection and classification

Landscapes

• Chemical & Material Sciences (AREA)
• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Plasma & Fusion (AREA)
• Analytical Chemistry (AREA)
• Health & Medical Sciences (AREA)
• General Health & Medical Sciences (AREA)
• Organic Chemistry (AREA)
• Life Sciences & Earth Sciences (AREA)
• Biochemistry (AREA)
• Genetics & Genomics (AREA)
• Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
• Gastroenterology & Hepatology (AREA)
• Medicinal Chemistry (AREA)
• Molecular Biology (AREA)
• Proteomics, Peptides & Aminoacids (AREA)
• Spectroscopy & Molecular Physics (AREA)
• Biophysics (AREA)
• General Physics & Mathematics (AREA)
• Immunology (AREA)
• Pathology (AREA)
• Manufacturing & Machinery (AREA)
• Computer Hardware Design (AREA)
• Microelectronics & Electronic Packaging (AREA)
• Power Engineering (AREA)
• Drying Of Semiconductors (AREA)

Priority Applications (3)

Applications Claiming Priority (6)

Publications (1)

Family

ID=27387343

Family Applications (1)

Country Status (5)

Cited By (11)

Families Citing this family (32)

Citations (5)

Family Cites Families (4)

• 2000
  • 2000-01-26 US US09/491,845 patent/US6582618B1/en not_active Expired - Lifetime
  • 2000-06-13 EP EP00946783A patent/EP1210724B1/en not_active Expired - Lifetime
  • 2000-06-13 DE DE60041408T patent/DE60041408D1/de not_active Expired - Lifetime
  • 2000-06-13 JP JP2001522570A patent/JP2003509839A/ja active Pending
  • 2000-06-13 WO PCT/US2000/016100 patent/WO2001018845A1/en not_active Ceased

Patent Citations (5)

Also Published As

Similar Documents

Legal Events