The waveequation

By ENDERS ROBINSON and DEAN CLARK

T he foundation of seismologyis the theory of wave motion,

a complicatedconceptthat is still - after centuriesof experi-
ments and speculationsby many of the very greatestscientists
- an area of active researchin many disciplines.Even simple
forms of wave motion are difficult to describeverbally; but,
ironically, the simplest type of wave is remarkably easy to
describe(and subsequentlyanalyze) mathematically.
This is one of thoseareaswhere,in the wordsof Nobel Prize
physicistStevenWeinberg,mathematicshas a “spooky” corre-
lation to the physicalworld. Although somenaturally occurring
crystalshaveperfectgeometricshapes,right trianglesare a purely
mathematical concept.They exist outside our ordinary experi-
enceof the physicalworld. Have you everfound a perfectlyright-
triangular rock, or blade of grassor leaf in your back yard or
on a field trip? Yet we rememberfrom elementarytrigonometry
(the mathematical analysisof the propertiesof triangles) that
the graph of the sine function - nothing more than the ratio
of two sidesof a right triangle - perfectly representscertain
periodicmotions,suchasthe (small) oscillationsof a pendulum.
This type of sinusoidalmotion is calledsimpleharmonicmotion.
The pure sine curve, u = sin x, is quite restricted.The value
of u can never be greater than 1 or lessthan -1 and x must
traversea distanceof 2t radians before one cycleof motion is
completed.Theselimitations are, however,not serious.The sine
function is easily tailored to representany regularly repeating
motion no matter what its height/depth (or amplitude), its fre-
quency of oscillation, or its value when it crossesa “starting”
point (often the x = 0 line). Such an all-purpose sine function
can be written, supposingu to be the disturbancecausedby the
motion. as
u = A sin 27r ?- - c
x T >
wherex is distanceand I is time Five graphsof u vs.x are shown
in Figure 1. The number A (chosento be positive) represents
the amplitude;the distancebetweenconsecutivecrestsis X (called
the wavelength);the quantity T is the period or the time it takes
the waveto completeone cycle.The crestof the wave movesa
distanceX in time T Since X is a distance and T a time the
quotient X/T equals the wave’s velocity - almost always ex-

14 GEOPHYSICS:THE LEADING EDGE OF EXPIGRATION JULY 1987

pressedsimply as V. Therefore, it’.k 1~t1xt.dand I I\ allowed to
tary, a wave crest sweepspast the fixed poml with a propaga-
tion velocity given by 1:
In seismic work we usually plot wave motion as a function
of time for fixed value of x and varying values of 1. This results
in sine curves with a vertical (see Figure 2) orientation, rather
than horizontal as in Figure 1.
There are other useful ways in which we can write a sine func-
tion to representwavemotion. Instead of wavelengthand period,
we can use wavenumber (k) and frequency (w) where

and then the sinusoidal wave may be expressed Figure 3. Illustration of f(x - vt).
u = A sin (kx - wt)
which representsa simple harmonic progressivewave. We can the quantity k (expressedin radians per meter) is the angular
also write this curve as wavenumber, related to cyclical wavenumber x by the equation
k = Zax.
u = A sin k(x - vt)
We can generalizethis result without much difficulty. We note
because v = h/T = w/k. that the quantity x - vt reproducesitself when t becomesI + t’
The quantity w (which is expressedin units of radians per sec- and x becomes x + vt' because
ond) representsangular frequency. It is related to cyclical fre- x + vt’ ~ vf t+ 1’) =x - vt.
quencyf(expressed in Hertz) by the equation w = 27rf. Likewise
Therefore any function of x - vt can be said to representa wave.
An effective way to illustrate this is to imagine a taut string
lying on the x axis. If the string is displaced in any way perpen-
dicular to the x axis, then the shape of the resulting curve can
be written u = f(x). If the displacements alter in such a way
that the pulse travels with velocity v in the positive x direction
without change of shape, the equation representing the pulse
at any time t will still be u = f(x) . . . . . provided that we move
the origin a distance vt in the positive direction (see Figure 3).
In referenceto the old origin, the equation of the pulse will have
x replaced by x - vt or u = f(x-vt).
We can therefore state that this is the general equation of a
wavewith constant shapetraveling in the positive direction with
velocity v. Furthermore, everywaveof this type must be expressi-

I‘ q
37 -_-_ - - - - - - -
4 ble in this form. Similarly, a wavegoing in the oppositedirection
(i.e. negative x) is representedby a function u = g(x + vt).
---- - - - - -
T
The first great insight into the mathematical analysisof waves
was made by the legendary Greek mathematician and religious
leader Pythagoras who is believed to have died about 500 BC.
He discoveredthat the pitch of a sound from a plucked string
Figure 1. Sinusoidal wave motion as a function of distancex.
depends upon the string’s length, and that harmonious sounds
are given off by strings whose lengths are in the ratio of whole
numbers. However, significant additional progresswas impossi-
*x
I I I I I I I 1 I ble until the invention of calculus, more than 2000 years later,
permitted the English mathematician Brook l’&ylor to make the
I b-4 first productive attempt at the quantification of wave motion.
Consider a stretchedstring with initial shapef(x). According
to basic differential calculus,the slope of the tangent line at any
point representsthe rate of changeof the functionf with respect
to x. This rate of change is the first derivative of fwith respect
tox; in turn, the rate of changeof the slope (or secondderivative
off with respect to x) representsthe curvature of the function.
Now consider the motion of any particular point on the
string. We have seen that the traveling wave can be represented
asf(x - vt) which at the point x = 0 becomesmerelyf(-vt). That
point is moving up and down approximately at right anglesto
the x axis. Back to basic calculus. The point’s up-and-down
velocity is given by the first derivative offwith respectto t and
the point’s acceleration is given by the second derivative off
with respect to t.
When the string is in its equilibrium position (horizontal along
the x axis), there is no net vertical force acting on any point on
the string. However, when the string is curved, the tension
in the string exerts a restoring force which attempts to move it
Figure 2. Sinusoidal wave motion as a function of time t. back to its equilibrium position. The more the curvature, the

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION JULY 1987 I5

 1‘ he \+l~~c’
., cquatlor I< arrequation m spaceand time coordinates
1V,K;and 1) whereagthe related dispersion equation is an equa-
flkjn in wavenumber and t’frequencycoordinates (k,, k,, k,, and
2). C;eophysicistsusually want to look at the data in the familiar
\pace-time display: but it is often advantageous to transform
the data mto the wavenumber-frequencydomain for computer
procc~slng.The basisfor the transformation from one domain
to the other is the mathematical operation known as the Fourier
Small ~wvaturemeans Large
C”rYaf”re means
transform.
small restoring force large
resfarlng force
(One of the best treatments of the Fourier transform, as it
Figure 4. Force is proportional to curvature of string. relatesto exploration geophysics,is given on pages 10.6and 10.7
of Roy Lindseth’s book Digital ProcessingofGeophysical Data:
A Review Figures 6, 7, 8, and 9 are adapted from this book.)
The Fourier transform allows us to separate a seismic trace
into its individual frequency components. If many traces were
thus “broken down ,” we could createseismicsectionscomposed
of traces of only one, identical, frequency. Let us now assume
that we have done this and examine three situations, in each of
which we will let frequency (w) be constant,
I + , As everygeophysicistknows, a wavefront is a curveof constant
phase. If one follows the samecrestof wavemotion, the line con-
Nearlystratght
t,mepath Strongly curved tome path
meanssmall acceleration meant large acceleralion
necting these points makes up a wavefront. In each of the three
following examples, the wavefront will be a straight line but in
Figure 5. Acceleration proportional to the amount of bend in each case with a different angle to the horizontal. When a
the time path of a point on the string. wavefront strikes a horizontal line, the movement in the hori-
zontal direction gives rise to the apparent horizontal velocity.
greater this restoring force (Figure 4). Taylor noted this and If the wavefront is perfectly horizontal, then all points on it strike
reasoned that the restoring force is proportional to curvature the horizontal surfacesimultaneouslyso the apparent horizontal
and, as a student of Sir Isaac Newton, he knew that force is velocity is infinite. A geophysicalexamplewould be a deepreflec-
proportional to acceleration (Figure 5). Thus he wrote the tion approachingthe surfacevertically from depth. On the other
equation hand, if the wavefront makes an angle to the horizontal, then
curvature = (Y acceleration the apparent horizontal velocity would be finite.
In the first case,our section is flat, i.e. the wavefront parallels
where 01is a constant of proportionality. Taylor could not fully the .u-axis(no dip is present), which means each trace is identi-
develop the properties of this equation because he had no cal ( Figure 6 ). As a result, any waveprofile in the x direction
knowledge of partial derivatives. But after they were invented, would be a constant. If we measurethe amplitudes of the wave
his speculation was confirmed. It also turned out that the con- along any horizontal time line and then plot them againstvalues
stant of proportionality was 1/v2. In modern mathematical of .Y (see the top of Figure), our graph would be a perfectly
notation, this equation is written straight horizontal line. Since there are no oscillations.the wave-
ah I a%
-=-2
a2 lJ* at
and it is known as the one-dimensional wave equation. When
generalized to three dimensions, it governs the melodies of
Pythagoras,the propagation of seismicenergythrough the earth,
and all other wave motion. The three dimensional waveequation
is
a54 a% ak 1 a%
--+x+az’=v’at”
8x2
If we now go back to one of our original equations for simple
harmonic motion- u = A sin (kx- wt) - and take secondpar-
tial derivativeswith respectto x and t, and then substitutethese
i
secondderivatives into the one-dimensional wave equation, we
obtain Figure 6.

-Ak2 sin(kx- tit) = $ (- Ati2) sin (kx - wf).

After canceling common factors, we discover that

k2 = -“,f_

which is called the dispersionequation for the one-dimensional

wave equation. It relates wavenumber and frequency. When
extended to three dimensions, this equation becomes

k: + k: + kl = $

where k = Jk: + k: + kf is the wavenumberand k,, k,,, k, are

Figure 7.
the wavenumbercomponents in the three coordinate directions.

16 GEOPHYSICS:THE LEADING EDGE OF EXPLORATION JULY 1987

length is infinite and the correspondtngw,l\cnt:lnt)r.l (h) I, ~et,~,
This wave motion is representedon the uavenumber-flequeni:!
domain by the single point (k,o) = (0,~) becausek z 0 and
d represents our constant frequent!

Next consider a section of’ traces with the \ame constant fre-
quency (w) but where the wavefront has a \mall amount ot’dip
( Figure 7 ). The motion along the horizontal time line w*ould
look quite different from that of the flat-wavefront section. Each
of the vertical tracesis now slightly out of phasewith its adjoin-
ing traces.As a result, motion in the x-direction is not constant,
(a)Seismic
sectionwithhigh-velocity (b) Fouriertransformwith the two
but will have a low-wavenumber sinusoidal character. The cor-
reflectionsand low-velocityground dispersionlines.
responding low wavenumber is paired with the frequency and roll. Zeroingout
becomes a single point (k,w) on the graph. B groundmll
If the amount of wavefront dip is increasedand the frequency I dispersionline
remains the same as in the precedingtwo cases,the wavenumber
increases (Figure 8). Evidently, for any given frequency, the
wavenumber is related directly to the amount of dip of the
wavefront.
Our x-t graphs representtime sections;that is they are plots
of time vs. distance.Since the ratio of distanceto time produces
velocity, it is seen that the slope of the wavefront line (time/
distance) is the reciprocal of velocity. Thus any sinusoidal wave
motion which has the same wavefront dip will have the same
apparent horizontal velocity.
In a fourth example, the wavefront dip is held constant for
(d) Seismic section without the (c) Fourier translorm with the
any chosen wavenumber(k). Since we are holding the wavefront ground1011. ground-rolldispersionline zeroed
dip constant, it follows that we are holding the velocity (v) con- out.
stant. As we know, the frequency (w) is specified by the disper-
Figure 10. Pie-slice filtering.
sion equation, w = kv. Now let us look at the plot of LLvs.
k for a constant velocity ( Figure 9 ). The dispersionequation
saysthat this plot is a straight line with slope v. That is if we of the wavefronts are transformed into a single dispersion line
which goes through the origin in the w-k plane. Thus if we want
change W, we see that the wavenumber (k) falls on a straight
to wipe out all the waveswith this apparent velocity, all we have
line as shown in the figure.
to do is zero-out the appropriate dispersionline. This is the basis
This allows us to do some startling things in data processing.
of velocity filtering, which is also called pie sliceor fan filtering.
All the wavefronts with a given velocity are spread all over the
routine seismictime section. In other words, each wavefront has In Figure IO, we seefour diagrams (a), (b), (c), and(d) which
a different intercept with the horizontal (x) axis. However, if we are connected in a clockwise fashion by operations A, B, and
transform this data into the frequency-wavenumberdomain, all C. Diagram (a) showsa seismicsectionwith two reflected events
and three ground-roll events. These events criss-crossand that
would make separation difficult in the x vs. I form. However,
let us now exploit the fact that the reflected events have high
apparent horizontal velocity, while the ground-roll eventshave
low velocity. With (a) as input, the Fourier transform A gives
us (b) as outTut. In (b) the two reflected events appear as one
dispersion line through the origin, whereas the three ground-
roll events appear as a separate dispersion line through the
origin. The fact that dispersionlines alwaysgo through the origin
means that the two types of eventshave been separated.That is,
diagram (b) has the appearanceof a fan, or a sliced-uppie, where
one pie slice contains the reflected events and another pie slice
the ground-roll events. In operation B, we let the computer eat
+ up (i.e. erase) the pie slice containing the ground-roll disper-
,
sion line, thereby giving diagram (c). With (c) as input, the
Figure 8. inverse Fourier transform operator C gives us (d) as output.
As expected, the unwanted ground-roll events do not appear
in diagram (d), and we are left with only the desired reflections.

T he use of frequency-wavenumberanalysis is one of the most

powerful tools of seismic data processing. As we have seen,
measurementof wavenumberas a function of frequencyprovides
a reliablemeansof separatingand measuringthe variousvelocity
components on a seismicsection. Other important seismicpro-
cessingoperations (including dip moveout as well as migration)
can make use of w-k analysis. All these processingmethods are
tied physically to the dispersionequation, which in turn follows
from Taylor’s inspired insight (in 1715) - spatial curvature is
proportional to temporal acceleration- that led to the original
Figure 9. formulation of the wave equation. &

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION JULY 1987 17

 A waveat a boundary:
Reflection,transmission/refraction,
and diffraction
By ENDERS ROBINSON and DEAN CLARK

“Giveme a place to stand.” of probability and the famous

Archimedes said, “a lever and principle of wavemotion which
a rock, and 1 will move the bears his name.
Earth. ” To explain Huygens’ princi-
He very likely was the first ple, let us go back to the previ-
scientistto speculateabout mov- ous article in this series (The
ing the Barth, but modern ex- wave equation July 1987) and
ploration geophysicists were recall the three-dimensional
probably the first to make such waveequation. In the caseof a
a daring idea the very underpin- homogeneousisotropicmedium
ning of their systematicinvesti- (meaningthe wavevelocityis the
gations.For, indeed,moving the same at all places and in all
Barth - admittedly on a very directions), the wave equation
small scale, nothing like what yields a particularly simple and
Archimedes had in mind - is beautiful solution: the spherical
precisely what we must do to wave.Not surprisingly,a spheri-
generate the seismic waveswe cal wave is one in which the
need to bring us information successivewavefronts of wave
about the subsurface. motion emanating from a point
The part of the wavein which source are concentric spheres
geophysicists are mostinterested centered on the source. If we
is the wavefront; i.e., a surface slice through the three-dimen-
over which the phase of the sional physical reality to pro-
traveling wavedisturbanceis the same If we know the position duce a two-dimensional cross-section,the sphereswill become
of a wavefrontat a certaintime we can find its subsequentposi- circles(Figure 1).
tion at any later time by meansof one of the most elegant and
fundamental conceptsin seismology- Huygens’ principle. H uygens’principledatesfrom 1690and thus significantlypre-
The greatDutch scientistChristiaan Huygens(16294695)was cedesthe development of the wave equation, but Huygens’
one of the most formidable intellects who followed in the insight was so profound that his principle actually makes use
immediate wake of Galileo. Huygens had first-rank, possibly of the so-calledGreen’s function in the subsequentlyconfirming
genius, talents in many disciplines- astronomy,optics, and mathematics.The basisof Huygens’ conceptis that each point
mathematicsin particular. His contributions ranged from the along a wavefront may be viewed as a point sourcethat pro-
most far-teachingtheoreticalspeculationsto extremelyimportant ducesa secondarysphericalwavelet (known as the Huygens’
advancesin contemporary technology.Among the latter were wavelet)which propagatesaway from the point in all directions.
the perfection of the pendulum clock and innovations in lens In Figure 2, curve AB representsthe instantaneousposition of
grinding (which led to his discoveryof the ringsaround Saturn). a wavefront.In order to find the location of the wavefrontafter
Among the former weresomeof the basicconceptsin the theory an interval of time (designatedby the customaryAf), we draw

38 GEOPHYSICS:THE LEADING EDGE OF EXPLORATION SEPTEMBER1987

 circles of radius v A1 (v, of course, is the wave velocity) about
a seriesof points in AB. The sum, or envelope,of these second-
ary wavelets is the new wavefront A’B’. The envelope is the
surface tangent to all of the secondary spheres.
This technique, known as Huygens’ construction, is one of
the most elegant productsof classicalphysics.But, as usual, the
physical reality is more complicated than the theory because
the logical conclusion of Huygens’ theory is that two, not just
one, disturbanceswould be propagated - one on each side of
the wavefront and traveling in opposite directions. However, it
is common knowledge that disturbance occursonly in the “for-
ward” direction of wave propagation and that the theoretical
“backward” disturbance, predicted by Huygens’ principle and
shown by dashed line A”B” in Figure 2, does not appear. Why
not? It took nearly 200 years to obtain a completely satisfac-
tory solution to the mystery. Fresnel took a major step toward
clearing up this thorny question when he proposed, in 1826,that
the backward wave did not occur becauseof destructive inter-
ference effects; 40 years later Kirchhoff developed an integral
solution of the wave equation which shows that the secondary
waveletsfrom the point sourceson the wavefrontdo destroyeach
other by mutual interference excepton the forward propagating
wavefront A’B’.
Huygens’ principle is particularly valuable when we want a
graphical way to explain the three fundamental ways in which

Figure 1. Spherical wave.

A ;A’

Radius =’ vt

Figure 2. f&ens’ construction. Figure 3. A wave at a boundary.

GEOPHYSICS:THE LEADING EDGE OF EXPLORATION SEPTEMBER 1987 39

 S
7 Source
’ Image source
k,
Figure 4. Reflection of a spherical wave.
Figure 5. Law of reflection.
the direction of a propagating wave can change: reflection,
refraction, and diffraction (Figure 3).
I n most cases,when a wave arrives at a boundary between two
different media, part of the wave’s energy is reflected back into
the original medium and part of it is transmitted into the new
medium but at a different direction (i.e., refracted). The study
of reflection and refraction is facilitated by the use of rays. In
an isotropic medium, a ray is a line that is everywhere perpen-
dicular to the successivewavefronts. While seismicenergy does
not travel only along raypaths, the greater part of the energy
does indeed follow them. (Some seismic energy would reach a
point by diffraction even if the raypaths between the point and
the energy source were blocked.) Raypaths therefore constitute
a very useful method of studying wave propagation. This
disciplineis called geometrical seismicsor geometrical acoustics.
It is the geophysical counterpart of the well known subject of
geometrical optics.
(There is an interesting facet of geometrical seismicswhich
does not appear in the textbooks on geometrical optics. Because
light occurs at such high frequencies and travels at such a high
speed, the waveforms of light are not measured as a function
of time As a result,geometricaloptics dealsonly with the spatial
paths of light rays. However, in seismic work, we do routinely
measurewaveformsand the seismicsectionis the resultingspace-
time representation.In geometrical seismics,therefore, we must
consider space-time paths in addition to purely spatial paths.
As we will see in these articles, this space-time feature adds a
whole new dimension to the classicray theory as found in books
on geometrical optics and it was a decisive factor, via
40 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION SEPTEMBER
Figure 6a. Snell’s law of refraction.
Figure 6b.
his study of light, which propelled Einstein to the theory of
relativity.)
LRt us initially apply Huygens’ principle to a wavefront that
is completely reflected from a plane surface and see how it
specifiesthe direction and curvature of the reflected wavefront.
Figure 4 illustratesthe reflection of a sphericalwave. The surface
ABC representsthe hypothetical position of the wavefront if
the plane reflecting surface were not present. When the wave
reachesP (the nearestpoint of the reflecting plane to the source
S), the point P becomes the origin of a secondary wavelet. At
immediately succeedinginstants, the adjacent points to the left
and right of point P are struck by the incident wavefront, and
they in turn also become sourcesof secondary wavelets. The
totality of the secondarywaveletsemitted by the successive points
on the reflecting plane has the spherical surface AB’C for its
envelope.We notice that the reflectedwavefront appearsto come
from some point S’ behind the reflecting plane; in fact, it’s a
simple matter to show that S’ is the mirror image of the source
point S.
N ow let’s use Huygens’ principle to derive the direction of a
reflected ray that is not perpendicular to the reflecting surface.
In Figure 5,O is the angle of incidence(by definition, the angle
that the incident ray makes with the normal to the reflecting
interface). The corresponding angle of reflection is denoted 0’.
The incident wavefront falls along the line AB, which is shown
in Figure 6 at the instant point B strikes the reflecting surface.
Some of the energy in the incident wave at point B will reflect
toward point D. As the incident wavefront continues toward the
interface, each point between B and C will successivelyserve
1987

Interface
High velocity v2
medium
Figure 7. Critical angle.
as a point source for secondary wavelets which will propagate
back into the original medium. When the incident wavefront
reaches point C, the reflected wavefront will be on line DC.
The speed of the reflected wave is the same as that of the
incident wave, making the length of line AB equal to that
of line CD. Trigonometry can then be used to establish one of
the cornerstonesof seismictheory; i.e., that 8 = 8’, or the angle
of incidence is equal to the angle of reflection.This is called
the law of reflection.
Of course,in most cases,all of the energywill not be reflected
from the interface but some will be transmitted into the second
medium. When the wave(not at verticalincidence)entersanother
medium with a different speed, it will change direction -- or
undergo refraction. Figure 6a illustrates Huygens’ construction
of a refracted wave.
We supposethat the velocity ~2in the lower medium is greater
than velocity y in the upper medium. When the incident wave
reachesthe interface at B, the wavelet radiating from B into the
lower medium travels faster than the wavelet moving from A
toward C. Wavelets from successiveportions of the wavefront
entering the lower medium will have longer radii in a given time
interval than those traveling in the “slower” upper medium.
Thus the refractedray will bend away from the vertical; the angle
of refraction I!&will be greater than the angle of incidence.
This concept is further illustrated in Figure 6b which shows
the refraction construction from Figure 6a overlain by the
hypothetical propagation of wavefront AB as it would have
occurred in the absence of interface BC. Figure 6b makes it
apparent that, if the lower medium is faster, the angle of refrac-
tion must be greater than the angle of incidence. Indeed, one
of the basic theorems of Euclidean geometry makes it apparent
that 0, could equal & only if wavefront AB maintained an iden-
tical direction of propagation as it passedthrough interface BC.
Huygens’ constructionfor refraction leadsto an obviousques-
tion. What is the mathematical relation between the incident
and refracted angles?(Geophysicists,of course, know that one
exists . becauseif it didn’t, seismic investigations would be
much, much harder.)
R efer back to Figure 6a. In the time At, that it takesa wavelet
to travel from point A to point C, a waveletin the lower medium
will travel from point B to point D. Sincedistanceequalsvelocity
times time we immediately see that
AC = v, At and BD = b At.
Since triangles ABC and BCD are borh right triangles and
GEOPHYSICS: THF I f,AIIIN(; EDGE OF EXPLORATION SEPTEMBER 1987 41
Source
-v--/so;
Interface
High velocity medium
Figure 8. Head wave in a refraction survey.
respectively contain angles 8, and B,, we see that
sin 8, = AC/BC and sin 0, = BD/BC.
Substituting for AC and BD, respectively,we get
sin 0, :m v, At/BC and sin 0, 7 y Al/BC.
Ehminating At:BC we get
<in fl,:‘v, = sin H2/\b,
which is known as the law of refractionor Snell’s law.
(Although this proof of Snell’s law seemsremarkably simple
to modern eyes, it somehow eluded mathematicians - unlike
the related law of reflection which was known at least as far
back as the Greeks - until relatively recent times. Science
labored for about 1500 years under a false theory, offered by
the famous Claudius Ptolemy of Alexandria, that the anglesof
incidence and refraction, and not their respective sines, were
mathematically joined. The first publication, a slightly incor-
rect one, of the modern law did not come until 1637. Surpris-
ingly, the author was not Willebrord Snell but RenCDescartes.
Snell, reportedly after yearsof work, discoveredthe correct ver-
sion of the law in 1621but did not publish it. Fellow Dutchman
Huygens, born three years after Snell’s death, saw this work -
which subsequentlydisappeared - and it was due to Huygens’
efforts that Snell wa$ given his deservedscientific immortality.
Snell had a second profound effect on geophysicalexploration
because he developed a method of determining distances by
trigonometric triangulation and thus was one of the most
important pioneers of scientificmapmaking. Snell’s law can atso
be derived, even more easily in fact, via differential calculusand
that derivation is a prime example of the astounding analytical
power of that mathematical tool. See Fermat and the principle
sf least time TLE, December 1986.)
In refraction from a slower medium to a faster medium (the
usual case in seismicexploration), the rays turn away from the
normal. What happens when sin 8, y > y? Snell’s law predicts
that sin 0, > 1, an impossibility. However, the theory does not
predict an impossibility. Instead, it predicts that suchrefraction
ir; impossible. In fact, when o1 = 8, with 0, defined by sin 0,
._ v,115,Snell’s law says that sin 02 = 1 and therefore 02 =
90 degrees: such an angle of incidence 0, is called the critical
angle and the refracted ray grazes the interface. See Figure 7.
For angles of incidence greater than the critical angle, there
is total internal reflection; i.e., all of the energy is reflected at
the boundary back into the slower medium. The exploration
technique called “amplitude with offset” makes use of this
property. Since total reflection occurs at anglesgreater than the
 OIOO-

0 200 -

03w-

04CQ-

0 500 -

06M)-

0 700 -

om-

0 900 -

1 000 -

1 loo-

Irn-

1 300 -

1 400 -

15m-

1.600 -

1700-

lrn-

t 900 -

2000-

2100-

2200- Figure10.Huygens’construction of a diffracted wavefront show-

2.340 - ing how the wavefront A’B’ curls in behind the obstacle.
2.400 -
obstaclehasa lengthapproximatelyequalto X, then the amount
2.5w -

2.6M) -
2.7w -
2.8W -
2.9w - wavelengthof 1.3m. That is comparableto room dimensions,
3cco-
Figure 9. Diffractions generatedby termination of three flat
reflectors.
The diffractioncurvaturebecomes smallerwith depth.
The diffraction bnach under the reflection (the backward
branch)is the inverseof the bmnch which extqds beyondthe
reflection(theforwardbranch).The cmstof the diffractioncurve
locates the diffracting point and the diffraction curvature
dependson the depth and the velocity abovethe diffracting
point.The &l&ion ampMudedecmases to one-halfat the point
where the reflection is tangent to the diffraction curve, and
diffraction-curveamplitudeis antisymmetricabout this point
of tangency.Amplitudesand waveshape are continuousat the
point of tangency.(CourtesyChevron Oil Co.)
criticalangle,reflectedwavesthat originatedfrom distantsources
will havegreateramplitude than thosefrom sourcesthat are
“inside” the critical angle.
T he wavewhich travelsalong the interface,after the incident
wavestrikesat the criticalangle,is knownasthe headwave.See
Figure 8. As the headwavetravelsalong the interface,it con-
tinuously feedsenergyback up into the slowermedium. This
escaping energyleavestheinterfaceat the sameangle,thecritical
angle. Naturally this escapingenergycan be detectedat the
surfaceby geophonesand this is the fundamentalprincipleof
seismicrefractionprospecting,the most important methodof
geophysical explorationfor petroleumin the 1920s.It is still used
but long agoyieldedits dominantpositionin explorationto the
refledtionmethod.
When wavespassaroundan obstacleor throughan aperture,
they tend to curl around the edgesso that the shadowof the
obstacleon the downstreamside is not sharplydefined. This
aspectof wavebehavioris calleddiffraction. Diffractedsound
wavescan be heardaround corners,and water wavesentering
a harborspreadinto theareabehindthe breakwater. The amount
of diffraction can be qualitativelydeterminedby the ratio of
thelinear dimensionsof the obstacleto the wavelengthX. If the
of diffraction is large(i.e., many waveseasilybend aroundthe
obstacle),so that the notion of a shadowzone becomesmean-
ingless. For example, a note at middle C (264 Hz) has a
so a sound from around a corner is audible. However,if the
obstaclehasdimensionsvery muchgreaterthan X, the diffrac-
tion into the hiddenregionbecomesnegligibleand the shadow
of the obstacleis relativelysharp.Geophysicists are confronted
with diffracted eventseveryday. A textbookseismicexample
wouldbe the caseof a sharpgeologicdiscontinuitysuchasthe
abrupt termination of a reflectorat a fault plane.We can see
the hyperbolicdiffraction curvegeneratedby the discontinuity
on the seismicsection,but the seismicdatais still sharpenough
to allow delineationof the fault plane (Figure 9).
Diffraction allowsus to accountfor the penetrationof wave
motion into regionsforbiddenby geometricalseismics. The law
of reflectionand Snell’s law are not requiredto hold for dif-
fractedrays.However,Huygens’principledoesremainin force.
Its explanationof the movementof diffractedwavesaroundan
obstacleis shownin Figure 10.
In fact, this use of Huygens’ principle, in particular in its
embodimentin Kirchhoff’ssolutionof thewaveequation,might
be interpretedasbasic In thissense,a reflectionmay be thought
of as the interferenceresultof diffraction from pointslying on
the reflector.In otherwords,eachsourceof a secondarywavelet
may be considereda diffractionsource. . . and Huygens’con-
structionrepresents the resultinginterferencepattern.Although
eachindividual diffractionpoint on the reflectingsurfacedoes
not obey the law of reflection(sincethe diffractedraysgo in
all directions),the resultingenvelopdoesyield a wavefrontobey-
ing the law of reflection.
I n conclusion,we hesitantlyoffer a conjectureaboutHuygens’
principle. . . hesitantlybecausesomeseismologists might con-
siderit anathema.Huygens’abilitieswerenot limitedto science.
He wasalso an accomplishedartist and examplesof his work
are reproducedat the beginningof this article; they appeared
in the original publicationof Huygens’Thift?deIa lumith and
representhis own graphicalconceptionof Huygens’principle.
Could it possiblybe that suchan elegantlybeautiful explana-
tion for an ever-presentnaturalphenomenonresultedmorefrom
Huygens’ artistic sensibilitiesthan from rigorous scientific
analysis? g

42 GEOPHYSICS:
THE LEADING EDGE OF exploration SEPTEMBER1987
 Elasticity:Stressand strain
By ENDERS ROBINSON and DEAN CLARK
E lasticity is the property that enablesa fluid or solid body to
resistchangein sizeand shapewhen an externalforceis applied
and to return to its original size and shape when the force is
removed.This conceptis a major building block in seismology
becauseit is the elasticpropertiesof rockswhich allow seismic
wavesto propagatethrough the earth.
The theory of elasticity is one of the major achievementsof
classicalphysics.Its architectsincludemany of the colossalscien-
tific figuresof the 17th, 18th and 18th centuries,among whom
are Robert Hooke, Gottfried Wilhelm Leibniz, JamesBernoulli
and his nephew Daniel Bernoulli, Leonhard Euler, Thomas
Young, Charles Augustin Coulomb, Augustin Louis Cauchy,
Claude Louis Marie Navier, Simeon Denis Poissonand George
Gabriel Stokes.The framework, which they and others con-
structedto quantify and analyzeelasticity,derivesfrom the basic
conceptsof classicalmechanicsknown as stressand strain and
the mathematical linkage known as Hooke’s Law.
Stressesare forcesper unit area that are transmitted through
a material, i.e., forcesexertedby one part of a body on a neigb-
boring part. Stressesthat act perpendicularly to a surfaceare
Figure 1.
16 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION FEBRUARY
normal stresses;thosethat act parallel to it are shear stresses.
As an example,considerthe forcethat actsat the baseof a col-
umn of rock at depth z (beneath the ground level) to support
the column (seeFigure 1). The weight of the column of cross-
sectionalarea AA is pgzAA, whereP is the constantdensityand
g is the accelerationof gravity. This weight must be balanced
by an upwardsurfaceforce a&4 distributedon the horizontal
surface element of area AA at depth z. Here we assumethat
thereare no vertical forceson the lateral surfacesof the column.
The quantity a., is thus the surfaceforce per unit area acting
perpendicularly to the horizontal surface & that is, uZrris a
normal stress.In equilibrium the opposing forcesmust be the
same, so
0.2 = P&z
This normal stressdue to the weight of the overlying rock or
overburdenis known as the lithostatic stress.
Verticalsubsurfaceareasalso receivenormal stresses. The nor-
mal stressacting in the x direction on a plane perpendicularto
the x direction is a,. The horizontal normal stresscomponents
Figure 2.
1988
urr and uYvcan include large-scaletectonic forces, in which case
u, # mYY # err. However,there are instancesin global geophysics
in which rock has been heated to sufficiently high temperatures
or was initially sufficiently weak so that the three normal stresses
are each equal to the weight of the overburden. When the three
stresses are equal, they are referred to as the oressure. This
balance between pressure and the weight of the-overburden is
called a lithostatic state of stress.Likewise, hydrostatic equilib-
rium can exist in the sea, where pressureforcesare exertedequally
in all directions and pressure increases linearly with depth.
Of course, forces can also act parallel to an area. Consider
the forces acting on the element of area AA lying in the plane
of a strike-slipfault (seeFigure 2). The normal compressiveforce
u&4 acting on the fault face is a consequence of both the
Figure 3.
Figure 4.
GEOPHYSICS:
weight 01 the overburden and the tectonic forces which push
the two sidesof the fault together. The tangential or shear force
oVr is the frictional resistancethat opposes the tectonic forces
The theory of elasticityis fundamental
to seismologybecauseit is the elastic
propertiesof rocks which allow seismic
wavesto propagatethroughthe earth.
driving the left lateral motion of the fault.
Both kinds of stressare involved in Figure 3 which is a model
of a zone of continental collision in which (as often happens)
THE LEADING EDGE OF EXPLORATION FEBRUARY 1988 17
 A

I
I

*
t
0

Figure 5.

P P’
0 (original position) w (displaced position)

I-L
*w
Originaf length Displakw7t

Figure 6.

P
l

+x-I
Before After
Figure 7.

18 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION FEBRUARY 1988

 P
0
Before
Figure 8.
a thin sheet of crystallinerock is overthrustupon adjacent conti-
nental rocks by means of a low-angle thrust fault. In Figure 3,
the thrust sheethas been put in position as a result of horizontal
tectonic forces. If the influence of gravity is neglected, the total
horizontal tectonic force Fr due to horizontal tectonic stressa,,
is u,HW (H is the thickness of the thrust sheet and W is the
width of the sheet). The total resisting shear stressFR is u,,LW
(L is the length of the thrust fault). Often the shear stressox=
is proportional to the normal stresspressingthe surfacestogether.
In such cases,a,, = cu,, where uZZis the vertical normal stress
acting on the base of the thrust sheet and the constant c is the
coefficient of friction. If we assumea,, has the lithostatic value
UZZ = pgH, then by setting FT = Fe, we find that
Uxx = wgL
This quantity is the tectonic stressrequired to emplace a thrust
sheet of length L.
The double subscript notation is necessarybecausethe stress
on a surface element in a solid body is not in general normal
(perpendicular) to that surface, but impacts the surface element
at an angle. However, the stresscan be described by separating
it into normal and tangential components by the use of appro-
priate coordinate axes. This is illustrated in Figure 4 where three
mutually perpendicular axes (the traditional x&z) are oriented
at point P. The stressesacting on the three planes normal to
the three axes, and which pass through P, are indicated. This
is one of the most conventional notations for stress.The symbol
u indicates stress; the first subscript refers to the direction of
the force component and the second subscript to the direction
of the normal to the element of area. Thus a stressnotation in
which both subscripts are identical, such as uZr, represents a
normal stress.A stressnotation with differing subscripts indi-
cates a shear stress.
T here are in theory, as seen in Figure 4, three normal stresses
and six shear stresses.However, for practical purposes there are
only three independent shear stressesbecauseu,, = u,,~,uxZ =
err, e,, = Gy. These equalities must hold becausethere can be
no net torque on the small cube, otherwise it would be spinning.
Strain quantifies the deformation or distortion that a body
undergoes due to the application of external forces. There are
two basic categories: normal strain and shear strain. The con-
ventional notation for strain is the symbol e and two subscripts.
Thus ~~~representsa normal strain and cry representsa shear
strain.
A geologic example illustrating normal and shear strain is
GEOPHYSICS: THE LEADING EDGE OF EXPLORATION FEBRUARY 1988 19
After
presentedin Figure 5. Figure 5a showsa family of linesthat could
model a geologic cross section of flat horizontal beds, Figure
5b shows the same model after a shearing distortion. The line
AB is initially normal to the family of lines (or parallel beds);
but after folding, this line makes the angle $ with the normal.
The angle II/ is called the angular shear and its tangent is defined
as the shear strain. The ratio of the increased length of each
line in Figure 5b to its original length in Figure 5a is defined
as the normal strain.
Since stressand strain are both ex-
pressed as ratios, they are often confused.
However, there are two important dif-
ferences between them: stressesare
related to force (in fact, one working
definition is that stressis a measurement
of a material’s internal resistanceto an
external force) whereasstrains deal strictly
with configuration; and stressesspecify
conditions at a particular instant,
whereasstrains compare conditions at
two different times.
For a more formal mathematicaldevelopmentof normal strain,
let us look at a simple situation, called homogeneous strain, in
which the strain is consistentthroughout the rock. Supposethat
the block of rock in Figure 6 has been stretched uniformly so
that its dimension in the x-direction has changed.The movement
U, of a rock grain initially at x is proportional to x. The pro-
portionality constant is AL/L where L is the original length of
the block and AL is the change in its length. Thus
u, = (AL/L)x
If the strain is not uniform, the proportionality constant will
vary from place to place; in this case the proportionality factor
is denoted as a kind of local AL/L which, when the displacement
is very small (nearly alwaysthe situation in geophysicalexplora-
tion), can be expressed
atdax
t.Z* =
(Obviously, some mildly complicated calculuswas involved. The
(Stress continued on p. 29)
(Stresscontinuedfrom p. 19)
particulars can be found in many of the standard textbooks.)
The number E,, called the normal strain, describesthe amount
of stretching in the x-direction. In general, there is also stretch-
ing in the y and z directions and those amounts, E,,,,and eZr,
are similarly established and defined as the normal strains in
the y and z directions.
To developa mathematical descriptionof shearstrain, we again
begin with the homogeneous caseby isolating a small cube (see
Figure 7) in undisturbed rock. When the rock is deformed, the
cube is transformed so that its initially rectangularcrosssection
has become a parallelogram. If the strain is symmetric with
respect to x and y, the total angle of shear is made up of two
equal parts, 8/2. From Figure 7b, we seethat the x displacement
U, is proportional to they coordinate, namely U, = [tan6’/21y.
When an angle is small, it is approximately equal to its tan-
gent, making u, = (0/2)y and K, = (0/2)x. This permitsthe shear
strainseXYand eyxto be defined asO/2. The displacementformu-
las then become u, = EDGY and U, = E,J
Now assumea slightly alteredsituation. Instead of having both
angles rotate inward toward the diagonal, let both move in the
same direction so that u, = - (0/2)y and u,, = (8/2)x (see Figure
8). In this case,the cube is simply rotated through the angle o/2.
There is no distortion so, by definition, there is no strain. Thus
we must make certain that our mathematicaldescriptionof strain
eliminates pure rotations such as this. This can be achieved by
defining shearstrain as the arithmetic averageof the fwo angles,
thus (becausesmall anglesare approximately equal to their tan-
gents, and tangents are equal to derivatives):
Ex!J= Eyx = (l/2) [au,/ar + &+/ax]
Nine strains (three normal and six shear) can exist in a three-
dimensional body, but by extending the reasoning which estab-
GEOPHYSICS:
lished that E,~ = +, it follows that only three shear strains are
independent.
S, mce stressand strain are both expressedas ratios, they are
often confused. However, there are two important differences
betweenthem: stressesare relatedto force - in fact, one working
definition is that stressis a measurementof a material’s internal
resistanceto an external force, whereas strains deal strictly with
configuration; and stressesspecify conditions at a particular
instant, whereas strains compare conditions at two different
times. This is emphasized becausethe “state of strain at a par-
ticular instant” is commonly discussed.This means the strain
relating the body at that instant to some earlier shape, which
is almost always the original undeformed shape. On the other
hand, the stresscondition at any particular instant can be speci-
fied completely by the force distribution at that instant; nothing
has to be known about prior force distribution.
Rocksare among the many substancesthat can be considered,
when the deformations are small, perfectly elastic.Thus we can
assume that elastic theory in general and Hooke’s Law (the
mathematical relationship of stressand strain) in particular are
applicable to the forces in and deformations of the earth. This
is quite fortunate becausethe essenceof Hooke’s Law, stress
is proportional to strain, is linear - a rarity among physical
phenomena(99 percent of which are nonlinear) and a greatcon-
venience mathematically (99 percent of the known techniques
are linear).
Seismic wavesoccur when the equilibrium of the particles in
the earth is disturbed. Hooke’s Law can be combined with
Newton’s Law of motion to analyzethe wavemotion. Thesecom-
putational manipulations and the physical inferences that can
be extracted from them will be the themes of the next two arti-
cles in this series. E
THE LEADING EDGE OF EXPLORATION FEBRUARY 1988 29
d~scovcryseemsrather incidental among an incredible number
of profound speculationsand solid accomplishments in a large
number of disciplines.
Hooke developedtheories about the wave motion of light and
the inverse-squarenature of gravitational force that anticipated
Huygens and Newton; he predicted steam engines, named the
basic unit of biology the cell, speculatedthat matter was com-
posed of atoms, discovered the hairspring which made small
chronometers possible,was the first to state that matter expands
when heated, was an ingenious instrument designer and experi-
mentalist and a gifted microscopist (the drawings in his famous
book Micrographia are admired for their artistic merit as well
as their scientificobservations).Finally, he wasa skilled architect,
playing a key role in the rebuilding of London after the Great
Fire of 1666. (Bethlehem Royal Hospital, which gave the term
bedlam to the language, was designed by Hooke.)
A lthough he devoted little time to the earth sciences, some
rank Hooke second only to Nicholaus Steno among the
geologists of the era. Hooke studied fossils, earthquakes, the
structureof crystalsand probably discoveredindependently (and
perhapsprior to Steno) the law of constancyof interfacial angles.
He speculatedabout evolution two centuriesbefore Darwin and
proposeda dynamic earth model three centuriesbefore the plate
tectonics revolution.
Hooke accomplished all of this by age 42. He hved another
25 years but did little more of importance, spending most 01
the time arguing over scientific priority - particularly with his
greatest contemporary, and bitterest enemy, Sir Isaac Newton.
It has been suggestedthat the extraordinarily nasty relationship
between Hooke and Newton was a major factor in the latter’s
nervous breakdown in 1692. However, Newton recovered and
his ideas dominated European science,to the detriment of the
reputations of Hooke and others, for the next two centuries.
Hooke’s discoveryof the relationshipbetweenstressand strain
probably dates from the 166Os,but he did not publish it until
1678.The mathematics of the time did not allow him to develop
the principle into the sophisticated form that makes it an in-
valuable tool in physics. This was done by others, primarily
Cauchy (1823) and Stokes (184.5), who produced the modern
theory that is studied today.
Hooke’s scientific work combined the generosity of the new
age and the parsimony of the past. He was active in promoting
schemesfor cooperative endeavor, such as weather records,and
he was lavish with fruitful suggestions. However, Hooke also
had an anxiety to wear the laurels of priority. He was the last
well-known author to use the time-honored device of securing
priority by preliminary announcement in an anagram. He first
gave Hooke’s law as the anagram ceiiinosssttuvtwo yearsbefore
he disclosedthe solution ut tensiosic vis in a published descrip-
tion of the experimental evidence supporting the law.
H ooke’s law says that the pulling power of a stretched string
is proportional to the displacement. This is the case of a body
subjected to deformation in a single direction. Hooke’s law
simply saysthat, for a linear body suchas a string, normal stress
u,, is proportional to normal strain exx, that is
The proportionality constant is called Young’s modulus
(named after the English scientist Thomas Young whose versa-
tility rivals that of Hooke, but that’s another and later story).
For most materials, Young’s modulus is of the order of a
megabar (10” dynes/cm’).
However, this simple form of Hooke’s law does not hold in
three dimensions. The concept known as the dilatation must be
developed before the 3-D form of the law can be derived.
GEOPHYSICS:THE
T irt’ \tressc\ and \tramj m a three-dlr7lensionaI body were
deccrrbed in the previous article (February 1988 TLE). There
arc rhrce normal stresses(uxx, u,,, uZZ)and six shear stresses.
However. there are only three independent shearstressesbecause
:I:, ‘I,,,~ 7,; - ill,, trv: = a,,,. The same reasoning shows
that rhere are three normal strainsand SIXshear strainsbut only
rhree Independent shear strains.
The development of a mathematical foundation for
seismology requires that the nine stresscomponents be linked
to the nine strain components for each small piece of rock. The
3-D form of Hooke’s law provides this relationship and gives
us a vital insight into the nature of wave motion in the earth
(or any other solid body).
If it IS assumed that the rock is isotropic, i.e., noncrystalline
10 that there are no preferred directions, the stresscomponents
must be related to the strain components in a way that does not
depend on the coordinate directions. This means the uXyand
I’,, must be related in only one form
QX, = (constant) cry
‘this constant is defined to be the Lame shear modulus cited
earlier. Actually, for mathematical convenience, this constant
is usually denoted by 2~ which makes
1’‘I’ = &e,, ox2 = 2pe,,, u.~~= 2pe,,
the mathematical form of the relationship betweenshear stresses
and shear strains.
An obvious initial assumption is that this form would hold
for the relationship between normal stressand strain (i.e., o,,
= Zpe,,). But this is not correct becausenormal strains result
in a change in area (2-D) or volume (3-D). This involvesdilata-
tion, the basic concept which brought mathematical immortality
to Gottfried Wilhelm Leibniz (16461716) as coinventor of
calculus.Although Leibniz is sometimescalled the most brilliant
intellect in an age when genius was fairly common, he was
practically forgotten in the final years of his life. And when he
was initially resurrected,it was as the prime target in Voltaire’s
merciless satire Candide (1758).
However, Leibniz ultimately reassumedan honored place in
this history of sciencebecausehis dilatation concept - not the
fluxions of Newton - became the foundation of modern
calculus. The Leibniz theory is superior because it introduces
multiple variablesat the outset. And, becauseit is basedon area
instead of Newton’s tangent lines, the fundamental tie with
integration (also based on area) becomes more apparent.
c onslder the rectangle with sides u and v (cross hatches) in
Figure I. The product uv representsthe area, and the symbolism
d (uv), called the differential of uv, representsan incremental
increasein the area (or, in the physicsof Leibniz, a small pulsa-
tion). The dotted portion of Figure 1 representsthe differential
d fuv).
‘ LEADING EDGE OF EXPLORATION AUGUST 1988 59
 I‘igurc 2 showsthat the differential is the simple sum of three -\rial\~~\ ,)t 1til5 equalron shoi<sthar the percentageincreasem
elements: the two slabsrrdv and vdu plus the tiny section in the area ICthe sum of the percentageincrease<of the two sides(which
corner, du rl~, or i\ a ier\ welcome discovery becauseit introduces linearity into
thr calculatton\ b> replacin! a product with a sum). The obvious
d(uv) = udv + vdu + du dv counterpart for this equation in ela\ticit! theory is

Leibniz, more or less by intuition, eliminated the last term and J (‘,I t C’,L
thus derived the fundamental equation of differential calculus.
It wasthe masterstrokeof genius,although it was not appreciated becaux the definition of normal strain (see February article)
at the time The formal proof was written by Cauchy more than is the ratio of change in length to the original length. Extension
a century later. This is often cited as the date that rigor (some to 3-D IS straightforward.
say rigor mortis) became an integral part of mathematics.
After elimination of the last term, the percentageincreasein A= eix + e,, + e:;
area can be obtained by dividing the differential by the original
area which say5that ihe dilatation (the ratio of the increasein volume
of a small piece of rock to the original volume) is equal to the
d(uv) = dv f du sum of the three normal strains.
UV V u Adding the dilatation to Hooke’s law resultsin a formula for
normal stresswith the basic form

the corner (Jrx = 2he,, + (constant)A

one slab \
This new constant is defined as the other previously mentioned
Lame constant. Hooke’s law for the three normal stressesthus
dv dv reads

u xx = 2pe,, + XA
UYY = +e,, XA f
u,z = 2pe,, f X A

The next article in this serieswill explain how the two types
of seismic wave motion, longitudinal and shear, are extracted
from Hooke’s law. g

60 GEOPHYSICS:THE LEADING EDGE OF EXPLORATION AUGUST 1988

 Elasticity:Cartesian fields of dilatation
T hroughouthistorythere hasalwaysbeen
a struggle in the human mind between the
discreteand the continuous.This difference
in viewpoint climaxed in the 17th century
because of a fundamental conflict in the
cosmologiesof Descartesand Newton. The
17th century was arguably the century of
scientific genius: Kepler, Napier, Bacon,
Galileo, Descartes, Fermat, Pascal,
Huygens, Hooke, Boyle, Leeuwenhoek,
Newton, and Leibniz.
Descartesand Newton standapart from
this pantheonof scientific immortals. The
former was the key figure in establishing
modern science;the latter endowed it with
conceptsthat were so consistentlysuccess-
ful that sciencesoongainedthe position of
respect-if not outright awe--that it has
held ever since.
Both Descartes and Newton saw the
necessityof introducing mathematicsinto
physics; and both are ranked with the
greatestmathematicians,even though both
treated mathematics as a tool to further
physical investigationsand not as an end in
itself.
Their major difference was in style.
Descartes gave total freedom to his im-
agination which usually outran contem-
porary experimental results and
mathematicalknowledge. Newton operated
at the other extreme, exercising iron con-
trol over his scientific intellect. This is per-
fectly exemplified in his declaration.
Hypotheses non jngo, usually translated,
“I do not form hypotheses.” Newton was
referring to the distinction between un-
founded hypotheses and experimental
evidence. In his Principia (Book 3, Rule
3), Newton wrote, “We are certainly not
to relinquish the evidence of experiments
for the sake of dreams and vain fictions of
our own devising.”
28 GEOPHYSICS:THE LEADING EDGE OF EXPLORATION
By ENDERS ROBINSON and DEAN CLARK
The major difference in the systems
proposed by Newton and Descartes was
causedby the controversialconceptof “ac-
tion at a distance.” Newtoniansciencecon-
ceded that interactionbetween discreteand
separatedparticles must be occurring. An
example-indeed, the fundamental ex-
ample upon which his entire systemrests-
is gravitational attraction; another is
electrical attraction between two charged
particles. Cartesian science. conversely,
did not allow the possibilityof discretepar-
ticles interacting through empty space; in-
stead, it filled all space with an ethereal
substancewhich acted on bodies (as a chip
of wood 1scarried about by an eddy in a
pool). In this system. the force of gravity
on a planet is but a manifestationof a much
different reality-~thc sweepingof the planet
through spaceby a Cartexianvortex.
Contemporary debate over these con-
trastingviews was intenseand wax put into
stark relief by Voltaire when hc wrote in
his Ottres Ang1ui.se.s.“A Frenchmanwho
arrives in London finds a great change in
philosophy, as cvcrything else. He left the
world full, he finds it empty. In Pdrir one
sees the universe composedof vortices of
subtle matler. In London one sees none of
this.”
The debate lastedalmost a century, but
Newtonian physics, about 1743, ultimately
gained universal acceptance in Europe’s
scientific community. However, this situa-
tion was not permanent. Descartes’ ideas
reemerged in a different guise in the mid-
dle of the 19th cenlury. The researchesof
Michael FaradayandJamesClerk Maxwell
on electromagnetismled to experimental
results which could only be explained by
the introductionof electromagneticfields.
The fields, like Cartesianvortices, t’illsd all
space. From a physical point of view. the
concept of a field is necessary in our ;I[-
JUNE IMY
tempt to visualize how one body affects
another. Insteadof assumingthat separated
bodies can interact without anything
transpiring through the finite distancebe-
tween them, we imagine that an object
creates (or is surroundedby) a field. Any
other object which contactsthis field is af-
fected by it. The type of field is determined
by the physicalcharacteristicsof the object
from which it emanates.
T he notion of a field is extremely impor-
tant in geophysics. Consider the top of a
stove with one burner turned on. The heat
flows from the burner to other points on
the top of the stove. The temperaturedis-
tributionon the stovetop representsa field,
and the temperature at any point on the
stove top is a function of the spatialcoor-
dinatesx,y and the time coordinate f. Be-
cause temperature is a scalar quantity
(possessingmagnitude but not direction).
the set of all temperaturesforms a scalar
field. The temperature distribution in the
solid earth representsa scalar field with
three spatial coordinatesand a time coor-
dinate. A familiar scalar field is a
topographicmap, the field being the eleva-
tion of the surfaceof the earth above mean
sea level.
The topographic map is also a con-
venient way to introducean importantcon-
cept, one that is familiar to everyone who
has climbed a mountain. This is the
gradient, abbreviatedgrad. The gradient is
the scalarfield manifestationof the mathe-
matical operationknown as differentiation
in calculus.As a climber goesup the moun-
tain, he experiencesthe rate of change of
elevation. The gradient is defined as the
vector (since it has both magnitude and
direction) which points in the steepest
direction (i.e., the direction in which the
From left: Title page of Descartes’ Principio Philosophiae, Amsterdam, 1644; page 153 of Principiu on vortices; figure showing
lines of force and equipotential surfaces from Maxwell’s A Treatise on Electricity and Magnetism, Vol. 1, Oxford, 1873. Photos
courtesy of the History of ScienceCollections, University of Oklahoma Libraries.
rate of changeis greatest). The magnitude
of this vector is the rate of change of the
elevation along this path. Thus, if the
climber continuallyfollows the gradient, he
will take the shortestpath to the top of the
mountain.
An importantproperty of the gradientis
that its direction is always perpendicularto
the contourcurve. This is extremely useful
becauseit meansthat the gradient is related
to the contour curve in a way that is com-
pletely independentof any particular coor-
dinate system. For example, if one
topographic map is made with respect to
true north and anotherwith respectto mag-
netic north, the gradient on each is the
same.
The mapof the gradientsat all the points
forms a vector field (because, of course,
each gradient is a vector). Geophysics
routinely deals with many different kinds
of vector fields, such as the flow of water
in the ocean,gravity fields, electromagnetic
fields, and seismic wavefields.
Two other forms of differentiation,
divergence and curl, are used on vector
wavefields. But, before these operations
can be defined, two conceptsfrom integral
calculus-the flux integral and the work in-
tegral-need to be introduced. The
remainderof this article defines the flux in-
tegral and divergence. The next article in
this serieswill define the work integraland
the curl. Ultimately, theseconceptswill be
integratedinto a discussionof Hooke’s law
to explain seismicwave propagation.
L?.et 5 consider a small surface, suchas a
postagestamp, and assumethat vectors in
the vector field cut the stampin the normal
(perpendicular) direction (Figure 1). The mationof the fluxes for each postagestamp
flux is defined as the product of the mag- (as the limit of the area of each stampbe-
nitude of the (average) vector and the area comes infinitesimal).
of the stamp. (Thus, flux is an attempt to If the surfaceis closedarounda volume,
quantify the flow of a field through a par- suchas an egg shell, the flux integral over
ticular area.) If the vectorscut throughthe the surfacearea yields the total flux of the
stampat an angle, the flux is defined as the field emerging out of the shell. The diver-
area times the componentof the magnitude gence of a vector is defined as the ratio of
of the vector in the direction of the normal the flux of the closedsurfacedivided by the
(Figure 2). In the extreme case, the vectors volume contained by the surface (as the
are all parallel to the stamp, so no vectors limit of the volume becomesinfinitesimal)
cut throughthe stampat all. Since, in this or
case, the magnitudein the direction of the
normal is zero, the flux is zero (Figure 3). divergenceof vector = flux/volume ,
The flux integralis a straightforwardex-
tension of this basic definition. Take any which is a scalar quantity since both flux
arbitrary surface, suchas a geologic inter- and volume are scalars.
face that has been bent and deformed, and A very important result is straightfor-
divide its total area into a lot of contiguous wardly developedfrom this definition. As-
postagestamps.The flux integral of a vec- sume that many small eggs are crowded
tor field through this interface is the sum- together inside a large egg. According to
Height of box is equal to length
(i.e., magnitude) of vector
Base of box is area of
postage stamp (shaded)
Figure 1. Case in which vector is normal (i.e., perpendicular) to the postagestamp.
Flux equals volume of box.
GEOPHYSICS:THE LEADING EDGE OF EXPLORATION JUNE 1989 29
 the definitionjust given, we see each small
egg satisfiesthe relationship
r--------
/
/ flux = divergence x volume
/ Height of box is equal to
normal component of vector Now add up all theseequationsover all
_---- the small eggs. The summationon the left
Normal component of vector is the outward flux through the large egg
since all the interior fluxes cancel, a result
that is mathematicallysound. The summa-
tion on the right is the sum of the diver-
gence times the volume of each small,
interior egg. Or, as Figure 4 shows, the
I flux out of a closed surface is equal to the
volume integral or summationof the diver-
Ggure 2. Case in which vector cuts postagestamp at an angle. Flux equals volume gences times the volume elements within
of box. the surface.(Known variouslyas the diver-
gence theorem, Gauss’s law, and Green’s
theorem. this link betweenareaand volume

Vector parallel to postage stamp

1 is of great importance in applied mathe-

matics.)

,/i-’ It
Figure 3. Case in which vector lies on
also maneuvers us into a position to
postagestamp. Flux equals zero.
adapt the fundamentalidea of Descartes-
a wavefield pervading all space-to seis-
mology. We commonly think of a seismic
wavefield extendingthroughthe subsurface
-I rock layers. Its simplestform would occur
in an unbounded elastic solid that is
homogeneous(the same at all points) and
On closed surface: Flux integral = sum of isotropic(the samein all directions).When
normal component of vector x surface element the stresseson this solid are not in equi-
librium. wave motion can result.
The unbalancedstressescause a small
particle to oscillate about its equilibrium
position. This displacement is a vector
(usually designated u) which has length
equalto the amountof movementanddirec-
tion equal to the direction of displacement.
Inside volume:
Both quantities vary as time varies. The
Sum of divergence of vector x volume element
vector field u at all points and all times
Divergence of vector makes up the seismic wavefield, a direct
applicationof the Cartesiandescriptionof
nature.
Divergenceis the vectorform of the cal-
culus conceptof dilatation (see Elusticit~:
Hooke’s kw. August 1988 TLE ). and this
relationshipis a key to the vector treatment
of wave motion.
Consider the 1-D case shown in Figure
Figure 4. The divergencetheorem illustrated: Both sums are the same. 5. All movement takes place along the x
axis. The coordinate of point P in its un-
displaced position is x. and its displaced
coordinateis x + U. Thus the quantityu is
P Displaced P
. . the displacement. In order to define the
strain at point P, we mustconsiderhow its

\\ \\ position to adjacent points has changed.

\ \ Displaced Point Q is very close to P and has coor-

dmate x + AX in its undisplacedposition.
0 \‘_Q l‘ Q Q’s displacementis u + Au.
-- i+L__~-i Au can be identified as the flux issuing
from the 1-D “volume” A X. so the diver-
X gence of u can be written as
div u = flux/volume = AU / Ax .
Volume Flux
but the last expressionis the normal strain
E,,. Therefore.
Figure 5. Displacementsof pointsPand Q. Both of the above lines should be super- div u = Exx
imposed becausePand Q lie on the same line, but the diagram is made in this way for
clarity. The 2-D caseis shownin Figure 6. The

30 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION JUNE 1Y8Y


Av-
AY- ,
Figure 6. Flux and volume in the 2-D case. Remember that Leibniz said that we can
forget the smallestarea (upper right rectangle) as it is an infinitesimal of higher order.
flux is the dotted area, making the total flux
AvAu + AxAv.The 2-D “volume” is ArAy.
Thus, the divergence of vector u (with
comnonentsu and 1’) is
div u = flux/volume
= (Ay Au + AX Av) I hr A.v
= Au i Ar -t Av I A!
However, the two ratios on the rtght
are, respectively, the normal strains &,x
and E,,. The sum of these strains is, by
definition, the dilatationA and is equivalent
to the divergenceof the displacementvec-
Au
tor II. Similar reasoning extends the result
to three dimensions.that is,
div u = Exr + E‘,._
, t E,, = A
In other words, div u gives the net
amountby which a small box is being al-
ternately compressedand stretched as a
particle oscillates. Divergence, then, rep-
resentsmotion in the normal directionsand
none in the tangential directions. A sub-
sequentarticle will show that if this small
box is in the earth. this dilatation
propagatesas wave motion. The resulting
GEOPHYSICS: THE LEADING EDGE OF EXPLORATION
waves are known as compressionalwaves
or longitudinalwavesor P waves, and they
are the most importanttype of wave used
in seismicexploration for petroleum.
This is a stunninglypractical resurrec-
tion of the Cartesianvortices,a conceptthat
once seemedpermanentlyexiled from re-
spectablescientific discourse.Voltaire, in
his preface to the French translation of
Newton’s Principia, wrote, *‘If there were
still somebody absurd enough to defend
subtle and twisted (screwformed) mat-
ter., .that producesgravity, one would say:
this man is a Cartesian;if he shouldbelieve
in monads, one would say he is a Leib-
nizian. But there are no Newtonian& as
there are no Euclideans.It is the privilege
of error to give its name to a sect.”
v oltaire’s harshjudgment did not meet
the testof time Today Euclideangeometry
is but a sect of the more general non-
Euclidean geometry, and Newtonian
physics is but a sect of quantumphysics.
And Newton’s gravitational attraction
through empty space is but a sect of
Einstein’s theory of gravitation (also
known as general relativity). In Einstein’s
theory, spaceitself is curvedby twistedand
screwformed matter, and this curvatureis
what causesgravitational attraction. This
makesEinstein, in Voltaire’s definition, a
Cartesian. L5.
JUNE 1989 31
 Elasticity:Cartesianfieldsof rotation
By ENDERS ROBINSON and DEAN CLARK
T he protracted struggle between the systemsof Descartesand
Newton transformed both of them into symbolic figures. One,
Newton, embodied the ideal of modern and successfulscience,
firmly basedupon experimentaldata which it subjectedto precise
mathematicaltreatment.The other, Descartes,symbolizeda reac-
tionary and fallacious attempt to subject science to metaphysics,
disregardingexperimentsand, indeed, replacingthem by fantastic
and unprovablehypothesesaboutthe behavior of matter. The lar-
ter view is, of course,the one propagatedby the Newtonians.The
Cartesiansstatedthe situationsomewhatdifferently.
The Cartesiansrecognizedthe great superiority of Newtonian
precision as compared to the Cartesian cosmology. But they
rejected outright Newtonian attraction becausethey saw instan-
taneousaction-at-distanceas an occult quality or, even worse, as
magic or miracle. They did not admit the existenceof a perfect]!
void space, that is, the existence of “nothing” through which
gravitationalattractionwas supposedto act. Descartes’ teaching5
denied the existenceof a void or vacuumand held that spatialex-
tensionand matter were identical.
Voltaire wrote: “Geometry, which Descarteshad, in a sense,
created was a good guide and would have shownhim a safe path
to physics. But at the end he abandonedthis guide and delivered
himself to the spirit of the system. From then on, his philosophy
becamenothing more than ingeniousromance.He createda world
that existed only in his imagination and filled it with vortices of
subtle matter, the speedof which some people even calculated.”
The last commentwas a jibe at Huygens, who, usingdatasupplied
by Roemer, was the first to attempt to calculate the actual speed
of light. Voltaire’s attack was probably prompted by Huygens’
nonacceptanceof Newton’s gravitationaltheory. Huygensthought
any theory that did not include a mechanicalexplanationwas fun-
damentally flawed. His own ideas about gravity were based on
Cartesianvortices.
Despite the disdain with which Voltaire and other Newtonians
treated Cartesianvortices, the idea was not so ridiculousas their
sneeringsuggests.Huygens was not the only great name who ac-
cepted it and attemptedto extend the theory. Varignon, Leibniz.
Kant, and Laplace were all influenced by the concept of cosmic
voriices
The wholesale shift of Descartes’ thinking back into the
forefront of physicsbegan with the work of LeonhardEuler. This
most prolific mathematicianof all time developedthe mechanics
of fluid media, thereby providing an idealizedmathematicalmodel
for the transmissionof action in a continuousmedium. His math-
ematics enabled those who thoughtof energy as somethingprop-
agatedacrossspaceto interpret this concept in a precise manner.
(;1:01’11\1’S;1(‘5 1 Ill: 11. \l)lh(, I~lX;l~ 01 ~51’1 Olb\l
Euler’s work formed the basis of a new metaphor-the field. A
field is understoodroughly as a regionof spacein which eachpoint
is characterizedby quantities(scalar or vector) that are functions
of the spacecoordinatesX, y, z and time r.
Faraday provided the experimentalevidencesuggestingthe ex-
istenceof the field in the first half of the 19th century. Maxwell,
shortly afterward, unified light, electricity, and magnetismwith a
single formulation-the electromagneticfield-that is now a cor-
nerstoneof physical theory.
I n electromagnetism.the easiestcircumstanceto treat is the case
In which nothing dependsupon time-the static case. All charges
are permanentlyfixed in space:or, if they do move. they move as
a steady flow in a circuit (so both the charge density, P, and the
current density,j, are constant).In this case, the electromagnetic
field breaksinto two fields that are not interconnected.The electric
field E and the magnetic field B are distinct phenomenaso long
3schargesand currentsare static.
The rlectrostaticfield, a vector field representedby a vector E
which does not changewith time is due to electric chargesrepre-
sentedby the charge density P. If a small region of spacecontains
positive charges, then the field vectors E will emanatefrom that
source: and if a small region containsnegativecharges, the field
vectors will flow to that sink. Therefore, the sourcepoints of the
field are those for which the divergenceis greater than zero, and
the sink pointsare thosefor which the divergenceis lessthan zero.
(Figure 1 maps a vector field of one sourceand one sink). Math-
ematically,the divergenceof the electrostaticfield E is
div E = PIE,
where the sourcesare representedby the charge density P, and E()
is the electric constant.This fomula is known as Gauss’s law and
is one of Maxwell’s four equations.It also relates to the modern
theory of seismicwave propagation(to which we shall return after
the introductionof a complementaryconcept).
If the density of an electric current (designatedj) through a
wire is constant, a staric magnetic field B will surroundthe wire
(Figure 2). Magnetic fields make loopsaroundthe currents.They
do not diverge from a source(there is an electric charge, but there
is no magnetic charge), so, in every case,
div B = 0 .
However, the magnetic field does have a rotation or curl that
is proportionalto the current density; that is,
IO% UO\ b\ll3l~K I’%)‘
Figure 1. Field lines (solid) and equipotentials (dashes)for two
equal and oppositepoint charges. The positivecharge is an iso-
lated source,and the negativecharge is an isolated sink.
curl B = cl0j
where h is the magnetic constant.
Curl is a conceptwhich can also be appliedto any vector field.
An electrostaticfield has sourcesbut no rotation, so in every case,
curl E = 0 .
To summarizetheseconcepts-which are of major importance
in geophysics-an electrostaticfield E may be describedas a vec-
tor field with a given divergence and zero curl, while a magneto-
static field B has a given curl and zero divergence.
Curl is one of two basic differential operatorsused on vector
fields (the other is divergence which was discussedin BASIC SEIS-
MOLOGY 5, June 1989 TLE). The curl is best explained via the so-
called work integral.
Work integral is a generic term which applies to any vector
field, but the name actually derives from the specialcasein which
the field is a force field. In physics,the quantityknown as “work”
is defined as the product of the force componentin the direction
of a path and the lengthof the path. When the path is straight, this
definition can be applieddirectly. However, in the caseof a curved
path, things are more complicated: the curve must be broken up
into a lot of small straight-linesegments;the work along each seg-
ment is computed; and all of these are summed. This summation
becomesthe work integral (in the limit when each segment be-
comes infinitesimal).
,--
-,
Thus, the work integral is a summationof contributionsalong
any arbitrary curve (not excluding a closedcurve) in a vector field.
tQ;$--J -’
A_
c2
-’
/
\
/
\
i
Figure 3. The coalescenceof two closed curves C, and C, into
a single closedcurve C. The contributions to the work integral
along the common segment cancel out becauseof the opposite
sensesof circulation along this segment.
GEOPHYSICS:
Figure 2. The magnetic field (two lines only illustrated) outside
of an infinitely long straight wire carrying a constant electric
current.
Each contribution is the product of the vector componentin the
direction of the path multiplied by the path length.
When the vector representsforce, the work integral is indeed
the total amountof work performed in transversingthat curve. But
the usefulnessof this integral is by no means confined to cases
where the vector field is a force field. For example, in fluid
dynamics,the vector may be velocity ( so it is a velocity field ).
If the curve is closed, the work integral actually defines the
circulation aroundthat curve. The work integral is a scalarquan-
tity because work is a scalar. The value of the work integral
depends,in general, on the curve C but not on the coordinatesys-
tem used in the evaluation.
Let Cl and Cz be two closed curves with a common segment
(Figure 3.) Let the work integral be taken around each curve so
that the common segmentis traversedin oppositedirections.Then
the contributionto the work integral for each curve is equal and
oppositefor the common segment.Thus, the sumof the two work
integralsis equal to the single work integral over the closedcurve
which consistsof the two original curves less their common seg-
ment
The work integral provides the means for defining the curl
(Figure 4). We take a point within a vector field u. At that point,
we take a fixed unit vector n and considerany small surfaceele-
ment with this fixed vector as the normal. This small surfaceele-
ment is boundby a small closedcurve C. We now define the curl:
Componentof curl u in fixed normal direction =
work integral of u around closedcurve/area within closed curve.
Figure 4. Geometry used to define the curl. The unit vector n
is the normal to the small surface area enclosedby the curve
C. The component (in the direction n) of the curl of the vector
u is equal to the quotient of the work integral of u around C
divided by the enclosedarea.
THE LEADING EDGE OF EXPLORATION NOVEMBER 1989 19
 Around curve: Work integral =
sum of tangential component of vector u x path element

Inside area: Flux integral =

sum of normal component of curl u x area element

Curl of vector u

Figure 5. Stokestheorem illustrated. Both sumsare the same.

Velocity vector u (with magmtude IJT

and in tangential direction)

J Small volume of water

r Circular arc _ utr

A wf
perpendicular l

to page
(i.e., straight up)

Figure 6. A small volume element of water. The element is Figure 7. A small paddle wheel in circularly rotating water.
rotating in a circle of radius r. The angular velocity vector w The inner paddle is hit with a velocity lessthan that hitting the
points in the direction of the axis of rotation, and the vector outer paddle, thereby making the paddle rotate.
has magnitude denoted by CO.

This definition holds in the limit when the area becomes in-
finitesimally small. The curl is a vector quantity.
A very important result can be developedfrom this definition.
Assume that many small areasare crowded together inside a large
area. According to the definition just given, we seethat each small
area satisfies
Work integral = normal componentof curl u x area element.
Now add up all theseequationsover all the small area elements.
The summationon the left is the work integral around the large
area (since all the interior contributionscancel out). The summa-
tion on the right is the sum of the normal component of curl II
times each small interior area element. Or, as Figure 5 shows, the
work integral arounda closed curve is equalto the surfaceintegral
(i.e., summation)of the normal componentsof the curl times the
area elementswithin the curve. (The surfaceintegral here is recog-
nized as the flux of the curl vector through the closed surface.)
This result is known as Srokesrheorem. which can be statedas:
Tile work integral of a vectoraround a closedcurve is equal to
the jlux integral of the curl of the vector through any sudace
boundedby the curve.
T he precedingdiscussionmay leave the reader with a feeling of
not knowing exactly what the curl of a vector is. The fact that the
curl has somethingto do with the work integral around a closed
path suggeststhat the curl somehowdescribesCartesian vortices
which are rotating, swirling, or curling around. An item taken
from fluid motion will help make theseimpressionsclearer. (This
example takes us back to the fundamentalwork of Euler on fluid
mechanics.)

20 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION NOVEMBER 1989

(Left) Rem! Descartes(1596-1650). French scientistand philosopher.His Cartesian systemestablishedthe ideal of mathematical
certitude in metaphysical demonstrations and, by brushing aside the then familiar scholastic subtleties, introduced modern
philosophyand scienceof thought. (Right) Sir Isaac Newton (1642-1727). English natural philosopherand mathematician. He is
credited with the invention of calculus, the formulation of the laws of motion and the law of universal gravitation, and the dis-
coveryof the spectrumof light.

Supposethat a bucketof water is rotatedaboutits axis at a con-
stantangularvelocity w. This velocity vectorpointsin the direc-
tion of the axis and hasmagnitudedenotedby o. At equilibrium,
all the water is rotatingat this angularvelocity,so that eachsmall
volume of water is undergoingcircular motion (Figure 6). Con-
sider a small elementof volume at radius r. In time I, this small
volume sweepsout an angle WI, so the circular distancttraversed
is wtr. Thusthemagnitudeof its velocityis thedistanceotr divided
by t, or WI. The directionof this velocity is the tangentialdirec-
tion. Denotethe velocity vectorwith this magnitudeand direction
asu.
Now let us find the work integral of the velocity around this
circle. Since the velocity vectoris tangentialto the circular path,
the full magnitudeor of the velocityis usedin computingthe work
integral. The lengthof the circular path is 2nr. Thus the value of
thework integralis magnitudetimesdistance,2nr2w.
Becauseall the water in the bucket is rotating at constantan-
gular velocity,we would expectthe curl of the velocity vectorto
be the sameeverywhere.This is indeedthe case;that is, curl u =
constant.Moreover,we would expectthat the curl would point in
the samedirectionas the axis of rotation,and this is alsothe case.
Now let us considerthe flux of the curl acrossthe given circle
of radiusr. This flux integralis equalto the summationof the curl
timesthe small areaelementsmaking up the circle. Since the curl
is a constant,this summationreducesto simply the magnitudeof
thecurl timestheareaof thecircle,
flux of curl u = 1curl u 1 (nr*).
Stokestheoremsaysthat the work integral and flux are equal or,
2nr*w = nr* 1curl u 1.
Sincebotho andcurl u pointin the samedirection,thisresultgives
(uponcancellationof nr2)
w = l/2 curl u.
If we put a small paddlewheel in the water, the small wheel
would startto spin becausethe impingingwater would exert a net
torqueon thepaddle(Figure 7). Moreover,the paddlewheelwould
rotate most quickly with its axis pointing in the directionof the
curl (namely,the directionof the axisof the bucket).Furthermore,
the paddlewheel rotateswith the sameangularvelocity for every
point in the bucket(which illustratesthe constantvalueof the curl
in thiscase).
The curl of any vectoru representsthe tangentialmotion of the
vectoru. This tangentialmotion is traditionallycalledthe rotation
vectorw anddefined
w= l/2 curl u
due to the reasoningillustratedin the bucket example. The curl
recordsthe directionand magnitudeof the maximum circulation
at a givenpoint.
T he vectorconceptsof div and curl were discussedin this series
becausethey are requiredto define a seismicwavefield. Particle
motion is describedby a displacementvectoru, which represents
the oscillationof a tiny rock particle. The wavefieldallows two
kindsof motion-motion in thenormaldirection(thestretchingand
shrinking of the particle) and motion in the tangential direc-
tion. Normal motion is given by dilatationwhich is equalto div u
and rotationalmotion is proportionalto curl u. When thesecon-
ceptsare combinedwith Hooke’s law (describedin a previousar-
ticle), it can be shownthat dilatationA and rotation o propagate
separately,asdistinctseismicwaveforms.E.

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION NOVEMBER 1989 21

 Elasticity:Equationsof motion
By ENDERSROBINSONand DEAN CLARK
0 ceanographers havelong saidthat our planetshouldbe called
Water, not Earth, becausedry land representslessthan a quarter
of planetarysurface.Applying this sameline of reasoningto the
structureof all matter,we shouldcall-according to theNewtonian
perspective-the whole universeVacuum. The starsare but tiny
isles in the ocean of interstellar “near vacuum.” Even the in-
dividual atomsmakingup matterare nothingbut small nucleiwith
surroundingelectionsenmeshedin a seaof vacuum.
But todayour view of “vacuum” is thatof theCartesians.This
omnipresentmedium,called “vacuum” sinceancienttimes, is by
no meansemptinessor nothing. Vacuum influenceseverythingit
surrounds.Every experimentin elementaryparticlephysicsdem-
onstratesthe interactionof subatomicparticleswith one another
and with the vacuum.
JohannWolfgang von Goethe,whosegreatliterary eminence
hasall but eclipsedhis (not insignificant)scientificcontributions,
wrote in Faust:
Let usfathom it, whatevermay befall,
In this, thy Nothing,may I$nd my All.
Goethe’s metaphorencompasses an insightof the great quan-
tum physicistPaul Dirac who realizedthat somephysicalobjects
revealthemselvesonly occasionally.An unexcitedatomin a min-
imum-energy statedoes not radiate and, consequently,remains
unobservableif not subjectedto any action. Eachelementarypar-
ticle is but a manifestationof its own sea. The particle is unob-
servableuntil its sea is acted upon in a definite way. When a
quantumof light getsinto this “Dirac sea,” the seacan ejectout
of itselfan electronof negativeenergy.A multitudeof conclusions
followed from Dirac’s insight, including the discovery of the
positronand otherantiparticles.
During the pasthalf century,the Dirac seahasturnedinto the
oceanknown asphysicalvacuumand Dirac himselfoncesaidthat
the problemof describingvacuumwasthe main one facingphysi-
cists. The currently prevailing description,the conceptof fields
pervadingall space,hasevolvedfrom the once-deridedCartesian
vortices.And, amongthe many scientificideasdependenton this
view, is themathematicalfoundationof the theoryof seismicwave
propagation.
The first two articleson elasticity(BasicSeismology3 and 4)
discussedthe mathematicsof stressand strain for a medium in
staticequilibrium. This article will elaborateon that foundationto
illustratewhy wave motioncan resultwhen the stresses on a solid
are not in equilibrium.
More than one mathematicaltechniquecan be.usedto support
this assumption.In this article, we will use the tool commonly
known as vectoranalysiswhich in large part was the creationof
one of the mostremarkablefiguresin modem science,J. Willard
Gibbs.
24 GEOPHYSICS:THE LEADING EDGE OF EXPLORATION JULY 1990
Gibbs now is regularlycited as the greatestscientistever pro-
ducedin the United States.In fact, shortly a&r his death, his-
torian Henry AdamscalledGibbs “the greatestof all Americans,
judgedby his rank in science.” However, during his life (1839-
1991), he was all but unknown to US scientific leaders. This
resultedfrom a combinationof curiouscircumstances:his own
casualinterestin recognition(e.g., he neverjoined the American
PhysicalSociety);a teachingstylethat was accessibleto only the
brightestgraduatestudents;much more interest,by the contem-
porary US scientificestablishment,in immediatelyuseful ideas
rather than highly theoreticalwork (e.g., the immensefame of
Gibbs’ contemporariesBell and Edison); and mathematicsfar
beyondthe abilitiesof mostwho readthe lightly regardedTrens-
actions of the ConnecticutAcademyof Sciencesin which he
published(in the 1870s)his first importantpapers.JamesClerk
Maxwell did recognizethe importanceof Gibbs’ ideas;Maxwell
personallymadea physicalmodelbasedon a Gibbs’ conceptand
sentit to him. However, Maxwell died beforehe couldconvince
other Europeanscientistsof the value of Gibbs’ thought.It was
not until the last few years of his life that Gibbs receivedrichly
deservedhonors(notablythe CopleyMedal of the Royal Society)
from the scientificcommunity.
A greaterhonor came half a century later. Albert Einstein,
shortlybefore his death, was askedto name the most powerful
thinkershe had known. “Lorentz,” Einstein answeredwithout
hesitation.Then, after some reflection, he added:“I never met
Willard Gibbs; perhaps,had I done so, I might have placedhim
besideLorentz.”
The rigorousmathematicsthat Gibbs developedfor thermo-
dynamicsand statisticalmechanicswere his greatestscientific
achievements.This article, though, is built around still another
Gibbsconceptof monumentalimportance-vectoranalysis.Gibbs
developedits presentform more than a centuryago-m Elements
of VectorAnalysis,first printedin 1881. Astonishingly,consider-
ing the surgical precisionwith which vector analysiscan treat
many complexphysicalsituations,this ideawasgreetedwith con-
siderableopenhositility. One reviewercall4 it “a hermaphrodite
monster,compoundedof the notationsof Hamilton and Grass-
man.” However, its usefulnesswas unanimouslyrecognizedby
the turn of the century.Today, vectoranalysissothoroughlyper-
meatesscientificliterature(e.g., mostof thearticlesin GEOPHYSICS
contain mathematicalnotation initiated by Gibbs) that mathe-
matican/historianEdna E. Kramer called this concept“indispen-
sible to every seriousstudentof physics.”
Vectors are a most convenient means for mathematically
describingand analyzing seismicwaves.
Consideran unboundedelasticsolidthatis homogeneous (same
at all points)and isotropic(samein all directions).If a disturban~
passesthroughthe material, the displacementof a small particle
 J. Willard Gibbs
whose equilibrium position is the point P(x,y,z) can be specified
at any instant as a vector u(x,y,z,t). This vector has its origin at
the point of equilibrium and pointsin the direction of particle dis-
placement. The length of the vector gives the amountof displace-
ment. As time varies, the length and direction of the vector alter
to representthe oscillationof a small particle aboutits equilibrium
point.
Adapting basic vector conceptsto the mathematicaltheory of
elasticity requires the introductionof vector fields.
A physical field is a quantity which dependsupon position in
space. The simplestpossiblephysical field is a scalar field; i.e.,
a field which is characterizedat each point by a single number.
(A good example in geophysicsis the potential field representing
the force of gravity. This field does not changewith time There
are scalar folds which do change with time Consider material,
like the solid earth, that hasken heatedat someplacesand cooled
at others. The temperatureof the body varies from point-to-point
in a complicatedway, and will be a function not only of position
but of time This is an example of what is known as a time-de-
pendent scalarfield.)
In geophysics,scalarfields are depictedby meansof contours,
which are imaginary surfaces (in 3-D) or lines (in 2-D) drawn
throughall points for which the field has the samevalue. Contour
lines originated, of course, on maps where they connect points
with the same elevation. But contour lines can also be used in
other areas, such as on a temperaturefield (where they are called
isothermalsurfacesor isotherms).
Vector fields, in contrast, are fields in which a vector is at-
tachedto each point in space. The flow of heat in the earth is an
example. If the temperatureis high at one placeandlow at another,
there is a flow of heat from the warmer place to the colder. Thus
heat flow is a quantity which has direction. A scalaris not a suf-
ficient mathematicaldescription. However, heat flow can be rep-
resentedat each point by a vector. This vector varies with both
position and time Its magnitudegives the amountof heatflowing
at any point at the designatedtime and its direction gives the
direction of the flow.
GEOPHYSICS: THE LEADING EDGE OF EXPLORATION JULY 1990 25
Johann Wolfgang von Goethe
When fields vary with time the variation can be obtainedby
taking the derivative (or rate of change)with respectto time Find-
ing the rate of changewith respectto position is trickier because
there are three coordinatesinsteadof one. However, it can be neat-
ly handledby the concept of the gradient (which is defined as a
vector which gives the rate of change of a field with respectto
position). The gradientvector (abbreviatedgrad) is perpendicular
to the contour line at which it originates, points uphill in the
steepestdirection, and possessesmagnitudeequal to the rate of
change in that direction. The gradient has two particularly valu-
able properties-it indicates the direction and amount of the
greatestrate of changeat any particularpoint, andit is independent
of any systemof coordinateaxes. (The gradient was more fully
discussedin Elasticity: Cartesianfields of dilatation, Basic Seis-
mology 5, IZE June 1989.)
Another key conceptin vector fields is that of the divergence
of a vector (abbreviateddiv). Divergence, unlike gradient, is a
scalar. To obtain a physicalinterpretationof divergence, consider
a small box in spacethat is subjectedto elasticwave motion. The
displacementof a small particIe (micrograin of sand)in that small
box is denotedby the vector u. AS this par&lt~ oscillates,each of
the six faces of the small box will undergo normal strain. The
divergenceof the vector Y is equalto the sumof the no& strains
in the three coordinatediiections. That is, div u gives the net
amountby which the small box is being alternativelystretchedand
compressed(in the normal directions) as the particles oscillate.
This equatesto the definition of dilatation given in Basic Seismol-
ogy 5 and, indeed, div u and dilatation A are merely different
words for the same scalar.
The last conceptthat must be introducedis the rotation of vec-
tor u, written curl II (although Europeanauthorssometimesuse
rot u), which is also a vector. (The curl is defined in Elasticity:
Cartesianfields of rotation, Basic Seismology6, TLE November
1989.) Considera particle that is undergoingonly shearstrain, no
normal strain; this particle will thus be performing some sort of
rotating motion in a plane at right angles to the direction of
propagation.
 These concepts-gradient, divergence, and curl--can hr c‘om
bined with ideaspresentedin earlier articlesin this seriesto derive
mathematicalexpressionswhich describe wave motion in elastic
solids.
When a body is in equilibrium position, no net force is acting
on any point in the body. However, when the body is deformed,
the stresses(due to internal elastic forces of the body itself) exert
restoring forces which attempt to reestablishequilibrium. The
more the distortion, the greater the restoring force. The English
mathematicianBrook Taylor (see Z?Wwave equation,Basic Seis-
mology 1, ZZJ?July 1987) discoveredthat the restoring force is
proportionalto spatialcurvaturein the caseof a vibrating string.
The greater the curvature,the greater the restoringforce.
But what is spatialcurvaturein the caseof a solid body?It can
be representedby secondderivativesof the particle displacement
u with respect to the spatial coordinates(x,y,z). The gradient,
divergence, and curl are each first derivatives; i.e., each repre-
sentsa slopeof one form or another. Secondderivativesrepresent
curvature, and there are only three independentsecondderiva-
tives:
grad div II
div curl u
curl curl u
The quantity div II (or dilatation) representsthe change in
volume of a small region (due to compressionand stretching,the
internal stresses).Thus the secondderivative, grad div u, is the
vector field which representsthe directional slope of the dilata-
tion. Since curl u involves no changein volume, the divergence
of curl is zero and this secondderivative need not be considered
further. Because
curl u = 20,
26 GEOPHYSICS:THE LEADING EDGE OF EXPLORATION JULY 1990
the final secondderivative is
curl curl u = 2 curl 0.
Now, returning to the reasoningof Taylor (force per unit of
volume is proportionalto spatial curvature), the relationshipcan
be written
force density =
(constant)grad div u + (different constant)curl
curl u
It turns out that the first constantis A + 2~ and the secondis
-p where A and p are Lame’s constants.This determinationof the
unknown constantsin the equationcomes from a direct applica-
tion of Hooke’s law (see Elasticity:Hook’s law, BasicSeismol-
ogy 4, ZZE August 1988). Thus, the equationbecomes
force density = (A + 2~) grad div u - p curl curl u .
However, Newton’s secondlaw of motion equatesforce to the
product of mass and acceleration. Thus the force density is a
productof massdensity (denotedby P and acceleration(written
as a secondderivative with respectto time). The aboveequation
becomes
P$ = (X t 2~) graddiv u - p curl curl u
which is known, in seismology,as the equationof motion. After
some mathematicalmanipulation,two wave equationscan be ex-
tracted from this expression.
First, the equation will be written in another form. The first
term on the right involveddiv u which is the dilation A. The second
term on the right involvescurl u, which is twice the rotation w
(that is, curl u = 20). By making these substituttons.chcaqua- wave velocity Q. These dilatational waves are also know as P-
tion of motion can be written waves, compressionalwaves, or longitudinalwaves.
2 If we then take the curl of the equationof motion, the term in-
P+= (X + 2/.l)gradd-2~c”rlw volving A vanishesand the result (after some manipulation)is

An advantageof this form of the equationof motion is that it rJlw =Pa28

explicitly displaysthe dilatationA and the rotationw. p at2
Dilatation A representsa measure of change in volume of a If we define fl as
small region without any rotation; and rotation representsmove-
ment of a small region withoutany changein volume. The respec- a=K
tive mathematicalexpressionsare curl A = 0, and div o = 0. P
Now, keeping those very helpful expressionsin mind, we take then this equation(in rectangularcoordinates)is
the divergenceof the equationof motion. The term involving w
vanishes,leaving
azw azw a’w -- I azw
z+av’+ai’= pz at2

(X+2p)divgradA=P 9 This is again recognizedas the three-dimensionalwave equation,

and the interpretationis that the rotation (r, propagateswith wave
The operator div grad is encounteredso often that it has its velocity /3. These rotational waves are also known as S-waves,
own symbol9’ and name, the Laplacian.By substitutingthis sym- shearwaves, or transversewaves.
bol anddefining a as The bottom line is that modem theory statesthat there are two
types of seismicbody waves. One type resultsfrom the normal
“=vX
pulsationsdue to the compressionalstretching/shrinkingof a con-
P tinuum of small rock particles; these generateP-waves traveling
the equationbecomes at velocity a. The secondtype results from tangentialpulsations
1 8’A due to the rotation of a continuumof small rock particles;these
V’A = generateS-wavestraveling at velocity p.
-Fat2
In this seriesof articles to this point, we have completedthe
In rectangularcoordinates,this equationis formidable task of deriving the equationof motion and thereby
obtaining the wave equationsfor compressionalwaves and for
#A a=A #A _-- 1 8’A shearwaves. A readerwho has followed the argumentup to now
z+ay+saz’- 0z at1
can feel a senseof accomplishment,for this material is usually
covered in the first weeks of a graduatecoursein seismologyat
We recognizethis (from the first article) as the three-dimensional a gooduniversity. The next articlesin this serieswill turn to some
wave equation, which says that the dilatation A propagateswith of the more practicalaspectsof seismicexploration.L

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION JULY 1990 27

 Sampling
andtheNyquistfrequency
By ENDERSROBWSONand DEAN CLARK

S wiss-bornLeonhardEuler (1707~83),who lived primarily in

Germany and Russia,was the mostprolific mathematicianof all
time Few people,in any branchof science,haveevenapproached
his astonishingpublicationrecord-a large percentageof whichis
of the first importanc4z
Mathematician/historianMorris Kline estimatesthat Euler
publishedoriginal researchat the incrediblerateof 800 pagesper
year during most of his long professionallife. A modem edition
of his collectedworksrequires74 volumes.
By the middle of the 18th century, Euler had discoveredand
publisheda large number of the practicalsolutionsto diiTerential
equationsthatare~o~nusedtodayaodalsoconaibutedkey
ideas toward the developmentof numerical methods for ap-
proximatesolutionsto thosewhich can’t be explicitly expressed.
This was less than 100 years aikerthe invention of calculusand
200 yearsbeforethe invention of the electroniccomputer!And it Hony Nyquist Leonhad Euler
representsonly a small fractionof Euler’s overall contributionto
mathematicsand physics.
Euler’s mostfamousdiscoveryis the compactformula per second.If we designateonespokeon thewheelasthereference
eri+ 1= 0 vector, this vectormakes 10 completerotationsin every sec4md.
Each completerotation (through360” or 2~. radians)represents
which relatesthe fundamentalnumberof calculus,the fundamen- one cycle, so the vector’s frequency-in this par&c&r case-is
tal number of trigonometry, the fundamentalcomplex number, 10 cyclesper second.Cycles per secondha$ its own term hertz
and the two fuodamental“natural” numbers.The impactof this (in honor of 19th century German physicistHeinrich Hertz), so
formulaon the scienti6cworld wasmonumental,bothmathemati- this vector’s frequencywould mostoften be written 10 Hz. (This
cally and psychologically.The 19th century American mathe- is still somewhatincompletebecauseit does not specify which
matician and astronomerBenjamin Peirce regardedit with awe, directionthe vectoris rotating.The normal mathematicalconven-
reportedlytelling his students:“Gentlemen, this is surelytrue. It tion givesa positivefrequencyto rotationin the counterclockwise
is absolutelyparadoxical;we canuotunderstandit, and we don’t directionand negativefrequencyin the clockwisedirection.Fre-
know what it means, but we have proved it, and thereforewe quencyis customarilydenotedby the symbolf.)
know it must be the truth.*’

T he famous formula cited above is, however,just a specific
result of the extraordinarilyimportantequationwhich Euler ex-
tractedfrom his pioneeringresearchin complexnumbersand in-
fiilite series:
eiot =costit+isinwf
When at = T in &is equation, it reducesto the result which
startledBenjaminPeirce(and many, many othersthen and now).
Thisgeneralfonnulaisthekeytotheuseofsinesandcos~of
the time variable t in muchdigital processingof seismicdata.
The idea of frequencyis basic to this approach,so it is ap-
propriateto carefully examine exactly what it means. Consider,
for example, a wheel that is rotatingat a rate of 10 revolutions
C losely relatedto the conceptof frequencyis that of period-
defmedas the shortesttime interval in which periodicmotionex-
actly repeats.In our rotatingvectorexample,the motion exactly
repeats10 times per second,so it is saidto have a period (usual-
ly denoted2) of l/10 second.PeriodTand -fare reciprc+
cals, meaningthat T = llfandf = l/T.
A complicationarisesat this point becausethere is frequency
and then there is frequency.The frequencydescribedearlier is
rotational frequency or the number of rotations made by a
referencevectorin a given time interval. However, anothervalu-
able conceptis angularfrequencyor the relationshipof the angle
betweenthe rotatingvector and a tied line (usually the x-axis).
Portunately,when angularfrequency(normally denotedo) is ex-

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION MARCH 1991 51

 pressed in radianspersecond,it is mathematically
tied to rotation- Y
al fresuencyby the straightforward formula
w= 2x5
With these&ii&ions established,it becomesapparentthat
Euler’s formula is the mathematicalexpressionfor the rotating
vector.This is diagrammedin Figure 1 which showsa vector
rotatingin the positive(counterclockwise) direction.The vector
sinot
has unit length and makesan angle tit with the x-axis which
automatically assigns valuesof cosit andsinat, respectively, to
thevector’sprojections onthecoordinate axes.If wecall thisplane
thecomplexz-plane(wherez = x + iy), Euler’s formulareveals
thatthe positionof therotatingvectorat any timeis e ial? If the
angularfrequencyis fixed at somepositivevalue, thenthis vec-
tor will-as timeincreases-rotatein a counterclockwise direction Flgure 1. Coordinates of the rotating vector.
andtraceout a unit circle in ther-plane. SeeFigure2.
This type of repeatedmotionwith fixed frequencyis called
simpleharmonicmotionand, as Euler’s formnlamakesclear, is
easilyand uniquelyexpressed mathematically by merely adding
the sineand cosinefunctions.
However,up until thispoint, e i@rhasbeentreatedas a con-
tinuousfunction-that is, the parametert wasallowedto assume
anyrealvalue. The mathematics is not sostraightforward when-
as routine4yhappensin modem signal analysis,particularlyin
explorationselamology-thefnnction is not continuousbut is
sampled at equallyspacedtlmes.Thisshiftfromcontinuous to dis-
creteis necessary sothatthe datamaybe efficientlyprocessed on
higb-spead, digital computers.
If we denotecontin~us timeby t, tbe samplingintervalby Ar,
and the discratetime Index by the iutegern, then the sampling
timepointsare givenby the fundamentalrelation
t = nAt.

A most
@ortaut consequence, in shiftingfrom the continuous
to thedlsuUe formsof a die function,is theintroductionof the
pawamsnOncalledaliasing.Aliasmgis not just a mathematical
extra&y. It is a common,everydayexperience but we aresoac-
customed to it thatwe are only vaguelyawareof it.
Supposewe seea moving objectfirst at pointA and then at
pointB, but our eyes-which blink openandshutmanytimesper Figure 2. The rut&lug vectortracesout the unit circle.
minute-do not seethe actualmotionbetweenA and B. Instead,
our mind interpolates the shortestdistancebetweenA andB.
Magiciansbasemanyof their sleight-of-hand tricksuponthe
Actual rotationand
alias@ phenomenon andit alsoexplainsthe “wheel paradox”in
Westernmovie scenesof stagecoaches or coveredwagonswhich apparentrotation
have large, spokedwheels. In these scenes,we see, as the
stagecoach picksup speed,the spokes turn fasterandfasteriu the
forwarddirection...andthen (eventhoughthe stagecoach is still
going forward)suddenlyreversedirectionand slow down to a
stop...and~ensuddenlyswitchbacktothe“right”direction,&c.
of course,a movie is not continuousbut, rather,a seriesof
still pictures;in otherwords,it samplesthe motionbeingphoto-
graphed.And, as a result, high ratesof wheel rotationcan be
aliasedto low fmquencies.
Bow doesthis happen?Considerthe caseof one partkukU
spokewhichcanbe npresentedby thevectore IOf. If thisfunc-
tionissampledattimest = nL, the vectorbecomes
ei4&.
At the slow qeed of wAt = r/6, our samplefunctiOn ia the
sequenceof vectors
eiw(r/6)n
~~v~areshowninFigure3forn=0,1,2.~
humanr&d registerswhatis seenasthe smallestmotionthataC-
UXIMS for the sampledobservations.Thus, in thiscase,the+
&n and the apparentmotionare the same. Figure 3. Case in which the apparent rotation ls lu the same
However,at fasterspeeds,themind will interpretdifferently. clke&m as the actual rotation.

52 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION MARCH 1991

 If wL = 11x16, the sequence of vectorsfor n = 0, 1,2 will ap-
pearasshownin Figure4. The mind will againregisterthesmal-
lest rotationthat accountsfor the observationsand, in this case,
thatwill be a negativerotationof x//6 insteadof a position rota-
tion of 11~16and this givesthe effectof reversemotioninstead
of forwardmotion.

T he cntical
. points, in keeping with the well knownexpression
“180” outof phase.“,occurfor eitherw = -Al or w = r/(Ar).
When w = s/(Af), the samplefunctionis a sequence of f 1 vec-
tors whichflop back and forth on the x-axis. See Figure5. The
apparentmotion,asa result,couldbe eitherway. Thesetwo fre-
quenciesareknown,respectively, astheupperandlowerNyquist
frequencies and the intervalbetweenthem is the Nyquistrange.
SeeFigure6.
This is an importantconceptbecause,as is illustratedby the
stagecoach wheelexample,any actualrotationalwaysappearsas
an apparentrotationwithin the Nyquistrange.In the casewhere
w = rl(6Ar), the actualfrequencyis within theNyquistrangeso
actualfrequencyandapparentfrequencywereidentical.However,
thefrequencyo = 1lr1(6&) is well outsidetheNyquistrangeso
Figure!4. Casein whiebthe apparentrotationis in the oppasite the apparentfrequencyis the aliasedversionwithin the range,
dire&ion as the actual rotation. namelyw = -or/. The combination of theNyquistrangeand
the fact thatone completerotationis 2a radiansleadsto the Ny-
quistrule: Add or subtractsome integer multiple of 2rl(At) from
Actual the actualfrequencyto obtainthe apparentor aliasedfrequency.
Cyclic frequencyf is w/2x, sothe cyclic Nyquistrangecon-
sistsof frequencies from-l/(2&) to 1/(2b). The reciprocalof the
samplingintervalAf is the samplingrate or samplingfrequency
fp In otherwords,the samplingfrequencyis one sampleper At.
Thus,in turn, the Nyquistfrequencyl/ (2t) is half the sampling
frequency.This allowsthe Nyquistrule to be rephrased as: If a
frequencyf isoutsidetheNyquistrange,thenits aliasis foundby
subtracting fromf an integralmultipleof the samplingfrequency
(i.e., the aliasedfrequencyisf - wwhere k is thatintegerwhich
putsthe aliasedfrequencyinto the Nyquistrange).
For example,a samplingintervalof 4 mscorresponds to a sam-
PositiveVector x
pling rate of 110.004which is 250 samplesper secondwhichis
expressed asfs = 250 Hz. The Nyquistfrequencyis half thesam-
pling frequencyor 125 Hz. The Nyquistrange,whichcoversan
intervalequalto the samplingfrequency,is -125 Hz to 125 HZ.
If theactualfrequencyis 425 Hz (whichis outsidetheNyquist
range),what wouldbe the apparentor aliasedfrequency? It can
be found, accordingto the Nyquistrule, by subtractingan ap-
propriateintegermultipleof the samplingfrequencyfrom the ac-
tual frequencysothatthedifferencelies in the Nyquistrange.In
Apparentrotationeitherway
this case,the appropriate multipleis 2 and the aliasedfrequency
is
z&Wti&bicb the apparent rotation can appear to
. 425 - 2(250) = - 75.

T he many termsin this articleprefacedby “Nyquist” wereso

samedto honor the remarkableSwedish-American communi~a-
tionsengineerHarry Nyquist(1889-1976)whowasbotha theorist
andinventor(well over 100patents)of greatdistinction.Although
virtually all of Nyquist’swork was donein the era whencorn-

-rrldrm 0
municationstechnologywas baaedon vacuumtubes,manyOfhis
ideasremain quiteusefulin themodernsilicon-chipworldof data
transferandprocessing.He is alsocreditedwith a key conceptin
Lower Nyquist UpperNyquist modeminformationtheoryI
frrsuency frW=ncy Of course, the major use of Nyquist’s work in geophysicsis
the elimination of alias frequencieson digitally recorded seismic
data. The Nyquist frequency is the highest frequencythat can be
obtainedfor a given samplinginterval. If the signalcontains higher
frequencies,then the spectrumof its sampledversion will exhibit
F’lguw 6. A sampled s&al Is alIased so that its fkequeneieslie a zone of aliased frequencies.This means, in practical terms, the
within the Nyqdst range. smaller the sampling interval, the larger the Nyquist interval. E

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION MARCH 1991

 BASIC SEISMOLOGY 9
The eikonal equation and the secret Pythagorean theorem
ENDERS A. ROBINSON, New York, New York, U.S.
DEAN CLARK, The Leading Edge, Tulsa, Oklahoma, U.S.
T he computation of traveltimes from a velocity function
is required in many seismic processing and modeling
schemes—in particular Kirchhoff depth migration and related
methods. One of the most popular methods uses the eikonal
equation, whose derivation goes back to the work of
Pythagoras about 2500 years ago.
Pythagoras, the Greek philosopher and mathematician,
was born on the small island of Samos in the Aegean Sea about
570 BCE, a century before the golden age of classical Greece.
He founded a philosophical and religious society in Croton
on the east coast of the tip of Italy about 532 BCE. His closest
followers lived permanently within the gates of the society,
had no personal possessions, were vegetarians, and followed
a strict code of secrecy.
Pythagoras, of course, is best known now for the
Pythagorean theorem in geometry—one of the most famous
and important of all mathematical formulas. However, other
mathematical discoveries are also attributed to Pythagoras, or
rather more generally to the Pythagoreans. A central belief of
Pythagoras and his followers was that “everything is num-
ber” which they described as a quantity that could be expressed
as a ratio of two integers (i.e., a rational number). The
Pythagoreans demonstrated that pitch could be represented
as a simple ratio of the lengths of equally tight strings. The
Pythagoreans are also credited with, by using their own the-
orems, disproving their own belief that all numbers are ratio-
nal. This was the famous demonstration that the square root
of two (as given by the length of the hypotenuse of an equi-
lateral right triangle with each leg equaling one) must be irra-
tional. According to legend, the Pythagoreans tried to keep
this discovery a secret, because it invalidated their entire rai-
son d’etre, and even killed the man who developed the proof.
The Pythagoreans knew that the sum of the angles of a
triangle is equal to two right angles and the generalization
that the interior angles of a polygon with n sides have a sum
of 2n-4 right angles and the sum of its exterior angles equals
four right angles. The Pythagoreans knew of the five regular
solids. It is thought that Pythagoras himself knew how to con-
struct the first three. It is unlikely that he would have known
how to construct the other two.
The Pythagoreans were also among the greatest
astronomers of their times. They taught that the Earth was a
sphere at the center of the universe, recognized that the orbit
of the Moon was inclined to the equator of the Earth, and were
among the first to realize that Venus the evening star was the
same celestial body as the morning star.
The famous Pythagorean theorem has a close relative,
here termed the “secret” Pythagorean theorem, that is the
basis of the modern differential eikonal equation. Figure 1
shows the right triangle BAO with horizontal leg OA, ver-
tical leg OB, hypotenuse AB, and altitude OC. The
Pythagorean theorem for this right triangle states that
OA2 + OB2 = AB2
or, in the familiar wording, the sum of the squares of the
0000 THE LEADING EDGE AUGUST 2003
Figure 1.caption
legs of a right triangle equals the square of the hypotenuse.
The “secret” Pythagorean theorem says that the sum of
squares of the reciprocals of the legs of a right triangle equals
the square of the reciprocal of the altitude:
OA-2 + OB-2 = OC-2
This strikes many, at first glance, as counterintuitive but
the proof is straightforward. Draw line OD collinear with OC
and with length that is the reciprocal of the length of OC (i.e.,
OD = 1/OC). Drop line DE perpendicular to the horizontal.
Then DE = OD sin θ. But sin θ = OC/OB. The key insight is
to recall that, by definition, (OD)(OC) = 1 and therefore
DE = OD (OC/OB) = OB-1
or, in words, the length of DE is the reciprocal of the length
of OB. Similarly the length of OE is the reciprocal of the
length of OA (i.e., OE = OA -1 ). Thus the “regular”
Pythagorean theorem for right triangle ODE, i.e.,
OE2 + DE2 = OD2
is the secret Pythagorean theorem for right triangle BAO, i.e.,
(1/OA)2 + (1/OB)2 = (1/OC)2
For example, if OA = 0.5000, OB = 0.8660, and OC =
0.4330, their respective reciprocals are OE = 2, DE = 1.1547,
and OD = 2.3094.
So, how is this secret Pythagorean theorem used in mod-
ern applied geophysics? In Figure 2, assume O is a point on
a wavefront at a given instant of time. Let line BCA be the
wavefront at subsequent time increment ∆t. The distance OA
is the horizontal space increment ∆x traveled by the wave-
front in time increment ∆t. Similarly, distance OB is the ver-
tical space increment ∆y traveled by the wavefront in the
same time increment. Since the wavefront is normal to the
raypath, distance OC is the space increment ∆s traveled
along the ray by the wavefront in time increment ∆t.
The secret Pythagorean theorem in terms of the legs ∆x
and ∆y and the altitude ∆s is
AUGUST 2003 THE LEADING EDGE 749
 Figure 2. caption

If we multiply this equation by (∆t)2, we obtain

and suddenly we are practically face-to-face with the very

useful geophysical concept of “slowness” which is defined
as the reciprocal of velocity. The apparent speed along the
horizontal direction is ∆x/∆t. Thus the apparent slowness
along the horizontal direction is ∆t/∆x. Similarly, the appar-
ent speed along the vertical direction is ∆y/∆t and the appar-
ent slowness along the vertical direction is ∆t/∆y. The actual
speed along the raypath direction is ∆s/∆t and thus the
actual slowness along the raypath direction is ∆t/∆s.
Therefore, the above equation can be interpreted as say-
ing the sum of squares of the apparent slownesses in the
coordinate directions equals the square of the actual slow-
ness in the raypath direction. Since the actual slowness in
the raypath direction is the reciprocal of seismic velocity v,
the equation may be written

Figure A1. caption

In the limit, as the increments become smaller, this equa-

tion becomes a differential equation, the so-called eikonal
equation. If we add the third spatial dimension, the eikonal
equation is
a c
d
In this equation, the function t(x,y,z) is the traveltime
(also called the eikonal) from the source to the point (x,y,z)
and 1/v(x,y,z) is the slowness (or reciprocal velocity) at that b
point. This is practically applied by using the eikonal equa-
tion to describe traveltime propagation in an isotropic Figure A2. caption
medium when the velocity function is known at all points
in space and, as an initial condition, the source or a partic- lection of old books. He also provided us with another proof
ular wavefront is specified. Furthermore, one must choose of the secret Pythagorean theorem which, like its more
one of the two branches of the solutions (either the wave famous relative, can be attacked from several directions.
going from the source or the wave going to the source). The In Figure A2, the Pythagorean theorem establishes that
eikonal equation then yields the traveltime field t(x,y,z) in a2 + b2 = c2. Furthermore, ab/2 = cd/2 because both equal
a heterogeneous medium, as required for migration and the area of the same triangle. The latter can be rearranged
other seismic processing needs. as c = ab/d. Squaring both sides and then substituting for
c2 leaves a2 + b2 = (ab/d)2. Finally, multiplying both sides
Appendix. The Pythagorean theorem has many proofs. by 1/a2b2 produces the secret Pythagorean theorem, 1/a2 +
Several are detailed in “Pythagoras and Jones” (TLE, 1985). 1/b2 = 1/d2. TLE
The most famous is Proposition 47 from Book 1 of Euclid’s
Elements. Its conclusion is shown in Figure A1 which orig- Corresponding author: ear11@columbia.edu
inally appeared in The English Euclide, Being the Firft SIX ELE-
MENTS of GEOMETRY, Tranflated out of the GREEK, with
Annotations and ufeful Supplements by Edmund Scarburgh
(Oxford, 1705). We are grateful to Chris Liner for letting us
copy this figure from one of the jewels of his personal col-

750 THE LEADING EDGE AUGUST 2003 AUGUST 2003 THE LEADING EDGE 0000
 Basic seismology 10: The King's Chamber
and seismic ray direction
ENDERS A. ROBINSON, Newburyport, Massachusetts, USA
DEAN CLARK, TLE Editor

When Pythagoras was about 18 years old, he went to the

island of Lesbos where he worked and learned from
Anaximander and Thales of Miletus. Thales had visited
Egypt and he recommended that Pythagoras go to Egypt.
Pythagoras arrived there around 547 BCE when he was in
his early twenties. He stayed in Egypt for 21 years and
learned a variety of things, probably including some geom-
etry that is routinely credited to the Greeks.
The evidence for this is imposing: the Great Pyramid of
Khufu, one of the largest buildings ever constructed. It was
erected at least 2500 years BCE and ranked as the world’s
tallest building for the next 43 centuries. Its volume stag-
gers the imagination. According to the Encyclopedia
Britannica, St. Peter’s in Rome, the cathedrals in Florence and
Milan, and Westminster Abbey and St. Paul’s in London Figure 1. The King's Chamber of the Great Pyramid.
could all fit inside at the same time. The precision of this
massive construction is impressive even by modern stan-
dards. Although thousands of presumably unskilled labor-
ers were involved, and the tools they used were very limited,
the lengths of sides of the base are within .05 percent of each
other (4 1⁄2 inches in 775 ft!).
This precision, on such a vast scale, suggests sophisti-
cated mathematical knowledge. Analysis of the geometry
of the King’s Chamber (a rectangular parallelepiped) inside
the Great Pyramid is further confirmation. The perfectly
aligned walls, ceiling, and floor are made of large polished
blocks of pink Aswan granite. These blocks are smooth to
within one-hundredth of an inch over 19-ft lengths. This
granite is so hard that experts say it would be extremely dif-
ficult to reach this accuracy today with the best machining
equipment. The chamber’s dimensions are length 1048 cm,
width 524 cm, and height 586 cm. Thus, modern dimensions
lead to the conclusion that the floor was designed to be a
Figure 2. The lower wedge of the King's Chamber.
double square since the length is exactly twice the width.
Was this the intent of the ancient Egyptian architects?
Yes, according to no less an authority than Isaac Newton(!)
who established to general assent that the floor of the King’s
Chamber was indeed 10 ǂ 20 royal cubits. With the royal
cubit as the distance parameter, the dimensions of the cham-
ber are length (a) = 20, width (b) = 10, and height (h) = 11.18
(Figure 1). For the volume to equal a double cube, the height
would have to equal the width. But it doesn’t; it exceeds the
width by 1.18 royal cubits. Why? The builders apparently
took extreme care to make the floor a double square so it’s
doubtful they just picked an arbitrary height. And, in fact,
they didn’t although it will require some arithmetic to
answer that question and confirm that the Egyptians were
pretty advanced in some areas of mathematics that weren’t
“finalized” for several thousand years.
Observe that the floor diagonal

.
Thus the chamber was designed so that its height—
Figure 3. The reconstructed King's Chamber.
(h) = = 11.18—is one half the length of the floor diag-
onal. This at least mathematically relates the awkward look-

MAY 2005 THE LEADING EDGE 485

 ing parameter 11.18 to another dimension but, again, we ask
why? Observe that the small side of the room has base
(b) = 10 and height (h) = . As a result the face diagonal
of the small side is .
The room diagonal (d) is the hypotenuse of the right trian-
gle with sides (f) = 15 and (a) = 20. The result is that
and that the triangle with sides f =
15, a = 20, d = 25 is a 3:4:5 right triangle. The harmonic pro-
portion of the room shows the intimate relationship between
1:2:3:4:5 and strongly implies that the designers of the King’s
Chamber knew the basic properties of right triangles at least
2000 years before Pythagoras.
The fascinating mathematics of the King’s Chamber can
be used to illustrate some mathematics of importance in
applied geophysics, even how the direction of a seismic ray
can be determined.
But first a word about vectors. A vector is specified by
magnitude and direction. For example, sailing orders might
be given by the vector expressed as 2 miles, ENE. This Figure 4. The directional derivative is the slope of the tangent line to
means sail 2 miles east-northeast. When a grid is imposed, the curve obtained by intersecting the surface with the vertical plane
though the direction line.
a vector may be specified by coordinates. If east is the x axis
and north is the y axis, then ENE represents an angle of π/8
radians. (There are π/2 radians in 90°, so there are π/8 radi-
ans in 90°/4.) Thus the x coordinate is 2 cos π/8 and the y
coordinate is 2 sin π/8 and the above sailing order in this
grid system would be the vector (1.74661, 0.97435). This
means sail 1.74661 miles east and then sail 0.97435 miles
north.
Two vectors can be multiplied in various ways. The most
common form is the dot product which is simply the sum
of the products of the respective coordinates. The dot prod-
uct of the vectors (1,2) and (3,4) is

(1 ǂ 3) + (2 ǂ 4) = 3 + 8 = 11

The first step in our application to geophysics is to

impose an (x, y, t) coordinate system on the lower wedge of
the chamber (Figure 2). In terms of a traditional 2D seismic
survey, the x-axis would be the horizontal axis, the y-axis
would be the vertical axis, the t-axis would be the travel- Figure 5. Case 1 (left) uses the least time, so BD is a raypath. In case 2
time axis, and the plane ABFE would be the traveltime sur- (right), the fictitious time is greater than the least time, so AD is not a
face t(x,y). Lines of constant time on the traveltime surface, raypath.
which represents the dependence of traveltime t on the hor-
izontal and vertical coordinates x and y, appear on the (x, vation of the surface along the line AB plus the rise in the
y) plane as contour lines (or level lines) which represent elevation of the surface along the line BD. In other words,
wavefronts. For example, line AB represents the wavefront
for traveltime of zero and line CD represents the wavefront FD = AB tan α + BD tan β
for traveltime of FD.
One of the fundamental concepts of calculus is that slope which is
is the rate at which an ordinate of a point on a line changes
with respect to a change in the abscissa. In other words, the FD = AB ∂t/∂x + BD ∂t/∂y
slope is the tangent of the angle of inclination of the line.
The angle of inclination (α) of line AB is zero, so its slope is Therefore the directional derivative is
tan α = 0. The angle of inclination (β) of line BF is not zero,
and its slope is tan β = FD/DB = h/b = 11.18/10 = 1.118. The tan δ = FD/AD = (AB/AD) ∂t/∂x + (BD/AD) ∂t/∂y
partial derivatives of a function give the rate of change of
a function in the directions parallel to the coordinate axes. If θ is the angle that the floor diagonal makes with the
Thus the partial derivatives of t(x, y) are ∂t/∂x = tan α = 0 x-axis, then AB/AD = cos θ and BD/AD = sin θ and the direc-
and ∂t/∂y = tan β = 1.118. tional derivative can be written
However, what is really needed is the rate of change of
traveltime t in an arbitrary direction—a directional deriva- tan δ = cos θ ∂t/∂x + sin θ ∂t/∂y
tive. In this case, if the specified direction is that of the floor
diagonal AD, the directional derivative is the slope tan δ of which can also be written as the dot product of two vectors
the room diagonal; i.e., tan δ = FD/AD = h/c = 11.18/22.36
= 0.5. tan δ = (cos θ, sin θ) ȱ (∂t/∂x , ∂t/∂y)
In developing a general formula for the directional deriv-
ative, we see that the height FD equals to the rise in the ele- The first vector on the right hand side is the unit vector

486 THE LEADING EDGE MAY 2005

in the required direction, namely
u = (cos θ, sin θ)
The second vector is called the gradient of t; that is
grad t = (∂t/∂x, ∂t/∂y)
As a result, the directional derivative can be written
more concisely as the dot product
tan δ = u ȱ grad t
The “robustness” of this definition of directional deriv-
ative is demonstrated by analyzing the King’s Chamber if
it were rebuilt as indicated in Figure 3. The traveltime sur-
face is the plane AGHE. This more complicated design has
a rise in elevation along the horizontal axis and along the
vertical axis.
The rise in the elevation of the surface is
DH = BG + JH = AB tan α + BD tan ȋ
which, as before, gives the required formula
tan δ = DH/AD = (AB/AD) tan α + (BD/AD) tan β
= u ȱ grad t
Today’s “modern” view of differentiability in two dimen-
sions, x and y, allows a function t(x, y) of two variables to be
differentiated at a point (x0, y0) if the surface it defines in (x,
y, t) space looks (in the limit) like a plane near the point. The
plane that t(x, y) resembles is called the tangent plane and thus
the formulas derived from the right angles in the King’s
Chamber apply for any differentiable function. Figure 4 illus-
trates the directional derivative.
So, how does this explain the direction a seismic ray is trav-
eling? Look at Figure 5. The medium is homogeneous and
isotropic, so the speed v is the same in all directions. Consider
two cases—case 1 in which the direction is given by line BD
and case 2 in which the direction is given by line AD.
The directional derivative for case 1 is tan β and the direc-
tional derivative for case 2 is tan δ. Thus case 1 represents the
case of the greatest directional derivative. In case 1, a particle
of energy travels distance BD in time FD. Time FD equals dis-
tance BD divided by speed v (i.e., FD = BD/v). In case 2, a par-
ticle of energy travels distance AD in time AD/v. Since
AD > BD, it follows that the (fictitious) time in case 2 is greater
than the (least) time in case 1. In fact, case 1 gives the mini-
mum time for all possible directions. Fermat’s principle states
that the actual path between two points taken by a ray is the
one which is traversed in the minimal time. (The modern, full
version of Fermat’s principle states that time must be extremal,
which means that the time can be either minimal or maximal
or at points of inflection.) Thus, case 1 must apply. In other
words, the ray goes in the direction with the greatest direc-
tional derivative.
For something to think about until the next paper, con-
vince yourself that the directional derivative is the greatest in
the direction of the gradient. (In other words, rays go in the
direction of the gradient of the traveltime). Use the equation
derived in this article (directional derivative = u ȱ grad t where
u is the unit direction vector and grad t is the gradient of the
traveltime). TLE
MAY 2005 THE LEADING EDGE 487
 Basic Seismology 11—Reflecting on the digital revolution
ENDERS A. ROBINSON, Newburyport, Massachusetts, USA
DEAN CLARK, TLE Editor

G eophysics in the early 1960s became the first science to

experience the digital revolution which today is universal,
from high definition television to the latest developments
in medicine. But the digital revolution did not result from
new knowledge; it resulted from new technology, namely
the digital computer. If the ancient Greeks had possessed
this technology, mathematicians might never have devel-
oped one of their most beautiful creations—calculus.
Eudoxus of Cnidos, who lived from 408 to 355 BCE,
studied with Archytas of Tarentum, a follower of Pythagoras.
Eudoxus visited Athens where he attended lectures by Plato,
spent over a year in Egypt where he studied astronomy with
the priests at Heliopolis, and then traveled to Cyzicus in the
Propontus where he established a school which proved very
popular and where he is believed to have introduced the
method of exhaustion.
The most famous application of the method of exhaus-
tion was to approximate the area or volume of a figure by
adding the areas or volumes of a sequence of inscribed fig-
ures within the given figure. The successive inscribed fig-
ures, in the limit, exhaust the given figure. Around 225 BCE,
Archimedes of Syracuse used the method of exhaustion to
determine areas and volumes of geometric figures. In other
words, Archimedes almost invented integral calculus about Figure 1. Polygon with twelve equal sides inscribed in a circle.
1900 years before Newton and Leibniz.
Archimedes used the method of exhaustion to determine This expression gives the area of the circle as one-half
one of the most important formulas in mathematics, the area of the product of the radius and circumference. The constant
of a circle. In Figure 1, a regular polygon (i.e., a polygon with π is the ratio of the circumference c to the diameter 2r of the
n equal sides) is inscribed in a circle. The polygon is made circle and substituting this definition into equation 3 yields
up of n triangles, one triangle for each side of the polygon the familiar closed expression for the area of the circle
and the area of the polygon is n times the area of one of the
triangles: (4)

(1) In summary, the approximation given by equation 1

where b and h are, respectively, the base and height of one
of the triangles. This equation gives a discrete approxima-
tion for the area of a circle and can approximate the true
value to any degree of accuracy if n is made large enough.
It is, in effect, a type of discrete algorithm that is required
for programming a mathematical operation on a digital
computer. If digital computers were available in the time of
Archimedes, it is conceivable that his efforts would have
stopped here. However, he continued to analyze equation
1 and rewrote it as
(2)
The term nb is the perimeter of the polygon. As n increases
and b simultaneously decreases (an interplay that is at the
heart of integral calculus), the area of the polygon approaches
the area of the circle. The perimeter becomes closer and
closer to the circumference c of the circle and the height h
of the triangle approaches the radius r of the circle. Thus,
Archimedes made the intuitive leap that the area of the
polygon would, as it approximated a circle to an increasing
degree, approach the value:
approaches the true value given by equation 4 in the limit,
as n tends to infinity and b tends to 0. Because equation 4
gives the exact value, it would appear more useful. However,
in order to use equation 4, it is necessary to know the numer-
ical value of π and until the time of Archimedes, it was not
known how to compute the value of this fundamental con-
stant. Empirical estimates, using methods such as wrapping
a cord around a cylindrical object or rolling a circular object
along a ruler, ranged from 3 to 3.2. Archimedes solved the
problem with another clever application of the theory of
exhaustion. He observed that the perimeter of a regular
polygon inscribed in a circle is smaller than the circumfer-
ence of the circle and, similarly, the perimeter of a regular
polygon circumscribed about a circle is larger than the cir-
cumference of the circle. Consequently, Archimedes realized
that he could calculate lower and upper limits on the cir-
cumference of the circle, from which he could estimate the
value of π, using equation 1 to determine the areas of these
two polygons. The greater the number of sides of the poly-
gons, the closer the limits would be. Archimedes did this
for polygons of 96 sides and found that
(5)

(3) Archimedes also anticipated differential calculus and

his method of finding the tangent to his spiral (the so-called

1030 THE LEADING EDGE OCTOBER 2005

 Archimedes spiral) was the only early “application” of dif-
ferential calculus. Other than finding the tangent to a cir-
cle, this is the only tangency result extant from antiquity. In
any case, the Greeks failed to recognize the connection
between area problems and tangency problems—now
known (as its name implies) as the monumentally impor-
tant fundamental theorem of calculus. This connection was
not discovered until the seventeenth century.
Can the Archimedes approach be applied to modern
geophysics? Indeed, it can determine the path of a seismic
ray. In the old days, a common method of hand migration
used a wavefront chart, which is a graph showing wave-
fronts and raypaths for an assumed vertical distribution of
velocity V(z). The raypaths are found by applying Snell’s
law. Each raypath is characterized by Snell’s parameter
p=sin i0/V0, where i0 is the angle of incidence and V0 is the
velocity at the horizontal datum plane z0=0. We can construct
the raypath chart by:
• dividing the medium into a large number of thin beds,
in each of which the veloc-
ity is constant
• for bed n, letting the thick-
ness (the vertical displace-
ment of the ray) be ∆zn
• letting the horizontal dis-
placement of the ray be ∆xn
• letting the traveltime be ∆tn
• letting in denote the angle of
incidence Figure 2. Raypath in thin bed.
Based on Figure 2, we quickly come up with the discrete
equations that would be used in a digital computer to obtain
the raypaths and the wavefronts:
(6)
However, in order to obtain a continuous solution, we
divide the last two equations by ∆z and take the limit as n
approaches infinity:
(7)
Integration then produces the values for x and t:
(8)
Equations 8 will generate an analytic solution for the ray-
paths and wavefronts. Raypaths are always perpendicular
1032 THE LEADING EDGE OCTOBER 2005
Figure 3. Wavefront chart for a linear increase of velocity with depth.
to the corresponding wavefronts. In the case of constant
velocity, the wavefronts are concentric circles and the ray-
paths are straight lines radiating from the center.
But the applied seismologist rarely has a case with a nice
constant velocity. In many situations, the velocity function
is the linear expression
V=V0+az
where V0 and a are constants. In this case, the wavefronts
are circles, but they are no longer concentric. The raypaths
are circles orthogonal to the wavefront circles. Figure 3
(taken from Sheriff's Encyclopedic Dictionary of Applied
Geophysics) is a wavefront chart showing the circular wave-
fronts and the associated circular raypaths. The chart acts
as an analog computer. For each pair of values T and sin i0
measured on the reflection seismogram, one pair of values
of x and z (which gives the underground position of the
reflector) can be read from the chart.
When the velocity functions are more complex than a
linear expression, it is very difficult or impossible to obtain
a closed-form solution by integration. As a result numeri-
cal analysis has to be used to approximate the integrals.
There was no easy way to compute the wavefront charts for
more complex velocity functions before the advent of the
digital computer. However with the lightning speed of the
digital computer in the 21st century, such approximations
are very quickly done and incredibly accurate. The analog
computers represented by the wavefront charts are no longer
needed for seismic migration.
In a sense, mathematics has come full circle. Archimedes
was the pioneer who took computing from discrete approx-
imations to exact analytic expressions. His work was the
beginning of the journey that led to the invention of calcu-
lus and, for the next few centuries, mathematicians spent
considerable effort looking for “closed form” solutions to
increasingly complicated integrals.
However, for most of the interesting problems in con-
temporary geophysics, the differential equations and/or
integral equations, although beautiful to see, are too diffi-
cult or unwieldy to solve analytically. The digital revolu-
tion takes us back (albeit inside the memory of the digital
computer) to numerical algorithms that use the type of dis-
crete approximations as originated by Archimedes. The
more things change, the more they stay the same. TLE
Corresponding author: endersrobinson@comcast.net
 Basic Seismology 12—Heron of Alexandria
and Fermat’s principle of least time
ENDERS A. ROBINSON and DEAN CLARK

H eron of Alexandria, a Greek math- can cover any distance in no time. For
ematician who lived approximately this reason, it was natural for him to use
10–70 CE and taught at the Museum the idea of minimum distance, for the
in Alexandria, wrote on the measure- idea of minimum time would not have
ment of geometric figures and appar- occurred to him.
ently had a flair for invention that was The bending of visual images of
centuries before his time. objects partially submersed in water
For example, he is credited with the was noted in antiquity. In modern ter-
first documented steam engine, the minology, when a ray of light passes
aeolipile or “wind ball” (Figure 1), which from one medium to another, it is bent
consists of a hollow sphere mounted on (refracted). In Figure 2, θ1 is the angle
a boiler by two pipes. The mounting that the incident ray makes with the
allows the sphere to rotate and, as the normal (at the boundary of two lay-
water boils, steam rises through the ers with different velocities) and θ2 is
pipes into the sphere, and ultimately the angle that transmitted ray makes
escapes from the sphere through two with the normal. In this example, light
bent outlet tubes (canted nozzles). The is traveling from a medium to a denser
escaping steam produces a rotary medium, and the ray of light bends
motion—a principle similar to that used toward the normal. The ancient Alex-
in jet propulsion. The steam engines of andrian astronomers intuitively re-
the 18th century were in part based on alized the need to correct for atmos
this design. Figure 1. Model of Heron’s steam engine in Smith pheric refraction in computing times
Heron also developed instruments College Museum of Ancient Inventions. of rising and setting heavenly bodies.
for measuring distances and roads, and Ptolemy, slightly younger than Heron,
is credited with the formula that gives the area of a triangle made measurements of the angles of incidence and refrac-
from its sides. If a, b, and c are the lengths of the sides of a tri- tion for the passage of light from air to water and tried to
angle, meaning the semiperimeter s is (a + b + c)/2, then find a mathematical relationship between the two angles.
Heron’s formula states the area of the triangle is However, this problem is harder that the reflection rela-
tionship and Ptolemy was unsuccessful.
In fact, progress was not made until Johannes Kepler
This formula is still used in land surveying. (1571–1630) showed in 1611 that for any given pair of mate-
Heron’s inventions are notable for another modern con- rials the ratio θ1/θ2 is (approximately) fixed for small angles.
cept, self-regulating feedback control systems. For example, In other words, the two angles are proportional when the
he invented a self-filling wine bowl which had a hidden float angles are small.
valve that automatically sensed the level of wine in a bowl. Willebrord Snell (1580–1626) performed a series of care-
Heron perhaps anticipated rotary drilling in his description ful refraction measurements and, by using his observational
of a machine, called the Cheirobalistra, which had a refined data, solved the problem in 1625 when he found that a lin-
screw-cutting technique that would enable it to bore a tunnel ear relationship exists between the sine of the incidence
through a mountain. He also envisaged an instrument called angle and the sine of the refraction angle, as expressed by
the hodometer for measuring distances traveled by wheeled the equation
vehicles.
But Heron’s most important contribution (at least to
anyone using reflection seismology) is that he established
the law of reflection—that the angle of incidence is equal to where n is a fixed number characteristic of the two media.
the angle of reflection. This is now known as Snell’s law and the fixed ratio n is called
About four centuries earlier, Aristotle had observed that the relative refractive index of the two media. Table 1 shows
most motion appears to be in either straight lines or circles. that the relative refraction index for air to water is 1.33. Of
The straight line is the shortest route between two points, while course, as we all know, Snell’s law is basic to understand-
the circle is the shortest periphery for a given area. Aristotle’s ing the paths taken by seismic waves. Note that for small
observation suggested to Heron a common generating prin- θ, sin θ ≈ θ. This explains Kepler’s version of the law. Snell
ciple, namely minimization. In his Catoptrica, Heron showed died in 1626, just a year after his discovery and did not pub-
that the path taken by a ray of light reflected from a plane mir- lish his results, so his result was not widely known.
ror is shorter than any other reflected path that might be René Descartes (1596–1650) discovered a law of refrac-
drawn between the source and point of observation. Heron tion about the same time and published it in 1637. Descartes
then applied his minimization principle to prove that in a mir- also provided a physical explanation to “fit” the law of
ror, the angle of reflection is equal to the angle of incidence. refraction to his mechanistic philosophy of nature. His mech-
In other words, were the angles to differ from each other, the anistic description of light propagation represented light as
distance that the light traveled would not be the least possi- ballistic particles or corpuscles and, due to a greater attrac-
ble. Heron, as did every one else until the modern scientific tion of the corpuscles by a denser medium, the light parti-
revolution, thought that light travels with infinite speed and cles must speed up when they pass into a denser medium.

556 THE LEADING EDGE MAY 2006

Figure 2. Refraction: A light ray is bent toward the normal (dotted line)
when it passes from a less dense to a more dense medium.
Figure 3. Descartes law of refraction.
Descartes’ derivation of his version of the law of refrac-
tion was based on two premises which we now know are
wrong—that a dense medium must transmit light more
effectively (i.e., with more velocity) than one which is not
as dense, and that the tangential component of his ray vec-
tor remains constant as the ray passes through a boundary.
Figure 3 shows how, according Descartes, a ray of light
Figure 4. Derivation of Fermat’s law of refraction. The raypath ADB is
slight displacement from the raypath distance ACB.
Figure 5. Blow-up of Figure 4 near the interface.
would act when passing into a denser medium
when v1 and v2 are the respective velocities.
Since Descartes assumed that the tangential
components are equal (his second false
premise), it follows immediately that
which leads to the (incorrect) Descartes law of
refraction
However, remember that Descartes’ also
assumed (wrongly) that v2>v1 and it follows
from this that θ2<θ1. This is the correct result
and correctly predicts that the light ray bends
toward the normal in passing into a denser
medium. If we (incorrectly) let n = v2/v1, we
see the two false premises in effect cancel each
other and Descartes’ result is the same as
Snell’s.
Pierre de Fermat, a contemporary of Descartes, contested
this result because he believed that light must travel slower
(not faster) in a denser medium. Fermat had studied the
works of the ancient mathematicians in the original Greek
and, consequently, was inspired to derive the law of refrac-
tion by using Heron’s assumption that that light always
prefers the shortest possible path. However, it was immedi-
MAY 2006 THE LEADING EDGE 557
 ately obvious to him that it couldn’t be shortest path, as
Heron reasoned for reflection, because the shortest path
between a point above water and a point below water would
be a straight line, and there would be no refraction (i.e., no
bending) at all. Explaining this difference caused Fermat to
make one of the most important conceptual leaps in the his-
tory of physics. His intuition told him that light moves slower
in the denser medium. Thus, instead of assuming that light
travels along a path that minimizes distance, Fermat reasoned
that light travels along the path that minimizes time. This is
now known as Fermat’s principle of least time.
Fermat’s conceptual leap was astounding because he had
to assume (1) that the speed of light is finite (which had not
yet been demonstrated), (2) that light has a fixed charac-
teristic speed in each substance, and (3), counterintuitively,
that the speed is lower in denser media (which is opposite
to the view of Descartes and counter to the physics of sound).
Furthermore, since calculus had not yet been invented,
Fermat also had to do pioneering work in mathematics to
derive the law of refraction based upon his principle of least
time. His method of minimization involves ingenuity that
is very close to differential calculus. Figure 4 shows two
neighboring paths, labeled old path and new path, from
point A to point B. Figure 5 shows the mathematics near the
interface. Note the perpendiculars DD’ and CC’. The upper
segment AD of the new path is shorter (by the amount dp)
558 THE LEADING EDGE MAY 2006
than the upper segment AC of the old path. In compensa-
tion, the lower segment DB of the new path is longer (by
the amount dq) than the lower segment CB of the old path.
The new path “gains” the time it would take to go the dis-
tance dp, but “loses” the time it would take to go the dis-
tance dq. Time is distance divided by velocity. Thus the net
time difference dt is dp/v1 – dq/v2. Some basic trigonometry
reveals that dp = dx sin θ1 and dq = dx sin θ2. As a result, the
net time difference can be written
and then rearranged as
The shortest possible time for light to travel from A to
B would occur when the difference between the two “legs”
of the path was 0. Setting dt/dx to 0 immediately leads to
which can be rearranged as Fermat’s law of refraction
Since Fermat assumed that v2<v1, θ2 must be smaller
than θ1 and, in accordance with Snell, a light
ray must bend toward the normal when pass-
ing into a denser medium.
Fermat published his findings in 1650 but
his derivation was not recognized for several
years and, in fact, ignited a major scientific
controversy because Fermat had not provided
a physical explanation for his assumption
that the light ray would automatically follow
a path of least time. Only a direct measure-
ment of the speed of light in two media with
different refractive indices would suffice. This
could not be done at the time and indeed was
not possible until the middle of the 19th cen-
tury.
Without this knowledge about the veloc-
ities involved, it could not be determined
whose derivation is correct and the greatest
physicists of the time (or any time) disagreed.
Hooke and Newton were among those who
believed that light travels faster in denser
media. Huygens agreed with Fermat that the
light travels slower in denser media and, as
will be shown in the next article in this series,
used this to derive the law of refraction based
on his wave-theory of light.
Perhaps, the clinching confirmation that
Fermat was correct came with calculus and
is one of the greatest demonstrations of the
power of this technique. It was described in
an earlier article in this series (February 1987)
and, although very familiar to most inter-
ested in the history of science, it is well worth
revisiting. TLE
Corresponding author: endersrobinson@comcast.net
 Basic Seismology 13—Huygens’ principle
ENDERS A. ROBINSON and DEAN CLARK
A ccording to a story, perhaps apocryphal, the great Dutch
scientist Christiaan Huygens (1629–1695) dropped a stone
into the canal next to his house, intently observed the cir-
cular wavelet that moved out across the water surface, and—
like Pythagoras and Plato—intuited that a world of
perfection was behind the imperfect visible world and this
perfect world was constructed of perfect mathematical and
geometric formulations. Of course, the wavelets that
Huygens supposedly observed as a boy were never perfectly
circular, but his mind held a clear understanding of a per-
fect circle and it can be argued that, in the spirit of Plato,
Huygens spent his life in uncovering the massively impor-
tant role played by the circle in science.
The Netherlands is a great seafaring nation, and
Huygens’ first scientific/technological contributions
involved improvements to the two most important naviga-
tional tools of the 17th century, the telescope and the clock.
Huygens, helped by his brother Constanign, designed a
telescope that was far superior to contemporary devices
and in March 1655 he discovered Saturn’s moon Titan. He
was also able to explain the curious extension of Saturn,
which had intrigued astronomers since Galileo first observed
it in 1610: Saturn was encircled by a ring, thin and flat,
nowhere touching, inclined to the ecliptic.
In that period, the central problem of navigation was
determining longitude. Longitude can, in effect, be measured
by time because if the difference in local time at two points
is known, the longitudinal distance between them can be
computed. However, in the first part of the 17th century, this
was not a practical option because the existing mechanical
clocks were not sufficiently accurate. Galileo had discovered
that a pendulum could be used as a frequency-determining
device for a clock and, although many consider him the
father of this scientific breakthrough, he never built such a
clock. Huygens is the true inventor of the pendulum clock
in which the escapement counts the swings and a driving
weight provides the push. In effect, the escapement is a
feedback regulator that controls the speed of this type of
mechanical clock. Huygens produced his first clock in
December 1656, and it was much more accurate than con-
temporary clocks. It is not an overstatement to contend that
this was one of the great technological breakthroughs in his-
tory. Pendulum clocks were the most accurate clocks in the
world for the next 300 years. The invention by Huygens of
the first accurate clock can be considered the beginning of
the modern world, which is based on science and technol-
ogy, because it permitted much more sophisticated experi-
ments and detailed measurements.
Huygens was able to construct such a clock because of
his investigations of the mathematics of the circle, and this
would lead him to additional discoveries that are of major
importance in modern geophysics.
Galileo believed that a pendulum is isochronic; in other
words, that the period of a pendulum does not depend on
the amplitude of its swing. Huygens, via mathematics, found
that a pendulum swinging through the arc of the circle is
not isochronic. It only appears isochronic when the length
of the arc is quite short relative to the length of the pendu-
lum. This property gives a clock with a long pendulum an
advantage over a clock with a short pendulum. However,
the pendulums of the early clocks were kept short and light
1252 THE LEADING EDGE OCTOBER 2006
Figure 1.
to minimize the amount of energy needed to keep them in
motion. As a result, the early pendulum clocks had very wide
pendulum swings, which decreased their accuracy.
Huygens, in turning to mathematics, originated what is
now known as differential geometry and the idea of cur-
vature, a value that shows how much a curve deviates from
a straight line.
Note the curve MPQ in Figure 1 and that, at P, PT is the
tangent line and PN is the normal line (i.e., the line per-
pendicular to the tangent). The figure also shows four cir-
cles (A–D), each tangent to the curve at P. Such circles are
called tangent circles and each has its center on the normal.
There is a unique tangent circle that fits curve MPQ at P bet-
ter than any other. This optimum-fitting circle (D, in this case)
may be described as the circle that “kisses” the curve and
it is known as the osculating (derived from the Latin for kiss-
ing) circle. The curvature (κ) of MPQ at P is defined as κ =
1/r, where r is the radius of the osculating circle (D). The
sharper the curve at P, the larger its curvature. The flatter
the curve at P, the smaller its curvature—-with the limit
being a curve with no curvature, namely a straight line that
would coincide with the tangent. In summary, the osculat-
ing circle is the circle that touches a curve (on the concave
side) and whose radius equals its curvature. Just as the tan-
gent is the line that best approximates a curve at a point,
the osculating circle is the circle that best approximates the
curve at the point.
Huygens then developed the theory of evolutes. The evo-
lute of a curve is the locus of the centers of the osculating
circles of the curve.
In Moby Dick (1851), Herman Melville wrote: “I was first
indirectly struck by the remarkable fact that all bodies glid-
ing along a cycloid will descend from any point in precisely
the same time.” The Latinized word for same time is tau-
tochrone and, in mathematics, the tautochrone problem con-
sists of finding the curve along which a bead placed
anywhere on the curve will fall to the bottom of the curve
in the same amount of time. The solution is a cycloid, a fact
first discovered by Huygens. The cycloid is intimately related

Figure 2.
Figure 3.
to the circle; a cycloid is the locus of a point on the rim of a
circle rolling along a straight line. In Horologium Oscillatorium
(1673), Huygens gives a complete mathematical description
of an improved pendulum clock and calls such a device a
cycloidal clock because its pendulum is forced to swing in
an arc of a cycloid. Huygens did this by suspending the pen-
dulum (made up of a bob on a wire string) at the cusp of
the evolute of the cycloid. The cycloidal clock was extremely
accurate, but unfortunately the movement caused an exces-
sive amount of friction.
Meanwhile, Robert Hooke,
another figure with a prominent role
in establishing the fundamentals of
geophysics, invented the anchor
escapement for a pendulum clock. The
anchor escapement required a smaller
angle of swing than the angle required
by the escapements of the early pen-
dulum clocks. Pendulum clocks
became so accurate that the cycloidal
clock quickly became passé. However,
Huygens made one more great con-
tribution to the measurement of time
when, in 1675, he built a chronometer
that used a balance wheel and a spi-
ral spring instead of a pendulum.
Figure 4.
Balance wheels and spiral
springs were the basis for
almost all watches until the
invention of the quartz crystal
oscillator in the twentieth cen-
tury.
Although the cycloidal
clock was of practical use for
only a limited time, it is his-
torically important because it
can be viewed as the first suc-
cessful design of an intricate
apparatus based on higher
mathematics. Heron of
Figure 5. Alexandria and Leonardo da
Vinci used mechanical princi-
ples in order to design their inventions, but the mathemat-
ics involved was essentially just Euclidean geometry which
dates from about 300 BCE. The introduction of higher math-
ematics to accomplish mechanical design gives Huygens a
claim to being the father of modern technology.
In Traité de la Lumière (1690), Huygens made one of the
great contributions to theoretical physics when he postulated
that light was a wave. Let us attempt to speculate on
Huygens’ thinking at the time. We can imagine that, in
working out the theory of evolutes, Huygens would have
asked questions like: What is the evolute of a circle? What
is the evolute of a straight line? He would have come up
with something like the diagram in Figure 2, which shows
that the evolute of circle is the center of the circle. In other
words, the osculating circle for circle C is circle C itself and,
similarly, that the evolute of a straight line is a point at infin-
ity. Huygens then would have logically progressed to con-
sidering diagrams such as those in Figure 3. The left diagram
shows how a spherical wavefront propagates. The right dia-
gram shows how a plane wavefront propagates. In other
words, in the spirit of Plato, Huygens used the perfect world
of circles to explain the intricacies of wave motion. In 1921
OCTOBER 2006 THE LEADING EDGE 1253
 the other downward. There are two envelopes—the enve-
lope that gives the upgoing reflected wavefront and the
envelope that gives the downgoing refracted wavefront.
The law of reflection and refraction can be derived by
analyzing the part of the incident wavefront that lies between
rays AC and BDF as point C contacts the interface (CF). Let
∆t represent the time increment it takes the wave to travel
from D to F which means that DF = ν1∆t. The two wavelets
emanating from point C are different because the wavelet
above the interface is a semicircle with radius ν1∆t and the
wavelet below the interface is a semicircle with radius ν2∆t.
The envelopes at time t + ∆t are given by the tangent lines
FG and FE.
θi between the incident wavefront CD and the interface
is the angle of incidence. θr between the reflected wavefront
FG and the interface is the angle of reflection. θt between
the refracted wavefront and the interface is the angle of
refraction. These angles are part of three right triangles
(CFD, CFG, and CFE) which have the common hypotenuse
CF. Thus the sines of these three angles have a common
denominator, that is

Figure 6.

at Belle Isle, Oklahoma, USA, J. C. Karcher was the first per-

son to record a seismic reflection line. Figure 4 is Karcher’s However, DF = CG = ν1∆t, and CE = ν2∆t which means
1921 migration diagram of the Viola interface as given by that the above equations can be written
an envelope of circular arcs. Each time a geophysicist does
a prestack depth migration, he or she is using a method
based upon the fundamental concept known as Huygens’
principle which may be summarized as follows: Because their sines are equal, it follows that the angle of
incidence is equal to the angle of reflection. It also follows
Given a wavefront at a given instant of time, each point that
on the wavefront emits a spherical wavelet (Figure 5). The
wavelets are in phase with the original wavefront and
propagate outward with the same speed. The wavelets
constructively interfere and their envelope forms a new which is Snell’s law—one of the mathematical underpin-
wavefront. In the same manner the envelope of wavelets nings of seismology and a marvelous proof of a relation-
from this new wavefront gives the next wavefront. ship that had been sought for centuries.
Thus the wavefront method of Huygens correctly, and
Huygens imagined that this process repeats itself as the with a dramatic simplicity and elegance, generates the laws
wave propagates. If the medium is homogeneous and of reflection and refraction. The work of Huygens on the
isotropic, the spherical wavelets may be constructed with telescope and the clock would secure his place in the annals
finite radii. On the other hand, if the medium is inhomo- of exploration geophysics. His principle of constructing
geneous, the wavelets will have infinitesimal radii, and the wave motion by the use of secondary wavelets makes
magnitudes of the radii will depend on the wave velocity Huygens one of the great pioneers of exploration geophysics
of the medium at the respective centers of the wavelets. and, indeed, one of the major figures in the creation of mod-
Calculus is needed to deal effectively with these infinitesi- ern science. TLE
mals.
Huygens used his wave theory to
finally establish the laws of refection
and refraction which, particularly the
latter, had been sought since ancient
times. In Figure 6, a plane interface sep-
arates the upper medium from the lower
medium. Assume the wave velocity in
the lower medium is greater than in the
upper medium, i.e. ν2>ν1.
Figure 6 shows a downgoing plane
wavefront (CD) in the upper medium
that is diagonally incident on the inter-
face. As each point on the wavefront
arrives at the interface, it behaves
according to Huygens’ principle and
emits two wavelets, one upward and

OCTOBER 2006 THE LEADING EDGE 1255

 Basic Seismology 14—Michael Faraday and the eikonal equation
ENDERS A. ROBINSON and DEAN CLARK
I t will probably surprise some to see the name of Michael
Faraday (1791–1867) in this series because he is rarely, if ever,
associated with seismology or mathematics. However, some
of Faraday’s concepts provide an effective analog for some
fundamentals of exploration seismology. It is also pleasant to
revisit Faraday’s life which is one of the most inspiring sto-
ries in the history of science and which itself has some lessons
that bear re-examination.
As has often been told, Faraday metamorphosed, in a
remarkably short time, from barely literate apprentice book-
binder into one of the world’s most influential scientists. His
ingenious experiments yielded some of the most significant
principles and inventions in scientific history. He developed
the first dynamo (in the form of a copper disk rotated between
the poles of a permanent magnet) which is the precursor of
modern dynamos and generators. He (and independently
Joseph Henry) discovered electromagnetic induction, and a
vast industry resulted from this work. During the 1830s, he
laid the foundations of classical electromagnetic field theory,
later fully developed mathematically by James Clerk Maxwell,
and his concept of a field based on lines of force is a funda-
mental principle of modern theoretical physics. Faraday made
so many discoveries during his 40-year career that the 1981
edition of the Encyclopedia Britannica has six separate entries
on the results of his work in addition to a long biography. The
combination of the amazing life and astonishing discoveries
prompted the famous novelist Aldous Huxley to write: “Even
if I could be Shakespeare, I think I should still choose to be
Faraday.”
Static or frictional electricity was known to the ancients.
Objects become charged in two ways, called positive and neg-
ative, governed by the rule that like charges repel and unlike
charges attract. A field of fixed charges, an electrostatic field,
is described by two physical quantities, the field strength vec-
tor E and the potential φ.
The field strength vector is defined as E = F/q where q is
the test charge and F is the force acting on this charge at the
given point in the field. Faraday proposed the use of lines of
force as a means to visualize an electric field. A line of force
is a straight or curved line whose tangent at each point coin-
cides with the direction of the field strength vector. Lines of
force begin at positive charges and end at negative charges.
The concept of a potential is now almost universally regarded
as representing a powerful advance in knowledge but, as will
be discussed later, that was not always true.
In the electrical case, the potential φ at a point is the work
per unit charge that would be necessary to carry a positive
test charge from infinity to that point. The work is then said
to be stored in the field as potential energy. Because potential
is a measure of work, potential is not directional. The poten-
tial at a point is represented by a real number (i.e., a scalar),
while the force acting on a test body would be a vector. How
is the concept of a potential used? The answer is that E can be
obtained from the potential by simply taking the gradient (usu-
ally designated by grad). Like a sheep in wolf’s clothing, a
gradient is a derivative in vector’s clothing. The basic equa-
tion which says that the negative gradient of the potential gives
the electric field is
-grad φ = E
24 THE LEADING EDGE JANUARY 2007
Figure 1. The field of a dipole shows mirror symmetry about the center
point of the dipole. The equipotential surfaces (dotted lines) are every-
where perpendicular to the lines of force (solid lines).
Any problem with specified charges can be solved by
first computing the potential, and then using this equation
to get the field. There is a physical significance to this equa-
tion. Because E is the gradient of a scalar potential, it fol-
lows from the mathematics of vector calculus that the curl
of E must vanish. This condition represents a fundamental
property of electrostatics; namely, the electric field in elec-
trostatics is curl-free. This equation can be used to give a
geometrical description of an electric field. First, the field
lines (that is, lines that are always tangent to the electric field
vector) are drawn. Next, the surfaces of equal potential are
drawn. Because the electric field vector is the negative gra-
dient of the potential, it follows that E is perpendicular to
the equipotential surface. If E were not, then E would have
a component in the surface and, in turn, that component
would be changing in the surface which means that the sur-
face would not be equipotential. Consequently, to avoid
this contradiction, the lines of force must be always per-
pendicular to the equipotential surface (Figure 1).
In common usage, the quotient of the change in eleva-
tion divided by the change in distance is often called the
gradient. For example, if the hill rises 1 m in a horizontal
distance of 5 m, the (common-usage) gradient would be 1
over 5. In mathematical language, this (common-usage) gra-
dient is actually the directional derivative in the direction
of the horizontal separation. In other words, when a sur-
veyor speaks of gradient, he often means directional deriv-
ative. In order for the directional derivative to be a gradient
(in the mathematical sense) the direction of the horizontal
separation must be in the steepest direction. It is useful to
keep this distinction between directional derivative and gra-
dient. The gradient is a two-dimensional vector that points
in the direction of the greatest slope. The dot product of a
gradient vector and an arbitrary direction vector gives the
rate of change of the function in that direction.
The concept of potential was first developed to solve
problems of gravitational attraction. In fact, a simple grav-
itational analogy is helpful in explaining potential. Over a
small region, gravity may be taken as uniform and parallel
(simply, down). We do work in carrying an object up a hill.
This work is stored as potential energy, and it can be recov-
ered by descending in any way we choose. An imaginary
terrain, with the accompanying topographic map, can be
used to visualize a potential function. Topographic maps pro-
vide information about elevation of the surface above sea
level by contour lines. Each point on a contour line has the
same elevation, so a contour line represents an equipoten-
tial curve. A set of contour lines tells the trained interpreter
the shape of the terrain: hills are represented by concentric
loops, stream valleys by vees; steep slopes have closely
spaced contour lines, gentle slopes have widely spaced con-
tour lines. The contour interval is the elevation difference
between adjacent contour lines. Using our gravitational field
analogy, the contour lines on a topographic map are lines
of constant elevation above sea level and hence of constant
gravitational potential energy. If we let a ball roll down a
mountain, the ball rolls down a path perpendicular to the
contour lines—i.e., down the steepest descent or negative
gradient. So if we could measure the contour lines before
releasing the ball, we could then predict the path it would
follow down the mountain. The downward path is the curve
of steepest slope or negative gradient.
This brings us (fortunately or unfortunately depending
upon individual predilection) back to the mathematics and
the link between Faraday’s theory about potential fields
and exploration seismology. Vectors are those quantities,
often denoted in bold face, that have a magnitude and direc-
tion. A fairly detailed description of the basic mathematical
manipulation of vectors can be found in Sheriff’s Encyclopedic
Dictionary of Applied Geophysics. The key point for this arti-
cle is that, in a Cartesian coordinate system, a two-dimen-
sional vector may be represented as the ordered pair of real
numbers. If i represents the unit vector in the x direction,
and j the unit vector in the y direction, then r = xi + yj. The
magnitude of the vector is as follows:
The dot product of two vectors v and w is the scalar value
v
w =vwcos θ = vw cos θ
where θ is the angle between v and w. Two nonzero vectors
are orthogonal (perpendicular) if and only if their dot prod-
uct is zero.
What is the counterpart of potential in seismic theory?
We assume, for simplicity, an isotropic medium. As devel-
oped in Basic Seismology 10 (TLE, 2005), the traveltime
function t(x,y) is analogous to the potential. This function
is a scalar function that represents the traveltime surface as
a function of x and y. The x axis is the horizontal axis and
the y axis is the vertical (depth) axis. The gradient of t(x,y)
is the vector function whose first component is the partial
derivative of t with respect to x and whose second compo-
nent is the partial derivative of t with respect to y. In other
words, the gradient of t is
Let u = (cos θ, sin θ) be the unit vector in a given direc-
tion. The directional derivative is the dot product
tan δ = u
grad t
where δ is the angle of elevation. An important application
of this expression for the directional derivative is in finding
the direction of steepest increase (or decrease) of the func-
tion t(x,y). The directional derivative is greatest in the direc-
tion when the dot product of u with the gradient vector is
greatest, which is exactly when u points in the same direc-
tion as the gradient. Another application is in finding the
direction of no increase (or decrease) of the function t(x,y).
The directional derivative is zero in the direction when the
dot product of u with the gradient vector is zero, which is
exactly when u points in the same direction as the contour
line. The value of the function neither decreases nor increases
along a contour curve. The gradient vector is perpendicu-
lar to the contour curve and the magnitude of the gradient
indicates the steepness of the slope.
What is the counterpart of line of force in seismic the-
ory? If the potential t(x,y) is the seismic traveltime function,
then its contour curves represent wavefronts. A vector field
is a rule that assigns a vector, in our case the gradient, grad
t(x,y), to each point. In visualizing a vector field we imag-
ine there is a vector extending from each point. Thus the
vector field associates a direction to each point. If the seis-
mic energy (for the most part) moves in such a manner that
its direction at any point coincides with the direction of the
gradient at that point, then the curve traced out is called a
seismic ray. A ray corresponds to the flow line (or the line
of force) in other disciplines.
A vector function can be used to represent a space curve.
Here for simplicity we use only two dimensions, but a third
dimension can always be added. Let s(t) denote arc length
measured along the curve. Suppose the curve is defined by
the equations x = x(s) and y = y(s). This curve can be repre-
sented by the vector r(s) = (x(s), y(s)). The vector extends
from the origin to the point on the curve. As s increases, the
tip of the vector traces out the curve. The derivative vector
τ of the particle is then given by
The derivative vector points in the direction of move-
ment and thus is tangent to the curve. The differential of arc
length is given by ds = dr, so the magnitude of the deriva-
tive vector τ is
Thus τ is the unit tangent vector to the space curve. The
requirement that the tangent τ and the gradient have the
same direction is
grad t(x,y) = n(x,y) τ(x,y)
where n(x,y) is some scalar-valued function. The flow lines
of the vector field n(x,y) t(x,y) are the same as those of grad
t(x,y), since the function n(x,y) does not affect the direction.
If we take the squared magnitude of the above equation, we
obtain
because τ is a unit vector. In Basic Seismology 9 (TLE, 2003),
the eikonal equation is given by
Thus we see that n(x,y) = 1/v(x,y), or, in other words, n
is the reciprocal of the seismic velocity. Strictly, velocity is
a vector, but in seismology the term velocity usually refers
to speed, that is, the propagation rate of a seismic wave with-
JANUARY 2007 THE LEADING EDGE 25
 out implying direction. Another word for speed is swiftness.
As a result n is called slowness. This we have shown that
grad t = nτ
where t is the seismic wavefront, n is the slowness, and τ is
the unit tangent to the seismic ray. This equation is the vec-
tor eikonal equation. The left side involves the wavefront;
the right side involves the ray; and the connecting link is
the slowness. At shallow depths, the slowness is large, so
the gradient is large and the wavefronts are closely spaced.
At large depths, the slowness is small, so the gradient is small
and the wavefronts are widely spaced. In the seismic case,
instead of equipotential surfaces we have wavefronts;
instead of lines of force, we have seismic rays; but the math-
ematics is the same.
This seems rather straightforward and, indeed, rather
simplistic to those with basic training in vector calculus. But
this was not always the case. Faraday’s initial presentation
of lines of force, on 3 April 1846, was greeted with ridicule
by his most eminent contemporaries. One went so far as to
publish the suggestion that the self-taught Faraday, who had
virtually no education in or knowledge of mathematics,
should leave theoretical physics to the properly trained.
This is a depressingly common scenario in science. Great
theoretical breakthroughs are routinely dismissed out of
hand by the contemporary experts, only to be later adopted
by younger scientists and rather quickly proved. For exam-
ple, we cite one of the great “handoffs” in the history of sci-
ence. Maxwell, 40 years younger than Faraday, published
the first of his three famous papers on electromagnetism in
26 THE LEADING EDGE JANUARY 2007
1855, the year that Faraday retired, and had his entire the-
ory worked out in less than a decade. Other examples of
initially scorned ideas, of more than minimal import to geo-
physics, are Fourier analysis and plate tectonics. (The very
eminent geophysicist J. Tuzo Wilson told one of the authors
that he had been “roundly booed at the AAPG convention”
after making a presentation in support of plate tectonics in
the early 1960s.)
This initial resistance by great scientists to some of the
greatest scientific breakthroughs is perhaps a bigger mys-
tery than those which Faraday unraveled.
Suggested reading. A recent, entertaining, and well written
account of Faraday’s remarkable life and career is The Electric
Life of Michael Faraday by Alan Hirshfield (Walker, 2006). In
Maxwell on the Electromagnetic Field (Rutgers University Press,
1997), Thomas K. Simpson provides an almost line-by-line
analysis of Maxwell’s 1855, 1861, and 1864 papers which devel-
oped the mathematical foundation of the theory that began with
Faraday’s work in the 1830s. Simpson also provides extensive
background on the lives and accomplishments of Faraday,
Maxwell, and Lord Kelvin (who did much of the initial work
on the mathematics used by Maxwell). The book has extensive
illustrations and commentaries designed to assist readers with
limited knowledge of mathematics and physics. It is also quite
witty in places, which would have delighted Maxwell whose
fondness for clever wordplay was well known during his life-
time. TLE
Corresponding author: Endersrobinson@comcast.net
 Fermat and the principle
of least time
By ENDERS ROBINSON and DEAN CLARK
A previous article in this series, Descartesas geophysicist
(TLE, August 1985) prompted a letter from Sven Treitel com-
menting that Descartes indeed had a role in the discovery of
the law of refraction (now generallycredited to Willebrord Snell)
but Descartes’ work had one major flaw - it essentially put
things exactly backwards!
Treitel is not, of course, the first astute observer to note this.
One of his predecessorswas Pierre de Fermat (born 1601 in
Beaumont-de-Lomagne, France; died 1665 near Toulouse) and
one of the results of this observation was Fermat’s Principle
which is basicto our understandingof the propagationof seismic
waves through the earth.
Fermat is best known for his famous (perhaps notorious)
“Last Theorem” in which he stated - actually scribbled in the
margin of a book - that he had found that the equation x”
+ y” = z” (n > 2) did not have a solution in whole numbers
and that the proof was simple but he did not have room to write
it in the available space. Fermat’s proof has never been found
and three centuriesof mathematicians have since failed, despite
great effort, to confirm his statement. Empirical evidence,
however, supports Fermat’s contention; computers have
established that the equation has no solution for the integers
between 2 and 125,100.(That was as of March, 1986, so the list
is probably longer by now.)
It is very possible that Fermat did discover such a proof for
he waswithout questionone of the outstandingmathematicians
in history. He invented analytic geometry independently of
Descartes(and extended it to three dimensions instead of stop-
ping at two), apparently intuited the basic principles of differen-
tial calculus before Newton, was at least a co-founder (with
Pascal and Huygens) of probability theory, and inauguratedthe
modern “theory of numbers.”
It took many years for Fermat to receive credit for much of
his amazing body of work because he was an amateur mathe-
matician - most of his formal education and all of his career
were in law - who devoted only his spare time to the subject.
He did not publish in normal mathematical channels of com-
munication; many of his ideas were circulated only in letters to
friends and these were not published until 1679, 14 years after
his death.
F ermat also contributed to physical theory, most importantly
with what is now called Fermat’s Principle or the principle of
least time This concept has had tremendous influence on the
development of physical thought in and beyond the study of
classicaloptics (to which Fermat first applied it). As previously
34 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION FEBRUARY
stated, it is absolutely indispensable to our current concept of
seismic wave propagation.
A variety of natural phenomena exhibit what might be called
the minimum principle, or its twin the maximum principle. These
principles find expressionin certain geometric statements. (For
example, a straight line is the shortest distance between two
Points on a plane or a circle enclosesthe largestarea of all closed
curvesof equal length on a plane). Many of theseexampleswere
known to the ancients. One story says that the Phoenician
princessDido obtained a grant from a North African chief, the
grant being for as much land as she could enclosein an ox hide.
Dido cut the hide into long, thin strips; tied the ends together;
and staked out the area upon which Carthage was built.
Hero of Alexandria is believed to be the first to apply the
minimum principle to light. (Almost nothing is known of’his
life, not even the century in which he lived. The accepted time
frame is sometime between 150 BC and 250 AD. Isaac Asimov
saysthat a lunar eclipsereferred to in Hero’s writings was visible
in Alexandria in 62 AD and thus guessesthat he was born about
20 AD.) In his study of the law of reflection, Hero explained
the reflection of light in terms of an analogy with balls bounc-
ing off a plane surface. For both a reflecting beam of light and
a bouncing ball, the angle of incidence is always equal to the
angle of reflection. Hero thus concluded that the path between
two fixed points that included a reflection from a mirror was
such that the path length was a minimum. Hero’s reason for
postulating this minimum-path principle was that Nature did
nothing in vain.
The related law of refraction, which is basic to geophysical
exploration, was experimentally studied by the Greek-Egyptian
astronomer Ptolemy in the first century AD. However, it was
not formulated for another 1,500 years. Johann Kepler, in his
study of optics in the early seventeenth century, made many
hypotheses- some shrewdand closeto the mark - concerning
the refraction of light. But nothing of the first magnitude came
out of this research.
T he law of refraction was discoveredsome years later by Snell
and, independently, by Descartes.Snell formulated it in or after
1621;it was the result of many years of experimentation as Well
as the study of Kepler’s book Ad VitellionemParalipomena
(1604) and Risner’s Optica (1606), both of which quote lbn al-
Haytham and Witelo.
Snell’s manuscript has disappeared but it had been studied
by Huygens who commented on it in his Diopticu (1703).
1987

L.et the eye 0 (in the all) recei\cs‘1 liuhr 13~ ,~rlginarlnp
at point R in a medium (perhaps water). The ran is
refracted at S on the surface of the medium. Then 0
observesthe point R as if it were point 1. on the line RM
drawn perpendicular to the surface, and the ratio I S ‘RS
is constant for all rayp.
Snell’s wording agrees with the present tbrmulation of the law.
As we know, the angle of incidence i is defined as the angle the
incident ray RS makeswith the normal to the surface at S. Thus
angle i is equal to the angle MRS in Figure I. .41so,the angle
of refraction r is defined as the angle that the refracted ray SO
makes with the normal to the surface at S. We see that angle
r is equal to angle MLS in Figure 1. We also see from Figure
1 that
Figure 1. bell’s construction (1621).
MS MS
sin i = - and sin r = -,
RS LS
so the ratio sin i/sin r is the same as the ratio LS/RS which,
according to Snell, is a constant. Snell’s law thus saysthat the
ratio of the sine of the refraction angle to the sine of the inci-
dent angle is a constant for all rays and this is the modern state-
ment of the law.
T he publication priority of the law of refraction remains with
Descartesin his Optics (1637) in which he stated his conclusion
but presentedno experimental verification. Descartes’discourse
on refraction is interesting in itself - even if it is wrong -
becauseit opens the door for the concept of the particle-wave
duality which is basic to modern physics.
Descarteswas led to his erroneousconclusion,later discovered
by Fermat and Treitel, by first considering the law of reflection
and harking back to Hero’s analogy between the motion of a
ball (which travels as a particle) and light (which, according to
modern science,travelsas a wave). He then analyzesthe motion
of the ball upon bouncing from a hard surface (Figure 2). The
ball, being impelled from A toward B, meets the surface of the
ground at point B and bounces.
In order to predict what direction the ball must now go,
Descartes describesa circle centered at B and passing through
A. Becausethe ball movesat a constant speed, the time it takes
to travel from B to another point on the circumference of the
circle must be the same time it took the ball to travel from A
GEOPHYSICS:
10 B. Descartes&a\\5 three straight lines AC, HB, and FE (each
perpendicular to CE and such that CB equals BE). Then he says
that the time it takes the ball to advance to the right from A
to B must be equal to the time it takes the ball to advance to
the right from B to some point on the line FE. The ball cannot
be 5imultaneously at a point on the line FE and at a point on
the circumference of the circle unless it is either at point D or
point F becausethese are the only points where the circle and
line intersect.Sincethe ground preventsthe ball from continuing
toward D, it must go to F. Descartesthen saysthat you can easily
see how reflection occurs, i.e., according to an angle which is
always equal to the one we call the angle of incidence, He con-
cludesthat, similarly, if a ray coming from point A falls to point
B on a flat mirror, it is reflected toward Fin such a manner that
the angle of reflection FBH is equal to the angle of incidence
.4BH.
Figure 2. Descartes’law of reflection.
Descartesnext treatsrefraction and again usesthe ball analogy.
He supposesthat a ball impelled from A to B does not meet
an unyielding surface at the CBE plane but rather a cloth (see
Figure 3). Becausethe cloth is weak and loosely woven, the ball
ruptures it . . but the ball consequently loses half its speed.
Again, Descartes separatesmotion into two components. The
vertical component can be changed in any manner through the
encounter with the cloth, but the horizontal component must
remain the same (because the cloth does not oppose the ball
going in this direction). A circle is again drawn along with three
vertical lines, but this time there is twice as much distance
between FE and HB as between HB and AC. (Or, the distance
BE is twice that of BC.) Becausethe ball loses half its speed
by going through the cloth, it takes twice as much time to reach
the circumference from the center B as it did to reach the center
from initial point A. And, since the ball loses none of its
rightward speed, it goes twice as far (the distance BE) to the
right as it did previously (the distance CB). Thus the ball must
go to point I, the point where the circumference and the verti-
cal line FE intersect.
Descartes now comes to his fatal misstep. He states:
Finally,inasmuchaslight follows in this respectthe same
lawsas the movementof the ball, when rays passobliquely
fromone transparent body to another, they are deflected
in such a manner that they make a smallerangleto the
normal for the body with the greater speed.
ln Figure 3, the speed I/, in the lower body is half the speed
V, in the upper body, so the distance BE is equal to twice the
distance CB. Angle CAB is the angle of incidence i, and angle
THE LEADING EDGE OF EXPL_QRATION FEBRUARY 1987
 EIB is the angle of refraction r. We have sin i = E,
BA
sin r = E ; and, since BA = BI, we have= = E = 2=
BI sin i BC
!?,but this is not correct!
VT
High-speedmedium
with velocityV,
cloth
CB\ E Descartes’ partisans,however,refusedto acceptFermat’s proof
\ I
Figure 3. Descartes’law of refraction.
A sgeophysicists,we know that the truth is just the opposite;
namely,that when rayspassobliquely from one transparentbody
to another, they are deflected so they make a larger angle to the
normal for the body with the greater speed. Thus the construc-
tion of Figure 2 is incorrect in the case of light rays (or seismic
rays or any wave motion), and the correct result would be
- sin r = -1 = - V,
sin i 2 V,
for transmission from the high-speed incident medium with
velocity Vi to the low-speed refracted medium with velocity V,
= r/i/2.
HOWcould Descartesmake sucha blunder in the face of obser-
rational evidence?His answer is:
But perhapsyou will be astonishedto find that the rays
of light havea largeranglefrom the normal in the air
than in the water, quite the opposite of a ball which has
a smaller angle from the normal in the air than in the
water.However,you will ceaseto find thisstrangeif you
recall the naturethat I attributed to light. The harder
and firmer are the small particles of a transparent body,
the more easily do they allow the light to pass.
Thus to get out of mistake 1 (that light moves like a particle),
he makesmistake2 (that light travelsfaster in water than in air).
The two mistakescancel and Descartes’statement of Sneil’s law
is all right, as such.
Descartes’major error was in acceptingthe analogy that light
trav$ as a particle but he camouflaged this mistake by making
anbther, that light travels faster in water than in air. As far as
history goes,he is batting SOO:he getssome credit for thederiva-
tion of the law of refraction but he is faulted for not knowing
that the velocity of light is smaller in material substancesthan
in a vacuum.
F ermat was very critical of Descartes’ work. He wasconvinced
that there was no merit at all in Descartes’ proof of the law of
refraction and, in fact, he came to regardDescartesas something
of a fraud. As a result, Fermat undertook his own derivation
of the law of refraction as a point of honor. He embarked on
a couFgeof researchwhich eliminated the bouncing ball analogy.
Like Descartes, he reached back to Hero of Alexandria . .
not for bouncing balls, but for the idea of minimum path.
I$owever, he altered it by postulating that the path of a light
ray connecting two fixed points is the one for which the time
of tmcht, not the length, is a minimum. His own words are:
“Je reconnoispremi&remenf_ la vPritPde ceprincipe, que la
36 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION
nature agit toujours par les voies les plus courtes.”
After severalyears of hard work, Fermat arrived at the law
of refraction and his resultsagreedwith Descartesup to the con-
stant of proportionality . . . but then theydiffered dramatically.
Descartes found that this constant was the ratio of the velocity
of the incident light to the velocityof the refractedlight. Fermat’s
finding was exactly the opposite, namely
sin r = V,
sin i vi
and therefore light must travelmore slowly in the densermedium.
Fermat’s result is correct.
and argued that his physical assumption regarding least time
wasfaulty. Fermat had little chance against what was then the
scientific establishment and he made a lessthan valiant defense
of his discovery,one that would in time be accepted as one of
the mostbasicphysical laws. Fermat’s last word on this matter
was expressedin a letter to one of Descartes’principal followers:
1donot pretendnor haveI everpretendedto be Nature’s
private confidant. She has obscureand secretways that I
have never pretended to penetrate. I have only offered
her a littlegeometrical assistanceon the subjectof refrac-
tion if it should be needed. But since you assure me,
Monsieur,that her affairs are in order without it and
thatsheis content to follow the course that M. Descartes
has prescribed, I surrender to you my supposedphysical
conquest and content myself with the purely geometrical
problem.
One should not judge Fermat too harshly by this “surrender.”
The theories of Descarteswere so entrenched that it took nearly
a hundred years before European sciencedropped them for the
obviously superior model of Isaac Newton. time though, was
on Fermat’s side and finally vindicated him, at least scientifi-
cally, long after his death.
(Vindication for Fermat for his practice of law is another
matter. Ultimate judgment there is likely to be harsh. It is known
that he once condemned a church official to be burned at the
stake. What could possibly have been the horrendous crime?
Fermat’s action in this case is stunning evidence that a mind
which can easily graspscientific conceptsfar in advanceof con-
temporary thinking can, in other matters,be a willing accomplice
to the most tinscientific prejudices of his time.)
F: lgure 4 summarizes the differences between Descartes and
Fermat. The velocity in the upper medium is twice the velocity
in the lower medium. In Descartes’ model, the upper distance
traveled (AB) is equal to the lower distance traveled (BI) and
both are equal to the radius of the constructed circle. It
immediately follows that the traveltime in the upper medium
is half the traveltime in the lower medium. DescartespostuIate$
that the horizontal component of velocity is the same in both
media and therefore, because the upper traveltime is half th&
lower traveltime,it follows that horizontal distanceh in theupper
medium is one-half the horizontal distance of the travel oath
in the lower medium.
Fermat argued that the distance traveled in the high-speed
(upper) medium should be increased,but only up to a certain
point. What is that point? It is the one wherethe total travel-
time is a minimum. This is representedby path API in Figure
4. As the point of intersectionon the interfaoemoves from pbint
B (the Descartessolution) to point P @tieFormat solution), the
traveltime from A to I would decreaseIt wouldbe at a minimum
at P and then start to increaseagain as the point of intersection
moved farther to the right.
But how did Fermat find this optimum point of leasttime
The answeris not knownbut it is possiblethat he did it by inWnt-
FEBRUARY 1987
 velocity V,
Upper (high-speed)
medium

Lower(low-speed)
mediumwith hall
the velocityof velocity V,
the uppermedium.

I
Figure4. Descartes’path ABI (incorrect)and Fermat’s least-time Figure 5. Derivation of Snell’s law.
path API (correct).

ing the basic underpinnings of differential calculusmany years by taking the derivative of this equation and setting it equal to
before Newton. zero. The derivative of i in respect to x is
Consider Figure 5. The traveltime between A and I, by way
x (c-x)
of point P, is
v, (a2 + ?)I’* v2 [fJ2 t (c-x)2]“2’
[ _ (a2 + 2)“2 + [LIZ+ (c-x)y* and if this is made equal to zero the result is Sell’s law (!)
VI V2 because
and according to the fundamentaltheory of maxima and minima .Y (C--X)
in differential calculus,the minimum value of I can be obtained (a2 t 2)“’ = sin01, and ,b2 + (c_x,2,1,2 = sin0;

GEOPHYSICS: THE LEADING EDGE OF exploration FEBRUARY 1987 37