The present invention relates to a technique applied to lithography and, more particularly, to a technique of determining the defocus direction of an image formed on a resist by an exposure apparatus for exposing a resist on a substrate to light via a pattern of a reticle.
In lithography of semiconductor manufacture in which a reticle pattern is projected and transferred onto a wafer by exposing it using exposure light, an apparatus for inspecting the exposure and defocus amount is used.
The conventional flow of lithography will be described.
A resist pattern is formed by applying a resist serving as a photosensitizer to a substrate such as a semiconductor wafer, projecting a mask pattern on the resist by exposure using an exposure apparatus, and then developing the resist.
The dimensions of the formed resist pattern are checked by a scanning electron microscope (measuring SEM or CD-SEM) with a measuring function. The conventional processing contents of the measuring SEM include, e.g., acquiring an electron beam image of an area containing parts with strict dimensional accuracy (step 1), measuring the dimensions (step 2), determining whether the dimensions meet the standards (step 3), and if not, changing the exposure of the exposure apparatus (step 4, the exposure correction amount is ΔE). For, e.g., a positive resist, if the resist dimensions are too large, the exposure is increased. If the resist width is too small, the exposure is decreased.
The relationship between a resist pattern and a film pattern after etching will be described next.
The shape of a resist pattern and the shape of a film pattern have a predetermined relationship if the etching conditions are same. To obtain a film pattern having a predetermined shape, the resist pattern must have a predetermined shape, too. In, e.g., starting a new process, a wafer called a FEM (Focus Exposure Matrix) is prepared by exposing a pattern while changing the focus and exposure in each shot (one exposure cycle). A focus and exposure to obtain a predetermined resist pattern shape are found by measuring the dimensions of the resist pattern in each shot and cutting the wafer to check its sectional shape. That is, a so-called “condition determination” is performed.
With this operation, the exposure (E0) and focus value (F0) to widen the margin are determined. Product wafers are exposed on the basis of these conditions. However, it may be sometimes impossible to obtain a resist pattern with a predetermined shape under the conditions (E0, F0) determined by the “condition determination” because of various process variations (e.g., a change in resist sensitivity, a variation in thickness of an antireflection film under the resist, or a drift of various kinds of sensors of the exposure apparatus). This is detected in dimension measurement (step 2). In the conventional technique, the change in resist shape caused by process variations is compensated for by correcting the exposure.
KLA-Tencor in U.S.A. announced, on Jun. 24, 2003, “MPX” that enables in-line focus/exposure monitoring as a new option of an overlay measuring apparatus “Archer” series. “MPX” can monitor the focus and exposure by analyzing a unique dual tone line end shortening (LES) target and accurately separating the focus and exposure. On the basis of the data, management of the defocus and exposure of the exposure apparatus, grasping the apparatus variation, and specifying the cause can be done quickly. KLA-Tencor states that users can suppress any decrease in yield associated with focus and save the cost of millions of dollars a year by using that option. Techniques related to such a measuring apparatus are disclosed in U.S. Pat. Nos. 5,629,772, 5,757,507, 5,790,254, 6,137,578, and 6,577,406.
In the prior arts, to detect and cope with process variations, dimensional values such as a line width are checked by using measuring SEM. If the dimensional values do not meet the standards, the exposure is corrected. However, this method has the problem of focal depth degradation described below.
When the exposure changes, the line width changes. On the other hand, the line width rarely changes even when the focus changes. However, when the focus changes, the sectional shape of the resist changes although the line width does not change. As described above, the change in sectional shape influences the shape of the film pattern after etching. For this reason, in the prior arts incapable of detecting a focus variation, poor film pattern shapes may be produced in large quantities after etching.
As described above, since displacement caused by defocus cannot be corrected by correcting only the exposure, the resist can have no normal sectional shape. In addition, since exposure is not executed at the center of the depth of focus, the depth may be insufficient, and poor film pattern shapes may be produced in large quantities after etching.
In “MPX”, the focus amount and exposure are estimated by analyzing a dual tone line end shortening target (to be referred to as a mark hereinafter). The mark includes a hollow grating mark and a solid grating mark, as shown in
It is an object of the present invention to provide a novel technique of determining the defocus direction of an image formed on a resist by an exposure apparatus based on a resist pattern formed through an exposure of the resist to the image.
In order to achieve the object, according to the first aspect of the present invention, there is provided a method of determining a defocus direction of a pattern image formed on a resist by an exposure apparatus for exposing a resist on a substrate to light via a pattern of a reticle, comprising an image capturing step of capturing a resist image of a resist pattern formed through an exposure of a resist to the pattern image by the exposure apparatus to obtain image data, an extracting step of extracting a feature of the image data to obtain feature data, and a determining step of determining the defocus direction based on the feature data.
According to the above aspect, the extracting step includes a differentiating step of differentiating the image data to obtain differential image data.
According to the above aspect, the extracting step includes a calculating step of calculating a power spectrum of the differential image data as the feature data.
According to the above aspect, in the determining step, the determining step determines the defocus direction based on the feature data and feature data obtained in advance from a resist pattern whose defocus direction is known.
According to the above aspect, the determining step detemines the defocus direction by projecting the feature data of the image data to probability space data which has been obtained by projecting the feature data obtained in advance to one of a Mahalanobis space and a Bayes space.
According to the above aspect, the probability space data is prepared with respect to each of two directions of defocus.
According to the above aspect, in the determining step determines the defocus direction based on two probability data obtained by projecting the feature data of the image data to each of the probability space data.
According to the above aspect, the feature data obtained in advance are obtained with respect to a plurality of resist patterns, the resist patterns being obtained with respect to a plurality of values of at least one of a defocus amount and a dose of the exposure.
According to the second aspect of the present invention, there is provided an apparatus for determining a defocus direction of a pattern image formed on a resist by an exposure apparatus for exposing a resist on a substrate to light via a pattern of a reticle, comprising a camera configured to capture a resist image of a resist pattern formed through an exposure of a resist to the pattern image by the exposure apparatus to obtain image data, and a processor configured to extract a feature of the image data and to obtain feature data, and to determine the defocus direction based on the feature data.
According to the above aspect, the processor is configured to differentiate the image data to obtain differential image data.
According to the above aspect, the processor is configured to calculate a power spectrum of the differential image data as the feature data.
According to the above aspect, the processor is configured to determine the defocus direction based on the feature data and feature data obtained in advance from a resist pattern whose defocus direction is known.
According to the above aspect, the processor is configured to determine the defocus direction by projecting the feature data of the image data to probability space data which has been obtained by projecting the feature data obtained in advance to one of a Mahalanobis space and a Bayes space.
According to the above aspect, the processor is configured to store the probability space data with respect to each of two directions of defocus.
According to the above aspect, the processor is configured to determine the defocus direction based on two probability data obtained by projecting the feature data of the image data to each of the probability space data.
According to the above aspect, the feature data obtained in advance are obtained with respect to a plurality of resist patterns, the resist patterns being obtained with respect to a plurality of values of at least one of a defocus amount and a dose of the exposure.
An exposure apparatus for exposing a resist on a substrate to light via a pattern of a reticle comprises one of the above-dscribed determining apparatuses.
A method of manufacturing a device comprises steps of exposing a resist on a substrate to light via a pattern of a reticle using an exposure apparatus as defined above, developing the exposed substrate; and processing the developed substrate to manufacture the device.
A device manufacturing method of the present invention comprises a step of determining a defocus direction of a pattern image formed on a resist by an exposure apparatus for exposing a resist on a substrate to light via a pattern of a reticle, using one of the above-described determining apparatuses.
According to the present invention, the defocus direction of the exposure apparatus can be determined from the resist pattern.
Other objects and advantages besides those discussed above shall be apparent to those skilled in the art from the description of a preferred embodiment of the invention which follows. In the description, reference is made to accompanying drawings, which form a part hereof, and which illustrate an example of the invention. Such example, however, is not exhaustive of the various embodiments of the invention, and therefore reference is made to the claims which follow the description for determining the scope of the invention.
The outline of the present invention will be described with reference to the accompanying drawings.
In the present invention, the defocus direction is estimated by using not only a mark edge interval such as an interval CD1 or CD2 shown in
An input/output (image capturing) means or step (input/output unit S701 in
A feature extracting means or step (feature extracting unit S703 in
Edge components are extracted by differentiating the waveform (image data) (differentiated image data is obtained). Additionally, to extract the difference in edge sharpness and edge interval between the positive focus waveform and negative focus waveform, the calculated differentiated waveform is Fourier-transformed to calculate the power spectrum.
A learning data calculating means or step (learning data calculating unit S706 in
Next, feature data corresponding to the outliers of each data group is removed, thereby preventing the outliers from influencing the statistic of the two feature data groups. For example, the feature data of a waveform having an exposure largely different from the optimum exposure is removed from learning data. Alternatively, the feature data of a waveform having a focus amount largely different from the optimum focus amount is removed from learning data.
A determining means or step (determining unit S702 in
The feature data of the unknown waveform is projected to the positive focus probability space and negative focus probability space. The probability that the unknown waveform is a positive focus waveform and the probability that the unknown waveform is a negative focus waveform are calculated. On the basis of the magnitude relationship between them, the defocus direction is determined.
The embodiments of the present invention will be described below in detail with reference to the accompanying drawings.
The embodiments to be described below are mere examples of a means for implementing the present invention and should appropriately be corrected or changed in accordance with various kinds of conditions or the arrangement of the apparatus to which the present invention is applied. The same reference numerals as in the above-described prior art denote the same elements in the following description.
An overlay inspection apparatus which is indispensable in practicing the present invention will be described first.
The overlay inspection apparatus measures the alignment accuracy or distortion of a semiconductor exposure apparatus called a stepper or scanner. As shown in
The overlay inspection apparatus was developed during the 1980s to meet the requirements of high accuracy and mass production along with size reduction of semiconductor devices and introduced in device makers on a full scale in the 1990s. Before introduction of the overlay inspection apparatus, inspection is done by visually reading a device called a vernier based on the same principle as that of calipers by using a microscope. Currently, overlay inspection apparatuses are available from makers in Japan and abroad.
The arrangement of the overlay inspection apparatus will be described next with reference to
Referring to
The principle of all commercially available current overlay inspection apparatuses employs “bright field illumination+image processing”, like the alignment optical system of an exposure apparatus as shown in
The resolving power of the alignment detection system is estimated (because the makers have not disclosed the optical specifications). When the magnification between the wafer and the CCD camera is ×100, and the pixel pitch of the CCD camera is 10 μm, the resolving power is 100 nm/pixel on the wafer surface. Hence, the above-described accuracy of 10 nm or less is achieved here probably by various image processing techniques.
A defocus direction determining method according to the first embodiment of the present invention will be described next.
Each block may be a device including a dedicated program or processor to implement the function to be described below or may be implemented by executing a control program to control specific hardware corresponding to the function.
The outline of this embodiment will be described first with reference to
An input/output unit S701 receives unknown waveform data which is obtained at unknown focus amount and exposure by the image capturing element 12 of the overlay inspection apparatus shown in
The positive focus learning waveform information and negative focus learning waveform information input from the input/output unit S701 in advance are learned as teach data. The learning result is held by a learning data storage unit. The learning data can be updated by teach data input at appropriate times. When the unknown waveform is input then, the defocus direction is determined on the basis of the learning result. The determination result is output from the input/output unit S701.
A feature extracting unit S703 receives the unknown waveform, positive focus learning waveform information, or negative focus learning waveform information from the input/output unit S701. Upon receiving the unknown waveform, the feature extracting unit S703 outputs the feature data to the determining unit S702. Upon receiving the positive focus learning waveform information, the feature extracting unit S703 registers the feature data in a positive focus feature data storage unit S704 together with the focus amount and exposure. Upon receiving the negative focus learning waveform information, the feature extracting unit S703 registers the feature data in a negative focus feature data storage unit S705 together with the focus amount and exposure.
A learning data calculating unit S706 removes target data from the positive focus feature data (S704) by using the focus amount range and exposure range which are registered in a specific data removal information storage unit S707 and outputs obtained feature data to a positive focus learning data storage unit S708.
The learning data calculating unit S706 also removes target data from the negative focus feature data (S705) by using the focus amount range and exposure range which are registered in the specific data removal information storage unit S707, and outputs the obtained feature data to a negative focus learning data storage unit S709.
The determining unit S702 receives the feature data of the unknown waveform calculated by the feature extracting unit S703, the feature data of the positive focus learning data (S704), and the feature data of the negative focus learning data (S705), and outputs the defocus direction of the unknown waveform to the input/output unit S701.
Each functional block will be described below in detail.
<Feature Extracting Unit S703>
As shown in
<Learning Data Calculating Unit S706>
The learning data calculating unit S706 is provided to remove the feature data of a waveform having a focus amount near the optimum focus amount and prevent overlap between the positive focus feature data group and negative focus feature data group, thereby improving the degree of separation of the two feature data groups.
More specifically, as shown in
Also, in step S901, feature data at all exposures corresponding to the focus amount range including the optimum focus amount registered in the specific data removal information storage unit S707 is removed from the negative focus feature data (S705), and the obtained feature data is output to the negative focus learning data storage unit S709.
<Determining Unit S702>
As shown in
In step S1002, the determining unit S702 receives the negative focus feature data group of the negative focus learning data (S709) and generates a Mahalanobis space as a negative focus probability space.
When the Mahalanobis space is used, the distance from the center of the data group can be assumed to be an occurrence probability belonging to the data group. This assumption does not hold in a Euclidean space. In addition, the calculation amount using a Mahalanobis space is smaller than the calculation amount using a Bayes space.
In step S1003, the feature data of the unknown waveform is projected to the positive focus probability space calculated in step S1001, thereby calculating the probability that the unknown waveform is a positive focus waveform. In step S1003, the feature data of the unknown waveform is also projected to the negative focus probability space calculated in step S1002, thereby calculating the probability that the unknown waveform is a negative focus waveform.
If the probability that the unknown waveform is a positive focus waveform, which is calculated in step S1003, is higher than the probability that the unknown waveform is a negative focus waveform in step S1004, information representing that the defocus direction of the unknown waveform is the positive direction is output to the input/output unit S701 (S1006). Otherwise, the processing advances to step S1005.
If the probability that the unknown waveform is a positive focus waveform, which is calculated in step S1003, equals the probability that the unknown waveform is a negative focus waveform in step S1005, information representing that the defocus direction of the unknown waveform is zero (optimum focus) is output to the input/output unit S701 (S1007). Otherwise, information representing that the defocus direction of the unknown waveform is the negative direction is output to the input/output unit S701 (S1008).
In the second embodiment, a Bayes space is used in place of the Mahalanobis space in steps S1001 and S1002 of the determining unit S702 of the first embodiment.
In the Bayes space, the distance from the center of the data group can be assumed to be an occurrence probability belonging to the data group, as in the Mahalanobis space. However, the Bayes space is a complete multidimensional normal distribution probability space based on the Bayes conditional probability. On the other hand, the Mahalanobis space is no multidimensional normal distribution in the strict sense, although each dimension is a normal distribution. For this reason, the occurrence probability when the Mahalanobis space is used contains a large error as compared to the occurrence probability when the Bayes space is used. Hence, the robustness of determination can be improved as compared to the first embodiment.
In the third embodiment, the learning data calculating unit S706 of the first embodiment has the following arrangement.
The feature data of a waveform having a focus amount largely different from the optimum focus amount corresponds to the outliers. An object of this embodiment is to remove the data from learning data to prevent the outliers from influencing the statistic of the positive focus feature data group and the statistic of the negative focus feature data group.
More specifically, in step S901 shown in
Also, in step S901, feature data at all exposures corresponding to the focus amount range including the maximum focus amount registered in the specific data removal information storage unit S707 is removed from negative focus feature data (S705), and the obtained feature data is output to a negative focus learning data storage unit S709.
In the fourth embodiment, a Bayes space is used in place of the Mahalanobis space in steps S1001 and S1002 of the determining unit S702 of the third embodiment. Even in this case, the robustness of determination can be improved as compared to the third embodiment due to the reason described in the second embodiment.
In the fifth embodiment, the learning data calculating unit S706 of the first embodiment has the following arrangement.
The feature data of a waveform having an exposure largely different from the optimum exposure corresponds to the outliers. An object of this embodiment is to remove the data from learning data to prevent the outliers from influencing the statistic of the positive focus feature data group and the statistic of the negative focus feature data group.
More specifically, in step S901 shown in
Also, in step S901, feature data at all focus amounts corresponding to the exposure range including the maximum exposure registered in the specific data removal information storage unit S707 is removed from negative focus feature data (S705), and the obtained feature data is output to a negative focus learning data storage unit S709.
In the sixth embodiment, a Bayes space is used in place of the Mahalanobis space in steps S1001 and S1002 of the determining unit S702 of the fifth embodiment. Even in this case, the robustness of determination can be improved as compared to the fifth embodiment due to the reason described in the second embodiment.
In the seventh embodiment, the learning data calculating unit S706 of the first embodiment has the following arrangement.
This embodiment has two objects.
The first object is to remove, from learning data, the feature data of a waveform having a focus amount near the absolute value of the optimum focus amount and prevent overlap between the positive focus feature data group and negative focus feature data group, thereby improving the degree of separation of the two feature data groups.
The second object is to remove, from learning data, the feature data of a waveform having a focus amount largely different from the optimum focus amount to prevent the outliers from influencing the statistic of the positive focus feature data group and the statistic of the negative focus feature data group, and to remove, from learning data, the feature data of a waveform having an exposure largely different from the optimum exposure to prevent the outliers from influencing the statistic of the positive focus feature data group and the statistic of the negative focus feature data group.
More specifically, in step S901 shown in
Also, in step S901, feature data at all exposures corresponding to the focus amount range including the optimum focus amount registered in the specific data removal information storage unit S707, feature data at all exposures corresponding to the focus amount range including the maximum defocus amount, and feature data at all focus amounts corresponding to the exposure range including the maximum exposure are removed from negative focus feature data (S705), and the obtained feature data is output to a negative focus learning data storage unit S709.
In the eighth embodiment, a Bayes space is used in place of the Mahalanobis space in steps S1001 and S1002 of the determining unit S702 of the seventh embodiment. Even in this case, the robustness of determination can be improved as compared to the fifth embodiment due to the reason described in the second embodiment.
As an effect common to the above-described embodiments, a conventional overlay inspection apparatus can be used without changing its hardware configuration.
When the above-described functions are added to the conventional overlay inspection apparatus, an overlay error, exposure error, and focus error can be measured as important performance of an exposure apparatus.
Hence, according to the above-described embodiments, exposure by the exposure apparatus can be done in an optimum focus state. More specifically, the conventional overlay inspection apparatus cannot determine the direction of focus error. However, according to the above-described embodiments, the direction can be determined. For this reason, the focus error can be reduced by, e.g., setting, in the exposure apparatus, an offset value corresponding to the focus error including the direction measured by the overlay inspection apparatus.
[Schematic Arrangement of Exposure Apparatus]
Referring to
The alignment optical system 27 sends, to a signal processing unit 27, an image signal obtained by photoelectrically converting an optical image of the mark on the wafer 24 by the CCD camera 12. The signal processing unit 27 calculates mark position information on the basis of the image signal. On the basis of the position information calculated by the signal processing unit 27, a central processing unit 28 positions the wafer stage 26 to correct the misalignment of the wafer. The central processing unit 28 also acquires, through the alignment optical system 27, the images of various marks to measure the overlay error, exposure error, and focus error and applies the method described in the above embodiments. In this way, an overlay error, exposure error, and focus error including a direction are measured.
[Device Manufacturing Method]
A semiconductor device manufacturing process using the exposure apparatus of the embodiment will be described next.
In step S3 (wafer manufacture), a wafer is manufactured using a material such as silicon. In step S4 (wafer process) called a preprocess, an actual circuit is formed on the wafer by the exposure apparatus by lithography using the mask and wafer. In step S5 (assembly) called a post-process, a semiconductor chip is formed from the wafer prepared in step S4. This step includes assembly processes such as assembly (dicing and bonding) and packaging (chip encapsulation). In step S6 (inspection), inspections including operation check test and durability test of the semiconductor device manufactured in step S5 are performed. A semiconductor device is completed with these processes and shipped in step S7.
The wafer process in step S4 has the following steps (
As many apparently widely different embodiments of the present invention can be made without departing from the spirit and scope thereof, it is to be understood that the invention is not limited to the specific embodiments thereof except as defined in the claims.
This application claims the benefit of Japanese Patent Application No. 2005-029829 filed on Feb. 4, 2005, which is hereby incorporated by reference herein in its entirety.
Number | Date | Country | Kind |
---|---|---|---|
2005-029829 | Feb 2005 | JP | national |