Quantum computer and quantum state control method of quantum computer

The quantum computer system leverages longitudinal relaxation and random gates to address the limitations of existing quantum computers, enabling them to solve real-world problems effectively and accurately without requiring error correction.

JP7590928B2Active Publication Date: 2024-11-27HITACHI LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
JP2021097298
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-06-10
Publication Date
2024-11-27
Estimated Expiration
2041-06-10

AI Technical Summary

Technical Problem

Current quantum computers, including NISQ computers and quantum annealing systems, lack effective algorithms for solving real-world problems and are limited by the assumption of maintaining quantum coherence, making it difficult to handle general problems.

Method used

A quantum computer system that utilizes longitudinal relaxation to transition from a high-energy state to a low-energy state, employing a Hamiltonian evolution method where the Hamiltonian is gradually changed over time intervals approximately equal to the longitudinal relaxation time, allowing the system to naturally relax to the ground state, and incorporates random gates to eliminate systematic errors.

Benefits of technology

Enables quantum computers to solve real-world problems without error correction, handling general problems and improving solution accuracy through variational methods and machine learning, while naturally correcting operational errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007590928000002
    Figure 0007590928000002
  • Figure 0007590928000003
    Figure 0007590928000003
  • Figure 0007590928000004
    Figure 0007590928000004
Patent Text Reader

Abstract

To provide a quantum computer which can handle real problems with a gate-type quantum computer, and a method for controlling the quantum state of the quantum computer.SOLUTION: A quantum computer comprises a quantum register for holding quantum bits, a control gate for performing operation upon the quantum register, and readout means for observing the state of the quantum register, gradually changes Hamiltonian H(t) for a prescribed time to repeat longitudinal relaxation to a ground state. A unitary computation determined by the Hamiltonian H(t) at each time of day is carried out by the control gate taking a time on the order of a longitudinal relaxation time, the quantum state is relaxed for each time on the order of the longitudinal relaxation time, and the initial state prepared in the ground state is evolved with time to the Hamiltonian ground state that is the problem to be solved.SELECTED DRAWING: Figure 1A
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to computer technology, and in particular to quantum computer technology. [Background technology]

[0002] In the IT society, there is no limit to the demands on computer performance, and there are high hopes for quantum computers to meet these demands. Quantum computers first drew attention to Shor's prime factorization algorithm, which demonstrated that quantum computers can process information faster than classical computers. Shor's algorithm was subsequently generalized into a phase estimation algorithm, leading to the invention of the HHL algorithm, which can solve simultaneous equations quickly. Another algorithm that raised expectations for quantum computers was Grover's search algorithm, which was generalized into an amplitude amplification algorithm and was fused with a phase estimation algorithm to develop into an amplitude estimation algorithm.

[0003] Quantum computers are highly promising algorithmically, but these algorithms are based on the assumption of error-free quantum computers. However, quantum computers use quantum bits, which are linear superpositions of |0> and |1> states, as their basic units, and in that sense are analog, so bit errors and phase errors can always occur. Although it is possible to realize an error-tolerant quantum computer by using quantum error correction, this requires the realization of a large-scale quantum computer with a sufficiently small error rate, and there is no prospect of this being realized.

[0004] A concept that has been gaining attention in recent years is the NISQ (Noisy Intermediate-Scale Quantum) computer. This is an attempt to operate a medium-scale quantum computer (with 50 to several hundred quantum bits) without error correction. It has been proven that NISQ computers can process faster than classical computers (quantum supremacy), but what has been proven is that they can generate random numbers through quantum interference, and the challenge is unique and not for solving real problems. In other words, algorithms for solving real problems using NISQ computers remain an unsolved issue.

[0005] There is also a technology called quantum annealing in computing that uses quantum technology. To distinguish it from quantum annealing, the above quantum computer is sometimes called a gate-type quantum computer.

[0006] Quantum annealing is a method to convert the problem so that the solution is the ground state of Hamiltonian H0, set the initial state to the ground state of Hamiltonian H1, which can easily realize the ground state, and gradually shift the Hamiltonian from H1 to H0 to finally obtain the ground state of H0. Quantum annealing is based on the adiabatic theorem, and the continuation of quantum coherence is a theoretical premise. In this method, the shape of H0 is determined in hardware, and it usually takes the form of a Hamiltonian called an Ising spin system. Therefore, problems that are easy to implement are limited, and it is extremely complex and difficult to implement general problems. In addition, Ising spin takes two states, +1 or -1, and does not assume a linear superposition state as a solution. In other words, it assumes a classical solution and cannot be applied to problems that have a quantum mechanical solution.

[0007] There is also a method that uses the Zeno effect, which is similar to quantum annealing. Like quantum annealing, it starts from the ground state of H1 and ultimately tries to obtain the ground state of H0. The difference is that repeated measurements are performed during the calculation process to determine the state at each time. The measurement method is based on a phase estimation algorithm, but other measurements can be used as long as they determine the state. However, no specific measurement methods other than the phase estimation algorithm have been clearly defined, and the time evolution is premised on the continuation of quantum coherence.

[0008] For example, Patent Document 1 discloses a conventional quantum computer system comprising a quantum register consisting of at least one qubit, a quantum unit having a control gate that operates on the quantum register and a readout gate that observes the state of the quantum register, a classical memory device, and a control device that can access the classical memory device, wherein the classical memory device stores quantum microcode, which is a sequence of operation instructions for the control gate or the readout gate, and the control device reads out the quantum microcode from the classical memory device to control the control gate or the readout gate. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] JP 2013-114366 A Summary of the Invention [Problem to be solved by the invention]

[0010] As mentioned above, there are no known algorithms for solving real problems with NISQ computers. Quantum annealing does not have the freedom to change the shape of the Hamiltonian in hardware, and in the case of the Ising spin type, it cannot be applied to quantum mechanical problems. The method using the Zeno effect has not been implemented for repeated measurements and remains a theoretical concept. In addition, both quantum annealing and the Zeno effect method are theoretically premised on maintaining quantum coherence, and it is difficult to realize these theories with actual quantum technology. As mentioned above, both methods have issues, and there is no method that can handle general problems. The challenge is to realize a quantum computer that can handle general problems.

[0011] Therefore, an object of the present invention is to enable a gate-type quantum computer to handle real problems. [Means for solving the problem]

[0012] A preferred aspect of the present invention is a quantum computer comprising a quantum register for holding quantum bits, a control gate for performing operations on the quantum register, and a readout means for observing the state of the quantum register, and which repeats longitudinal relaxation to the ground state by gradually changing a Hamiltonian H(t) over a predetermined time, wherein a unitary operation determined by the Hamiltonian H(t) at each time is performed by the control gate over a time period of about the longitudinal relaxation time, the quantum state is relaxed every time period of about the longitudinal relaxation time, and an initial state prepared at the ground state is evolved over time to the ground state of the Hamiltonian of interest.

[0013] Another preferred aspect of the present invention is a quantum state control method for a quantum computer comprising a quantum register for holding quantum bits, a control gate for performing operations on the quantum register, and a readout means for observing the state of the quantum register, the quantum bits being controlled by the control gate, the quantum bits being controlled by the control gate, and a Hamiltonian H(t) being gradually changed over a predetermined time period to repeat longitudinal relaxation to the ground state, a unitary operation determined by the Hamiltonian H(t) at each time being performed by the control gate over the predetermined time period, the quantum state being relaxed at each predetermined time period, and an initial state prepared as the ground state being evolved over time to the ground state of the Hamiltonian of interest. Effect of the Invention

[0014] Gate-type quantum computers will be able to handle real problems. [Brief description of the drawings]

[0015] [Figure 1A] FIG. 2 is a block diagram showing a hardware configuration of the present embodiment. [Figure 1B] A diagram showing the principle of the present embodiment, (a) described in terms of an energy diagram, (b) an example of a specific quantum circuit. [Diagram 2] A conceptual diagram showing the relationship between the Bloch sphere and quantum bits. [Diagram 3] FIG. 4 is an image diagram of a frequency distribution obtained by the method of the present embodiment. [Figure 4A] A circuit diagram in which the commutative gates in Figure 1B(b) are swapped so that two qubit gates are not consecutive. [Figure 4B] A circuit diagram in which a random gate is selected for each division of an RZ gate, and the selected random gate and its inverse gate are inserted between all gates in the division. [Figure 4C] A circuit diagram that combines each gate with the random gates and inverse gates before and after it into a single gate. [Figure 4D]A circuit diagram in which a random gate is selected for each division of the RY gate, and the selected random gate and its inverse gate are added. [Figure 4E] A circuit diagram that combines each gate with the random gates and inverse gates before and after it into a single gate. [Diagram 5] A circuit diagram of an RZZ gate broken down into RI, RZ, and RCZ gates. [Figure 6] Circuit diagram for determining the ground state of the hydrogen molecule. [Figure 7] 1 is a flow chart showing a series of steps according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] The embodiments will be described in detail with reference to the drawings. However, the present invention should not be interpreted as being limited to the description of the embodiments shown below. It will be easily understood by those skilled in the art that the specific configuration can be changed without departing from the concept or purpose of the present invention.

[0017] In the configurations of the embodiments described below, the same parts or parts having similar functions are denoted by the same reference numerals in different drawings, and duplicated explanations may be omitted.

[0018] When there are multiple elements having the same or similar functions, they may be described by using the same reference numerals with different subscripts. However, when there is no need to distinguish between multiple elements, the subscripts may be omitted.

[0019] The designations "first," "second," "third," and the like in this specification are given to identify components, and do not necessarily limit the number, order, or content. Furthermore, numbers for identifying components are used in different contexts, and a number used in one context does not necessarily indicate the same configuration in another context. Furthermore, a component identified by a certain number does not prevent the component from also having the function of a component identified by another number.

[0020] In order to facilitate understanding of the invention, the position, size, shape, range, etc. of each component shown in the drawings, etc. may not represent the actual position, size, shape, range, etc. Therefore, the present invention is not necessarily limited to the position, size, shape, range, etc. disclosed in the drawings, etc.

[0021] All publications, patents, and patent applications cited herein are hereby incorporated by reference in their entirety.

[0022] As used herein, elements referred to in the singular are intended to include the plural unless the context clearly indicates otherwise.

[0023] The following is a summary of the contents disclosed in the embodiment. In a gate-type quantum computer, the Hamiltonian is initialized to the ground state of H1, and the Hamiltonian is gradually changed from H1 to H0. The Hamiltonian at each time is allowed to continue for approximately the relaxation time, and the gate operation determined by each Hamiltonian is repeated during that time, during which each Hamiltonian is longitudinally relaxed to its ground state (sequential initialization). The longitudinal relaxation of each Hamiltonian is repeated for each duration as described above, leading to the desired ground state. In addition to this, the following method is used in combination. (1) Repeatedly guiding the system to the desired ground state and measuring it to obtain a distribution of measured values. (2) We set a trial function (variation function) that realizes the distribution of the obtained measurements, and improve the accuracy of the solution by using the calculus of variations. Machine learning is also used to determine the trial function. (3) A random gate and its inverse gate are inserted between the gates (original gates) of the circuit determined by the Hamiltonian, and the random gate is integrated with the original gate to randomize the original gate and eliminate systematic errors. EXAMPLES

[0024] Physical systems generally have the tendency to transition from a high-energy state to a low-energy state. This is called longitudinal relaxation. This embodiment aims to operate quantum computing using longitudinal relaxation as a driving force. This principle is essentially different from that of ordinary quantum computers. Ordinary quantum computers are based on the premise that there is no relaxation and the state is maintained (so-called coherent), and research and development is being carried out with this premise in mind.

[0025] As mentioned above, physical systems naturally have the tendency to transition from a high-energy state to a low-energy state. In other words, physical systems have an inherently incoherent nature. This embodiment makes good use of this incoherent nature to operate a quantum computer.

[0026] 1A shows an example of the hardware configuration of a quantum computer according to an embodiment. As disclosed in Patent Document 1, for example, the hardware of the quantum computer 1000 functions as a processor or memory using a quantum register 1001 that holds quantum bits, a control gate 1002 that operates the quantum register 1001, and a readout means 1003 that observes the state of the quantum register 1001. Such a quantum computer 1000 is generally controlled by a classical computer 2000.

[0027] Below, we will explain the embodiment with concrete examples. The problem to be solved is converted into a Hamiltonian form so that the solution is the ground state of the Hamiltonian H0. A Hamiltonian is defined for each problem to be solved. For a combinatorial optimization problem, the Ising spin Hamiltonian H0 = -Σ jk J jk Z j Z k -Σ j h j Z j (See A. Lucas, arXiv:1302.5843v3) where Z j is the Z component of the Pauli spin matrix, J jk and jis a coefficient that is determined by the problem. In quantum chemistry, pharmaceuticals, and quantum many-body systems, the Hamiltonian H0 of each system is given, and finding the ground state is also a fundamental problem.

[0028] In this embodiment, an NISQ computer, which is a gate-type quantum computer not equipped with the above-mentioned error correction function, is used as the quantum computer 1000. This embodiment can be implemented by executing the process described below in an NISQ computer known at the time of filing and hardware included in the concept thereof.

[0029] To obtain the ground state of H0, first set the state of quantum computer 1000 to the ground state of Hamiltonian H1. H1 is, for example, H1=-BΣ j X j Here, X j is the X component of the Pauli spin matrix, and B is an appropriate coefficient. Quantum bits can be intuitively understood in terms of spin. |0> of a quantum bit corresponds to up spin, and |1> corresponds to down spin. H1= -BΣ j X j corresponds to a spin system with a uniformly applied transverse magnetic field, and B corresponds to the magnitude of the transverse magnetic field. The spins are in the ground state when they are aligned with the magnetic field, and this is easy to achieve. That is, it is easy to achieve the ground state of the Hamiltonian H1. This is easy because of longitudinal relaxation. However, it is not easy to obtain the ground state of H0 instead of H1. This is because there are generally many locally lowest energy states in H0, and there is a high probability that the spins will fall into these states during longitudinal relaxation.

[0030] Therefore, we first prepare the system in the ground state of Hamiltonian H1, gradually change the Hamiltonian, and longitudinally relax it to the ground state determined by the Hamiltonian each time. If the Hamiltonian is only changed slightly, the ground state will be almost the same, and it is possible to lead to the correct ground state with a high probability. The longitudinal relaxation to the ground state determined by the Hamiltonian each time can also be said to be initialization each time.

[0031] As described in Patent Document 1, problem setting and control in the quantum computer 1000 can generally be performed by a classical computer (von Neumann type computer) 2000. That is, the problem to be solved is set by the classical computer 2000 that controls the quantum computer 1000. Based on the problem set by the classical computer 2000, the quantum bit of the quantum register 1001 is prepared in the ground state of the Hamiltonian H1 by the control gate 1002 as the initial state setting of the quantum computer 1000. This operation is realized by controlling the gate based on the control program 2001 implemented in the classical computer 2000. A configuration example of the classical computer 2000 is omitted since it is also described in Patent Document 1.

[0032] Fig. 1B is a diagram showing the principle of this embodiment, in which Fig. 1B(a) is expressed in terms of an energy diagram, and Fig. 1B(b) is an example of a specific quantum circuit of a quantum computer.

[0033] Let us express the Hamiltonian as H(t) = s(t)H0+ [1 - s(t)]H1 using the parameter s(t). s(t) is s(t0) = 0 at time t = t0, and s(t0+T) = 1 at time t = t0+T, and s(t) gradually changes from 0 to 1. The Hamiltonian at t = t0 is H(t0) = H1, and its ground state is |ψ(t0)> (see Figure 1B(a)).

[0034] The circuit in Figure 1B has a Hamiltonian H(t j ) = s(t j )H0+ [1 - s(t j )]H1(where H0= -Σ jk J jk Z j Z k -Σ j h j Z j , H1 = -BΣ j X j ) If the Hamiltonian changes, the quantum circuit will also change. In this embodiment, the above Hamiltonian is used as an example for explanation.

[0035] For example, H1 = -BΣ j X j The ground state of is |ψ(t0)>=|++…+>. Here, |+> = (|0>+|1>) / √2. At time t=t0, the state prepared at |ψ(t0)> is gradually changed. In the time interval from t=t0 to t=t1=t0+δt1, the Hamiltonian is set to H(t1) = s(t1)H0+ [1 - s(t1)]H1 and the state is allowed to evolve over time. The unitary operator representing the time evolution is given by U1= exp[-iH(t1) δt1] according to the Schrödinger equation. Here, the Planck constant is set to h / 2π=1. δt1 is set to approximately the longitudinal relaxation time. Through longitudinal relaxation (relaxation to the ground state 100 in FIG. 1B(a)), the state becomes approximately the ground state |ψ(t1)> of H(t1) = s(t1)H0+ [1 - s(t1)]H1.

[0036] Similarly, in the time interval from t = t1 = t0 + δt1 to t = t2 = t1 + δt2, the Hamiltonian is set to H(t2) = s(t2)H0 + [1 - s(t2)]H1 to evolve the state over time. The unitary operator representing the time evolution is given by U2 = exp[-iH(t2) δt2]. δt2 is set to approximately the longitudinal relaxation time. Through longitudinal relaxation (relaxation to the ground state 200 in Figure 1B(a)), the state roughly becomes the ground state |ψ(t2)> of H(t2) = s(t2)H0 + [1 - s(t2)]H1. This is repeated below. For infinitesimal time δt j is T=Σ j n δt j Assume that time T is divided into n parts so that the following is satisfied. After n repetitions, the state is approximately H(t n ) = H(t0+T) = ground state of H0 |ψ(t n )>.

[0037] Changes in the Hamiltonian H are reflected in the unitary operation based on that Hamiltonian, and these changes are reflected in the control gate, which is the subject of the operation. However, the longitudinal relaxation used in this embodiment does not accompany unitary operations. Even if you try to perform unitary operations, a natural phenomenon of transition from a high energy state to a low energy state is added. This is longitudinal relaxation.

[0038] The longitudinal relaxation time depends on the system being studied and can be obtained by simulation or experiment. The value of the longitudinal relaxation time is roughly determined for a given system, but it varies by an order of magnitude for different systems, and can range from microseconds to tens of seconds or more.

[0039] Above, δt j At each time instant, the Hamiltonian H(t j ) = s(t j )H0+ [1 - s(t j )]H1 and the unitary operator U j = exp[-iH(t j ) δt j ] and described how the time evolution takes place.

[0040] In summary, we use the parameter s(t) to define the Hamiltonian H(t) = s(t)H0+[1-s(t)]H1, where s(t) is the time domain from t=t0 to t=t. n =t0+T, from s(t0=0 to s(t n ) = 1, the initial state is the ground state of H1 |ψ(t0)>, and from time t = t0 to t = t1 = t0 + δt1, the unitary operation U1 = exp[-iH(t1) δt1] is performed, from time t = t1 to t = t2 = t1 + δt2, the unitary operation U2 = exp[-iH(t2) δt2] is performed, ... and δt j (j=1, 2, …, n) is the longitudinal relaxation time and T=Σ j n δt j and δt j The state is relaxed at each time step, and the initial state |ψ(t0)> is evolved to |ψ(t0+T)>.

[0041] Next, we will show how gates are operated on a quantum computer using the quantum circuit diagram in Figure 1B(b) as an example. Figure 1B(b) shows the case of a 4-qubit system. An example of the Hamiltonian is the previously mentioned H0 = -Σ jk J jk Z j Z k -Σ j h j Z j , H1 = -BΣ j X j First, the state is |ψ(t -1 )>=|0…0> (input state 111 in Fig. 1B(b)), and each qubit is rotated along the Y axis R Y By applying (π / 2) (Y-axis rotation gate 112 in FIG. 1B(b)), the ground state of H1 |ψ(t0)> = |++…+> = [Π j R Y (π / 2) j ]|0…0> (the state at t=t0 and its position 113 in FIG. 1B(b)).

[0042] Figure 2 is a conceptual diagram showing the relationship between the Bloch sphere and a quantum bit. Y (π / 2) can be understood graphically using the Bloch sphere shown in Figure 2. The Bloch sphere illustrates a quantum bit with the north pole as |0> and the south pole as |1>. |+> = (|0>+|1>) / √2 is expressed as (x,y,z) = (1,0,0). Therefore, to get from |0> to |+>, we need to rotate it by π / 2 around the Y axis. R Y (π / 2) represents this operation. In addition, if we write the Y-axis rotation mathematically, Y is the Y component of the Pauli spin matrix, R Y (θ) = exp(-iθY / 2). Similarly, R X (θ) = exp(-iθX / 2), R Z (θ) = exp(-iθZ / 2) , R I (θ) = exp(-iθ / 2).

[0043] |ψ(t0)> = |++…+> is H1 = -BΣ j X jis an eigenvalue of -B, and H1|ψ(t0)> = -B|ψ(t0)>. If we keep the Hamiltonian at H1, then |ψ(t)> = exp[-iBΣ j X j (t-t0)]|++…+>. exp[-iBΣ j X j (t-t0)] corresponds to the X-axis rotation on the Bloch sphere, and even if the vector (1,0,0) is rotated around the X-axis, it remains at (1,0,0), so if the Hamiltonian is kept at H1, the state will remain at |ψ(t0)>.

[0044] In the calculation of this embodiment, the Hamiltonian is set to H(t1) = s(t1)H0+ [1 - s(t1)]H1 during the time interval from t=t0 to t=t1=t0+δt1. H(t1) = s(t1)(-Σ jk J jk Z j Z k -Σ j h j Z j ) + [1 - s(t1)](-BΣ j X j ) and written as unitary operators: U1 = exp(-i{s(t1)(-Σ jk J jk Z j Z k -Σ j h j Z j ) + [1 - s(t1)]( -BΣ j X j )} δt1) The argument of the exponential function contains the X and Z operators and is non-commutative, but if we divide the time interval δt1 into N parts and make it infinitesimal, we can break it down into individual products. That is, U1 ≒ Π p=1 N {Π j exp(-i[1 - s(t1)]( -BX j ) δt1 / N) ×Π j exp[-is(t1)( -h j Zj ) δt1 / N] × Π jk exp[-is(t1)( -J jk Z j Z k ) δt1 / N]} p It is.

[0045] θ t =2[1 - s(t1)]( -B) δt1 / N, θ j = 2s(t1)( -h j ) δt1 / N, θ jk = 2s(t1)( -J jk ) δt1 / N, R ZZ (θ jk ) = exp(-iθ jk Z j Z k / 2), then Π j exp(-i[1 - s(t1)]( -BX j ) δt1 / N) is the X-axis rotation of each quantum bit R X (θ t ) (X-axis rotating gate 114 in FIG. 1B(b)), Π j exp[-is(t1)( -h j Z j ) δt1 / N] is the Z-axis rotation of each quantum bit R Z (θ j ) (Z-axis rotating gate 115 in FIG. 1B(b)), Π jk exp[-is(t1)( -J jk Z j Z k ) δt1 / N] is the R ZZ (θ jk ) (R in Figure 1B(b) ZZ Gate 116). R ZZ If the system can directly realize the ZZ interaction, the gate can be used. If the system cannot directly realize the ZZ interaction, the controlled phase gate can be used. jk = I - (IZ j )(IZ k ) / 2, this can be realized using a CZ gate and a Z-axis rotary gate.

[0046] In this way, U1 is the X-axis rotating gate, the Z-axis rotating gate, and the R ZZ If we can break it down into gates, we can arrange them in a sequence to construct a quantum circuit. Figure 1B(b) illustrates this. As mentioned above, we can divide the Hamiltonian H(t) into coefficients h j 1-qubit operator and coefficient J jk and the coefficient h j Unitary operations in terms and coefficient J jk The unitary operations in the terms are respectively one-qubit gates (114, 115) and multi-qubit gates (R ZZ The calculation was performed using gate 116).

[0047] In this embodiment, the coefficient h j Term and coefficient J jk The Hamiltonian H(t)=Σ j h j Z j +Σ jk J jk Z j Z k The unitary operation U = exp[-iH(t) δt] for the time interval δt determined by is divided into N parts and U = Π q=1 N {Π j [exp(-ih j Z j δt / N)]Π jk [exp(-iJ jk Z j Z k δt / N)} q Then, exp(-ih j Z j δt / N) and the 1-qubit gate corresponding to exp(-iJ jk Z j Z k The operation is realized by the operation of a two-qubit gate, which is a multi-qubit gate corresponding to the multi-qubit gate (δt / N).

[0048] In U1, δt1 is divided into N, so the same gate operation is repeated N times. Each gate operation is digital when viewed individually, but when viewed on average over the entire time interval δt1, the Hamiltonian is H(t1) = s(t1)( -Σ jk J jk Z j Z k -Σ j h j Z j ) + [1 - s(t1)]( -BΣ j X j ) is obtained. δt1 is set to approximately the longitudinal relaxation time, so that the system will generally relax to the ground state of H(t1) after the passage of time δt1. The reason for dividing it into N is to reduce the effect of each gate operation and average the system response. For this purpose, the setting value of N is preferably N ≒ 10 or more, but this is not necessarily strict and can be determined taking into account the longitudinal relaxation time of the system and the time required for each gate. U2, …, U n The method for achieving this is the same as that for U1, and we use H(t2), …, H(t n ) is the basis.

[0049] Through the above gate processing, |ψ(t n )>. n )> in the Z basis, we obtain eigenvalues ​​of +1 and -1 corresponding to |0> and |1>. In problems such as combinatorial optimization problems where the solution state is |0> or |1>, the measured value becomes the solution candidate as is. By repeating the above calculations and measurements, multiple solution candidates can be obtained, and these can be examined one by one to select the optimal solution.

[0050] In quantum chemistry and quantum many-body problems, the solution is not |0>, |1>, but a superposition state |ψ(t n )>=Σ i a i For an m-qubit system, i = 0, …, 2. m - 1. If we measure |ψ(t n )>=Σ i a i The wave packet collapses to one of |i>. The probability that the wave packet collapses to |i> at each measurement is |ai | 2 By repeating the calculation and measurement, |a i | 2 That is, the distribution of |ψ(t n )>=Σ i a i |i> can be roughly calculated. |a i | 2 Since we only got a i However, the solutions to quantum chemistry and quantum many-body problems have symmetry, and a i Even if we cannot accurately measure the phase of |a i | 2 If you can find the answer, you can say that you have obtained the answer. i | 2 from a including phase i A method for restoring the above will be described in Example 3. EXAMPLES

[0051] In the first embodiment, it has been described that the calculation and measurement are repeatedly performed. By repeatedly performing the calculation and measurement, in a problem where the solution state is |0> or |1>, multiple solution candidates are obtained, and when the solution state is |ψ(t n )>=Σ i a i In the problem where |i>, |a i | 2 The distribution of was obtained. In addition to this, the purpose of repeating the calculation and measurement is to reduce the influence of operational errors.

[0052] The solution state is |ψ(t n )>=Σ i a i In the case of |i>, the wave packet collapses into one of |i> during measurement, but due to initialization errors, gate errors, and measurement errors, |ψ(t n )>=Σ i a i There is a possibility that the wave packet may be collapsed to |j>, which is not a component of |i>. However, if the errors are sufficiently random, it is possible to collapse the wave packet to the original |i> as the average value by increasing the number of samples.

[0053] This is shown in Figure 3. Only the peak gives the solution, and the width of the peak is based on the motion error. In other words, it is possible to eliminate the motion error through multiple measurements.

[0054] For this method of elimination to work well, the errors must be sufficiently random and there must be no systematic errors (coherent errors). Therefore, a random gate and its inverse gate are added in pairs to eliminate systematic errors on average. In addition, the random gate and its inverse gate are integrated separately with the original gate to prevent an increase in the number of gates.

[0055] 4A to 4E show an example of a procedure for randomizing each gate of the quantum circuit of FIG. 1B(b). FIG. 4A is a circuit diagram in which commutative gates in FIG. 1B(b) are exchanged to prevent two-qubit gates from being consecutive. FIG. 4B shows a circuit diagram in which R Z A random gate is selected for each gate division, and the selected random gate and its inverse gate are inserted between all gates in the division. Figure 4C shows a single gate in which each gate and the random gate and inverse gate inserted before and after it are integrated. Figure 4D shows a single gate in which the random gate and the inverse gate inserted before and after each gate are integrated. Y A random gate is selected at each gate interval, and the selected random gate and its inverse gate are added. Figure 4E is a diagram in which each gate and the random gate and inverse gate inserted before and after it are integrated into a single gate. Figure 1B(b) and Figures 4A-4E perform the same task, but the systematic errors are canceled out on average in the quantum circuit of Figure 4E.

[0056] The procedure makes use of the following property, where the Pauli matrix σ i ∈{X, Y, Z} and a rotation gate R i (θ)∈{R X (θ), R Y (θ), R Z (θ)} are collectively expressed using the subscript i. i = iR i (π), σ i ±1 / 2 = exp(±iπ / 4)R i (±π / 2) and σi and R i (θ) is commutative. When i≠j≠k, σ i R j (θ)σ i = R j (-θ), σ i 1 / 2 R j (θ)σ i -1 / 2 = R k (ε ijk θ), σ i -1 / 2 R j (θ)σ i 1 / 2 = R k (-ε ijk θ), where ε XYZ =ε YZX =ε ZXY =+1, ε XZY =ε ZYX =ε XZY =-1.

[0057] Step 1: Make sure that the two-qubit gates are not arranged consecutively as much as possible. In Figure 1B(b), R ZZ (θ 12 ) Gate and R ZZ (θ 23 ) gate, and R ZZ (θ 34 ) Gate and R ZZ (θ 23 ) gates are lined up in a row (116). R ZZ (θ 23 ) Gate and 115 R Z (θ j ) gate is commutative, so we exchange it (R in Figure 4A). ZZ gate 116b and Z-axis rotating gate 115').

[0058] Step 2: 2.1: R in the circuit X , R Y , R Z In Figure 4A, we select the gate with the most X and R Z are the same number. So R Z Let be the maximum number of gates. 2.2: The gate selected in 2.1 is R j (In 2.1, R Z ) The circuit is R j For each division, {I, σ i , σ i -1 ,σ j 1 / 2 ,σ j -1 / 2} and insert the selected random gate and its inverse gate between all gates in each section. Since the random gate and its inverse gate are paired, this circuit is equivalent to the original circuit (Figure 4B). j A two-qubit gate is placed in the partition R ii If (i ≠ j), we define a random gate pair as {I, σ j ,σ i -1} In FIG. 4B, R Z The gates are the divisions. For example, the two-qubit gate in the division from 115' to the Z-axis rotation gate 125' is R ZZ gate, i = j. Therefore, the random gate is {I, Z, Z -1 , Z 1 / 2 , Z -1 / 2}. 2.3: The selected random gate and its reverse gate are integrated with the original gates at the rear and front, respectively (Fig. 4C). 114, 115', and 125' in Fig. 4B become gate 114', which is an integrated gate of X-axis rotating gate and random gate, gate 115'', which is an integrated gate of Z-axis rotating gate and random gate, and gate 125'', which is an integrated gate of Z-axis rotating gate and random gate in Fig. 4C.

[0059] Step 3: R not selected in 2.1 j Select the gate with the most number of gates in and perform step 2 in the same way. X However, the limited space available on the paper allows X is not displayed sufficiently. Therefore, in the example of Figure 4, step 3 is omitted.

[0060] Step 4: The remaining R not selected in 2.1 and step 3 j Repeat step 2 for R. Y It becomes. R Y There is an R in the gate division. ZZ Therefore, the random gate in 2.2 is {I, Y, Y -1} (FIG. 4D). The integration in 2.3 results in FIG. 4E. 112, 114', 115'', 116a, and 116b in FIG. 4D are gates 112', 114', and 115' in which the Y-axis rotation gate and the random gate are integrated, and gates 114'', 114'', and 115'', in which the Y-axis rotation gate and the random gate are integrated, are gates 115'' and 115'', in which the R ZZ The gate and the random gate are integrated to form gates 116a' and 116b'.

[0061] Step 5: This completes the randomization of the gates. You can repeat steps 2-4 to randomize further. j The selection remains the same.

[0062] Through the above transformation, 112, 114, 115, and 116 in Figure 1B(b) become 112', 114'', 115''', 116a and 116b' in Figure 4E. Comparing the two, we can see that the gates have been randomized. This randomization is performed each time the circuit is operated and measured. As a result, systematic errors are eliminated on average. By randomizing the gates, a function to offset systematic errors is embedded in the circuit itself, but the offsetting effect can be further improved if randomization is performed each time the circuit is operated and measured.

[0063] With the systematic errors eliminated, the peaks are now in the correct positions in Figure 3.

[0064] FIG. 1B(b) and FIG. 4A to E are R ZZ This is a circuit diagram when the gate can be directly implemented. ZZ If the gate cannot be directly implemented, use CZ jk = I - (IZ j)(IZ k This is realized using a CZ gate and a Z-axis rotating gate based on the relationship:

[0065] Figure 5 shows the Z j Z k = I+Z j +Z k -2CZ jk According to the R in Figure 1B(b), ZZ (θ) gate rewritten. Phase R based on identity gate I I Since (θ) = exp(-iθ / 2) does not affect the measured value, it is usually omitted from the circuit diagram, but this phase must be taken into consideration when implementing the circuit, so it is clearly shown in Figure 5. EXAMPLES

[0066] In Example 1, the solution is a linear superposition state |ψ(t n )>=Σ i a i In quantum chemistry and quantum many-body systems, the challenge is to obtain the state of a many-body electron system, and this is exactly the case when |ψ(t n )>=Σ i a i In particular, finding the ground state is a fundamental problem.

[0067] Electrons are fermions and therefore have different properties from quantum bits; however, fermions can only be in two states: either occupying a certain state or not. This property is similar to quantum bits, which can only be in the |0> and |1> states. Therefore, it is possible to represent fermions using quantum bits according to certain transformation rules.

[0068] According to the transformation rule, the Hamiltonian of m many-electron systems is generally

[0069]

number

[0070] (A. Kandala, et al., Nature 549, 242 (2017)), where f j is a coefficient determined by the overlap of electron orbits, etc., P j is composed of the tensor product of m I and Pauli operators. For example, the ground state of a hydrogen molecule is a two-electron system, and sufficient accuracy can be achieved by considering the 1s state of each of the two atoms. When converted to a quantum bit, the Hamiltonian is H0 = f0I1I2 + f1Z1Z2 + f2Z1I2 + f3I1Z2 + f4X1X2 (see N. Moll, et al., Quantum Sci. Technol.3, 030503 (2018).).

[0071] To obtain the ground state of H0, we simply perform the calculations shown in Figure 1B with H1 = -B(X1 + X2).

[0072] Figure 6 shows a circuit diagram for obtaining the ground state of a hydrogen molecule. The term f0 is omitted here because it only provides an offset. R X (-θ) = Y -1 / 2 R Z (θ)Y 1 / 2 From the relationship of R XX (θ) = (Y1 -1 / 2 Y2 -1 / 2 ) R ZZ (θ) (Y1 1 / 2 Y2 1 / 2 ) so R XX Gate is R ZZ It can be converted to an operator (R ZZ R realized by gate and Y-axis rotating gate XX The randomization of the circuit in FIG. 6 can be performed in the same manner as in FIG. 4A to FIG. 4E. ZZ The gate can be further converted into a CZ gate and a rotary gate as in the case of FIG.

[0073] Using the circuit in Figure 6, multiple calculations are performed to obtain |a i | 2 As a result of obtaining the distribution of, for example, |a 00 | 2 ≒ 0.5, |a 11 |2 ≒ 0.5. Here, the subscript i is expressed in binary. In this case, the solution is |ψ> = a 00 |00> + a 11 |11> can be predicted. Therefore, we use the variational method to minimize the expectation value E = <ψ|H0|ψ> by finding the parameter a 00 , a 11 If we calculate the phase, we get a i In this example, |a 00 | = |a 11 By adding the condition |, the variation parameters (amount of calculations) can be reduced.

[0074] The calculus of variations is the parameter a i Change the value of a little at a time to calculate E, and select the value that gives the smallest E. i is the final solution. For a small number of electron systems, as in this example, the calculus of variations is easy, but as the number of electrons increases, the amount of calculation required for the calculus of variations increases. In this case, it is effective to use a quantum computer to prepare |ψ>, and a method called Variable Quantum Eigensolver (VQE) corresponds to this.

[0075] In this example, we use the test function |ψ> = a 00 |00> + a 11 |11>. When the number of electrons increases, predicting the trial function itself can be difficult. In that case, it is effective to use various machine learning methods. For example, the method called the restricted Boltzmann machine predicts |ψ> = Σ i a i a in |i> i In expressing i , d j , W ij are the variational parameters, v = (v1, v2, …) are the variables of the area called the visible layer, and h = (h1, h2, …) are the variables of the area called the hidden layer. E(v,h) = -Σ i c i v i - Σ j d j hj - Σ ij W ij v i h j Define a virtual energy function E(v,h) determined by a i is expressed by E(v,h), where v i , h j = ±1. The partition function Z = Σ vh If we use exp[-E(v,h)], then p(v,h) = exp[-E(v,h)] / Z is the quantity that corresponds to the occurrence probability of (v,h). If we take a trace with respect to the hidden layer variable h, then a(v) = Σ h p(v,h). i Let i be the binary representation of i. m-1 i m-2 …i0, and then {0, 1} displayed as {+1, -1} is v m-1 v m-2 …If v0, then a(v) is a i That is, |a(v)| 2 = |a i | 2 By variational calculation, we find i , h j , W ij If we decide, the state |ψ> = Σ i a i |i> is obtained. In this way, |ψ> = Σ i a i If we use |i> as a trial function and perform further variation calculations to minimize E = <ψ|H0|ψ>, we can get a more accurate |ψ> = Σ i a i |i> is obtained.

[0076] To describe the machine learning method in a more general way, "Let the variational function |ψ(λ)> be determined by the parameter λ, and expand |ψ(λ)> in |j> to obtain |ψ(λ)> = Σ j a j (λ)|j>, then |a j (λ)| 2 In the case of the restricted Boltzmann machine, ci , d j , W ij corresponds to λ, and a j (λ) is a(v) = Σ h Equivalent to p(v,h).

[0077] And, a calculated with the parameter λ j (λ) to a j Let |ψ>=Σ be the initial value. j a j Let |j> be the variation function. Parameter a j By repeatedly calculating E while gradually changing a, we performed a variational calculation to find the minimum value of the expected value E = <ψ|H0|ψ>, and found the minimum value of E. j is output as the final solution.

[0078] FIG. 7 shows a series of steps as a block diagram. A problem is set (S701), a circuit is created (S702), and gates and random gates are integrated as described in FIG. 1B to FIG. 6 (S703). The ground state |ψ(t0+T)> of the Hamiltonian H0, which is the subject of the problem, is measured in the Z basis, and the quantum register 1001, the control gate 1002, and the readout means 1003 are operated to repeat calculations and measurements to obtain |ψ(t0+T)> (S704), and a frequency distribution of the measured values ​​of |ψ(t0+T)> is obtained (S705). When calculating using the calculus of variations, an appropriate trial function is determined (S706), and the ground state of the Hamiltonian H0 is made estimable by variation calculation (S707). As mentioned above, machine learning (S708) may be used to determine the trial function.

[0079] The method of this embodiment realizes quantum computing by using the longitudinal relaxation phenomenon as a driving force. The ground state corresponds to the correct state, and the excited state corresponds to the incorrect state. Transitions to the excited state that may occur during a calculation correspond to operational errors. Since the excited state longitudinally relaxes to the ground state, operational errors are naturally corrected. In other words, error correction is performed according to the natural properties of physical systems. Therefore, the method of this embodiment also works with gate-type quantum computers that do not implement quantum error correction, such as NISQ computers.

[0080] According to this embodiment, quantum computing driven by longitudinal relaxation makes it possible to handle real problems with a gate-type quantum computer (NISQ computer) that does not have an error correction function. In other words, it is possible to make a quantum computer with noise in its operation applicable to general real problems without error correction.

[0081] In addition, because it is a gate type, any unitary operation is possible (universal quantum computer), and general problems can be handled without being limited to combinatorial optimization problems. Furthermore, (1) because a distribution of measured values ​​is obtained, the solution for each quantum bit does not need to be a binary value of 0 or 1, and it can handle quantum mechanical problems in which the solution is a linear superposition state of |0> and |1>. (2) The accuracy of the solution can be improved by combining it with the calculus of variations and machine learning. (3) Systematic errors can be eliminated, which has the effect of improving the accuracy of the solution. [Explanation of symbols]

[0082] 100: Relaxation to the ground state 111: Input state 112: Y-axis rotating gate 112': Y-axis rotating gate and random gate integrated 113: State and position at t=t0 114: X-axis rotating gate 114': A gate that combines an X-axis rotating gate and a random gate 114: A gate that combines the 114' gate and a random gate 115: Z-axis rotating gate 115': Z-axis rotating gate 115'': Z-axis rotating gate and random gate integrated 116:R ZZ gate 116a:R ZZ gate 116a':R ZZ Gate 116b:R, which combines a gate and a random gate ZZ gate 116b':R ZZ A gate that combines a gate and a random gate 117:R ZZ R realized by gate and Y-axis rotating gate XX gate 125': Z-axis rotating gate 125'': Z-axis rotating gate and random gate integrated 200: Relaxation to the ground state

Claims

1. A quantum computer comprising: a quantum register for holding a quantum bit; a control gate for performing an operation on said quantum register; and a readout means for observing a state of said quantum register, said quantum computer repeating longitudinal relaxation to a ground state by gradually changing a Hamiltonian H(t) for a predetermined time, A quantum computer characterized in that the control gate performs a unitary operation determined by the Hamiltonian H(t) at each time over a time period of approximately the longitudinal relaxation time, relaxes the quantum state every time period of approximately the longitudinal relaxation time, and evolves the initial state prepared in the ground state into the ground state of the Hamiltonian of interest over time.

2. The quantum computer in which a unitary operation U = exp[-iH(t)δt] determined by a Hamiltonian H(t) is performed by the control gate consisting of a one-qubit gate and a multi-qubit gate, Using the parameter s(t), we define the Hamiltonian H(t) = s(t)H 0 +[1-s(t)]H 1 year, s(t) at time t=t 0 From t=t n =t 0 +T between s(t 0 )=0 to s(t n )=1, The initial state is H 1 The ground state of |ψ(t 0 )> and time t=t 0 From t=t 1 =t 0 +δt 1 Then, the unitary operation U 1 = exp[-iH(t 1 ) δt 1 ] at time t=t 1 From t=t 2 =t 1 +δt 2 Then, the unitary operation U 2 = exp[-iH(t 2 ) δt 2 ] to..., δt j (j=1, 2, …, n) is the longitudinal relaxation time and T=Σ j n δt j Fulfilling δt j The state is relaxed every time to the initial state |ψ(t 0 )> to |ψ(t 0 2. The quantum computer according to claim 1, wherein the quantum state is evolved in time such that ...

3. The Hamiltonian H(t) is expressed by coefficient h j 1-qubit operator and coefficient J jk Decompose it into the sum of two qubit operators, The coefficient h j A unitary operation determined by the one-qubit operator is performed by the one-qubit gate, and the coefficient J jk A unitary operation determined by the two-qubit operator is performed by the multi-qubit gate, 3. The quantum computer according to claim 2, wherein operations are realized by gate operations.

4. The coefficient h j The operator and the coefficient J jk The Hamiltonian H(t)=Σ given by the sum of the operators j h j Z j +Σ jk J jk Z j Z k The unitary operation U = exp[-iH(t)δt] for the time interval δt determined by is divided into N parts and U = Π q=1 N {Π j [exp(-ih j Z j δt / N)]Π jk [exp(-iJ jk Z j Z k δt / N)} q year, The exp(-ih j Z j δt / N) and the one-qubit gate corresponding to exp(-iJ jk Z j Z k 5. The quantum computer according to claim 3, wherein an operation is realized by an operation of a two-qubit gate, which is the multi-qubit gate corresponding to δt / N.

5. Said |ψ(t 0 +T)> is measured in the Z basis, and the above |ψ(t 0 +T)>, and repeat the calculation and measurement to obtain |ψ(t 0 +T)>, and use this to obtain the Hamiltonian H 0 3. The quantum computer according to claim 2, wherein the basis state of said quantum computer is estimable.

6. The quantum computer according to claim 1, characterized in that the gates are randomized by classifying 1-qubit gates by rotation axis, dividing a series of quantum circuits based on 1-qubit gates of a certain rotation axis, selecting a random gate for each division, inserting the selected random gate and its inverse gate between all gates in each division, and integrating the random gate and its inverse gate separately with gates of the original circuit.

7. 7. The quantum computer according to claim 6, wherein the random gate is commutative with a one-qubit gate of the certain rotation axis.

8. The certain rotation axis is defined as the i axis, and σ i If we define the Pauli matrix as {I, σ i , σ i -1 } or {I, σ i , σ i -1 σ i 1 / 2 , σ i -1 / 2 7. The quantum computer according to claim 6, wherein:

9. Let the variation function be |ψ>=Σ j a j |j> and parameter a j The initial value of |a j | 2 is determined to match the frequency distribution, and the expected value E=<ψ|H 0 |ψ> is calculated, and the parameter a j Repeat the calculation to find E by gradually changing a j 6. The quantum computer according to claim 5, wherein the quantum computer outputs the following as a final solution:

10. Let us assume that the variational function |ψ(λ)> is determined by the parameter λ, and expand |ψ(λ)> in |j> to obtain |ψ(λ)> = Σ j a j (λ)|j>, then |a j (λ)| 2 6. The quantum computer according to claim 5, wherein the parameter λ is determined so that the frequency distribution coincides with said frequency distribution.

11. a obtained from the above parameter λ j (λ) to a j As the initial value of |ψ>=Σ j a j |j> is the variation function, and the expectation value E=<ψ|H 0 A variation calculation was performed to find the minimum value of |ψ>, and the minimum value of E was given by a j 11. The quantum computer according to claim 10, wherein the quantum computer outputs as a final solution:

12. 10. The quantum computer according to claim 9, wherein the generation of the variation function |ψ> is performed by VQE.

13. A quantum state control method for a quantum computer comprising a quantum register for holding quantum bits, a control gate for performing an operation on the quantum register, and a readout means for observing a state of the quantum register, the method comprising the steps of: controlling the quantum bits by the control gate, The control gate controls the quantum bit, gradually changing the Hamiltonian H(t) over a predetermined time period, and repeating longitudinal relaxation to the ground state; A unitary operation determined by the Hamiltonian H(t) at each time is performed over a predetermined time by the control gate, thereby relaxing the quantum state at each predetermined time. A method of controlling the quantum state of a quantum computer in which an initial state prepared in the ground state is evolved over time to the ground state of the target Hamiltonian.

14. The quantum state control method of a quantum computer according to claim 13, wherein the quantum computer is a NISQ computer.

15. The quantum state control method for a quantum computer according to claim 13, wherein the predetermined time corresponds to a longitudinal relaxation time.

Citation Information

Patent Citations

  • Quantum computer system, control method and program for quantum computer system

    JP2013114366A

  • Estimating an energy level of a physical system

    WO2019220122A1