Quasi-probabilities on Fermionic Phase Spaces

Ninnat Dangniam, Christopher Jackson, Carlton Caves

Center for Quantum Information and Control, University of New Mexico

Christopher Ferrie

Centre for Quantum Software and Information, University of Technology Sydney

SQuInT, 23 February 2018

Why Quasi-probability?

"Resources"
Negativity
Contextuality
Bell non-locality
Entanglement
Interference
Quantum speed-up
Tools
Quasi-probability
Covariance matrix
Stabilizer formalism
Matrix product state
Monte Carlo
Computational complexity

Contextuality
Bell non-locality
Entanglement
Interference
Quantum speed-up

Covariance matrix
Stabilizer formalism
Matrix product state
Monte Carlo
Computational complexity

Negativity

Quasi-probability

"Resources"

Tools

Why Quasi-probability?

Positive quasi-probabilities

Monte Carlo

Classical simulation

Why Quasi-probability?

Monte Carlo

Classical simulation

Negative quasi-probabilities

No classical simulation

Why Quasi-probability?

Efficiently representable

positive quasi-probabilities

Negativity is a necessary but not sufficient quantum resource

Free states/ Operations Wigner function Discrete Wigner function
(Odd dimensions only)
Positive pure states Gaussians (Hudson thm)


 
Stabilizer states (Discrete Hudson thm)
 
(Non-convex-Gaussian) positive mixed states e.g. Single-photon-added thermal states

 
"Bounded universal"
states
Operations Quadratic bosonic Hamiltonians Cliffords
Measurements Gaussians Paulis

Q functions are always positive

Q(\alpha) = \langle \alpha|\rho|\alpha \rangle

Q function of a photon number state

"In evaluating the possibility of a classical explanation of an experiment, one must consider the negativity of not just the representation of [states] but of measurements as well, and one must look at representations other than that of Wigner."

This Talk

\rho
E

State

\mu_{\rho}(\Omega) = \text{Tr}[\rho F(\Omega)]

Measurement

\zeta_{E_k}(\Omega) = \text{Tr}[E_k D(\Omega)]

Quasi-probability Representations

Outcome k

Hermitian

Hermitian

Quantum experiment

\text{Tr}[A
A = \int d\Omega \, \text{Tr}[AF(\Omega)]
A
A
F(\Omega)
]

Frame

D(\Omega)

Dual frame

F(\Omega)
D(\Omega)

Frames and Their Duals

\text{Tr}[A
A = \int d\Omega \, \text{Tr}[A|\alpha \rangle \langle \alpha|]
A
A
|\alpha \rangle \langle \alpha|
]

Frame

D(\Omega)

Dual frame

F(\Omega)
D(\Omega)

Frames and Their Duals

Frame for the

Q function

Frame for the P function

\rho
E

State

\mu_{\rho}(\Omega) = \text{Tr}[\rho F(\Omega)]

Measurement

\zeta_{E_k}(\Omega) = \text{Tr}[E_k D(\Omega)]
P(E_k|\rho) = \text{Tr}[E_k \rho] = \int d\Omega \, \mu_{\rho} (\Omega) \zeta_{E_k} (\Omega)

Quasi-probability Representations

Born rule as phase space average

Outcome k

Hermitian

Hermitian

Quantum experiment

The frame or the dual frame must contain a non-positive operator

Quasi-probability Representations

\rho
E

State

Measurement

Quantum experiment

Fermions

Wigner function Discrete Wigner function
(Odd dimensions only)
Positive pure states Gaussians (Hudson thm)
 
Stabilizer states (Discrete Hudson thm)
 
(Non-convex-Gaussian) positive mixed states e.g. Single-photon-added thermal states

 
"Bounded universal"
states
Operations
 
Cliffords
Measurements Gaussians Paulis

Free states/ Operations

Quadratic bosonic Hamiltonians

Reversible free operations simply permute points of the phase space

U(g) F(\Omega) U^{\dagger}(g) = F(g \cdot \Omega)

G-covariant frame

Free operations should not increase "resource" (negativity)

The Wigner frame is self-dual and covariant w.r.t. all quadratic bosonic Hamiltonian evolutions

\{ a_j, a^{\dagger}_k \} = \delta_{jk}
H = \sum_{jk}

Quadratic Fermionic Hamiltonians

Free Fermionic Evolution

\{ a_j, a_k \} = 0
\alpha_{jk} a_j^{\dagger}a_k
+
\beta_{jk} a_j^{\dagger} a_k^{\dagger}
+ \text{H.C.}

Number-preserving

Squeezing

\{ c_{\alpha}, c_{\beta} \} = 2 \delta_{\alpha \beta}

Majorana operators

a_j = \displaystyle{\frac{c_{2j-1} + ic_{2j}}{2}}

Quadratic Fermionic Hamiltonians

Free Fermionic Evolution

H = \frac{i}{4} \sum_{\alpha\beta,\alpha < \beta}
h_{\alpha\beta}
[c_{\alpha},c_{\beta}]

Antisymmetric

\{ a_j, a^{\dagger}_k \} = \delta_{jk}
\{ a_j, a_k \} = 0

Majorana operators

\{ c_{\alpha}, c_{\beta} \} = 2 \delta_{\alpha \beta}
a_j = \displaystyle{\frac{c_{2j-1} + ic_{2j}}{2}}

Quadratic Fermionic Hamiltonians

Free Fermionic Evolution

H = \frac{i}{4} \sum_{\alpha\beta,\alpha < \beta}
h_{\alpha\beta}
[c_{\alpha},c_{\beta}]

Antisymmetric

\vec{c}(t) = e^{\bold{h}t} \vec{c}(0)
\bold{h}
\text{SO(2n)}
\in

Group of rotations

in 2n dimensions

\begin{pmatrix} c_1 \\ c_2 \\ \vdots \\ c_{2n} \end{pmatrix}
| \Omega \rangle = U(g) |0 \rangle

The group action on the vacuum define an SO(2n)-covariant Q Function

Orbit

Coset space

Number-preserving

\{\Omega \} = \text{SO}(2n)/\text{U(n)}

Fermionic Gaussian states

Phase space

=
Q(\alpha) = \langle \Omega|\rho|\Omega \rangle

Q function

S^2 \simeq \text{SO}(3)/\text{SO}(2)
|n\rangle
\theta
\phi
|0\rangle

Guiding Example

Q functions on a sphere for spin-j systems

Spherical Q Function

R(\eta,\theta,\phi) = R_Z(\phi) R_Y(\theta) R_Z(\eta)
|\langle n|m \rangle|^2 = \frac{(1+\cos\theta)}{2}^{2j}

Quasi-probability representation for spin-j systems

|n\rangle
\theta
\phi
|0\rangle
n
m

Spherical Q Function

R_{n \to m}
R(\eta,\theta,\phi) = R_Z(\phi) R_Y(\theta) R_Z(\eta)
|\langle n|m \rangle|^2 = \frac{(1+\cos\theta)}{2}^{2j}

Quasi-probability representation for spin-j systems

|n\rangle
\phi
|0\rangle
\theta
0
0

Spherical Q Function

R_Y
R_Z
R(\eta,\theta,\phi) = R_Z(\phi) R_Y(\theta) R_Z(\eta)
|\langle n|m \rangle|^2 = \frac{(1+\cos\theta)}{2}^{2j}

Quasi-probability representation for spin-j systems

|n\rangle
\phi
|0\rangle
R_Z
0
0
\theta

Spherical Q Function

R(\eta,\theta,\phi) = R_Z(\phi) R_Y(\theta) R_Z(\eta)
|\langle n|m \rangle|^2 = \frac{(1+\cos\theta)}{2}^{2j}

Quasi-probability representation for spin-j systems

|n\rangle
\phi
|0\rangle
\theta
R_Y
0
0
|0\rangle
|0\rangle
|0\rangle
R_{Y^+}
|0\rangle
R_{Z^+}
R_{Y^-}
R_{Z^-}

Fermionic Q Function

\text{U}(4)

2 commuting

two-mode squeezings

|\langle \Omega|\Omega' \rangle|^2 = (\frac{1+\cos \theta^+_{12}}{2}) (\frac{1+\cos \theta^+_{34}}{2})

Number-preserving

R_{Y^+}
R_{Z^+}
R_{Y^-}
R_{Y^-}
R_{Y^-}
R_{Z^-}
R_{Z^-}
R_{Z^-}

Jordan-Wigner transform to 4-qubit quantum circuit

|0\rangle
|0\rangle
|0\rangle
R_{Y^+}
|0\rangle
R_{Z^+}
R_{Y^-}
R_{Z^-}

Fermionic Q Function

\text{U}(4)

commuting

two-mode squeezings

|\langle \Omega|\Omega' \rangle|^2 = (\frac{1+\cos \theta^+_{12}}{2}) (\frac{1+\cos \theta^+_{34}}{2})

Number-preserving

R_{Y^+}
R_{Z^+}
R_{Y^-}
R_{Y^-}
R_{Y^-}
R_{Z^-}
R_{Z^-}
R_{Z^-}
\lfloor \frac{n}{2} \rfloor

Jordan-Wigner transform to 4-qubit quantum circuit

|\langle \Omega|\Omega' \rangle|^2 = |\langle 0| U^{\dagger}(\Omega) U(\Omega') |0 \rangle|^2
= |\langle 0| U(\Omega^{-1}\Omega') |0 \rangle|^2
= |\langle 0| U(Kg^{-1}g'K') |0 \rangle|^2

U(n)-bi-invariant function 

Fermionic Q Function

\in \text{Span}\{ \lfloor \frac{n}{2} \rfloor + 1 \, \, \text{zonal spherical functions} \}
K = \text{U}(n)

From the Q function, we can generate a continuous family of G-covariant frames and dual frames that gives a unique self-dual "Wigner function"

F^{(s)}(\Omega) = \int d\Omega' \,
\text{Tr} [F^{(s)}(\Omega) F^{(1)}(\Omega')]
|\Omega' \rangle \langle \Omega'|

P

Q

Wigner

s

-1

0

1

Stratonovich-Weyl axioms

Fermionic Wigner Function

From the Q function, we can generate a continuous family of G-covariant frames and dual frames that gives a unique self-dual "Wigner function"

P

Q

Wigner

s

-1

0

1

Stratonovich-Weyl axioms

= \text{Tr} [U(\Omega) F^{(s)}(0) U^{\dagger}(\Omega) U(\Omega') F^{(-s')}(0) U^{\dagger}(\Omega')]
= \text{Tr} [F^{(s)}(0) U(\Omega^{-1}\Omega') ^{(-s')}F(0) U^{\dagger}(\Omega^{-1}\Omega')]

U(n)-bi-invariant function!

\text{Tr} [F^{(s)}(\Omega) F^{(s')}(\Omega')]

Fermionic Wigner Function

4-mode fermionic Gaussian state

Fermionic Wigner Function

4-mode fermionic Gaussian state

  • Closest analogue, but may not be a fair comparison to bosonic quasi-probability representations
    • The continuous family of bosonic quasi-probability representations are not covariant under the full quadratic Hamiltonian evolution
    • There is no analogue of the Heisenberg group for fermions

Why Negativity?

  • Scaling of negativity and simulatability
  • Negativity of non-Gaussian states 
  • Connection to existing quasi-probability representations for fermions

Future Directions

Summary

  • We define quasi-probability representations in terms of frames and dual frames.
  • We obtain a continuous family of SO(2n)-covariant quasi-probability representations from fermionic Gaussian states.
  • We prove the uniqueness of the self-dual "Wigner function".
  • The construction applies to phase spaces that are symmetric homogeneous spaces and more.

ND, "Quantum Phase Space Representations and Their Negativities", University of New Mexico, PhD Dissertation (2018)

Fermion phase spaces

By Ninnat Dangniam

Fermion phase spaces

SQuInT, Santa Fe, 23 February 2018

  • 541