Indium phosphide growth monitoring system and method based on multi-parameter learning

Abstract

The present invention relates to the field of semiconductor material technology, and in particular to an indium phosphide growth monitoring system and method based on multi-parameter learning, the steps of which include: collecting temperature information and solid-liquid interface height information in a single crystal furnace, and collecting indium phosphide crystal growth characteristic information at the same time; obtaining an axial temperature gradient index, a radial temperature gradient index, a temperature fluctuation index, and a thermal field symmetry index in the single crystal furnace according to the temperature information; obtaining a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information; constructing a multi-parameter learning model, obtaining crystal control information according to the multi-parameter learning model, and controlling the indium phosphide crystal growth process through the crystal control information. The present invention improves the growth efficiency and crystal quality of indium phosphide crystals by quantifying the thermal field distribution and crystal growth characteristics in a single crystal furnace.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor materials, and in particular to an indium phosphide growth monitoring system and method based on multi-parameter learning. Background Art

[0002] As a second-generation semiconductor material, indium phosphide (InP) has many advantages such as high saturated electron drift velocity, strong radiation resistance, good thermal conductivity, and high photoelectric conversion efficiency. It is widely used in optical communications, high-frequency millimeter-wave devices, and optoelectronic integrated circuits.

[0003] At present, the main growth methods of InP single crystal materials are liquid-enclosed Czochralski method (LEC), vertical gradient solidification method (VGF) and vertical Bridgman method (VB). Among them, the vertical Bridgman method is one of the main process technologies for growing low dislocation density InP single crystals.

[0004] In a vertical Bridgman single crystal furnace, the thermal field distribution in the single crystal furnace is the key single crystal growth condition. Reasonable control of the temperature gradient in the furnace and the solid-liquid interface during the single crystal growth process, especially the axial and radial temperature distribution in the furnace, is crucial to the growth of InP single crystals. Good thermal field symmetry is the prerequisite for establishing the temperature gradient, which has a significant impact on the growth morphology, growth size and crystal quality of the crystal. Existing InP growth monitoring methods usually ignore the influence of thermal field symmetry on the growth of InP single crystals, resulting in too many defects inside the crystal. Summary of the invention

[0005] In order to overcome the defects and shortcomings of the prior art, the present invention provides an indium phosphide growth monitoring system and method based on multi-parameter learning, which quantifies the thermal field distribution in the single crystal furnace and the crystal growth characteristics, thereby improving the growth efficiency and crystal quality of indium phosphide crystals.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention provides an indium phosphide growth monitoring method based on multi-parameter learning, comprising the following steps:

[0008] S1, collecting temperature information and solid-liquid interface height information in the single crystal furnace, and collecting indium phosphide crystal growth characteristic information;

[0009] S2. Acquire an axial temperature gradient index, a radial temperature gradient index, a temperature fluctuation index, and a thermal field symmetry index in the single crystal furnace according to the temperature information;

[0010] S3, obtaining a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information;

[0011] S4, constructing a multi-parameter learning model, inputting the axial temperature gradient index, the radial temperature gradient index, the temperature fluctuation index, the thermal field symmetry index and the crystal growth characteristic index into the multi-parameter learning model to obtain crystal control information, and controlling the indium phosphide crystal growth process through the crystal control information.

[0012] As a preferred technical solution, obtaining the axial temperature gradient index in step S2 includes the following specific steps:

[0013] S201, acquiring the collected temperature information in the single crystal furnace;

[0014] S202, substituting the temperature information into an axial temperature gradient index calculation formula to calculate the axial temperature gradient index, the axial temperature gradient index calculation formula is:

[0015]

[0016] Where ΔT p Indicates the temperature difference of the pth group of adjacent sensors arranged along the same axis, ΔL p represents the distance difference between the pth group of adjacent sensors arranged along the same axis, P represents the number of sensors arranged along the same axis, K AT Represents the axial temperature gradient index.

[0017] As a preferred technical solution, obtaining the radial temperature gradient index in step S2 includes the following specific steps:

[0018] S203, acquiring the collected temperature information in the single crystal furnace;

[0019] S204, substituting the temperature information into a radial temperature gradient index calculation formula to calculate the radial temperature gradient index, the radial temperature gradient index calculation formula is:

[0020]

[0021] Where ΔT q Indicates the temperature difference of the qth group of adjacent sensors arranged in the same radial direction, ΔL q represents the distance difference between the qth group of adjacent sensors arranged along the same axial direction, Q represents the number of sensors arranged along the same radial direction, and K RT Represents the radial temperature gradient index.

[0022] As a preferred technical solution, obtaining the temperature fluctuation index in step S2 includes the following specific steps:

[0023] S205, sorting the temperature information collected by the i-th sensor according to the collection time to obtain a first temperature sequence of the i-th sensor;

[0024] S206, traverse the first temperature sequence of the i-th sensor to obtain a set of temperature maximum points of the first temperature sequence, wherein the set of temperature maximum points is:

[0025] Timax={Timax(t)|Timax(t)>Ti(t+1)&Timax(t)>Ti(t-1),t=2,3,4,···,n};

[0026] Where T i max (t) indicates that the temperature information collected by the i-th sensor at time t is the temperature maximum point, T i (t+1) represents the temperature information collected by the i-th sensor at time t+1, T i (t-1) represents the temperature information collected by the i-th sensor at time t-1, & represents the AND operation, n represents the number of collections, T i max represents the set of temperature maximum points;

[0027] S207, traverse the first temperature sequence of the i-th sensor to obtain a set of temperature minimum points of the first temperature sequence, wherein the set of temperature minimum points is:

[0028] Timin={Timin(t)|Timin(t)<Ti(t+1)&Timin(t)<Ti(t-1),t=2,3,4,···,n};

[0029] Where T i min (t) indicates that the temperature information collected by the i-th sensor at time t is the temperature minimum point, T i (t+1) represents the temperature information collected by the i-th sensor at time t+1, T i (t-1) represents the temperature information collected by the i-th sensor at time t-1, & represents the AND operation, n represents the number of collections, T i min Represents the set of temperature maximum points;

[0030] S208, merging the temperature maximum point set and the temperature minimum point set to obtain a second temperature sequence of the i-th sensor;

[0031] S209, substituting the second temperature sequence of the i-th sensor into a temperature fluctuation coefficient calculation formula to calculate the temperature fluctuation coefficient of the i-th sensor, the temperature fluctuation coefficient calculation formula is:

[0032]

[0033] Where ΔT j Indicates the temperature difference of the jth group of adjacent temperature information in the second temperature sequence, Δt j represents the time difference of the jth group of adjacent temperature information in the second temperature sequence, M represents the number of temperature information included in the second temperature sequence, and U i represents the temperature fluctuation coefficient of the i-th sensor;

[0034] S210, calculating the temperature fluctuation index in the single crystal furnace according to the temperature fluctuation coefficient of the sensor, the temperature fluctuation index calculation formula is:

[0035]

[0036] Where U i represents the temperature fluctuation coefficient of the i-th sensor, N represents the number of sensors in the single crystal furnace, K TF Indicates the temperature fluctuation index in the single crystal furnace.

[0037] As a preferred technical solution, obtaining the thermal field symmetry index in step S2 includes the following specific steps:

[0038] S211, dividing the temperature information collected by two sensors at symmetrical positions along the central axis of the single crystal furnace at the same horizontal height into a symmetrical temperature sequence;

[0039] S212, substituting the symmetric temperature sequence into the temperature symmetry coefficient calculation formula to calculate the symmetry coefficient of each symmetric temperature sequence, the symmetry coefficient calculation formula is:

[0040]

[0041] Where T st represents the temperature information collected by the first sensor in the sth symmetrical temperature sequence at time t, T st ′ represents the temperature information collected by the second sensor in the sth symmetrical temperature sequence at time t, n represents the number of collections, R s represents the symmetry coefficient of the sth symmetric temperature sequence;

[0042] S213, calculating the thermal field symmetry index in the single crystal furnace according to the symmetry coefficient of the symmetrical temperature sequence, the thermal field symmetry index calculation formula is:

[0043]

[0044] Where R srepresents the symmetry coefficient of the sth symmetric temperature sequence, S represents the number of symmetric temperature sequences, K TFS Represents the thermal field symmetry index in the single crystal furnace.

[0045] As a preferred technical solution, step S3 includes the following specific steps:

[0046] S301, acquiring the operation time of the single crystal furnace, and acquiring the solid-liquid interface height information at the same time, wherein the solid-liquid interface height information includes initial solid-liquid interface height information and current solid-liquid interface height information, and acquiring the indium phosphide crystal growth characteristic information at the same time, wherein the indium phosphide crystal growth characteristic information includes current carrier concentration information and intrinsic carrier concentration information;

[0047] S302, substituting the single crystal furnace operation time, the solid-liquid interface height information and the indium phosphide crystal growth characteristic information into a crystal growth characteristic index calculation formula to calculate the crystal growth characteristic index, the crystal growth characteristic index calculation formula is:

[0048]

[0049] Where t r represents the operation time of the single crystal furnace, H represents the current solid-liquid interface height information, H0 represents the initial solid-liquid interface height information, V std represents the standard growth rate of the crystal, ρ represents the current carrier concentration information, ρ′ represents the intrinsic carrier concentration information, α represents the adjustment factor, K CG Represents the crystal growth characteristic index.

[0050] As a preferred technical solution, the multi-parameter learning model in step S4 is constructed based on a BP neural network, and the multi-parameter learning model comprises an input layer, a hidden layer and an output layer. The input of the input layer is the axial temperature gradient index, the radial temperature gradient index, the temperature fluctuation index, the thermal field symmetry index and the crystal growth index. The output of the output layer is the crystal control information, and the crystal control information comprises the relative rotation speed and relative lifting speed between the crystal and the crucible.

[0051] It should be noted that the adjustment factor is obtained as follows: 5000 sets of single crystal furnace operation time, solid-liquid interface height information and indium phosphide crystal growth characteristic information are collected to distinguish whether the indium phosphide crystal growth rate and crystal quality meet the production requirements, and the single crystal furnace operation time, solid-liquid interface height information and indium phosphide crystal growth characteristic information are substituted into the crystal growth characteristic index calculation formula for calculation, and the calculated crystal growth characteristic index and the distinction result are simultaneously imported into the fitting software, and the optimal adjustment factor that meets the distinction accuracy of the distinction result is output.

[0052] The present invention also provides an indium phosphide growth monitoring system based on multi-parameter learning, comprising:

[0053] An information collection module is used to collect temperature information and solid-liquid interface height information in the single crystal furnace, and to collect growth characteristic information of indium phosphide crystals;

[0054] A multi-parameter calculation module, used to obtain an axial temperature gradient index, a radial temperature gradient index, a temperature fluctuation index and a thermal field symmetry index in the single crystal furnace according to the temperature information;

[0055] A growth characteristic index calculation module, used to obtain a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information;

[0056] A multi-parameter learning model construction module is used to construct a multi-parameter learning model, input the axial temperature gradient index, the radial temperature gradient index, the temperature fluctuation index, the thermal field symmetry index and the crystal growth characteristic index into the multi-parameter learning model to obtain crystal control information, and control the indium phosphide crystal growth process through the crystal control information;

[0057] A control module is used to control the operation of the information acquisition module, the multi-parameter calculation module, the growth characteristic index calculation module and the multi-parameter learning model construction module.

[0058] An electronic device of the present invention comprises: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes an indium phosphide growth monitoring method based on multi-parameter learning by calling the computer program stored in the memory.

[0059] A computer-readable storage medium of the present invention stores instructions, and when the instructions are executed on a computer, the computer is enabled to execute an indium phosphide growth monitoring method based on multi-parameter learning.

[0060] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0061] The present invention quantifies the thermal field distribution and crystal growth characteristics in a single crystal furnace by acquiring the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index, thermal field symmetry index and crystal growth characteristic index in the single crystal furnace. At the same time, by constructing a multi-parameter learning model to optimize the relative rotation speed and relative lifting speed between the crystal and the crucible, the growth process of the indium phosphide crystal is precisely controlled, and the growth efficiency and crystallization quality of the indium phosphide crystal are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0063] Figure 1 It is a schematic diagram of the overall process of the indium phosphide growth monitoring method based on multi-parameter learning of the present invention;

[0064] Figure 2 It is a structural schematic diagram of the single crystal furnace in the indium phosphide growth monitoring method based on multi-parameter learning of the present invention;

[0065] Figure 3 It is a structural schematic diagram of the multi-parameter learning model in the indium phosphide growth monitoring method based on multi-parameter learning of the present invention;

[0066] Figure 4 It is a schematic diagram of the structure of the indium phosphide growth monitoring system based on multi-parameter learning of the present invention. DETAILED DESCRIPTION

[0067] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0068] Example 1

[0069] like Figure 1 As shown, this embodiment provides an indium phosphide growth monitoring method based on multi-parameter learning, which specifically includes the following steps:

[0070] S1: Collect the temperature information and solid-liquid interface height information in the single crystal furnace, and collect the growth characteristic information of indium phosphide crystal at the same time;

[0071] like Figure 2 As shown, during the growth process of indium phosphide crystals, the relative movement of the heater and the crucible in the single crystal furnace is controlled, thereby controlling the height information of the solid-liquid interface, so that the melt is gradually crystallized to complete the single crystal growth. At the same time, by rotating the crucible at a certain speed, the melt flow is improved and the melt is heated more evenly, thereby growing indium phosphide crystals with a lower dislocation density.

[0072] S2: Obtaining the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index and thermal field symmetry index in the single crystal furnace according to the temperature information;

[0073] During the growth of InP crystals, the axial temperature gradient can affect the solute distribution and crystallization rate by affecting the convection intensity of the melt. The axial temperature gradient index is used to quantify the uniformity of the axial temperature distribution in the single crystal furnace to ensure the stability and repeatability of the axial temperature gradient during crystal growth. The specific steps for obtaining the axial temperature gradient index are as follows:

[0074] S201: Acquire the collected temperature information in the single crystal furnace;

[0075] S202: Substituting the temperature information into the axial temperature gradient index calculation formula to calculate the axial temperature gradient index. The axial temperature gradient index calculation formula is:

[0076]

[0077] Where ΔT p Indicates the temperature difference of the pth group of adjacent sensors arranged along the same axis, ΔL p represents the distance difference between the pth group of adjacent sensors arranged along the same axis, P represents the number of sensors arranged along the same axis, K AT represents the axial temperature gradient index;

[0078] The uniform radial temperature gradient helps to reduce the internal stress of the indium phosphide crystal during its growth process and reduce the generation of defects such as dislocations and grain boundaries. At the same time, the radial temperature gradient can affect the crystallization rate and quality by affecting the shape of the solid-liquid interface. The radial temperature gradient index includes the following specific steps:

[0079] S203: Acquire the collected temperature information in the single crystal furnace;

[0080] S204: Substituting the temperature information into a radial temperature gradient index calculation formula to calculate a radial temperature gradient index. The radial temperature gradient index calculation formula is:

[0081]

[0082] Where ΔT q Indicates the temperature difference of the qth group of adjacent sensors arranged in the same radial direction, ΔL q represents the distance difference between the qth group of adjacent sensors arranged along the same axial direction, Q represents the number of sensors arranged along the same radial direction, and K RT represents the radial temperature gradient index;

[0083] During the growth of InP crystals, the melt in the single crystal furnace is in a supercooled state. Excessive temperature fluctuations will cause the InP crystals to grow rapidly along multiple crystal directions in a short period of time, inducing the appearance of twins and polycrystals. Therefore, the temperature fluctuation index is used to quantify the temperature stability in the single crystal furnace. The temperature fluctuation index is obtained by the following specific steps:

[0084] S205: sorting the temperature information collected by the i-th sensor according to the collection time to obtain a first temperature sequence of the i-th sensor;

[0085] S206: traverse the first temperature sequence of the i-th sensor to obtain a set of temperature maximum points of the first temperature sequence, where the set of temperature maximum points is:

[0086] Timax={Timax(t)|Timax(t)>Ti(t+1)&Timax(t)>Ti(t-1),t=2,3,4,···,n};

[0087] Where T i max (t) indicates that the temperature information collected by the i-th sensor at time t is the temperature maximum point, T i (t+1) represents the temperature information collected by the i-th sensor at time t+1, T i (t-1) represents the temperature information collected by the i-th sensor at time t-1, & represents the AND operation, n represents the number of collections, T i max Represents the set of temperature maximum points;

[0088] S207: Traverse the first temperature sequence of the i-th sensor to obtain a set of temperature minimum points of the first temperature sequence. The set of temperature minimum points is:

[0089] Timin={Timin(t)|Timin(t)<Ti(t+1)&Timin(t)<Ti(t-1),t=2,3,4,···,n};

[0090] Where T i min (t) indicates that the temperature information collected by the i-th sensor at time t is the temperature minimum point, T i (t+1) represents the temperature information collected by the i-th sensor at time t+1, T i (t-1) represents the temperature information collected by the i-th sensor at time t-1, & represents the AND operation, n represents the number of collections, T i min Represents the set of temperature maximum points;

[0091] S208: merging the temperature maximum point set and the temperature minimum point set to obtain a second temperature sequence of the i-th sensor;

[0092] S209: Substituting the second temperature sequence of the i-th sensor into the temperature fluctuation coefficient calculation formula to calculate the temperature fluctuation coefficient of the i-th sensor. The temperature fluctuation coefficient calculation formula is:

[0093]

[0094] Where ΔT j Indicates the temperature difference of the jth group of adjacent temperature information in the second temperature sequence, Δt j represents the time difference of the jth group of adjacent temperature information in the second temperature sequence, M represents the number of temperature information included in the second temperature sequence, and U i represents the temperature fluctuation coefficient of the i-th sensor;

[0095] S210: Calculate the temperature fluctuation index in the single crystal furnace according to the temperature fluctuation coefficient of the sensor. The temperature fluctuation index calculation formula is:

[0096]

[0097] Where U i represents the temperature fluctuation coefficient of the i-th sensor, N represents the number of sensors in the single crystal furnace, K TF Indicates the temperature fluctuation index in the single crystal furnace;

[0098] A symmetrical thermal field can ensure that the crystal is uniformly thermally affected during growth, thereby reducing stress concentration and defect formation during crystal growth. A symmetrical thermal field also helps maintain the stability and fluidity of the melt. Obtaining the thermal field symmetry index includes the following specific steps:

[0099] S211: dividing the temperature information collected by two sensors at symmetrical positions along the central axis of the single crystal furnace at the same horizontal height into a symmetrical temperature sequence;

[0100] S212: Substituting the symmetrical temperature sequence into the temperature symmetry coefficient calculation formula to calculate the symmetry coefficient of each symmetrical temperature sequence. The symmetry coefficient calculation formula is:

[0101]

[0102] Where T st represents the temperature information collected by the first sensor in the sth symmetrical temperature sequence at time t, T st ′ represents the temperature information collected by the second sensor in the sth symmetrical temperature sequence at time t, n represents the number of collections, R srepresents the symmetry coefficient of the sth symmetric temperature sequence;

[0103] S213: Calculate the thermal field symmetry index in the single crystal furnace according to the symmetry coefficient of the symmetrical temperature sequence. The thermal field symmetry index calculation formula is:

[0104]

[0105] Where R s represents the symmetry coefficient of the sth symmetric temperature sequence, S represents the number of symmetric temperature sequences, K TFS Represents the thermal field symmetry index in the single crystal furnace.

[0106] S3: obtaining a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information;

[0107] During the growth of InP crystals, the position change of the solid-liquid interface is closely related to the growth rate, growth mode and growth quality of the crystal. By collecting and analyzing the height information of the solid-liquid interface, the dynamic characteristics of crystal growth can be obtained. At the same time, the carrier concentration information reflects the distribution of carriers inside the crystal, which directly affects the electrical and optical properties of the crystal. Therefore, by obtaining the crystal growth characteristic index through the solid-liquid interface height information and carrier concentration information, the growth status of InP crystals can be quantitatively evaluated. Obtaining the crystal growth characteristic index includes the following specific steps:

[0108] S301: Acquire the single crystal furnace operation time, and simultaneously acquire the solid-liquid interface height information, the solid-liquid interface height information includes the initial solid-liquid interface height information and the current solid-liquid interface height information, and simultaneously acquire the indium phosphide crystal growth characteristic information, the indium phosphide crystal growth characteristic information includes the current carrier concentration information and the intrinsic carrier concentration information;

[0109] S302: Substituting the single crystal furnace operation time, the solid-liquid interface height information and the indium phosphide crystal growth characteristic information into a crystal growth characteristic index calculation formula to calculate the crystal growth characteristic index. The crystal growth characteristic index calculation formula is:

[0110]

[0111] Where t r represents the operation time of the single crystal furnace, H represents the current solid-liquid interface height information, H0 represents the initial solid-liquid interface height information, V std represents the standard growth rate of the crystal, ρ represents the current carrier concentration information, ρ′ represents the intrinsic carrier concentration information, α represents the adjustment factor, K CG Represents the crystal growth characteristic index.

[0112] S4: construct a multi-parameter learning model, input the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index, thermal field symmetry index and crystal growth characteristic index into the multi-parameter learning model to obtain crystal control information, and control the indium phosphide crystal growth process through the crystal control information;

[0113] The multi-parameter learning model is based on BP neural network, such as Figure 3 As shown, the multi-parameter learning model includes 1 input layer, 1 hidden layer and 1 output layer. The input layer includes 5 nodes. The input of the input layer is axial temperature gradient index, radial temperature gradient index, temperature fluctuation index, thermal field symmetry index and crystal growth index. The hidden layer includes 7 nodes. The activation function of the hidden layer is ReLU activation function. The output layer includes 2 nodes. The output of the output layer is crystal control information. The crystal control information includes the relative rotation speed and relative lifting speed between the crystal and the crucible. The activation function of the output layer is linear activation function. The learning rate (learning_rate) of the multi-parameter learning model is 0.001, the loss function (loss) is MSE mean square error, the optimization algorithm (optimizers) is Adam, the number of training rounds (epochs) is 500, and the batch size (batch_size) is 64.

[0114] Example 2

[0115] like Figure 4 As shown, this embodiment provides an indium phosphide growth monitoring system based on multi-parameter learning, including:

[0116] An information collection module is used to collect temperature information and solid-liquid interface height information in the single crystal furnace, and to collect growth characteristic information of indium phosphide crystals;

[0117] A multi-parameter calculation module is used to obtain the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index and thermal field symmetry index in the single crystal furnace according to the temperature information;

[0118] A growth characteristic index calculation module is used to obtain a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information;

[0119] A multi-parameter learning model building module is used to build a multi-parameter learning model, input the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index, thermal field symmetry index and crystal growth characteristic index into the multi-parameter learning model to obtain crystal control information, and control the indium phosphide crystal growth process through the crystal control information;

[0120] The control module is used to control the operation of the information acquisition module, the multi-parameter calculation module, the growth characteristic index calculation module and the multi-parameter learning model construction module.

[0121] In this embodiment, the multi-parameter calculation module is used to obtain the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index and thermal field symmetry index in the single crystal furnace according to the temperature information, wherein obtaining the axial temperature gradient index includes the following specific steps:

[0122] Obtain the collected temperature information in the single crystal furnace;

[0123] Substitute the temperature information into the axial temperature gradient index calculation formula to calculate the axial temperature gradient index. The axial temperature gradient index calculation formula is:

[0124]

[0125] Where ΔT p Indicates the temperature difference of the pth group of adjacent sensors arranged along the same axis, ΔL p represents the distance difference between the pth group of adjacent sensors arranged along the same axis, P represents the number of sensors arranged along the same axis, K AT represents the axial temperature gradient index;

[0126] Obtaining the radial temperature gradient index includes the following specific steps:

[0127] Obtain the collected temperature information in the single crystal furnace;

[0128] Substitute the temperature information into the radial temperature gradient index calculation formula to calculate the radial temperature gradient index. The radial temperature gradient index calculation formula is:

[0129]

[0130] Where ΔT q Indicates the temperature difference of the qth group of adjacent sensors arranged in the same radial direction, ΔL q represents the distance difference between the qth group of adjacent sensors arranged along the same axial direction, Q represents the number of sensors arranged along the same radial direction, and K RT represents the radial temperature gradient index;

[0131] Obtaining the temperature fluctuation index includes the following specific steps:

[0132] The temperature information collected by the i-th sensor is sorted according to the collection time to obtain the first temperature sequence of the i-th sensor;

[0133] Traverse the first temperature sequence of the i-th sensor to obtain the temperature maximum point set of the first temperature sequence. The temperature maximum point set is:

[0134] Timax={Timax(t)|Timax(t)>Ti(t+1)&Timax(t)>Ti(t-1),t=2,3,4,···,n};

[0135] Where T i max (t) indicates that the temperature information collected by the i-th sensor at time t is the temperature maximum point, T i (t+1) represents the temperature information collected by the i-th sensor at time t+1, T i (t-1) represents the temperature information collected by the i-th sensor at time t-1, & represents the AND operation, n represents the number of collections, T i max Represents the set of temperature maximum points;

[0136] Traverse the first temperature sequence of the i-th sensor to obtain the temperature minimum point set of the first temperature sequence. The temperature minimum point set is:

[0137] Timin={Timin(t)|Timin(t)<Ti(t+1)&Timin(t)<Ti(t-1),t=2,3,4,···,n};

[0138] Where T i min (t) indicates that the temperature information collected by the i-th sensor at time t is the temperature minimum point, T i (t+1) represents the temperature information collected by the i-th sensor at time t+1, T i (t-1) represents the temperature information collected by the i-th sensor at time t-1, & represents the AND operation, n represents the number of collections, T i min Represents the set of temperature maximum points;

[0139] Merge the temperature maximum point set and the temperature minimum point set to obtain the second temperature sequence of the i-th sensor;

[0140] Substitute the second temperature sequence of the i-th sensor into the temperature fluctuation coefficient calculation formula to calculate the temperature fluctuation coefficient of the i-th sensor. The temperature fluctuation coefficient calculation formula is:

[0141]

[0142] Where ΔT j Indicates the temperature difference of the jth group of adjacent temperature information in the second temperature sequence, Δt j represents the time difference of the jth group of adjacent temperature information in the second temperature sequence, M represents the number of temperature information included in the second temperature sequence, and U irepresents the temperature fluctuation coefficient of the i-th sensor;

[0143] The temperature fluctuation index in the single crystal furnace is calculated according to the temperature fluctuation coefficient of the sensor. The temperature fluctuation index calculation formula is:

[0144]

[0145] Where U i represents the temperature fluctuation coefficient of the i-th sensor, N represents the number of sensors in the single crystal furnace, K TF Indicates the temperature fluctuation index in the single crystal furnace;

[0146] Obtaining the thermal field symmetry index includes the following specific steps:

[0147] The temperature information collected by two sensors at symmetrical positions along the central axis of the single crystal furnace at the same horizontal height is divided into a symmetrical temperature sequence;

[0148] Substitute the symmetrical temperature sequence into the temperature symmetry coefficient calculation formula to calculate the symmetry coefficient of each symmetrical temperature sequence. The symmetry coefficient calculation formula is:

[0149]

[0150] Where T st represents the temperature information collected by the first sensor in the sth symmetrical temperature sequence at time t, T st ′ represents the temperature information collected by the second sensor in the sth symmetrical temperature sequence at time t, n represents the number of collections, R s represents the symmetry coefficient of the sth symmetric temperature sequence;

[0151] The thermal field symmetry index in the single crystal furnace is calculated according to the symmetry coefficient of the symmetrical temperature sequence. The calculation formula of the thermal field symmetry index is:

[0152]

[0153] Where R s represents the symmetry coefficient of the sth symmetric temperature sequence, S represents the number of symmetric temperature sequences, K TFS Represents the thermal field symmetry index in the single crystal furnace.

[0154] In this embodiment, the growth characteristic index calculation module is used to obtain the crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information, wherein obtaining the crystal growth characteristic index includes the following specific steps:

[0155] Obtaining the single crystal furnace operation time, and obtaining the solid-liquid interface height information at the same time, the solid-liquid interface height information includes the initial solid-liquid interface height information and the current solid-liquid interface height information, and obtaining the indium phosphide crystal growth characteristic information at the same time, the indium phosphide crystal growth characteristic information includes the current carrier concentration information and the intrinsic carrier concentration information;

[0156] The single crystal furnace operation time, the solid-liquid interface height information and the indium phosphide crystal growth characteristic information are substituted into the crystal growth characteristic index calculation formula to calculate the crystal growth characteristic index. The crystal growth characteristic index calculation formula is:

[0157]

[0158] Where t r represents the operation time of the single crystal furnace, H represents the current solid-liquid interface height information, H0 represents the initial solid-liquid interface height information, V std represents the standard growth rate of the crystal, ρ represents the current carrier concentration information, ρ′ represents the intrinsic carrier concentration information, α represents the adjustment factor, K CG Represents the crystal growth characteristic index.

[0159] In this embodiment, a multi-parameter learning model construction module is used to construct a multi-parameter learning model, and the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index, thermal field symmetry index and crystal growth characteristic index are input into the multi-parameter learning model to obtain crystal control information, and the indium phosphide crystal growth process is controlled by the crystal control information, wherein the multi-parameter learning model is constructed based on a BP neural network, and the multi-parameter learning model includes an input layer, a hidden layer and an output layer, the input of the input layer is the axial temperature gradient index, radial temperature gradient index, temperature fluctuation index, thermal field symmetry index and crystal growth index, and the output of the output layer is the crystal control information, and the crystal control information includes the relative rotation speed and relative lifting speed between the crystal and the crucible.

[0160] The above-mentioned parameters and steps for each unit module to achieve corresponding functions in the indium phosphide growth monitoring system based on multi-parameter learning of the present invention can refer to the parameters and steps in the embodiment of the indium phosphide growth monitoring method based on multi-parameter learning above, and will not be repeated here.

[0161] Example 3

[0162] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements the above-mentioned indium phosphide growth monitoring method based on multi-parameter learning. It should be noted that all computer programs of the indium phosphide growth monitoring method based on multi-parameter learning are implemented in C language, wherein the information acquisition module, the multi-parameter calculation module, the growth characteristic index calculation module, the multi-parameter learning model construction module and the control module are all controlled by a remote server.

[0163] Example 4

[0164] This embodiment provides a computer-readable storage medium having a rewritable computer program stored thereon;

[0165] When the computer program is executed on a computer device, the computer device is enabled to execute the above-mentioned indium phosphide growth monitoring method based on multi-parameter learning.

[0166] For example, the computer readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

Claims

1. A method for monitoring indium phosphide growth based on multi-parameter learning, characterized in that: The steps include: S1, collecting temperature information and solid-liquid interface height information in the single crystal furnace, and collecting indium phosphide crystal growth characteristic information; S2. Acquire an axial temperature gradient index, a radial temperature gradient index, a temperature fluctuation index, and a thermal field symmetry index in the single crystal furnace according to the temperature information; S3, obtaining a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information; S4, constructing a multi-parameter learning model, inputting the axial temperature gradient index, the radial temperature gradient index, the temperature fluctuation index, the thermal field symmetry index and the crystal growth characteristic index into the multi-parameter learning model to obtain crystal control information, and controlling the indium phosphide crystal growth process by the crystal control information; The step S3 comprises the following specific steps: S301, acquiring the operation time of the single crystal furnace, and acquiring the solid-liquid interface height information at the same time, wherein the solid-liquid interface height information includes initial solid-liquid interface height information and current solid-liquid interface height information, and acquiring the indium phosphide crystal growth characteristic information at the same time, wherein the indium phosphide crystal growth characteristic information includes current carrier concentration information and intrinsic carrier concentration information; S302, substituting the single crystal furnace operation time, the solid-liquid interface height information and the indium phosphide crystal growth characteristic information into a crystal growth characteristic index calculation formula to calculate the crystal growth characteristic index, the crystal growth characteristic index calculation formula is: ； In the formula Indicates the running time of the single crystal furnace. Indicates the current solid-liquid interface height information, represents the initial solid-liquid interface height information, represents the standard crystal growth rate, Indicates the current carrier concentration information, Indicates the intrinsic carrier concentration information, represents the adjustment factor, Represents the crystal growth characteristic index.

2. The indium phosphide growth monitoring method based on multi-parameter learning according to claim 1, characterized in that: Obtaining the axial temperature gradient index in step S2 includes the following specific steps: S201, acquiring the collected temperature information in the single crystal furnace; S202, substituting the temperature information into an axial temperature gradient index calculation formula to calculate the axial temperature gradient index, the axial temperature gradient index calculation formula is: ； In the formula Indicates the first The temperature difference of adjacent sensors in a group, Indicates the first The distance difference between adjacent sensors in a group, Indicates the number of sensors arranged along the same axis. Represents the axial temperature gradient index.

3. The indium phosphide growth monitoring method based on multi-parameter learning according to claim 1, characterized in that: Obtaining the radial temperature gradient index in step S2 includes the following specific steps: S203, acquiring the collected temperature information in the single crystal furnace; S204, substituting the temperature information into a radial temperature gradient index calculation formula to calculate the radial temperature gradient index, the radial temperature gradient index calculation formula is: ； In the formula Indicates the same radial arrangement The temperature difference of adjacent sensors in a group, Indicates the first The distance difference between adjacent sensors in a group, Indicates the number of sensors arranged along the same radial direction, Represents the radial temperature gradient index.

4. The indium phosphide growth monitoring method based on multi-parameter learning according to claim 1, characterized in that: Obtaining the temperature fluctuation index in step S2 includes the following specific steps: S205, The temperature information collected by the sensors is sorted according to the collection time, and the A first temperature sequence of sensors; S206, traverse The first temperature sequence of the sensor is obtained to obtain a set of temperature maximum points of the first temperature sequence, and the set of temperature maximum points is: ； In the formula Indicates The sensors in The temperature information collected at any moment is the temperature maximum point. Indicates The sensors in The temperature information collected at all times, Indicates The sensors in The temperature information collected at all times, Representation and operation, Indicates the number of collections. Represents the set of temperature maximum points; S207, traverse The first temperature sequence of the sensor is obtained to obtain a set of temperature minimum points of the first temperature sequence, and the set of temperature minimum points is: ； In the formula Indicates The sensors in The temperature information collected at any moment is the temperature minimum point. Indicates The sensors in The temperature information collected at all times, Indicates The sensors in The temperature information collected at all times, Representation and operation, Indicates the number of collections. represents the set of temperature maximum points; S208, merging the temperature maximum point set and the temperature minimum point set to obtain the first a second temperature sequence of sensors; S209, The second temperature series of the sensor is substituted into the temperature fluctuation coefficient calculation formula to calculate the The temperature fluctuation coefficient of each sensor is calculated as follows: ； In the formula Indicates the second temperature series The temperature difference of adjacent temperature information of a group, Indicates the second temperature series The time difference of adjacent temperature information of a group, represents the amount of temperature information contained in the second temperature sequence, Indicates Temperature fluctuation coefficient of each sensor; S210, calculating the temperature fluctuation index in the single crystal furnace according to the temperature fluctuation coefficient of the sensor, the temperature fluctuation index calculation formula is: ； In the formula Indicates The temperature fluctuation coefficient of each sensor is Indicates the number of sensors in the single crystal furnace, Indicates the temperature fluctuation index in the single crystal furnace.

5. The indium phosphide growth monitoring method based on multi-parameter learning according to claim 1, characterized in that: Obtaining the thermal field symmetry index in step S2 includes the following specific steps: S211, dividing the temperature information collected by two sensors at symmetrical positions along the central axis of the single crystal furnace at the same horizontal height into a symmetrical temperature sequence; S212, substituting the symmetric temperature sequence into the temperature symmetry coefficient calculation formula to calculate the symmetry coefficient of each symmetric temperature sequence, the symmetry coefficient calculation formula is: ； In the formula Indicates The first sensor in a symmetrical temperature sequence is The temperature information collected at all times, Indicates The second sensor in a symmetrical temperature sequence is The temperature information collected at all times, Indicates the number of collections. Indicates The symmetry coefficient of a symmetric temperature series; S213, calculating the thermal field symmetry index in the single crystal furnace according to the symmetry coefficient of the symmetrical temperature sequence, the thermal field symmetry index calculation formula is: ； In the formula Indicates The symmetry coefficient of a symmetric temperature series, represents the number of symmetric temperature sequences, Represents the thermal field symmetry index in the single crystal furnace.

6. The indium phosphide growth monitoring method based on multi-parameter learning according to claim 1, characterized in that: The multi-parameter learning model in step S4 is constructed based on a BP neural network, and the multi-parameter learning model comprises an input layer, a hidden layer and an output layer. The input of the input layer is the axial temperature gradient index, the radial temperature gradient index, the temperature fluctuation index, the thermal field symmetry index and the crystal growth index. The output of the output layer is the crystal control information, and the crystal control information comprises the relative rotation speed and relative lifting speed between the crystal and the crucible.

7. An indium phosphide growth monitoring system based on multi-parameter learning, which is implemented based on the indium phosphide growth monitoring method based on multi-parameter learning according to any one of claims 1 to 6, characterized in that: The system comprises: An information collection module is used to collect temperature information and solid-liquid interface height information in the single crystal furnace, and to collect growth characteristic information of indium phosphide crystals; A multi-parameter calculation module, used to obtain an axial temperature gradient index, a radial temperature gradient index, a temperature fluctuation index and a thermal field symmetry index in the single crystal furnace according to the temperature information; A growth characteristic index calculation module, used to obtain a crystal growth characteristic index according to the solid-liquid interface height information and the indium phosphide crystal growth characteristic information; A multi-parameter learning model construction module is used to construct a multi-parameter learning model, input the axial temperature gradient index, the radial temperature gradient index, the temperature fluctuation index, the thermal field symmetry index and the crystal growth characteristic index into the multi-parameter learning model to obtain crystal control information, and control the indium phosphide crystal growth process through the crystal control information; A control module is used to control the operation of the information acquisition module, the multi-parameter calculation module, the growth characteristic index calculation module and the multi-parameter learning model construction module.

8. An electronic device comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; characterized in that the processor executes the indium phosphide growth monitoring method based on multi-parameter learning as described in any one of claims 1 to 6 by calling the computer program stored in the memory.

9. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer is caused to execute the indium phosphide growth monitoring method based on multi-parameter learning as described in any one of claims 1 to 6.

Images