JP2007505322A - Line profile asymmetry measurement - Google Patents

Abstract

  Aspects of the invention provide methods and mechanisms for measuring symmetry / asymmetry of arrays of microelectronic features. One embodiment of the present invention provides a method for measuring asymmetry of a three-dimensional structure in a microelectronic device. With this method, light is directed to the array of microelectronic features of the microelectronic device. The light illuminates a portion of the array that includes the entire length and width of the plurality of microelectronic features. Light scattered back from the array is detected under conditions selected from the group consisting of one or more reflection angles, one or more wavelengths, or combinations thereof. The method also includes examining one or more characteristics of the backscattered light by performing operations including examining data from the reflection reflection angle.

Description

  The present invention relates to optical verification of microelectronic devices, and in particular to measurement of line profile asymmetry using scatterometry.

(Mutual citation of related applications)
This application is a continuation of US patent application Ser. No. 10/086339 (Publication No. 2002/0149782) entitled “Line Profile Asymmetry Measurement Using Scatterometry” filed on Feb. 28, 2002. U.S. Provisional Patent Application No. 60 / 502,444 entitled “Line Profile Asymmetry Measurement” filed on Sep. 12, 2003 and filed on Mar. 2, 2001. Claims priority based on US Provisional Patent Application No. 60 / 273,039 entitled "Process Quality Certification by Measuring Line Profile Asymmetry Using Scatterometry". The entirety of each of these applications is incorporated herein by reference.

  Note that the following discussion refers to a number of publications by author and year of publication, and because of the recent publication date, a particular publication should not be considered prior art with respect to the present invention. The commentary on such publications in this application is given for a more complete understanding and that such publications are prior art for the purpose of patentability determination and Should not be interpreted.

  The manufacture of microelectronic devices is a complex procedure that uses a variety of equipment for the various process steps involved. First, a lithographic process transfers the image being created to a photosensitive material known as a photoresist. This image in the photoresist then serves as a mask in the next patterning process known as etching. Etching is the process of transferring the resist image to a suitable material such as polycrystalline silicon. The etched material is then flooded with insulating material, planarized if necessary, and the entire process begins again.

  Throughout the process, the device being created should be essentially symmetric in every step, i.e. correctly manufactured transistor gates should have equal left and right sidewalls, as well as equal left and right corner rounding, etc. Will have other features like. If an error occurs during processing, this desired symmetry will be compromised and, as a result, the integrity or functionality of the device will also be compromised. If the asymmetry is fairly severe, the device will not work at all.

  The present invention relates to performing a symmetry / asymmetric measurement by scatterometry. Scattering is an optical verification technique that is well suited for measuring symmetry or asymmetry for microelectronic devices. By analyzing the light scattered from the array of microelectronic features, a line profile measurement can be made. In particular, scatter measurements, measured at +45 degrees and -45 degrees from a position perpendicular to the surface, that is perpendicular to the surface, are ideal for symmetric / asymmetric measurements because the reflectance characteristics of the line profile can change at these angles. However, the corner angle is not necessarily required to detect asymmetry. In order to increase the sensitivity of this effect, the array of features should be arranged in a specific orientation known throughout the specification and claims as a general conical configuration, i.e. the wave vector of the illumination beam One of them that does not remain parallel to the plane of symmetry.

  Prior art approaches typically use “conventional” scattering. These are measurements intended to measure surface roughness, scratches, pitching, and the like. However, the present invention is based on the physical properties of diffraction, and measurements in the present invention always occur for periodic features (such as line / space gratings).

  Previous work on scatterometry used techniques for measuring line profiles in resist and etched materials. C. J. et al. Raymond et al., “Characterization of resist and etched line profiles using scattering measurements”, Integrated Circuit Metrology, Inspection and Process Control XI, Proc. SPIE 3050 (1997). Embodiments of the present invention provide a technique for measuring asymmetric line profiles (eg, unequal sidewall angles).

  The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate one or more embodiments of the invention and, together with the description, serve to explain the principles of the invention. . The drawings are only for purposes of illustrating one or more preferred embodiments of the invention and are not to be construed as limiting the invention.

(Detailed explanation)
Overview Aspects of the present invention provide methods and mechanisms for measuring symmetry / asymmetry of arrays of microelectronic features. One embodiment of the present invention provides a method for measuring asymmetry of a three-dimensional structure in a microelectronic device. With this method, light is directed to the array of microelectronic features of the microelectronic device. The light illuminates a portion of the array that includes the entire length and width of the plurality of microelectronic features. Light scattered back from the array is detected under conditions selected from the group consisting of one or more reflection angles, one or more wavelengths, or combinations thereof. The method also includes examining one or more characteristics of the backscattered light by performing operations including examining data from the reflection reflection angle.

  A method for measuring line profile asymmetry in a microelectronic device according to another embodiment of the invention includes directing light at an angle of incidence relative to the array of microelectronic features of the microelectronic device. . The light scattered back from the array is detected at an additional angle with respect to the incident angle. One or more characteristics of the detected light are compared to an asymmetric model that includes a single feature profile having a top surface, a bottom surface, and a midline in cross section. The midline extends between the top surface and the bottom surface and is perpendicular to the bottom surface, and the cross section is asymmetric about the midline.

Method of Scattering Measurement A line profile can be measured by analyzing electromagnetic radiation scattered from an array of microelectronic features. In some preferred embodiments, the scatterometer measures, for example at +45 degrees and -45 degrees from a position perpendicular to the surface, that is, the reflection characteristics of the line profile change at these angles. This has been found to be particularly well-suited for symmetry / asymmetry measurements. To increase the sensitivity of this effect, it is preferable to arrange the array of features in a specific orientation known as a general conical configuration. Nonetheless, scatterometers according to other embodiments measure at a non-residual angle.

  The measurement of the scatterometry can be performed at any residual angle, +/− 45 degrees is an example, but a suitable pair of complementary angles is for example about +/− 0.00001 ° to about +/−. A useful embodiment is in the range of approximately 0 ° to approximately +/− 90 °, such as 80 °, ie a useful embodiment is a scatter measurement at an additional angle of approximately +/− 0.00001 ° to approximately +/− 47 °. (Reflectance cannot be measured at an additional angle with respect to an incident angle of 0 °, so 0.00001 ° is arbitrarily chosen as the nominal angle in this application, ie any other nominal angle is sufficient. Will). Scatterometry measurements can be performed at several angles or a series of angles. In addition, the measurement at each angle can include a single wavelength of radiation (such as a laser) or include radiation consisting of several wavelengths or a broad range of radiation (such as a white light source). Can do. It would be possible to measure only the luminance of the radiation, or to measure luminance and phase in a coordinated manner, similar to the ellipsometry measurement.

  The optimal electromagnetic radiation source will depend on the nature and size of the grating. Nevertheless, for the sake of clarity, in the following discussion, electromagnetic radiation is generally referred to as light. Regardless of the light source used or the method by which it was measured, assuming that the array is oriented in a general conical configuration, comparing data from the co-angles will determine if there is asymmetry. Can show instantly. If the light measurements are the same without any further analysis, the profile is symmetric. Conversely, if the light measurements are different, the profile is asymmetric. In general, the more sensitivity is used, the better the measurement sensitivity. This makes angular scatterometers (scatterometers that scan through angles) more suitable for measuring these profile asymmetries than spectral scatterometers (scatterometers that scan through wavelengths). In some embodiments, the scatterometer can scan through the entire range of angles and the entire range of wavelengths.

Application examples of the method for measuring the residual angle scattering of the present invention include, but are not limited to, the following.
-An array of wafer platforms to the optical system, such as an array by a lithography tool (stepper or scanner) or an array in a lithography process-An optical, such as an array by a lithography tool (stepper or scanner), or an array in a lithography process Wafer alignment to system-Identification of lens aberrations present in lithography tool or process-General diagnosis of lithography tool or process imaging performance-Measurement of temperature uniformity of drying process / station-Resist spin coaster, or Measuring thickness uniformity of spin processing-Measuring developer process / station uniformity-Etching tool or process characterization-Planarization tool or process characterization-Metallization tool or process Characterization-control of any of the above processes

  In the most general aspect, one purpose of semiconductor processing is to create an essentially symmetric device (eg, transistor gate). Indeed, it is rare to create a device that is intentionally not symmetric or asymmetric. For this purpose, the lithographic patterning process aims for symmetry, especially with respect to the footings at the bottom of the lines and equivalent sidewalls. Similarly, the etch process also encourages to produce symmetrical features, in this case mostly against the sidewalls of the line. In either control of these processing steps, the measurement technique must be able to detect asymmetry and preferably measure any asymmetry present (such as unequal left and right sidewalls).

  Scatter measurement is an optical metric based on the analysis of light scattered from a periodic array of features. In the strict physical sense, the light “scattered” from this periodic sample is actually due to diffraction, but from a general point of view, for purposes of explanation, we will scatter it. Call it. When a series of periodic features (known as diffraction gratings) are illuminated by a light source, the scattered / diffracted light reflection properties depend on the structure and configuration of the features themselves. Therefore, by analyzing the “signature” of scattering, the shape and dimensions of the diffraction grating can be identified.

Diffraction can actually produce a number of different “orders”, or light beams that are scattered from its features. In modern semiconductor production geometries, the period of features is small, so there is typically only one diffraction order. This order is known as the “specular” or “zero order” order and is the light beam most frequently used in scatterometry techniques. One of the more common methods of analyzing light scattering using specular reflection orders is to change the angle of incidence of an illumination light source (usually a laser). As shown in FIG. 1, the scattering “sign” is measured as the incident angle θ _i is varied and the detector moves in conjunction with the angle θ _n to measure the diffractive power of the specular order. It is this scattering signature known as the angular signature that contains information about the diffractive structure such as the thickness of the grating, and the width of the grating lines. When correctly measured, this angular signature can also contain information about any asymmetry present in the grid lines as well. By measuring through the co-inclination (both positive and negative with respect to the normal), a sign that is asymmetric can be obtained if the line is asymmetric. Conversely, if the line profile is actually symmetric, the measured sine will also be symmetric. However, for the purpose of comparison, if a suitable theoretical diffraction model is available, no additional angle is necessary and the “inverse” problem (see below) can be implemented.

  The method of scatter measurement is often described in two parts, typically known as “forward” and “inverse” problems. In the simplest sense, the forward problem is the measurement of scatter signatures, and the inverse problem is the analysis of the signatures to provide meaningful data. Over the years, many types of scatterometers have been validated, such as C.I. J. et al. Raymond et al., “Measuring sub-wavelength photoresist gratings using optimal scattering measurements”, Journal of Vacuum Science and Technology B 13 (4), pp. 1484-1495 (1995), S.A. Columbe et al., “Ellipsometric scattering measurements in measurements below 0.1 μm”, Integrated Circuit Metrology, Inspection and Process Control XII, Proc. SPIE 3332 (1999), Z.M. R. Hatab et al., “Characteristics of groove depth in 16 megabit dynamic random access memory using two-dimensional diffraction analysis”, Journal of Vacuum Science and Technology B 13 (2), pp. 174-182 (1995), and X. Ni et al., “Specular reflection spectroscopic scattering measurement in DUV lithography”, Proc SPIE 3677, pp. 159-168 (1999). Nonetheless, the most widely studied is the transformation of the corners or “2-θ” (due to the two θ variables shown in FIG. 1), where, as previously mentioned, a scattering signature is obtained. In order to change the incident angle. Preference is given to this type of scatterometer, but it is not essential for measuring the asymmetry of the line profile. Note that the scanning optical system of FIG. 1 allows this angular scatterometer to measure both positive and negative angles from normal incidence (0 degrees) up to approximately +/− 47 degrees. Should.

  Several different approaches have been explored to solve the inverse problem. C. J. et al. Raymod et al. (1995), supra, R.M. H. Krukar, Ph. D. Dissertation, University of New Mexico (1993), J. MoI. Bischoff et al., Proc SPIE 3332, pp. 526-537 (1998), and I.I. J. et al. Kallioniemi et al., Proc SPIE 3743, pp. 33-40 (1999). The most common method is model-based analysis because the optical response of the diffraction grating can be simulated exactly from Maxwell's equations. These approaches rely on comparing the measured scattering signature with the signature generated from the theoretical model. Both differential and integral models were searched. Because these diffraction models are computationally intensive, in general, standard regression techniques cannot currently be used without introducing errors due to the performance of the regression, but those with small or acceptable errors If so, a regression approach could be used. In general, however, the model is used a priori to generate a series of signatures corresponding to individual iterations of various grating parameters such as grating thickness and grid line width. When all parameters are repeated over a range of values, the resulting set of signatures is known as a signature library. When the scatter signature is measured, it is compared against the library to find the closest match. In order to identify the closest match, a standard Euclidean distance criterion is used that minimizes the mean square error (MSE) or the root mean square error (RMSE). The modeled signature parameter found to be closest to the measured signature is interpreted as the measured signature parameter. The scatterometer in some embodiments preferably includes analysis software based on error minimization.

  In previous studies, scatterometry was performed in C.I. J. et al. Raymond et al. (1995), supra, and C.I. Baum et al., “Measurement of resist line width and profile using scatterometry”, critical dimension (CD) measurement and photoresist sample profile as in SEMATECH AEC-APC Conference, Vail, Colorado (September 1999). For the characterization of Bushman et al., "Scatter measurements for process monitoring of polycrystalline silicon gate etch," Process, Equipment, and Materials Control in Integrated Circuit Manufacturing III, Proc. SPIE 3213 (1997), C.I. Baum et al., “Measurement of Scattering Measurements for Post-etch Polycrystalline Silicon Gate Metrics”, Integrated Circuit Metrology, Inspection and Process Control XIII, Proc. SPIE 3677, pp. 148-158 (1999), and C.I. Raymond et al., “Scatter measurement for measurement of metal features”, Integrated Circuit Metrology, Inspection and Process Control XIV, Proc. SPIE 3998, pp. Used to characterize the profile of etched materials such as polycrystalline silicon and metals as in 135-146 (2000). Since this technology has demonstrated high speed, non-destructive, and excellent accuracy, it is an attractive alternative to other metrics used in mainstream semiconductor manufacturing. In particular, the scatter measurement is asymmetric because it can quickly indicate whether there is some asymmetry on the grid line (without performing the inverse problem), as demonstrated. It is completely applicable to measurement.

  When considering whether to consider symmetry in the measured diffraction efficiency of the specular reflection (zero order) scattering signature, both the input and output fields are decomposed into S and P components for the input boundary of the lattice problem. It is convenient to do this (in this case the xy-plane). FIG. 2 shows the geometry of these components relative to the angular scan direction (showing scans from both positive and negative angular regions). Note that the entrance plane shown in this figure is the page itself, and no reference has been made regarding the orientation of the grating relative to this entrance plane. From the figure, it can be seen that there is a phase difference in the S-polarized component when the beam moves from one half of the angular region to the other half. This phase difference is one reason that an asymmetric angular sine can be created from an asymmetric line profile.

  The orientation of the grating relative to the entrance plane is another consideration in measuring the asymmetry of the sample. 3A and 3B show two orientations, known as a conical configuration and a conventional configuration, respectively. According to the first principle, scanning parallel to the grating vector (the so-called “normal” or “conventional” configuration shown in FIG. 3A) is the only case that never combines the S and P modes of the total electromagnetic field. (For example, M. Moharam et al., “Formulation for stable and efficient implementation of exact coupled wave analysis of binary lattices”, J. Opt. Soc. Amer. A, Vol. 12). , pp. 1068-1076 (May 1995) (see equation (48)). In the general cone scattering problem, if the incident illumination is in a completely P-polarized state, the connectivity in question demonstrates that both the S and P components can be observed in the (all) output field. Similarly, both the S and P components can be observed in the (all) output field even when the incident illumination is in a completely S-polarized state.

  The scattering problem is linear, so the superposition principle applies. When using mixed spectral states for the input wave, the input field can be decomposed into S and P components, and the problem can be solved separately, and then the resulting output field can be superimposed with complex amplitudes . The S component of the total output field consists of contributions from both the S and P parts of the input field due to the perfect connectivity of the problem. The same explanation can be applied to the P component of the entire output field. The superposition is done with complex amplitudes, so the field components in the S-polarization state arising from the S and P parts of the input field show interference effects. This means that the relative phase difference between the S and P components of the entire input field can be converted into an amplitude difference in the S and P components of the entire output field. At the beginning of this year, in any case where coupling is present, the asymmetry of the output diffraction efficiency is expected. It should also be noted that in a strict conical scan (the wave vector of the illumination beam remains parallel to the plane of symmetry of the structure), the symmetric structure does not produce any coupling. Therefore, in this case, symmetry in the measured diffraction efficiency is expected. Measured S and only for asymmetric structures or general cone scans where both the S and P components are present in the input beam (the wave vector of the illumination beam does not remain parallel to the symmetry plane of the structure) Allow for asymmetry in P diffraction efficiency.

  In order to introduce this concept of asymmetry, the grid line that produces the asymmetry scatter signature is considered a simple photoresist line profile as shown in FIGS. 4 (a)-(c). FIG. 4 (a) shows a perfectly symmetric profile in which both wall angles are equal to 90 degrees. In FIG. 4 (b), the angle of the right wall is changed to 80 degrees, while in FIG. 4 (c), the opposite case is shown (left is 80 degrees and right is returned to 90 degrees). Have been). FIG. 5 shows the angular scatter signature associated with each of these profiles and measured through the co-inclination. As can be seen, the symmetric profile provides a symmetric scattering signature for both polarizations. However, the asymmetric profile shows considerable asymmetry for both polarizations. In practice, as a result of the asymmetry of the profile, the sine appears distorted or “tilted”. Furthermore, comparison of sine data at 80/90 degrees and 90/80 degrees shows an interesting result—inversion of the sidewall angle results in sine inversion. Physically, this reversal is the same as rotating the wafer 180 degrees, thereby swapping the positive and negative areas of the scan, so the results are consistent. These figures also show the advantages of angular scatter measurements to identify the presence of asymmetry, since a simple visual inspection of the signature can prove that the profile is asymmetric.

(Comparison of asymmetric models)
In other embodiments, asymmetries can be identified by performing a solution to the inverse problem, such as performing model comparisons either by regression or by using library comparisons, for example. Let's go. This may be advantageous if, for example, only “one-sided” (positive or negative) angles are present, or if the system is a spectral scatterometer operating at a fixed angle.

  Figures 19-24 show some structures where model comparison may be useful. Each of these figures is a cross section of a feature, which will be referred to as a feature profile. In some embodiments, the feature may be a line of a diffraction grating, whose cross section will be substantially perpendicular to the longitudinal axis of the line (not shown). Some of the illustrated feature profiles, eg, FIGS. 19 and 20, are single line profiles. Other embodiments, such as FIGS. 21 and 22, are stacked or multilayer diffractive structures that may comprise two or more features. For example, FIG. 21 can be thought of as a feature profile with a first line profile G superimposed on a second line profile H, where in FIG. 21 an asymmetric single line profile I is instead symmetric. Overlying the single line profile H.

  Each of the model feature profiles of FIGS. 19-24 is asymmetric. Turning first to FIG. 19, the feature contour 100 includes a bottom surface 102, a top surface 104, and left and right side walls 106 and 108, respectively. An ideal symmetry would have a top surface 104 parallel to the bottom surface 102 that intersects the parallel sidewalls 106 and 108 at right angles. In FIG. 19, the left side wall 106 is vertical, but the right side wall 108 is tilted. Thus, the feature profile 100 is asymmetric about a midline Z that extends between the top surface 102 and the bottom surface 104 and is perpendicular to the bottom surface. In FIG. 19, the midline Z is located equidistant from the leftmost point of the feature (side wall 106) and the rightmost point of the feature (where the side wall 108 joins the bottom surface 102). There is no midline perpendicular to the bottom surface 102 in which the feature profile 100 is symmetric. The single line profile 110 of FIG. 21 also includes a bottom surface 112, a top surface 114 parallel to the bottom surface, and two sidewalls 116 and 118. Neither side wall 116 nor 118 is vertical, but the left side wall 116 is inclined at one angle with respect to the normal and the right side wall 118 is inclined at another angle with respect to the normal. Therefore, the feature profile 110 is asymmetric about the midline Z.

  Multi-layer features will include, in one layer, a symmetric feature profile such as line H in FIGS. 21 and 22, and an asymmetric feature profile such as line G in FIG. 21 and line I in FIG. . The lower line profile J in FIG. 23 is not a perfect rectangle but is symmetric-a midline (not shown) perpendicular to the center of the bottom of the line results in two symmetric halves I will. Regardless of the symmetry of the feature profile of one layer of the structure in FIGS. 21-23, the entire feature profile is asymmetric. In other embodiments, the model feature profile may have two or more asymmetric single layer feature profiles that are asymmetric. For example, FIG. 24 shows a bilayer feature where both the upper feature profile I and the lower feature profile K are asymmetric about a vertical midline (not shown).

  Many of the model feature profiles of FIGS. 19-24 include at least three different corners. As a result, the left two corners of the profile can be right angles, but the corners A and B are different from each other and neither is a right angle. Some model diagrams, such as line K of the multilayer diffractive structure of FIGS. 20 and 24, have four different corners in the cross section of the feature. In FIG. 20, each of the corners included in the feature profile, ie, corners C, D, E, and F, is different from the others. In some superimposed or multilayer diffractive structures, at least one of the superimposed features, and optionally two or more, have at least three different angles in the cross section of the line.

  In one embodiment of the present invention, a theoretical library of single or multi-layer diffractive structures and corresponding simulated or theoretical diffractive signals such as diffractive signatures is generated to produce single or multi-layer theoretical diffractive structures. The theoretical diffraction signature based on is compared with the measured diffraction signature. This can be done in any number of different ways. In one approach, an actual library of theoretical output signals is generated based on the parameters assigned to the variables. This library can be generated prior to the actual measurement of the diffraction signature, or it can be generated in the process of matching the measured diffraction signature with the theoretical diffraction signature. Thus, as used in this application, the theoretical library is a library generated independently of the measured diffraction signature, and the theoretical "best estimate" and results of the measured undercut multilayer geometry A library generated based on the calculation of the theoretical diffraction signature that results in the identification of the best match by iterative comparison with the structure with varying parameters. This library can optionally be pruned by removing signals that can be accurately represented by interpolation from other signals in the reference set. Similarly, a library index can be generated by associating each signature with one or more indexing functions and ordering the indexes based on the magnitude of their correlation. Methods for the construction or generation of this type of library and their optimization are well known in the art.

  In one approach, a strict theoretical model based on Maxwell's equations is used to calculate a prediction of the optical signal characteristics of a diffractive structure, such as a diffraction signature, as a function of the diffractive structure parameters. In this process, a set of trial values of the diffractive structure parameters can be selected and a computer-representable model of the diffractive structure, including its optical material and geometry, can be built based on these values. Numerically simulate the electromagnetic interaction between the diffractive structure and the illumination radiation to calculate the prediction of the diffraction signature. Any of a variety of adaptive optimization algorithms can be used to adjust the parameter values of the diffractive structure, and the process is repeated iteratively to determine the difference between the measured and predicted diffractive structures. , So that the best match can be obtained. US Published Patent Application US 2002/0046008 discloses one database method for structure identification, while US Published Patent Application US 2002/0038196 discloses another method. . Similarly, US published patent application US 2002/0135783 also discloses various theoretical library techniques, as disclosed in US published patent application US 2002/0038196.

  Generation of libraries from model patterns is disclosed in numerous references, particularly US Patent Application Publication Nos. 2002/0035455, 2002/0112966, 2002/0131040, 2002/0131055, and 2002/00165636. As is well known in the art. Early references on these methods can be found in R.A. H. Krukar, S .; S. H. Naqvi, J. et al. R. McNeil, J. et al. E. Franke, T .; M.M. Mimczyk, and D.M. R. Hush, “A Novel Diffraction Technique for Metrics of Etched Silicon Grating”, OSA Annual Meeting Technical Digest, 1992 (Optical Society of America, Washington, DC, 1992), Vol. 23, p. 204, and R.A. H. Krukar, S .; M.M. Gaspar, and J.A. R. McNeil, “Wafer Testing Using Scattered Light, and Critical Dimension Estimation”, Machine Vision Applications in Character Recognition and Industrial Inspection, Donald P. D'Amato, Wolf-Ekehard Blanz, Byron E .; Dom, Sugar N .; Srihari, Proc SPIE, 1661, pp 323-332 (1992).

  Other approaches to matching, including real-time regression analysis, can be used as well. These methods are known in the art and are used to “best fit” such as diffraction signatures based on model permutations such as permutations in single line or multi-layer diffractive structures. A theoretical diffraction signal can be identified. In a technique generally described as iterative regression, one or more simulated diffraction signatures are compared with the measured diffraction signature, thereby creating an error signal difference, and then another Calculate the simulated diffraction signature and compare it with the measured diffraction signature. This process is repeated or repeated, ie, regressed, until the error is reduced to a specific value. One method of iterative regression is non-linear regression, which can optionally be performed in “real time” or “on the fly” mode. Various iterative regression algorithms well known to those skilled in the art can be applied to the interpretation of the measured diffraction signature by comparison with a simulated diffraction signature based on the profile of the model structure.

  In addition to the parameters associated with single or multi-layer patterns as disclosed in this application, other diffractive structure parameters that can be utilized in the theoretical library are the grating period, the structure's parameters, including the parameters of its various layers. Material parameters of the substrate on which the structure is placed, such as material parameters, film thickness, and the refractive index of the film under the structure, and weights due to the relative contribution of the CD, structure and substrate at a particular location It includes any parameter that can be modeled, including elements such as various weights or averages such as.

  In yet another embodiment, short-period structures can be modeled, and the results can be utilized, for example, by regression or model comparison. As used in this application, the term “short-period structure” is short enough that two or more full lengths and full widths of the structure can be included in the area illuminated by the light source of the scatterometer used. Includes a three-dimensional structure having a length. For example, if the area illuminated by the intended scatterometer is about 40 μm wide, the short-period structure will have a longitudinal length of less than 40 μm (eg, in a direction perpendicular to the k vector of the grating). , And a sufficiently short lateral spacing to include at least two features (eg, a distance along the k-vector of the line grid between adjacent lines of the line grid). The length of the short-periodic structure is such that the length of the short-periodic structure is such that at least two, preferably three or more of the short-periodic structures are spaced apart from each other and still fit in the longitudinal direction within the irradiation area. Desirably less than half of The length of the short period structure is preferably short enough to be a relevant parameter for incident illumination.

  In one example of an embodiment, each line of the line grid model can be defined as a series of short lines arranged in a vertical direction, rather than a single long line. For example, each of the short lines can be about 5-20 μm long, about 0.2-1 μm wide and about 0.5-2 μm wide. The period between the parallel spectral lines can be about 0.5-2 μm. If the scatterometer incident radiation extends into a circular region having a diameter of about 40 μm, many full lengths and full widths of the short lines of this structure will be included in the irradiation.

  Figures 25-27 schematically illustrate short period structures according to selected embodiments of the present invention. In FIG. 25, each of the short period structures comprises an array of features, each of which includes a first post or hole that overlies another post or hole. The axis of the first post or hole (not shown) in each of these features is offset from the axis of the second post or hole, resulting in a bilayer feature including a “staircase” shape. In a preferred embodiment, the post is elliptical in shape and is preferably a long elongated ellipse, thereby providing the greatest solution for the corner angle analysis. The array will be arranged as a series of lines of features that are periodic in the X direction but need not be periodic in the Y direction. In one useful model, the array is a regular array of posts or holes that are periodic in both the X and Y directions. In FIG. 26, a first series of linear features is placed on top of a second series of linear features and distorted in the x and y orientation relative to the second series of linear features, so that the structure Are offset or include “staircase” features as in FIG. FIG. 27 also includes a first series of linear features that are placed over the second series of linear features and distorted in the x and y orientation with respect to the second series of linear features. Unlike FIGS. 25 and 26, the cross-section of at least one straight feature is asymmetric and provides at least three different interior angles. In the particular implementation shown in FIG. 27, similar to lines I and J in FIG. 23, both the first and second features are asymmetric. In one useful embodiment, in the lateral dimension (relative to the device plane), the dimensions of the structure are different and preferably quite different. For example, in FIG. 26, the short lines are considerably longer (in the X direction of FIG. 25) than their width spread (in the Y direction of FIG. 25). While FIGS. 25-27 illustrate a simple circular or rectangular structure, the method according to other embodiments of the present invention can use any three-dimensional structure, but it can be a repetitive or repetitive structure. Is preferred.

  Some implementations of the present invention use a theoretical model based on the three-dimensional structure of an array of short period structures. Due to the large number of variables in such a structure, it is complex to calculate a 3D model, but it is still necessary to generate a model and use the scatterometry technique described above for an actual 3D structure. This model can be used for comparison and analysis purposes with acquired data. Also, such 3D models can and are intended to utilize a variety of algorithms and methods designed to simplify model calculations.

  In yet another embodiment, the asymmetry in the array of three-dimensional short-period features is measured by measuring the co-angle (both positive and negative with respect to the normal), preferably the co-angle θ (also with respect to the vertical). Can be determined by measuring over the entire range (both positive and negative). If the three-dimensional structure is asymmetric, a sign that is asymmetric can be obtained. Conversely, if the three-dimensional structure is actually symmetric, the measured signature will also be symmetric.

  The usefulness of the present invention in identifying asymmetry in a three-dimensional structure, for example, by comparison of the residual angles over a range, is shown schematically in FIGS. FIG. 28 is a graph of angular scatter measurement signatures (symmetric across the full angle) of a first series of rectangular three-dimensional linear structures placed on a second series of rectangular structure structures as shown in FIG. It is. In FIG. 28, for a single feature overlaid, the solid line shows no offset, the dashed line shows the 25 nm offset of these single features, and the dotted line shows the 50 nm offset. . S-polarization measurements and P-polarization measurements are symmetric about an angle of 0 degrees where there is no offset (solid line). With a 25 nm offset (dashed line), each profile (such as the S data profile or P data profile) is “distorted” by about 0 degrees so that each plot of S and P data is asymmetric. As the asymmetry of the three-dimensional structure increases, the asymmetry in the resulting plot increases accordingly, so the asymmetry is greater at the 50 nm offset (dotted line) than at the 25 nm offset (solid line). FIG. 29 is a graph of an angular scatter measurement sine of an elliptical “post-on-post” three-dimensional structure similar to that of FIG. 25, where the first series of elliptical posts is Place it on a second series of similarly shaped posts. The solid line in FIG. 29 indicates that there is no offset for the first and second series of posts, the broken line indicates the 25 nm offset, and the dotted line indicates the 50 nm offset. As shown in FIG. 28, the degree of asymmetry in the S data and in the P data correlates with the degree of asymmetry of the three-dimensional structure.

  Therefore, in accordance with an embodiment of the present invention, taking a measurement of the scattering measurement of the short period structure at the retraction angle requires a relatively very small computing power. Modeling a three-dimensional structure is a rigorous calculation and not all cone models with the simplest structure can be easily obtained in a reasonable time by current computing devices and programs. However, embodiments of the present invention that measure at a reciprocal angle can identify asymmetry by examining the symmetry of the collected data, which is simpler or easier from a computational point of view. .

  Therefore, the scattering measurement can be applied particularly to the structure on the three-dimensional structure, and the measurement of the scattering measurement of the diffraction order of the zeroth order or the specular reflection is easily affected by the displacement of the arrangement in the continuous three-dimensional structure layer. This deviation (also called offset) in the three-dimensional structural layer results in an asymmetric line profile that can be measured using a scatterometer in the appropriate measurement orientation. As can be seen, the sign changes when an offset is introduced, which is a positive sign for typical measurement sensitivity. In one embodiment, the measurement orientation is identified empirically based on the inherent nature of the most important three-dimensional measurement (eg, whether the most important measurement is in the x, y, or z direction). Can do. Therefore, in addition to changing the angle θ over a certain range (and measuring the corresponding covariance in each case), the angle Φ (rotation angle) is changed and the optimum angle Φ in the three-dimensional measurement performed is performed. It is also possible to specify.

  The following example identifies a process in which a scatterometry approach according to aspects of the present invention has good sensitivity in measuring feature asymmetry and therefore symmetric results are desirable, such as lithography and etch processing Shows that can be used for. Comparison with other measurement techniques such as AFM and cross-sectional SEM shows good consistency.

Industrial Applicability The present invention is further illustrated by the following non-limiting examples.

  To evaluate the feasibility of performing asymmetric profile measurements using scatterometry, three different sample formats were investigated (Examples 1-3). The first set of samples consisted of three wafers of photoresist lines on a metal substrate. The second sample was a single wafer of etched polycrystalline Si. The third set was also a single wafer of lattice lines printed on 193 nm photoresist. In each sample set, raw scatter signatures were obtained by performing measurements across the positive and negative angles in a conical scan orientation. For each sample set, an appropriate scatterometry library is generated, including independent left and right deformations in the sidewalls, as well as other parameters such as CD and thickness.

  Example 4 illustrates the use of the present invention to measure the arrangement of two consecutive layers on a semiconductor wafer.

Example 1 Photoresist Line on Metal Substrate The line width of this sample set was nominally 250 nm wide. The top-down stack configuration consisted of a patterned photoresist on the ARC on a TiN layer followed by a thick AlCu layer (which effectively serves as a substrate).

  The raw signature from this sample set showed considerable asymmetry. FIG. 6 shows one signature from this data set with the positive and negative halves of the angular scan superimposed on top of each other (“symmetric”). Obviously, the two halves are not the same, as the figure shows. In practice, they differ in terms of reflectivity by more than 5% at some angles, and at some angles the sign structures are different as well. This is a manifestation of profile asymmetry since measurements were made in a conical lattice orientation.

  The raw signature from this data set matched well with the model. The sidewall angle results from these wafer measurements can be seen in FIG. Recall that the library anticipated independent fluctuations in the left and right sidewall angles. In addition to the scatterometry measurements, AFM data from the same location is visible on this plot. Both AFM and scatterometer results show that there is indeed a difference in sidewall angles and that it is in the range of 1-2 degrees. Both of the data similarly recognize that the left sidewall angle is steeper than the right sidewall angle. The scatterometry data shows that the left and right angles move in unison, i.e., the entire width of the line does not change, but rather "wobbles" 1-2 degrees around the entire wafer. This effect could be due to the correlation between the left wall angle parameter and the right wall angle parameter, but the examination of the modeled sine data would leave one sidewall angle fixed. When the other side wall angle can be changed, the left and right side wall angle parameters are clarified to be completely different.

Example 2 Etched Polycrystalline Silicon Line The line width in this sample set was in the range of 150-300 nm. The stack is composed of patterned (etched) polycrystalline Si on oxide on a Si substrate. The raw signature from this sample set showed a slight asymmetry when measured in a conical configuration. FIG. 8 illustrates one sine such that both the positive and negative halves of the sine are “symmetric” so as to exhibit this asymmetry.

  To obtain a comparison, cross-sections of the wafers used for these scatterometry measurements were made and measured by SEM to identify the line sidewall angles. FIG. 9 shows the result of comparing the measured values of the left and right sidewall angles of the two techniques. As the figure shows, both tools convey some degree of side wall asymmetry, and generally the left wall angle is smaller. Further, the correlation of the sidewall angle between the two methods is good, and the same inclination is shown here and there.

Example 3-193 nm Photoresist Line The last sample set examined was a single wafer of BARC layer, polyester layer, oxide layer, and 193 nm photoresist line printed on a silicon substrate. The nominal size of the features on this wafer was 180 nm line.

  In this wafer, the signature data was only loosely asymmetric when measured in the conical mode. FIG. 10 shows the S and P polarizations in one of these signs that are “mirrored” back to itself. This asymmetry was much weaker than the sine asymmetry observed from the previous sample.

  The left and right sidewall angle data in one row from this wafer can be seen in FIG. Measurements made at the same location by the AFM are included on the plot of this figure. Both measurement techniques are in good agreement for the overall magnitude of the wall angle. AFM data shows more asymmetric measurements, but is generally consistent with the data from the scatterometer. From this same row, a comparison of the CD measurements obtained by the scatterometer and AFM can be seen in FIG. As the figure shows, the agreement between the AFM and scatterometer measurements is very good. The average difference between these two approaches is 2.43 nm.

Example 4-Continuous layer alignment measurement The alignment of two consecutive layers on a semiconductor wafer is important in the final performance of the manufactured device. This arrangement (also called superposition) is so important that there are dedicated tools to perform this one task. These tools are based on measuring images of special array labels printed on each layer. The semiconductor industry is moving toward increasingly smaller dimensions, however, a large amount of uncertainty surrounds the ability of these tools to provide the required measurement solution.

  Scattering measurement is a technique well suited for overlay measurement. By using the lattice structure on the lattice, the measurement value of the scattering measurement of the diffraction order of the 0th order or the specular reflection is easily affected by the deviation of the arrangement in the continuous lattice layer. This movement in the lattice layer (also referred to as offset) results in an asymmetric line profile, which is in the proper measurement orientation, and preferably (but not essential), both the positive and negative angles. ) Can be measured using a scatter meter capable of measuring.

  FIG. 13 represents an image of a grid-on-grid profile that can be used for overlay measurements. An error in aligning two successive layers results in a shift between the grid line and the asymmetric line profile. FIG. 14 shows the results demonstrating sensitivity to offset or overlay errors in measurements performed in a conventional orientation (see FIG. 3 (a)). When an offset is introduced, the sign changes, which is a positive sign in general measurement sensitivity. However, as shown in FIG. 15, in a conventional scan, the signature that occurs when a +/- offset of the same magnitude is brought in is not unique. Thus, conventional scanning is less desirable than cone scanning.

  Repeating the practice with conical scanning results in the signature shown in FIGS. For conventional scanning, FIG. 16 shows that the sine changes with the offset, while FIG. 17 shows that the change is now unique between the (left / right) offsets. Also note the symmetry about 0 degrees.

  Thus, studies have shown that both the specular (zero order) order conventional scan and the cone scan are both sensitive to offsets, but only the cone scan provides a unique signature for left / right deviation. . Existing scatterometry methods for evaluating overlay include the use of first and higher orders, and therefore require special measurement hardware to measure higher orders. See, for example, Sohail Naqvi et al., “Diffraction technique for monitoring and controlling lithographic processes”, JVSTB 12 (6) (November 1994) (Moire interference technique using higher orders), Bischoff et al., “Optical diffraction based overlay measurement”, Proc. SPIE Vol. 4344, pp. See 222-233 (2001) (first order measurement of the lattice in the lattice).

  By replacing the reactants described generally or individually and / or manipulating the conditions of the invention towards the conditions used in the examples, the examples can be repeated successfully as well.

  Although the invention has been described in detail with particular reference to these preferred embodiments, other embodiments can achieve the same results. Variations and modifications of the present invention will be apparent to those skilled in the art, and the appended claims are intended to cover all such modifications and equivalents. The entire disclosure of all cited references, applications, patents and publications referred to above are hereby incorporated by reference into this application.

It is a block diagram of the angle scatterometer used in the embodiment of the present invention. Fig. 4 shows the measurement geometry of the angular scatterometry used in an embodiment of the present invention. The measurement orientation of so-called conventional scattering measurement is shown. The measurement orientation of so-called cone scattering measurement is shown. A symmetrical resist profile is shown. An asymmetric resist profile is shown. An asymmetric resist profile is shown. 5 is a graph of angle sine data corresponding to the profiles of FIGS. FIG. 6 is a graph of (symmetric) angular scatter measurement signatures from a metal resist wafer. FIG. It is a graph of the side wall angle which arises from the wafer 5 of the sample set of a resist on metal. Figure 2 is a graph of (symmetric) angular scatter measurement signatures from an etched polycrystalline silicon wafer. 4 is a comparison of the left wall angle between a scattering measurement and a cross-section SEM in an etched polycrystalline silicon wafer. 4 is a comparison of the right side wall angle between a scattering measurement and a cross-sectional SEM in an etched polycrystalline silicon wafer. FIG. 4 is a graph of (symmetric) angular scatter measurement signatures from a 193 nm resist wafer. FIG. Comparison of left wall angle between scattering measurement and cross-sectional SEM for a 193 nm resist wafer. Comparison of right wall angle between scatterometry and cross-section SEM for a 193 nm resist wafer. It is a comparison of CD measurement of AFM and scattering measurement in a 193 nm resist wafer. FIG. 5 shows an image of a lattice-on-grid profile that can be used to measure overlay misalignment. 14 is a graph of angular scatter measurement signatures in the profile of FIG. 13 using a conventional (non-conical) scan. FIG. 6 is a graph of (non-unique) angular scatter measurement signatures at left and right offsets using conventional scanning. FIG. 15 is a graph of angular scatter measurement signatures in the profile of FIG. 14 using a cone scan. FIG. 6 is a graph of (unique) angular scatter measurement signatures at left and right offsets using a cone scan. Fig. 3 shows an asymmetric single line model used in the prior art, where the acute angles are equal to each other and the obtuse angles are also equal to each other, so that the cross section of each line provides only two different angles. Fig. 4 illustrates an asymmetric single line model of an embodiment of the present invention, where two angles are right angles, angle A is obtuse and angle B is acute, so the cross section provides three different angles. . Fig. 4 shows an asymmetric single line model of an embodiment of the present invention, where all four interior angles are different, angles C and F are acute angles, and angles E and D are obtuse angles. Fig. 3 shows a line superposition asymmetric model of an embodiment of the present invention, where line H is rectangular and line G is a parallelogram that is offset on both sides with respect to the sidewall arrangement and is not rectangular. FIG. 4 shows a line superposition asymmetric model of an embodiment of the present invention, where line H is rectangular and line I is offset on one side with respect to the sidewall arrangement, but not on the other side. Furthermore, the cross section of line I provides three different corners. FIG. 4 illustrates a line superposition asymmetric model of an embodiment of the present invention, where line I is offset on one side with respect to line J with respect to the arrangement of sidewalls, but not on the other side; Further, each cross section of lines I and J provides three different corners. Fig. 3 shows a line superposition asymmetric model of an embodiment of the present invention, where line I is offset on both sides with respect to line J with respect to the arrangement of sidewalls, where the cross section of line I has three different interior angles And the cross section of line K provides four different interior angles. FIG. 4 is an illustration that can be used as a model in an embodiment of the present invention, where a first series of three-dimensional posts are placed on and into a second series of posts. In contrast, the post is distorted in the x and y orientations, so that the post is offset or includes stepped features. FIG. 4 is an illustration that can be used as a model in an embodiment of the present invention, where a first series of three-dimensional linear structures are placed on a second series of linear structures and the second series of linear structures; Distorted in the x and y orientation with respect to the linear structure, so that the structure is offset or includes stepped features. FIG. 4 is an illustration that can be used as a model in an embodiment of the present invention, where a first series of three-dimensional linear structures are placed on a second series of linear structures and the second series of And at least one cross-section of the linear structure is not rectangular and provides at least three different interior angles. FIG. 25 is a graph of angular scatter measurement signatures (symmetric over the entire range of co-angles) of a first series of rectangular three-dimensional linear structures placed on a second series of rectangular structure structures, such as FIG. Yes, with respect to the first and second series of structures, the solid line indicates the case where there is no offset, the broken line indicates the offset of 25 nm, and the dotted line indicates the offset of 50 nm. A first series of three-dimensional elliptical posts are placed over a second series of similarly shaped posts (as shown in FIG. 26) of the post structure on an elliptical post (over the entire range of co-angles). (Symmetric) graph of angular scatter measurement signature, where for the first and second series of structures, the solid line indicates no offset, the dashed line indicates 25 nm offset, and the dotted line indicates An offset of 50 nm is shown.

Claims (33)

Directing light onto a array of microelectronic features of a microelectronic device to irradiate a portion of the array including a total length and width of a plurality of the microelectronic features;
Detecting light scattered back from the array at a condition selected from the group consisting of one or more reflection angles, one or more wavelengths, or combinations thereof; and
Asymmetry of the three-dimensional structure in the microelectronic device comprising the step of examining one or more properties of the return scattered light by performing an operation comprising examining data from the reflection reflection angle How to measure sex. The method of claim 1, wherein the directing step includes directing light at a substantially single wavelength. The method of claim 1, wherein the directing step includes directing light at a plurality of wavelengths. The method of claim 1, wherein the examining step includes comparing light intensities. The method of claim 1, wherein the examining further comprises comparing phases. The method of claim 1, wherein the examining step further comprises comparing a ratio of light amplitude to light phase. The method of claim 1, 2, or 3, wherein the step of directing comprises directing light to the array of microelectronic features in a general conical configuration. The method according to claim 1, 2, or 3, wherein the step of directing and the step of detecting are performed by an angular scatterometer. The method according to claim 1, 2, or 3, wherein the step of directing and the step of detecting are performed by a spectral scatterometer. The method according to claim 1, 2, or 3, wherein the examining step includes decomposing the return scattered light into S and P components with reference to the incident surface. The method according to claim 1, 2, or 3, wherein the detecting step includes detecting specular reflection order diffracted light. Detecting the asymmetry selected from the group consisting of asymmetry within a single layer of the microelectronic device and asymmetry within multiple layers of the microelectronic device using the results of the examining step;
The method of claim 1, 2, or 3, further comprising: Controlling the manufacturing process if the result of the comparing step indicates asymmetry in the array;
The method of claim 12, further comprising: Directing light to an array of microelectronic features of a microelectronic device at an angle of incidence relative to the array;
Detecting the light scattered back from the array at an additional angle with respect to the incident angle; and
An asymmetric model comprising one or more characteristics of the detected light in a cross-section comprising a single feature profile having a top surface, a bottom surface, and a midline extending between the top surface and the bottom surface and perpendicular to the bottom surface And measuring the line profile asymmetry in the microelectronic device, wherein the cross-section is asymmetric about the midline. The method of claim 14, wherein the cross section of the single feature profile includes at least three different interior angles. The method of claim 14, wherein the cross-section of the single feature profile includes four different interior angles. The single feature profile is a first line, and the model includes a second line feature profile overlying the first line;
The method of claim 14, wherein at least one of the first line and the second line includes at least three different interior angles on a cross section of the line. 18. The method of claim 17, wherein at least one of the first line and the second line has four different interior angles on a cross section of the line. The first line and the second line both include at least three different interior angles on the cross-section of each of the first line and the second line. 18. The method according to 17. The method of claim 17, wherein the first line has at least one sidewall offset from a second sidewall of the second line. The first line has a first side wall aligned with the first side wall of the second line and a second side wall offset from the second side wall of the second wall. 21. The method of claim 20, wherein: The method of claim 14, wherein the directing step includes directing light at a substantially single wavelength. The method of claim 14, wherein the directing step includes directing light at a plurality of wavelengths. The method of claim 14, wherein the comparing step includes comparing light intensities. The method of claim 14, wherein the comparing step further comprises comparing phases. 15. The method of claim 14, wherein the comparing step further comprises comparing a ratio of light amplitude to light phase. 15. The method of claim 14, wherein the step of directing includes directing light to the array of microelectronic features in a general conical configuration. The method of claim 14, wherein the directing and detecting are performed by an angular scatterometer. The method of claim 14, wherein the directing and detecting steps are performed by a spectral scatterometer. The method according to claim 14, wherein the comparing includes decomposing the return scattered light into S and P components with respect to an incident surface. 15. The method of claim 14, wherein the detecting step comprises detecting specular reflection order diffracted light. Using the result of the comparing step to detect asymmetry selected from the group consisting of asymmetry in a single layer of the microelectronic device and asymmetry in multiple layers of the microelectronic device. The method according to claim 14, wherein: 33. The method of claim 32, further comprising controlling a manufacturing process if the result of the comparing step indicates asymmetry in the array.

Images