Salvador Workshop 2021

Qubitization of Field Theories
Paulo Bedaque
A. Alexandru, A. Carosso, Andy
Sheng,…
Date
Classical numerical quantum
field theory/many-body physics:
iS[ϕ]
∫ Dϕ e [ϕ]
⟨ ⟩=
𝒪
∫ Dϕ e iS[ϕ]
𝒪
−SE[ϕ] N
∫ Dϕ e [ϕ] 1 a
∑
⟨ ⟩= ≈ [ϕ ]
𝒪
𝒪
∫ Dϕ e −SE [ϕ] N a=1
𝒪
mani
real (cl
Large amount of theory and

practice:
✤ What can and cannot be

computed
✤ How many gates (time) is
V
required, sometimes cost ~e
✤ Large collection of
algorithms
✤
iS[ϕ]
∫ Dϕ e [ϕ]
⟨ ⟩=
𝒪
∫ Dϕ e iS[ϕ]
𝒪
iS is purely imaginary:
the mother of all sign problems
−SE[ϕ]
∫ Dϕ e [ϕ]
⟨ ⟩=
𝒪
∫ Dϕ e −SE [ϕ]
𝒪
SE may be complex (chemical
V
potential) probably has cost O(e ):
sign problem
Direct diagonalization:
V 3
exponential cost ~ (2 )
−SE[ϕ]
∫ Dϕ e [ϕ]
⟨ ⟩=
𝒪
∫ Dϕ e −SE [ϕ]
𝒪
SE may be complex (chemical
V
potential) probably has cost O(e ):
QCD equation of state, Hubbard
model away from 1/2-filling, …
iS is purely imaginary:
Quantum numerical quantum
| ψ⟩ = α | 0⟩ + β | 1⟩
qubit:
iπ
− 2 σzΔt
e unita
gates
Few (very clever) algorithms

doing “weird” stuff
−iHΔt −iHΔt
| ψ(0)⟩ e e ⋯ | ⟨ψ(t) | | ψ(t)⟩ |2
𝒪
}
Troter formula
but Quantum Physics is easy !

1. Encode the Hilbert space into

qubits
2. Prepare the initial state
3. Encode the hamiltonian into
quantum gates
4. Find something suitable to
measure
1. Encode the Hilbert space into

qubits
2. Prepare the initial state
3. Encode the hamiltonian into
quantum gates
4. Find something suitable to
measure
Field theories have infinite
dimensional Hilbert spaces
but
Quantum computers have finite
N
registers ~e
Discretize space (lattice)

bosonic theories: discretize field
space
Field theories have infinite
dimensional Hilbert spaces
but
Quantum computers have finite
N
registers ~e
Example: nuclear physics

(protons and neutrons, spin up
| q1⟩ | 0000⟩
and down)| 1010⟩
| q2⟩
| q3⟩
vacuum neutron u
| q4⟩ proton dow
| q1⟩ and down)
| q2⟩
{
| q3⟩ site 1
| q4⟩ # of qubits ~ V
Hilbert space dimension 4
.
.
.
| q1⟩ | 0000⟩ | 1010⟩
| q2⟩
site N
| q3⟩
vacuum neutron u
| q4⟩ proton dow
| q1⟩ and down)
| q2⟩
{
| q3⟩ site 1 e −iΔtH0
| q4⟩
e −iΔtV
.
. Local ham
. polynomial #
| q1⟩ # gates
| q2⟩ exponentia
site V e −iΔtH0
| q3⟩
| q4⟩
This kind of encoding does not
work for bosons:
occupation number n=0,1,2,3,
…
Condensates ?
Technical complication
Naive truncations break symmetries of the theory: no (space) continuum limit
continuum limit:
g0(a), g1(a), . . . g̃0(a), g̃1(a), . . .
} g0(a → 0) ∼
1
log(Λa)
}
1
g4(a → 0) ∼ 4
a
SU(3) gauge theory
U
at each link: a a
iλ A
U U
Ψ(U) = Ψ(e )
U
infinite-dimensional 8-dimensional
space space
finest discretization: SU(3)

S(1080) (“Valentiner
group”)
S(1080) gauge theory
U
1080
∑
U U Ψ(U) = ψiUi
i=1
U
1080-dimensional
space (11 qubits)
2 1 2
∑ ∑
US = − □ − □
p p
g0 p
2 g1 p
2
U U
No continuum limit.
There are no g0, g1 for fine
enough lattices
2 1 2
∑ ∑
US = − □ − □
p p
g0 p
2 g1 p
2
U U
U
S(1080)
Monte Ca
extrapolate to the same SU(3) improved
continuum limit Monte Carlo
SU(3)
Monte Carlo
2.2
2.1
2.0
S(1080)
1.9
++
1.8 E
SU(3)--athenodorou
glueball mass
SU(3)--us
1.7 S
1.6
S1080 Mo
Chen
1.5
0.0 0.2 0.4 0.6 0.8 1.0 1.2
O(3) σ-model
in 1+1D it is asymptotically
free, like QCD
1
2g ∫
2 μ
S = 2 d x ∂μn . ∂ n
unit vecto
at each point:
n Ψ(n)
infinite-dimensional 2-dimensional
2
space space: S
O(3) σ-model
free, like QCD
1
2g ∫
2 μ
S = 2 d x ∂μn . ∂ n
unit vecto
at each point:
{
rotation invariant: O(3)
n Ψ(n)
20-dimensional discrete: 20 points
space
O(3) σ-model
free, like QCD
1
2g ∫
2 μ
S = 2 d x ∂μn . ∂ n
unit vecto
at each point:
{
n = (x1, x2, x3) Ψ = ψ0 + ψi xi + ψij xi xj + ⋯

2 2 2
x1 + x2 + x3 =1
infinite-dimensional
O(3) σ-model
free, like QCD
1
2g ∫
2 μ
S = 2 d x ∂μn . ∂ n
unit vecto
at each point:
{
n = (x1, x2, x3) Ψ = ψ0 + ψiσi

2 2 2
σ1 + σ2 + σ3 =3
4-dimensional 4-p
“fuzzy
O(3) σ-model
free, like QCD
1
2g ∫
2 μ
S = 2 d x ∂μn . ∂ n
unit vecto
at each point:
{
n = (x1, x2, x3) Ψ = ψ0 + ψiσi

2 2 2
σ1 + σ2 + σ3 =3
4-dimensional 4-p
“fuzzy
Fuzzy O(3) σ-model
4 V
Hilbert space: (C )
∑[ ]
η
HΨ = ηg [σi(x), [σi(x)], Ψ] ± 2 σi(x)σi(x + 1)Ψ
2
x
g
2
∇ 2
∂n . ∂n ≈ (nx+1 − nx) = 2 − 2nx+1 . nx
exact O(3) invariance

σ-model is exactly solvable
fuzzy model can be “solved” by
Fuzzy O(3) σ-model
site 1 {
site 2 {
−iΔtK −iΔtV1 −iΔtV2 −iΔtV3
e e e e
3-site
simulation:
Fuzzy O(3) σ-model
a1 aN
| Ψ⟩ = tr[A ⋯ . . . A ] | a1⋯aL⟩
a1 a2
ai = 1,2,3,4
i = 1,...ba
variational parameters A ij
1. find energy gap Δ and

correlation length 1/m
2. adjust η so Δ=m (Lorentz
symmetry)
3. Δ(L) is determined by phase
O(3) σ-model
IR(asymptotically free)
UV
} ~M ~ΛQCD
} ~1/a
continuum O(3) sigma model

Antiferromagnetic
fuzzy O(3) σ-model L << 1/m: pert
L >> 1/m:
Luscher formula
Ferromagnetic fuzzy
O(3) σ-model L << 1/m: pert
L >> 1/m:
pert. theory
Generalizations
O(5), O(7), … are running now

different “commutative”
truncation of O(3) is running
now
O(4)=SU(2) x SU(2): chiral
model
SU(2) gauge theory is
reminiscent of chiral models
SU(3)? Quarks?
Summary
(Trotterized) time evolution mimics real time evolution

local hamiltonian can lead to exponential improvement on finite density/real time
calculations
encoding bosonic theories is tricky: preserve some symmetries to recover the
continuum limit
fuzzy sphere construction works for the σ-model; what about other models?

Salvador Workshop 2021

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Salvador Workshop 2021

Uploaded by

Copyright:

Available Formats

Qubitization of Field Theories

Large amount of theory and

✤ What can and cannot be

Few (very clever) algorithms

but Quantum Physics is easy !

1. Encode the Hilbert space into

1. Encode the Hilbert space into

Discretize space (lattice)

Example: nuclear physics

g0(a), g1(a), . . . g̃0(a), g̃1(a), . . .

finest discretization: SU(3)

n = (x1, x2, x3) Ψ = ψ0 + ψi xi + ψij xi xj + ⋯

n = (x1, x2, x3) Ψ = ψ0 + ψiσi

n = (x1, x2, x3) Ψ = ψ0 + ψiσi

exact O(3) invariance

1. find energy gap Δ and

continuum O(3) sigma model

O(5), O(7), … are running now

(Trotterized) time evolution mimics real time evolution

You might also like