Вы находитесь на странице: 1из 5

Discretization and the Continuum Hypothesis: A Creative Opposition

Adam Taylor, EGM 6611

The continuum hypothesis in mechanics, at the foundation of classical solid and


fluid mechanics, is now known to be an entirely approximative assumption
employed to simplify and make tractable various mathematical models of nature.
This realization has occurred relatively recently in the history of science. The debate
is almost as old as natural philosophy itself, with a rich record dating to antiquity;
the ancient atomists, including the likes of Epicurus and Lucretius, believed nature
to be nothing else than the sum of a huge number of discrete fundamental particles
bouncing around in a void. However, the Platonic idea that the void is, by definition,
nothing, and that nothing cannot be said to exist, was adopted by Aristotle, whom
went on to describe a physical model of the universe which essentially served as the
canonical model in the western world for the better part of two millennia; and so,
“nature abhors a vacuum” served as a rule to even the likes of the revolutionary
scientists Galileo, Pascal, and Descartes, who saw no reason to believe that matter
wasn’t omnipresent and everywhere continuous.

Einstein (1905) is often cited as giving the first widely accepted theoretical proof of
the existence of atoms and molecules by way of his studies into Brownian motion;
the following physical revolution made certain the idea that, at its most fundamental
level, nature is essentially discrete. All matter is composed of atoms, which in turn
are composed of subatomic particles, but mostly empty space. The continuum
hypothesis, however, has proved indispensible in mechanics for a number of
reasons. For many physical systems, it is reasonable to completely ignore the
molecular and atomic make-up of the structure under analysis; mathematically, the
assumption of a continuum allows for infinitesimal divisions of an object, permitting
the application of analysis and governing differential equations with continuous
solutions. Discrete mechanics requires either the interactions of mathematical
points of matter under the influence of continuous field, or of the application of
statistical methods. One might think that the mechanics of the discrete and the
continuous therefore naturally provide a clean partition of mechanics as a whole.
This is not the case. Maugin (2014) describes much of his work in Generalized
Continuum Mechanics as lying “between two a priori antagonistic occupations,”
being the discretization and continualization of specific physical problems. Instead
of these occupations being mutually disjoint, this author describes the two activities
as a “two-level back-and-forth mutual enriching,” as a dialectic approach which
motivates forward expansion in the single field of GCM. There have been numerous
attempts to blur the lines between these two approaches. Attempts have been made
to show that fundamentally discrete models (those that include a specific number of
discrete particles) obey the governing equations of continuum mechanics. On the
other hand, models that are fundamentally based in the continuum hypothesis have
included generalizations that phenomenologically mimic the microstructure of the
material. We will here recount some basic results of two specific examples, one for
each of these scenarios.

Irving and Kirkwood: The Statistical Equations of Hydrodynamics

In 1950, J. H. Irving and John G. Kirkwood published a seminal paper in which the
authors derived the classical (continuum) equations of hydrodynamics from the
principles of statistical particle mechanics; in particular,

The Equation of Continuity,


r ( r, t ) = -Ñr × [ r (r, t)u(r, t)]
¶t
the Equation of Motion,


¶t
[ ru] + Ñr × [ ruu] = X + Ñr × s

the Equation of Energy Transport,


E + Ñr × [ Eu + q - u ×s ] = 0
¶t

These are the classical equations derived from the classical theory of fluid dynamics,
where of course the fluid material is treated under the continuum hypothesis. The
authors treat a statistical-mechanical system of N molecules, each with 3
translational degrees of freedom. The instantaneous state of such a system is
described as a point in a 6N-dimensional phase space, representing the 3N
coordinates of the particles contained in position vectors (R1, R2..., RN ) and the 3N-
componenets of their momenta vectors ( p1, p2 ..., pN ) . It is derived that for an
arbitrary dynamical variable a ( R1...RN , p1...pN ) , the rate of change of the expectation
value of a is
¶ N
p
a ; f = å k ×ÑRka - ÑRkU ×Ñpka ; f
¶t k=1 mk

Note the use of the ordinary inner product within the phase space (there will be a
total of 6N integrations). This relation is then employed in showing that the
governing equations mentioned hold true for the statistical system of particles
analyzed in the phase space. We give as an example the derivation of the continuity
equation. Setting
N
a = å m jd (R j - r)
j=1
where d is the Dirac Delta-Function, we see that the following equality holds.
pk
×ÑRka - ÑRkU ×Ñpka = pk × ÑRkd (Rk - r) = -Ñr × [ pkd (Rk - r)]
mk

Then, from the derivation for the rate of change of expectation value of a , we see
that
¶ ¶ N
r ( r, t ) = a ; f = å -Ñr × éë pkd ( Rk - r )ùû ; f = -Ñr × [ r (r, t)u(r, t)]
¶t ¶t k=1

Then it is proved that the continuity equation is satisfied for the system of discrete
particles when analyzed statistically through the method of a phase space. The other
equations mentioned are proved valid by Irving and Kirkwood in a similar manner.
This result is our first example of a mutual enriching between the processes of
“discretization” and “continualization” as described by professor Maugin. These
results are comforting, in a sense. The general governing equations discussed, while
formulated as differential equations of functions of continuous variables, in fact
express laws of nature which are empirically verifiable. Since it is known that nature
is in a sense discontinuous or discrete, it is quite well that our discrete mathematical
models obey these laws as accurately as those that are constructed under the
continuum hypothesis.

Mindlin: Micro-Structure in Linear Elasticity:

We have presented an example of a mathematical model of discrete molecules being


made to obey the equations of continuum mechanics; we now seek to present an
example of the opposite effect, a model derived under the continuum hypothesis, yet
which accounts for microscopic effects. Mindlin (1963) presents an elegant model of
a linear- elastic, three-dimensional continuum which exhibits properties of a crystal
lattice. The inclusion of the phenomena correlated to this microstructure, (labeled
as micro-displacement, micro-deformation, micro-rotation, micro-strain ect.) are
included in the theory by way of the idea of a “unit cell.” These are often, in other
literature, referred to as Representative Elementary Volumes (REVs) and represent
the smallest volumes of the model which are representative of a specific behavior on
a macroscopic scale.

For a material volume V with material coordinates Xi and spatial coordinates xi , we


have the usual continuum displacement components

ui = xi - Xi

However, in each material particle (point) there is thought to be a “micro-volume” V’


with corresponding micro-scale material and spatial coordinates parallel to the
macroscopic coordinates with an origin fixed in the “particle.” There is therefore a
corresponding micro-displacement defined as

u'i = x'i - X 'i

Then, as Mindlin describes, we have, along with the usual linear-elastic strain e ij
there exists a micro-strain y (ij ) , the symmetric part of the micro-deformation tensor
y ij (using Mindlin’s notation). Corresponding to these definitions are the relative
deformation g ij = ¶iu j - y ij and the micro-deformation gradient xijk = ¶iy jk .

It can be imagined that the constitutive equations for this system will vary
considerably from the classical, “purely macroscopic” linear-elastic continuum.
Mindlin derives a potential energy density function from the above motion-
deformation components, which requires 18 independent coefficients, of the
following form:

1 1 1 1
W = le ii e jj + me ij e ij + b1g iig jj + b2g ijg ij + b3g ijg ji + g1g ii e jj
2 2 2 2

( ) 1
2
1
+g2 g ij + g ji e ij + a1 xiik xkjj + a2 xiik x jkj + a3 xiik x jjk + a4 xijj xikk + a5 xijj xkik
2
1 1 1 1 1
+ a8 xiji xkjk + a10 xijk xijk + a11 xijk x jki + a13 xijk xikj + a14 xijk x jik + a15 xijk xkji
2 2 2 2 2

The strains on different scales will of course correspond to different stresses


obtainable by way of this function; we see that, if

¶W ¶W ¶W
t ij = t ji = , s ij = , mijk =
¶e ij ¶g ij ¶xijk

we can derive the elastic constitutive equations of this micro-structure exhibiting


linear-elastic material, analogous to the generalized Hooke’s Law:

t pq = ld pqe ii + 2 me pq + g1d pqg ii + g2 (g pq + g qp )


s pq = g1d pqe ii + g2e pq + b1d pqg ii + b2g pq + b3g qp
m pqr = a1 ( xiipd qr + xriid pq ) + a2 ( xiiqd pr + xirid pq ) + a3 xiird pr + a4 x piid qr + a5 ( xqiid pr + xipid qr )
( )
+a8 xiqid pr + a10 x pqr + a11 xrpq + xqrp + a13 x prq + a14 xqpr + +a15 xrqp

These constitutive equations describe a continuum with macroscopic material


behavior which will clearly differ substantially from a classical linear-elastic
material; the differences are caused by the attempt to model the continuum as if it
had a specific microstructure.
It is clear that the notions of “discretization” and “continualization” appear, at first
glance, to be fundamentally antagonistic. The imagination strains to conceive of a
material which exists as a continuum containing discrete material points with their
own physical properties; Maugin himself introduces the idea of “a grain-like
structure in a continuum” but playfully hints that such a notion may be oxymoronic
or heretical. However, we have seen examples which show that such ideas can be
fruitful. We have seen from Irving and Kirkwood that fundamentally discrete
systems, as all those in nature truly are, must also obey the governing equations
employed in continuum mechanics. We have also seen an example of a continuum-
model which attempts to simultaneously capture the influence of a microstructure.
Approaches of the latter type can be thought to be pragmatic, in a sense; the model
is specifically created to capture desired phenomenon regardless of possible issues
of philosophical consistency. If it is true that there exist no true continua in nature,
then the continued adoption of the continuum hypothesis in the first place shows
this same kind of pragmatism. And if this is the case, is there any more reason to
object to modeling microstructure in GCM than there is to the employment of a
classical Euler-Bernoulli beam?

Works Cited:

1. A. Einstein, Investigations of the Theory of Brownian Movement, (Dover, 1956).

2. Gerrard A. Maugin, Some Remarks on Generalized Continuum Mechanics,


Mathematics and Mechanics of Solids, 2014

3. J. H. Irving, John. G. Kirkwood, The Statistical Mechanical Theory of Transport


Processes. IV. The Equations of Hydrodynamics, The Journal of Chemical Physics, 18,
6, 817-829 (1950)

4. Mindlin, R.D., Micro-structure in Linear Elasticity, Arch. Rat. Mech. Anal. 16,51-78
(1964)

Вам также может понравиться