Scribd Logo
Upload

Welcome to Scribd!

  • An Introduction To Quantum Monte Carlo - Tobochnik Gould

    Uploaded by

    rr2012
    31 views5 pages

    Original Title

    An Introduction to Quantum Monte Carlo - Tobochnik Gould

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    31 views5 pages

    An Introduction To Quantum Monte Carlo - Tobochnik Gould

    Uploaded by

    rr2012

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 5
    COMPUTER ‘SIMULATIONS interest in numerical calculations of many-particle quantum and mechanical systems of both zero and finite temperature. For example, high-temperature super- conductivity is an area of special interest and our ‘understanding of low-dimensional magnetic materials has benefited from such an analysis. High-energy theorists also use the same techniques to understand phase transitions in various field theories. ‘Our goal in this column is to introduce several concepts necessary for understanding quantum Monte Carlo methods. Further discussion of these methods will be the subject of future columns. Since quantum Monte Carlo is based in large part on the estimation of path inte- grals, we will first introduce such integrals in the context of the principle of least action in classical dynamics. We will then generalize the path-integral formalism to problems in quantum mechanics. The action S can be defined as' s=[ua=fr-na, wo where the Lagrangian L = T — Vis written in terms of the kinetic energy T and the potential energy ¥. The value of S depends on the particular path of the particle. The principle of least action is that 6s: @ That is, the action is an extremum for the classical path between fixed beginning and end points, (xo) and (x,t). ‘The principle of least action has been used mostly as a basis for new formulations of classical and quantum ‘mechanics and few introductory textbooks discuss prob- Jems that use the principle of least Action to solve any physical problems. However, we can gain more insight into the least-action principle by actually drawing trajec- tories and computing the action associated with each trajectory. Then by trial and error we can find the trajectory that minimizes S. Such a program was suggested in Ref. 2. We have written a program that uses a mouse as an input to allow students to draw position versus time trajectories for a single particle in a one-dimensional Potential V(x). The action is approximated by replacing. the integral over a path by the sum | rn many areas of physics and chemistry there is much “Ton Tobochnik is asistant profesor of physics ond computer science at Kalamazoo College, Kalamazoo, M1 $9007 and Kenneth Mulder ise sph ‘moreat Kalamazoo College. Harvey Goulds profesor of physics at Clark nisersiy, Worcester, MA 01610. An Introduction To Quantum Monte Carlo Jan Tobochnik, Harvey Gould, and Kenneth Mulder 53 (Lm — 109) © where the time interval is divided into N equal steps of width At= (t—%)/N, the position x, is the position of the particle. In our program the minimum value of Ar is determined by the number of pixels in the horizontal direction. The velocity u, is approximated by 4y-1)/At. “ See Problem 1 for suggestions for further study. Introductions to path-integral methods in quantum mechanies can be found in Ref. 3. We focus on finding the ground-state energy of a single particle in a one- ‘dimensional potential V(x). This problem is conceptually simple bu sufficiently complex to provide the background for the techniques to be discussed in later columns. To in- troduce our notation, we briefly summarize the path- integral formalism. In the traditional formulation of quantum mechan- ics, the probability amplitude (x1) satisfies the Schro- dinger equation: Cet) an, where H is the Hamiltonian. If we assume that the Spectrum of H is discrete, the probability amplitude for having particle at pointxat timer given that it was local- ized at point xp at time fy can be expressed as BV ) K(x tetote) =D bn(2)ba (xe (6), where ¢, satisfied the eigenvalue equation Hé, = E,4,. ‘The total probability amplitude (x,t) can be written as, J axoKeatitotW ote fost (1) ‘The superposition principle can be expressed in terms of K Youn est) = dn Rent) XKxvhitelos SHS (8) We see that K can be interpreted as a propagator. We assume that (6) applies for imaginary values of ¢ and write r= it with 7 real. In this way we can write K(x 700) = 5 4,294, (oe mre revses, mabe 80a wasnenmsase vow areas ve pees as COMPUTER SIMULATIONS and K (45,0) +49 (xY'e "700 0 if the ground-state energy Ey and the first excited state are well separated from each other. ‘Quantum mechanics can also be formulated in terms of the action. The main result of the path-integral formalism is that K can be expressed as? Kxtztotg) = Se (a0) The sum in (10) is over all trajectories and not only the trajectory that minimizes the classical action. To use a computer to evaluate K and hence the ground-siate energy numerically, we need to convert the sum over all trajectories in (10) into a simpler form. We divide the time interval between fg and f= fy into NV intervals, and generalize (8) as K(X,5Xote) Jesediy eT KG. 580, AD where K(X). 1% ASK 10h6.3%p4)) im (11) is evaluated for small Ar. We can show that in the limit a m_\* iat 655180) (=) ft Capek y (ks PI M(t) _ yxy}. (a2: ECE) -r]} om Sic (1) a (2) nae he epnsn of oil tory integrals, we let € = / At and combine (11) and (12) K(xx0.Ne) (55)"Jarren xo |-$ 5 EC +ras}f ay where xy =x. From (9) and (13) we see that if we set x= x)= %y and take Ne large, we can find the ground- state energy and wave function. ‘The form of (13) is that of a multidimensional integral with the displacement variable x, atociated with the time 7. The sequence xp ry is a posible path and the integral in (13) is over al paths. We see that the kinet- ic energy term, m(x,,., — %,)*/2¢, encourages smooth paths with small differences from one point to the next. Note that the form of K(xxNe) in (13) is sila tothe partition function in statistical mechanics, with /f playing the role of an inverse temperature. The effect of the above analysis is to convert a quantum mechanical problem for a single particle in an external potential into statistical mechanics problem for NV “atoms” on a ring connected by nearest-neighbor springs with spring constant, m/e. Each atom also feels the ame external potential V. The label j denotes the order of the atoms in the ring, and the variables x, are displacements from the spring equilibrium positions rather than the positions of the atoms relative to a fixed origin. The choice of xp in the agrument of K in (13) is ar- bitrary. Because of the periodic boundary condition, X=, No point in the closed path is unique, and the same path that contributes to K at x» contributes equally to K at x, etc. Since all the atoms are equivalent, we can use averages overall the displacements of the atoms to compute the ground-state wave function: Thus, if we set Xo =x and integrate over x in (13), the right-hand side becomes a partition function. ‘As we make explicit in the following, we can use standard Monte Carlo techniques including the Metropo- lis algorithm to “simulate” the motion of N atoms on a ring and hence to estimate (13).? However, itis important to remember that the atoms are a product of our analysis and that it is impossible to simulate the actual path of a ‘quantum mechanical particle. A summary of the proce- dure is given below: (0) Choose the parameters V, the number of atoms, and ¢, the time interval, such that Ne>1. It is also necessary to choose 6, the maximum trial change in the displacement of an atom, and M, the total number of Monte Carlo steps per atom. Choose units such that m and # are equal to unity. (1) Choose an initial configuration for the atom displacements x,. If you do not know the approximate shape of the ground-state probability amplitude, give the om 03 02 of curve represents the exact ground-state probability density. +x, random values within a range of displacements where the potential is not too large. (2) To implement the Metropolis algorithm, choose an atom at random and change its displacement from x, to x, =4,+6(Q2r—1), where r is a random number uniformly distributed on the unit interval. Compute the change in the energy E given by aseg (y+ ey If AE

    You might also like