Discover millions of ebooks, audiobooks, and so much more with a free trial

Only $11.99/month after trial. Cancel anytime.

An Introduction to the Theory of the Boltzmann Equation
An Introduction to the Theory of the Boltzmann Equation
An Introduction to the Theory of the Boltzmann Equation
Ebook347 pages2 hours

An Introduction to the Theory of the Boltzmann Equation

Rating: 2 out of 5 stars

2/5

()

Read preview

About this ebook

Boltzmann's equation (or Boltzmann-like equations) appears extensively in such disparate fields as laser scattering, solid-state physics, nuclear transport, and beyond the conventional boundaries of physics and engineering, in the fields of cellular proliferation and automobile traffic flow. This introductory graduate-level course for students of physics and engineering offers detailed presentations of the basic modern theory of Boltzmann’s equation, including representative applications using both Boltzmann’s equation and the model Boltzmann equations developed within the text. It emphasizes physical aspects of the theory, and it represents a practical resource for researchers and other professionals. The problems following each chapter are intended as learning examples, and they frequently extend and generalize the text material. Additional editorial features include a historical introduction, references, and subject and author indexes. 1971 edition.

LanguageEnglish
Release dateDec 27, 2012
ISBN9780486143828
An Introduction to the Theory of the Boltzmann Equation

Related to An Introduction to the Theory of the Boltzmann Equation

Titles in the series (100)

View More

Related ebooks

Mechanical Engineering For You

View More

Related articles

Reviews for An Introduction to the Theory of the Boltzmann Equation

Rating: 2 out of 5 stars
2/5

1 rating0 reviews

What did you think?

Tap to rate

Review must be at least 10 words

    Book preview

    An Introduction to the Theory of the Boltzmann Equation - Stewart Harris

    EQUATION

    CHAPTER 1

    statistical mechanical

    preliminaries

    1–1 THE MICROSCOPIC DESCRIPTION

    Equilibrium and nonequilibrium statistical mechanics have as one of their objectives the determination of the basic thermo-fluid properties of systems which are characterized by having a large number of degrees of freedom. The description is on the molecular level and is usually given in terms of the spatial coordinate and momentum of each of the molecules that constitute the system of interest. This is to be contrasted with the corresponding thermodynamic description, which for a one-component system requires but three variables, only two of which are independent. For a fluid mechanical description we must specify in addition to the three thermodynamic variables the velocity field, stress tensor, and heat flux vector. The last two quantities are not usually treated as independent variables, however, so that a minimal fluid mechanical description requires five independent variables. The statistical mechanical description of the same system requires 6N variables, where N is the number of molecules in the system, typically about 10²³. As mentioned above, these variables are usually taken to be the 3N spatial coordinates, q1, q2, ... , qN ≡ {q}, and the 3N conjugate momenta, p1, p2, ... pN ≡ {p}, of the constituent molecules. Here we have already simplified matters by considering the spatial specification of a molecule to be completely determined by a point in space, so that we have neglected orientation effects, vibrations, and, in fact, all phenomena which are related to the structure of the molecules. We will generally restrict ourselves to structureless particles throughout this book.

    The equations of motion of the molecular system we are considering are given most conveniently in terms of the Hamiltonian of this system, which we denote as H({q}, {p}), where

    H({q},{q}) = E ({q},{q}) (1–1)

    In the absence of external fields, E({q},{p}) is the total energy, kinetic energy plus any potential energy due to intermolecular forces, of the system. In terms of H, the Hamiltonian formulation of the equations of motion for the system is

    The state of the system at any time can be specified by a single point in the 6N dimensional space (called the gamma space, or Γ space) having the mutually orthogonal axis q1, q2, ... , pN-1, pN. This space is an example of what is more generally called a phase space; a point in the phase space of a particular system determines the phase, that is, state, of that system. If we know what the intermolecular forces are, then given any point in the gamma space (initial data), we can determine from Equation (1–2), in theory, the state of the system at any later or earlier time. Thus Equation (1–2) provides us with the modus operandi for calculating the trajectories of a particular system in the phase space, provided we are supplied with some initial conditions.

    1–2 RELATIONSHIP BETWEEN MICROSCOPIC AND

    MACROSCOPIC DESCRIPTIONS

    We will accept the following statement as a basic postulate of both equilibrium and nonequilibrium statistical mechanics.

    Postulate

    All intensive macroscopic properties of a given system can be described in terms of the microscopic state of that system.

    Thus any intensive macroscopic property which we may observe by measurement, Gobs, can be also written as G({q},{p}) and determined directly from a knowledge of the microscopic state. In most cases it may be a difficult task to determine the proper function G, but it is nevertheless postulated that for every intensive macroscopic property there exists a G. For the common thermo-fluid macroscopic properties we shall consider here we will always be able to explicitly display the corresponding G.

    In measuring any macroscopic quantity the associated physical measurement is never instantaneous, but rather must always be carried out over some finite time τ. This interval may be made very small by refinement of observational technique, but it will always remain finite. The quantity we observe is thus not the instantaneous value, G({q}, {p}), but rather a time-averaged value of G, where if t0 is the time the observation is begun

    If the proper value of G associated with a particular macroscopic quantity is known, then the observed value can be calculated from Equation (1–3), using the equations of motion given by Equation (1–2), provided that the microscopic state of the system is known at some given time. In practice such a calculation is, of course, impossible. We are not able to determine the microscopic state at any given time, and even were such a specification possible, we would be faced with an overwhelming number of equations to solve. Thus any attempt to put Equation (1–3) to practical use will involve an insurmountable computational problem, due primarily to the large number of degrees of freedom of the systems which we typically encounter.

    1–3 THE GIBBS ENSEMBLE

    from Equation (1–3), we turn to another method of determining Gobs. This is the method of ensembles, originally due to Gibbs. We consider in the minds eye a large number of systems, η, which macroscopically are equivalent to the system we are considering. The thermo-fluid properties of each of the η replicas are then equal to those of the system of interest. The microscopic description is not specified, and we would expect that this will differ greatly among the replica systems in the absence of such specification, since there are large numbers of microscopic states corresponding to any given macroscopic state. This collection of systems is referred to as an ensemble.

    Each system in the ensemble can be represented by a point in the phase space. If we let η → ∞ these points become quite dense in the phase space, and we can describe their distribution throughout the phase space by a density function which is a continuous function of {q}, {p}. We can also normalize this function so that it is a probability density in the phase space; the normalized probability density function will be denoted by FN({q},{p},t) = FN. FN will be symmetric in the qi and pi.

    Specifically we will have

    Although FN is itself a probability, it changes in time in a completely deterministic manner. If we know the dependence of FN on the phase at a particular time, then by solving Hamilton’s equations, Equation (1–2), we can determine FN at any later (or earlier) time.

    An ensemble average of the macroscopic property G({q},{p}) can be defined in the following manner:

    (τ) exists is also referred to as the ergodic statement.)

    Postulate

    The above postulate states that we can consider ensemble averages rather than time averages as a basis for determining macroscopic properties from the microscopic description. Thus we are led to study in detail the properties of the probability density FN. As a final remark we note that the validity of the above postulates, here as elsewhere, must be ultimately measured in the absence of rigorous proof by how well the predictions based on them compare with experiment.

    1–4 THE LIOUVILLE EQUATION

    We are interested in determining how FN evolves in time due to the natural motion of each ensemble member in the phase space. The Hamilton equations, Equation (1–2), determine how the phase of each system changes in time, and so we would expect that they will play a central role in determining the temporal behavior of FN as well.

    Let us consider the change, dFN, in the value of FN at the phase point {x} ≡ {q}, {p} at the time t which results from arbitrary, infinitesimal changes in these variables. We will have

    This expression must be valid for all infinitesimal changes {x} → {x + dx}, so that we can specify that this infinitesimal change be just that which follows the trajectory of the system in the phase space. We then have (dqi/dti, (dpi/dti, and for this case Equation (1–6) can be rewritten as

    In the above equation ∂FN/∂t is the local change of FN, that is, the change at the point {x}, and dFN/dt is the total change of FN along the trajectory in the neighborhood of {x}. Liouville’s theorem now states that

    The statement of Liouville’s theorem, Equation (1–8), is referred to as Liouville’s equation. Before proving this result let us first consider its implications. In words, Liouville’s theorem states that along the trajectory of any phase point the probability density in the neighborhood of this point remains constant in time. From this it follows that an incremental volume about the phase point x, whose surface S(x) is defined by a particular set of phase points, is also invariant in time. Thus as the point x, moves along its trajectory, the volume dx defined by S(x) changes in shape, but its total volume remains constant. This is easily seen as a consequence of the fact that Hamilton’s equations have unique solutions, so that there can be no intersection of the trajectories of separate ensemble members in the phase space. The points inside dx can never cross the surface S(x), then, since this would mean that their trajectories intersect those of the points which define S(x). Since both FN and the number of points inside dx remain constant in time, we can conclude that the volume of dx must also remain unchanged. A formal way of stating the above is to say that the transformation of the phase space into itself induced by the flow described by Hamilton’s equations preserves volumes.

    Since FN remains constant along a trajectory in the phase space, any function of FN will also have this property. Of particular interest will be the function

    We will return to this function later when we discuss the ideas of reversibility and irreversibility. In this context we mention here that Liouville’s equation is an example of a reversible equation. By this we mean that the transformation t →−t leaves the form of the equation unaltered so that if FN({q(t)},{p(t)},t) is a solution so is FN({q(−t)},{−p(-t)}, −t).

    1–5 PROOF OF LIOUVILLE’s THEOREM

    We consider here a simple proof of Liouville’s theorem. Consider an arbitrary but fixed volume V in the phase space, and let its surface be denoted as S. The number of phase points inside V at the time t is ηv, where

    Since V is fixed, we have

    We can also obtain an expression for dηv/dt, the net rate of change in the number of phase points inside V, by equating this quantity to the net flow of phase points passing through S. } denotes the 6NNN is the unit normal at S, then we have

    points out of S. The surface integral in the above equation can be written as a volume integral by making use of a generalization of Gauss’s theorem, so that

    where ∇x is the gradient operator in the phase space. Since V was arbitrarily chosen, we can combine Equations (1–9) and (1–11) to obtain the following result:

    Writing out the gradient operator and flow vector we then have,

    From Hamilton’s equations we see that the last term is identically zero:

    so that Equation (1–12) reduces to Liouville’s equation

    which completes our proof of Liouville’s theorem.

    At this point one might reasonably wonder why we have gone to all the trouble of introducing the Gibbs ensemble and ensemble density FN. We are certainly still faced, in the presence of the Liouville equation, with an impossible computational problem due to the overwhelmingly large number of degrees of freedom present. This question will be answered in the next chapter, where we consider a more tractable description which we obtain directly from the ensemble formalism.

    References

    The standard reference on Hamiltonian mechanics is:

    1. H. Goldstein, Classical Mechanics. Reading, Mass.: Addison-Wesley, 1959.

    The statistical mechanical preliminaries are elaborated upon in:

    2. T. L. Hill, Statistical Mechanics. New York: McGraw-Hill, 1956.

    3. A. I. Khinchin, Mathematical Foundations of Statistical Mechanics. New York: Dover, 1949.

    4. R. Tolman, The Principles of Statistical Mechanics. Oxford, England: Oxford University Press, 1938.

    5. G. E. Uhlenbeck and G. Ford, Lectures in Statistical Mechanics. Providence, R.I.: American Mathematical Society, 1963.

    6. R. L. Liboff, Theory of Kinetic Equations. New York: Wiley, 1968.

    Problems

    1–1. The cannonical distribution function of equilibrium statistical mechanics is FN = Ae-βH, where A and β are constants and H is the Hamiltonian of the system being described. Determine the constant A in terms of β and H.

    1–2. Show that FN = Ae-βH is a stationary solution of Liouville’s equation (A, β, and H are defined in Problem 1–1).

    1–3. Consider the trajectory of a specified system in the gamma space. On this trajectory we label points Γ0, Γ1, ..., Γn with n → ∞ in such a way that it takes the system the same amount of time, say τ;, to move from each Γi to Γi(t), for any function G of the phase coordinates on this trajectory. By replacing the integral which appears in the time average by a sum, show that this time average is equal to the ensemble average when the ensemble is prepared so that the states of the ensemble members are given by the Γi.

    1–4. Prove that the solution to Hamilton’s equations are unique for a given Hamiltonian; that is, if two trajectories in the gamma space pass through Γ(t + Δt) corresponding to the state of the system at time t + Δt, then they both must have originated from the same Γ(t).

    1–5. An incremental volume in the gamma space, dΓ, is transformed into the volume dΓ′ through a flow described by Hamilton’s equations. As discussed in Section 1–4, dΓ = dΓ′. Prove this for a system of noninteracting particles by showing that the Jacobian of the transformation is unity.

    1–6. Write Liouville’s equation for a system which is acted on by a time-independent external field.

    1–7. The Liouville equation can

    Enjoying the preview?
    Page 1 of 1