Imperial/TP/2020/MC/01 Supersymmetry and Quantum Computation P. Marcos Crichigno Blackett Laboratory, Imperial College, Prince Consort Rd., London, SW7 2AZ, U.K. and Institute for Theoretical Physics, University of Amsterdam, Science Park 904, Postbus 94485, 1090 GL, Amsterdam, The Netherlands The interplay between supersymmetry and classical and quantum computation is discussed. First, it is shown that the problem of computing the Witten index of 2 quantum mechanical systems is #P-complete and therefore intractable. Then, the notions ofN supersymmetry ≤ in the space of qubits and supersymmetric quantum circuits are introduced and some of their properties discussed. In particular, it is shown that these define a nontrivial subclass of quantum algorithms with robustness properties typical of supersymmetric systems. Concrete examples, including the supersymmetric SYK model and fermion hard-core models are discussed. Some applications and open questions are suggested. Introduction. Finding the ground states of a physi- of qubits, we then define the notion of supersymmetric cal system is hard. This is true not only in a practical quantum circuits, and use these to design quantum al- sense but also in a formal, computational, sense. A pro- gorithms associated to certain supersymmetric physical totypical example is that of a classical spin-glass system observables. for which the problem of finding the ground state is NP- The intuition from the physics of supersymmetric sys- hard [1]. Similarly, finding the ground state of a quantum tems is that this should define a nontrivial subclass Hamiltonian is QMA-hard [2, 3]. This implies, in particu- of quantum algorithms, with advantageous properties lar, that there is no efficient algorithm for finding ground (e.g., invariance under certain deformations) over non- states, assuming standard conjectures in computational supersymmetric ones, and which capture highly non- complexity theory. A number of deep connections be- trivial problems of both physical as well as mathematical tween the theory of computational complexity and sta- interest. As we discuss, this intuition bears out. tistical mechanics systems have been pointed out in, e.g., We emphasize that this definition of supersymmetry [4{8]. holds for any system of qubits and does not require Supersymmetry is a symmetry relating bosonic and supersymmetry to be realized at a fundamental level in fermionic states of a system. Supersymmetric systems nature. In particular, any implementation of a quan- are often more amenable to analysis and various ana- tum computer can be made supersymmetric in this sense. lytic and exact results are possible. Indeed, the ground states{or some of their properties{can be found analyti- = 2 quantum mechanics. The Hilbert space of N cally in various nontrivial supersymmetric systems. This any quantum mechanical theory can be decomposed as raises the question of the computational complexity asso- = B F where each factor refers to the subspace of H H ⊕H ciated to the ground states of supersymmetric systems. bosonic and fermionic states. These are distinguished by Although many exact results are known in specific su- the operator ( 1)F , acting as +1 on bosonic states and − persymmetric systems, we show that the ground state as 1 on fermionic states. By definition, in a theory with − problem for supersymmetric theories remains computa- = 2 supersymmetry there exists a complex Grassmann N tionally hard. operator , sending states in B into states in F and Q H H Supersymmetry was first proposed as a possible sym- vice versa, and satisfying the algebra [9{12], metry of relativistic quantum field theory. However, its 2 2 H = ; y ; = ( y) = 0 ; (1) arXiv:2011.01239v2 [quant-ph] 25 May 2021 applications extend to a number of areas in mathematics fQ Q g Q Q including, most famously, Morse theory, mirror symme- where H is the Hamiltonian of the system and try, and generalized complex geometry. ( 1)F ; = 0. One says an operator is bosonic or In this letter, we focus on supersymmetric quantum fermionicf − Qg if it commutes or anticommutes with ( 1)F , mechanics [9{12] and bring the attention to the inter- respectively. The supercharge is thus a fermionic opera-− play between supersymmetry and the theory of quantum tor and the Hamiltonian bosonic. It follows directly from computation. As we discuss, a natural setting for incor- (1) that the spectrum of supersymmetric systems is pos- porating supersymmetry in quantum computation is the itive semidefinite, E 0, and that a state Ω has E = 0 ≥ j i fermionic model of quantum computation [13, 14]. Defin- iff Ω = y Ω = 0. All such states, which may be ing supersymmetry operators in the fermionic model, one bosonicQ j i or fermionic,Q j i are called supersymmetric ground can then map these to qubit space via a standard spin- states. A crucial property of states with E > 0 is that 1 2 Jordan-Wigner transformation or its generalizations. they are paired: for every such bosonic state there is a Having defined the action of supersymmetry in the space corresponding fermionic state with the same energy. This 1 2 is not necessarily the case for supersymmetric ground This can be thought of as the insertion of operators in states. In fact, a quantity of particular interest is the the Euclidean path integral of the theory, with periodic Witten index, defined as the difference in the number boundary conditions for fermions along a compactified of bosonic and fermionic supersymmetric ground states Euclidean time direction τ = it. An important property [12]: of the observables (5) and (6) is that they are invariant under exact deformations, B F F nE=0 nE=0 = Tr ( 1) ; (2) H I ≡ − − k k + k ; (7) where in the last equality one uses the fact that states O !O E with E > 0 are paired and thus do not contribute to as can be easily checked. For the former, this follows the trace. The Witten index gives a lower bound on the from properties of the supersymmetric ground state total number of supersymmetric ground states via the and, for the latter, from cyclicity of the trace. Thus, B F these observables are sensitive only to the cohomology inequality nE=0 + nE=0 . In particular, if = 0 the system must have supersymmetric≥ jIj ground states.I 6 class of supersymmetric operators. These robustness 2 properties will be relevant to our discussion of quantum Since = 0, supersymmetry defines the Z2-graded complexQ of vector spaces, computation below. F B F B Computational complexity of supersymmetric systems. C : Q Q Q ; (3) H −!H −!H −!H Let us briefly review relevant concepts of complexity the- and the Euler characteristic of C coincides with the Wit- ory (see, e.g., [15, 16]). The complexity class P is the class ten index. It is this topological nature of the Witten of decision problems (with a \yes/no" answer) which can index that makes it a robust quantity and, in some situ- be solved by a deterministic Turing machine in polyno- ations, easily calculable. mial time. The class NP is the class of decision problems Computing general physical observables in supersym- for which the problem instances which give \yes" can be metric systems can be as formidable a task as in non- checked in polynomial time. The complexity class #P supersymmetric systems. However, there are a subset is the set of counting problems associated to decision of physical observables, \supersymmetric observables," problems in NP. For example, whereas the problem of which have special properties and can often be computed deciding if a boolean formula has a satisfying instance exactly, the Witten index being an example. These relate is a problem in NP, the problem of counting how many to an important set of operators called supersymmetric, satisfying instances it has is a problem in #P. A problem or -closed, operators. A bosonic operator is -closed H is said to be #P-hard if it is at least as hard as any if Q O Q problem in #P or, more precisely, if any problem in #P can be reduced to H in polynomial time. A problem is [ ; ] = 0 : (4) said to be #P-complete if it is #P-hard and belongs to Q O the class #P. Among these, -exact operators are defined as those Q The Witten index (2) can sometimes be computed ex- which can be written as = ; Ψ0 , for some fermionic E fQ g actly and with little dynamical information. In particu- Ψ0. By nilpotency, all -exact operators are -closed. lar, supersymmetry ensures that the contribution of all Not all -closed operators,Q however, are necessarilyQ - Q Q states with E > 0 in the trace in (2) cancel out and exact; whether or not this is the case is determined by thus the index can be computed with no knowledge of the -cohomology of operators. Two supersymmetric Q the supersymmetric ground states themselves. Further- operators and 0 are said to be in the same cohomol- O O more, under certain conditions the index is invariant un- ogy class if 0 = + . The analogous definitions hold der small, supersymmetric, deformations of the system for fermionicO -closedO E and exact operators, exchanging Q [12] which can sometimes be exploited to bring the sys- commutators and anticommutators. tem to a weakly coupled point, where the Witten index An important set of physical observables is given by can be efficiently computed in perturbation theory. Al- the correlation function of supersymmetric operators in though this is often the case in specific supersymmetric a supersymmetric ground state: systems, we show next there can be no efficient algorithm for computing the Witten index for generic supersymmet- Ω (t ) (t ) Ω ; (5) hO1 ···OniΩ ≡ h j O1 1 ···On n j i ric systems, assuming standard conjectures in computa- tional complexity.