Sunteți pe pagina 1din 19

Candidate Number: 419422

Project Number: A&L08


Project Title: DNA as quantum measurement apparatus
Supervisor: Vlatko Vedral
Word Count: 8423 (without appendices), 12040 (with appendices)
1
Abstract
It has been proposed that the proton constituting the hydrogen bond between the two
strands of the DNA double helix could exhibit coherent quantum dynamics on a biolog-
ical timescale. In this report, I present and investigate a Hamiltonian which describes
the proton as a two state quantum system energetically coupled to a fermionic bath of
environmental electrons, as found in DNA. Firstly, I apply a polaron transformation to
examine the nature of the proton-electron coupling. I then use an open quantum systems
approach to computationally model the behaviour of the two-site system, considering and
comparing both Markovian and non-Markovian dynamics. I nd that the thermalising
behaviour of the two state system depends sensitively on the relative conguration of the
chemical potential of the bath and the resonant energies of proton transfer.
1. INTRODUCTION
In Darwins theory, you just have
to substitute mutations for slight
accidental variations (just as quan-
tum theory substitutes quantum
jump for continuous transfer of en-
ergy).
Erwin Schrodinger, What is
Life?, 1944.
Historically, the many mechanisms and man-
ifestations of life have been a subject of wonder
and speculation for professional physicists, seem-
ing to defy many of the established thermody-
namic principles about the physical universe[1].
We have moved beyond a nave vitalism, but biol-
ogy has still proved a fertile eld of study for those
interested in complexity and out-of-equilibrium
thermodynamic systems. There has also been a
good deal of recent interest in looking for explicit
manifestations of quantum mechanics in biolog-
ical systems, with advances made in our under-
standing of the atomic-scale mechanisms under-
lying photosynthesis[24] and avian navigation[5
7]. It is also tempting for a quantum information
scientist to look to biology for the possibility that
the process of evolution has developed a setting
for coherent quantum dynamics which are con-
trollable on a macroscopic biological timescale[8].
Previous study has been done in the context of
DNA[9]: Schrodinger, in his publication What is
Life? explored the relationship between the fun-
damental discreteness of both Darwinian muta-
tions and quantum mechanics[1]. In this work,
he argued that there must be a discrete physi-
cal mechanism encoding the mutations of evolu-
tionary theory, and that the discreteness of en-
ergy levels in quantum mechanics could provide
this function. It has long been understood that
the proton of the cross-strand hydrogen bond can
be modelled by an energetically weighted two-
site Hamiltonian[10]. This system allows for the
proton to quantum tunnel through the ener-
getic barrier between the sites. It is thought that
a single tunnelling event could instigate a bio-
chemical change, bringing about a macroscopic
base pair mutation. This is a signicant genetic
event; it is exciting that explicit quantum me-
chanics could contribute to the rate at which this
happens. Further, it has been commented that
this close relationship between quantum mechan-
ical phenomena and a macroscopically measur-
able event in the context of DNA provides an
ideal setting to discuss more abstract interpreta-
tive problems of quantum mechanics associated
with decoherence[11].
Previous work has investigated the extent
to which interstrand proton dynamics aect
longitudinal electronic properties, such as the
conductance[12], but in this project I ask the op-
posite question: to what extent do the electrons
aect proton tunnelling dynamics, bringing about
genetic mutation and collapsing any coherent su-
perposition of the hydrogen bond state? I use a
similar Hamiltonian to [12], with site energies and
tunnelling amplitudes associated with the proton,
a contribution of self-energies from the electronic
modes and a number-displacement coupling term.
Having introduced the relevant background on
quantum mechanics in section 2, I go on in sec-
tion 3 to apply a polaron-type Lang-Firsov uni-
tary transformation to this Hamiltonian with rea-
soning motivated by analogy to the extensively
studied spin-boson model[13]. The ambition is
to move into a new reference frame in which the
coupling term is recast, with coupling dynamics
being absorbed into a new renormalised coherent
hopping rate, allowing for further analytical work.
Following this, in section 4, I use the theory
of open quantum systems[14] to pursue a compu-
tational approach to understanding the dynam-
2
ics. Having dened and considered the density
matrix of the two-site system and bath together,
I nd the von Neumann equation for the time
evolution of the density matrix directly from the
Schrodinger equation. I then derive a master
equation for the two-site system and then per-
form, present and analyse computational simula-
tions of the dynamics of the system. During this
derivation, I make certain assumptions about the
nature of the coupling between the proton and
the bath, among which are the Born approxima-
tion, that of weak coupling between the system
and bath, and the Markov approximation, that
the dynamics of the system density matrix are
strictly local in time.
In section 5, I then use the theory of time con-
volutionless projection operators[14] to nd the
leading order non-Markovian contributions to the
dynamics, for which historical congurations of
the system aect the future dynamics of the sys-
tem through excitations of the environment. I
perform and present computational simulations
of the proton dynamics in the non-Markovian
case and make a comparison to the Markovian
dynamics. I hope to analyse the dynamics of
a spin-fermion system and gain understanding
of whether sustained quantum mechanical eects
can occur in the hydrogen bond in DNA. Conclu-
sions are presented in section 6.
2. BACKGROUND
The contents of this section can be found in
any standard reference book on quantum infor-
mation, such as [15].
The quantum state of a system can be de-
scribed as a wavevector: a rank 1 object in
an n-dimensional Hilbert space, often written in
Diracs bra-ket notation as | . There exists a
corresponding dual vector (|)

= |. These
bra and ket wavevectors can be expressed as
explicit column or row vectors upon choice of a
physical basis. These being objects from abstract
linear algebra, it is possible to dene products of
these, such as the outer product:
| | =
_

2
_
_

2
_
=
_

1

1

1

2

2
_
.
(1)
This is an object of rank 2 and dimension m-
by-n. Similarly, there exists the tensor product,
namely:
| | = | =
_

2
_

2
_
=
_
_
_
_

2
_
_
_
_
(2)
This is an object of rank 1 and dimension m
n = 4.
For an ensemble of quantum objects, each with
a well dened individual wavevector |
i
, it is pos-
sible to a dene a tensor product of each of their
ket wavevectors and hence create a wavevector for
the entire system,
| =

i
|
i
(3)
An ensemble in which each object has a well-
dened state is known as a pure state, but a
physical theory must be able to deal mathemati-
cally with statistical ensembles of these. For this,
it is possible to use the density matrix , a rank
2 object of dimension n
2
, dened for a statistical
mixture of pure states as:
=

j
p
j
|
j

j
| . (4)
This is simply the sum over the outer products
of each bra state vector with each ket state vector,
weighted according to their statistical presence.
This is a powerful tool for describing the state of
a general quantum ensemble.
The diagonal elements of the density matrix
correspond to the probabilities of nding the sys-
tem in each state; the o-diagonal elements are
only present in superposition states, and are a
measure of the extent to which the system devi-
ates from a classical probability distribution. For
a physical system, the probabilities for nding the
system in any state should be normalised to one.
Hence the density matrix has the mathematical
property that tr() = 1, independent of basis. It
is also notable that the density matrix is always
Hermitian and complex.
One further useful property of the density ma-
trix is its relation to physical expectation values;
the trace of the product of an operator and the
density matrix will return the expectation value
of the operator:
tr(A) =

i
p
i

i
| A|
i
= A . (5)
3
It is possible to dene a density matrix asso-
ciated with any component of the system, which
can be extracted by an operation known as the
partial trace. If there exists a basis |j
B
that spans
a subspace B, then a partial trace over B is de-
ned as:
tr
B
() =
d

j=1
j
B
| |j
B
=
S
. (6)
This produces a lower dimension density ma-
trix associated with the remainder of the total
system, in this case labelled S. This operation will
prove useful in the theory of open quantum sys-
tems, when it becomes necessary to dissociate the
density matrix of the two-site system from that
of the environment: I will have to trace out the
density matrix of the environment over a basis
which spans the possible states of the bath.
3. THE HAMILTONIAN AND THE
POLARON MODEL
3.1. The Hamiltonian
I will present the model Hamiltonian in terms
of fermionic creation and annihilation operators
of the proton and the electron states; mathemat-
ical operators which respectively increase or de-
crease the number of particles in their associated
state by one. They obey the fermionic canonical
anticommutation relations, given by:
_
d
i
, d
j
_
=
_
d

i
, d

j
_
= 0,
_
d

i
, d
j
_
=
ij
. (7)
These relations imply physical results consis-
tent with Paulis exclusion principle: fermions
cannot share the same quantum numbers, so the
consecutive application of either operator to the
same state returns zero. Similarly, the product of
operators d

j
d
j
for a specic state j returns the
mutually commuting set of number operators n
j
,
with eigenvalues zero or one.
The Hamiltonian I studied is as follows:
H =

i=1,2
E
i
c

i
c
i
+

i=1,2

k
i
(
i,k
i
)d

i,k
i
d
i,k
i
+V (c

1
c
2
+ c

2
c
1
) +

i=1,2

k
i
g
i,k
i
c

i
c
i
(d
i,k
i
+ d

i,k
i
).
(8)
I have used c
i
as the operator associated with
a proton at a site i, and d
i,k
i
as the operator for
an electron at site i in mode k
i
. The summations
here are over i, the proton sites, and k
i
, the elec-
tronic modes. There are four energetic contribu-
tions to the dynamics to consider: the self energy
E
i
associated with each proton site, the self en-
ergy
i,k
i
associated with each electronic mode at
each site, a tunnelling energy V for the proton
to hop from one site to the next and a coupling
term, proportional to a coupling energy g
i,k
i
, the
occupancy of a given proton site and the displace-
ment of the electrons from that site, intended to
model the Coulomb attractive force. There is also
, a constant chemical potential which maintains
conservation of electron number in the bath.
3.2. The Polaron Transformation
Hamiltonians of this form have been studied
thoroughly in the case of a two-state system in-
teracting with a massless bosonic bath[13], but
there is relatively little in the literature investigat-
ing the dynamics in a fermionic environment[16].
With reasoning by analogy to the bosonic case,
I applied a polaron transformation; a unitary
change of reference frame that displaces the elec-
trons at a site on the condition that a proton is
present. This may allow a rephrasing of the prob-
lem in terms of a quasiparticle, the polaron,
a coupling of charge to mechanical displacement
through the polarisation of the medium.
This transformation has an associated unitary
operator; the objective is to generate an electronic
displacement on the condition of the presence of
a proton, so the generator must be proportional
to the the number operator of the proton and
the momentum operator of the electron. I conse-
quently used the following unitary operator with
an associated transformation energy
i,k
i
:
U = exp
_

i=1,2

k
i

i,k
i
c

i
c
i
(d

i,k
i
d
i,k
i
)

. (9)
I applied this to each component of the Hamil-
tonian, making use of the Hadamard Lemma and
the fermionic anticommutation relations (see Ap-
pendix A) to derive the complete transformed
Hamiltonian, presented here with general trans-
formation energy
i,k
i
and associated denitions:
4

H =UHU

i
E
i
c

i
c
i
+ V ( c
1

c
2
+ c
2

c
1
) +

i,k
i

i,k
i
d

i,k
i
d
i,k
i
+

i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)

_
g
i,k
i
n
i
cos

+
i,k
i

i,k
i
sin

j,k
j
2
j,k
j
n
j
(d
j,k
j
d

j,k
j
)[
i,k
i

i,k
i
(1 cos

g
i,k
i
sin

]
_
,
(10)
c
j
= c
j
exp
_

k
j

j,k
j
(d
j,k
j
d

j,k
j
)
_
, (11)
where c
j
is the transformed proton operator,
=

l,k
l
4
2
l,k
l
n
l
, and n
j
= c

j
c
j
is the number
operator for proton site j.
1
Having derived periodic commutation rela-
tions for the fermionic terms and constructed the
transformed Hamiltonian, I found that the cou-
pling term is not absorbed into the tunneling term
as in the spin-boson case, but instead remains
with a dependance on the transformation energy.
It can be seen from equation (10) that no choice
of transformation energy will eliminate this term.
This approach could nonetheless be of use. In-
troducing the polaron transformation makes it
possible to tune the distribution of free energy
between the coupling term and the tunnelling
term by the choice of
i,k
i
[17]. Further analytical
progress would use perturbation theory to anal-
yse the Hamiltonian, with the renormalised tun-
nelling rate as the small parameter. The clear
objective is to make the remaining coupling term
as small as possible, so as to minimise the size
of the eective coupling and allow further consid-
eration of the dynamics of the remaining compo-
nents.
4. MARKOVIAN OPEN QUANTUM
SYSTEMS
4.1. Background
In this section, I derive the results necessary
for a master equation driven analysis of this sys-
tem, following [14], chapter 3.
1
It is worth noting that the appearance of in the denom-
inators of terms in (10) does not mean that the expres-
sion features inverse number operators. The terms are
dened as power series, and an expansion of the numer-
ator will reveal that only positive powers of n
i
appear.
It is much more convenient to carry out this
analysis in the interaction picture: For a system
with a solvable Hamiltonian H
0
representing the
energetic contribution of the uncoupled systems,
and a Hamiltonian H
I
(t) representing the cou-
pling dynamics, there exists a total Hamiltonian
H:
H = H
0
+ H
I
(t). (12)
By introducing the unitary time evolution op-
erators U
0
(t), U
I
(t) and U(t) as the propagators
from t
0
to t of the Hamiltonians H
0
, H
I
(t) and H,
respectively, it is possible to use U
0
(t) to trans-
form all of the quantum mechanical objects from
the Schrodinger picture into interaction space
in which the operators carry the dynamics of the
unperturbed system:
A
I
(t) = U

0
(t)A
S
(t)U
0
(t), (13)
|
I
(t) = U

0
(t) |
S
(t) = U
I
(t) |
S
(t
0
) , (14)
i
d
dt
U
I
(t) = H
I
(t)U
I
(t). (15)
By substituting this result into the
Schrodinger equation it is now clear that
the wavevector dynamics now only depend
on the interaction Hamiltonian. It is possible
to construct the density matrix as dened in
equation (4) and take a time derivative to give:

I
(t) = U
I
(t)
S
(t
0
)U

I
(t), (16)
d
dt

I
(t) = i
_
H
I
(t),
I
(t)

. (17)
This equation of motion for the density matrix
is known as the von Neumann equation, and has
equivalent integral form:

I
(t) =
I
(0) i
_
t
0
ds
_
H
I
(s),
I
(s)

. (18)
5
Assuming that the bath starts in a steady
state, it is now possible to insert the integral state-
ment into equation (17) and take the trace over
the reservoir to nd:
d
dt

S
(t) =
_
t
0
ds tr
B
_
H
I
(t),
_
H
I
(s), (s)

.
(19)
At this point it is necessary to make the rst
approximation: that the coupling is weak and so
the interaction has negligible eect on the envi-
ronment. This Born approximation means that
the density matrix can be separated into a tensor
product, such that tr
B
((t)) tr
B
(
S
(t)
B
) =

S
(t), to give:
d
dt

S
(t) =
_
t
0
ds
_
H
I
(t),
_
H
I
(s),
S
(s)

. (20)
This is now a integro-dierential equation for
(t), for which its present dynamics depends on all
previous congurations in time. To simplify this,
it is possible to make the Markov approximation,
rst replacing the integrand (s) by the current
state (t). The second part of the approximation
is to substitute s with t s and allow the upper
limit on the integral to go to innity. This is ac-
ceptable if the integrand disappears for s
B
,
the bath correlation time. Physically, this means
that there is a coarse-grained time resolution on
the analysis, in which the systems dynamics are
resolved but bath excitation is not. This leaves
a fully Markovian quantum master equation, for
which only the present state of the system deter-
mines the dynamics:
d
dt

S
(t) =
_

0
ds
_
H
I
(t),
_
H
I
(t s),
S
(t)

.
(21)
In order to apply this to specic systems, the
Schrodinger picture coupling Hamiltonian must
be decomposed into contributions from the sys-
tem and the environment, denoted by A

and B

,
respectively:
H
I
=

. (22)
The system contribution A

is now decom-
posed by a projection onto the resonances =

of the system, with and

each being the


eigenenergies of the two state system. This pro-
jection gives the contributions A

(), which share


an eigenbasis with the system:
A

()

=
()A

), (23)
where () = | | is a projection operator
onto the subspace associated with system eigenen-
ergy . As a consequence of this decomposition,
each projected coupling contribution will simply
pick up a scalar complex exponential when time-
evolving in the interaction picture. It is now pos-
sible to reconstruct a coupling Hamiltonian in the
interaction picture in a convenient form:
H
I
(t) =

,
e
it
A

() B

(t), (24)
where B

(t) = U

0
(t)B

U
0
(t) is the coupling con-
tribution from the bath in the interaction picture.
The quantum master equation can now be con-
structed using the interaction picture coupling
Hamiltonian above and equation (21):
d
dt

S
(t) =

,
e
i(

)t

()

_
A

()
S
(t)A

) A

)A

()
S
(t)
+ hermitian conjugates
_
, (25)

()
_

0
dse
is
B

(t)B

(t s), (26)
B

(t)B

(t s) tr
B
[B

(t)B

(t s)
B
].
(27)
Equations (27) and (26) are the correlation
functions of the bath and the one-sided Fourier
Tranform thereof, respectively.
There is one further approximation to make:
the secular approximation, related to the rotat-
ing wave approximation of quantum optics. Es-
sentially, the assumption is that the terms oscil-
lating with frequency |

| will not be resolved


on the timescale of the relaxation term, such that
only the stationary terms for which =

will
contribute dynamically. This leaves:
d
dt

S
(t) =

()
_
A

()
S
(t)A

)
A

)A

()
S
(t) + h.c.
_
. (28)
6

() can be written in terms of real and


imaginary parts, giving a rewritten master equa-
tion with associated denitions:
d
dt

S
(t) =i[H
LS
,
S
(t)] +D(
S
(t)), (29)

() =
1
2

() + iS

(), (30)
H
LS
=

,
S

()A

()A

(), (31)
D(
S
) =

()
_
A

()
S
A

()

1
2
{A

()A

(),
S
}
_
. (32)
The structure of Equation (29) gives some
physical insight: H
LS
, the Lamb shift Hamil-
tonian, is proportional to the real part of the
Fourier Transform of the correlation function and
contributes the unitary time evolution associated
with the coupling-induced renormalisation of the
system energies. D(
S
), the dissipator, is propor-
tional to the imaginary part of the Fourier Trans-
form of the bath correlation function and leads to
decoherent behaviour.
4.2. Markovian DNA
This analysis can now be applied to the Hamil-
tonian of equation (8). This Born-Markov anal-
ysis includes assumptions of both weak coupling
and a memory-less bath. Neither of these are def-
initely physically justiable: the electron-proton
Coulomb repulsion will not always be weak, and
the bath is likely to be small enough such that his-
torical congurations of the two-site system will
have a signicant inuence on future dynamics.
However, I am interested in pursuing a master-
equation driven analysis of this system with little
previous work to develop, so it is reasonable to
start with a possibly restrictive model and then
rene any assumptions later.
It is a familiar exercise to solve the two state
unperturbed Hamiltonian to nd its eigenergies
and eigenstates in the site basis:
E

=
1
2
_
(E)
2
+ V
2
, (33)
| =
1
_
2E
2

+ E

E
_
V/2
E/2 + E

_
, (34)
where E = E
2
E
1
, the dierence in proton
site energies. The system and bath contribu-
tions to the coupling Hamiltonian are seen, in the
Schrodinger picture, to be:
A
i
= c

i
c
i
, B
i
=

k
i
g
i,k
i
(d
i,k
i
+ d

i,k
i
). (35)
I constructed the objects dened in equation
(23) by projecting the system contributions of the
coupling Hamiltonian onto the energy dierences

:
0
, 0,
0
, where
0
= E
+
E

is the res-
onant energy of the proton. I also transformed
the bath coupling contributions into the interac-
tion picture:
A
i
(
0
) =| | | A
i
| , (36)
A
i
(0) =

=
| | | A
i
| , (37)
B
i
(t) =

k
i
g
i,k
i
(e
i
i
t
d
i
+ e
i
i
t
d

i
). (38)
Then, with the interaction picture contribu-
tions B
I
(t) and the assumption that the
B
is
a stationary state of the bath Hamiltonian, and
hence that the correlation functions are station-
ary in time, I found the correlation function:
B

i
(s)B
j
(0) =

k
j

ij

k
i
k
j
g
2
j,k
j
[n(
j
)e
i
j
t
+ n(
j
)e
i
j
t
], (39)
with
j
=
j,k
j
, using d

i,k
i
d
i,k
i
= n(
i,k
i

), the Fermi-Dirac distribution. This is often


expressed as n(x) = 1/[1 + exp(x)], where =
(k
b
T)
1
, and has the property 1 n(x) = n(x).
I now assume a continuum limit, replacing
the sum over discrete mode couplings with a fre-
quency integral over a general spectral density
J(f):

k
i
g
2
i,k
i

_

0
dfJ(f)(f
i,k
i
). (40)
It is worth noting that by making the environ-
mental bath a continuous distribution of states,
the chemical potential will now always also be
the Fermi energy associated with the Fermi-Dirac
distribution. From hereon the two are synony-
mous. Taking the Fourier transform to nd
ij
as dened in equation (26), I found:
7

ij
() = [J() +J(+)]n()
ij

k
i
k
j
. (41)
It is clear that the bath correlation function
processes frequencies relative to the chemical po-
tential . The contributing factors to the dy-
namics are the occupancy of the resonant states,
which lie from the Fermi level, and their cou-
pling strengths J(). Transitions on resonance
with a frequency
0
can stimulate transitions ei-
ther by exciting an electron from the Fermi level
to the state with energy +
0
or exciting a hole
from the state with energy
0
to the Fermi
level. Both of these transitions have the occu-
pancy factor n(
0
); this is a result of the dynam-
ical equivalence of electrons and holes.
There is no imaginary part to
ij
: it is com-
posed of real functions and takes real arguments.
Consequently, the Lamb-shift Hamiltonian H
LS
of
equation (31) is zero, so there is only dissipation
present in the interaction picture master equa-
tion.
I then simulated the time evolution of the den-
sity matrix in this model, using Matlabs quan-
tum optics toolbox [18]. I used the widely used
Drude-Lorentz spectral density[19] to describe the
electron-proton coupling, which gives physically
reasonable results, although further work can be
done in investigating the precise nature of cou-
pling in the context of DNA. This spectral den-
sity has a characteristic coupling strength and
a spectral peak at = :
J() =

2
+
2
. (42)
4.3. Data and Analysis
My choice of physical parameters for the sim-
ulations was determined by an order of magni-
tude comparison with the spectral lines of liquid
water; I decided that having physical energies of
order 1000cm
1
would provide reasonably phys-
ical results[20]. The only parameter not of this
order is the coupling magnitude = 1cm
1
. This
was to ensure the validity of the Born approxima-
tion in the simulation; I do not claim that this is
necessarily physical.
The generated data show clear dissipative dy-
namics; in Figure 1 the state approaches a clas-
sical thermal equilibrium distribution over diag-
onal elements of the density matrix. The o-
0 1 2 3 4 5
x 10
8
0
0.2
0.4
0.6
0.8
1
Time (s)
O
c
c
u
p
a
n
c
y
FIG. 1: These are the absolute values of the density
matrix in the Schrodinger picture and site basis plotted
against time, green being
11
, red
22
and blue |
12
| =
|
21
|. The proton was initially in the more energetic
site 2, ie with
22
= 1, with T = 77K, with =
8000/cm,
0
= 5099/cm, = 1/cm and = 1000/cm.
diagonal elements contribute interesting dynam-
ics; the density matrix will thermalise to a classi-
cal diagonal distribution in the energy basis, but
in the site basis there are signicant coherent dy-
namics. There is a jump in the coherences when
the states initially exchange amplitude, implying
that there are both quantum coherent and classi-
cal hopping contributions to the proton trans-
fer.
It is also evident that the system equilibrates
to a superposition state in the site basis, with
non-zero coherences. This is sensible; the den-
sity matrix is diagonal in the energy basis in the
nal thermal state, so there will be o-diagonal
elements when viewed in any other basis. This is
true for general Hermitian density matrices and is
only untrue in the special case of the maximally
mixed state, proportional to the identity, which
is diagonal in all bases. This state is seen in the
data: the o-diagonal terms go to zero when the
site populations are equal, suggesting that the
system passes through a maximally mixed state
on the way to equilibrium.
It is possible to identify two clear dynamical
regimes in Figure 2, depending on the chemical
potential.
Bath congurations for which the chemical po-
tential is greater than the resonant energy
0
of
the two state system stimulate population trans-
fer until the two state system has a Boltzmann
8
0 1 2 3 4 5
0
0.2
0.4
0.6
0.8
1
/
0
O
c
c
u
p
a
n
c
y
FIG. 2: These are the occupancies of the second,
high energy site plotted against /
0
, evaluated at
t = 5 10
9
s in blue, t = 3.9 10
8
s in green and
t = 10
6
s in red, being the chemical potential and

0
the dierence in energy eigenvalues of the two-site
system. This was performed at T = 300K. The red
line coincides with a Boltzmann-distribution value of
0.0097. There is a clear dynamical change at =
0
.
population distribution at the same temperature
as the ambient bath. This is physically sensible;
both states which are
0
above or below the Fermi
level of the bath are present, and these transi-
tion states can stimulate transitions in the two
state system in either direction until a dynamic
equilibrium is reached. It is worth noting that
baths with chemical potentials closer to the reso-
nant energy of the two state system will approach
this equilibrium point more quickly, with a peak
in population transfer at
0
+ . This is
also physically reasonable considering the shape
of the spectral density; the coupling spectral den-
sity has a peak energy at = , and the transition
state with this energy is sampled by a Markovian
bath with a Fermi level at =
0
+ . The rate
of equilibration will generally depend on the cou-
pling strengths of the transition states and it is for
this reason that the transient populations of Fig-
ure 2 resemble the Drude-Lorentz spectral density
distribution of equation (42).
The second dynamical regime is for when the
chemical potential is less than the resonant fre-
quency; there is a clear discrete change at =
0
in the simulation output. In this regime, the ef-
fect of the coupling is to again remove coherences
and transfer the proton occupancy into a ther-
mally equilibrated Gibbs state. A state for which
the chemical potential is less than the resonant
frequency
0
does this on a slower timescale, how-
ever, than a conguration for which the chemical
potential is a similar amount above the resonant
frequency. In this model, this is because there is
no state
0
below the Fermi level to contribute to
transitions; it is only interactions with the state

0
above the Fermi level that can stimulate popu-
lation transfer. Having lost fully half of the mech-
anisms which previously contributed to dynami-
cal evolution, it is sensible that the system would
equilibrate at a slower timescale in this regime.
This is still not the complete picture. It would
make sense for equilibration times to increase as
the chemical potential becomes smaller than
0
,
since there will be fewer electrons states within

0
of the Fermi level to stimulate transitions. I
do not think, however, that it is a discrete change
that occurs at precisely =
0
, as suggested by
the data, but a more general principle that if there
are fewer electron states energetically local to the
Fermi level then any transitions will take longer.
It is a relic of the assumptions previously made
that only electron states which are exactly
0
away from the chemical potential can participate
in transitions. This likely comes from the Markov
approximation: if the bath correlation functions
only depend on the current state of the system,
it is eectively proportional to a delta function
over historical times, (t t

). This delta func-


tion contributes to a very sharp resonance in fre-
quency space, such that only electron states ex-
actly on resonance relative to the Fermi level can
participate in transitions. It is not physically rea-
sonable that no other electrons can stimulate this
transition, neither is it sensible that the dynamics
and excitations of the electron bath are not taken
into account. This model does, however, cap-
ture some immediate qualitative behaviour of the
electron-proton interaction, and give some idea of
the timescales over which decoherence and popu-
lation transfer occur, given a set of parameters.
5. NON-MARKOVIAN APPROACHES
5.1. Background
With a view to achieving greater verisimilitude
for the model, I looked into various established
methods for avoiding making the Markov approx-
imation. If this can be achieved then it is possible
to incorporate a bath memory which allows pre-
vious proton states to contribute to present pro-
ton dynamics. This occurs by exchanging infor-
9
mation with the bath: the two-site-system leaves
a dynamical imprint on the bath, which, by virtue
of having a non-zero correlation function, re-
members this information and hence allows it to
aect future dynamics. This stands in contrast to
the Markovian case, in which information about
the two-site-system is immediately dissipated into
the bath, with no possibility of aecting future
dynamics.
In the context of DNA, it is likely that the
dynamics will be highly non-Markovian since the
bath of electrons local to the proton will be rela-
tively small. It can hence be expected that bath
excitations and dynamical changes will contribute
signicantly to the local environment, making the
Markovian assumption of a large, static bath look
increasingly unrealistic.
There is an established method to evaluate
non-Markovian eects to arbitrary orders in cou-
pling strength with time-convolutionless projec-
tion operator techniques. I follow the derivation
of the results in Appendix B, worked through fully
in Chapter 9 of [14]. Here I quote the lowest-order
generator of dynamics K
2
(t), with associated def-
initions:

B
=

i
1
,i
2
_
t
0
dt

12
[

1, [

2,
S
]]
+ i
12
[

1, {

2,
S
}]
_

B
, (43)

12
(t, t

) =Re tr
B
{B
i
1
(t)B
i
2
(t

)
B
}, (44)

12
(t, t

) =Im tr
B
{B
i
1
(t)B
i
2
(t

)
B
}, (45)

1 =A
i
1
(t),

2 = A
i
2
(t

), (46)
recalling the denition of A

and B

in equation
(22). This method requires an integral over the
history of the whole system to evaluate the dy-
namics at a given point. This dependance on past
congurations is central to the non-Markovian na-
ture of this analysis, with the bath correlation
functions determining how environmental excita-
tions will act as a memory for the system. Like
the Markovian case, I will work in the continuum
limit, such that the sums over electronic modes in
B
i
become integrals over frequency with an asso-
ciated spectral density J(), as in equation (40).
5.2. Non-Markovian DNA
Initially, I used the same Drude-Lorentz spec-
tral density J() as in the Markovian case, but
the frequency integrals encountered when evalu-
ating the correlation functions converged far too
slowly to prove numerically feasible. Instead, I
used a spectral density that provided similar dy-
namics, but converged faster for high frequencies
[21]:
J
2
() =

e
/
. (47)
In this case,
12
(t, t

) and
12
(t, t

) become:

12
(s) =
_

0
d

e
/
cos(

s), (48)

12
(s) =
_

0
d

e
/
tanh
_

sin(

s),
(49)
where s = t

t and

= . This spectral
density proved to be more numerically feasible,
and I used Mathematica to nd an analytical so-
lution for
12
and gain insight into the structure
of
12
. As in the Markovian case, I have no reason
to select this specic spectral density over others,
but it should still give a good general picture of
the dynamics. I again used Matlab to simulate
the dynamics of the density equation, given this
new equation of motion.
5.3. Data and Analysis
Realising that the oscillations of the bath cor-
relations become faster with a higher chemical po-
tential, and hence increase the time taken to eval-
uate the frequency integrals of equations (49) and
(48), I chose to keep at a constant low value and
instead vary the energies of the two-site system
with respect to when performing an analysis.
It is for this reason that the energy parameters
in these simulations are smaller than those in the
Markovian analysis; I do not claim these are phys-
ical energies found in DNA, but the relative con-
guration of energies is identical, so I only expect
this to aect the timescales over which population
transfer occurs and not the fundamental proper-
ties of the dynamics.
It is also worth noting that the spectral den-
sities (by denition) peak in frequency space at
= then exponentially decay; this corresponds
to a similar exponential decay with time of the
correlation functions. To ease the computational
load, I imposed a historical cut-o t
hist
=
1
as the time integrated over in equation (43), a
time by which the bath correlation functions be-
come negligibly small. This is a measure of the
10
0 0.5 1 1.5 2 2.5
x 10
8
0
0.2
0.4
0.6
0.8
1
Time (s)
O
c
c
u
p
a
n
c
y
FIG. 3: These are the absolute values of the density
matrix in the Schrodinger picture and site basis plotted
against time, blue being
11
, green
22
and red |
12
| =
|
21
|. The proton was initially in the state
22
= 1 at
T = 77K, with = 70/cm,
0
= 51/cm, = 0.01/cm
and = 10/cm.
memory, the length of time for which past con-
gurations will inuence future congurations of
the system. This simulation proved to be much
more computationally challenging, placing a limit
on the number of simulations achievable.
Like the Markovian case, it is clear from Figure
3 that the system approaches a Boltzmann ther-
mal equilibrium in the dynamical range of >
0
,
with a non-zero coherence term, implying that the
equilibrium state is a superposition over the site
basis. Again, as in the Markovian case, this is
physical; a thermal state is diagonal in the en-
ergy eigenbasis but will not generally be diagonal
in any other basis. It is worth noting that there is
a non-zero coherence term when the site occupan-
cies are equal: the system does not pass through
a maximally mixed state, unlike the Markovian
case.
The simulations in the second energetic
regime, that of <
0
, did not produce physically
meaningful data. The generated density matrices
had negative eigenvalues and hence could not be
interpreted in any reasonable physical way. It is
possible that by only taking the leading order ex-
pansion of K(t), the complete dynamics of the
system are not captured, and it is necessary to
make a fourth or higher order expansion to gain
an understanding of the nature of the dynamics.
I produced simulations for several relative con-
gurations of and
0
, but the large amount
1 2 3 4
x 10
11
0.494
0.495
0.496
0.497
0.498
0.499
0.5
0.501
0.502
Time (s)
O
c
c
u
p
a
n
c
y
FIG. 4: These are the absolute values of the density
matrix in the Interaction picture and site basis plotted
against time, blue being
11
, green
22
and red |
12
| =
|
21
|. This was performed at T = 77K, with =
30/cm,
0
= 21.5/cm, = 0.01/cm and = 10/cm.
of time needed for each simulation made it pro-
hibitively dicult to produce data for a distribu-
tion of values of /
0
. This makes it dicult to
resolve the question of whether the discontinuity
between dynamical regimes of Figure 2 is a relic
of the Markov approximation or a physical phe-
nomenon.
It is possible to see in Figure 4 the ne-
structure time evolution of the density matrix.
There is clear oscillatory behaviour with a con-
stant frequency which was not present in the
Markovian case but was present in the non-
Markovian simulations. This is the interaction
with the bath being resolved on a dynamical
timescale; the oscillations have a frequency of
the order of the bath characteristic energy =
30cm
1
, and hence period of order (c)
1

1.1 10
11
s. This is a fundamental character-
istic of a non-Markovian analysis, that the time
resolution is such that bath correlation memory
eects are resolved and contribute to the dynam-
ics of the complete system. This is part of the
reason why a non-Markovian approach is more
computationally demanding, since there is a well-
dened upper limit on the length of possible dis-
crete timesteps.
6. CONCLUSION
The initial results from a purely analytical ap-
proach to modelling DNA hydrogen bonds as a
11
spin-fermion system suggest that the problem is
not simplied by application of a Lang-Firsov po-
laron transformation. The polaron approach does
allow the possibility of tuning the strength of the
coupling by a careful choice of transformation en-
ergy. Further study into dierent transformations
may produce interesting results, although I sug-
gest that the periodic nature of fermionic com-
mutation relations will ensure that the coupling
term will never be totally absorbed into the renor-
malised tunnelling energy.
Using a master equation driven computational
approach does give some insight into the dynam-
ics and the nature of the spin-fermion coupling.
The transient dynamics of the system are heavily
dependant on the relationship between the chem-
ical potential and the resonant energy of the two-
site system. For a conguration in which the
chemical potential is higher than the character-
istic energy
0
, the system dephases quickly to a
thermal state. In the regime of <
0
, the sys-
tem equilibrates more slowly to a thermal state,
owing to the smaller number of energetically ac-
tive electron states available to participate in the
transition. It can be expected that DNA, in which
there are relatively few local electrons for each
proton, will exist in this small bath regime.
I suggest that this discrete break between
regimes is not physical; when the chemical poten-
tial is small then there are comparatively few
electrons available to stimulate transitions, so it
is sensible that the dynamics will evolve on longer
timescales. It is not reasonable, however, to nd
a discontinuity at exactly =
0
, as there is
nothing fundamentally dierent about the bath-
proton relationship on either side of this energetic
boundary. Unfortunately, this cannot be demon-
strated using the available data. It is likely that
a thorough non-Markovian analysis will yield in-
structive results but this has proved to be more
computationally challenging. I have successfully
applied this in the regime for which >
0
, but
could not perform a meaningful simulation in the
small bath regime.
It would also be instructive to nd estimates,
empirical or theoretical, for values that the phys-
ical variables involved would take. This has been
an abstract modelling investigation, but apply-
ing these results more specically to the mech-
anism of base-pair mutation would take further
work in quantitatively describing the local envi-
ronment of the hydrogen bond, which could vary
signicantly according to base-pair. It is clear
from the work done, however, that the dynamics
of the spin-fermion model depend sensitively on
bath size, especially in the regime most likely to
be of relevance hydrogen bonding in DNA.
7. ACKNOWLEDGEMENTS
Thanks go to Ross Dorner, John Goold and
Felix Pollock for insight into the nature of many-
body fermion problems.
[1] E. Schrodinger, What is Life? (Cambridge Uni-
versity Press, 1944).
[2] G. R. Fleming, J. L. Martin, and J. Breton, Na-
ture 333, 190 (1988).
[3] G. Panitchayngkoon, D. Hayes, K. A. Fransted,
and G. S. Engel, PNAS 107 (2010).
[4] G. S. Engel and G. R. Fleming, Nature 446
(2007).
[5] M. Arndt, T. Juman, and V. Vedral, HFSP
Journal 3, 386 (2009).
[6] E. Gauger, E. Rieper, J. J. L. Morton, S. C.
Benjamin, and V. Vedral, Sustained quantum co-
herence and entanglement in the avian compass,
arXiv:0906.3725v5 (2009).
[7] S. Johnsen and K. Lohmann, Physics Today 61,
29 (2008).
[8] D. G. Cory, M. D. Price, W. Maas, E. Knill,
R. Laamme, Z. W. H., H. T. F., and S. S. S.,
Phys. Rev. Lett. 81, 2152 (1998).
[9] E. Rieper, J. Anders, and V. Vedral, Quantum
entanglement between the electron clouds of nu-
cleic acids in dna, arXiv:1006.4053v2 (2010).
[10] P.-O. Lowdin, Rev Mod Phys 35, 724 (1963).
[11] D. Home and R. Chattopadhyaya, Phys. Rev.
Lett. 76 (1996).
[12] C.-M. Chang, A. H. Castro Neto, and A. R.
Bishop, Chemical Physics 303, 189 (2004).
[13] R. Dorner, J. Goold, L. Heaney, T. Farrow, P. G.
Roberts, J. Hirst, and V. Vedral, Quantum coher-
ent contributions in biological electron transfer,
arXiv:1111.1646v2 (2011).
[14] H.-P. Breuer and F. Petruccione, The theory of
open quantum systems (Oxford University Press,
2007).
[15] M. A. Nielsen and I. L. Chuang, Quantum Com-
putation and Quantum Information (Cambridge
University Press, 2004).
[16] X. Zhao, W. Shi, L.-A. Wu, and T. Yu, Fermionic
stochastic schrodinger equation and master equa-
tion: An open system model, arXiv:1203.2220v1
(2012).
[17] D. P. S. McCutcheon and A. Nazir, J. Chem Phys
12
135, 13 (2011).
[18] S. M. Tan, J. Opt. B: Quantum Semiclass. Opt.
1 (1999).
[19] J. Zhu, s. Kais, P. Rebentrost, and A. Aspuru-
Guzik, J. Chem Phys B 115, 21 (2010).
[20] J.-J. Max and C. Chapados, J. Chem. Phys. 131
(2009).
[21] A. J. Leggett, S. Chakravarty, A. T. Dorsey,
M. P. A. Fisher, A. Garg, and W. Zwerger, Rev.
Mod. Phys. 59, 1 (1987), URL http://link.
aps.org/doi/10.1103/RevModPhys.59.1.
APPENDIX A: POLARON DETAILS
Here I present the transformation of the Hamiltonian H, with the unitary transformation U, with
all relevant algebraic derivation:
H =

i=1,2
E
i
c

i
c
i
+ V (c

1
c
2
+ c

2
c
1
) +

i=1,2

k
i
(
i,k
i
)d

i,k
i
d
i,k
i
+

i=1,2

k
i
g
i,k
i
c

i
c
i
(d
i,k
i
+ d

i,k
i
), (A1)
U = exp
_

i=1,2

k
i

i,k
i
c

i
c
i
(d

i,k
i
d
i,k
i
)

. (A2)
I will make extensive use of the Hadamard Lemma:
e
A
Be
A
= B +
_
A, B

+
1
2!
_
A,
_
A, B

+
1
3!
_
A,
_
A,
_
A, B

, (A3)
valid if A, B C
mm
. I trasform the proton operators c
j
rst:
c
j
=Uc
j
U

,
=c
j
+

k
i

i,k
i
(d

i,k
i
d
i,k
i
)[c

i
c
i
, c
j
]
+
1
2!

i,l

k
i
,k
l

i,k
i

l,k
l
(d

l,k
l
d
l,k
l
)(d

i,k
i
d
i,k
i
)[c

l
c
l
, [c

i
c
i
, c
j
]] + ...
=c
j
(1 +

k
i

j,k
i
(d
j,k
i
d

j,k
i
) +
1
2!

k
i
,k
l

j,k
i

j,k
l
(d
j,k
i
d

j,k
i
)(d
j,k
l
d

j,k
l
) + ...)
=c
j
exp(

k
j

j,k
j
(d
j,k
j
d

j,k
j
)),
(A4)
having used the fermionic commutation relation [c

i
c
i
, c
j
] =
ij
c
j
to compress the commutators.
The conjugate operator follows immediately:
c
j

= c

j
exp(

k
j

j,k
j
(d
j,k
j
d

j,k
j
)). (A5)
I now transform as completely as possible the rst two expressions in the Hamiltonian, which dene
the isolated proton system:

i=1,2
E
i
Uc

i
U

Uc
i
U

+ V (Uc

1
U

Uc
2
U

+ Uc

2
U

Uc
1
U

) =

i=1,2
E
i
c

i
c
i
+ V ( c
1

c
2
+ c
2

c
1
). (A6)
I will now transform the coupling term in the Hamiltonian:
Ug
i,k
i
c

i
c
i
(d
i,k
i
+ d

i,k
i
)U

= g
i,k
i
c

i
c
i
U((d
i,k
i
+ d

i,k
i
)U

, (A7)
13
realising that the proton number operator c

i
c
i
= n
i
is left unchanged under this transformation. It
is most instructive to consider the rst two terms of the Hadamard expansion of the term U((d
i,k
i
+
d

i,k
i
)U

:
U(d
i,k
i
+ d

i,k
i
)U

=(d
i,k
i
+ d

i,k
i
) +
_

j,k
j

j,k
j
n
j
(d

j,k
j
d
j,k
j
), d
i,k
i
+ d

i,k
i

+
1
2!
_

l,k
l

l,k
l
n
l
(d

l,k
l
d
l,k
l
),
_

j,k
j

j,k
j
n
j
(d

j,k
j
d
j,k
j
), d
i,k
i
+ d

i,k
i

+ ...
=(d
i,k
i
+ d

i,k
i
) +

j,k
j
2
j,k
j
n
j
(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)
+
1
2!
_

l,k
l
2
l,k
l
n
l
(d

l,k
l
d
l,k
l
),

j,k
j
2
j,k
j
n
j
(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)

+ ...
=(d
i,k
i
+ d

i,k
i
) +

j,k
j
2
j,k
j
n
j
(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)

1
2!

j,k
j

l,k
l
4
j,k
j
n
j

l,k
l
n
l
(d
i,k
i
+ d

i,k
i
)
jl
+ ...
=(d
i,k
i
+ d

i,k
i
) +

j,k
j
2
j,k
j
n
j
(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)

1
2!

l,k
l
4
2
l,k
l
n
l
(d
i,k
i
+ d

i,k
i
) + ...,
(A8)
where extensive use has been made of the canonical fermionic anticommutation relations to evaluate
the commutators, and that for a fermionic number operator, n
2
l
= n
l
. There is a clear recursion
relation emerging: the second order commutator is a multiple of the zeroth order term, by a factor
= 4

l,k
l

2
l,k
l
n
l
. This recursion relation, that the (n + 2)
th
commutator is a multiple of the n
th
by a factor , holds for both even and odd n, allowing us to construct the transformed electronic
contribution to the coupling Hamiltonian:

d
i,k
i
+

i,k
i
= (d
i,k
i
+ d

i,k
i
)(1

l,k
l
4
2
l,k
l
n
l
2!
+

l,k
l

,k

l
16
2
l,k
l
n
l

2
l

,k

l
n
l

4!
+ ...)
+

j,k
j
2
j,k
j
n
j
(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)(1

l,k
l
4
2
l,k
l
n
l
3!
+

l,k
l

,k

l
16
2
l,k
l
n
l

2
l

,k

l
n
l

5!
+ ...)
= (d
i,k
i
+ d

i,k
i
) cos(

) +

j,k
j
2
j,k
j
n
j
(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)
sin(

,
(A9)
using the polynomial expansions of sine and cosine, and, once again, with = 4

l,k
l

2
l,k
l
n
l
. The
fully transformed coupling term is thus:

i,k
i
g
i,k
i
Uc

i
c
i
(d
i,k
i
+ d

i,k
i
)U

i,k
i
g
i,k
i
n
i
_
cos() +

j,k
j
2
j,k
j
n
j
sin()

(d

j,k
j
d
j,k
j
)
_
(d
i,k
i
+ d

i,k
i
).
(A10)
I will now transform the electronic self-energy term:
14
U

i,k
i
(
i,k
i
)d

i,k
i
d
i,k
i
U

i,k
i
(
i,k
i
)
_
d

i,k
i
d
i,k
i
+
_

j,k
j

j,k
j
n
j
(d

j,k
j
d
j,k
j
), d

i,k
i
d
i,k
i

+
1
2!
_

l,k
l

l,k
l
n
l
(d

l,k
l
d
l,k
l
),
_

j,k
j

j,k
j
n
j
(d

j,k
j
d
j,k
j
), d

i,k
i
d
i,k
i

+ ...
_
=

i,k
i
(
i,k
i
)
_
d

i,k
i
d
i,k
i

j,k
j

j,k
j
n
j

ij
(d
i,k
i
+ d

i,k
i
)
+
1
2!
_

l,k
l

l,k
l
n
l
(d

l,k
l
d
l,k
l
),

j,k
j

j,k
j
n
j

ij
(d
i,k
i
+ d

i,k
i
)

+ ...
_
=

i,k
i
(
i,k
i
)
_
d

i,k
i
d
i,k
i

i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)

1
2!
_

l,k
l

l,k
l
n
l
(d

l,k
l
d
l,k
l
),
i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)

+ ...
_
=

i,k
i
(
i,k
i
)
_
d

i,k
i
d
i,k
i

i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)

1
2!

l,k
l

l,k
l
n
l

i,k
i
n
i
(d

l,k
l
d
l,k
l
)(d
i,k
i
+ d

i,k
i
) + ...
_
.
(A11)
The rst two orders of the expansion are recognisable from the expansion of the coupling term.
The nested higher-order commutators will be identical, so the same recursion relation applies, that
the (n + 2)
th
order term is equal to the n
th
with a multiplicative factor of = 4

l,k
l

2
l,k
l
n
l
. It
is very important to note, however, that the factorial terms associated with the Hadamard expansion
are mismatched, so the associated orders to cancel between the two Hamiltonian contributions. Still,
it is still possible to reconstruct the electronic self-energy contribution using a similar method:

i,k
i
(
i,k
i
)

d
i

d
i
=

i,k
i
(
i,k
i
)
_
d

i
d
i

i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)
_
1

l,k
l
4
2
l,k
l
n
l
3!
+ ...
_
+

l,k
l

l,k
l
n
l
(d

l,k
l
d
l,k
l
)(d
i,k
i
+ d

i,k
i
)
_

1
2!
+

l,k
l
4
2
l,k
l
n
l
4!
+ ...
__
=

i,k
i
(
i,k
i
)
_
d

i
d
i

i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)
sin(

+ 2

j,k
j

i,k
i
n
i

j,k
j
n
j
(cos

1)

(d

j,k
j
d
j,k
j
)(d
i,k
i
+ d

i,k
i
)
_
=

i,k
i
(
i,k
i
)
_
d

i
d
i
+
_

sin(

+ 2

j,k
j

j,k
j
n
j
(cos

1)

_
(d

j,k
j
d
j,k
j
)
_
(d
i,k
i
+ d

i,k
i
).
(A12)
I am now in a position to present the fully transformed Hamiltonian, by a summation of all of the
individual transformed contributions:
15

H =UHU

i
E
i
c

i
c
i
+ V ( c
1

c
2
+ c
2

c
1
) +

i,k
i

i,k
i
d

i,k
i
d
i,k
i
+

i,k
i
n
i
(d
i,k
i
+ d

i,k
i
)

_
g
i,k
i
n
i
cos

+
i,k
i

i,k
i
sin

j,k
j
2
j,k
j
n
j
(d
j,k
j
d

j,k
j
)[
i,k
i

i,k
i
(1 cos

g
i,k
i
sin

]
_
,
(A13)
with and c
j
dened above. The transformation can be tuned to eliminate leading-order terms
in the residual transformed coupling term, by setting the transformation strength parameter
i,k
i
=
g
i,k
i
/(
i,k
i
). For this value of
i,k
i
, the transformed Hamiltonian takes the slightly simpler form:

H = UHU

i=1,2
E
i
c

i
c
i
+ V ( c
1

c
2
+ c
2

c
1
) +

i=1,2

k
i
(
i,k
i
)d

i,k
i
d
i,k
i
+

i=1,2

k
i
_
cos

+
sin

j=1,2

k
j
2
g
j,k
j

j,k
j

n
j
(d
j,k
j
d

j,k
j
)[
(1 cos


sin

]
_
g
i,k
i
n
j
(d
i,k
i
+ d

i,k
i
),
(A14)
now with = 4

l,k
l
g
2
l,k
l
/(
l,k
l
)
2
n
l
.
16
APPENDIX B: NON-MARKOVIAN THEORY
This section follows the work of [14], chapter
9.
When trying to examine the dynamics of a
subsystem
S
of an open system, it is instructive
to dene a projection superoperator P acting on
the full density matrix such that:
P = tr
B
{}
B
=
S

B
. (B1)
It has a complementary superoperator Q. de-
ned such that:
Q = P, (B2)
with the properties P +Q = I, P
2
= P, Q
2
=
Q and PQ = QP = 0.
It is worth making some assumptions about
the environmental reference state of the bath: I
will claim that it is a time-independant stationary
Gibbs state, and that odd moments of the interac-
tion Hamilltonian with respect to it vanish, that
is:
tr
B
_
H
I
(t
1
)H
I
(t
2
)...H
I
(t
2
n + 1)
B
_
= 0, (B3)
or, following on from this:
PL(t
1
)L(t
2
)...L(t
2
n + 1)P = 0. (B4)
It is possible to apply the project operator
P and its conjugate Q to the Liouville-Von-
Neumann equation to produce equations of mo-
tion for the relevant and irrelevant parts of
the total density matrix :
P

t
(t) =PL(t)(t)
=PL(t)P(t) + PL(t)Q(t)
(B5)
Q

t
(t) =QL(t)(t)
=QL(t)P(t) + QL(t)Q(t),
(B6)
where the identity operator I = P + Q has
been inserted between the Liouvillian and the
density matrix in each case. It is possible to solve
the equation of motion for the irrelevant part of
the density matrix Q(t):
Q(t) =G(t, t
0
)Q(t
0
)
+
_
t
t
0
dsG(t, s)QL(s)P(s),
(B7)
using the denition of the propagator:
G(t, s) T

exp
_

_
t
s
ds

QL(s

), (B8)
T

here being the time-ordering chronological


operator, and the time derivative of G(t, s) being:

t
G(t, s) = QL(t)G(t, s). (B9)
Now, with the assumption that the entire sys-
tem starts in a separable state, such that P(t
0
) =
(t
0
), then the term proportional to Q(t
0
) will
vanish, since QP = 0, giving:
Q(t) =
_
t
t
0
dsG(t, s)QL(s)P(s). (B10)
Replacing this back into the equation of mo-
tion for the relevant part P(t), and using the
expression for the time derivative of G(t, s) above:
P

t
(t) =PL(t)P
+
2
_
t
t
0
dsPL(t)G(t, s)QL(s)P(s).
(B11)
Now, using the assumption that odd moments
of H
I
with respect to the bath vanish, the rst
term can be removed, giving:
P

t
(t) =
2
_
t
t
0
dsPL(t)G(t, s)QL(s)P(s)
=
_
t
t
0
dsK(t, s)P(s),
(B12)
where the memory kernel K(t, s) is a superop-
erator, dened as:
K(t, s) =
2
PL(t)G(t, s)QL(s)P. (B13)
17
The ambition of the time-convolutionless oper-
ator method is to remove the dependance in (21)
on previous density matrix states (s). It is pos-
sible to relate this object to the current state (t)
by a simple relationship:
G(t, s) = T

exp
_

_
t
s
ds

L(s

, (B14)
where T

is the antichronological time-


ordering operator. Using this object, the exact
solution of the irrelevant part of the density ma-
trix can be re-expressed:
Q(t) =
_
t
t
0
dsG(t, s)QL(s)PG(t, s)(P +Q)(t)
=(t)(P +Q)(t),
(B15)
having dened:
(t) =
_
t
t
0
dsG(t, s)QL(s)PG(t, s), (B16)
such that, assuming that 1 (t) has an in-
verse:
Q(t) = [1 (t)]
1
(t)P(t). (B17)
It is now possible to reinsert this, as before,
into the equation of motion for the relevant part
of the density matrix, to give:

t
P(t) = K(t)P(t), (B18)
with
K(t) = PL(t)[1 (t)]
1
P. (B19)
The above is exact and now local in time. The
next task is to apply a perturbative expansion to
(t) and K(t), to allow further analysis to take
place. Realising that [1 (t)]
1
as a geometric
series and applying a expansion to (t) and K(t)
in powers of :
[1 (t)]
1
=

n=0
[(t)]
n
(B20)
K(t) =

n=0
PL(t)[(t)]
n
P =

n=1

n
K
n
(t)
(B21)
(t) =

n=1

n
(t). (B22)
Comparing equal powers of to nd each con-
tribution K
n
(t) to the rst two orders:
K
1
(t) = PL(t)P (B23)
K
2
(t) = PL(t)
1
(t)P. (B24)
Recalling that odd consecutive moments of
L(t) vanish, and using the denition of (t) and
zeroth order expansions of G(t, s) and G(t, s) to
nd
1
(t):
K
1
(t) = 0 (B25)
K
2
(t) =
_
t
0
dt1PL(t)L(t
1
)P, (B26)
leaving K
2
(t) as the rst generally non-zero
contribution to K(t). A second-order terminated
master equation for the reduced density matrix

S
(t) gives:
d
dt

S
(t) =
2
_
t
0
ds
_
H
I
(t),
_
H
I
(s),
S
(t)

.
(B27)
It is now possible to apply this to a specic
coupling Hamiltonian H
I
, expressed as:
H
I
=

k
F
K
Q
k
. (B28)
Now dening real and imaginary parts of the
bath correlation functions and a convenient nota-
tion for the contribution F
k
to H
I
:

01
= Rtr
B
{Q
i
0
(t)Q
i
1
(t
1
)
B
} (B29)

01
= Rtr
I
{Q
i
0
(t)Q
i
1
(t
1
)
B
} (B30)

0 = F
i
0
(t),

1 = F
i
1
(t
1
). (B31)
18
Finally, it is now possible to write the second-
order contribution to the time convolutionless
generator K(t) as:
K
2
(t)
S

B
=

i
0
,i
1
_
t
0
dt
1
_

01
[

0, [

1,
S
]] + i
01
[

0, {

1,
S
}]
_

B
.
(B32)

S-ar putea să vă placă și