Method for determining properties of a semiconductor layer structure

Abstract

The invention relates to a method for determining at least one property of a semiconductor layer structure, comprising the following method steps: A) providing the semiconductor layer structure; B) providing reference data containing, for at least one reference location point of a reference layer structure, a plurality of wavelength-dependent reference intensities of radiation reflected and / or emitted by the reference layer structure; C) providing at least one reference property of the reference layer structure at the reference location point, the reference property comprising at least a layer thickness and / or a porosity; D) measuring the semiconductor layer structure at at least one measurement location point and generating measurement data, which contains a plurality of wavelength-dependent measurement intensities of radiation reflected and / or emitted by the semiconductor layer structure for the measurement location point on the semiconductor layer structure, the number of wavelength-dependent measurement intensities being lower than the number of wavelength-dependent reference intensities and / or the measurement data having spectrally lower resolution than the reference data; E) determining the property of the semiconductor layer structure at the measurement location point according to the reference intensities and the reference property and the measurement intensities, the determined property of the semiconductor layer structure comprising a layer thickness and / or a porosity.

Description

[0001] Methods for determining the properties of a semiconductor layer structure

[0002] Description

[0003] The invention relates to a method for determining properties of a semiconductor layer structure.

[0004] Semiconductor layer structures are semiconductor components with one or more layers that are used in semiconductor technology to manufacture electronic components.

[0005] Semiconductor wafers, especially those made of silicon, represent cost-intensive components in the production of photovoltaic components and account for up to 60% of the price of a solar cell. For this reason, it is desirable to achieve the highest possible material yield in the production of semiconductor wafers.

[0006] The production of semiconductor wafers typically begins with the purification and crystallization of semiconductor material to create a high-purity crystal block, also known as an ingot. This crystal block is then sliced ​​into wafers and processed into the aforementioned semiconductor wafers. The crystal block can be sliced ​​using saws, which allows for good dimensional accuracy of the semiconductor wafers. However, valuable semiconductor material is lost as a result of the swarf.

[0007] Epitaxial growth of semiconductor wafers using vapor deposition (VPD) allows for a reduction in material losses during wafer fabrication. In this process, a carrier gas containing the chemical precursors of the semiconductor layer to be produced is passed over the substrate. These precursors react at the substrate surface, and a homogeneous thin film is formed through appropriate process control. To enable reuse of the so-called seed wafer, the grown layer must be removable. For this purpose, the seed wafer is prepared before the epitaxial growth process in such a way that a separating layer forms, allowing for easy separation from the seed wafer. As a first step, the seed wafer is porosified, resulting in a stacked system of several thin layers with varying porosity.

[0008] In a multi-layer system, the uppermost of these layers may, for example, have a lower porosity than the layer below it and is therefore referred to as a low porosity layer (LPL), while the layer below it is referred to as a high porosity layer (HPL).

[0009] To create layers with varying porosities, an etching process is often used. Depending on the process parameters, such as the bath composition, etching time, and applied electric field, the etching process can be controlled, allowing the different porous properties of the layers to be adjusted as needed. This is often referred to as porous etching. The etching process removes material and creates pores or cavities in the layer, thereby increasing the porosity.

[0010] As a result of a subsequent sintering process, the pores on the surface of the LPL close, creating a flat surface for the epitaxial growth of semiconductor material, especially silicon, from the gas phase.

[0011] Simultaneously, the sintering process enlarges the pores of the HPL layer to such an extent that the pore walls practically disappear. This type of cavity formation enables the subsequent separation of the epitaxially produced semiconductor wafer from the substrate and the reuse of the substrate for further growth cycles.

[0012] The properties of LPL and HPL are crucial for obtaining high-quality semiconductor wafers that detach easily from the substrate. Therefore, quality assurance of the semiconductor layer structure is essential for efficient production. Optical spectrophotometric data analysis is a common method for obtaining information about the structure and properties of thin films. The necessary measurements are typically performed with high-precision spectrophotometers. However, such measurements are time-consuming, especially when spatially resolved data on the semiconductor layer structure is required, particularly across the entire layer.

[0013] To achieve sufficiently high production volumes, it is desirable to implement quality assurance through inline production monitoring, ideally subjecting all manufactured layer structures to quality inspection. To prevent such inline production monitoring measures from becoming a bottleneck in production, the required measurement and testing times should not exceed the production cycle time and should therefore be completed quickly.

[0014] In addition to the need to monitor the manufacturing process of semiconductor wafers and their layer structures, there is also a need to verify the quality of already manufactured solar cells with two or more layers. This applies particularly to so-called heterojunction (HJT) solar cells, which comprise layers of amorphous silicon (a-Si) and transparent conductive oxide (TCO), or tunnel oxide passive contact (TOPCon) solar cells with polycrystalline silicon. Given the high market value of multilayer solar cells, there is a strong economic interest in being able to reliably and quickly ensure that they exhibit the desired properties.

[0015] A similar need for rapid and comprehensive quality assurance exists in the production of so-called perovskite solar cells or perovskite-silicon tandem solar cells, which are gaining increasing importance due to their high efficiency and cost-effective production potential. Perovskite solar cells are also based on multilayer structures, in which the perovskite absorber layer and the adjacent functional layers, especially the electron layer and the hole transport layer, must each exhibit high structural and functional quality. However, during the application of these layers, e.g., by spin coating or blade coating, local inhomogeneities, defects, or differences in the thickness and properties of the layers can occur, which can significantly affect the performance and long-term stability of the solar cell.To ensure process stability and minimize reject rates, inline-capable quality assurance is also required here, enabling spatially resolved, fast and non-destructive testing of the layer properties - without exceeding the cycle time of solar cell production.

[0016] The invention is therefore based on the objective of proposing a method with which at least one property of a semiconductor layer structure can be determined spatially resolved, quickly and with high accuracy.

[0017] The problem is solved by a method according to claim 1. Advantageous further developments are each the subject of the dependent subclaims.

[0018] The method according to the invention serves to determine at least one property of a semiconductor layer structure and comprises the following process steps:

[0019] A - Providing the semiconductor layer structure;

[0020] B - Providing reference data which, for at least one reference location of a reference layer structure, contain a plurality of wavelength-dependent reference intensities of radiation reflected and / or emitted by the reference layer structure;

[0021] C - Providing at least one reference property of the reference layer structure at the reference location point, wherein the reference property includes a layer thickness and / or a porosity;

[0022] D - Measuring the semiconductor layer structure at at least one measurement point and generating measurement data which, for the measurement point on the semiconductor layer structure, contain a plurality of wavelength-dependent measurement intensities of radiation reflected and / or emitted by the semiconductor layer structure, wherein the number of wavelength-dependent measurement intensities is lower than the number of wavelength-dependent reference intensities and / or wherein the measurement data have a lower spectral resolution than the reference data;

[0023] E - Determining the property of the semiconductor layer structure at the measurement location point as a function of the reference intensities and the reference property and the measurement intensities, wherein the determined property of the semiconductor layer structure includes a layer thickness and / or a porosity.

[0024] A key insight underlying the invention is that the optical properties of a semiconductor layer structure depend on the layer properties of the semiconductor layer structure itself. Therefore, by measuring the reflected and / or emitted radiation, it is fundamentally possible to deduce the layer properties.

[0025] In particular, the radiation emitted and / or reflected by the semiconductor layer structure is light radiation, the intensity of which can be attributed to specific layer properties, namely porosity and / or layer thickness. For example, the semiconductor layer structure can be excited to photoluminescence and / or electroluminescence, and the light radiation emitted in this process exhibits different wavelength-dependent intensities across its spectral range. Likewise, the semiconductor layer structure can be exposed to light radiation and reflect it in such a way that the reflected light radiation exhibits different wavelength-dependent intensities across its spectral range.

[0026] A second finding is that the measurement of emitted and / or wavelength-dependent radiation can be performed with a comparatively small number of measurement intensities and / or with a low spectral resolution of the measurement intensities, and yet it is still possible to determine the property of the semiconductor layer structure to be determined with high accuracy. For this, however, reference data with a higher spectral resolution and / or with a greater number of reference intensities must be provided, which are specifically assigned to the reference properties present at the reference location.

[0027] In a simple embodiment, for example, the measurement intensities and the reference intensities at identical wavelengths or wavelength ranges can be compared to allow conclusions to be drawn about the property to be determined if there is sufficient similarity between the measurement intensities and the reference intensities. The reference layer structure can therefore be considered a semiconductor layer structure whose properties are referred to as reference properties, which are comparable to the layer properties of the measured semiconductor layer structure to be determined.

[0028] In one possible embodiment, the reference layer structure is a sub-region of the semiconductor structure provided in process step A. Here, the reference intensities can be determined with high resolution before or during process step B, at least at the reference point. In another possible embodiment, the reference layer structure is a sub-region of a sample specimen on which the reference data are acquired. In particular, the sample specimen can originate from the same production batch as the semiconductor substrate from process step A. However, it is also within the scope of the invention that the sample specimen can originate from a previous production batch. In the simplest application, both the reference data and the reference properties can be determined metrologically.

[0029] In particular, the reference properties at the reference location can include a reference layer thickness and / or a reference porosity. It is also conceivable that the reference properties at the reference location contain one or more optical constants, which are determined, in particular, through measurement or simulation before the procedure is carried out. Specifically, the reference layer structure can exhibit a non-constant layer thickness profile. Studies have shown that the influence of optical constants on the reference and measurement data is discernible with a non-constant layer thickness profile.

[0030] The low spectral resolution of the measurement data enables measurements to be performed quickly and at a large number of measurement points, in order to determine the layer properties of the semiconductor layer structure, namely the porosity and / or the layer thickness, across the entire surface and with spatial resolution. Within the scope of the invention, the term "spatially resolved" means that the property of the semiconductor layer structure can be determined at a selectable measurement point on the surface of the semiconductor layer structure and can be specifically assigned to this measurement point. For example, the spatial resolution of a measurement system by which the measurement data can be acquired is 1024 x 1024 pixels or 3000 x 3000 pixels, preferably with a pixel size between 20 pm / pixel and 200 pm / pixel.

[0031] Within the scope of the invention, the wavelength-dependent intensities generally correspond to a plurality of scalar intensity values, each of which can be assigned a wavelength in the spectrum of the emitted radiation. Spectral resolution refers to the smallest possible detectable difference between the wavelengths of the intensities of the emitted and / or reflected radiation. High spectral resolution therefore means that the smallest possible difference between two wavelengths at which an intensity value can be measured is small. Conversely, low spectral resolution means that the smallest possible difference between two wavelengths at which an intensity value can be measured is large.

[0032] The wavelength-dependent reference intensities can be available with a spectral resolution of 2 nm, particularly in a spectrum between 300 nm and 1100 nm. In contrast, the wavelength-dependent intensities can be available with a resolution of 50 nm, also particularly in a spectrum between 300 nm and 1100 nm. Preferably, the wavelength-dependent reference intensities and / or the measured wavelength-dependent intensities are determined by measuring discrete wavelength-dependent intensity values. Considering the aforementioned spectral resolutions, the reference data can contain up to 400 reference intensities in total. The measurement data can contain only three, four, or five measurement intensities in total.

[0033] It is within the scope of the invention that the reference data and the measurement data differ from each other not only in their spectral resolution. Additionally or alternatively, the number of wavelength-dependent measurement intensities in the measurement data can be higher than the number of wavelength-dependent reference intensities in the reference data. To improve the comparability between the measured data and the stored reference data, it can be provided that the data are calibrated or scaled before the actual property determination. In this process, the measurement data and / or reference data are subjected to normalization, linear scaling, or other mathematical transformation to bring them into a comparable range of values.This allows for a more precise mapping of measurement intensities to stored reference properties and reduces the influence of systematic deviations, for example, due to different measurement conditions, instrument characteristics, or sample surfaces. Calibration can be performed using defined calibration standards or by an automated comparison of statistical characteristics, such as mean, variance, and spectral shape.

[0034] In a simple embodiment, the reference data can be determined at least at the reference location using a spectrometer. The measurement data, on the other hand, can be acquired using imaging with a multispectral measurement system and / or a hyperspectral measurement system, at least at the measurement location.

[0035] One difference between a multispectral and a hyperspectral measurement system can be seen in the fact that the multispectral measurement system is designed to capture a comparatively small number of broad wavelength bands, while the hyperspectral measurement system is designed to capture a comparatively large number of narrow wavelength bands.

[0036] In multispectral measurement, the semiconductor layer structure can be selectively illuminated with light from a defined number of wavelength ranges sequentially, allowing the multispectral measurement system to be configured for camera-based detection of wavelength-dependent measurement intensities. In hyperspectral measurement, an image can be captured by decomposing reflected light into wavelength ranges using filters and / or a prism. Both multispectral and hyperspectral measurement systems can utilize homogeneous white illumination of the semiconductor layer structure, for example, using an illumination dome.

[0037] The semiconductor layer structure can comprise at least partially monocrystalline silicon. The semiconductor layer structure can comprise at least two layers, wherein a first layer has a lower porosity than a second layer. The specific layer property can include a porosity and / or layer thickness of the first layer and / or a porosity and / or layer thickness of the second layer. It is also conceivable that the semiconductor layer structure comprises more than two layers, in particular three layers, preferably more than three layers.

[0038] In an advantageous embodiment of the invention, the reference data provided in process step B are entered into a first model before process step E, and the reference property of the first model is output in process step C. Alternatively, the reference property provided in process step C is entered into a first model before process step E, whereupon the first model outputs the reference data in process step B.

[0039] According to the further development described above, the reference layer structure can be defined by the first model, which can be used to map a general relationship between the layer properties and the optical properties of the semiconductor layer structure. In particular, spectrally high-resolution reference intensities can be input into the first model, which then outputs the corresponding reference properties. As described above, a first model can alternatively be designed such that the reference property is input into the first model, and the first model outputs the corresponding reference data.

[0040] The first model can therefore be designed in at least two different ways, which differ in their possible inputs and outputs. According to the first design, at least one reference intensity can be inputted into the first model, and a reference property can be output by the first model. According to the second design, at least one reference property can be inputted into the first model, and a reference intensity can be output by the first model.

[0041] By using the first model, it is unnecessary to measure both the reference properties and the reference intensities. Instead, only the reference properties or the reference intensities can be measured, and the other quantity can be determined model-wise using the first model. For this purpose, the first model can comprise a physical model, in particular a simulation model, and / or an analytical model and / or a trained model based on artificial intelligence, in particular a neural network.

[0042] It is recognized that the optical constants of one or more layers of semiconductor layer structures influence the measurable layer thicknesses and / or porosities. Therefore, it is advantageous that the reference properties also include at least one optical constant at the reference location and can likewise be incorporated into the first model. This makes it possible to define, and in particular train, the first model in such a way that the metrological influence of the optical constants on the measurable layer properties at the measurement location can be taken into account. Specifically, it is possible for the first model to output a property of the semiconductor layer structure at the measurement location where the metrological influence of the optical constant at that location is compensated. The at least one optical constant can, in particular, be determined by ellipsometry at the reference location.

[0043] In an advantageous further training program, the procedure includes the following procedural step:

[0044] DO – Measuring the reference layer structure at the reference location and generating comparison data. This data contains a plurality of wavelength-dependent comparison intensities of radiation reflected and / or emitted by the reference layer structure at the reference location, with the comparison intensities having the same number and / or spectral resolution as the measurement intensities. A second model is defined at least as a function of the comparison data and the reference property at the reference location. In process step E, the property of the semiconductor layer structure at the measurement location is determined by inputting the measurement data into the second model, and the second model outputs the layer property at the measurement location. In particular, process step DO is performed before or during process step E.The aforementioned further development process is based on the fact that the reference point serves both to acquire spectrally high-resolution reference intensities and to acquire a comparison dataset. Crucially, the comparison intensities must be available with the same number and / or spectral resolution as the measurement intensities. This has the advantage that the second model can be defined based on data with the same information density as the data used to determine the properties of the semiconductor layer structure. This is particularly advantageous when the second model is based on artificial intelligence and is trained using the comparison data and the reference properties at the reference point. In this case, the input in process step E, in the form of the measurement intensities, is in the same format as the data used to train the second model.

[0045] While the first model specifies a relationship between the property of the reference layer structure and reference intensities, the second model specifies a relationship between the property of the reference layer structure and the comparison intensities. Preferably, the second model comprises an analytical model and / or a model based on artificial intelligence, in particular a neural network.

[0046] It is within the scope of the invention that the comparison data are provided using the same means as those used to acquire the measurement intensities, in particular by means of a multispectral measurement system and / or a hyperspectral measurement system. It is also within the scope of an advantageous embodiment that the comparison data based on the reference layer structure are provided using the same means as those used to provide the reference data. In particular, a spectrometer can be used for this purpose, whose spectral resolution is set lower and / or only a portion of the measured intensities is used to determine the comparison data.

[0047] Provided that the reference properties at the reference location also take into account at least one optical constant, the second model can advantageously be defined, and in particular trained, in such a way that its metrological influences are taken into account when determining the properties of the semiconductor layer structure at the measurement location.

[0048] In an advantageous further development, the reference data and the reference properties are provided at a plurality of mutually adjacent reference points, and the comparison intensities at these reference points are determined. The second model is defined, at least as a function of the comparison intensities and the reference properties at the reference points, such that in process step E, the second model takes into account a neighborhood relationship between properties at two adjacent measurement points. Preferably, the reference points of the plurality of reference intensities, on which the second model is defined as a function, are located along a line or in a planar arrangement.

[0049] The further development described above makes it possible to determine the property of the semiconductor layer structure not solely based on the measurement intensities at a single measurement point. Rather, it reveals that the properties of the semiconductor layer structure typically change continuously between multiple measurement points. Therefore, considering the property of the semiconductor layer structure at a neighboring measurement point can be used to support the determination of the property at that specific measurement point. Preferably, the property to be determined at a measurement point in process step E can thus be ascertained as a function of the property at a neighboring measurement point. For this purpose, the property at a measurement point can, for example, be initially determined and then corrected as a function of the property at a neighboring measurement point.Similarly, the property at the adjacent measurement point can be used to verify the plausibility of the determined property at that point or to assign a reliability value to the determined property. A further advantage is that determining properties at adjacent measurement points leads to continuous results, which are easier to interpret, especially when imaging the semiconductor layer structure.

[0050] If the second model is analytical, the neighborhood relationship can be specified as a function of a parameter, which, for example, represents the intensity difference between two measured intensities at two adjacent measurement points and preferably must not exceed or fall below a predefined threshold. If the second model is based on artificial intelligence, it can directly consider the neighborhood relationship for determining the property of the semiconductor layer structure.

[0051] Preferably, the property of the semiconductor layer structure determined in process step E at the measurement point is fed back into the second model and process step E is repeated at a further measurement point, wherein the layer property at the further measurement point is output by the second model as a function of a neighborhood relationship to the layer property at the measurement point.

[0052] It is advantageous that a previously determined property of the semiconductor layer structure at one measurement point is fed back into the second model so that it can be considered when determining the property of the semiconductor layer structure at another measurement point. If process step E is performed using a computing unit, this can reduce the required computing power, since the property of the semiconductor layer structure at the further measurement point can be determined not explicitly, but with the aid of the previously determined property and a neighborhood relationship.

[0053] In a further advantageous development, the reference properties provided in process step C are synthetically generated at a plurality of reference locations and fed into the first model. In process step B, the first model outputs the reference data, in particular with spectrally high-resolution reference intensities, and in process step DO, the comparison data, in particular with spectrally low-resolution comparison intensities.

[0054] As described above, the first model can essentially be a model that specifies a relationship between the layer properties of a semiconductor layer structure at a reference point and the optical properties of the semiconductor layer structure. By providing a synthetically generated reference layer structure, the first model can serve to generate a large amount of data, which can be used in particular for defining, and preferably training, the second model. The synthetically generated reference layer structure can be viewed as a set of synthetically generated data that represent the reference layer properties. In particular, the synthetically generated data includes at least multispectral measurement data and / or reflectance spectra and / or thickness and / or porosity properties.

[0055] In a further advantageous development, the second model is defined depending on the reference properties, the reference data, and the comparison data. In process step E, the property of the semiconductor layer structure is determined spatially resolved at a large number of measurement points by inputting the measurement data into the second model, which then outputs the layer property spatially resolved for the measurement points.

[0056] Preferably, the reference properties can be located along a one-dimensional trace and / or a two-dimensional region on the reference layer structure and can have an arbitrarily adjustable spatial distribution. By providing a large amount of high-resolution reference data and low-resolution comparison data, the second model can be defined, and in particular trained, to output the layer property for any desired measurement point in process step D.

[0057] In another variant of the process, the reference data provided in process step B, with high-resolution spectral reference intensities, and the comparison data provided in process step DO, with low-resolution spectral comparison intensities, are synthetically generated and fed into a first model. This model outputs the reference properties at the reference locations in process step C. The second model is defined, as previously described, based on the reference properties, the reference data, and the comparison data. In process step E, the property of the semiconductor layer structure is determined with spatial resolution at a large number of measurement locations by inputting the measurement data into the second model, which then outputs the layer property with spatial resolution for the measurement locations.The method variant described above differs from the previously described further development in that the reference data and the comparison data are synthetically generated. Specifically, the reference data and / or the comparison data can be in the form of a two-dimensional image that represents the wavelength-dependent intensity distributions underlying the high-resolution reference data and / or the low-resolution comparison data. Using the first model, the reference properties can be synthetically determined and used to define the second model.

[0058] In an advantageous embodiment, the layer structure comprises at least partially monocrystalline silicon, and the semiconductor layer structure comprises two epitaxially grown layers, wherein a first layer has a higher porosity than a second layer. Alternatively, the semiconductor layer structure can comprise at least one perovskite layer and at least two functional layers, wherein a first of these layers has a different optical property, in particular transmission, absorption, or reflection behavior, than a second of these layers.

[0059] One advantage is that the implementation of the method is not limited to a specific type of semiconductor layer structure, but is particularly applicable to epitaxially grown layers as well as to those that can form or already form part of a perovskite solar cell.

[0060] In an advantageous embodiment, at least process step D is carried out on different, and in particular sequentially produced, layers of the semiconductor layer structure during its fabrication. It is advantageous that the process can be carried out at different times and process stages during the fabrication of semiconductor layer structures to enable quality control. In particular, the measurements according to process step D and, especially, the determination of the properties of the semiconductor layer structure according to process step E are thus performed layer by layer and, in particular, with a time offset from one another. The invention also relates to a device for carrying out the process according to the invention or an advantageous embodiment thereof.The device comprises a receptacle configured to provide a semiconductor layer structure and at least one storage unit configured to provide reference data which contains, for at least one reference location of a reference layer structure, a plurality of wavelength-dependent reference intensities of radiation reflected and / or emitted by the reference layer structure and at least one reference property of the reference layer structure at the reference location, wherein the reference property comprises at least a layer thickness and / or a porosity.The device comprises at least one measuring unit by means of which the semiconductor layer structure is provided at at least one measuring point for generating measurement data. This data contains, for the measuring point on the semiconductor layer structure, a plurality of wavelength-dependent measurement intensities of radiation reflected and / or emitted by the semiconductor layer structure. The number of wavelength-dependent measurement intensities is lower than the number of wavelength-dependent reference intensities, and / or the measurement data has a lower spectral resolution than the reference data. The device further comprises an evaluation unit for determining the property of the semiconductor layer structure at the measuring point as a function of the reference intensities and the measurement intensities. The determined property of the semiconductor layer structure includes a layer thickness and / or a porosity.In particular, the statements regarding the features and achievable technical advantages of the method are transferable to the device and apply accordingly.

[0061] The advantages and further developments of the invention are explained below with reference to exemplary embodiments and the figures.

[0062] They show

[0063] Figure 1 shows a semiconductor layer structure in side view a) and

[0064] Top view b); Figure 2 Process steps of a first method for determining a property of the semiconductor layer structure;

[0065] Figure 3 Process steps of a second method for determining a property of the semiconductor layer structure;

[0066] Figure 4 Process steps of a third method for determining a property of the semiconductor layer structure;

[0067] Figure 5 Process steps of a fourth method for determining a property of the semiconductor layer structure.

[0068] Figure 1 shows a semiconductor substrate 1, which is a crystalline silicon ingot on which a semiconductor layer structure 2 is arranged, produced by an electrochemical etching process. The semiconductor layer structure 2 comprises a first layer 3 and a second layer 4. The first layer 3 has a higher porosity than the second layer 4 and serves as a separating layer so that the second layer 4 can be detached from the semiconductor substrate 1. Typically, the first layer 3 is referred to as the high porosity layer (HPL) and the second layer 4 as the low porosity layer (LPL).

[0069] In order for the upper layer 4 to be separated from the semiconductor substrate 1 as a semiconductor wafer and used to manufacture a semiconductor component, such as a solar cell, it is necessary that both the upper layer 4 and the lower layer 3 exhibit the desired quality with regard to their porosity and thickness, and that this quality can be determined spatially resolved, quickly, and with high accuracy. The semiconductor layer structure 2 can, in a manner not described in detail here, comprise a perovskite layer and one or more adjacent functional layers and be intended for the production of a so-called perovskite solar cell.

[0070] The following implementation variants of a method are based on the understanding that the optical properties of the semiconductor layer structure depend on the properties of layers 3 and 4. In particular, it is possible to excite the semiconductor layer structure to emit and / or reflect light radiation and to deduce the properties of the semiconductor layer structure by measuring and evaluating the spectrum of this light radiation. Typically, measuring and evaluating spectral data involves long measurement times and is unsuitable for inline process monitoring. However, the method variants described here make it possible to shorten the required measurement times and achieve high accuracy for spatially resolved measurements.

[0071] A first embodiment of the method is illustrated in views a) and b) of Figure 2.

[0072] View a) of Figure 2 shows the structure shown in Figure 1 in a top view, with the upper layer 4 obscuring the underlying layer 3 and the semiconductor substrate 1. The semiconductor layer structure is divided into several points, which are shown as squares for easier understanding and include a reference point 5 and a measurement point 6. For greater generality, the coordinates of the reference point 5 are denoted as x0, y0 and the coordinates of the measurement point 6 as x1, y1. One aspect of carrying out the method variants described here is that the reference point 5 serves as the starting point for determining the layer properties at measurement point 6 with spatial resolution and in a short time.

[0073] For this purpose, reference data are first determined at the reference location r(xO, yO) using a spectrometer. As shown in Figure a), these data comprise a plurality of wavelength-dependent reference intensities. The reference data r(xO, yO) exhibit 400 discretely distributed reference intensities over a wavelength range between 300 nm and 1100 nm.

[0074] Furthermore, the semiconductor layer structure at measurement point 6 is measured, and measurement data q(x1, y1) are determined, which contain a plurality of wavelength-dependent measurement intensities for measurement point 6. The measurement data q(x1, y1) exhibit a significantly lower spectral resolution compared to the reference data, whereby in the embodiment shown here, a number of 4 discretely distributed reference intensities are available over a wavelength range between 300 nm and 1100 nm. In addition, reference properties at the reference location pi are determined. Pi, Phpi (xO, yO) are provided, containing the layer thicknesses and porosities of layers 3 and 4. The reference properties pipi, p Ppi (xO,yO) can be determined using another reference measuring device or derived from an estimate or a computational model. The reference properties can, in particular, include optical constants at the reference location point in order to define and, in particular, train model 7 in such a way that the influence of the optical constants on the measurable intensities and the resulting layer thicknesses and porosities at the measurement location point can be taken into account.

[0075] As illustrated by dashed arrows in view b) of Figure 2, the reference data r(xO, yO) and the reference properties pipi, Phpi (xO,yO) are used to define a model 7 which specifies a relationship between wavelength-dependent intensities and the porosities and thicknesses of the semiconductor layer structure 2.

[0076] By entering (solid arrow) the measurement data at the measurement point q(xO, yO), a comparison with the reference data r(xO, yO) is carried out using model 7 and the layer property at the measurement point pi is determined. P i, ph P i (x1, y1) output.

[0077] Model 7 is a mapping rule that approximates the wavelength-dependent distribution of the reference intensities, thus making it easy to compare the measurement data q(x1, y1) with r(x0, y0) and to check whether the measurement intensities and the reference intensities are sufficiently similar or identical. Depending on the specific type of Model 7, this allows conclusions to be drawn as to whether the layer properties at measurement point 6 are the same as at reference point 5, or to what extent they differ.

[0078] One advantage of the approach shown here is that the measurement data q(x1, y1) at the first measurement point have a comparatively low data density and can be acquired quickly. This facilitates the implementation of the procedure within the framework of inline quality control. This advantage is particularly pronounced when the reference data r(x0, y0) are acquired with a spectrometer and exhibit a comparatively high data density. As further explained below, it is even possible to completely forgo the measurement-based acquisition of the reference data and generate it synthetically, thereby further reducing the required measurement times for the actual measurements to be performed on the semiconductor layer structure.

[0079] A second embodiment of a method for determining the layer properties is illustrated in views a) and b) of Figure 3.

[0080] As described in view a) of Figure 2, view a) of Figure 3 shows a semiconductor layer structure in plan view. As already explained in connection with view a) of Figure 2, reference data r(xO, yO) and measurement data q(x1, y1) are provided. Additionally, comparison data q(xO, yO) are provided at the reference location. These contain a plurality of wavelength-dependent comparison intensities that have the same spectral resolution as the measurement data q(x1, y1) at the measurement location. In particular, the comparison data q(xO, yO) can be acquired using the same means as those used to determine the measurement data q(x1, y1), for example, a multispectral measurement system.

[0081] In contrast to the method explained in connection with Figure 2, the method shown in Figure 3 uses two separate models 7, 8 with which the layer properties at the measurement point pi are determined. P i, ph Pi (x1, y1) can be determined.

[0082] For this purpose, the reference data r(xO, yO) at the reference location are inputted into a first model 7, which is a physical model that specifies a general relationship between spectrally high-resolution intensities and the layer properties. The output of the first model 7 is the reference properties at the reference location pi. P i, p Ppi (xO, yO).

[0083] The reference properties at the reference location point pi P i, pn PiThe data at the reference location q(xO, yO) and the comparison data at the reference location q(xO, yO) are used to train a second model (dashed arrows), which provides a general relationship between low-resolution spectral intensities and the layer properties. The measurement data q(x1, y1) at the measurement location are fed into the second model 8, whereupon this model calculates the layer properties at the measurement location pi. P i, pn Pi outputs (x1,y1).

[0084] The advantage of the method shown in Figure 3 is that the second model is trained with reference data that has the same spectral resolution as the measurement data at the measurement point used to determine the properties of the semiconductor layer structure at that point. This improves the quality of the second model and enables a more accurate determination of the properties of the semiconductor layer structure at measurement point 6.

[0085] A third embodiment of a method for determining the layer properties is illustrated in views a) and b) of Figure 4.

[0086] As described in view a) of Figure 3, view a) of Figure 4 shows a semiconductor layer structure in plan view. In contrast to the previous explanations, reference data are provided at a plurality of reference locations r(xO..., yO...), with the corresponding reference locations 5, 5', 5"... arranged along a continuous track on the semiconductor layer structure. The comparison data q(xO..., yO...) are also provided for a plurality of reference locations.

[0087] In addition, measurement data are acquired at a first and a second measurement point q(x1, y1) and q(x2, y2), respectively. The second measurement point 9 has the coordinates x2, y2 and is adjacent to the first measurement point 6. The measurement data at the second measurement point q(x2, y2) have the same spectral resolution as the measurement data at the first measurement point q(x1, y1).

[0088] Similar to what was already explained in connection with view a) of Figure 3, the reference data r{xO..., yO...) are entered into the first model 7, whereupon this model outputs the layer properties p^p^YxO..., yO...) for the majority of the reference locations 5, 5', 5".... The layer properties Pi p i,p hThe reference data pi'(xO..., yO...) and the comparison data q(xO..., yO...) are used to train the second model 8. It is advantageous that the reference points are adjacent to each other, as this allows the second model 8 to be trained additionally taking into account the proximity of the layer properties at neighboring points (dashed arrows).

[0089] By inputting the measurement data at the first measurement point q(x1,y1), the second model 8 can thus determine the layer properties at the first measurement point pi. P i, p Ppi Output (x1,y1). Additionally, the layer properties at the first measurement point pi are displayed. P i, pn P i (x1,y1) is fed back (dashed arrow). This allows the second model to determine the layer properties pi. P i, p PpiConsider (x1, y1) at the first measurement point if measurement data is entered at a second measurement point q (x2, y2) and the layer properties at this second measurement point pi p i, p Ppi (x2,y2) taking into account the neighborhood relationships to the layer properties at the first measurement point pi P i, p Ppi Output (x1,y1).

[0090] A fourth embodiment of a method for determining the layer properties is illustrated in Figure 5.

[0091] In the method shown in Figure 5, a synthetically generated distribution of layer properties Pi is used. p i,Pi ipix, y) as input for the first model 7, which in this embodiment is a physical model that generally specifies a relationship between layer properties and wavelength-dependent intensities. It is important that the synthetically generated distribution of the layer properties is planar, so that the layer properties are available for any coordinate. The synthetically generated distribution can be generated in any way imaginable, for example, by taking into account previous investigations of fabricated layer structures or by estimation.

[0092] The output of the first model 7 includes, on the one hand, spectrally high-resolution reference data r(x,y) and, on the other hand, spectrally low-resolution measurement data q(x,y), both of which are also spatially resolved. The reference data and the measurement data r(x,y) and q(x,y) serve to define the second model 8 by performing a consistency check 10, establishing a relationship between the reference data and measurement data r(x,y) and the layer properties Pipi,p to be determined. hpi (x,y) is to be specified. Upon completion of the supervised learning process, reference data at a reference location point r(x0,y0) as well as measurement data available across the entire surface q(x,y) are entered into the second model 8 and the layer properties are spatially resolved for any arbitrary point on the semiconductor layer structure Pipi,p hpi (x,y) output.

[0093] Alternatively, the reference data and the measurement data r(x,y) and q(x,y) can be acquired metrologically and / or synthetically generated, rather than being output by the first model 7. These can then be entered into the first model 7 to determine layer properties Pi. p i,Phpi*(. x >y zu They can be determined and, on the other hand, used to define the second model 8. Furthermore, the explanations regarding the embodiment shown in Figure 5 apply accordingly.

Claims

Claims 1. Method for determining at least one property of a semiconductor layer structure, comprising the following method steps: A - Providing the semiconductor layer structure; B - Providing reference data which, for at least one reference location of a reference layer structure, contain a plurality of wavelength-dependent reference intensities of radiation reflected and / or emitted by the reference layer structure; C - Providing at least one reference property of the reference layer structure at the reference location point, wherein the reference property includes at least one layer thickness and / or one porosity; D - Measuring the semiconductor layer structure at at least one measurement point and generating measurement data which, for the measurement point on the semiconductor layer structure, contain a plurality of wavelength-dependent measurement intensities of radiation reflected and / or emitted by the semiconductor layer structure, wherein the number of wavelength-dependent measurement intensities is lower than the number of wavelength-dependent reference intensities and / or wherein the measurement data have a lower spectral resolution than the reference data; E - Determining the property of the semiconductor layer structure at the measurement location point as a function of the reference intensities and the reference property and the measurement intensities, wherein the determined property of the semiconductor layer structure includes a layer thickness and / or a porosity.

2. The method of claim 1, wherein the reference data provided in process step B are inputted into a first model and the first model outputs the reference property in process step C or The reference property provided in process step C is entered into a first model, and the first model outputs the reference data in process step B.

3. Method according to claim 1, wherein the first model comprises a physical model, in particular a simulation model, and / or an analytical model and / or a model based on artificial intelligence, in particular a neural network.

4. Method according to claim 2 or 3, comprising one method step DO - Measuring the reference layer structure at the reference location point and generating comparison data which, for the reference location point on the reference layer structure, contain a plurality of wavelength-dependent comparison intensities of radiation reflected and / or emitted by the reference layer structure, wherein the comparison intensities of the comparison data are available with the same number and / or spectral resolution as the measurement intensities, and a second model is defined at least as a function of the comparison data and the reference property at the reference location point, and in process step E, the property of the semiconductor layer structure at the measurement location point is determined by inputting the measurement data into the second model, and the second model outputs the layer property at the measurement location point.

5. The method of claim 4, wherein the second model comprises an analytical model and / or a trained model based on artificial intelligence, in particular a neural network.

6. The method of claim 4, wherein the reference data and the reference property are provided at a plurality of mutually adjacent reference locations and the comparison data are determined at the reference locations, and the second model is defined at least as a function of the comparison data and the reference properties, such that the second model in process step E takes into account a neighborhood relationship between properties of the semiconductor layer structure at two adjacent measurement locations.

7. Method at least according to claim 6, wherein the reference location points of the plurality of reference intensities, depending on which the second model is defined, are arranged along a line or in a planar arrangement.

8. Method according to claim 6 or 7, wherein the property of the semiconductor layer structure determined in process step E is fed back into the second model at the measurement location and Procedure step E is repeated at another measurement point, whereby the layer property at the further measurement point is output by the second model depending on a neighborhood relationship to the layer property at the measurement point.

9. Method at least according to claim 4, wherein the reference properties provided in process step C are synthetically generated at the reference location points and are input into a first model, and the first model outputs the reference data with spectrally high-resolution reference intensities in process step B and the comparison data with spectrally low-resolution comparison intensities in process step DO.

10. Method at least according to claim 4, wherein The reference data provided in process step B with spectrally high-resolution reference intensities and the comparison data provided in process step DO with spectrally low-resolution comparison intensities are synthetically generated and entered into a first model, and the first model outputs the reference properties at the reference location points in process step C.

11. Method according to claim 9 or 10, wherein the second model is defined depending on the reference properties, the reference data and the comparison data, and in method step E the determination of the property of the semiconductor layer structure at a plurality of measurement points is carried out with spatial resolution by inputting the measurement data into the second model and the second model outputting the layer property with spatial resolution for the measurement points.

12. Method according to one of the preceding claims, wherein the wavelength-dependent reference intensities and / or the measured wavelength-dependent intensities are determined by measuring discrete wavelength-dependent intensity values.

13. Method according to claim 12, wherein at least the wavelength-dependent measurement intensities are captured by imaging using a multispectral camera and / or a hyperspectral camera.

14. Method according to one of the preceding claims, wherein the semiconductor layer structure comprises two epitaxially grown layers, wherein a first layer has a higher porosity than a second layer, or the semiconductor layer structure comprises at least one perovskite layer and at least two functional layers, wherein a first of these layers has different optical properties, in particular transmission, reflection or absorption properties, than a second of these layers.

15. Method according to one of the preceding claims, wherein at least process step D is carried out during the production of a semiconductor layer structure on different and, in particular, successively produced layers of the semiconductor layer structure.

Images