Method for adapting depth structure model data to coherence tomography measurement data and coherence tomography method

The method adapts deep structure model data to coherence tomography measurement data by optimizing reflectivity coefficients, enabling detailed characterization of layer structures beyond standard OCT resolution limits, achieving non-destructive, high-resolution imaging and material parameter determination.

WO2024231427A1PCT designated stage expired Publication Date: 2024-11-14FRIEDRICH SCHILLER UNIV JENA
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
PCT/EP2024/062672
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-08
Filing Date
2024-05-08
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

Optical coherence tomography (OCT) is limited in its depth resolution by the bandwidth of the incident light, making it difficult to resolve structures below this resolution and accurately characterize layer transitions in samples.

Method used

A computer-implemented method that adapts deep structure model data to coherence tomography measurement data by deriving layer thicknesses and transitions, extracting spectrally resolved reflectivity coefficients, and recursively optimizing model reflectivity coefficients to match measured data, allowing for detailed characterization of layer structures beyond the standard resolution limits.

Benefits of technology

Enables non-destructive, high-resolution imaging and characterization of layer structures with nanometer precision, improving upon traditional methods by providing detailed material parameters and reducing noise in measurements, effectively surpassing the limitations of standard OCT techniques.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure EP2024062672_14112024_PF_FP_ABST
    Figure EP2024062672_14112024_PF_FP_ABST
Patent Text Reader

Abstract

The invention relates to a method for adapting depth structure model data (M) to coherence tomography measurement data of a sample (P) having a plurality of layers (Si) composed of layer materials in a layer construction in a depth direction (z). According to the method, a plurality of layers (S1, S2) having layer thicknesses (D1, D2) and a plurality of layer transitions (s1, s2) upstream of the respective layers (S1, S2) in the depth direction (z) are derived from the depth structure (r(z)). Furthermore, model reflectivity coefficients (r1_mod(ω), r2_mod(ω)) are adapted to the respective layer-transition-specific, spectrally resolved reflectivity coefficients (r1_exp(ω), r2_exp(ω)) by way of a value that optimizes the adaptation being calculated in each case for the at least one parameter (σ1, d1; σ2, d2) of one of the model layer transitions (ms1, ms2). On the basis of the value calculated for the at least one parameter (σ1, d1; σ2, d2), depth structure model data (M) can be obtained, the resolution of which in the depth direction (z) goes beyond the resolution predefined by a CT measurement on which the spectral complex field reflectivity (r(ω)) is based.
Need to check novelty before this filing date? Find Prior Art

Description

Method for fitting deep structure model data to coherence tomography measurement data and coherence tomography methods [1] The present invention relates to a computer-implemented method for adapting deep structure model data to coherence tomography measurement data of a sample and to a method for examining a sample by means of optical coherence tomography. [2] Optical coherence tomography (OCT) has established itself as an optical microscopy technique for non-destructive three-dimensional imaging. OCT is based on the backscattering of incident light at an interface of a sample to be examined. The interface is also referred to as the layer transition. Information about the sample to be examined can be obtained, for example, by evaluating the interference of the light (back)scattered at different interfaces. In addition to spatial superposition, a low temporal coherence of the superimposed light is required. See "Optical coherence tomography - principles and applications" AF. Fercher et al., Rep. Prog. Phys. 66 239 (2003) for a basic introduction to OCT. [3] The spectral range for OCT examinations was determined within the framework of the so-called XCT (XUV optical coherence tomography) using extreme ultraviolet (XUV) sources (e.g., synchrotron-based or laser-based) to the spectral range from 30 eV to 530 eV. See, for example, US 7,656,538 B2 length coherence tomography") for laser-based XCT. Due to the large usable frequency bandwidths in the XUV, XCT enables axial resolution in the range of, for example, a few tens of nanometers. This opens up new applications in the non-destructive investigation of, for example, (multilayer) coatings such as optical or XUV mirrors, functional axial structures in solar cells or axially structured semiconductors (graphene-based electronics), as well as biological membrane layer structures. In particular, XCT enables three-dimensional imaging of near-surface structures of thick samples. For example, in the investigation of structured semiconductors, XCT represents an interesting alternative to scanning electron microscopy (SEM) or transmission electron microscopy (TEM), since the latter usually requires sample preparation, which is often destructive. [4] In the scientific publication “Laboratory setup for extreme ultraviolet coherence tomography driven by a high-harmonic source”, J. Nathanael et al., Rev. Sci. Instrum. 90, 113702 (2019) an exemplary setup for the detection of XCT spectra is presented. using a nanostructured sample as an example, as well as an exemplary data processing of the XCT spectra, are described. Further introductory explanations of spectroscopic XCT are also disclosed in "Coherence tomography with broad bandwidth extreme ultraviolet and soft X-ray radiation," S. Skruszewicz et al., Applied Physics B (2021) 127:55. [5] How an exemplary XCT algorithm in combination with a one-dimensional phase retrieval (PR) algorithm can derive the deep structure of a sample from the autocorrelation signal is explained in detail in the scientific publication “Optical coherence tomography with nanoscale axial resolution using a laser-driven high-harmonic source", S. Fuchs et al., Vol. 4, No. 8 / August 2017 / Optica and the associated supplement (“Supplementary information”). [6] The scientific publication “Characterization of encapsulated graphene layers using extreme ultraviolet coherence tomography”, F. Wiesner et al., Vol. 30, No.18 / 29 August 2022 / OpticsExpress 32267, further discloses how reflectivity ratios following an XCT deep structure analysis can be used for the model-based extraction of interface parameters. [7] The inventors have recognized that OCT is limited in its resolution in the depth direction by the bandwidth of the incident light, but structures that cause the backscattering of the incident light can have a size in the depth direction that is below the OCT resolution given by the bandwidth. [8] One aspect of this disclosure is based on the object of specifying a method that can resolve structures of a sample within the scope of coherence tomography (CT) that are below the resolution given by the bandwidth of the coherence tomography. A further object is to derive parameters of such structures and, based on them, to adapt a deep structure model of the sample. Furthermore, aspects of the disclosure can be based on the object of more closely characterizing layer transitions in a sample with a layered structure. [9] At least one of these objects is achieved by a method according to claim 1 or by a method according to claim 10. Further developments are specified in the subclaims.

[0010] In one aspect, a computer-implemented method for adapting Deep structure model data to coherence tomography (CT) measurement data of a sample stored in a Depth direction in a layer structure has a plurality of layers (Si) of layer materials, the following steps: - Providing CT measurement data in a processor, wherein the CT measurement data comprise a depth structure r(z) uniquely derived from a spectral complex field reflectivity, wherein the depth structure r(z) is based on a resolution in the depth direction that is predetermined by a CT measurement underlying the spectral complex field reflectivity, - (in particular by data processing based on the depth structure r(z) with the processor,) deriving from the depth structure r(z) a plurality of layers with layer thicknesses and a plurality of layer transitions that are located upstream of the respective layers in the depth direction, (It is noted that the positions / area of ​​the layers and layer transitions in the depth direction result from the layer thicknesses of the layers in the layer sequence.) - (in particular by data processing based on the depth structure r(z) with the processor,) extracting layer transition-specific, spectrally resolved reflectivity coefficients from the depth structure r(z) for the majority of layer transitions, - Providing deep structure model data in the processor, wherein the deep structure model data models the layer structure of the sample and is initialized for the plurality of derived layers based on the layer thicknesses and the layer materials belonging to the layers, and a plurality of model layer transitions whose spectrally resolved model reflectivity coefficients can be modeled layer transition-specifically based on at least one parameter, and - (in particular by data processing with the processor) adapting the model reflectivity coefficients to the respective layer transition-specific, spectrally resolved reflectivity coefficients by calculating a value optimising the adaptation for at least one parameter of one of the model layer transitions, and - (in particular by data processing with the processor,) adapting the deep structure model data on the basis of the value calculated for the at least one parameter (and in particular storing the adapted deep structure model data in a memory and / or outputting the adapted deep structure model data on a display).

[0011] In some embodiments of the method, before adapting a model reflectivity coefficient of a layer transition, the model reflectivity coefficients preceding in the depth direction can be adapted to the layer transition-specific, spectrally resolved reflectivity coefficients.

[0012] In some embodiments of the method, the deep structure model data can be adjusted layer by layer in the depth direction with respect to the layer transitions. Additionally or alternatively, the value calculated for the at least one parameter can be below the resolution specified by the CT measurement data in the depth direction.

[0013] In some embodiments, the model reflectivity coefficients for adapting the deep structure model data to the CT measurement data can be optimized recursively in several steps with regard to layer transitions located deeper in the sample, wherein, in particular, starting from the surface of the sample, the layer structure can be reconstructed and the deep structure model data can be updated layer by layer by - in a first step, parameters of a surface layer (such as roughness and thickness of the surface layer) are optimized by minimizing a deviation between the measured and modeled reflectivity coefficient of the surface layer by varying the parameters, and the values ​​of the surface layer parameters are fixed in the deep structure model data for further optimization, - in a second step, parameters of the second layer transition in the depth direction, such as a roughness and a thickness of a buried oxide layer, are optimized by minimizing a deviation function between the measured and modeled reflectivity coefficient of the second layer transition by varying the parameters with a fixed parameter of the surface layer, and the values ​​of the parameters of the second layer transition are fixed in the deep structure model data for further optimization, and - In subsequent steps, the parameters of the nearest layer transition in the depth direction are optimized (and, in particular, fixed for subsequent adjustments). Optionally, the sum of the thicknesses determined for the surface layer and the subsequent layer transitions and the thicknesses of the layers determined from the depth structure r(z) can be kept constant relative to the depth structure r(z).

[0014] In some embodiments of the method, layer transition-specific, spectrally resolved reflectivity coefficients can be extracted by: - Identifying a depth range in the deep structure associated with a layer transition r(z), in particular by sequential spatial filtering of the deep structure r(z), and - Calculating the layer transition-specific, spectrally resolved reflectivity coefficients by a Fourier transform of the field reflectivity r(z) restricted to the respective layer transition.

[0015] In some embodiments of the method, the at least one parameter of a model reflectivity coefficient of a layer transition may be selected from the group of parameters comprising a layer transition thickness, a roughness, a spectral reflection or transmission value, a material density, a material type, a material composition, a stoichiometric ratio of a material composition and / or a material transition gradient of a material composition.

[0016] In some embodiments of the method, the layer transition may comprise an interface or an intermediate layer delimited by two interfaces between two layers of the same material or an interface or an intermediate layer delimited by two interfaces between two layers of different materials.

[0017] In some developments, the method may further comprise a first step of adapting the initialized deep structure model data to the coherence tomography measurement data, which comprises optimizing a model reflectivity coefficient that is assigned to an input-side interface of the sample and represents a first modelable layer transition, in particular a modelable surface layer, by adjusting it with a reflectivity coefficient.

[0018] In some developments, the method may further comprise an output step in which the adapted deep structure model data and in particular the specified values ​​of the at least one parameter for the layer transitions are output by the processor on an input and output device (and alternatively or additionally stored by the processor in a memory).

[0019] In some embodiments of the method, the value optimizing the adaptation for a layer transition can be calculated by minimizing a deviation function, which is dependent on the at least one parameter, between the layer transition-specific, spectrally resolved reflectivity coefficient assigned to the layer transition and the model reflectivity coefficient assigned to the layer, in particular iteratively, in the context of a parameter scan.

[0020] In a further aspect, a method for examining a sample having a plurality of layers in a layer structure in a depth direction by means of optical coherence tomography comprises the following steps: - Carrying out a CT measurement on the sample, whereby a spectral complex field reflectivity is generated as CT measurement data by a phase-sensitive measurement and evaluation of an optical coherence signal or a spectrally resolved measurement of an intensity reflectivity of the sample, carried out in particular within the framework of an artifact-free coherence tomography data analysis, and a derivation of phase information by an iterative phase retrieval process, - Deriving a unique depth structure r(z) by Fourier transforming the spectral complex field reflectivity with a processor, whereby the depth structure is based on a resolution in the depth direction that is specified by the CT measurement, - (in particular by data processing based on the depth structure r(z) with the processor,) deriving layers and layer thicknesses of the sample taking into account a dispersion correction based on known layer materials of the sample, - generating deep structure model data with the processor, wherein the deep structure model data models the layer structure of the sample and is initialized for the plurality of derived layers based on the layer thicknesses and the layer materials belonging to the layers, and a plurality of model layer transitions whose spectrally resolved model reflectivity coefficients can be modeled in a layer transition-specific manner based on at least one parameter, and - Carrying out a computer-implemented method, in particular as described above, for adapting the deep structure model data to the CT measurement data of the sample, whereby the deep structure model data are extended in particular by parameters of model layer transitions.

[0021] In some embodiments of the method, the CT measurement on the sample can be carried out using XUV radiation and, in particular, the spectral width of the XUV radiation can allow a resolution of the CT measurement in the depth direction to a range of a few 10 nanometers and the computer-implemented method can identify structures associated with the transition layers with nanometer-level resolution.

[0022] In some embodiments of the method, the CT measurement on the sample can be performed for a plurality of lateral regions, in particular to detect lateral inhomogeneities of layer structure parameters present within a layer. In particular, radiation can be irradiated successively onto the lateral regions, so that a spectral complex field reflectivity is generated for each of the lateral regions of the plurality of lateral regions for performing the computer-implemented method for adapting the deep structure model data to the CT measurement data of the sample.

[0023] The concepts described here relate to the depth structure r(z) of a sample (also referred to as axial sample structure). As can be seen from the publications mentioned above, in the context of coherence tomography, a depth structure r(z) - i.e. a "reflective" response function - results from the measurement of the backscattering of the incident light. The depth structure r(z) represents a distribution of the complex reflectivity (derived from the measurement) along the depth direction z. The depth structure r(z) (i.e. the axial distribution of the complex reflectivity) can be evaluated in order to describe the measured sample with a layer structure model. The evaluation can, for example, result in axial interface positions, i.e. those given in the depth direction; an interface is also referred to herein as a layer transition. An interface / layer transition can, for example,have a specific interface reflectivity (e.g., increased compared to neighboring "homogeneous" material regions). A (material) layer can be assigned to the region between two neighboring interface positions in the layer structure model, whereby the position and extent of the layer in the depth direction are determined by the position of the interfaces. Thus, layer parameters of "thick" layers of a sample's structure can be derived from the depth structure r(z) (derived from the measurement). The layer structure model assigned to the sample can be constructed from the position and extent of the layers.

[0024] The concepts described herein can, among other things, have the following advantages over the state of the art in generating models of a sample (such as electron microscopy) or avoid corresponding disadvantages of the state of the art: non-destructive interaction with the sample, essentially no need for destructive sample preparation, low heat input during the measurement, lateral resolution in the nanometer range and high sampling rates, especially in XCT. Advantages of the laser-based XCT methods disclosed herein as examples can be transferred, at least in part, to OCT methods in the visible or infrared spectrum.

[0025] Using the computer-implemented method disclosed herein for fitting deep structure model data to CT measurement data of a sample, a (essentially non-destructive) XCT examination can measure the layer structure and composition of a sample with a spatial depth resolution that can be equivalent to or even higher than that of electron microscopy. At the same time, the disclosed methods can enable the determination of the roughness of external and internal interfaces.

[0026] Sample characterization based on the computer-implemented method disclosed herein, using a novel, model-based stepwise optimization for the reconstruction of a layer structure, which can be combined with, among other things, broadband extreme ultraviolet (EUV) and soft X-rays for CT measurements, can provide information about local material parameters within the framework of a high-precision deep structure analysis with nanometer resolution. Such determinable material parameters include, for example, a stoichiometric ratio of a material composition, a thickness, and / or a roughness of a layer transition, whether at a surface (external interface) or at an internal interface.In general, based on the method disclosed herein, conclusions can be drawn about existing material phases in a sample and their spatial dimensions, for example a thickness of existing native surface and interfacial oxide layers.

[0027] A further advantage is that a depth-directed limitation of the Fourier transform (“truncated Fourier transform”) during the extraction of spectrally resolved reflectivity coefficients of layer transitions – as a kind of filtering process – can reduce the noise in the layer-transition-specific measurement data. In the frequency domain, the truncated Fourier transform acts like a sharp frequency filter, so that only the modulation of the depth structure r(z) corresponding to a layer transition considered within an optimization step is evaluated. The proposed method thus offers the advantage of determining physically measurable quantities more precisely through individual sequential optimization steps and thus being able to incorporate them into the sample model for further optimization steps. This is conceptually an essential Difference to, for example, an OCT brute-force optimization approach with completely free parameters and only one measurement variable, the OCT depth structure.

[0028] Disclosed herein are concepts that allow aspects of the prior art to be improved, at least in part. In particular, further features and their usefulness will become apparent from the following description of embodiments with reference to the figures. The figures show: Fig. 1 is a schematic representation of an exemplary measurement setup for a coherence tomography measurement with broadband extreme ultraviolet (EUV) and soft X-rays, Fig. 2 is a flow chart illustrating an exemplary method for examining a sample by means of a coherence tomography measurement and its exemplary evaluation of measurement data according to the invention, Fig. 3 is a flow chart illustrating an exemplary calculation of a parameter of a modeled layer transition in the optimization of a spectrally resolved model reflectivity coefficient with respect to a measured reflectivity coefficient and Fig. 4 a sketch to explain an adapted deep structure model.

[0029] Essential for the concepts described herein is that a measured complex field reflectivity r(o) contains information regarding a structure of a sample comprising layers as well as phase information regarding spectrally resolved reflectivity coefficients n exp(co), each of which is associated with a layer transition.

[0030] To adapt deep structure model data to CT measurement data, the inventors use in particular the possibility of experimentally measured reflectivity coefficients of the layer transitions n exp(o) recursively compare the reflectivity coefficients n mod(co) resulting from the deep structure model data. In particular, the inventors recognized that dispersion and absorption / transmission in overlying, previously irradiated layers affect the radiation reflected from a layer transition. Based on this, a recursive calculation of the reflectivities of the layer transitions is proposed for the creation of the deep structure model data. This calculation begins at the incident side (sample surface) and continues layer by layer into the sample. This procedure simplifies the calculations required for Optimization of deep structure model data. In other words, the inventors propose a recursive optimization of the deep structure model data, which, based on the physical reflection process in the sample, reduces the dependence of the individual layer reconstructions on each other not only to the level of modeling the model reflectivity coefficients n mo d(®), but also in the sequence of reconstruction, especially in the context of an iterative adaptation.

[0031] Furthermore, the inventors have recognized that for the reconstruction of parameters of a layer transition, the influences of roughness and layer thickness on the reflectivity of a layer transition can be distinguished. This has a particular impact on the modeling of the reflectivities of the layer transitions and their optimization with respect to the measurement data.

[0032] In general, the starting point of the method disclosed herein is an OCT / XCT measurement, which, in addition to deep structure information, provides phase information about the reflections at layer transitions for the evaluation of the measurement data, for example, in the form of a spectral complex field reflectivity r(o) of the sample. Examples of OCT measurement methods that capture phase information include: (1) OCT measurement methods that measure the phase of the sample reflectivity in an interferometric setup relative to a reference object of known phase and reflectivity, e.g. in the context of a spatial or spectral scanning process. (2) Fourier domain OCT measurement methods (especially XCT measurement methods) that have been supplemented by a phase evaluation, e.g., with a one-dimensional phase retrieval (PR) algorithm.

[0033] The analysis aspects described herein relate in particular to the so-called Fourier-domain OCT variant (FD-OCT / FD-XCT). This variant differs from the time-domain OCT variant, which is implemented, for example, within the framework of axial scanning of the sample, e.g., by moving a reference mirror. The time-domain OCT variant is not or only with difficulty technically feasible for XCT, so XCT is usually implemented as spectrometer-based OCT (also referred to as FD-OCT) or alternatively as swept-source OCT—as explained below in connection with Fig. 1. In FD-OCT, the actual measured variable is the spectrum reflected from the sample, which is modulated and carries information about the axial structure of the sample (depth information).

[0034] Fig. 1 schematically shows an XCT measurement setup 1 for XUV coherence tomography. The setup comprises a radiation source 3, for example, a femtosecond laser system for generating pulsed laser radiation with a central wavelength of 800 nm, pulse durations in the range of several tens of femtoseconds, and repetition rates in the kHz range. In an optical parametric amplifier, the wavelength can be shifted into the infrared range. The infrared laser pulses with a pulse energy in the mJ range are focused in a source chamber 5, e.g., in argon gas, for High Harmic Generation (HHG). By changing the delay between the pump and signal pulses in the optical parametric amplifier, XUV radiation with a quasi-continuous spectrum can be generated, which propagates along an optical beam path 7 of the XCT measurement setup.

[0035] The XUV radiation generated in the source chamber 5 is focused by a toroidal mirror 9 (e.g. f = 1 m) onto a sample P in a sample chamber 11. From the XUV generation point, the optical beam path 7 of the XUV radiation runs in a vacuum, whereby differential pumping stages in a sample chamber 11 create pressures in the range of, for example, ~ 10 8 mbar. Infrared radiation components can be filtered by thin aluminum and zirconium membranes 13.

[0036] Furthermore, the sample P and the beam path 7 can be displaced relative to each other (for example, by moving the sample P using a controllable sample holder, in particular in a (lateral) direction orthogonal to the normal of the sample surface or to the direction of incidence of the XUV radiation). This makes it possible to perform measurements on different lateral regions (positions) of the sample P. In other words, the beam path 7 can enter the sample P at different positions on the surface, allowing different regions of a layer to be scanned and, for example, lateral inhomogeneities of sample parameters to be detected (for example, a layer thickness, a roughness of a layer transition, etc.).

[0037] Radiation reflected from the sample P is measured by a spectrometer 15. For example, the radiation focused in one plane by a cylindrical mirror 17 can be efficiently spectrally characterized by a 1200-line grating 19 in the other plane. The recorded spectrum contains interference due to reflection from a layer structure of the sample P, which makes it difficult to characterize the sample P with regard to the Layer structure in an evaluation unit 21.

[0038] To separate the incident and outgoing radiation, sample P is illuminated at an angle. When evaluating the acquired spectrum, the inevitable absorption of the XUV radiation propagating within sample P plays a significant role, particularly in the XUV range. The strength of the absorption (and thus the transmission of the radiation to a deeper layer transition) depends on the frequency of the radiation and the materials present in sample P and must therefore be considered when modeling the reflectivities of layer transitions present in sample P in the deep structure model.

[0039] The evaluation unit 21 may comprise a central processing unit (CPU) programmed to execute calculations related to adapting deep structure model data to coherence tomography measurement data by executing instructions stored in program code. The CPU may comprise one or more microprocessors in conjunction with one or more memory elements 21B, as well as an input and output device 21C (e.g., a touchscreen). The memory element 21B may store one or more microprocessor-readable instructions (program code), CT measurement data, and deep structure model data and provide them to the microprocessor 21A for data processing.For example, the microprocessor 21A can perform a Fourier transformation of the CT measurement data, derive layers, layer thicknesses, and layer transitions, in particular their positions in the depth direction, from a deep structure r(z), extract spectrally resolved reflectivity coefficients, initialize and update deep structure model data, model spectrally resolved model reflectivity coefficients based on at least one parameter in a layer transition-specific manner, and optimize deep structure model data using values ​​determined for the parameters. In particular, the microprocessor 21A can, for example, be configured to adapt model reflectivity coefficients to layer transition-specific, spectrally resolved reflectivity coefficients by the microprocessor 21A calculating a value that optimizes the adaptation for a parameter of a model layer transition.In addition, the evaluation unit 21 can comprise a controller for controlling various components of the XCT measurement setup 1 in order to perform desired actions during the CT measurement, such as setting parameters of the XUV radiation, an impact position and an angle of incidence on the sample P, as well as reading out, accessing and / or sending measurement data sets of the spectrometer 15 and processing the measurement data sets and model data sets, for example identifying maxima in the acquired spectra, etc.

[0040] For further details of an exemplary XCT measurement setup 1, please refer to the publications mentioned above.

[0041] CT measurement methods are based on the evaluation of interference in the spectrum acquired with the spectrometer 15, for example. The interference is due, on the one hand, to the position of the surface of the sample P in relation to the layer structures lying deeper in the sample P, and, on the other hand, to the positions of the layer structures relative to one another. In particular, OCT / XCT evaluation methods have been developed which allow a clear reconstruction of the axial layer structure even for complex interference (see, for example, the publication by S. Fuchs et al. mentioned above). With so-called phase retrieval (PR)-XCT algorithms, not only the axial structure of the sample but also the complete spectral phase can be derived from the reconstruction. The phase contains information, for example, about the absorption and dispersion of the materials in the sample, as well as the incidence geometry and the refractive index of the dominant material in a layer.In order to derive the axial layer structure of the sample from the reflected spectrum, a Fourier transformation of the acquired spectrum into the spatial domain (here, for example, the depth direction z) can be performed.

[0042] The complex reflectivities caused by layer transitions can be described in the form of spectrally resolved reflectivity coefficients (exemplary modeled reflectivity coefficients are displayed on the input and output device 21C in Figure 1). These layer transition-specific reflectivities lead to additional phase terms that can be derived from the CT measurement data and used according to the invention to adapt deep structure model data.

[0043] With reference to Figures 2 to 4, an exemplary, particularly recursive, extension of the deep structure analysis of spectral CT measurement data obtained using OCT, here, for example, XCT, is disclosed below. Deep structure model data are adapted to the CT measurement data with a spatial depth resolution that preferably exceeds a resolution in the depth direction specified by the CT measurement data.

[0044] The depth structure analysis shown in Fig. 2 refers to the investigation of an axially structured sample P with a layer structure consisting of several layers as well as regions between the layers, which are referred to herein as layer transitions. For example, a layer transition can be the transition between two layers of different materials A and B. Such a transition can be achieved, for example, by a gradient in the density distribution of materials A and B in the region of the layer transition can be characterized. The layer transition can, for example, extend in an axial section of the sample P that is not resolved by a conventional CT measurement. Furthermore, a thin layer of another material C (material C between two "boundary layers") can form in the region of the layer transition. Such a layer transition can exist between two layers of the same material (material sequence ACA) or of different materials (material sequence ACB). Such a thin (intermediate or surface) layer can also extend in an axial section of the sample P that is not resolved by a conventional CT measurement.

[0045] Fig. 2 schematically shows sample P, onto which XUV / HHG radiation 31 is irradiated at an angle and from which radiation 33 is reflected according to a spectral complex field reflectivity r(o). For the CT measurement, sample P, with regard to the optical properties and the irradiated area, essentially has two homogeneous thick layers S1, S2 as well as two homogeneous layer transitions in the form of a thin surface layer s1 and a thin intermediate layer s2, located upstream of the respective layers S1, S2 in the depth direction z. If the refractive index changes in the depth direction z, e.g., at the transition between layers S1 and s2, this leads to (partial) backscattering or reflection of the incident radiation. In sample P, reflections on the front and back sides of the thin layers s1, s2 are schematically indicated by arrows.

[0046] In a generally described sample, layer transitions in the form of thin layers Sj border thick layers Sj, which can be directly resolved in the OCT signal, so that the layer structure of the sample is formed by an alternating sequence of thin layers Sj with layer thicknesses dj and thick layers Sj with layer thicknesses Dj. The layer transitions represent interfaces with a reflectivity to which reflectivity coefficients rj(co) can be assigned. The spectral complex field reflectivity r(o) of a sample with n layers is given by the sum of the mutually independent reflectivities rj(co) at the corresponding depths Zj and the real parts of the wavenumber AZ(CD):

[0047] As schematically summarized in the left half of Fig. 2 for the known CT analysis, the intensity reflectivity R(o) (also referred to herein as the OCT / XCT signal) is measured during the XCT measurement 101 with the spectrometer 15. From the intensity reflectivity R(o), the spectral complex field reflectivity r(o) can be derived, for example, using the aforementioned phase recovery method (step 103). The OCT / XCT signal already contains the information on the depth structure r(z) (resulting purely from a CT analysis) (as explained above, the depth structure r(z) is a distribution of the complex reflectivity of the sample (derived from the measurement) along the depth direction z), which can be calculated, for example, by Fourier transformation of the spectral complex field reflectivity r(o) of the sample P (step 105).Furthermore, by incorporating dispersion information 41 (for example, the available knowledge of the materials of the thick layers SI, S2 of the layer structure of the sample P), the z-positions of the thick layers SI, S2 and their layer thicknesses D1, D2 can be derived (step 107). These parameters essentially represent a result 43 of known OCT evaluation methods (also referred to herein as "pure OCT depth structure (parameters)").

[0048] For the resolution of layer transitions in the layer structure of the sample P, generally the acquisition of layer transition-specific information, the inventors take advantage of the fact that layer transitions that are not resolved in the pure OCT depth structure are characterized by parameters such as a layer transition-specific roughness, a very small thickness in the nanometer range, a layer transition-specific material composition (e.g. stoichiometric ratio of the materials AB or the formation of an oxide as material C), etc. These parameters determine the layer transition-specific reflectivities rj(o) and can be used to model the layer transitions.

[0049] An exemplary procedure for adapting deep structure model data M to the CT measurement data of the sample P while modeling layer transitions is shown schematically in the right half of Fig. 2.

[0050] From the depth structure r(z), the z-positions of the layer transitions (in Fig. 2, the thin layers s1, s2) that precede the respective layers SI, S2 in the depth direction z can be derived. Furthermore, layer-transition-specific, spectrally resolved reflectivity coefficients n exp (®), ß e Xp (®), r3 e Xp (®)... from the depth structure r(z), or the spectral complex field reflectivity r(o), respectively (step 109). For example, a depth range Azl corresponding to a thin layer sl assigned to the depth structure r(z). In general, the processor can, for example, detect maxima in the depth structure r(z) by sequential spatial filtering of the depth structure r(z). In other words, it can perform an algorithm-based determination / derivation of depth ranges Azl, Az2 around a signal peak in the depth structure r(z) (see also Fig. 4). By a Fourier transformation of the depth structure (field reflectivity) r(z) limited to the respective layer transition Sj, the processor can determine the (layer transition-specific) spectrally resolved reflectivity coefficient q assigned to the layer transition. exp (®) calculate.

[0051] Based on the result 43 from the evaluation of the XCT signal and the reflectivity coefficients q exp(®) a deep structure model is formed (step 111). The deep structure model models the layer structure of the sample P and is provided to the processor P in the form of deep structure model data M. In the example of Fig. 2, the deep structure model is initialized for the (thick) layers SI, S2 derived from the deep structure r(z) based on the layer thicknesses D1, D2 and the (known) layer materials assigned to the thick layers SI, S2.

[0052] Furthermore, the deep structure model includes model layer transitions msj, which are characterized by spectrally resolved model reflectivity coefficients q m od(o). The model reflectivity coefficients q mod(o) can be modeled specifically for each layer transition based on at least one parameter. In Fig. 2, roughness parameters <51, <52 and thicknesses dl, d2 are indicated as exemplary parameters for modeling the model layer transitions msi, ms?. Examples of further parameters of model layer transitions include a spectral reflection and / or transmission value, a material density, a material type, a material composition, a stoichiometric ratio of a material composition, and / or a material transition gradient of a material composition. For example, a sample material (such as an oxide layer) for a model layer transition can be derived from the measured reflectivity.

[0053] Fig. 2 further shows how, in an exemplary recursive optimization, the deep structure model data M are adapted layer by layer in the depth direction z with respect to the layer transitions sl, s2 to the CT measurement data.

[0054] For example, if a surface layer is present on the sample P, as shown in Fig. 2, reflectivity coefficients n exp (®) can be calculated. Accordingly, a value corresponding to the input-side interface of the sample P The model reflectivity coefficient n mod(o) can be modeled based on the parameters layer thickness dl, material type, and roughness c 1 . The model reflectivity coefficient n mod(o) characterizes the first modelable layer transition msi, in particular a modelable surface layer of the sample P. Approaches for modeling the reflectivity of a surface layer are known; see also the following explanations for modeling an intermediate layer. In a first step 121 of recursively adapting the initialized deep structure model data M to the CT measurement data, the model reflectivity coefficient n mod(o) is determined by comparing it with the reflectivity coefficient ri exp(o) and setting values ​​for the parameter(s) are optimized. Fig. 2 shows exemplary optimized parameters GI and dl as the result 51 of the first step 121, which are incorporated into the deep structure model data M (updating them) and form the basis for the calculation of a model reflectivity coefficient r2 mod(o) with respect to the propagation of the radiation up to the model layer transition ms?.

[0055] In a second step 123 of recursive fitting of the initialized deep structure model data M to the CT measurement data, a model reflectivity coefficient r2 mod(o) corresponding to layer s2 of sample P is modeled using the parameters layer thickness d2, material type, and roughness G2. The model reflectivity coefficient r2 mod(o) characterizes the second modelable layer transition ms2. Approaches to modeling the reflectivity of such an intermediate layer, for example via the reflectivities of the front and back of the intermediate layer as well as their distance, are known, see for example “Material-specific imaging of nanolayers using extreme ultraviolet coherence tomography”, F. Wiesner et al., Optica, 8(2):230-238, 2021. Refractive indices and susceptibilities of the materials involved can be found, for example, in “Lo -ene gy x-ray interaction coefficients: Photoabsorption, scattering, and reflection: E= 100 -2000 ev z= 1-94”, BL Henke et al., Atomic data and nuclear data tables 27(1): 1-144, 1982. Further modeling, for example for gradual material transitions between two layers, is known from “Influence of surface and interface roughness on x-ray and extreme ultraviolet reflectance: A comparative numerical study”, Y. Esashi et al., OSA Continuum, 4(5): 1497-1518, 2021. Further methods for reflection simulation / modeling are known, for example, from “Surface studies of solids by total reflection of x-rays”, LG Parratt, Physical review, 95(2):359, 1954 or. . Schick et al., Computer Physics Communications, 185(2):651-660, 2014.

[0056] There is another comparison, this time of the model reflectivity coefficient r2 mod(o) with the reflectivity coefficient r2 exp(o) for setting values ​​for the parameters of the second modelable layer transition ms2. Fig. 2 shows, as an example, optimized parameters G2 and d2 as a result 53 of the second step 123. These are again incorporated into the deep structure model data M and, together with the optimized parameters GI and dl, form the basis for calculating a model reflectivity coefficient of a subsequent layer transition (generally all subsequent layer transitions), in particular with regard to the propagation of the radiation up to the subsequent model layer transition, and for comparing it with a corresponding experimentally determined reflectivity coefficient (step 125). The procedure is carried out for all identified layer transitions, so that the deep structure model data M is expanded layer by layer with parameters of the model layer transitions.

[0057] In other words, before fitting a model reflectivity coefficient rj mod(co) of a layer transition j, the model reflectivity coefficients n< preceding in the depth direction z are <j mod(co) der vorausgehenden Schichtübergänge k (k<j) an die schichtübergangsspezifischen, spektral aufgelösten Reflektivitätskoeffizienten n<<j exp (o) - adjusted stepwise in depth direction z.

[0058] Furthermore, Fig. 2 illustrates a continuous check (step 127) of the deep structure model data M, in particular the modeled parameters 55 (illustrated in Fig. 2 with parameters ai and di) with the higher-level information from the measured deep structure r(z). The check results, for example, in an adjustment of the thickness Dl in light of the modeled layer thicknesses dl. The modeled layer thicknesses dl, d2 of the layer transitions are below the resolution of the CT measurement, but can thus be taken into account in the reconstruction of the layer structure of the sample P.

[0059] In summary, the knowledge of the layer transition-specific, spectrally resolved reflectivity coefficients n exp (o), obtained from the depth structure r(z), enables a selective optimization of parameters when modeling layer transitions in the depth direction z.

[0060] Fig. 3 shows a schematic flow diagram illustrating an example calculation of a parameter of a model layer transition in the optimization of a spectrally resolved model reflectivity coefficient. The calculation of the parameter values occurs embedded in the sequential, recursive optimization for each layer transition, for example in a subroutine executed with the processor.

[0061] Based on the XCT evaluation (result 43) of the CT measurement data (including phase reconstruction), the reflectivity coefficients rj exp(®) of the layer transitions, for example the reflectivity coefficients of individual thin oxide layers, are known. Parameter values ​​of a modeled layer transition j, for example the thickness dj and the roughness qj of an individual layer (or other parameters such as material density, etc.), are now to be calculated. The deep structure model data M are included in a modeling 61 of the reflectivity coefficients q mod(co). For example, the transmission of radiation through the sample P up to the layer transition j, i.e. "above" the layer transition, is calculated according to the already optimized parameters of the deep structure model. Analytical formulas for modeling the reflection of a thin layer with reflection coefficients on the front and back of the thin layer, taking into account polarization, components of the wave vector in the layers above and below, angle of incidence, and roughness (e.g.can be taken into account by the Nevot-Croce factor) etc. are known.

[0062] The subroutine shown in Fig. 3 concludes with a comparison of the reflectivity coefficient q obtained from the measurement exp (®) with the modeled reflectivity coefficient rj mod(o), where the modeled reflectivity coefficient rj mod(o) is varied using a parameter scan 63 of the parameters modeling the layer transition j. For the evaluation of the comparison, for example, a deviation function such as the squared deviation X 2 normalized with the measurement error a(co), can be minimized within the parameter scan 63. Alternatively, absolute or logarithmic deviations can be used.

[0063] The reflection is repeated for a multitude of combinations of the parameters dj, qj to be determined until a minimum in the deviation function is reached. For the calculation, for example, the parameter space can be discretized so that a multitude of parameter combinations can be calculated and the minimum of the deviation function can be automatically searched for. Alternatively, optimization algorithms such as downhill simplex, Levenberg-Marquardt, simulated annealing, and genetic algorithms can be used to determine the parameters. The subroutine in Fig. 3 outputs the parameters 65 optimized for layer transition j.

[0064] Fig. 4 illustrates the results obtained using the computer-implemented method described herein for fitting deep structure model data to CT measurement data during a non-destructive characterization of a metallic layer system investigated by XCT. Sample P, shown schematically in Fig. 4, consists of an intermetallic A12Cu substrate 71 and a solid Al coating 73. In addition to quantifying a depth profile 75 of sample P with nanometer resolution in the depth direction z, a reconstruction of layer thickness parameters and roughness parameters of layer transitions, in this case of a surface oxide layer 77 and an interfacial oxide layer 79, was performed using the novel model-based stepwise optimization explained herein.

[0065] In Fig. 4, measured reflectivity coefficients I rj exp l for the two layer transitions are plotted against the energy E (or CD) as a solid line 81, and the corresponding error intervals 83 are plotted using dotted lines. In the middle energy range shown, there is a substantial overlap with the modeled reflectivity coefficients 84 ( | rj mod I ) (dashed lines).

[0066] Fig. 4 also shows Table 91 with the sample parameters of the layer transitions / interlayers (thicknesses and roughnesses in nanometers) obtained from the non-destructive XCT measurement as part of the extended structural analysis. The resolution of the resulting values ​​is in the range of a few nanometers and is significantly improved compared to the resolution of pure OCT analysis. Table 91 also lists the results of surface-sensitive measurement methods, for example, an electron microscopy (TEM) and an atomic force microscopy (AFM), to validate the obtained sample parameters.

[0067] It can be seen that the non-destructive XCT measurement resolved the layer structure (Al layer thickness Dl, thickness dl of the surface oxide layer, thickness d2 of the oxide layer at the substrate interface) as well as the composition (oxides) with essentially the same spatial resolution as electron microscopy (see TEM values). In addition, the non-destructive XCT measurement, using extended structural analysis, enabled the determination of a roughness Rq0 of the outer interface and a roughness Rql of the inner interface. The roughness values ​​are comparable to the AFM values.

[0068] It is explicitly emphasized that all the inventions disclosed in the description and / or claims Characteristics as separate and independent of each other for the purpose of the original Disclosure as well as for the purpose of limiting the claimed invention should be viewed independently of the feature combinations in the embodiments and / or the claims. It is explicitly stated that all range specifications or specifications of groups of units disclose every possible intermediate value or subgroup of units for the purpose of the original disclosure as well as for the purpose of limiting the claimed invention, in particular also as a limit of a range specification.

Claims

Patent claims 1. Computer-implemented method for adapting deep structure model data (M) to coherence tomography (CT) measurement data of a sample (P) which has a plurality of layers (Si) of layered materials in a depth direction (z) in a layer structure, comprising the steps: - Providing CT measurement data in a processor, wherein the CT measurement data comprise a depth structure (r(z)) uniquely derived from a spectral complex field reflectivity (r(o)), wherein the depth structure (r(z)) is based on a resolution in the depth direction (z) that is predetermined by a CT measurement underlying the spectral complex field reflectivity (r(o)), - deriving from the depth structure (r(z)) a plurality of layers (SI, S2) with layer thicknesses (Dl, D2) and a plurality of layer transitions (sl, s2) which are located in front of the respective layers (SI, S2) in the depth direction (z), - Extraction of layer transition-specific, spectrally resolved reflectivity coefficients (n exp(co), r2 e xp(o)) from the depth structure (r(z)) for the majority of layer transitions (sl, s2), - Providing deep structure model data (M) in the processor, wherein the deep structure model data (M) model the layer structure of the sample (P) and are initialized for the plurality of derived layers (Sl, S2) based on the layer thicknesses (Dl, D2) and the layer materials belonging to the layers (Sl, S2), and a plurality of model layer transitions (msl, ms2), whose spectrally resolved model reflectivity coefficients (n mod(o), ß mod(o)) can be modeled layer transition-specifically based on at least one parameter (ol, dl; o2, d2), - Adjusting the model reflectivity coefficients (n mod(o), ß mod(o)) to the respective layer transition-specific, spectrally resolved reflectivity coefficients (n exp(®), r2 exp(o)), by calculating a value optimising the adjustment for at least one parameter (ol, dl; o2, d2) of one of the model layer transitions (msl, ms2), and - adjusting the deep structure model data (M) based on the value calculated for the at least one parameter (ol, dl; o2, d2).

2. Computer-implemented method according to claim 1, wherein, before adapting a model reflectivity coefficient of a layer transition, the model reflectivity coefficients preceding each other in the depth direction (z) are adapted to the layer transition-specific, spectrally resolved reflectivity coefficients.

3. Computer-implemented method according to claim 1 or 2, wherein the deep structure model data (M) are adapted layer by layer in the depth direction (z) with respect to the layer transitions (sl, s2) and / or wherein the value calculated for the at least one parameter (G 1 , dl; o2, d2) lies below the resolution in the depth direction (z) specified by the CT measurement data.

4. Computer-implemented method according to one of the preceding claims, wherein, in order to adapt the deep structure model data (M) to the CT measurement data, the model reflectivity coefficients (n mod(co), ß mod(co)) are optimized recursively in several steps with regard to layer transitions (sl, s2) lying deeper in the sample (P), wherein, in particular, starting from the surface of the sample (P), the layer structure is reconstructed and the deep structure model data (M) are updated layer by layer by - in a first step (121), parameters (ol, dl) of a surface layer such as a roughness and a thickness of the surface layer (sl) are optimized by determining a deviation (X 2 ) between measured and modelled reflectivity coefficient of the surface layer is minimised by varying the parameters (ol, dl), - the values ​​of the parameters (ol, dl) of the surface layer (sl) are fixed in the deep structure model data (M) for further optimization, - in a second step (123), parameters (o2, d2) of the second layer transition (s2) in the depth direction (z), such as a roughness and a thickness of a buried oxide layer, are optimized by determining a deviation function (X 2 ) between measured and modelled reflectivity coefficient of the second layer transition (s2) is minimised by varying the parameters (o2, d2) with fixed parameters (ol, dl) of the surface layer, - the values ​​of the parameters (o2, d2) of the second layer transition are fixed in the deep structure model data for further optimization, - in further steps (125) the parameters of the layer transition closest in the depth direction (z) are successively optimized, and - optionally the sum of the thicknesses (dl, d2) determined for the surface layer (sl) and the subsequent layer transitions (s2) and the thicknesses (Dl, D2) of the depth structure (r(z)) certain layers (SI, S2) is kept constant with respect to the deep structure (r(z)).

5. Computer-implemented method according to one of the preceding claims, wherein layer transition-specific, spectrally resolved reflectivity coefficients (n exp (®), r2 exp(o)) can be extracted by - identifying a depth range (Azl) in the depth structure (r(z)) associated with a layer transition (si), in particular by sequential spatial filtering of the depth structure (r(z)), and - Calculating the layer transition-specific, spectrally resolved reflectivity coefficients (n exp(co), r2 exp(o)) by a Fourier transform of the field reflectivity (r(z)) restricted to the respective layer transition (sl, s2).

6. Computer-implemented method according to one of the preceding claims, wherein the at least one parameter (ol, dl; o2, d2) of a model reflectivity coefficient (ri mod(o), r2 mod(o)) of a layer transition (si) is selected from the group of parameters comprising - a layer transition thickness (di), - a roughness (oi), - a spectral reflection or transmission value, - a material density, - a type of material, - a material composition, - a stoichiometric ratio of a material composition and / or - a material transition gradient of a material composition.

7. Computer-implemented method according to one of the preceding claims, wherein the layer transition (sl, s2) - an interface or an intermediate layer defined by two interfaces between two layers of the same material or - comprises an interface or an intermediate layer delimited by two interfaces between two layers of different materials.

8. A computer-implemented method according to any one of the preceding claims, further comprising: - a first step of adapting the initialized deep structure model data (M) to the coherence tomography (CT) measurement data comprising: Optimizing a model reflectivity coefficient (n mod(o)) which is assigned to an input-side interface of the sample (P) and represents a first modelable layer transition (sl), in particular a modelable surface layer, by adjusting it with a reflectivity coefficient (n exp (o)), and / or - an output step in which the adapted deep structure model data (M) and in particular the specified values ​​of the at least one parameter (ol, dl; o2, d2) for the layer transitions (sl, s2) are output by the processor on an input and output device (2 IC).

9. Computer-implemented method according to one of the preceding claims, wherein the value optimizing the adaptation for a layer transition (sl, s2) is calculated by determining a deviation function (X 2) between the layer transition-specific, spectrally resolved reflectivity coefficients (n exp (®), ß _ex P (®)) and the model reflectivity coefficient (n mod(o), ß mod(o)) assigned to the layer, in particular iteratively, is minimized within the framework of a parameter scan.

10. Method for examining a sample (P) which has a plurality of layers (Si) in a layer structure in a depth direction (z) by means of optical coherence tomography (CT) with the steps: - Carrying out a CT measurement on the sample (P) (step 101), whereby a spectral complex field reflectivity (r(o)) is generated as CT measurement data by - a phase-sensitive measurement and evaluation of an optical coherence signal or - a spectrally resolved measurement of an intensity reflectivity (R(o)) of the sample (P), carried out in particular within the framework of an artifact-free coherence tomography data analysis, and a derivation of phase information by an iterative phase retrieval method (step 103), - Deriving a unique depth structure (r(z)) by Fourier transforming the spectral complex field reflectivity (r(o)) with a processor (step 105), wherein the Depth structure (r(z)) is based on a resolution in the depth direction (z) which is determined by the CT measurement, - deriving layers (SI, S2) and layer thicknesses (Dl, D2) of the sample (P) taking into account a dispersion correction based on known layer materials of the sample (P) (step 107), - generating deep structure model data (M) with the processor (step 109), wherein the deep structure model data (M) model the layer structure of the sample (P) and are initialized for the plurality of derived layers (S1, S2) based on - the layer thicknesses (Dl, D2) and the layer materials belonging to the layers (SI, S2), and - a plurality of model layer transitions (msl, ms2), whose spectrally resolved model reflectivity coefficients (n mod(o), ß mod(o)) can be modeled layer transition-specifically based on at least one parameter (ol, dl; o2, d2), and - Carrying out a computer-implemented method for adapting the deep structure model data (M) to the CT measurement data of the sample (P) according to one of the preceding claims, whereby the deep structure model data (M) are extended in particular by parameters (ol, dl; o2, d2) of model layer transitions (msi).

11. The method according to claim 10, wherein the CT measurement on the sample (P) is carried out using XUV radiation and in particular the spectral width of the XUV radiation limits a resolution of the CT measurement in the depth direction to a range of a few tens of nanometers and the computer-implemented method identifies structures associated with the transition layers with a resolution in the nanometer range.

12. The method according to claim 10 or 11, wherein the CT measurement on the sample (P) is carried out for a plurality of lateral regions in order to detect, in particular, lateral inhomogeneities of parameters of the layer structure present in a layer, and wherein, in particular, radiation is irradiated successively onto the lateral regions so that a spectral complex field reflectivity (r(o)) is generated for each of the lateral regions of the plurality of lateral regions for carrying out the computer-implemented method for adapting the deep structure model data (M) to the CT measurement data of the sample (P) according to one of the preceding claims.

Citation Information

Patent Citations

  • Short-wavelength coherence tomography

    US7656538B2