JPH03171301A - Material synthetic control system - Google Patents

Abstract

PURPOSE:To control the material synthesization at a microscopic level by evaluating microscopically the synthesization performed at the atomic/molecular levels and estimating the physical properties in real time based on the result of evaluation. CONSTITUTION:A means 11 is provided to synthesize a material showing the optional physical properties together with a means 17 which observes and evaluates in real time the property of a material noted at present, and a simulation means 18 which checks logically the physical properties of a relevant system based on a proper logical model. Then a material is synthesized at a speed lower than the simulation and at the same time the simulation is carried out with the information obtained by the means 17 used as the starting value. The result of the synthesization if carried on under the conditions is estimated together with a way of correction presumed for the result of estimation. Thus the synthetic conditions are controlled in a feedback system. As a result, the synthesizing process is optimized so as to produce the satisfactory materials with highest efficiency and to develop a new material.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention relates to a substance synthesis control system, and is particularly applicable to fields that require materials with controlled structures on the atomic and molecular scale. For example, advanced semiconductor device manufacturing methods, synthesis of catalyst materials with special structures, functional materials, energy materials, optical materials, and superconducting materials that utilize the properties of materials due to their microscopic structures. , related to the development of magnetic materials, etc., and related to control systems for substance design, material design, molecular design, drug design, alloy design, etc.

(Conventional technology) Synthesizing substances according to their field of use is one of the fundamental objectives of engineering, and chemistry and physics have made great contributions as methods for this purpose.

What is created as a result of this basically supports our lives.

Until now, synthesis methods have mainly utilized the macroscopic properties of substances. Of course, the mechanism was based on microscopic objects, but for example, reactions were caused in a reactor to obtain the average product. For this reason, the focus of process control has been how uniform the average value can be obtained and how much dispersion can be suppressed. Of course, it is an important issue to bring the average value to the desired point, but this is rather
There was a tendency to let nature take its course. However, in recent years, methods for synthesizing substances have made tremendous progress, and it has recently become possible to synthesize substances whose structures are controlled on the m-scale, such as the atomic and molecular scales. To give a concrete example of such a synthesis method, for example, molecular beam epitaxy (MB
E) Crystal growth method may be mentioned. MBE uses a molecular beam that is as uniform as possible to line up atoms and molecules one by one on a target, and is, in a sense, one of the ultimate materials synthesis methods.

Unlike traditional chemical synthesis methods, in this type of synthesis technology, it is important to control the synthesis conditions. In the traditional method, the control parameters for the successive conditions are relatively rough;
Synthetic methods used to work around Avogadro's number, but these new synthetic methods require control at the atomic level, and are also synthetic methods that allow this. However, although control is possible, the optimal conditions are often unknown. Nowadays, there is a strong tendency to search for the optimal solution empirically by varying various conditions. The optimal conditions obtained in this way become a database and serve as guidelines for future synthesis efforts.

In the above-mentioned MBE method, controlling factors include the intensity of the molecular beam, the temperature of the substrate, and the type and concentration of residual gas. With these, it is possible to control a certain elongation speed of the crystal/thin film.

On the other hand, significant progress has been made in recent years in substance evaluation technology. In particular, temporal resolution and spatial resolution have improved dramatically compared to conventional methods. As a means of investigating the structure of substances (how atoms and molecules are arranged), conventional transmission electron microscopy (TEM), Yasashi IE type [-S
iJII ffl (SEM): In addition, evaluation methods that can further distinguish between individual atoms and molecules have been devised and realized. For example, scanning tunneling microscopy (STM)
If we use low-speed electron diffraction (LEED), we can capture individual atoms with sufficient spatial resolution, and if we use low-speed electron diffraction (LEED), we can see how the atoms on the surface of a material are arranged.
You can check it directly. Moreover, the time required for evaluation is becoming instantaneous, and as a result, in-situ observation and evaluation are becoming possible. Of course, the evaluation method does not only need to be one that determines the atomic structure on a microscopic scale; the electronic state is also an important factor to consider. In particular, the electronic state of the solid surface governs the reactivity of the solid, and is also important for semiconductors. In the case of insulators, dielectric properties and optical properties are specified.

To measure and evaluate the electronic state on the surface of a solid, various methods such as Auger electron spectroscopy (AES) are used to excite electrons near the solid surface and use spectroscopic means such as electrons and light emitted due to this. You can check it out. Depending on the measurement method, it is becoming possible to investigate the electronic state only in the very vicinity of the surface, or to investigate the electronic state even deeper into the interior of a solid. This means that the electronic state can be investigated along the direction perpendicular to the surface. Furthermore, using STM, it is also possible to know the local electronic state at a point from its current/voltage characteristics (such a method is called STS). This allows us to capture changes in the electronic bear in the direction parallel to the surface. Furthermore, it is also possible to investigate the electronic state by focusing only on the very vicinity of the surface and avoiding affecting the surface state as much as possible, or by gently bringing the metastable molecule close to the solid surface and allowing it to interact with the surface state. This can be done by spectroscopy of the electrons emitted from this (metastable molecular separation spectroscopy: MDS). Actual measurements of bulk electronic states will continue to be important, but for example, the most basic measurement of the Fermi surface is performed using magnetoresistive methods, which have been used for a long time. In addition to measurements such as Shuvinnikov-Do Haas effect, Tohas-van Alphwen effect, cyclotron resonance, etc., recently, the positron annihilation method and computer tomography (CT), which is popular in medical equipment, have been combined to Methods to directly obtain three-dimensional images of surfaces have also been proposed and implemented, and are beginning to provide new knowledge.

Along with electronic states, magnetic properties are also important. Thermal neutron scattering experiments are probably the most serious way to investigate magnetic properties, whether in the bulk or on the surface. This makes it possible to observe spin waves and identify magnetic order.

Now, with improvements in semiconductor devices and the emergence of revolutionary architectures, the processing power of electronic computers is increasing at an almost unprecedented rate.

This trend has accelerated especially since the advent of supercomputers, and now computers boasting IOGFLOPS processing power are not uncommon. This means that large-scale numerical calculations that were previously infeasible can be performed.
Detailed simulations have actually been carried out using this method. A simulation is a numerical experiment conducted based on an assumed theoretical model, and is also a mock experiment. Therefore, it is decisive what kind of model to adopt.
No results can be obtained other than the results that are supposed to be removed from the assumed model. However, progress has been made in the most fundamental sciences for understanding nature, such as physics, and the basic equations are known beyond doubt. For this reason, there is a need to solve this problem, even if it is by force. Specifically, the various properties exhibited by matter are basically described, understood, and predicted at the same time according to the principles of quantum mechanics. Therefore, the basic equation, Schrödinger's equation, can be solved using the required system (boundary conditions). However, since this equation is generally a nonlinear partial differential equation, it is difficult to solve it numerically. However, we approximated numerically non-essential parts while taking care not to affect the results as much as possible, and with some limitations,
Thanks to the invention of a linear calculation method and the idea of a calculation method from a completely different perspective, it has recently begun to be solved experientially using high-speed electronic computers. Once the electronic state is known, the total energy of the system can be calculated. By calculating the total energy for various atomic configurations, it is possible to discuss and predict structural stability. Furthermore, since electronic states basically define the properties of materials, many of the transport phenomena in metals can be understood and predicted from their band structures or Fermi surface structures. Furthermore, the optical properties exhibited by semiconductors and insulators are basically caused by interband transitions. By discussing this issue, there is a growing momentum to design optical materials.

With the development of condensed matter theory and the development of computers, it has become possible to quantitatively and non-empirically investigate real systems, rather than extremely idealized states. The third field following chemistry, theoretical physics and theoretical chemistry,
Computational physics and computational chemistry have emerged as new academic fields, and direct engineering applications of materials design, such as material design, alloy design, drug design, molecular design, and semiconductor design, are beginning to emerge as a reality. Although ``material design'' is merely a direct recasting of the goals of ``engineering,'' it is methodologically completely new, and the possibility of realization is very high. Currently, attempts at ``material design'' are being made in many fields.

The various elements mentioned above, such as material synthesis, material evaluation, and understanding/prediction of physical properties, are currently being developed only in their respective forms. For this. At present, a more efficient system for synthesizing, evaluating, and predicting substances that combines the strengths of each has not been achieved.

Let us take the design and development of semiconductors as an example of how these technologies are currently used. Of course, technologies such as material synthesis, evaluation, and prediction are not used here, but the methodologies are very similar, so it can be used as an example of an existing technology.

Now, the following tasks are usually performed in the design and development of semiconductor devices. First, the necessary basic technology is examined based on the design specifications. For example, if 0.5 μm lithography is required, there will be an optical system for cutting it, development of resist materials, etc., but here we will assume that these elemental technologies are already in place. Therefore simulators (this includes process simulators, device simulators,
There are various circuit simulators, batan simulators, etc.
It is also assumed that it is fully equipped with a resist simulator, etc.). Based on the design specifications, for example, by making full use of a process simulator, an impurity profile is determined so as to exhibit desired characteristics, and ion implantation conditions (dose amount, accelerating voltage, order of ion implantation, etc.) to provide this are determined. Ion implantation and the like are actually performed according to these process parameters. Of course, this is the characteristic exactly as simulated.
To check whether it shows the SIM
Analyze and evaluate the design later using S and XPS, and reflect the results in the next design development process, that is, simulation.

However, in the case of ion implantation, while performing ion implantation, we do not check whether the parameters are correct or not, and we do not check whether the impurity profile is really as expected, and we almost miss the target. However, there is no way to correct it (there is no way to know that you are off target in the first place). In such a method, the first big problem is preparing a simulation that gives correct results. Furthermore, even if the correct result is given in the assumed case, it is difficult to guarantee the correctness under all conditions. Furthermore, since the conditions of synthesis are not known, it is difficult to efficiently produce high-quality products.

(Problem to be solved by the invention) As mentioned above, with the progress of technology in recent years, atoms and
Based on synthetic technology that can synthesize substances while controlling the microscopic scale such as the molecular scale, in-situ observations that can evaluate the atomic and molecular scale in real time, and theoretical models, Simulations using high-speed electronic computers are now possible.

However, currently each technology is relatively isolated,
Although results and evaluations are being integrated, it has not yet reached the point where feedback is provided through physical property predictions. This makes it difficult not only to search for new materials, but also to efficiently synthesize high-quality materials.

The present invention was created in view of the above points, and aims to systemize the latest technologies by organically combining them, optimize the synthesis process, and produce high-quality materials most efficiently and create new materials. The purpose is to provide a substance synthesis control system that enables the development of materials.

[Structure of the invention] (Means for solving the problem) In order to achieve the above object, the present invention provides a means for combining substances exhibiting arbitrary physical properties, and a method for measuring the properties of the substance of interest in real time. a means for observing and evaluating the irrsitu within the
It is equipped with a simulation tool that can theoretically investigate the physical properties of the system based on an appropriate theoretical model, and is capable of synthesizing materials at a slower speed compared to the simulation execution method, while obtaining the results obtained from the evaluation method. Using the obtained information as a starting value, perform a simulation on the same I, r, and find out what will happen if the synthesis continues under the same conditions.
A substance synthesis control system is constructed in which it estimates how modifications should be made from here to obtain the desired compound and controls the synthesis conditions in a feedback manner accordingly. .

In the present invention, the means for synthesizing substances in the above-mentioned substance synthesis control system is one that can control the structure on a microscopic scale such as an atomic or molecular scale, and is not concrete at the current state of the art. Examples include molecular beam epitaxial growth (MBE), pyrolytic vapor deposition (MOCVD) using organometallic compounds, vacuum evaporation, sputtering, ion cluster beam (ICB), Langmuir-Blodgett ( It is characterized by having a synthesis means using L B), etc.

In the present invention, the method for evaluating substances in the above-mentioned substance synthesis control system is one that allows for in-situ (11 sltu) observation of desired physical properties during actual patent application, and is a method that can be used at the current state of the art. , Low-energy electron diffraction (LEED), reflection high-speed electron diffraction (RHEED), extended X-ray absorption fine structure (EXAFS), and surface wide-area X-ray absorption fine structure are methods that can determine the atomic structure at the microscopic level. Structure (S)
EXAFS), transmission electron microscope (TEM), scanning electron microscope, i (SEM), scanning tunneling microscope m (STM)
), atomic force microscope (AFM), field ion microscope (F
Examples of methods for measuring electronic states include photoelectron spectroscopy (PES, , metastable molecular separation spectroscopy (MDS), etc.; for vibrational spectra, there are methods such as infrared absorption (IR) and Raman spectroscopy;1 methods for understanding magnetic properties include electron spin resonance (ESR) and nuclear magnetic resonance. (NMR),
It is characterized by having specific examples such as scattering experiments using thermal neutrons, and secondary ion mass spectrometry (SIMS) and ion neutralization method (INS) for further elemental analysis.

In the present invention, as a substance synthesis method in the above-mentioned substance synthesis control system, a plurality of apparatuses based on the same principle are used, and a very small number of synthesis systems among these are injected using fraudulent equipment.
It is characterized in that it evaluates Sltu, performs simulation based on this information, and reflects this result on all synthesis devices.

Furthermore, in the present invention, when the substance synthesis method in the substance synthesis control system described above has one or more apparatuses based on the same principle, and any one of them is to be evaluated, the evaluation means The system is characterized in that it is evaluated using evaluation methods based on different principles, and the system is theoretically simulated based on the complementary information obtained from these evaluation methods, and the results are reflected in all synthesis equipment. do.
Finally, in the present invention, in the substance synthesis control system described above, a plurality of independent high-speed computers are provided, and a group of process parameters are prepared based on the information obtained from the evaluation device. Simultaneously run the computers in parallel by setting different process parameters, compare the results, and select the optimal process parameters based on predetermined procedures or criteria. is characterized in that it is reflected on the synthesis device.

(Function) The method of synthesizing a substance exhibiting arbitrary physical properties as used in the present invention means a synthesis method controlled on a microscopic scale, that is, a method that can synthesize a substance with the required properties if sufficient control is performed. do. In the present invention, we control this material synthesis by predicting physical properties through simulations based on a theoretical model.
Based on this, process parameters are controlled. However, in the present invention, the simulation is not performed separately and used as a database as in the past. In the present invention, the simulation is not performed independently of the synthesis situation at that time, but the system being synthesized is minutely simulated using various techniques.
aPI.J [value] and use it as a starting value for the simulation, or use it as a control variable of the simulation port, such as the temperature and pressure of the system, to perform the simulation. This allows the synthesis process to be tracked at any time and a feedback filter is applied. Thereby, it is possible to stably synthesize a desired substance. Of course, if the synthesis were to be carried out completely ideally, such a special kll
! API/evaluation should not be necessary. As long as the starting values for the process parameters are determined strictly, it's fine, but as the synthesis progresses, the results will always deviate from the initial expectations. Synthetic over-manipulation is not done ideally or imaginatively,
This is because there are factors that oscillate the necessary system. However, J.
If the current state of synthesis using the F value device is accurately captured, a prediction system can be activated using this as a new starting point.

The prediction system can not only know what will happen if synthesis continues with the current process parameters, but also predict what parameters should be changed from now on to achieve the original plan. In this way, it is possible to absorb the amount that deviates from the initial estimate and obtain the target material as a result.

(Length Examples) Below, length examples of the present invention will be described with reference to the drawings.

EXAMPLE I We will describe as an example the synthesis of semiconductors or metals with a superlattice structure using the MBE method, evaluation of this using the LEED method or RHEED method, and the control of MBE at 1,000 aPIU through computer experiments. System configuration first
As shown in the figure. 11 is an MBE device, and the MBE device 11 is attached with an electron gun 12 for RHEED and an H#J device 13 for RHEED. The observation device 13 has a screen 14 and a camera 15 that captures an image of the screen.

The MBE device 11 also includes a quadrupole mass spectrometer 16.
is installed. The structural analysis computer 17 outputs the R}fEED control signal and also outputs the RHEED view A.
Structural analysis is performed by importing data obtained from the p+ device 13 and mass spectrometer 16. The structural data obtained by the horizontal angle q analysis computer 17 is sent to a child DPJ computer, ie, a simulator 18. The simulator 18 performs predictions of the properties of the wall that will be obtained if the crystal growth continues, based on the external human power data and the configuration data. The simulation results are sent to the control computer 19.

The control computer 19 has a pre-given setting of 1 type.
, 1 and the data sent from the emulator 18,
Control signals for controlling various growth conditions are sent to the MBE apparatus 11.

The MBE apparatus 11 is for growing a crystal by applying a molecular beam beam from a molecular beam source to a target cutting plate in a high vacuum. In this method, the entire system is first evacuated to reduce residual gas as much as possible. At the same time, in order to remove impurity gas coming from the inner wall of the vacuum container,
It can only be made by baking at ℃). A molecular beam source is obtained by vaporizing component elements by heating a crucible with a small hole in the top called a Knudsen cell.

To create a superlattice, molecular beam sources of several component elements must be prepared. When having multiple Knudsen cells like this, a shutter is provided between the opening and the substrate.
It is necessary to control this so that any molecular beam hits the substrate. The temperature of the Knudsen cell is O. l'
The temperature can be controlled within a range of approximately C, and the intensity of the molecular beam is adjusted by controlling the temperature. The intensity of the molecular beam is measured by a quadrupole mass spectrometer 16 placed near the substrate.

Fourfold! ! ! i! Since the mass spectrometer 16 can detect molecules with extremely high sensitivity, it can measure even a weak molecular beam intensity that grows one atomic layer per second. The molecular beam must travel straight toward the substrate. For this reason, it is necessary to avoid unnecessary thermal noise. Therefore, the Knudsen cell is usually surrounded by a metal wall cooled with liquid nitrogen.

Similarly, the substrate itself also prevents heating due to molecular beam collisions.
It is designed to maintain a constant temperature.

With such a device, the growth rate can be slowed down to one atomic layer per second, so there is plenty of time for control. This allows superlattice growth that is controlled down to the number of atomic layers. Furthermore, since the entire process is carried out at low temperatures, the structure is less prone to blurring due to thermal diffusion compared to conventional methods.

For this purpose, a superlattice structure with ideal interfaces can be created.

The LEED method uses the wave nature of electrons to cause diffraction through the atomic arrangement of the crystal surface, and obtains information such as the mutual relationship between the intensities of the incident light and the diffracted light, and the angular relationship between the incident light and the diffracted light. By collecting information about various reciprocal lattice rods, we are trying to learn their atomic arrangements. Since the energy of the incident electron beam used in normal LEED is about 1 to 300 eV, the de Broglie wavelength is about 0.7 to 12 A, which is about the same as the interatomic distance. Moreover, since the incident energy is small, the scattering cross section of the atoms is large, and because of this, only 2 or 3 atoms reach from the surface, so the diffraction pattern reflects only the atomic arrangement near the surface.

Theoretical analysis of the LEED method is performed using dynamic diffraction theory that takes into account the effects of multiple scattering. However, primarily, a strong spot can be obtained for a diffraction image with a reciprocal lattice rod that satisfies the Bragg reflection condition, that is, the change in wave number vector between the incident electron beam and the reflected electron beam is equal to the reciprocal lattice vector. .

The influence of multiple scattering is applied to the cusp Mj as a deviation from this. Now, it is quite a difficult problem to uniquely determine the atomic arrangement in the length space from the bang in the reciprocal lattice space. Although the estimated atomic configuration can be limited to a certain range, in general, several atomic configurations in real space are assumed, these model structures are Fourier transformed, and the electron diffraction images are directly generated. The final structure is determined by calculating and comparing this with the actually measured batan.

In the RHEED method, the spatial resolution is limited because the LEED method uses a low-energy electron beam, whereas the resolution can be increased by using a high-speed electron beam. However, when this is directly applied to the surface of a solid, electrons will penetrate into the interior of the solid, and scattering will occur not only on the surface but also inside the solid. In order to avoid this, in the RHEED method, an electron beam is incident on the solid surface, and only the reflected waves are observed, and only information from the surface is picked up. The information obtained is the diffraction pattern of the electron beam, so it is basically the same as LEED, but since the electron beam is shot into the surface, a clean RHEED pattern can be obtained when the surface is flat, but the surface is dirty. and(
Alternatively, if the monoatomic layer does not grow neatly and there are irregularities), streaks occur in the spots of the RHEED pattern. This allows M B E b'i. It is possible to estimate the flatness of a medium solid surface by 11 times. In other words, when looking at the RHEED pattern without MBE growth, at some point a beautiful spot image is obtained, then as the MBE growth progresses, a streak appears at that spot, and after a while, the streak disappears again and a beautiful spot appears. You can get it. When this beautiful spot is obtained, a flat crystal surface is obtained. Of course, it is also possible to find the unit vector of a two-dimensional lattice from the analysis of the diffraction pattern, and in RHEED, reflections include information not only about surface atoms but also about two to three atomic layers below the surface, so we can analyze this. This allows us to know the positions of atoms near the surface.

On the other hand, if the interactions between atoms are determined by some method, MBE simulation can be performed according to the classical molecular dynamics method. Of course, once the processing power of computers catches up, it will be possible to simulate the elementary processes of MBE by determining the instantaneous electronic states based on quantum mechanical and non-empirical methods and calculating the forces acting on atoms. can. Since the current atomic arrangement is known from the LEED/RHEED analysis, we use this as a starting value, and based on the theoretical model, we can determine what will happen at the atomic level if MBE continues as it is based on data such as the current beam intensity, degree of vacuum, and substrate temperature. It is predicted that the structure will change. Therefore, if there is a desired atomic structure, it is estimated from the simulation results what process parameters should be selected to approximate it.

Here, we will discuss in more detail the case of analyzing and predicting lp1 using a classical theoretical model. First, use some method to express the interaction between atoms as a function of their distance. Specifically, a functional form is assumed, and the parameters included therein are determined to match the cohesive energy height 7Ipj value, lattice vibration spectrum, etc. There are a number of assumed functional forms, so how to choose one is a difficult problem, but in the case of a simple semiconductor, the following VFF! ! Good results can be obtained by assuming the potential) jf is known.

V-aΣ(lrl r'l dz)2+βΣ(
θ1, 100 14k) 2 + VoHere, ``1 is the position vector of atom i, dll is the atom? The equilibrium interatomic distance between and j, θ1, is the angle formed by atoms i, j, and k, and θ0■, is its equilibrium value.

Actual measurement data may be used for these equilibrium values. Since vo means the depth of the potential, it can be estimated from the measured value of cohesive energy. α. β is a parameter that determines the stiffness of the lattice, but if we assume an appropriate cluster and calculate the molecular vibration of this cluster, the frequency is a function of these parameters, and the corresponding experimental results (Raman scattering It can be determined by the method of least squares from experiments (experiments, infrared absorption experiments, etc.). The atomic structure is determined according to the interatomic interactions determined in this way, and this stable solution is obtained by solving the classical Newtonian equation of motion (or an equation physically equivalent to this) rrx d2r/dt2--gradV (r). You can decide from m is the mass of each atom. Since Newton's equation of motion is a second-order differential equation related to time, it is necessary to provide appropriate initial conditions, but this can be done by numerically integrating on a computer using actual measurements as described above. It is also possible to incorporate considerations such as the plate being fixed and the substrate not being unnecessarily heated due to collisions with molecular beams in the form of a rising condition for the differential equation. Note that the temperature of the system (near the growth surface by MBE) can be estimated from the kinetic energy of atoms by assuming the law of equal distribution of energy. Since the number of falling atoms can be estimated from the actual aFI value of the flow rate of the molecular beam, it is possible to simulate MBE growth based on the actual δPI value in a paleogenetic manner.

From such a simulation, the process parameter dependence of the MBE process can be determined immediately. It is possible to predict how Kisho Bocho will progress if MBE growth continues as it is, and if this is not desirable, it is possible to change the parameter values to bring it in a more desirable direction. Furthermore, by changing the process parameter values several times to find out what type of crystal growth will occur from the current atomic arrangement, and selecting the most appropriate (desired) value from among these, we can improve efficiency. <MBE growth is −==
It is J-Noh. Of course, MB is also used during simulation execution.
EM length is being carried out simultaneously.

Therefore, if simulations are performed all the time, actual measurement results cannot be taken into account. Therefore, if you import actual measurement results at appropriate time intervals, perform a simulation, investigate as much as you can within the allowed time, control the parameters, and once that is completed, import the actual measurement values again and repeat the simulation. good. Of course, you can use a preset value for this appropriate time interval, but
It is also possible to recognize a skewer in which the simulation result and the actual measurement value deviate by a certain degree, clear the simulator at that stage, import the actual measurement value, and restart the simulation from there.

Example 2 Synthesis of a semiconductor or metal with a superlattice structure by MOCVD method, examining its electronic state using photoelectron spectroscopy (PES) or Auger electron spectroscopy (AES), and calculating the electronic state using a computer. Next, an example of a system that controls synthesis by applying feedback will be described while comparing with the following. The overall system configuration is shown in Figure 2. 2
1 is an MOCVD apparatus, 22 is a raw material gas supply system, and M
An Auger spectrometer 23 is attached to the OCVD apparatus 21. 24 is a computer for Auger process angle q analysis, 25 is a simulator, and 26 is a control computer for making judgments and controlling the number of capsules. The basic configuration of the system is shown in Figure 1.

The MOCVD method, which uses chemical reactions on the surface of organic metals to grow crystals, is attracting attention as being the easiest to control among CVD methods. On the other hand, there are many theoretical predictions regarding superlattices, and control on the atomic order has become essential. In particular, the technology to uniformly grow 11 atoms is extremely difficult! It is essential. This is possible through lli atomic layer epitaxy (A
LE) There is a law. In contrast to the usual MBE method, which grows crystals by bombarding molecular beams, this method takes advantage of the selectivity of the reaction of molecules on the solid surface, and in principle makes ALE possible. That is. In order to realize ALE with MBE, we must continue to observe this in detail and, for example, LEE.
Methods such as operating the shutter of the molecular beam source are used to detect when the D pattern changes. However, by taking advantage of the specificity of reactions on solid surfaces, it is possible to automatically control growth (for example, by growing only a monoatomic layer and stopping that growth). A specific example of this is the specificity of photoexcitation reactions near solids. As described above, the MOCVD method has excellent characteristics that MBE does not have, but the large number of equipment parameters is preventing its widespread use.

For this purpose, detailed measurement and evaluation of physical properties is essential.

The AES method uses the Auger process, in which inner-shell electrons are ejected by electron collision, and electrons (usually valence electrons) are released when the holes created here relax without electromagnetic radiation. This method directly examines the density of banded bears. Since this electronic state extremely strongly depends on the element it originates from, it is also possible to perform elemental analysis based on this. By the way, it is quite difficult to investigate the Auger process itself theoretically. Since the Auger process is fundamentally a many-body problem, it is impossible to solve it directly using current condensed matter theory. However, since the electronic state obtained by the AES method reflects the local density of states, the electronic state of the system including Table 1 is calculated and the angle q is calculated based on the electron state density spectrum obtained from this. Is possible. To do this, assume a certain atomic arrangement (or you can calculate the energy of the entire system ab initio and use the atomic arrangement that gives this minimum value without assuming it), and then calculate the Schrödinger equation for this system. I made a mistake and found the density of states directly.

By comparing this with experiment, it can be determined whether the assumed atomic arrangement was correct. Although electronic state calculations can be performed ab initio, the ab initio method here means that no empirical, or ad hoc, parameters are introduced, and it is inevitable that some kind of approximation will be used. Calculate using. For example, it is unavoidable at present to approximate a many-electron problem integrally or to search for a ground state in a restricted space. Of course, from the standpoint of theoretical physics, approximations are made to extract as much of the essence of the skewer as possible, and even with these approximations, results are still directly comparable to experiments. Although such slight approximations are unavoidable, theoretical results can still be obtained that can be compared with experiments.

Optical properties of semiconductors, such as optical absorption spectra, may be related to elementary excitations unique to solids (exciton, polariton, plasmon, etc.), but basically they can be understood and predicted beyond their band structure. . Optical properties are extremely important physical quantities when semiconductors are used in optical devices, but calculations of electronic states and associated predictions are important in order to design them as desired. In the current state of physical theory, it is difficult to first give optical properties and then determine the electronic state (band structure) that matches them, but conversely, it is difficult to theoretically estimate optical properties from the electronic states. It's relatively easy. Of course, there should be a one-to-one correspondence between the electronic state and the optical property, but this is the same situation as the integrands not being the same just because the integral values are the same.

Specifically, the optical absorption spectrum is the imaginary part ε2 of the dielectric function.
14 from (ω), which is ε2 (ω)-1/ω2fΣ2/ (2 π) 3l < vkl p
l vh> B 6 [E (k)
-E (k') --)5 ω d
It is k. The sum is taken at point K, which is occupied by electrons, and point k', which is not occupied by electrons. p is the momentum operator. This formula does not take temperature effects into account.

H ε2 (ω) at the degree of inhibition has the Fermi component 4i of the electron.
need to be considered. Since the wave function ■ is determined as a system of eigenfunctions of Schrödinger's equation, it can be obtained numerically by solving the electronic state with a computer, and the optical absorption spectrum of this system can be theoretically understood from the above equation. Compare this with the optical characteristics in the design (1-),
Feedback can be applied so that the desired substance is synthesized.

Other optical properties can also be determined using a similar method.

It is known that in a semiconductor superlattice, electronic states form subbands with finer horizontal structures than normal band structures. The subband structure is directly controlled by the superlattice thickness, and this structure gives semiconductor superlattices a characteristic optical absorption spectrum. This is because the optical absorption spectrum in this system depends on state transitions between subbands.

By the way, electron mobility is an important parameter that determines the switching speed of a device. The higher the mobility of electrons, the higher the speed of switching operation. This makes GaAs more attractive than the currently widely used silicon. On the other hand, in a semiconductor superlattice, modulation doping causes band bending due to space charge loss, which allows conduction electrons and impurity ions to exist in different 1φ, and this causes the effect of Coulomb decay. is significantly weakened. Therefore, it is possible to create a device capable of high-speed operation. This was the first reason why semiconductor superlattices attracted attention. As can be seen from the above reasons, the mobility of electrons in a semiconductor superlattice largely depends on the thickness of the superlattice, the amount of doping, and its spatial profile. For this reason, it is important to obtain desired characteristics while adjusting these. Now, the electron mobility μ is ideally given by the following equation assuming the Boltzmann method.

μ1eτ/m r−′−f (vk−V1′)” EQ (k, k
−) dΩ / v t ・E where τ is the relaxation time, Q (k, k) is the scattering probability, ■, is the velocity of the electron with wave number vector k, E is the applied home, m
is the mass of the electron, and e is the elementary charge.

To calculate this integral or scattering probability, we need a self-consistent wave function and an energy fixed value. Scattering centers may include impurity ions, lattice imperfections, and scattering from ascending planes of the superlattice.

By incorporating these effects, the scattering probability can be evaluated, and the electron mobility can be predicted by IIF+.

Furthermore, using the example of a semiconductor superlattice, if for some reason the synthesis equipment does not control well and the thickness of a certain layer becomes thicker than expected, the normal superlattice In the production method, there was no choice but to scrap this lot, but as shown in the Honjo example, we calculated the electronic state in the superlattice structure that had become even thicker, and from this, we could calculate, for example, light absorption. If the spectrum is within the required specifications, calculate this, calculate the deviation between the current system and the required amount, and then determine how thick the superlattice structure should be made to be on top of this. The process parameters required for this can be estimated by the simulator 25. In this way, it is possible to finally obtain a product that meets the design specifications.

Of course, it is not possible to create the desired product by examining only optical properties or electron mobility.

Even if these properties are desirable, such substances cannot exist stably if the structure itself is unstable.

As mentioned earlier, structural stability can be discussed from the electronic state, so we will simulate whether 7i can exist stably while maintaining the desired properties when the current synthesis is continued. That is also important. From a broader perspective, there are optimized parameters, that is, a good solution in terms of optical properties, but when combined with structural stability, it is important to select within the range that is the best possible. .

This embodiment can achieve this.

Embodiment 3 Next, a third embodiment will be described in which a plurality of synthesis devices operating in a row of dishes are provided.

The system constructed by the present invention is essentially an integrated system for controlling the synthesis of substances as shown in Figures 1 and 2, but the synthesis method that can be controlled at the atomic and molecular level has an industrial meaning. It is said that productivity is extremely poor. Therefore, no matter how well the synthesis method is controlled, which is said to have poor productivity, the quality of the synthesized substance will improve, but the productivity, or throughput, will not improve much. It is not economical to provide a Jf number device and a prediction system for each such synthesizer. By the way, if the synthesizer can be controlled at the atomic and molecular level, when operating it in parallel, the exact same operation, that is, synthesis is possible if the conditions are broken. There is no need to provide a system for every component.

? If there are multiple synthesis devices based on the same principle, some of them are designated as representative devices and an evaluation device is attached to them. Basically, this can be done with one device. FIG. 3 shows the system configuration of such an embodiment. In the case of figure, multiple MB
E device 31■, 312,..., 31. are installed in parallel. For example, as shown in Example 1, LEED/R
When crystal growth is performed by the MBE method while the solid surface can be harvested using HEED, there is only one merging device, MB in the example in Figure 3.
A LEED evaluation device 32 is provided in the E device 311. With this L E E D Fl-value device 32, L E E D
/RHEED pattern is taken and analyzed by the angle analyzer 33.
A molecular dynamics simulation is performed using the simulator 34 based on the results of the angle analysis. Based on the simulation results, the control computer 35 makes decisions and controls variables, and sends control signals to all MBE devices 31.
Send to 1-317.

If the control is sufficient, the same LE can be used from other MBE devices.
An ED/RHEED pattern should be obtained. Therefore, it is sufficient to sample from one place. Quantitative molecular dynamics simulations performed on computers, of course, always give the same results based on the same human data. Therefore, even if this is done using a number calculator, it is just wasteful.

Control the process parameters of all equipment according to the results obtained from the simulation.

This mucilage allows for highly productive and controlled synthesis.

Furthermore, since it is necessary to prepare only one set of the tr value device and the simulator for pre-All, it is also excellent in economical efficiency. Of course, if necessary, it is also possible to perform a simulation taking into consideration the variation in the combining ability between synthesizers, the accuracy of evaluation by the evaluation device, and the like. Although it is quite difficult to evaluate the variation itself and take it into account, it is possible to estimate it in a region where the results are not so sensitive to human parameters. In other words, a permissible range is determined and process parameters are determined within that range. This result can be fed back to all synthesis devices. Originally, if we try to contain substances with high precision,
Regions whose results are not sensitive to process parameters must be selected and merged. Such areas often exist. Therefore, there is no problem as long as the variation in performance between the combining devices is smaller than this.

In this way, production capacity can be increased by operating a single synthesis device in parallel, even if the synthesis device has poor productivity, and control can also be performed in a single device.

Embodiment 4 Next, an embodiment will be described in which there are a plurality of synthesis devices operating in parallel, and a plurality of different evaluation devices are provided in addition to these devices.

There are various experimental methods for analyzing solid surfaces.

Some experiments characterize the electronic state near the surface, and others provide results that reflect the positions of atoms on the surface. The theoretical framework for analyzing this must also directly reflect the physical processes that correspond to each experience. However, from a broader perspective, all these phenomena are closely related.

For example, the electronic state of the surface naturally depends on the arrangement of atoms near the surface.
This is determined by the position of the atom, and the position of the atom must be determined so that the energy of the entire system is minimized, so its electronic state has a direct effect on this. In other words, in physical terms, Kuchinen W is only one possibility according to the ``principle of least action,'' which states that out of various possibilities, only the one that minimizes a certain naughty child is realized. You are choosing your gender. For this reason, self-consistent cross-sections are always required, but what can be obtained by various revaluation techniques is only a cross-section with such a self-consistent angle t. For this reason, the only description of the system of loss must be one that can be obtained comprehensively using various J/F value methods. The Pre-All method has similar aspects. It would be possible to express a single nature in a consistent manner using both the classical molecular dynamics method and the quantum-based description method.

As described in Example 3, synthetic methods controlled on an atomic/molecular scale are generally inferior in terms of isoproducibility. Therefore, it is required to operate multiple synthesis devices in parallel, but when evaluating this state using an evaluation device, different phenomena are evaluated rather than using the same evaluation method. A possible method is to attach different evaluation devices to the synthesis device so that the same physical state can be observed using various observation means. The information obtained is multifaceted,
The description of the system will be more complete. Specifically, let us consider the case where a superlattice structure is obtained by growing a semiconductor thin film using the MBE method similar to the previous example.

In this case, in order to directly investigate the position of the atom, there are methods such as LEED and RHEED that measure it in reciprocal space, methods that measure it directly in real space such as STM and AFM, and short-distance methods such as STM and AFM. SEXAFS etc. are becoming old as things that need to be finely cut down on order. From the same atomic arrangement, we must find the results of measurements that give the same atomic position as a result. In many cases, different experiments give complementary results. Therefore, it is often difficult to determine all the parameters necessary for computer simulation from a single experiment, and similarly, it is difficult to completely determine the control factors of a synthesizer from a single experiment. It is particularly difficult to reconcile large experimental findings involving electronic states with experimental results regarding atomic positions.

In order to theoretically follow the elementary process of M B E excursion, we must first understand the atomic arrangement on the solid surface. This allows us to theoretically know the electronic state on the solid surface. Of course, in an ideal system, it should be possible to determine the atomic energy by minimizing the total energy of the system, but in the current system, the situation may not be ideal due to contamination, etc. do not have. Therefore, it is necessary to know the current atomic and molecular arrangement, including details of adatoms and molecules. This includes LEED, RHE as mentioned above.
ED, STM, AFM, SEXAFS, etc. are used. Now, the electronic state, especially the level of solid H on the surface, is often determined by very delicate factors. Recently, adsorption on simple metal surfaces′
It has been revealed that the decrease in the function I due to alkali metal atoms is a phenomenon that occurs when the electric current between the adatom and the base metal is pushed out into the vacuum region. The charge distribution and therefore the electronic state are determined. In the elementary process of MBE, it is thought that the irradiated molecules cause some kind of chemical change with the parent phase, so the relative positional relationship of the surface states is crucially important. Even within the range of experimental errors in atomic arrangement, it is quite possible that it will make a large contribution to the unit relationship. For this reason, it is important to know the electronic state experimentally. For this purpose, it is important to know the average electronic state by using optical electron spectroscopy or XPS-UPS, and at the same time, it is important to know the electronic state directly by STS. It is also important to obtain the local density of states of .

FIG. 4 shows the system configuration of the present embodiment. In this embodiment, the plurality of synthesis devices include an MBE device 41. ~41.
This shows the case where . For evaluation of films obtained by these devices, a LEED/RHEED evaluation device 42,
SEXAFS evaluation device 43 and LEELS evaluation device 4
Use 4.

Of course, more evaluation methods other than these may be provided, and if technically possible, a single synthesis device may be equipped with a plurality of evaluation devices. In this case, it is naturally impossible to observe simultaneously, but it is possible to process in a time-sharing manner.

Information about the surface obtained from these evaluation devices is sent to analysis devices 45, 46 and 4 according to each original raccoon.
7. The LEED/RHEED pattern angle q analyzer 45 performs this Fourier analysis on the pattern in the reciprocal lattice space to obtain the atomic arrangement in the global region. SEXAFS slam analysis device 46
From this, it is possible to determine the viscosity separation between atoms, especially the crystallinity of the adsorbent from the solid surface, according to the multidispersion 8L theory. Depending on the LEELS baton analysis device 47, the density of state curve of Chiko can be directly calculated. 1. There are many. Empirically, this gives us sufficient information about the appearance;
First, we set up a model that can produce consistent viscous fruit. 111 with the atomic arrangement and the position of the sucker, the Schrödinger equation becomes the angle 'l For example, many electrons are missing due to unitary approximation of the child problem, problems that arise due to not being able to take completeness when expanding the true wave function with known basis functions, etc. It is possible that the situation cannot be reproduced.

Therefore, it is necessary to create a model that satisfies all of these requirements. There are many ways to create a model, but the one with the best chance of success is to avoid destroying the physical image as much as possible. In other words, do not introduce empirical parameters unnecessarily.
, Basically, assuming that the unitary approximation is almost valid, we use physical values for direct Coulomb energy and kinetic energy. However, since the physical validity of the Tonko correlation term, which directly represents the many-body effect, is often unclear, it is not evaluated using well-known interpolation formulas, but rather is parameterized to some extent and compared to actual measurements. It is possible to leave some room for matching. Furthermore, since there is no need to limit the number of theoretical models to one, it is also possible to prepare several theoretical models that also include empirical parameters. Some people choose a theoretical model from among these that can reproduce the current evaluation results without contradiction as much as possible, and adopt that theoretical model.

In this way, while minimizing the discrepancy between theory and reality, it is possible to use the simulator 48 to simulate how the thin film will grow as a result of continued MBE growth using the current 713 process parameters. be. In addition, by changing several process parameters and examining in real time how each of them grows, these results are compared with the currently synthesized one, and the control computer 49 selects the optimal process parameters. , this is the MBE device 411-41. will be given feedback.

Since we are considering bringing empirical parameters into the theoretical model, we first fit these model parameters using the simulator 48, and confirmed that the current 7E Theory 1 model reproduces the experimental values well. Then, predictions for various process parameters must be made within a limited time. An example of the procedure is shown in FIG.

In this case, the scale of the simulation is expected to be very large, and the amount of calculation to be processed is enormous, but there is a good chance that future ultra-high-speed computers will be able to process this in real time.

Example 5 A failed example incorporating multiple simulations will be described below.

In the present invention, based on the information obtained from the JF value device, we use a computer to examine how the synthesis will change from the current situation in real time according to a theoretical model, and then synthesize the optimal parameters. The idea is to learn about this in advance and feed it back into the synthesis in real time. This goes beyond the question of what will happen if the current situation continues as it is. It is necessary to simulate several cases by varying the process parameters for the seven material situations. These simulations are of course completely independent, so if multiple computers are provided, it is possible to run them in parallel and process them at the same time. This means that the computer running the simulation is parallel = 1
``Calculating machine (parallel processor or heckle processor)
Whether it is or is not is a completely different question.

FIG. 6 shows the system configuration of this failed example. This also uses an MBC device 61, which is equipped with a LEED evaluation device 62 and a LEED baton analysis device 63. Further, the current number of simulators 641 to 64n, into which common structural data from the 414 analysis device 63 and different external human power data are taken in, are also provided.

The simulation I15 results are sent to the control computer 65 in the same way as in the previous embodiments.

Here, the theoretical models used in the simulators 64, to 64n are the same, and the human data used for simulation is also the same because it is based on evaluation of the current state of the material. Simulations are performed by adding different process parameters to this. For example, when synthesizing at the current temperature, heating at 1°C/min,
It is possible to simultaneously investigate cases such as cooling at a rate of 0.5°C/min. The obtained results are compared with each other in the control computer 65, and the one closest to the specified specifications of the currently synthesized substance is selected, and this is fed back to the synthesis device.

Note that if you have multiple computers and can perform simulations at the same time, it is not necessarily necessary to perform calculations according to the same theoretical model.

For example, when there are several levels of precision in a theoretical model, there are two levels: one is to use molecular dynamics or the Monte Carlo method based on astronomical mechanics, and the other is to use non-empirical electronic theory to investigate the atomic arrangement. Based on Schrödinger's formula
Assuming that there is a level in which the placement of atoms is determined by minimizing the total energy of the system, the former has an overwhelming advantage over the latter in terms of computational complexity, but the theoretical validity of the former is questionable. and the reliability of the results pales in comparison to the latter. In such cases, it is especially important to prevent calculations from being completed in time and the synthesizer to be unable to be controlled in real time.
For the time being, I will use a method with a small amount of 1 nose and apply feedback between 1H and 2R, but at the same time 1:
You can do the 51 calculations later, but you can continue the simulation with a more reliable theoretical model using another 21 calculation machine.
It is possible to confirm its validity, or if a deviation occurs, to change the process parameters in a direction to absorb the deviation as appropriate. In this way, the emotion of emulation can be increased. Since there are various levels of theoretical models derived from condensed matter theory, simulations can be performed according to each level.

The system configuration shown in FIG. 6 allows such control.

In the description of the husband example above, composition elimination, evaluation method, child Ul
Although we have listed specific methods for each method and shown the system structure, this is of course to make the explanation easier to understand. If these are combined and systemized according to Masaaki Hon's policy, they are included in the present invention. Further, it is possible to configure a system by combining various embodiments described so far, and these should naturally be included in the present invention.

Note that, since the present invention is a system configuration or technology itself, it does not depend on whether or not devices for synthesizing, evaluating, and predicting substances are integrated.

In other words, it may be an integrated device, and if it cannot be integrated, it is sufficient to configure it as a system.

[Effects of the Invention] As described above, according to the present invention, synthesis is performed at the atomic/molecular level. Material synthesis can be controlled at the microscopic level by visually evaluating IL and predicting physical properties in real time based on the results.

This makes it possible to specifically design materials with desired properties according to design specifications, just like designing a car or television, that is, "material design." In addition to this, it will become possible to synthesize new substances, which is expected to have extremely large effects, and will lead to the creation and development of completely new fields.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the system configuration of Embodiment 1 of the present invention, FIG. 2 is a diagram showing the system configuration of Embodiment 2 of the present invention, and FIG. 3 is a diagram illustrating the system configuration of Embodiment 3 of the present invention. , Fig. 4 is a diagram showing the system configuration of the fourth embodiment of the present invention, Fig. 5 is a diagram showing an example of the simulation procedure in the large embodiment 4, and Fig. 6 is a diagram showing the system configuration of the large embodiment 5 of the present invention. FIG. 11... MBE device, 12... RHEED electron gun, 13... RHEED river observation device, 14... Screen, 15... Camera, 16... Quadrupole mass spectrometer, 17... Structural analysis computer, 18... Simulator, 19... Control computer, 21...
・MOCVD apparatus, 22... Raw material gas supply system, 23.
... Auger spectrum mouth meter, 24 ... Auger analysis computer, 25 ... simulator, 26 ...
- Control computer, 311-317... MBE device, 32... LEED evaluation device, 33... LEED
Pattern analysis device, 34...Simulator, 35...
Control computer, 411-41. ...MBE device, 42...LEED evaluation plow device, 43...SEXAF
S; #Do value device, 44... LEELS evaluation device, 45
...LEED batan angle analysis device, 46...SEXA
FS pattern evaluation device, 47...LEELS baton evaluation device, 48...Simulator, 49...Control computer, 61...MBE device, 62...LEED
Evaluation device, 63... LEED pattern angle analysis device, 6
4, ~64. ...Simulator, 65...Control computer.

Claims (3)

[Claims] (1) A means of synthesizing a substance, a means of observing and evaluating the properties of the synthesized substance on the spot in real time, and using the information obtained from this evaluation means as a starting point based on an appropriate theoretical model. , predicts what will happen if synthesis continues under the same conditions, and estimates how to make modifications from here to obtain the desired substance.It has an execution speed that is sufficiently fast compared to the synthesis speed of the substance. A substance synthesis control system comprising: a simulation means; and a means for feeding back the results of the simulation means to the synthesis means to control synthesis conditions. (2) The means for synthesizing the substance is a molecular beam epitaxial growth (MBE) device, a vapor phase growth method using an organometallic compound (
MOCVD) equipment, vacuum evaporation equipment, sputtering equipment, ion cluster beam (ICB) equipment, Langmuir
The substance synthesis control system according to claim 1, which is any one of Blodgett (LB) film forming apparatuses. (3) The evaluation means are low-energy electron diffraction (LEED), reflection high-energy electron diffraction (RHEED), and extended X-ray absorption fine structure (EXAFS).
and surface extensive X-ray absorption fine structure (SEXAFS), transmission electron microscopy (TEM), scanning electron microscopy (SEM)
), means for evaluating atomic structure using scanning tunneling microscopy (STM), atomic force microscopy (AFM) or field ion microscopy (FIM), photoelectron spectroscopy (PES, XPS, UPS), Auger electron spectroscopy (AES) , high-resolution electron energy loss spectroscopy (H
REELS, LEELS), means for measuring electronic states using Penning ionization electron spectroscopy or metastable molecular separation spectroscopy (MDS), means for measuring vibrational spectra using infrared absorption (IR) or Raman spectroscopy, electron spin Means of evaluating magnetic properties using resonance (ESR), nuclear magnetic resonance (NMR) or thermal neutron scattering experiments, or secondary ion mass spectrometry (SIMS) or ion neutralization techniques (INS) The substance synthesis control system according to claim 1, which is a means for performing an analysis.