Dynamics of a Quantum Dissipative System Coupled with an Oscillator ()
1. Introduction
Path integral methods constitute an interesting and extended part of mathematical physics and there is considerable effort in their development [1] . The use of the central limit theorem in path integral methods is of interest as it leads to the solution of the sign problem [2] - [12] appearing in quantum physics. That solution is applicable to various systems even beyond quantum mechanics (see the conclusions in [6] ).
Here, we study the dissipative dynamics of a quantum mechanical system, which is coupled with an oscillator via an interaction term of the form
, where s is the coordinate of the system and x is one of the oscillators. The dissipation on the system is modelled via coupling the system’s particle with a harmonic heat bath of inverse temperature
. Proceeding we first path integrate over the bath and the oscillator and obtain a path integral expression for the reduced system’s density matrix, which includes the bath and oscillator’s influence functionals. This kind of expression is well known to show a highly oscillatory phase leading to failure of numerical methods for their evaluation, namely the Monte Carlo method. That problem is known as the sign problem. We can solve that problem via extracting an alternative expression for the propagator called sign solved propagator where the oscillations are controlled. The whole method based on the use of the central limit theorem was developed by the author and was applied to other models in previous papers [2] - [12] . So, in that way, we can derive the time evolution of the system’s density matrix. In fact, here we study the time evolution of the position of a particle in a symmetric double well. We have chosen a double well potential as it incorporates tunnelling effects in the whole dynamics and behaves as a two-level system. In the final applications, we consider the particle interacting with only the harmonic bath, with only the oscillator or with both of them. Systems similar to the present one have been studied exhaustively. See for example [13] [14] [15] and references there. Other methods of study include the use of generalized Langevin equations or master equations [13] . However, the present path integral approach combined with the solution of the sign problem gives exact closed results from a fully quantum mechanical point of view.
The present paper proceeds as follows. In Section 2, we give the system and its Hamiltonian, consider the path integral that describes it and further path integrated over the bath and give the form of the corresponding influence functional. In Section 3, we derive the influence function of the interaction of the system with an oscillator. In Section 4, we solve the sign problem to study the time evolution of the system’s density matrix. In Section 5, we give results of the theory in the case of a double well coupled with a bath, or with an oscillator or with both of them and suppose initially a Gaussian wavefunction. In Section 6, we present our conclusions and finally, in Appendix, we solve the sign problem in the case of a density matrix.
2. Model Description
Since in the present paper, we consider the dynamics of a particle coupled on the one hand with a harmonic bath, which models a dissipative environment and on the other with a harmonic oscillator, the full Hamiltonian has the form:
(1)
is the system’s Hamiltonian given by:
(2)
The oscillator’s one has the form:
(3)
supposing time independent parameters and coupling of the form
. That coupling can appear if we assume that the system deforms the oscillator’s potential. Then, we obtain the present effective interaction.
Finally, the harmonic bath’s Hamiltonian has the form:
(4)
which has a linear coupling. s may be interpreted as a reaction coordinate coupled to a large number of harmonic bath degrees of freedom. Equation (4) includes counterterms quadratic in s, which renormalize the system potential. That ensures that important potential features such as the barrier height do not change with the coupling strength.
Quantum mechanical observables of the system can be obtaining after tracing the full density matrix
over the bath and the oscillator. i.e.
(5)
Moreover, we assume that the interaction of all the three subsystems is switched on at time
. i.e. we assume that the density matrix at the initial time is:
(6)
where
is the system’s initial density matrix,
is the bath’s one and
the oscillator’s initial one.
The whole dynamics can be extracted via path integrating over the bath, the oscillator and the system. At first, we consider the integration over the bath. Its coordinates appear in the Hamiltonian (4). We assume it to be at inverse temperature
. Then, according to standard methods [15] , we can obtain a corresponding influence functional in the form:
(7)
where we set:
(8)
where
is the spectral density. It incorporates the characteristics of the bath pertaining to the dynamics of the reaction coordinate corresponding to the system.
In the next section, we path integrate the oscillator Hamiltonian.
3. Integration over the Oscillator
We consider the Hamiltonian (3) of a harmonic oscillator of time independent frequencies and masses. We proceed to path integrate the Hamiltonian (3) via standard methods. We set:
(9)
(10)
to obtain the free of mass terms Hamiltonian:
(11)
where we set:
(12)
If
is the propagator corresponding to the Hamiltonian (3) and
is the one corresponding to the Hamiltonian (11), then we can easily check that they obey the relation:
(13)
Therefore, we can concentrate our attention on the propagator of the Hamiltonian (11). It can be calculated via standard path integration. To proceed towards the integration, we perform the canonical transformation:
(14)
(15)
(16)
The
function depends on s through the differential Equation (32) (see below). i.e. s is supposed to be a function of time and to describe the coordinate of the double well. Moreover, during the present section’s evaluation, it is fixed. The above transformation is canonical since it preserves the Poisson brackets and therefore it preserves the volume element in phase space. As the present, transformation involves the generic time redefinition (16), we give more details. The
time slices discrete form of
involves the times
, where the time step is
. Now, on integrating the path integral expression on the momentums, it becomes (see below as well):
(17)
Then, under the transformations (14)-(16), the time step becomes
, where we have symmetrised the expression in order to avoid any preference of the one time over the other. So, we conclude that the path differential measure takes the form:
(18)
and the discretized action appearing in Equation (17) is:
(19)
Therefore, on using the expansions
, we find that the propagator
is related with the transformed one via:
(20)
where on switching to a phase space path integral, we get:
(21)
where we set:
(22)
and we have used the notation:
(23)
(24)
(25)
Now, we impose constrain on
by setting the global time-dependent term multiplying the
terms in Equation (21) equal to a constant. i.e.
(26)
where we set:
(27)
Further, the integration with respect the
variables is performed and we find the propagator
(28)
where
is the solution of the differential equation:
(29)
and
(30)
In (28), we have set
and
.
So, finally, we can obtain the propagator
from Equations (9) and (10). It has the form:
(31)
The differential Equation (29) takes the form:
(32)
and
(33)
Therefore, the propagator can be derived from the system of Equations (31)-(33). We have to solve the differential Equation (32) with the variable s as parameter, evaluate Equation (33) and then apply (31). We notice that in Equations (3) and (31)-(33) instead of the square function
there could appear any function of s.
Now, we assume that the harmonic oscillator is initially at the state:
(34)
Then,
in Equation (6) becomes:
(35)
and the influence functional describing the effect of the oscillator on the system is:
(36)
where we set:
(37)
and
(38)
(39)
So, the effect of the bath on the system is described by the influence functional (7)-(8) while the effect of the oscillator by the influence functional (36). Now, we turn our attention to the system.
4. Solution of the System’s Sign Problem
According to standard path integral methods, as well as the discussion in the previous sections, the system’s density matrix at time t is going to have the path integral representation:
(40)
where we have set
,
and
.
is the number of time slices in the path integral and
is the influence functional which describes the possible interactions of the system. Expression (40) is highly oscillatory and therefore standard Monte Carlo techniques fail to confront it. We can bypass the whole point if we interpret the Hamiltonians as random variables and apply the central limit theorem on the phase of Equation (40). In that way, we obtain an oscillation free expression called sign solved density matrix.
Now, we observe that the influence functional has the form:
(41)
(see Equations (7), (8) and (36)). It has a product form as the bath and the harmonic oscillator are not directly coupled.
The bath influence functional
in its discrete form is:
(42)
The matrix elements
are given in [15] . As we can observe there,
of them, corresponding to
, are of order
and the rest ones are of order
. So, for N large enough, there are positive constants such that:
(43)
(44)
Further, expression (31) is bounded and therefore its matrix elements are bounded as well. So,
is bounded. As we prove in Appendix, a theory similar to the sign solved propagator theory of Ref. [7] and Ref. [12] applies. Eventually, we obtain the sign solved influence functional expression:
(45)
Now, we observe that the primed and unprimed variables of h in Equation (45) and therefore of I and R have a diagonal form and therefore
and
. So, eventually, the influence functional given by Equations (7) and (8) takes the diagonal form:
(46)
where we set:
(47)
We have evaluated the time integrals as the position variables are diagonal.
In the present paper, we use the ohmic spectral density:
(48)
where
is a cutoff frequency.
Moreover, by setting
to correspond either to
or to
, the propagator (31) becomes:
(49)
and according to Equation (36), the influence functional R has the diagonal form:
(50)
where we set:
(51)
and
(52)
(53)
Here,
(54)
and
(55)
Finally, we obtain:
(56)
In the next section, we proceed to an application.
5. Application to a Symmetric Double Well
In the present section, we proceed to an application of the above theory. We consider the symmetric double well potential:
(57)
and assume to have prepared a Gaussian wave packet centered on the right well with wavefunction:
(58)
and energy
. Due to tunneling, the level
splits into the levels
and
with corresponding wavefunctions:
(59)
(60)
and energy differences:
(61)
To apply the theory of the previous section, we use the initial density matrix:
(62)
and insert it in Equation (56).
Now, we are in position to generate the time evolution of the system’s density matrix. In that way, the evolution of the system’s observables under the preparation (58) can be studied.
We consider the matrix elements:
(63)
Then,
(64)
In the expectation values in Equation (56), we choose as sampling functions the expressions (58)-(60), so that
where the i is the same as in Equations (63) and (64). Then,
.
is the tunnelling frequency.
So, the expectation value of the position has the form:
(65)
We observe that the expectation value of the position is closely related with the inversion of a corresponding two-level system. In fact, as we can conclude from the above analysis the present double well potential can be interpreted as a two-level system.
Throughout the section, we consider the system initially in the state (62).
In Figure 1, we plot the expectation value
in the case of the system interacting with only the harmonic bath. So, we use the influence functional
. In that case, the matrix elements (63) are Gaussian and we obtain:
(66)
As the time increases, the bath causes decrease of the amplitude of the tunnelling oscillations.
In Figure 2(a) and Figure 2(b), we consider the system interacting with just the oscillator. So, we use the influence functional
. At small times there appear extra oscillations, besides the tunnelling’s ones. Moreover, for fixed time the absolute value
decreases as
increases (see the initial wavefunction (34) of the oscillator and the influence functional (50)).
In Figure 3, we consider the full system. So, we use the influence functional
. There appears a combination of the effects described in the cases in Figure 1 and Figure 2. Oscillations at small times and decrease of the amplitude as time increases. We should expect such a result as the full influence functional is the product of the influence functionals corresponding to the system’s interaction with just the bath or the oscillator (see Equation (41)).
6. Conclusions
In the present paper, we study the dynamics of a quantum mechanical system interacting linearly with a bath and quadratically with an oscillator. In our study, we use influence functional methods derived previously and concerning the interaction of systems with baths as well as methods on the dynamics of oscillators and combine them with methods on the solution of the sign problem due to the author. As an application, we have considered a symmetric double well interacting with a bath or with an oscillator, or with both of them, and study the time evolution of the relevant density matrix. We focus on a double well potential as it incorporates tunnelling effects on the whole dynamics and behaves as a two-level system.
Figure 1. Mean position of the dissipative double well system. We use
(solid),
(dashed) and
(dotted). We have set:
,
,
,
,
. Here, according to Equation (61),
.
(a)(b)
Figure 2. Mean position of a double well system coupled with an oscillator. In (a), we consider the mean position as a function of
; while in (b), we consider the position as a function of
for certain values of
. We use:
,
,
,
,
,
. Here, according to Equation (61),
.
Figure 3. Mean position in the following cases. A double well: Dashed-dotted. A dissipative double well: Solid. A double well coupled with an oscillator: Dotted. A double well coupled with a harmonic bath and an oscillator: Dashed. The parameters have the values:
,
,
,
,
,
,
,
,
,
. Here, according to Equation (61),
.
In conclusion, the present methods are capable for giving closed expressions on various systems’ density matrix time evolution and therefore interesting relevant dynamical information can be gained.
Appendix: Sign Solved Density Matrix
According to the solution of the sign problem for expression (40),
(A1)
We have used the functions:
(A2)
(A3)
In (A2) and (A3), we use appropriate sampling functions. In the primed f and g in (A1), we use a primed sampling function for the variances.
In calculations, we are interested in integrals of the form
, where
,
are appropriate functions. So, we consider the expression
corresponding to the term in Equation (A1) involving the
and
functions after that integration. We intent to prove that only the diagonal term in the last expression in Equation (A1) can give the exact result as
because
tends to zero. Then, we obtain Equation (45). To prove that, we diagonalize and integrate the Gaussian products
and
. We notice that according to Equation (50) the R is bounded (see Equation (41)).
According to Equations (42)-(44), the bath influence functional
has the form:
(A4)
where for N large enough,
(A5)
(A6)
We want to bound the expression
appropriately. We proceed via performing on the terms composed of the f functions, the change of variables:
(A7)
and similarly, the change of variables:
(A8)
on the terms composed of the g functions.
We have set:
(A9)
Similar primed transformations apply to the case of primed variables.
Then, we obtain:
(A10)
and
are appropriate time dependent functions. We have set:
(A11)
and
(A12)
where
,
. The matrices in Equation (A10) correspond to the symmetric matrices:
(A13)
where we set:
(A14)
and similarly for the primed variables. Moreover,
(A15)
has the form:
(A16)
Depending on the s being primed, unprimed or mixed we use primed, unprimed or mixed variables in the coefficients (A15) of the quadratic forms in Equation (A10). For example, for the term corresponding to
in Equation (A10), we use in the expression (A15) the product
.
Now, we study the matrices (A13, A14). We observe that on the one hand,
(A17)
and on the other for each determinant,
(A18)
is a Chebyshev polynomial of the second kind of order
. More particularly let the numbers
,
, be the roots of the equation
. They are simple, real roots and
,
. Then, the eigenvalues of the matrices
are going to be given by the expressions:
(A19)
(A20)
Further, the diagonal quadratic forms
can be diagonalized simultaneously with the quadratic forms corresponding to the matrices
. So, we conclude that the eigenvalues of the quadratic forms:
(A21)
are going to have the form:
(A22)
(A23)
where
are appropriate real numbers with
.
Further, for N large enough, the expressions corresponding to the matrices
in (A10) are perturbation terms. So, eventually the full matrices on the two exponentials will have eigenvalues:
(A24)
(A25)
According to the whole above discussion after a Gaussian integration, we obtain:
(A26)
where we set:
(A27)
(A28)
The constants
,
depend on the form of the functions
and
as well as
.
Finally, we infer that since the terms in the curly brackets in Equation (A26) tend to zero as
, the first term in eq. (A1) is exact as
and corresponds to the sign solved time evolution of the density matrix.