Local certification of programmable quantum devices of arbitrary high dimensionality

11/21/2019
by   Kishor Bharti, et al.
0

The onset of the era of fully-programmable error-corrected quantum computers will be marked by major breakthroughs in all areas of science and engineering. These devices promise to have significant technological and societal impact, notable examples being the analysis of big data through better machine learning algorithms and the design of new materials. Nevertheless, the capacity of quantum computers to faithfully implement quantum algorithms relies crucially on their ability to prepare specific high-dimensional and high-purity quantum states, together with suitable quantum measurements. Thus, the unambiguous certification of these requirements without assumptions on the inner workings of the quantum computer is critical to the development of trusted quantum processors. One of the most important approaches for benchmarking quantum devices is through the mechanism of self-testing that requires a pair of entangled non-communicating quantum devices. Nevertheless, although computation typically happens in a localized fashion, no local self-testing scheme is known to benchmark high dimensional states and measurements. Here, we show that the quantum self-testing paradigm can be employed to an individual quantum computer that is modelled as a programmable black box by introducing a noise-tolerant certification scheme. We substantiate the applicability of our scheme by providing a family of outcome statistics whose observation certifies that the computer is producing specific high-dimensional quantum states and implementing specific measurements.

READ FULL TEXT VIEW PDF
POST COMMENT

Comments

There are no comments yet.

Authors

page 2

08/02/2021

Large-scale quantum machine learning

Quantum computers promise to enhance machine learning for practical appl...
12/16/2020

Programmable Quantum Annealers as Noisy Gibbs Samplers

Drawing independent samples from high-dimensional probability distributi...
01/26/2021

Quantum machine learning models are kernel methods

With near-term quantum devices available and the race for fault-tolerant...
08/04/2020

A divide-and-conquer algorithm for quantum state preparation

Advantages in several fields of research and industry are expected with ...
02/25/2022

Short Paper: Device- and Locality-Specific Fingerprinting of Shared NISQ Quantum Computers

Fingerprinting of quantum computer devices is a new threat that poses a ...
01/21/2021

Noisy intermediate-scale quantum (NISQ) algorithms

A universal fault-tolerant quantum computer that can solve efficiently p...
11/28/2019

Poq: Projection-based Runtime Assertions for Debugging on a Quantum Computer

In this paper, we propose Poq, a runtime assertion scheme for debugging ...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

Methods

The main tool we use in this work to show that anti-hole inequalities admit robust self-testing is Theorem 1, shown in BRVWCK19 , which provides a sufficient condition for a graph to be self-testable. This result relies crucially on the rich properties of a powerful class of mathematical optimisation models, known as Semidefinite programs (SDPs) (see Appendix A

). SDPs constitute a vast generalisation of linear optimisation models where scalar variables are replaced by vectors and the constraints and objective function are affine in terms of the inner products of the vectors. Equivalently, collecting all pairwise inner products of these vectors in matrix, known as the Gram matrix, an SDP corresponds to optimising a linear function of the Gram matrix subject to affine constraints. Analogously to linear programs, to any SDP there is an associated a dual program whose value is equal to the primal under reasonable assumptions. Next, we single out certain properties of primal-dual solutions that are of relevance to his work. A pair of primal dual optimal solutions (

) with no duality gap (i.e. ), satisfies strict complementarity if the and give a direct sum decomposition of the underlying space. Furthermore, an optimal dual solution with rank is dual nondegenerate if the tangent space at of the manifold of symmetric matrices with rank equal to together with the linear space of matrices defining the SDP span the entire space of symmetric matrices. In this work we focus on the Lovász theta SDP, see () in Appendix B. The proof of our main result involves two main steps. First, we construct a dual optimal solution of () for an odd-cycle graph by providing an explicit mapping between the Gram vectors of a primal () optimal solution of an odd-cycle graph and the Gram vectors of a dual optimal solution of the complement graph; see Theorem 2 for details. Next, once we construct a dual optimal solution, we show in Theorem 3 that it satisfies the non-degeneracy conditions given in (7). By Theorem 1 , this shows that anti-hole inequalities admit robust self-testing. Details of the proofs can be found in Appendix C. Additionally, we show that not all graphs admit self-testing by providing a counter example of such a graph (see Appendix F). The overall scheme for determining whether a graph is self-testable (equivalently whether the primal optimal solution of the Lovász theta SDP corresponding to that graph is unique or not) is provided in the form of a flowchart in Figure 3.

Figure 3: Flowchart for determining (non)uniqueness of primal solution(s). The refers to the fact that one can still hope to arrive at a definitive answer by restarting the algorithm using a different dual optimal solution (if it exists).

Acknowledgements.— We thank the National Research Foundation of Singapore, the Ministry of Education of Singapore, MINECO Project No. FIS2017-89609-P with FEDER funds, and the Knut and Alice Wallenberg Foundation for financial support. A significant part of the research project Local certification of programmable quantum devices of arbitrary high dimensionalitywas carried out during “New directions in quantum information” conference organized by Nordita, the Nordic Institute for Theoretical Physics.

References

  • (1) Mayers, D. & Yao, A. Self testing quantum apparatus. Quantum Info. Comput. 4, 273–286 (2004). URL http://dl.acm.org/citation.cfm?id=2011827.2011830.
  • (2) Mayers, D. & Yao, A. Quantum cryptography with imperfect apparatus. In Proceedings 39th Annual Symposium on Foundations of Computer Science (Cat. No. 98CB36280), 503–509 (IEEE, 1998).
  • (3) Šupić, I. & Bowles, J. Self-testing of quantum systems: a review. arXiv preprint arXiv:1904.10042 (2019).
  • (4) Reichardt, B. W., Unger, F. & Vazirani, U. Classical command of quantum systems. Nature 496, 456 (2013).
  • (5) Popescu, S. & Rohrlich, D. Generic quantum nonlocality. Phys. Lett. A 166, 293–297 (1992). URL http://www.sciencedirect.com/science/article/pii/037596019290711T.
  • (6) McKague, M., Yang, T. H. & Scarani, V. Robust self-testing of the singlet. J. Phys. A 45, 455304 (2012).
  • (7) Bugliesi, M., Preneel, B., Sassone, V. & Wegener, I. Automata, Languages and Programming: 33rd International Colloquium, ICALP 2006, Venice, Italy, July 10-14, 2006, Proceedings, vol. 4051 (Springer, 2006).
  • (8) Tsirel’son, B. S. Quantum analogues of the bell inequalities. the case of two spatially separated domains. J. Soviet Math. 36, 557–570 (1987).
  • (9) Summers, S. J. & Werner, R. Bell’s inequalities and quantum field theory. i. general setting. J. Math. Phys. 28, 2440–2447 (1987).
  • (10) Bell, J. S. On the einstein podolsky rosen paradox. Physics (Long Island City, N.Y.) 1, 195–200 (1964). URL https://link.aps.org/doi/10.1103/PhysicsPhysiqueFizika.1.195.
  • (11) Bharti, K. et al. Robust self-testing of quantum systems via noncontextuality inequalities. Phys. Rev. Lett. 122, 250403 (2019). URL https://link.aps.org/doi/10.1103/PhysRevLett.122.250403.
  • (12) Kochen, S. & Specker, E. P. The problem of hidden variales in quantum mechanics. J. Math. Mech. 17, 59–87 (1967).
  • (13) Cabello, A., Severini, S. & Winter, A. Graph-theoretic approach to quantum correlations. Phys. Rev. Lett. 112, 040401 (2014). URL https://link.aps.org/doi/10.1103/PhysRevLett.112.040401.
  • (14) Clauser, J. F., Horne, M. A., Shimony, A. & Holt, R. A. Proposed experiment to test local hidden-variable theories. Phys. Rev. Lett. 23, 880–884 (1969). URL https://link.aps.org/doi/10.1103/PhysRevLett.23.880.
  • (15) Klyachko, A. A., Can, M. A., Binicioğlu, S. & Shumovsky, A. S. Simple test for hidden variables in spin-1 systems. Phys. Rev. Lett. 101, 020403 (2008). URL https://link.aps.org/doi/10.1103/PhysRevLett.101.020403.
  • (16) Cabello, A., Danielsen, L. E., López-Tarrida, A. J. & Portillo, J. R. Basic exclusivity graphs in quantum correlations. Physical Review A 88, 032104 (2013).
  • (17) Chudnovsky, M., Robertson, N., Seymour, P. & Thomas, R. The strong perfect graph theorem. Annals of mathematics 51–229 (2006).
  • (18) Bharti, K., Arora, A. S., Kwek, L. C. & Roland, J. A simple proof of uniqueness of the kcbs inequality. arXiv preprint arXiv:1811.05294 (2018).
  • (19) Gottesman, D. as it has been referenced in the following blog [http://www.scottaaronson.com/blog/?p=284; accessed 13-apr-2017 (2004).
  • (20) Broadbent, A., Fitzsimons, J. & Kashefi, E. Universal blind quantum computation. In 2009 50th Annual IEEE Symposium on Foundations of Computer Science, 517–526 (IEEE, 2009).
  • (21) Fitzsimons, J. F. & Kashefi, E. Unconditionally verifiable blind quantum computation. Physical Review A 96, 012303 (2017).
  • (22) Aharonov, D., Ben-Or, M., Eban, E. & Mahadev, U. Interactive proofs for quantum computations. arXiv preprint arXiv:1704.04487 (2017).
  • (23) Aharonov, D., Ben-Or, M. & Eban, E. Interactive proofs for quantum computations. Tech. Rep. (2008).
  • (24) Mahadev, U. Classical verification of quantum computations. In 2018 IEEE 59th Annual Symposium on Foundations of Computer Science (FOCS), 259–267 (IEEE, 2018).
  • (25) Kleinmann, M., Guehne, O., Portillo, J. R., Larsson, J.-Å. & Cabello, A. Memory cost of quantum contextuality. New Journal of Physics 13, 113011 (2011).
  • (26) Fagundes, G. & Kleinmann, M. Memory cost for simulating all quantum correlations from the peres–mermin scenario. Journal of Physics A: Mathematical and Theoretical 50, 325302 (2017).
  • (27) Cabello, A., Gu, M., Gühne, O. & Xu, Z.-P. Optimal classical simulation of state-independent quantum contextuality. Physical review letters 120, 130401 (2018).
  • (28) Budroni, C. Contextuality, memory cost and non-classicality for sequential measurements. Philosophical Transactions of the Royal Society A 377, 20190141 (2019).
  • (29) Cabello, A., Gu, M., Gühne, O., Larsson, J.-Å. & Wiesner, K. Thermodynamical cost of some interpretations of quantum theory. Physical Review A 94, 052127 (2016).
  • (30) Malinowski, M. et al. Probing the limits of correlations in an indivisible quantum system. Physical Review A 98, 050102 (2018).
  • (31) Knuth, D. E. The sandwich theorem. The Electronic Journal of Combinatorics 1, 1 (1994).
  • (32) Lovász, L. On the shannon capacity of a graph. IEEE Trans. Inf. Theory 25, 1–7 (1979).
  • (33) Alizadeh, F., Haeberly, J.-P. A. & Overton, M. L. Complementarity and nondegeneracy in semidefinite programming. Math. Program. 77, 111–128 (1997).
  • (34) Thinh, L. P., Varvitsiotis, A. & Cai, Y. Structure of the set of quantum correlators via semidefinite programming. arXiv preprint arXiv:1809.10886 (2018).

Appendix A Semidefinite programming basics

A semidefinite program (SDP) is given by an optimisation problem of the following form

(P)

where denotes the cone of real positive semidefinite matrices and . The corresponding dual problem is given by

(D)

A pair of primal-dual optimal solutions () with no duality gap (i.e. ), satisfies strict complementarity if

(4)

Lastly, an optimal dual solution is called dual nondegenerate if the linear system in the symmetric matrix variable

only admits the trivial solution .

Central to this work is the Lovász theta SDP corresponding to a graph , whose primal formulation is:

()

and the dual formulation we use is given by:

()

Appendix B Robust self-testing

To prove our main result we use the following definitions from BRVWCK19 . A non-contextuality inequality is a self-test for the realisation  if:

  1. achieves the quantum supremum ;

  2. For any other realisation that also achieves , there exists an isometry such that

    (5)

Furthermore, a non-contextuality inequality is an -robust self-test for if it is a self-test, and furthermore, for any other realisation satisfying

there exists an isometry such that

(6)

The proof of our main result hinges on the following theorem (first introduced in BRVWCK19 ):

Theorem 1.

Consider a non-contextuality inequality . Assume that

  1. There exists an optimal quantum realisation such that

    and for all , and

  2. There exists a dual optimal solution for the SDP () such that the homogeneous linear system

    (7)

    in the symmetric matrix variable only admits the trivial solution .

Then, the non-contextuality inequality is an -robust self-test for .

Appendix C Self-testing anti-hole inequalities

The anti-hole non-contextuality inequalities are given by for all . The quantum bound for the anti-hole inequalities, i.e., the Lovász theta number for is knuth1994sandwich . A canonical quantum ensemble which achieves the quantum value for the anti-hole inequalities corresponding to odd is given by dimensional quantum state and projectors. Explicitly, the quantum state is

(8)

but the description of the projectors is more involved CDLP13 . Let us denote the -th component of corresponding to projector as . For and ,

(9)
(10)
(11)

for and

(12)
(13)

For the anti-hole non-contextuality inequalities, the ensemble described above achieves the quantum value and satisfies the first condition of Theorem 1. It remains to establish the existence of a dual optimal solution for the SDP corresponding to the Lovász theta number of anti-hole graphs such that the conditions in (7) are satisfied. Towards this goal, we first proceed to provide the explicit form of the dual optimal solution.

Theorem 2.

Let be the unique optimal solution for . Then,

is a dual optimal solution for Another useful expression for is given by:

(14)

where is the vector of all ones of length ,

and maps an -dimensional vector and outputs the corresponding circulant matrix.

Proof.

It was shown in BRVWCK19 that admits a unique optimal solution . For any , the map taking (modulo ) is an automorphism of (i.e., a bijective map that preserves adjacency and non-adjacency). In particular, this implies that is circulant and furthermore, constant along each band. Specifically, all diagonal entries of are equal, and as , it follows that

(15)

Analogously, for a pair of indices with we have that . Moreover, be feasibility of we have that . Thus has the correct value and it remains to show that is feasible. Next, by feasibility of we have that , when in . Thus, by definition of we have that for all edges of . Finally we show that . Indeed,

where we used (15) and that (see Theorem 8 of lovasz1979shannon ). To finish the proof we note that

(16)

where the second last equality follows from the constraint that for and the last equality follows by substituting . ∎

Finally, we show that the dual optimal solution satisfies the conditions in 7.

Theorem 3.

The dual optimal solution , corresponding to the complement of an odd-cycle graph satisfies the conditions in 7.

Proof.

We show that for any odd the only symmetric matrix satisfying

(17)

is the matrix , where here refers to an edge in the graph. Barring the constraint, the rest already guarantee that there are at most potentially non-zero entries in the matrix (not counting the repeated entries) corresponding to graph. Let the first row of be . We fill the rest of the potential non-zero slots in with . For example for , we have

(18)

For notational convenience, let , where , and . In the rest of this section we use the notation to denote where is an integer. The linear equation corresponding to implies that

(19)

For , the linear equations corresponding to and imply

(20)

Now consider the linear equations corresponding to . These equations along with 20 imply

(21)

Using 21 along with 19 we have

(22)

Finally, using the equations corresponding to again and 22 we have

(23)

implying that , as desired. ∎

Appendix D Dimension for optimal violation of anti-hole inequalities

Theorem 4.

Given an anti-hole non-contextuality inequality with an odd number of measurement events, the quantum system achieving the optimal quantum bound must be at least dimensional.

Proof.

The value of a non-contextuality inequality achievable within quantum theory is equal to the Lovász theta number of the underlying graph and admits the SDP formulation (). The lower bound on the dimension of a quantum system achieving the optimal quantum bound is the rank of the unique primal optimal matrix

where is the all-ones vector of length , is the circulant function that takes as input a dimension vector and outputs a matrix with the input vector as its top row and every subsequent row being one place right shifted modulo and . Since

is real, its rank over complex field is the same as over real field, and equals to the number of nonzero eigenvalues (with multiplicity). Furthermore, a lower bound on the rank of

is given by the rank of the lower right block matrix (the circulant portion). The eigenvalues of a circulant matrix can be calculated easily using the circulant vector. A few lines of algebra yields the following expression for the eigenvalues of the lower right block matrix,

for and denotes the Lovász theta number for the holes with odd One can see that unless or Thus, the rank of the circulant matrix is for all odd values of Thus, the lower bound on the rank of the optimal feasible matrix is which is same as the lower bound on the dimension of the desired quantum system. ∎

Appendix E Complex versus Real SDPs

Lemma 5.

Consider a real SDP

that admits a unique optimal solution witnessed by a dual nondegenerate optimal solution . Then, the SDP considered over the complex numbers, i.e.,

()

still admits a unique optimal solution, where and denotes the set of Hermitian positive semidefinite matrices.

Proof.

First, we show that the study of a complex SDP can be reduced to an equivalent real SDP. This fact is well known but we provide a brief argument for completeness. Indeed, for any feasible solution , the constraint is equivalent to two constraints on its real and imaginary part, namely: and . Furthermore, checking whether is Hermitian PSD is equivalent to

Based on these observations we define the realification of as the following SDP over the real numbers:

()
s.t.

Clearly, the solutions of are in bijection with the solutions of the realification, and thus, to show that has a unique solution it suffices to show that has a unique solution. Bringing into standard SDP form we arrive at the formulation:

(24)
s.t.

whose dual is to minimize the function over all satisfying

We conclude the proof by showing that is a dual nondegenerate optimal solution for the realification. First, by dual feasibility we have for appropriate scalars . Setting and all other dual variables to zero, we have established feasibility. Second, to show optimality note that is optimal for the realification, and furthermore, . Lastly, to check nondegeneracy consider a symmetric matrix satisfying

(25)

and

(26)

Now, constraint (25) is equivalent to

Furthermore, using (26), from the third equation we get , from the first one we get and from the second one . Summarizing, for all we have that

As is dual nondegenerate it has the property that for any :

Putting everything together we get .

Appendix F Not all non-contextuality inequalities admit self-testing

We have proved that all fundamental non-contextuality inequalities admit self-testing. A natural question is whether every non-contextuality inequality with separation between corresponding non-contextual hidden variable bound and quantum bound admits self-testing. Below we provide an explicit non-contextuality inequality which shows that the answer is negative. In graph theoretical terms, we identify a non-perfect graph whose Lovász theta SDP admits multiple primal optimal solutions. We make crucial use of the following result (alizadeh, , Theorem 5) to determine the (non)uniqueness of primal optimal under strict complementarity, see also thinh2018structure .

Theorem 6.

Let be a pair of primal and dual optimal solutions satisfying strict complementarity.Then, uniqueness of implies that is dual nondegenerate.

The exclusivity graph of our counter-example is shown in Figure 4. The corresponding canonical non-contextuality inequality is given by

(27)

whose quantum bound is equal to .

Figure 4: The above exclusivity graph corresponds to the canonical non-contextuality inequality with minimal number of measurement events which doesn’t admit self-testing.

Consider the pair of primal-dual optimal solutions

(28)

where and

(29)

where , , and . Since and , strict complementarity holds. Using Theorem 6, the uniqueness of implies dual nondegeneracy. To determine dual nondegeneracy for we (once again) resort to solving system of linear equations. The symmetric variable matrix is given by

(30)

Solving for the linear systems of equations , we get , , , , , . For example, if we set , we can get a consistent assignment of , from to , which isn’t all zero. Hence, the dual solution is degenerate, which together with strict complementarity implies that the primal is not unique. Thus the non-contextuality inequality in (27) does not admit self-testing.

We also report that we found several other non-perfect graphs (and equivalently non-contextuality inequalities) which do not admit self-testing. Identifying the exact classes of graphs which admit self-testing will be interesting but we leave that as an open question.