WO2006023489A2 - Cyclic error compensation in interferometry systems - Google Patents

Abstract

The invention features a method including: (i) providing an interference signal S(t) from two beams directed along different paths, wherein the signal S(t) is indicative of changes in an optical path difference n{tilde over (L)}(t) between the different paths, where n is an average refractive index along the different paths, {tilde over (L)}(t) is a total physical path difference between the different paths, and t is time; (ii) providing one or more coefficients representative of one or more errors that cause the signal S(t) to deviate from an ideal expression of the form A.sub.l cos(.omega..sub.Rt+.phi.(t)+.zeta..sub.l), where A.sub. l and.zeta..sub.l are constants,. omega.. sub.R is an angular frequency difference between the two beams before being directed along the different paths, and.phi.(t)=nk{tilde over (L)}(t), with k=2.pi./.lambda. and.lambda, equal to a wavelength for the beams; (iii) calculating a linear combination of values of the signal S(t); and (iv) reducing the effect of the deviation of S(t) from the ideal expression on an estimate of {tilde over (L)}(t) using an error signal generated from the coefficients and one or more error basis functions derived at least in part from the linear combination of values of the signal S(t).

Description

CYCLIC ERROR COMPENSATION IN INTERFEROMETRY

SYSTEMS

BACKGROUND

This invention relates to interferometers, e.g., displacement measuring and dispersion interferometers that measure displacements of a measurement object such as a mask stage or a wafer stage in a lithography scanner or stepper system, and also interferometers that monitor wavelength and determine intrinsic properties of gases. Displacement measuring interferometers monitor changes in the position of a measurement object relative to a reference object based on an optical interference signal. The interferometer generates the optical interference signal by overlapping and interfering a measurement beam reflected from the measurement object with a reference beam reflected from the reference object.

In many applications, the measurement and reference beams have orthogonal polarizations and different frequencies. The different frequencies can be produced, for example, by laser Zeeman splitting, by acousto-optical modulation, or internal to the laser using birefringent elements or the like. The orthogonal polarizations allow a polarizing beam splitter to direct the measurement and reference beams to the measurement and reference objects, respectively, and combine the reflected measurement and reference beams to form overlapping exit measurement and reference beams. The overlapping exit beams form an output beam that subsequently passes through a polarizer. The polarizer mixes polarizations of the exit measurement and reference beams to form a mixed beam. Components of the exit measurement and reference beams in the mixed beam interfere with one another so that the intensity of the mixed beam varies with the relative phase of the exit measurement and reference beams. A detector measures the time-dependent intensity of the mixed beam and generates an electrical interference signal proportional to that intensity. Because the measurement and reference beams have different frequencies, the electrical interference signal includes a "heterodyne" signal portion having a beat frequency equal to the difference between the frequencies of the exit measurement and reference beams. If the lengths of the measurement and reference paths are changing relative to one another, e.g., by translating a stage that includes the measurement object, the measured beat frequency includes a Doppler shift equal to 2vnp I λ , where v is the relative speed of the measurement and reference objects, λ is the wavelength of the measurement and reference beams, n is the refractive index of the medium through which the light beams travel, e.g., air or vacuum, and p is the number of passes to the reference and measurement objects. Changes in the relative position of the measurement object correspond to changes in the phase of the measured interference signal, with a 2π phase change substantially equal to a distance change L_RT of λ /(≈p) , where L_RT is a round-trip distance change, e.g., the change in distance to and from a stage that includes the measurement object.

Unfortunately, this equality is not always exact. Many interferometers include nonlinearities such as what are known as "cyclic errors." The cyclic errors can be expressed as contributions to the phase and/or the intensity of the measured interference signal and have a sinusoidal dependence on the change in optical path length pnL_RT . For example, a first order harmonic cyclic error in phase has a sinusoidal dependence on (2πpnL_RT) I λ and a second order harmonic cyclic error in phase has a sinusoidal dependence on 2 (2πpnL_RT ) / λ . Additional cyclic errors may include higher order harmonic cyclic errors, negative order harmonic cyclic errors, and sub-harmonic cyclic errors. Cyclic errors can be produced by "beam mixing," in which a portion of an input beam that nominally forms the reference beam propagates along the measurement path and/or a portion of an input beam that nominally forms the measurement beam propagates along the reference path. Such beam mixing can be caused by ellipticity in the polarizations of the input beams and imperfections in the interferometer components, e.g., imperfections in a polarizing beam splitter used to direct orthogonally polarized input beams along respective reference and measurement paths. Because of beam mixing and the resulting cyclic errors, there is not a strictly linear relation between changes in the phase of the measured interference signal and the relative optical path length pnL between the reference and measurement paths. If not compensated, cyclic errors caused by beam mixing can limit the accuracy of distance changes measured by an interferometer. Cyclic errors can also be produced by imperfections in transmissive surfaces that produce undesired multiple reflections within the interferometer and imperfections in components such as retroreflectors and/or phase retardation plates that produce undesired ellipticities in beams in the interferometer. For a general reference on the theoretical cause of cyclic error, see, for example, CW. Wu and R.D. Deslattes, "Analytical modelling of the periodic nonlinearity in heterodyne interferometry," Applied Optics, 37, 6696-6700, 1998.

In dispersion measuring applications, optical path length measurements are made at multiple wavelengths, e.g., 532 nm and 1064 nm, and are used to measure dispersion of a gas in the measurement path of the distance measuring interferometer. The dispersion measurement can be used to convert the optical path length measured by a distance measuring interferometer into a physical length. Such a conversion can be important since changes in the measured optical path length can be caused by gas turbulence and/or by a change in the average density of the gas in the measurement arm even though the physical distance to the measurement object is unchanged. In addition to the extrinsic dispersion measurement, the conversion of the optical path length to a physical length requires knowledge of an intrinsic value of the gas. The factor F is a suitable intrinsic value and is the reciprocal dispersive power of the gas for the wavelengths used in the dispersion interferometry. The factor F can be measured separately or based on literature values. Cyclic errors in the interferometer also contribute to dispersion measurements and measurements of the factor F . In addition, cyclic errors can degrade interferometric measurements used to measure and/or monitor the wavelength of a beam. The interferometers described above are often crucial components of scanner systems and stepper systems used in lithography to produce integrated circuits on semiconductor wafers. Such lithography systems typically include a translatable stage to support and fix the wafer, focusing optics used to direct a radiation beam onto the wafer, a scanner or stepper system for translating the stage relative to the exposure beam, and one or more interferometers. Each interferometer directs a measurement beam to, and receives a reflected measurement beam from, a plane mirror attached to the stage. Each interferometer interferes its reflected measurement beams with a corresponding reference beam, and collectively the interferometers accurately measure changes in the position of the stage relative to the radiation beam. The interferometers enable the lithography system to precisely control which regions of the wafer are exposed to the radiation beam. hi practice, the interferometry systems are used to measure the position of the wafer stage along multiple measurement axes. For example, defining a Cartesian coordinate system in which the wafer stage lies in the x-y plane, measurements are typically made of the x and y positions of the stage as well as the angular orientation of the stage with respect to the z axis, as the wafer stage is translated along the x-y plane. Furthermore, it may be desirable to also monitor tilts of the wafer stage out of the x-y plane. For example, accurate characterization of such tilts may be necessary to calculate Abbe offset errors in the x and y positions. Thus, depending on the desired application, there may be up to five degrees of freedom to be measured. Moreover, in some applications, it is desirable to also monitor the position of the stage with respect to the z- axis, resulting in a sixth degree of freedom.

SUMMARY

Among other aspects, the invention features electronic processing methods that characterize and compensate cyclic errors in interferometric data. Because cyclic errors are compensated electronically, the interferometry system that produces the data has greater tolerance to optical, mechanical, and electronic imperfections that can cause cyclic errors, without sacrificing accuracy. The compensation techniques are especially useful for interferometric data used to position microlithographic stage systems. hi part, the invention is based on the realization that prior values of a main interferometric signal can be used to reduce the effect of cyclic errors on an estimate of a length being measured by the interferometry system (e.g., a length indicating the position of a stage). Any of a number of signal transformations such as a quadrature signal or a Fourier transform can be derived from linear combinations of the prior values of the main interferometric signal. These derived signal transformations can then be used to generate one or more error basis functions that represent one or more cyclic error terms in the main interferometric signal. Appropriate amounts of each error basis function (e.g., as determined by a coefficient for each error basis function) form an error signal that is subtracted from a signal from which the measured length is obtained. The corresponding reduction in cyclic errors increases the accuracy of the measured length.

For example, the prior values of the main interferometric signal can be used to calculate an estimate for a quadrature signal for the main interferometric signal.

Algebraic combinations of such signals can yield error basis functions in the form of sinusoidal functions whose time-varying arguments correspond to particular cyclic error terms. In embodiments in which the interferometer beams have a heterodyne frequency splitting, one may also calculate the quadrature signal of the heterodyne reference signal, and the error basis functions may be derived from algebraic combinations the main signal, the reference signal, and the quadrature signals of the main and reference signals.

The error basis functions are used to isolate particular cyclic error terms in the main signal and characterize coefficients representative of each cyclic error term (e.g., its amplitude and phase). For example, algebraic combinations of the error basis functions and the main signal and its quadrature signal can move a selected cyclic error term to zero-frequency, where low-pass filtering techniques (e.g., averaging) can be used to determine its amplitude and phase. Such coefficients are stored. Thereafter, a superposition of the error basis functions weighted by the stored coefficients can be used to generate an error signal that can be subtracted from the main signal to reduce the cyclic errors therein and improve its accuracy.

The technique is particularly useful when the Doppler shift is small relative to the heterodyne frequency because the frequency of each cyclic error term is nearly equal to that of primary component of the main signal, in which case the estimate for the quadrature signal of the main signal is more accurate. This is an especially important property because it is precisely when the frequencies of the cyclic error terms are near that of the primary component of the main signal that the cyclic error terms are most problematic because they cannot be removed by frequency filtering techniques. Furthermore, at small Doppler shifts, one or more of the cyclic error frequencies maybe within the bandwidth of a servo system used to position a stage based on the interferometric signal, in which the case the servo loop may actually amplify the cyclic error term when positioning the stage. Small Doppler shifts are actually quite common in microlithographic stage systems, such as when searching for an alignment mark, scanning in an orthogonal dimension to the one monitored by the interferometric signal, and changing stage direction. Moreover, at small Doppler shifts, selecting an integral relationship between the sampling rate of the detector and the heterodyne frequency (e.g.,

6:1) yields an especially simple formula for the quadrature signal.

In addition, at small Doppler shifts, the main signal is nearly periodic with the heterodyne frequency, in which case prior data can be used to generate the error signal.

As a result, correction of the main signal can be accomplished with only a single real-time subtraction of the error signal from the main signal, significantly reducing the computation time associated with the correction and thereby reducing data age errors in any servo system for position a microlithography stage.

Error basis functions can be derived from other linear combinations of prior values of the main interferometric signal besides a quadrature signal including a Fourier transform of the main interferometric signal. In the case of a Fourier transform, the resulting error signal is subtracted from a complex signal from which the measured length is obtained, as described in more detail below.

We now summarize various aspects and features of the invention. In general, in one aspect, the invention features a first method including: (i) providing an interference signal S(t) from two beams directed along different paths, wherein the signal <S(t) is indicative of changes in an optical path difference nL{t) between the different paths, where n is an average refractive index along the different paths, L{t) is a total physical path difference between the different paths, and t is time;

(ii) providing coefficients representative of one or more errors that cause the signal S(t) to deviate from an ideal expression of the form A₁ cos(ω_R? + φ(f ) + ζ₁ ) , where A₁ and ζ₁ are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(t) = nkL(t) , with k = 2π/λ andλ equal to a wavelength for the beams; (iii) calculating a linear combination of values of the signal S(t) ; and (iv) reducing the effect of the deviation of S(t) from the ideal expression on an estimate of L(t) using an error signal generated from the coefficients and one or more error basis functions derived at least in part from the linear combination of values of the signal S(t). Embodiments of the first method may include any of the following features.

The method may further include directing the two beams along the different paths and measuring the interference signal S (t) . For example, at least one of the beams may be directed to reflect from a movable measurement object before producing the interference signal S (t) . Furthermore, the beam directed to contact the movable measurement object may reflect from the measurement object multiple times before producing the interference signal S(t) . Also, the beams may be directed to reflect from different locations of the movable measurement object before producing the interference signal S(t).

The errors may correspond to spurious beam paths.

The deviation may be expressed as where

p=l, 2, 3..., and m is any integer not equal to p, and where the provided coefficients include information corresponding to at least some of A_{m p} and C,_{m p} .

The angular frequency difference ω_R may be non-zero. The method may further include: providing a reference signal S_R (t) = A_R cos (ω_Rt + ζ_R ) , where A_R and ζ_R are constants; and calculating a quadrature reference signal S_R (t) based on the signal S_R (t) , wherein the error basis functions are derived from the signals S(t), S(t), S_R (t), and S_R (t) . For example, the method may further include measuring the reference signal S_R (t) based on an output from a common source of the two beams. Furthermore, the method may include generating reference signals S_R (t) and S_R (t) from a lookup table indexed synchronously with sampling of S(t).

Calculating the quadrature signal S(t) may include calculating the quadrature

signal S(t) based on the expression

where τ >0 and ω_M is an instantaneous rate of change of a phase of the interference signal S(t) . For example, calculating the quadrature signal S(t) may further include approximating ω_M according to ω^ ≈ ω_R + dφ(t)/dt , where φ(t) in the expression for ω _M is determined from the interference signal S(t) assuming the deviation of S(t) from the ideal expression is negligible. Alternatively, calculating the quadrature signal -S'(r) may further include approximating ω_M according to ω_M ≈ ω_R . In the latter case, calculating the quadrature signal S(t) may include calculating the quadrature signal S(t)

according to the simplified expression for

τ = (π + 6πN)/3ω_R , where N is a non-negative integer.

Calculating the quadrature reference signal S_R (t) may include calculating the quadrature reference signal S_R (t) based on the expression

For example,

calculating the quadrature reference signal S_R (t) may include calculating the quadrature reference signal S_R (t) according to the simplified expression

where N is a non-negative

integer.

The interference signal S[t) may be provided at a data rate that is an integer multiple of ω_R /2π . Each of the error basis functions may correspond to a function that includes one or more leading sinusoidal terms having a time-varying argument that corresponds to a time- varying argument of an error term that represents a portion of the deviation of S(t) from the ideal expression. For example, the error basis functions may correspond to one or more pairs of sine and cosine functions having an argument whose time-varying component has the form ω_Rt + (m/p)φ(t) , wherep is a positive integer and m is an integer not equal top. Alternatively, the error basis functions may correspond to one or more pairs of sine and cosine functions having an argument whose time-varying component has the form (m/ p) φ(t) , wherep is a positive integer and m is an integer not equal top.

A leading term for a function corresponding to one of the error basis functions may have the same values of m andp as a time-varying argument of one of the error terms.

Each of the error basis functions may be proportional to the function that includes one or more leading sinusoidal terms.

Each of the pairs of sine and cosine functions may correspond to imaginary and real parts, respectively, of a complex signal.

The error basis functions may correspond to multiple pairs of the sine and cosine functions. For example, the error basis functions may include multiple pairs of the sine and cosine functions from a family of the sine and cosine functions with

{(/, = \_{i ϊn} = -l),(p = \,m = 0),(p = \,m = 2),(p = \,m = 3),and(p = 2,m = \)} .

The method may further include generating the error basis functions from the signals S(t) and S(t) .

The method may further include generating the error basis functions from the signals S (t) , S(t), S_R (t) , and S_R (t) . For example, the error basis functions may be generated from algebraic combinations of the signals S(t), S(t), S_R (t), and S_R (t) . The method may further include generating the error signal S_ψ (t) . For example, the error signal t_ψ (t) may be generated from a superposition of the error basis functions weighted by the coefficients representative of the errors.

The method may further include determining a value for the optical path difference nL(t) from the interference signal S(t) after the effect of its deviations are reduced.

The quadrature signal S(t) may be calculated from the interference signal S(t) based on prior values of S(t) according to the approximation S(t) ≈ S(t - 2πN/ω_R ), where N is a positive integer. The error basis functions used to generate the error signal SL (t) may be derived from prior values of the signals S(t) and S (t) according to the approximations

S (t) ≈ S (t - 2πN/ω_R ) and S (t) ≈ S (t - 2πM/ω_R ) , where N and M are positive integers.

The angular frequency difference may satisfy ω_R > 100 - dφ(t)/cfr . Furthermore, it may satisfy ω_R > 500 • dφ(t)/dt .

In general, in another aspect, the invention features a second method. The second method is for estimating one or more coefficients representative of one or more errors that cause an interference signal iS(t) from two beams directed along different paths to deviate from an ideal expression of the form A₁ cos(ω_Rt + φ(t) + ζi ) , wherein the signal ■S'(t) is indicative of changes in an optical path difference nL[t) between the different paths, where n is an average refractive index along the different paths, L(t) is a total physical path difference between the different paths, t is time, A₁ and ζ, are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(ϊ) = nkL(t), with k = 2π/λ andλ equal to a wavelength for the beams. The third method includes: (i) calculating a calculating a linear combination of values of the signal S(t) ; and (ii) calculating an estimate for the coefficients based on filtering a signal derived at least in part from the linear combination of values of the signal S(t).

Embodiments of the second method may further include any of the following features.

Calculating the linear combination of values of the signal S(t) may include calculating a quadrature signal S (t) based on values of the signal S(t). Alternatively, calculating the linear combination of values of the signal S(t) may include calculating a transform signal D_q (t₁ ) based on values of the signal S{t) , for a frequency value determined by q, updated at a time t₁. For example, the transform signal D ( t₁ ) may include a sample of a discrete Fourier transform of the signal S (t) .

The method may further include: providing a reference signal S_R (t) = A_R cos(ω_Rt + ζ_R ) , where ω_R is non-zero and A_R and ζ_R are constants; and calculating a quadrature reference signal S_R (t) based on the signal S_R (t) , wherein the estimate for the coefficients is based on the signals S (t) , S (t) , S_R (t) , and S_R (t) .

Calculating the estimate for the coefficients may include generating error basis functions derived from the signals S(t) and S(t) .

Calculating the estimate for the coefficients may include generating error basis functions derived from the signals S(t), S(t), S_R (t), and S_R (t) . For example, the error basis functions may correspond to one or more pairs of sine and cosine functions having an argument whose time-varying component has the form ω_Rt + (tn/p)ψ(t) , where p is a positive integer and m is an integer not equal to p. Furthermore, the error basis functions may be generated from algebraic combinations of the signals S(t), S(t),

S_R (t), and S_R (t) . Also, the error basis functions may correspond to multiple pairs of the sine and cosine functions. For example, the error basis functions may include multiple pairs of the sine and cosine functions from a family of the sine and cosine functions with {(p = l,m = -\),(p = l,m = 0),(p = l,m = 2),(p = l,m = 3),and(p = 2,m = l)} . Calculating the estimate for the coefficients may include low-pass filtering algebraic combinations of the error basis functions and the signals S[t) and S(t). Calculating the estimate for the coefficients may include low-pass filtering algebraic combinations of the error basis functions and the signals S(t) and S(t).

For example, the low-pass filtering may include using a Butterworth filter. Embodiments of the second method may further include any of the features described above in connection with the first method.

In another aspect, the invention features an apparatus including a computer readable medium which during operation causes a processor to perform either of the first or second methods. In general, in another aspect the invention features a first apparatus including: (i) an interferometry system which during operation directs two beams along different paths and provides an interference signal S(t) from the two beams, wherein the signal S(t) is indicative of changes in an optical path difference nL(t) between the different paths, where n is an average refractive index along the different paths, L{t) is a total physical path difference between the different paths, and t is time, wherein imperfections in the interferometry system produce one or more errors that cause the signal S{t) to deviate from an ideal expression of the form A_\ cos(ω_Rt + φ(t) + ζ₁ ) , where Aχ and ζ, are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(t) = nkL(t), with k = 2π/λ andλ equal to a wavelength for the beams; and (ii) an electronic processor which during operation receives the interference signal S(^) from the interferometry system, receives one or more coefficients representative of the one or more errors, calculates a linear combination of values of the signal S(t), and reduces the effect of the deviation of S(t) from the ideal expression on an estimate of Z(t) using an error signal S_ψ (t) generated from the coefficients and one or more error basis functions derived at least in part from the linear combination of values of the signal S(t) . Embodiments of the first apparatus may include features corresponding to any of the features described above in connection with the first method.

In general, in another aspect, the invention features a second apparatus including: (i) an interferometry system which during operation directs two beams along different paths and provides an interference signal S(t) from the two beams, wherein the signal S(t) is indicative of changes in an optical path difference nZ(t) between the different paths, where n is an average refractive index along the different paths, L(t) is a total physical path difference between the different paths, and t is time, wherein imperfections in the interferometry system produce one or more errors that cause the signal S(t) to deviate from an ideal expression of the form A₁ cos(ω_Rt + φ(t) + ζ_] ) , where A₁ and ζj are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(t) = nkL(t), with k = 2π/λ andλ equal to a wavelength for the beams; and (ii) an electronic processor which during operation receives the interference signal S(t) from the interferometry system, calculates a linear combination of values of the signal 5(t), and calculates an estimate for one or more coefficients representative of the one or more errors based on filtering a signal derived at least in part from the linear combination of values of the signal S(t).

Embodiments of the second apparatus may include features corresponding to any of the features described above in connection with the second method.

In another aspect, the invention features a lithography system for use in fabricating integrated circuits on a wafer, the system including a stage for supporting the wafer, an illumination system for imaging spatially patterned radiation onto the wafer, a positioning system for adjusting the position of the stage relative to the imaged radiation, and either the first or second apparatus for monitoring the position of the wafer relative to the imaged radiation.

In another aspect, the invention features a lithography system for use in fabricating integrated circuits on a wafer, the system including a stage for supporting the wafer, and an illumination system including a radiation source, a mask, a positioning system, a lens assembly, and either the first or second apparatus, wherein during operation the source directs radiation through the mask to produce spatially patterned radiation, the positioning system adjusts the position of the mask relative to the radiation from the source, the lens assembly images the spatially patterned radiation onto the wafer, and the apparatus monitors the position of the mask relative to the radiation from the source.

In a further aspect, the invention features a beam writing system for use in fabricating a lithography mask, the system including a source providing a write beam to pattern a substrate, a stage supporting the substrate, a beam directing assembly for delivering the write beam to the substrate, a positioning system for positioning the stage and beam directing assembly relative one another, and either the first or second apparatus for monitoring the position of the stage relative to the beam directing assembly.

In another aspect, the invention features a lithography method for use in fabricating integrated circuits on a wafer, the method including supporting the wafer on a moveable stage, imaging spatially patterned radiation onto the wafer, adjusting the position of the stage, and monitoring the position of the stage using either the first or second method.

In a further aspect, the invention features a lithography method for use in the fabrication of integrated circuits including directing input radiation through a mask to produce spatially patterned radiation, positioning the mask relative to the input radiation, monitoring the position of the mask relative to the input radiation using either the first or second method, and imaging the spatially patterned radiation onto a wafer.

In yet another aspect, the invention features a lithography method for fabricating integrated circuits on a wafer including positioning a first component of a lithography system relative to a second component of a lithography system to expose the wafer to spatially patterned radiation, and monitoring the position of the first component relative to the second component using either the first or second method.

In a further aspect, the invention features a method for fabricating integrated circuits, the method including a foregoing lithography method. In another aspect, the invention features a method for fabricating integrated circuits, the method including using a foregoing lithography system.

In yet another aspect, the invention features a method for fabricating a lithography mask, the method including directing a write beam to a substrate to pattern the substrate, positioning the substrate relative to the write beam, and monitoring the position of the substrate relative to the write beam using either the first or second method.

As used herein "algebraic combinations" means combinations of operands (e.g., real or complex numbers including values of signals) according to one or more algebraic operations (e.g., addition, subtraction, multiplication, and division).

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. In case of conflict with publications, patent applications, patents, and other references mentioned incorporated herein by reference, the present specification, including definitions, will control.

Other features, objects, and advantages of the invention will be apparent from the following detailed description.

DESCRIPTION OF DRAWINGS

FIG. Ia is a schematic diagram of a processing unit for generating cyclic error basis functions and characterizing cyclic error coefficients based on a main interference signal S(t) and a reference signal S_R (t) . FIG. Ib is a schematic diagram of a processing unit for generating an error signal

<S_ψ (t) from the cyclic error basis functions and characterized coefficients and using the error signal to reduce cyclic errors in the main interference signal S(t) . FIG. Ic is a schematic diagram of an exemplary measurement system for generating cyclic error basis functions and characterizing cyclic error coefficients based on a complex measurement signal.

FIG. Id is a schematic diagram of an error estimator for the measurement system of FIG. Ic.

FIG. Ie is a schematic diagram of a processing unit for the error estimator of FIG. Id.

FIG. 2 is a schematic diagram of an M^th order digital filter for use in low-pass filtering algebraic combinations of the main signal, the reference signal, their quadrature signals, and the error basis functions to yield the cyclic error coefficients.

FIG. 3 is a schematic diagram of an interferometry system including a high- stability plane mirror interferometer (HSPMI).

FIG 4 is a schematic diagram of an embodiment of a lithography tool that includes an interferometer. FIG 5(a) and FIG 5(b) are flow charts that describe steps for making integrated circuits.

FIG 6 is a schematic of a beam writing system that includes an interferometry system.

Like reference symbols in the various drawings indicate like elements.

DETAILED DESCRIPTION

Embodiments include an electronic cyclic error compensation (CEC) procedure for compensation of cyclic error effects in interferometry applications, such as heterodyne interferometry. In preferred embodiments, the compensation is achieved for low slew rates of a plane mirror measurement object attached to a stage or attached to a reference system with associated interferometers attached to a stage. When optical techniques are used to eliminate and/or reduce the amplitudes of certain cyclic errors such as sub- harmonic cyclic errors to < 0.05 nra (3σ), the remaining cyclic errors of the harmonic type with amplitudes of 0.5 nm or less can be treated as having constant amplitudes with fixed offset phases and the required accuracy of the cyclic error compensation for a remaining cyclic error term is approximately 10 % to meet a compensated cyclic error budget of 0.05 nm (3σ) or less. Further, the number of cyclic error terms that need to be compensated electronically are typically a small number, e.g., of the order of 3. In preferred embodiments, the processing operations of CEC at high digital processing rates can be limited to a single add operation, whereas the remaining processing operations, which require additions, subtractions, multiplications, and divisions, can be performed at lower rates using prior values of the interference signal.

Typically, cyclic error effects in heterodyne interferometry can be eliminated by filtering the heterodyne signal in frequency space, e.g., using Fourier spectral analysis, when the Doppler shift frequencies can be resolved by the frequency resolution of a phase meter used to determine the heterodyne phase. Unfortunately, such filtering techniques cannot be used to eliminate cyclic error effects at low slew rates of the stage (including, e.g., zero speed of the stage) when the corresponding Doppler shift frequencies cannot be distinguished from the frequency of the primary signal. Further complications arise with cyclic error frequencies are within the bandwidth of the servo system, in which case the cyclic errors can be coupled directly into the stage position through the servo control system, and even amplify the error in the stage position from a desired position.

Specific details of embodiments of the CEC are described further below. In one approach, the CEC procedure processes real time-sampled values of a digitized measurement signal (DMS) generated by an analog-to-digital-converter (ADC).

Advantages of this "DMS approach" include a cyclic error correction signal that maybe generated in a "feed forward mode," where the feed forward mode can involve a simple discrete transform based on a translation in time and need not require a spectral analysis or the use of a discrete transform such as a discrete Fourier transform, such as a fast Fourier transform (FFT). Likewise, conjugated quadratures of the main interference signal and the reference signal can be generated by simple discrete transforms and need not require the use of a discrete transform such as a discrete Hubert transform. Moreover, the feed forward mode can reduce the number of computer logic operations that are required at the very highest compute rates and can thereby reduce errors in data age that are introduced by incorporation of CEC. In another approach, the CEC procedure processes complex values of a complex measurement signal (CMS) generated by a discrete Fourier transform (DFT) module after the ADC module. Advantages of this "CMS approach" include the ability to update the DFT (and the CEC computations) at a lower rate (e.g., 10 MHz) than the ADC sampling rate (e.g., 120 MHz). A reduction in the CEC update rate enables a simplified hardware architecture. For example, a reduction in CEC update rate by a factor of 12 can result in a hardware savings of greater than a factor of 12. The CMS approach also eliminates cyclic errors that are due to finite arithmetic precision of the samples generated by the ADC and of the DFT coefficients and calculations in the DFT module. The CMS approach is also less subject to noise than the DMS approach due to the number of samples and the window function used by the DFT module.

Another advantage of both the DMS approach and the CMS approach is that the cyclic error coefficients can be characterized at Doppler shift frequencies for which the phase meter cannot distinguish between the cyclic error frequencies from the frequency of the primary component of the interference signal. Furthermore, the cyclic error coefficients can be characterized and used for compensation over a range of Doppler shift frequencies that is small relative to heterodyne frequency, which is a range over which the cyclic error coefficients are typically frequency independent, thereby simplifying the cyclic error correction. Prior to describing preferred embodiments of the CEC, it is useful to classify cyclic errors according to properties with respect to the cyclic error amplitudes. Three classifications are set out with reference to a standard high stability plane mirror interferometer (HSPMI). They are: Type 1 - Constant amplitude cyclic errors; Type 2 - Variable amplitude cyclic errors; and Type 3 - Intermittent cyclic errors. The amplitudes of Type 1 cyclic errors are independent of the orientation of a plane mirror measurement object of a plane mirror interferometer.

The amplitudes of Type 2 cyclic errors are dependent on the orientation of the plane mirror measurement object with a relative variability similar to that experienced by the amplitude of the primary component of the respective electrical interference signal, e.g., < ± 20 %. The amplitudes of Type 3 cyclic errors are nominally zero except when the reflecting surface of the plane mirror measurement object is parallel to within ≈ 50 μrad of a conjugate image of a reflecting or partially reflecting surface, e.g., the reference mirror, of the interferometer. Examples of Type 1 cyclic errors are the harmonic cyclic errors generated by polarization mixing in the source of an input beam to the heterodyne interferometer, polarization mixing that is produced by a polarization beam-splitter having finite extinction ratios that splits the input beam into reference and measurement beams, and polarization mixing that is produced by a silver coated cube corner retroreflector such as commonly used in a high stability plane mirror interferometer (HSPMI). The amplitudes of Type 1 cyclic errors are typically ;$ 0.25 run using techniques such as described in U.S.

Patent Application Serial No. 10/174,149 entitled "Interferometry System and Method Employing an Angular Difference in Propagation Between Orthogonally Polarized Input Beam Components" to Peter de Groot and Henry A. Hill and filed July 17, 2002, in U.S. Patent No. 6,201 ,609 Bl entitled "Interferometers Utilizing Polarization Preserving Optical Systems" to Henry A. Hill, and in U.S. Patent No. 6,163,379 entitled "Interferometer with Tilted Waveplates for Reducing Ghost Reflections" to Peter de Groot, the contents of which is incorporated herein by reference.

Examples of Type 2 cyclic errors are the harmonic cyclic errors generated by spurious beams due to unwanted reflections by certain surfaces. The typical amplitude of a Type 2 cyclic error is 0.06 nm for a respective surface having a reflectivity of 0.0025. The amplitude of a particular Type 2 cyclic error will typically vary by g ± 20 % as the stage mirror is scanned through a range of for example ± 500 μrad in orientation with a physical separation of « 0.7 m between the polarizing beam-splitter of an interferometer and plane mirror measurement object.

Examples of Type 3 intermittent cyclic errors are sub-harmonic cyclic errors that have relatively large amplitudes, e.g. 2 nm, when the conditions for Type 3 cyclic errors are met. The Type 3 cyclic errors can be eliminated or reduced to below 0.025 nm (3σ) by optical techniques where elements of the interferometer are rotated or tilted to eliminate or reduce certain cyclic non-linear errors such as described in U.S. Patent Application Publication No. 2003/0038947 entitled "Tilted Interferometer" to Henry A. Hill, the contents of which are herein incorporated in its entirety by reference. The elimination of Type 3 cyclic errors by the optical techniques considerably reduces the task left for CEC in achieving a compensated cyclic error residual of 0.025 nm (3σ) or less. Of course, in further embodiments of the invention, sub-harmonic cyclic errors (such as the Type 3 half-cycle error) may also be compensated, as described further below.

The variability of Type 2 cyclic errors expressed in terms of a displacement is typically ≤ ± 0.010 nm in amplitude. As a consequence, Type 2 cyclic errors can be

treated as cyclic errors having constant amplitudes with constant offset phases in implementation of CEC that has a compensated cyclic error budget of 0.01 nm (3σ) or less.

The number of cyclic error terms that need to be compensated when the Type 3 cyclic errors are eliminated or reduced by the optical techniques are low, e.g. of the order of 3, for a compensated cyclic error budget of 0.05 nm (3σ) or less. A particular cyclic error term after elimination of Type 3 cyclic errors may comprise one or more Type 1 and/or Type 2 cyclic errors. Nonetheless, in further embodiments, the CEC may also be used to compensate Type 3 cyclic errors. Additional material describing quantifying and correcting cyclic errors are described in commonly owned U.S. Patent No. 6,252,668, U.S. Patent No. 6,246,481, U.S. Patent No. 6,137,574, U.S. Patent No. 6,747,744, and U.S. Patent No. 6,806,961, the entire contents each of which are incorporated herein by reference.

We now describe the DMS approach for the CEC that operates in a feed forward mode of operation, in which a cyclic error correction signal S_ψ (t) is subtracted from a corresponding electrical interference signal S(t) of an interferometer to produce a compensated electrical interference signal. The phase of the compensated electrical interference signal is then measured by a phase meter to extract relative path length information associated with the particular interferometer arrangement. Because cyclic error effects have been reduced, the relative path length information is more accurate. As a result, the compensated electrical interference phase can be used to measure and control through a servo control system the position of a stage, even at low slew rates including a zero slew rate where cyclic error effects can otherwise be especially problematic. The DMS approach is also described in published U.S. application no. 10/616,504 (publication number US 2004/0085545 Al).

Referring to FIGS. Ia and Ib, in a preferred embodiment, the CEC comprises two processing units. One processing unit 10 determines cyclic error basis functions and factors relating to the amplitudes and offset phases of cyclic errors that need be compensated. A second processing unit 60 of CEC generates cyclic error correction signal S_ψ (/) using the cyclic error basis functions and the factors relating to the amplitudes and offset phases determined by first processing unit 10. The first processing unit 10 of CEC for the first embodiment is shown schematically in Fig. Ia and the second processing unit 60 of CEC of the first embodiment is shown schematically in Fig. Ib. Referring now to FIG,. Ia, an optical signal 11 from an interferometer is detected by detector 12 to generate an electrical interference signal. The electrical interference signal is converted to digital format by an analog to digital converter (ADC) in converter/filter 52 as electrical interference signal S(t) and sent to the CEC processor. For example, the ADC conversion rate is a high rate, e.g., 120 MHz. In the present embodiment, we focus on a particular set of four cyclic error terms that are compensated at low slew rates. Adaptation to compensate for a different set of cyclic errors will be evident to one skilled in the art based on the subsequent description. The electrical interference signal S(t) comprising the four cyclic error terms can be expressed in the form

where

φ_R is the phase of a reference signal S_R (t) with dφ_R /dt = ω_R corresponding to 2π times the frequency difference of the measurement beam and reference beam components of the input beam to the interferometer; A₁ and ζ₁ are the amplitude and offset phase, respectively, of the primary component of the electrical interference signal; ε_₁ , ε₀ , ε₂ , and ε₃ are the amplitudes for the cyclic error terms; ζ_-1 , ζ₀ , ζ₂ , and ζ₃ are the offset phases of the cyclic error terms;

φ = 4kL (7)

for a plane mirror interferometer such as a HSPMI (which involves two passes of the measurement beam to the measurement object); k is a wavenumber corresponding to wavelength λ of beam 11; and L is the one way physical length of the measurement path relative to the one way physical length of the reference path of the interferometer. Cyclic error amplitudes ε_₁ , ε₀ , ε₂ , and ε₃ are much less than the A₁ , i.e. < (1/50) A₁ . An example of the frequency difference ω_R/2π is 20 MHz. Note that there is generally a set of cyclic error terms whose phases are independent of ψ_R . This set of cyclic error terms has been omitted from Equation (2) because they are eliminated by a high pass filter in converter/filter 52.

The factors relating to amplitudes ε_p and offset phases ζ_p of the cyclic error terms and the time dependent factors of the cyclic error terms are generated using measured values of both S(t) and reference signal S_R (t) . The factors relating to amplitudes ε_p and offset phases ζ_p are determined and the results transmitted to a table

40 for subsequent use in generation of the cyclic error correction signal SL (f) . The time dependent factors of the cyclic error terms are obtained by application of simple discrete transforms based on trigonometric identities and properties of conjugated quadratures of signals.

Optical reference signal 13 is detected by detector 14 to produce an electrical reference signal. The optical reference signal can be derived from a portion of the input beam to the interferometer. Alternatively, the electrical reference signal can be derived directly from the source that introduces the heterodyne frequency splitting in the input beam components (e.g., from the drive signal to an acousto-optical modulator used to generate the heterodyne frequency splitting). The electrical reference signal is converted to a digital format and passed through a high pass filter in converter/filter 54 to produce reference signal S_R (t) . Reference signal S_R (t) in digital format is written as

where A_R and ζ_R are the amplitude and offset phase, respectively, of the reference signal. The ADC conversion rate in 54 for S_R(t) is the same as the ADC conversion rate in 52 for S(t) . The quadrature signal S_R(t) of S_R(t) written as

is generated by electronic processing using measured values of S_R (t) according to the formula

where 1/τ is the ADC conversion rate of reference signal S_R (t) in 54. For the example of a frequency difference for ω_R /2π = 20 MHz and an ADC conversion rate l/τ of 120 MHz, Equation (10) reduces to a particularly simple form

Reference signal S_R(t) and quadrature signal S_R(t) are conjugated quadratures of reference signal S_R(t) . The quadrature signal S_R(t) is generated by processor 16 using Equation (11) or Equation (10) as appropriate. The quadrature signal S(t) of S(t) is generated by processor 56 using the same processing procedure as that described for the generation of quadrature signal S_R (t) . Accordingly, for the example of a ratio of 1/τ and ω_R/2π equal to 6,

where

Equation (12) is valid when the stage slew rate is low, e.g., when φ changes insignificantly over the time period 2τ . Signal S(t) and quadrature signal S(t) are conjugated quadratures of signal S(t) . Notably, the integral relationship between 1/τ and ω_R /2π allows the generation of feed forward values S'(t) and S'(t) of S(t) and S(t) , respectively, and S_R' (t) and S_R' (t) of S_R (t) and S_R (t) , respectively, are according to the formulae

where m is an integer such that the error in the phases of feed forward signals with respect to corresponding phases of signals is less than predetermined values set by an end use application. In other words, prior values of the main interference signal and the reference signal can be used to generate the quadrature signals and subsequent error basis functions. In other embodiments in which the ratio between 1/τ and ω_R /2π is an integer different from 6, Equations (14)-(17) are modified accordingly.

Using algebraic combinations of the signals S(t) , S(t) , S_R(t) , and S_R(t) , processing unit 10 generates cyclic error basis functions, which are sine and cosine functions that have the same time-varying arguments as the cyclic error terms, and then uses the cyclic error basis functions to project out respective cyclic error coefficients from

S(t) and S(t) by low-pass filtering (e.g., averaging). The cyclic error basis functions for S_ψ0 (t), for example, are especially simple and correspond to the reference signal and its quadrature signal S_R (t) , and S_R (t) . In other words, to process the signals for information about the cyclic error term ε₀ cos(φ_R +ζ₀) , signals S_R(t) and S_R(t) are used as time dependent factors in the representation of the cyclic error term ε₀ cos(φ_R + ζ₀) .

To better understand the representation, it is beneficial to rewrite cyclic error term ε₀ cos (φ_R + ζ₀ ) in terms of the time dependent functions cos (φ_R + ζ_R ) and sin (φ_R + ζ_R ) with the results

Conjugated quadratures S(t) and S(t) and conjugated quadratures S_R (t) and S_R (f) are transmitted to processor 20 wherein signals ∑_o (t) and ∑_o (t) are generated. Signals ∑o (') and ∑o (t) are given by the equations

Using properties of conjugated quadratures and certain trigonometric identities, e.g. , cos γ + sin γ = 1 , we have

Notably, the generation of signals ∑₀ (t) and ∑₀ (t) project the coefficients associated with S_ψ0(t) to zero frequency, where low-pass filtering techniques can be used to determine them. Thus, signals ∑₀ (t) and ∑₀ (t) are transmitted to low pass digital filters in processor 24, e.g., low pass Butterworth filters, where coefficients A_R A₀ and A_RB_Q are determined.

For a Butterworth filter T_n ( x) of order n , the corresponding outputs of the low pass digital filters for inputs ∑₀ [t) and ∑₀ (t) are

where θ(x) denotes a term of the order of x , ω_c is the - 3 dB angular cutoff frequency, and ω_D = dφ/dt .

The terms on the right hand sides of Equations (28) and (29) with factors A_RA₁ are the sources of the largest errors and accordingly determine the specifications of n and the minimum ratio for ω_D /ω_c that can be used when the outputs of processor 24 are stored in table 40. For a fourth order Butterworth filter, i.e., n = 4 , and a minimum ratio for ω_D /ω_c = 7 , the error terms on the right hand side of Equations (28) and (29) will generate errors that correspond to ≤ 0.010 nm (3 σ ). When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D/ω_c is satisfied, the outputs A_R A₀ and A_RB₀ of the low pass filters in processor 24 are stored in table 40 and in processors 26 and 28 under the control of signal 72 (from processor 70). Notably, in this preferred embodiment, ω_D can vary by factors such as 2 or more during the period associated with output values of A_RA_Q and A_RB_Q that are stored in table 40.

Quadratures S_R and S_R are transmitted to processor 22 and the value for A_R is determined in processor 22 by a procedure similar to that used in processor 20 and processor 24 using the formulae

The value of the order n need only be for example 2. Values of A_R are transmitted to table 40 and stored under the control of signal 72.

Alternatively, detector 14, converter/filter 54, and processor 16 can be replaced by a lookup table synchronized to the reference signal. This can potentially reduce uncertainty caused by noise on the reference signal. If the value of A_R is normalized to unity then some equations can be simplified and processor 22 can be removed. The values for A_RA₀ , A_RB₀ , S(t), S_R (t) S_R (t), S_R (t), and A_R are

transmitted to processor 26 and the values of A_RA_Q , A_RB₀ , and A_R are stored in processor 26 under the control of signal 72 for the generation of conjugated quadratures S₁ (t) and S₁ (t) where

The values for A_RA₀ , A_RB₀ , ∑_o (t) , and ∑_o (t) are transmitted to processor 28 and the values of A_R A₀ and A_RB₀ are stored in processor 28 under the control of signal 72 for the generation of conjugated quadratures ∑₁(t) and ∑₁(t) where

Signals S_R(t) , S_R(t) , ∑₁ (t) , and ∑₁ (t) are transmitted to processor 30 for the generation of conjugated quadratures ∑_₁ (t) and ∑_₁ (t) where

Signals ∑_-1 (t) and ∑_-1 (t) correspond to the cyclic errror basis functions for S_ψ-1(t) in that the leading terms of ∑_-1 (t) and ∑_-1 (t) are sinusoids with the same time- varying argument as that of S_ψ-1 (t) .

Coefficients A_R A₁A_₁ and -A_R A₁B_-1 are next determined through digital low pass filters, e.g., low pass Butterworth filters, in processor 32 where

The input signals for the digital filters are Σ₄ (t) and Σ₄ (t) . Input signals Σ₄ (t) and Σ₄ (t) are generated in processor 32 using signals S₁ , S₁ , ∑_-1 (t) , and ∑_-1 (t) according to the formulae

Equations (39) and (40) are written in terms of A_₁ and B_₁ using Equations (37) and (38) as

Signals Σ₄ (t) and Σ₄ (t) are sent to low pass digital filters in processor 32, e.g.,

low pass Butterworth filters, where coefficients 2A_R ²A₁A_₁ and -2A_R A₁B_₁ are determined. For a Butterworth filter T_n (x) of order n , the corresponding outputs of the low pass digital filters for inputs Σ₄ (t) and Σ₄ (t) are

The terms on the right hand sides of Equations (43) and (44) with factors A_R ² A₁ ²

are the sources of the largest errors and accordingly determine the specifications of n and the minimum ratio for ω_D/ω_c that can be used when the outputs of processor 32 are stored in table 40. For a fourth order Butterworth filter , i.e., n = 4 , and a minimum ratio for ω_D/ω_c = 3.5 , the error terms on the right hand side of Equations (43) and (44) will generate errors that correspond to < 0.010 nm (3 σ ). The outputs 2A_R A₁A_₁ and

-2AR²A₁B_-1 of low pass filters of processor 32 are divided by 2 to generate A_R A₁A_₁ and -A_R ²A₁B_₁ as the outputs of processor 32. When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D/ω_c is satisfied, the outputs A_R A₁A_₁ and -A_R ²A₁B_₁ of processor 32 are stored in table 40 and in processor 34 under the control of signal 72.

Signals S₁ (t) and S₁ (t) are transmitted to processor 30 for the generation of conjugated quadratures ∑₂ (t) and ∑₂(t) where

Signals ∑₂(t) and ∑₂(t) correspond to the cyclic error basis functions for S_ψ2(t) in that the leading terms of ∑₂(t) and ∑₂(t) are sinusoids with the same time-varying argument as that of S_ψ2(t).

Signals S₁, S₁, Z₁(O, ∑i(t), ∑₂(t), and ∑₂(t) and coefficients A_R ² A₁A_-1, and are -A_R ² A₁B_₁ transmitted to processor 34 and coefficients A_R A₁A_-1, and -A_R A₁B_₁ stored in processor 34 under the control of signal 72 for generation of conjugated quadratures Σ₃(t) and ∑₃(t) where

Signals Σ₃ (t) and Σ₃ (t) correspond to the cyclic error basis functions for S_ψ3 (t) in that the leading terms of Σ₃ (t) and Σ₃ (t) are sinusoids with the same time- varying argument as that of S_ψ3 (t) .

Coefficients A_RA₁ ² A₂ and -A_RA₁ ² B₂ (where A₂ and B₂ are cyclic error coefficients for S_ψ2 and are given explicitly by Equations (67) and (68), respectively, further below) are next determined through digital low pass filters, e.g., low pass Butterworth filters, in processor 38. The input signals for the digital filters are Σ₅ (t) and

Σ₅ (t) , respectively. The input signals are generated in processor 38 using signals Si , S₁ , Σ₂ (t) , and Σ₂ (t) according to the formulae

The expansions of Σ₅ (t) and Σ₅ (t) , given by Equations (51) and (52), respectively, in terms of cyclic error and non-cyclic error terms are

where A₂ and B₂ are given by Equations (67) and (68), respectively, shown further below.

Signals Σ₅ (t) and Σ₅ (t) are sent to low pass digital filters in processor 38, e.g.,

low pass Butterworth filters, where coefficients A_RA₁ ²A₁ and -A_R A₁ ²B₂ are determined. For a Butterworth filter T_n (x) of order n , the corresponding outputs of the low pass digital filters for inputs Σ₅ (t) and Σ₅ (t) are

The terms on the right hand sides of Equations (55) and (56) with factors A_RA_\ are the sources of the largest errors and accordingly determine the specifications of n and the minimum ratio for ω_D /ω_c that can be used when the outputs of processor 38 are stored in table 40. For a fourth order Butterworth filter , i.e., n = A , and a minimum ratio forcθ_£,/ω_c = 7 , the error terms on the right hand side of Equations (55) and (56) will generate errors that correspond to < 0.010 nm (3 σ). The outputs A_RAχ A₂ and

-A_nA ²B₂ of low pass filters of processor 38 are the outputs of processor 38. When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to a>_D/ω_c is satisfied, the outputs A_R A₁ A₂ and -A_R A_\ B₂ of processor 38 are stored in table 40 under the control of signal 72. Coefficients A_R ² Aχ A₂ and -A_n ²A₁ ³B₂ (where A₂ and B₂ are cyclic error coefficients for S_ψ3 and are given explicitly by Equations (69) and (70), respectively, further below) are next determined through a digital low pass filter, e.g., a low pass Butterworth filter, in processor 36. The input signals for the digital filters are Σ₆ (t) and

Σ₆ (t) , respectively. The input signals are generated in processor 36 using signals S₁ , S₁ , Σ₃ (t) , and Σ₃ (t) according to the formulae

The expansions of Σ₆ (t) and Σ₆ (t) given by Equations (57) and (58), respectively, in terms of cyclic error and non-cyclic error terms are

where A₃ and B₃ are given by Equations (69) and (70), respectively.

Signals Σ₆ (t) and Σ₆ (t) are sent to low pass digital filters in processor 36, e.g.,

low pass Butterworth filters, where coefficients A_R A₁ A₃ and -A_R AfB₃ are determined. For a Butterworth filter T_n [x) of order n , the corresponding outputs of the low pass digital filters for inputs Σ₆ (t) and Σ₆ (t) are

The terms on the right hand sides of Equations (61) and (62) with factors A_n ²A₁ are the sources of the largest errors and accordingly determined the specifications of n and the minimum ratio for ω_D /ω_c that can be used when the outputs of processor 36 are stored in table 40. For a fourth order Butterworth filter , i.e., n = 4 , and a minimum ratio foτω_D/ω_c = 3.5 , the error terms on the right hand side of Equations (61) and (62) will generate errors that correspond to ≤ 0.010 ran (3 σ ). The outputs A_n ²A₁ ³A₃ and

-A_n ²A ³B₃ of low pass filters of processor 36 are the outputs of processor 36. When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D /2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to (o_D/ω_c is satisfied, the outputs A_n ²A ³A₃ and -A_n ²A₁ B₃ of processor 36 are stored in table 40 under the control of signal 72.

Finally, quadratures S and S are transmitted from processors 52 and 56 respectively to processor 18 for the purpose of determining a value for A₁ . First, a signal S (t) S (t) + S (t) S (t) is generated where

The signal of Equation (63) is sent to a low pass digital filter in processor 18, e.g., a low pass Butterworth filter, where the coefficient A₁ is determined. For a Butterworth filter T_n (x) of order n , the corresponding outputs of the low pass digital filter for the signal of Equation (63) is

The accuracy required for the determination of A₁ is approximately 0.5% in order to limit errors generated in the computation of cyclic error signals S_ψ to ≤ 0.010

run (3 σ ). Therefore the error terms ε_-1 ² , ε₀ ² , ε₂ ² , and ε₂ ² on the right hand side of Equation (64) are negligible. The terms on the right hand side of Equation (64) of the

form are the sources of the largest Doppler shift frequency dependent errors

and accordingly determine the specifications of n and the minimum ratio for ω_D/ω_c that can be used when the output of processor 18 is stored in table 40. For a second order Butterworth filter , i.e., n = 2 , and a minimum ratio for ω_D /ω_c = 3.5 , the Doppler shift frequency dependent error terms on the right hand side of Equation (64) will generate errors that correspond to < 0.010 ran (3 σ ). The output A_{ ² of the low pass filter of processor 18 is the output of processor 18. When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D/ω_c is satisfied, the output A² of processor 18 is stored in table 40 under the control of signal

72.

Referring now to FIG. Ib, processor 60 generates the compensating error signal S_ψ . With respect to generating signal 5_ψ , it is beneficial to rewrite the ε_₁ , ε₀ , ε₂ , and ε₃ cyclic error terms of S_ψ in terms of the highest order time dependent terms of

∑_-1 (t) , ∑_-1 (t) , S_R, S_R, Σ₂ (t) , Σ₂ (t) , Σ₃ (t) , and Σ₃ (t) , i.e., cos(φ_R-φ-ζ₁+2ζ_R), sin(φ_R-φ-ζ₁+2ζ_R) , cos(φ_R+ζ_R), sin(φ_R+ζ_R), cos(φ_R+2φ+2ζ₁-ζ_R), sin (φ_R+2φ+2ζ₁-ζ_R), cos(φ_R + 3φ + 3ζ₁ -2ζ_R), and sin (φ_R + 3φ + 3ζ₁ - 2ζ_R ) as

Equation (65) for S_ψ is next written in the form

where A_₁ , B_₁ , A₀, and B₀ are given by equations(37), (38), (22), and (23), respectively, and

Compensation error signal S_ψ is generated in processor 44 using Equation (66), the coefficients transmitted from table 40 as signal 42, and the signals ∑_₁ (t) , ∑_₁ (t) ,

S_R , S_R , Σ₂ (t) , ∑₂ {t) , Σ₃ (t) , and Σ₃ (t) (which comprise the cyclic error basis functions) under control of control of signal 74 (from processor 70). Explicitly,

In other words, the compensation error signal is generated form a superposition of the error basis functions weighted by the cyclic error coefficients. The compensating signal iS_ψ is subtracted from signal S in processor 46 under control of signal 76 (from processor 70) to generate compensated signal S - S^ . Control signal 76 determines when signal S is to be compensated. The phase φ = AkL is then extracted from the compensated signal with a subsequent processor (not shown) to, for example, provide a more accurate measurement of the distance L. In the presently preferred embodiment, the error compensation signal S_ψ (t) is subtracted from prior values of the signals S(t) . For example, feedforward signal S'(t) , (as described in Equation (14)) may replace signal S(t) . The delay, m, of the feedforward signal is chosen to be equal to the processing delay in calculating the error basis functions and subsequently S_ψ (t) from signal S(t) . In this manner, S'(t) and S_ψ (t) represent the same time of input signal S(t) .

In further embodiments, the cyclic error coefficients may be stored and updated at a lower data rate than that used to generate the cyclic error basis functions from the feedforward values. In such cases, the stored values for the cyclic error coefficients may used for the calculation of the cyclic error basis functions as necessary. Of course, in yet further embodiments, the coefficients and/or the error basis functions can be calculated in real time, without the use of the feedforward signals. An important assumption in the preferred embodiment of the cyclic error compensation described above is that the Doppler shift frequency dφ{t)/dt is small enough relative to the heterodyne frequency ω_R , that the quadrature signal S (t) can be approximated (in analogy to Equation (10) by the expression:

or the simpler expression given by Equation (12). In further embodiments, the quadrature signal S(t) may be more accurately calculated according to:

where ω_M is the instantaneous rate of change of the phase of the main interference signal S(t) , which can be determined with sufficient accuracy by assuming that the cyclic error contributions to 5(t) are negligle.

Also, in further embodiments, the cyclic error compensation technique may be used for cyclic error terms different from those explicitly described in Equations (3)-(6). In particular, using algebraic combinations of the signals S(t) , S(t) , S_R (t) , and S_R (0 , a processing unit can generate cyclic error basis functions, which are sine and cosine functions that have the same time-varying arguments as the cyclic error terms that need to be compensated, and then use the cyclic error basis functions to project out respective cyclic error coefficients from S(t) and S(t) by low-pas filtering (e.g., averaging). For example, to determine the coefficients for a half-cycle cyclic error of the form:

one can calculate cyclic error basis functions for the half-cycle cyclic error as follows. First calculate signals Σ₇' (t) and Σ₇ (t) as:

Notably, the leading term of Σ₇ (t) is cos(φ/2 + ζ₁ /2 - ζ_R /2)| , and the leading term of

Σ₇ (t) is |sin (φ/2 + ζ₁ /2 -ζ_R /2)| . Zero phase crossings in Σ₇ (t) and Σ₇ (t) are then

measured to remove the absolute value operation and define Σ₇ (t) and Σ₇ (t) , which have leading terms cos(φ/2 + ζ₁/2 -ζ_R /2) and sin (φ/2 + ζ₁ /2 -ζ_R /2) , respectively. Half-cycle error basis functions ∑_1/2 (t) and ∑_1/2 (t) are then generated as:

where Σ,/₂ (t) and ∑y₂ (0 ^{nave as me}i^r leading terms sine and cosine functions, respectively, whose time-dependent argument is the same as that of S ty₂) • ^⁰ determine the coefficients for S /y₂\, one projects half-cycle error basis functions onto

S(t ) and S(t) to move the half-cycle cyclic error component to zero frequency, for example:

Low-pass filtering (e.g., with the Butterworth filter) of Σ₈ (t) and Σ₈ (t) then yield the half-cycle error coefficients in analogy to the extraction of the previously described cyclic error coefficients. In particular, the leading terms following the low-pass filtering are

A_RAy₂ cos[c,y₂ - ζ, /2 -ζ_R /2) and A_RAy₂ sin(ζy₂ - ζ, /2 -ζ_R /2), respectively.

We now describe the CMS approach for the CEC, in which a cyclic error correction signal D_ψ (t) is subtracted from a corresponding DFT signal D(t) of the electrical interference signal S(t) to produce a compensated DFT signal. FIG. Ic shows a simplified schematic diagram of a measurement using the CMS approach. The optical interference signal 111 is received and amplified by photoelectric receiver 112. The resulting electrical interference signal 113 is filtered by lowpass filter (LPF) 114 producing filtered signal 115. The LPF 114 is designed to prevent harmonics of the interference signal 111 from being aliased into the frequency range of interest. Filtered signal 115 is digitized by ADC 116, to produce digitized measurement signal 117. A typical ADC for a high perfoπnance displacement measuring interferometer may have 12 bits of resolution at sampling rates of 120 MHz. The digitized measurement signal 117 is processed by phase meter 120 (described below) to produce outputs magnitude 125 and phase 127 which represent the digitized measurement signal 117 as a transform. The magnitude output 125 is used for status and diagnostic purposes. The phase output 127 is used by position calculator 130 which is fully described in published U.S. Application No. 10/211,435 (publication number US 2003/0025914 Al), incorporated herein by reference. Position calculator 130 calculates measured position 131 and estimated speed 133. The measured position 131 is filtered by digital filter 136, which is fully described in U.S. Patent 5,767,972, incorporated herein by reference, to generate filtered position signal 137. Filtered position signal 137 represents the desired measurement of the distance !,.

Phase meter 120 includes a Discrete Fourier Transform (DFT) processor 122, a cyclic error compensation (CEC) calculator 140, and a Coordinate Rotation by Digital Computer (CORDIC) converter 124. Signals 123, 143, 145, and 147 are complex values, which consist of both a real component and an imaginary component, as a + jb , where a is the real component, b is the imaginary component, and j is √-1 . (The symbol i is sometimes used in the literature instead of j .) Other representations of complex or quadrature values can be used, and may be expressed using other symbols such as, for example, / and Q , or X and Y , or A and A . Complex values may be converted from rectangular (real and imaginary) representation to polar (magnitude and phase angle) representation. The numeric representation of the digital signals may be integer, fractional, or floating point.

The DFT processor 122 converts a series of consecutive samples of digitized measurement signal 117 into a complex measurement signal 123 representing a transform of the digitized measurement signal 117 at a selected center frequency of DFT processor 122. The center frequency is determined by control circuitry (not shown) and the estimated speed 133 is determined by position calculator 130.

An exemplary DFT processor 122 is a 72-point windowed DFT performed at a 10 MHz rate. Since the DFT calculation is being updated in time, the complex measurement signal 123 is expressed as a function of time D(t) • This 10 MHz update rate provides 83% overlap of samples of the digitized measurement signal 117 between one DFT calculation and the next for an ADC sampling rate of /_$■ = 120 MHz. A typical window function is the Blackman window, which reduces errors due to the discontinuities at the beginning and end of the series of digitized measurement signal samples used for the DFT.

The CEC calculator 140 calculates and compensates for certain of the cyclic errors. CEC error estimator 144 (described in more detail below with reference to FIG. Id) calculates complex error compensation signal 145. Optional delay 142, and other delays (not shown) in CEC calculator 140 may be used to match the processing delay of the various calculations. Adder 146 combines delayed complex measurement signal 143 with complex error compensation signal 145 to produce compensated complex measurement signal 147, in which the certain cyclic error signals are substantially reduced.

CORDIC converter 124 converts the compensated complex measurement signal 147 to magnitude 125 and phase 127.

Referring to FIG. Id, the CEC error estimator 144 includes two processing units. One processing unit 148 determines error basis functions and complex factors relating to the amplitudes and offset phases of the certain cyclic errors that need be compensated. A second processing unit 204 generates complex error compensation signal Z)_ψ (t) 145 using the error basis functions and complex factors relating to the amplitudes and offset phases determined by first processing unit 148.

The first processing unit 148 for one embodiment is shown schematically in Fig. Id and the second processing unit 204 of this embodiment is shown schematically in Fig. Ie. These processing units are incorporated into the architecture shown in Fig Ic that may also include any of a variety of other techniques such as a glitch filter (as described in published U.S. Application No. 10/211,435), dynamic data age adjustment (as described in U.S. Patent No. 6,597,459, incorporated herein by reference), and digital filtering as described in 5,767,972.

Note that, as in the DMS approach, the set of cyclic error terms whose phases are independent of φ_R has been omitted from Equation (2) because, in this CMS approach, they are eliminated by LPF (Lowpass filter) 114 and the bandpass filtering inherent in DFT processor 122.

Note that there are conditions where the cyclic error terms ε₂ or ε₃may be aliased and appear at or near the frequency of the measurement signal. Let f_R be the reference frequency ( f_R = ω_R /2π ), f_D be the Doppler shift frequency ( f_D = ω_D /2π ), and f_M be the resulting measurement frequency. When f_D = -(2/3)/_Λ ,

Also, when f_D = - (1/2) f_R ,

In both cases, the negative frequency is indistinguishable from the corresponding positive frequency, and measurement accuracy is affected.

Complex factors relating to amplitudes z_p and offset phases ζ^ of the four cyclic error terms and time dependent factors of the cyclic error terms are generated using processed values D(t) 123 from DFT processor 122. The factors are stored in registers 162, 176, 186, and 192 for subsequent use in generation of the cyclic error correction signal Z)_ψ (t) 145. The time dependent factors of the cyclic error terms are obtained by application of digital transforms based on trigonometric identities and properties of complex signals.

DFT processor 122 calculates the complex DFT of the digitized measurement signal 117 as:

(83) for 0 < q < N - 1 , where τ = 1 / f_s , W_n is a window function centered at and

t_] is the time at which the DFT calculation is updated. When q is selected by control circuitry (not shown) as an integer approximately equal to N/_MZ/_S, to correspond to the center frequency of the primary component of the digitized measurement signal. A typical value for N is 72, and a typical window function W_n is the Blackman window function.

The equation for the DFT is arranged so that the center of the sequence of the N consecutive samples of the digitized measurement signal 117 (n = 35.5 in this case) corresponds to zero phase and the center of the window function. Therefore, since the window function and DFT coefficients (the the complex exponential or equivalently the "cos+ j sin " factors shown below) are symmetrical about the center, and window weighting is zero at the ends, a change in the value of q , has a reduced effect on the phase of the result.

The transform signal D_q (t_\ ) is updated at a rate f_v that is lower than the rate f_s at which the signal S{t) is sampled. In this example, t₁ = /At₁ (where / is an integer and At₁ ≡ \l fu is the update interval) and f_υ = 10 MHz.

Alternatively, the DFT equation can be "folded" to reduce the number of multiplication operations that are performed and calculated as:

for 0 < q < ZN - 1 , where x_n (t, ) for n = 0, ..., N - 1 are N consecutive samples (i.e., t = nτ ) of S(t + t_λ ) , and W_n is centered at n = 0. In this example, the larger range of q yields a more finely spaced resolution of 1/8 bin to reduce amplitude variations (or "picket fence" effect) as the frequency changes from one bin to the next. The DFT function is equivalent to a mixing and a filtering operation. The mixing is a result of multiplying the input data by the complex exponential or equivalently the " COS+ j sin " factor. The filtering is a result of the summation and the window function W_n .

Since the window function W_n makes terms zero outside the range of summation, the DFT expression can be written as a sum over all n. In a simplified expression for an "unfolded" DFT (as in equation (83)), the digitized measurement signal S{nτ + t₁ ) 177 has been approximated by its primary term A₁ cos(φ_R (nτ + t₁ ) + φ(nτ + t₁ ) + ζ₁ ) , where explicit sampling (t = nτ ) has been included:

where φ_R (nτ + t₁ ) = ω_Rnτ + ω_Rt₁ and φ(nτ + t₁) = ω_Dnτ + ω_Dt₁ + φ₀ is the sampled

version of the Doppler phase shift φ(t) = (ύ_Dt + φ₀, and is the phase of

the DFT with q ≈ Nf _M I f_s , yielding ψ_c(n,q) ≈ 2πf_Mnτ = ω_Mnτ . (The q in D_q(t_l) has been suppressed so that D(t₁) ≡ D_q (t₁ ) .) Using trigonometric identities, equation (85) can be expanded to (with time dependent arguments temporarily suppressed):

The terms containing φ_R + φ + ζ, + φ_c are high frequency sinusoids varying with n that are filtered out in the summation including the window function W_n that covers many cycles. The constant 1/2 may be dropped for convenience. The terms containing φ_R" + φ + ζ₁ - φ_c that are slowly varying in the summation over the window remain:

Since ω_M ≈ ω_R + ω_o the terms of the sinusoids including n drop out.

When the update interval Δt₁ ≡ 1 / f_U is given by Δt₁ = m / f_R (where m is an integer) the value of t, is a multiple of 1 / f_R . Under this condition, for low stage velocities (i.e., ω_D ≪ ω_R ) ω_Rt₁ ≈ m2π and DFT calculation becomes:

The DFT result is a complex value which can represent either a positive or a negative frequency. Equation (88) could be written as a sum, rather than a difference, with appropriate changes in sign in other equations. This choice of sign is not important, provided that the choices result in a reduction in cyclic error.

Equation (88) can be combined with equations (1) to (6). This results in:

where φ(r, ) is the value of the Doppler phase shift q>(t₁ ) = co^f, + φ₀ that is updated with t₁ . For convenience, in the equations (98) - (116) below, r, is written simply as t and φ(t₁) is written simply as φ . Also, complex measurement signal D[t) 123 and complex error compensation signal D_ψ (t) 145 are assumed to be updated at the rate f_υ such that D (t) ≡ D (t₁ ) and D_ψ (t) ≡ D_ψ (t, ) .

Figure Id shows a schematic diagram of CEC error estimator 144. In the following discussions the product of two complex signals frequently occurs. The following identities are used:

The next step is the processing of signals for information about the cyclic error term D_ψ0. The signal D(i) is sent to LPF (Lowpass Filter) 160, for example an IIR

(Infinite Impulse Response) Butterworth digital filter, an FIR (Finite Impulse Response), or CIC (Cascaded Integrator Comb) digital filter as described by Hogenauer (An Economical Class of Digital Filters for Decimation and Interpolation; E. B. Hogenauer; IEEE Transactions on Acoustics, Speech, and Signal Processing; VoI ASSP-29, No 2, April 1981, p 155-162, incorporated herein by reference). The CIC filter has the advantages in this implementation of simple design (using only integer addition) and decimation by large ratios. The implementation of an LPF for a complex signal uses two identical real LPF functions, one is used for the real component, and one is used for the imaginary component. The use of digital functions ensures precise matching of amplitude and phase response of the two filters. For an LPF T_n (x) of order n, the complex output C₀ is approximately:

where O(x) denotes a term of the order of x , ω_c is the - 3 dB angular cutoff frequency, and ω_D = dφ/dt .

The term on the right hand sides of Equation (98) with factor A_\ is the source of the largest error and accordingly determines the specifications of n and the minimum ratio for ω_D/ω_c that can be used when the outputs of LPF 160 are stored in register 162. For a fourth order LPF, i.e., n = 4 , and a minimum ratio for G)₁, /ω_c = 7 , the error terms on the right hand side of Equation (98) will generate errors that correspond to < 0.010 nm

(3 σ ). When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D /ω_c is satisfied, the output C₀ of the LPF 160 is stored in register 162 as C_0Λ under the control of signal 161.

An advantage of this approach is that ω^, can vary by factors such as 2 or more during the period when output values of C₀ are stored in register 162.

The values for C₀^ and D(t) are transmitted to processor 164 for the generation of complex signal E₁ if) where

Signal E₁ is sent to processor 168, which calculates ∑_-1 as the complex conjugate of ∑₁.

Signal E₁ is sent to processor 180, which calculates S₂.

Signal E₂ is divided by two and sent to LPF 190, as described earlier for LPF

160.

For an LPF T_n (x) of order n, the complex output C₄ is approximately:

where O(x) denotes a term of the order of x , ω_c is the -3 dB angular cutoff frequency, and ω_D = dφ/dt . The term on the right hand sides of Equation (102) with factors A₁ ² is the source of the largest error and accordingly determines the specifications of n and the minimum ratio for ω_D/ω_c that can be used when the outputs of processor 190 are stored in register 192. For a fourth order LPF , i.e., n = 4 , and a minimum ratio for ω_D / ω_C = 3.5 , the error terms on the right hand side of Equation (102) will generate errors that correspond to ≤ 0.010 nm (3 σ ). When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D /ω_c is satisfied, the output C₄ of LPF 190 is stored in register 192 as C_4R under the control of signal 161. Signals ∑₁ , ∑₂ , and C_4R are sent to processor 200 which calculates ∑₃. Signal

∑_3A is calculated by combining signals C_4R and Z₁ using multiplier 194:

Signal ∑_3B is calculated by combining signals ∑, and ∑₂ using multiplier 202:

Signal ∑₃ is calculated by combining signals ∑_{3 A} and ∑_3B using constant multiplier 196 and subtracter 198:

Signal ∑₅ is calculated by combining signals ∑_₁ and ∑₂ using multiplier 182:

Signal Σ₅ is sent to LPF 184, as described earlier for LPF 160.

For an LPF T_n (x) of order n, the complex output C₅ is approximately:

where O(x) denotes a term of the order of x , ω_c is the - 3 dB angular cutoff frequency, and ω_D = dφ/dt . The term on the right hand sides of Equation (107) with factor A^ is the source of the largest errors and accordingly determines the specifications of n and the minimum ratio for ω_D /ω_c that can be used when the outputs of LPF 184 is stored in register 186.

For a fourth order LPF , i.e., n = 4 , and a minimum ratio for ω_D/(ύ_c = 7 , the error terms on the right hand side of Equation (107) will generate errors that correspond to < 0.010 nm (3 σ ). When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D /ω_c is satisfied, the output C₅ of LPF 184 is stored in register 186 as C_5R under the control of signal 161.

Signal ∑₆ is calculated by combining signals ∑_₁ and ∑₃ using multiplier 172:

Signal Σ₆ is sent to LPF 174, as described earlier for LPF 160.

For an LPF T_n (x) of order n, the complex output C₆ is approximately:

where O(x) denotes a term of the order of x , ω_c is the - 3 dB angular cutoff frequency, and ω_D = dφ/dt . The term on the right hand sides of Equation (107) with factor A * is the source of the largest errors and accordingly determines the specifications of n and the minimum ratio for ca_D /ω_c that can be used when the outputs of LPF 174 is stored in register 176.

For a fourth order LPF , i.e., n = 4 , and a minimum ratio for ω_D/ω_c = 3.5 , the error terms on the right hand side of Equation (109) will generate errors that correspond to g

0.010 nm (3 σ ). When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D/ω_c is satisfied, the output C₆ of LPF 174 is stored in register 176 as C_6R under the control of signal 161. The magnitude squared of signal D(J) is calculated by multiplier 152:

This may be simplified and implemented as:

Alternate methods to calculate a value equivalent to the result of equation (111) are possible, for example the magnitude output 125 from CORDIC converter 124 could be squared, or signal E₁ could be used instead of Z)(O •

When equation (111) is evaluated using the terms of D(O from equations (89) to (94), the following result is obtained:

The signal C₁ (t) is sent to LPF (Lowpass Filter) 154 as described earlier for LPF 160.

For an LPF T_n (x) of order s, the output C₁ is approximately:

The accuracy required for the determination of C₁ is approximately 0.5% in order to limit errors generated in the computation of cyclic error signals S_ψj to < 0.010 ran

(3 σ ). Therefore the error terms ε_₁ ² , ε₀ ² , ε₂ , and ε₂ ² on the right hand side of Equation (113) are negligible. The terms on the right hand side of Equation (113) of the

form I are the sources of the largest Doppler shift frequency dependent errors

and accordingly determine the specifications of n and the minimum ratio for ω_D/ω_c that can be used when the output of LPF 154 is held in register 156, providing signal Ci R. For a second order LPF , i.e., n = 2 , and a minimum ratio for ω_D/ω_c = 3.5 , the Doppler shift frequency dependent error terms on the right hand side of Equation (113) will generate errors that correspond to £ 0.010 nm (3 σ ). When the stage is moving at a speed such that the corresponding Doppler shift frequency ω_D/2π is 10 to 100 times greater than the bandwidth of the stage servo control system and the requirement with respect to ω_D/ω_c is satisfied, the output C₁ of LPF 154 is stored in register 156 as C_{1 R} under the control of signal 161.

Processor 204 calculates compensating signal D_ψ(t) as shown in FIG. Ie and

Equation (114).

This can be expanded using the most significant terms from the previously expressed equations (98), (113), (102), (107), and (109) for the C^ terms, and equations

(100), (101), and (105) for the ∑_N terms to:

and simplified to:

Note that the calculated result in equation (116) is identical to D_ψ (t) as described in equations (90) to (94). Control signal 211 enables the output of D_ψ (t) when conditions indicate that a valid compensation signal is present. Control signal 211 is generated by control circuitry (not shown) that may consider the speed of motion, the signal level, the cyclic error level, the duration of motion, and possible other conditions to determine the validity or usefulness of D_ψ (t) . It may also be desirable to enable each of the four cyclic error terms separately, for example if the ε₃term was found to be insignificant for a specific implementation, the calculation could be disabled to reduce noise contributed by the calculation. When a £>_ψ (t) term or output is disabled, zero is substituted.

The calculated D_ψ (t) is subtracted from signal D(t) by subtractor 146, resulting in a measurement signal with substantially reduced cyclic errors. Using either the DMS approach or the CMS approach, other techniques can be incorporated to compensate for some of the sources of cyclic errors. For example, the dominant error term is typically the unshifted cyclic error component, S^_Q , or equivalently D_ψ0 , which stays at constant phase and frequency regardless of stage motion. This term arises from the presence of both optical frequencies in either the reference arm or the measurement arm of the displacement measuring interferometer or both. This occurs, for example, if the optical frequencies of the light source are not perfectly separated into orthogonal linear polarization states.

Though the cyclic error compensation techniques are described using the double pass plane mirror interferometer by way of example, they can be applied to any two- frequency, displacement measuring interferometer in which the cyclic error term which does not experience Doppler shift is dominant.

Compensation for this unshifted term can be achieved by momentarily blocking the beam in the reference path of the interferometer and then in the measurement path of the interferometer at two different times, during, for example, interferometer installation; measuring and recording the complex amplitude of the measurement signal in the beam- blocked condition; adding the two different complex amplitudes so obtained to obtain a compensation coefficient representing the unshifted term that is present when neither beam is blocked; then using the complex coefficient to compensate for the unshifted term during operation of the electronic compensation techniques described herein with the automatic coefficient updating disabled.

The technique of beam blocking can either be manual or automated using, for example, motorized shutters in the reference and measurement paths of the interferometer which are controlled by the same computer that reads the complex amplitudes in the beam-blocked condition; calculates the sum; and stores the coefficient for use in the electronic compensation techniques described herein.

Finally, it is noted that, if desired, the accuracy of the cyclic error correction can further be improved to higher order in the amplitude of the cyclic error coefficients by iterating the compensation of the main signal S(t) . In other words, for each subsequent iteration the compensated signal for the main signal is used to generate corresponding cyclic error basis functions and determine a higher order correction to each of the cyclic error coefficients.

Referring now to FIG. 2, a block diagram is shown for an M^th order digital filter used in the cyclic error compensation described above. The figure is in the "Direct Form I" representation standard to the digital signal processing community. The input discrete time series is x(n) and the output discrete time series is y(ή) . The z^"1 operation indicates a one-sample delay. A time-domain representation of the filter takes the form:

The coefficients α, and b_t are selected to produce the desired properties for the filter. For the case of the Butterworth filter, the a_t and b_t coefficients are selected to produce the frequency filtering given by Equations 28, 29, 43, 44, 55, and 56. Furthermore, other embodiments of the cyclic error compensation may implement different low-pass filtering schemes to yield the coefficients of low-frequency terms. The Butterworth filter, and other low-pass digital filters, are well known in the art. See, for example: Oppenhiem, A.V., Schafer, R.W., and , J.R. Buck, "Discrete-Time Signal Processing", Upper Saddle River, NJ: Prentice Hall, 1999; and Proakis, J.G., and , D.G. Manolakis, "Digital Signal Processing: Principles, Algorithms, and Applications", New York, NY: MacMillan, 1992.

Depending on the embodiment, the compensation technique described above can be implemented in hardware or software, or a combination of both. The technique can be implemented in computer programs using standard programming techniques following the method and figures described herein. Program code is applied to input data to perform the functions described herein and generate output information. The output information is applied to one or more output devices such as the servo control system.

Each program may be implemented in a high level procedural or object oriented programming language to communicate with a computer system, or the programs can be implemented in assembly or machine language, if desired. In any case, the language can be a compiled or interpreted language. Moreover, the program can run on dedicated integrated circuits preprogrammed for that purpose.

Each such computer program may be stored on a storage medium or device (e.g., ROM or magnetic diskette) readable by a general or special purpose programmable computer, for configuring and operating the computer when the storage media or device is read by the computer to perform the procedures described herein. The computer program can also reside in cache or main memory during program execution. The compensation technique can also be implemented as a computer-readable storage medium, configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner to perform the functions described herein. Now referring to FIG. 3, an interferometry system including a high stability plane mirror interferometer (HSPMI) 411 is shown for optical generating the main interference signal. The HSPMI 411 includes a polarization beam-splitter 330, a retroreflector 332, quarter wave phase retardation plates 334 and 336, and a plane mirror reference object 342. Input beam 422 is a two-component beam. The two components have different frequencies and are orthogonally plane polarized. The different frequencies can be produced in source 415, for example, by laser Zeeman splitting, by acousto-optical modulation, or internal to the laser using birefringent elements or the like. HSPMI 411 splits input beam 422 into two components. One component, shown as first and second pass measurement beams 322 and 324, reflects from measurement object 490 twice before exiting HSPMI 411. The other component, shown by first and second pass reference beams 328 and 327, reflect from reference mirror 342 twice before exiting HSPMI 411. The exiting beam components overlap and form output beam 423.

An electrical interference signal 352 is generated by the detection of output beam 423 in detector 420. Detector 420 includes a polarizer to mix the reference and measurement beam components of output beam 423 with respect to polarization. Electrical interference signal 352 contains a heterodyne interference signal corresponding to main interference signal S(t) .

In further embodiments, the interferometry system may be different than that shown in FIG 3. In particular, the cyclic error compensation technique is applicable to other types of interferometers as well. For example, the main interference signal S(t) may be produced by an interferometry system that may include any of single and/or multiple pass interferometers, passive interferometers, dynamic interferometers, and dispersion interferometers. Furthermore, the interferometry system may monitor one or more degrees of freedom, each of which may produce a corresponding main interference signal S(t) , which may be compensated for cyclic errors as disclosed herein. Furthermore, the degree(s) of freedom monitored by the interferometry system may include any of changes in distance to a measurement object, changes in relative distance between two measurement objects, changes in the angular orientation of a measurement object, and changes in the direction of the input beam. Examples of dynamic interferometers are described in U.S. Patent Application Serial No. 10/226,591 filed August 23, 2002 and entitled "DYNAMIC INTERFEROMETER CONTROLLING DIRECTION OF INPUT BEAM" by Henry A. Hill. Examples of passive zero shear interferometers are described in U.S. Patent Application Serial No. 10/207,314, entitled "PASSIVE ZERO SHEAR

INTERFEROMETERS," filed July 29, 2002, by Henry A. Hill. Examples of angular displacement interferometers are described in: U.S. Patent Application Serial No. 10/226,591 entitled "DYNAMIC INTERFEROMETER CONTROLLING DIRECTION OF INPUT BEAM," filed August 23, 2002; U.S. Provisional Application 60/314,345 filed August 22, 2001 and entitled "PASSIVE ZERO SHEAR INTERFEROMETERS USING ANGLE SENSITIVE BEAM-SPLITTERS," both by Henry A. Hill, and U.S. Patent Application Serial No. 10/272,034 entitled "INTERFEROMETERS FOR MEASURING CHANGES IN OPTICAL BEAM DIRECTION" and filed October 15, 2002 by Henry A. Hill and Justin Kreuzer. Alternatively, or additionally, interferometry systems may include one or more differential angular displacement interferometers, examples of which are also described in U.S. Patent Application Serial No. 10/272,034. Examples of interferometry systems for measuring more than one degree of freedom and for reducing beam shear are described in U.S. Patent Application Serial No. 10/352,616 filed January 28, 2003 and entitled "MULTIPLE-PASS INTERFEROMETRY" by Henry A. Hill and U.S. Patent Application Serial No. 10/351,708 filed January 27, 2003 and entitled

"MULTI-AXIS INTERFEROMETER" by Henry A. Hill. Other forms of multiple pass interferometers are described in an article entitled "Differential interferometer arrangements for distance and angle measurements: Principles, advantages and applications" by C. Zanoni, VDI Berichte Nr. 749, 93-106 (1989). Examples of two- wavelength dispersion interferometers are described in U.S. Patent No. 6,219,144 Bl entitled "APPARATUS AND METHOD FOR MEASURING THE REFRACTIVE INDEX AND OPTICAL PATH LENGTH EFFECTS OF AIR USING MULTIPLE-PASS INTERFEROMETRY" by Henry A. Hill, Peter de Groot, and Frank C. Demarest and U.S. Patent No. 6,327,039 Bl by Peter de Groot, Henry A. Hill, and Frank C. Demarest. Because of the cyclic error compensation, the interferometry systems described herein provide highly accurate measurements. Such systems can be especially useful in lithography applications used in fabricating large scale integrated circuits such as computer chips and the like. Lithography is the key technology driver for the semiconductor manufacturing industry. Overlay improvement is one of the five most difficult challenges down to and below 100 nm line widths (design rules), see, for example, the Semiconductor Industry Roadmap, p.82 (1997).

Overlay depends directly on the performance, i.e., accuracy and precision, of the distance measuring interferometers used to position the wafer and reticle (or mask) stages. Since a lithography tool may produce $50-100M/year of product, the economic value from improved performance distance measuring interferometers is substantial. Each 1% increase in yield of the lithography tool results in approximately $lM/year economic benefit to the integrated circuit manufacturer and substantial competitive advantage to the lithography tool vendor. The function of a lithography tool is to direct spatially patterned radiation onto a photoresist-coated wafer. The process involves determining which location of the wafer is to receive the radiation (alignment) and applying the radiation to the photoresist at that location (exposure).

To properly position the wafer, the wafer includes alignment marks on the wafer that can be measured by dedicated sensors. The measured positions of the alignment marks define the location of the wafer within the tool. This information, along with a specification of the desired patterning of the wafer surface, guides the alignment of the wafer relative to the spatially patterned radiation. Based on such information, a translatable stage supporting the photoresist-coated wafer moves the wafer such that the radiation will expose the correct location of the wafer.

During exposure, a radiation source illuminates a patterned reticle, which scatters the radiation to produce the spatially patterned radiation. The reticle is also referred to as a mask, and these terms are used interchangeably below. In the case of reduction lithography, a reduction lens collects the scattered radiation and forms a reduced image of the reticle pattern. Alternatively, in the case of proximity printing, the scattered radiation propagates a small distance (typically on the order of microns) before contacting the wafer to produce a 1 : 1 image of the reticle pattern. The radiation initiates photo-chemical processes in the resist that convert the radiation pattern into a latent image within the resist. Interferometry systems are important components of the positioning mechanisms that control the position of the wafer and reticle, and register the reticle image on the wafer. If such interferometry systems include the features described above, the accuracy of distances measured by the systems increases as cyclic error contributions to the distance measurement are minimized. In general, the lithography system, also referred to as an exposure system, typically includes an illumination system and a wafer positioning system. The illumination system includes a radiation source for providing radiation such as ultraviolet, visible, x-ray, electron, or ion radiation, and a reticle or mask for imparting the pattern to the radiation, thereby generating the spatially patterned radiation. In addition, for the case of reduction lithography, the illumination system can include a lens assembly for imaging the spatially patterned radiation onto the wafer. The imaged radiation exposes resist coated onto the wafer. The illumination system also includes a mask stage for supporting the mask and a positioning system for adjusting the position of the mask stage relative to the radiation directed through the mask. The wafer positioning system includes a wafer stage for supporting the wafer and a positioning system for adjusting the position of the wafer stage relative to the imaged radiation. Fabrication of integrated circuits can include multiple exposing steps. For a general reference on lithography, see, for example, J. R. Sheats and B. W. Smith, in Microlithosraphy: Science and Technology (Marcel Dekker, Inc., New York, 1998), the contents of which is incorporated herein by reference. Interferometry systems described above can be used to precisely measure the positions of each of the wafer stage and mask stage relative to other components of the exposure system, such as the lens assembly, radiation source, or support structure. In such cases, the interferometry system can be attached to a stationary structure and the measurement object attached to a movable element such as one of the mask and wafer stages. Alternatively, the situation can be reversed, with the interferometry system attached to a movable object and the measurement object attached to a stationary object. More generally, such interferometry systems can be used to measure the position of any one component of the exposure system relative to any other component of the exposure system, in which the interferometry system is attached to, or supported by, one of the components and the measurement object is attached, or is supported by the other of the components.

An example of a lithography scanner 1100 using an interferometry system 1126 is shown in Fig. 4. The interferometry system is used to precisely measure the position of a wafer (not shown) within an exposure system. Here, stage 1122 is used to position and support the wafer relative to an exposure station. Scanner 1100 includes a frame 1102, which carries other support structures and various components carried on those structures. An exposure base 1104 has mounted on top of it a lens housing 1106 atop of which is mounted a reticle or mask stage 1116, which is used to support a reticle or mask. A positioning system for positioning the mask relative to the exposure station is indicated schematically by element 1117. Positioning system 1117 can include, e.g., piezoelectric transducer elements and corresponding control electronics. Although, it is not included in this described embodiment, one or more of the interferometry systems described above can also be used to precisely measure the position of the mask stage as well as other moveable elements whose position must be accurately monitored in processes for fabricating lithographic structures (see supra Sheats and Smith Microlithography: Science and Technology).

Suspended below exposure base 1104 is a support base 1113 that carries wafer stage 1122. Stage 1122 includes a plane mirror 1128 for reflecting a measurement beam 1154 directed to the stage by interferometry system 1126. A positioning system for positioning stage 1122 relative to interferometry system 1126 is indicated schematically by element 1119. Positioning system 1119 can include, e.g., piezoelectric transducer elements and corresponding control electronics. The measurement beam reflects back to the interferometry system, which is mounted on exposure base 1104. The interferometry system can be any of the embodiments described previously. During operation, a radiation beam 1110, e.g., an ultraviolet (UV) beam from a UV laser (not shown), passes through a beam shaping optics assembly 1112 and travels downward after reflecting from mirror 1114. Thereafter, the radiation beam passes through a mask (not shown) carried by mask stage 1116. The mask (not shown) is imaged onto a wafer (not shown) on wafer stage 1122 via a lens assembly 1108 carried in a lens housing 1106. Base 1104 and the various components supported by it are isolated from environmental vibrations by a damping system depicted by spring 1120. In other embodiments of the lithographic scanner, one or more of the interferometry systems described previously can be used to measure distance along multiple axes and angles associated for example with, but not limited to, the wafer and reticle (or mask) stages. Also, rather than a UV laser beam, other beams can be used to expose the wafer including, e.g., x-ray beams, electron beams, ion beams, and visible optical beams.

In some embodiments, the lithographic scanner can include what is known in the art as a column reference. In such embodiments, the interferometry system 1126 directs the reference beam (not shown) along an external reference path that contacts a reference mirror (not shown) mounted on some structure that directs the radiation beam, e.g., lens housing 1106. The reference mirror reflects the reference beam back to the interferometry system. The interference signal produce by interferometry system 1126 when combining measurement beam 1154 reflected from stage 1122 and the reference beam reflected from a reference mirror mounted on the lens housing 1106 indicates changes in the position of the stage relative to the radiation beam. Furthermore, in other embodiments the interferometry system 1126 can be positioned to measure changes in the position of reticle (or mask) stage 1116 or other movable components of the scanner system. Finally, the interferometry systems can be used in a similar fashion with lithography systems involving steppers, in addition to, or rather than, scanners.

As is well known in the art, lithography is a critical part of manufacturing methods for making semiconducting devices. For example, U.S. Patent 5,483,343 outlines steps for such manufacturing methods. These steps are described below with reference to FIGS. 5(a) and 5(b). FIG. 5(a) is a flow chart of the sequence of manufacturing a semiconductor device such as a semiconductor chip (e.g., IC or LSI), a liquid crystal panel or a CCD. Step 1151 is a design process for designing the circuit of a semiconductor device. Step 1152 is a process for manufacturing a mask on the basis of the circuit pattern design. Step 1153 is a process for manufacturing a wafer by using a material such as silicon.

Step 1154 is a wafer process which is called a pre-process wherein, by using the so prepared mask and wafer, circuits are formed on the wafer through lithography. To form circuits on the wafer that correspond with sufficient spatial resolution those patterns on the mask, interferometric positioning of the lithography tool relative the wafer is necessary. The interferometry methods and systems described herein can be especially useful to improve the effectiveness of the lithography used in the wafer process.

Step 1155 is an assembling step, which is called a post-process wherein the wafer processed by step 1154 is formed into semiconductor chips. This step includes assembling (dicing and bonding) and packaging (chip sealing). Step 1156 is an inspection step wherein operability check, durability check and so on of the semiconductor devices produced by step 1155 are carried out. With these processes, semiconductor devices are finished and they are shipped (step 1157).

FIG. 5(b) is a flow chart showing details of the wafer process. Step 1161 is an oxidation process for oxidizing the surface of a wafer. Step 1162 is a CVD process for forming an insulating film on the wafer surface. Step 1163 is an electrode forming process for forming electrodes on the wafer by vapor deposition. Step 1164 is an ion implanting process for implanting ions to the wafer. Step 1165 is a resist process for applying a resist (photosensitive material) to the wafer. Step 1166 is an exposure process for printing, by exposure (i.e., lithography), the circuit pattern of the mask on the wafer through the exposure apparatus described above. Once again, as described above, the use of the interferometry systems and methods described herein improve the accuracy and resolution of such lithography steps.

Step 1167 is a developing process for developing the exposed wafer. Step 1168 is an etching process for removing portions other than the developed resist image. Step 1169 is a resist separation process for separating the resist material remaining on the wafer after being subjected to the etching process. By repeating these processes, circuit patterns are formed and superimposed on the wafer.

The interferometry systems described above can also be used in other applications in which the relative position of an object needs to be measured precisely. For example, in applications in which a write beam such as a laser, x-ray, ion, or electron beam, marks a pattern onto a substrate as either the substrate or beam moves, the interferometry systems can be used to measure the relative movement between the substrate and write beam.

As an example, a schematic of a beam writing system 1200 is shown in FIG. 6. A source 1210 generates a write beam 1212, and a beam focusing assembly 1214 directs the radiation beam to a substrate 1216 supported by a movable stage 1218. To determine the relative position of the stage, an interferometry system 1220 directs a reference beam 1222 to a mirror 1224 mounted on beam focusing assembly 1214 and a measurement beam 1226 to a mirror 1228 mounted on stage 1218. Since the reference beam contacts a mirror mounted on the beam focusing assembly, the beam writing system is an example of a system that uses a column reference. Interferometry system 1220 can be any of the interferometry systems described previously. Changes in the position measured by the interferometry system correspond to changes in the relative position of write beam 1212 on substrate 1216. Interferometry system 1220 sends a measurement signal 1232 to controller 1230 that is indicative of the relative position of write beam 1212 on substrate 1216. Controller 1230 sends an output signal 1234 to a base 1236 that supports and positions stage 1218. In addition, controller 1230 sends a signal 1238 to source 1210 to vary the intensity of, or block, write beam 1212 so that the write beam contacts the substrate with an intensity sufficient to cause photophysical or photochemical change only at selected positions of the substrate.

Furthermore, in some embodiments, controller 1230 can cause beam focusing assembly 1214 to scan the write beam over a region of the substrate, e.g., using signal 1244. As a result, controller 1230 directs the other components of the system to pattern the substrate. The patterning is typically based on an electronic design pattern stored in the controller. In some applications the write beam patterns a resist coated on the substrate and in other applications the write beam directly patterns, e.g. , etches, the substrate.

An important application of such a system is the fabrication of masks and reticles used in the lithography methods described previously. For example, to fabricate a lithography mask an electron beam can be used to pattern a chromium-coated glass substrate. In such cases where the write beam is an electron beam, the beam writing system encloses the electron beam path in a vacuum. Also, in cases where the write beam is, e.g., an electron or ion beam, the beam focusing assembly includes electric field generators such as quadrapole lenses for focusing and directing the charged particles onto the substrate under vacuum, hi other cases where the write beam is a radiation beam, e.g., x-ray, UV, or visible radiation, the beam focusing assembly includes corresponding optics and for focusing and directing the radiation to the substrate.

A number of embodiments of the invention have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the invention. Accordingly, other embodiments are within the scope of the following claims.

Claims

WHAT IS CLAIMED IS:

1. A method comprising: providing an interference signal S(t) from two beams directed along different paths, wherein the signal S(t) is indicative of changes in an optical path difference nL(t) between the different paths, where n is an average refractive index along the different paths, L(t) is a total physical path difference between the different paths, and t is time; providing one or more coefficients representative of one or more errors that cause the signal S (t) to deviate from an ideal expression of the form A₁ cos(ω_Rt + φ(t) + C,_\ ) , where A₁ and ζ₁ are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(t) = nkL{t), with k = 2π/λ andλ equal to a wavelength for the beams; calculating a linear combination of values of the signal S(t) ; and reducing the effect of the deviation of S[t) from the ideal expression on an estimate of L(t) using an error signal generated from the coefficients and one or more error basis functions derived at least in part from the linear combination of values of the signal S(t) .

2. The method of claim 1, wherein the linear combination comprises a quadrature signal S(t) based on values of the signal S(t) .

3. The method of claim 2, wherein the quadrature signal S(t) is based on two values of the signal S(t) .

4. The method of claim 2, wherein the quadrature signal S(t) is updated at a rate at which the signal S(t) is sampled.

5. The method of claim 1, wherein the linear combination comprises a transform signal D_q (t, ) based on values of the signal S(t) , for a frequency value determined by q, updated at a time t_\.

6. The method of claim 5, wherein the transform signal D_q [t_λ ) comprises a sample of a discrete Fourier transform of the signal S(t) .

7. The method of claim 6, wherein

where τ = \/f_s, and W_n is a

window function centered at

8. The method of claim 7, further comprising selecting q as an integer approximately equal to Nfulfs and storing the transform signal D_q {t₁ ) under conditions for which ^_M can be estimated as f_M ≈ (ω_R + dφ(t)/cfr)/2π .

9. The method of claim 6, wherein the transform signal D_q (/, ) is updated at a rate lower than a raters at which the signal S(t) is sampled.

10. The method of claim 9, wherein the transform signal D_q [t_χ ) is updated at a rate/_R/w based on a time-shifted signal S(t + lm/ f_R ) to yield an updated transform signal, where / and m are positive integers, and q is an integer approximately equal to Nf_M/f_S.

11. The method of claim 10, further comprising storing the transform signal D (ti ) under conditions for which f_M can be estimated as f_M ≈ ( ω_R + dφ(t)/ dt)/2π .

12. The method of claim 6, wherein the discrete Fourier transform includes a window function.

13. The method of claim 6, wherein the discrete Fourier transform is folded.

14. The method of claim 2, wherein the error signal comprises a time-domain signal S_ψ (t) .

15. The method of claim 14, wherein reducing the effect of the deviation of S(t) from the ideal expression on an estimate of L{t) comprises subtracting the error signal from S(t) .

16. The method of claim 1, wherein the error signal comprises a signal D _ψ (t₁ ) updated at a time t₁.

17. The method of claim 16, wherein reducing the effect of the deviation of S(t) from the ideal expression on an estimate of L(t) comprises subtracting the error signal from a discrete Fourier transform of values of S(t) .

18. The method of claim 1 , further comprising directing the two beams along the different paths and measuring the interference signal S (t) .

19. The method of claim 18, wherein at least one of the beams is directed to reflect from a movable measurement object before producing the interference signal

S(t).

20. The method of claim 19, wherein the beam directed to contact the movable measurement object reflects from the measurement object multiple times before producing the interference signal S(t) .

21. The method of claim 19, wherein the beams are directed to reflect from different locations of the movable measurement object before producing the interference signal S(t).

22. The method of claim 1, wherein the errors correspond to spurious beam paths.

23. The method of claim 1, wherein the deviation can be expressed as

, where p=1, 2, 3..., and w is any integer not equal to

p, and where the provided coefficients comprise information corresponding to at least some of A_{m ,p} and ζ_m,p .

24. The method of claim 2, wherein ω_R is non-zero.

25. The method of claim 24, further comprising: providing a reference signal S_R (t) = A_R cos(ω_Rt + ζ_R ) , where A_R and ζ_R are constants; and calculating a quadrature reference signal S_R (t) based on the signal S_R (t) , wherein the error basis functions are derived from the signals S(t), S(t), S_R (t), and S_R (ή .

26. The method of claim 25, further comprising measuring the reference signal S_R (t) based on an output from a common source of the two beams.

27. The method of claim 25, further comprising generating reference signals S_R (t) and S_R (t) from a lookup table indexed synchronously with sampling of S(t).

28. The method of claim 2, wherein calculating the quadrature signal S(t) comprises calculating the quadrature signal S(t) based on the expression

, where τ >0 and ω_M is an

instantaneous rate of change of a phase of the interference signal S(t).

29. The method of claim 28, wherein calculating the quadrature signal S(t) further comprises approximating ω_M according to <o_M ≈ ω_R + dφ(t)/dt , where φ(f) in the expression for ω_M is determined from the interference signal S(t) assuming the deviation of S(t) from the ideal expression is negligible.

30. The method of claim 28, wherein calculating the quadrature signal S(t) further comprises approximating ω_M according to ω_M ≈ ω_R .

31. The method of claim 30, wherein calculating the quadrature signal S(t)

comprises calculating the quadrature signal S(t) according to the simplified expression where N is a non-negative

integer.

32. The method of claim 25, wherein calculating the quadrature reference signal S_R (t) comprises calculating the quadrature reference signal S_R (t) based on the

expressio S_R (t) = (cot ω_Rτ) S_R (t - 2τ) _: — S_R (t - τ) , where τ >0.

33. The method of claim 32, wherein calculating the quadrature reference signal S_R (t) comprises calculating the quadrature reference signal S_R (t) according to the

simplified expression

where N is a non-negative integer.

34. The method of claim 1, wherein the interference signal S(t) is provided at a data rate that is an integer multiple of ω_R /2π .

35. The method of claim 1, wherein each of the error basis functions corresponds to a function that includes one or more leading sinusoidal terms having a time- varying argument that corresponds to a time-varying argument of an error term that represents a portion of the deviation of S(t) from the ideal expression.

36. The method of claim 35, wherein the error basis functions correspond to one or more pairs of sine and cosine functions having an argument whose time-varying component has the form ω_Rt + (m/p)φ(t) , where p is a positive integer and m is an integer not equal to p.

37. The method of claim 36, wherein a leading term for a function corresponding to one of the error basis functions has the same values of m andp as a time-varying argument of one of the error terms.

5 38. The method of claim 37, wherein each of the error basis functions is proportional to the function that includes one or more leading sinusoidal terms.

39. The method of claim 35, wherein the error basis functions correspond to one or more pairs of sine and cosine functions having an argument whose time-varying o component has the form (/w/ p) φ (t) , where p is a positive integer and m is an integer not equal to p.

40. The method of claim 39, wherein a leading term for a function corresponding to one of the error basis functions has the same values of m and p as a time- varying 5 argument of one of the error terms.

41. The method of claim 40, wherein each of the error basis functions is proportional to a linear combination of samples of the function that includes one or more leading sinusoidal terms. 0

42. The method of claim 39, wherein each of the pairs of sine and cosine functions corresponds to imaginary and real parts, respectively, of a complex signal.

43. The method of claim 36, wherein the error basis functions correspond to 5 multiple pairs of the sine and cosine functions.

44. The method of claim 43, wherein the error basis functions comprise multiple pairs of the sine and cosine functions from a family of the sine and cosine functions with {(p = 1,m = -1),(p = 1,m = 0),(p = 1,m = 2),(p = 1,m = 3),and (p = 2,m = 1)} . 0

45. The method of claim 2, further comprising generating the error basis functions from the signals S(t) and S(t) .

46. The method of claim 25, further comprising generating the error basis functions from the signals S(t), S(t), S_R (t), and S_R (t) .

47. The method of claim 46, wherein the error basis functions are generated from algebraic combinations of the signals S(t), S(t), S_R (t) , and S_R (t).

48. The method of claim 14, further comprising generating the error signal

MO-

49. The method of claim 48, the error signal 5_ψ (t) is generated from a superposition of the error basis functions weighted by the coefficients representative of the errors.

50. The method of claim 1, further comprising determining a value for the optical path difference nL(t) from the interference signal S{t) after the effect of its deviations are reduced.

51. The method of claim 2, wherein the quadrature signal S(t) is calculated from the interference signal S(t) based on prior values of S(t) according to the approximation S(t) ≈ S (t - 2πN/ω_R ) , where N is a positive integer.

52. The method of claim 14, wherein the error basis functions used to generate the error signal S_ψ (t) are derived from prior values of the signals S(t) and S (t) according to the approximations S(t) ≈ S(t - 2πN/ω_R ) and S (t) ≈ 5 (t - 2πM/ω_R ) , where N and M are positive integers.

53. The method of claim 1, wherein ω_R > 100 • dφ(t)/dt .

54. The method of claim 1, wherein ω_R > 500 • dφ(t)/dt .

55. An apparatus comprising a computer readable medium which during operation causes a processor to perform the method of claim 1.

56. An apparatus comprising: an interferometry system which during operation directs two beams along different paths and provides an interference signal S(t) from the two beams, wherein the signal S(t) is indicative of changes in an optical path difference nL(t) between the different paths, where n is an average refractive index along the different paths, L{t) is a total physical path difference between the different paths, and t is time, wherein imperfections in the interferometry system produce one or more errors that cause the signal S(t) to deviate from an ideal expression of the form

A₁ cos(ω_Rt + φ(t) + ζ₁ ) , where A₁ and ζ₁ are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(t) = nkL(t), with k = 2π/λ andλ equal to a wavelength for the beams; and an electronic processor which during operation receives the interference signal S{t) from the interferometry system, receives one or more coefficients representative of the one or more errors, calculates a linear combination of values of the signal S{t), and reduces the effect of the deviation of S(t) from the ideal expression on an estimate of L[t) using an error signal S_ψ (t) generated from the coefficients and one or more error basis functions derived at least in part from the linear combination of values of the signal S(ή.

57. A lithography system for use in fabricating integrated circuits on a wafer, the system comprising: a stage for supporting the wafer; an illumination system for imaging spatially patterned radiation onto the wafer; a positioning system for adjusting the position of the stage relative to the imaged radiation; and the apparatus of claim 56 for monitoring the position of the wafer relative to the imaged radiation.

58. A lithography system for use in fabricating integrated circuits on a wafer, the system comprising: a stage for supporting the wafer; and an illumination system including a radiation source, a mask, a positioning system, a lens assembly, and the apparatus of claim 56, wherein during operation the source directs radiation through the mask to produce spatially patterned radiation, the positioning system adjusts the position of the mask relative to the radiation from the source, the lens assembly images the spatially patterned radiation onto the wafer, and the apparatus monitors the position of the mask relative to the radiation from the source.

59. A beam writing system for use in fabricating a lithography mask, the system comprising: a source providing a write beam to pattern a substrate; a stage supporting the substrate; a beam directing assembly for delivering the write beam to the substrate; a positioning system for positioning the stage and beam directing assembly relative one another; and the apparatus of claim 56 for monitoring the position of the stage relative to the beam directing assembly.

60. An apparatus comprising: an interferometry system which during operation directs two beams along different paths and provides an interference signal S (t) from the two beams, wherein the signal S(t) is indicative of changes in an optical path difference nL(t) between the different paths, where n is an average refractive index along the different paths, L(t) is a total physical path difference between the different paths, and t is time, wherein imperfections in the interferometry system produce one or more errors that cause the signal S(t) to deviate from an ideal expression of the form

A₁ cos (ω_Rt + φ(t) + ζ₁ ), where A₁ and ζ₁ are constants, ω_R is an angular frequency difference between the two beams before being directed along the different paths, and φ(t) = nkL{t), with k = 2π/λ andλ equal to a wavelength for the beams; and an electronic processor which during operation receives the interference signal

S(t) from the interferometry system, calculates a linear combination of values of the signal <-» (*), and calculates an estimate for one or more coefficients representative of the one or more errors based on filtering a signal derived at least in part from the linear combination of values of the signal S(t) .

61. A lithography method for use in fabricating integrated circuits on a wafer, the method comprising: supporting the wafer on a moveable stage; imaging spatially patterned radiation onto the wafer; adjusting the position of the stage; and monitoring the position of the stage using the method of claim 1.

62. A lithography method for use in the fabrication of integrated circuits comprising: directing input radiation through a mask to produce spatially patterned radiation; positioning the mask relative to the input radiation; monitoring the position of the mask relative to the input radiation using the method of claim 1; and imaging the spatially patterned radiation onto a wafer.

63. A lithography method for fabricating integrated circuits on a wafer comprising: positioning a first component of a lithography system relative to a second component of a lithography system to expose the wafer to spatially patterned radiation; and monitoring the position of the first component relative to the second component using the method of claim 1.

64. A method for fabricating integrated circuits, the method comprising the lithography method of claim 61.

65. A method for fabricating integrated circuits, the method comprising the lithography method of claim 62.

66. A method for fabricating integrated circuits, the method comprising the lithography method of claim 63.

67. A method for fabricating integrated circuits, the method comprising using the lithography system of claim 57.

68. A method for fabricating integrated circuits, the method comprising using the lithography system of claim 58.

69. A method for fabricating a lithography mask, the method comprising: directing a write beam to a substrate to pattern the substrate; positioning the substrate relative to the write beam; and monitoring the position of the substrate relative to the write beam using the interferometry method of claim 1.