Method, system and medium for controlling manufacture process having multivariate input parameters

Abstract

A method, system, and medium of modeling and/or for controlling a manufacturing process is disclosed. In particular, a method according to embodiments of the present invention includes calculating a set of predicted output values, and obtaining a prediction model based on a set of input parameters, the set of predicted output values, and empirical output values. Each input parameter causes a change in at least two outputs. The method also includes optimizing the prediction model by minimizing differences between the set of predicted output values and the empirical output values, and adjusting the set of input parameters to obtain a set of desired output values to control the manufacturing apparatus. Obtaining the prediction model includes transforming the set of input parameters into transformed input values using a transformation function of multiple coefficient values, and calculating the predicted output values using the transformed input values.

Description

FIELD OF THE INVENTION

The present invention relates to a method, system and medium for modeling and controlling processes. More specifically, the present invention relates to modeling and controlling semiconductor-processing equipment that has multivariate input parameters.

BACKGROUND OF THE INVENTION

In manufacturing products that include precision discrete parts (e.g., microelectronic chips on silicon substrates), controlling manufacturing processes plays a crucial role. Controlling such processes may require, among other things, monitoring the characteristics of manufactured parts (e.g., processed wafers, hereinafter referred to as outputs) and adjusting input parameters accordingly. By adjusting the values of the input parameters, different types of outputs can be produced and the characteristics of the outputs can also be controlled.

For automating the control of the manufacturing processes, a mathematical model of the processing equipment can be used. One example of such a model is called a predictive model. This model is used to predict the future output values (e.g., the characteristics of products) based on historical information (e.g., input parameter values and the corresponding output qualities).

One such predictive model is an offset technique, which is illustrated in FIG. 1. In particular, the values of a number of input parameters 101 are received by an input/output dependency model 103, which calculates a predicted output value y₁^Pred 105 based on the input values. A corrector 109 then compares the predicted value y₁^Pred with an actual output value y₁^a 107 for the given values of the input parameters. If the predicted and actual output values are similar to each other within a certain range, no change is made to the input/output dependency model 103. If the predicted and actual output values are different (e.g., outside the range) from each other, the predictor input/output dependency model 103 is modified by adjusting an offset value (O₁) 111 based on the magnitude of the difference.

In equipment that has more than one output, at least some of the outputs may include mutual (shared) inputs. This means the output values of the equipment are not completely independent from each other (e.g., changing an input to adjust a given output may unintentionally change the characteristics of other outputs). In a conventional modeling technique, each output has its own correction system as if the output values are independent from each other. Because the dependencies between the different outputs are not accounted for by the conventional technique, it does not always lead to accurate predictions. In addition, adjusting one offset of one output can affect other outputs.

SUMMARY OF THE INVENTION

Embodiments of the present invention advantageously overcome the above-described shortcomings of the aforementioned techniques. More specifically, embodiments of the present invention provide a system, method and medium for controlling semiconductor-processing equipment that has multivariate input parameters and outputs.

Embodiments of the present invention minimize the effects of outputs being interdependent from each other. This is achieved by providing input parameter transformations having transformation coefficients. The coefficients are obtained by minimizing a score function. This, in turn, allows accurate models to be obtained. Using the models, highly precise control of manufacturing equipment is accomplished.

In particular, an example method according to embodiments of the present invention includes the steps of identifying at least one input that causes a change in at least two of a plurality of outputs, storing values of the identified inputs and corresponding empirical output values, and calculating and storing predicted output values, based on, in part, the values of the identified inputs. The example method may further include the steps of calculating a set of transform coefficients by minimizing a score equation that is a function of, in part, differences between one or more of the empirical output values and their corresponding predicted output values, and calculating one or more input values for one or more desired output values based on, in part, the calculated set of transform coefficients.

BRIEF DESCRIPTION OF THE DRAWINGS

The detailed description of the present application showing various distinctive features may be best understood when the detailed description is read in reference to the appended drawings in which:

FIG. 1 is a diagram showing a conventional offset model;

FIG. 2 is a diagram illustrating processing equipment;

FIG. 3 is a diagram illustrating a model of the processing equipment shown in FIG. 2 in accordance with embodiments of the present invention;

FIG. 4 is a block diagram illustrating various components of embodiments of the present invention;

FIG. 5 is a flow chart illustrating processing steps of embodiments of the present invention;

FIG. 6 is a diagram illustrating a CMP process;

FIG. 7 is a block diagram representation of an example embodiment of a computer configured to perform embodiments of the present invention; and

FIG. 8 is a diagram illustrating an example of a memory medium that can be used for storing computer programs of embodiments of the present invention.

DETAILED DESCRIPTION

Embodiments of the present invention generally provide systems, methods and mediums for creating one or more adaptive process models to mathematically represent multivariate input parameter systems. The present invention is particularly applicable in a manufacturing process such as manufacturing and/or processing semiconductor wafers. In particular, the present invention relates to modeling techniques as used by equipment involved in the manufacturing of semiconductor wafers. A general overview of embodiments of the present invention is provided below. It will be followed by a specific example implementation of the present invention.

Before discussing embodiments of the present invention, FIG. 2 shows a simplified graphical representation of processing equipment 205 with input parameters 201 and outputs 203. Examples of processing equipment include etcher tools, deposition tools, chemical mechanical planarization (CMP) tools, etc. The processing equipment 205 can include one or more tools. Depending upon the values of the input parameters 201, different processes can be achieved. For instance, in a deposition tool, different types of layers can be deposited on a wafer and/or the thickness of the layer can be varied.

As a general overview of embodiments of the present invention, in FIG. 3, the processing equipment 205 has a set of input parameters 301, a set of predicted outputs 303, and a prediction model 305 therebetween (replacing the processing equipment of FIG. 2). The overall goal of the prediction model is to minimize differences between the predicted output values and empirically collected output values (i.e., the actual output values). Once the prediction model is optimized (e.g., the differences between the predicted and actual output values have been minimized), the model can then be used in setting input parameters based on desired output values. In other words, for a given set of desired output values, the model can be used in a reverse fashion to calculate the input parameter values that would cause output values close to the desired output values. The calculated input parameter values are also known as recipes.

In embodiments of the present invention, the step of obtaining the predictive model can be divided into two steps. The first is to transform the values of the input parameters 301 into transformed input values 307. The second is to use the transformed input values 307 in calculating predicted output values 303.

With respect to the transformation, input parameter values (X₁, X₂, X₃) along with coefficient vector {right arrow over (P)} are transformed into (X′₁, X′₂, and X′₃) by transform functions ψ₁, ψ₂, and ψ₃. Examples of transformation functions include:

X′₁=PX₁; X′₂=PX₂ (In this example, the value of {right arrow over (P)} is identical for both X₁ and X₂.)

2) X′₁=P₁₁X₁+P₁₂X₁²; X′₂=P₂₁X₁²+P₂₂X₂²+P_crossX₁X₂ (In this example, P₁₁, P₁₂, P₂₁, P₂₂ and P_cross can have different values.)

The coefficient values are calculated by the steps of: a. collecting historical information on input parameter values and actual output values; b. creating a score function based on the collected information; and c. finding the coefficient values that minimize the score function, S_p.

The above steps are described by making references to semiconductor processing tools. As such, the step of collecting the historical information entails a set of data points for processing a number of wafers. In particular, input parameter values and actual output values for a number of wafers that have been processed by the processing equipment would be collected. This collection would then be used in the next step of minimizing the score function. Here, the score function, S_p, is:

$S_p=\sum _{i,k}{W_{i,k}\left(y_{\mathrm{actual}}^{\mathrm{ik}}-y_{\mathrm{predicted}}^{\mathrm{ik}}\left({\stackrel{->}{X}}^{i^{\prime }}\left({\stackrel{->}{X}}^i,\stackrel{->}{P}\right)\right)\right)}^2$ where:

• i—number of wafer;
• k—number of output;
• y_actual—an actual output value;
• y_predicted—a predicted output value, as calculated based on transformed inputs for a particular wafer i ({right arrow over (X)}′ⁱ);
• {right arrow over (X)}′ⁱ=(X′₁ⁱ, X′₂ⁱ, X′₃ⁱ) is the transformed input vector, calculated on the base of the actual input;
• and {right arrow over (X)}ⁱ=(X₁ⁱ, X₂ⁱ, X₃ⁱ) for wafer i together with the transformation parameters {right arrow over (P)}. This calculation is performed using the following transformation functions:ψ₁(X₁, X₂, X₃,{right arrow over (P)}); ψ₂(X₁, X₂, X₃,{right arrow over (P)}); and ψ₃(X₁, X₂, X₃,{right arrow over (P)}).The next step, as noted above, is to minimize the score S_p, i.e., to find {right arrow over (P)} values that provide the minimum of S_p (S_pmin)

The above-described steps calculate an optimal {right arrow over (P)} (i.e., a vector of coefficients for input transformation functions) such that the prediction model of the present invention provides the closest possible predicted outputs to the actual outputs. In a processing model with multivariate input parameters, when the score is minimized, the negative effect of the interdependencies between output values on the model accuracy would also be minimized.

Now turning to describe an example implementation of the embodiments described above, as shown in FIG. 4, the example implementation includes a number of components: an input transformer 401, an input-output dependency model 403, a corrector 405 and a storage device 407. All these components can be implemented in hardware, firmware, software and/or any combination thereof.

These components are further explained by also referring to FIG. 5. In particular, the historical information (i.e., y_a^ik,{right arrow over (X)}ⁱ) is stored into the storage device 407. The corrector 405 then retrieves the historical information (y_a^ik,{right arrow over (X)}ⁱ) from the storage device 407 (step 501). Since the retrieved historical information contains raw input parameter values, the information is sent to the input transformer 401 along with coefficients {right arrow over (P)} (step 503). The coefficient {right arrow over (P)} can be stored in the storage device 407 or in the corrector 405.

The input transformer 401, upon receiving the information from the corrector 405, calculates transformed input parameter values {right arrow over (X)}′ⁱ (step 505). Once the transformed input parameter values are calculated, the input transformer 401 sends the transformed input values to the corrector 405.

The corrector 405, upon receiving the transformed input parameter values from the input transformer 401, sends the transformed input parameter values to the input/output dependence model 403. The input/output dependency model 403 then calculates predicted output parameter values y_pred (step 507). The corrector 405 then calculates the score S_p, and sets a new {right arrow over (P)} (a vector of parameters of input transformation functions) in order to minimize the score S_p (step 509). These steps can be repeated until an optimum {right arrow over (P)} that yields a minimal score S_p is obtained, and return the optimum {right arrow over (P)}. Each time new data is obtained, a new score from new data is created and a new optimum {right arrow over (P)} value is calculated. This newly calculated vector {right arrow over (P)} could be used for transforming the input values, meaning: {right arrow over (P)}_new≡{right arrow over (P)}_optimum.

In embodiments of the present invention, the optimum coefficients can be combined with the most recent vector such that:{right arrow over (P)}_new≡{right arrow over (P)}_previous+K({right arrow over (P)}_optimum−{right arrow over (P)}_previous) wherein K<1.

As a new set of data points arrives, a new optimum {right arrow over (P)} can be recalculated.

Once a set of coefficients is calculated, a set of input values can be obtained (e.g., a recipe) for a desired set of output values. More specifically, from a set of desired values, a set of transformed input values, {right arrow over (X)}′ⁱ, can be obtained by reversing the predictive model (e.g., the input/output dependence model 403). The transformed input values can then be reverse transformed using the coefficients {right arrow over (P)} to obtain the input value to produce the desired output values.

In the above-described embodiments, the raw input values are transformed using the calculated coefficients. The transformation is required to account for the dependencies among input parameters as graphically illustrated in FIG. 6. More specifically, a surface of a wafer having five regions with varying degrees of roughness is to be polished by a CMP process. The goal is to achieve a flat surface depicted by a dotted line in FIG. 6. In conventional techniques, one region would be polished without regard to the other regions. However, polishing one region can affect the polishing of another region (e.g., when an offset is applied in region 1 in order to bring the height in region 1 down to the broken line, the height in region 2 is also influenced by the changes of region 1). Using the embodiments of the present invention, these dependencies are accounted for.

An example embodiment of the computer in which embodiments of the present invention operate (e.g., the various components described in FIG. 4) is described below in connection with FIGS. 7-8. FIG. 7 illustrates a block diagram of one example of the internal hardware 713 of a computer configured to perform embodiments of the present invention. A bus 756 serves as the main information highway interconnecting various components therein. CPU 758 is the central processing unit of the internal hardware 713, performing calculations and logic operations required to execute embodiments of the present invention as well as other programs. Read only memory (ROM) 760 and random access memory (RAM) 762 constitute the main memory. Disk controller 764 interfaces one or more disk drives to the system bus 756. These disk drives are, for example, floppy disk drives 770, or CD ROM or DVD (digital video disks) drives 766, or internal or external hard drives 768. These various disk drives and disk controllers are optional devices.

A display interface 772 interfaces display 748 and permits information from the bus 756 to be displayed on display 748. Communications with external devices, such as the other components of the system described above, occur utilizing, for example, communication port 774. Optical fibers and/or electrical cables and/or conductors and/or optical communication (e.g., infrared, and the like) and/or wireless communication (e.g., radio frequency (RF), and the like) can be used as the transport medium between the external devices and communication port 774. Peripheral interface 754 interfaces the keyboard 750 and mouse 752, permitting input data to be transmitted to bus 756. In addition to these components, the internal hardware 713 also optionally includes an infrared transmitter and/or infrared receiver. Infrared transmitters are optionally utilized when the computer system is used in conjunction with one or more of the processing components/stations/modules that transmit/receive data via infrared signal transmission. Instead of utilizing an infrared transmitter or infrared receiver, the computer system may also optionally use a low power radio transmitter 780 and/or a low power radio receiver 782. The low power radio transmitter transmits the signal for reception by components of the production process, and receives signals from the components via the low power radio receiver. The low power radio transmitter and/or receiver are standard devices in industry.

Although the computer in FIG. 7 is illustrated having a single processor, a single hard disk drive and a single local memory, the analyzer is optionally suitably equipped with any multitude or combination of processors or storage devices. For example, the computer may be replaced by, or combined with, any suitable processing system operative in accordance with the principles of embodiments of the present invention, including sophisticated calculators, and hand-held, laptop/notebook, mini, mainframe and super computers, as well as processing system network combinations of the same.

FIG. 8 is an illustration of an example computer readable memory medium 884 utilizable for storing computer readable code or instructions. As one example, medium 884 may be used with disk drives illustrated in FIG. 7. Typically, memory media such as floppy disks, or a CD ROM, or a digital video disk will contain, for example, a multi-byte locale for a single byte language and the program information for controlling the modeler to enable the computer to perform the functions described herein. Alternatively, ROM 760 and/or RAM 762 illustrated in FIG. 7 can also be used to store the program information that is used to instruct the central processing unit 758 to perform the operations associated with various automated processes of the present invention. Other examples of suitable computer readable media for storing information include magnetic, electronic, or optical (including holographic) storage, some combination thereof, etc.

In general, it should be emphasized that the various components of embodiments of the present invention can be implemented in hardware, software or a combination thereof. In such embodiments, the various components and steps would be implemented in hardware and/or software to perform the functions of embodiments of the present invention. Any presently available or future developed computer software language and/or hardware components can be employed in such embodiments of the present invention. For example, at least some of the functionality mentioned above could be implemented using Visual Basic, C, C++, or any assembly language appropriate in view of the processor(s) being used. It could also be written in an interpretive environment such as Java and transported to multiple destinations to various users.

The many features and advantages of embodiments of the present invention are apparent from the detailed specification, and thus, it is intended by the appended claims to cover all such features and advantages of the invention which fall within the true spirit and scope of the invention. Further, since numerous modifications and variations will readily occur to those skilled in the art, it is not desired to limit the invention to the exact construction and operation illustrated and described, and accordingly, all suitable modifications and equivalents may be resorted to, falling within the scope of the invention. For instance, output values can be transformed similar to the transform performed on the input parameters, and operations can be performed on the transformed output values similar to those performed on the transformed input parameters.

Claims

1. A method for controlling a manufacturing apparatus, the method comprising: calculating, by a computer system, a set of predicted output values; obtaining, by the computer system, a prediction model based on a set of input parameters, the set of predicted output values and empirical output values, wherein obtaining the predication model comprises transforming the set of input parameters into transformed input values using a transformation function having a plurality of coefficient values;optimizing, by the computer system, the prediction model by minimizing differences between the set of predicted output values and the empirical output values, wherein minimizing the differences comprises collecting historical information on the set of input parameters and the empirical output values; creating a score function on the collected information; and determining coefficient values that minimize the score function to minimize the difference; andadjusting, by the computer system, the set of input parameters using the optimized prediction model to obtain a set of desired output values to control the manufacturing apparatus.

2. The method of claim 1, wherein obtaining the prediction model comprises: transforming the set of input parameters into transformed input values using a transformation function having a plurality of coefficient values; andcalculating predicted output values using the transformed input values.

3. The method of claim 1, wherein the score function is

4. The method of claim 3, further comprising: collecting additional empirical output values and corresponding input values;calculating a new set of coefficients {right arrow over (P)}new; andusing the new set of coefficients as the optimal value of {right arrow over (P)}.

5. The method of claim 3, further comprising: collecting additional empirical output values and corresponding input values;calculating a new set of coefficients as {right arrow over (P)}new≡{right arrow over (P)}previous+K({right arrow over (P)}optimum−{right arrow over (P)}previous),

6. A manutacturing apparatus, comprising: a memory device storing a set of input parameters and empirically-collected output values; a processing device, coupled to the memory device, to receive the set of input parameters and to calculate a set of predicted output values, wherein the processing device is configured to calculate a prediction model to minimize differences between the set of predicted output values and the empirically collected output values;calculate the predication model to minimized the difference comprises the processing device to transform the set of input parameters into transformed input values using a transformation function having a plurality of coefficient value, to collect historical information on the set to input parameters and the empirical output values, to create a score function based on the collected information, and to determine coefficient values that minimize the score function to minimize the differences and to obtain a set of desired output value to control manufacturing apparatus.

7. The manufacturing apparatus of claim 6, wherein the processing device is configured to use the prediction model in a reverse fashion to set a new set of input parameters to obtain a set of desired output values.

8. The manufacturing apparatus of claim 6, further comprising a tool that is controlled by the processing device, wherein the set of input parameters is a recipe for the tool.

9. The manufacturing apparatus of claim 6, wherein the tool is at least one of a etcher tool, a deposition tool, or a chemical mechanical planarization tool.

10. A non-transitory computer-readable medium for storing instructions that when executed a computer cause the computer to perform a method for predicting output characteristics of a device produced by a manufacturing apparatus, the method comprising: calculating a set of predicted output values;obtaining a prediction model based on a set of input parameters, the set of predicted output values, and empirical output values, wherein obtaining the predication model comprises transforming the set of input parameters into transformed input values using a transformation function having a plurality of coefficient values;optimizing the prediction model by minimizing differences between the set of predicted output values and the empirical output values, wherein minimizing the differences comprises collecting historical information on the set of input parameters and the empirical output values; creating a score function on the collected information; and determining coefficient values that minimize the score function to minimize the difference; andadjusting the set of input parameters using the optimized prediction model to obtain a set of desired output values to control the manufacturing apparatus.

11. The medium of claim 10, wherein obtaining the prediction model further comprise: calculating predicted output values using the transformed input values.

12. The medium of claim 11, wherein the score function is

13. The medium of claim 12, wherein the method further comprises: collecting additional empirical output values and corresponding input values;calculating a new set of coefficients {right arrow over (P)}new; andusing the new set of coefficients as the optimal value of {right arrow over (P)}.

14. The medium of claim 12, wherein the method further comprises: collecting additional empirical output values and corresponding input values;calculating a new set of coefficients as {right arrow over (P)}new≡{right arrow over (P)}previous+K({right arrow over (P)}optimum−{right arrow over (P)}previous),

15. A system for controlling a manufacturing apparatus, comprising: a memory device storing an input that causes a change in at least two of a plurality of outputs, and empirical output values;means for calculating predicted output values based on the input; and means for minimizing effects of the at least two of the plurality of outputs that are interdependent from each other, wherein the means for minimizing effects comprises means for collecting historical information on the input parameter and the empirical output values, means for creating a score function based on the collected information and means for calculating a set of transform coefficients that minimize the score function to minimize the effects on the empirical output values and the predicted output values to obtain a set of desired output values to control the manufacturing apparatus.

16. The system of claim 15, further comprising means for calculating an input value for one or more desired output values based on the calculated set of transform coefficients.

17. The system of claim 15, further comprising means for identifying the input that causes the change in the at least two of the plurality of outputs.

18. An apparatus, comprising: a storage device configured to store historical information including input parameter data and empirical output data;a corrector coupled to the storage device to receive the historical information from the storage device;an input transformer to receive the historical information and a set of coefficients {right arrow over (P)} from the collector, wherein the input transformer is configured to calculate transformed input parameters {right arrow over (X)}′i and to send the transformed input parameters {right arrow over (X)}′i to the corrector;an input-output dependency model to receive the transformed input parameters {right arrow over (X)}′i from the corrector, wherein the input-output dependency model is configured to calculate predicted output parameter values ypred;the corrector is configured to calculate a score equation and a new set of coefficients P to minimize the score equation; and the corrector is configured to obtain a set of input values for a set of desired output values.

19. The apparatus of claim 18, wherein the corrector is configured to obtain the set of input values for the set of desired output values by obtaining a set of transformed input parameters {right arrow over (X)}′i by reversing the input-output dependency model and by obtaining the set of input values to produce the set of desired output values by reversing the transformed input parameters {right arrow over (X)}′i using the new set of coefficients {right arrow over (P)}.