arXiv:1302.5843v3 [cond-mat.stat-mech]

Ising formulations of many NP problems

arXiv:1302.5843v3 [cond-mat.stat-mech]

Ising formulations of many NP problems

?

?

?

?

?

?

?

?

?

?

?

?

?

Ising model?

Quantum Version?

How is related to TSP?

Ising model?

Quantum Version?

How is related to TSP?

Statistical Mechanics

Partition Functions

Analitycal Mechanics

Hamiltonians

Angular Momentum

Energy

Canonical Coordinates

Spin

Magnetic Momentum

Quantum Mechanics

Electromagnetism

NOT REALLY!

Do I need SOME PHYSICS BACKGROUND?

Thermodinamics

Statistical Mechanics

Partition Functions

Analitycal Mechanics

Hamiltonians

Do I need SOME PHYSICS BACKGROUND?

Angular Momentum

Energy

Canonical Coordinates

Spin

Magnetic Momentum

Quantum Mechanics

Electromagnetism

JUST A LITTLE

Thermodinamics

outline

An ising model recap

Ludmila Augusta Soares Botelho

History time: around 100 years ago

Is there is a very simple model to describe ferromagnetism?

Is there a phase transition in the magnetism of a linear sequence of little magnetic moments where only neighbors are energetically coupled?

(Ferro)magnetism

Wilhelm Lenz

{

{

{

$$z$$

History time: around 100 years ago

(Ferro)magnetism

Wilhelm Lenz

 Ernst Ising

?

!

Wilhelm Lenz

 Ernst Ising

?

 Model

It does not have phase transition!

... what about some quantum mechanics?

... more dimensions?

History time: around 100 years ago

( the best time)

Why do we care about Statistical Mechanics?

Physics time

( the best time)

Physics time

Why do we care about Statistical Mechanics?

It relates the

 microscopic 

with the

MACROSCOPIC

Physics time

Why do we care about Statistical Mechanics?

It relates the

 microscope 

with the

MACROSCOPE

  • Few assumptions

Canonical Ensemble

(More precisely, Statistical Mechanics)

S

R

~

~

~

  • Probability of particular microscopic state

$$P_j = \frac{e^{-\beta E_j}}{Z} $$

$$Z = \sum_{\sigma}e^{-\beta E_\sigma}$$

ustandssumme

  • Partition function 

(T, V, N)

Physics time

  • Few assumptions

Canonical Ensemble

(More precisely, Statistical Mechanics)

S

  • Probability of particular microscopic state

$$P_j = \frac{e^{-\beta E_j}}{Z} $$

$$Z = \sum_{\sigma}e^{-\beta E_\sigma}$$

ustandssumme

  • Partition function 

(T, V, N)

  • Free Energy
  • Systems Total Energy
  • Heat Capacity

Connection with Thermodynamics

  • Entropy
  • Magnetic Susceptibility
  • Magnetization

R

quiz: why the "-" signal?

Systems energy

$$\mathcal{H} = -J \sum_{i} s_i s_{i+1} - h\sum_i^N s_i$$

  • Hamiltonian

{

interaction

{

external field

$$Z=?$$

$$z$$

$$h$$

$$\cdots$$

$$\cdots$$

{

Free Energy

Magnetization

Partition Function

1D

Systems energy

$$\mathcal{H} = -J \sum_{i\neq j} s_i s_j - h\sum_i^N s_i$$

  • Hamiltonian

{

interaction

{

external field

$$Z=?$$

$$z$$

$$h$$

$$\cdots$$

$$\cdots$$

{

Free Energy

Entropy

Magnetization

Partition Function

 $$Z=\sum_{\{s_i\}} e^{\beta J \sum s_i s_{i+1} + \beta h \sum s_i }$$

 $$Z=\sum_{\{s_i\}} e^{\beta J \sum s_i s_{i+1}}$$ 

 $$e^{+ \beta h \sum s_i }$$

No external field! $$h=0$$

Boundary Conditions: Open

$$\{s_1,s_2,s_3,...,s_N\}\to \{s_1,\mu_1,\mu_2,...,\mu_{N-1}\}$$

$$\mu_1 = s_1 s_2,\; \mu_2 = s_2s_3 ,...,\; \mu_{N-1}=s_{N-1}s_N$$

"coordinate transformation"

$$s_i=\pm 1$$

$$\mu_i=\pm 1$$

$$=\sum_{\{s1,...,\mu_N\}} e^{\beta J \sum \mu_i}$$ 

$$=2\prod_{i=1}^{N-1}\sum_{\mu_i=\pm 1} e^{\beta J \mu_i}$$ 

 $$Z=2(2\cosh{\beta J})^{N-1}$$

...

Boundary Conditions: Periodic

$$s_{N+1}=s_1$$

$$s_i=\pm 1$$

 $$Z=\sum_{\{s_i\}} e^{\beta J \sum s_i s_{i+1} + \beta h \sum s_i }$$

 $$T(s_i,s_{i+1})=\exp\left[\beta J  s_i s_{i+1} + \frac{\beta h}{2}  (s_i+s_{i+1})\right] $$

 $$=\sum_{\{s_i\}} \prod_{i=1}^N T(s_i,s_{i+1})$$

T= \begin{bmatrix} \exp{\beta(J+h)} & \exp{-\beta J} \\ \exp{-\beta J} & \exp{\beta(J-h)} \end{bmatrix}

 $$Z=\mathrm{tr}(T)^N$$

transfer matrix

$$\mathrm{tr}(A)=\sum_i A_{i,i}$$

$$\mathrm{tr}(A)^2=\sum_i A_{i,j} A_{j,i}$$

$$\mathrm{tr}(A)^3=\sum_i A_{i,j} A_{j,k} A_{k,i}$$

 $$\mathrm{tr}(A^N)=\sum_{\{s_i\}} \prod_{i=1}^N A(s_i,s_{i+1})$$

...

SYMMETRY

$$U^{-1}TU=D$$

$$U^{-1}=U ^\dagger$$

$$Z=\mathrm{tr}(U^{-1}DU)^N$$

$$=\mathrm{tr}(D)^N$$

$$=\lambda_+^N + \lambda_-^N $$

$$\lambda_\pm = e^{\beta J}\cosh{\beta h} \pm \sqrt{e^{-2\beta J}+ e^{2\beta J} \sinh{\beta h}}$$

$$Z=\lambda_+^N \left[1+\left(\frac{\lambda_-}{\lambda_+}\right)^N \right]$$

Connection with thermodinamics

$$(N \rightarrow \infty)$$

$$=\left[e^{\beta J}\cosh{\beta h} + \sqrt{e^{-2\beta J}+e^{2 \beta J}\sinh ^2}{\beta h}  \right]$$

$$Z=\lambda_+^N  = \lambda_{\text{max}}^N$$

  • Free Energy

$$F = -T\ln{Z}$$

  • Magnetization

$$h \rightarrow 0$$

$$M = -\frac{\partial F}{\partial h}$$

$$M \rightarrow 0$$

$$M = \frac{1}{Z}\sum_{\{s_i \}}\mathcal{M}e^{\beta E}$$

$$\mathcal{M} = \sum_{n=1}^Ns_n$$

$$P(\text{all up}) = \frac{e^{\beta J(N-1)}}{2(\cosh{\beta J}^{N-1} )}$$

$$ E(\text{all up}) = -J\sum_{i\neq j} s_i s_j =-J (N-1)$$

$$P(\text{some state}) = \frac{e^{-\beta E_{\text{of that state}}}}{Z}$$

$$\left(\beta = \frac{1}{k_B T}\right)$$

CORRELATION FUNCTION

$$\langle f_x \rangle = \sum_x f_x P(x)$$

$$\langle s_n s_{n+r}\rangle =\frac{1}{Z} \sum_{\{s_i\}} s_n s_{n+r} e^{-\beta \mathcal{H}}$$

$$\langle s_n s_{n+r}\rangle =\tanh^r(\beta J)$$

$$\langle s_n s_{n+r}\rangle =\tanh^r(J/T)$$

$$=\exp\{r \ln [\tanh(J/T)]\}$$

$$=e^{-\frac{r}{\xi}}$$

$$\xi=-\frac{1}{\ln \tanh{(J/T)}}$$

What about more complex systems?

$$s_i = \pm 1, 0$$

$$s_i$$

?

transfer matrix

  • Potts Model

What about more complex systems?

$$s_i = \pm 1, 0$$

transfer matrix

?

\{

?

  • Curie-Weiss model

What about more complex systems?

$$s_i = \pm 1, 0$$

$$s_i$$

?

2D

transfer matrix

  • Onsager - 1943
  • Sherrington-Kirkpatrick

Phase transition depends on dimensionality

lower critical dimension

(\(h= 0\))

  • \(h \neq 0\)

?

- It does has phase transition!

- Kramers & Wannier

- Ising Spin Glass

$$\mathcal{H} = - \sum_{i, j} J_{i,j} s_i s_{j} - h\sum_i^N s_i$$

$$\mathcal{H} = -J \sum_{i} s_i s_{i+1} - h\sum_i^N s_i$$

  • Difficulty Increases with Dimensions
  • Mean-Field Approximation

What about more complex systems?

Open problem!

quantum mechanics

$$\mathcal{H} = -J \sum_{i} s_i s_{i+1} - h\sum_i^N s_i$$

$$\mathcal{H} = -J \sum_{i} \sigma_i^z \sigma_{i+1}^z $$

Pauli Matrices

Reduces to classical version

$$\mathcal{H} = -J \sum_{i} \sigma_i^z \sigma_{i+1}^z - h\sum_i^N \sigma_i^x$$

\sigma^x= \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}
\sigma^y= \begin{pmatrix} 0 & -i \\ i & 0 \end{pmatrix}
\sigma^z= \begin{pmatrix} 1 & 0 \\ 0 & -1 \end{pmatrix}

$$E = -J \sum_{i} \sigma_i \sigma_{i+1} $$

\{

transverse field

energy levels  \( \sigma_i =\pm 1 \)

diagonal!

\{
  • It has phase transition!

-Ferromagnetic vs Paramagnet

$$ \sigma_i^\alpha = I^{\otimes i-1} \otimes \sigma^\alpha I^{\otimes N-i}  $$

quantum mechanics

$$\mathcal{H} = -J \sum_{i} \sigma_i^z \sigma_{i+1}^z - h\sum_i^N \sigma_i^x$$

$$\mathcal{H} = - \frac{1}{2} \sum_{i}^N \left[ J_x \sigma_i^x \sigma_{i+1}^x +J_y \sigma_{i}^y \sigma_{i+1}^y+  J_z \sigma_i^z \sigma_{i+1}^z - h\sigma_i^z\right]$$

$$\mathcal{H} = - \frac{1}{2} \sum_{i}^N \left[ J_x \sigma_i^x \sigma_{i+1}^x +J_y \sigma_{i}^y \sigma_{i+1}^y+  - h\sigma_i^z\right]$$

$$J_x = \left( \frac{1+ \gamma}{2} \right)$$

$$J_y = \left( \frac{1- \gamma}{2} \right)$$

- Tensor Networks

Quantum Many-Body Problems

  • Heisenberg Model
  • XY Model
  • 1D Quantum Ising Model

magnet

Thank you!

Dziękuję Ci!

Obrigada!

धन्यवाद!

 Teşekkür ederim

An ising model recap

By ludmilaasb

An ising model recap

  • 152