Hybrid measurement system and method for measuring in thin films

Abstract

The present invention relates to a hybrid measurement system and method for measuring in a thin film. Specifically, the present invention provides a measurement method and system for online measurement of one or more parameters of a thin film in a structure traveling on a production line. First measurement data and second measurement data are provided from a plurality of measurement sites on the thin film being measured, wherein the first measurement data corresponds to a first type of measurement from a first selected group of a relatively small number of measurement sites, and the second measurement data corresponds to a second type of optical measurement from a second group of a significantly higher number of measurement sites. Processing the first measurement data is used to determine at least one value of at least one parameter of the thin film in each measurement site of the first group. The at least one parameter value is used to interpret the second measurement data, thereby obtaining data indicating the distribution of the values ​​of the at least one parameter within the measurement sites of the second group.

Description

[0001] This application is a divisional application of Chinese patent application No. 201680057884.1, whose application date is August 4, 2016 and invention name is “Hybrid measurement system and method for measurement in thin films”. Technical Field

[0002] The present invention belongs to the field of measurement technology and relates to a measurement system and method, which can be particularly used to control the manufacturing process of a structure / sample containing a thin film, such as a semiconductor wafer. Background Art

[0003] The advanced semiconductor industry is continuing to integrate thinner and multi-layer films of new material compositions (e.g., SiGe, HKMG, 3D FinFet, etc.) This is one way to continue performance growth to ensure that advanced nodes (such as 1X and beyond) continue to follow Moore's Law.

[0004] Process control of ultra-thin films in the 1-2 nm range requires total measurement uncertainty (TMU1 - relative accuracy measure) of the order of 0.1 Å and fast matching (Fleet Matching). Optical metrology techniques such as ellipsometry, currently the workhorse for high throughput thin film measurements, are reaching their performance limits. As an example, the ability to accurately identify film thickness variations of 0.1 Å is fundamentally limited by crosstalk with other parameters of the thin film stack (thickness of other layers, variable material composition) and by limited optical sensitivity. Summary of the invention

[0005] There is a need in the art for new methods for monitoring / measuring film parameters, particularly for films with variable parameters such as, for example, material concentration. Moreover, the monitoring technology should preferably be able to effectively process control the manufacture of film-based structures performed on a production line. In other words, the measurement technology should be effectively operable in a real-time (or online or on-tool) measurement mode. In this regard, it should be noted that the terms "on-line" or "on-tool" as used herein refer to an alternative measurement mode of measurement by a separate measurement tool operating in an offline mode.

[0006] Optical metrology techniques are based on the measurement of the properties of light reflected by a sample. For thin film measurements, they rely on the contrast between the optical properties of different films to model, deconvolute and extract the film thickness alone, and are often extended to profile measurements of periodic structures using the properties of scattered light (scatterometry, also known as optical critical dimension (OCD)). While this approach is fast and well suited for advanced process control (APC), it relies on assumptions about optical properties and geometry. In particular, the performance of ultra-thin films with variable composition is often subject to errors due to the inability to distinguish the effects of changes in optical properties from changes in thickness. This is particularly true in the context of thin film measurements. Figure 1 It is schematically shown in , which shows an illustration of optical measurements and data interpretation of planar or periodic structures.

[0007] The figure shows the on-tool (or online) and offline measurement schemes in a self-explanatory manner. In the online measurement scheme, spectral measurements (e.g., polarization-based measurements) are applied to a sample (e.g., a semiconductor wafer), and a fitting procedure is applied to the measured spectral data, and upon identifying the best fitting spectral data from the library, the corresponding structural parameters (e.g., CD, height, SWA, etc.) are determined. Also as shown, a spectral library is established using an offline measurement scheme. The offline measurement scheme utilizes model creation (modeled structure on the target) using input data such as film stack information, film optical properties, and geometry information. The model can then be optimized using various known methods for creating / updating a spectral library, i.e., theoretical spectral data.

[0008] X-ray methods come in different forms. X-ray photoelectron spectroscopy (XPS) is particularly useful for thin films and measures the spectral response of photoelectrons generated by X-ray irradiation of a target. Such photoelectrons rapidly recombine within the material, with the net effect that only electrons generated in the surface ~10nm of the material are able to escape the stack and be detected. Signals from different atomic species / molecules are recorded as a function of energy and attributed to different layers of the stack, consistent with physical interactions between the X-rays, electrons and the different materials. This is done in a controlled manner. Figure 2 Schematically shown in , it shows a diagram of XPS measurement and measurement data in a self-explanatory manner. Therefore, XPS measurement can independently extract the composition (relative number of atoms of different types) and layer thickness of multilayer ultra-thin films. However, XPS technology does not provide the desired throughput required for mass production (HVM). In other words, XPS measurement is actually not suitable for the online measurement mode of structures performed on the production line for production process control.

[0009] refer to Figure 3A and 3B , which shows the impact of measurement uncertainty on HVM. Figure 3AMeasurements are shown that are subject to precision (Type A) errors, which can generally be minimized by reducing the number of floating parameters. Figure 3B Measurements subject to accuracy (Type B) errors are shown. Any measurement is subject to different sources of error. Depending on the specific case of model-based metrology, Type A errors are easily quantified based on random noise-accuracy statistical analysis. This type of error is usually minimized by reducing the number of model parameters that may vary (and correspondingly increasing the stability of the values ​​of key parameters when repeated measurements are made). Type B errors (accuracy indicators) are usually caused by crosstalk between key parameters (floating) and some fixed parameters (the entire corresponding physical sample parameters actually change). Such fixed parameters can include optical properties of the material. In an HVM environment, such errors are largely hidden and can lead to incorrect wafer placement. As Figure 3B As shown, comparison with a reference can help establish a relative calibration to "true" (arrow G) during recipe setup. However, unintentional (unknown) changes in some fixed model parameters during HVM can cause corresponding changes in the measured values; this invalidates the original relationship to "true", potentially triggering erroneous wafer placement.

[0010] A "reference level" measurement is defined as being invariant to unintentional changes in other parameters caused by normal process variations in HVM, while being ultra-sensitive to the measurement parameter in question. Examples of such measurements can be considered X-ray fluorescence (XRF), which provides the total number of atoms of a particular species, regardless of atoms from other species present in the measured sample, or XPS, which detects atomic species present in the surface layers of the sample, regardless of the stacking depth beyond ~10nm. Ideally, in-line metrology would provide "reference level" measurements. In practice, there is always a trade-off between measurement speed and performance, and in many cases, resolving this issue is beneficial in favor of higher throughput, higher risk metrology.

[0011] Hybrid metrology is the practice of combining two or more metrology tools that measure the same or similar structures. Data is shared between tool sets in a complementary manner to improve metrology performance and enable measurement of complex structures that cannot be measured with adequate performance by any individual tool set. To date, most hybrid metrology work has combined non-reference metrology (CD-SEM and OCD).

[0012] The present invention provides a new hybrid metrology solution for thin film measurement based on a combination of X-ray based techniques and optical methods. This is a combination of slow and accurate reference level metrology (XPS) with fast and precise non-reference (optical), a combination of XPS and optics to allow the measurement of composition on thin films and structures.

[0013] The inventive technique provides a so-called Hybridable Smart Sampling (HESS) measurement technique. This is related to the fact that tighter process windows require better control of process variability on a sample (e.g., a wafer) as well as "hot spot" detection and edge monitoring. Higher measurement sampling rates are desired, as well as performance at higher quality levels. To mitigate the tradeoff between speed and performance, the inventive technique explores a hybrid combination of XPS and optical methods, where XPS uses a "sparse sampling plan" and optical uses a "dense full-wafer sampling". The inventive technique extends XPS performance to full-wafer maps with improved throughput provided by optical methods.

[0014] More specifically, the present invention extends measurements with a limited sample set to larger sampling (optical measurements). This technique is generally applicable to situations where there are a few (relatively few) high-accuracy (trusted, slow) measurements and a large number (relatively large number) of fast optical measurements. Such relatively slow high-accuracy measurements may include one or more of the following techniques: XPS, X-ray fluorescence (XRF), transmission electron microscopy (TEM), focused ion beam (FIM), secondary ion mass spectrometry (SIMS). Relatively fast measurements are optical measurements, which may be based on, for example, spectral reflectance measurements, polarization spectral reflectance measurements, etc.

[0015] It should be understood that for the purposes of this disclosure, "relatively slow" and "relatively fast" measurements and corresponding measurement systems are distinguished from each other in that "slow" measurements require more time to collect and interpret measurement data, but provide higher accuracy than those of "fast" measurements. Therefore, in the present invention, "relatively slow" and "relatively fast" measurements are applied to a much smaller number N of measurement locations / points / locations and a much larger number M of measurement locations / points / locations, respectively, such that M=N+K (K is an integer starting from 1).

[0016] Preferably, the measurement positions are selected so that N positions L1 ... L are measured slowly (usually at least two such positions). n Included in the M positions of rapid measurement, namely L1, ..., L n , L'1, L'2, ..., L' k middle.

[0017] The measurement sites / positions for at least the fast measurements are preferably selected to be substantially uniformly distributed within the structure under measurement. For example, considering a typical semiconductor wafer (i.e., a disk-like structure), the "fast" measurement positions are substantially uniformly distributed along the radius of the wafer (i.e., including positions in the central region and the peripheral region).

[0018] Fast measurements are trained for high accuracy slow measurements. Training on a structure (wafer) basis can eliminate several major causes of optical tool inaccuracy, making extrinsic and intrinsic calibration issues and wafer loading tilt no longer relevant. In fact, the measurement optics can be simple, where only raw spectral collection and training are required, eliminating requirements such as special alignment, NA issues, and no tool-to-tool matching of the optics is required. Using multiple measurement channels and polarization-based measurements can be beneficial, but they do not require matching from system to system.

[0019] Thus, the technique of the present invention eliminates most of the inaccuracies of calibration, tool matching, alignment, etc. (which usually require highly specialized best known methods). In addition, it can eliminate modeling. Furthermore, it eliminates the risk of training a design of experiments (DOE) that does not adequately include the measured sample (because without a DOE, the sample itself is the DOE). The technique provides an extension of any reference metrology (slow) to a larger coverage using a measured spectrum that acts as a "virtual reference" after training on a wafer.

[0020] Therefore, according to one broad aspect of the present invention, there is provided a measurement method for online measurement of one or more parameters of a thin film in a structure traveling on a production line. The method comprises:

[0021] providing first measurement data and second measurement data from a plurality of measurement sites on the film being measured, wherein the first measurement data corresponds to a first type of measurement from a first selected group of a relatively small number of measurement sites, and the second measurement data corresponds to a second type of optical measurement from a second group of a significantly higher number of measurement sites;

[0022] processing the first measurement data and determining at least one value of at least one parameter of the thin film in each measurement site of the first set;

[0023] The second measurement data are interpreted using the at least one parameter value, thereby obtaining data indicative of a distribution of values ​​of the at least one parameter within the second set of measurement sites.

[0024] Preferably, the second set of measurement sites are selected to be substantially uniformly distributed within the film being measured.

[0025] Preferably, the measurement sites of the second group include the measurement sites of the first group.

[0026] In some embodiments of the present invention, the above process of interpreting the second measurement data using the at least one parameter value includes: identifying a matching set between the measurement sites in the first and second measurement data; creating a temporary trained recipe for the matching set, and using the temporary trained recipe to interpret the second measurement data.

[0027] In some embodiments, the first set of measurement sites are selected to enable measurement of minimum and maximum values ​​of at least one parameter of interest. In some instances, the selection is based on prior knowledge of a changing shape of at least one parameter of a thin film on a structure.

[0028] In some embodiments, the methods of the present invention are used for measurements in semiconductor wafer structures.

[0029] As described above, the first type of measurement and the second type of measurement differ at least in the time required to collect and interpret the measurement data, wherein the first type of measurement is a relatively slow measurement and the second type of measurement is a relatively fast measurement. Preferably, the first type (slow) measurement is characterized by a higher accuracy than the second type of measurement.

[0030] In some embodiments, the first type of measurement comprises XPS. The second type of measurement is an optical measurement and may comprise a spectral reflectance measurement.

[0031] According to another broad aspect of the present invention, there is provided a measurement method for online measurement of one or more parameters of a thin film in a structure travelling on a production line, the method comprising:

[0032] applying XPS measurements to a first selected group of a relatively small number of measurement sites in the thin film and providing data from the XPS measurements;

[0033] applying the optical measurement to a second set of measurement sites including a significantly higher number of measurement sites than the first set and providing optically measured data;

[0034] processing data from the XPS measurements and determining a value of at least one parameter of the thin film in each measurement site of the first set;

[0035] The optically measured data are interpreted using the values ​​of the at least one parameter, thereby obtaining data indicative of a distribution of the values ​​of the at least one parameter within the film, the distribution being indicative of the quality of a process applied to the structure in production.

[0036] According to yet another broad aspect of the present invention, there is provided a system for online measurement of one or more parameters of a film in a structure traveling on a production line. The system comprises:

[0037] a first measurement device configured and operable to perform a first type of measurement according to a first sparse sampling plan and to provide a first measured representation of at least one parameter of the thin film at a first selected set of a relatively small number of measurement sites in the thin film;

[0038] a second optical measurement device configured and operable to perform a second type of measurement according to a second dense sampling plan and to provide a second measured representation of at least one parameter of the thin film in a second set of a significantly higher number of measurement sites; and

[0039] A control system configured to process first and second measurement data, the processing comprising processing the first measurement data and determining at least one value of at least one parameter of the film in each measurement site of the first group; and interpreting the second measurement data using the at least one parameter value to obtain data indicating the distribution of the values ​​of the at least one parameter within the measurement sites of the second group. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to better understand the subject matter disclosed herein and to illustrate how it can be implemented in practice, embodiments will now be described by way of non-limiting examples only, with reference to the accompanying drawings, in which:

[0041] Figure 1 is a schematic diagram of the general principle of a known scheme for optical measurements, such as using an OCD tool based on spectral reflectometry;

[0042] Figure 2 is a schematic diagram of the general principle of a known protocol for XPS measurements;

[0043] Figure 3A and 3B The effect of measurement uncertainty on HVM is shown;

[0044] Figures 4A to 4C The principle of the present invention is schematically shown, wherein Figure 4A A block diagram of a hybrid measurement system of the present invention is shown; Figure 4B A flow chart showing a measurement scheme used in such a system; Figure 4C Examples of the setup of measurement sites / locations for sparse and dense sampling plans are given;

[0045] Figures 5A to 5F Measurements on a single layer film are exemplified for the SiON case, where Figure 5A Shown is a design of experiments (DOE) set of 5 wafers measured over time by XPS; Figure 5B XPS and optical images of the measurement sites are shown; Figure 5CThe HESS-SiON layer thickness measured by the OCD tool, HESS and XPS, and the N_Dose composition measured by HESS and XPS are shown; Figure 5D-5E HESS topography images of five DOE wafers are shown, respectively for SiON layer thickness and N-dose composition; Fig. 5F The accuracy of XPS, thickness measured by OCD tool and HESS measurement are shown;

[0046] Fig. 6A and 6B An example is given of measurements in a multilayer film stack where Fig. 6A shows the results of using a standard optical measurement with two tool sets; and Figure 6B A HESS measurement with two tool sets is shown; and

[0047] Fig. 7A and Figure 7B Measurements on a planar stack comprising an ultra-thin high-k (HK) on top of an ultra-thin interlayer (IL) dielectric layer are illustrated. DETAILED DESCRIPTION

[0048] Figure 1 and Figure 2 The general principles of known approaches for "fast" (online) optical measurement and measurement data interpretation and "slow" (offline) X-ray measurement and measurement data interpretation, respectively, are schematically shown. Figure 3A and 3B The effect of measurement uncertainty on HVM is shown.

[0049] As described above, the present invention provides a thin film measurement technique suitable for online measurement, and in particular for process control in manufacturing thin film-containing structures.

[0050] refer to Figures 4A to 4C , which schematically illustrates the principle of the present invention. Figure 4A A block diagram of a hybrid measurement system 100 of the present invention is shown; Figure 4B A flow chart 200 showing a measurement scheme used in such a system; Figure 4C The arrangement of measurement sites / locations for sparse and dense sampling plans is illustrated.

[0051] The measurement system 100 is configured to perform in-line measurements of thin film containing structures W traveling on a production line, and can be configured for process control of a production process continuously applied to the structures by a particular processing tool. More specifically, the present invention is for process control used in the semiconductor industry, and therefore, is exemplified below with respect to this specific application. However, it should be understood that the principles of the present invention are not limited to this particular application.

[0052] Furthermore, the present invention is particularly useful for measuring thin film parameters, such as unpatterned layers, and is therefore exemplified below for measurements on the topmost thin film layer on a wafer. However, it should be understood that the techniques of the present invention can also be used for measurements in patterned layers.

[0053] Furthermore, the invention is exemplified below as using XPS and optical measurements, as slow and fast measurements, respectively. However, as mentioned above, these are non-limiting examples of relatively slow and relatively fast measurement techniques, respectively, distinguished from each other by the factors defined above.

[0054] Thus, the measurement system 100 is configured as a so-called automatic measurement system in the sense that it is associated with a production line processing tool arrangement applied to a flow of wafers traveling on a production line (e.g., is integrated with such a tool arrangement). The measurement system comprises main structural components such as a first measurement device 102 configured to apply a first type of measurement (relatively slow measurement, such as XPS measurement) to a measurement site on a wafer and a second optical measurement device 104 configured to apply a second type of measurement (relatively fast measurement, such as optical measurement) to a measurement site. A control system 106 is also provided in the system, which receives and processes first and second measurement data MD1 and MD2 corresponding to the slow and fast measurements, and generates output data indicative of structural parameters, such as related to process control.

[0055] The control system 106 may be a separate computer system connectable (via wired or wireless signal transmission) to the outputs of the measuring devices 102 and 104 for receiving data indicative of / corresponding to the corresponding measurement data therefrom; or may be part of either of the measuring devices 102 and 104; or the data processing means of the computer system 106 may be distributed between these measuring devices. The control system 106 thus typically includes data input and output means 106A, 106B; memory means 106C; and a data processor and analyzer 106D; and may also include appropriate communication / data formatting means.

[0056] A controller 108 is also provided in the system 100 in association with the measuring devices 102 and 104 for operating these measuring devices to implement the predetermined first and second measurement plans relative to the selected positions / locations on the film. The controller 108 may or may not be part of the control system 106 (tool); and / or the software / hardware module of the controller may be part of one or both of the internal controllers of the measuring devices.

[0057] The measurement scheme 200 implemented by the system 100 utilizes a combination of first and second measurement plans, respectively, a so-called "small sampling plan" (or "sparse sampling plan") 202 and a "large sampling plan" (or "dense sampling plan") 204. The implementation of these measurement modes is controlled by the controller 108 as described below, and the corresponding measurement data is processed and analyzed by the corresponding modules / processors of the processor tool 106D.

[0058] The sparse sampling plan 202 is applied by relatively slow high accuracy measurements, which in the present example consist of XPS measurements performed by the measurement device 102. Therefore, N positions (usually several positions, for example N=2, positions L1 and L2) on the thin film layer of the wafer W are selected and XPS measurements are applied to these N positions. The N positions for XPS measurements are preferably set at a relatively low spatial frequency (i.e., a large distance between them). The data MD1 of the XPS measurements are provided and processed to determine one or more parameters of the thin film in each of the N positions, such as thickness and / or specific material concentration. For the dense sampling plan measurement data MD2 (optical data), the parameters of the thin film so determined are used in a "learning" / "training" mode.

[0059] The dense sampling plan 204 is based on optical measurements applied to a relatively large number M (M>>N) of measurement locations / sites on the wafer MS arranged at a relatively high spatial frequency. Preferably, the M locations of the dense sampling plan 204 include the N locations of the sparse sampling plan 202 and an additional K locations, i.e., M=N+K. Data MD2 measured by the dense sampling plan are provided, and one or more desired parameters of the film determined in each of the N locations are used to interpret (by a fitting procedure) the optical measurement data MD2, thereby determining the distribution of one or more film parameters within the structure.

[0060] It should be noted that a "training" mode or sparse sampling scheme can be applied to some similar structures for "learning" optical data interpretation. For example, training using optical measurements such as XPS measurements can be applied to 5 wafers (using 9-point maps in each wafer), and this measurement data MD1 is used to determine the thickness and nitrogen dose (Ndose) parameters in the uppermost layer (thin film) of the wafer. Then, the optical measurements of the entire wafer (e.g., 49-point map) are interpreted.

[0061] The controller 108 is configured to operate the measurement of appropriately selected positions / locations applied to the sparse sampling plan and the dense sampling plan. For example, the second set of M positions selected for the fast measurement mode are substantially evenly distributed on the structure. Moreover, as described above, the M positions measured by the "fast" device 104 are selected to include the N positions measured by the "slow" device 102.

[0062] For the N positions, they are preferably selected so that the minimum and maximum values ​​of the parameter of interest in these positions can be measured.To this end, the controller 108 utilizes certain structure-related data previously provided, indicating the shape of the variation of the parameter of interest on the structure.

[0063] The following are some specific but non-limiting examples of the measurement techniques of the present invention.

[0064] Figures 5A to 5F Measurements on a single film, SiON case, are illustrated as an example. Figure 5A Shown is a design of experiments (DOE) set of 5 wafers measured by XPS over a period of several weeks; Figure 5B A 5-point XPS pattern (N=5) and a 49-point optical pattern (M=49) are shown; Figure 5C HESS-SiON layer thickness (designated as OCD, HESS, and XPS) and nitrogen dose (N_Dose) composition (HESS and XPS) are shown; Figure 5D-5E HESS topography images of five DOE wafers are shown for SiON layer thickness and N-nitrogen (N-Dose) composition; Fig. 5F It is shown that the XPS / XPS_OCD (HESS) accuracy is 10 times better than non-hybrid optical measurement accuracy of SiON stacks.

[0065] In this example, a 5-wafer DOE (Design of Experiments) of SiON thin films deposited on silicon substrates is considered. The thickness and N dose vary between wafers based on process separation. XPS is the reference toolset for SiON. In a series of repeated measurements over several weeks ( Figure 5A , where graph H1 represents the variability fingerprint on the wafer), the film and composition were measured, and the slow outgassing of nitrogen and film growth due to continued slow oxidation at the dielectric / substrate interface was also measured.

[0066] The 9-point map data MD1 measured using the XPS device 102 is combined with the 49-point full-wafer map data MD2 measured using the optical measurement device. Figure 5B Such a diagram is shown in . Figure 5C It is shown that even when the optical properties (related to nitrogen dose) are fixed in the model due to low sensitivity, similar measurements using optical techniques alone (designated as OCD plots) show a lower ability to identify the DOE conditions for thickness. As mentioned above, the term Hybridable Smart Sampling (HESS) as used herein refers to a method of combining / blending low (sparse) sampling measurements (instances from a reference tool) with high (dense) sampling measurements (instances from a workhorse tool) to allow for a cost-effective yet high-performance metrology solution. The HESS method (HESS plot) shows essentially the same accurate measurements of thickness and N-dose components as the XPS method alone.

[0067] The correspondence between HESS and XPS patterns is shown by Figure 5D and 5E This is further confirmed by the topography image of the 5 DOE wafers shown. The image shows circles representing the measured XPS points, which are then extended to the entire wafer by HESS.

[0068] In addition to the unexpected accuracy improvement provided by HESS over standard optical methods, the inventors have estimated the precision improvement to be about 10 times, with HESS having similar precision to XPS (about 0.9% thickness variation and about 0.2% nitrogen dose variation). Fig. 5F These results are shown in : For SiON stacks, the XPS / XPS_OCD (HESS) accuracy is 10 times better than the non-hybrid optical measurement accuracy.

[0069] Figure 6A-6B Measurements in a multilayer thin film stack are illustrated. In this example, the case of an ultra-thin oxide residue on top of a stack of SiN and SiO2 layers is considered. Fig. 6A Results using standard optical measurements are shown with two tool sets showing the differences between the two tools and relative to a reference level XPS measurement. Figure 6B A HESS measurement is shown where both tool sets show a good match to XPS, a correct wafer map, and improved tool matching. It is thus shown that by applying the principles of HESS to both sparse XPS data and dense optical data, not only the overall wafer map is improved, but also the match between the two HESS tool results and the match to the reference is improved. When using reference quality data in hybrid metrology, improvements in measurement performance for multiple attributes can be observed, such as accuracy and matching in this example.

[0070] refer to Fig. 7A and Figure 7B , illustrates measurements on a planar stack containing an ultra-thin high-k (HK) on top of an ultra-thin interlayer (IL) dielectric layer. This example presents the most challenging layer within the RMG stack. The challenge typically faced by the optical toolset is to properly separate between ultra-thin high-k and interlayer dielectric films (small thickness, relatively similar optical properties). This type of measurement is well within the capabilities of XPS, where the photoelectron signals from SiO and HfO are uncorrelated. In an 8-wafer DOE, XPS recognized the DOE intent and correctly quantified the thickness of both layers ( Fig. 7A ). For XPS, using only 5 points per wafer, the HESS method was applied using full-wafer optical measurements and resulted in similar measurement performance while reducing the measurement throughput overhead ( Figure 7BAs shown in the figure, in the 8-wafer DOE, XPS recognizes the DOE intent and correctly quantifies the thickness of HfO2 and IL ( Fig. 7A ); HESS on planar target sampling (XPS for OCD and 5 sites for FWM) yields similar measurement performance to XPS: XPS (sparsely sampled sites) tracks HESS thickness of HfO2 and IL ( Figure 7B ).

[0071] The inventors have shown a variety of applications (SiON, residual oxide, and HK / IL) where the HESS method of the present invention is used to combine sparse / limited XPS sampling with high optical on-wafer sampling to provide a cost-effective and high performance measurement solution.

[0072] Therefore, the present invention provides a new method called the HESS method, which involves a "combined" or hybrid method in which a relatively slow / accurate measurement tool / device and a relatively fast, less accurate tool / device are used. The XPS measurement device is an example of such a slow / accurate tool. The XPS technique itself is capable of measuring the entire wafer map with sufficient performance, but it takes too much time to do this. Therefore, considering the application to online measurement of structures traveling on the production line, the present invention uses XPS as little as possible. The small sample set includes some (in some instances, a minimum of 5) points to collect the minimum and maximum values ​​of the parameters of interest generated by the manufacturing process. For example, if it is known that the cross-wafer variation of a particular parameter of interest is generally U-shaped, the center will be the lowest point and the edge will be the highest point, so the XPS measurement is applied to these points. In the case of an M shape, the center and mid-radius points are selected as the small sample set. No matter what XPS measures, this is a small sample set.

[0073] The second rapid measurement tool is optical, such as a spectroscopic reflectometer. The corresponding spectra (center and edge, or center and mid-radius) are identified in the spectroscopic reflectometry graph (optical measurement data) and the hose spectrum is trained on the XPS values ​​to obtain the trained recipe.

[0074] Then, the entire spectral reflectance map is analyzed and all spectra are interpreted with the trained recipe. For a "small sample" subset of spectra, the same values ​​(as trained) are obtained back, and for the other spectra, a neural network interpolation / extrapolation of the trained values ​​is obtained.

[0075] In the example above, a "small" or sparse sampling plan provides a 5 point (or 9 point) map measured by XPS, and a "large" or dense sampling plan provides a "full wafer map" measured by a spectroscopic reflectometer. The "large map" is analyzed to select dies from it that also have equivalents to the "small map" (for example, if XPS measures 4 points in the center and at the edge, then only the 4 points at the center and the edge are selected from the large map that includes all the dies). The optical data is interpreted with a modeless recipe, for example trained on 5 and interpreted on 65. In other words, this is a "natural extrapolation" (based on spectral data). A recipe is established based on the wafer. It should be noted that this approach can be combined with standard modeless techniques and provides an intermediate way, where training with one wafer may then only require verification and retraining, or another wafer may also be added to the training set, and so on.

[0076] It should also be noted that this technique utilizes "on-the-fly training." In effect, a small sampling plan is applied with a first measurement tool, then a large sampling plan data is acquired with a second measurement tool, and the interpretation of the data from the second measurement tool includes the following: identifying a sample set that matches between the measurements of the first and second tools, training is performed relative to this sample set to create a temporary trained recipe, which is used to interpret the measurement data from the large sample set of the second measurement tool, and the temporary trained recipe is no longer needed. It should be understood that this technique is completely different from standard training, where the DOE is designed, trained, and saved as a "unique" recipe for all other wafers. In contrast, in the present invention, this is a recipe for each wafer obtained through "on-the-fly training."

Claims

1. A method for measuring one or more parameters of a thin film in a structure traveling on a production line, the method comprising: applying X-ray photoelectron spectroscopy (XPS) measurements to a first selected group of a small number of measurement sites in a thin film disposed on a structure and providing data from the XPS measurements; applying the optical measurement to a second set of measurement sites comprising a higher number of measurement sites of the first set and providing optically measured data; processing the data of the XPS measurements and determining a value of at least one parameter of the film in each measurement site of the first set; as well as interpreting the optically measured data using at least one parameter value, thereby obtaining data indicative of a distribution of values ​​of the at least one parameter within the film, the distribution being indicative of the quality of a process applied to the structure in production; The process of interpreting the optically measured data using at least one parameter value comprises: identifying a matching set between measurement sites in the XPS measured data and the optically measured data; and A temporary training recipe is created for the matching set, and the optically measured data is interpreted using the temporary training recipe.

2. A system for in-line measurement of one or more parameters of a thin film in a structure traveling on a production line, the system comprising: A first measuring device, wherein the first measuring device is configured to: performing a first type of measurement according to a first sparse sampling plan, and providing a first measured representation of at least one parameter of a thin film at a first selected set of a small number of measurement sites in the thin film, the thin film being arranged on a structure; A second optical measuring device, wherein the second optical measuring device is configured as: performing a second type of measurement according to a second dense sampling plan, and providing second measurement data representative of at least one parameter of the film in a second set of a higher number of measurement sites; as well as A control system, the control system being configured to: processing the first measurement data and determining at least one value of at least one parameter of the thin film in each measurement site of the first set; and interpreting second measurement data using at least one parameter value and obtaining data indicative of a distribution of values ​​of the at least one parameter within the second set of measurement sites; Wherein, in order to interpret the second measurement data using the at least one parameter value, the control system is configured as follows: identifying a matching set between measurement sites in the first measurement data and the second measurement data; and A temporary training recipe is created for the matching set, and the second measurement data is interpreted using the temporary training recipe.

3. The system according to claim 2, comprising: A controller configured to: for selected measurement sites, control the first measurement device and the second measurement device to perform the measurements according to the sparse sampling plan and the dense sampling plan, respectively, wherein the measurement sites of the second group are selected to be substantially uniformly distributed within the film being measured.

4. The system according to claim 3, wherein: The controller is configured to select the second set of measurement sites that includes the first set of measurement sites.

5. The system according to claim 3, wherein: The controller is configured to select measurement sites of the first group so that a minimum value and a maximum value of the at least one parameter in the selected measurement sites can be measured.

6. The system according to claim 5, wherein: The controller is configured to select the first set of measurement sites using data of a changing shape of at least one parameter of the thin film across the structure.

7. The system according to claim 2, wherein: The first and second measurement devices differ from each other at least in the time required to collect and interpret measurement data, the first type of measurement being a relatively slow measurement and the second type of measurement being a relatively fast measurement.

8. The system according to claim 7, wherein: The first type of measurement performed by the first measuring device is characterized by a higher accuracy compared to the second type of measurement.

9. The system according to claim 2, wherein: The first type of measurement device is configured to perform X-ray Photoelectron Spectroscopy (XPS) measurements.

10. The system according to claim 2, wherein: The second type of measuring devices includes spectroscopic reflectometers.

11. The system according to claim 2, wherein: The at least one parameter of the thin film includes a thickness of the thin film.

12. The system according to claim 2, wherein: The structure includes a semiconductor wafer.

Images