COMPILATION METHOD OF A QUANTUM CIRCUIT ON A QUANTUM PROCESSOR WITH IONS TRAPPED IN A TRAP
Patent Information
- Application Number
- DE602019062170
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2018-12-26
- Filing Date
- 2019-12-23
- Publication Date
- 2024-11-20
- Estimated Expiration
- 2039-12-23
AI Technical Summary
Existing methods for compiling quantum circuits on trapped ion quantum processors are not automated, leading to inefficiencies and limitations in handling circuits with more than four qubits due to exponential time and memory requirements, and require manual programming which can be suboptimal.
The method involves dividing the quantum circuit into layers of specific types, decomposing parts into phase polynomials, and recomposing them directly into Molmer-Sorensen entangled gates, allowing for automated and efficient compilation by grouping and transforming gates to reduce the number of complex operations.
This approach enables the compilation of quantum circuits in a completely automated and optimized manner, improving efficiency and speed, and facilitates scaling up to larger numbers of qubits by reducing the number of complex gates and enhancing precision.
Description
Technical field
[0001] The invention relates to the field of methods for compiling a quantum circuit on a trapped-ion quantum processor. Prior art
[0002] It is known from a first prior art Compiling quantum algorithms for architectures with multi-qubit gates, Martinez et al. a method for compiling a quantum circuit on a quantum processor allowing to obtain a compiled quantum circuit comprising a minimum of entangling quantum gates.
[0003] The method proposes to read all the gates of a quantum circuit and to construct unitary operations, i.e. a unitary matrix of size 2 N< x 2 N< , with N the number of qubits in the quantum circuit. Using an optimizer, entangling gates that can be compiled on a trapped-ion quantum processor are added. At the same time, gates with local phase shifts parameterized at different angles are added to converge the unitary matrix. As long as the unitary matrix does not converge, a new entangling gate is added to the quantum circuit.
[0004] Although this method is optimal in terms of result, in fact, the circuit obtained includes the smallest number of entangling quantum gates, it is exponential in time and memory, a unitary matrix of size 2 N< x 2 N< having to be calculated at each increment. This method cannot therefore be applied to quantum circuits with more than four qubits.
[0005] Another known method is described in the document Basic circuit compilation techniques for an ion-trap quantum machine, Maslov and Use of global interactions in efficient quantum circuit constructions, Maslov and Nam.
[0006] This method proposes to program certain blocks of the quantum circuit in a language readable by a trapped-ion quantum processor. Some blocks of the quantum circuit are rewritten using entangling quantum gates that can be compiled on a trapped-ion processor.
[0007] However, this method is not automatic, since the programming is done by a programmer. Thus, the risk is that the programmer will not detect certain blocks of the quantum circuit to be rewritten. This method is therefore not optimal. Technical problem
[0008] These known solutions do not allow for the automated compilation of a quantum circuit, while allowing for rapid and efficient compilation of the entire quantum circuit.
[0009] The invention at least partially overcomes the drawbacks of the prior art and optimizes the compilation of quantum circuits on trapped-ion quantum processors.
[0010] To this end, according to the invention, it will be interesting to divide the quantum circuit into layers of different types. The layers of the quantum circuit are divided so as to comprise only quantum gates of a certain predetermined type. This division makes it possible, on the one hand, to be able to compile these layers according to a compilation process specific to the quantum gates contained in these layers. The application of specific compilation processes makes it possible to greatly increase the efficiency of the compilation method. On the other hand, the division makes it possible to group several quantum gates in the same layer in order to apply a single compilation process for a plurality of gates, thus making it possible to significantly increase the speed of compilation of the entire quantum circuit.
[0011] The compilation method according to the invention therefore allows the compilation of a quantum circuit in a completely automated and optimized manner, both in terms of efficiency and in terms of compilation speed.
[0012] There is provided a method of compiling a quantum circuit according to claim 1.
[0013] According to another aspect, there is provided a method of compiling a quantum circuit on a trapped ion quantum processor comprising: a step of decomposing all or part of the quantum circuit into one or more phase polynomials, a step of directly recomposing said all or part of the quantum circuit in the form of Molmer-Sorensen entangling quantum gates without passing into the form of CNOT quantum gates, a step of grouping the recomposed quantum gates to obtain a compiled circuit.
[0014] Furthermore, the decomposition of all or part of the quantum circuit into phase polynomials makes it possible to rewrite these parts of the quantum circuit directly in the form of Molmer-Sorensen entangling quantum gates. In other words, it is not necessary to recompose these parts of the quantum circuit in the form of other CNOT-type gates. The method therefore makes it possible to limit the number of steps during the recomposition, thus increasing, on the one hand, the efficiency of its implementation, and, on the other hand, its compilation speed.
[0015] The features set out in the following paragraphs may optionally be implemented. They may be implemented independently of each other or in combination with each other:
[0016] Preferably, the step of compiling the entangling layers also transforms the quantum gates of these entangling layers so that: all or at least some of these collective or entangling quantum gates apply simultaneously to at least three qubits, advantageously apply simultaneously to the majority of qubits, even more advantageously apply simultaneously to all qubits.
[0017] Thus, according to the invention, it will be interesting to transform the quantum circuit so that only quantum gates executable by the quantum processor remain, and to automatically share some of these quantum gates with several qubits, and preferably with a majority or even all of the qubits. This sharing makes it possible, on the one hand, to make the compilation very efficient, and, on the other hand, the sharing makes it possible to limit the number of complex quantum gates in this quantum circuit, thus making it easier to scale up when the number of qubits in the circuit is increased because it is then sufficient to also share these quantum gates on the additional qubits.
[0018] The compilation method according to the invention therefore allows the compilation of a quantum circuit in a completely automated and efficiency-optimized manner.
[0019] Preferably, the step of compiling the entangling layers comprises: a decomposition of each entangling layer into a phase polynomial, a direct recomposition in the form of quantum gates of the third category and the fourth category of quantum gates without going through a form of quantum gates of the first category of quantum gates.
[0020] Thus, the decomposition of all or part of the quantum circuit into phase polynomials makes it possible to rewrite these parts of the quantum circuit directly in the form of Molmer-Sorensen entangling quantum gates. In other words, it is not necessary to recompose these parts of the quantum circuit in the form of other CNOT-type gates. The method therefore makes it possible to limit the number of steps during the recomposition, thus increasing, on the one hand, the efficiency of its implementation, and, on the other hand, its compilation speed.
[0021] The local layer compilation step includes: a calculation of local layer local matrices to be applied to each qubit, a decomposition of each local layer local matrix into a sequence of quantum gates of a rotation along the Z axis, a rotation along the X axis, a rotation along the Z axis, a sub-decomposition of the rotation along the X axis into a sequence of quantum Hadamard gates, a rotation along the Z axis, Hadamard, a sub-sub-decomposition, of the Hadamard gate into a sequence of quantum gates of a rotation along the Z axis, a rotation along the X axis, a rotation along the Z axis, all three rotations by an angle of π or π / 2, a mutualization between qubits, of the rotation gates along the X axis by an angle of π / 2.
[0022] This decomposition of the local layers makes it possible to obtain quantum gates with a rotation along the X axis having a fixed angle. The X-axis rotation gates therefore no longer depend on the local matrices, which makes it possible to share the X-axis rotation quantum gates with a majority of qubits in the local layers, and preferentially with all qubits in the local layers. Furthermore, such a decomposition makes it possible to obtain a quantum circuit recomposed from the local layers which contains only two collective X-axis rotation gates, which greatly improves the accuracy of the process.
[0023] Preferably, the method comprises: a preliminary step of transforming the quantum circuit so that it no longer contains anything other than: * a first predetermined category of two-qubit quantum gates, * and / or one-qubit quantum gates.
[0024] Thus, this preliminary step makes it possible to transform the circuit so that it only contains quantum gates on which the steps of the method according to the invention can be implemented.
[0025] Preferably, during the compilation step of the entangling layers: the number of collective or entangling quantum gates of the third category of quantum gates in the compiled entangling layers is minimized with priority over the minimization of quantum gates of the fourth category.
[0026] Since the entangling and collective gates of the third category are the most expensive in terms of execution time and process precision, minimizing their number as a priority makes it possible to obtain a quantum circuit that is less expensive in terms of execution time and more precise during its compilation.
[0027] Preferably, the first predetermined category of two-qubit quantum gates comprises: * two-qubit diagonal quantum gates, * CNOT quantum gates, the second predetermined category of one-qubit quantum gates includes: * one-qubit diagonal gates, the third predetermined category of N-qubit collective or entangling quantum gates includes: * N-qubit Molmer-Sorensen entangling gates, * X-axis rotating collective gates, the fourth predetermined category of one-qubit quantum gates includes: * Z-axis local phase shifts.
[0028] The quantum gates of the four predetermined categories therefore correspond to the quantum gates that can be executed on a trapped-ion quantum processor. Brief description of the drawings
[0029] Other characteristics, details and advantages of the invention will appear on reading the detailed description below, and on analyzing the attached drawings, in which: Fig. 1 [ Fig. 1 ] shows a flowchart of the main steps of a method for compiling a quantum circuit on a trapped-ion quantum processor according to one embodiment of the invention. Fig. 2a [ Fig. 2a ] shows a representation of a quantum circuit comprising two-qubit quantum gates of the first predetermined category and one-qubit quantum gates, before the separation of the quantum circuit into local and entangling layers. Fig. 2b [ Fig. 2b ] shows a directed acyclic graph of the quantum circuit of the Figure 2b , used, in a particular embodiment of the invention, to carry out the separation of the quantum circuit into local and entangling layers. Fig. 3a [ Fig. 3a] shows a pseudo code for extracting local layers according to one embodiment of the invention. Fig. 3b [ Fig. 3b ] shows a pseudo code for extracting the entangling layers according to one embodiment of the invention. Fig. 4 [ Fig. 4 ] shows a pseudo code of a Gaussian elimination used for the recomposition of the entangling layers according to an embodiment of the invention. Fig.5 [ Fig. 5 ]. shows an example of recomposition of an entangling layer into a quantum circuit comprising only collective or entangling quantum gates with N qubits of a third predetermined category and quantum gates with one qubit of a fourth predetermined category Fig. 6 [ Fig. 6] shows another example of recomposing an entangling layer into a quantum circuit comprising only collective or entangling quantum gates with N qubits of a third predetermined category and single-qubit quantum gates of a fourth predetermined category. Fig. 7 [ Fig. 7 ] shows an example of a quantum circuit resulting from the compilation of a local layer. Description of the embodiments
[0030] Reference is now made to the figure 1 , in which the main steps of a method for compiling a quantum circuit on a trapped-ion quantum processor are illustrated.
[0031] In the obtaining step S1, a quantum circuit is obtained. The quantum circuit is advantageously composed of two-qubit quantum gates of a first predetermined category of quantum gates, and / or one-qubit quantum gates.
[0032] More precisely, the quantum circuit is advantageously composed of CNOT quantum gates and any diagonal quantum gates with two qubits, forming the quantum gates of the first predetermined category, and / or quantum gates with one qubit.
[0033] A transformation step S2 may optionally be applied to the quantum circuit. The transformation step S2 is applied when the quantum circuit is composed of quantum gates other than those of the first predetermined category or of one-qubit quantum gates. The transformation step S2 comprises a transformation of the quantum circuit so that this quantum circuit now only comprises quantum gates of the first predetermined category of quantum gates, and / or one-qubit quantum gates.
[0034] In the case where the quantum circuit comprises only quantum gates of the first predetermined category of quantum gates, and / or single-qubit quantum gates, the separation step S3 is implemented directly after the obtaining step S1.
[0035] A separation step S3 is then implemented. The separation step S3 comprises the separation of the quantum circuit into local layers and entangling layers. The local layers comprise only one-qubit quantum gates. The entangling layers comprise only two-qubit quantum gates of the first category and / or one-qubit quantum gates of a second category.
[0036] Single-qubit quantum gates of the second category are any diagonal single-qubit quantum gates.
[0037] The separation step S3 includes the generation of a directed acyclic graph associated with the quantum circuit.
[0038] There Figure 2a illustrates an example of a quantum circuit with three qubits q0, q1 and q2, comprising fourteen quantum operations, or fourteen quantum gates. The Figure 2b illustrates the directed acyclic graph corresponding to the quantum circuit of the Figure 2a . the directed acyclic graph allows us to represent the dependencies between quantum operations. Thus, on this graph, the vertices represent quantum gates 1 to 14, and the edges represent the dependencies between quantum gates.
[0039] THE Figures 2a and 2b are purely illustrative, a quantum circuit comprising a number of qubits greater than 10 qubits, preferably greater than 20 qubits and preferably greater than 30 qubits, is used. A quantum circuit comprising a much larger number of quantum gates may also be used.
[0040] From the directed acyclic graph, the separation of the quantum circuit into local layers and entangling layers is performed. More precisely, the separation is done by successively building the largest local layer from the last entangling layer. This is repeated until all quantum gates are contained in a local or entangling layer.
[0041] The separation can be done sequentially, i.e. starting from gate 1 to gate 14.
[0042] There Figure 3aillustrates a pseudo code implemented for separating the circuit into local layers. In this pseudo code, the get_succ function returns the first gate succeeding quantum gate g on qubit q. In other words, the quantum gates succeeding quantum gate g on qubit q are grouped into a local layer, as long as these quantum gates are local quantum gates, i.e., 1-qubit quantum gates.
[0043] There Figure 3b illustrates a pseudo code implemented for the separation of the circuit into entangling layers. In this pseudo code, the function get_pred returns the first gate preceding gate g on qubit q and the function successors_ent returns the list of entangling quantum gates succeeding g, for all qubits.
[0044] That is, entangling layers can include one-qubit quantum gates, such as one-qubit diagonal quantum gates, in addition to two-qubit entangling gates, which cannot be included in local layers.
[0045] Steps S4 and S5 illustrate the compilation of the entangling layers. Steps S4 and S5 aim to obtain a quantum circuit comprising only quantum gates of a third category and a fourth category of quantum gates. The quantum gates of the third category and the fourth category of quantum gates correspond to quantum gates executable on a trapped-ion quantum processor.
[0046] Notably, quantum gates in the third category include N-qubit Molmer-Sorensen entangling gates and collective X-axis rotation gates. Quantum gates in the fourth category include local Z-axis (or Z-axis rotation) phase shifts.
[0047] The S4 entangling layer decomposition step involves decomposing the entangling layers into phase polynomials.
[0048] More precisely, at this stage of the process, the entangling layers comprise only two-qubit quantum gates of the first category and / or one-qubit quantum gates of the second category, i.e. CNOT quantum gates and / or one-qubit diagonal quantum gates.
[0049] Any circuit containing such quantum gates can be written in the form: U x 1 … x n > = e ip x 1 … x n h x 1 … x n > ,
[0050] Where h is linear and p is of the form: P x 1 … x n = Σθ i .f i x 1 … x n ,
[0051] And where the fi are linear functions.
[0052] In order to determine h , a matrix H comprising 0s and 1s, of dimension NxN is constructed, with N the number of qubits in the entangling layer. The matrix H is initialized to the identity, that is to say in a matrix comprising a diagonal of 1.
[0053] In order to determine linear functions fi , a hash table F is built. The hash table F hashes the elements of the matrix H towards floats.
[0054] A plurality of execution models is then defined. In particular, when reading the quantum gates of the entangling layer, when: the quantum gate is a CNOT quantum gate from qubit i to qubit j, then H[ j ] = H[ j ] XOR H[ i ],with XOR corresponding to the bitwise XOT of two bit sequences, the quantum gate is a rotation along Z of angle a on qubit i, then: ∘ if F[ H[ i ] ] is empty then F[ H[ i ] ] = a, ∘ otherwise, F[ H[ i ] ] = F[ H[ i ] ] + a, the quantum gate is a controlled phase of angle a from qubit i to qubit j, then: ∘ F H i XOR H j + = − a / 2 , ∘ F H i + = a / 2 , ∘ F H j + = a / 2 .
[0055] Once these execution models are implemented, the matrix H is decomposed into a sequence of operations corresponding to a sequence of quantum gates by means of Gaussian elimination.
[0056] There figure 4 illustrates the pseudo code implemented for Gaussian elimination applied to the matrix H.
[0057] Gaussian elimination allows to obtain a sequence O of operations which decompose the matrix H.Each operation in this sequence O is a set of CNOT gates sharing the same control qubit. These gates are commonly called CNOT fan-out gates. Gaussian elimination is applied to each row 0 to N of the matrix H, respectively at each iteration J=0 to J=N-1. By iterative application of Gaussian elimination to each row of the matrix H, a CNOT fan-out sequence of linear size equal to N is obtained.
[0058] Then, a recomposition step of the S5 entangling layers is performed. This recomposition step of the S5 entangling layers includes the recomposition of the entangling layers so that they comprise only Molmer-Sorensen entangling quantum gates, collective quantum gates with rotation along the X axis, quantum gates with local phase shifts along the Z axis.
[0059] This recomposition is performed directly. More precisely, this means that the quantum circuit comprising only Molmer-Sorensen entangling quantum gates, collective quantum gates with rotation along the X axis, and quantum gates with local phase shifts along the Z axis is obtained directly after the decomposition step S4, without going through another form of quantum gate (in particular CNOT quantum gates).
[0060] When the CNOT fan-out sequence was obtained in the decomposition step S4, a linear number of tables H0, H1, ..., HN, corresponding to the matrix H at each iteration j=0 to j=N of the Gaussian elimination, is obtained.
[0061] A plurality of execution models are determined for the recomposition step. In particular: if a line Hk[ j ] appears in the matrix F, a quantum gate with angle Z rotation F[ Hk[ j ] ]is added on qubit j after the CNOT fan-out number k, If an input F[ b ] does not appear anywhere in the sequence of Hi, with Hi belongs to the tables H0, H1, ..., HN : ∘ and if there exists a table Hk and two qubits c and t such that Hk[c] XOR Hk[t] == b, with b a chosen table, then, the phase F[ b ] is inserted via a heuristic described below, ∘ if no such table exists Hk, the phase is inserted at the end of the sequence O of operations such that: ▪ a set of CNOT quantum gates sharing the same target qubit (commonly called a CNOT fan-in) is inserted having as control qubits all those for which a bit equal to 1 is present in b except for one qubit y, this last qubit y being the target qubit, ▪ a quantum gate with rotation along Z of angle is inserted F[ b ] on qubit y, ▪ insertion of the same CNOT fan-in.
[0062] The remaining phases are inserted according to a heuristic using quantum gates of the fan-out type PHASES. A fan-out PHASE is a set of phase-controlled quantum gates of the form: ∘ otherwise a new fan-out PHASE of a single phase from c to t with as angle F[ b ] is inserted.
[0063] The same pattern is performed with t, if t is already in use.
[0064] There Figure 5 is an example of recomposing a CNOT fan-out quantum gate having two target qubits, into a quantum circuit comprising only Molmer-Sorensen entangling quantum gates, collective quantum gates rotating along the X axis and local phase shifts along the Z axis.
[0065] There figure 6illustrates a recomposition of fan-out PHASES into a quantum circuit comprising only Molmer-Sorensen entangling quantum gates, collective quantum gates rotating along the X axis and local phase shifts along the Z axis.
[0066] When recomposing circuits, the number of Molmer-Sorensen entangling gates and collective quantum gates rotating along the X axis is minimized as a priority, as these gates are the most expensive in terms of execution time and quantum circuit accuracy.
[0067] The number of Molmer-Sorensen entangling gates and collective quantum gates rotating along the X axis can be constant.
[0068] Steps S6 to S9 then illustrate the compilation of the local layers. The compilation of the local layers aims to obtain a quantum circuit comprising only quantum gates with rotation along the Z axis and one rotation along the X axis. Advantageously, the quantum gates with rotation along the X axis are collective, that is to say that it is possible to share a quantum gate with rotation along the X axis with several qubits of the quantum circuit, and even more advantageously with all the qubits.
[0069] The compilation of local layers includes a step of decomposing the local layers S6. The step of decomposing the local layers S6 includes a calculation of the local matrices U to be applied to each qubit. These local matrices U are decomposed into quantum gates with rotation along the X axis, along the Z axis and along the X axis. The ZXZ decomposition allows to obtain three angles a, b and c such that: U = Rz(a)Rx(b)Rz(c), where Rz and Rx correspond to single-qubit rotations about the Z and X axes.
[0070] A first sub-decomposition step S7 is then implemented. The first sub-decomposition step S7 includes the sub-decomposition of the X-axis rotation quantum gates into Hadamard P, Z-axis rotation and Hadamard quantum gates. P
[0071] Then, a second step of sub-decomposition S8 is implemented. The second step of sub-decomposition S8 uses the identity Rx(b) = P Rz(b), where P is a Hadamard gate, and P= Rz(π / 2) Rx(π / 2) Rz(π / 2). Each local matrix can thus be formulated as follows: U = Rz a + π / 2 Rx π / 2 Rz b + π Rx π / 2 Rz c + π / 2
[0072] Finally, a pooling step S9 is implemented. The pooling step collectivizes the X-axis rotation quantum gates to all qubits. Indeed, the Rx of the formula do not depend on the local matrix U because they have a fixed angle π / 2. Thus, from each local matrix U1...UN, a quantum circuit comprising Z-axis rotation gates and X-axis collective rotation gates can be obtained, as illustrated in figure 7 .
[0073] The compilation method according to the invention comprises a grouping step S10, during which the compiled local and entangling layers are concatenated to obtain the quantum circuit compilable on the trapped-ion quantum processor.
[0074] This quantum circuit therefore only includes gates that can be executed by the trapped-ion quantum processor. More precisely, this quantum circuit includes collective quantum gates with rotation along the X axis, Molmer-Sorensen entangling quantum gates, and quantum gates with local phase shifts along the Z axis.
[0075] A further step of reducing the size of the quantum circuit can also be performed. In particular, a maximum of two quantum gates with collective rotations along the X axis can follow each other without redundancy. It is thus possible to merge quantum circuits containing more than two rotations.
[0076] This quantum circuit therefore only includes gates that can be executed by the trapped-ion quantum processor. More precisely, this quantum circuit includes collective quantum gates with rotation along the X axis, Molmer-Sorensen entangling quantum gates, and quantum gates with local phase shifts along the Z axis.
[0077] A further step of reducing the size of the quantum circuit can also be performed. In particular, a maximum of two quantum gates with collective rotations along the X axis can follow one another without redundancy. It is thus possible to merge quantum circuits containing more than two collective rotations along the X axis and then re-decompose them using a pre-selected method, thus producing a more concise, compilable quantum circuit.
[0078] By way of example, a comparative table showing the results of the different methods of the prior art compared to the results of the method according to the invention is described below. [Table 1] Circuit Method (1) Method (2) Method (3) Method (4) QFA-5 30 23 52 36 QFA-10 135 53 112 76 QFA-15 315 83 172 116 QFA-20 570 113 232 156 QFA-n 3^(n(n-1) / 2) 6n - 7 12n - 8 8n - 4 QFT-5 10 7 16 16 QFT-10 45 17 36 36 QFT-15 105 27 56 56 QFT-20 190 37 76 76 QFT-n n(n-1) / 2 2n - 3 4n - 4 4n - 4
[0079] The first column "method (1)" presents the results of the first described prior art. The second column "method (2)" presents the results of the first described prior art, in which the compilation is performed manually, by making reductive assumptions such as the semi-locality of the Molmer-Sorensen gates or the decrease in the interaction strength of the qubits between them during the application of a Molmer-Sorensen gate. The third column "method (3)" corresponds to the same implementation, without the reductive assumptions described above. The fourth column "method (4)" corresponds to the method according to the invention.
[0080] By result we mean the number of quantum gates included in the quantum circuit compiled by the associated method.
[0081] The first line corresponds to an implementation of a quantum Fourier transform on a quantum circuit with 5, 10, 15, 20 qubits. The next line corresponds to this same implementation for a quantum circuit with n qubits.
[0082] The third line corresponds to an implementation of a quantum Fourier adder on a quantum circuit with 5, 10, 15, 20 qubits. The next line corresponds to this same implementation for a quantum circuit with n qubits.
[0083] It therefore appears that quantum circuits compiled by the first method include an exponential number of quantum gates, thus increasing the execution time and decreasing the circuit accuracy.
[0084] On the other hand, the results of columns 2 and 3 are not experimentally verifiable, given that the method is manual.
[0085] Thus, the method according to the invention makes it possible to generate more concise quantum circuits than those generated by the prior art, while proposing a completely automated method. List of cited documents Non-patent literature
[0086] For all useful purposes, the following non-patent element(s) is (are) cited: nplcit1: Compiling quantum algorithms for architectures with multi-qubit gates, Martinez et al; nplcit2: Basic circuit compilation techniques for an ion-trap quantum machine, Maslov et Use of global interactions in efficient quantum circuit constructions, Maslov and Nam.
Claims
1. Method for compiling a quantum circuit for a trapped-ion quantum processor, comprising: - an optional prior step of transforming the quantum circuit so that it contains only: * a predetermined first category of two-qubit quantum gates, * and / or one-qubit quantum gates, - a step of separating said quantum circuit into: * local layers comprising solely one-qubit quantum gates, * entangling layers comprising solely: ∘ two-qubit quantum gates of the first category, ∘ and / or one-qubit quantum gates of a predetermined second category, - a step of compiling local layers, comprising a calculation of local matrices of local layers to be applied to each qubit, - a step of compiling entangling layers, separate from the step of compiling the local layers, - the step of compiling the entangling layers transforming the quantum gates of these entangling layers so that they contain only: * N-qubit collective or entangling quantum gates of a predetermined third category, * one-qubit quantum gates of a predetermined fourth category, - a step of grouping together the compiled local layers and the compiled entangling layers in a compiled quantum circuit, and characterised in that: - the quantum circuit comprises a number of qubits greater than 10 qubits, preferably greater than 20 qubits and preferably greater than 30 qubits, and - the step of compiling the local layers also comprises: - a decomposition of each local matrix of a local layer into a sequence of quantum gates with a rotation on the Z axis, a rotation on the X axis and a rotation on the Z axis, - a sub-decomposition of the rotation on the X axis into a sequence of Hadamard quantum gates, with a Hadamard rotation on the Z axis, - a sub-sub-decomposition of the Hadamard gate into a sequence of quantum gates with a rotation on the Z axis, a rotation on the X axis and a rotation on the Z axis, all three rotations by an angle of π or π / 2, - a mutualisation between qubits of the gates with rotation on the X axis by an angle of π / 2.
2. Method for compiling a quantum circuit for a trapped-ion quantum processor according to claim 1, wherein the step of compiling the entangling layers also transforms the quantum gates of these entangling layers so that: - all or at least some of these collective or entangling quantum gates are applied simultaneously to at least three qubits, advantageously are applied simultaneously to the majority of the qubits, and even more advantageously are applied simultaneously to all the qubits.
3. Method for compiling a quantum circuit for a trapped-ion quantum processor according to one of claims 1 or 2, wherein the step of compiling the entangling layers comprises: - a decomposition of each entangling layer into a phase polynomial, - a direct recomposition in the form of quantum gates of the third category and of the fourth category of quantum gates without passing through a form of quantum gates of the first category of quantum gates.
4. Method for compiling a quantum circuit for a trapped-ion quantum processor according to one of claims 1 to 3, characterised in that it comprises: - a prior step of transforming the quantum circuit so that it contains only: * a predetermined first category of two-qubit quantum gates, * and / or one-qubit quantum gates.
5. Method for compiling a quantum circuit for a trapped-ion quantum processor according to one of claims 1 to 4, characterised in that, during the step of compiling the entangling layers: - the number of collective or entangling quantum gates of the third category of quantum gates in the compiled entangling layers is minimised as a priority with respect to the minimisation of the quantum gates of the fourth category.
6. Method for compiling a quantum circuit for a trapped-ion quantum process according to one of claims 1 to 5, wherein: - the predetermined first category of two-qubit quantum gates comprises: * two-qubit diagonal quantum gates, * CNOT quantum gates, - the predetermined second category of one-qubit quantum gates comprises: * one-qubit diagonal gates, - the predetermined third category of N-qubit collective or entangling quantum gates comprises: * N-qubit Molmer-Sorensen entangling gates, * collective gates with rotation on the X axis, - the predetermined fourth category of one-qubit quantum gates comprises: * local phase shifts on the Z axis.
