Reference
arXiv:1410.2143v2
(Special thanks to Hal Tasaki and Tohru Koma)
Introduction
Very new concept proposed by Wilczek
PRL 2012
PRL 2012
PRL 2013
P. Coleman, Nature 493, 166 (2013)
What is “2me crystal”?
Crystalline order in the temporal direcLon
Ordinary crystals: crystalline order in spa2al direcLons
x
y z
space x density
ρ(x)
Ordinary crystals
Lme t observable
O(t)
Time crystals
Special RelaLvity:
(Some) symmetry between space and 2me!
We know crystals exist.
Then why not Lme crystals?
Spontaneous Symmetry Breaking (SSB)
Can the Lme translaLon be spontaneously broken in a similar manner?
Our space itself is homogeneous.
Laws of physics do not depend on posiLon.
But, the state realized in nature (e.g., crystals) is not homogeneous.
SpaLal translaLon is spontaneously broken!
Wilczek’s idea
t Supercurrent J(x,t)
t Supercurrent J(x,t)
But! Obvious problems…
• Time dependence vs. eigenstates (“ground state”) X(t) = ei H t X(0) e− i H t
<X(t)> = ei E t <X(0)> e− i E t = <X(0)>
• State with steady moLon perpetual mo2on machine!
1st kind: produce work without input energy (violate 1st law of TD)
2nd kind: convert thermal energy to work with 100% efficiency (violate 2nd law of TD) 3rd kind: completely eliminates fricLon, dissipaLve forces, to maintain moLon.
Feynman ratchet Overbalanced
wheel
ObjecLons by Bruno and Nozières
PRL 2013
EPL 2013 Wilczek’s superconduncLng soluLon
with rotaLng kink is not the lowest energy state
OscillaLng phenomena i n nonequilibrium
• Josephson effect (potenLal gradient)
OscillaLng phenomena i n nonequilibrium
• Skyrmion crystals (temperature gradient)
Mochizuki et al, Nature Materials (2014)
OscillaLng phenomena i n nonequilibrium
• Belousov–ZhaboLnsky reacLon
QuesLon to address
Can the Lme translaLon be spontaneously broken in the thermal equilibrium?
In par2cular, in the ground state?
Our Answer: No, it cannot be.
(at least within the current framework of quantum mechanics and thermodynamics.)
cf. Classical Lme crystal
• For Lagrangians, like
‐ ℒkin = ½(∂tn)2 − ½(∇n)2 + k0 n·∇× n
‐ ℒkin = ½(∂tϕ)2 − ¼(∂i2ϕ)2 + k02 ½(∂iϕ)2
energy is minimized by k = k0 and ω = 0 (FFLO).
• Can we do the same for Lme derivaLves?
‐ ℒkin = ½(∂tn)2 − ω0 (nx∂t ny−ny∂t nx) − ½(∇n)2
‐ ℒkin = ¼(∂t2ϕ)2 − ω02 ½(∂tϕ)2 − ½(∂iϕ)2
‐ ℒkin = ¼(∂tϕ)4 − ω02 ½(∂tϕ)2 − ½(∂iϕ)2
‘energy’ is minimized by k = 0 and ω = ω0 (?)
• Quan2za2on is nontrivial, leading to different QM.
Shapere&Wilczek, “Branched Quan2za2on”, PRL (2012)
Definition of Time Crystals
Definition by
expectation value of order parameter
<O(x,t)>
How do we usually define SSB?
Take a field ϕi (x,t) such that
ϕi ‘= UϕiU‐1 = Dij ϕj U = ei ε Q, D = ei ε T
If <ϕi ‘(x,t)> ≠ <ϕj(x,t)> or <[Q, ϕi(x,t)]> = T <ϕi(x,t)> ≠ 0, we say the symmetry is broken.
• <[Sx, (‐1)iSy(xi,t)]> = i <(‐1)iSz(xi,t)> ≠ 0 (anLferromagnet)
• <[Px, ρ(x,t)]> = i ∂x<ρ(x,t)> ≠ 0 (crystal)
• <[H, O(x,t)]> = i ∂t<O(x,t)> = 0 (?) (Lme crystal)
But SSB is more subtle.
Easy to prove <[Q, ϕi(x,t)]> = 0 quite generally.
Symmetry of the canonical ensemble
• At a finite temperature, we define <ϕi> = tr(ϕi e−H/T) / tr(e−H/T)
• But…
tr(ϕi e−H/T)
= tr(Uϕie−H/TU‐1 ) U = ei ε Q = tr(UϕiU‐1e−H/T) [H, Q] = 0
= Dij tr(ϕje−H/T) UϕiU‐1 = Dij ϕj meaning that <[Q, ϕi(x,t)]> = 0.
Symmetry of the exact ground state
• explain crystals on blackboard
• exact spectrum of Heisenberg AiLferromagnet
Anderson’s tower of states
C. Lhuillier, arXiv: cond‐mat/0502464
“Symmetry breaking state” is not an eigenstate of H, except when order parameter
commutes with Hamiltonian.
The GS does not show SSB
| >
+| >
+| >
+| >
+| > + …
|GS > = ∫
0 ad
dx |x > =
Schrodinger cat state
Physical states
Each state has a disLnct expectaLon value of OP
they cancel a~er sum.
Symmetry breaking field
Explicitly break the symmetry of H by h: H’ = H – h ∫ddx ϕi(x,t).
H’ does not commute with Q.
Symmetry of canonical ensemble / GS is lost. limh limV <[Q, ϕi(x,t)]> can be nonzero.
• H’ = H – h Σi (‐1)i Sz(xi) (anLferromagnet)
• H’ = H – h ∫ddx cos(G x)ρ(x,t) (crystal)
• H’ = H – h ∫ddx cos(Ω t)ρ(x,t) (Lme crystal)
| >
+| >
+| >
+| >
+| > + …
SB field picks up a physical state!
|GS > = ∫
0 ad
dx |x > =
Difficulty for Lme crystal
Recap
• H’ = H – h Σi (‐1)i Sz(xi) (anLferromagnet)
• H’ = H – h ∫ddx cos(G x)ρ(x,t) (crystal)
• H’ = H – h ∫ddx cos(Ω t)ρ(x,t) (Lme crystal?)
Time‐dependent external field invalidates stat‐mech.
• ‘energy’ is ill‐defined.
• What is ‘ensemble’ under the field?
• Note: Floquet formalism gives the spectrum but does not help.
cf. Argument by Bruno and Nozières
SomeLmes, Lme‐dependence of the external field can be eliminated by switching to the rotaLng frame.
H’ = H – Ω L
They discussed stat‐mech under applied field using this frame.
• Even if their argument is true, it is not fully general.
• Not sure if their argument makes sense..
Definition by
long-range order <O(x,t)O(x’,t’)>
Long‐Range Order (LRO)
• AsymptoLc behavior of <O(x,t)O(x’,t)>: lim|x’‐x|∞<O(x,t)O(x’,t)> = σ2 > 0
• LRO measures how well two separated points are correlated.
• Symmetry breaking field is not needed.
We can use the symmetric&unique ground state without the expectaLon value of OP <O(x,t)> = 0.
• Theorem: LRO SSB (the inverse hasn’t been proven) See Koma‐Tasaki 1993, 1994
The GS does not show SSB
| >
+| >
+| >
+| >
+| > + …
|GS > = ∫
0 ad
dx |x > =
Schrodinger cat state
Physical states
Each state has a disLnct expectaLon value of OP
they cancel a~er sum.
The GS may have the long‐range order!
| >
+| >
+| >
+| >
+| > + …
|GS > = ∫
0 ad
dx |x > =
Schrodinger cat state
Physical states
All states have the same long range order!
they don’t cancel!
LRO for Lme crystals
As x – x’ increases:
• <Sz(xi) Sz(xj)> (– 1)i – j σ2 (anLferromagnet)
• <ρ(x) ρ(x’) > ΣG σG2cos G (x – x’) (crystal) We define TC by analogy:
• <O(x,t) O(x’,t’) > f(x – x’, t – t’) (Lme crystal) When f(x, t) is a nontrivial funcLon of t,
we call it a Lme crystal.
Remember that we use symmetric state:
• the unique GS at T = 0: |GS >
• the canonical ensemble at T>0: <ϕi> = tr(ϕi e−H/T)/tr(e−H/T)
Integrated version of LRO
• AnLferromagnet
<Sz(xi) Sz(xj)> (– 1)i – j σ2
Φ = Σi(– 1)i Sz(xi) limV∞< Φ2 >/V2 = σ2 ≠ 0
• Crystal
<ρ(x) ρ(x’) > ΣG σG2cos G (x – x’)
ΦG = ∫0 a ddx ρ(x) ei G x
limV∞< ΦGΦ–G >/V2 = σG2 ≠ 0
• Time crystal
<O(x,t) O(x’,t’) > f(x – x’, t – t’)
ΦG(t)=∫0 a ddx O(x,t) ei G x
limV∞< ΦG(t)Φ–G(t’)>/V2 = fG(t – t’)
Absence of time-dependent LRO
at T = 0
Outline
• Statement fG(t ) defined by LRO
– limV∞< e i H t ΦG e– i H t Φ–G>/V2 = fG(t)
– ΦG = ∫0 a ddx O(x) ei Gx
is 2me independent.
• AssumpLon locality of the Hamiltonian
(can be relaxed to a power‐low decaying interacLon)
• Approach double commutator introduced in Horsch‐von der Linden (1988)
Sketch of the proof
• For simplicity, we focus on G = 0.
• Consider two HermiLan operators A and B:
• We want to show:
• Step 1/3
|a+b| ≤ |a|+|b| X(1) − X(0)
= ∫0 1 ds (dX(s)/ds)
bound the integrand Many thanks to
Koma‐Tasaki
arXiv: 1410:2143
!!
Sketch of the proof
• step 2/3
• step 3/3
Schwarz ineq.
|a∙b| ≤ |a||b|
H|0 > = E0|0 >
< 0|[A,[H, A]]|0 >
= < 0|(A(H A – A H) – (H A – A H)A)|0 >
= 2 < 0|A (H – E0) A|0 >
Many thanks to Koma‐Tasaki
arXiv: 1410:2143
[a(x),b(y)] δ(x ‐ y)
k[ ˆA, [ ˆH, ˆA]]k = O(V 3−2) = O(V )
Absence of time-dependent LRO
at T > 0
Outline
• Statement fG(t) defined by LRO
– limV∞< e i H t ΦG e– i H t Φ–G>/V2 = fG(t)
– ΦG = ∫0 a ddx O(x) ei Gx
is 2me independent.
• AssumpLon locality of the Hamiltonian
(can be relaxed to a power‐low decaying interacLon)
• Approach Lieb‐Robinson bound (1972)
Sketch of the proof
Step 1/2: bound for gAB(t)
Sketch of the proof
Step 2/2: relaLon to fAB(t)
f(t) is Lme‐indep.
Summary
• Many difficulLes in characterizing Time Crystal based on the expectaLon value of order parameter.
• We instead define TC by the asymptoLc behavior of the correlaLon funcLon.
• We showed generally that the LRO cannot depend on Lme in the equilibrium or in the ground state.
• ‘Time crystal’ seems impossible.