Measuring local critical dimension uniformity of an array of two-dimensional structure elements

Abstract

A method, system, and non-transitory computer readable medium for measuring local critical dimension uniformity of an array of two-dimensional structure elements, the method can include obtaining an optical spectrographic analysis spectrum of the acquired array; feeding the optical spectrographic analysis spectrum of the acquired array to a trained machine learning process, wherein the trained machine learning process is trained to map the optical spectrographic analysis spectrum to an average critical dimension (CD) and a local critical dimension uniformity (LCDU); and outputting, by the trained machine learning process, the average CD and the LCDU of the array.

Description

[0001] Cross-references

[0002] This application claims priority to U.S. Provisional Patent No. 62 / 980,335, filed February 23, 2020, which is incorporated herein by reference. Background Technology

[0003] Semiconductor devices are manufactured using highly complex processes, including a photolithography stage during which a photoresist layer is exposed to extreme ultraviolet (EUV) radiation to form a pattern within the photoresist. This patterned photoresist can then be used as an etching mask to form nanoscale structural elements.

[0004] A two-dimensional array of structural elements (such as vias) should consist of identical structural elements. However, the photolithography stage (and any other manufacturing stage, such as etching) is not ideal, and the structural elements may be misaligned.

[0005] Local critical size uniformity (LCDU) represents the variation in critical size (CD) between different structural elements and is three times the standard deviation of the average CD of the structural elements.

[0006] LCDU can significantly impact manufacturing yield and should be closely monitored.

[0007] LCDU can be measured using a CD scanning electron microscope (CD-SEM) to measure the CD of individual structural elements and then estimate the LCDU. This process is very lengthy and in practice limited to measuring a few structural elements at high magnification—this can reduce the accuracy of LCDU measurements and may also damage the sample being tested.

[0008] There is an increasing need to provide a safe and accurate method for measuring arrays of two-dimensional structural elements in LCDUs. Summary of the Invention

[0009] A non-transitory computer-readable medium may be provided for a system, method, and an LCDU for storing instructions for measuring an array of two-dimensional (2D) structural elements. Attached Figure Description

[0010] To understand the invention and how it can be practiced, preferred embodiments will now be described by way of non-limiting example only with reference to the accompanying drawings.

[0011] Figure 1 An example of one method is shown;

[0012] Figure 2 An example of one method is shown;

[0013] Figure 3 An example of a portion of the system and via array is shown;

[0014] Figure 4 An example of an array of through holes is shown, and sixteen through holes are also shown;

[0015] Figure 5 It shows different arrays, optical spectral analysis spectra, various average CD measurements and LCDU measurements, processors implementing machine learning processes, and mappings generated by machine learning processes. Detailed Implementation

[0016] Numerous specific details are set forth in the following detailed description to provide a thorough understanding of the invention. However, those skilled in the art will understand that the invention can be practiced without these specific details. In other instances, well-known methods, processes, and components have not been described in detail so as not to obscure the invention.

[0017] The concluding section of this specification specifically points out and explicitly claims protection for the subject matter considered to be in this invention. However, the organization and operation of the invention, as well as its objects, features, and advantages, are best understood by referring to the following detailed description in conjunction with the accompanying drawings.

[0018] It is understood that, for the sake of simplicity and clarity, the elements shown in the figures are not necessarily drawn to scale. For example, for clarity, the dimensions of some elements may be enlarged relative to others. Furthermore, where deemed appropriate, reference numerals may be repeated in the figures to indicate corresponding or similar elements.

[0019] Any reference in this specification to any system, method, or non-transitory computer-readable medium shall be modified as necessary to apply to any other system, method, or non-transitory computer-readable medium. For example, any reference to a system shall be modified as necessary to apply to methods executable by that system and to non-transitory computer-readable media that can store instructions executable by that system.

[0020] Since at least one embodiment of the present invention can be largely implemented using electronic components and circuits known to those skilled in the art, details will be explained only to the extent deemed necessary by the foregoing in order to understand and evaluate the basic concepts of the invention and to avoid obscuring or departing from the teachings of the invention.

[0021] Any numbers or values ​​shown below should be considered as non-limiting examples.

[0022] The phrase "A based on B" can mean that A is based solely on B, or that A is based on B and one or more other components and / or parameters and / or information. "Based on" implies that the calculation of A is influenced by B and / or that the value of A is a function of the value of B.

[0023] The term "acquire" can include generating, receiving, etc. For example, receiving the optical spectral analysis spectrum of an array of two-dimensional (2D) structural elements, or generating the optical spectral analysis spectrum of an array (e.g., by irradiating the array with broadband radiation, collecting the signal emitted from the array due to the irradiation, and processing the signal).

[0024] The specification and / or drawings may relate to a processor. A processor may be processing circuitry, a portion of processing circuitry, a virtual machine core, etc. Processing circuitry may be implemented as a central processing unit (CPU), a graphics processing unit (GPU), and / or one or more other integrated circuits, such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), fully custom integrated circuits, etc., or combinations of these integrated circuits.

[0025] Any combination of any steps of any method shown in the instruction manual and / or accompanying drawings may be provided.

[0026] Any combination of any subject matter that may be provided for any claim.

[0027] Any combination of systems, units, components, processors, and sensors shown in the specification and / or drawings may be provided.

[0028] Some examples mentioned in the specification and accompanying drawings involve through-holes. Through-holes are non-limiting examples of 2D structural elements. Circuitry can be considered as a one-dimensional structural element.

[0029] A fast and accurate method can be provided for measuring an array of two-dimensional structural elements in an LCDU.

[0030] The method can be used for automated process control (APC) during high-volume manufacturing (HVM) for manufacturing processes that may include EUV lithography and / or ETCH processes.

[0031] The method may include using a machine learning process trained to map the optical spectral analysis spectrum of an array of 2D structural elements to the average CD and LCDU values ​​of the array.

[0032] The training process may include obtaining average CD and LCDU values ​​for different arrays as measured by CD-SEM. Different arrays may vary from one another based on their manufacturing process conditions (e.g., differences in irradiation conditions such as dose, focus, irradiation direction, and / or any other manufacturing process conditions) in order to provide an effective mapping for arrays manufactured under various manufacturing process conditions.

[0033] For each array, the measurement should include measuring the CD of a significant number (e.g., at least 100, 150, 200, 300, 400, etc.) of the 2D structural elements of the array. This significant number exceeds the number needed to determine only the average CD – because measurements are required to capture CD values ​​spanning an integer three times the standard deviation of the average CD of the structural elements.

[0034] Measurement of the CD of a significant number of 2D structural elements makes up for the gap between the number of 2D structural elements measured by CD-SEM (within a reasonable time period) (usually only a few) and the huge number of 2D structural elements (even hundreds, thousands, tens of thousands and more) whose CD is measured by optical spectral analysis.

[0035] Before calculating the average CD and LCDU, the CD values ​​of structural elements can be filtered (e.g., by performing outlier removal).

[0036] Outliers can be irregular contours (e.g., if the contour of a via is circular, then a contour that deviates from the circle can be considered to have an irregular contour), and vias merged in CD-SEM images.

[0037] Outlier removal significantly improves the accuracy of the method—because LCDU is highly sensitive to outliers. In statistics, an outlier is a data point that is significantly different from other observations. What constitutes an outlier and / or what causes a significant difference can be determined and / or defined in any way.

[0038] The machine learning process was trained to build a model based on the average CD and LCDU values ​​of different arrays measured by CD-SEM and the optical spectral analysis spectra of different arrays.

[0039] The training process can be a supervised training process, in which optical spectral analysis spectra of different arrays are fed as inputs to the machine learning process, while the average CD and LCDU values ​​are provided as the expected outputs of the machine learning process.

[0040] The training process can be performed offline and / or on a Design of Experiments (DOE) wafer.

[0041] Optical spectral analysis spectra can be obtained via HVM, whether in standalone (SA) or integrated metrology (IM) systems.

[0042] After training the machine learning process (to provide the trained machine learning process), the average CD and LCDU of the measurement array of the 2D structural element can be determined by: (a) obtaining the optical spectral analysis spectrum of the acquired measurement array of the 2D structural element, (b) feeding the optical spectral analysis spectrum of the measurement array of the 2D structural element to the trained machine learning process, and (c) receiving the average CD and LCDU of the measurement array of the 2D structural element from the trained machine learning process.

[0043] The proposed method is expected to be more accurate and faster than known model-based optical spectral analysis methods that require calculating high-precision models of the measurement array of 2D structural elements—because such models are difficult to compute and highly sensitive to the (unknown) positions of the 2D structural elements with different CD values ​​within the array. For example, assuming the array comprises N rows and M columns of vias, and the vias have K different CD values, the positions of these K vias with different CD values ​​can affect the output of model-based optical spectral analysis.

[0044] The following methods can be performed by a computer or a spectrometer, a system that may include a spectrometer, a system with irradiation and collection paths, etc.

[0045] Figure 1 An example of a method 100 for measuring an array of 2D structural elements is shown.

[0046] Method 100 can begin with step 110 to obtain the optical spectral analysis spectrum of the array.

[0047] Step 110 may be followed by step 120, in which the acquired optical spectral analysis spectrum of the array is fed into a trained machine learning process, wherein the trained machine learning process is trained to map the optical spectral analysis spectrum to average critical size (CD) and local critical size uniformity (LCDU) values.

[0048] Step 120 can be followed by step 130, which outputs the average CD and LCDU of the array through a trained machine learning process.

[0049] Figure 2 The method for training a machine learning process is explained.160

[0050] Method 160 can begin with steps 170 and 180.

[0051] Step 170 may include obtaining optical spectral analysis spectra of different arrays.

[0052] Step 180 may include obtaining average CD and LCDU values ​​for different arrays, the average CD and LCDU values ​​being measured by CD scanning electron microscopy (CD-SEM).

[0053] After removing abnormal CD values ​​from the array, the array average CD and LCDU values ​​of different arrays can be calculated.

[0054] After steps 170 and 180, step 190 can be performed to establish a mapping (or a model of the mapping) between the optical spectral analysis spectrum and the average CD and LCDU values ​​through a machine learning process.

[0055] Step 190 may include multiple iterations, each iteration may include:

[0056] a. Select one of the multiple arrays.

[0057] b. Feed the optical spectral analysis spectrum of the array into the machine learning process.

[0058] c. Generate machine learning process outputs related to the array.

[0059] d. Compare the output of the machine learning process associated with the array with the average CD and LCDU values ​​of the array (or apply a cost or error function). The average CD and LCDU values ​​are measured by CD scanning electron microscopy (CD-SEM).

[0060] e. Configure the machine learning process based on the results of step (c) and jump to step (a) to select a new array.

[0061] The optical spectral analysis spectrum of the array can be fed as input to the machine learning process, and the average CD and LCDU values ​​can be provided as the desired output of the machine learning process.

[0062] LCDU accuracy may require obtaining the average CD and LCDU of an array with a significant number of 2D structural elements.

[0063] Figure 3 Approximately thirty through holes are shown in the system 100, sample 90, and through-hole array portion 200.

[0064] System 100 may be a spectrometer and includes an irradiation path that irradiates sample 90 with broadband irradiation (which may include radiation of different frequencies or different “colors”) – as seen from different types of arrows emitted from broadband source 102, passing through irradiation optics (e.g., lenses 104 and 106) and impacting an array of vias (see array section 200 – which includes vias 202(1,1)-202(5,6)).

[0065] Broadband radiation is reflected and / or returned and / or scattered from the array by collecting optical devices (such as lenses 112 and 114) onto color separator 116, which separates different colors so that different colors are represented by different pixels of the radiation impact detector 120.

[0066] Non-limiting examples of spectrometers may include, for example, spectrometers manufactured by Nova Measuring Instruments Ltd., and may be (or may be included in) (for example) integrated metrology systems such as ASTERA, NOVA i570, NOVA i550, NOVA i500, etc., and / or standalone systems: NOVA T500, NOVA T550, NOVA T600, etc. An example of a spectrometer is shown in U.S. Patent No. 6,657,736, which is incorporated herein by reference.

[0067] The detection signal forms an optical spectral analysis spectrum (or is processed to provide an optical spectral analysis spectrum), which is fed to processor 122, which can implement a machine learning process to provide the average CD and LCDU of the array.

[0068] Figure 4 An example of an array 211 of through holes is shown - in which 16 through holes (4×4 through holes) are shown and represented as 211(5,5)-211(5,8), 211(6,5)-211(6,8), 211(7,5)-211(7,8) and 211(8,5)-211(8,8).

[0069] Vias 211(5,6) and 211(7,7) are outliers - and if array 211 is used to train a machine learning process - their CDs are ignored when calculating the average CD and LCDU of array 211.

[0070] exist Figure 4 In this context, the CD of a through hole is the average value between the x-axis CD of the through hole (e.g., CDx 231(5,6), CDx231(5,8), and CDx 231(7,7)) and the y-axis CD of the through hole (e.g., CDy 232(5,6), CDy 232(5,8), and CDy 232(7,7)).

[0071] Figure 5 The training process is shown, during which the machine learning process 220 establishes a mapping 222 based on N optical spectra 214(1)-214(N) of N arrays 211(1)-211(N) and N average CDs 212(1)-212(N) and N LCDUs 213(1)-213(N) of N arrays 211(1)-211(N).

[0072] Any components used to achieve the same function are arranged to be effectively "associated" in order to achieve the desired functionality. Therefore, any two components combined here to achieve a particular function can be considered "associated" with each other to achieve the desired functionality, regardless of the architecture or intermediate components. Similarly, any two components so associating can also be considered "operably connected" or "operably coupled" with each other to achieve the desired functionality.

[0073] Furthermore, those skilled in the art will recognize that the boundaries between the above operations are merely illustrative. Multiple operations may be combined into a single operation; a single operation may be distributed across other operations, and operations may be performed with at least partial overlap in time. Moreover, alternative implementations may include multiple instances of the operations, and the order of the operations may be varied in various other implementations.

[0074] Furthermore, for example, in one implementation, the illustrated examples can be implemented as circuitry located on a single integrated circuit or within the same device. Alternatively, these examples can be implemented as any number of separate integrated circuits or separate devices interconnected with each other in a suitable manner.

[0075] Furthermore, for example, the example or a portion thereof may be implemented as a software or code representation of a physical circuit, or may be converted into a logical representation of a physical circuit, for example, implemented in any suitable type of hardware description language.

[0076] However, other modifications, changes, and substitutions are also possible. Therefore, the specification and drawings are considered illustrative rather than restrictive.

[0077] In the claims, any reference symbols in parentheses should not be construed as limiting the claims. The word “comprising” does not exclude the presence of other elements or steps besides those listed in the claims. Furthermore, the terms “a” or “an” as used herein are defined as one or more. Moreover, the use of introductory phrases such as “at least one” and “one or more” in the claims should not be construed as implying that the introduction of another claim word by the indefinite article “a” or “an” limits any particular claim word containing such an introduction to include only one such invention, even if the same claim contains the introductory phrase “one or more” or “at least one” and the indefinite article, such as “a” or “an.” The same applies to the use of definite articles. Unless otherwise stated, terms such as “first” and “second” are used to arbitrarily distinguish the elements described by these terms. Therefore, these terms are not necessarily intended to indicate the time or other priority of these elements. The fact that certain measures are recited in mutually different claims does not mean that combinations of these measures cannot be advantageously used.

[0078] While certain features of the invention have been described and illustrated herein, many modifications, substitutions, variations, and equivalents will now occur to those skilled in the art. Therefore, it should be understood that the appended claims are intended to cover all such modifications and variations falling within the true spirit of the invention.

[0079] The terms “comprising,” “including,” “having,” “forming,” and “consistently composed of” are interchangeable. For example, any method may include at least the steps included in the drawings and / or specification, or may include only the steps included in the drawings and / or specification.

Claims

1. A method for measuring the local critical dimension uniformity of an array of two-dimensional structural elements, the method comprising: Obtain the optical spectral analysis spectrum of the acquired array, wherein the two-dimensional structural element is a through hole; The acquired optical spectral analysis spectra of the arrays are fed into a trained machine learning process, wherein the trained machine learning process is trained to map the optical spectral analysis spectra to an average critical size CD and a local critical size uniformity LCDU; wherein the trained machine learning process is trained by feeding the following: a) optical spectral analysis spectra of different arrays; b) the average CD and LCDU of different arrays, the average CD and LCDU being measured by CD scanning electron microscopy (CD-SEM); and wherein each of the different arrays includes at least a defined number of vias, the defined number being determined based on three times the standard deviation of the average critical size of the vias of the array; and The average CD and LCDU of the array are output through the trained machine learning process.

2. The method according to claim 1, wherein, The average CD and LCDU of the different arrays are calculated after removing abnormal CD values ​​of the arrays.

3. The method according to claim 1, wherein, The optical spectral analysis spectrum is the input to the machine learning process, and the average CD and LCDU of the different arrays are the expected outputs of the machine learning process.

4. The method according to claim 1, wherein, Each of the different arrays includes at least one hundred and fifty two-dimensional structural elements.

5. The method according to claim 1, wherein, Each of the different arrays includes at least four hundred two-dimensional structural elements.

6. The method of claim 1, further comprising training the machine learning process to map the optical spectral analysis spectrum to average CD and uniformity LCDU values.

7. The method according to claim 1, wherein, Obtaining the optical spectral analysis spectrum of the acquired array involves irradiating the array with broadband radiation and collecting the signals emitted from the array as a result of the irradiation.

8. A non-transitory computer-readable medium for measuring the local critical dimensional uniformity of an array of two-dimensional structural elements, the non-transitory computer-readable medium storing instructions for: Obtain the optical spectral analysis spectrum of the acquired array, wherein, The two-dimensional structural element is a through hole; The acquired optical spectral analysis spectra of the arrays are fed into a trained machine learning process, wherein the trained machine learning process is trained to map the optical spectral analysis spectra to average critical size CD and local critical size uniformity LCDU values; wherein the trained machine learning process is trained by feeding the following: a) optical spectral analysis spectra of different arrays; b) average CD and LCDU of different arrays, the average CD and LCDU being measured by CD scanning electron microscopy (CD-SEM); and wherein each of the different arrays includes at least a defined number of vias, the defined number being determined based on three times the standard deviation of the average critical size of the vias of the array; and The average CD and LCDU of the array are output through the trained machine learning process.

9. The non-transitory computer-readable medium according to claim 8, wherein, The average CD and LCDU of the different arrays are calculated after removing abnormal CD values ​​of the arrays.

10. The non-transitory computer-readable medium according to claim 8, wherein, The optical spectral analysis spectrum is the input to the machine learning process, and the average CD and LCDU of the different arrays are the expected outputs of the machine learning process.

11. The non-transitory computer-readable medium according to claim 8, wherein, Each of the different arrays includes at least one hundred and fifty two-dimensional structural elements.

12. The non-transitory computer-readable medium according to claim 8, wherein, Each of the different arrays includes at least four hundred two-dimensional structural elements.

13. The non-transitory computer-readable medium of claim 8, storing instructions for training the machine learning process to map the optical spectral analysis spectrum to average CD and uniformity LCDU values.

14. The non-transitory computer-readable medium according to claim 8, wherein, The acquired optical spectral analysis spectral storage is used to irradiate the array with broadband radiation and collect the signals emitted from the array due to the irradiation.

15. A system for measuring the local critical dimensional uniformity of an array of two-dimensional structural elements, the system comprising a processor configured to: Receive the acquired optical spectral analysis spectrum of the array, wherein, The two-dimensional structural element is a through hole; The acquired optical spectral analysis spectra of the arrays are fed into a trained machine learning process, wherein the trained machine learning process is trained to map the optical spectral analysis spectra to an average critical size (CD) and a local critical size uniformity (LCDU); wherein the trained machine learning process is trained by feeding the following: a) optical spectral analysis spectra of different arrays; b) the average CD and LCDU of different arrays, the average CD and LCDU being measured by CD scanning electron microscopy (CD-SEM); and wherein each of the different arrays includes at least a defined number of vias, the defined number being determined based on three times the standard deviation of the average critical size of the vias of the array; The average CD and LCDU of the array are output through the trained machine learning process.

Images