JP2007080963A - Method and unit for forming thin film - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a unit to form a thin film which is formed in its stabilized quality. <P>SOLUTION: In a crystal growth step, the thin film is formed using a thin film forming apparatus 11. The characteristic value of the thin film formed in this step includes a process error w. In a measuring step, measured characteristic of the formed thin film is measured with a measuring unit 12. The measured characteristic value measured in this step includes a measuring error ν. In a virtual crystal growth step, characteristic value of the thin film formed with a virtual thin film forming apparatus 13 is obtained and an arithmetic characteristic value is computed. In a measuring error removing step, a forecasted characteristic value wherein only the measuring error ν among the process errors w and measuring errors ν is removed from the measuring characteristic value can be computed with a measuring error removing unit 14, on the basis of the computed characteristic value and the measured characteristic value. Moreover, in a apparatus condition determining step, the new apparatus condition of the thin film forming apparatus 11 is determined with a condition setting unit 15 based on the forecasted characteristic value. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a thin film forming unit that feedback-controls the characteristic value of a thin film to be formed next based on the characteristic value of the formed thin film, and brings the characteristic value of the thin film close to a command value.

  In a conventional thin film forming unit, a thin film is formed on a substrate by a thin film forming apparatus, and a characteristic value of the formed thin film is measured. In the thin film forming unit, the apparatus conditions for the next thin film are determined based on the measurement result.

  FIG. 9 is a block diagram schematically showing the configuration of a conventional thin film forming unit 1 according to the first technique. The condition setting unit 2 determines apparatus conditions based on the input characteristic values. The thin film forming apparatus 3 is, for example, a molecular beam epitaxial apparatus (MBE apparatus), and forms a thin film based on apparatus conditions. The apparatus conditions are, for example, the temperature of the material of the molecular beam to be irradiated, the temperature of the substrate, and the irradiation time of the molecular beam. The measurement unit 4 measures characteristic values such as the film thickness, composition ratio, and light emission characteristics of the thin film, and acquires the measurement results. The thin film forming unit 1 acquires, for example, a measurement result of one lot of substrates. The condition setting unit 2 determines a new apparatus condition based on the measurement result and the inputted characteristic value so that the next one lot of thin film is formed into a thin film having the inputted characteristic value. The thin film forming apparatus 3 forms a thin film on the substrate based on the new apparatus conditions.

  In the second conventional technique, the target value of the parameter to be controlled based on the setting parameter is calculated using a model function corresponding to the state change in the heat treatment vessel, and the calculation result is fed back to the process condition, thereby forming the film There is a heat treatment apparatus and a film forming method for improving the reproducibility of thickness.

JP 2002-343726 A

  When the characteristic value of the thin film formed by the conventional thin film forming apparatus 3 is measured by the measuring unit 4, the measurement result is measured based on the resolution of the measuring instrument of the measuring unit 4 with respect to the input characteristic value. Errors and process errors that occur in the process of forming the thin film are included. The actual characteristic value of the formed thin film is a characteristic value obtained by adding a process error to the input characteristic value, and the measurement result includes a measurement error in addition to the actual characteristic value. Therefore, when the characteristic value of the thin film is feedback-controlled based on the measurement result, the thin film is formed so that the characteristic value including the measurement error is close to the input characteristic value. Therefore, the actual characteristic value of the formed thin film does not converge to the input characteristic value even if the thin film characteristic value is repeatedly feedback controlled based on the measurement result, and the quality of the formed thin film is poor. It is stable and leads to low yield. Also, by removing both measurement error and process error, the actual characteristic value of the formed thin film does not converge to the input characteristic value as in the case where measurement error is included, and the quality of the formed thin film is reduced. It becomes unstable and yield is low.

  In the film forming method of the second conventional technique, a statistical model function such as an autoregressive moving average model is created for a known periodic process variation, and a value estimated by this model function is set as a process condition. The model function is updated based on the actual value of the film thickness after film formation. However, actual process variations include irregular changes, and the film thickness includes measurement errors and the like. Therefore, in order to provide accurate feedback to the target, it is necessary to separate the process variation and measurement error and then estimate the true state of the process and reflect it in the apparatus setting conditions. Furthermore, in order to improve the accuracy of state estimation, it is desirable to create a model function according to the physics of film formation.

  An object of the present invention is to provide a thin film forming method and a thin film forming unit in which the quality of the formed thin film is stable.

The present invention is a film forming process in which a thin film is formed using a thin film forming apparatus, the film forming process in which a thin film includes a process error in its characteristic value;
A measurement process for measuring a measurement characteristic value of a thin film to be formed, wherein the measurement characteristic value is measured in a state including a measurement error; and
Based on the apparatus conditions of the thin film forming apparatus in the film forming step, a calculation step for calculating a calculation characteristic value of the thin film to be formed by calculation,
Based on the calculation characteristic value calculated in the calculation step and the measurement characteristic value measured in the measurement step, a predicted characteristic value obtained by removing only the measurement error from the measurement characteristic value out of the process error and the measurement error is calculated. A measurement error elimination process;
And an apparatus condition determining step for determining a new apparatus condition of the thin film forming apparatus in the film forming process when the thin film is formed next based on the predicted characteristic value calculated in the measurement error removing process. This is a method for forming a thin film.

  According to the present invention, in the film forming process, a thin film is formed using a thin film forming apparatus. The characteristic value of the thin film formed at this time includes a process error. In the measurement step, the measurement characteristic value of the thin film to be formed is measured. The measurement characteristic value measured at this time includes a measurement error. In the calculation step, the calculation characteristic value of the thin film to be formed is obtained by calculation. In the measurement error removal step, a predicted characteristic value obtained by removing only the measurement error from the measurement characteristic value out of the process error and the measurement error can be calculated based on the calculation characteristic value and the measurement characteristic value. Further, in the apparatus condition determining step, a new apparatus condition for the thin film forming apparatus is determined based on the predicted characteristic value.

  In the measurement error elimination process, the present invention further estimates the film formation state in the film formation process based on the measured characteristic value and the calculated characteristic value, and based on the estimated film formation state in the film formation process. Thus, the arithmetic expression used to calculate the arithmetic characteristic value in the arithmetic step is corrected.

  According to the present invention, in the measurement error removal step, the film formation state in the film formation step is further estimated based on the measurement characteristic value and the calculation characteristic value, and the arithmetic unit is calculated based on the estimated film formation state. Thus, the arithmetic expression used for calculating the characteristic value can be corrected.

The present invention provides a thin film forming apparatus for forming a thin film with a process error included in its characteristic value;
A measuring means for measuring the characteristic value of the formed thin film in a state including a measurement error;
Based on the apparatus conditions of the thin film forming apparatus when forming a thin film with the thin film forming apparatus, an arithmetic device that obtains a characteristic value of the thin film to be formed by calculation,
Measurement error elimination that calculates the predicted characteristic value in which only the measurement error is removed from the measurement characteristic value among the process error and measurement error based on the characteristic value calculated by the arithmetic unit and the characteristic value measured by the measuring means And a device condition determining means for determining a new device condition of the thin film forming device when the thin film is next formed based on the predicted characteristic value calculated by the measurement error removing device. Is a unit.

  According to the present invention, the thin film forming apparatus forms a thin film, and the measurement characteristic value of the formed thin film is measured by the measuring means. The process error is included in the characteristic value of the thin film formed by the thin film forming apparatus, and the measurement error is included in the measurement characteristic value measured by the measuring unit. The arithmetic device obtains the characteristic value of the thin film to be formed by calculation. The measurement error removing unit calculates a predicted characteristic value obtained by removing only the measurement error from the process characteristic value and the measurement characteristic value based on the calculation characteristic value and the measurement characteristic value. The apparatus condition determining means can determine a new apparatus condition of the thin film forming apparatus based on the predicted characteristic value.

  Further, in the present invention, the measurement error removing means further estimates the film forming state of the thin film forming apparatus based on the measured characteristic value and the calculated characteristic value, and based on the estimated film forming state of the thin film forming apparatus, Based on the characteristic value measured by the measuring means and the characteristic value calculated by the arithmetic device, the arithmetic expression used for calculating the arithmetic characteristic value by the arithmetic device is corrected.

  According to the present invention, the measurement error removing means further estimates the film forming state of the thin film forming apparatus based on the measured characteristic value and the calculated characteristic value, and based on the estimated film forming state of the thin film forming apparatus. The arithmetic expression used for calculating the arithmetic characteristic value by the arithmetic device can be corrected.

Moreover, the present invention is a molecular beam epitaxial apparatus in which the thin film forming apparatus forms a thin film by irradiating a substrate with a plurality of molecular beams.
The arithmetic device includes a molecular dose calculator that calculates a molecular dose based on the molecular beam intensity of each molecular beam irradiated on the substrate, a composition ratio calculator that calculates a composition ratio of a thin film based on the molecular dose, And a characteristic value calculation unit for calculating a calculation characteristic value of the thin film based on the composition ratio.

  According to the present invention, a thin film can be formed with a molecular beam epitaxial (abbreviation MBE) apparatus. In the present invention, the molecular dose calculator calculates the molecular dose irradiated onto the substrate by each molecular beam, and the composition ratio calculator calculates the composition ratio based on the molecular dose. The characteristic value calculation unit calculates the calculated characteristic value of the thin film based on the composition ratio. Thereby, the characteristic value of the thin film can be calculated.

In the present invention, the apparatus conditions include the temperature T of the vapor deposition source of each molecular beam, the temperature increase width α of each vapor deposition source, and the irradiation time t of each molecular beam.
The molecular dose calculation unit is based on the constant A, the Arrhenius coefficient B, and the apparatus conditions determined for each material of the molecular beam.

The molecular dose is calculated by calculating the molecular beam intensity F _{x according} to the equation (1), and further calculating the molecular dose by integrating the molecular beam intensity F _x with the irradiation time t.
According to the present invention, the molecular dose of each molecular beam of the MBE apparatus can be calculated.

  According to the present invention, it is possible to calculate a predicted characteristic value obtained by removing only a measurement error from a process error and a measurement error from the measurement characteristic value. In the apparatus condition determination step, a new apparatus condition is determined based on the calculated predicted characteristic value. Therefore, since the characteristic value of the thin film formed based on the new apparatus conditions does not include measurement errors as in the conventional technique, the deviation from the desired characteristic value is smaller than in the conventional technique. Therefore, the quality of the formed thin film can be stabilized and the yield can be improved. Further, the quality of the semiconductor element including the thin film and the apparatus including the semiconductor element can be stabilized and the yield can be improved. In addition, the predicted value characteristic value includes only a measurement error and includes a process error. Therefore, the characteristic value of the thin film subjected to repeated feedback control converges to a desired characteristic value. The characteristic value obtained by adding the measurement error to the desired characteristic value and the characteristic value obtained by subtracting the process error from the desired characteristic value do not converge as in the conventional case, and can be approximated by the desired characteristic value.

  According to the present invention, the film formation state in the film formation step is estimated based on the calculation characteristic value and the measurement characteristic value, and the calculation formula used in the calculation step is corrected based on the film formation state. Therefore, even if the film formation state changes over time or due to environmental conditions, the arithmetic expression used in the calculation process is corrected following the change in the film formation state. As a result, the calculated characteristic value becomes a calculated characteristic value that takes into account fluctuations in the film formation state, and becomes a characteristic value that approximates the characteristic value of the formed thin film as compared with the case where the corrected characteristic value is not followed and corrected. Since the predicted characteristic value is calculated based on such a calculated characteristic value and the measured characteristic value, the predicted characteristic value can be brought closer to a desired characteristic value, and determined based on the predicted characteristic value. The thin film formed according to the apparatus conditions can be brought close to a desired characteristic value.

  According to the present invention, it is possible to calculate a predicted characteristic value obtained by removing only a measurement error from a process error and a measurement error from the measurement characteristic value. The device condition determining means determines a new device condition based on the predicted characteristic value. Therefore, since the characteristic value of the thin film formed based on the new apparatus conditions does not include measurement errors as in the conventional technique, the deviation from the desired characteristic value is smaller than in the conventional technique. Thereby, the quality of the formed thin film can be stabilized, and the yield can be improved. Further, the quality of the semiconductor element including the thin film and the apparatus including the semiconductor element can be stabilized and the yield can be improved. Therefore, the characteristic value of the thin film subjected to repeated feedback control converges to a desired characteristic value. A characteristic value obtained by adding a measurement error to a desired characteristic value and a desired characteristic value do not converge to a process error as in the past, and can be approximated by a desired characteristic value.

  According to the present invention, the film formation state of the thin film forming apparatus is estimated based on the calculation characteristic value and the measurement characteristic value, and the calculation formula used in the calculation apparatus is corrected based on the film formation state. Even if the film formation state changes over time, the arithmetic expression used in the arithmetic device is corrected following the change in the film formation state. As a result, the calculated characteristic value becomes a calculated characteristic value that takes into account fluctuations in the film formation state, and becomes a characteristic value approximated to the characteristic value of the actually formed thin film as compared with the case where the corrected characteristic value is not corrected. Since the predicted characteristic value is calculated based on such a calculated characteristic value and the measured characteristic value, the predicted characteristic value can be brought closer to a desired characteristic value, and determined based on the predicted characteristic value. The thin film formed according to the apparatus conditions can be brought close to a desired characteristic value.

  According to the present invention, the molecular characteristic calculator, the composition ratio calculator, and the characteristic value calculator can calculate the calculated characteristic value of the thin film formed by the MBE apparatus, and the thin film forming unit can be realized by the MBE apparatus. it can.

  According to the present invention, calculation of molecular dose can be realized. The composition ratio of each material can be calculated based on the molecular dose, and the characteristic value of the thin film can be calculated based on the composition ratio. Therefore, by calculating the molecular dose, it is possible to calculate the calculation characteristic value of the thin film.

  Hereinafter, a plurality of embodiments for carrying out the present invention will be described with reference to the drawings. Portions corresponding to the matters described in the preceding forms in each embodiment are denoted by the same reference numerals, and overlapping description may be omitted. When only a part of the configuration is described, the other parts of the configuration are the same as those described in the preceding section. In addition to the combination of parts specifically described in each embodiment, the embodiments may be partially combined as long as the combination is not particularly troublesome.

  FIG. 1 is a block diagram showing the function of the thin film forming unit 10 according to the first embodiment of the present invention divided into blocks. The thin film forming unit 10 is a unit for forming a thin film on a conveyed substrate, and is a unit for bringing the thin film close to a characteristic value input by a user. The thin film forming unit 10 measures and calculates the characteristic value of the thin film, feedback-controls the characteristic value of the formed thin film based on the measurement result and the calculation result, and uses the characteristic value of the formed thin film to the user. Approaches the entered characteristic value. The characteristic value includes, for example, the composition ratio, the film thickness, and the optical characteristics of the material constituting the thin film. The thin film forming unit 10 is included in a part of a unit for manufacturing a semiconductor element, for example, and is responsible for one step of the manufacturing process of the semiconductor element. The thin film forming unit 10 basically includes a thin film forming device 11, a measuring unit 12, a virtual thin film forming device 13 (hereinafter sometimes referred to as “virtual device 13”), a measurement error removing unit 14, and a condition setting unit. 15 is included.

  The thin film forming apparatus 11 is an apparatus that mounts a substrate transported by a transport device (not shown) and forms a thin film by crystal growth of a plurality of materials on the substrate based on the device conditions. The apparatus conditions include, for example, the temperature of the substrate to be controlled and the material for crystal growth, and the time for crystal growth. The thin film forming apparatus 11 is, for example, a CVD apparatus or an MBE apparatus, and an MBE apparatus is used in the present embodiment. In the MBE apparatus, a substrate transported by a transport apparatus (not shown) is placed, and each material filled in a plurality of crucibles, in this embodiment, three materials of indium (In), gallium (Ga), and aluminum (Al). This is an apparatus for forming a thin film by irradiating a substrate with a molecular beam of a material and growing each of the materials on the substrate. Below, the case where an MBE apparatus is applied to the thin film formation apparatus 11 is demonstrated. However, the present invention is not limited to the case where an MBE apparatus is used, and may be a CVD apparatus as described above. The thin film forming apparatus 11 includes a temperature control unit 16 that controls the temperature of the substrate to be placed and the temperatures of the three materials In, Ga, and Al based on the apparatus conditions, and each molecular beam on the substrate to be placed. It includes a crystal growth portion 17 for crystal growth by irradiation.

  The measuring unit 12 which is a measuring means is electrically connected to the measurement error removing unit 14 and has a function of measuring the characteristic value of the thin film formed by the crystal growing unit 17. The measured characteristic value of the thin film includes at least the same kind of characteristic value as the characteristic value to be input. For the measurement unit 12, for example, an X-ray diffractometer that measures the composition ratio, an SEM device that measures the film thickness, and a photoluminescence measurement device that measures the wavelength are used. A characteristic value measured by the measurement unit 12 (hereinafter sometimes referred to as “measurement characteristic value”) is transmitted to the measurement error removal unit 14. The substrate on which the thin film is formed is measured by the measuring unit 12 and then transferred to an apparatus for a next process in a semiconductor manufacturing process (not shown).

  The virtual device 13, which is an arithmetic device, is electrically connected to the measurement error removing unit 14, and has a function of calculating the characteristic value of the thin film to be formed based on the device conditions. The virtual device 13 includes a virtual setting unit 18 and a virtual crystal growth unit 19. The virtual setting unit 18 has a function of setting a material model for crystal growth. The virtual crystal growth unit 19 has a function of calculating a characteristic value of a thin film using an arithmetic expression when a thin film is formed by crystal growth of a material model on a substrate model. In the present embodiment, there is a function of calculating the characteristic value of the thin film using the arithmetic expression when the thin film is virtually formed by the MBE apparatus. The calculated characteristic value (hereinafter sometimes referred to as “calculated characteristic value”) is transmitted to the measurement error removing unit 14.

  A measurement error removal unit 14 that is a measurement error removal unit is electrically connected to the condition setting unit 15. The measurement characteristic value includes a process error caused by process variation when the thin film forming apparatus 11 forms a thin film, and a measurement error caused by the resolution of the measuring instrument of the measurement unit 12 and the reproducibility of the measurement location. The measurement error removal unit 14 has a function of predicting and calculating a characteristic value obtained by removing only the measurement error from the measurement characteristic value based on the measurement characteristic value and the calculation characteristic value. This predicted characteristic value may be referred to as a predicted characteristic value. Further, the measurement error removing unit 14 has a function of correcting the calculation formula used in the virtual crystal growth unit 19 based on the measurement characteristic value and the calculation characteristic value. When the crystal growth unit 19 virtually grows a crystal on the next substrate model, the virtual crystal growth unit 19 calculates a calculation characteristic value using the corrected calculation formula. The measurement error removal unit 14 has a function of transmitting the predicted characteristic value to the condition setting unit 15.

  The condition setting unit 15 which is an apparatus condition determining means is electrically connected to the thin film forming device 11 and the virtual device 13, and is a thin film characteristic value (hereinafter referred to as "input characteristic value") input by a user through an operation unit (not shown). The thin film forming apparatus 11 has a function of setting the apparatus conditions. The condition setting unit 15 transmits the set apparatus conditions to the thin film forming apparatus 11 and the virtual apparatus 13. The temperature control unit 16 of the thin film forming apparatus 11 controls the temperature of the substrate to be placed and the material of each molecular beam based on the apparatus conditions. The crystal growth unit 17 forms a thin film by irradiating the substrate with the molecular beam of each of the materials for a growth time under apparatus conditions to grow the crystal. Further, the virtual device 13 calculates a calculation characteristic value based on the device condition. The condition setting unit 15 has a function of correcting the apparatus condition set this time according to the difference between the predicted characteristic value and the input characteristic value, and determining and setting the next apparatus condition. The thin film forming apparatus 11 forms the next thin film based on the apparatus conditions. In other words, the thin film forming apparatus 11 feedback-controls the characteristic of the thin film formed based on the predicted characteristic value, and brings the characteristic value of the thin film formed next closer to the input characteristic value. Similarly, the virtual device 13 calculates the characteristic value of the thin film to be formed next based on the device conditions.

  FIG. 2 is a flowchart showing the procedure of the thin film forming process of the thin film forming unit 10. In the thin film forming process, the characteristic value of the thin film is feedback-controlled based on the measured characteristic value obtained by measuring the formed thin film by the measuring unit 12 and the calculated characteristic value calculated by the virtual device 13. When the power of the thin film forming unit 10 is turned on, the manufacturing process starts and the process proceeds to step s1. In step s1, which is a characteristic value input process, the user inputs an input characteristic value R through the operation unit. When the input characteristic value R is input, the process proceeds from step s1 to step s2. In step s2, which is an apparatus condition determining step, the condition setting unit 15 determines apparatus conditions such as the molecular beam material, the substrate temperature, the material temperature, and the crystal growth time based on the input characteristic value R. The condition setting unit 15 transmits the determined apparatus conditions to the thin film forming apparatus 11 and the virtual apparatus 13, and sets these apparatuses to the apparatus conditions. Further, various setting constants are input to the measurement error removing unit 14 and determined. When the device is set as a device condition, the process proceeds from step s2 to step s3 and step s6. Step s3 is a process executed by the thin film forming apparatus 11, and step s6 is a process executed by the virtual apparatus 13, respectively. These processes are executed simultaneously. However, it is not limited to being executed simultaneously, and may be executed separately. First, formation of a thin film and measurement of characteristic values of the thin film forming apparatus 11 will be described.

  In step s3, which is a temperature control process, the substrate transported to the thin film forming apparatus 11 is controlled to a temperature set in the apparatus conditions, and the molecular beam material is controlled to a temperature set in the apparatus conditions. In the present embodiment, the temperature of each material is controlled to the temperature set in the apparatus conditions by controlling the temperature of the cell of the MBE apparatus. When the temperature is controlled by the apparatus conditions, the process proceeds from step s3 to step s4.

  In step s4, which is a crystal growth process, the substrate is irradiated with a molecular beam of each material for the growth time (irradiation time) set in the apparatus conditions, and a thin film is formed on the substrate. When a thin film is formed by the thin film forming apparatus 11, the actual characteristic value of the formed thin film does not become the input characteristic value R due to process variations in the manufacturing process such as changes in temperature and the state of the apparatus. Is a characteristic value including a process error w. The thin film formed in step s4 is formed in a state including this process error w. When the thin film is formed, the process proceeds from step s4 to step s5.

  In step s5, which is a measurement process, the measurement characteristic value of the formed thin film is measured by the measurement unit 12, and this measurement characteristic value is transmitted to the measurement error removal unit 14. In addition to the actual characteristic value of the formed thin film, the measurement characteristic value measured in the measurement process includes a measurement error ν caused by the resolution of the measurement device of the measurement unit 12 and the reproducibility of the measurement location. Accordingly, the measurement characteristic value of the formed thin film includes two errors of the process error w and the measurement error ν with respect to the input characteristic value R. When the measurement characteristic value is transmitted to the measurement error removing unit 14, the process proceeds from step s5 to step s8.

  Next, calculation of the characteristic value of the thin film by the virtual device 13 will be described. In step s6, which is a model setting process, the virtual device 13 sets a material model based on the type of material of the molecular beam and the temperature of the material input by the user through an operation unit (not shown). When the material model is set, the process proceeds from step s6 to step s7.

  In step s7, which is a virtual crystal growth process, a calculation characteristic value is calculated based on the material model set in step s6. Below, the calculation method of a calculation characteristic value is demonstrated. The virtual crystal growth unit 19 calculates Equation (2) based on the virtual model of the substrate and material determined in Step s6 and their temperature states.

Y is a calculation characteristic value, H is a transfer function, and X is a coefficient indicating how the thin film is formed, that is, a coefficient indicating the process state. The process state X is a coefficient for causing the calculation characteristic value to follow the process fluctuation and the fluctuation of the characteristic value caused by disturbance or the like that occur in the thin film formation process. In other words, the process state X is a coefficient for calculating a calculation characteristic value including the process error w from an ideal characteristic value based on the apparatus conditions. The transfer function H is a transfer function for converting process fluctuations, disturbances, and the like into observable physical property values, and is a transfer function for converting into characteristic values of a thin film in the present embodiment. The transfer function H is a function given by a physical property value used for calculation of the calculation characteristic value Y.

In the present embodiment, for example, consider a case where the thin film Al composition ratio Y _Al is used as a calculation characteristic value. First, the molecular beam intensity F _x is given by Equation (1) as a constant A, an Arrhenius coefficient B, a temperature T of the material cell, a temperature rise α of the cell, and an irradiation time t.

The subscript x of the molecular beam intensity F _x indicates the material of the molecular beam, and in this embodiment, it is In, Ga, and Al as described above. Next, the molecular dose F _{x — sum} is given by an expression obtained by integrating the molecular beam intensity F _x with the irradiation time t, and is given by the expression (3) when the molecular beam is irradiated on the substrate from time t1 to time t2.

The subscript x of the molecular dose F _{x_sum} indicates the material of the molecular beam, and in this embodiment, it is In, Ga, and Al as described above. Since the Al composition ratio Y _Al is proportional to the molecular dose of the target material divided by the sum of the molecular doses of the respective materials, this molecular dose F _{x — sum} and the process state X are used to give the equation (4).

The Al composition ratio Y _Al is the product of the process state X obtained by dividing the molecular dose of Al by the sum of the molecular doses of the respective materials In, Ga and Al. Comparing equation (4) and equation (2), the transfer function H _Al when calculating the _Al composition ratio Y _Al is given by equation (5).

The transfer function H _Al is obtained by dividing the molecular dose of Al by the sum of the molecular dose of each material n, Ga and Al. The virtual device 13 calculates the calculation characteristic value Y _Al based on the transfer function H _Al and the process state X. The calculated calculation characteristic value Y _Al is transmitted to the measurement error removing unit 14, and when transmitted, the process proceeds from step s7 to step s8.

In step s8, which is a filter gain calculation step, the measurement error removal unit 14 corrects the process state X based on the measurement characteristic value measured by the measurement unit 12 and the calculation characteristic value calculated by the virtual device 13. Is calculated, that is, the filter gain K for feedback control of the process state X is calculated. The filter gain K is a gain for bringing the process state X used in the arithmetic expression of the virtual device 13 closer to the _true process state X _true . The true process state X _true is a process state in which the error due to the measurement error ν is removed from the process state based on the measurement calculation value, and is an actual process state in consideration of only the thin film forming process of the formed thin film.

Filter gain K is given by the variance of the estimated process states X _est and the error X _true -X _est is an estimated value of the true process state X _true true process conditions X _{tru e} is minimized. By calculating the filter gain K so that the variance of the error X _true −X _est is minimized, the estimated process state X _est calculated using the filter gain K as described above is brought closer to the _true process state X _true . be able to. Hereinafter, a method for calculating the filter gain K using the Kalman filter method will be described in detail.

First, a method for calculating the estimated process state _Xest will be described. In the thin film formation process, and calculates the estimated process states X _est for each measurement, and the estimated process states X _est to follow the variation of the true process conditions X _true. Therefore, the value of the estimated process state X _est varies with the number of measurements. Hereinafter, a calculation method of the estimated process state Xest at the tth measurement (t = 1, 2, 3,...) _{Will be} described. The number of times of measurement t is synonymous with the number of times the characteristic value of the thin film formed by the measurement unit 12 is measured. The process conditions of the measurement t th thin film X _t, measured t-th from the measurement t + 1 th process condition when moved in the X coefficients representing the time course of _t A _t, the process variations as w _t, process state X _{t + 1} is given by equation (6).

Process state X _{t + 1} is the product of the measurement t th process state X _t and the coefficient A _t, is given by the formula plus the variation w _t of the process. Factor A _t, unless aging, its value is 1.

Assuming that the estimated process state X _{t-1 / t-1} of the measurement t−1 is based on this, the calculation process state X _{t / t-1} of the measurement t time calculated by the equation (6) is 7).

When the calculation process state X _{t / t−1} is corrected based on the measurement characteristic value Y _t and the calculation characteristic value Y _{t / t−1} , the corrected process state X _{t / t} is obtained by using the filter gain K. It is given by equation (8).

The corrected process state X _{t / t} includes the calculation process state X _{t / t−1} and the difference Y _t −Y _{t / t−1} between the measured characteristic value Y _t and the calculated characteristic value Y _{t / t−1.} It is expressed as the sum of products with the filter gain K. The corrected process state X _{t / t} is an estimated value of the true process state, and is the estimated process state X _est for the t-th measurement.

Since the calculation characteristic value is calculated based on the equation (2), when the equation (2) is substituted into the equation (8), the estimated process state X _{t / t} is given by the equation (9).

The transfer function H _t is a transfer function determined based on the apparatus conditions at the time of measurement t, and the filter gain K _t is the filter gain K of measurement t. The estimated process state X _{t / t} includes the calculation process state X _{t / t−1} and the difference Y _t −H _t X _{t / t} between the measured characteristic value Y _t and the calculated characteristic value H _t × X _{t / t−1.} It is represented by the sum of products of ₋₁ and the filter gain K. The estimated process state X _{t / t} is calculated by this equation.

Next, a calculation method of the filter gain K will be described based on the estimation process state _Xest . The filter gain K is determined to a value that minimizes the variance E [(X _true −X _est ) ² ] of the estimation error X _true −X _est between the _true process state X _true and the estimated process state X _est . Here, E [A] represents the expected value of A. Assuming that the true process state X _true of the t-th measurement is X _t and the estimated process state X _est is X _{t / t} , these estimated errors X _err are expressed by Equation (10) based on Equation (9). Is calculated.

Estimation error X _err the difference calculation from the true process state X _t process state X _{t / t-1,} and the measured characteristic value Y _t and calculating characteristic values _{H t X t / t-1} Y t -H t X is a value obtained by subtracting the product of the _{t / t-1} and the filter gain K _t.

Here measured characteristic value Y _t, the characteristic values of the thin film formed in a true process conditions, that is, the actual characteristic values of the formed thin film in which measurement error ν is applied, the true process state X _t And the transfer function H _t and the measurement error ν _t are given by equation (11). The measurement error ν _t is a measurement error in the measurement at the measurement t.

The measurement characteristic value Y _t is a value obtained by adding a measurement error ν _t to the product of the transfer function H _t and the true process state X _t .
Substituting this measured characteristic value Y _t into equation (10) yields equation (12).

The estimated error X _err is obtained by subtracting the product of the filter gain Kt and the transfer function Ht from 1 and the filter gain from the product of the true process state X _t and the operation process state X _{t / t−1.} This is a value obtained by subtracting the product of K _t and the estimation error ν _t .
Based on this estimated error X _err , the variance E [X _err ² ] is given by equation (13).

The variance E [X _err ² ] of the estimated error X _err is ₁ in the variance E [(X _t −X _{t / t−1} ) ² ] between the true process state X _t and the operation process state X _{t / t−} 1. Product of the product of the filter gain K _t and the transfer function H _t subtracted from the square of the product, and the product of the square of the filter gain K _t and the variance E [ν _t ² ] of the measurement error ν _t Is the sum of

Distributed E [X _err ^2] Since a function of filter gain K _t, the variance E [X _err ^2] is minimized, the differential value of the filter gain K _t is 0 for distributed E [X _err ^2] That is, it is necessary to satisfy Expression (14).

When formula (14) is expanded based on formula (13), formula (15) is given.

Thus, the filter gain K _t is given by Expression (15) using the transfer function H _t , the variance E [(X _t −X _{t / t−1} ) ² ] and [ν _t ² ]. Substituting equation (15) into equation (13) gives equation (16).

If E [ν _t ² ] is eliminated from equation (16) using equation (15), equation (17) is given.

The variance E [(X _t −X _{t / t−1} ) ² ] is given by equation (18) based on equations (6) and (7).

Measurement error removal section 14, based on the equation (18) and (15), calculates the filter gain K _t. When the Kalman filter method is used, it is assumed that process variation and measurement error follow a normal distribution. Therefore, E [w _t ² ] is a setting constant representing the variance of the process variation w _t and is a constant input in advance by the user through the operation unit, and is given by Expression (19). E [ν _t ² ] is a setting constant representing the variance of the measurement error ν _t and is a constant input in advance by the user through the operation unit, and is given by Expression (20).

フィルタゲインＫ_tが演算されると、ステップｓ８からステップｓ９へ移行する。

When the filter gain K _t is calculated, the process proceeds from step s8 to step s9.

In step 9, which is a predicted characteristic value calculation step, the measurement error removing unit 14 calculates a predicted characteristic value Y _est based on the filter gain K _t calculated in step s8. Here, the calculation method of the prediction characteristic value Y _{est at} the t-th measurement (t = 1, 2, 3,...) _{Will be} described. First, the filter gain K _t calculated in step s8 is substituted into equation (8), and the estimated process state X _{t / t} for the t-th measurement is calculated. Thus, the estimated process state X _{t / t} from which the measurement error is removed can be calculated. Next, based on the estimated process state X _{t / t} , the t-th transfer function H _t measured and the equation (2), Y _{t / t} that is the predicted characteristic value Y _est for the measurement _t is calculated. ) Is obtained.

The predicted characteristic value Y _{t / t} is a product of the transfer function H _t and the estimated process state X _{t / t} and is a characteristic value from which measurement errors are removed. This way to calculate the predicted characteristic value Y _est, the program proceeds from step s9 to step s10.

In step s10 which is a re-execution confirmation process, it is determined whether or not to re-execute thin film formation. When the formation of the thin film is finished due to power being turned off, the thin film forming process is finished. When the thin film formation is re-executed, the calculated predicted characteristic value Y _est is transmitted to the condition setting unit 15, and the process proceeds from step s10 to step s11.

In step s11, which is a process state calculation step, the calculation process state X _{t + 1 / t} of the measurement t + 1 time is calculated using Expression (7) based on the estimated process state X _{t / t} . This calculation process state X _{t + 1 / t} is transmitted to the virtual device 13. The virtual device 13 calculates the calculation characteristic value Y _{t + 1 / t} using the calculation process state X _{t + 1 / t} at the measurement t + 1. When the computation process state X _{t + 1 / t} is transmitted to the virtual device 13, the process returns from step s11 to step s2.

When returning from step s11 to step s2, the condition setting unit 15 determines the apparatus conditions for the next thin film formation based on the input characteristic value R and the predicted characteristic value Y _est . Based on the determined apparatus conditions, the processes after step s3 proceed. The characteristic value of the thin film formed in this way is feedback-controlled, and the characteristic value of the formed thin film is brought closer to the input characteristic value R.

  In the present embodiment, the crystal growth process corresponds to a film formation process, and the virtual crystal growth process corresponds to a calculation process. The filter gain calculation step and the predicted characteristic value calculation step correspond to the measurement error removal step.

Hereinafter, the flow of calculation of the filter gain K _t , the predicted characteristic value Y _{t / t} and the estimated process state X _{t / t} will be specifically described. First, in step s2, the user operates the operation unit with a measurement error variance E [ν _t ² ], a process variation variance E [w _t ² ] and a coefficient A _t , and an arithmetic process that is a predicted value for the first measurement count. _{Predetermined} values are input as the state X _1/0 and the variance E [(X ₁ −X _1/0 ) ² ]. Here, the measurement error variance E [ν _t ² ] and the process variation variance E [w _t ² ] are tuning parameters of the filter gain K _t . When the value of the measurement error variance E [ν _t ² ] is set large, the filter gain K _t decreases, and when the value of the process variation variance E [w _t ² ] is set large, the filter gain K _t increases. For example, when the measurement accuracy is low, the value of the measurement error variance E [ν _t ² ] is set large. Further, for example, when the process fluctuation frequently occurs, the value of the process fluctuation variance E [w _t ² ] is set large. In addition, the calculation process state X _1/0 and the variance E [(X ₁ −X _1/0 ) ² ], which are the predicted values of the first measurement, are determined based on the characteristics at the time of condition determination.

In step s8, the filter gain K ₁ is given by equation (22) based on the input value.

フィルタゲインＫ_tが演算されると、ステップｓ８からステップｓ９へ移行する。

When the filter gain K _t is calculated, the process proceeds from step s8 to step s9.

In step s9, an estimated process state X _1/1 is calculated based on the filter gain K ₁ given by equation (22) and equation (8), and the estimated process state X _1/1 is given by equation (23). .

When the predicted characteristic value Y _1/1 is calculated based on the estimated process state X _1/1 given by the expression (23) and the expression (21), the predicted characteristic value Y _1/1 is given by the expression (24). .

When the predicted characteristic value Y _1/1 is calculated in this way, the process proceeds from step s9 to step s10, and when it is determined that re-enforcement is performed in step s10, the process proceeds to step s11.

In step s11, based on the estimated process state X _1/1, calculation process state X _2/1 is calculated, the arithmetic process state X _2/1 is given by Equation (25).

When the calculation process state X _2/1 is calculated and transmitted to the virtual device 13, the process returns from step s11 to step s2. When the apparatus conditions are determined again in step s2, and the process proceeds to step s8 through crystal growth and virtual crystal growth, in step s8, the estimation error X is based on the variance E [(X ₁ −X _1/0 ) ² ]. calculates the _err of variance E [X _err ^2]. The variance E [X _err ² ] of the estimation error is given by equation (26).

Based on the variance E [X _err ² ] given by the equation (26) and the equation (18),
When the variance E [(X ₂ −X _2/1 ) ² ] is calculated, the variance E [(X ₂ −X _2/1 ) ² ] is given by Equation (27).

Based on the variance E [(X ₂ −X _2/1 ) ² ] given by this equation (27), the filter gain K ₂ is calculated. Based on the filter gain K ₂ , the predicted characteristic value Y _2/2 and the calculation process state X _3/2 are calculated in each step, and the process returns to step s2. Similarly, based on the filter gain K _t (t = 3, 4, 5...), The predicted characteristic value Y _{t / t} and the calculation process state X _{t + 1 / t} are calculated in each step, and the thin film Move the characteristic value closer to the input characteristic value.

  Below, the effect which the thin film formation unit 10 of this Embodiment has is demonstrated. According to the thin film forming unit 10 of the present embodiment, a predicted characteristic value obtained by removing only the measurement error ν from the process error w and the measurement error ν can be calculated from the measurement characteristic value. In the device condition determination step, a new device condition is determined based on the calculated predicted characteristic value. Therefore, since the characteristic value of the thin film formed based on the new apparatus conditions does not include the measurement error ν as in the conventional technique, the deviation from the input characteristic value R is smaller than that in the conventional technique. Therefore, the quality of the formed thin film can be stabilized and the yield can be improved. Further, the quality of the semiconductor element including the thin film and the apparatus including the semiconductor element can be stabilized and the yield can be improved. In the predicted value characteristic value, only the measurement error ν is removed and the process error w is included. Therefore, the characteristic value of the thin film subjected to repeated feedback control can be close to the input characteristic value R. The input characteristic value R can be approximated by the input characteristic value R without converging to the characteristic value obtained by adding the measurement error ν to the input characteristic value R and the characteristic value obtained by subtracting the process error w from the input characteristic value R.

  According to the thin film formation unit 10 of the present embodiment, the process state in the process state calculation step is estimated based on the calculation characteristic value and the measurement characteristic value, and is used in the virtual crystal growth step based on this process state. The process state X of the calculation formula (2) is corrected. Even if the process state changes with time, the process state X of the calculation formula (2) used in the calculation step is corrected following the change in the process state. As a result, the calculation characteristic value becomes a calculation characteristic value that takes into account variations in the process state, and becomes a characteristic value approximated to the characteristic value of the actually formed thin film as compared with the case where the correction is not performed by following the calculation characteristic value. Since the predicted characteristic value is calculated based on the calculated characteristic value and the measured characteristic value, the predicted characteristic value can be made closer to the input characteristic value R, and determined based on the predicted characteristic value. The thin film formed according to the apparatus conditions can be brought close to the input characteristic value R.

  According to the thin film forming unit 10 of the present embodiment, a predicted characteristic value obtained by removing only the measurement error ν from the process state and the measurement error ν can be calculated. The condition setting unit 15 determines a new device condition based on the predicted characteristic value. Therefore, since the characteristic value of the thin film formed based on the new apparatus condition does not include the measurement error ν as in the conventional technique, the deviation from the input characteristic value R is smaller than that of the conventional technique. Thereby, the quality of the formed thin film can be stabilized, and the yield can be improved. Further, the quality of the semiconductor element including the thin film and the apparatus including the semiconductor element can be stabilized and the yield can be improved. Therefore, the characteristic value of the thin film subjected to repeated feedback control can be close to the input characteristic value R. The input characteristic value R can be approximated by the input characteristic value R without converging to the characteristic value obtained by adding the measurement error ν to the input characteristic value R and the characteristic value obtained by subtracting the process error w from the input characteristic value R.

  According to the thin film forming unit 10 of the present embodiment, the process state is estimated by the measurement error removing unit 14 based on the calculation characteristic value and the measurement characteristic value, and the calculation used in the virtual device 13 is based on the process state. The process state X of equation (2) is corrected. Even if the process state changes with time, the process state X of the calculation formula (2) used in the calculation step is corrected following the change in the process state. As a result, the calculation characteristic value becomes a calculation characteristic value that takes into account variations in the process state, and becomes a characteristic value approximated to the characteristic value of the actually formed thin film as compared with the case where the correction is not performed by following the calculation characteristic value. Since the predicted characteristic value is calculated based on the calculated characteristic value and the measured characteristic value, the predicted characteristic value can be made closer to the input characteristic value R, and determined based on the predicted characteristic value. The thin film formed according to the apparatus conditions can be brought close to the input characteristic value R.

According to the thin film forming unit 10 of the present embodiment, the filter gain K is determined so that the variance between the _true process state X _true and the estimated process state X _est is minimized. Therefore, the predicted characteristic value calculated by the estimated process state _Xest calculated using the filter gain K is a characteristic value obtained by removing the measurement error ν from the measurement characteristic value, and is a characteristic value including the process error w. The characteristic value of the thin film that does not include the measurement error ν, that is, the true characteristic value, is formed with the process value w included in the characteristic value. Therefore, the characteristic value of the formed thin film can be made closer to the input characteristic value R by including the process error w in the predicted characteristic value.

In the present invention, since the filter gain K is determined so that the variance between the _true process state X _true and the estimated process state X _est is minimized, the estimated process state X _est estimated based on the filter gain K is determined. Is close to the _true process state X _true . Therefore, based on the estimated process state X _est , the calculated predicted characteristic value Y _est approximates the true characteristic value that is the original characteristic value. Since the apparatus conditions are set based on the predicted characteristic value Y _est and the input characteristic value R, the characteristic value of the formed thin film can be approximated to the input characteristic value R.

  According to the thin film forming unit 10 of the present embodiment, the estimated process state is calculated for each number of measurements, and the next calculation process state is calculated based on the estimated process state. The virtual device 13 calculates the next calculation characteristic value using this calculation process state. That is, the calculation characteristic value is calculated by causing the calculation process state to follow the true process state every number of times of measurement. The calculation characteristic value calculated in this way is closer to the true characteristic value than when the calculation process state does not follow the true process state. As described above, since the predicted characteristic value is calculated based on the calculated characteristic value approximated by the true characteristic value, it is possible to calculate the predicted characteristic value that is more approximate to the true characteristic value.

  According to the thin film forming unit 10 of the present embodiment, the measurement error ν is removed using the Kalman filter method. Thereby, the calculation of the predicted measurement value can be realized, the characteristic value of the thin film formed by this calculation can be stabilized, and the quality of the thin film can be stabilized.

  FIG. 3 is a block diagram showing the functions of the thin film forming unit 10A according to the second embodiment of the present invention divided into blocks. The thin film forming unit 10A according to the second embodiment of the present invention is similar in configuration to the thin film forming unit 10 according to the first embodiment. Accordingly, the thin film forming unit 10A will be described only with respect to the configuration different from the actual thin film forming unit 10 of the first embodiment, and the same configuration will be denoted by the same reference numeral and description thereof will be omitted.

  The virtual device 13A is electrically connected to the measurement error removing unit 14, and has a function of calculating a characteristic value of a thin film to be formed based on device conditions and creating a virtual model of the thin film. The virtual device 13A includes a virtual setting unit 18, a molecular dose calculation unit 20, a composition ratio calculation unit 21, and a characteristic value calculation unit 22.

  The molecular dose calculation unit 20 is electrically connected to the virtual setting unit 18 and the composition ratio calculation unit 21 and is based on the virtual model of the substrate and the virtual model of the material set by the apparatus conditions and the virtual setting unit 18. It has a function to calculate. The molecular dose calculated by the molecular dose calculation unit 20 is transmitted to the composition ratio calculation unit 21. The composition ratio calculation unit 21 is electrically connected to the characteristic value calculation unit 22 and has a function of calculating the composition ratio of the thin film based on the molecular dose calculated by the molecular dose calculation unit 20. The composition ratio of the thin film calculated by the composition ratio calculation unit 21 is transmitted to the characteristic value calculation unit 22.

  The characteristic value calculator 22 calculates the calculated characteristic value of the thin film based on the composition ratio of the thin film calculated by the composition ratio calculator 21. The characteristic value of the thin film is a characteristic value that can be calculated based on the composition ratio of the material such as the film thickness and the oscillation wavelength of light. The calculated characteristic value calculated by the characteristic value calculation unit 22 is transmitted to the measurement error removal unit 14. The characteristic value calculation unit 22 calculates a calculation characteristic value based on an arithmetic expression, similarly to the virtual device 13A of the first embodiment. This arithmetic expression is also corrected by the measurement error removing unit 14 as in the thin film forming unit 10A of the first embodiment.

  FIG. 4 is a flowchart showing the procedure of the thin film forming process of the thin film forming unit 10A. In the thin film forming process, the characteristic value of the thin film to be formed is feedback-controlled based on the measured characteristic value measured by the measurement unit 12 and the calculated characteristic value calculated by the virtual device 13A. The thin film forming process of the thin film forming unit 10A of the second embodiment is similar to the thin film forming process of the thin film forming unit 10A of the first embodiment. Therefore, the description of the thin film forming process will be made only with respect to the procedure different from the thin film forming process of the thin film forming unit 10A of the first embodiment, and the same procedure will be denoted by the same reference numeral. Omitted.

  When the substrate model and the material model are set in step s6, the process proceeds to step a1. In step a1, which is a molecular dose calculation step, the molecular weight calculation unit calculates the molecular dose irradiated to the substrate based on the virtual model of the substrate and material determined in step s6 and their temperature states. The molecular beam intensity Fx of each material is given by the above equation (1) as a constant A, an Arrhenius coefficient B, a temperature T of the material evaporation source cell, a temperature rise width α of the cell, and an irradiation time t.

The subscript x of the molecular beam intensity F _x indicates the material of the molecular beam, and in this embodiment, it is In, Ga, and Al as described above. The molecular dose F _{x — sum} is given by an equation obtained by integrating the molecular beam intensity F _x with the irradiation time t, and is given by the above equation (3) when the molecular beam is irradiated on the substrate from the time t1 to the time t2.

The subscript x of the molecular dose F _{x_sum} indicates the material of the molecular beam, and in this embodiment, it is In, Ga, and Al as described above. When the molecular dose of each material is calculated, the process proceeds from step a1 to step a2.

In step a2, which is a composition ratio calculation step, the composition ratio calculation unit 21 calculates the composition ratio of the thin film based on the molecular dose of each material. For example the composition ratio R _Al of Al, In, Ga and Al molecules dose F _{In_sum,} when the F _{Ga_sum} and F _{Al_sum,} the composition ratio _{R x} of each material is given by equation (17). The composition ratio R _x is an arithmetic composition ratio not including the process state X, unlike the _Al composition ratio Y _Al given by the expression (4) in the first embodiment.

各材料の組成比を演算すると、ステップａ２からステップａ３へ移行する。

When the composition ratio of each material is calculated, the process proceeds from step a2 to step a3.

In step a3, which is a characteristic value calculation step, a calculated characteristic value is calculated based on the composition ratio calculated in step a2. The characteristic value calculator 22 calculates characteristic values such as film thickness and light oscillation wavelength based on the molecular dose calculated in step a1 and the material composition ratio _Rx calculated in step a2. As an example, when a semiconductor is manufactured by crystal growth, the oscillation wavelength λ of light oscillated from the semiconductor will be specifically described. The calculation characteristic value Y which is the oscillation wavelength λ of light is given by the equation (29).

Here, h is the Planck constant, c is the speed of light, and E _g is the band gap. Here, the band gap E _g is a value determined based on the composition ratio. The calculation characteristic value is calculated using this equation (29). When the characteristic value is calculated, the process proceeds from step a3 to step s8.

  FIG. 5 is a graph showing the deviation between the characteristic value of the formed thin film and the input characteristic value R. In FIG. 5, the characteristic value and the input characteristic value R of the formed thin film are photoluminescence wavelengths, and FIG. 5 shows the deviation of the characteristic value for each batch. In FIG. 5, the vertical axis indicates the deviation, and the horizontal axis indicates the batch number, so-called lot number. In batch numbers 1 to 20 (hereinafter sometimes referred to as “conventional group”) in FIG. 5, characteristic values of the thin films formed by the conventional thin film forming unit 10A are shown, and batch numbers 21 to 40 ( Hereinafter, “this working group”) shows characteristic values of the thin film formed by the thin film forming process of the present embodiment. In the conventional group, a batch number t is given to the thin film formed at the measurement t-th time, and in this working group, a batch number (t + 20) is given to the thin film formed at the measurement t-th time.

  In the conventional group, feedback control is performed so that the characteristic value of the thin film converges to the input characteristic value R, but the deviation varies even if the number of feedback control is repeated. Therefore, the characteristic value of the thin film varies, and the quality of the thin film is not stable. Further, since the deviation does not converge toward 0, it cannot be said that a thin film having a characteristic value approximate to the input characteristic value R is formed stably. On the other hand, in the present embodiment, the deviation converges to 0 each time the number of feedback control is repeated as compared with the conventional group. Compared with the conventional group, there is little variation in the characteristic value of the thin film, and the quality of the thin film is stable. Further, since the deviation converges to 0, a thin film having a characteristic value that approximates the input characteristic value R is stably formed.

  FIG. 6 is a graph showing the frequency distribution of the deviation between the characteristic value of the formed thin film and the input characteristic value R. In FIG. 6, the vertical axis represents frequency and the horizontal axis represents deviation. In FIG. 6, the alternate long and short dash line indicates the frequency of deviation of the conventional group, and the solid line indicates the frequency of deviation of the present working group. In the conventional group, it is distributed in the range from deviation-3 to deviation-3, and in this working group, it is distributed in the range from deviation-1 to deviation-1. From this result, it can be seen that the characteristic value of the thin film is formed to follow the input characteristic value R in the present working group than in the conventional group.

  FIG. 7 is a graph showing the characteristic values of the thin film formed by the thin film forming unit 10A at each time. FIG. 8 is a graph showing the characteristic values of the thin film formed by the thin film forming unit 10A at each time. FIG. 7 is a graph showing the time change of the true characteristic value 30, the measured characteristic value 31, and the estimated characteristic value 32 when the process variation is large and the measurement error is small, and the vertical axis indicates the deviation of the characteristic value. The horizontal axis indicates time. FIG. 8 is a graph showing the time changes of the true characteristic value 33, the measured characteristic value 34, and the estimated characteristic value 35 when the process variation is small and the measurement error is large. The vertical axis indicates the deviation of the characteristic value. The axis indicates the time. As shown in FIG. 7, with the process variation, the true characteristic value 30, the observed characteristic value 31, and the estimated characteristic value 32 vary greatly, and these characteristic values 30, 31, 32 are affected by the process variation. I can see that. On the other hand, as shown in FIG. 8, when the process variation is small and the measurement error is large, the measurement characteristic value 34 varies greatly with the measurement error. However, the estimated characteristic value 31 is the measurement error. Has been removed, so it is not affected by measurement errors and does not vary much. The estimated characteristic value 35 follows the true characteristic value 33. In this way, it is possible to perform feedback control so that only the measurement error is removed and the characteristic value of the formed thin film is brought close to the true characteristic value.

  Below, the effect which 10A of thin film formation units show | plays is demonstrated. According to the thin film formation unit 10A of the present embodiment, the molecular weight calculation unit, the composition ratio calculation unit 21 and the characteristic value calculation unit 22 can calculate the calculation characteristic value of the thin film formed by the MBE apparatus. The unit 10A can be realized by an MBE device.

According to the thin film forming unit 10A of the present embodiment, the calculation of the molecular dose F _{x_sum} can be realized. The composition ratio _Rx of each material can be calculated based on the molecular dose _{Fx_sum,} and the characteristic value Y of the thin film can be calculated based on the composition ratio _Rx . Therefore, by calculating the molecular dose _{Fx_sum} , it is possible to calculate the calculation characteristic value Y of the thin film.

  According to the thin film forming unit 10A of the present embodiment, the same effects as the thin film forming unit 10A of the first embodiment are produced.

  In the present embodiment, although an MBE apparatus is used for the crystal growth unit 17, it is not limited to the MBE apparatus. A chemical vapor deposition (substantially CVD) apparatus may be used instead of the MBE apparatus. Table 1 shows the apparatus conditions in this case.

  As an apparatus condition, a gas flow rate is used instead of the cell temperature. The control target of the MBE apparatus is the molecular beam intensity, whereas the control target of the CVD apparatus is the gas flow rate. The measurement items of the MBE apparatus and the CVD apparatus mean characteristic values to be measured, and both are the film thickness or the oscillation wavelength λ.

  The case where a CVD apparatus is used will be described in detail. The film thickness Y of the virtual model in the virtual apparatus 13 is expressed as follows, assuming that the gas flow rate per unit time is F (t), the transfer function is H, and the process state is X. (30).

Therefore, the transfer function H is given by equation (31).

Using this transfer function H, the thin film forming unit 10, 10 </ b> A performs the thin film forming process, so that the thin film forming unit 10, 10 </ b> A can be realized even when the CVD apparatus is used for the crystal growth unit 17. it can. However, even when a CVD apparatus is used, the characteristic value is not limited to the film thickness, and may be the film formation amount per unit time and the light oscillation wavelength λ.

It is a block diagram which divides | segments and shows the function of the thin film formation unit 10 which is the 1st Embodiment of this invention for every block. 4 is a flowchart illustrating a procedure of a thin film forming process of the thin film forming unit 10. It is a block diagram which divides and shows the function of 10 A of thin film formation units of the 2nd Embodiment of this invention for every block. It is a flowchart which shows the procedure of the thin film formation process of 10 A of thin film formation units. It is a graph which shows the deviation of the characteristic value of the formed thin film, and the input characteristic value R. It is a graph which shows the frequency distribution of the deviation of the characteristic value of the formed thin film, and the input characteristic value R. It is a graph which shows the characteristic value of the thin film formed by 10 A of thin film formation units at each time. It is a graph which shows the characteristic value of the thin film formed by 10 A of thin film formation units at each time. It is a block diagram which shows roughly the structure of the thin film formation unit 1 of the conventional 1st technique.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10,10A Thin film formation unit 11 Thin film formation apparatus 12 Measurement part 13 Virtual thin film formation apparatus 14 Measurement error removal part 15 Condition setting part 20 Molecular weight calculation part 21 Composition ratio calculation part 22 Characteristic value calculation part

Claims (6)

A film forming process in which a thin film is formed using a thin film forming apparatus, wherein the thin film is formed in a state in which a process error is included in the characteristic value;
A measurement process for measuring a measurement characteristic value of a thin film to be formed, wherein the measurement characteristic value is measured in a state including a measurement error; and
Based on the apparatus conditions of the thin film forming apparatus in the film forming step, a calculation step for calculating a calculation characteristic value of the thin film to be formed by calculation,
Based on the calculation characteristic value calculated in the calculation step and the measurement characteristic value measured in the measurement step, a predicted characteristic value obtained by removing only the measurement error from the measurement characteristic value out of the process error and the measurement error is calculated. A measurement error elimination process;
And an apparatus condition determining step for determining a new apparatus condition of the thin film forming apparatus in the film forming process when the thin film is formed next based on the predicted characteristic value calculated in the measurement error removing process. Method for forming a thin film.   In the measurement error elimination process, the film formation state in the film formation process is further estimated based on the measurement characteristic value and the calculation characteristic value, and the film formation state in the estimated film formation process is estimated in the calculation process. 2. The thin film forming unit according to claim 1, wherein an arithmetic expression used for calculating the arithmetic characteristic value is corrected. A thin film forming apparatus for forming a thin film with a process error included in the characteristic value;
A measuring means for measuring the characteristic value of the formed thin film in a state including a measurement error;
Based on the apparatus conditions of the thin film forming apparatus when forming a thin film with the thin film forming apparatus, an arithmetic device that obtains a characteristic value of the thin film to be formed by calculation,
Measurement error elimination that calculates the predicted characteristic value in which only the measurement error is removed from the measurement characteristic value among the process error and measurement error based on the characteristic value calculated by the arithmetic unit and the characteristic value measured by the measuring means And a device condition determining means for determining a new device condition of the thin film forming device when the thin film is next formed based on the predicted characteristic value calculated by the measurement error removing device. unit.   The measurement error removing means further estimates the film forming state of the thin film forming apparatus based on the measured characteristic value and the calculated characteristic value, and is measured by the measuring means based on the estimated film forming state of the thin film forming apparatus. 3. The thin film forming unit according to claim 2, wherein an arithmetic expression used for calculating the arithmetic characteristic value by the arithmetic device is corrected based on the obtained characteristic value and the characteristic value calculated by the arithmetic device. The thin film forming apparatus is a molecular beam epitaxial apparatus that forms a thin film by irradiating a substrate with a plurality of molecular beams.
The arithmetic device includes a molecular dose calculator that calculates a molecular dose based on the molecular beam intensity of each molecular beam irradiated on the substrate, a composition ratio calculator that calculates a composition ratio of a thin film based on the molecular dose, The thin film forming unit according to claim 3, further comprising a characteristic value calculation unit that calculates a calculation characteristic value of the thin film based on the composition ratio. The apparatus conditions include the temperature T of each molecular beam deposition source, the temperature rise α of each deposition source and the irradiation time t of each molecular beam,
The molecular dose calculation unit is based on the constant A, the Arrhenius coefficient B, and the apparatus conditions determined for each material of the molecular beam.

Calculating the molecular beam intensity F _x by equation (1), further by integrating the irradiation time t the molecular beam intensity F _x, the thin film forming unit according to claim 5, wherein computing the molecular dose.

Images