FIELD OF INVENTION

This invention relates to an optical surface profiler. More particularly, this invention relates to an optical non-contact surface profiler capable of (i) surface profiles of transparent layers on light-absorbing surfaces, (ii) surface profilesof light-absorbing surfaces through transparent layers, (iii) thickness profiles of transparent layers on light-absorbing surfaces, and (iv) surface profiles of light-absorbing surfaces.

BACKGROUND

With the increasing competition in the semiconductor industry and the expense of producing integrated circuit devices, there is a need for quality assurance equipment to identify defective substrates as early in the fabrication process aspossible. Surface profile equipment capable of resolving at a microscopic precision have been adapted for performing such quality assurance.

There are optical surface profilers available capable of measuring the surface profile of an opaque substrate. There also are spectrophotometric and ellipsometric devices available capable of measuring the thickness of a transparent layer on anopaque substrate at a single position on the sample. However, because the integrated circuit fabrication process includes adding layers of transparent film to light-absorbing or opaque substrates, there also is a need for an apparatus capable ofmeasuring the surface profile of a transparent film on a light-absorbing or opaque substrate. Similarly, there is a need for an apparatus capable of measuring the subsurface profile of light-absorbing or opaque material through the transparent layer. Similarly there is a need to measure the film thickness profile of a transparent film on a light-absorbing or opaque substrate.

SUMMARY OF THE INVENTION

It is an object of the invention to provide an optical profiler for (i) determining the surface profile of a transparent layer on a light-absorbing or opaque substrate and (ii) determining the sub-surface profile of a light-absorbing or opaquesurface through a transparent layer.

It is another object of the invention to provide an optical profiler for determining the thickness profile of a transparent layer on a light-absorbing or opaque substrate.

It is another object of this invention to provide a method for determining the index of refraction of a transparent layer on a light-absorbing or opaque substrate.

Briefly, a preferred embodiment of the present invention includes a specially configured Linnik microscope with several interchangeable narrowband spectral filters in the illumination path in combination with a photosensing array and a computer.

By alternatively configuring the microscope of the profiler in interferometric mode (Linnik reference channel unshuttered) and spectrophotometric mode (Linnik reference channel shuttered), the profiler gathers phase data from an interferencepattern and reflectance data from a reflectance pattern, respectively. Such data is used to determine the surface profile of a transparent layer on a light-absorbing or opaque substrate and the sub-surface profile of a light-absorbing or opaquesubstrate through a transparent layer.

The phase data and reflectance data contains information about the surface of the transparent layer, the layer thickness, and the surface of the substrate. To extract the surface profile of the transparent layer from the interference pattern, acorrection is required for each point in the profile which removes the contribution to the phase due to the transparent layer thickness.

The phase correction value is calculated for each point of the profile from a film thickness profile and the optical properties of the film and the substrate. The film thickness profile is calculated from the reflectance data. The surfaceprofile of the transparent layer is then calculated by subtracting the phase correction value from the phase data and then converting the corrected phase information into a surface profile.

The sub-surface profile of the light-absorbing material beneath the transparent layer is found by subtracting the film thickness profile from the surface profile.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic of the non-contact surface profiler including a Linnik interferometric objective.

FIG. 2 is a flow chart of the phase measurement program of the computing device for the non-contact surface profiler.

FIG. 3 is a flow chart of the film thickness measurement program of the computing device for the non-contact surface profiler.

FIGS. 4a-4d illustrate alternative surface profile results for a light-absorbing material beneath a transparent film having an unknown index of refraction.

FIG. 5 is a sample surface profile of an opaque substrate.

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT

Referring to FIG. 1, the non-contact surface profiler includes a microscope equipped with a dual beam interferometric objective attachment 12. Although a Linnik interferometric attachment 12 is preferred, other interferometric attachments mightbe used.

The profiler using the Linnik attachment 12 includes a light source 14 and a filter wheel 16. The filter wheel 16 includes spectral filters 17 each of which is transparent to a particular wavelength of light and is opaque to all otherwavelengths. Light emitted from the light source 14 passes through a collimating lens 18, then through the selected filter 17 on the filter wheel 16. By rotating the wheel 16, the different spectral filters are interchanged so that light of a selectedwavelength passes through the filter.

The light passing through the filter 17 is incident upon a beamsplitter 20 which divides the incident light. A portion of the light is transmitted through the beamsplitter 20 and focused by a reference objective lens 22 onto a reference mirror24. The light then reflects off the reference mirror 24, passes back through the reference objective lens 22 and returns to the beamsplitter 20.

The other half of the light beam is reflected by the beamsplitter 20 (instead of being transmitted) and is focussed by a sample objective lens 26 onto the sample S to be profiled.

The sample may, for example, be an intergrated circuit layer F on a semiconductor wafer W at an intermediate stage of the fabrication process. The light reflects off the sample S, passes back through the sample objective lens 26 and returns tothe beamsplitter 20.

The beams returning from the sample and the reference mirror recombine and interfere at the beamsplitter 20. Images are formed of the interference pattern of the reference mirror surface and the sample surface by a lens 30 at a photo sensingdevice 28. Alternatively the reference channel 22, 24 may be shuttered so that the lens 30 forms only the image of the sample surface at the photosensing device 28.

The photo sensing device 28 is a solid state photodetector array linked to a computer 32 via a multi-bit A/D converter (not shown) which preferably has 12 bit or greater precision. The computer 32 processes the data from the photosensing deviceand additionally may format and display raw or processed data in various formats. (i.e. FIG. 5).

The FIG. 1 embodiment with the Linnik interferometer is the preferred embodiment. The Linnik geometry easily allows the use of high magnification microscope objectives. Additionally, the reference channel of the Linnik 12 may be shutteredeasily to allow the profiler to be used in two modes. The profiler can be used to obtain interferometric data (reference portion unshuttered) or to obtain standard microscope image reflectance data (reference portion shuttered). Other possibleembodiments include the Mireau or Michelson interferometric objectives and/or a rotatable grating in the illumination path instead of spectral filters.

Measuring Surface Profiles

For applications where the sample consists only of an opaque material, multi-wavelength phase measuring interferometry (MWPMI) techniques may be used to determine a surface profile. If the sample consists of a thin transparent layer over anopaque substrate then MWPMI is used in combination with imaging spectrophotometry to determine the surface profile. By using the profiler of FIG. 1 with the reference portion unshuttered, the light incident on the photosensing device 28 forms aninterference pattern from the light reflected from the sample S and the light reflected from the reference mirror 24. By rotating the filter wheel 16, interference patterns are obtained for different illumination wavelengths.

The interference pattern that is formed represents the phase difference between the two interfering wavefronts. By using phase measuring interferometry (PMI) techniques, this phase difference may be determined very accurately.

To perform PMI requires that the position of the sample be changed relative to the reference mirror. This can be achieved by moving either the reference mirror 24 or the sample S. A piezo-electric transducer 40 is included in the profiler 10 tomove the reference mirror 24. As the reference mirror is moved, the phase difference, and thus the intensity distribution across the interference pattern changes. The photosensing device 28 is used to detect the intensity across the image of thesurface at the different reference mirror positions. In operation the mirror is moved to vary the phase of the beam reflected from the reference mirror 24 in one-quarter wavelength steps.

Various PMI algorithms may be implemented in the computing device 32 to compute the surface profile. Standard PMI is done at one wavelength. Using visible light the maximum measurable step height at a single wavelength has been approximately1500 angstroms. Step height refers to the height difference between adjacent sampling points (i.e. points on the sample surface corresponding to the adjacent photodetectors in a photodetector array). By interchanging spectral filters to takemeasurements at several wavelengths, the maximum measurable step height may increase to approximately 200,000 angstroms.

Referring to FIG. 2, a flow chart of the phase measurement software is listed illustrating the operation of the profiler for multiwavelength phase measuring interferometry. At step 50, the light intensity across the photosensing device 28 ismeasured. At step 52, the reference mirror 24 is moved by the piezoelectric transducer to shift the phase of the reflected light by 90 degrees. At step 54, it is determined whether the phase has been shifted four times such that measurements have beenmade at phase shifts of 0, 90, 180, and 270. Once measurements are made at these four phases, step 56 is reached. At this point it is determined whether sets of the phase measurement have been made at 3 wavelengths. If not, step 58 is performedcausing the filter wheel to be advanced. Once measurements are made for three wavelengths, data has been stored for the intensity distribution I(X,Y) incident at the photosensing device 28 at each wavelength and reference mirror position. This data isused in step 60 to calculate fringe visibility using the formula: ##EQU1## where .gamma.=fringe visibility

I.sub.n =measured intensity at reference mirror position n.

In step 62 the fringe visibility is tested to see if it is greater than a predetermined noise threshold. If not, then a mask value is assigned in step 64. Otherwise the phase value is calculated.

Phase value is calculated at step 66 using the four sets of intensity measurements.

The phase at each detector point (x,y) and wavelength, .lambda., is:

If the sample is an opaque material with no transparent layers, phase value relates to the surface profile by the following equation: ##EQU2## where: d(x,y) is the surface profile; and

m(x,y) is an integer describing the 2.pi. ambiguity inherent in phase measuring interferometry.

The 2.pi. ambiguity is resolved by making phase measurement at more than one wavelength. This is the technique of multi-wavelength PMI. If two wavelengths, .lambda..sub.1, and .lambda..sub.2 are used, m.sub.1 (X,Y,.lambda.) and m.sub.2(x,y,.lambda.) are assumed to be equal. The two measured phase values are then subtracted to obtain

Solving for the surface profile, we find

However, we have magnified the error in the profile by the factor .lambda./(.lambda..sub.2 -.lambda..sub.1). The precision of the single wavelength PMI is restored by substituting the profile obtained from Eq. 5 into Eq. 3 and using theknowledge that m(x,y,.lambda.) is an integer. Once m(x,y,.lambda.) is known, Eq. 3 is solved for d(x,Y). This technique may be expanded to any number of wavelengths obtaining greater range and noise immunity advantages.

FIG. 5 represents a surface profile formatted as a 3 dimensional plot for a portion of a sample opaque substrate.

If the sample contains a transparent layer over a light-absorbing or opaque material, then a phase correction factor which accounts for the presence of the transparent film is necessary to compute the surface profile.

The phase change upon reflection from a transparent film on an light absorbing substrate (e.g. SiO.sub.2 on Silicon) can be calculated if the film thickness and optical constants of the film and substrate are known. The analysis can be found inM. Born & E. Wolf (1980) Principles of Optics. The equation for determining the phase change upon reflection .delta..sub.r is: ##EQU3## where r.sub.23 =.rho..sub.23 e.sup.i.phi.23 is the reflectivity of the film-substrate boundary, r.sub.12 is thereflectivity of the air-film boundary, and .beta. given by

where n is the refractive index of the film and t is the film thickness.

In order to use this analysis to determine the phase change upon reflection the film thickness must be known. The film thickness is measured by shuttering the reference channel and using imaging spectrophotometry. Once the film thickness isdetermined the phase change upon reflection is calculated using equation 6 for every point along the profile.

The corrected phase .PHI. (X,Y) is given by

where .phi.(x,y) is the phase determined by MWPMI and .delta..sub.r is the phase change upon reflection. The surface profile d(x,y) is then determined by using .PHI.(x,y) in place of .phi.(x,y) in equation 5.

MEASURING FILM THICKNESS PROFILES

For applications where only the film thickness profile is desired Imaging Spectrophotometric Profiling (ISP) is used. ISP is done with the reference portion of the profiler 10 shuttered. Referring to FIG. 1, light emitted from the source 14passes through a collimating lens 18 then through a filter 17 on filter wheel 16. The light passing through the filter 17 is incident upon the beamsplitter 20. Because the reference portion is shuttered, the light transmitted through the beamsplitteris not used.

Part of the light beam, however, is reflected by the beamsplitter 20 and focused by a sample objective lens 26 onto the sample S. For film thickness measurement, the sample is a transparent film F on a light-absorbing or opaque substrate. Aportion of the light is reflected from the film surface and reflected back through the optical system. The other portion is refracted and transmitted through the film then reflected from the light-absorbing surface under the film back through theoptical system. The reflected beams interfere and return through the sample objective lens 26, to the beamsplitter 20 and then pass through the imaging lens 30 and focus on the photosensing device 28. The photosensing device 28 measures the intensityprofile of the reflected image incident upon its surface. From the intensity data a reflectance profile of the sample is calculated. To determine the thickness of the Film F, intensity data is measured at multiple wavelengths by the interchange ofdifferent filters.

Referring to FIG. 3, a flow chart of the film thickness measurement software is listed. At step 70, the light intensity across the photosensing device 28 is measured. At step 72, it is determined whether measurements have been made for apredetermined number of wavelengths. If not, step 74 is executed and the filter wheel 16 is rotated for passing light of a different wavelength. Once measurements have been taken for all wavelengths, step 76 is performed in which the reflectance ateach wavelength is calculated. The following formula is used to calculate reflectance from the measured intensity taking into account detector gains and offsets:

where: A and B are constants for each wavelength and detector array location.

A is an intensity measurement with no sample in place at wavelength .lambda.. B is determined experimentally by measuring the intensity of a sample having known reflectance at wavelength .lambda. and then solving for B.

The reflectance of a thin transparent film is given by ##EQU4## (see Born and Wolf). R(.lambda., t) is at a maximum or minimum value when

where

and m is an integer. m decreases by 1 between each extrema as the wavelength is increased.

The film thickness is obtained from a set of reflectance values by applying a curve fitting algorithm to find the wavelengths where the reflectance extrema occur. These wavelengths are then used to solve Eq. 12 for the film thickness. For twoadjacent extrema, the film thickness is given by: ##EQU5## where the a and b subscripts refer to the first and second extrema wavelengths respectively with .lambda..sub.b greater than .lambda..sub.a.

By applying the above method to each element in the photosensor array, a film thickness profile is obtained. If less than two extrema occur within the wavelength range examined, Eq. 12 cannot be solved for both m and t, and another algorithmmust be used. A least squares fit of the measured reflectance to theoretical reflectance is an alternative algorithm.

Determining the Index of Refraction Of A Transparent Film

To determine the film thickness as discussed above, knowledge of the index of refraction of the film must be known. For some materials the index of refraction is well behaved varying little from standard values (e.g. silicon-dioxide). Thus astandard value may be used for the index of refraction. Other materials (e.g. silicon-nitride) have indices which vary widely for the same material under different conditions. For such materials with indices that vary widely, the following method maybe used for determining the index of refraction.

By using a test structure of (i) a transparent film having an unknown index of refraction on (ii) a light-absorbing or opaque substrate having a known surface geometry, the transparent film index may be determined with the non-contact surfaceprofiler. The test structure requires only that there be a change in transparent film thickness over a region of known substrate geometry (see FIG. 4a). The test structure is then measured using the non-contact surface profiler assuming a typical indexof refraction. If the substrate profile through the film shows a trench or a step (as in FIG. 4b and FIG. 4c) where the geometry should be flat the assumed index is incorrect. The correct index of refraction is determined by iteratively calculating thesurface profile of the substrate through the transparent film with varying film indices of refraction until the substrate geometry measured through the transparent film matches the expected geometry (as in FIG. 4d).

While a preferred embodiment of the machine has been illustrated and described, the invention is not limited to the embodiment illustrated here. For example, there are numerous interferometric configurations (Linnik, Mireau, Michelson), manytechniques for PMI (phase-stepping or integrating bucket), and different ways of achieving reference shuttering depending on the interferometer geometry (shuttering, defocussing the reference, averaging out the fringes, etc). The scope of the inventionis intended to be determined by reference to the claims and their equivalents interpreted in light of prior art.

* * * * *