JP2002064199A - Optimum condition deciding method for device simulation parameter and optimum condition assisting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an optimum condition deciding method, being reasonable and of high possibility, of simulation parameter for improved precision in a semiconductor device simulation, and to provide an optimum condition assisting device for deciding an optimum condition for the simulation parameter. SOLUTION: A large amount of simulation parameters modeling a physical phenomenon inside a semiconductor, related to a semiconductor devise simulation are taken as adjustment/control items (control factor and level), and statistical process is performed about the information data for the control factor. Thus an SN ratio and a level-specific SN ratio average value, as well as a sensitivity and a level-specific sensitivity average value are acquired, and a level at which the level-specific SN ratio average value is maximum is examined for each control factor, with the control factor and such control factor of combination of levels being an optimum condition of the simulation parameter.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a semiconductor device simulation technology, and more particularly to a method for determining optimum conditions of device simulation parameters for obtaining a highly accurate device simulation result, and an optimum condition support used for determining the optimum conditions. Related to the device.

[0002]

2. Description of the Related Art With the recent miniaturization of semiconductor devices, device simulation technology has become indispensable. Therefore, improvement in device simulation accuracy has been desired. The device simulator models various physical phenomena centering on the electrical properties inside the semiconductor and calculates based on the models, and variables constituting these models are called device simulation parameters. Then, in order to make the device simulation result as close as possible to the actually measured value, that is, to improve the device simulation accuracy, it is necessary to set a plurality of device simulation parameter values to appropriate values.

Conventionally, a plurality of device simulation parameters are independently adjusted and controlled each time a simulation of each semiconductor device is executed according to the intuition and experience of a device simulation user, and a combination of optimum states is created, thereby achieving a device simulation. An optimization method of realizing the optimum condition of the parameter has been adopted.

[0004]

However, setting the values of the device simulation parameters is a very difficult task because there are a plurality of device simulation parameters and a plurality of corresponding semiconductor device structures. Therefore, the conventional method based on the intuition and experience of the user is a trial-and-error method in which the value of the device simulation parameter is set every time the device simulation is executed, and thus requires a great deal of time and effort.

[0005] Further, since the model is complicated with miniaturization, it is not easy to set values of device simulation parameters.

Further, according to the conventional method, since the approach to optimizing the device simulation parameters is not objective, it is difficult to say that optimal conditions have been set with high accuracy.

SUMMARY OF THE INVENTION The present invention has been made in view of such problems, and a method for determining optimum conditions for device simulation parameters, which is rational and highly accurate in improving the accuracy of semiconductor device simulation, and for determining such optimum conditions. It is an object of the present invention to provide an optimum condition support device.

[0008]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention relates to a method for determining optimum conditions for device simulation parameters for modeling a physical phenomenon inside a semiconductor in a semiconductor device simulation. Parameters as control factors, for each of the control factors, device simulation parameter values corresponding to a predetermined number of levels;
Setting a model formula, applying a plurality of the device simulation parameters having mutually different level combinations for the control factors, creating an experiment plan for performing a device simulation on a plurality of semiconductor device structures, and A simulation result is obtained by performing a device simulation based on the plurality of semiconductor devices, and actual characteristics corresponding to the simulation results in the plurality of semiconductor device structures are associated with predetermined signal factors. And the S / N ratio for each control factor and the S / N ratio average value for each level based on the plurality of semiconductor device structure information,
The level at which the average SN ratio for each level is the maximum is examined for each of the control factors, and the condition of the device simulation parameter corresponding to the level is determined as the optimum condition for each of the control factors. .

[0009] In the semiconductor device simulation,
In the method for determining optimum conditions of device simulation parameters modeling a physical phenomenon inside a semiconductor, the device simulation parameters are used as control factors, and for each of the control factors, values of device simulation parameters corresponding to a predetermined number of levels, Setting a model formula, applying a plurality of the device simulation parameters having mutually different level combinations for the control factors, creating an experiment plan for performing a device simulation on a plurality of semiconductor device structures, and A simulation result is obtained by performing a device simulation based on the plurality of semiconductor device structures, and actual characteristics corresponding to the simulation results in the plurality of semiconductor device structures are associated with predetermined signal factors, and the respective signal characteristics are obtained. Based on a plurality of actual characteristic data as factors and a plurality of the semiconductor device structure information, an SN ratio and an S / N average value for each control factor and a sensitivity and an average sensitivity value for each level are obtained for each control factor. In addition to examining the level when the different SN ratio average value is the maximum for each control factor, examining the level when the level-specific sensitivity average value is zero for each control factor, the level corresponding to those levels It is characterized in that the condition of the device simulation parameter is determined as the optimum condition for each control factor.

[0010] Further, in the semiconductor device simulation,
In the method for determining optimum conditions of device simulation parameters modeling a physical phenomenon inside a semiconductor, the device simulation parameters are used as control factors, and for each of the control factors, values of device simulation parameters corresponding to a predetermined number of levels, Set a model formula, create an experimental plan for a preliminary experiment for device simulation execution applying a plurality of the device simulation parameters having different level combinations for the control factors, based on the experimental plan, A simulation result is obtained by performing a device simulation on a plurality of semiconductor device structures, and an actual characteristic corresponding to the simulation result on the plurality of semiconductor device structures corresponds to a predetermined signal factor. Only, based the actual characteristics of the data of a plurality of semiconductor device structures is the signal factor,
The SN ratio for each control factor and the average SN ratio for each level, and the sensitivity and the average sensitivity for each level are determined, and the level when the average SN ratio for each level is maximized is examined for each control factor,
By examining the level at which the average sensitivity value for each level is zero for each of the control factors, and determining the conditions of the device simulation parameters corresponding to those levels as the optimum conditions of the device simulation parameters for each control factor. A plurality of device simulation parameters having a higher influence than the device simulation parameters are narrowed down, and the optimum conditions of the device simulation parameters obtained by the preliminary experiment are set as new control factors, and a predetermined control factor is set for each of the new control factors. Setting device simulation parameter values and model formulas corresponding to the number of levels, and executing device simulation using the plurality of device simulation parameters having different combinations of levels for the new control factor And a simulation result is obtained by performing a device simulation on the plurality of semiconductor device structures, and an actual characteristic corresponding to the simulation result on the plurality of semiconductor device structures is determined by a predetermined signal factor. The SN ratio and the SN ratio average value for each control factor, and the sensitivity and the level for each control factor, based on the data of the plurality of actual characteristics as the signal factors and the plurality of pieces of semiconductor device structure information. A sensitivity average value is obtained, and a level when the level-specific SN ratio average value is maximized is checked for each control factor, and a level when the level-specific sensitivity average value is zero is checked for each control factor. And determining the conditions of the device simulation parameters corresponding to those levels as optimal conditions for each control factor. And the.

[0011] Further, in the semiconductor device simulation,
In an optimum condition support apparatus for determining an optimum condition of a device simulation parameter modeling a physical phenomenon in a semiconductor, information data of a control factor of the device simulation parameter and information of a device simulation parameter set for each control factor are provided. When information data for each level indicating a value is input, a first storage unit that stores the information data, and when information data of a level corresponding to a plurality of experiments and the control factors are input, these are stored. A second storage unit for storing information data,
When a simulation result obtained by the experiment and information data of actual characteristics measured corresponding to a signal factor are input, a third storage unit for storing the information data, and the first storage And performing a statistical operation based on the information data stored in the second storage unit and the third storage unit, thereby obtaining an S / N ratio and an average S / N ratio for each experiment, and a sensitivity and a sensitivity for each level. A statistical operation unit for calculating and outputting an average value.

[0012]

According to the method for determining the optimum conditions of the device simulation parameters, the information data of the control factors and the device simulation set for each control factor for a large number of adjustment / control items (control factors and levels) in the semiconductor device simulator. By performing statistical processing on information data for each level indicating the value of the parameter,
The SN ratio and the average SN ratio for each level, and the sensitivity and the average sensitivity for each level are obtained. Then, when the average SN ratio for each level is maximum or when the average sensitivity for each level is zero, the level is checked for each control factor, and the condition of the device simulation parameter corresponding to that level is determined for each control factor. Since the optimum manufacturing conditions are set, it is possible to quickly and easily determine the optimum conditions for the device simulation parameters with high accuracy from a large number of adjustment / control items. In addition, since the approach to optimization has objectivity, by executing a device simulation using the optimum conditions of the obtained device simulation parameters, the accuracy of semiconductor device simulation is improved, and the development of a further high-quality LSI is improved. And so on.

Further, the above-mentioned optimal condition support apparatus executes complicated statistical processing based on information data on a large amount of adjustment / control items (control factors and levels) in a semiconductor device simulator. This is effective in simplifying the processing for determining the conditions.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, the principle of the method for determining optimum conditions for device simulation parameters according to the present invention will be described. Note that this method will be described in order for each processing step.

(Step 1) First, from among device simulation parameters (hereinafter referred to as parameters) necessary for improving the device simulation accuracy, the electrical characteristics of the semiconductor device structure to be obtained are influenced. Adopt brazing parameters.

(Step 2) Next, the plurality of parameters determined in Step 1
Parameters (for the sake of explanation, let n parameters be α₁~ Α _nToss
Items to be adjusted and controlled for each of
Factors affecting the A (hereinafter referred to as control factors) A ~
N is determined, and a plurality of m adjustments are made for each of the control factors A to N.
・ By setting control conditions (hereinafter referred to as levels)
Fig. 1 shows the relationship between the selection of control factors and their levels as shown in Fig. 1.
Create a table. Note that the levels of all control factors are the same as m.
It is not necessary to set m = 2 or m = 3, for example.
You.

For example, if the parameter α ₁ is a parameter related to mobility in consideration of inter-carrier scattering, the content A1 corresponding to the level 1 of the control factor A is changed to a specific parameter related to mobility (for example, 10). , The contents A2 corresponding to the level 2 are replaced with the contents of specific mobility-related parameters (for example, 2
0), and similarly set up to the m-th level. Further, if the parameter alpha ₂ is a parameter of the mobility relationship in consideration of the vertical electric field, specific numerical values representing a specific speed, the contents B2 corresponding to level 2 content B1 corresponding to the level 1 of the control factor B The contents of the parameters related to the mobility are set in the same manner up to the m-th level. However, the contents N _{1 to} N _m corresponding to the respective levels of the control factor N are set so that a clear difference can be obtained in the accuracy of the simulation result. Further, a plurality of parameters (for example, α ₁ ,
α ₂ ) and the corresponding level contents may be set.
For example, if the parameter II is related to the setting of the level as well as the parameter I relating to the mobility in consideration of the inter-carrier scattering, both of them should be used as one control factor and the content of each level should be set specifically. It may be. And
For the remaining parameters α _{3 to} α _n , control factors and their corresponding levels are set in the same manner.

(Step 3) Next, based on the initial information set in steps 1 and 2, an experiment plan for obtaining the optimum conditions of the parameters as shown in FIG. 2 is set. For example, level 1 is assigned to the control factors A to N in the first experiment No. 1, and levels 1 and 2 are appropriately assigned to the control factors A to N in the second experiment No. 2. Assignment and the remaining experiments No. 3 to No. k are appropriately selected and assigned from Level 1 to Level m.
The level combination is set to be different between experiments. That is, as shown in the orthogonal table of FIG. 2, levels 1 to m are assigned in association with a plurality of k experiments No. 1 to No. k and control factors A to N.

In the orthogonal table, the level of each experiment is given so that each experiment is orthogonal. The orthogonal table for that is:
The size (number m of control factors and its level) is determined (for example, L4, L8, L9, L12, L18...), And an orthogonal table is selected according to the number and level of the selected control factors. That is, at the time when the orthogonal table is selected, the level of the control factor in each experiment is determined.

(Step 4) Next, semiconductor device structures E ₁ to E _q for performing a device simulation are prepared. Then, a device simulation of each semiconductor device structure is executed in accordance with the conditions (control factors and levels) of each experiment in the experiment plan. By performing k experiments of No. 1 to No. k, k types of simulation results can be obtained by simulation based on one semiconductor device structure. Further, p points to be used as data are prepared from a simulation result based on one semiconductor device structure. This is a point taken out as a sample at the time of evaluation in determining the optimum condition. Further, data actually measured based on each semiconductor device structure and each semiconductor device structure for each of E ₁ to E _q (hereinafter, referred to as actual measured values)
_Is prepared, and the measured values are set as signal factors M ₁₁ to M _qp . Here, E ₁ to E _q have different semiconductor device structures. However, for example, different gate voltages may be used in the case where the simulation results are IV characteristics.

Also, with respect to the actually measured values, the same sample points as in the simulation results are used for the respective semiconductor device structures E ₁ to E
_Take p out of each _q . Since the simulation results are compared with the measured values, the sample points must be the same.

Thus, for one experiment number, y _{11
Qy qp} , that is, p × q simulation results are obtained. Then, the signal factors M ₁₁ to
Simulation results corresponding to M _{q p} and the semiconductor device structure E ₁ ~ E _q is the associating y ₁₁ ~ y _qp. This is shown in FIG.

As described above, by setting conditions of a plurality of parameters to be controlled for a plurality of semiconductor device structures and performing a device simulation, it is possible to quantitatively know the influence of the parameters. Further, by setting actual measurement values as signal factors, a target of a simulation result based on each semiconductor device structure is set, and the present method has variation information of the simulation result with respect to the target value.

(Step 5) Next, the S / N ratio for each experiment is obtained by performing statistical processing on the simulation results y ₁₁ to y _qp for each experiment shown in FIG. Statistic processing on the measured values y ₁₁ to y _qp of Experiment No. 1 shown in FIG. 3 will be described as a representative. First, the following equation (1) shows the measured value y 11 of FIG.
By applying the ₁₁ ~ y _qp, obtaining the total variation S _T for experiments No.1.

[0025]

(Equation 1)

Further, a repetition number (also referred to as an effective divisor) r for Experiment No. 1 is obtained based on the following equation (2).

[0027]

(Equation 2)

Here, q is the number of the semiconductor device structures shown in FIG.

Further, the variation “regression variation” S _X of the proportional term of the signal for Experiment No. 1 is determined based on the following equation (3).

[0030]

(Equation 3)

Further, by substituting the value S _e obtained based on the following equations (4) and (5) and V _e into the following equation (6), the characteristic value for Experiment No. 1, ie, the SN ratio Find η ₁ ^* .

[0032]

(Equation 4)

[0033]

(Equation 5)

[0034]

(Equation 6)

Then, the arithmetic processing of the above equations (1) to (6) is performed for the remaining experiments No. 2 to No. k.
The SN ratios η ₂ ^{* to} η _k ^* corresponding to Experiment Nos. _{2 to} No. _k are determined. Therefore, as shown in FIG.
All the SN ratios η ₁ ^{* to} η _k ^* for each of No. _{1 to} No. _k are obtained.

(Step 6) Next,
SN ratio η₁ ^*~ Η_k ^*Water for each of the control factors A to N based on
Average value according to quasi (hereinafter referred to as average SN ratio according to level) μA1,
μA2,…, μAm, μB1, μB2,…, μBm,…, μN
Calculate 1, μN2, ..., μNm. For example, in FIG.
The experiment number using level 1 of the control factor A is N
o.1, No. 2, No. 3,..., No. 6, respectively
SN ratio is η₁ ^*, Η_Two ^*, Η_Three ^*, ..., η ₆ ^*Because
SN ratio average value μA1 for each level related to level 1 of control factor A
Is μA1 = (η₁ ^*+ Η_Two ^*+ Η_Three ^*+ ... + η₆ ^*) / ΖA1 and
Become. ΖA1 is the level 1 for the corresponding control factor A.
This is the number of experiments performed, for example, in the case of FIG.
The experiments performed using control factor A and level 1 were No. 1 to No.
Since it is 6, the value of ζA1 is six. Similarly, SN ratio by level
When the average value μA2 is obtained, level 2 is used for control factor A.
No. 7 to No. 12
You. Therefore, the average SN ratio μA2 for each level is
SN ratio η at the bar₇ ^*, Η₈ ^*, ..., η₁₂ ^*Using,
μA2 = (η₇ ^*+ Η₈ ^*+ ... + η₁₂ ^*) / ΖA2
A2 was performed using level 2 for the relevant control factor A
The number of experiments. For example, in the case of FIG.
Experiments performed using quasi 2 were No. 7 to No. 12, so
The value of 値 A2 is six. The same applies to control factor B.
The average SN ratio μB1 for each level is the control factor B in FIG.
Check the number of the experiment performed using level 1 in
Can be calculated using the SN ratio of the experiment number
You.

In this way, the average SN ratio for each level of each of the control factors A to N is obtained. .., ΜAm, μB
1, μB2, ..., μBm, ..., μB1, μB2, ..., μBm
Are associated with the control factors A to N in correspondence with the levels 1 to m, as shown in the table of FIG. ..., ΜAm, μB
1, μB2, ..., μBm, ..., μN1, μN2, ..., μNm
Indicates an objective basis for obtaining the optimum condition for each of the control factors A to N. For example, the average SN ratios μA1 and μA of the control factors A in the auxiliary table of FIG.
2, ..., in MyuAm, shows that if μA1 becomes maximum, the contents A1 level 1 for controlling factor A is the optimum condition of the control factor A, i.e. the parameter alpha ₁ . Also, the average value μB1 for each level corresponding to the control factor B,
If μB2 has the maximum value among μB2,..., μBm, it indicates that the content B2 of level 2 is the optimum condition for the control factor B. Similarly, for the remaining control factors C to N, the content of the level corresponding to the maximum average SN ratio for each level indicates the optimum condition of the parameter.

[0038] Thus, by referring to the level A1~Nm in FIG based on the optimum level of each control factor A to N, the optimum conditions of the parameters alpha ₁ to? _N and the control factor A to N Can be determined.

Furthermore, if there is no large difference in the average value of the S / N ratio for each level between the levels of the respective control factors, it indicates that the relevant control factor is not an important factor affecting the accuracy of device simulation. . Therefore, when an experiment or analysis for further optimization is performed, such an unimportant control factor is removed, and it becomes an objective judgment data for narrowing down the important control factor.

(Step 7) Steps 1 to 6 correspond to preliminary experiments for obtaining parameters which particularly affect the simulation result from a large number of parameters. Therefore, a process for determining the optimum condition of each parameter in more detail (hereinafter referred to as a narrowed-down experiment with respect to a preliminary experiment)
I do.

First, if there is no significant difference in the average value for each level among the respective levels of the respective control factors obtained in the process 6, they are not important factors affecting the simulation result, and are therefore excluded. The conditions (values) of the parameters relating to such control factors are known conditions such as the application of the current conditions, while only the control factors that greatly contribute to the simulation result are selected. Thereby, the narrowing down of the optimum condition is realized.

For example, if C to H of the control factors A to N are important factors, the relationship between the control factors C to H and the levels 1 to m is determined as shown in FIG. FIG. 5 corresponds to FIG. 1, and contents C1,..., Cm,..., H1,... Of new levels 1 to m corresponding to the respective control factors C to H.
, Hm are also set in the same manner as in FIG.

(Step 8) Next, a large number of adjustment / control items
From the (device simulation parameters), make an experiment plan using important factors that particularly affect the simulation results. When the important factors are already known and the number is so small that there is no need to narrow down, the experiment may be started from the following example. FIG.
In the same manner as in the experiment plan, a plurality of experiments No. 1 to No. n
By assigning the levels 1 to m in association with and the control factors C to H, an experiment plan as shown in the orthogonal table of FIG. 6 is created.

For example, as shown in FIG.
Level 1 is assigned to the control factors C to H in o.1, Levels 1 and 2 are appropriately assigned to the control factors C to H in the second experiment No. 2, and the remaining experiment Nos. .3 to No. n are set so that the combination of levels differs between experiments, such as by appropriately selecting and allocating from levels 1 to m. In the orthogonal table, the level of each experiment is given so that each experiment is orthogonal.
That is, at the time when the orthogonal table is selected, the level of the control factor in each experiment is determined.

(Step 9) Next, in accordance with the conditions (control factors and levels) of each experiment in the experiment plan of FIG.
By executing No. n, n device simulations are executed. Then, the signal factor, which is an actually measured value, is associated with the simulation result corresponding to the semiconductor device structure.

(Step 10) Next, by performing statistical processing on the simulation results using the above equations (1) to (6), the SN ratio η _{1 for} each experiment as shown in the right column of FIG. 6 is obtained.
Seeking ～Ita _n, further levels by SN ratio average value of the SN ratio for each level 1~m in each control factor C to H based on these SN ratio _{η 1 ~η n σC1, σC2,} ···, σCm, ..., σH
1, σH2, ..., σHm are determined.

[0047] Further, based on a simulation result for each respective experiment by performing arithmetic processing of the following equation (7), determined the sensitivity S ₁ to S _n of each experiment, further, these sensitivity S ₁
Level 1 at each control factor C H based on ~ S _n
ＭC1, δC2,..., ΔCm,..., ΔH1, δH2,
..., δHm is obtained.

[0048]

(Equation 7)

The average value of the S / N ratio and the average value of the sensitivity for each level are calculated as shown in FIG.
~ H and the level.

(Step 11) Next, the level of the average SN ratio for each level obtained in this way and the level of the average sensitivity per level closest to zero are examined for each control factor.
, The level content of each control factor is set as the optimal condition of the parameter. When deciding the optimum condition, it is easy to understand if the average SN ratio for each level and the average sensitivity for each level are graphed.

(Step 12) Next, the fact that the present invention is a method for determining optimum conditions for parameters with high accuracy will be described. That is, it is verified whether or not the levels of these control factors are the optimal conditions.

It is assumed that the level having the maximum value for each control factor among the average values of the SN ratios for each level obtained in step 11 is determined as the optimum condition of the parameter. It is assumed that the optimum conditions of the parameters can be determined to be level 1, level 2 for the control factor E, level 1 for the control factor F, level 3 for the control factor G, and level 3 for the control factor H. Note that such optimum conditions are defined as C3 D1 E2 F1 G3
It is represented as H3. Among these control factors, it is assumed that the variation between the levels of the average SN ratio for each level in each of the factors D, F, and G is larger than the other factors. in this case,
Although three half factors are selected out of the six factors C to H, it is preferable to select approximately half of all factors.

Based on the average SN ratio σD1 for each level of the control factor D at level 1, the average SN ratio σF1 for each level of F at level 1, and the average SN ratio σG3 for each level of G at level 3, When the average σAV is estimated, σAV = (σD1−T) + (σ
F1−T) + (σG3−T) + T. In addition, with respect to the condition of the currently applied parameter (current condition) with respect to the optimal condition, the process average σAV ′ is obtained in the same manner as in the case of the optimal condition.
Ask for. Here, T is a value obtained by adding the SN ratio average value for each level of one factor for all levels and dividing the sum by the number of levels. By taking the difference between σAV and σAV ′, it is possible to obtain a gain (improvement degree) that would be obtained when the optimum condition was applied instead of the current condition. The gain obtained by estimating the process average is Δ
ηg (= σAV-σAV ').

(Step 13) A device simulation is performed using the optimum conditions C3 D1 E2 F1 G3 H3 of the parameters determined in the step 11 (referred to as confirmation experiments).
Obtain simulation results. Further, a device simulation is performed using the current conditions of the parameters, and a simulation result is obtained. From the obtained simulation results, the above-described equations (1) to (6) are applied to determine the SN ratio ηb when using the optimum conditions and the SN ratio ηc when using the current conditions. Then, the gain Δη = ηb−ηc is calculated.

At this time, the closer the gain Δηg obtained by the estimation and the value of the gain Δη actually obtained from the simulation result using the optimum condition and the current condition are, the higher the accuracy of the optimum condition is. Become. Specifically, Δη
= Δηg ± 30%.

As described above, according to the parameter optimum condition determining method of the present invention described in accordance with the steps 1 to 11, the parameter of a highly accurate parameter can be selected from an extremely large number of adjusted / controlled items (control factors and levels). Optimum conditions can be easily obtained. Further, by examining whether or not the gain has been reproduced in accordance with steps 12 and 13, it is possible to obtain a more reliable parameter optimum condition. or,
Since the approach to optimization has objectivity, it is possible to develop a higher-quality LSI by utilizing a device simulator using the optimum conditions of the obtained parameters.

Next, a specific embodiment to which the principle of the present invention is applied will be described. The process will be described in the order of processing in accordance with the description of the principle of the invention, and "process x" in the description of the specific example will be referred to as "process x *" in the description of the principle.

(Step 1 *) Vd-Id of semiconductor device
The description will be made on the assumption that a device simulation for obtaining characteristics is performed. Simulation results (Vd
−Id characteristic),
Parameters in mobility model considering vertical electric field
I, parameter II in the mobility model considering the vertical electric field, parameter III in the mobility model considering the vertical electric field, parameter IV in the mobility model considering the vertical electric field, parameter V in the mobility model considering the vertical electric field, Seven parameters, i.e., parameter VI in the mobility model considering the vertical electric field and parameter VII in the mobility model considering the vertical electric field, were used as control factors, and the corresponding number of levels was 3.

(Step 2 *) Next, the control factor A is set as a parameter I in the mobility model taking the vertical electric field into consideration, and the second level of the control factor A is used as the parameter currently used value A2 (current value). Based on this, the value A3, which was slightly increased, was set as the third level, and the value A1, which was slightly reduced, was set as the first level.

Similarly, the control factor B is set as a parameter II in the mobility model considering the vertical electric field, and the second level of the control factor B is set to the value B of the parameter currently used.
2 (current value) and a slightly larger value B3 based on it
The value of the level B1, which was slightly reduced, was set as the first level.

The control factor C is a parameter III in the mobility model considering the vertical electric field, and the control factor D is a parameter IV in the mobility model considering the vertical electric field.
The control factor E is a parameter V in the mobility model considering the vertical electric field, the control factor F is a parameter VI in the mobility model considering the vertical electric field, and the control factor G is a parameter VII in the mobility model considering the vertical electric field.
The level was determined for each control factor as shown in FIG.

(Step 3 *) Next, an experimental plan for a preliminary experiment as shown in FIG. 2 was prepared. In this specific example, as shown in the orthogonal table of FIG. 9, 18 types of experiments were performed, and levels Nos. 1 to 3 corresponding to experiments No. 1 to No. 18 and control factors A to G were assigned. In this specific example, since the number of levels was set to 3 in all seven control factors,
Level number 2 indicating the experimental design in the leftmost column in the orthogonal array of FIG. 9
Are not used.

(Step 4 *) Next, a semiconductor device structure for performing a device simulation is prepared. The semiconductor device structure E ₁ is “a semiconductor device having a gate length I”. The semiconductor device structure E ₂ is “a semiconductor device having a gate length of II”.

As described above, the semiconductor device structures to be subjected to the device simulation are determined from E ₁ to E ₆ .

A device simulation of one semiconductor device structure was performed in 18 experiments N shown in FIG.
It is executed under the conditions of o.1 to No.18. Simulation results are Vd-Id characteristics, since a curve, such as approximately 10, the horizontal axis direction (drain voltage) 20 divided sampling points were each evaluated (y ₁₁ ~ y _qp).
Further, since the characteristics obtained by the different semiconductor device structures E ₁ to E _q are different, E ₁ to E ₅ shown in FIG.
_Eq . Here, _assuming that a difference between voltages applied to the gates of E ₁ to E _q (E ₁ is a gate voltage of 5 [V], E ₂ is a gate voltage of 4 [V],...), A relationship as shown in FIG. 11 is obtained. The data collection method and evaluation procedure are the same. Step 4
Explaining using the symbols used in the above, q = 6 semiconductor device structures, and p = 20 points taken out as samples. Therefore, the signal factor in the specific example is p × q = 6 ×
20 = 120, and the corresponding simulation result is also 120. The simulation results for each experiment number are as shown in FIG.

(Process 5 *) Next, the S / N ratios η ₁ ^{* to} η _k ^* for each experiment were obtained by performing statistical processing on the simulation results and actual measurement values for each experiment shown in FIG. As a representative example of the processing for the experimental result of Experiment No. 1, first, by applying the above-described Equation (1) and performing the arithmetic processing of the following Equation (8), the total variation S _{T1 of} Experiment No. _{1 is obtained.}
I asked.

[0067] ^{_{S T1 = 2.87E-6 2 +}} 6.80E-6 2 + ... + 1.69E-5 2 + 8.69E-6 2 + 2.83E-5 2 + ... + 7.65E-5 2 + ... + 2.07E -5 ² + 7.62E-5 ² +… + 3.79E-4 ² = 1.8282E-8 ・ ・ ・ (8)

Next, the number of repetitions r ₁ of Experiment No. ₁ was obtained by performing the calculation of the following expression (9) by applying the above expression (2).

R ₁ = 2.79E-6 ² + 6.73E-6 ² +... + 1.60E-5 ² + 8.77E-6 ² + 2.98E-5 ² +... + 8.01E-5 ² +... + 2.11E -5 ² + 7.96E-5 ² + ... + 4.04E-4 ² = 1.20971E-8 (9)

Next, S _X1 for Experiment No. 1 is obtained by performing the calculation of the following equation (10) by applying the above equation (3), and applying the above equation (4) by applying the above equation (4). )) To obtain S _e1 for Experiment No. 1, and further calculate
V _e1 was obtained by performing the calculation of the following equation (12) by applying (5).

S _X1 = (2.87E-6 × 2.79E-6 + 6.80E-6 × 6.73E-6 + ... + 1.69E-5 × 1.60E-5 + 8.69E-6 × 8.77E-6 + 2.83E-5 x 2.98E-5 + ... + 7.65E-5 x 8.01E-5 + 2.07E-5 x 2.11E-5 + 7.62E-5 x 7.96E-5 + ... + 3.79 E-4 × 4.04E-4) ² / 1.20971E-8 = 1.79538E-8 ・ ・ ・ (10)

[0072]

[0073]

Then, the SN ratio η ₁ ^* for Experiment No. ₁ was obtained by performing the calculation of the following expression (13) by applying the expression (6).

Η ₁ ^* = 10 × Log {(1.79538E-8 − 2.73798E-12) /1.20971E-8/2.73798E-12} = 117.96 (13)

Further, the same calculation is performed for the remaining experiments No. 2 to No. 18 to obtain the respective SN ratios η ₂ ^*
~ Η ₁₈ ^* was determined. Therefore, as shown in the right column of FIG. 13, the SN ratios η ₁ ^{* to} η ₁₈ ^* for each of Experiments No. 1 to No. ₁₈ were obtained.

(Step 6 *) Next, the SN ratio η₁ ^*~ Η₁₈ ^*Based on
The average SN ratio by level for each of the control factors A to G
Was. For each level corresponding to level 1 of control factor A shown in FIG.
As a representative of the calculation of the SN ratio average value μA1, the control factor
Experiments using level 1 in A were No. 1, No. 2,
 No.3, No.10, No.11, No.12
Therefore, as shown in the following equation (14),
Corresponding SN ratio η₁ ^*, Η_Two ^*, Η_Three ^*, Η _Ten ^*, Η₁₁ ^*, Η₁₂ ^* 
By calculating the average of the
Ask.

ΜA1 = (117.96 + 113.85 + 109.23 + 112.4 + 113.19 + 111.75) /6=113.0633 (14)

Then, the averages μA2 and μA3 for the levels 2 and 3 of the control factor A and the remaining control factors B
Average value μB1 by level for each level 1 to level 3
, ΜB3,..., ΜG1,..., ΜG3 are similarly calculated to obtain average values for each level of control factors A to G as shown in FIG. Was determined.

FIG. 15 shows the auxiliary table of FIG. 14 in which the control factors A to
The graph is a graph of the average S / N ratio for each of the levels 1 to 3 for each G.

As is clear from FIG. 15, since the SN ratio at the level 2 for the control factor A is the maximum, it is shown that the level 2 is the optimum condition for the control factor A, and the level 1 for the control factor B. It is known that the optimum conditions are level 2 for control factor C, level 1 for control factor D, level 1 for control factor E, level 1 for control factor F, and level 2 for control factor G. it can. That is, the level contents A2, B1, C2, D1, E of FIG.
1, F1 and G2 can be applied to the semiconductor device structure E
When performing a device simulation of ₁ ~ E _6, the optimal condition parameters.

(Step 7 *) Next, from the results obtained in Steps 1 to 6, a control factor which particularly has a large effect on the simulation result (Vd-Id characteristic) is selected, and the optimum conditions for the parameters are determined in more detail. I will do it.

First, it can be determined from FIG. 15 that a factor having a large difference between the levels of the average values for each level in each control factor is a factor having a large effect on the simulation result. Therefore, from FIG. 15, the control factor A (the parameter I in the mobility model considering the vertical electric field) and the control factor B
(Parameter I in mobility model considering vertical electric field
It was determined that four factors of I), control factor E (parameter V in the mobility model considering the vertical electric field), and control factor G (parameter VII in the mobility model considering the vertical electric field) were important.

As for the control factor A, the level was newly determined mainly around A2 because the SN ratio near A2 was large. As for the control factor B, a level near B1 was newly set because the SN ratio of B1 was the largest. As for the control factor E, a level near E1 was newly set because the SN ratio of E1 was large. Regarding the control factor G, a level near G2 was newly set because the SN ratio of G2 was the largest. FIG. 16 shows newly set control factors and their levels.

(Step 8 *) Next, an experiment plan shown in FIG. 17 was created. First, the semiconductor device structures E ₁ to E ₆ prepared in the preliminary experiment were prepared. By performing nine device simulations for one semiconductor device structure, device simulation using nine different parameter conditions is performed, and the control factors A, B, E, G, and Experiment N based on three levels
o.1 to No. 9 and levels 1 to 3 corresponding to control factors A, B, E, and G were assigned. Then, device simulations for six types of semiconductor device structures were executed according to the conditions (control factors and levels) for each experiment in the experiment plan of FIG.

(Step 9 *) Next, execution is performed based on the experimental plan.
The simulation result obtained is converted to a signal factor M₁~ M ₂₀And half
Conductor device structure E₁~ E₆In FIG.
It was shown to. In the simulation results,
The target sample point is the case of the preliminary experiment shown in FIG.
20 points in the same way as for
The same 20 points were evaluated.

(Step 10 *) Next, by performing statistical processing applying the above-mentioned equations (1) to (7), the SN ratio η and sensitivity S for each experiment as shown in FIG. With respect to the S / N ratio η and the sensitivity S, the average S / N ratio for each level and the average sensitivity for each level as shown in FIG. 20 were obtained. For example, assuming that the statistical processing based on the simulation result of the experiment No. 1 shown in FIG. 18 is described as a representative, first, a square of the following equation (15) to which the above equation (1) is applied is given. Find the sum _ST1 .

[0088] ^{_{S T1 = 2.81E-6 2 +}} 6.74E-6 2 + ... + 1.65E-5 2 + 8.48E-6 2 + 2.95E-5 2 + ... + 7.78E-5 2 + ... + 2.08E -5 ² + 7.79E-5 ² +… + 3.86E-4 ² = 1.8997E-8 ・ ・ ・ (15)

Next, the repetition number r ₁ is obtained by the calculation of the following equation (16) to which the above equation (2) is applied.

R ₁ = 2.79E-6 ² + 6.73E-6 ² +... + 1.60E-5 ² + 8.77E-6 ² + 2.98E-5 ² +... + 8.01E-5 ² +... + 2.11E -5 ² + 7.96E-5 ² + ... + 4.04E-4 ² = 1.20971E-8 (16)

Next, the following equation (17) applying the above equations (3) to (5)
S _X1 , S _e1 , and V _e1 are obtained by the calculations in (19).

S _X1 = (2.81E-6 × 2.79E-6 + 6.74E-6 × 6.73E-6 + ・ ・ ・ + 1.65E-5 × 1.60E-5 + 8.48E-6 × 8.77E-6 + 2.95E-5 x 2.98E-5 + ... + 7.78E-5 x 8.01E-5 + 2.08E-5 x 2.11E-5 + 7.79E-5 x 7.96E-5 + ... + 3.86 E-4 × 4.04E-4) ² / 1.20971E-8 = 1.82515E-8 ・ ・ ・ (17)

[0093]

[0094]

Further, the calculation of the following equation (20) applying the above equation (6) and the following equation (21) applying the above equation (7) yields S
The N ratio η ₁ and the sensitivity S ₁ were determined.

Η ₁ = 10 × Log {(1.82515E-8 − 6.2647E-12) /1.20971E-8/6.2647E-12} = 113.8157 (20)

S ₁ = 10 × Log {(0.07455E-8−6.2647E-12) /1.20971E-8} = 1.7845 (21)

The SN ratios η _{2 to} η ₉ and the sensitivities S _{2 to} S ₉ of the remaining experiments No. _{2 to} No. ₉ were similarly obtained by statistical processing, and finally shown in the right column of FIG. Thus, the SN ratio and sensitivity for each experiment were determined.

Next, based on the SN ratios η _{1 to} η ₉ , the average SN ratio σ for each of the control factors A, B, E, and G is determined, and the control is performed based on the sensitivities S _{1 to} S _9. Factors A, B,
The average sensitivity value δ for each level of E and G was determined.

[0100] Level 1 SN for control factor A
As a representative of the ratio average value σA1, in FIG. 17, the experiments corresponding to level 1 of the control factor A are No. 1 and No.
2, No. 3, so σA1 = (113.8157+
109.2988 + 116.6034) / 3 = (η ₁ + η ₂ + η ₃ ) / 3 =
113.239. On the other hand, if the sensitivity average δA1 of the level 1 for the control factor A is represented, δA1 = (S ₁ + S ₂
+ S ₃ ) /3=1.29517. Then, the residual SNR average value for each level and the sensitivity average value for each level were obtained in the same manner, and the result as shown in FIG. 20 was obtained.

FIG. 21 is a table showing the levels of S shown in the auxiliary table of FIG.
The N ratio averages are grouped for each of the control factors A, B, E, and G,
This is a graph for each level 1 to 3. Also, FIG.
2 is a graph obtained by classifying the average sensitivity values by level shown in the auxiliary table of FIG. 20 for each of the control factors A, B, E, and G, and for each of the levels 1 to 3.

(Step 11 *) Next, among the average values of the S / N ratios obtained by the levels thus obtained, the level having the maximum value for each control factor was determined as the optimum manufacturing condition. Therefore, as is clear from FIGS. 20 to 22, the optimum conditions are level 3 for the control factor A, level 3 for the control factor B, level 1 for the control factor E, and level 2 for the control factor G. I was able to decide. Note that such an optimum condition is represented as A3 B3 E1 G2.

(Step 12 *) Next, the fact that the present invention is a method for determining optimum manufacturing conditions with high accuracy will be described. That is, it is verified whether or not the levels of these control factors are the optimal conditions.

Among these control factors, two control factors A and B are selected as control factors having large fluctuations in the SN ratio.
Average SN ratio value σA3 for each level of level 3 of control factor A
When the process average σAV is estimated based on the average value σB3 of the control factor B and the level 3 of the control factor B, σAV = σA3 + σB3−T = 1
14.604 + 115.691-112.82345 = 117.472 (db). In addition, the process average σAV * when the levels of the current control factors B and G are used for the optimal condition is estimated. Since the second level is the current condition for all of the control factors B and G, σAV = σA2 + σB2-T = 110.627 + 110.307
-112.82345 = 108.111 (db). Therefore, the estimated gain Δηg is Δηg = 117.472−108.111 = 9.361.
(db). In addition, T = 112.82345 was used for all experiments (9
Is the average value of η in one experiment).

(Step 13 *) Next, using the optimum conditions of the control factors (parameters) determined in the step 11 *, a device simulation of the semiconductor device structures E ₁ to E ₆ is executed to obtain a simulation result. At the same time, a device simulation using the current conditions is also performed, and a simulation result is obtained. By applying the equations (1) to (6), the SN ratio ηb of the simulation result under optimal conditions
When the SN ratio ηc of the simulation result under the current condition was obtained, ηb = 118.054 (db), ηc = 108.903 (db)
And the gain Δη = 118.054−108.903 = 9.151 (db).

Here, when the estimated gain Δη * was compared with the gain Δη calculated by the confirmation experiment, it was confirmed that the values were very close. That is, it has been proved that the present invention is a method for determining an optimum condition of a parameter with high accuracy.

Next, an embodiment of an optimum condition support device for determining an optimum condition of a parameter will be described with reference to FIG.

This optimum condition support device is constituted by a microcomputer system, and an operator can input various data from an input device 2 such as a keyboard connected to a central control unit 1 of the system. The central control unit 1 is configured to perform arithmetic processing on the input data and output the arithmetic result to an output device 3 such as a display. The central control unit 1 includes first, second, third, and fourth storage units 4, 5, 6, and 7, and a statistical operation unit 8. First
The storage unit 4 stores parameters α1 to αn, control factors A to N, and level 1 as shown in FIG.
~ M and the information data of the level contents A1 ~ Nm,
These information data are subjected to filing processing corresponding to the orthogonal table in FIG. 2 and stored. The second storage unit 5 stores an experiment No. as shown in FIG.
1 to No. 1; When information data of a level corresponding to k and control factors A to N are input, the information data is subjected to filing processing corresponding to the orthogonal table of FIG. 2 and stored.
Third storage unit 6, the operator semiconductor device structure E ₁ as shown in FIG. 3 via the input device 2 ~ E _q and a signal factor M
₁₁ to M _qp and the measured values Y ₁₁ to Y for each experiment.
_When the information data of _pq is input, the information data is subjected to filing processing corresponding to the orthogonal table of FIG. 2 and stored.

The statistical operation section 8 includes first to third storage sections 4 to
(1) to (7) based on the information data stored in
2 and the sensitivity for each experiment as shown in the right column of FIG. 2 and the sensitivity for each experiment as shown in the right column of FIG. Then, the average value of the SN ratio for each level and the average value of the sensitivity for each level as shown in FIGS. 4 and 7 are obtained, and the data of the calculation results are transferred to the fourth storage unit 7. The fourth storage unit 7 performs a filing process corresponding to an auxiliary table as shown in FIGS. 4 and 7 and stores the data of the calculation result transferred from the statistical calculation unit 8.

Then, the operator inputs the first
When the output of the data stored in the fourth storage units 4 to 7 is commanded, the output device 3 outputs these data in the form of the orthogonal table or in the form of a graph as shown in FIGS. I do.

As described above, the optimum condition support apparatus performs a statistical process on a huge amount of information data and outputs a result indicating the optimum manufacturing conditions. Therefore, the optimum condition support apparatus is an extremely effective means for engineers involved in the field of semiconductor device simulation. And can greatly contribute to the improvement of semiconductor device simulation accuracy.

[0112]

The effects obtained by typical ones of the inventions disclosed by the present application will be described as follows.

According to the method for determining optimum conditions for device simulation parameters of the present invention, device simulation parameters in device simulation are a large amount of adjustment / control items (control factors and levels), and statistical processing of information data for these is performed. By performing, the SN ratio and the SN ratio average value by level, and the sensitivity and the sensitivity average value by level are obtained, and the SN ratio average value by level is maximum,
Since the level when the average sensitivity per level is close to zero is checked for each control factor and the condition of the parameter corresponding to that level is set as the optimum condition for each control factor, a large number of adjustment / control items , It is possible to quickly and easily determine an optimum condition with high accuracy. In addition, since the approach to optimization has objectivity, executing device simulation using the obtained optimal conditions of the device simulation parameters leads to improvement of device simulation accuracy, and development of higher quality LSI. And the like, and so forth, and so forth, which exhibit extremely excellent effects.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram for explaining the principle of a method for determining an optimum condition of a device simulation parameter according to the present invention.

FIG. 2 is an explanatory diagram for explaining the principle of a method for determining an optimum condition for device simulation parameters according to the present invention.

FIG. 3 is an explanatory diagram for explaining the principle of the method for determining an optimum condition of a device simulation parameter according to the present invention.

FIG. 4 is an explanatory diagram for explaining the principle of the method for determining an optimum condition for device simulation parameters according to the present invention.

FIG. 5 is an explanatory diagram for explaining the principle of the method for determining an optimum condition for device simulation parameters according to the present invention.

FIG. 6 is an explanatory diagram for explaining the principle of the method for determining an optimum condition for device simulation parameters according to the present invention.

FIG. 7 is an explanatory diagram for explaining the principle of the method for determining an optimum condition for device simulation parameters according to the present invention.

FIG. 8 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 9 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 10 is an explanatory diagram for explaining a specific example in which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 11 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 12 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition of a device simulation parameter according to the present invention is applied.

FIG. 13 is an explanatory diagram for describing a specific example to which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 14 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 15 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 16 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 17 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 18 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 19 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters of the present invention is applied.

FIG. 20 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 21 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 22 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 23 is an explanatory diagram for explaining a specific example to which the method for determining an optimum condition for device simulation parameters according to the present invention is applied.

FIG. 24 is a block diagram showing a configuration of an embodiment of an optimum condition support device.

[Description of Signs] 1 Central control unit 2 Input device 3 Output device 4 First storage unit 5 Second storage unit 6 Third storage unit 7 Fourth storage unit 8 Statistical calculation Department

Claims (4)

[Claims] 1. A method for determining an optimum condition of a device simulation parameter for modeling a physical phenomenon in a semiconductor in a semiconductor device simulation, wherein the device simulation parameter is a control factor, and each control factor has a predetermined number of levels. Setting a value of the device simulation parameter corresponding to a level, a model formula, applying a plurality of the device simulation parameters having different level combinations for the control factors, and executing a simulation on a plurality of semiconductor device structures. A simulation result is obtained by performing a device simulation based on the experiment plan, and actual characteristics corresponding to the simulation results in the plurality of semiconductor device structures are prepared. Correspondence to a predetermined signal factors, the data of a plurality of actual characteristics the a respective signal factor and,
Based on the plurality of semiconductor device structure information, the SN ratio for each control factor and the average SN ratio for each level are obtained, and the level when the average SN ratio for each level is maximized is examined for each control factor, Determining a condition of the device simulation parameter corresponding to the level as an optimum value of the device simulation parameter for each control factor. 2. A method for determining optimum conditions of device simulation parameters for modeling a physical phenomenon inside a semiconductor in a semiconductor device simulation, wherein the device simulation parameters are control factors, and a predetermined number of levels are set for each of the control factors. Setting a value of the device simulation parameter corresponding to a level, a model formula, applying a plurality of the device simulation parameters having mutually different level combinations for the control factors, and executing a simulation for a plurality of semiconductor device structures. An experiment plan is created, and a simulation result is obtained by performing a device simulation based on the experiment plan, and actual characteristics corresponding to the simulation results in the plurality of semiconductor device structures are obtained. In association with the predetermined signal factors, the data of a plurality of actual characteristics the a respective signal factor and,
Based on a plurality of the semiconductor device structure information, an SN ratio and an average SN ratio for each level, and a sensitivity and an average sensitivity value for each level are obtained for each control factor, and the level at which the average SN ratio for each level is maximum. Is checked for each of the control factors, and the level at which the average sensitivity per level is zero is checked for each of the control factors, and the conditions of the device simulation parameters corresponding to those levels are determined for the device for each of the control factors. A method for determining an optimum condition of a device simulation parameter, which is determined as an optimum value of a simulation parameter. 3. A method for determining an optimum condition of a device simulation parameter for modeling a physical phenomenon inside a semiconductor in a semiconductor device simulation, wherein the device simulation parameter is a control factor, and a predetermined number of levels are set for each of the control factors. An experiment for a preliminary experiment for executing a device simulation applying a plurality of the device simulation parameters having different level combinations with respect to the control factors, by setting a value of the device simulation parameter corresponding to a level and a model formula. A plan is created, and a simulation result is obtained by performing a device simulation for a plurality of semiconductor device structures based on the experiment plan, and a simulation result is obtained for the plurality of semiconductor device structures. Is associated with a predetermined signal factor, and based on data of the actual characteristics of the plurality of semiconductor device structures as the signal factors, the SN ratio for each control factor and the SN for each level are determined. A ratio average value, and a sensitivity and a sensitivity average value for each level are obtained, and the level when the SN ratio average value for each level is maximized is checked for each of the control factors. By examining the level for each of the control factors and determining the conditions of the device simulation parameters corresponding to those levels as the optimal conditions of the device simulation parameters for each control factor, a large number of combinations of the device simulation parameters can be obtained. The device simulation parameters having a higher influence were narrowed down, and the device Optimum conditions of the simulation parameters are set as new control factors. For each of the new control factors, device simulation parameter values and model expressions corresponding to a predetermined number of levels are set. A simulation result is obtained by creating an experiment plan for executing a device simulation applying the plurality of device simulation parameters having the combination of levels, and performing a device simulation on the plurality of semiconductor device structures; The actual characteristics corresponding to the simulation result in the device structure are associated with a predetermined signal factor, and data of a plurality of actual characteristics as the respective signal factors,
Based on a plurality of the semiconductor device structure information, an SN ratio and an average SN ratio for each level, and a sensitivity and an average sensitivity value for each level are obtained for each control factor, and the level at which the average SN ratio for each level is maximum. Is checked for each of the control factors, and the level at which the average sensitivity per level is zero is checked for each of the control factors, and the conditions of the device simulation parameters corresponding to those levels are determined for the device for each of the control factors. A method for determining an optimum condition of a device simulation parameter, which is determined as an optimum value of a simulation parameter. 4. An optimum condition support apparatus for determining an optimum condition of a device simulation parameter modeling a physical phenomenon in a semiconductor in a semiconductor device simulation, comprising: information data of a control factor of the device simulation parameter; When information data for each level indicating the value of the device simulation parameter set in the control factor is input, a first storage unit for storing the information data includes a plurality of experiments and a level corresponding to the control factor. When the information data is input, a second storage for storing the information data is performed.
When a simulation result obtained by the experiment and information data of actual characteristics measured corresponding to a signal factor are input, a third storage unit that stores these information data, By performing a statistical operation based on the information data stored in the first storage unit, the second storage unit, and the third storage unit, the S / N ratio for each experiment and the S / N ratio average value for each level, and A statistical operation unit for calculating and outputting a sensitivity and a sensitivity average value for each level;