Decomposition of Quantum Gates

Transcription

1 Decomposition of Quantum Gates With Applications to Quantum Computing Dean Katsaros, Eric Berry, Diane C. Pelejo, Chi-Kwong Li College of William and Mary January 12, 215

2 Motivation Current Conclusions and Schemes Another Important Scheme Future Directions

3 Motivation Qubit Classical computers store information in bits, vs qubits in a Quantum computer

4 Motivation Qubit Classical computers store information in bits, vs qubits in a Quantum computer Quantum Gates Quantum gates are similar to logic gates in classical computing, in that they are used to manipulate a quantum system

5 Motivation Qubits and Quantum Gates have Mathematical Realizations

6 Motivation Qubits and Quantum Gates have Mathematical Realizations Qubits are vectors

7 Motivation Qubits and Quantum Gates have Mathematical Realizations Qubits are vectors Quantum Gates are Unitary Matrices

8 Motivation Qubits are Quantum Systems

9 Motivation Qubits are Quantum Systems Letting apple i, apple 1i be apple two measureables, the vector (qubit) a 1 b = a + b represents the superposition a i + b 1i 1

10 Motivation Qubits are Quantum Systems Letting apple i, apple 1i be apple two measureables, the vector (qubit) a 1 b = a + b represents the superposition a i + b 1i 1 We concatenate 2+ qubits into multi-qubit quantum ensembles via tensor products:

11 Motivation Qubits are Quantum Systems Letting apple i, apple 1i be apple two measureables, the vector (qubit) a 1 b = a + b represents the superposition a i + b 1i 1 We concatenate 2+ qubits into multi-qubit quantum ensembles via tensor products: apple a b apple c d = 2 3 ac 6ad 4bc5 bd

12 Motivation Quantum Computing Letting apple i, apple 1i be apple two measureables, a qubit a 1 b = a + b represents the superposition a i + b 1i 1 We concatenate 2+ qubits into multi-qubit quantum ensembles via tensor products: apple a b apple c d = 2 3 ac 6ad 4bc5 bd This 2-qubit system has 4 measureables, represented by the basis vectors of C 4.

13 Motivation Example. Consider further apple a b apple c d = 2 3 ac 6ad 4bc5 bd

14 Motivation Example. Consider further apple a b apple c d = 2 3 ac 6ad 4bc5 bd I The basis vectors, corresponding to physical measureables, of the above bipartite or joint quantum state are e 1 = 6 4 5, e 2 = 6 4 5, e 3 = , and e 4 =

15 Example. I We use the physicists notation; i = , 1i = , 1i = , 11i =

16 motivation Example. I We use the physicists notation; i = , 1i = , 1i = , 11i = So, ac ad bc bd 3 5 = ac i + ad 1i + bc 1i + bd 11i

17 Motivation Question: How Many Measureables does a 64-qubit multipartite system have?

18 Motivation Some other operations Let A, B 2 M 2. The tensor product of A and B is A B = apple a11 B a 12 B a 21 B a 22 B. The direct sum of A and B is defined as A B = apple A B, where 2 M 2.

19 Motivation Quantum Gates reign things in An n-qubit system has 2 n measureable states, and a classical computer has to deal with each of these...

20 Motivation Quantum gates reign things in An n-qubit system has 2 n measureable states, and a classical computer has to deal with each of these... A Quantum computer uses Quantum, orunitary, Gates (Unitary matrices) to handle these n-qubit systems in a single operation.

21 Motivation Definition AmatrixU 2 M n (C) isunitaryifu U = U U = I where denotes the conjugate transpose. Important Properties U is invertible and U 1 = U The rows and columns of U are orthonormal

22 Motivation-Example Quantum Gates in 1 qubit Hadamard Gate The Hadamard gate, H, is a commonly used gate where H = 1 p 2 apple Pauli Matrices x = apple 1 1 y = z = apple i i apple 1 1

23 Motivation The set of Quantum Gates a quantum computer can generate directly determines its capability.

24 Motivation The set of Quantum Gates a quantum computer can generate directly determines its capability. Obviously, we do not want to limit our systems possible operations...

25 Motivation The set of Unitary Gates a quantum computer can generate directly determines its capability. Obivously, we do not want to limit our systems possible operations... We can do even better: How can we not only allow for all operations, but have an e cient generating set of simple unitaries?

26 A Decomposition Scheme-2 Qubit Case Experimentalists are working on possible physical manifestations in the 1-4 qubit cases. 2 Qubits Corresponds to 4-by-4 Unitaries There are two types of gates that are easy to implement 1-control gates Free-gates

27 A Decomposition Scheme-2 Qubit Case Experimentalists are working on possible physical manifestations in the 1-4 qubit cases. 2 Qubits Corresponds to 4-by-4 Unitaries There are two types of gates that are easy to implement 1-control gates Free-gates Experimentalists find these to be simple to implement.

28 Decomposition of Quantum Gates-2 Qubit Case 1-Control Gates (1V )=I 2 V (V )=V I v 11 v 12 (V ) = 6 1 4v 21 v (V 1) = 6 v 11 v v 21 v 22

29 Decomposition of Quantum Gates-2 Qubit Case Free-Gates (V ) =V I 2 = ( V )=I 2 V = 2 3 v 11 v 12 6 v 11 v 12 4v 21 v 22 5 v 21 v v 11 v 12 6v 21 v 22 4 v 11 v 12 5 v 21 v 22

30 A Decomposition Scheme-2 Qubit Case Whats the di erence? Consider a 2-qubit Vector State q = a i + a 1 1i + a 2 1i + a 3 11i.

31 A Decomposition Scheme-2 Qubit Case Whats the di erence? Consider a 2-qubit Vector State q = a i + a 1 1i + a 2 1i + a 3 11i. Operating on this system with a free-gate, ( V ), yields (I V )(q) = i V (a i + a 1 1i)+ 1i V (a 2 i + a 3 1i).

32 A Decomposition Scheme-2 Qubit Case Whats the di erence? Consider a 2-qubit Vector State q = a i + a 1 1i + a 2 1i + a 3 11i. Operating on this system with a free-gate, ( V ), yields (I V )(q) = i V (a i + a 1 1i)+ 1i V (a 2 i + a 3 1i). Operating on this system with a 1-control gate, (1V), yields (I V )(q) =a i + a 1 1i + 1iV (a 2 i + a 3 1i).

33 A Decomposition Scheme-2 Qubit Case Whats the di erence? Consider a 2-qubit Vector State q = a i + a 1 1i + a 2 1i + a 3 11i. Operating on this system with a free-gate, (*V), yields (I V )(q) = i V (a i + a 1 1i)+ 1i V (a 2 i + a 3 1i). Operating on this system with a 1-control gate, (1V), yields (I V )(q) =a i + a 1 1i + 1iV (a 2 i + a 3 1i). 1-controls, or controlled gates, in general, are named so because they act solely on some of the components of a multi-partite state, and leave the rest alone (computationally expensive!)

34 A Decomposition Scheme-2 Qubit case Previous Result(Li, Roberts, Yin) Using control gates, one can decompose an arbitrary n-by-n n unitary matrix into a product of at most 2 unitary matrices I P-unitary matrices are (1V ), (V ), (V 1), and v 11 v 12 4 v 21 v

35 A Decomposition Scheme-2 Qubit case Previous Result(Li, Roberts, Yin) Using control gates, one can decompose an arbitrary n-by-n n unitary matrix into a product of at most 2 unitary matrices I P-unitary matrices are (1V ), (V ), (V 1), and i.e. n k = n! k!(n k)! v 11 v 12 4 v 21 v

36 A Decomposition Scheme-2 Qubit case Previous Result(Li, Roberts, Yin) One can decompose an arbitrary n-by-n unitary matrix into a n product of at most 2 P-unitary matrices I P-unitary matrices are (1V ), (V ), (V 1), and i.e v 11 v 12 4 v 21 v n k = n! k!(n k)!. For 4-by-4, at most 6 unitary matrices. For 8-by-8, at most 14 unitary matrices. etc.

37 A Decomposition Scheme-2 Qubit case The above decomposition scheme heavily utilized control gates. The next step was to introduce free-gates into the decomposition, and achieve a lowest possible cost.

38 A Decomposition Scheme-2 Qubit case The above decomposition scheme heavily utilized control gates. The next step was to introduce free-gates into the decomposition, and achieve a lowest possible cost. More Important Results (Li, Pelejo) In the 4-by-4 case, 3 1-control gates is enough for any unitary

39 A Decomposition Scheme-2 Qubit case The above decomposition scheme heavily utilized control gates. The next step was to introduce free-gates into the decomposition, and achieve a lowest possible cost. More Important Results (Li, Pelejo) In the 4-by-4 case, 3 1-control gates is enough for any unitary We can always freely transform a 4-by-4 1-control gate into a (1V) gate

40 A Decomposition Scheme-2 Qubit case The above decomposition scheme heavily utilized control gates. The next step was to introduce free-gates into the decomposition, and achieve a lowest possible cost. More Important Results (Li, Pelejo) In the 4-by-4 case, 3 1-control gates is enough for any unitary We can always freely transform a 4-by-4 1-control gate into a (1V) gate A decomposition scheme was developed and extended to all n, as well as a recursive formula giving the number of free and k-control gates that could be used to decompose an arbitrary unitary.

41 A Decomposition Scheme-2 Qubit case The above decomposition scheme heavily utilized control gates. The next step was to introduce free-gates into the decomposition, and achieve a lowest possible cost. More Important Results (Li, Pelejo) In the 4-by-4 case, 3 1-control gates is enough for any unitary We can always freely transform a 4-by-4 1-control gate into a (1V) gate A decomposition scheme was developed and extended to all n, as well as a recursive formula giving the number of free and k-control gates that could be used to decompose an arbitrary unitary. We want(ed) to further reduce the number of controls!

42 Current Scheme Questions: How many gates are necessary, and, specifically, how many 1-control gates are necessary and su cient? What is the most e unitaries? cient scheme for decomposing general 1-control gates are a metaphoric cost in a decomposition!

43 Current Scheme How should one attack the problem? What can we do for free that simplifies the problem, or gives telling information about our candidate?

44 Current Scheme How should one attack the problem? What can we do for free that simplifies the problem, or gives telling information about our candidate? (?)

45 Current Scheme How should one attack the problem? What can we do for free that simplifies the problem, or gives telling information about our candidate? (?) Switch focus from finding ways to decompose a matrix, to finding out what must be true if the matrix can be written as a product of free gates, free gates and a single 1-control gate, etc.

46 Current Scheme Example in 4-by-4 case If a matrix M can be decomposed using only free gates, it can be written as M = A B, Where A and B are 2-by-2 unitary matrices.

47 Current Scheme Example in 4-by-4 case If a matrix M can be decomposed using only free gates, it can be written as M = A B, Where A and B are 2-by-2 unitary matrices. (?) This requires that each block be a scalar multiple of some unitary!

48 Current Scheme Example in 4-by-4 case If a matrix M can be decomposed using only free gates, it can be written as M = A B, Where A and B are 2-by-2 unitary matrices. (?) This requires that each block be a scalar multiple of some unitary! If M can be decomposed using free gates, and a single 1-control, then it can be written as M =(A B)(I 2 W )(E F ), Where A, B, W, E, F all unitary.

49 Current Scheme Recall the Singular Value Decomposition For any matrix A 2 M n, there is a unitary equivalence of A yielding a diagonal matrix, with entries the singular values of A

50 Our Scheme Recall the Singular Value Decomposition For any matrix A 2 M n, there is a unitary equivalence of A yielding a diagonal matrix, with entries the singular values of A Example. M = apple i i.

51 Current Scheme Recall the Singular Value Decomposition For any matrix A 2 M n, there is a unitary equivalence of A yielding a diagonal matrix, with entries the singular values of A Example. M = apple i i. Its singular value decomposition yields the factorization, M = U V = apple i i apple 1 1 apple 1 1.

52 Sidenote Not Gate The unitary matrix apple 1 1 is known as the Not Gate. I It is important-a class of controlled gates utilizes its properties Ex., the CNOT Gate is

53 Current Scheme Number of necessary 1-control gates apple M11 M We let M = 12 be a general 4 4 unitary matrix. By the M 21 M 22 SVD, there exist unitary U and V such that V M 11 U = C = diag(c 1, c 2 ). So where S = diag(s 1, s 2 ). (I 2 V ) M (I 2 U) = apple C SU VS VCU, Our Scheme revolves around the values of c 1 and c 2

54 Current Scheme Free Decomposition (Theorem) Given a 4 by 4 unitary matrix M = apple M11 M 12 M 21 M 22 =(I 2 V ) apple C SU VS VCU (I 2 U), Letting C = diag(c 1, c 2 ). Then, M is a product of free gates if and only if c 1 = c 2 and s 1 UV + c 1 UV and s 1 c 1 V are scalar matrices. I i.e., for a given unitary, check three things, and you ll know whether controlled gates are needed for decomposition!

55 Current Scheme One 1-Control and Free Gates (Theorem) Again, take a unitary and write it as M = apple apple M11 M 12 =(I M 21 M 2 V ) C SU 22 VS VCU (I 2 U). Then, M is a product of free gates and one 1-control gate if and only if either, (i) c 1 = c 2 and C, S, U, and V are simultaneously unitarily diagonalizeable. (ii) c 1 6= c 2 2 (, 1) and V, U are both scalar matrices. (iii) C and S are rank 2

56 Current Scheme 2 1-Control and Free Gates(?) We know that a unitary can be written as a product of free gates and two 1-control gates when c 1 = c 2 2 (, 1) and U, V are not simultaneously diagonalizeable. This is incomplete, c 1 6= c 2 and?

57 Another Scheme The Result of Kraus and Cirac-see [1] The authors proved that every U 2 SU(4) can be written as U =(A 1 A 2 )(exp(i(d x x x +d y y y +d z z z ))(B 1 B 2 ) with A 1, A 2, B 1, B 2 2 SU(2), d x, d y, d z 2 R.

58 Another Scheme We also know that any U 2 SU(4) is decomposable using at most three 1-control gates-[6]. We wish to know whether the two di erent schemes can be used in combination. i.e. SVD is not computationally expensive-when is it better? Can this be used to find conditions where two 1-controls are su cient? Insight into the general case

59 Future Directions Comparison of the two Schemes. Utility of Di erent Schemes Relative to Di erent Physical Manifestations. Find a quantitative operation on a matrix which determines which scheme is most e cient. Higher qubits.

60 Acknowledgements Eric Berry, Diane C. Pelejo, Dr. Chi-Kwong Li The NSF for their support through the EXTREEMS-QED grant William and Mary Mathematics

61 References B. Kraus and J.I. Cirac, Optimal Creation of Entanglement Using a Two Qubit Gate, Kazuyuki FUJII, Hiroshi OIKE and Tatsuo SUZUKI, More on the Isomorphism SU(2) SU(2) and SO(4), Stephen Bullock and Igor Markov. An Arbitrary Two-qubit Computation In 23 Elementary Gates. arxiv:quant-ph/2112 Chi-Kwong Li, Rebecca Roberts, Xiaoyan Yin Decomposition of unitary matrices and quantum gates arxiv: Chi-Kwong Li, Diane Pelejo Decomposition of quantum gates arxiv: M. Nakahara and T. Ohmi, Quantum Computing: From Linear Algebra to Physical Realizations