Sie sind auf Seite 1von 21

Entanglement in

Classical Magnets
Samantha Bauman, 31 July 2015
WISE REWU Program 2015
(Women in Science & Engineering; Research Experience for Women
Undergraduates)

Mentor: Dr. Carlos Bolech


Department of Physics
University of Cincinnati

Abstract
This paper reports the application of continuous matrix product states (cMPS)
to evaluate the one-dimensional quantum Ising Model. The purpose of cMPS is to
solve continuous systems in one dimension, which the Ising Model qualifies as. The
Boson and Fermion models have already been evaluated with cMPS; we took one
step further by applying it to the Ising Model. The process occurred in three steps.
First, the Ising Model was evaluated using the exact Hamiltonian equation. Then,
cMPS was recreated for the Boson Model to ensure that it was working properly.
Laslty, cMPS was applied tot the Ising Model itself. Although its development still in
progress, cMPS appears to be effective for the Ising Model.

Acknowledgements
I would like to sincerely thank my mentor, Dr. Carlos Bolech, for involving me
in his research this summer and taking the time to help me to understand the
different concepts involved in the project. I truly appreciate the opportunity that he
gave me to have a real research experience. I would also like to thank the graduate
assistant that I was working with, Sangwoo Chung, who was an immense help in
teaching me about some of the advanced math used.
I would also like to take this time to mention what a wonderful opportunity
the WISE REWU Program was been for me. I just completed my first year in the
engineering program at the University of Cincinnati and currently have four more
years in the program, and being involved in research was an experience I was not
expecting to have so early. The program also taught me some valuable professional
skills, including how to write technical reports and press releases and how to give
presentations. I also learned what to expect in graduate school and what steps I
should take if it is something I would like to pursue. This and other experiences
within the WISE Program has convinced me to continue with the ACCEND

(Accelerated Engineering Degree) Program which will allow me to graduate in four


years with not only my Bachelors degree in Electrical Engineering, but my Master of
Science as well. Most importantly to me, I was able to meet a unique and talented
group of women involved in various scientific and engineering fields that I can now
look at as my peers.
Lastly, I would like to thank the support in funding from the University of
Cincinnati and the Defense Advenced Research Projects Agency Optical Lattice
Emulator (DARPA OLE) program.

Table of Contents
Cover Page 1
Abstract

Acknowledgements

Table of Contents

Outline of Report5
Problem Statement
Introduction

Background

Research Methods

Exact Diagonalization of the Quantum Ising Model 8


The Boson Model 9
The Quantum Ising Model

10

Results and Discussion 11


Exact Diagonalization of the Quantum Ising Model 11
The Boson Model 15
The Quantum Ising Model

15

Conclusion 19
References 20

Outline of Report
Problem Statement
The purpose of the project and a brief statement of the methods are
mentioned.

Introduction
The primary focus is to discuss the motivation behind the research that gives
the reader some background. The two main concepts of the research are
also introduced.

Background

An in-depth explanation of the models and methods used is given, primarily


the Quantum Ising Model and continuous matrix product state (cMPS). The
concepts behind the models are also explained for the reader.

Research Methods

This portion is split into three separate parts. The first part describes the
Ising Model being analyzed using exact diagonalization. The second portion
is concerned with the use of cMPS to evaluate the Boson Model and the last
part describes how cMPS was used with the Ising Model.

Results and Discussion

The results obtained from the different portions of the research are described
and analyzed in detail here. Just as with the Research Methods section, this
portion is split into three parts that correspond to the different Methods that
were used.

Conclusions
A summary of the outcomes of the research as of 31 July, 2015. It also briefly
states what the next steps will be in order to progress in the research.

References

A list of the sources referred to throughout the period of the research and the
writing of this report.

Problem Statement
The purpose of this research is to further develop the analysis method known
as continuous matrix product state (cMPS) in the context of a known model; in this
case, the one dimensional quantum Ising model.

Introduction

From closing a refrigerator door to using a computer, magnetism plays a very

important role in our society. Magnets play an integral part of many pieces of
advanced technologies, some of these include MRI machines in the medical field,
maglev trains in transportation, and in the hard drives of computers. However,
there is still a lot that that can be learned about how magnetism works on an atomic
scale. Unfortunately, there are not many methods that can easily evaluate the
minute details of magnets. Continuous matrix product state (cMPS) is a new
method that can give us more details. By using a model that is already well known,
the Ising Model in this case, cMPS can be further developed within its context with
the hope of its eventual use to analyze models that are more complex than the
Ising.

Background
In materials that can be magnetized, each atom has a magnetic moment,
also referred to as a spin. This is the result of the individual spins of the electron
that compose the atom. In the Quantum case, which is what we are concerned
with, the magnetic moments can only be in one of two directions along the same
axis (e.g. up and down). For most materials, the magnetic moments are not aligned
in parallel but have random orientations. This is a case of antiferromagnetism and
is what occurs in materials that can be magnetized but are not. The magnetic
torques from each atom cancel out with each other (e.g. an up spin will cancel out
with a down spin) and leave the material with no macroscopic magnetic force.

However, there is also the case of ferromagnetism in which the atoms align in the
same direction. Then the individual magnetic torques do add up with each other to
create a macroscopic magnetic force that can be felt, thus giving magnets.
The Ising Model that is being used to develop cMPS has three main
parameters: it is one dimensional, quantum, and in the presence of a transverse
magnetic field. Looking at a one dimensional model, atoms on one side of the
model cannot have direct interactions with each other. The Ising Model investigates
how the interactions of forces between local atoms influence a macroscopically
large amount of atoms to spin in alignment with each other, creating a case of
ferromagnetism. In this case, the model is also in the presence of a transverse
magnetic field, meaning that the field is oriented perpendicular to the axes of the
atomic spins. As the strength of the transverse field increases, the spins are more
likely to align in the same direction of the field.
Due to the size of the matrices involved in the Hamiltonian equation that
describes the Ising Model, it is difficult evaluate the Hamiltonian equation for an
amount of atoms greater than 20. Since even small amounts of material can have
much more than 1023 atoms, 20 is a very low number and so is largely inaccurate
for determining minute details about systems such as the Ising Model. One of the
ways to solve this is to use matrix product state (MPS). MPS is a method of analysis
that has been found to accurately determine the energies of systems similar to and
including the Ising Model when electrons in those systems are in their ground state
or low energy states. The goal of this project was to develop continuous matrix
product state (cMPS), a method very similar to MPS. The two methods are similar,
but MPS is used to investigate models of solids while cMPS is used for fluid and
gaseous models. The primary difference in these is that models that cMPS will
analyze will contain atoms that are not evenly distributed and that are constantly in

motion. Although this is not true for the Ising Model, it was chosen because it has
already been well analyzed using MPS. The data that was collected for it through
MPS will allow us to have known values to compare with results from cMPS. Since
cMPS is still relatively new, it is important that we have a strong basis to compare it
with so that it can be developed further. The eventual goal of this research is that
the foundation for cMPS will be strong enough for its use with other, more complex
models that we do not yet know much about.

Research Methods
Exact Diagonalization of the Quantum Ising Model
Before developing cMPS within the context of the Ising Model, it was
necessary to determine what was to be expected. MATLAB was used to
evaluate the Hamiltonian function that describes the energies of the one
dimensional Ising Model in an exact diagonalization process. The basis of the
Hamiltonian function are the two matrices, x and z, which are the spin
configurations that represent the magnetic moments of the atoms and the
transverse magnetic field, respectively.
N1

H ( h )=J
i

x
i

x
i+1

( J N 1 )h i

(1)

J represents the local forces between the atoms and N represents the
number of atoms that are being calculated for in the equations. The function
represents a circular chain of single atoms that are in a magnetic field that is
perpendicular to the axes of the spins of the atoms. The ground state energy
was evaluated as a function of the strength of the transverse magnetic field
(h). Due to the size of the matrices being evaluated, only 12 atoms were able
to be calculated without exceeding the memory.
The order parameters were also calculated for x and z in order to
reproduce the transitions that are found when h/J = 1. After initial trials, it

was found that the Hamiltonian function required the addition of a small,
almost negligible, magnetic field in order to stabilize the results of the order
parameter calculations. This magnetic field is represented by the term

h i
'

in Equation 2.
N1

H ( h )=J
i

x
i

x
i+1

( J N 1 )h h xi
i

z
i

'

(2)

Using the eigenvectors representing the ground state energies that are
found with Equation 2, the order parameters were calculated for x and z
using Equations 3 and 4 respectively.
N

ix >
i

z
i

>
i

e gs ( ixe gs )
e gs e gs

(3)

e gs ( zi e gs )
e gs e gs

(4)

The Boson Model


In order to ensure that the computations involved in cMPS will work
accurately when applied to the Ising Model, the code was initially recreated
for the Boson Model. As cMPS has already been developed with the Boson
Model, it allowed an accurate set of data for comparison in order to
determine that the code was working properly. The code consisted of three
main files: Main.cpp, cMPS.cpp, and nrutil.cpp. The nrutil.cpp file was created
using a simulated annealing method which minimizes multidimensional
functions.1 Main.cpp connected the three files and served as the basis for
running the program. It set the parameters for the functions used in both
cMPS.cpp and Main.cpp. cMPS.cpp contained all the computations that were

exclusive to the Boson Model as well as some functions that allowed for ease
of reading and usage by those who would edit the program.
All of the calculations for continuous matrix product state were
contained in the cMPS.cpp file. cMPS begins with two matrices that are then
used in various calculations to find the energy of the system. The values
contained within these matrices are run through the simulated annealing
functions in order to try to minimize the final output of the matrix
calculations, in this case, energy. The matrices that were used for the Boson
Model were B, a simple DxD diagonal matrix, and R, a complex DxD matrix,
where D was a parameter determined by the programmer and varied from 2
to 8. The energy of the system was given by the addition of ke and ie, which
are represented by the Equations 5 and 6.

, R ] ) ]
ke=Tr [ e TL ( [ Q , R ] [ Q
R ) ]
ie=c Tr [ e TL ( RR R

(5)
(6)

The energy was found as a function of varying c from 0 to 100. The


variable L represents the length of the material being looked at and Q and T
were given by Equations 7 and 8 respectively.

1
Q=iB R R
2

(7)

R R

T =Q I + I Q+

(8)

Once results from the code matched the known values from Boson
Model, confirming the accuracy of the program, the cMPS.cpp file could be
modified for the Ising Model.

The Quantum Ising Model


The code files that were used for the Ising Model had the same set up
as those that were used to replicate the Boson Model analysis. The three
main files were Main.cpp, cMPS.cpp, and nrutil.cpp. The main calculations

10

that were used to evaluate the energies of the Ising Model were stored in the
cMPS.cpp file.
Just as in the Boson Model, the computations began with two main
matrices, A and P. The values in A and P were determined randomly and then
run through the simulated annealing functions in order to find the lowest
output of the calculations. Initially, the matrix P was used to compose a
matrix R that was then used in the calculations, but after several trials, it was
decided that the code worked more efficiently when R was set to a nonchanging matrix described by Equation 9 where c is a positive real number,

ID/2 is a

D D
x
2 2

identity matrix, and S+ =

[ ]

0 1 .
0 0

+
I D S

(9)

R=c
The final energy is the sum of two terms:

c 0 E 0c 1 E 1 . E0 and E1 are

described by Equations 10 and 11 respectively, where T is the same one that


is described in Equation 8.

]
E0=Tr[ e TL(R R)

(10)

~
aQ

E1=Tr[e TL( R I + I R )( e I D )( R I + I R ) ]
2

(11)

The terms E0 and E1 are multiplied by c0 and c1 where c1 = J, the

magnitude of the magnetic forces between atoms and


g is the magnitude of the transverse magnetic field.

11

c 0=2 J(1g )

where

12

Results and Discussion


Exact Diagonalization of the Quantum Ising Model
-0.8

Ground State Energy as a Function of h

-1
-1.2

Energy

-1.4
-1.6
-1.8
-2
-2.2
0

Exact Diagonalization: N = 12
N=4
Analytic Formula
0.5

1.5

Figure 1: The ground state energy of the system is plotted as a function of the strength of the
transverse field. The Hamiltonian equation was evaluated for both 4 and 12 atomic spins and
plotted with the exact values from an analytic formula.

In the above graph, the ground state energies from the Hamiltonian
equation give values that are slightly less than those that are found using an
analytic formula. It is also very distinct that the values from only four atomic
spins had a larger margin of error than the values from twelve atomic spins.

13

< x> vs. h

Order Parameter (< x>)

<x> (N = 4)
< > (N = 8)

0.8

<x> (N = 12)

0.6
0.4
0.2
0
0

0.5

Strength of Transverse Magnetic Field (h)

1.5

Figure 2: The order parameter for x is evaluated as a function of the strength of the
transverse magnetic field. From left to right, they represent models in which there were 4, 8,
and 12 atoms.

As in the case of the energy (Figure 1), 12 atoms was the largest
amount that were able to be analyzed using the Hamiltonian equation
without running out of memory. With a significantly larger amount of atoms,
the order parameter of x makes a very drastic change from about 1 to
exactly 0 at h = 1. Not enough atoms were in the model to show this change,
but the transition point does become significantly closer to 1 as the number
of atoms increases. Although there was only a small difference between the
exact results and the results found using a small amount of atoms in Figure 1,
Figure 2 shows that a large amount of atoms are required to understand
certain details of the Ising Model.

14

vs. h
z

1
0.9
0.8

N=4
N=8
N = 12

< z >

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

0.5

1.5

2.5

Figure 3: The order parameter for z is evaluated as a function of the strength of the
transverse magnetic field, similar to Figure 2. The same number of atoms that were used in
Figure 2 for the order parameter of x were also used here: 4, 8, and 12.

Similar to the order parameter of x, the order parameter of z also


shows a transition around h = 1. Again, the transition becomes more
apparent as the number of atoms used increases. This further supports the
need for developing a model that can give accurate values for a large amount
of atoms.

15

The Boson Model

Energy vs c

3.5

Energy

2.5

2
Exact Points
D=2
D=4
D=6
D=8

1.5

0.5
0

10

20

30

40

50

60

70

80

90

100

Figure 4: The energy of the Boson Model was evaluated as a function of c, the interaction
strength between atoms. The computations were run for matrices of size D x D where D = 2,
4, 6, and 8 and compared with the exact values of the model.

The values found using the developed code were within an acceptable
range of the exact values, demonstrating its accuracy for matrices where D
was sufficiently large (roughly D = 4). The negligible error confirmed the
proficiency of the code for cMPS and signified that it was ready to by applied
to the Ising Model.

The Quantum Ising Model


Results were obtained for the cMPS computation of the Ising Model that
corresponds to Figures 1, 2, and 3. The research is still undergoing, so the
following results are only preliminary results.

16

Ground State Energy vs. g


-1
-1.5
Energy

-2 D=4

Exact

-2.5
-3
0

0.5

1.5

2.5

Strength of Transverse Magnetic Field (g)

Figure 5: The energy of the system was evaluated for varying strengths of the transverse
magnetic field. These values were obtained using cMPS.

The values for the energy are almost exactly what was expected. The values
also appear to stay consistent as the transverse magnetic field increases.
While running the code, the energy value is the final output that we are trying
to minimize, so all the work was primarily concerned with first matching up
calculated energy with the expected values.

17

<z> vs. g
0
-0.2
-0.4

Order Parameter
D=4 (<z>)

Exact
-0.6
-0.8
-1
0 0.5 1 1.5 2 2.5

Strength of Transverse Magnetic Field (g)

Figure 6: Using the results obtained from evaluating the Ising Model using cMPS, the order
parameter of z was calculated. The value z was calculated for each energy obtained and
plotted as a function of the strength of the transverse magnetic field.

The order parameters of x and z were more often than not trickier to find
accurate values of compared to the energy values. The value of z show a
gradual transition from 0 to -1 when g = 1. This transition was captured by
Figure 6.

18

<x> vs. g
1.0
0.8
0.6
D=4
Order Parameter (<x>)

0.4
0.2
0.0
0 0.5 1 1.5 2 2.5

Strength of Transverse Magnetic Field (g)


Figure 7: Just as in Figure 6, the order parameter of x was calculated using the results
obtained from evaluating the Ising Model with cMPS. The value x was plotted as a function of
the strength of the transverse magnetic field.

The value that we have been the most concerned about has actually been x.
According to previous findings, the order parameter makes a sharp transition
from 1 to 0 when the transverse magnetic field is equivalent to 1. That trend
is almost perfectly displayed here. The order parameter of x is actually what
we have been struggling the most with to minimize correctly. These latest
results are a great improvement compared to what we have been previously
getting, suggesting that we should be able to accurately evaluate the Ising
Model using cMPS as predicted.

19

Conclusion
Although cMPS is still being optimized for the Ising Model, we are coming very
close to the final results. Figures 5 and 7 are the best examples of this. Figure 7
especially allowed us to see that cMPS can give us accurate information about the
Ising Model where exact diagonalization cannot due to its inability to evaluate a
large amount of atoms. The next steps to take are going to be primarily geared
toward fine tuning the coding. A main component of cMPS involves begin the
program at one value and allowing it to run until the value is minimized as much as
possible. Something that may be wrong is simply that the program is not
minimizing the value as much as it can, so some more work there may be
necessary. Other changes to cMPS may include playing with the initial matrices that
are provided, particularly matrices A and P that were mentioned in the Quantum
Ising Model section of the Research Methods. All in all, the results that we are
seeing from cMPS are looking very promising. The limits of cMPS are only beginning
to be tested. The work that has been done and is still being done with the Ising
Model is just another step to finding out new, pivotal information about other
complex models.

20

References
Byron, Frederick W., and Robert W. Fuller. Mathematics of Classical and Quantum Physics.
Reading,
Massachusetts: Addison-Wesley Publishing Company, 1969. Print.
F. Verstraete and J. I. Cirac. Continuous Matrix Mroduct States for Quantum Fields. Physical
Review Letters 104,
190405 (2010).
Fitzpatrick, Richard. The Ising Model. 29 March 2009. Web. <
http://farside.ph.utexas.edu/teaching/329/
lectures/node110.html>
Griffiths, David J. Introduction to Quantum Mechanics. Upper Saddle River, New Jersey:
Prentice Hall, Inc., 1995.
Print.
Leon, Steven J. Linear Algebra With Applications, 7th Edition. Upper Saddle River, New
Jersey: Prentice Hall, Inc.
2006. Print.
Magic of Magnetism. Northeastern University. Web. <
http://www.northeastern.edu/sunlab/mom/
technology.html>
Sangwoo S. Chung, Kuei Sun, and C. J. Bolech. Matrix product ansatz for Fermi fields in one
dimension. Physical
Review B 91, 121108 (2015).
Vid Stojevic, Jutho Haegeman, I. P. McCulloch, Luca Tagliacozzo, and Frank Verstraete.
Conformal data from finite
entanglement scaling. Physical Review B 91, 035120 (2015).

21

Das könnte Ihnen auch gefallen