JP4329474B2 - Thin film formation method - Google Patents

Description

  The present invention relates to a thin film forming method for forming a thin film on a main surface of a semiconductor substrate.

  Conventionally, there is a CVD method (chemical vapor deposition method) as a method for forming a thin film on the surface of a substrate, which is applied to a thin film forming apparatus for forming an epitaxial layer on the main surface of a semiconductor substrate.

  As such a thin film forming apparatus, as shown in FIGS. 2A and 2B, there is a thin film forming apparatus 1 including a single wafer type vapor phase growth apparatus 2 (see, for example, Patent Document 1). The vapor phase growth apparatus 2 includes a reaction furnace 3 in which a semiconductor substrate (not shown) is disposed. A disc-shaped susceptor 20 on which a semiconductor substrate is placed is rotatably provided inside the reaction furnace 3. In addition, for example, four source gas supply ports 3d that open in the gas flow direction X are provided at one end of the reaction furnace 3, and a gas discharge port 3e is provided at the other end. A purge gas supply port (not shown) for supplying purge gas to a space below the susceptor 20 is provided below the source gas supply port 3d.

  In addition, a raw material gas supply device 4 is connected to the reaction furnace 3 via a raw material gas supply port 3d. The source gas supply device 4 includes a gas supply pipe 40 that forms a gas flow of a mixed gas containing the source gas and the carrier gas along the surface of the semiconductor substrate. One end of the gas supply pipe 40 is connected to gas sources 41 and 42 of the source gas and carrier gas, and the flow rates of the source gas and carrier gas are controlled by mass flow controllers 43 and 44, respectively. The other end of the gas supply pipe 40 is divided into four parallel flow paths 40a to 40d, and the ends of the flow paths 40a to 40d are connected to the source gas supply port 3d. Among these four flow paths 40a to 40d, the inner two flow paths 40a and 40b and the outer two flow paths 40c and 40d supply mixed gas at the same flow rate, respectively, The flow rate distribution on the semiconductor substrate in the direction Y substantially orthogonal to the gas flow direction X is controlled by adjusting the flow rate on the outside and the flow rate on the outside by the valves 45 and 46.

By the way, in order to form a thin film having a uniform film thickness by the single wafer type vapor phase growth apparatus 2 as described above, the flow rate of the carrier gas and the semiconductor substrate each time the components of the reaction furnace 3 are replaced for maintenance. It is necessary to change the values of a plurality of parameters such as the flow rate distribution and the purge gas flow rate in accordance with the reaction furnace 3 after the parts replacement.
JP 2000-68215 A

  However, in order to reduce the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate and bring it closer to the ideal film thickness distribution, conventionally, trial and error are repeated while gradually changing parameters that affect the film thickness distribution. Gradually approaching the ideal film thickness distribution. For this reason, a large number of trial semiconductor substrates used during trial and error are required, and a long time is required for trials, which is one of the factors that reduce the operating rate of the apparatus.

  The subject of this invention is providing the thin film formation method which can form the thin film with a small film thickness distribution easily.

The thin film formation method of the present invention is a thin film formation method of forming a thin film on the main surface of a semiconductor substrate by vapor phase growth,
A thin film is formed on the main surface of separate semiconductor substrates in the first reactor by applying a plurality of set value groups obtained by changing and combining values of a plurality of parameters that change the film thickness distribution of the thin film to be formed. A data creation step of creating film thickness distribution data consisting of a plurality of film thickness distributions corresponding to the plurality of set value groups by measuring a film thickness distribution of each formed thin film,
A film thickness distribution measuring step of applying a set of set values in the second reactor to form a thin film on the main surface of the semiconductor substrate and measuring the film thickness distribution M of the thin film;
The predicted optimum set value group C that can minimize the distance d1 between the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate and the ideal film thickness distribution I in the second reactor is the film thickness. A predicted optimum set value group calculating step of calculating from distribution data and an actual value of the film thickness distribution M;
Applying the predicted optimum set value group C, and forming a thin film on the main surface of the semiconductor substrate by vapor phase growth in the second reactor ,
The predicted optimum set value group calculation step includes:
A first process of calculating a virtual film thickness distribution S1 of a thin film to be formed on the main surface of the semiconductor substrate based on the film thickness distribution data when applying the set of set value groups;
A second process of calculating a virtual film thickness distribution S2 of a thin film to be formed when another set of set value groups is applied, based on the film thickness distribution data;
A film thickness variation distribution H obtained by changing the parameter from the one set of set value groups to the other set of set value groups is obtained from a difference between the virtual film thickness distribution S2 and the virtual film thickness distribution S1, and A third process for predicting an expected film thickness distribution R when the other set of set value groups is applied by adding a film thickness change distribution H to the film thickness distribution M;
The second process and the third process are performed a plurality of times while changing the other set of set values in the second process, and the distance d2 between the expected film thickness distribution R and the ideal film thickness distribution I is minimized. A set value group to be obtained is set as the predicted optimum set value group C, and
Following the third process,
Applying the predicted optimum set value group C, forming a thin film on the main surface of the semiconductor substrate by vapor phase growth in the second reactor, and measuring the film thickness distribution M ′ of the thin film;
A fifth process for calculating a distance d3 between the film thickness distribution M ′ and the ideal film thickness distribution I;
A virtual film thickness distribution S1 ′ of a thin film to be formed on the main surface of the semiconductor substrate when the predicted optimum set value group C is applied, and a thin film to be formed when another set of set value groups is applied. A virtual film thickness distribution S2 ′ of the first processing based on the film thickness distribution data;
A film thickness change distribution H ′ obtained by changing the parameter from the predicted optimum set value group C to the other set of set values is determined from a difference between the virtual film thickness distribution S2 ′ and the virtual film thickness distribution S1 ′. And a seventh process for predicting an expected film thickness distribution R ′ when the other set of set values is applied by adding the film thickness change distribution H ′ to the film thickness distribution M ′. ,
An eighth process of calculating a distance d4 between the expected film thickness distribution R ′ and the ideal film thickness distribution I;
The other set of set value groups in the sixth process is changed and the sixth process to the eighth process are performed a plurality of times to obtain a set value group C ′ that minimizes the distance d4. And a ninth process for resetting the set value group C ′ as the predicted optimum set value group C when the distance d4 _min and the distance d3 satisfy a predetermined criterion .

According to the present invention, the predicted optimum set value group C that can minimize the distance d1 between the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate and the ideal film thickness distribution I in the second reactor . Is uniquely calculated from the film thickness distribution data and the actually measured value of the film thickness distribution M, and by applying the predicted optimum set value group C and performing vapor phase growth, the film thickness distribution on the main surface of the semiconductor substrate is calculated. Small thin films can be formed. Therefore, unlike the prior art, trial and error in which the parameter value is changed each time a thin film is formed is not repeated, so that the consumption of the semiconductor substrate can be suppressed and the time required for the trial can be shortened. That is, a thin film having a small film thickness distribution can be easily formed.

Here, the film thickness variation distribution between the film thickness formed when a certain set value group is applied and the film thickness formed when another set value group is applied is related to the difference between the reactors. It is known from many experiments that they are consistent.
Therefore, the predicted optimum set value group calculation step in the thin film forming method of the present invention,
A first process of calculating a virtual film thickness distribution S1 of a thin film to be formed on the main surface of the semiconductor substrate based on the film thickness distribution data when applying a set of set value groups;
A second process for calculating a virtual film thickness distribution S2 of a thin film to be formed when another set of set value groups is applied, based on the film thickness distribution data;
A film thickness change distribution H by changing a parameter from one set of set value groups to another set of set value groups is obtained from the difference between the virtual film thickness distribution S2 and the virtual film thickness distribution S1, and the film thickness change amount A third process for predicting an expected film thickness distribution R when another set of set value groups is applied by adding the distribution H to the film thickness distribution M;
A set value group in which the distance d2 between the expected film thickness distribution R and the ideal film thickness distribution I is minimized by performing the second process and the third process a plurality of times while changing another set of set value groups in the second process. look, shall be the predicted optimal set value group C this.
Thereby , as shown in FIG. 1, it is possible to calculate an optimum set value group that is expected to minimize the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate. Therefore, the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate can be easily minimized.

The predicted optimum set value group calculating step in the thin film forming method of the present invention is performed after the third process.
Applying a predicted optimum set value group C, forming a thin film on the main surface of the semiconductor substrate by vapor phase growth in the second reactor , and measuring the film thickness distribution M ′ of the thin film;
A fifth process for calculating a distance d3 between the film thickness distribution M ′ and the ideal film thickness distribution I;
The virtual film thickness distribution S1 ′ of the thin film to be formed on the main surface of the semiconductor substrate when the predicted optimum set value group C is applied, and the thin film to be formed when another set of set value groups are applied. A sixth process of calculating the virtual film thickness distribution S2 ′ based on the film thickness distribution data;
A film thickness change distribution H ′ obtained by changing the parameter from the predicted optimal set value group C to the other set of set values is obtained from the difference between the virtual film thickness distribution S2 ′ and the virtual film thickness distribution S1 ′. A seventh process for predicting an expected film thickness distribution R ′ when the other set of set value groups is applied by adding the film thickness variation distribution H ′ to the film thickness distribution M ′;
An eighth process for calculating a distance d4 between the expected film thickness distribution R ′ and the ideal film thickness distribution I;
By changing another set of set value groups in the sixth process and performing the sixth to eighth processes a plurality of times, a set value group C ′ that minimizes the distance d4 is obtained, and the distance d4 _min and the distance at this time If the d3 and satisfies a predetermined criterion, Ru and a ninth process to reset the set value group C 'as expected optimal setting value group C.
As a result , even if a measurement error occurs in the film thickness distribution actual measurement step or a prediction error occurs in the predicted optimum set value group calculation step, the semiconductor device can be obtained by resetting the expected optimum set value group C. The film thickness distribution of the thin film formed on the main surface of the substrate can be minimized.

Here, when forming an epitaxial wafer for an individual semiconductor element in which it is important that the film thickness is constant in each place, the ideal film thickness distribution I is preferably a constant value T.
When forming an epitaxial wafer for an integrated circuit in which surface flatness is important, the ideal film thickness distribution I is preferably a distribution obtained by subtracting the thickness distribution of the semiconductor substrate from a constant value T ′.

  The parameters in the thin film forming method of the present invention are as follows: the flow rate of the carrier gas that forms the atmosphere in the reaction furnace; the flow rate of the purge gas that purges the space below the susceptor that supports the semiconductor substrate; and the flow direction of the carrier gas It is preferable that the gas flow rate distribution in the reactor in the substantially orthogonal direction.

According to the present invention, the predicted optimum set value group C that can minimize the distance d1 between the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate and the ideal film thickness distribution I in the second reactor . Is uniquely calculated from the film thickness distribution data and the actually measured value of the film thickness distribution M, and by applying the predicted optimum set value group C to perform vapor phase growth, the film thickness distribution is small on the main surface of the semiconductor substrate. A thin film can be formed. Therefore, unlike the prior art, trial and error in which the parameter value is changed each time a thin film is formed is not repeated, so that the consumption of the semiconductor substrate can be suppressed and the time required for the trial can be shortened. That is, a thin film having a small film thickness distribution can be easily formed.

  Hereinafter, embodiments of a thin film forming method according to the present invention will be described with reference to the drawings. In the present embodiment, the thin film forming method will be described as a method of forming a silicon epitaxial layer on the main surface of a semiconductor substrate made of silicon single crystal.

An example of the thin film formation apparatus 1 used for the thin film formation method of this invention is shown in FIG. As shown in this figure, the thin film forming apparatus 1 includes a vapor phase growth apparatus 2.
The vapor phase growth apparatus 2 is a single wafer type in the present embodiment, and includes a disk-shaped susceptor 20.
A counterbore 20a is formed on the upper surface of the susceptor 20, and a silicon single crystal substrate is placed in the counterbore 20a. A support member 21 is in contact with the lower surface of the susceptor 20, thereby supporting the susceptor 20. A rotation drive device (not shown) is connected to the support member 21, and the susceptor 20 is rotated by driving the rotation drive device.
The susceptor 20 is disposed inside the reaction furnace 3. The top wall 3a and the bottom wall 3b of the reaction furnace 3 are made of translucent quartz, and a heating device (not shown) that radiates toward the inside of the reaction furnace 3 through the top wall 3a and the bottom wall 3b. Are arranged above and below the reaction furnace 3.

The side wall 3c of the reaction furnace 3 has a plurality of source gas supply ports 3d for supplying a source gas together with a carrier gas, a purge gas supply port (not shown) for supplying a purge gas for purging a space below the susceptor 20, and a reaction A gas discharge port 3e for discharging the gas in the furnace 3 is formed. The source gas supply port 3d and the gas discharge port 3e are located at one end and the other end of the reaction furnace 3, and all the source gas supply ports 3d are open in the gas flow direction X. The purge gas supply port is located below the source gas supply port 3d. In the present embodiment, four source gas supply ports 3d and two gas discharge ports 3e are formed in the side wall 3c. Further, trichlorosilane or the like can be used as the source gas, H ₂ gas or the like can be used as the carrier gas, and H ₂ gas or the like can be used as the purge gas.
A raw material gas supply device 4 for supplying a raw material gas is connected to the raw material gas supply port 3d, and a purge gas supply device (not shown) for supplying a purge gas is connected to the purge gas supply port.

The source gas supply device 4 includes a gas supply pipe 40 that forms a gas flow of a mixed gas composed of source gas and carrier gas along the surface of the silicon single crystal substrate.
One end of the gas supply pipe 40 is connected to the gas sources 41 and 42 of the source gas and the carrier gas, and the flow rates of the source gas and the carrier gas can be controlled by the mass flow controllers 43 and 44, respectively. Yes. Here, the flow rate of the carrier gas is one of the parameters for changing the film thickness distribution of the silicon epitaxial layer. When the flow rate of the carrier gas is increased, the thickness of the silicon epitaxial layer formed in the central portion of the silicon single crystal substrate is increased. Tend to be larger.
The other end of the gas supply pipe 40 is divided into four parallel flow paths 40a to 40d, and the ends of the flow paths 40a to 40d are connected to the source gas supply port 3d. Among these four flow paths 40a to 40d, the inner two flow paths 40a and 40b and the outer two flow paths 40c and 40d supply mixed gas at the same flow rate, respectively, The flow rate distribution on the silicon single crystal substrate in the direction Y substantially orthogonal to the gas flow direction X can be controlled by adjusting the flow rate on the outside and the flow rate on the outside by the valves 45 and 46. Here, this flow rate distribution is one of the parameters for changing the film thickness distribution of the silicon epitaxial layer. Specifically, the ratio of the flow rates of the outer flow paths 40c, 40d to the flow rates of the inner flow paths 40a, 40b. Increases, the film thickness of the silicon epitaxial layer formed in the central portion of the silicon single crystal substrate tends to be small, and the film thickness of the silicon epitaxial layer formed in the outer peripheral portion tends to increase.

  The purge gas supply device can control the flow rate of the purge gas. Here, the flow rate of the purge gas is one of the parameters for changing the film thickness distribution of the silicon epitaxial layer. When the flow rate of the purge gas increases, the film thickness of the silicon epitaxial layer formed on the outer peripheral portion of the silicon single crystal substrate decreases. Tend to be.

  Next, the thin film formation method according to the present invention will be described. The thin film forming method in the present embodiment is a method of forming a silicon epitaxial layer having a uniform film thickness in the reaction furnace 3 (hereinafter referred to as the reaction furnace 3P for convenience) after replacement of parts such as the susceptor 20. .

First, a silicon epitaxial layer is formed on the main surfaces of a plurality of silicon single crystal substrates in a reaction furnace 3 (hereinafter referred to as a reaction furnace 3Q for the sake of distinction from the reaction furnace 3P for convenience).
Specifically, first, a silicon single crystal substrate is placed in the counterbore 20a of the susceptor 20. Next, the mounted silicon single crystal substrate is heated to 1100 to 1200 ° C. by the heating device, and the susceptor 20 and the silicon single crystal substrate are rotated by the rotation driving device via the support member 21. In this state, a raw material gas is supplied into the reaction furnace 3Q with a predetermined composition and flow rate together with a carrier gas, thereby vapor-phase-growing the silicon epitaxial layer. At this time, a plurality of parameters for changing the film thickness distribution of the silicon epitaxial layer, in this embodiment, the flow rate of the carrier gas and the ratio of the flow rates of the outer flow paths 40c and 40d to the flow rates of the inner flow paths 40a and 40b. And a set value group that combines the flow rate of the purge gas is changed for each silicon single crystal substrate. Specifically, the values of these parameters are changed by controlling the mass flow controllers 43 and 44, the valves 45 and 46, and the purge gas supply device.

  Next, by measuring the film thickness distribution of each formed silicon epitaxial layer, as shown in FIG. 3, film thickness distribution data including a plurality of film thickness distributions corresponding to a plurality of set value groups is created ( Data creation process). Thereby, the film thickness distribution corresponding to every set value group can be obtained complementarily. This film thickness distribution data is data unique to the reactor 3Q, and even when the same set value group is applied to the reactor 3P, the film thickness distribution of the formed silicon epitaxial layer does not necessarily match. . Further, in FIG. 3, the film thickness distribution when the flow rate of the purge gas is constant and the flow rate of the carrier gas and the ratio of the flow rates of the outer flow paths 40c and 40d to the flow rates of the inner flow paths 40a and 40b are changed. Only the changes in are enlarged and shown in detail.

  Next, a set of set values is applied in the reactor 3P to form a silicon epitaxial layer on the main surface of the silicon single crystal substrate, and the film thickness distribution M of the silicon epitaxial layer is measured (film thickness distribution measurement). Process). This film thickness distribution measurement step may be performed before the data creation step.

ここで、ｎは予想膜厚分布Ｒ及び理想膜厚分布Ｉから抽出した点（位置）の個数、ｉは抽出した点の番号であり、同じ添え字ｉが付されたＲｉとＩｉとは同じ位置の膜厚を示すものとする。 Next, an expected optimum set value group C that can minimize the distance d1 between the film thickness distribution of the silicon epitaxial layer formed on the main surface of the silicon single crystal substrate and the ideal film thickness distribution I in the reaction furnace 3P. The search is performed based on the film thickness distribution data and the actually measured value of the film thickness distribution M (predicted optimum set value group calculation step).
Specifically, first, when applying the set of set values in the reactor 3Q, the virtual film thickness distribution S1 of the silicon epitaxial layer to be formed on the main surface of the silicon single crystal substrate is represented by the film thickness distribution data. (First process).
Next, a virtual film thickness distribution S2 of the silicon epitaxial layer to be formed when another set of set value groups is applied in the reaction furnace 3Q is calculated based on the film thickness distribution data (second process).
Next, the film thickness change distribution H by changing the parameter from the one set of set value groups to the other set of set value groups is obtained from the difference between the virtual film thickness distribution S2 and the virtual film thickness distribution S1. Ask. The film thickness change distribution H is a constant value regardless of the difference between the reaction furnace 3P and the reaction furnace 3Q. Next, the film thickness variation distribution H is added to the film thickness distribution M actually measured in the film thickness distribution actual measurement step (third process). Thereby, as shown in FIG. 1, the expected film thickness distribution R in the case of applying the other set of set values in the reaction furnace 3P is expected.
Thereafter, the other set of set values in the second process is changed, and the second process and the third process are performed a plurality of times. At this time, every time the third process is performed, the distance d2 between the expected film thickness distribution R and the ideal film thickness distribution I is calculated, and the next set value set is determined so that the distance d2 becomes smaller. Then, as shown in FIG. 4B, setting is made when the distance d2 between the expected film thickness distribution R and the ideal film thickness distribution I increases, that is, when the distance d2 becomes minimum, regardless of which parameter is changed. A value group is set as a predicted optimum set value group C. The distance d2 between the expected film thickness distribution R and the ideal film thickness distribution I can be defined as follows, for example.

Here, n is the number of points (positions) extracted from the expected film thickness distribution R and the ideal film thickness distribution I, i is the number of the extracted points, and Ri and Ii with the same subscript i are the same. The film thickness at the position shall be indicated.

  Then, the calculated optimum set value group C is applied, and a silicon epitaxial layer is formed on the main surface of the silicon single crystal substrate in the reaction furnace 3P (thin film forming step). Thereby, a silicon epitaxial layer having a film thickness distribution similar to the expected film thickness distribution R, that is, a film thickness distribution M ′ close to the ideal film thickness distribution I is formed. However, the film thickness distribution M ′ of the silicon epitaxial layer formed at this time can be made closer to the ideal film thickness distribution I due to the measurement error in the film thickness distribution actual measurement process, the prediction error in the predicted optimum set value group calculation process, and the like. There is still a possibility.

Therefore, in the present embodiment, before the thin film formation step, the predicted optimum set value group C is reset to a better set value group as follows.
That is, first, the predicted optimum set value group C is applied to form a silicon epitaxial layer on the main surface of the silicon single crystal substrate in the reaction furnace 3P, and the film thickness distribution M ′ is actually measured (fourth process).
Next, a distance d3 between the film thickness distribution M ′ and the ideal film thickness distribution I is calculated (fifth process).
Next, when applying the predicted optimal set value group C, applying the virtual film thickness distribution S1 ′ of the silicon epitaxial layer to be formed on the main surface of the silicon single crystal substrate and another set of set value groups The virtual film thickness distribution S2 ′ of the silicon epitaxial layer to be formed is calculated based on the film thickness distribution data (sixth process).
Next, the film thickness variation distribution H ′ by changing the parameter from the predicted optimal set value group C to another set of set value groups is obtained from the difference between the virtual film thickness distribution S2 ′ and the virtual film thickness distribution S1 ′. By calculating and adding this film thickness change distribution H ′ to the film thickness distribution M ′, an expected film thickness distribution R ′ when another set of set value groups is applied is predicted (seventh process), A distance d4 between the expected film thickness distribution R ′ and the ideal film thickness distribution I is calculated (eighth process).
Then, another set of set value groups in the sixth process is changed, and the sixth to eighth processes are performed a plurality of times to obtain a set value group C ′ that minimizes the distance d4, and the distance d4 _{min at} this time And the distance d3 satisfy the predetermined criterion, the set value group C ′ is reset as the predicted optimum set value group C (ninth process). Here, as a predetermined standard that the distance d4 _min and the distance d3 should satisfy, for example, there are the following standards.

  According to the above thin film formation method, the predicted optimum set value group C that can form a silicon epitaxial layer having a small film thickness distribution on the main surface of the silicon single crystal substrate can be uniquely calculated. Therefore, unlike the conventional case, since the trial and error of changing the parameter value every time the silicon epitaxial layer is formed is not repeated, the consumption of the silicon single crystal substrate can be suppressed and the time required for the trial can be shortened. That is, a silicon epitaxial layer having a small film thickness distribution can be easily formed.

  Further, by resetting the predicted optimum set value group C, the film thickness distribution M ′ can be made as close as possible to the ideal film thickness distribution I. Therefore, unlike the conventional case shown in FIG. The search for the set value group does not end when the film thickness distribution of the silicon epitaxial layer is within the predetermined range. Accordingly, it is possible to form a silicon epitaxial layer having a better film thickness distribution as compared with the conventional case.

  In the above embodiment, the silicon epitaxial layer is formed on the main surface of the silicon single crystal substrate by the thin film forming method according to the present invention. However, the thin film is formed on the main surface of another semiconductor substrate. It's also good.

It is a figure for demonstrating the calculation method of expected film thickness distribution. (A) is a top view which shows schematic structure of a thin film formation apparatus, (b) is a longitudinal cross-sectional view of a vapor phase growth apparatus. It is a figure for demonstrating film thickness distribution data. (A) is a figure for demonstrating the search method of the conventional optimal setting value group, (b) is a figure for demonstrating the search method of the optimal setting value group in this invention.

Explanation of symbols

3 (3P, 3Q) Reactor 20 Susceptor X Flow direction of carrier gas Y Direction substantially orthogonal to gas flow direction (substantially orthogonal direction)

Claims (4)

A thin film forming method for forming a thin film on a main surface of a semiconductor substrate by vapor phase growth,
A thin film is formed on the main surface of separate semiconductor substrates in the first reactor by applying a plurality of set value groups obtained by changing and combining values of a plurality of parameters that change the film thickness distribution of the thin film to be formed. A data creation step of creating film thickness distribution data consisting of a plurality of film thickness distributions corresponding to the plurality of set value groups by measuring a film thickness distribution of each formed thin film,
A film thickness distribution measuring step of applying a set of set values in the second reactor to form a thin film on the main surface of the semiconductor substrate and measuring the film thickness distribution M of the thin film;
The predicted optimum set value group C that can minimize the distance d1 between the film thickness distribution of the thin film formed on the main surface of the semiconductor substrate and the ideal film thickness distribution I in the second reactor is the film thickness. A predicted optimum set value group calculating step of calculating from distribution data and an actual value of the film thickness distribution M;
Applying the predicted optimum set value group C, and forming a thin film on the main surface of the semiconductor substrate by vapor phase growth in the second reactor ,
The predicted optimum set value group calculation step includes:
A first process of calculating a virtual film thickness distribution S1 of a thin film to be formed on the main surface of the semiconductor substrate based on the film thickness distribution data when applying the set of set value groups;
A second process of calculating a virtual film thickness distribution S2 of a thin film to be formed when another set of set value groups is applied, based on the film thickness distribution data;
A film thickness variation distribution H obtained by changing the parameter from the one set of set value groups to the other set of set value groups is obtained from a difference between the virtual film thickness distribution S2 and the virtual film thickness distribution S1, and A third process for predicting an expected film thickness distribution R when the other set of set value groups is applied by adding a film thickness change distribution H to the film thickness distribution M;
The second process and the third process are performed a plurality of times while changing the other set of set values in the second process, and the distance d2 between the expected film thickness distribution R and the ideal film thickness distribution I is minimized. A set value group to be obtained is set as the predicted optimum set value group C, and
Following the third process,
Applying the predicted optimum set value group C, forming a thin film on the main surface of the semiconductor substrate by vapor phase growth in the second reactor, and measuring the film thickness distribution M ′ of the thin film;
A fifth process for calculating a distance d3 between the film thickness distribution M ′ and the ideal film thickness distribution I;
A virtual film thickness distribution S1 ′ of a thin film to be formed on the main surface of the semiconductor substrate when the predicted optimum set value group C is applied, and a thin film to be formed when another set of set value groups is applied. A virtual film thickness distribution S2 ′ of the first processing based on the film thickness distribution data;
A film thickness change distribution H ′ obtained by changing the parameter from the predicted optimum set value group C to the other set of set values is determined from a difference between the virtual film thickness distribution S2 ′ and the virtual film thickness distribution S1 ′. And a seventh process for predicting an expected film thickness distribution R ′ when the other set of set values is applied by adding the film thickness change distribution H ′ to the film thickness distribution M ′. ,
An eighth process of calculating a distance d4 between the expected film thickness distribution R ′ and the ideal film thickness distribution I;
The other set of set value groups in the sixth process is changed and the sixth process to the eighth process are performed a plurality of times to obtain a set value group C ′ that minimizes the distance d4. A thin film forming method comprising: a ninth process for resetting the set value group C ′ as the predicted optimum set value group C when the distance d4 _min and the distance d3 satisfy a predetermined criterion . The ideal film thickness distribution I are thin film forming method according to claim 1 Symbol mounting, characterized in that a constant value T. The ideal film thickness distribution I is a thin film forming method according to claim 1 or 2, characterized in that the distribution obtained by subtracting the thickness distribution of the semiconductor substrate from a constant value T '. The parameters include a flow rate of a carrier gas that forms an atmosphere in the reaction furnace, a flow rate of a purge gas that purges a space below the susceptor that supports the semiconductor substrate, and a direction substantially orthogonal to the flow direction of the carrier gas. thin film forming method according to claim 1 or 2, characterized in that the gas flow rate distribution of the reactor.

Images