Path integral formulation: Difference between revisions

From SklogWiki
Jump to navigation Jump to search
m (Added a mention of quantum hard spheres in the introduction.)
(Cite tidy.)
Line 1: Line 1:
The '''Path integral formulation''' is an elegant method by which [[quantum mechanics | quantum mechanical]] contributions can be incorporated  
The '''Path integral formulation''' is an elegant method by which [[quantum mechanics | quantum mechanical]] contributions can be incorporated  
within a classical simulation using Feynman path integrals (Refs. 1-7). Such simulations are particularly  applicable to light atoms and molecules such as [[hydrogen]], [[helium]], [[neon]] and [[argon]], as well as quantum rotators such as [[methane]] and [[water]]. From a more idealised point of view path integrals are often used to study [[quantum hard spheres]].
within a classical simulation using Feynman path integrals (see the [[Path integral formulation#Additional reading|additional reading ]] section). Such simulations are particularly  applicable to light atoms and molecules such as [[hydrogen]], [[helium]], [[neon]] and [[argon]], as well as quantum rotators such as [[methane]] and [[water]]. From a more idealised point of view path integrals are often used to study [[quantum hard spheres]].
==Principles==
==Principles==
In the path integral formulation the canonical [[partition function]] (in one dimension) is written as (Ref. 4 Eq. 1)
In the path integral formulation the canonical [[partition function]] (in one dimension) is written as  
(<ref  name="Berne">[http://dx.doi.org/10.1146/annurev.pc.37.100186.002153  B. J. Berne and ­D. Thirumalai "On the Simulation of Quantum Systems: Path Integral Methods", Annual Review of Physical Chemistry '''37''' pp. 401-424 (1986)]</ref>  Eq. 1)
:<math>Q(\beta, V)= \int {\mathrm d} x_1 \int_{x_1}^{x_1} Dx(\tau)e^{-S[x(\tau)]}</math>  
:<math>Q(\beta, V)= \int {\mathrm d} x_1 \int_{x_1}^{x_1} Dx(\tau)e^{-S[x(\tau)]}</math>  
where <math>S[x(\tau)]</math> is the Euclidian action, given by (Ref. 4 Eq. 2)
where <math>S[x(\tau)]</math> is the Euclidian action, given by (<ref  name="Berne"> </ref> Eq. 2)
:<math>S[x(\tau)] = \int_0^{\beta \hbar} H(x(\tau))</math>
:<math>S[x(\tau)] = \int_0^{\beta \hbar} H(x(\tau))</math>
where <math>x(\tau)</math> is the path in time <math>\tau</math> and <math>H</math> is the [[Hamiltonian]].
where <math>x(\tau)</math> is the path in time <math>\tau</math> and <math>H</math> is the [[Hamiltonian]].
This leads to (Ref. 4 Eq. 3)
This leads to (<ref  name="Berne"> </ref> Eq. 3)
:<math>Q_P = \left( \frac{mP}{2 \pi \beta \hbar^2} \right)^{P/2} \int ... \int {\mathrm d}x_1... {\mathrm d}x_P e^{-\beta \Phi_P (x_1...x_P;\beta)}</math>
:<math>Q_P = \left( \frac{mP}{2 \pi \beta \hbar^2} \right)^{P/2} \int ... \int {\mathrm d}x_1... {\mathrm d}x_P e^{-\beta \Phi_P (x_1...x_P;\beta)}</math>
where the Euclidean time is discretised in units of  
where the Euclidean time is discretised in units of  
Line 13: Line 14:
:<math>x_t = x(t \beta \hbar/P)</math>
:<math>x_t = x(t \beta \hbar/P)</math>
:<math>x_{P+1}=x_1</math>
:<math>x_{P+1}=x_1</math>
and (Ref. 4 Eq. 4)
and (<ref  name="Berne"> </ref> Eq. 4)
:<math>\Phi_P (x_1...x_P;\beta)= \frac{mP}{2\beta^2 \hbar^2} \sum_{t=1}^P (x_t - x_{t+1})^2 + \frac{1}{P}  \sum_{t=1}^P  V(x_t)</math>.
:<math>\Phi_P (x_1...x_P;\beta)= \frac{mP}{2\beta^2 \hbar^2} \sum_{t=1}^P (x_t - x_{t+1})^2 + \frac{1}{P}  \sum_{t=1}^P  V(x_t)</math>.


where <math>P</math> is the Trotter number. In the Trotter limit, where <math>P \rightarrow \infty</math> these equations become exact. In the case where <math>P=1</math> these equations revert to a classical simulation. It has long been recognised that there is an isomorphism between this discretised quantum mechanical description, and the classical [[statistical mechanics]] of polyatomic fluids, in particular flexible ring molecules (Ref. 3), due to the periodic boundary conditions in imaginary time. It can be seen from the first term of the above equation that each particle <math>x_t</math> interacts with is neighbours <math>x_{t-1}</math> and <math>x_{t+1}</math> via a harmonic spring. The second term provides the internal potential energy. Thus in three dimensions one has the ''density operator''
where <math>P</math> is the Trotter number. In the Trotter limit, where <math>P \rightarrow \infty</math> these equations become exact. In the case where <math>P=1</math> these equations revert to a classical simulation. It has long been recognised that there is an isomorphism between this discretised quantum mechanical description, and the classical [[statistical mechanics]] of polyatomic fluids, in particular flexible ring molecules<ref>[http://dx.doi.org/10.1063/1.441588      David Chandler and Peter G. Wolynes "Exploiting the isomorphism between quantum theory and classical statistical mechanics of polyatomic fluids", Journal of Chemical Physics '''74''' pp. 4078-4095 (1981)]</ref>, due to the periodic boundary conditions in imaginary time. It can be seen from the first term of the above equation that each particle <math>x_t</math> interacts with is neighbours <math>x_{t-1}</math> and <math>x_{t+1}</math> via a harmonic spring. The second term provides the internal potential energy. Thus in three dimensions one has the ''density operator''


:<math>\hat{\rho} (\beta) = \exp\left[ -\beta \hat{H} \right]</math>
:<math>\hat{\rho} (\beta) = \exp\left[ -\beta \hat{H} \right]</math>
Line 36: Line 37:
:<math>\langle K_P \rangle = \frac{3NP}{2\beta}- \langle U_{\mathrm {spring}}  \rangle    </math>
:<math>\langle K_P \rangle = \frac{3NP}{2\beta}- \langle U_{\mathrm {spring}}  \rangle    </math>
==Harmonic oscillator==
==Harmonic oscillator==
The density matrix for a harmonic oscillator is given by (Ref. 1 Eq. 10-44)
The density matrix for a harmonic oscillator is given by (<ref>R. P. Feynman and A. R. Hibbs "Path-integrals and Quantum Mechanics", McGraw-Hill, New York (1965) ISBN 0-07-020650-3</ref> Eq. 10-44)


:<math>\rho(x',x)= \sqrt{ \frac{m \omega}{2 \pi \hbar \sinh \omega \beta \hbar} } \exp \left( - \frac{m \omega}{2  \hbar (\sinh \omega \beta \hbar)^2 }  \left( (x^2 + x'^2 ) \cosh \omega \beta \hbar - 2xx'\right)\right)</math>  
:<math>\rho(x',x)= \sqrt{ \frac{m \omega}{2 \pi \hbar \sinh \omega \beta \hbar} } \exp \left( - \frac{m \omega}{2  \hbar (\sinh \omega \beta \hbar)^2 }  \left( (x^2 + x'^2 ) \cosh \omega \beta \hbar - 2xx'\right)\right)</math>  
Line 44: Line 45:
*[http://dx.doi.org/10.1119/1.1715108 L. Moriconi "An elementary derivation of the harmonic oscillator propagator", American Journal of Physics '''72''' pp. 1258-1259  (2004)]
*[http://dx.doi.org/10.1119/1.1715108 L. Moriconi "An elementary derivation of the harmonic oscillator propagator", American Journal of Physics '''72''' pp. 1258-1259  (2004)]
==Wick rotation and imaginary time==
==Wick rotation and imaginary time==
One can identify the [[Temperature#Inverse_temperature | inverse temperature]], <math>\beta</math> with an imaginary time <math>it/\hbar</math> (see Ref. 7 &sect; 2.4).
One can identify the [[Temperature#Inverse_temperature | inverse temperature]], <math>\beta</math> with an imaginary time <math>it/\hbar</math> (see <ref>M. J. Gillan "The path-integral simulation of quantum systems" in "Computer Modelling of Fluids Polymers and Solids" eds. C. R. A. Catlow, S. C. Parker and M. P. Allen,  NATO ASI Series C '''293''' pp. 155-188 (1990) ISBN 978-0-7923-0549-1</ref> &sect; 2.4).
*[http://dx.doi.org/10.1103/PhysRev.96.1124 G. C. Wick "Properties of Bethe-Salpeter Wave Functions", Physical Review '''96''' pp. 1124-1134 (1954)]
*[http://dx.doi.org/10.1103/PhysRev.96.1124 G. C. Wick "Properties of Bethe-Salpeter Wave Functions", Physical Review '''96''' pp. 1124-1134 (1954)]


==Rotational degrees of freedom==
==Rotational degrees of freedom==
In the case of systems having (<math>d</math>) rotational [[degree of freedom | degrees of freedom]] the [[Hamiltonian]] can be written in the form (Ref. 8 Eq. 2.1):
In the case of systems having (<math>d</math>) rotational [[degree of freedom | degrees of freedom]] the [[Hamiltonian]] can be written in the form (<ref  name="Marx">[http://dx.doi.org/10.1088/0953-8984/11/11/003 Dominik Marx and Martin H Müser "Path integral simulations of rotors: theory and applications", Journal of Physics: Condensed Matter '''11''' pp. R117-R155  (1999)]</ref> Eq. 2.1):
:<math>\hat{H} = \hat{T}^{\mathrm {translational}} + \hat{T}^{\mathrm {rotational}}+ \hat{V}</math>
:<math>\hat{H} = \hat{T}^{\mathrm {translational}} + \hat{T}^{\mathrm {rotational}}+ \hat{V}</math>


where the rotational part of the kinetic energy operator is given by (Ref. 8 Eq. 2.2)
where the rotational part of the kinetic energy operator is given by (<ref  name="Marx"> </ref> Eq. 2.2)


:<math>T^{\mathrm {rotational}} = \sum_{i=1}^{d^{\mathrm {rotational}}} \frac{\hat{L}_i^2}{2\Theta_{ii}}</math>
:<math>T^{\mathrm {rotational}} = \sum_{i=1}^{d^{\mathrm {rotational}}} \frac{\hat{L}_i^2}{2\Theta_{ii}}</math>


where <math>\hat{L}_i</math> are the components of the angular momentum operator, and <math>\Theta_{ii}</math> are the moments of inertia. For a rigid three dimensional asymmetric top the kernel is given by (Ref. 9 Eq. 5):
where <math>\hat{L}_i</math> are the components of the angular momentum operator, and <math>\Theta_{ii}</math> are the moments of inertia. For a rigid three dimensional asymmetric top the kernel is given by (<ref>[http://dx.doi.org/10.1103/PhysRevLett.77.2638  M. H. Müser and B. J. Berne "Path-Integral Monte Carlo Scheme for Rigid Tops: Application to the Quantum Rotator Phase Transition in Solid Methane", Physical Review Letters '''77''' pp. 2638-2641 (1996)]</ref> Eq. 5):


:<math>\rho(\omega,\omega'; \beta/P) = \sum_{JM\tilde{K}} \left( \frac{2J+1}{8\pi^2}\right) d_{MM}^J (\tilde{\theta'} )  
:<math>\rho(\omega,\omega'; \beta/P) = \sum_{JM\tilde{K}} \left( \frac{2J+1}{8\pi^2}\right) d_{MM}^J (\tilde{\theta'} )  
Line 92: Line 93:
*[http://dx.doi.org/10.1063/1.1316105 C. Chakravarty and R. M. Lynden-Bell "Landau free energy curves for melting of quantum solids", Journal of Chemical Physics '''113''' pp. 9239-9247 (2000)]
*[http://dx.doi.org/10.1063/1.1316105 C. Chakravarty and R. M. Lynden-Bell "Landau free energy curves for melting of quantum solids", Journal of Chemical Physics '''113''' pp. 9239-9247 (2000)]
==References==
==References==
#R. P. Feynman and A. R. Hibbs "Path-integrals and Quantum Mechanics", McGraw-Hill, New York (1965) ISBN 0-07-020650-3
<references/>
#R. P. Feynman "Statistical Mechanics", Benjamin, Reading, Massachusetts, (1972) ISBN 0-201-36076-4  Chapter 3.
==Additional reading==
#[http://dx.doi.org/10.1063/1.441588      David Chandler and Peter G. Wolynes "Exploiting the isomorphism between quantum theory and classical statistical mechanics of polyatomic fluids", Journal of Chemical Physics '''74''' pp. 4078-4095 (1981)]
*R. P. Feynman "Statistical Mechanics", Benjamin, Reading, Massachusetts, (1972) ISBN 0-201-36076-4  Chapter 3.
#[http://dx.doi.org/10.1146/annurev.pc.37.100186.002153  B. J. Berne and ­D. Thirumalai "On the Simulation of Quantum Systems: Path Integral Methods", Annual Review of Physical Chemistry '''37''' pp. 401-424 (1986)]
*[http://dx.doi.org/10.1103/RevModPhys.67.279 D. M. Ceperley "Path integrals in the theory of condensed helium", Reviews of Modern Physics '''67''' 279 - 355 (1995)]
#[http://dx.doi.org/10.1103/RevModPhys.67.279 D. M. Ceperley "Path integrals in the theory of condensed helium", Reviews of Modern Physics '''67''' 279 - 355 (1995)]
*[http://dx.doi.org/10.1080/014423597230190 Charusita Chakravarty "Path integral simulations of atomic and molecular systems", International Reviews in Physical Chemistry '''16''' pp. 421-444 (1997)]
#[http://dx.doi.org/10.1080/014423597230190 Charusita Chakravarty "Path integral simulations of atomic and molecular systems", International Reviews in Physical Chemistry '''16''' pp. 421-444 (1997)]
# M. J. Gillan "The path-integral simulation of quantum systems" in "Computer Modelling of Fluids Polymers and Solids" eds. C. R. A. Catlow, S. C. Parker and M. P. Allen,  NATO ASI Series C '''293''' pp. 155-188 (1990) ISBN 978-0-7923-0549-1
#[http://dx.doi.org/10.1088/0953-8984/11/11/003 Dominik Marx and Martin H Müser "Path integral simulations of rotors: theory and applications", Journal of Physics: Condensed Matter '''11''' pp. R117-R155  (1999)]
#[http://dx.doi.org/10.1103/PhysRevLett.77.2638  M. H. Müser and B. J. Berne "Path-Integral Monte Carlo Scheme for Rigid Tops: Application to the Quantum Rotator Phase Transition in Solid Methane", Physical Review Letters '''77''' pp. 2638-2641 (1996)]
==External links==
==External links==
*[http://www.smac.lps.ens.fr/index.php/Programs_Chapter_3:_Density_matrices_and_path_integrals Density matrices and path integrals] computer code on SMAC-wiki.
*[http://www.smac.lps.ens.fr/index.php/Programs_Chapter_3:_Density_matrices_and_path_integrals Density matrices and path integrals] computer code on SMAC-wiki.
[[Category: Monte Carlo]]
[[Category: Monte Carlo]]
[[category: Quantum mechanics]]
[[category: Quantum mechanics]]

Revision as of 12:55, 9 July 2009

The Path integral formulation is an elegant method by which quantum mechanical contributions can be incorporated within a classical simulation using Feynman path integrals (see the additional reading section). Such simulations are particularly applicable to light atoms and molecules such as hydrogen, helium, neon and argon, as well as quantum rotators such as methane and water. From a more idealised point of view path integrals are often used to study quantum hard spheres.

Principles

In the path integral formulation the canonical partition function (in one dimension) is written as ([1] Eq. 1)

where is the Euclidian action, given by ([1] Eq. 2)

where is the path in time and is the Hamiltonian. This leads to ([1] Eq. 3)

where the Euclidean time is discretised in units of

and ([1] Eq. 4)

.

where is the Trotter number. In the Trotter limit, where these equations become exact. In the case where these equations revert to a classical simulation. It has long been recognised that there is an isomorphism between this discretised quantum mechanical description, and the classical statistical mechanics of polyatomic fluids, in particular flexible ring molecules[2], due to the periodic boundary conditions in imaginary time. It can be seen from the first term of the above equation that each particle interacts with is neighbours and via a harmonic spring. The second term provides the internal potential energy. Thus in three dimensions one has the density operator

which thanks to the Trotter formula we can tease out , where

and

The internal energy is given by

The average kinetic energy is known as the primitive estimator, i.e.

Harmonic oscillator

The density matrix for a harmonic oscillator is given by ([3] Eq. 10-44)

Related reading

Wick rotation and imaginary time

One can identify the inverse temperature, with an imaginary time (see [4] § 2.4).

Rotational degrees of freedom

In the case of systems having () rotational degrees of freedom the Hamiltonian can be written in the form ([5] Eq. 2.1):

where the rotational part of the kinetic energy operator is given by ([5] Eq. 2.2)

where are the components of the angular momentum operator, and are the moments of inertia. For a rigid three dimensional asymmetric top the kernel is given by ([6] Eq. 5):

where are the Euler angles, is the Wigner D-matrix and are the eigenenergies.

Techniques

Path integral Monte Carlo

Path integral Monte Carlo (PIMC)

Path integral molecular dynamics

Path integral molecular dynamics (PIMD)

Centroid molecular dynamics

Centroid molecular dynamics (CMD)

Ring polymer molecular dynamics

Ring polymer molecular dynamics (RPMD)

Contraction scheme

Normal mode PIMD

Grand canonical Monte Carlo

A path integral version of the Widom test-particle method for grand canonical Monte Carlo simulations:

Applications

References

Additional reading

External links