Film formation system and film formation method for wafer

Abstract

To provide an information processor and operation detector, capable of recognizing more complicated gesture operations of a user.SOLUTION: A deposition system includes a deposition device and a computer. The deposition device has a deposition chamber capable of fitting a plurality of deposition species. The computer has a calculation region for predicting the time required for deposition when arrangement of the deposition species is set on the basis of a calculation model including an Ising model or a QUBO.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a film forming system. Mu and and a method for forming a film on a wafer. [Background technology]

[0002] 2. Description of the Related Art A film forming apparatus for forming a multilayer film on a wafer is known (for example, see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 5650760 Summary of the Invention [Problem to be solved by the invention]

[0004] The film forming apparatus has a plurality of film forming species for forming each layer constituting the multilayer film. The film forming species is, for example, a raw material used when stacking the multilayer film, for example, a target material used in a sputtering method. The film forming apparatus transports the wafer to a position facing a predetermined film forming species for each layer to be formed, and performs film formation. The required thickness of each layer is different, and the film forming rate is different for each film forming species. Therefore, the time required for film formation is different for each layer, and film formation of a specific layer may be the rate limiting factor for throughput. Changing the arrangement of the film forming species in the film forming apparatus may prevent the wafer from being in a film formation standby state, reduce the transport time, and improve the throughput.

[0005] As the number of layers in a multilayer film increases, the number of deposition species also increases, resulting in an infinite number of combinations of deposition species arrangements. Until now, the deposition species arrangements were often set based on the experience of the operator. However, further improvement of throughput is required.

[0006] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a film formation system, a factory system, and a method for forming a film on a wafer, which can reduce the time required for film formation. [Means for solving the problem]

[0007] In order to solve the above problems, the present invention provides the following means.

[0008] (1) A film formation system according to a first aspect includes a film formation apparatus and a computer, the film formation apparatus having a film formation chamber capable of accommodating a plurality of film formation species, and the computer having a calculation domain for predicting the time required for film formation when the arrangement of the film formation species is set based on a calculation model including an Ising model or a QUBO.

[0009] (2) In the film formation system according to the above aspect, the Ising model or the QUBO may have elements whose number corresponds to the product of the number of the film formation species and the number of the film formation chambers, and each of the elements may be a binary variable.

[0010] (3) In the film forming system according to the above aspect, the energy function of the Ising model or the QUBO may include an energy function corresponding to a throughput of the film forming apparatus.

[0011] (4) In the film forming system according to the above aspect, the Ising model or the QUBO may use values ​​corresponding to the transport time of the wafer on which the film is to be formed and the processing time in each process as parameters for determining the interaction parameters between the elements.

[0012] (5) In the film forming system according to the above aspect, the Ising model or the QUBO may be defined on the condition that a process in any of the steps can be replaced with a process in another chamber.

[0013] (6) In the film formation system according to the above aspect, the film formation apparatus may further have a process chamber for performing a process other than film formation, and the processing in the process chamber may be defined as a step that cannot be replaced with processing in another chamber in the Ising model or the QUBO.

[0014] (7) In the film formation system according to the above aspect, the Ising model or the QUBO may have a constraint, and the constraint may reflect an upper limit on the number of film formation species of the same type among the film formation species.

[0015] (8) In the film formation system according to the above aspect, the Ising model or the QUBO may have a constraint, and the constraint may reflect an upper limit on the number of slots for the film formation species in the film formation chamber.

[0016] (9) In the film forming system according to the above aspect, the Ising model or the QUBO further includes, as the constraint, an additional element whose number corresponds to the sum of the number of slots in the film forming chambers, and the additional element may be a binary variable.

[0017] (10) In the film formation system according to the above aspect, the Ising model or the QUBO may have a constraint, and the constraint may reflect a condition that film formation species that may be simultaneously formed on a wafer are placed in the same film formation chamber.

[0018] (11) In the film formation system according to the above aspect, the Ising model or the QUBO may have a constraint, and the constraint may reflect a condition that the times to reach the end of the life of the plurality of film formation species are approximately constant.

[0019] (12) In the film forming system according to the above aspect, the Ising model or the QUBO has a constraint, and the Ising model or the QUBO further has an additional element representing a wafer process and a processing time of the process as the constraint, and the additional element may be a binary variable.

[0020] (13) In the film forming system according to the above aspect, the Ising model or the QUBO has a constraint, and the Ising model or the QUBO further has an additional element representing the loading time of the wafer as the constraint, and the additional element may be a binary variable.

[0021] (14) In the film formation system according to the above aspect, the computer may further have a memory area for storing, as training data, an arrangement of a specific film formation species and a time required for film formation in the case of the specific film formation species arrangement, and a learning area for determining interaction parameters between elements of the Ising model or the QUBO by machine learning based on the training data and designing a computational model.

[0022] (15) In the film formation system according to the above aspect, the memory area may reacquire, as teacher data, the arrangement of film formation species used for the prediction in the calculation area and the time required for film formation predicted in the calculation area.

[0023] (16) In the film forming system according to the above aspect, the machine learning may be performed by a factorization machine.

[0024] (17) A factory system according to a second aspect includes a plurality of devices and a computer, and the computer has a computational domain that predicts, based on a computational model including an Ising model or a QUBO, a time required for processing when the plurality of devices are arranged in a plurality of areas in which the devices can be installed.

[0025] (18) In the factory system according to the above aspect, the computer may further have a memory area that stores, as training data, the placement of the device in the area and the time required for processing in that placement, and a learning area that determines interaction parameters between elements of the Ising model or the QUBO by machine learning based on the training data and designs a computational model.

[0026] (19) A third aspect of the present invention relates to a method for forming a film on a wafer, the method comprising the steps of: designing a computational model including an Ising model or a QUBO model using an arrangement of a plurality of film formation species in a film formation apparatus as an element; predicting, based on the computational model, a time required for film formation for an arrangement of film formation species; and, if the predicted time is equal to or shorter than a predetermined time, applying the arrangement of film formation species as an arrangement of film formation species in the film formation apparatus.

[0027] (20) In the wafer deposition method according to the above aspect, the step of designing the computational model may include a step of acquiring, as training data, a specific arrangement of deposition species and a time required for deposition in the case of the specific arrangement of deposition species, and a step of determining interaction parameters between elements by machine learning based on the training data.

[0028] (21) In the method for forming a film on a wafer according to the above aspect, when the time predicted based on the calculation model is equal to or longer than a predetermined time, the method may further include a step of reacquiring, as teacher data, the arrangement of the film formation species and the time required for film formation when the film formation species is arranged. Effect of the Invention

[0029] The film formation system, factory system, and wafer film formation method according to the present invention can reduce the time required for film formation. [Brief description of the drawings]

[0030] [Figure 1] 1 is a schematic diagram of a film forming system according to a first embodiment. [Diagram 2] This is an image of the Ising model, QUBO. [Diagram 3] 1 shows an example in which the arrangement of film formation species in a film formation chamber is reflected in the state of elements in a calculation model. [Figure 4] This is an example showing a state in which the arrangement of film formation species in a film formation chamber is reflected in the calculation model and additional elements are added as constraints. [Diagram 5] FIG. 2 is a schematic diagram of a computing element that performs annealing calculations in a calculation domain. [Figure 6] FIG. 1 is a schematic diagram of a spin photodetector. [Figure 7] FIG. 1 is a schematic diagram of a film forming system according to a second embodiment. [Figure 8] FIG. 1 is a flow diagram of a method for designing a computational model. [Figure 9] FIG. 13 is a schematic diagram of a factory system according to a third embodiment. [Figure 10] 4 is a time chart showing all steps of a film forming apparatus in Example 1 and Comparative Example 1. [Figure 11] FIG. 13 is a Gantt chart showing the results of throughput in Example 1. [Figure 12] FIG. 13 is a Gantt chart showing the results of throughput in Comparative Example 1. [Figure 13] 4 is a time chart showing all steps of a film forming apparatus in Example 2 and Comparative Example 2. [Figure 14] FIG. 13 is a Gantt chart showing the results of throughput in Example 2. [Figure 15] FIG. 13 is a Gantt chart showing the results of throughput in Comparative Example 2. [Figure 16] 11 is a time chart showing all steps of a film forming apparatus in Example 3 and Comparative Example 3. [Figure 17] FIG. 13 is a Gantt chart showing the results of throughput in Example 3. [Figure 18] FIG. 13 is a Gantt chart showing the results of throughput in Comparative Example 3. [Figure 19] FIG. 13 is a diagram showing the results of throughput in Example 4. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0031] The present embodiment will be described in detail below with reference to the drawings as appropriate. The drawings used in the following description may show characteristic parts in an enlarged scale for the sake of convenience in order to make the features of the present embodiment easier to understand, and the dimensional ratios of each component may differ from the actual ones. The materials, dimensions, etc. exemplified in the following description are merely examples, and the present invention is not limited thereto, and may be appropriately modified and implemented within the scope of the present invention.

[0032] "First embodiment" 1 is a schematic diagram of a film formation system 300 according to a first embodiment. The film formation system 300 includes a film formation apparatus 100 and a computer 200.

[0033] The film forming apparatus 100 includes, for example, a plurality of film forming chambers 10, a process chamber 20, a transfer chamber 30, and a load lock chamber 40.

[0034] The film forming apparatus 100 may be, for example, a film forming apparatus that employs a single-wafer processing method in which wafers are processed one by one. The film forming apparatus 100 is a film forming apparatus that uses a sputtering method, a film forming apparatus that uses a chemical vapor deposition method, or the like.

[0035] The film formation chamber 10 is a chamber in which a film is formed on a wafer. For example, a plurality of film formation species 1 can be installed in one film formation chamber 10. Each of the film formation chambers 10 shown in FIG. 1 has eight film formation species 1. By having a plurality of film formation species 1, one film formation chamber 10 can form a plurality of layers. The film formation species in each film formation chamber 10 may be different from each other, or the film formation chambers 10 may have a plurality of film formation species made of frequently used materials.

[0036] The process chamber 20 is a chamber in which processes other than film formation are performed. Examples of processes other than film formation include heating and cooling of the substrate, partial peeling of the formed film, and chemical reaction processes such as oxidation. The transfer chamber 30 is a chamber in which wafers are transferred between the respective chambers. The load lock chamber 40 is a chamber that connects the outside with the vacuum region, and is sometimes called a vacuum reserve chamber. For example, a wafer that enters the transfer chamber 30 from the load lock chamber 40 is transferred to one of the film formation chambers 10 or the process chamber 20 in the order of the process.

[0037] The computer 200 determines the arrangement of the film forming species 1 in the film forming apparatus 100. The arrangement of the film forming species 1 means which film forming species 1 is arranged in which chamber. The computer 200 may also determine the order of transport of wafers in the film forming apparatus 100 and the timing of loading wafers into the film forming apparatus 100.

[0038] The computer 200 has a calculation domain 201. The calculation domain 201 includes a calculation model based on the Ising model or QUBO. The calculation domain 201 predicts a time required for film formation when the film formation species 1 is arranged in a specific arrangement. The arrangement of the film formation species 1 can be determined based on the calculation result of the computer 200.

[0039] The Ising model is a model that predicts the overall stable state when multiple elements interact with each other and a forcing force is applied to each element. Figure 2 shows an image of the Ising model. The Ising model has multiple elements E that interact with each other through a forcing force F. Each element E consists of a spin s. The spin s can be either up or down. The element E is represented by a binary variable. Depending on the setting of the forcing force F, the stable state can be a state in which adjacent spins s are in equilibrium, or an antiparallel state. The forcing force F is called the interaction parameter.

[0040] The Ising model is expressed by the following energy function (cost function).

[0041]

number

[0042] where σ i , σ j is an input variable, which can be +1 or -1. σ i , σ j corresponds to the state of spin s in Figure 2. J ij is the interaction parameter. J ij corresponds to the forcing force F in Figure 2. h i is a parameter associated with external factors.

[0043] QUBO (Quadratic Unconstrained Binary Optimization) is a computational model that can be converted to an equivalent Ising model. QUBO can be applied to computational models in the same way as the Ising model. QUBO is expressed by the following energy function (cost function).

[0044]

number

[0045] Here, q i , q j is an input variable, which is either 1 or 0. i , q j corresponds to the state of spin s in the Ising model. Q ij is the interaction parameter in QUBO. Q ij corresponds to the forcing force F in the Ising model.

[0046] Below, we will explain how to design a computational model using the example of applying QUBO. QUBO can be converted to be equivalent to the Ising model, so the application of the Ising model can be done in the same way as QUBO.

[0047] The arrangement of the deposition species 1 is the state of spin s in the element E of the QUBO (q i , q jSpecifically, the number of deposition species N seed and the number of chambers N ch The number of elements E(X i,j ) are prepared. Each element E(X i,j ) state x i,j is defined as "1" when the jth film formation species is placed in the ith chamber, and as "0" otherwise.

[0048] FIG. 3 is an example of the state when the arrangement of film formation species 1 is applied to element E. In the example of FIG. 3, there are film formation species a to f. Film formation species a to f are arranged in chambers A and B, respectively. Film formation species a, c, d, and f are arranged in the first chamber A, and film formation species b and e are arranged in the second chamber B. When this state is applied to element E, the number of elements E becomes 12, which is the product of the number of film formation species (6) and the number of chambers (2). The 1st, 3rd, 4th, and 6th in the first chamber A where the film formation species are arranged become "1", and the 2nd and 5th in the second chamber B become "1". The formula of the QUBO energy function is as follows.

[0049]

number

[0050] H t indicates a value corresponding to the time required for film formation (throughput). The QUBO interaction parameter Q (coercive force F) is defined, for example, as follows:

[0051] The time required for film formation (throughput) Ht is expressed by the following formula.

[0052]

number

[0053] N stack is the total number of steps in the deposition equipment. Ht is expressed by tracing the steps in the deposition equipment in order.

[0054] h t is expressed in the following four cases depending on the type of the i-th and i-1-th processes.

[0055] In the case where the i-th process and the i-1-th process are both depositions using deposition species that can change the chamber to be placed, h t is expressed by the following formula: "The chamber in which the film formation species is arranged can be changed" means that the arrangement location of the film formation species is not determined to be a specific chamber, and the film formation species can be arranged in another chamber.

[0056]

number

[0057] Here, t_deliver j,k is the transport time from the jth chamber to the kth chamber, t_process stacki is the processing time of the i-th step. These times can be obtained, for example, by actual measurement or simulation.

[0058] In the case of a process in the process chamber 20 in which a film is formed using a film forming species that can be changed in the chamber in which the i-th process is placed and a film is formed using a film forming species that cannot be changed in the chamber in which the i-1-th process is placed (for example, the film forming species is fixed in a certain chamber) or a process other than film formation, h t is expressed by the following formula:

[0059]

number

[0060] In the case of a process in the process chamber 20 in which the i-th process is performed with a film formation species that cannot be changed (for example, the film formation species is fixed in a certain chamber) or a process other than film formation, in the case of a film formation with a film formation species that can be changed in the chamber in which the i-1-th process is performed, h t is expressed by the following formula:

[0061]

number

[0062] In the case where the i-th process and the i-1-th process are performed in a process chamber 20 in which the chamber in which the film formation species is used or a process other than film formation is performed and the chamber in which the film formation species is used cannot be changed (for example, the film formation species is fixed in a certain chamber), h t is expressed by the following formula:

[0063]

number

[0064] Constraints may be added to the energy function according to various constraints in film formation. For example, constraints are set so that the energy of the state of the QUBO element corresponding to a pattern that does not satisfy the constraints is large. Although it is not necessary to set constraints, setting constraints increases the processing speed and accuracy of calculations.

[0065] For example, if there is only one film formation species made of a specific material and film formation species made of the same material are not allowed to be placed in multiple chambers, the total number of film formation species that can be placed in each chamber at the same time is 1. This condition is applied to each film formation species. That is, in FIG. 3, if film formation species a is placed in chamber A, film formation species a cannot be placed in chamber B. The following constraint may be added to express this constraint.

[0066]

number

[0067] When the number of chambers containing the jth film forming species is exactly 1, H A is 0. On the other hand, in other cases, H Ais a value greater than 0, which acts to make the state of high energy in the QUBO energy function. λ is a coefficient and is a value greater than 0. The absolute value of λ is adjusted depending on the value you want to optimize.

[0068] 1 has eight film formation species 1. In other words, eight or more film formation species 1 cannot be set in each film formation chamber 10.

[0069] In order to express the above constraints, we first define an additional element (Y i,k ) is added. Y i,k The state of is "1" when k deposition species are placed in the i-th chamber, and "0" otherwise. i,k ) is added. i,k This state indicates that four deposition species are placed in the first chamber A and two deposition species are placed in the second chamber B. The restriction on the placement of deposition species 1 due to the upper limit on the number of slots in each chamber can be expressed as the following constraint.

[0070]

number

[0071] N slot,i represents the number of slots in the i-th chamber. The first term on the right side is an additional element (Y i,k ) is 0 when the sum of the states is exactly 1, and is greater than 0 otherwise. The second term on the right hand side is the element E(X i,j ) and additional element (Y i,k ) state consistency, and X i,j The sum of the states of Y i,k The value is 0 if the product of the state and the number of slots k is the same, and is greater than 0 otherwise.

[0072] w jis the number of slots used for the j-th film formation species. When only one slot is used for film formation, the w of the film formation species in that film formation is set to 1. When a film formation species uses multiple slots, for example, when multiple types of materials are simultaneously sputtered to form a predetermined film.

[0073] When forming a layer using multiple film-forming species simultaneously, the film-forming species must be placed in the same film-forming chamber 10. Therefore, placing film-forming species that may be simultaneously formed on a wafer in the same film-forming chamber is also a constraint. In this case, multiple film-forming species are regarded as one film-forming species, and the value of w is the number of film-forming species. The simultaneous film-forming is, for example, simultaneous sputtering.

[0074] If you want to simultaneously improve the throughput of wafers with different film formation processes, you need to set the above H t Design and add them together to create a new H t For example, if you want to simultaneously improve the throughput of wafer 1 and wafer 2, which have different film formation processes, you can use the following formula: t may also be used.

[0075]

number

[0076] H t1 , H t2 are Ht for wafer 1 and wafer 2, respectively. w t1 , w t2 is a weight. It is possible to adjust the weight of a wafer for which it is desired to give priority to improving the throughput so that the weight is large.

[0077] For example, the above two constraints H A , H B The energy function when is added is expressed as follows:

[0078]

number

[0079] Furthermore, when the film forming species 1 is replaced at the end of its life, the film forming apparatus 100 is stopped during that period. The replacement of the film forming species 1 takes time. Reducing the number of times the film forming species 1 is replaced increases the production efficiency of the film forming apparatus 100. A constraint term may be set so that the times to reach the end of their life are approximately constant. "Approximately constant" means that the times to reach the end of their life are within a 10% fluctuation range based on the time to reach the end of the life of a certain film forming species 1.

[0080] For example, the wafer transport order may be added as a constraint of the Ising model or QUBO. When reflecting the wafer transport order in the calculation model, an additional element (Z m、t ) is added to the additional element Z m、t is represented by process m in the film formation equipment of the wafer and time t at which the above process is executed. The process of the wafer in the film formation equipment is the flow of film formation and processing from when the wafer is put into the load lock chamber LL until it is finally transported to the load lock chamber. When process m is executed at time t, the additional element Z m、t State of z m、t is set to "1", and otherwise is set to "0".

[0081] In addition to the order of wafer transport in a process of a film-forming device, the invention may also be extended to the order of wafer transport in processes before and after a film-forming process. For example, the invention may also be extended to the order of wafer transport in a process of a single-sheet manufacturing device such as an etching device or a resist coater.

[0082] Also, for example, a constraint of the Ising model or QUBO may be added to the fact that a new wafer is inserted into the deposition chamber midway. When this is reflected in the calculation model, an additional element (A t、w ) is added. t、w is expressed by the timing t when the wafer is inserted into the deposition chamber and the wafer insertion position w. When the wafer w is inserted at time t, the element A t、w State a t、wis set to 1, and otherwise to 0.

[0083] In the calculation area 201, an annealing calculation is performed based on the designed calculation model. The annealing calculation is a calculation method that fluctuates the states x and y of the spin s of each element E to find the optimal state. The annealing calculation predicts the data (x) of the arrangement of the film-forming species 1 that makes H smaller and the time (throughput) required for film formation at that time.

[0084] The annealing calculation can be performed using a calculation method such as simulated annealing, evolutionary algorithm, tabu search, etc.

[0085] Annealing calculations can be performed on a dedicated machine (annealing machine) that specializes in Ising model and QUBO calculations. For example, machines such as quantum annealing machines (D-wave, NEC), coherent Ising machines (NTT), simulated bifurcation machines (Toshiba), digital annealers (Fujitsu), and CMOS annealers (Hitachi) can be used.

[0086] A quantum gate type computer may be used for the annealing calculation. For example, if the QAQA (Quantum Approximate Optimization Algorithm) is used, the Ising model and QUBO can be calculated using a quantum gate type computer.

[0087] 5 is a schematic diagram of a processing element 50 that performs the annealing calculation in a calculation domain 201. The processing element 50 includes a calculation unit 51, a communication unit 52, and a control unit 53.

[0088] The calculation unit 51 is a part that performs the above-mentioned annealing calculation. The calculation unit 51 includes a quantum computer. The quantum computer is a general term that includes the annealing calculation and quantum gate type computers. The quantum computer includes, for example, optical elements, superconducting elements, etc. The calculation of the quantum computer is performed, for example, by optical elements, superconducting elements, etc.

[0089] Calculations using optical elements can be performed at room temperature. However, optical fibers over a kilometer in length must be operated at high speed, and there are technical challenges such as miniaturization, high-speed calculations, and the production of squeezed light. Calculations using superconducting elements require ultra-low temperature technology of around 20 millikelvin to reduce noise.

[0090] The communication unit 52 is responsible for communication between the calculation unit 51 and the control unit 53. Signals in the communication unit 52 are transmitted using, for example, metal wiring or optical wiring.

[0091] When the calculation unit 51 includes a superconducting element, it is preferable to use optical wiring to transmit signals in the communication unit 52. This is because if metal wiring is used, it is difficult to maintain the ultra-low temperature of the calculation unit 51 due to heat conduction between the calculation unit 51, which is in an ultra-low temperature environment, and the control unit 53, which is in a room temperature environment.

[0092] When optical wiring is used for transmitting signals in the communication unit 52, a photoelectric conversion element for converting an optical signal into an electric signal is provided near the control unit 53. The photoelectric conversion element is, for example, a spin photodetector.

[0093] 6 is a schematic diagram of a spin photodetector. The spin photodetector 60 has a magnetization free layer 61, a magnetization fixed layer 62, and a nonmagnetic layer 63 sandwiched between them. The magnetization free layer 61 and the magnetization free layer 62 include a ferromagnetic material. The magnetization M61 of the magnetization free layer 61 is easier to reverse than the magnetization M62 of the magnetization fixed layer 62. When the intensity of the light L irradiated to the magnetization free layer M61 changes, the magnetization of the magnetization free layer 61 tilts. The light L irradiated to the magnetization free layer M61 is, for example, circularly polarized light.

[0094] When the relative angle between the magnetization M61 of the magnetization free layer 61 and the magnetization M62 of the magnetization fixed layer 62 changes, the resistance value in the stacking direction of the spin photodetector 60 changes. That is, the spin photodetector 60 can convert the change in intensity of the light L into an electrical signal, which is the resistance value of the spin photodetector 60. For example, when the intensity of the light L is changed at high speed to change the magnetization direction of the magnetization free layer 61 at high speed, the spin photodetector 60 can generate a high-frequency signal.

[0095] The magnetization free layer 61 is, for example, a ferrimagnetic material. A ferrimagnetic material is a magnetic material that exhibits ferromagnetism as a whole, but has an antiferromagnetic microscopic magnetic structure. Since there is a difference in magnetic moment between the magnetization oriented in a first direction in the microscopic magnetic structure and the magnetization oriented in a second direction opposite to the first direction, the material exhibits ferromagnetic properties as a whole. When the ferrimagnetic material is irradiated with sufficient light energy from the outside, the combination of the antiferromagnetic magnetic structure is reversed, and the ferromagnetic properties exhibited as a whole are reversed.

[0096] A ferrimagnetic material is, for example, an alloy of an element with 4f electrons in its outermost shell and an element with 3d, 4d, and 5d electrons in its outermost shell. An example of a ferrimagnetic material is GdCoFe. A ferrimagnetic material may be an oxide such as a ferrite material. A ferrite material exhibits ferrimagnetism when it is combined with elements with different magnetic moments as 3d elements.

[0097] The magnetization free layer 61 may have a laminated structure of a ferromagnetic layer and a ferrimagnetic layer made of a ferrimagnetic material. The ferromagnetic layer may be one generally used in magnetoresistance effect elements. The ferromagnetic layer may be, for example, a single metal such as Co, Fe, or Ni, an alloy containing these, or a Heusler alloy. The magnetization state of the ferrimagnetic layer changes when light is irradiated. The magnetization of the ferromagnetic layer changes with the change in the magnetization state of the ferrimagnetic layer. The ferromagnetic layer is mainly responsible for the resistance change associated with the magnetoresistance effect of the spin photodetector 60. The ferrimagnetic layer is mainly responsible for the magnetization reversal of the spin photodetector 60 by light. By dividing the functions of the ferromagnetic layer and the ferrimagnetic layer, the high-speed operation and high output of the spin photodetector 60 can be achieved at the same time. It is preferable that the ferromagnetic layer is located closer to the nonmagnetic layer 63 than the ferrimagnetic layer.

[0098] A spin conduction layer may be inserted between the magnetization free layer and the ferrimagnetic layer. The spin conduction layer transmits the spin current generated when the ferrimagnetic material is reversed to the ferromagnetic layer. As a result, magnetization reversal of the ferromagnetic layer is induced. When the spin photodetector 60 has this configuration, magnetization reversal on the order of femtoseconds is possible. The spin conduction layer is made of a material with high spin conduction properties. For example, metals such as copper and silver, semiconductors such as silicon, germanium, and gallium arsenide, and two-dimensional materials such as graphene can be used as the spin conduction layer.

[0099] The magnetization free layer 61 is, for example, a ferromagnetic material. In general, a ferromagnetic material is used for the magnetization free layer 61 in a magnetoresistance effect element. However, in order to function as the spin photodetector 60, the incident optical signal must be polarized. If the optical signal is not polarized, the optical signal has the effect of heating the ferromagnetic material. In this case, the magnetization direction of the ferromagnetic material cannot be arbitrarily oriented by the optical signal. If the optical signal is polarized, the magnetization direction of the ferromagnetic material can be arbitrarily oriented by the polarization of the optical signal. If the magnetization free layer 61 is a ferrimagnetic material, the heating and cooling processes are required. If the magnetization free layer 61 is a ferromagnetic material, the heating and cooling processes are not required, so the optical signal can be converted into an electric signal faster than if it is a ferrimagnetic material.

[0100] The communication unit 52 transmits the calculation result of the calculation unit 51 to the control unit 53 as an optical signal, and the spin photodetector converts the optical signal into an electrical signal.

[0101] The control unit 53 controls the calculation unit 51. The control unit 53 transmits signals from the outside to the calculation unit 51 and transmits signals from the calculation unit 51 to the outside. The control unit 53 includes, for example, a high-frequency measuring device, an oscillator, etc. The control unit 53 transmits the calculation results in the calculation unit 51 to the film forming apparatus 100.

[0102] The throughput can be improved by actually replacing the arrangement of the film formation species 1 in the film formation apparatus 100 based on the data (x) of the arrangement of the film formation species 1 predicted by the annealing calculation.

[0103] "Second embodiment" The film formation system according to the second embodiment is different from the first embodiment in that the interaction parameter Q is calculated using machine learning. 7 is a schematic diagram of a film formation system 301 according to the second embodiment. The film formation system 301 includes a film formation apparatus 100 and a computer 210.

[0104] The computer 210 has a memory area 202 , a learning area 203 , and a calculation area 201 .

[0105] The memory area 202 stores, as training data, the time required for film formation when the film formation species 1 are arranged in a specific arrangement in the film formation apparatus 100. The arrangement of the film formation species 1 refers to which film formation chamber the film formation species 1 is arranged in and of which material. The time required for film formation (throughput) when the film formation species 1 are arranged in a specific arrangement is based on, for example, actual measurement. The throughput for each arrangement of the film formation species 1 is obtained by experiment and is acquired in the memory area 202 as training data. The greater the amount of training data, the higher the accuracy of the calculation model described below.

[0106] In the learning area 203, interaction parameters between elements are determined by machine learning based on teacher data, and a computation model is designed. Below, a method for designing a computation model will be described using an example in which QUBO is applied to the computation model.

[0107] 8 is a flow diagram of a method for designing a computational model. First, the learning area 203 acquires training data from the memory area 202 (first step S1). The training data can be acquired by actual measurement or simulation. In the case of actual measurement, a film formation process is performed in a pattern in which a plurality of film formation species 1 are arranged, and the experimental results of the throughput at that time are used as the training data. In the case of simulation, a film formation process is simulated in a pattern in which a plurality of film formation species 1 are arranged, and the calculated results of the throughput at that time are used as the training data.

[0108] The arrangement pattern of the film-forming species 1 in the training data may be selected randomly, or in order to improve the accuracy of machine learning, several arrangement patterns that are expected to have significantly different throughputs may be provided, or the respective patterns may be combined.

[0109] Next, the training data is applied to the QUBO (second step S2). The arrangement of the film-forming species 1 in the training data is determined by the state of spin s in the element E of the QUBO (q i , q j )

[0110] Next, the learning area 203 applies multiple training data to QUBO and performs machine learning (third step S3). If the number of training data is small, machine learning may be performed using a factorization machine. The factorization machine is a type of machine learning model expressed by the following formula.

[0111]

number

[0112] w and v are fitting parameters obtained by machine learning, and correspond to the QUBO interaction parameter (Q). When K is increased, the number of fitting parameters changes, and is adjusted appropriately according to the complexity of the problem.

[0113] After machine learning, an interaction parameter (Q) is determined based on the training data (fourth step S4).

[0114] Once the interaction parameter (Q) is determined, the calculation model is determined. Here, constraints are added to the calculation model as necessary. The constraints are, for example, those presented in the first embodiment. This calculation model is sent to the calculation domain 201.

[0115] The calculation area 201 predicts the time required for film formation when a new arrangement of the film formation species 1 is set, based on a designed calculation model.

[0116] First, an annealing calculation is performed on the calculation region 201 from the designed interaction parameter (Q) (fifth step S5). Data (x) of the arrangement of the film-forming species 1 that makes H smaller by the annealing calculation is output.

[0117] Next, the throughput for the arrangement of the film-forming species 1 found in the calculation area 201 is obtained (sixth step S6). The throughput can be obtained by actual measurement or simulation. For example, in the case of actual measurement, the throughput is found by changing the arrangement of the film-forming species 1 found in the calculation area 201 and conducting a film-forming experiment. In the case of simulation, the throughput is calculated by simulating the film-forming process in the arrangement of the film-forming species 1 found in the calculation area 201.

[0118] Next, it is determined whether the acquired throughput has reached the target value (seventh step S7). If the throughput has reached or reached the target value, the computer 200 outputs that value. The arrangement of the film-forming species 1 used in the calculation can sufficiently shorten the time required for film formation. On the other hand, if the throughput has not reached or reached the target value, the result is sent to the memory area 202 and is reacquired as training data. Then, the above process is repeated until the throughput reaches or reaches the target value.

[0119] Then, once the arrangement of the film-forming species 1 in the film-forming apparatus 100 has been determined, film formation on the wafer is performed. The multilayer film to be formed is, for example, a magnetoresistance effect element. The magnetoresistance effect element has at least two ferromagnetic layers and a nonmagnetic layer sandwiched between them. In many cases, each of the ferromagnetic layers has a multi-layer structure. The nonmagnetic layer may be a conductor, a semiconductor, or an insulator, for example, an oxide film or a nitride film.

[0120] The above describes the embodiments of the present invention in detail with reference to the drawings. However, each configuration and combination thereof in each embodiment is merely an example, and addition, omission, substitution, and other modifications of the configuration are possible without departing from the spirit of the present invention.

[0121] "Third embodiment" The factory system according to the third embodiment is an expansion of the film forming system according to the first embodiment to a factory system for the entire factory. Fig. 9 is a schematic diagram of the factory system according to the third embodiment. The factory system 1000 has a computer 200 and a plurality of devices 500. Within the factory, there are a plurality of installation areas 501 in which the devices 500 can be installed.

[0122] If each of the tools 500 is allocated to an available area 501 for installation in order to reduce the transfer time between the tools 500 and improve the throughput, the number of combinations will be enormous as the number of tools 500 and the number of available areas 501 for installation of the tools increase. Therefore, by associating the film-forming species 1 in the first embodiment with the tool 500 and the chamber 10 with the available area 501 for installation of the tool, the computer 200 can calculate the optimal arrangement of each of the tools 500 in the available area 501 for installation of the tool using a calculation flow similar to that of the first embodiment, thereby improving the throughput. The number of slots in the chamber 10 may be set to correspond to the number of tools 500 that can be installed in the available area 501 for installation of each tool.

[0123] According to the factory system of this embodiment, the throughput of the entire factory can be improved.

[0124] Example 1 In Example 1, the film formation system 300 shown in FIG. 1 was used to determine the time required for film formation when the arrangement of film formation species was set.

[0125] The film formation apparatus of Example 1 has three chambers ch-A, ch-B, and ch-C and a transfer chamber. Wafers can be transported between the three chambers via the transfer chamber. Each of ch-A and ch-B is equipped with eight slots in which any film formation species can be placed. Ch-C is a chamber for performing processes other than film formation, and film formation species cannot be placed therein.

[0126] There are six types of film formation. Film formation types a, b, c, d, and e use one slot. Film formation type f uses four slots. In other words, film formation with film formation type f is performed by simultaneous sputtering from four slots.

[0127] [Table 1]

[0128] After being loaded into the load lock chamber LL (see Figures 1 and 11), the wafer undergoes the steps shown in the time series chart in Figure 10 below from left to right, and is finally transported to the load lock chamber LL. The above steps constitute the entire process in the film formation apparatus. Each block in Figure 10 represents a process, and among the blocks, those with an alphabet written inside are film formation processes, and the alphabet indicates the type of film used in film formation. The film formation process is carried out in ch-A or ch-B. On the other hand, block 1001 is a process other than film formation, and is carried out in ch-C. The processing time for each process was measured and determined to be the width of each block in Figure 10.

[0129] The transfer times between each chamber or between each chamber and the load lock chamber LL were measured and were given in the table below (unit: seconds).

[0130] [Table 2]

[0131] Next, the optimal arrangement was obtained according to the calculation flow shown in the first embodiment. A and H B was added to the energy function, and λ = 2000. The annealer calculation was performed by simulated annealing.

[0132] The Annealer calculation output the following table layout. The Annealer calculation took 26 seconds.

[0133] [Table 3]

[0134] The throughput with the above-mentioned arrangement of the film forming species was obtained by simulation. As a result, the total process time was 59 minutes and 24 seconds. Figure 11 is a time series chart (Gantt chart) of Example 1 divided by chamber.

[0135] Comparative Example 1 In Comparative Example 1, the film formation species were randomly arranged. Film formation species a, b, c, d, e, and f were randomly arranged in ch-A and ch-B as shown in the table below so as not to exceed the upper limit of the number of slots in each chamber.

[0136] [Table 4]

[0137] The throughput in the above arrangement was calculated by simulation to be 1 hour, 8 minutes, and 3 seconds. Fig. 12 is a time series chart (Gantt chart) of Comparative Example 1 divided by chamber.

[0138] The throughput of Example 1 was improved compared to Comparative Example 1. This is believed to be because Example 1 shown in FIG. 11 has fewer transfers and transfer times than Comparative Example 1 shown in FIG.

[0139] Example 2 Example 2 is an example in which the number of chambers and the number of film-forming species are increased compared to Example 1, resulting in an enormous number of combinations. The film-forming apparatus of Example 2 has six chambers, ch-A, ch-B, ch-C, ch-D, ch-E, and ch-F, and a transfer chamber. Wafers can be transported through the transfer chamber. The number of slots for each chamber is shown in the table below.

[0140] [Table 5]

[0141] Each of ch-A, ch-B, ch-C, and ch-D has five slots in which any film formation species can be placed.Ch-E and ch-F are chambers for processes other than film formation, and film formation species cannot be placed in them.

[0142] From actual measurements, the transfer time between each chamber was uniformly 50 seconds, and the transfer time between the load lock chamber LL and each chamber was 44 seconds.

[0143] There are 13 types of film formation: a, b, c, d, e, f, g, h, i, j, k, l, and m, and the number of slots used is all 1.

[0144] After being loaded into the load lock chamber LL, the wafer undergoes the steps shown in the time series chart in FIG. 13 from left to right, and is finally transported to the load lock chamber. The entire process in the film formation device is the steps described above. Each block in FIG. 13 represents a process. Among the blocks, those with an alphabet inside are film formation processes, and the alphabet indicates the film type used in film formation. The film formation process is performed in ch-A, ch-B, ch-C, or ch-D. Meanwhile, block 2001 and block 2002 are processes other than film formation, with block 2001 being performed in ch-E and block 2002 being performed in ch-F. The processing time of each process was measured and was determined as the width of each block in FIG. 13.

[0145] The optimal arrangement was obtained according to the calculation flow of the first embodiment. A and H B was added to the energy function, and λ = 2000. The annealer calculation was performed by simulated annealing.

[0146] The Annealer calculation output the following table layout. The Annealer calculation took 54 seconds.

[0147] [Table 6]

[0148] The throughput in the above arrangement was calculated by simulation to be 1 hour 19 minutes 11 seconds. Fig. 14 is a time series chart (Gantt chart) of Example 2 divided by chamber.

[0149] Comparative Example 2 In Comparative Example 2, the deposition species were randomly arranged. The deposition species a, b, c, d, e, f, g, h, i, j, k, l, and m were randomly arranged in ch-A, ch-B, ch-C, and ch-D as shown in Table 7 below, so as not to exceed the upper limit of the number of slots in each chamber.

[0150] [Table 7]

[0151] The throughput in the above arrangement was calculated by simulation to be 1 hour 31 minutes 41 seconds. Fig. 15 is a time series chart (Gantt chart) of Comparative Example 2 divided by chamber.

[0152] The throughput of Example 2 was improved compared to Comparative Example 2. This is believed to be because the number of conveyances in Example 2 shown in FIG. 14 was smaller than that in Comparative Example 2 shown in FIG.

[0153] Example 3 Example 3 is an example in which the number of chambers is increased even more than in Example 2, resulting in an enormous number of combinations. The deposition apparatus of Example 3 has eight chambers, ch-A, ch-B, ch-C, ch-D, ch-E, ch-F, ch-G, and ch-H, and a transfer chamber. Wafers can be transported through the transfer chamber. The number of slots for each chamber is shown in the table below.

[0154] [Table 8]

[0155] Each of ch-A, ch-B, ch-C, ch-D, ch-E, ch-F, and ch-G has five slots in which any film formation species can be placed. Ch-H is a chamber for processes other than film formation, and film formation species cannot be placed in it.

[0156] From actual measurements, the transfer time between each chamber was uniformly 50 seconds, and the transfer time between the load lock chamber LL and each chamber was 44 seconds.

[0157] There are 35 types of film formation species, and the number of slots used is 1 in each case.

[0158] After being loaded into the load lock chamber LL, the wafer undergoes the steps shown in the time series chart in FIG. 16 from left to right, and is finally transported to the load lock chamber. The entire process in the film formation apparatus is the steps described above. Each block in FIG. 16 represents a process. Among the blocks, those with numbers written inside are film formation processes, and the numbers indicate the film formation type used in film formation. Film formation processes are performed in ch-A, ch-B, ch-C, ch-D, ch-E, ch-F, or ch-G. Meanwhile, block 3001 is a process other than film formation, and is performed in ch-H. The processing time of each process was measured and was determined to be the width of each block in FIG. 16.

[0159] The optimal arrangement was obtained according to the calculation flow of the first embodiment. A and H B was added to the energy function, and λ = 4000. The annealing calculation was performed using a simulated bifurcation machine.

[0160] The Annealer calculation output the following layout in the table. The Annealer calculation took 231 seconds.

[0161] [Table 9]

[0162] The throughput in the above arrangement was calculated by simulation to be 3 hours 11 minutes 38 seconds. Fig. 17 is a time series chart (Gantt chart) of Example 2 divided by chamber.

[0163] Comparative Example 3 In Comparative Example 3, the film formation species were randomly arranged. Thirty-five types of film formation species were randomly arranged in ch-A, ch-B, ch-C, ch-D, ch-E, ch-F, and ch-G as shown in Table 10 below, without exceeding the upper limit of the number of slots in each chamber.

[0164] [Table 10]

[0165] The throughput in the above arrangement was calculated by simulation to be 3 hours 27 minutes 28 seconds. Fig. 18 is a time series chart (Gantt chart) of Comparative Example 2 divided by chamber.

[0166] The throughput of Example 3 was improved compared to Comparative Example 3. This is believed to be because the number of conveyances in Example 3 shown in FIG. 17 was smaller than that in Comparative Example 2 shown in FIG.

[0167] The arrangement of all the film-forming species in Example 3 and Comparative Example 3 is approximately 2.88×10 25 There are 100 combinations, and if one were to try to simulate the throughput for all the arrangements of the film-forming species 1 by exhaustive simulation, it would take more than 1000 days even using a supercomputer with a computing performance of 200 petaflops, so it is not realistic to find an arrangement that improves throughput by exhaustive simulation. On the other hand, in this embodiment, the annealer calculation can be performed in 231 seconds, so it is possible to find an arrangement that improves throughput in a short time.

[0168] Example 4 In Example 4, the throughput was determined when two wafers were simultaneously loaded into a film formation apparatus and films were formed in parallel.

[0169] The film formation equipment and the wafer process are the same as those in Example 2. There are wafers 1 and 2 which are subjected to the same process, and the arrangement of the film formation species in the film formation equipment and the timing of inputting each wafer into the film formation equipment are optimized according to the calculation flow (see FIG. 8) shown in the second embodiment.

[0170] The aim was to reduce the time from when the preceding wafers start to be loaded until processing of the two wafers is completed, with the target value being set at 1 hour and 30 minutes.

[0171] First, 20 pieces of training data were collected. The deposition species placement and wafer loading timing were randomly selected, and throughput data was collected based on actual measurements.

[0172] Next, machine learning was performed using the collected training data with a factorization machine. The factorization machine's K was set to 8. The factorization machine's K is a parameter that is adjusted depending on the complexity of the problem. w and v were obtained by machine learning, and were converted into the QUBO interaction parameter (Q), and the computational model was determined.

[0173] Next, the calculation model was input to an annealing machine to perform annealing calculations. A simulated bifurcation machine was used as the annealing machine. The annealing calculations output data (x) for the arrangement of the film-forming species 1.

[0174] Next, based on the data (x) on the arrangement of the film-forming species 1, the arrangement of the film-forming apparatus and the timing of wafer input were actually set, and the throughput was obtained by actual measurement.

[0175] The throughput obtained by actual measurement was added as training data, and the above process was repeated.

[0176] Figure 19 shows the throughput obtained by repeating the above process. By repeating the above process, the time was gradually reduced, and after the 10th iteration, the throughput was improved to within the target value of 1 hour and 30 minutes. [Explanation of symbols]

[0177] 1 film formation species, 10 film formation chamber, 20 process chamber, 30 transfer chamber, 40 load lock chamber, 100 film formation device, 200 computer, 201 storage area, 202 learning area, 203 calculation area, 300 film formation system, 500 device, 1000 factory system, E element, F forcing force, s spin, 1001, 2001, 2002, 3001 process other than film formation

Claims

1. The method includes: The film forming apparatus has a film forming chamber in which a plurality of film forming species can be placed, The computer has a calculation domain for predicting a time required for film formation when an arrangement of film formation species is set based on a calculation model including an Ising model or a QUBO model, A film formation system, wherein the Ising model or the QUBO has a number of elements corresponding to the product of the number of film formation species and the number of film formation chambers, and each of the elements is a binary variable.

2. The film forming system according to claim 1 , wherein the energy function of the Ising model or the QUBO includes an energy function corresponding to a throughput of the film forming apparatus.

3. 3. The film forming system according to claim 1, wherein the Ising model or the QUBO uses values ​​corresponding to a transport time of a wafer on which a film is formed and a processing time in each process as parameters for determining an interaction parameter between the elements.

4. The film forming system according to any one of claims 1 to 3, wherein the Ising model or the QUBO is defined under the condition that a process in any step can be replaced with a process in another chamber.

5. The film forming apparatus further includes a process chamber for performing a process other than film formation, The film forming system according to claim 4 , wherein a process in the process chamber is defined as a step that cannot be replaced with a process in another chamber in the Ising model or the QUBO.

6. The Ising model or the QUBO has constraints, The deposition system according to claim 1 , wherein the constraint reflects an upper limit on the number of deposition species of the same type among the deposition species.

7. The Ising model or the QUBO has constraints, The deposition system of claim 1 , wherein the constraint reflects an upper limit on the number of slots for the deposition species in the deposition chamber.

8. The Ising model or the QUBO further includes, as the constraint, additional elements whose number corresponds to the sum of the number of slots of the deposition chambers; The deposition system of claim 7 , wherein the additive factor is a binary variable.

9. The Ising model or the QUBO has constraints, The deposition system of any one of claims 1 to 8, wherein the constraint reflects a condition that deposition species that may be simultaneously deposited on a wafer are located in the same deposition chamber.

10. The Ising model or the QUBO has constraints, 10. The film forming system according to claim 1, wherein the constraint reflects a condition that the times to reach the end of the lifespan of the plurality of film forming species are substantially constant.

11. The Ising model or the QUBO has constraints, The film forming system according to any one of claims 1 to 10, wherein the Ising model or the QUBO further has an additional element representing a wafer process and a processing time of the process as the constraint, and the additional element is a binary variable.

12. The Ising model or the QUBO has constraints, The film forming system according to any one of claims 1 to 11, wherein the Ising model or the QUBO further has an additional element representing a loading time of the wafer as the constraint, and the additional element is a binary variable.

13. the computer includes a storage area for storing, as training data, a specific arrangement of film-forming species and a time required for film formation in the case of the specific arrangement of film-forming species; A learning area in which interaction parameters between elements of the Ising model or the QUBO are determined by machine learning based on the teacher data, and a computational model is designed; The deposition system according to any one of claims 1 to 12, further comprising:

14. The film forming system according to claim 13 , wherein the storage area reacquires, as teacher data, the arrangement of film forming species used for prediction in the calculation area and the time required for film formation predicted in the calculation area.

15. The deposition system of claim 13 or 14, wherein the machine learning is performed by a factorization machine.

16. A step of designing a calculation model including an Ising model or a QUBO model using the arrangement of a plurality of film formation species in a film formation apparatus as an element; predicting a time required for film formation for a given arrangement of film formation species based on the calculation model; and applying the film formation species arrangement as a film formation species arrangement of the film formation apparatus when the predicted time is equal to or shorter than a predetermined time. A method for forming a film on a wafer, wherein the Ising model or the QUBO has a number of elements corresponding to the product of the number of film formation species and the number of film formation chambers in the film formation apparatus, and each of the elements is a binary variable.

17. The step of designing the computational model comprises: acquiring, as teacher data, a specific arrangement of film-forming species and a time required for film formation in the case of the specific arrangement of film-forming species; The method for forming a film on a wafer according to claim 16 , further comprising a step of determining an interaction parameter between elements by machine learning based on the training data.

18. 18. The method of forming a film on a wafer according to claim 17, further comprising the step of reacquiring, as teacher data, the arrangement of the film forming species and the time required for film formation when the time predicted based on the calculation model is equal to or longer than a predetermined time.

Images