Professional Documents
Culture Documents
E. Guzmán-González‡
Departamento de Física
Universidad Autónoma Metropolitana-Iztapalapa
PO Box 55-534, 09340, CDMX, México.
E. Serrano-Ensástiga§
Centro de Nanociencias y Nanotecnología, Universidad Nacional Autónoma de México
arXiv:2202.01973v1 [quant-ph] 4 Feb 2022
Holonomic quantum computation makes use of non-abelian geometric phases, associated to the evolution of
a subspace of quantum states, to encode logical gates. We identify a special class of subspaces, for which a
sequence of rotations results in a non-abelian holonomy of a topological nature, so that it is invariant under
any 𝑆𝑂(3)-perturbation. Making use of a Majorana-like stellar representation for subspaces, we give explicit
examples of topological-holonomic (or toponomic) NOT and CNOT gates.
It is well known that cyclic evolution of a quantum state [𝜓] tioned above. To this end, we need to synthesize a number of
in the projective Hilbert space ℙ() ≡ ℙ of a physical sys- results, some of them quite new.
tem gives rise to a geometric phase, which is invariant under The first ingredient we will need is a formulation of the
time-reparametrizatons of the curve 𝐶(𝑡) traced in ℙ [1] (we Wilczek-Zee (WZ) effect, that parallels the treatment of the
denote by [𝜓] points in ℙ, i.e., equivalence classes of normal- abelian case in [8]. In that reference, given a curve [𝜓𝑡 ],
ized states |𝜓⟩ in the Hilbert space that only differ by an 0 ≤ 𝑡 ≤ 𝑇 , traced by a state in ℙ, one chooses arbitrarily but
overall phase). The non-abelian generalization of this effect, smoothly a phase for each 𝑡, obtaining a lift |𝜓𝑡 ⟩ in . In terms
due to Wilczek and Zee [2], involves the cyclic evolution of of the latter, the geometric phase assigned to [𝜓𝑡 ] is given by
a 𝑘-dimensional degenerate subspace Π of the Hilbert space
, in which case the phase factor mentioned above gets re- 𝜑geo = 𝜑tot − 𝜑dyn
placed by a unitary 𝑘 × 𝑘 matrix 𝑈 which acts in Π — as in the 𝑇
abelian case, 𝑈 is reparametrization invariant. This latter fact = arg⟨𝜓(0)|𝜓(𝑇 )⟩ + 𝑖 𝑑𝑡 ⟨𝜓(𝑡)|𝜕𝑡 |𝜓(𝑡)⟩ , (2)
∫0
has prompted the suggestion of using such holonomies [3] in
the implementation of logical gates in quantum computing, as where the two terms on the right hand side are known as the
it provides a certain level of immunity from noise effects [4, 5]. total and dynamical phase, respectively — it is easily seen that
Specifying the above to the case of a spin-𝑠 system, so that 𝜑geo does not depend on the particular lift |𝜓𝑡 ⟩ chosen, so it is
= ℂ𝑁 , 𝑁 = 2𝑠 + 1, the simplest realization of a cyclic a property of the curve [𝜓𝑡 ] in ℙ.
evolution of a state in ℙ is through a sequence of rotations 𝑅𝑡 The above physical quantum states [𝜓] correspond to rays
in 𝑆𝑂(3) (with 𝑅0 being the identity operation), in , and the set of such rays is the projective space ℙ. Sim-
ilarly, 𝑘-dimensional subspaces in , of the type that ap-
[𝜓𝑡 ] = [𝐷(𝑠) (𝑅̃ 𝑡 )|𝜓0 ⟩] , (1)
pear in the WZ effect, form the Grassmannian Gr(𝑘, 𝑁) (note
as in the familiar example of a precessing spin, where 𝐷(𝑠) (𝑅̃ 𝑡 ) that Gr(1, 𝑁) = ℙ) and the cyclic evolution of such a sub-
is the spin-𝑠 irreducible representation of the lift 𝑅̃ 𝑡 of 𝑅𝑡 in space corresponds to a closed curve Π(𝑡), in Gr(𝑘, 𝑁), with
𝑆𝑈 (2) — said lift is unique, if one specifies that 𝑅̃ 0 = 𝐼 0 ≤ 𝑡 ≤ 𝑇 and Π(0) = Π(𝑇 ). Given an arbitrary or-
(rather than −𝐼). For a general state [𝜓0 ], cyclicity implies that thonormal basis {|𝜙𝑖 (𝑡)⟩}, 𝑖 = 1, … , 𝑘 in Π(𝑡), with |𝜙𝑖 (0)⟩
𝑅𝑡 must start and end at the identity of 𝑆𝑂(3), 𝑅0 = 𝑅𝑇 = 𝐼, not necessarily equal to |𝜙𝑖 (𝑇 )⟩ (since a given plane may be
but when [𝜓0 ] has a nontrivial discrete rotational symmetry spanned by many different bases), define the matrix 𝑄 with
group Γ = {𝑅(0) ≡ 𝐼, 𝑅(1) , … , 𝑅(𝑝) }, 𝑅𝑡 might also end at any entries 𝑄𝑖𝑗 = ⟨𝜙𝑖 (0)|𝜙𝑗 (𝑇 )⟩, and the Wilczek-Zee connection
of the symmetry rotations 𝑅(𝑖) ∈ Γ. In this latter case, it has (𝑡), with entries 𝑖𝑗 (𝑡) = ⟨𝜙𝑖 (𝑡)|𝜙̇ 𝑗 (𝑡)⟩, the dot denoting time
been shown in [6], that when |𝜓0 ⟩ is anticoherent [7], meaning derivative. The polar part of 𝑄 is given by 𝑃 = 𝑊 𝑉 † , where
its spin expectation value vanishes, the geometric phase ac- 𝑄 = 𝑊 𝐷𝑉 † is the singular value decomposition of 𝑄, with
quired is of a topological character, and is thus invariant under 𝑊 , 𝑉 unitary, and 𝐷 diagonal. Then to the curve Π(𝑡) we
arbitrary (not necessarily small) smooth perturbations. The assign, following [9] (see also [10]), the unitary holonomy
principal aim of the present work is a generalization of these
considerations to the case of the non-abelian holonomies men- 𝑈geo = 𝑃 𝐹 −1 , (3)
2
𝐷(𝑠) (𝑅𝑚 )|𝜙𝑖 ⟩ ≠ |𝜙𝑖 ⟩, but the plane they generate after the ro-
tation might be identical with the original one, 𝑅𝑚 (Π) = Π.
Clearly, to be able to describe efficiently loops in Gr(𝑘, 𝑁),
we need an intrinsic characterization of a 𝑘-plane, that brings
to the forefront its transformation properties under rotations.
For the case 𝑘 = 1, i.e., when dealing with physical states in
ℙ, there is such a characterization, due to Majorana [13] (see
also [14, 15]). It is well known that spin-1/2 states can be rep-
resented as points on the Bloch sphere, essentially via their
spin expectation value,
[𝜓] ↦ 𝑛̂ = 2⟨𝜓|𝐒|𝜓⟩ . (10)
It is also a fact that the quantum states of a system of 2𝑠 qubits,
that are symmetric under permutations of the qubits, are in 1-
to-1 correspondence with spin-𝑠√states, e.g., the 2-qubit sym-
metric state (| + −⟩ + | − +⟩)∕ 2 corresponds to the spin-1 FIG. 1. Majorana constellation of the spin-2 state |𝜒⟩ of Eq. (12),
state |𝑠 = 1, 𝑚 = 0⟩ — a similar correspondence holds for any forming a regular tetrahedron with one vertex on the 𝑥-axis.
spin-𝑠 state |𝜓⟩,
|𝜓⟩ ↦ |𝜓1 ⟩ ∨ |𝜓2 ⟩ ∨ … ∨ |𝜓2𝑠 ⟩ , (11) in [16], albeit in a more complicated form. A general spin-𝑠
where the |𝜓𝑖 ⟩ are qubit states and the ∨-product denotes a 𝑘-plane can be uniquely represented by a family of constel-
symmetrized tensor product, e.g., |𝜓1 ⟩ ∨ |𝜓2 ⟩ ≡ |𝜓1 ⟩ ⊗ lations, of various spins, the values of which depend on both
√ 𝑠 and 𝑘, each of which is weighted by a complex number. If
|𝜓2 ⟩ + |𝜓2 ⟩ ⊗ |𝜓1 ⟩)∕ 2. Of course, a 1-to-1 correspon-
one arbitrarily orders the constellations, one may treat these
dence can always be established, in an infinity of ways, be-
complex weights as a pseudo-spinor, and assign to it a “spec-
tween vector spaces of equal dimension — the special prop-
tator” constellation, so that the above 𝑘-plane representation
erty of the particular correspondence mentioned above is that
is purely visual. There is a way of doing this such that un-
it commutes with the action of rotations, namely, given a spin-
der rotations, the Majorana constellations rotate rigidly, while
𝑠 state |𝜓⟩ = (𝜓1 , … , 𝜓𝑁 ), one may rotate it by 𝑅 ∈ 𝑆𝑂(3) by
the spectator constellation remains invariant. Thus, any rota-
left-multiplying it with 𝐷(𝑠) (𝑅), or, one may represent |𝜓⟩ as
tional symmetries of the 𝑘-plane are conveniently encoded in
in (11), transform each of the qubits with the spin-1/2 matrix
its multiconstellation.
𝐷(1∕2) (𝑅), and map the result back into a spin-𝑠 state with the
As a concrete example of the above general procedure, con-
inverse mapping — the results of the two sequences of opera-
sider the spin-2 2-plane ΠNOT ∈ Gr(2, 5), generated by
tions are the same. Since each state in the r.h.s. of (11) can be
represented by a point on the unit sphere, |𝜓⟩ itself can be rep- 1 √
resented by an unordered (because of the symmetrization) set |𝜓1 ⟩ = √ (1, 0, 0, 2, 0) (13)
of 2𝑠 points (stars) on the sphere, its Majorana constellation. 3
1 √
When |𝜓⟩ is transformed by 𝐷(𝑠) (𝑅) in , its constellation |𝜓2 ⟩ = √ (0, − 2, 0, 0, 1) . (14)
rigidly rotates by 𝑅 in physical space. To appreciate the enor- 3
mous advantage furnished by this representation of rays in ,
consider the spin-2 state The Majorana constellations of these states are both regular
(√ √ √ √ √ √ √ ) tetrahedra, antipodal to each other, with one vertex on the 𝑧-
3−2 6 3+ 6 1 3− 6 4+ 2 axis (see image on the left in Fig. 2). Following the procedure
|𝜒⟩ = , , √ , , √ , detailed in [16], we find that the vector space inside which
12 6 2 2 6 4 6
(12) the spin-2 2-planes are naturally embedded (this is the Plücker
and ponder whether it has any rotational symmetries, i.e., embedding, see, e.g., p. 43 of [17], or p. 110 of [18]) is 10
whether there exists any rotation 𝑅 such that 𝐷(2) (𝑅)|𝜒⟩ = dimensional, and splits into spin-3 and spin-1 subspaces. For
𝑒𝑖𝛼 |𝜒⟩, 𝛼 ∈ ℝ, so that [𝜒] ∈ ℙ is invariant under 𝑅 — ad- ΠNOT itself we find the components
( ))
mittedly not an easy question. But then, taking a look at the 1 ( √ √ √ )(
corresponding Majorana constellation, shown in Fig. 1, one ΠNOT = − 2, 0, 0, 5, 0, 0, 2 0, 0, 0 , (15)
3
realizes (e.g., by verifying that all stars are equidistant) that it
forms a regular tetrahedron, and one immediately concludes where we have grouped together the components of each spin
that [𝜒] is invariant, e.g., under a rotation about the 𝑥-axis by multiplet. Since the spin-1 component vanishes in this case,
2𝜋∕3. the multiconstellation of ΠNOT only has a spin-3 constella-
Could it be that a similar stellar representation exists for tion, shown on the right in Fig. 2 — this is a regular octahe-
spin-𝑠 𝑘-planes? Indeed it does, as it has been recently shown dron, which fully represents the rotational symmetries of the
4
FIG. 2. Left: Majorana constellations of the states |𝜓1 ⟩ (red), |𝜓2 ⟩ FIG. 3. Majorana constellations of the states |𝜓𝑎 ⟩ (red), |𝜓𝑏 ⟩ (blue),
(blue) — see Eqs. (13), (14). Right: Majorana-like spin-3 constella- which form an alternative basis of ΠNOT .
tion of the plane ΠNOT , generated by |𝜓1 ⟩, |𝜓2 ⟩ — it forms a regular
octahedron. The spin-1 component of ΠNOT is zero, hence, its spec-
tator constellation consists of a single star at the north pole (yellow).
the holonomy associated to the sequence of rotations 𝑅2 (𝑡) = [3] B. Simon, “Holonomy, the quantum adiabatic theorem, and
𝑅2𝜋𝑡̂𝐳∕5 , 0 ≤ 𝑡 ≤ 1, with 𝑅2 (1) another symmetry rotation of Berry’s phase,” Phys. Rev. Lett., vol. 51, pp. 2167–2170, 1983.
ΠCNOT . [4] P. Zanardi and M. Rasetti, “Holonomic quantum computation,”
Phys. Lett. A, vol. 264, pp. 94–99, 1999.
Concluding this short letter we point out the necessity to [5] P. Solinas, P. Zanardi, and N. Zanghí, “Robustness of non-
refine our search for anticoherent planes in the spin-𝑠 Hilbert abelian holonomic quantum gates against parametric noise,”
space. The ones presented above, as well as several others that Phys. Rev. A, vol. 70, p. 042316, 2004.
we are aware of, were found with a variety of ad hoc methods [6] P. Aguilar, C. Chryssomalakos, E. Guzmán-González, L. Han-
otel, and E. Serrano-Ensástiga, “When geometric phases turn
that fail to convey a satisfactory geometrical picture of their
topological,” J. Phys. A: Math. Theor., vol. 53, no. 6, p. 065301,
locus inside the corresponding Grassmannian — we are cur- 2020. arXiv:1903.05022.
rently pursuing such an understanding in the hope of realizing [7] J. Zimba, “Anticoherent spin states via the Majorana representa-
additional noise-tolerant logical gates. tion,” Electronic Journal of Theoretical Physics, vol. 3, no. 10,
pp. 143–156, 2006.
[8] N. Mukunda and R. Simon, “Quantum kinematic approach
to the geometric phase: I. General formalism,” Ann. Phys.,
vol. 228, pp. 205–268, 1993.
[9] D. Kult, J. Åberg, and E. Sjöqvist, “Noncyclic geometric
ACKNOWLEDGEMENTS changes of quantum states,” Phys. Rev. A, vol. 74, p. 022106,
2006.
[10] E. Sjöqvist, “Geometric phases in quantum information,” In-
CC, LH, and ESE would like to acknowledge partial finan- tern. J. Quant. Chem., vol. 115, pp. 1311–1326, 2015.
cial support from the DGAPA PAPIIT project IN111920 of [11] D. Varshalovich, A. Moskalev, and V. Khersonskii, Quantum
UNAM. ESE would also like to thank DGAPA UNAM for a Theory of Angular Momentum. World Scientific, 1988.
postdoctoral fellowship. [12] R. Pereira and C. Paul-Paddlock, “Anticoherent subspaces,”
Jour. Math. Phys., vol. 58, p. 062107, 2017.
[13] E. Majorana, “Atomi orientati in campo magnetico variabile,”
Nuovo Cimento, vol. 9, pp. 43–50, 1932.
[14] I. Bengtsson and K. Życzkowski, Geometry of Quantum States
∗
chryss@nucleares.unam.mx (2nd Ed.). Cambridge University Press, 2017.
†
hanotel@correo.nucleares.unam.mx [15] C. Chryssomalakos, E. Guzmán-González, and E. Serrano-
‡
edgar.guzman@correo.nucleares.unam.mx Ensástiga, “Geometry of spin coherent states,” J. Phys. A: Math.
§
edensastiga@ens.cnyn.unam.mx Theor., vol. 51, no. 16, p. 165202, 2018. arXiv:1710.11326.
[1] M. V. Berry, “Quantal phase factors accompanying adiabatic [16] C. Chryssomalakos, E. Guzmán-González, L. Hanotel, and
changes,” Proceedings of the Royal Society of London A: Mathe- E. Serrano-Ensástiga, “Stellar representation of multipartite an-
matical, Physical and Engineering Sciences, vol. 392, no. 1802, tisymmetric states,” Commun. Math. Phys., vol. 381, pp. 735–
pp. 45–57, 1984. 764, 2021. arXiv:1909.02592.
[2] F. Wilczek and A. Zee, “Appearance of gauge structure in simple [17] I. R. Shafarevich, Basic Algebraic Geometry 1. Springer, 2013.
dynamical systems,” Phys. Rev. Lett., vol. 52, pp. 2111–2114, [18] N. Jacobson, Finite-Dimensional Division Algebras Over
Jun 1984. Fields. Springer, 2010.