Liquid-cooled quantum computing server optimization method and system

Abstract

The application discloses a liquid cooling-based quantum computing server optimization method and system, relates to the technical field of quantum computing, and discloses the liquid cooling-based quantum computing server optimization method and system, which acquires quantum state and liquid cooling state data in real time, dynamically generates optimization parameters by using an AI model, and cooperatively adjusts quantum error correction operation and a liquid cooling circulation system, solves the technical problems of low heat dissipation efficiency and large error correction resource consumption in the traditional scheme, and improves system stability and computing efficiency.

Description

Technical Field

[0001] This application relates to the field of quantum computing technology, and in particular to a method and system for optimizing liquid-cooled quantum computing servers. Background Technology

[0002] Quantum computing servers rely on hardware such as superconducting qubits, ion traps, or photonic qubits to achieve parallel computing. However, qubits are extremely sensitive and easily affected by factors such as temperature fluctuations, electromagnetic noise, and quantum decoherence. This necessitates an operating environment that meets extremely low temperatures and strong electromagnetic shielding requirements. Existing optimization schemes for quantum computing servers mainly employ combinations of traditional air cooling with quantum error correction codes or single-phase immersion liquid cooling with static error correction codes. While these schemes offer improvements in heat dissipation performance and system overhead, they still have many limitations. For example, single-phase immersion liquid cooling significantly increases the difficulty of quantum chip packaging due to the high viscosity of the coolant; traditional quantum error correction codes consume a large number of qubit resources and have poor real-time performance in error correction operations; and independently designed heat dissipation and computing architectures suffer from high system redundancy and low energy efficiency. These technical shortcomings severely restrict the stability and computational efficiency of quantum computing servers, especially during long-term operation and large-scale quantum computing tasks.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this application is to provide a liquid-cooled quantum computing server optimization method and system, which aims to improve the heat dissipation efficiency and error correction real-time performance of the quantum computing server, as well as reduce system resource consumption.

[0005] To achieve the above objectives, this application proposes an optimization method for a liquid-cooled quantum computing server, the method comprising:

[0006] An initial dataset is acquired by a sensor network deployed on a quantum chip, the initial dataset including quantum state data and liquid-cooled state data;

[0007] Based on the initial dataset, it is processed by a neural network-based dynamic parameter optimization model to obtain an optimized parameter dataset, which includes error correction code parameter data and liquid cooling parameter data.

[0008] The error correction code parameter data is input into the dynamic error correction code controller to adjust the quantum error correction operation, and the liquid cooling parameter data is input into the liquid cooling system controller to adjust the liquid cooling cycle.

[0009] In one embodiment, the step of acquiring the initial dataset via a sensor network deployed on a quantum chip includes:

[0010] Quantum state data is acquired in real time by a quantum sensor and output. The quantum state data includes the decoherence time of the qubit and the error rate of the quantum gate operation.

[0011] The liquid cooling status data is collected in real time by a liquid cooling sensor and output. The liquid cooling status data includes coolant temperature, liquid level and pressure data.

[0012] The quantum state data and the liquid-cooled state data are combined to output the initial dataset.

[0013] In one embodiment, the step of processing the initial dataset using a neural network-based dynamic parameter optimization model to obtain an optimized parameter dataset includes:

[0014] Based on the quantum state data and liquid-cooled state data in the initial dataset, the time series data is processed by a time series neural network model to output the dynamic characteristic data of the qubit.

[0015] Based on the dynamic characteristic data of the qubit, the quantum gate operation error rate is predicted by a random simulation method, and the predicted quantum gate operation error rate data is output.

[0016] The optimized parameter dataset is generated based on the predicted quantum gate operation error rate data and quantum state data.

[0017] In one embodiment, the step of generating the optimized parameter dataset based on the predicted quantum gate operation error rate data and quantum state data includes:

[0018] Based on the predicted quantum gate operation error rate data, high-stability qubits and low-stability qubits are classified to output qubit classification data;

[0019] Based on the quantum bit classification data, the ratio of physical quantum bits to auxiliary quantum bits is allocated to output error correction code parameter data;

[0020] Based on the liquid cooling status data, the heat dissipation requirements are matched, and liquid cooling parameter data is output.

[0021] The error correction code parameter data and liquid cooling parameter data are merged to output the optimized parameter dataset.

[0022] In one embodiment, the steps of inputting the error correction code parameter data into a dynamic error correction code controller to adjust the quantum error correction operation, and inputting the liquid cooling parameter data into a liquid cooling system controller to adjust the liquid cooling cycle include:

[0023] Input the error correction code parameter data into the dynamic error correction code controller, and adjust the surface code distance and measurement cycle parameters;

[0024] The liquid cooling parameter data is input into the liquid cooling system controller to adjust the liquid cooling cycle.

[0025] In one embodiment, the step of inputting the error correction code parameter data into the dynamic error correction code controller and adjusting the surface code distance and measurement period parameters includes:

[0026] Based on the error correction code parameter data, determine the adjustment value of the surface code distance;

[0027] Reconstruct the topology of the qubit based on the adjusted values;

[0028] Phase compensation operations are inserted based on the reconstructed topology to optimize quantum error correction.

[0029] In one embodiment, the step of inputting the liquid cooling parameter data into the liquid cooling system controller to adjust the liquid cooling cycle includes:

[0030] Based on the liquid cooling parameter data, the boiling phase transition of the coolant on the surface of the quantum chip is controlled.

[0031] Based on the liquid cooling parameter data, the gaseous coolant is converted into liquid reflux through the condenser;

[0032] Based on the liquid cooling parameter data and gravity, a pump-free self-circulation system can be achieved.

[0033] In one embodiment, the method further includes:

[0034] Quantum gate operation results are measured using quantum state tomography, and quantum gate fidelity data is output.

[0035] The quantum gate fidelity data is input into the neural network-based dynamic parameter optimization model to iteratively update the optimization parameter dataset.

[0036] In one embodiment, the step of measuring the quantum gate operation results using quantum state tomography and outputting quantum gate fidelity data includes:

[0037] Obtain quantum state data after quantum gate operations;

[0038] Based on the quantum state data, quantum gate fidelity data is generated using a quantum state tomography algorithm.

[0039] Furthermore, to achieve the above objectives, this application also proposes a liquid-cooled quantum computing server optimization system, which includes: a memory, a processor, and a liquid-cooled quantum computing server optimization program stored in the memory and executable on the processor. The liquid-cooled quantum computing server optimization program is configured to implement the steps of the liquid-cooled quantum computing server optimization method.

[0040] The liquid-cooled quantum computing server optimization method and system proposed in this application solves the technical problems of low heat dissipation efficiency and high error correction resource consumption in traditional solutions by acquiring quantum state and liquid cooling state data in real time, dynamically generating optimization parameters using an AI model, and coordinating the quantum error correction operation and liquid cooling circulation system, thereby improving system stability and computing efficiency. Attached Figure Description

[0041] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart illustrating an embodiment of the liquid-cooled quantum computing server optimization method of this application.

[0044] Figure 2 For this application Figure 1 A detailed flowchart of step S100;

[0045] Figure 3 For this application Figure 1 A detailed flowchart of step S200;

[0046] Figure 4 For this application Figure 3 A detailed flowchart of step S230;

[0047] Figure 5 For this application Figure 1 Detailed flowchart of step S300;

[0048] Figure 6 For this application Figure 5 A detailed flowchart of step S310;

[0049] Figure 7 For this application Figure 5 A detailed flowchart of step S320;

[0050] Figure 8 This is a flowchart illustrating another embodiment of the liquid-cooled quantum computing server optimization method of this application;

[0051] Figure 9 For this application Figure 8 Detailed flowchart of step S400;

[0052] Figure 10 This is a schematic diagram of a structure provided for an embodiment of the liquid-cooled quantum computing server optimization system of this application.

[0053] Explanation of icon numbers:

[0054] 10. Memory; 20. Processor.

[0055] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0056] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0057] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0058] In existing technologies, quantum computing servers rely on hardware such as superconducting qubits, ion traps, or photonic qubits to achieve parallel computing. However, the extreme sensitivity of qubits necessitates extremely low temperatures and strong electromagnetic shielding for their operation. Traditional optimization schemes combine air cooling with quantum error-correcting codes or single-phase immersion liquid cooling with static error-correcting codes. However, these schemes suffer from problems such as high coolant viscosity, difficulty in quantum chip packaging, error correction overhead exceeding 30% of the total number of qubits, and high system redundancy. For example, during the operation of a superconducting quantum chip, the liquid cooling system cannot match the dynamically changing heat dissipation requirements in real time, while the static error-correcting code cannot adjust resource allocation according to the stability differences of the qubits, resulting in reduced energy efficiency.

[0059] To address the aforementioned issues, the inventors discovered that the independent operation of the heat dissipation and error correction systems in existing technologies is a key factor leading to resource waste. By analyzing the correlation between qubit decoherence time and liquid cooling status data, they proposed inputting real-time monitoring data into a unified model for collaborative optimization. Furthermore, considering the coupling effect between quantum gate operation error rate and coolant temperature fluctuations, a dynamic adjustment mechanism was designed to balance error correction overhead and heat dissipation efficiency. Ultimately, this resulted in a technical approach that collects multi-dimensional data through a sensor network, generates joint optimization parameters using an AI model, and simultaneously controls the error correction and liquid cooling systems.

[0060] Based on this, embodiments of this application provide an optimization method for a liquid-cooled quantum computing server, referring to... Figure 1 The liquid-cooled quantum computing server optimization method includes steps S100 to S300, wherein:

[0061] Step S100: Obtain an initial dataset through a sensor network deployed on the quantum chip, the initial dataset including quantum state data and liquid-cooled state data;

[0062] Step S200: Based on the initial dataset, process it through a neural network-based dynamic parameter optimization model to obtain an optimized parameter dataset, which includes error correction code parameter data and liquid cooling parameter data;

[0063] Step S300: Input the error correction code parameter data into the dynamic error correction code controller to adjust the quantum error correction operation, and input the liquid cooling parameter data into the liquid cooling system controller to adjust the liquid cooling cycle.

[0064] In this embodiment, the sensor network refers to a multi-type sensor array covering the surface of the quantum chip and the liquid cooling pipes. Specifically, it can be implemented using a combination of quantum sensors and liquid cooling sensors. The quantum sensors are used to capture the decoherence time of the qubits and the error rate of quantum gate operations, while the liquid cooling sensors are used to monitor the temperature, level, and pressure data of the coolant. The neural network-based dynamic parameter optimization model refers to a hybrid algorithm framework that integrates time series prediction and stochastic simulation. Specifically, it can be implemented using a combination of neural networks and Monte Carlo methods. By analyzing the correlation between the dynamic characteristics of the quantum state and the liquid cooling parameters, optimized parameters that balance error correction efficiency and heat dissipation requirements are generated. The dynamic error correction code controller refers to a hardware module that can reconfigure the topology of the qubits. Specifically, it can be implemented using programmable logic devices. Phase compensation operations are inserted based on the surface code distance adjustment value to optimize error correction performance. The liquid cooling system controller refers to a control unit that supports boiling phase change and pump-free self-circulation. Specifically, it can be implemented using a combination of microfluidic chips and a condenser. Dynamic heat dissipation is achieved by regulating the coolant phase change process.

[0065] In this embodiment, a quantum sensor collects the decoherence time and quantum gate operation error rate of the qubits in real time, while a liquid-cooled sensor simultaneously acquires data on coolant temperature, level, and pressure, forming a multi-dimensional initial dataset. This dataset is input into a neural network-based dynamic parameter optimization model. The time-series neural network first extracts the dynamic features of the qubits and then predicts the quantum gate operation error rate through random simulation. Based on the prediction results, the model classifies the qubits into high-stability and low-stability categories, and accordingly allocates the ratio of physical qubits to auxiliary qubits to generate error correction code parameters. Simultaneously, it combines the real-time liquid cooling status with heat dissipation requirements to generate liquid cooling parameters. The error correction code parameters drive the dynamic error correction code controller to adjust the surface code distance and reconstruct the qubit topology. The liquid cooling parameters control the coolant to undergo a controlled boiling phase transition on the surface of the quantum chip. The gaseous coolant is converted to liquid by a condenser and then self-circulates under gravity.

[0066] In this embodiment, data fusion is achieved through a sensor network, and an AI model is used to simultaneously optimize error correction and heat dissipation parameters. For example, when the decoherence time of a qubit shortens, the model automatically increases the liquid cooling circulation rate and reduces the proportion of auxiliary qubits, thereby reducing system energy consumption while maintaining error correction capability. Furthermore, boiling phase change cooling improves heat transfer efficiency compared to traditional single-phase liquid cooling, while dynamically adjusting the surface code distance avoids the waste of fixed error correction resources. Thus, in scenarios with fluctuating qubit stability, dynamically adjusting the surface code distance can reduce the occupancy rate of auxiliary qubits. Simultaneously, adjusting the coolant boiling intensity based on real-time heat dissipation requirements avoids excessive cooling energy consumption in traditional liquid cooling systems. Moreover, by eliminating redundant modules in the independent control architecture, the overall system energy efficiency ratio is improved, solving the problem of low resource utilization in existing technologies.

[0067] In one feasible implementation, refer to Figure 2 Step S100 includes steps S110 to S130, wherein:

[0068] Step S110: Real-time acquisition of quantum state data through a quantum sensor, and output of the quantum state data, which includes the decoherence time of the qubits and the error rate of the quantum gate operation;

[0069] Step S120: Real-time liquid cooling status data is collected through a liquid cooling sensor, and the liquid cooling status data is output. The liquid cooling status data includes coolant temperature, liquid level and pressure data.

[0070] Step S130: Merge the quantum state data and the liquid-cooled state data to output the initial dataset.

[0071] In this embodiment, the quantum sensor refers to a device used to monitor the dynamic characteristics of qubits. Specifically, it can be implemented using a superconducting quantum interference device (QI) or a Josephson junction sensor. It is used to capture the decoherence time and quantum gate operation error rate of the qubits, thus providing basic data for subsequent error correction code adjustment. The liquid cooling sensor refers to a device used to monitor the operating status of the cooling system. Specifically, it can be implemented using distributed temperature sensors, capacitive level sensors, and piezoresistive pressure sensors. It is used to acquire real-time data on coolant temperature, level, and pressure, thus providing a basis for the dynamic control of the liquid cooling cycle. The initial dataset refers to a structured data set formed by fusing quantum state data and liquid cooling state data. This merging can be achieved through timestamp alignment and data format standardization, providing multi-dimensional input for the neural network-based dynamic parameter optimization model.

[0072] In this embodiment, quantum sensors are deployed in key areas of the quantum chip, such as the gaps between qubit arrays or near control circuitry, to acquire decoherence time and quantum gate operation error rate in real time. Decoherence time is obtained by measuring the relaxation time and dephase time of the qubits, while the quantum gate operation error rate is calculated by statistically analyzing the state deviation after quantum gate operations. Liquid-cooled sensors are integrated into the coolant circulation pipeline or storage tank; for example, temperature sensors are placed at the coolant inlet and outlet, a liquid level sensor is installed on the side wall of the storage tank, and a pressure sensor is embedded in the circulation loop. Quantum state data and liquid-cooled state data are transmitted to the central processing unit through a unified data interface. A synchronous sampling mechanism is used during data merging to ensure the temporal consistency of the two types of data. The resulting initial dataset can simultaneously reflect the dynamic error characteristics of quantum computing and the real-time heat dissipation status of the liquid cooling system.

[0073] In this embodiment, the coordinated deployment of quantum sensors and liquid-cooled sensors enables simultaneous monitoring of multiple parameters of the quantum computing environment and heat dissipation status. Furthermore, by acquiring pressure data in real time, this solution can predict changes in coolant flow resistance in advance, avoiding a decrease in heat dissipation efficiency due to local boiling. Additionally, by acquiring the error rate of quantum gate operations online, the timeliness of error correction code parameter adjustment is significantly improved.

[0074] Through the above technical solution, this application solves the problem of lag in heat dissipation control caused by the single sensor type in single-phase immersion liquid cooling, and overcomes the real-time deficiency of traditional quantum error correction codes that rely on offline error data. The synchronous acquisition and fusion of quantum state data and liquid cooling state data enables the subsequent neural network-based dynamic parameter optimization model to accurately correlate quantum computing errors with heat dissipation fluctuations. For example, when an abnormal rise in coolant temperature is detected, the system can combine the synchronously acquired quantum gate operation error rate data to determine whether it is necessary to prioritize adjusting liquid cooling parameters to suppress decoherence rate, or prioritize optimizing error correction code parameters to compensate for existing computational errors. This reduces system redundancy and avoids resource waste caused by independent heat dissipation and computing architectures.

[0075] In one feasible implementation, refer to Figure 3 Step S200 includes steps S210 to S230, wherein:

[0076] Step S210: Based on the quantum state data and liquid-cooled state data in the initial dataset, process the time series data through a time series neural network model to output the dynamic feature data of the qubit;

[0077] Step S220: Based on the dynamic characteristic data of the qubit, predict the quantum gate operation error rate through a random simulation method, and output the predicted quantum gate operation error rate data;

[0078] Step S230: Based on the predicted quantum gate operation error rate data and quantum state data, generate the optimized parameter dataset.

[0079] In this embodiment, the time-series neural network model refers to an algorithm that uses recurrent neural networks or long short-term memory networks to extract features from time-series data of quantum states and liquid cooling states. Specifically, it can be implemented using a multi-layer gated recurrent unit structure to capture the dynamic correlation between qubit decoherence time and coolant temperature changes. The stochastic simulation method refers to a technique based on Monte Carlo methods or Markov chains to probabilistically model the error rate of quantum gate operations. Specifically, it can be implemented using Bayesian inference combined with noise parameter distribution to quantify the coupled influence of qubit stability and the heat dissipation capacity of the liquid cooling system. The optimized parameter dataset refers to a joint configuration set containing error correction code parameters and liquid cooling parameters. Specifically, it can be implemented using a dynamic programming algorithm to perform multi-objective optimization of qubit classification results and heat dissipation requirements to simultaneously adjust the error correction operation intensity and cooling cycle efficiency.

[0080] In this embodiment, quantum state data and liquid-cooled state data are input into a time-series neural network model. A sliding window mechanism is used to extract temporal features from historical data. For example, cross-modal correlation analysis is performed between the decoherence time of the qubit and the coolant pressure change, outputting a feature vector reflecting the dynamic stability of the qubit. Subsequently, based on the dynamic feature data of the qubit, a probabilistic prediction of the quantum gate operation error rate is made using stochastic simulation methods. For example, a probability distribution model of the error rate changing with time and environment is established based on the qubit's decoherence decay curve and the range of liquid-cooled temperature fluctuations. Finally, by combining the predicted quantum gate operation error rate data and real-time quantum state data, an optimized dataset containing error-correcting code parameters and liquid-cooling parameters is generated. For example, the surface code distance is dynamically adjusted according to the error rate threshold, and the circulation flow rate is matched based on the coolant temperature gradient.

[0081] In this embodiment, the proposed solution utilizes joint modeling of temporal neural networks and stochastic simulations to track the stability changes of qubits in real time, thereby dynamically optimizing error correction code parameters and liquid cooling parameters. Furthermore, by coupling quantum gate error prediction with liquid cooling state data, this solution achieves synergistic optimization of error correction overhead and heat dissipation efficiency. This allows for real-time adjustment of error correction code parameters based on changes in qubit dynamic stability, reducing redundant occupancy of auxiliary qubits. Simultaneously, the linkage control between liquid cooling parameters and error correction operations reduces coolant circulation energy consumption. For example, when the qubit decoherence time shortens due to temperature fluctuations, the system can automatically increase the surface code distance and accelerate coolant circulation, maintaining the accuracy of quantum gate operations while avoiding excessive consumption of physical qubit resources.

[0082] In one feasible implementation, refer to Figure 4 Step S230 includes steps S231 to S234, wherein:

[0083] Step S231: Based on the predicted quantum gate operation error rate data, classify high-stability qubits and low-stability qubits to output qubit classification data;

[0084] Step S232: Based on the quantum bit classification data, allocate the ratio of physical quantum bits to auxiliary quantum bits to output error correction code parameter data;

[0085] Step S233: Based on the liquid cooling status data, match the heat dissipation requirements and output liquid cooling parameter data;

[0086] Step S234: Merge the error correction code parameter data and liquid cooling parameter data to output the optimized parameter dataset.

[0087] In this embodiment, high-stability qubits and low-stability qubits refer to the classification of qubits based on their stability levels according to predicted quantum gate operation error rate data. This can be achieved using threshold comparison or cluster analysis to identify low-stability qubits that require priority allocation of error correction resources. The ratio of physical qubits to auxiliary qubits refers to the dynamic adjustment of the quantity relationship between physical qubits used for computational tasks and auxiliary qubits used for error correction. This can be achieved using a dynamic resource allocation algorithm to reduce redundant occupation of auxiliary qubits while ensuring error correction capability. Matching heat dissipation requirements refers to adjusting the operating parameters of the liquid cooling system based on the coolant temperature, level, and pressure parameters in the liquid cooling status data. This can be achieved using a closed-loop control algorithm or PID control strategy to ensure that the heat dissipation capacity of the liquid cooling system is adapted to the heat load of the quantum chip in real time.

[0088] In this embodiment, during quantum computing, quantum gate operation error rate data is used to evaluate the stability differences of each qubit. By setting an error rate threshold or using unsupervised clustering methods, qubits are divided into high-stability and low-stability categories. For low-stability qubits, the system automatically increases the allocation ratio of auxiliary qubits to enhance error correction capabilities, while reducing the use of auxiliary resources for high-stability qubits. Simultaneously, the liquid cooling system dynamically adjusts the coolant flow rate or the distribution of the phase transition region based on real-time collected parameters such as coolant temperature and pressure, ensuring that the heat dissipation requirements of different areas of the quantum chip are precisely met. The error correction code parameters and liquid cooling parameters are integrated through a data fusion module to form a unified optimized parameter dataset, providing a basis for coordinated control of subsequent error correction operations and liquid cooling cycles.

[0089] Compared to existing technologies, traditional methods employ a fixed ratio of physical to auxiliary qubits, resulting in insufficient error correction resources for low-stability qubits and wasted resources for high-stability qubits. This proposed solution reduces the total number of auxiliary qubits used while maintaining error correction effectiveness through dynamic classification and ratio adjustment. Regarding liquid cooling control, existing technologies rely on preset static heat dissipation parameters, which cannot adapt to the dynamically changing thermal load of the quantum chip during operation. This solution, however, achieves precise matching of heat dissipation efficiency and computational demands through real-time data matching. Through these technical solutions, this application effectively reduces the proportion of auxiliary resources required for quantum error correction operations, avoiding resource redundancy issues caused by fixed ratio allocation. Simultaneously, the liquid cooling system can dynamically adjust its operating parameters according to the actual thermal load of the quantum chip, reducing energy loss during coolant circulation. The synergistic optimization of these two types of parameters further reduces the overall system redundancy and improves the energy efficiency ratio during quantum computing task execution.

[0090] In one feasible implementation, refer to Figure 5 Step S300 includes steps S310 to S320, wherein:

[0091] Step S310: Input the error correction code parameter data into the dynamic error correction code controller and adjust the surface code distance and measurement cycle parameters;

[0092] Step S320: Input the liquid cooling parameter data into the liquid cooling system controller to adjust the liquid cooling cycle.

[0093] In this embodiment, the dynamic error correction code controller refers to a device that adjusts the quantum error correction operation parameters in real time. Specifically, it can be implemented using hardware logic circuits based on FPGA or ASIC, used to dynamically optimize the allocation of error correction resources according to the stability classification results of the qubits. The surface code distance refers to the range of topological correlation between physical qubits and auxiliary qubits in the quantum error correction code. Specifically, it can be achieved by adjusting the encoding scale of the logical qubits, and its value directly affects the balance between error correction capability and resource consumption. The measurement period parameter refers to the synchronous measurement frequency of the auxiliary qubits during quantum error correction. Specifically, it can be controlled by a variable clock signal to control the timing of the measurement circuit, used to match the quantum decoherence rate. The liquid cooling system controller refers to a device that regulates the phase change and circulation of the cooling medium. Specifically, it can be achieved through the coordinated operation of a microfluidic valve array and a pressure sensor to realize pump-free self-circulating heat dissipation. The liquid cooling circulation refers to the process of the coolant boiling and absorbing heat on the surface of the quantum chip and then condensing and flowing back. Specifically, it can be achieved through a gravity-driven two-phase flow path design to reduce system complexity.

[0094] In this embodiment, after the surface code distance adjustment value in the error correction code parameter data is input into the dynamic error correction code controller, the physical layout topology of the qubits is reconfigured. For example, the surface code distance in the low-stability qubit region is increased to enhance error correction capability, while the distance in the stable region is reduced to decrease auxiliary bit occupancy. The measurement period parameter is dynamically matched according to the real-time predicted value of the quantum gate operation error rate. For example, the measurement interval is shortened when the decoherence time is shortened to improve the real-time performance of error correction. After the liquid cooling parameter data is input into the liquid cooling system controller, the boiling phase transition region of the coolant is precisely controlled by a microfluidic valve. For example, the boiling intensity is increased in the hot spot region of the quantum chip to improve local heat dissipation efficiency. At the same time, the condenser reflux rate is automatically adjusted according to the pressure data to maintain the stability of the pump-free circulation.

[0095] It is understandable that traditional static error correction code schemes use fixed surface code distances and measurement periods, leading to redundant error correction resources being wasted in high-stability qubit regions. Dynamic adjustment mechanisms, however, can differentiate parameters based on qubit classification results, reducing the overall occupancy of auxiliary qubits. Existing single-phase immersion liquid cooling relies on mechanical pumps to drive circulation, which not only increases packaging complexity but also suffers from flow lag due to coolant viscosity. In contrast, a pump-free self-circulating scheme based on phase change control, driven by gravity and pressure, can reduce system energy consumption and improve thermal response speed. Thus, this application can optimize error correction resource allocation based on the dynamic characteristics of qubits, reduce the proportion of auxiliary qubits, and simultaneously reduce the packaging difficulty of quantum chips through phase change-driven liquid cooling circulation, achieving a simultaneous improvement in heat dissipation efficiency and system energy efficiency.

[0096] In one feasible implementation, refer to Figure 6 Step S310 includes steps S311 to S313, wherein:

[0097] Step S311: Based on the error correction code parameter data, determine the adjustment value of the surface code distance;

[0098] Step S312: Reconstruct the topology of the qubit based on the adjustment value;

[0099] Step S313: Insert a phase compensation operation based on the reconstructed topology to optimize quantum error correction.

[0100] In this embodiment, the surface code distance refers to the dimensional parameter of the two-dimensional planar structure composed of physical qubits in the error-correcting code. Specifically, it can be achieved by adjusting the arrangement ratio of physical qubits to auxiliary qubits, for example, adjusting the surface code distance from 5 to 7 to enhance error correction capability. The qubit topology refers to the connection relationship between physical qubits and auxiliary qubits on the quantum chip. Specifically, it can be achieved by reconfiguring the coupling links between qubits, for example, rearranging the originally dispersed auxiliary qubits into a ring structure. The phase compensation operation refers to the correction operation performed on phase errors caused by changes in topology during quantum gate operations. Specifically, it can be achieved by inserting a single-qubit rotation gate at a specific angle or adjusting the control parameters of a two-qubit gate, for example, adding a Z-axis rotation operation between adjacent qubits to eliminate phase shift.

[0101] In this embodiment, the dynamic error correction code controller first analyzes the logical qubit stability index and physical qubit error rate in the error correction code parameter data, and determines the adjustment value of the surface code distance according to a predefined threshold rule. For example, when the error rate of quantum gate operation is detected to exceed the threshold, the surface code distance is increased from the initial value of 5 to 7 to enhance the error correction capability. Subsequently, based on the adjusted surface code distance, the qubit topology is reconstructed by reallocating the positions of physical qubits and auxiliary qubits. For example, two of the eight physical qubits originally used for computation are converted into auxiliary qubits and rearranged into a honeycomb structure. Finally, in the reconstructed topology, phase compensation operations are automatically inserted according to the changes in the coupling strength between qubits. For example, controlled phase gates are added between adjacent qubits to offset the quantum state phase deviation caused by topology reconstruction, thereby optimizing the fault tolerance performance of the error correction code.

[0102] In this embodiment, by dynamically adjusting the surface code distance and topology, the number of auxiliary qubits can be reduced while maintaining error correction effectiveness. Furthermore, by inserting phase compensation operations, quantum state distortion caused by structural changes is effectively eliminated, achieving dynamic optimization of quantum error correction resources and reducing the number of physical qubits required for the error-correcting code. Simultaneously, the phase compensation operation improves the accuracy of the error correction operation. Specifically, the dynamic adjustment of the surface code distance allows the system to adaptively balance error correction capability and resource overhead based on the real-time error rate, while the phase compensation operation after topology reconstruction reduces the accumulation of quantum state distortion caused by structural changes, thereby improving the overall reliability and error correction efficiency of the quantum computing process.

[0103] In one feasible implementation, refer to Figure 7 Step S320 includes steps S321 to S323, wherein:

[0104] Step S321: Based on the liquid cooling parameter data, control the coolant to boil and undergo a phase transition on the surface of the quantum chip;

[0105] Step S322: Based on the liquid cooling parameter data, the gaseous coolant is converted into liquid reflux through the condenser;

[0106] Step S323: Based on the liquid cooling parameter data and gravity, a pump-free self-circulation is achieved.

[0107] In this embodiment, the boiling phase change of the coolant on the quantum chip surface refers to the absorption of heat through the phase change process from liquid to gas on the quantum chip surface. This can be achieved using low-boiling-point coolants such as fluorinated liquids or liquid nitrogen, thereby increasing the heat absorption efficiency per unit volume of coolant. The condenser converts the gaseous coolant into a liquid reflux state through a heat exchange device, specifically a finned condenser or a microchannel condenser. This phase change circulation enables efficient recycling of the coolant. Pump-free self-circulation refers to coolant circulation without relying on a mechanical pump. This can be achieved through gravity-driven circulation combined with natural convection formed by the density difference between the two phases, reducing system vibration and electromagnetic interference by eliminating the mechanical pump.

[0108] In this embodiment, when the liquid cooling system controller receives liquid cooling parameter data, it first adjusts the coolant flow rate and pressure based on the real-time temperature data of the quantum chip, causing the coolant to reach the critical boiling condition on the surface of the quantum chip. During the boiling phase transition, the coolant absorbs the heat generated by the quantum chip and transforms into a gaseous state. Subsequently, the gaseous coolant enters the condenser through pipes. The condenser exchanges heat with an external cooling medium, such as circulating water or air, to re-condense the gaseous coolant into a liquid state. The liquid coolant naturally flows back to the surface of the quantum chip under the action of gravity, forming a closed-loop circulation system without mechanical pump drive. This circulation process achieves dynamic matching between heat dissipation capacity and the thermal load of the quantum chip by precisely controlling the boiling phase transition point and condensation efficiency.

[0109] Compared to existing technologies, traditional single-phase immersion liquid cooling relies on forced circulation of high-viscosity coolant, requiring complex sealing structures and mechanical pump drives, which are prone to vibration interference and electromagnetic noise. This solution improves heat dissipation efficiency through two-phase boiling phase change, employs a pumpless self-circulating design to eliminate mechanical vibration sources, and simplifies the coolant circulation path. Existing independent cooling systems require additional circulation pumps and buffer devices, resulting in high system redundancy. This solution achieves a compact heat dissipation architecture through the synergistic effect of gravity drive and phase change circulation. Through these technical solutions, this application effectively reduces the packaging difficulty of quantum chips, avoids the wetting effect of high-viscosity coolant on precision quantum devices, and improves heat dissipation efficiency per unit area through phase change cooling. The pumpless self-circulating design reduces system vibration interference sources, which is beneficial for maintaining the coherence time of qubits. Dynamically matched heat dissipation control can adapt to transient thermal load changes in the quantum chip, avoiding increased quantum gate operation error rates caused by localized overheating.

[0110] In one feasible implementation, refer to Figure 8 The implementation method also includes steps S400 to S500, wherein:

[0111] Step S400: Measure the quantum gate operation results using quantum state tomography and output quantum gate fidelity data;

[0112] Step S500: Input the quantum gate fidelity data into the neural network-based dynamic parameter optimization model and iteratively update the optimization parameter dataset.

[0113] In this embodiment, quantum state tomography refers to reconstructing the complete information of quantum gate operation results by measuring the quantum state density matrix. Specifically, it can be implemented by processing the projection data under the measurement basis using the maximum likelihood estimation algorithm, which is used to quantify the accuracy of quantum gate operations. The dynamic parameter optimization model based on a neural network can be implemented using a time-series neural network combined with a gradient descent algorithm, used to adjust the error correction code parameters and liquid cooling parameters based on real-time feedback.

[0114] In this embodiment, after the quantum gate operation is completed, quantum state tomography is used to obtain the state distribution data of the qubits. A fidelity index is generated by calculating the difference between the actual and theoretical states. This fidelity data is transmitted to a neural network-based dynamic parameter optimization model as a constraint condition for the objective function during model training. The model generates a new set of optimized parameters based on the current fidelity and historical data, such as adjusting the surface code distance or coolant flow rate parameters, and synchronizes the updated parameters to the dynamic error correction code controller and the liquid cooling system controller. This process is executed cyclically at a preset period, forming a closed-loop feedback mechanism.

[0115] In this embodiment, the proposed solution uses real-time measurement of fidelity data to drive the AI ​​model to dynamically optimize parameters, improving the matching degree between error correction resource allocation and heat dissipation requirements. This achieves synergistic optimization of quantum error correction operations and the liquid cooling system, solving the problems of high error correction overhead and high system redundancy in traditional solutions. Through a closed-loop feedback mechanism, the utilization rate of qubit error correction resources is improved, while the energy consumption of the liquid cooling system is dynamically controlled within the minimum required range.

[0116] In one feasible implementation, refer to Figure 9 Step S400 includes steps S410 to S420, wherein:

[0117] Step S410: Obtain quantum state data after quantum gate operation;

[0118] Step S420: Based on the quantum state data, generate quantum gate fidelity data using a quantum state tomography algorithm.

[0119] In this embodiment, quantum state tomography refers to a method of reconstructing the quantum state density matrix by measuring multiple projections of a quantum system. Specifically, it can be implemented using maximum likelihood estimation or Bayesian inference algorithms, and is used to accurately quantify state deviations after quantum gate operations. Quantum gate fidelity data refers to a quantitative index characterizing the similarity between actual quantum gate operations and ideal operations. Specifically, it can be obtained by calculating the overlap integral between the actual quantum state and the target quantum state, and is used to assess the degree of error accumulation during quantum computing.

[0120] In this embodiment, after the quantum gate operation is completed, the state information of the qubit is captured by the sensor network to form quantum state data. This data is input into the quantum state tomography algorithm module, which reconstructs the density matrix of the quantum state by solving a system of linear equations. Compressed sensing technology is used during the reconstruction process to reduce the number of measurements; for example, data reconstruction is completed by combining random basis vector measurements with sparsity constraints. The generated quantum gate fidelity data is fed back to a neural network-based dynamic parameter optimization model, triggering dynamic updates to the optimization parameter dataset.

[0121] In some specific implementations, the quantum state tomography algorithm can employ a neural network-based variational quantum state estimation method, which approximates the true quantum state distribution by training parameterized quantum circuits. The selection of measurement basis vectors can be dynamically adjusted through optimization algorithms, for example, selecting the optimal measurement sequence based on the decoherence time of the current qubit.

[0122] Compared to existing technologies, traditional methods rely on fixed-period quantum process tomography, which consumes significant computational resources and lacks real-time feedback. This solution combines quantum state tomography with a dynamic optimization system, establishing closed-loop control through online-generated fidelity data to reduce the time spent by auxiliary qubits while maintaining measurement accuracy. Through this technical solution, this application achieves real-time monitoring and quantitative evaluation of quantum gate operation quality, providing precise error distribution information for adjusting dynamic error correction code parameters. This technique effectively solves the resource waste problem caused by error feedback delays in traditional static error correction schemes, reducing the proportion of auxiliary qubits used through a closed-loop optimization mechanism while improving the synergistic efficiency of the liquid cooling system and error correction operations.

[0123] In the embodiments of this application, the liquid-cooled quantum computing server optimization method acquires quantum state and liquid cooling state data in real time, uses an AI model to dynamically generate optimization parameters, and coordinates the quantum error correction operation and the liquid cooling circulation system. This solves the technical problems of low heat dissipation efficiency and high error correction resource consumption in traditional solutions, and improves system stability and computing efficiency.

[0124] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the liquid-cooled quantum computing server optimization method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0125] This application also provides an optimized system for a liquid-cooled quantum computing server, referenced in [reference]. Figure 10 The liquid-cooled quantum computing server optimization system includes: a memory 10, a processor 20, and a liquid-cooled quantum computing server optimization program stored on the memory 10 and executable on the processor 20. The liquid-cooled quantum computing server optimization program is configured to implement the steps of the liquid-cooled quantum computing server optimization method.

[0126] The liquid-cooled quantum computing server optimization system provided in this application, employing the liquid-cooled quantum computing server optimization method described in the above embodiments, can improve the heat dissipation efficiency and error correction real-time performance of the quantum computing server, as well as reduce system resource consumption. Compared with the prior art, the beneficial effects of the liquid-cooled quantum computing server optimization system provided in this application are the same as those of the liquid-cooled quantum computing server optimization method provided in the above embodiments, and other technical features of the liquid-cooled quantum computing server optimization system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0127] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0128] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. An optimization method for a liquid-cooled quantum computing server, characterized in that, The method includes: An initial dataset is acquired by a sensor network deployed on a quantum chip, the initial dataset including quantum state data and liquid-cooled state data; Based on the initial dataset, it is processed by a neural network-based dynamic parameter optimization model to obtain an optimized parameter dataset, which includes error correction code parameter data and liquid cooling parameter data. The error correction code parameter data is input into the dynamic error correction code controller to adjust the quantum error correction operation, and the liquid cooling parameter data is input into the liquid cooling system controller to adjust the liquid cooling cycle; The step of processing the initial dataset using a neural network-based dynamic parameter optimization model to obtain an optimized parameter dataset includes: Based on the quantum state data and liquid-cooled state data in the initial dataset, the time series data is processed by a time series neural network model to output the dynamic characteristic data of the qubit. Based on the dynamic characteristic data of the qubit, the quantum gate operation error rate is predicted by a random simulation method, and the predicted quantum gate operation error rate data is output. Based on the predicted quantum gate operation error rate data and quantum state data, the optimized parameter dataset is generated; The step of generating the optimized parameter dataset based on the predicted quantum gate operation error rate data and quantum state data includes: Based on the predicted quantum gate operation error rate data, quantum bit classification data is output; the quantum bit classification data specifically includes high-stability quantum bits and low-stability quantum bits. The ratio of physical qubits to auxiliary qubits is determined based on the qubit classification data to output error correction code parameter data; Based on the liquid cooling status data, the heat dissipation requirements are matched, and liquid cooling parameter data is output. The error correction code parameter data and liquid cooling parameter data are merged to output the optimized parameter dataset.

2. The liquid-cooled quantum computing server optimization method as described in claim 1, characterized in that, The step of acquiring the initial dataset through a sensor network deployed on a quantum chip includes: Quantum state data is acquired in real time by a quantum sensor and output. The quantum state data includes the decoherence time of the qubit and the error rate of the quantum gate operation. The liquid cooling status data is collected in real time by a liquid cooling sensor and output. The liquid cooling status data includes coolant temperature, liquid level and pressure data. The quantum state data and the liquid-cooled state data are combined to output the initial dataset.

3. The liquid-cooled quantum computing server optimization method as described in claim 1, characterized in that, The steps of inputting the error correction code parameter data into the dynamic error correction code controller to adjust the quantum error correction operation, and inputting the liquid cooling parameter data into the liquid cooling system controller to adjust the liquid cooling cycle include: Input the error correction code parameter data into the dynamic error correction code controller, and adjust the surface code distance and measurement cycle parameters; The liquid cooling parameter data is input into the liquid cooling system controller to adjust the liquid cooling cycle.

4. The liquid-cooled quantum computing server optimization method as described in claim 3, characterized in that, The step of inputting the error correction code parameter data into the dynamic error correction code controller and adjusting the surface code distance and measurement cycle parameters includes: Based on the error correction code parameter data, determine the adjustment value for the surface code distance; Reconstruct the topology of the qubit based on the adjusted values; Phase compensation operations are inserted based on the reconstructed topology to optimize quantum error correction.

5. The liquid-cooled quantum computing server optimization method as described in claim 3, characterized in that, The steps of inputting the liquid cooling parameter data into the liquid cooling system controller to adjust the liquid cooling cycle include: Based on the liquid cooling parameter data, the boiling phase transition of the coolant on the surface of the quantum chip is controlled. Based on the liquid cooling parameter data, the gaseous coolant is converted into liquid reflux through the condenser; Based on the liquid cooling parameter data and gravity, a pump-free self-circulation system can be achieved.

6. The liquid-cooled quantum computing server optimization method as described in claim 1, characterized in that, The method further includes: Quantum gate operation results are measured using quantum state tomography, and quantum gate fidelity data is output. The quantum gate fidelity data is input into the neural network-based dynamic parameter optimization model to iteratively update the optimization parameter dataset.

7. The liquid-cooled quantum computing server optimization method as described in claim 6, characterized in that, The step of measuring the quantum gate operation results using quantum state tomography and outputting quantum gate fidelity data includes: Obtain quantum state data after quantum gate operations; Based on the quantum state data, quantum gate fidelity data is generated using a quantum state tomography algorithm.

8. A liquid-cooled quantum computing server optimization system, characterized in that, The liquid-cooled quantum computing server optimization system includes: a memory, a processor, and a liquid-cooled quantum computing server optimization program stored in the memory and executable on the processor, wherein the liquid-cooled quantum computing server optimization program is configured to implement the steps of the liquid-cooled quantum computing server optimization method as described in any one of claims 1 to 7.

Images