MIT OpenCourseWarehttp://ocw.mit.edu 5.74 Introductory Quantum Mechanics II Spring 2009 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.5-1 5.1. TIME-CORRELATION FUNCTIONS Time-correlation functions are an effective and intuitive way of representing the dynamics of a system, and are one of the most common tools of time-dependent quantum mechanics. They provide a statistical description of the time-evolution of a variable for an ensemble at thermal equilibrium. They are generally applicable to any time-dependent process for an ensemble, but are commonly used to describe random (or stochastic) and irreversible processes in condensed phases. We will use them in a description of spectroscopy and relaxation phenomena. This work is motivated by finding a general tool that will help us deal with the inherent randomness of molecular systems at thermal equilibrium. The quantum equations of motion are deterministic, but this only applies when we can specify the positions and momenta of all the particles in our system. More generally, we observe a small subset of all degrees of freedom (the “system”), and the time-dependent properties that we observe have random fluctuations and irreversible relaxation as a result of interactions with the surroundings. It is useful to treat the environment (or “bath”) with the minimum number of variables and incorporate it into our problems in a statistical sense – for instance in terms of temperature. Time-correlation functions are generally applied to describe the time-dependent statistics of systems at thermal equilibrium, rather than pure states described by a single wavefunction. Statistics Commonly you would describe the statistics of a measurement on a variable A in terms of the moments of the distribution function, P(A), which characterizes the probability of observing A between A and A+dAA dA A P A Average: =∫() (5.1) A2Mean Square Value: =∫dA A 2 PA () . (5.2) Similarly, this can be written as a determination from a large number of measurements of the value of the variable A: N A = 1 ∑Ai (5.3)Ni=1 A2 1 N 2=∑ Ai . (5.4)Ni=1 Andrei Tokmakoff, MIT Department of Chemistry, 2/24/20095-2 The ability to specify a value for A is captured in the variance of the distribution 2A2σ 2 = − A (5.5) The observation of an internal variable in a statistical sense is also intrinsic to quantum mechanics. A fundamental postulate is that the expectation value Aˆof an operator =ψ Aˆ ψ is the mean value of A obtained over many observations. The 2probability distribution function is associated with ψ dr . To take this a step further and characterize the statistical relationship between two variables, one can define a joint probability distribution, P(A,B), which characterizes the probability of observing A between A and A+dA and B between B and B+dB. The statistical relationship between the variables can also emerges from moments of P(A,B). The most important measure is a correlation function CAB = AB − A B (5.6) You can see that this is the covariance – the variance for a bivariate distribution. This is a measure of the correlation between the variables A and B, that is, if you choose a specific value of A, does that imply that the associated values of B have different statistics from those for all values. To interpret this it helps to define a correlation coefficient r = CAB . (5.7)σσAB ρ can take on values from +1 to −1. If r = 1 then there is perfect correlation between the two distributions. If the variables A and B depend the same way on a common internal variable, then they are correlated. If no statistical relationship exists between the two distributions, then they are uncorrelated, r = 0, and AB = A B . It is also possible that the distributions depend in an equal and opposite manner on an internal variable, in which case we call them anti-correlated with r = −1. Equation (5.6) can be applied to any two different continuous variables, but most commonly these are used to describe variables in time and space. For the case of time-correlation5-3 functions that we will be investigating, rather than two different internal variables, we will be interested in the value of the same internal variable, although at different points in time. Equilibrium systems For the case of a system at thermal equilibrium, we describe the probability of observing a variable A through an equilibrium ensemble average A . Classically this is A =∫ p∫ pq , ) (5.8)d dq A( ,;t)ρ(pqwhere ρ is the canonical probability distribution for an equilibrium system at temperature T −βH ρ=e (5.9)Z Z is the partition function and β=kBT. In the quantum mechanical case, we can write A =∑pn n A n (5.10) n where pn= e−βEn / Z (5.11) Equation (5.10) may not seem obvious, since it is different than our earlier expression A = aaA = Tr(ρA). The difference is that in the present case, we are ∑ n * m mn ,nm dealing with a statistical mixture or mixed state, in which no coherences (phase relationships) are present in the sample. To look at it a bit more closely, the expectation value for a mixture A =∑pk ψk A (5.12)ψk k can be written somewhat differently as an explicit sum over N statistically independent molecules 1 N ()i * ()iA = ∑∑(an )am n A m (5.13) i=1,N nm By statistically independent, we mean that the molecules interact only with their surroundings and not with each other. They have no knowledge of each other. Then, the sum over molecules is just the ensemble averaged value of the expansion coefficients = ∑ * A aa nAm (5.14)n m ,nm5-4 We can evaluate this average recognizing that these are complex numbers, and that the equilibrium ensemble average of the expansion coefficients is equivalent to phase averaging over the expansion coefficients. Since at equilibrium all phases are equally probable 12π *1* aa = aad φ= a a ∫2π e−iφnmdφnm (5.15)n m ∫ nm n m0 02π 2π where a = a eiφnandφnm =φn−φm. The integral in (5.15) is quite clearly zero unless φn =φm,n n giving −βE * aa = pn =en (5.16)n m Z Of course, even at equilibrium the expectation value of A for a member of ensemble as a function of time At. Although the behavior of At may be well-defined and periodic, for mixed i()()istates it generally is observed to fluctuate randomly: The fluctuations are about a mean value A. Given enough time, we expect that one