DE102012025551B3 - Method for spatially resolved measurement of inclination of surface area of reflective surface of object, involves evaluating reflected color structure in beam cross section of reflected light beam in image plane - Google Patents

Abstract

A collimated light beam having beam cross-section is generated with a color structure and directed to reflecting surface such that to-be-measured surface is completely within area of beam cross-section of illuminating beam. An intensity distribution over beam cross section of light reflected at the surface portion of illumination beam in image plane is detected by optical detector assembly (15,17a,17b). A reflected color structure in beam cross section of reflected light beam in image plane, is evaluated for determining and spatially resolving the inclination of the surface area. An independent claim is included for device for spatially resolved measurement of inclination of surface area of reflective surface of object.

Description

Technical application

The present invention relates to a method for the spatially resolved measurement of the inclinations of a surface area of a specular surface of an object in which a collimated illumination beam having a beam cross section is generated and directed onto the specular surface so that the surface area to be measured is completely within the beam cross section of the illumination beam. an intensity distribution over the beam cross-section of the reflected on the surface region illumination beam detected in at least one image plane with an optical detector assembly and location of the optical detector arrangement spatially resolved the inclinations of the surface area are determined. The invention also relates to a device for carrying out the proposed method.

The method is suitable for the spatially resolved inclination measurement of arbitrary reflecting surfaces and can be used above all for the topography determination of these surfaces in the millimeter, micrometer and nanometer range. An important field of application is, for example, the topography of semiconductor wafers, as required for quality assurance in semiconductor manufacturing.

State of the art

The proposed method and the associated device are based on the so-called "Makyoh" or "magic mirror" method. In this case, a collimated illumination beam is directed onto a reflecting surface and the intensity distribution of the illumination beam reflected at the surface is recorded in a plane perpendicular thereto. This shot is also known as the Makyoh picture. Unevennesses of the measured surface act as local convex or concave mirrors, which locally focus or defocus the light. Accordingly, intensity fluctuations occur in the recorded image plane at a distance L from the surface, which contain information about the topography of the surface. This information is initially qualitative only. The Makyoh image can thus be used to detect defects or imperfections, as described, for example, in S. Hahn et al., "Characterization of mirror-like wafer surfaces using the magic mirror technique," J. Cryst. Growth 103, 423-432 (1990).

Various approaches are known for obtaining quantitative results from this measurement method. One of these approaches is based on a geometrical-optical model in which the Makyoh image is considered, to some extent, as the Laplace image of the surface. The topography can then be determined by solving a Poisson equation. However, this model only applies to small distances L. For a high sensitivity of the inclination measurement and thus a high resolution, however, large distances L of the image plane to the surface are necessary. Therefore, this approach must find a compromise between model validity and resolution. Another approach uses an iterative method to calculate the phase of the reflected wavefront and the reflectivity of the surface. The basic idea of this wave-optical approach is a determination of the phase of the reflected wavefront by recording the Makyoh image in at least two different distances L from the surface. An iterative phase determination algorithm can then be used to determine the phase of the wavefront and thus the topography of the surface. However, this is a complex process that requires long computing times, as described, for example, in Z.J. Laczik, "Quantitative Makooh Topography", in: Optical Engineering 39 (9), 2562 (2000).

From JE Lukács et al., "Subpixel detection of a grid's node position for optical diagnostics", in: Thin Solid Films 516 8082-8086 (2008), a method for measuring the inclination of a specular surface is known in which structured illumination for The inclination measurement is used to obtain quantitative results. In this method, a grating structure is projected onto the surface via an optical grating. From Riesz et al., "Realization of quantitative Makyoh topography using a Digital Micromirror Device," in: Proc. Of SPIE Vol. 6616 66160L-1 (2007), also the projection of other binary intensity patterns is known, which are generated by an electronically controllable DMD (digital micromirror device). In these cases of structured illumination, by analyzing the distortion of the projected pattern after reflection, the inclination of the surface at the positions of the points identifiable in the projected structure, hereinafter referred to as label points, is calculated. However, in the case of a strong distortion, the original structure is no longer clearly comprehensible, so that the considered label points can no longer be unambiguously assigned for lack of distinctness. Especially at a large distance L, which is required for the detection of minute inclinations occurs for existing larger inclinations a correspondingly large shift of the considered points and thus the risk of overlap. The measuring range must therefore be restricted accordingly with this technique. In addition, the lateral resolution is limited, for example, in the projection of a grid structure by the discrete label points, since in addition to the bright label dots and dark intermediate areas (webs of the grid) are required.

From A. von Finck et al., "Makyoh Imaging for Characterizing Reflective Surfaces", in: tm - Technisches Messen 2009 76: 1, 26-33, a method is known in which linear gray filters are used in the beam path of the illumination radiation, to thereby obtain a multi-level intensity coding of the illumination radiation. Different areas of the beam cross-section projected onto the surface can then be identified on the basis of their brightness value. This facilitates the assignment of individual points of the projected pattern. However, when superposition of partial beams by the reflection, the brightness values can add up and are therefore no longer separable from one another in the recorded image. Furthermore, this method requires surfaces with a homogeneous reflectivity, since the reflection coefficient of the surface at the appropriate point has an influence on the intensity of the reflected partial beam. A measurement of, for example, structured semiconductor wafers with a reflection coefficient varying over the surface is then no longer possible. But also for surfaces with homogeneous reflectivity, this approach has disadvantages. The intensity of the reflected beam or beam is measured in a defined range (eg a pixel of a CCD) as a brightness value. The measured brightness also changes when the beam is locally focused or defocused by the surface. Depending on the magnitude of the change in the topography within the order of magnitude of a coded beam, this results in an inaccuracy in the identification of the beams, which are indeed coded via the intensity and distinguished by evaluating the brightness values.

The object of the present invention is to provide a method and a device for the spatially resolved measurement of inclinations of a surface area of a specular surface of an object, with which specular plane surfaces over a large area, d. H. also with diameters of the surface area to be measured of ≥ 300 mm, and can be measured quickly and inexpensively with a derived from the slope measurement height resolution in the nanometer range. The method and the device should in particular also be suitable for the topography determination of semiconductor wafers.

Presentation of the invention

The object is achieved with the method and the device according to claims 1 and 16. Advantageous embodiments of the method and the device are the subject of the dependent claims or can be found in the following description and the embodiments.

In the proposed method, a collimated illumination beam is generated and directed onto the specular surface so that the surface area to be measured is completely within the beam cross section of the illumination beam. The surface area to be measured may also be the entire reflecting surface, provided that it has sufficient planarity for the proposed measurement. With an optical detector arrangement, an intensity distribution over the beam cross section of the illumination beam reflected at the surface area is detected in at least one plane perpendicular to the propagation direction of the reflected illumination beam, referred to in the present application as the image plane. The inclinations at different positions or points of the surface area are then determined from the measured data of the optical detector arrangement and either the known distance of the image plane (s) to the surface or the known distance of several image planes from one another. The proposed method is characterized in that the collimated illumination beam is generated with a color structure over the beam cross section and directed to the specular surface, wherein the optical detector array is a reflected color structure obtained by the reflection at the surface in the beam cross section of the reflected illumination beam in the image plane recorded and evaluated for the determination of the inclinations of the surface area. In this case, a color structure is understood to be a spatial dependence of the wavelength or color of the illumination beam over the beam cross section, in which the wavelength or color varies over the beam cross section. Color means a wavelength or the superposition of two or more wavelengths with possibly different intensities. Such a color structure represents a color pattern and is also referred to below as a color pattern.

In the proposed method a distinctness of adjacent label points is thus achieved via the color pattern generated, so that even after a change of the color pattern by the reflection on the surface, the individual label points can still be clearly identified or assigned to their original position in the color pattern. In the method, a defined color pattern is thus projected onto the surface to be measured or to the surface area to be measured, wherein different points of the pattern are preferably labeled or identified via the evaluation of the wavelength or color instead of exclusively using the latter as in the latter prior art Intensity. The wavelength encoding of the collimated illumination beam avoids those occurring in the prior art in the context of structured illumination Overlay problems. A superimposition of individual label points in the image plane of the reflected illumination beam or a crossing of the associated sub-beams is unproblematic, since each label point of different wavelengths can be detected independently of the other label points and thus traceability of the associated sub-beam and thus a direction determination of this sub-beam is possible. An arrangement of the respectively detected label point in the originally projected pattern is thus possible in a simple and unambiguous manner. The proposed method is also independent of reflectivity differences of the object surface, as long as light of the corresponding wavelength is not completely absorbed by the object material. By solving the overlay problem, the maximum selectable distance L of the image plane from the object surface is no longer limited by the distance at which the label points overlap at a given maximum inclination of the surface region. A larger possible distance L thus allows a smaller still measurable inclination. The number of distinguishable label points is as large as the number of different wavelengths available through the illumination. If there is knowledge about the maximum expected inclination differences and thus maximum shift of the label point, the pattern consisting of different label points can be limited to a certain area and then repeated accordingly. Ie. Label points of the same coding (= wavelength) can occur several times, if one knows that there is no overlapping or crossing of indistinguishable label points in the area.

With the method and the associated device, a simple structure and a fast and convenient method for realizing a quantitative Makyoh sensor are available. By generating a collimated illumination beam having a sufficiently large beam diameter, in conjunction with the wavelength or color coding, a large-scale inclination measurement of a specular surface, for example a semiconductor wafer with a diameter of 300 mm, can be performed. The evaluation of the measurement data of the detector arrangement for the determination of the spatially resolved inclination can then be carried out by known methods, as are known from the publications mentioned in the introduction, which use structured illumination for the inclination measurement.

The device for carrying out the proposed method accordingly comprises a lighting unit with which an illumination beam with a suitable color structure can be generated, and a collimator unit connected downstream of the lighting unit, which collimates the illumination beam generated by the lighting unit. The collimated illumination beam represents the measuring or output beam of the device which is directed at the surface area to be measured. The device comprises an optical detector arrangement with which the reflected color structure in the beam cross section of the measurement beam reflected at the surface area can be detected in at least one image plane perpendicular to the propagation direction of the measurement beam reflected at the surface area.

The optical detector arrangement is one or more spatially resolving detectors with which, if appropriate in combination with corresponding color filters, the color structure in the beam cross section of the reflected illumination or measurement beam in the image plane can be detected. This may be one or more CCD chips or CMOS chips, which are arranged perpendicular to the optical axis of the reflected illumination beam. Each different wavelength or each different wavelength range of the illumination beam used must be able to be detected. This can be done for example by different CCD chips with corresponding upstream color filters for each wavelength or each wavelength range or by a CCD chip with an upstream filter wheel, which includes the corresponding color filter. When using a color structure with three different wavelengths or colors, in particular red, green and blue, a 3-CCD camera, a normal RGB camera or a CCD chip with filter wheel with corresponding filters for red, green and blue can be used , In order to take images of the reflected light at different distances from the image plane of the surface to be measured, either a detector or a camera can be used, which is displaced axially to the optical axis of the reflected illumination beam or a fixed arrangement of several cameras at different distances from each other , It is also possible to use an adjustable camera lens. The distance of the image plane to the surface to be measured can then be varied by changing the lens distance.

The color pattern used for illumination structuring or reflected by the sample surface is composed of regions of different wavelength or color, for example from a spectrum or individual points of different wavelengths. These label points can be detected according to their different wavelengths and separated from each other. Overlapping label points can be assigned at any time. This is independent of the reflectivity of the object surface, through which only the intensity of the reflected wavelength components, not however, the wavelength itself changes. Thus, the method allows the measurement of the surface of z. As structured semiconductor wafers, which may consist of areas with different materials and correspondingly different reflectivity. For example, different monochromatic light sources in the illumination device or narrow-band dichroic filters can be used for the generation of individual wavelengths or very narrow-band wavelength ranges of the illumination. By evaluating the deformation of the encoded pattern in the recorded image plane versus the known composition of the pattern in the irradiated illumination beam, the inclination of the measured surface area at different points of the surface area can be calculated.

In the proposed method, it is also possible to generate a larger number of colors by overlaying different wavelengths combined with an intensity modulation, via which the label points can then be correspondingly identified. When using a lighting with, for example, three wavelengths or wavelength ranges, z. B. red, green and blue, any colors of the RGB color space can be generated by superposition of the three wavelengths or wavelength ranges and corresponding intensity modulation. For evaluation, the RGB image of the reflected illumination or measuring beam recorded by the detector arrangement is z. B. transferred to the HSI color space (Hue, Saturation, Intensity) and the hue used to identify the label points. With overlapping of label points, however, a clear assignment is not always possible in every case. For topographies with slight differences in inclination, however, the distance L of the image plane to the object surface can be chosen so that there is no overlay or overlap of label punk. For relatively flat surfaces such. As polished semiconductor wafers, this aspect may be true. Compared to a multi-level pure intensity coding then there is still the advantage that a clear identification of the label points based on the different colors is possible in a simple manner. In the case of binary intensity coding, on the other hand, if the pattern is greatly distorted, it may no longer be possible to decide to which original position (eg row and column in a matrix arrangement of label points) the label point must be assigned, since the label points do not differ intrinsically. Compared to a binary pure intensity coding according to the prior art, the method still has the following advantages. There is a larger number of distinguishable label points available. In the case of pure prior art intensity coding, the maximum number of distinguishable label points that a pattern may contain is typically limited by the number N _B of distinguishable brightness levels of the detector array (eg, CCD sensor). The proposed method uses a color coding, ie a label point is determined by the wavelength and the brightness value. This makes (N _B ) ^N distinguishable label points available. N is the number of wavelengths available and distinguishable by the detector array. Another advantage is the solution of the already mentioned problem of the dependence of the intensity of a detector pixel on local focusing or defocusing of the corresponding beam. Focusing or defocusing changes the intensity per area, but not hue and saturation (in the HSI color space). If, for example, three different wavelengths are available, one obtains as many distinguishable label points as there are different colors in the HSI color space for I = const. gives.

Thus, if the problem described is relevant, the above number (N _B ) ^N of label points used may be constrained by one dimension of the HSI color space to become independent. Since a combination of wavelength coding and intensity coding is used, the method in this embodiment is generally no longer independent of the reflectivity of the surface, but a homogeneous reflectivity is a prerequisite.

Preferably, in the latter embodiment, a deep-sharp pattern (eg, a sine pattern) is used. Since in practice no ideal point light source can be realized, the image of the pattern has a finite depth of field. In order to achieve the greatest possible lateral resolution, it is advisable to image the pattern sharply on the surface to be measured. Depending on the size of the distance L of the image plane to the surface, the image is then recorded on the detector out of focus. When using large distances L and thus blurred image, deep-focus patterns are advantageous. The use of sinusoidal patterns in intensity has z. B. the advantage that the phase of the sine does not change with blurred image, only the contrast again lower.

In a further advantageous embodiment of the method, a phase shift method for identifying label points is used. Phase shift methods are z. B. from interferometry or triangulation known. In this case, the intensity is modulated in each pixel in time, wherein for the evaluation then several shots are made sequentially. Due to the a priori known frequency modulation can be in each pixel phase be determined, which serves to identify the label point. A variety of phase shift techniques are known (eg, three-shift, four-shift, least-square, etc.) which differ in the number of shots required and the size of the shift.

For the method proposed here, the phase shift method is preferably used in the manner described below. The values of the modulation are realized by means of the intensity, while the different recordings are not made consecutively, but at the same time in different wavelengths. The pattern of each of the wavelengths used is therefore shifted in phase relative to the other wavelengths. Realized is the z. By projecting a sinusoidal intensity pattern in one direction in one wavelength, while the same sinusoidal intensity pattern is out of phase in the other wavelengths in said direction. When using a sine pattern, at least three images with shifted sine are required in order to be able to determine the phase position of the pixels in one direction. The phase values and subsequent phase unwrapping then uniquely determine the coordinates of the label points in one direction. To determine the full position of the label points, the same procedure is also performed in the y direction. Thus, in this example, at least six shots with shifted sine intensity patterns are required, which are obtained simultaneously through the use of at least six distinguishable wavelengths or wavelength ranges. If, for example, only three different colors (eg RGB) are available, at least two shots are required in succession. With this embodiment of the method, an independence of reflectivity differences of the surface is achieved over the aforementioned embodiment. Furthermore, this embodiment also solves the already mentioned problem of the dependence of the intensity of a detector pixel on local focusing or defocusing of the corresponding beam.

In the proposed method, a recording of the reflected illumination beam is made in an image plane having a certain distance L to the object surface. By determining the displacements Δd of the positions of the label points in the image recording relative to the positions of the label points in reflection on a perfectly flat object surface, the deflection angle of the partial beams corresponding to the label points and thus the inclination angle α of the surface at the positions of the label points can be calculated via a geometrical calculation : tan (2α) = Δd / L. The positions of the label points upon reflection on a perfectly flat object surface can be obtained by a corresponding reference measurement on such a surface. The distance L must be known and calibrated accordingly. For very small inclinations, such as surfaces of polished semiconductor wafers, large distances L have to be selected (eg more than 10 m) in order to be able to detect the beam deflections. In this case, a very accurate inclination determination (<0.1 arcsec) is possible despite coarse accuracy in the determination of L (accuracy in the millimeter range). The condition for this is that L is very large against Δd. In general, the same shift Δd of a label point can be achieved both by an inclination of the object surface and by a different inclination in combination with a height difference on the object surface. With small expected height differences and small angles of inclination (L very large against Δd), however, the sensitivity of the described inclination calculation to height changes is negligible.

Alternatively, the inclinations of the surface can also be determined by taking at least two images in two different image planes of the reflected illumination beam, which have different distances L ₁ and L ₂ to the object surface, instead of a reference image. The inclination α at a point on the surface is then calculated by the change in position Δd of a recording point of the recording at a distance L ₁ for recording at a distance L ₂ and the distance ΔL = L ₁ -L _{2 of} the two image planes: tan (2α) = Δd / .DELTA.L. A reference recording is no longer necessary in this case (cf. 5b ).

When recording at least two images at different distances from the object surface, it is advantageous to choose one of the two distances negative, preferably L ₁ = -L ₂ . A negative value of the distance L means an image plane which, starting from the camera or the detector, does not lie before but behind the object surface. The choice of a negative distance L is possible virtually (virtual image plane) by the imaging by means of a lens. As a result, an enlargement of the vertical measuring range is achieved. The distance L or ΔL must be selected according to the expected inclination angle of the surface to be examined. He has to be big enough to be able to detect the smallest required inclinations. At a large distance L or ΔL, the displacements .DELTA.d of the label dots are correspondingly large and may under certain circumstances no longer be detected since they no longer reach the aperture of the detector unit or no longer hit the detector chip. Therefore, always a compromise for the size of the distance L must be found. The maximum occurring Surface slopes limit the size of L, which in turn restricts the smallest detectable surface tilt. By selecting L ₂ = -L ₁ , the measuring range is doubled. The distance of the two image planes is then ΔL = L ₁ - L ₂ = 2L ₁ . The resulting shift of a label point from image plane 1 to image plane 2 at a surface slope of α is then Δd = 2L ₁ tan (2α). In both recordings, however, there are maximum position changes of Δd = | L ₁ | tan (2α _maximum ) compared to a flat reference (L = 0). In general, you can increase the vertical range when | L ₁ - L ₂ | = | ΔL | > | L ₁ | and o. B. d. A. | L ₁ | ≥ | L ₂ |.

In a further advantageous embodiment of the proposed method, images are made at more than two different distances L to the object surface. This achieves a further increase in the accuracy of the inclination measurement. By determining the changes in the positions of the label dots in more than two image planes in the reflected illumination beam, the inclination of the straight line passing through the label point in the different planes can also increase the accuracy of the slope of the line passing through the label point in the different planes be determined.

The topography (relative height) of the surface or of the surface area can be determined from the spatially resolved inclinations of the measuring surface area determined by the method. The tilt data (x and y component of the slope at each point) provides the gradient field of the surface. To calculate the height, the following equation must then be solved: grad (h (x, y)) = ( _{nx '} (x, y), _ny' (x, y)) ^T , where h (x, y) the relative height at each point represents x, y and n _{x '} and n _y' corresponding to the _{x '} and _y' components of the slope at each point x, y. To calculate h (x, y), appropriate integration algorithms are used. Local integration methods can be used, for example direct integration by means of a trapezoid rule, or global integration methods, for example based on the least-squares method using z. As the Fourier transform or cosine transform. Examples of global integration methods are the method according to Frankot-Chellappa or the Poisson-Solver method.

As a light source of the illumination unit, a point light source is preferably used in the proposed method and the associated device, for example. An LED, a laser or other lamp in combination with a pinhole. For example, color filters can be used for the illumination structuring. Thus, for example, a filter can be used which transmits a certain pattern of different wavelengths when illuminated with white light. This may, for example, be a glass plate with color dots or color areas that allow different wavelengths or wavelength ranges to be transmitted at different points or in different areas. It can z. B. Bayer filter, interference filter or linear variable color filter (LVF) can be used or in a modification and wavelength-splitting elements such. B. diffraction gratings or prisms.

For the illumination structuring also electrically controllable light modulators can be used. With these light modulators, it is possible in a simple manner to also project temporally variable patterns. The process gets thereby an additional temporal component. The electronically controllable light modulators are preferably displays which consist of an arrangement of controllable pixel structures and can structure a light beam correspondingly (digital light processing). Examples of such light modulators are DMD (Digital Micromirror Display), LCD (Liquid Crystal Display) and in particular LCoS (Liquid Crystal on Silicon) displays. These displays are each irradiated with light of different wavelength or wavelength ranges (DMD, LCoS) or irradiated (LCD), whereby a spatial structuring of the light of a wavelength or a wavelength range is possible. With LCoS displays and LCDs, it is also possible to perform an intensity modulation in each pixel in addition to the spatial modulation. This is possible because liquid crystals are used both in LCoS displays and in LCDs, which can be aligned differently by electrical control. Thus, an intensity modulation of the light can be achieved by the use of polarization. Even with the use of DMDs, patterns with different levels of brightness can be created by tilting the mirrors in different frequency on / off and taking the camera shot with a sufficiently long exposure time. In order to generate a pattern consisting of different wavelengths, then either a plurality of light sources of different wavelengths or wavelength ranges are used, each of which irradiate one of the light modulators, or a white light source is, for example, by means of dichroic filters or wavelength-separating prisms (X-cube) in different Wavelengths or wavelength ranges are split, which then in turn in each case a light modulator is irradiated. The partial beams of the different wavelengths or wavelength ranges are structured by the different light modulators and finally combined again into a beam, in particular laid exactly one above the other. This can, for example, again with dichroic filters or wavelength-separating Prisms take place. In this way, a color-structured illumination beam is generated, the structure of which can also be changed at any time by corresponding activation of the light modulators. DMDs, LCDs and LCoS displays are currently available in resolutions up to 1920 px × 1200 px. They are controlled by suitable control units (controllers) z. In connection with a computer.

The use of electronically controllable light modulators also makes it possible to illuminate different regions of the object surface in succession, thereby additionally avoiding overlays of label points or overlapping of light rays. Thus, for example, only as many label points can be scanned one after the other over the object surface, point by point, as different and distinguishable wavelengths are available. For each appropriate time then a recording is made. The positions of the independent label points can then be determined without the influence of other label points. This results in a scanning optical measuring method in which no mechanical movements are required. Especially for highly accurate measurements, for example of topographies in the nanometer range, mechanical movements are always a challenge that can be avoided with this embodiment of the method.

In a very advantageous embodiment of the proposed method, an intelligent and data-efficient illumination structuring is made possible with the use of electronically controllable light modulators. The pattern, d. H. the number and distribution of the label points can be adapted dynamically to the examined surface area. Thus, for example, initially a relatively coarse pattern, d. H. directed little label points at long intervals on the surface area to be measured and made a corresponding recording to evaluate the inclinations. After evaluating the measured inclination changes of label points to their neighboring points, it can then be determined at which points short-wave topography changes occur at the object surface. In these areas, correspondingly more label points can be projected with shorter distances, ie greater lateral resolution, since a larger lateral resolution is required there in order to fully capture the topography, whereas in areas in which the topography changes only minimally no greater lateral resolution is required. Thus, it is possible to achieve a number and distribution of label points adapted to the respective surface. The advantage is an efficient number of measurement or data points, which must be stored and evaluated and avoid redundancy. This also results in an increase in the measuring speed and correspondingly larger throughputs. In one possible embodiment, a conventional Makyoh image can be recorded for the first analysis, in which unstructured illumination is used. Based on this Makyoh image, it is then possible to decide what the optimal, most efficient lighting structuring for the surface in question looks like. The traditional Makyoh image is the intensity image obtained with unstructured homogeneous illumination. To a certain extent, this image can be considered as the Laplace image of the surface topography. It thus represents, to a certain extent, a derivative of the surface inclination and has sensitivity to the surface curvature. Thus, it is ideal for detecting large changes in topography and thus optimal for setting an efficient illumination structuring. If, for example, on a polished semiconductor certain defects such. As "dimples" are quantified, the conventional Makyoh image can be used to detect the "dimples" and then projected label points with a suitably small mutual distance at the interesting areas, so that the slope and ultimately the topography of the "Dimples" to determine.

The lateral resolution can also be improved in the method by taking several pictures in succession with a laterally shifted pattern. The label points must be chosen large enough to be detected with the pixels of the CCD or detector. Thus, for example, a label point can consist of a plurality of pixels of the light modulator or of the light modulators. As measuring points at which the inclination is calculated, z. B. the boundary points between four label points are used. The lateral resolution is then given by the distance of the measuring points to each other. The lateral resolution can now be increased by taking a further image in which the projected pattern consisting of the label points is shifted laterally by, for example, one pixel of the light modulators. The lateral resolution in this case is the length corresponding to a displacement of a pixel of the light modulator on the object surface. In the case mentioned above, this length is smaller than that of a label point consisting of several pixels.

A lateral displacement of the pattern can also be achieved by the use of filters with variable refractive index in the beam path of the illumination beam. These filters are introduced to achieve lateral displacement of the pattern projection. With a lateral shift of the or the structure-generating filter or light modulators, the same effect be achieved. However, this has the disadvantage that a mechanical displacement is required, which is always a challenge for high-precision measurements. The filters of variable refractive index which are preferably used consist of a material which changes its refractive index by applying an electric field (electro-optical Kerr effect). By appropriate configuration can be achieved that a striking the plate light beam at the interfaces of air to the plate and plate to the air is refracted so that an offset of the light beam results perpendicular to the optical axis, wherein the direction of the light beam is maintained. The size of the offset is dependent on the applied electric field.

An improvement in the lateral resolution can also be achieved by a fixed lateral offset of light modulators to one another. The projected pattern then consists of overlapping label points, which, however, can be detected independently of each other due to their different wavelengths. The lateral resolution can thereby be improved to a size less than the distance that corresponds to the pitch of adjacent pixels of the light modulator on the object surface. In addition, filters with a variable refractive index can also be used here in order to be able to electrically change the offset between the light modulators. Such a filter is required behind each of the corresponding light modulators.

The use of electronically controllable light modulators also makes it possible to create special patterns for calibrating the measuring arrangement and to evaluate them accordingly. For example, an orientation of the object surface which is as vertical as possible to the optical axis of the incident illumination beam can be advantageous in order to obtain the lowest possible global tilting in the measured data. For this purpose, for example, a label point can be projected onto the object surface and the reflection can be detected by two recordings at different distances L. The object can now be tilted manually or automatically until the position of the label point is in the same position in both images. This allows the object surface at the point where the label point was projected, with high accuracy perpendicular to the optical axis of the device or measuring arrangement, d. H. be aligned with the incoming illumination beam.

In one embodiment of the proposed method and the associated device is used as a lighting unit, a projector for color projection, for example. A DLP, LCD, LCoS, SxRD, D-ILA projector. Depending on the design of the projector, the objective of the projector may have to be exchanged or adapted and, if necessary, modifications made in the beam path. In particular, 3-panel LCoS projectors are particularly well suited to the requirements, as they combine the advantages of LCD and DLP projectors and avoid their disadvantages. No color modulation is required for color representation (unlike DLP projectors). They have a uniform pixel structure and high contrast values. Such projectors are commercially available at low cost. A control of the projector and thus the light modulators is easily possible via a conventional graphics card of a computer, with any color structures can be generated.

In the proposed method and the associated device a point-like light source should be realized in order to obtain the best possible collimation of the illumination beam. For this purpose, preferably a spatial filter is used, which, for example, consists of a pinhole through which the illumination beam is focused. With the spatial filter higher frequencies of the wavefront are filtered out, d. H. so rays outside the focus. By this measure, a very well-collimated illumination beam, so in good approximation, a flat wavefront can be obtained.

For example, a spherical mirror or a corresponding lens construction can be used to generate the collimated illumination beam. Using a mirror for collimation requires more space. For this, mirrors can usually be manufactured with better surface accuracy than lenses. When using a lens, the use of an error-corrected lens, for example, an achromatic lens or an aspherical lens is recommended. The collimated illumination beam preferably strikes the object surface almost perpendicularly. However, this is not necessary in every case. Thus, an arrangement is also possible in which the collimated illumination beam strikes the surface at an angle ≠ 90 ° and is correspondingly reflected at the same angle to the surface normal and detected accordingly.

In the proposed method and apparatus, a collimated illumination beam having a color pattern or a color structure is projected onto the object surface. It is a Makyoh inclination-measuring method, in which the topography (height of the surface) can be calculated in a secondary step from the inclination data. The method can be used to measure reflective surfaces. In this context, mirroring means that a ray of light falling on the surface is reflected according to the principle of angle of incidence equal to the angle of reflection. to Obtaining the slope data uses the corresponding color-structured or color-coded illumination, in which different areas or points differ in part in their wavelength. The tilt data is determined by determining the position of different label points in one or more image planes perpendicular to the optical axis of the reflected illumination beam. With the method and the associated device, a large field of view can be realized, which can be measured over the entire area in a spatially resolved manner. The method enables a highly accurate inclination measurement with <0.1 arcsec. The projected label points can be uniquely identified due to the wavelength encoding, even when superimposing individual label points. This identification is also independent of the reflectivity of the sample surface. When used with electronically controllable light modulators for lighting structuring results in a considerable flexibility in the survey.

With the method and the device, a powerful 3D sensor can be realized, which is suitable for the investigation of reflective surfaces of all kinds. Thus, for example, mirrors, polished metal surfaces, glass surfaces, reflective paint or plastic surfaces and, in particular, bare and structured semiconductor wafers can be measured with high accuracy. Due to the high accuracy of the inclination determination, the method is particularly suitable for determining the surface accuracy and flatness of a plane surface. In addition, the method and the device for detecting and quantifying defects z. As scratches, cracks or polishing marks. Due to the ability to realize a large field of view, the method and apparatus are suitable for all applications where large area topography determination with high accuracy is required.

Areas of application are particularly in the field of semiconductor manufacturing for quality control of semiconductor wafers. Advantages include a high-precision measurement of topographies of semiconductor wafers ("bow", "warp", "flatness" and in particular "nanotopography"), high throughputs and a simple and cost-effective implementation. The method is also relatively robust to external vibration, as there is no direct sensitivity to the height of the object. The method can be used as an integrated measurement technique and serve for the qualification of production equipment and processes, for example of CMP processes (chemical-mechanical planarization). Furthermore, applications in a variety of other areas are possible, for example. Ophthalmology, especially for the optical, non-contact examination of the cornea.

Brief description of the drawings

The method and the device will be explained in more detail below with reference to exemplary embodiments in conjunction with the drawings. Hereby show:

1 a first example of a construction of the proposed device;

2 a second example of a construction of the proposed device;

3 a third example of a construction of the proposed device;

4 another example of a structure of the lighting device;

5 two examples of the choice of image levels in the detection; and

6 an example of a lateral offset of the color pattern.

Ways to carry out the invention

The proposed device for the full-surface measurement of the inclinations of a specular surface is composed of a wavelength-coded illumination structuring illumination unit, a collimator unit for producing a collimated illumination beam projected onto the specular surface, and an optical detector unit with which the reflected illumination beam is converged several image planes is detected spatially and color resolved. 1 shows a first example of the structure of such a device. The lighting unit consists of a point light source 1 , whose output beam via a first collimator lens 2 is collimated. In the beam path of this collimated light beam is a filter 19 used for lighting structuring, for example. An LVF. Through this filter 19 a color structure is generated over the beam cross section of the illumination beam, so that different regions of the beam cross section have partly different colors or wavelengths. After the illumination structuring, the light beam with a lens 6 through a spatial filter 7 focused, in this example, a pinhole to filter out of focus rays and thus higher frequencies of the wavefront of the illumination beam. This can be done after reflection at the beam splitter 8th with the further collimator lens 11 a well-collimated illumination beam is generated, which in this case is approximately perpendicular to the surface of the sample 12 is directed.

As a sample plate 14 a plate with the best possible flatness is used, for example a polished silicon plate in order to be able to measure semiconductor wafers as samples. In the case of a sample plate with a low surface quality, ie an uneven support surface, there is otherwise the risk that, when examining thin samples, such as the aforementioned semiconductor wafers, the sample and thus its surface will be deformed in accordance with the surface of the support. The topography of the sample tray would therefore push through a thin sample. For the measurement of semiconductor wafers therefore a sample plate with high flatness is required, the diameter of which is in the size of the diameter of the examined semiconductor wafer in order to realize a full-surface support. In the case of a nanotopography measurement on a semiconductor wafer is a soft pad 13 between sample 12 and sample plates 14 required. This may, for example, be a suitable cloth. The reason for this is that according to SEMI definition of the nanotopography measurement, the semiconductor wafer must be in an unreferenced state. When placed on a soft surface, the topography of the semiconductor wafer is not referenced by the support surface, that is not fully equalized by its own weight. Rather, the topography of the semiconductor wafer remains in a decisive for the nanotopography analysis spatial wavelength range, because it expresses itself accordingly in the soft pad. Possible minimal unevenness of the sample tray as well as particles, for example dust between sample tray and semiconductor wafer, are not critical since they press into the soft base accordingly instead of influencing the topography of the semiconductor wafer.

The on the surface of the sample 12 reflected illumination beam in turn passes through the collimator lens 11 and is over the beam splitter 8th separated from the original lighting beam. The reflected illumination beam is then transmitted through a lens 15 , For example, the lens of a camera used for detection, in the present example by a beam splitter 16 on two detector chips 17a . 17b , imaged a different distance to the lens 15 have and thus depict different image planes in the reflected illumination beam. Different image planes are understood as image planes at different distances from the sample surface. Instead of the two detector chips and only one detector chip can be used, which is then moved to perform the recordings between the two recording positions.

The reading out of the detector chips 17a . 17b via the corresponding camera electronics in conjunction with a PC 18 , The inserted lens 15 is respectively positioned so that - in the case of a perfectly flat sample surface - a parallel beam path of the reflected illumination beam results, in which the detector chips 17a . 17b are arranged. Since each different wavelength range of the color structure of the illumination beam must be detected, either detector chips with corresponding upstream color filters, detector chips with an upstream filter wheel or corresponding wavelength-sensitive detectors can be used, as has already been explained elsewhere. By evaluating the deformation or change of the coded color pattern in the respective distance of the image plane to the sample surface, the inclination of the sample surface can then be calculated spatially resolved.

2 shows a further example of an embodiment of the proposed device, which is in the construction of the illumination device and the collimation of the structure of 1 different. In this example, electronically controllable light modulators are used for the illumination structuring 5a . 5b and 5c used, in particular LCoS displays. In this example, three LCoS displays are used. The white light of the light source 1 is through an X-Cube 4 , ie a dichroic prism, split into three beams of different wavelength ranges. Each beam is structured separately by an LCoS display. Subsequently, the three rays are again through the same X-Cube 4 united. The polarizing beam splitter 3 is used to linearly polarize the light, which is required for the intensity modulation by liquid crystals in the LCoS displays. It also serves to decouple the light reflected from the LCoS displays. This decoupling works because the polarization direction has rotated by 90 ° after reflection on the LCoS display and now at the polarizing beam splitter 3 no longer transmitted but reflected. Instead of a polarizing beam splitter so-called. Wiregrid polarizers can be used. The control of the light modulators 5a . 5b . 5c done via the PC 18 , Again, the color-structured illumination beam is initially again through a spatial filter 7 focused to filter out higher frequencies of the wavefront.

In the example of 2 For example, for the collimation to produce the collimated illumination beam, an array of multiple lenses is formed 9 . 10 . 11 used in series. The lenses have a clever combination of different refractive indices, focal lengths and positions. As a result, the best possible correction of known lens defects and, on the other hand, a small footprint of the structure can be achieved. In a large field of view of, for example, 300 mm diameter is a correspondingly large collimating lens 11 required. For manufacturing reasons, there is a lower limit on the ratio of focal length to Diameter of the lens. If only one lens were used, then the structure would have a length of at least the focal length of the lens. By using a corresponding lens configuration as in 2 the space requirement can be reduced.

Objective telecentric lenses are also suitable for collimation 21 as exemplified in the construction of 3 is shown. These telecentric lenses are available with correspondingly large apertures. A sufficient size of the telecentric screen is required. There are also telecentric lenses available, which have a possibility for vertical, coaxial lighting coupling. As a result, no additional beam splitter for the beam splitting between incident and reflecting illumination beam longer needed because the illumination beam can be coupled directly laterally into the lens, as in 3 is indicated. A telecentric lens also has the advantage that sample-injected stray light is filtered away.

In the construction of 3 becomes a projector as a lighting unit 20 used, whose objective lens has been changed to focus on the focusing lens 6 in the spatial modulator 7 to enable. The control of this projector 20 to generate the color pattern again takes place via the schematically indicated in the figure PC 18 ,

4 shows a further example of an embodiment of the illumination device, which differs from the illumination device of 2 through a larger number of X-Cubes 4 and electronically controllable light modulators 5a - 5f different. This turns the white light of the light source in this example 1 split into a correspondingly larger number of different wavelength ranges, which are each structured with the associated light modulators. Thus, more different wavelengths are available and color structures with an even greater number of distinguishable colors can be generated. As indicated in the figure by the dots, if necessary, even more X-cubes 4 be used with associated light modulators.

5a schematically shows the geometric relationships in an embodiment in which a recording of the reflected on the sample surface measurement or illumination beam is made at a distance L to the sample surface. The figure shows a partial beam 100 of the incident illumination beam, one opposite a reference plane 200 an area inclined by an angle α 120 the sample surface, the partial beam reflected at the area 110 as well as the considered label point 130 in the picture plane 300 , By determining the displacement Δd of the label point with respect to the position with a perfectly flat sample surface or a photograph at L = 0, the deflection angle 2α of the partial or label beam and thus the inclination angle α of the sample surface at the viewing area can be calculated by geometrical calculation.

5b schematically shows the geometric relationships in a further embodiment in which the inclination of the sample surface from two images at different calibrated distances L ₁ and L _{2 is} calculated. In the example of 5b If one of the distances is negatively chosen, here L ₂ = -L ₁ . As a result, an enlargement of the vertical measuring range is achieved. The figure shows the two image planes 310 . 320 at a distance L ₁ or L ₂ = -L ₁ to the reference plane 200 , Here is a shift Δd of the label point 150 in the first picture plane 310 opposite the position of the same label point 140 in the second picture plane 320 measured. From this shift, in turn, the inclination angle α of the sample surface at the viewing area 120 be calculated. As compared with 5a can be seen, the embodiment of the 5b a higher sensitivity, since the measurable displacement Δd at the same inclination angle α is doubled.

6 finally illustrates the enlargement of the lateral resolution in a lateral displacement of the projected color structure, as has already been explained in connection with an advantageous embodiment of the method. In this case, for example, two images are taken in succession with a laterally shifted pattern. In 6 For this purpose, two mutually shifted identical color patterns are shown, in which the distinguishable label points 35 of the first and distinguishable label points 45 of the second, identical pattern are respectively quadratic surfaces, each composed of a plurality of pixels of the light modulators. As measuring points 30 . 40 serve in this example in each case the boundary points between four label points. This is in the figure with the measuring points 30 for the first pattern and the measurement points 40 indicated for the shifted second pattern. By lateral displacement of the projected pattern (indicated in the figure by the second pattern) by z. B. a pixel of the light modulators, the lateral resolution is increased over a projection without displacement. The lateral resolution in this case corresponds to the length corresponding to a displacement of a pixel of the light modulator on the sample surface. In the present case, this length is smaller than that of a label point. As described elsewhere, the same result can be achieved by having one group of light modulators acting to produce a first pattern shifted to another group of light modulators used to produce a second, distinguishable pattern. Thus, only one shot is required to achieve lateral resolution enhancement.

As can be seen from the foregoing description and embodiments, the proposed apparatus and method is a Makyoh-type construction. In contrast to the original Makyoh method, the new method offers the possibility of a quantitative evaluation. Any structures can be used as the color structure. An example is a pixel pattern in which respective adjacent pixels have different wavelengths or colors. For example, the individual pixels can then be used as label points. It goes without saying that areas of the same wavelength or color encompassing several pixels can also be used.

LIST OF REFERENCE NUMBERS

1

Point light source

2

collimator lens

3

polarizing beam splitter

4

dichroic prism (X-Cube)

5a

Light modulator 1

5b

Light modulator 2

5c

Light modulator 3

5d

Light modulator 4

5e

Light modulator 5

5f

Light modulator 6

5g

Light modulator 7

6

lens

7

spatial filter, z. B. pinhole

8th

beamsplitter

9

lens

10

lens

11

collimator lens

12

sample

13

soft pad

14

sample tray

15

lens

16

beamsplitter

17a

Detector chip 1

17b

Detector chip 2

18

PC

19

Filter for lighting structuring

20

projector

21

telecentric lens

30

Measuring points of the first sample

35

Label points of the first pattern

40

Measuring points of the second pattern

45

Label points of the second pattern

100

Partial beam of the illumination beam

110

reflected sub-beam

120

inclined surface area

130

label point

140

label point

150

label point

200

reference plane

300

image plane

310

Image plane at L ₁

320

Image plane at L ₂

Claims (27)

Method for the spatially resolved measurement of the inclinations of a surface area of a specular surface of an object, in which - a collimated illumination beam having a beam cross section is generated and directed onto the specular surface so that the surface area to be measured is completely within the beam cross section of the illumination beam, - an intensity distribution via the beam cross-section of the illumination beam reflected at the surface area in at least one image plane with an optical detector arrangement ( 15 . 17a . 17b ) and - from measurement data of the optical detector arrangement ( 15 . 17a . 17b ) spatially resolved the slopes of the surface area are determined, characterized in that the collimated illumination beam is generated with a color structure over the beam cross-section and directed to the specular surface, wherein with the optical detector array ( 15 . 17a . 17b ) detects a reflected color structure in the beam cross section of the reflected illumination beam in the image plane and is evaluated for determining the inclinations of the surface area. A method according to claim 1, characterized in that the color structure via color filter ( 19 ) or wavelength splitting optical elements is generated. A method according to claim 1, characterized in that the color structure via a plurality of electrically controllable light modulators ( 5a . 5b . 5c ) is generated, which are irradiated with light of different wavelengths or wavelength ranges and their output beams are superimposed. A method according to claim 1, characterized in that the color structure with a projector ( 20 ) is produced. Method according to claim 3 or 4, characterized in that the light modulators ( 5a . 5b . 5c ) or the projector ( 20 ) are controlled so that different areas of the surface area to be measured are illuminated in succession with a corresponding subarea of the color structure and the respectively reflected color structure with the detector array ( 15 . 17a . 17b ) is detected. Method according to one of claims 3 to 5, characterized in that the light modulators ( 5a . 5b . 5c ) or the projector ( 20 ) are so controlled that in regions of the surface area to be measured with greater changes in inclination, a grid of points which are distinguishable by their wavelength or color is produced with a smaller spacing of the points in the color structure than in areas with conversely smaller inclination changes. A method according to claim 6, characterized in that the areas of the surface area to be measured with stronger inclination changes by taking a Makyoh image without illumination structuring or a spatially resolved inclination measurement with a coarse grid are determined by their wavelength or color distinguishable points in the color structure in advance. Method according to one of claims 1 to 7, characterized in that for an increase in the lateral resolution of the tilt measurement, the color structure one or more laterally laterally displaced and the respective reflected color structure in the beam cross section of the reflected illumination beam in the image plane is detected and evaluated. A method according to claim 8, characterized in that the lateral displacement of the color structure by use and control of one or more filters with variable refractive index in the beam path of the illumination beam is achieved. Method according to one of claims 1 to 4, characterized in that a phase shift method is used, in which an assignment of points or areas in the detected reflected color structure by determining its phase position is possible, wherein the required phase shift by an intensity modulation in different wavelengths or wavelength ranges is realized. Method according to one of claims 1 to 10, characterized in that the inclinations of the surface area based on a change in the reflected color structure with respect to a recording at a distance of L = 0 to the specular surface or a reference recording by reflection of the illumination beam on a perfectly flat object surface will be determined. Method according to one of claims 1 to 10, characterized in that the reflected color structure in the beam cross section of the reflected illumination beam is detected in at least two different image planes and the inclinations of the surface area based on a change in the reflected color structure between the at least two image planes and the mutual distance of the image planes be determined. A method according to claim 12, characterized in that the two different image planes are selected at intervals L ₁ and L ₂ to the specular surface, for which applies: | L ₁ - L ₂ | = | ΔL | > | L ₁ | and | L ₁ | ≥ | L ₂ |. A method according to claim 12, characterized in that the two different image planes are selected at intervals L ₁ and L ₂ to the specular surface, for which applies: L ₁ = -L ₂ . Method according to one of claims 1 to 14, characterized in that the topography of the surface area is calculated from the determined inclinations. Device for the spatially resolved measurement of inclinations of a surface area of a specular surface of an object with - a lighting unit with which an illuminating beam with a color structure can be generated across the beam cross section, - a collimator unit connected downstream of the illumination unit ( 9 . 10 . 11 . 21 ), which collimates the illumination beam generated by the illumination unit, whereby a collimated illumination beam is obtained as a measurement beam of the device, which can be directed onto a surface area to be measured, and - an optical detector arrangement ( 15 . 17a . 17b ), with which a reflected color structure in the beam cross-section of the measuring beam reflected at the surface area can be detected in at least one image plane. Apparatus according to claim 16, characterized in that between the illumination unit and the collimator unit ( 9 . 10 . 11 . 21 ) a pinhole as a spatial filter ( 7 ) is arranged for the illumination beam. Apparatus according to claim 16 or 17, characterized in that at least one beam splitter unit ( 16 ) is arranged in the beam path of the reflected measuring beam, which divides the reflected measuring beam in at least two partial beam paths, and the detector arrangement ( 15 . 17a . 17b ) several detectors ( 17a . 17b ), which are arranged in different partial beam paths of the reflected measuring beam. Device according to claim 18, characterized in that the detectors ( 17a . 17b ) are designed and arranged so that they capture different image planes. Device according to claim 18, characterized in that the detectors ( 17a . 17b ) are formed so that they respectively detect the different wavelengths or wavelength ranges of the reflected measuring beam. Device according to one of claims 16 to 20, characterized in that the illumination unit a plurality of light sources ( 1 ) of different wavelengths or wavelength ranges. Device according to one of claims 16 to 21, characterized in that the illumination unit comprises one or more devices for color separation or color structuring. Apparatus according to claim 22, characterized in that the means for color separation or color structuring color filter ( 19 ) are. Apparatus according to claim 22, characterized in that the means for color separation or color structuring optical elements ( 4 ) are for wavelength splitting. Device according to one of claims 16 to 24, characterized in that the illumination unit a plurality of electrically controllable light modulators ( 5a . 5b . 5c ) which are irradiated with light of different wavelengths or wavelength ranges and whose output beams are superimposed. Device according to one of claims 16 to 20, characterized in that the lighting unit by a projector ( 20 ) is formed for color projection. Device according to one of claims 16 to 26, characterized in that one or more filters with variable refractive index are arranged in the beam path of the illumination beam, with which a lateral displacement of the color structure on the surface region can be achieved.

Images