Anti-error quantum compiler

By compiling quantum algorithms on quantum hardware and applying deterministic error correction protocols, the problem of error correction on multi-qubit quantum circuits is solved, and the stability and efficiency of quantum computing are improved.

CN119072698APending Publication Date: 2024-12-03Q CTRL PTY LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380034369.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-03-11
Filing Date
2023-02-28
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing quantum computers encounter difficulties in applying dynamic error suppression technology on multi-qubit quantum circuits, making it difficult to effectively correct errors without consuming additional resources, especially on the entire quantum circuit of multiple qubits.

Method used

By compiling quantum algorithms on quantum hardware using multiple qubits, determining the quantum relationship between qubits and between qubits and the external environment, applying deterministic error correction protocols specific to quantum circuits and hardware, optimizing quantum gates parameters to maximize fidelity and reduce errors in measurement results through multiple measurements and error characterization.

Benefits of technology

Effective error correction on multi-qubit quantum circuits is achieved, which improves the success probability of quantum algorithms and the stability of quantum hardware, reduces the error rate, and avoids the consumption of additional resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119072698A_ABST
    Figure CN119072698A_ABST
Patent Text Reader

Abstract

A method for generating an error-resistant quantum circuit according to a quantum algorithm is provided herein. The method determines characteristics of quantum hardware used to implement the quantum algorithm and informs the deterministic dynamic error suppression routine to implement. The method also takes into account a compiled quantum circuit, as the compiled quantum circuit provides information about the context of operations performed on the qubits, which is an important factor of the fidelity of the quantum gate and the final success of the algorithm. The method then applies deterministic error correction using the characteristics and the context.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Australian Provisional Patent Application No. 2022900597 filed on March 11, 2022, the contents of which are incorporated herein by reference in their entirety. Technical Field

[0003] The present disclosure relates to compiling quantum algorithms onto quantum hardware using error correction. Background Art

[0004] Quantum computers offer tremendous opportunities to solve problems that are intractable with existing classical hardware based on binary bits. Quantum computers use quantum bits ("qubits"), which can maintain a superposition of states and can become entangled.

[0005] However, a major obstacle on the path to widespread use of quantum computers is that qubits are relatively unstable, meaning that the quantum state of the qubit is lost too quickly for most practical calculations. Quantum error correction codes, such as surface codes, consume additional resources to prevent errors and use more qubits than are currently available, making them difficult to implement.

[0006] Dynamic error suppression is another method to stabilize quantum states on qubits. This involves a separate operation that is effective in error correction without the need for additional qubits. When it involves idle qubits (memory), it can be called dynamic decoupling, or when applied to the active manipulation of one or more qubits, it is called a dynamically corrected gate. Dynamic error suppression is deterministic because it does not require additional resources, such as having to combine multiple measurement runs to gain benefits. Although dynamic error suppression can be successfully performed on a single qubit or a pair of isolated qubits, it is difficult to apply dynamic error suppression on an entire quantum circuit with multiple qubits implementing a quantum algorithm.

[0007] Any discussion of documents, acts, materials, devices, articles of manufacture or the like contained in this specification is not to be taken as an admission that any or all of these matters formed part of the prior art base or were common general knowledge in the field relevant to the present disclosure as they existed before the priority date of each of the appended claims.

[0008] Throughout this specification, the word "comprise" or variations such as "comprises" or "comprising" should be understood to imply the inclusion of stated elements, integers or steps or groups of elements, integers or steps, but not the exclusion of any other elements, integers or steps or groups of elements, integers or steps. Summary of the invention

[0009] Methods for generating error-resistant quantum circuits from quantum algorithms are provided herein. The method determines characteristics of the quantum hardware used to implement the quantum algorithm and informs a deterministic dynamic error suppression routine to be implemented. The method also takes into account compiled quantum circuits because the compiled quantum circuits provide information about the context of operations performed on qubits, which is an important factor in the fidelity of quantum gates and the ultimate success of the algorithm. The method then uses the characteristics and the context to apply deterministic error correction.

[0010] A method for generating an error-corrected quantum circuit comprises:

[0011] Compiling quantum algorithms onto quantum hardware using multiple qubits to generate quantum circuits;

[0012] determining quantum relationships between the plurality of qubits and between the plurality of qubits and an external environment as a result of implementing the quantum circuit on the quantum hardware; and

[0013] The quantum relationships between the plurality of qubits and between the plurality of qubits and an external environment are used to determine a deterministic error correction protocol specific to the quantum circuit and specific to the quantum hardware for each qubit in the plurality of qubits and for interactions between the plurality of qubits.

[0014] In some embodiments, the method further comprises: performing a plurality of measurements on qubits of the quantum hardware to determine an error representation of the quantum hardware based on a topology of the quantum hardware.

[0015] In some embodiments, based on the topology of the quantum hardware, the error characterization includes one or more of the following: individual gate errors, decoherence processes, measurement result errors, and crosstalk between qubits.

[0016] In some embodiments, compiling includes, for each of a plurality of logical quantum operations in the quantum algorithm, selecting one or more quantum blocks to implement the quantum algorithm on the quantum hardware, wherein each of the quantum blocks implements a logical function of the logical quantum operation.

[0017] In some embodiments, the method further comprises determining an error correction protocol for each of the quantum blocks, the error correction protocol comprising error correction for errors identified in the error characterization and being specific to the one or more quantum blocks selected to implement the quantum algorithm.

[0018] In some embodiments, the method further comprises executing the quantum algorithm on the quantum hardware using the selected quantum block and the deterministic error correction protocol to obtain a qubit measurement result.

[0019] In some embodiments, the method further comprises reducing a measurement error in the qubit measurement by applying an error reduction operation based on the error characterization.

[0020] A method for characterizing errors in quantum hardware comprising a plurality of qubits, the method comprising: determining groups of qubits based on a topology of the quantum hardware to minimize inter-group correlations; performing a plurality of measurements to approximate a confusion matrix for individual qubit readings; and repeating the plurality of measurements for the groups of qubits to determine the error characterization of the hardware.

[0021] In some embodiments, the groups of qubits include qubit pairs having low cross coupling between the pairs.

[0022] In some embodiments, the error characterization includes one or more of individual gate fidelity, noise spectrum, and crosstalk strength.

[0023] In some embodiments, the method further comprises automating the determination of the error characterization using machine learning.

[0024] In some embodiments, the method further comprises using the qubit to optimize parameters of a quantum gate to maximize the fidelity of the quantum gate.

[0025] In some embodiments, fidelity is maximized for gates having qubits in the set.

[0026] A method for compiling a quantum algorithm onto quantum hardware having a local gate set comprises repeatedly performing the following steps:

[0027] executing a first compilation process to map the quantum algorithm onto the quantum hardware and the local gate set; and

[0028] An output of the first compilation process is used as an input for performing a second compilation process different from the first compilation process to maximize usage of low-error qubits as identified by the error characterization process.

[0029] A method for quantum error correction for quantum hardware includes:

[0030] dividing qubits of the quantum hardware into groups based on the topology of the quantum hardware;

[0031] partitioning the qubits into blocks based on quantum circuits compiled onto the quantum hardware;

[0032] applying driven single-qubit operations depending on the quantum circuit compiled onto the quantum hardware to correct quantum errors arising during identity operations; and

[0033] A time shift of the driven single qubit operation is determined to correct cross-coupling of each of the groups and each of the blocks during an identity operation and to suppress decoherence.

[0034] In some embodiments, the method further includes redefining the non-identity operation by time-domain modulation of a control signal for implementing the non-identity operation to correct the error process identified by the error characterization.

[0035] In some embodiments, the method further comprises modulating the control signal among the plurality of qubits in time in a continuous or discontinuous manner.

[0036] In some embodiments, the method further comprises assembling the block from a library of error-corrected quantum gates based on the error characterization.

[0037] A method for error correction compilation to correct error sources identified during an error characterization process includes:

[0038] receiving a compiled circuit compatible with a hardware device topology and gate set;

[0039] replacing quantum logic gates in the compiled circuit having uncorrected errors exceeding an acceptable threshold with error-corrected alternatives derived using the error characterization process; and

[0040] The definition of the circuit is output in a format for hardware execution with all alternative gate definitions, including identity and non-identity operations.

[0041] In some embodiments, the method further comprises:

[0042] Identify error-prone non-identity gates;

[0043] Replace error-prone non-identity gates with automatically calibrated gates;

[0044] replacing an identity gate in the compiled quantum circuit with an error-corrected identity gate that includes an error correction protocol;

[0045] An output error-robust circuit including the autocalibrated gates and the error-corrected identity gates is assembled into a circuit representation for execution on the quantum hardware.

[0046] A method for correcting errors in quantum measurement results of a quantum circuit on quantum hardware comprises:

[0047] receiving measurement results;

[0048] shrinking a tensor based on the group of qubits using an approximate confusion matrix from an error representation;

[0049] converting the tensor into a probability distribution; and

[0050] correcting the error in the quantum measurement result as indicated by the probability distribution; and

[0051] Output measurement result-error-correction result.

[0052] In some embodiments, the method further comprises: partitioning the quantum circuit into qubit groups based on a topology of the quantum hardware to minimize inter-group correlations; and rearranging the ordering of the tensor based on the qubit groups.

[0053] In some embodiments, the method further comprises adding a dimension to the tensor for unmeasured qubits.

[0054] In some embodiments, the method further includes applying a neural network to the measurements based on the error representation to recover correlations ignored by the tensor relief.

[0055] An Internet server is configured to:

[0056] providing a network interface comprising an input element for inputting a representation of a quantum algorithm and quantum hardware;

[0057] receiving, via the network interface, an incomplete description of the quantum algorithm and the quantum hardware;

[0058] Performing error characterization or retrieving error characterization previously performed for target quantum hardware;

[0059] generating a decorated quantum circuit description that implements the quantum algorithm on the quantum hardware, wherein the decorated quantum circuit description includes quantum logic blocks that perform error correction based on the representation;

[0060] outputting the quantum circuit description; and

[0061] Raw output data is received from the quantum hardware and analyzed using a complete characterization.

[0062] In some embodiments, outputting the quantum circuit description comprises:

[0063] The decorated quantum circuit description is sent to a user through the network interface; or the decorated quantum circuit description is sent to the quantum hardware to be executed on the quantum hardware.

[0064] In some embodiments, receiving and analyzing the raw output data includes: receiving the raw output data from the user through the network interface for analysis; or receiving the raw output data from the quantum hardware for analysis.

[0065] Optional features provided above with reference to one of the methods or Internet servers are equivalent optional features of the other methods and Internet servers. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 Dynamically decoupled (DD) patterns are demonstrated. (a, b) Two patterns that achieve DD for a single qubit are shown. As long as the idle time between two bit flips is equal to the remaining idle time, the effects of dephasing are suppressed. (c) Since each bit flip reverses the ZZ effect, applying both flips simultaneously leaves the ZZ coupling unchanged. Therefore, while the pattern in (c) suppresses the T2 effect on both qubits, it does not mitigate the ZZ effect. (d) Combining the two patterns in (a) and (b) achieves ZZ mitigation. Since the X-gate (bit flip) is not instantaneous, the asymmetry parameter δ can be optimized for optimal ZZ cancellation.

[0067] Figure 2 An example of a compiled circuit is shown. On the left is a snapshot of a 7-qubit algorithm that has been compiled to native gates available on this device. The disclosed method takes as input the compiled circuit and the device connectivity, as shown on the right, and returns a decorated circuit with the optimal DD embedded in it.

[0068] Figure 3 An example of DD embedding is provided. The input circuit at the top is the compiled circuit, and the bottom is the output circuit of the disclosed automated scheme, located in the original circuit, in which the optimal DD structure is embedded. The embedded DD structure includes X operations at a specific time as calculated by an error correction protocol specific to the input circuit and quantum hardware. It is worth noting that this figure has been simplified for ease of explanation, which means that the time (horizontal) location of the DD operations may not be accurate.

[0069] Figure 4 Examples of groups of four qubits per group are shown, which minimizes inter-group correlations in the measurement result readout.

[0070] Figure 5 The success probability of the disclosed method is shown. A certain number of qubits are prepared in the |1> state, as given on the x-axis, and the qubits are measured. The graph shows the success probability of measuring the correct bit string. The shallow circuit depth ensures that decoherence does not affect the success probability. Due to readout errors, the success probability decreases rapidly with increasing number of qubits. However, the disclosed tensor mitigation protocol appropriately mitigates the readout errors.

[0071] Figure 6 An example of a neural network architecture for error correction of measurements on a 7-qubit quantum computer is shown.

[0072] Figure 7 A comparison of the error in measurements (represented by the Hellinger loss) averaged across 2000 random probability distributions on a 7-qubit quantum computer is shown.

[0073] Figure 8a and 8b Different compilation methods using QFT circuits are compared. Method A is a random method, therefore, 50 instances of the method are presented in the figure. The total number of CX gates and the total circuit duration are used as comparative features. As shown, the hybrid method generally finds shorter circuits with a lower number of entangled gates. Although this figure does not prove it, the hybrid method is better at avoiding low-quality qubits. In the above case, the SPAM and CX errors for Q3 are above the device average. For circuits with less than 11 qubits, the hybrid method completely avoids this qubit. As a comparison, for a 10-qubit QFT, 70% of the circuits produced by method A contain problematic qubits.

[0074] Fig. 9 We show that we can partition the quantum circuit so that full device calibration can be done in parallel, which requires a fixed number of steps based on the device connectivity. The heavy hexagonal topology can be tuned in just four steps, regardless of the device size.

[0075] Fig.10 The success probability of the Bernstein-Vazirani algorithm on a 16-qubit device is shown. The black data corresponds to the best compilation available from the hardware provider, and the grey data has a compilation similar to the black data, as well as the latest DD scheme and measurement error mitigation available from the hardware provider. The purple data shows the performance using the software tools in this document.

[0076] Fig.11 Methods for generating error-corrected quantum circuits are demonstrated.

[0077] Fig.12 Methods for characterizing errors in quantum hardware comprising multiple qubits are demonstrated.

[0078] Fig.13 Methods for compiling quantum algorithms onto quantum hardware with local gate sets are demonstrated.

[0079] Fig.14 Demonstrated methods for quantum error correction for quantum hardware.

[0080] Fig.15 A method for error correction compilation is presented.

[0081] Fig.16 A method for correcting errors in quantum measurement results is demonstrated.

[0082] Fig.17 Demonstrates an Internet server. DETAILED DESCRIPTION

[0083] As described above, the methods disclosed herein relate to quantum algorithms implemented by multiple qubits on specific quantum hardware. Unlike classical digital computers, where electronic properties are abstracted into stable binary bit values, quantum hardware cannot provide this level of abstraction without producing unrealistic inaccuracies. This means that each quantum hardware has significantly different properties in terms of qubits. Specifically, for each different quantum hardware, and even for each circuit running on the hardware, the relationship between qubits and the relationship between qubits and the environment are different. Therefore, hardware-agnostic error correction will lead to unsatisfactory results. Further, error correction that does not take into account the compiled circuit structure will lead to high error rates.

[0084] Dynamic error suppression can be used to stabilize individual qubits or improve the fidelity of quantum logic gates that isolate a pair of qubits. An example protocol involves stabilizing qubits during idle time periods - this is called dynamic decoupling. During this time period, additional rotations can be applied to the qubit state to eliminate accumulated errors. However, in multi-qubit quantum circuits, the idle time of multiple qubits is not the same, and the qubits also interact with each other. Therefore, it is difficult to apply dynamic decoupling alone, and error correction depends on hardware characteristics and compiled circuits.

[0085] Compilation in this context means mapping a quantum algorithm onto quantum hardware, taking into account the available physical operations and connectivity between different qubits. Such quantum hardware typically has a local set of gates that may be specific to the quantum hardware, so the compiled quantum circuit uses the available set of gates, taking into account the possible introduction of additional gates. Compilation may also involve optimization of quantum circuits, for example reducing the number of gates by exploiting mathematical relationships between gates.

[0086] Similar to algorithms on classical computers, quantum algorithms are step-by-step procedures where each step can be executed on a quantum computer. Each step is an operation on the quantum state in the algorithm. During compilation, these steps are mapped to qubit operations, called gates. Examples of quantum algorithms include Shor's algorithm, variational quantum eigensolvers, Grover's algorithm, and even quantum error correction, but many other quantum algorithms can be implemented using the methods disclosed herein. Note that the methods disclosed herein are suitable for a variety of different qubit types, including nitrogen vacancies, trapped ions, silicon qubits, superconducting qubits, etc.

[0087] It should also be noted that the methods disclosed herein are performed by a classical computer, that is, a classical digital processor as the execution software performs the compilation, processes the error measurements, optimizes and selects gates, determines the error correction protocol, and processes the output measurements, etc.

[0088] The usefulness of a quantum computer ultimately depends on its ability to successfully implement meaningful quantum algorithms. An ideal implementation of a quantum algorithm is not possible on current quantum hardware, however, being able to run the algorithm with a high probability of success may provide a route to quantum advantage in the near future. The quality of a quantum algorithm is not only affected by the quality of individual building blocks (e.g., the quality of qubits, gates, and measurement results), but also by global device and algorithmic characteristics (such as device topology, multi-qubit correlations, and algorithmic structure). The present disclosure provides an automated method for optimizing different components that ultimately affect the quality of algorithm implementation on a given quantum hardware, and incorporates deterministic error suppression techniques. The disclosed method enables improvements across different algorithms and quantum devices. SUMMARY OF THE INVENTION

[0090] The disclosed protocol is a program for enhancing the performance of quantum algorithms on quantum hardware based on a deterministic error correction method. Once the user provides a quantum circuit through a website or an application programming interface (API), the disclosed protocol pre-performs a series of optimization and error correction steps and outputs the measured results of the quantum algorithm.

[0091] The workflow of the protocol is as follows −

[0092] ● Accept a circuit - Given a quantum circuit, first convert it into a suitable intermediate representation. This enables a robust and hardware-agnostic compiler.

[0093] • Front-end compilation - a series of compiler passes are used to reduce the depth (gate count) of the quantum circuit.

[0094] ● Hardware-specific back-end compilation - logical operations in the circuit are mapped to native gates available on the hardware. Additional hardware-specific compilation passes are performed to account for device topology, gate errors, parallel gate crosstalk, etc.

[0095] ● Reduce Idle and Crosstalk Errors by DD Decoration - Dynamic Decoupling (DD) sequences are incorporated at the algorithm level to mitigate various idle errors, including phase shift and ZZ crosstalk, based on the characteristics of the device and circuit returned by the compiler. This disclosure is about how to automatically incorporate the best DD sequence into the algorithm for execution, rather than the specific sequence design.

[0096] ● Gate error reduction through full-system AI-driven automated calibration - AI-driven optimizers are used to perform pulse-level gate calibration throughout the quantum computer, including optimizing multiple gates in parallel. These optimized pulses are used to execute higher fidelity gates used in the circuit. There are advantages to using AI to perform full-system calibration and gate optimization automation. The process can be automated prior to user interaction. In one example, all relevant error-prone gates are replaced with these optimized gates.

[0097] ●Execute the algorithm on the quantum hardware. The final definition of the circuit is output in a format compatible with hardware execution. This can be achieved by returning "new" commands to the user for execution, or by connecting directly to the hardware backend (locally or in the cloud).

[0098] ● Reduce measurement errors through efficient AI-enhanced readout error mitigation - The measured probability distribution according to the algorithm executed on the quantum device contains measurement errors. These errors are corrected in post-processing using an efficient AI-enhanced measurement error mitigation protocol. This disclosure provides a method for efficient error suppression as part of the overall workflow.

[0099] ● Return solutions with increased likelihood of algorithm success - Returns an "answer" from a quantum algorithm execution with increased confidence in success.

[0100] These processes can be integrated into local hardware systems or can be connected to machines accessible in the cloud via an API. There is no direct overhead for the user - maximum performance is achieved in a single pass through the tool chain. This is an advantage over other known circuit-level error suppression methods.

[0101] These methods can be used for direct algorithm execution or to improve the performance of quantum error correction (QEC). Enhancing QEC with deterministic circuit-level error suppression through this protocol is an advantage. Another advantage is that the performance of quantum algorithms can be improved without full quantum error correction.

[0102] These approaches are relevant to any quantum computing architecture and are not specific to a single qubit type.

[0103] Background on Noise and Error Processes

[0104] Quantum systems are subject to various types of noise. These noise processes affect the building blocks of quantum circuits, such as gates and measurement outcomes. These operations can be optimized to reduce the effects of noise and imperfections. However, standard gate calibration techniques typically keep the spectator (non-participating) qubits in the ground state. When applying algorithms, more complex context-dependent noise processes may appear, leading to different types of errors.

[0105] Single-qubit decoherence

[0106] During a multi-qubit algorithm, at any given moment some qubits will be idle. Even with perfect gates and measurements, for wide and deep circuits, idle errors alone can completely randomize the quantum circuit, rendering any quantum algorithm useless.

[0107] The quantum state of a single qubit can be represented by a complex superposition of its internal states, A|0>+Bexp{iφ}|1>, where A 2 +B 2 ≤1. While quantum gates intentionally direct quantum states to desired targets, quantum noise and unwanted couplings may also direct quantum states. For example, a noisy process may be due to energy loss to the environment, where the qubit state |1> decays to |0>. This process is characterized by a typical time scale known as T1. This type of error can only be mitigated by shortening the total duration of the circuit. However, as hardware quality improves, T1 times become longer, and for moderately sized algorithms this process is usually not a bottleneck.

[0108] Another noise process is phase shifting. In this process, the phase φ is disturbed and the quantum state loses its phase information. This process is characterized by a typical time scale called T2. In the presence of nearby actuators, T2 can be much shorter than T1, thus dominating the performance of the entire algorithm.

[0109] Crosstalk

[0110] While the two processes described above are due to coupling between the system and the environment, idle errors are also caused by unwanted coupling between nearby qubits. The most common idle error caused by unwanted coupling is called ZZ-error. In this process, a qubit acquires an unwanted phase that depends both on its state and on the states of nearby qubits. This unwanted coupling exists between pairs of coupled qubits and has a constant static part, which depends only on the system parameters, and a dynamic part, which depends on the drive applied during the circuit.

[0111] Gate Error

[0112] Universal quantum computing can be implemented with arbitrary single-qubit gates and two-qubit entangled gates. Both single-qubit gates and two-qubit entangled gates are typically implemented by illuminating the qubit with a shaped wave packet (pulse). In addition to the T1 and T2 errors mentioned above, the gate is also prone to other types of errors. The gate implementation is characterized by its pulse parameters. The ideal gate implementation has specific ideal pulse parameters, and any deviation from these parameters, known as over-rotation, detuning, or other types of deviations, will result in reduced performance. Such deviations are caused by unsuccessful calibration (optimization) processes, external time-varying noise, or system parameter drift. Another type of error is control error, and the pulse parameters may deviate due to the nonlinear transfer function of the control chain and the amplitude / phase noise in the pulse generator. These deviations can lead to coherent gate errors similar to over-rotation. Qubit decoherence further reduces gate fidelity. Although phase shift errors can usually be alleviated by optimizing pulse shaping, errors caused by qubit relaxation can only be reduced by faster gates. However, fast gates are prone to another type of error, where the system is excited into states outside the computational space, which can degrade the fidelity of the circuit.

[0113] Measurement results and reading errors

[0114] Measuring a qubit projects it to an eigenvalue of an observable. Typically, the result is a projection of the qubit to a ground state or an excited state, and ideally, repeated measurements of the same observable should yield the same result. Measurement error can occur due to noise in the readout chain, qubit backactions, qubit relaxation, and correlations between multi-qubit measurements. Measurement error results in a large quantum error correction overhead.

[0115] Overview of Dynamic Decoupling of Device Awareness Algorithms

[0116] Dynamical decoupling (DD) is an open-loop quantum control technique that aims to suppress errors on idle qubits. In its simplest form, DD is implemented by a periodic sequence of control pulses (gates), the net effect of which is to approximately eliminate the undesired coupling. Most DD protocols are designed to combat this T2 process, but there are different approaches to designing DD protocols for specific error reduction tasks. For example, the ZZ DD protocol can mitigate errors. However, the protocols for mitigating ZZ errors can vary dramatically from algorithm to algorithm and from device to device, depending on the device topology, connectivity, and the parallel capabilities of different layers of the algorithm. There is a large literature describing different ways to design DD sequences for different physical setups.

[0117] The present disclosure provides methods for designing protocols that are both efficient and automatically aware of hardware and circuit context and are suitable for general algorithms and a variety of quantum hardware. The ability to build a well-designed protocol for a given algorithm can make it useful.

[0118] Automated DD Solution

[0119] Basic DD Mode

[0120] The disclosed automated scheme combines quantum circuit and device topology ( Figure 2 ) as input and returns a new circuit in which the optimal DD protocol is embedded in the original circuit.

[0121] The guiding principle of DD is to replace idle time periods with identity operations consisting of one or more qubit manipulations that produce no net change to the qubit state. Figure 1 Such examples occur in (a, b). In these examples, the idle period is replaced by two applications of the X-gate (bit flips). Two consecutive bit flips are equivalent to no operation. However, each application of the X-gate reverses the direction of the T2-related (random) phase diffusion. Therefore, by matching the idle time between two bit flips with the remaining idle time, the effect of the T2 error is suppressed (statistically canceled). The statistical cancellation effect is better as the intermediate idle time t is shorter. Therefore, if the total idle time is long, it may be beneficial to replace it with several repetitions of the basic pattern. Since the application of the X-gate is not ideal, there is an optimal number of such repetitions. This number depends on the quality of the qubit (T2 time and X-gate error) and the strength of the ZZ coupling.

[0122] For a pair of coupled qubits, there is a ZZ idle error in addition to the T2 error. A single bit flip reverses the direction of the ZZ coupling. Therefore, applying bit flips (such as Figure 1 (c) will leave the ZZ coupling unchanged. To include ZZ mitigation, the two DD schemes must be relatively displaced. Since the application of the X-gate is not instantaneous, the asymmetry parameter δ can be optimized to achieve the best ZZ elimination. Figure 1 This pattern is a basic building block of the method, as shown in (d). In the case of more than two qubits coupled to each other, more complex patterns can be derived, and different sequences of operations can be used to eliminate the alternating coupling (such as XX).

[0123] The basic pattern containing a set of completely idle qubits is a relatively easy task, which only requires finding the best delta parameter and the best number of times the pattern is repeated. However, finding the best embedding of this pattern in any algorithm ( Figure 2 ) and there is no clear solution.

[0124] Automatically embed DD mode in any circuit

[0125] In this section, the present disclosure provides details of an automated scheme for embedding DD modes in general circuits. The method is designed to eliminate as many T2 and ZZ errors as possible (which is usually impossible to eliminate completely) while minimizing the number of X gates. Due to the individual errors and heating effects of each X gate, extremely dense DD schemes (including many operations in a short period of time) may do more harm than good.

[0126] Figure 3 An example is shown with the input circuit on the top and the compiled circuit with embedded DD on the bottom.

[0127] The automation protocol is as follows:

[0128] 1. Given a device topology, the protocol partitions the qubits into N groups by solving a graph coloring problem. For most hardware providers, N=2 means that all qubits belong to one of two possible groups. The qubits in one group will experience a delta-shifted DD, while the qubits in the other group will experience a non-shifted 2t-2tDD structure. The distinction between the two groups is shown in Figure 1 (d).

[0129] 2. Given a compiled circuit, the protocol slices the circuit into non-overlapping ‘global blocks’ that cover the entire circuit. A block is the maximum time period during which the set of idle qubits remains unchanged. Each block is characterized by an initial and final time [t i ,t f ] and sets of idle qubits.

[0130] 3. Given the block structure, the total number of DDX gates can be calculated. Using this number and the X-gate errors obtained from the backend, the protocol can calculate the X-index. If this index is below a given threshold, the protocol keeps the block structure unchanged and skips step 4. Otherwise, the protocol moves to step 4.

[0131] 4. The protocol partitions the circuit into non-overlapping groups of connected qubits of maximum size four. The protocol then applies the block slicing described in 2 to each group separately to find a new block structure ("local block"). This approach sacrifices ZZ correction of some links to significantly reduce the total number of X-gates. The grouping of qubits is not necessarily unique and is optimized to reduce the number of X-gates in the circuit and ZZ correct mismatches. Given the device topology and the set of active qubits, the optimal grouping construction is fully automated.

[0132] 5. Given a block structure (global or local), the protocol enforces a set of block merging rules to reduce the number of X-gates. The merging rules treat short blocks (less than twice the X-gate duration), blocks of qubits with connectivity 1, and consecutive blocks with some similarity rules in different ways.

[0133] 6. Given the final block structure, the protocol applies the optimal number of DD modes (adapted to the block duration) to the group of all idle qubits in the block as determined in 1. A qubit is considered idle or driven only after it first participates in the circuit.

[0134] Examples of input and output circuits are shown in Figure 3 The scheme is based on the fact that within each global block, the DD pattern optimally reduces the effects of both T2 and ZZ. However, in reality, since the gates in the circuit are not synchronized and have different durations, some blocks may be too short to implement any DD, or two blocks may be too short to implement the full pattern. Moreover, since the X-gates are not ideal, perfect embedding of the DD pattern is not enough, and a DD scheme with extreme loads of the X-gates may perform poorly. Steps 3 and 4 aim to find a compromise that optimizes the performance of a specific algorithm on a specific device.

[0135] The above procedure is hardware agnostic and fully automated. Given a compiled circuit, device topology and backend data, it returns a circuit in which the optimal DD is embedded. This method does not require pulse-level access. As long as the API of the quantum hardware accepts timing scheduling of gates, this method is applicable. The ideal δ-parameter and the optimal number of mode applications can be found in preliminary experiments, which are also automated.

[0136] Overview of Scalable Measurement Error Mitigation

[0137] An arbitrary quantum state is measured by projecting it onto an observable. This task can be reduced to finding its probability distribution over an exponentially large set of basis states. The quantum circuit is executed many times to generate the statistical data needed to construct the probability distribution. Due to the compounding of errors in the reading of each individual qubit, the overall measurement error grows exponentially with the number of qubits.

[0138] The readout assignment error of a single qubit can be represented by a confusion matrix, namely (C 1Q )-

[0139]

[0140] Among them C ij = P[i|j] is the probability of assigning |i> to the measured state |j>. For a single qubit, the ideal distribution can be obtained by the inverse of the confusion matrix, that is, And the obtained quasi-probability distribution is normalized. However, this method is prone to several scalability issues -

[0141] The size of the confusion matrix for the entire system With the quantum bit (N q ) grows exponentially. q = 42, storing C in a sparse format requires 580PiB, which is 120 times the capacity available on Fugaku, the fastest supercomputer currently available.

[0142] ● Estimating the confusion matrix requires Calibration experiment.

[0143] ●Since the measured distribution itself has There are different possible outcomes, so In the case of , using only M samplings will produce a poor estimate of the confusion matrix.

[0144] ●For multiple qubits, taking the inverse of the confusion matrix is ​​usually ill-defined and a time-consuming process.

[0145] Automatic measurement result correction

[0146] Tensor Relief

[0147] The disclosed method solves the above problem by partitioning the quantum device into smaller groups of qubits, which can then be efficiently corrected through tensor algebra. The groups are automatically selected in a way that minimizes the inter-group correlations between measurement results, based on the topology of a given quantum computer. Figure 4 An example of such a grouping is shown.

[0148] The disclosed method ensures that all combined groups span the entire quantum computer (or at least all measured qubits in a quantum circuit). g Let it be the number of groups, and i∈{k∈N:k≤N g {G i} represents a set of groups such that With {G i The confusion matrix corresponding to each group in} is {C i}, and each C i The size of the matrix is ​​given by Given.

[0149] N q The measured probability vector for qubits has elements, namely P i against Where P i N represents the measurement i q - bit binary representation of probability. The automated procedure is as follows -

[0150] 1. Dimension The measured probability vector of is shaped into a tensor of dimension (2×2×…×2) such that each dimension i of the tensor corresponds to each qubit index i in the bit string.

[0151] 2. Rearrange the order of the qubits in the tensor to match G i corresponds to the appropriate qubit in .

[0152] 3. If all qubits are not measured, then for the qubits that are measured, they belong to the same group G i All the unmeasured qubits in the tensor add another dimension to the tensor.

[0153] 4. Reshape the rearranged tensor into dimension Set this reshaped tensor to

[0154] 5. i tensors Each index is associated with the appropriate confusion matrix C i The reciprocal of . Set as elements and

[0155]

[0156] 6. Tensor Reshape into (2×2×...×2) and remove the additional dimensions added in step 3.

[0157] 7. The output of step 6 may contain negative elements. Convert this quasi-probability distribution into the closest probability distribution.

[0158] The number of calibration experiments for this method is given by Given. Figure 4 In the example provided, only 16 calibration experiments are required to fully characterize a 16-qubit device, as compared to 65,536 calibration experiments required for the full confusion matrix approach. In addition, since the disclosed method estimates only the distribution across 16 states, the number of shots required to achieve high accuracy is significantly reduced. Since only a small matrix needs to be inverted, the speed and accuracy of the protocol execution are improved.

[0159] from Figure 5 The effectiveness of the disclosed mitigation can be seen in . The disclosed form is directly applicable to sparse distributions. This results in a scalable, fast and competitively accurate protocol. Finally, the method processes the output probabilities through a custom neural network as described below, incorporating a small correction for long-range / inter-group correlations.

[0160] Neural Network (NN) for Multi-Qubit Correlations

[0161] The disclosed method uses the output of the above-mentioned tension-modulated mitigation as input to a neural network (NN). The goal of the NN is to recover previously ignored correlations (at the boundaries of the group), taking into account nonlinear effects and contextual errors that cannot be captured in the confusion matrix form. Since the method is looking for small modifications to the identity transformation, the method uses a ResNet-like architecture of the NN. The number of adjustable parameters in the NN is proportional to the number of qubits (N q ) in sublinear proportion. This approach minimizes the amount of training data required. Figure 6 An example of a neural network with 7 qubits is shown.

[0162] Disclosed herein is an automated method that seamlessly incorporates both the disclosed tensor mitigation method and the disclosed neural network method. Figure 7 It is shown that the neural network provides additional improvements over and above those provided by the subset tensor mitigation protocol. Furthermore, the trained network continues to provide improvements for at least up to 15 days without additional training data. The disclosed method is also suitable for mitigating expected errors of observables.

[0163] Quantum Compilation

[0164] Quantum compilers are used to find more efficient ways to execute programs on quantum devices. Quantum compilers use a series of "compiler passes" to handle the mapping of circuits to the device gate set and topology, and simplify the circuit instructions. This is done by removing unnecessary gates, reducing circuit depth, and moving gates to qubits with lower error rates. Simplifying circuit logic is the first step to mitigate the effects of noise during program execution - fewer operations reduce the chance of error, or certain error-prone operations or qubits can be removed from the circuit.

[0165] However, a delicate balance is found between these compiler passes in order to compile the circuit in an optimal way. For example, moving a single-qubit gate to a low-error qubit requires inserting a SWAP operation in the circuit. These SWAP gates increase the circuit depth by being composed of two-qubit gates (e.g., CNOT, CZ), which are also generally more prone to ZZ errors. Therefore, while some errors from single-qubit gates are mitigated, new sources of errors from two-qubit gates are introduced. For any given algorithm and device, there is a trade-off of diminishing returns between each compiler pass.

[0166] Crosstalk-aware compilation

[0167] As emphasized above, finding the right series of compiler passes is not trivial and depends on both the device and the algorithm. There are several open source quantum compilers that can serve as a basis for automating this process. By writing custom plugins to TKET (https: / / github.com / CQCL / tket) and qiskit (https: / / qiskit.org / ), the compiler flow can be customized based on the error characteristics of the device. The workflow of the "vanilla" program is as follows:

[0168] 1. Run an initial "gate pass" to map the circuit to device gates.

[0169] 2. Use the specified "Topology Transfer" to map the circuit to the device topology. This may require inserting SWAP gates in the circuit.

[0170] 3. Place the circuit in the desired format using the specified "routing pass". This will involve moving the circuit instructions to qubits with smaller gate errors. This may also involve inserting SWAP gates.

[0171] 4. Run a "SWAP-reduce pass" to iteratively minimize the number of SWAP gates in the circuit.

[0172] 5. Run another gate pass.

[0173] 6. Run a specified series of isomorphic reductions on the circuit ("simplification pass"). This can include removing redundant gates, minimizing two-qubit operations, and replacing certain gate sequences with simpler known definitions.

[0174] 7. Use "two-qubit gate rerouting" to avoid running certain gates in parallel.

[0175] 8. Run a final gate pass and topology pass before returning to the circuit.

[0176] While this vanilla approach is often successful, some algorithms may require more customized methods to maximize error-mitigation. Important among these are the topology, routing, and simplified transfers that are run on the circuit.

[0177] Specifically, the behavior of the topological and routing passes depends heavily on circuit characteristics such as the depth, density, and number of two-qubit gates. One undesirable consequence of these passes is the insertion of too many SWAP gates. For N SWAP gates, 3N two-qubit gates are required to implement them. Excessive rerouting can easily hurt performance compared to the original circuit placement. Additionally, the behavior of the simplification passes depends heavily on the format of the circuits. While two circuits may be isomorphic, differences in their gate operations may result in different simplifications.

[0178] To alleviate this problem, this paper discloses a hybrid compilation routine that alternates between different topologies, routings, and simplified passes. More specifically, the disclosed method compiles the initial circuit using three different procedures:

[0179] 1. Run the vanilla method using topology pass A, routing pass A, and simplified pass sequence A ("routine A").

[0180] 2. Run the vanilla method using topology pass B, routing pass B, and simplified pass sequence B ("routine B").

[0181] 3. Run the vanilla method "M" times in sequence, passing the output of routine A into routine B and back again until they merge ("hybrid routine").

[0182] 4. Avoid parallelizing specific gates when necessary.

[0183] Taking a hybrid approach, the disclosed method alternates between two different sets of passes that by themselves may have only modest effectiveness. However, by running back and forth, the disclosed method effectively exposes it to different initial conditions (topology, format, and gates) each time. This allows the disclosed method to extract as much benefit as possible from each pass. The final step is performed in the presence of strong crosstalk, in which case the quality of applying both gates simultaneously is much lower than if the two gates were implemented conservatively separately.

[0184] Once the disclosed method has run all three routines, the method uses a heuristic to select the "best" based on the number of two-qubit operations in each circuit. This method selects the best compiled circuit every time, regardless of the algorithm and device. Figure 8 provides a comparison of three different QFT routines. Ultimately, the selected circuit will be embedded in the above-mentioned DD mode using a set of customized "decoupled transfers".

[0185] Automated fully automated device door optimization and calibration

[0186] Quantum gates are the building blocks of any quantum circuit. Consistently maintaining a high quality (low error) base set of gates is a challenge faced by all hardware vendors. Reducing gate errors can be achieved by manipulating a set of accessible experimental control probes. The details depend on the specifics of the hardware.

[0187] The disclosed method can provide a low-level implementation of quantum gates based solely on measured system responses, without requiring a priori knowledge of the specific device model or its underlying error processes.

[0188] This black-box optimization protocol is based on a set of repeated experiments in which the disclosed method extracts the gate error from the error-amplification protocol. One such protocol fits the estimated average distortion to the number of gate repetitions N. For each value of N, the method applies N gate repetitions under test N times on two or more different initial states. 批 Different implementations. For example, a method evaluates its quality by applying a specific CX gate implementation on the initial states |+0> and |+1>.

[0189] After each run, a full state tomography is performed and an estimate of the distortion relative to the ideal target state is calculated. An effective measure of per-gate error is extracted by applying a linear fit to the average distortion as a function of N, which provides a measure of gate error in the low error limit.

[0190] The above procedure constitutes a cost function for assigning a quality measure to a particular gate implementation. It is not the only way to determine the error, and measurement routines such as randomized benchmarking, gate set tomography, etc. can be incorporated to construct the cost function. An automated optimization procedure is designed to interact with the system to produce better and better gate implementations. More details can be found in Y. Baum, M. Amico, S. Howell, M. Hush, M. Liuzzi, P. Mundada, T. Merkh, A. Carvalho, and M. J. Biercuk, Experimental deep reinforcement learning for error robust gate-set design on a superconducting quantum computer, PRX Quantum, 2, 040324 (2021). Specifically, the method uses a form of optimization, such as deep reinforcement learning, to design a set of universal error-robust quantum logic gates at runtime without knowing a specific Hamiltonian model of the system, its controls, or its underlying error process. More specifically, the method applies closed-loop control techniques, such as autonomous deep reinforcement learning, to discover the best control signals that drive state evolution through iterative interaction with a quantum computer; this provides new waveforms that define a universal gate set that outperforms human-designed solutions. Because the control design process involves direct hardware measurements, all error sources, including leakage, phase shift, crosstalk, RF signal miscalibration, and distortion are captured in the optimized control, even if the human operator is aware of these error sources.

[0191] The above procedure is applied to a pair of qubits. In general, different pairs that do not share qubits can be optimized in parallel. If there are strong crosstalk effects, it may be beneficial to avoid optimizing adjacent pairs in parallel. The method automatically splits the device into a set of qubit pairs that can be optimized in parallel. Such a split is Fig. 9 As depicted, the sixteen pairs are divided into four groups due to the strong crosstalk effect. This is an example grouping based on the connectivity of the example device. The gates between the pairs in each group are optimized in parallel.

[0192] Additional characterization routines may be run to understand device error sources and noise processes to directly or indirectly inform the gate optimization routine. This may include noise characterization or Hamiltonian parameter estimation procedures that help inform the definition of the optimization cost function.

[0193] Overview of overall benefits

[0194] The disclosed method is implemented as a software package and is agnostic to the hardware and algorithm type. Given an arbitrary algorithm and hardware properties, such as device topology, connectivity, and backend data on qubit quality, the disclosed method enhances the performance of the algorithm by employing the methods disclosed herein. Other preliminary experiments, such as crosstalk analysis or noise spectrum analysis, can also be automatically added to further improve performance. The method includes optimized system-aware compilation, optimal DD embedding, full device gate calibration / optimization, and scalable measurement error mitigation. Fig.10 The success probability of the Bernstein-Vazirani algorithm on a 16-qubit device is compared. The automated software tool is compared with the best compilation, DD scheme, and measurement error mitigation provided by the hardware provider.

[0195] Method for generating error-corrected quantum circuits

[0196] Fig.11 A method 1100 for generating an error-corrected quantum circuit is shown. The method 1100 includes compiling 1101 a quantum algorithm onto quantum hardware using a plurality of qubits to generate a quantum circuit. As discussed above, this involves mapping the quantum algorithm onto a native gate set of the quantum hardware and replacing the native gates with error-corrected gates using previous characterization measurements.

[0197] The method 1100 then includes determining 1102 quantum relationships (e.g., couplings) between the plurality of qubits and quantum relationships (e.g., couplings) between the plurality of qubits and an external environment as a result of implementing the quantum circuit on the quantum hardware. For example, this may involve determining pairs of gates that exhibit crosstalk with each other. This may also involve measuring results to characterize errors that are expected to occur as a result of interactions with the environment. Such quantum relationships may result in errors that are specific to a given quantum hardware, and to a quantum circuit as generated by a compilation step, because the compilation defines which gates are close to each other, and the actions applied to adjacent qubits, resulting in potential crosstalk. Additional details have been described above.

[0198] The method 110 then includes using 1103 the quantum relationships between the plurality of qubits and between the plurality of qubits and an external environment to determine a deterministic error correction protocol for each qubit in the plurality of qubits and for interactions between the plurality of qubits. Figure 1 and 2 As explained, deterministic error correction includes identity gates and time offsets to optimize correction of ZZ errors, as well as correction of optimized error-resistant quantum gates on qubit pairs and individual qubits.

[0199] Methods used to characterize the error

[0200] Fig.12 A method 1200 for characterizing errors in quantum hardware comprising a plurality of qubits is presented. This can be viewed as Fig.11 1102 in . Method 1200 can provide measured system responses without prior knowledge of a specific device model or its underlying processes. That is, method 1200 is part of a black box optimization approach in which the actual quantum hardware is treated as a black box, where little is known about how the quantum hardware is constructed.

[0201] The method 1200 includes determining 1201 a group of qubits based on the topology of the quantum hardware to minimize inter-group correlations, and then performing 1202 a plurality of measurements to approximate a confusion matrix for individual qubit readings. The method 1200 includes repeating the plurality of measurements for the group of qubits to determine the error characterization of the hardware. In some examples, the above procedure is applied to a pair of qubits. In general, different pairs that do not share qubits can be optimized in parallel, such as Fig. 9 shown.

[0202] Methods used for compilation

[0203] Fig.13A method 1300 for compiling a quantum algorithm onto quantum hardware having a local gate set is presented. The method 1300 comprises repeatedly performing the following steps: performing 1301 a first compilation process to map the quantum algorithm onto the quantum hardware and the local gate set; and using the output of the first compilation process as 1302 an input for performing a second compilation process different from the first compilation process to maximize the use of low-error qubits as identified by an error characterization process, as explained above with reference to 8a and 8b.

[0204] Methods for quantum error correction

[0205] Fig.14 A method 1400 for quantum error correction of quantum hardware is presented, as explained above with reference to the DD mode. The method 1400 includes partitioning 1401 qubits of the quantum hardware into groups based on the topology of the quantum hardware. The method 1400 then includes partitioning 1402 the qubits into blocks based on a quantum circuit compiled onto the quantum hardware. The method 1400 further includes applying 1403 driven single-qubit operations to correct quantum errors occurring during identity operations depending on the quantum circuit compiled onto the quantum hardware. Finally, the method 1400 includes determining 1404 a time shift of the driven single-qubit operations to correct cross-coupling and suppress decoherence for each group in the group and each block in the block during the identity operation, as explained above with reference to the DD mode. Figure 1 and 2 Explained.

[0206] Using the methods previously described, additional error correction can be achieved using error-resistant optimized gates. More specifically, a method for gate-set optimization of quantum hardware is provided. The method includes performing measurements on the quantum hardware to determine a quality measure of the gate implementation on the quantum hardware. The method then includes applying an optimization procedure to iteratively adjust the gate characteristics and repeating the measurements to gradually reduce the errors of the quantum gates. The gate characteristics may involve waveforms that control the qubits in the quantum hardware. The method may further include performing an optimization procedure to define a universal gate set. Further, the measurement results may represent a variety of error sources, such as leakage, phase shift, crosstalk, rf signal miscalibration and distortion. Even further, the measurement results and optimization procedures can be applied to the quantum hardware without knowing the details of the quantum hardware, such as the actual Hamiltonian of the physical device.

[0207] Method for error correction compilation

[0208] Fig.15A method 1500 for error correction compilation to correct error sources identified in an error characterization process is presented. The method includes receiving 1501 a compiled circuit compatible with the error characterization and a hardware device topology and gate set; and replacing 1502 quantum logic gates in the compiled circuit having uncorrected errors exceeding an acceptable threshold with error-corrected replacements derived using the error characterization process. Finally, the method 1500 includes outputting 1503 a definition of the circuit with all replacement gate definitions, including identity and non-identity operations, in a format for hardware execution.

[0209] Method for correcting errors in quantum measurement results

[0210] Fig.16 A method 1600 for correcting errors in quantum measurement results of a quantum circuit on quantum hardware is presented. The method 1600 includes receiving 1601 a measurement result; and shrinking a tensor based on the qubit group using 1602 an approximate confusion matrix from an error characterization. The method 1600 then includes converting 1603 the tensor into a probability distribution; and correcting 1604 the error in the quantum measurement result as indicated by the probability distribution. Finally, the method 1600 includes outputting 1605 a measurement result - an error-corrected result.

[0211] Internet Server

[0212] Fig.17 An Internet server 1700 is shown, which is a computer system connected to the Internet to receive data from and send data to various other computer systems via the Internet. Specifically, the Internet server 1700 is connected to user client computers and interface computers of quantum hardware via the Internet. The Internet server 1700 is a classical digital electronic computer system and includes a processor 1701, a program memory 1702, a data memory 1703, and a communication port 1704.

[0213] The program memory 1702 is a non-volatile computer-readable medium and stores program codes thereon. When the program codes are executed by the processor 1701, the processor performs the method disclosed herein, including Figure 11-16 The method disclosed in .

[0214] The Internet server 1700 is configured by program code installed on the program memory 1702 to provide or expose a network interface. The network interface includes input elements for inputting representations of quantum algorithms and quantum hardware. The network interface may be a graphical user interface with graphical input elements for data input or file upload, or may be an application programming interface (API) with syntactic input elements such as XML or JSON fields.

[0215] The Internet server 1700 is thus configured to receive the quantum algorithm and the incomplete description of the quantum hardware via the network interface. The quantum algorithm and the incomplete description may be provided by the same user, by different users, or the incomplete description may be provided by a supplier or provider of the quantum hardware. The description is considered "incomplete" because it contains an abstract description of the quantum hardware (such as a local gate set), but does not contain an error-characterized description of a single gate or circuit block.

[0216] Thus, internet server 1700 is further configured to perform error characterization for the target quantum hardware or retrieve previously performed error characterization (e.g., from data storage 1703). Error characterization may involve communicating with the quantum hardware to initiate qubits with quantum values, perform quantum operations, and receive measurement results to characterize errors as described above.

[0217] The internet server 1700 is further configured to generate a decorated quantum circuit description that implements a quantum algorithm on quantum hardware. "Decorated" in this context means that the quantum circuit includes quantum gates and a control protocol for error correction. Further, the decorated quantum circuit description includes quantum logic blocks that are error corrected based on the error characterization described above as gate calibration.

[0218] The internet server 1700 is further configured to output a quantum circuit description. This may be output to a user so that the user can send the description to quantum hardware for implementation. In other instances, the internet server 1700 sends the description directly to the quantum hardware, such as an API of a hardware provider, for immediate implementation.

[0219] Once the quantum algorithm has been executed on the quantum hardware, the Internet server 1700 receives and analyzes the raw output data from the quantum hardware using the complete representation. This involves the measurement error correction described above.

[0220] Thanks to the methods disclosed herein, deterministic error correction protocols are applied to quantum hardware, and error correction gates are used together with measurement result error correction. Together, these measures lead to a significant reduction in error rates that was previously unattainable.

[0221] Those skilled in the art will appreciate that, without departing from the broad general scope of the present disclosure, various changes and / or modifications may be made to the embodiments described above. Therefore, the embodiments of the present invention should be considered in all aspects as illustrative and not restrictive.

Claims

1. A method comprising: Receive incomplete descriptions of quantum algorithms and quantum hardware; Performing error characterization or retrieving error characterization previously performed for target quantum hardware; generating an initial quantum circuit description for implementing the quantum algorithm on the quantum hardware; generating a decorated quantum circuit description based on the initial quantum circuit description, wherein the decorated quantum circuit description includes a quantum logic block that is error corrected based on the error representation; wherein the error characterization is specific to the initial quantum circuit description and specific to the quantum hardware; outputting the quantum circuit description; and Raw output data is received from the quantum hardware and analyzed using a complete characterization.

2. The method of claim 1 , wherein the step of generating an initial quantum circuit outputs neighboring qubits and actions applied to the neighboring qubits, and the error characterization indicates crosstalk between the neighboring qubits due to the actions applied to the neighboring qubits.

3. The method according to claim 1 or 2, further configured to perform the following steps: Compiling quantum algorithms onto quantum hardware using multiple qubits to generate quantum circuits; determining quantum relationships between the plurality of qubits and between the plurality of qubits and an external environment as a result of implementing the quantum circuit on the quantum hardware; and The quantum relationships between the plurality of qubits and between the plurality of qubits and an external environment are used to determine a deterministic error correction protocol specific to the quantum circuit and specific to the quantum hardware for each qubit in the plurality of qubits and for interactions between the plurality of qubits.

4. The method of any one of the preceding claims, further configured to perform the step of performing a plurality of measurements on qubits of the quantum hardware to determine an error characterization of the quantum hardware based on the topology of the quantum hardware.

5. The method according to claim 3 or 4, wherein the error characterization comprises one or more of the following based on the topology of the quantum hardware: individual gate errors, decoherence processes, measurement result errors and crosstalk between qubits.

6. The method according to any one of claims 3 to 5, wherein compiling comprises, for each of a plurality of logical quantum operations in the quantum algorithm, selecting one or more quantum blocks to implement the quantum algorithm on the quantum hardware, wherein each of the quantum blocks implements a logical function of the logical quantum operation.

7. A method according to any one of claims 3 to 6, further configured to perform the step of determining an error correction protocol for each of the quantum blocks, the error correction protocol comprising error correction for errors identified in the error characterization and being specific to the selected one or more quantum blocks to implement the quantum algorithm.

8. The method according to any one of claims 3 to 7, further configured to perform the step of executing the quantum algorithm on the quantum hardware using the selected quantum blocks and the deterministic error correction protocol to obtain a qubit measurement result.

9. A method according to any one of claims 3 to 8, further configured to perform the step of reducing a measurement error in the qubit measurement by applying an error reduction operation based on the error characterization.

10. The method of any one of the preceding claims, further configured to characterize errors in quantum hardware comprising a plurality of qubits by performing the following steps: determining groups of qubits based on the topology of the quantum hardware to minimize correlations between groups; performing multiple measurements to approximate the confusion matrix of individual qubit readings; and The plurality of measurements are repeated for the group of qubits to determine the error characterization of the hardware.

11. The method of claim 10, wherein the group of qubits comprises qubit pairs having low cross coupling between pairs.

12. The method of claim 10 or 11, wherein the error characterization comprises one or more of individual gate fidelity, noise spectrum, and crosstalk strength.

13. The method according to any one of claims 10 to 12, further configured to perform the step of automating the determination of the error characterization using machine learning.

14. The method according to any one of claims 10 to 13, further configured to perform the step of optimizing parameters of a quantum gate using the qubits to maximize the fidelity of the quantum gate.

15. A method according to any one of claims 10 to 14, wherein for gates having qubits in the group, fidelity is maximized.

16. The method of any one of the preceding claims, further configured to compile a quantum algorithm onto quantum hardware having a local gate set by repeatedly performing the following steps: executing a first compilation process to map the quantum algorithm onto the quantum hardware and the local gate set; and An output of the first compilation process is used as an input for performing a second compilation process different from the first compilation process to maximize usage of low-error qubits as identified by the error characterization process.

17. The method according to any one of the preceding claims, further configured to perform quantum error correction of quantum hardware by performing the following steps: dividing qubits of the quantum hardware into groups based on the topology of the quantum hardware; partitioning the qubits into blocks based on quantum circuits compiled onto the quantum hardware; applying driven single-qubit operations depending on the quantum circuit compiled onto the quantum hardware to correct quantum errors arising during identity operations; as well as A time shift of the driven single qubit operation is determined to correct cross-coupling of each of the groups and each of the blocks during an identity operation and to suppress decoherence.

18. The method of claim 17, further configured to perform the step of redefining the non-identity operation by time-domain modulation of a control signal for implementing the non-identity operation to correct the error process identified by the error characterization.

19. The method of claim 17 or 18, further configured to perform the step of modulating the control signal in a plurality of qubits in time in a continuous or discontinuous manner.

20. The method according to any one of claims 17 to 19, further configured to perform the step of assembling the block from a library of error-corrected quantum gates based on an error characterization.

21. The method of any one of the preceding claims, further configured to perform error correction compilation to correct the error sources identified during the error characterization process by performing the following steps: receiving a compiled circuit compatible with a hardware device topology and gate set; replacing quantum logic gates in the compiled circuit having uncorrected errors exceeding an acceptable threshold with error-corrected replacements derived using the error characterization process; as well as The definition of the circuit is output in a format for hardware execution with all alternative gate definitions, including identity and non-identity operations.

22. The method according to claim 21, further configured to perform the following steps: Identify error-prone non-identity gates; Replace error-prone non-identity gates with automatically calibrated gates; replacing an identity gate in the compiled quantum circuit with an error-corrected identity gate that includes an error correction protocol; An output error-robust circuit including the autocalibrated gates and the error-corrected identity gates is assembled into a circuit representation for execution on the quantum hardware.

23. A method according to any one of the preceding claims, further configured to correct errors in quantum measurements of a quantum circuit on quantum hardware by performing the following steps: receiving measurement results; shrinking a tensor based on the group of qubits using an approximate confusion matrix from an error representation; converting the tensor into a probability distribution; and correcting the error in the quantum measurement result as indicated by the probability distribution; and Output measurement result-error-correction result.

24. The method according to claim 23, further configured to perform the following steps: partitioning the quantum circuit into groups of qubits based on the topology of the quantum hardware to minimize correlations between groups; and The ordering of the tensor is rearranged based on the groups of qubits.

25. The method of claim 24, further configured to perform the step of adding dimensions to the tensor for unmeasured qubits.

26. The method according to any one of claims 23 to 25, further configured to perform the step of applying a neural network to the measurements based on the error characterization to recover correlations ignored by tensor relief.

27. An Internet server configured to: providing a network interface comprising an input element for inputting a representation of a quantum algorithm and quantum hardware; receiving, via the network interface, an incomplete description of the quantum algorithm and the quantum hardware; Performing error characterization or retrieving error characterization previously performed for target quantum hardware; generating a decorated quantum circuit description that implements the quantum algorithm on the quantum hardware, wherein the decorated quantum circuit description includes quantum logic blocks that perform error correction based on the representation; outputting the quantum circuit description; and Raw output data is received from the quantum hardware and analyzed using a complete characterization.

28. The Internet server of claim 27, wherein outputting the quantum circuit description comprises: sending the decorated quantum circuit description to a user via the network interface; or The decorated quantum circuit description is sent to the quantum hardware for execution on the quantum hardware.

29. The Internet server of claim 27 or 28, wherein receiving and analyzing the raw output data comprises: receiving the raw output data from the user via the network interface for analysis; or The raw output data is received from the quantum hardware for analysis.