Wafer plane detection of lithographically significant contamination photomask defects

Wafer plane detection for significant etching contamination of photomask defects

This application is a partial continuation of the US patent application No. 11/622,432 filed by Wihl et al. on 01/11/2007 entitled "METHOD FOR DETECTING LITHOGRAPHICALLY SIGNIFICANT DEFECTS ON RETICLES", the US patent application It is incorporated herein by reference in its entirety for various purposes.

As the density and complexity of integrated circuits (ICs) continue to increase, detecting photomask patterns has gradually become more challenging. Each new generation of ICs has denser and more complex patterns that currently meet and exceed the optical limits of etching systems. To overcome these limitations, various resolution enhancement techniques (RET) such as optical proximity correction (OPC) have been introduced. For example, OPC helps to overcome some diffraction limitations by modifying the mask pattern so that the resulting printed pattern corresponds to the original desired pattern. Such modifications may include perturbation of the size and edges of the main IC features (ie, printable features). Other modifications involve adding serifs to the corners of the pattern and/or providing nearby sub-resolution assist features (SRAF) that are not expected to produce printed features and are therefore referred to as non-printable features. These non-printable features are expected to offset pattern disturbances that would otherwise occur during the printing process.

The addition of non-printable features to a mask with a dense population of printable features further complicates mask detection. Moreover, non-printable features and printable features have different effects on printed patterns. Therefore, the defects corresponding to these features will have different etching significance, and some of these defects can be ignored during inspection.

Provide a novel method and system for inspecting photomasks to identify significant etching contamination defects. The detection can be performed without having a separate reference image provided from a database or another die. The inspection techniques described herein involve capturing one or more test images of a photomask and using specific etching and/or resist models to construct corresponding test "simulated" images. These test simulation images simulate the printable and/or resist patterns of the tested photomask. Moreover, these initial test images are used in parallel operations to generate "composite" images. These images represent a defect-free mask pattern. The synthesized images are then used to generate reference simulation images similar to the test simulation images but without the significant contamination defects of the etching. Finally, compare the reference simulation images with the test simulation images to identify the significant etching contamination defects on the photomask.

In some embodiments, a method for detecting a photomask to identify significant etch contamination defects involves providing a photomask with one or more printable features and one or more non-printable features, generating a test transmission image, and A test reflection image, and a test simulation image is constructed by applying an etching system model to the test images. An example of a non-printable feature includes the sub-resolution assist feature (SRAF). In certain embodiments, at least one non-printable feature is larger than at least one printable feature. In some embodiments, the test transmission image is aligned with the test reflection image. The method also involves constructing synthetic reference images from the test images, and the synthetic reference images are used to construct reference simulation images from the synthetic reference images by applying the etching system model. The synthetic reference images are constructed by removing defects from the test images, which in turn are transformed into reference simulation images without the significant etched contamination defects. The method then starts by comparing the test simulation images with the reference simulation images in order to identify the significant etching defects. The application of the etching system model makes it possible to capture the etching effect corresponding to the non-printable features in the test simulation image and the reference simulation image. This detection method can be implemented without using a reference die or a reference database (ie, a separate reference image provided from the reference die or the reference database).

In some embodiments, an etching system model includes one or more of the following parameters: the numerical aperture of the etching system and/or the detection device, the wavelength of the etching system and/or the detection device, and the etching system And/or the lighting conditions of the testing equipment. In some embodiments, an inspection method also involves providing a model of a photoresist system to be used in the etching transfer and using the model to construct the test simulation images and the reference simulation images. The photoresist system model can include various resist formation process characteristics. In some embodiments, a method involves classifying the etching significant contamination defects.

In some embodiments, the reference simulation image includes a reference simulation transmission image and a reference simulation reflection image, and the test simulation image includes a test simulation transmission image and a test simulation reflection image. In these embodiments, identifying the significant etching contamination defects may involve comparing the reference simulated transmission image with the test simulated transmission image and comparing the reference simulated reflection image with the test simulated reflection image.

In some embodiments, constructing a test simulation image involves constructing a limited bandwidth mask pattern that includes only one linear term. This simplification of the mask pattern function can help speed up the detection process. However, if additional accuracy is required, a limited bandwidth mask pattern can also include non-linear (quadratic) terms. Constructing a test simulation image may involve separating printable and non-printable features based on the light intensity distribution of the limited bandwidth mask pattern. In some embodiments, a method also involves constructing a geometric map based on the limited bandwidth mask pattern to classify geometric features into the following categories: edges, corners, line ends, and other features. The geometric map can be used to classify geometric features while identifying the significant etching contamination defects. Moreover, identifying significant etching contamination defects may involve applying different detection thresholds for at least two different geometric feature types of the geometric image.

In some embodiments, a method also involves constructing a feature map based on the test simulation images. The feature map may include a plurality of image parts each having a corresponding mask error enhancement factor (MEEF). The MEEF can then be used, for example, to automatically adjust the detection threshold of each image part based on the corresponding MEEF value during the identification of the significant defects of etching contamination. In some embodiments, a process also involves providing a user-defined detection threshold for identifying one of the significant etch contamination defects.

In some embodiments, a system for detecting a photomask to identify significant etching contamination defects is provided. The system includes at least one memory and at least one processor. The at least one memory and the at least one processor are assembled State to perform the following operations: generate test images of the mask, construct test simulation images by applying an etching system model to the test images, construct synthetic reference images from the test images by removing defects from the test images , Constructing a reference simulation image by applying the model of the etching system to the synthetic reference images, and comparing the test simulation images with the reference simulation images to identify the significant etching contamination defects.

These and other aspects of the present invention will be further described below with reference to the drawings.

In the following description, a large number of specific details are listed in order to provide a thorough understanding of the present invention. The present invention can be implemented without some or all of these specific details. In other cases, the conventional process operations are not described in detail so as not to unnecessarily obscure the present invention. Although the present invention will be described in conjunction with specific embodiments, it should be understood that the present invention is not intended to be limited to these embodiments.

Introduction The novel methods and systems described in this article are used to inspect photomasks to identify significant etching defects. Inspection involves capturing one or more test images of the inspected mask and using these images to develop at least two separate simulated images, one of which does not have significant etching defects and is used as a reference. These simulated images consider various characteristics of the etching system and in some embodiments the characteristics of the photoresist material. This model-based method focuses on etching significant defects and ignores many other defects that are not printed during the etching process and are not significant defects.

In some embodiments, a specially configured inspection system is used to capture two or more light intensity images of the inspected mask. Multiple test images represent different optical characteristics of the photomask, for example, transmission and reflection images acquired at different wavelengths. These images are then aligned and processed together to restore a "limited bandwidth mask pattern" which is a functional representation of the test image, which in some embodiments contains only linear terms. Then, an etching system model containing various characteristics of a stepper and/or scanner is used to generate a simulated test image, such as an aerial image or a photoresist image. In a set of parallel operations, the test images are processed to eliminate any defects from the images. The resulting image is called a "composite" image. Similarly, the etching system models are applied to the synthetic images and the test images to generate synthetic simulation images, and the synthetic simulation images are compared with the test simulation images to obtain various defect information. In particular, since the etching system model is applied to both the test image and the reference image, the defect information relates to significant etching defects.

There is no need to obtain a separate reference image from a die-to-die or database. An example of a photomask inspection system is Starlight available from KLA Tencor in Milpitas, Canada.^TMDetection Systems. This system is designed to capture high-resolution transmission and reflection images of a mask and combine the two images to generate a composite reference mask image. The composite reference image has no defects (for example, a contamination pattern), even when the actual photomask has defects. The composite reference mask image is then used to generate composite reference transmission and reflection images, and the composite reference transmission and reflection images are compared with the captured test images to determine defects. Although this system can be successfully used to identify various photomask defects, its purpose is not to consider wafer printability defects or to focus particularly on etching significant defects. In other words, this system does not consider the defects of etching and printing. As mentioned above, not all defects have an impact on the resulting print pattern, and therefore, some of these defects can and should be ignored. For example, a contamination defect in a molybdenum-silicon area of a flat field has almost no effect on printable wafers. Likewise, many OPC feature defects have no or little etching significance. However, it is difficult and often impossible for inspectors to distinguish between etched significant defects and etched insignificant defects. Modern etching systems use advanced stepper illumination distribution and rely on the partially coherent nature of imaging behavior, making it a very difficult task for inspectors to manually distinguish defects.

The novel methods and systems described herein automatically consider the printability (or non-printability) characteristics provided in the etching and/or photoresist system model. Moreover, some of the proposed detection techniques do not require a separate reference image to be provided from a database or another die. In a production environment, this provides a significant advantage because a separate reference image is usually not available. Finally, some of the proposed methods implement automatic mask error enhancement factor (MEEF) enhancement in areas with high MEEF. Now we need to further explain some inventive aspects based on the detection system and technology.

<b>Examples of detection systems</b>

1A is a simplified schematic diagram of a typical etching system 100 according to some embodiments. The typical etching system can be used to transfer a mask pattern from a mask M to a wafer W. As shown in FIG. Examples of these systems include scanners and steppers, and more specifically one of the TWINSCAN systems available from ASML of Veldhoven, the Netherlands (for example, TWINSCAN NXT:1950i step-and-scan system). Generally speaking, an illumination source 103 guides a light beam through an illumination lens 105 to a mask M located in a mask plane 102. The illumination lens 105 has a numerical aperture 101 on that plane 102. The value of the numerical aperture 101 affects which defects on the photomask are etched notable defects and which are not etched notable defects. A part of the light beam passing through the mask M forms a patterned light signal, which is guided through the imaging optics 153 and onto a wafer W to initiate pattern transfer.

FIG. 1B provides a schematic diagram of a detection system 150 according to some embodiments. The detection system has an imaging lens 151 a with a relatively large numerical aperture 151 b on a main reticle plane 152. The illustrated inspection system 150 includes a microscope magnification optic 153 designed to provide, for example, 60 to 200X magnification for enhanced inspection. The numerical aperture 151b on the main mask plane 152 of the inspection system is usually much larger than the numerical aperture 101 on the main mask plane 102 of the etching system 100, which often results in a difference between the test detection image and the actual printed image. Each of these optical systems (100, 150) induces different optical defects in the generated images that are considered and compensated for in the novel detection techniques described herein.

The novel detection technique can be implemented on various specially configured detection systems (such as the specially configured detection system schematically illustrated in FIG. 1B). The system 150 includes an illumination source 160 that generates a light beam, which is guided through the illumination optics 161 to a mask M in the main mask plane 152. Examples of light sources include lasers or filtered lamps. In one instance, the source is a 193 nm laser. As explained below, the detection system 150 has a numerical aperture 151a on the main mask plane 152, which may be larger than the numerical aperture of a main mask plane of the corresponding etching system (e.g., element 101 in FIG. 1A). The mask M to be detected is placed on the main mask plane 152 and exposed to the source. The patterned image from the mask M is guided through a set of magnifying optical elements 153, and the set of magnifying optical elements project the patterned image onto a sensor 154. The signal captured by the sensor 154 can be processed by a computer system 173 or more commonly by a signal processing device, which can include an analog-to-digital converter configured to convert from The analog signal of the sensor 154 is converted into a digital signal for processing. The computer system 173 can be configured to analyze the intensity, phase, and/or other characteristics of the sensed beam. The computer system 173 can be configured (for example, with programmed commands) to provide a user interface (for example, on a computer screen) for displaying the obtained test images and other detection features. The computer system 173 may also include one or more input devices (for example, a keyboard, a mouse, and/or a joystick) for providing user input such as a constantly changing detection threshold. In some embodiments, the computer system 173 is configured to perform the detection techniques described in detail below. The computer system 173 usually has one or more processors coupled to an input/output port and one or more memories via a suitable bus or other communication mechanism.

Therefore, such information and program instructions can be implemented on a specially configured computer system. Therefore, this system includes program instructions/computer code that can be stored on a computer-readable medium for performing various operations described herein. Examples of machine-readable media include (but are not limited to): magnetic media (such as hard disks, floppy disks, and tapes); optical media (such as CD-ROM disks); magneto-optical media (such as optical disks); It is a hardware device (such as read-only memory device (ROM) and random access memory (RAM)) that stores and executes program instructions. Examples of program instructions include both machine code generated by a compiler and files containing higher-level code that can be executed by the computer using an interpreter. In some embodiments, a system for detecting a photomask includes at least one memory and at least one processor, and the at least one memory and the at least one processor are configured to perform the following operations: generating includes a test A test image of a transmission image and a test reflection image, a test simulation image is constructed by applying a model of the etching system to the test images, a synthetic reference image is constructed from the test images by removing defects from the test images, Constructing reference simulation images of the synthetic reference images by applying the model of the etching system to the synthetic reference images, and comparing the test simulation images with the reference simulation images to identify the significant etching defects. Examples of inspection systems include a specially configured TeraScan^TM The DUV inspection system and a specially configured Teron 600 series main mask defect inspection system are available from KLA-Tencor in Milpitas, California.

<b>Check</b><b>Examples of test methods</b>

FIG. 2 illustrates a process flow diagram corresponding to an example of a method for inspecting a photomask to identify significant defects in etching. This method can be used to detect various types of photomasks including one or more printable features and one or more non-printable features. For example, a photomask made of a transparent fused silica blank with a pattern defined by a metallic chromium adsorption film can be used. Generally speaking, this method can be used to inspect various memory substrates such as main photomasks, photomasks, semiconductor wafers, phase shift masks, and embedded phase shift masks (EPSM). As mentioned above, non-printable features may include various optical proximity correction (OPC) features for compensating for imaging errors caused by diffraction. One type of these features is the Sub-Resolution Auxiliary Feature (SRAF). In certain embodiments, at least one non-printable feature is larger than at least one printable feature.

Once the photomask is provided for the inspection process, for example, the photomask is placed on a testing platform of the inspection system, the process can begin at 202 by generating one or more test images of the photomask. For example, the mask can be illuminated to capture two or more light intensity images under different lighting and/or collection conditions. In a specific embodiment, a transmitted light intensity image and a reflected light intensity image are captured. In other embodiments, two or more reflection images or two or more transmission images are generated to simultaneously illuminate the mask with different wavelengths. For example, if the photomask material causes the transmission to be a strong function of the wavelength of the illuminating light, two different but close wavelengths can be used to generate a pair of transmission images each with a different transmission level.

The captured test images are generally aligned in operation 204. This alignment may involve matching the optical properties of the detection system to multiple test images. For example, in the case of transmission and reflection images, a certain adjustment can be made to these images to compensate for the difference in the optical paths of two separate signals. The alignment adjustment may depend on the specific geometry of one of the detection systems used.

Once aligned, the test images can be processed at 206 to recover a limited bandwidth mask amplitude function. This function is sometimes called a limited bandwidth mask pattern. In one method, the partially coherent optical imaging can be modeled as the sum of two or more coherent systems. This is explained in more detail in U.S. Patent Application No. 11/622,432, which is based on The purpose of explaining operation 206 is incorporated herein by reference. In particular, the Hopkins equation for partially coherent imaging can be used to form a transmission cross coefficient (TCC) matrix. Then, this matrix can be decomposed into corresponding eigenvectors that serve as the core of the coherent system. The weighted sum of the eigenvalues from the intensity base values of each of the iso-modulation systems produces an image intensity that can be used to represent the intensity of the transmission signal. In some embodiments, a linear term called only the amplitude function of the limited bandwidth mask can be used to represent the reflection and transmission intensities of the test images. In some embodiments, a limited bandwidth mask pattern also includes non-linear terms that specifically require additional accuracy. An example of this function is presented in Equation 1.

<maths><img file="TW201218293A_D0001.tif" /></maths>

in<i>a</i>_<i>R</i>It is the complex reflection amplitude that masks the difference between the foreground hue and the background hue;<i>I</i>_<i>T</i><i>(x,y)</i>Describe the transmission intensity image using a mask of the detection system;<i>c</i>_<i>T</i>It is the complex transmission amplitude of the background tone of the mask (for example, in a quartz and chromium binary mask<i>C</i>_<i>T</i>Describe the nature of the chromium pattern);<i>a</i>_<i>T</i>It is the complex transmission amplitude that masks the difference between the foreground hue and the background hue (for example, using the same mask as above<i>a</i>_<i>T</i>Describe the optical properties of the difference between quartz and chromium;<i>c</i>_<i>T</i>and<i>a</i>_<i>T</i>Of course it varies with the nature of the material layer described);<i>I</i>_<i>R</i><i>(x,y)</i>Describe the reflection intensity image using a mask of the detection system;<i>C</i>_<i>R</i>Is the complex reflection amplitude of the background tone of the mask and<i>a</i>_<i>R</i>It is the complex reflection amplitude that masks the difference between the foreground hue and the background hue;<i>Re(x)</i>Express<i>x</i>The real part of<i>P(x,y)</i>Define the mask pattern of the inspected mask;<i>E</i>_<i>i</i>and<i>λ</i>_<i>i</i>Respectively refer to the eigenvectors and eigenvalues of the associated elements of a transmission cross-coefficient (TCC) imaging matrix associated with the inspection tool;<i>D</i>_<i>i</i>Tie<i>E</i>_<i>i</i>and<i>Re(x)</i>ofDC gain. Limited bandwidth mask pattern<i>M(x,y)</i>It is caused by a function called a "recovery core" (<img file="TW201218293A_D0002.tif" /><i>λ</i>_<i>i</i><i>D</i>_<i>i</i><i>E</i>_<i>i</i>(<i>x</i>,<i>y</i>)) Revolving mask pattern<i>P(x,y)</i>definition. Therefore, the limited bandwidth mask pattern is the mask pattern function<i>P(x,y)</i>One in modified form.

The limited bandwidth mask pattern is then used in operation 208 to generate test simulation images such as test aerial images and/or test photoresist images. Provide etching and/or photoresist system models for this purpose. An etching system model may include the numerical aperture of the etching and inspection system, the wavelength used in the etching and inspection system, the lighting conditions used in the etching and inspection system, and other etching and inspection parameters. For example, as further explained in US Patent Application No. 11/622,432, the basic core of the etching system model can be adjusted to eliminate any drop in the passband of the etching system caused by the limited bandwidth nature of the restored mask pattern. drop. Use a function that defines a set of modified homophonic bases (<i>F</i>_<i>i</i><i>(x,y))</i>To describe the modified TCC matrix used to implement one of the limited bandwidth mask patterns in the etching system model. A TCC matrix used in one of the etching systems can involve many items. However, since most of the light levels are represented by the previous items, only these items can be used to obtain an accurate estimate. Therefore, when necessary, a censored model that significantly reduces the computational burden can be used. A user can obtain a desired accuracy by using as many items in the level number as required for each specific mask inspection application. In general, the application of an etching system model enables the capture of the etching effect of non-printable features in the test simulation image.

In some embodiments, constructing a test simulation image involves separating non-printable features and printable features based on the light intensity distribution of the limited bandwidth mask pattern. A method may also involve constructing a geometric map based on a limited bandwidth mask pattern to classify geometric features into one or more geometric feature types such as edges, corners, and line ends. Moreover, one of the processes of identifying significant etching defects can be enhanced by applying different detection thresholds to different geometric feature types of the geometric image.

Etching and/or photoresist modeling substantially eliminates insignificant etching defects from the resulting simulated image. These defects (also known as "nuisance defects") have almost no effect on printed patterns. For the purpose of this disclosure, etching significant defects are defined as those defects that are etched significant in the final printed pattern. In other words, some defects ("nuisance defects"), although present in the mask, have no significant effect on the printing pattern transferred to a photoresist layer. Examples include defects that are so small (or etch-insensitive on one of the patterns) that they are essentially irrelevant. In addition, a defect can be formed in a relatively defect-insensitive portion of the substrate. In some cases, a defect can be formed on an auxiliary or OPC feature (or other resolution enhancement feature) but does not reflect the final printed pattern. Therefore, a significant etching defect exists on the mask and can cause a significant effect in the etching transfer pattern. These significant etching defects can cause problems related to circuit malfunction, sub-optimal performance, and so on.

In some embodiments, a detection process also includes building a feature map based on the test simulation image. The feature map may include a plurality of image parts each having a corresponding mask error enhancement factor (MEEF). For example, a typical photomask contains a pre-corrected image of the final IC that is magnified fourfold. Although this factor helps reduce the sensitivity of the pattern to imaging errors, the small-size features of modern IC circuits (for example, 22nm and below) are negatively affected by light beam scattering during etching exposure. Therefore, the MEEF sometimes exceeds 1. For example, the dimensional error on a wafer exposed by a 4X photomask can be greater than a quarter of the dimensional error on the photomask. In some embodiments, the MEEF value capture in a feature map is used to automatically adjust the detection threshold of the corresponding image area. For example, regions with larger MEEF values can be detected more carefully than regions with lower MEEF values. This operation can be implemented in an automated mode. A process may also involve providing a user-defined detection threshold for identifying the significant etching defects.

As mentioned above, a separate reference image provided from a database or another die is usually not available. A detection process can include a series of parallel operations 210 and 212 for generating internal reference images from the test images, which can be used as input for the aligned test images and the restored mask pattern. These images are sometimes called composite images. Various algorithms can be used to generate synthetic images, for example, Starlight available from KLA Tencor in Milpitas, Canada^TMsystem. The system is configured to sample various geometric features on a mask during the calibration of the system and construct a database of transmission and reflection properties of these features. Later, during inspection, when scanning a test mask, the system constructs a composite image based on the information in the database. A similar technique is described in Emery's US Patent No. 6,282,309 filed on August 28, 2001, which is fully incorporated herein by reference for the purpose of describing the synthetic imaging algorithm. For example, the algorithm can be based on the assumption that the transmitted signal and the reflected signal (at the same detection point) always complement each other without defects. In other words, the sum of the transmitted signal and the reflected signal does not change along the scan without defects. Therefore, any observed deviation in the sum signal can be regarded as a defect and excluded from the composite image.

Once synthesized reference images are generated in operation 210, these images are further processed at 212 to generate reference simulated images such as a reference aerial image and/or a reference test image. Operation 212 is similar to operation 208 performed on the test image described above. However, since the synthesized reference images do not include defects, the corresponding reference simulated images are produced without significant etching defects.

At 214, the test simulation images are compared with the reference simulation images to identify significant etching defects. In some embodiments, the reference simulation images include a reference simulation transmission image and a reference simulation reflection image, and the test simulation images include a test simulation transmission image and a test simulation reflection image. In this embodiment, comparison is performed in two groups, that is, the reference simulated transmission image is compared with the test simulated transmission image, and the reference simulated reflection image is compared with the test simulated reflection image. In some embodiments, the classification identifies significant etch defects.

In some embodiments, the detection is also applicable to multiple tone masks. One example of these masks is the third of a pattern with a darkest area (for example, a chromium-domain opaque area) and a quartz or brightest area, and a gray-scale area with a darkness between the two Tone mask. These grayscale areas can be obtained in a variety of ways (for example, using EPSM materials, etc.). In this case, treat the mask as two different masks that are analyzed separately. For example, the same model as described above can be used to process a three-tone mask. However, the three-tone mask can be regarded as a mask with a background pattern (for example, chrome) and a grayscale pattern regarded as the foreground (for example, EPSM material). The images can be processed using the same equations and process operations as described above. A second analysis is performed on the mask using the EPSM material as the background pattern and the brightest pattern (e.g., quartz) regarded as the foreground. Alignment can be easily achieved because each of the materials has substantially different properties that exhibit different edge effects that can be used to align the images. The mask patterns can then be summed and then compared with the reference in the die-to-die or die-to-database comparison to verify the correctness of the wafer pattern through the process window and identify significant etching defects.

<b>in conclusion</b>

Although the above-mentioned invention has been described in considerable detail for the purpose of clear understanding, it is obvious that certain changes and modifications can be implemented within the scope of the attached patent application. It should be noted that there are many alternative ways of implementing the process, system, and equipment of the present invention. Therefore, the embodiments of the present invention should be regarded as illustrative rather than restrictive, and the present invention should not be limited to the details given herein.

<p>100. . . Etching system</p><p>101. . . Numerical aperture</p><p>102. . . flat</p><p>103. . . Illumination source</p><p>105. . . Illumination lens</p><p>150. . . Detection Systems</p><p>151a. . . Imaging lens</p><p>151b. . . Numerical aperture</p><p>152. . . Main mask plane</p><p>153. . . Imaging optics</p><p>154. . . Sensor</p><p>160. . . Illumination source</p><p>161. . . Illumination optics</p><p>173. . . computer system</p><p>M. . . Photomask</p><p>W. . . Wafer</p>

1A is a simplified schematic diagram of an etching system for transferring a mask pattern from a mask to a wafer according to some embodiments.

Figure 1B provides a schematic diagram of a photomask inspection device according to some embodiments.

FIG. 2 illustrates a process flow diagram corresponding to an example of a method for detecting a photomask to identify significant etching contamination defects.