Tale exta: Classically quantum

Schödinger equation
1d classical statistics
Continuous: Landau Hamiltonian
Discrete: 1d Ising model
Find equivalence in two ways
Path integral
Transfer matrices
To illustrate what theory is
 Schrödinger equation

Wave function: |ψi (ψ(x))

Obeys evolution equation:
i~ ∂ψ
∂t = Ĥψ
Stationary solutions: energy levels
En ψn = Ĥψn
Today’s concern: single variable x
p̂ 2
Hamiltonian: Ĥ = 2m + U(x̂)
Kinetic and potential energy
Hamiltonian in x-representation: (p̂→ − ~∂x )
~2 ∂ 2
Ĥ = − 2m ∂x 2 + U(x)

Singe-variable Schrödinger equation

 2 2 
~ ∂
En ψn (x) = − 2m ∂x 2 + U(x) ψn (x)
 (Quantum) statistics

Consider a system (quantum or classical)

that can be in a set of states.
Energies of the states: En
The sys is in thermal equilibrium
Probability to be in the state n: pn ∝ exp(−En /kB T ).

Partition function: ultimate answer

P
Z= n exp(−En /kB T )

Free energy: F = −kB T ln Z , Average energy:

hE i = −∂(ln Z )/∂((kB T )−1 )
Average = derivative with respect to conjugated force for
∂F
varaible x add E = −fx x. hxi = ∂fx
.
 1d Ising model

ultimately discretized!
nodes labelled by i, two ”spin
directions”σi = ±1
”Magnetic field”: −Hσ
Interaction between ”spins”:
per bond −Jσσ 0
Consider 1d lattice
Partition function:
(β =P 1/kB T , h = βH, K = βJ)
Z = {σi } exp (−βHeff ) =
P  P P 
{σi } exp K <ij> σi σj + h i σi

A 1922 phd exercise

 Landau Hamiltonian
continuous in space (y ) and variable(x, magnetization)
Coarse-graining of discrete models
→ x(y )
Two parts
Free energy density as function
of localR magnetization
βF = dy Ũ(x(y ))
Gradient term:
energy fine for inhomogenuity
 2
βF = dy K2 dy dx
R

Common application: 2nd order phase transition

Landau Hamiltonian
  2 
K dx
+ 2t x 2 + ux 4
R
βF = dy 2 dy

t > 0 symmetric solution x = 0 p

t < 0 symmetry breaking x = ± −t/u
 From Q to C: slicing

p̂ 2
one-particle Hamiltonian: Ĥ = + U(x).
2m

Partition function: Z = Tr exp −β Ĥ
Let’s slice it into small pieces:
Mesh in ”Imaginary time”: 0 < τi < β~, τi+1 − τi = dτ .
→:Z =      
Tr{exp −Ĥdτ /~ · exp −Ĥdτ /~ · exp −Ĥdτ /~ · . . . }
 
Slice operator: exp Ĥdτ /~ → W (x, x 0 )
A kind of discrete Schrodinder equation:
ψi+1 (x) = W (x, x 0 )ψi (x 0 ) dx 0
R
 
in the limit dτ → 0: exp −Ĥdτ /~ → 1 − Ĥdτ /~;
 
~ ∂2
−~ ∂ψ(x)
∂τ = − 2m ∂x 2 + U(x) ψ(x)
 Use of slicing?

OperatorR multiplication = extra integration:

ÂB̂ → dx1 A(x, x1 )B(x1 , x 0 )
RQ
Thus: Z = i dxi δ(x0 − xN )
W (xN , xN−1 ) . . . W (xi , xi−1 ) . . . W (x1 , x0 ).
Approximate commutation of
 tiny slices:
 H = H1 + 
H2 ;
dτ dτ dτ


exp − ~ (H1 + H2 ) ≈ exp − ~ Ĥ1 · exp − ~ Ĥ2
Let’s see what we have: W (x, x 0 ) = W1 (x, x 0 )W2 (x 0 );
W2 (x) = exp(−U(x)dτ /~)
0
 2   
exp −i p(x−x )
R dp p dτ
W1 (x, x 0 ) = 2π~ exp − 2m ~ ~
 0 2
q
= exp − m(x−x 2~dτ
) m~
2πdτ .
 Path integral
It is now    
RQ Q m(xi −xi−1 )2 U(xi−1 )dτ
Z= i dxi δ(x 0 − x N ) i exp − 2~dτ exp − ~

smooth x: (xi − xi−1 )2 /dτ →ẋ 2 dτ

Taking limit dτ → 0

Path integral representation

R β~ n 2 o
S mẋ
RQ

Z= τ dxτ exp − ~ ; S = 0
dτ 2 + U(x(τ ))

we sum up over closed paths

each path: a configuration of classical 1d field
 From C to Q: 1d lattice model

Take a rather general model

variable x(can be discrete) defined in N nodes
Energy per node: En (x): weight W2 (x)
Energy per bond: Eb (x, x 0 ): weight W1 (x, x).
Partition function:
dxi W1 (xN , xN−1 )W2 (xN−1 ) . . . W1 (xi+1 , xi )W2 (xi ) . . . W2 (x1 , x0 ).
R Q
Z = i

Suppose we’ve integrated till node i obtaining ψi (xi ).

R
On next step: ψi+1 (xi+1 ) = dxi W1 (xi+1 , xi )W2 (xi )ψ(xi )
Consider eigenvalue equation:k numbers eigenvalues
W (k) ψk (x) = dx 0 W1 (x, x 0 )W2 (x 0 )ψk (x 0 )
R

Partition function is given by the eigenvalues

 Eigenvalues <=> eigenenergies
(k) N
P
Periodic boundary conditions: Z = k (W )
In the limit N → ∞:

P.f. given by the highest eigenvalue

ln Z /N = ln W (0)

Compare with Q:
e −βEk
P
Finite temperature: Z = k
Limit T → 0 ln Z = −E0 β
(if) we can make continuous limit (y -cont.coordinate)
W1 (x, x 0 ) →
 exp(−A(x − x 0 )2 ); 
Z = exp (−E β))
RL   2
dx
E = 0 dy A dy + U(x(y ))

Full analogy
Example C → Q: Ising model in external field

Let’s symmetrize
p eigenvalue equation: p
W (k) ψk (x) = W2 (x) dx 0 W1 (x, x 0 ) W2 (x 0 )ψk (x 0 )
R

For Ising: x = σ = ±, W2 = exp(σh),W2 = exp(K σσ 0 )

Eigenvalue equation:

e −K
   K +h  
ψ+ e ψ+
W =
ψ− e −K e K −h ψ−

p
Eigenvalues: W0 , W1 = e K cosh h ± e 2K sinh2 h + e −2K .
Eigenfunctions: easy to find
 Averages in general 1d model

Very quantum!
Average in one node: hf (xi )i?
dxf (x)ψ02 (x)
R
Through ”ground state” wf. ψ0 (x): hf (xi )i =
Correlation in two nodes separated by N bounds: hhf (xi )f (xi+N )ii.

From ”ground” to excited”:

N
W (k)
X 
hhf (xi )f (xi+N )ii = |fk0 |2
W (0)

dxψk∗ (x)f (x)ψ0 (x).

R
Matrix elements: fk0 =
Correlation length ξ: corr ' exp(−N/ξ) at N → ∞
ξ = 1/ ln(W (0) /W (1) )
Is defined by the first excited state
 Correlations in Ising model

Apply these reasonings to 1d Ising model

Have solution, can look at correlations of magnetization:
corr (N) ≡ hhσi σi+N ii
Only two states: single exponent:corr ' exp(−N/ξ)
Correlation
 length: √ 
−1 cosh h+ sinh2 h+e −4K
ξ = ln √
cosh h− 2
sinh h+e−4K

Some limits:
Low temperatures (K  1), h = 0:
ξ = e 2K exponentially long.
Low temperatures, but e −2K  h  1
ξ = h−1 .
 Qubit physics

Some dimensionless units, ~ = 1

Two - well potential : (  U)
U(x) = U(x 2 − 1)2 + x
Strong barrier: U  1/m
w.f. Localized in each well p
oscillator spectrum with ω0 ' 8U/m.
In fact: quantum tunneling
 between
 the barriers
8 U
with amplitude ∆ ' exp − 3 ω0
2 × 2 Hamiltonian:
 
 ∆
Hqubit =
∆ −
√
Enegies: E± = 2 + ∆2
Delocalized at  = 0, || ' ∆
Localized at ||  ∆
 Translating it into 1d
This maps into
 a Landau Hamiltonian 
2
m ∂x
βF = dy 2 ∂y + U(x 2 − 1)2 + x
R

What does the qubit maps into?

Two wells =
two degenerate magnetizations, domains
Quantum tunneling = domain wall
Thickness: ' ω0−1
Q. amplitude - exp. small Boltzmann weight
 - m.f. lifting the degeneracy
√ 
Free energy βF = ln E− = ln 2 + ∆ 2
Magnetization: hxi = √ 
2 +∆2
Correlation length: √
ξ = (E+ − E− )−1 = 2/ 2 + ∆2
Two limits:
ξ = 2||−1 and ξ = 2∆−1 .