Machine learning based modeling method and system for magnetic response of nanomolecular devices

Abstract

The application discloses a machine learning-based nanomolecular device magnetic response modeling method and system, relates to the technical field of electromagnetism, and comprises an electromagnetic response testing module, an electromagnetic response modeling module and a parameter adjusting module.The electromagnetic response testing module is used for detecting the current and voltage oscillation frequency between the reference magnetic layer and the free magnetic layer of a nanometer oscillator and detecting the external magnetic field strength of the nanometer oscillator.The electromagnetic response modeling module is used for predicting the precession frequency shift state of the free magnetic layer under the current adjustment and predicting the voltage oscillation frequency change of the nanometer oscillator when the external magnetic field strength changes.The parameter adjusting module is used for adjusting the direct current spin polarization current and the external magnetic field strength.The electromagnetic response testing module comprises a conductive electrode, a current detection unit, a voltage detection unit and a spectrum analysis module, and the application has the characteristics of efficient adjustment.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of magnetic response detection, in particular to a magnetic response modeling method and system for nanomolecular devices based on machine learning. BACKGROUND

[0002] A spin-torque nano-oscillator is a super-small magnetic oscillation source for converting a direct current spin polarization current into an adjustable microwave signal. The electron spin angular momentum (magnetic moment) generated by the direct current spin polarization current is transmitted to a nanometer-sized small magnetic needle in a free magnetic layer in the oscillator, so that the small magnetic needle moves at a high speed in an external magnetic field. The frequency of the adjustable microwave signal determines the communication wave band, and the power determines the transmission distance. The nano-oscillator is often used in the form of an array as a microwave communication source.

[0003] Before the nano-oscillator is put into use, the frequency and the power need to be adjusted separately to match the use scene. At this time, an external magnetic field is applied to the nano-oscillator. The precession frequency of the free magnetic layer can be changed by adjusting the strength of the external magnetic field, and then the frequency of the microwave signal is changed. The spin torque can be adjusted by adjusting the direct current spin polarization current, and then the resistance change amplitude caused by the magnetic vibration is changed, so as to change the power.

[0004] However, when the current is adjusted, the change of the magnetic moment will also change the total magnetic field acting on the magnetic needle, thereby causing a frequency shift. At each adjustment, the strength of the external magnetic field needs to be adjusted immediately after the adjustment to fix the precession frequency, so as to maintain the controllability of the power adjustment process. Due to the difference in the microstructure of each nano-oscillator, the power needs to be adjusted for several times to reach a suitable power, resulting in a long adjustment process, especially when the parameters of a batch of nano-oscillators are calibrated. Therefore, it is necessary to design an efficient nano-molecular device magnetic response modeling method and system based on machine learning. SUMMARY

[0005] The application aims to provide a nano-molecular device magnetic response modeling method and system based on machine learning to solve the problems in the background.

[0006] In order to solve the above technical problems, the application provides the following technical scheme: a nano-molecular device magnetic response modeling method and system based on machine learning, comprising an electromagnetic response test module, an electromagnetic response modeling module and a parameter adjustment module. The electromagnetic response test module is used for detecting the current and voltage oscillation frequency between the reference magnetic layer and the free magnetic layer of the nano-oscillator, and detecting the strength of the external magnetic field of the nano-oscillator. The electromagnetic response modeling module is used for predicting the precession frequency shift state of the free magnetic layer under the current adjustment, and predicting the voltage oscillation frequency change of the nano-oscillator when the strength of the external magnetic field changes. The parameter adjustment module is used for adjusting the direct current spin polarization current and the strength of the external magnetic field.

[0007] According to the technical solution, the electromagnetic response test module includes a conductive electrode, a current detection unit, a voltage detection unit, a spectrum analysis module, a micro-Hall probe, and a magnetic field strength calculation module. The conductive electrode is electrically connected with the current detection unit and the voltage detection unit. The spectrum analysis module is electrically connected with the voltage detection unit. The micro-Hall probe is electrically connected with the magnetic field strength calculation module. The conductive electrode is used to lead out signals between the reference magnetic layer and the free magnetic layer, to realize the access and detection of the current and the voltage. The current detection unit is used to monitor the direct current spin polarization current size injected into the nano oscillator in real time. The voltage detection unit is used to collect the output voltage signal of the nano oscillator, to reflect the voltage oscillation caused by the magnetic resistance change. The spectrum analysis module is used to analyze the voltage signal in the time domain, to calculate the main frequency of the high-frequency oscillation, i.e., the precession frequency of the free magnetic layer, and to measure the power intensity of the main frequency point to obtain the power. The micro-Hall probe is used to detect the local external magnetic field strength around the nano oscillator. The magnetic field strength calculation module is used to convert the Hall voltage signal output by the micro-Hall probe into the actual external magnetic field strength value.

[0008] The electromagnetic response modeling module includes a model marking module, a data acquisition module, a precession frequency prediction module, a magnetic response prediction module, and a machine learning module. The model marking module is electrically connected with the data acquisition module. The data acquisition module is electrically connected with the spectrum analysis module and the magnetic field strength calculation module. The precession frequency prediction module and the magnetic response prediction module are electrically connected with the data acquisition module and the current detection unit. The machine learning module is electrically connected with the precession frequency prediction module and the magnetic response prediction module. The model marking module is used to find and mark the model of each nano oscillator. The data acquisition module is used to collect voltage, current, frequency, and magnetic field strength data. The precession frequency prediction module is used to predict the precession frequency change caused by the current change of the nano oscillator of the current model. The magnetic response prediction module is used to predict the voltage oscillation frequency change when the external magnetic field strength changes for the nano oscillator of the current model. The machine learning module is used to correct the prediction result of the precession frequency according to the actual precession frequency when the external magnetic field strength changes.

[0009] The parameter adjustment module includes a current control module, a magnetic field strength control module, a total power calculation module, and a power difference calculation module. The current control module is electrically connected with the precession frequency prediction module. The magnetic field strength control module is electrically connected with the magnetic response prediction module. The total power calculation module is electrically connected with the spectrum analysis module and the power difference calculation module. The current control module is used to control the direct current spin polarization current. The magnetic field strength control module is used to control the external magnetic field strength. The total power calculation module is used to calculate the total power of the nano oscillator array. The power difference calculation module is used to calculate the power that needs to be adjusted.

[0010] According to the technical solution, the working method of the system is:

[0011] S0, when first detected, each nanometer oscillator model is marked, so that each group of conductive electrodes connects the reference magnetic layer and the free magnetic layer of each nanometer oscillator;

[0012] S1, the direct current spin polarization current size in each nanometer oscillator is randomly adjusted, the time domain analysis is performed on the collected voltage signal, the main frequency of high frequency oscillation, that is, the precession frequency of the free magnetic layer is calculated, and the precession frequency is directly adjusted by adjusting the external magnetic field strength around each nanometer oscillator, so that the shift of the precession frequency is repaired after each current adjustment;

[0013] S2, the relationship between the precession frequency change amount and the current change amount when the current is adjusted each time, and the relationship between the precession frequency change amount and the magnetic field strength change amount when the external magnetic field strength is adjusted each time are recorded, the function relationship between the precession frequency change amount and the current change amount is fitted after batch data is obtained, the function relationship between the precession frequency change amount and the magnetic field strength change amount is fitted, and then the change of the precession frequency is predicted before the current and the magnetic field strength are adjusted;

[0014] S3, the machine learning model is used in combination with the two fitted function relationships to directly calculate the magnetic field strength to be adjusted before the current is adjusted, the magnetic field strength is directly adjusted according to the calculated magnetic field strength during adjustment, and the fitted function relationship is modified according to the actual precession frequency after the magnetic field strength is adjusted;

[0015] S4, when a new nanometer oscillator matrix needs to be adjusted in frequency and power, the frequency consistency is adjusted according to the fitting formula in S2, and the total power is calculated and the frequency and power of part of the nanometer oscillators are adjusted.

[0016] According to the technical solution, in S2, the function relationship between the precession frequency and the current change amount is fitted:

[0017] S2-1, according to the calculation formula of the precession frequency and the total effective magnetic field: , wherein is the precession frequency of the free magnetic layer, is the gyromagnetic ratio, is the total effective magnetic field, and , wherein is the external magnetic field strength, is the spin torque equivalent magnetic field, is the sum of other fixed magnetic fields, wherein , is the direct current spin polarization current size, so that the external magnetic field strength is unchanged This is the theoretical situation. In reality, due to the different lattice defects and interface roughness of each nano-oscillator, the microscopic magnetic state and thermal fluctuation state are different under different currents. The amount of change in current is constantly changing. With precession frequency It cannot follow a linear relationship and requires fitting with a nonlinear function;

[0018] S2-2, For the same type of nano-oscillator, its and The relationship has similarity, let the fit be... polynomial form of degree ,in The regression coefficients were obtained by fitting historical data to obtain the same type of nano-oscillator. and The correspondence is used to convert the sample input into a multinomial feature matrix, and the least squares method is used to solve the parameter vector to obtain the magnitude of the regression coefficients.

[0019] According to the above technical solution, the specific method for fitting the functional relationship between the precession frequency and the change in magnetic field strength in S2 is as follows: S2-3, based on the total effective magnetic field calculation formula in S2-1, when When fixed fixed, and It is a linear relationship, that is and It is a linear relationship, therefore each Each corresponds to a linear relationship with a proportional coefficient, which is obtained by detecting the historical data of various nano-oscillators of the same model. The size of the array is obtained ,in For different sizes The quantity is determined based on the detection results of the change in magnetic field strength and precession frequency, and the result is obtained at a certain time. Below and The proportionality coefficient, i.e. ,in , This is the proportionality coefficient.

[0020] According to the above technical solution, in step S3, the specific method of correction is as follows: the actual detected change in precession frequency after adjusting the external magnetic field strength is... Statistical analysis was performed and compared with the fitting results in S2-3 to obtain the error ratio. Collect each time And averaged to obtain Next time In the calculation of the original Subtract the above from Then calculate.

[0021] According to the above technical scheme, the S4 is specifically: by adjusting the external magnetic field intensity to make the frequency of each nanometer oscillator consistent, adjusting by using the fitting formula in S2-3, then detecting the total power of the nanometer oscillator in the current array, obtaining the difference according to the total power requirement, and calculating the average power of the total power requirement distributed to each nanometer oscillator, then calculating the difference between the power of each nanometer oscillator and the average power, adjusting the power of several nanometer oscillators by changing the current, so that the total power reaches the total power requirement, each time the current is changed to adjust the power, the change value of the current to the frequency is calculated according to the fitting formula in S2-2, and then the frequency is corrected according to the fitting formula in S2-3, and the nanometer oscillators without power adjustment do not need to adjust the frequency because the frequency is consistent.

[0022] According to the above technical scheme, in S4, the power of several nanometer oscillators is adjusted by changing the current, specifically: S4-1, first query the best power range of this type of nanometer oscillator If the total power requirement Is greater than the sum of the power of all nanometer oscillators , it means that the power of part of the nanometer oscillators needs to be increased, otherwise the power of part of the nanometer oscillators needs to be reduced, determine the increase and decrease, calculate the average power of the nanometer oscillator And arrange the power of each nanometer oscillator From small to large, when increasing, select several nanometer oscillators with the smallest power to increase the power, when reducing, select several nanometer oscillators with the largest power to reduce the power, and when adjusting the power, ensure that the power of all nanometer oscillators is within Range, and make the adjusted .

[0023] Compared with the prior art, the beneficial effects achieved by the present application are: in the present application, when adjusting the direct current spin polarization current each time, an electrode is connected between the reference magnetic layer and the free magnetic layer of the nanometer oscillator, the frequency of the output voltage oscillating with time is detected, the precession frequency of the free magnetic layer is calculated, after collecting a large amount of data, the precession frequency shift state of the free magnetic layer under current adjustment is predicted by using the method of machine learning, and the magnetic response characteristics of the nanometer oscillator under the external magnetic field are predicted, the function relationship between the change degree of the external magnetic field intensity and the change degree of the current is established, so that the frequency and power of each nanometer oscillator are efficiently adjusted individually.

[0024] In the power adjustment, first, the frequency of each nano-oscillator is made consistent by adjusting the external magnetic field strength, then the total power demand of the nano-oscillator array is determined, the current total power is detected, the difference is obtained, and only the power of a number of nano-oscillators with the largest power difference needs to be adjusted to maintain the total power to reach the total power demand, compared with the adjustment mode that the power of each nano-oscillator is completely consistent, the power adjustment steps are greatly shortened. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application, and are used to explain the application together with the embodiments of the application, and do not constitute a limitation on the application. In the drawings:

[0026] Figure 1 is a schematic diagram of the overall module structure of the application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0028] Please refer to Figure 1 The application provides a technical solution: a nano-molecular device magnetic response modeling method and system based on machine learning, which comprises an electromagnetic response test module, an electromagnetic response modeling module, and a parameter adjustment module. The electromagnetic response test module is used to detect the current and voltage oscillation frequency between the reference magnetic layer and the free magnetic layer of the nano-oscillator, and detect the external magnetic field strength of the nano-oscillator. The electromagnetic response modeling module is used to predict the precession frequency shift state of the free magnetic layer under the current adjustment, and predict the voltage oscillation frequency change of the nano-oscillator when the external magnetic field strength changes. The parameter adjustment module is used to adjust the direct current spin polarization current and the external magnetic field strength.

[0029] The electromagnetic response test module comprises a conductive electrode, a current detection unit, a voltage detection unit, a spectrum analysis module, a micro-Hall probe, and a magnetic field strength calculation module; the conductive electrode is electrically connected with the current detection unit and the voltage detection unit; the spectrum analysis module is electrically connected with the voltage detection unit; and the micro-Hall probe is electrically connected with the magnetic field strength calculation module; the conductive electrode is used to lead out signals between the reference magnetic layer and the free magnetic layer, to realize the access and detection of the current and the voltage; the current detection unit is used to monitor the direct current spin polarization current size injected into the nano oscillator in real time; the voltage detection unit is used to collect the output voltage signal of the nano oscillator, to reflect the voltage oscillation caused by the change in the magnetic resistance; the spectrum analysis module is used to analyze the voltage signal in the time domain, to calculate the main frequency of the high-frequency oscillation, that is, the precession frequency of the free magnetic layer, and to measure the power intensity of the main frequency point to obtain the power; and the micro-Hall probe is used to detect the local external magnetic field strength around the nano oscillator, and the magnetic field strength calculation module is used to convert the Hall voltage signal output by the micro-Hall probe into the actual external magnetic field strength value.

[0030] The electromagnetic response modeling module comprises a model marking module, a data acquisition module, a precession frequency prediction module, a magnetic response prediction module, and a machine learning module; the model marking module is electrically connected with the data acquisition module; the data acquisition module is electrically connected with the spectrum analysis module and the magnetic field strength calculation module; the precession frequency prediction module and the magnetic response prediction module are electrically connected with the data acquisition module and the current detection unit; and the machine learning module is electrically connected with the precession frequency prediction module and the magnetic response prediction module; the model marking module is used to find and mark the model of each nano oscillator; the data acquisition module is used to collect voltage, current, frequency, and magnetic field strength data; the precession frequency prediction module is used to predict the precession frequency change caused by the change in the current according to the current model of the nano oscillator; the magnetic response prediction module is used to predict the voltage oscillation frequency change when the external magnetic field strength changes according to the current model of the nano oscillator; and the machine learning module is used to correct the prediction result of the precession frequency according to the actual precession frequency when the external magnetic field strength changes.

[0031] The parameter adjustment module comprises a current control module, a magnetic field strength control module, a total power calculation module, and a power difference calculation module; the current control module is electrically connected with the precession frequency prediction module; the magnetic field strength control module is electrically connected with the magnetic response prediction module; the total power calculation module is electrically connected with the spectrum analysis module and the power difference calculation module; the current control module is used to control the direct current spin polarization current; the magnetic field strength control module is used to control the external magnetic field strength; the total power calculation module is used to calculate the total power of the nano oscillator array; and the power difference calculation module is used to calculate the power that needs to be adjusted.

[0032] The working method of the system is as follows:

[0033] S0. During the initial test, each nano-oscillator model is marked so that each set of conductive electrodes connects the reference magnetic layer and the free magnetic layer of each nano-oscillator.

[0034] S1. Randomly adjust the magnitude of the DC spin polarization current in each nano oscillator, perform time-domain analysis on the acquired voltage signal, calculate the main frequency of its high-frequency oscillation, i.e. the precession frequency of the free magnetosphere, and then adjust the external magnetic field strength around each nano oscillator to directly adjust the precession frequency, so that the deviation of the precession frequency is corrected after each current adjustment.

[0035] S2. Record the relationship between the precession frequency change and the current change each time the current is adjusted, and the relationship between the precession frequency change and the magnetic field strength change each time the external magnetic field strength is adjusted. After obtaining batch data, fit the functional relationship between the precession frequency change and the current change, and fit the functional relationship between the precession frequency change and the magnetic field strength change, so as to predict the change of precession frequency before adjusting the current and magnetic field strength.

[0036] S3. Using a machine learning model combined with the fitted two functional relationships, the magnetic field strength to be adjusted is calculated directly before adjusting the current. During adjustment, the magnetic field strength is adjusted directly according to the calculated magnetic field strength, and the fitted functional relationship is corrected based on the actual precession frequency after adjusting the magnetic field strength.

[0037] S4. When a new nano-oscillator matrix needs to adjust its frequency and power, the frequency is adjusted for consistency according to the fitting formula in S2, and the total power is calculated and the frequency and power of some nano-oscillators are adjusted.

[0038] In S2, the functional relationship between the precession frequency and the change in current is fitted:

[0039] S2-1. Based on the calculation formula for precession frequency and total effective magnetic field: ,in The precession frequency of the free magnetosphere. The gyroscope magnetic ratio, The total effective magnetic field, and ,in The external magnetic field strength, The equivalent magnetic field is the spin torque. It is the sum of other fixed magnetic fields, where , The magnitude of the DC spin-polarized current, therefore, depends on the external magnetic field strength. Under the condition of no change This is the theoretical situation. In reality, due to the different lattice defects and interface roughness of each nano-oscillator, the microscopic magnetic state and thermal fluctuation state are different under different currents. The current changes with time The precession frequency Cannot follow the linear relationship, need nonlinear function fitting;

[0040] S2-2, for the same type of nano oscillator, the The relationship with Similarity, set fitting as Polynomial form, Where Is the regression coefficient, by fitting historical data to obtain the The corresponding relationship with The sample input is converted into a polynomial feature matrix, and the least squares method is used to solve the parameter vector to obtain the size of the regression coefficient;

[0041] In S2, the specific method of fitting the function relationship between the precession frequency and the magnetic field intensity change is: S2-3, according to the total effective magnetic field calculation formula in S2-1, when Fixed Fixed, And Is a linear relationship, that is And Is a linear relationship, so each Each corresponds to a linear relationship of a proportionality coefficient, by detecting the size of each In the history of the same type of nano oscillator, an array Where Is the number of different sizes And according to the detection results of the magnetic field intensity change and the precession frequency, the proportionality coefficient of Under a certain And Is obtained, that is Where , Is the proportionality coefficient;

[0042] By collecting each precession frequency, current, and magnetic field intensity change in history, the personalized parameter relationship of the same type of nano oscillator is obtained, which is adjusted to be more accurate and more in line with the actual situation.

[0043] In S3, the specific way of correction is: the actual detected precession frequency change value After adjusting the external magnetic field intensity is counted and compared with the fitting results in S2-3 to obtain the error proportion Collect each And take the average to get Next time The original is subtracted from the above and then calculated again.

[0044] S4 is specifically: by adjusting the external magnetic field intensity to make the frequency of each nano oscillator consistent, adjusting by using the fitting formula in S2-3, then detecting the total power of the nano oscillators in the current array, obtaining the difference according to the total power requirement, and calculating the average power of the total power requirement distributed to each nano oscillator, and then calculating the difference between the power of each nano oscillator and the average power, by changing the current to adjust the power of some nano oscillators, so that the total power reaches the total power requirement, each time the current is changed to adjust the power, the change value of the current to the frequency is calculated according to the fitting formula in S2-2, and then the frequency is corrected according to the fitting formula in S2-3, and the nano oscillators without power adjustment do not need to adjust the frequency because the frequency is consistent.

[0045] In S4, the power of some nano oscillators is adjusted by changing the current, which is specifically: S4-1, first query the best power range of this type of nano oscillator , if the total power requirement is greater than the sum of the power of all nano oscillators , it means that the power of part of the nano oscillators needs to be increased, otherwise the power of part of the nano oscillators needs to be reduced, determine the increase and decrease, calculate the average power of the nano oscillators , and arrange the power of each nano oscillator from small to large, increase the power of the nano oscillators with the smallest power, and reduce the power of the nano oscillators with the largest power when needed, and adjust the power to ensure that the power of all nano oscillators is within , and the adjusted .

[0046] The present application detects the frequency of the high-frequency oscillation of the output voltage with time when adjusting the direct current spin polarization current each time, calculates the precession frequency of the free magnetic layer, collects a large amount of data, uses the method of machine learning to predict the precession frequency offset state of the free magnetic layer under current adjustment, and predicts the magnetic response characteristics of the nano oscillator under the external magnetic field, directly establishes a functional relationship between the change degree of the external magnetic field intensity and the change degree of the current, so as to efficiently adjust the frequency and power of each nano oscillator individually.

[0047] In the power adjustment, first, the frequency of each nano-oscillator is made consistent by adjusting the external magnetic field intensity, then the total power demand of the nano-oscillator array is determined, the current total power is detected, the difference is obtained, only the power of several nano-oscillators with the largest power difference needs to be adjusted, and the total power reaches the total power demand, compared with the adjustment mode that the power of each nano-oscillator is completely consistent, the power adjustment steps are greatly shortened.

[0048] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0049] Finally, it should be noted that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art will still be able to modify the technical solutions described in the foregoing embodiments or make equivalent replacements to some technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A machine learning-based method for modeling the magnetic response of nanomolecular devices, characterized by: include: S0. During the initial test, each nano-oscillator model is marked so that each set of conductive electrodes connects the reference magnetic layer and the free magnetic layer of each nano-oscillator. S1. Randomly adjust the magnitude of the DC spin polarization current in each nano oscillator, perform time-domain analysis on the acquired voltage signal, calculate the free magnetosphere precession frequency, and then adjust the external magnetic field strength around each nano oscillator to directly adjust the precession frequency. S2. Record the relationship between the precession frequency change and the current change each time the current is adjusted, and the relationship between the precession frequency change and the magnetic field strength change each time the external magnetic field strength is adjusted. After obtaining the batch data, fit the functional relationship between the precession frequency change and the current change, and fit the functional relationship between the precession frequency change and the magnetic field strength change. S3. Using a machine learning model combined with the fitted two functional relationships, the magnetic field strength to be adjusted is calculated directly before adjusting the current. During adjustment, the magnetic field strength is adjusted directly according to the calculated magnetic field strength, and the fitted functional relationship is corrected based on the actual precession frequency after adjusting the magnetic field strength. S4. When a new nano-oscillator matrix needs to adjust its frequency and power, the frequency is adjusted for consistency according to the fitting formula in S2, and the total power is calculated and the frequency and power of some nano-oscillators are adjusted.

2. The method for modeling the magnetic response of nanomolecular devices based on machine learning according to claim 1, characterized in that: In step S2, the functional relationship between the precession frequency and the change in current is fitted: S2-1. Based on the calculation formula for precession frequency and total effective magnetic field: ,in The precession frequency of the free magnetosphere. The magnetic ratio of the gyroscope. The total effective magnetic field, and ,in The external magnetic field strength, The equivalent magnetic field is the spin torque. It is the sum of other fixed magnetic fields, where , The magnitude of the DC spin-polarized current, therefore, depends on the external magnetic field strength. Under the condition of no change This is the theoretical situation; in actual practice... The amount of change in current is constantly changing. With precession frequency It cannot follow a linear relationship and requires fitting with a nonlinear function; S2-2, For the same type of nano-oscillator, its and The relationship has similarity, let the fit be... polynomial form of degree. ,in The regression coefficients were obtained by fitting historical data to obtain the same type of nano-oscillator. and The correspondence is used to convert the sample input into a multinomial feature matrix, and the least squares method is used to solve the parameter vector to obtain the magnitude of the regression coefficients.

3. The method for modeling the magnetic response of nanomolecular devices based on machine learning according to claim 2, characterized in that: In S2, the specific method for fitting the functional relationship between the precession frequency and the change in magnetic field strength is as follows: S2-3, according to the total effective magnetic field calculation formula in S2-1, when When fixed fixed, and It is a linear relationship, that is and It is a linear relationship, therefore each Each corresponds to a linear relationship with a proportional coefficient, which is obtained by detecting the historical data of various nano-oscillators of the same model. The size of the array is obtained ,in For different sizes The quantity is determined based on the detection results of the change in magnetic field strength and precession frequency, and the result is obtained at a certain time. Below and The proportionality coefficient, i.e. ,in , This is the proportionality coefficient.

4. The method for modeling the magnetic response of nanomolecular devices based on machine learning according to claim 3, characterized in that: In S3, the specific method of correction is as follows: the actual detected change in precession frequency after adjusting the external magnetic field strength is... Statistical analysis was performed and compared with the fitting results in S2-3 to obtain the error ratio. Collect each time And averaged to obtain Next time When calculating, the original Subtract Then perform the calculation.

5. The method for modeling the magnetic response of nanomolecular devices based on machine learning according to claim 4, characterized in that: Specifically, S4 involves: adjusting the external magnetic field strength to ensure consistent frequencies for each nano-oscillator, using the fitting formula in S2-3 for adjustment, then detecting the total power of the nano-oscillators in the current array, obtaining the difference based on the total power requirement, calculating the average power allocated to each nano-oscillator based on the total power requirement, and then calculating the difference between the power of each nano-oscillator and the average power. By changing the current, the power of several nano-oscillators is adjusted to ensure the total power meets the total power requirement. Each time the current is changed to adjust the power, the change in frequency due to the current is calculated according to the fitting formula in S2-2, and the frequency is corrected according to the fitting formula in S2-3.

6. The method for modeling the magnetic response of nanomolecular devices based on machine learning according to claim 5, characterized in that: In step S4, adjusting the power of several nano-oscillators by changing the current specifically involves: S4-1, firstly, finding the optimal power range for this type of nano-oscillator. If the total power demand Greater than the sum of the power of all nano-oscillators This means that the power of some nano-oscillators needs to be increased, and vice versa. The impact on the average power of the nano-oscillators after determining the increase or decrease will be determined. Calculations were performed, and the power of each nano-oscillator was... Arranged from smallest to largest, when increasing power, select the few nano-oscillators with the lowest power and increase their power; when decreasing power, select the few nano-oscillators with the highest power and decrease their power. When adjusting the power, ensure that the power of all nano-oscillators is within the range specified in the original text. Within the range, and make the adjusted .

7. A machine learning-based magnetic response modeling system for nanomolecular devices, characterized in that: This system can be applied to any one of the machine learning-based magnetic response modeling methods for nanomolecular devices according to claims 1-6. The system includes an electromagnetic response testing module, an electromagnetic response modeling module, and a parameter adjustment module. The electromagnetic response testing module is used to detect the current and voltage oscillation frequencies between the reference magnetic layer and the free magnetic layer of the nano oscillator, and to detect the external magnetic field strength of the nano oscillator. The electromagnetic response modeling module is used to predict the precession frequency shift state of the free magnetic layer under current adjustment and to predict the voltage oscillation frequency change of the nano oscillator when the external magnetic field strength changes, using machine learning methods. The parameter adjustment module is used to adjust the DC spin polarization current and the external magnetic field strength. The electromagnetic response testing module includes conductive electrodes, a current detection unit, a voltage detection unit, a spectrum analysis module, a micro Hall probe, and a magnetic field strength calculation module. The conductive electrodes are used to extract signals between the reference magnetic layer and the free magnetic layer to achieve the access and detection of current and voltage. The current detection unit is used to monitor the magnitude of the DC spin polarization current injected into the nano oscillator in real time. The voltage detection unit is used to collect the output voltage signal at both ends of the nano oscillator to reflect the voltage oscillation caused by the change in magnetoresistance. The spectrum analysis module is used to perform time-domain analysis on the voltage signal to calculate its high-frequency oscillation main frequency, i.e., the free magnetic layer precession frequency, and to measure the power intensity at the main frequency point to obtain the power. The micro Hall probe is used to detect the local external magnetic field strength around the nano oscillator. The magnetic field strength calculation module is used to convert the Hall voltage signal output by the micro Hall probe into the actual external magnetic field strength value. The electromagnetic response modeling module includes a model marking module, a data acquisition module, a precession frequency prediction module, a magnetic response prediction module, and a machine learning module. The model marking module is used to find and mark the model of each nano-oscillator. The data acquisition module is used to collect voltage, current, frequency, and magnetic field strength data. The precession frequency prediction module is used to predict the precession frequency change caused by the change in current based on the current model of the nano-oscillator. The magnetic response prediction module is used to predict the voltage oscillation frequency change when the external magnetic field strength changes based on the current model of the nano-oscillator. The machine learning module is used to correct the predicted precession frequency based on the actual precession frequency when the external magnetic field strength changes. The parameter adjustment module includes a current control module, a magnetic field strength control module, a total power calculation module, and a power difference calculation module. The current control module is used to control the DC spin polarization current, the magnetic field strength control module is used to control the external magnetic field strength, the total power calculation module is used to calculate the total power of the nano-oscillator array, and the power difference calculation module is used to calculate the power that needs to be adjusted.

Images