JP2002117018A - Composite calculating method for thermal fluid analysis - Google Patents

Abstract

The present invention relates to a thermo-fluid analysis used for a simulation of a cooling structure of a heat-generating LSI on a substrate, and provides a complex analysis method in which required parts are calculated with high accuracy and high speed as a whole. . By using an analysis method suitable for high-speed calculation and an analysis method suitable for high-precision calculation, partial correction coefficient data is created by high-precision calculation, and the whole system is calculated at high speed using the data. Thermal design time by thermal fluid simulation can be shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for generating heat L on a substrate.
The present invention relates to a thermal fluid simulation of a cooling structure of an SI package, a heat treatment apparatus for manufacturing a semiconductor, and the like, and more particularly to a high-speed and high-accuracy calculation method of a combined heat transfer analysis method for performing analysis by combining different types of analysis methods.

[0002]

2. Description of the Related Art As a conventional composite heat transfer analysis method, for example, "Research on Machinery", Vol. 43, No. 1, pp. 8-14 (issued in January 1991), "Supercomputer in Mechanical Engineering" As described in “2.3 Complex heat transfer analysis” in “Advanced use of heat”, a single general-purpose program that can handle all phenomena such as heat conduction, heat transfer, and radiation has been developed and used for complex calculations. Had been In addition, as described in the 29th Japan Heat Transfer Symposium Proceedings, pp. 376-377 (issued in May 1992) as "Development of large-scale heat transfer analysis system using inter-university network" One heat transfer problem is calculated by distributing the calculation results to a large number of computers, and the calculation results are mutually transferred as discrete numerical data to obtain data at a required position by an interpolation method to perform a composite calculation.

[0003]

In the first prior art, all functions to be subjected to complex heat transfer analysis are included in one general-purpose program, so that the entire program becomes very large. Has a problem that the applicable range is reduced due to the computer capacity.
The second prior art has a problem that the calculation speed is limited by the slowest calculation when performing calculations in a distributed manner. Further, all of the distributed calculation data must have the same structure.
There was a problem that the analysis of the dimensional structure data could not be combined and calculated.

[0004] It is an object of the present invention to provide an analysis method in which calculation results can be obtained at high speed with high accuracy as a whole when analysis is performed by combining different types of analysis methods. is there. Another object of the present invention is to provide an analysis method that enables a composite calculation of analysis methods having different data structures.

[0005]

In order to achieve the above object, a plurality of analysis methods including an analysis method suitable for high-speed calculation and an analysis method suitable for high-precision detailed calculation are prepared in advance, and a problem is solved. Then, the time change of the temperature field and the velocity field is calculated by the high-speed calculation for the whole system. Next, a high-precision calculation is performed only on a portion to be calculated in detail, and correction coefficient data for high-speed calculation of the entire system is created from the result. By repeating the calculation of the whole system again using the correction coefficient data, the required part in the analysis result of the whole system is made highly accurate, and the whole system is calculated at high speed. .

[0006] From discrete numerical data of a solution obtained by one analysis method, a continuous solution expression is derived by an approximate calculation expression such as a polynomial approximation, and the discrete expression required for another analysis method is derived from the calculation expression. Heat transfer analysis method that transmits data between multiple analysis methods by obtaining typical numerical data.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to FIGS. FIG. 1 shows an example of a wiring board on which a number of heating LSI packages are mounted. Inside the housing box of electronic devices such as computers and telephone exchanges,
Various electronic components such as a finless LSI package 2, a finned LSI package 3, and a capacitor are mounted on the wiring board 1, and the wiring boards 1 are housed in a housing box in a stacked manner at a narrow interval. . In order to prevent malfunction due to temperature rise of the heat-generating LSI package during operation,
Cooling air (arrow 4) flows from the side of the wiring board 1 with a fan to cool the LSI packages 2 and 3 so that the temperature of the LSI packages 2 and 3 is kept below the allowable temperature.

In the thermal design of the cooling structure, numerical simulations of thermal fluid phenomena such as flow, heat conduction, convection, and radiative heat transfer are performed by complex analysis. FIG. 2 shows a calculation flow chart of the first embodiment of the present invention for the thermal fluid simulation method.

Step 1: A calculation condition is given. This provides structural data and operating conditions as shown in FIG.

Step 2: The temperature distribution, velocity distribution, and heat flux distribution of the entire system are calculated at high speed by a high-speed analysis method. In the case of an air-cooled substrate as shown in FIG. 1, the nodal point method is a typical high-speed analysis method, but a finite volume method with coarse mesh division or a finite element method of two-dimensional calculation may be used. FIG. 3 shows calculation mesh conditions in the node method. In the nodal method, the flow temperature is calculated using the nodal point 5, the element 6 connecting the nodal point 5, and the cooling fan 7. The nodal method assumes a parabolic distribution in the thickness direction of the air layer and does not calculate the distribution (2D structural data).
The total number of nodes is reduced from 20 to about 1000, and high-speed calculation is possible. The calculation results of the temperature distribution and the velocity vector distribution for the example of FIG. 1 are shown in FIGS. In this example, since the calorific value is fixed, the calculation result is only a steady solution, but if the calorific value changes with time, the temporal change in temperature or speed also becomes the calculation result. All calculation results are saved in a data file.

Step 3: The user designates an area to be calculated with high accuracy in part by looking at the result of step 2. For example, consider a case where a portion surrounded by a broken line 8 shown in FIG. 3 is calculated with high accuracy.

Step 4: A high accuracy analysis method is used to calculate only a given portion with high accuracy in detail. An example of the high-precision analysis is the finite volume method, and the number of detailed calculation nodes (three-dimensional structure data) is 10,000 or more. At this time,
As the boundary conditions around the given portion, the results of the high-speed analysis of the entire system in step 2 are used. In the data transmission that sends the data of the high-speed analysis result of the whole system to the high-precision analysis, first, a continuous calculation formula is obtained by approximation calculation such as polynomial approximation from discrete numerical data on temperature and speed with respect to the coordinates of the analysis result in step 2. Ask for.

In the case of the example shown in FIG. 1, a parabolic distribution formula is given in the thickness direction of the air layer from the analysis model. Note that it is not necessary to use one continuous calculation formula, and a large number of calculation formulas obtained by dividing the area may be used. However, it is important that the formula be a smoothly continuous formula with respect to coordinates and time. Next, high-precision analysis is performed by obtaining temperature and speed numerical data on calculation nodes required for high-precision analysis from the calculation formula. FIG. 6 shows the calculation results of the high precision analysis superimposed on the analysis results of the entire system. Naturally, the result of the high-precision calculation is different from the result of only the high-speed calculation in step 2, and correction coefficient data for converting the high-speed calculation result into the high-precision calculation result is created based on a comparison between the two.

Step 5: Using the correction coefficient data of step 4, the calculation of the whole system is redone by a high-speed analysis method, and is stored in a data file. As a method of using the correction coefficient data, assuming that the value of the correction coefficient data is multiplied by the temperature and speed of the calculation result of the high-speed analysis,
A solution in which the temperature and speed variables multiplied by the values of the correction coefficient data in the high-speed analysis satisfy the conservation laws of mass, momentum, and energy. When the calculation of the entire system is redone, the calculation result may be slightly corrected based on the calculation result of step 4. For example, the size of the cooling fin attached to the LSI package may be slightly modified.

Step 6: Steps 3 to 5 are repeated, and a result is output when a final solution is obtained.

As described above, performing high-speed calculation of the entire system first and performing high-precision calculation only on a necessary portion can drastically reduce the calculation time as compared with the conventional method. In addition, performing high-speed calculation of the entire system first and performing high-precision calculation only on a necessary portion is consistent with a normal design procedure, and thus has an advantage that it is easy to design using a thermal fluid simulation. In addition, since discrete numerical data of one analysis method is converted into a continuous calculation formula and numerical data of another analysis method is created, regardless of the size, geometric structure, data structure, etc. of the calculation mesh of the analysis method. In general, data can be transmitted, so that any analysis method can perform complex calculations.

A second embodiment of the present invention will be described with reference to FIGS. FIG. 7 shows a heat treatment apparatus for manufacturing a semiconductor.
The semiconductor wafer 11 is inserted between a large number of lamp heaters 9 and 10 on the ceiling surface and the floor surface to perform heat treatment. The periphery is a reflection surface 12, and there is a ring plate 13 around the wafer 11. To prevent contamination, the wafer 11 is placed on a support jig 15 in a quartz glass tube 14 and heat-treated. Gas flows in the quartz glass tube 14 as indicated by an arrow 16. In this case, the surface state changes (such as formation of a film) on the wafer surface according to the heat treatment time, and the thermophysical property value such as the emissivity changes. Appropriate process conditions for performing a predetermined heat treatment on the wafer under such conditions are calculated by thermal fluid simulation. FIG. 8 shows a calculation flowchart. Only parts different from the first embodiment will be described.

Step 2: The temperature distribution, the radiant heat flux distribution, etc. of the whole system are calculated by the thermo-fluid analysis program.
In this case, it is assumed that there is no change in the surface state of the wafer due to the heat treatment. FIG. 9 shows a calculation result of a temperature distribution after heating the wafer for 20 seconds.

Step 3: Normally, basic data on the correspondence between the heat treatment temperature and the heat treatment time of the wafer and the change in the surface state has been clarified. For example, FIG. 10 shows data on the relationship between the oxide film thickness on the wafer and the emissivity for infrared light having a wavelength of 3 μm. Using the temperature change data of step 2 and its basic data, the temperature change of the wafer when only the surface property value of the wafer changes at each time is calculated. Here, unlike the calculation in step 2, the entire system is not calculated, and only the temperature change of the wafer is calculated for the heat flux to the wafer using the solution in step 2. Using this calculation result, correction coefficient data for converting the calculation result of step 2 into the calculation result of step 3 is created.

Step 4: Using the created correction coefficient data, the temperature distribution and the like of the entire system are calculated by the same thermal fluid analysis program as in step 2.

The thermo-fluid analysis program having the function of calculating the time change of the material property value is a special calculation program.
Even if you use a normal calculation program without the calculation function,
Advanced calculations including temporal changes in material properties can be performed.

[0022]

According to the present invention, the design time by the thermal fluid simulation can be shortened, and a composite calculation can be performed by any analysis method. In addition, it is possible to calculate a function not included in the calculation program by using a normal calculation program. As a result, there is an effect that the design cost can be reduced.

[Brief description of the drawings]

FIG. 1 is a perspective view showing mounting of a heating LSI on a substrate.

FIG. 2 is a calculation flow chart according to the first embodiment of the present invention.

FIG. 3 is a diagram showing calculation node conditions for the high-speed calculation in FIG. 1;

FIG. 4 is a diagram showing a calculation result of a temperature distribution in the high-speed calculation in FIG. 1;

FIG. 5 is a diagram showing a calculation result of a speed vector distribution in the high-speed calculation of FIG. 1;

FIG. 6 is a diagram showing a calculation result of a velocity vector distribution in which the calculation result of the high-precision calculation in FIG. 1 is superimposed on the analysis result of the entire system.

FIG. 7 is a vertical sectional view showing a semiconductor heat treatment apparatus.

FIG. 8 is a calculation flowchart showing a second embodiment of the present invention.

FIG. 9 is a diagram showing a calculation result of the temperature distribution in FIG. 7;

FIG. 10 is a characteristic diagram showing data on a relationship between an oxide film thickness on a wafer and an emissivity.

[Explanation of symbols]

1. Wiring board, 2. Finless LSI package, 3.
Finned LSI, 4 ... cooling air, 8 ... area for accurate calculation, 9, 10 ... lamp, 11 ... wafer.

Claims (1)

[Claims] 1. A combined analysis method for numerically analyzing the temperature distribution of a solid, the temperature distribution of a fluid, and the flow velocity distribution by a computer by using a plurality of different analysis methods in combination. The analysis of the whole system is performed by the following, then the analysis of the partial region is performed by an analysis method suitable for high-precision calculation, and correction coefficient data for the analysis result of the whole system is created from the analysis result of the partial region, and the correction coefficient A combined calculation method of thermo-fluid analysis characterized by re-analyzing the whole system while using data.