A STUDENT’S GUIDE TO DIMENSIONAL

A N A LY S I S

This introduction to dimensional analysis covers the methods, history, and

formalization of the field, and provides physics and engineering
applications. Covering topics from mechanics, hydro- and electrodynamics
to thermal and quantum physics, it illustrates the possibilities and
limitations of dimensional analysis. Introducing basic physics and fluid
engineering topics through the mathematical methods of dimensional
analysis, this book is perfect for students in physics, engineering, and
mathematics. Explaining potentially unfamiliar concepts such as viscosity
and diffusivity, this text includes worked examples and end-of-chapter
problems with answers provided in an accompanying appendix, which help
make this text ideal for self-study. Long-standing methodological problems
arising in popular presentations of dimensional analysis are also identified
and solved, making it a useful text for advanced students and professionals.

DON S. LEMONS is Professor Emeritus of Physics at Bethel College in

North Newton, Kansas, and has served as an assistant editor of the
American Journal of Physics. He is a member of the American Physical
Society and his research focuses primarily on plasma physics.
A STUDENT’S GUIDE TO
DIMENSIONAL ANALYSI S
Don S. Lemons
Bethel College, Kansas
 University Printing House, Cambridge CB2 8BS, United Kingdom

One Liberty Plaza, 20th Floor, New York, NY 10006, USA

477 Williamstown Road, Port Melbourne, VIC 3207, Australia

4843/24, 2nd Floor, Ansari Road, Daryaganj, Delhi – 110002, India

79 Anson Road, #06–04/06, Singapore 079906

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of education,

learning, and research at the highest international levels of excellence.

www.cambridge.org

Information on this title: www.cambridge.org/9781107161153

10.1017/9781316676165

© Don S.Lemons 2017

This publication is in copyright. Subject to statutory exception and to the provisions of relevant
collective licensing agreements, no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 2017

Printed in the United Kingdom by TJ International Ltd. Padstow Cornwall

A catalogue record for this publication is available from the British Library.

Library of Congress Cataloging-in-Publication Data

Names: Lemons, Don S. (Don Stephen), 1949– author.

Title: A student’s guide to dimensional analysis / Don Lemons.

 Description: New York : Cambridge University Press, 2017. | Includes bibliographical
references and index.

Identifiers: LCCN 2016040385 | ISBN 9781107161153 (hardback)

Subjects: LCSH: Dimensional analysis. | BISAC: SCIENCE / Mathematical Physics.

Classification: LCC TA347.D5 L46 2017 | DDC 530.8–dc23 LC record available at

https://lccn.loc.gov/2016040385

ISBN 978-1-107-16115-3 Hardback

ISBN 978-1-316-61381-8 Paperback

Cambridge University Press has no responsibility for the persistence or accuracy of URLs for
external or third-party Internet Web sites referred to in this publication and does not guarantee
that any content on such Web sites is, or will remain, accurate or appropriate.
 Contents
Preface
Acknowledgments

1 Introduction
1.1 Dimensional Homogeneity
1.2 Dimensionless Products
1.3 Dimensional Formulae
1.4 The Rayleigh Algorithm
1.5 The Buckingham π Theorem
1.6 The Number of Dimensions
1.7 The Number of Dimensionless Products
1.8 Example: Pressure of an Ideal Gas
1.9 A Mistake to Avoid
Essential Ideas
Problems

2 Mechanics
2.1 Kinematics and Dynamics
2.2 Effective Dimensions
2.3 Imposed Dimensions
2.4 Example: Hanging Spring-Mass System
2.5 Example: Hanging, Stretched Cable
2.6 Asymptotic Behavior
2.7 Example: Speed of Sound
 2.8 Example: Side Window Buffeting
2.9 Example: Two-Body Orbits
Essential Ideas
Problems

3 Hydrodynamics
3.1 Fluid Variables
3.2 Example: Water Waves
3.3 Surface Tension
3.4 Example: Largest Water Drop
3.5 Viscosity
3.6 Example: Hydraulic Jump
3.7 Example: Equilibrium Pipe Flow
3.8 Example: Non-Equilibrium Pipe Flow
3.9 Scale Models
Essential Ideas
Problems

4 Temperature and Heat

4.1 Heat Transfer
4.2 The Dimensions of Temperature and Heat
4.3 Conduction and Convection
4.4 Example: Cooking a Turkey
4.5 Diffusion
4.6 Making the Equations Dimensionless
4.7 Example: Growth of Arctic Ice
4.8 Example: Stack Effect
Essential Ideas
 Problems

5 Electrodynamics and Plasma Physics

5.1 Maxwell’s Equations
5.2 Example: Oscillations of a Compass Needle
5.3 Example: Radiation from an Accelerating Charge
5.4 Example: Child’s Law
5.5 Plasma Physics
5.6 Example: Plasma Oscillations
5.7 Example: Pinch Effect
Essential Ideas
Problems

6 Quantum Physics
6.1 Planck’s Constant
6.2 Example: Blackbody Radiation
6.3 Example: Spectral Energy Density of Blackbody Radiation
6.4 Example: Bohr Model
6.5 Atomic Units
6.6 Example: Quantum Ideal Gas
6.7 Example: Quantized Radiation from an Accelerating Charge
Essential Ideas
Problems

7 Dimensional Cosmology
7.1 The Fundamental Dimensional Constants
7.2 Eddington-Dirac Number
7.3 Example: Gravitational-Electrostatic Mass
7.4 Example: Classical Electron Radius
 7.5 Classical Scales
7.6 Fine Structure Constant
7.7 The 15 Scales
7.8 Planck Scale
7.9 Example: Planck Mass
7.10 The 22 Quantities
Essential Ideas
Problems

Appendix: Answers to Problems

References
Index
 Preface
I was charmed when as a young student I watched one of my physics
professors, the late Harold Daw, work a problem with dimensional analysis.
The result appeared as if by magic without the effort of constructing a
model, solving a differential equation, or applying boundary conditions. But
the inspiration of the moment did not, until many years later, bear fruit. In
the meantime my acquaintance with this important tool remained partial
and superficial. Dimensional analysis seemed to promise more than it could
deliver.
Dimensional analysis has charmed and disappointed others as well. Yet
there is no doubt that a deep understanding of its methods is useful to
researchers in a number of fields. Attesting to this fact is that dimensional
analysis is the subject of several good, graduate-level monographs. Even so,
dimensional analysis is often ignored at the introductory level except when
teachers admonish their students to “check the units of their result” and
warn them against “adding apples and oranges.”
The problem for teachers and students is that dimensional analysis has
no settled place in the physics curriculum. The mathematics required for its
application is quite elementary – of the kind one learns in a good high
school course – and its foundational principle is essentially a more precise
version of the rule against “adding apples and oranges.” Yet the successful
application of dimensional analysis requires physical intuition – an intuition
that develops only slowly with the experience of modeling and
manipulating physical variables. But how much intuition is required?
I have written A Student’s Guide to Dimensional Analysis in the belief
that the simple techniques of dimensional analysis can deepen our
understanding and enhance our exploration of physical situations and
processes at the introductory level. Thus, this text is designed for students
who are taking or have taken an entry-level, mathematically oriented,
university physics course. More experienced students and professionals
may also find this text useful.
One elementary application of dimensional analysis is to a simple
pendulum, that is, to a compact weight swinging from the end of a string.
How long does the pendulum take to complete one cycle of its motion? It
seems that this interval Δt might depend on the mass of the weight m, the
acceleration of gravity g, the length of the string l, and the maximum angle
of its swing θ. The adjective “simple” in “simple pendulum” means that the
size of the weight is negligibly small compared to the length of the
pendulum l. Therefore, we search for a function of m, l, g, and θ that
produces Δt. We note that the International System or SI unit of the period
Δt is the second, the SI unit of mass m is the kilogram, of length l is the
meter, and of acceleration g is a meter divided by a second squared. The
angle θ is dimensionless. We quickly discover that is the only
combination of m, l, and g that produces a quantity whose SI unit is a
second. Therefore, it must be that
where f(θ) is an as yet undetermined, dimensionless function of θ.
This result is typical of dimensional analysis. Much is learned but
something is left unlearned. The analysis teaches us that the period Δt of a
simple pendulum is directly proportional to the square root of its length
and does not at all depend on its mass m. We could experimentally verify
these findings, but dimensional analysis suggests that limited time might be
better spent exploring the undetermined function f(θ). However, this
example misleads as well as informs. Not many applications of dimensional
analysis are this simple, nor are we usually aware, as we probably are in
this case, of the result beforehand.
Dimensional analysis is most revealing when we know something but
not everything about a situation or process. Then dimensional analysis
builds on that something we already know. We might, for instance, be
familiar with the equations that describe a certain process but not have the
skill or the time to solve them in the usual way. Or we might wish to expand
our preliminary knowledge of a solution before completely determining its
form. Then again, we might know only the category of the problem (that is,
mechanics, thermodynamics, or electrodynamics) and the variable we want
to determine (that is, an oscillation period, a pressure drop, or an energy
loss rate). In all these cases, dimensional analysis constrains how the
relevant physical variables and constants work together to produce the
result sought.
The techniques of dimensional analysis could be presented in one
well-chosen example, occupying one or two pages of text. Then one might,
with some reason, expect to apply these techniques to new problems as they
arise. But this expectation is bound to remain unrealized. In order for
dimensional analysis to be fruitful, it must first be cultivated. Exploring the
method’s motivation, its history, and its formalization are steps in this
cultivation. Examining simple applications is another. Such are the aims of
Chapter 1 “Introduction.”
Chapters 2 through 6 take up more complex but still introductory
examples of dimensional analysis grouped into several subject areas:
Mechanics, Hydrodynamics, Thermal Physics, Electrodynamics, and
Quantum Physics. These examples illustrate the possibilities and limitations
of dimensional analysis. As appropriate, I will explain concepts (such as
surface tension, viscosity, and diffusivity) that may not be familiar to all
readers.
In the final chapter, Chapter 7 “Dimensional Cosmology,” I use
dimensional analysis to take a few steps in the direction of uncovering the
dimensional structure of our world. The result, preliminary and partial as it
is, brings us close to the boundary separating the known from the unknown.
Dimensional analysis has no settled place in the physics curriculum
because it fits easily in any number of places. We can use it in elementary
fashion to recall the shape of a formula or to reaffirm our understanding of
an already solved problem. Or we can use it to push forward into territory
unknown to us or uncharted by anyone.
Sometimes dimensional analysis fails us, but it is unlikely to do so
without announcing its failure and suggesting a better way to proceed. Does
the method produce no result? Then we have left out a crucial variable or
constant. Is the result uninformative? Then we have included too many
variables and constants.
A Student’s Guide to Dimensional Analysis is designed to guide
readers to an understanding of the motivation, methods, and exemplary
applications of dimensional analysis, its scope and powers as well as its
limitations. But I also hope the text will recreate for you the charm and
magic that first attracted me to the subject years ago.
 Acknowledgments
One of the rewards of writing a book is receiving responses to its initial
drafts from an obliging set of friends. I have been very fortunate in my
“first responders.” Ralph Baerlein and Galen Gisler read the whole text,
suggested many improvements, and kept me from making a number of
mistakes. Ikram Ahmed, Christopher Earles, Rick Shanahan, and David
Watkins all reviewed parts of the text and gave me good advice. Jeff Martin
suggested the problem “cooking a turkey.” I offer a heartfelt “thanks” to
them all.
My participation in a series of seminars on similarity methods,
organized by Susan G. Sterrett of Wichita State University in the academic
year 2014–15, renewed a youthful interest in dimensional analysis. As
usual, Cambridge University Press has provided me with excellent editorial
services.
 1
Introduction
◈
 1.1 Dimensional Homogeneity
Most physical variables and constants have dimensions. A mass, a distance,
and a time have, respectively, the dimensions mass, length, and time.
Dimensions are often taken for granted. They slip beneath our notice. Yet
dimensions are important elements of the way we think about the physical
world.
Imagine an animal that grows in size while keeping roughly the same
shape. Galileo (1584–1642) reasoned that the weight of the animal
increases with its volume in direct proportion to the third power l3 of a
characteristic length l, say, for an elephant, the length of its foreleg. Because
animal limbs push and pull across a cross-section, the strength of an animal
increases with the cross-sectional area of its limbs, that is, in direct
proportion to l2. Thus the animal’s strength-to-weight ratio changes in
proportion as l2/l3, that is, as 1/l or l−1. Therefore, larger animals are less
able to support their weight than smaller ones are. Galileo illustrated this
conclusion by comparing the relative strength of dogs and horses –
creatures with roughly the same shape.

A small dog could probably carry on his back two or three dogs of his
own size; but I believe that a horse could not carry even one of his own
size.

[1]

If Galileo had not thought dimensionally, he could not have made this
interesting argument.
 By the time of Isaac Newton (1643–1727) scientists had begun to think
in terms of combinations of different dimensions. For instance, the
dimension of speed is length divided by time, the dimension of acceleration
is length divided by time squared, and, according to Newton’s second law,
the dimension of force is mass times length divided by time squared.
Newton regarded mass, length, and time as primary, fundamental
dimensions and combinations of these as secondary, derived ones. [2]
One of the first things a physics student learns is that one should not
add, subtract, equate, or compare quantities with different dimensions or
quantities with the same dimension and different units of measure. For
instance, one cannot add a mass to a length or, for that matter, 5 meters to 2
kilometers. This rule against what is sometimes called “adding apples and
oranges” means that every term that is added, subtracted, equated, or
compared in every valid equation or inequality must be of the same
dimension denominated in the same unit of measure. This is the principle of
dimensional homogeneity.
The principle of dimensional homogeneity is nothing new. Scientists
have long assumed that every term in every fully articulated equation that
accurately describes a physical state or process has the same dimension
denominated in the same unit of measure. However, it was not until 1822
that Joseph Fourier (1768–1830) expressed this principle in a way that
allowed important consequences to be derived from it. [3]
 Symmetry under Change of Units
Behind the principle of dimensional homogeneity is a symmetry principle.
Symmetry principles tell us that something remains the same as something
else is changed. Here the something that remains the same is the form of the
equation or inequality and the things that are changed are the units in which
the dimensions of its terms are expressed. Thus, if we change the unit of
length from meters to kilometers and the form of the equation does not
change, that equation observes this particular symmetry and is, at least in
this regard, dimensionally homogeneous. If we change all of its units and
the form of the equation does not change, this equation is fully
dimensionally homogeneous.
An equation can be useful without being dimensionally homogeneous.
For instance,

s = 4.9t2
(1.1)
correctly describes the downward displacement s
denominated in meters of an object falling freely from rest for a period of
time t denominated in seconds. Yet (1.1) is not invariant with respect to
changes in its units of measure. Compare (1.1) with

(1.2)

in which we have parameterized the acceleration of gravity with the symbol

g. This equation is now symmetric with respect to changes in all its units of
measure. It is fully articulated and dimensionally homogeneous.
 We are concerned in this text with relations among dimensional
variables, for instance, s and t, and dimensional constants, such as g, that
are symmetric with respect to changes in units and, therefore, dimensionally
homogeneous. The principle of dimensional homogeneity and its
consequences are foundational to the theory of dimensional analysis.
 1.2 Dimensionless Products
Consider the vertical position y of a freely falling object at time t. We know
that

(1.3)

where y − yo is the object’s displacement from its initial position yo, υyo is
its initial velocity, g is the magnitude of the gravitational acceleration, and
our coordinate system is oriented so that y becomes more negative as the
object falls and time advances. Equation (1.3) observes the principle of
dimensional homogeneity. For the dimension of y − yo is length; the
dimensions of υyo and t are, respectively, length/time and time so that the
dimension of their product υyot is length; and the dimension of gt2 is
length/time2 multiplied by time2 or, again, length. Furthermore, (1.3)
contains no dimensional constants masquerading as dimensionless numbers
– as does (1.1).
One consequence of the dimensional homogeneity of (1.3) is that
dividing each of its terms by gt2 produces an equation,

(1.4)

that relates one dimensionless combination or “product” (y − yo)/gt2 to

another υyo/gt. This transformation of a dimensionally homogeneous
equation, from a relation among dimensional variables and constants to a
relation among dimensionless products, can always be realized.
 Consider, for instance, the Stefan-Boltzmann law, according to which
the density of radiant energy E in a cavity of volume V whose walls are at a
temperature T is described by

(1.5)

where kB is Boltzmann’s constant, c is the speed of light, and h is Planck’s

constant. Each of the variables E, V, and T and each of the constants kB, c,
and h are dimensional quantities. If equation (1.5) is dimensionally
homogeneous (and it is), it may take the form

(1.6)

of a dimensionless product equal to a dimensionless

number C. In this case C = 8π5/15.
 Dimensional Analysis
We have, in these two examples, turned dimensionally homogeneous
relations among dimensional variables and constants, (1.3) and (1.5), into
relations among one or more dimensionless products, (1.4) and (1.6). We
shall soon learn a way to reverse this process. We will first use an
algorithm, the Rayleigh algorithm, to discover the dimensionless products
relevant to a particular state or process. When only one dimensionless
product is found, the only way it can form a dimensionally homogeneous
equation is for this product to equal some dimensionless number as in (1.6).
When two or more dimensionless products are found, as in (1.3), they must
be related to one another by some function, as in (1.4). The Rayleigh
algorithm does not determine these numbers and these functions but merely
finds the dimensionless products.
 1.3 Dimensional Formulae
Every dimensional variable or constant assumes values in the form of a
number times a unit of measure – for instance, 5 kilograms or 16 meters.
Furthermore, every unit of measure makes its dimension known. A meter
per second and a kilometer per hour are both a length/time while a metric
ton and a kilogram are both masses. We need to know the dimension, more
precisely the dimensional formula, of every relevant dimensional variable
and constant in order to dimensionally analyze a state or process. For this
purpose we use the symbol M to stand for the dimension mass, L to stand
for the dimension length, and T to stand for the dimension time. The
notation [x] means “the dimension of x.” Therefore, [m] = M and [g] = LT−2
are dimensional formulae. Not every dimensional formula can be expressed
in terms of only M, L, and T, but many can be.
The dimensional formula of a product of factors is the product of the
dimensional formula of each factor. Thus [ma] = [m][a]. For convenience,
we define the dimensional formula of a dimensionless number to be 1.
Therefore, [π] = 1 and so [9.8 ⋅ m/s2] = [9.8] ⋅ [m/s2] = [m] ⋅ [s−2] = LT−2.
 1.4 The Rayleigh Algorithm
John William Strutt (1842–1919), also known as Lord Rayleigh,
successfully applied dimensional analysis to a number of problems over a
long career. He dimensionally analyzed the strength of bridges, the velocity
of waves on the surface of water, the vibration of tuning forks and drops of
falling water, the color of the sky, the decay of charge on an electrical
circuit, the determinants of viscosity, and the flow of heat from a hot object
immersed in a cool stream of water. Rayleigh prefaced his 1915 summary
of these applications of the principle of dimensional homogeneity (known
to him as “the principle of similitude”) with these words,

I have often been impressed with the scanty attention paid even by
original workers to the great principle of similitude. It happens not
infrequently that results in the form of “laws” are put forth as novelties
on the basis of elaborate experiments, which might have been
predicted a priori after a few minutes’ consideration.

[4]

While our applications of dimensional analysis may require more than “a

few minutes consideration,” Rayleigh’s method of applying “the principle
of similitude” is simple and direct. We adopt it, as have many others, with
only slight modification.
 A Marble on the Interior Surface of a Cone
To illustrate Rayleigh’s method, imagine a small marble of mass m rolling
in a circle of radius R on the interior surface of an inverted cone defined by
an angle θ as illustrated in Figure 1.1. We wish to know how the time Δt
required for the marble to complete one orbit is determined by m, R, and θ.
The acceleration of gravity g may also enter into the relation we seek.
Gravity, after all, is one of the two forces that keep the marble on the cone’s
surface. The intermolecular forces of the material composing the cone and
the marble also determine, in some degree, the period Δt, but we ignore
these forces because we believe their effect is adequately accounted for by
assuming the marble stays on the surface of the cone. By including some
variables and constants in our analysis and excluding others we construct a
model of the marble’s motion.

Figure 1.1. Marble on the interior surface of a cone defined by angle θ.

 Rayleigh’s Algorithm
Rayleigh’s method of dimensional analysis identifies the dimensionless
products one can form out of the model variables and constants, in this case
Δt, m, R, g, and θ. Each dimensionless product takes a form ΔtαmβRγgδθε
determined by the Greek letter exponents α, β, γ, δ, and ε or, somewhat
more simply, by the form ΔtαmβRγgδ and the exponents α, β, γ, and δ. After
all, the angle θ, whether denominated in radians or degrees, is proportional
to a ratio of an arc length to a radius, that is, a ratio of one length to another.
While angles have units (degrees or radians), their units are dimensionless.
The key to Rayleigh’s method of finding dimensionless products is to
require that the product ΔtαmβRγgδ be dimensionless. Since

(1.7)

the product ΔtαmβRγgδ is dimensionless when

T: α − 2δ = 0,
(1.8a)
M: β = 0,
(1.8b)
and

L: γ + δ = 0.
(1.8c)
The three equations (1.8a), (1.8b), and (1.8c) constrain the
four unknowns, α, β, γ, and δ, to a family of solutions β = 0, γ = − α/2, and
δ = α/2 parameterized by α. [The symbols T, M, and L preceding equations
(1.8) identify the source of each constraint.] Therefore, (Δtg1/2/R1/2)α is
dimensionless for any α, which means that Δtg1/2/R1/2, as well as θ, is
dimensionless.
Once we know the dimensionless products that can be formed out of
the model’s dimensional variables and constants, we know they must be
related to one another by an undetermined function, that is, in this case
expressed by

(1.9)

where f(θ) is a dimensionless function of the dimensionless “product” θ.

This is as far as dimensional analysis per se takes us. A more detailed,
dynamical study reveals that .
 The Rayleigh Algorithm Modified
Note that equations (1.8) are solved by a family of solutions parameterized
by a non-vanishing exponent α and also that the identity of the
dimensionless product Δtg1/2/R1/2 is independent of the value of α. The
exponent α, introduced in the term Δtα, seems superfluous, and indeed, it is
– as long as we know that Δt, the variable whose expression we seek, the
variable of interest, remains in the dimensionless product. In this case no
harm is done by freely choosing α. In particular, choosing α = 1 is
equivalent to determining the three remaining exponents β, γ, and δ as those
that make ΔtmβRγgδ dimensionless. Then β = 0, γ = − 1/2, and δ = 1/2. This
solution again produces the dimensionless product Δtg1/2/R1/2. Henceforth
we adopt the practice of including the variable of interest with an exponent
of 1 as the first factor in the dimensionless product.
Observe that this analysis, issuing as it does in Δtg1/2/R1/2 = f(θ), is a
significant advance on knowing only that Δt, m, R, g, and θ are related to
one another by an unknown function, say, by Δt = h(m, R, g, θ). For
suppose that empirically determining the function f(θ) in (1.9) requires 10
pairs of Δtg1/2/R1/2 versus θ data. Since 10 pairs of data determine how one
term depends on one other (the others remaining constant), 104 pairs of data
are required to determine how one variable Δt depends on the four
dimensional variables and constants m, R, g, and θ. Thus, 104 pairs of data
are required to determine the function in Δt = h(m, R, g, θ). The Rayleigh
algorithm reduces the effort required by a factor of 1,000!
 1.5 The Buckingham π Theorem
In 1914, Edgar Buckingham (1867–1940) proved, in formal algebraic
detail, a theorem we have, thus far, merely illustrated – a theorem usually
referred to as the Buckingham π theorem or sometimes, more simply, as the
π theorem. [5] The π theorem may be divided into two conceptually distinct
parts. First,

If an equation is dimensionally homogeneous, it can be reduced to a

relationship among a complete set of independent dimensionless
products.
[6]

A set of dimensionless products is complete if and only if all possible

dimensionless products of the dimensional variables and constants can be
expressed as a product of powers of the members of this set. The members
of this set are independent if and only if none of them can be expressed as a
product of powers of the other members.a The symbol π in the phrase π
theorem refers to members of a complete set of independent dimensionless
products. Buckingham denoted these dimensionless products by π1, π2 …
Thus, for example, in the marble on the interior of a cone problem,
π1 = Δtg1/2/R1/2 and π2 = θ. The second part of the π theorem consists of the
following statement.

The number of complete and independent dimensionless products Np is

equal to the number of dimensional variables and constants NV that
 describe the state or process minus the minimum number of
dimensions ND needed to express their dimensional formulae. Thus,

Np = NV − ND.
(1.10)
Statement (1.10) is the most common expression of the
π theorem.
 1.6 The Number of Dimensions
Most dimensional analysts adopt M, L, and T as dimensions appropriate for
mechanical processes and states. We did so in describing the marble on the
interior surface of a cone. In that case, ND = 3. Furthermore, since Δt, m, R,
g, and θ describe the marble’s motion, NV = 5. Therefore, according to
(1.10) NP = NV − ND, 2(=5 − 3) complete and independent dimensionless
products should be produced. By applying the Rayleigh algorithm we find
these to be Δtg1/2/R1/2 and θ. The set Δtg1/2/R1/2 and θ is complete because
every possible dimensionless product of Δt, m, R, g, and θ can be expressed
as a product of some power of Δt2R/g times some power of θ. And its
members are independent because Δt2R/g and θ are not powers of each
other.
But the minimum number of dimensions needed to express the
dimensional formulae of the NV dimensional variables and constants is not
always 3, as it is in this example. Neither is the identity of the minimum
number of dimensions necessarily M, L, and T – as they often are in
mechanical problems. Rather, the dimensions required are, in
Buckingham’s words, the “arbitrary fundamental units [dimensions] needed
as a basis for the absolute system.” [7] And only when we can ensure that
ND is the minimum number of dimensions needed can we depend on
NP = NV − ND to be observed. Otherwise, NP = NV − ND remains a mere
“rule of thumb” – often observed but sometimes not. We will learn how to
recognize the minimum number of dimensions in Section 2.2.
 1.7 The Number of Dimensionless
Products
Note that the more dimensionless products, π1, π2, … πNP, produced, the
less determined the state or process described by the dimensional model.
After all, if only one product π1 is produced, the result sought assumes a
form f(π1) = 0 whose solution π1 = C is in terms of a single undetermined
dimensionless number C. However, if two dimensionless products, π1 and
π2, are produced, these are related by g(π1, π2) = 0 whose solution
π1 = h(π2) leaves a function h(π2) of a single variable undetermined. And if
three dimensionless products, π1, π2, and π3, are produced, these are related
by a function j(π1, π2, π3) = 0 whose solution π1 = k(π2, π3) leaves a function
k(π1, π2) of two variables undetermined.
It is clear that in order to more completely determine a state or process,
we need to minimize the number NP of complete and independent
dimensionless products. According to the rule of thumb NP = NV − ND, we
do this by minimizing NV (the number of dimensional variables and
constants that describe the model) and, assuming we have such freedom, by
maximizing ND (the minimum number of dimensions in terms of which
these variables and constants can be expressed.) However, minimizing NV
and maximizing ND are not straightforward tasks. Both require skill and
judgment – the same kind of skill and judgment needed to construct a
model of a physical state or process.
 1.8 Example: Pressure of an Ideal Gas
Many of these ideas are illustrated in the dimensional analysis of how the
pressure p of an ideal gas depends on quantities that describe its state. The
pressure a gas exerts on its container walls is the average rate per unit area
with which its molecules collide with and transfer momentum to the wall.
The ideal gas model treats these molecules as randomly and freely moving,
massive, point particles whose instantaneous collisions with other particles
and with the walls conserve their energy.
Therefore, we believe the ideal gas pressure p should depend on the
number density of the gas molecules N/V where N is the number of gas
molecules contained in volume V, the mass of each of the molecules m, and
their average or characteristic speed . These parameters should be
sufficient, since they are the elements out of which the momentum of the
gas particles, the rate at which they collide with the wall, and their energy
are composed. To include other variables or constants such as, for instance,
the acceleration of gravity g would be to introduce extraneous
dimensionless products and make our result not so much inaccurate as
uninformative. For convenient reference, we collect these symbols, their
descriptions, and their dimensional formulae in Table 1.1.

Table 1.1
Symbol Description Dimensional
Formula

p Pressure ML−1T−2
 Symbol Description Dimensional
Formula

N/V Number density L−3

m Molecular mass M

Characteristic speed LT−1

Note that we have included the number of particles N in volume V only

in the combination N/V. For this reason, we have 4(=NV) variables: p, N/V,
m, and . Since they are expressed in terms of 3(=ND) dimensions, M, L,
and T, the rule of thumb NP = NV − Nd predicts 1 (=4 − 3) dimensionless
product.
Recall that in executing the Rayleigh algorithm, here and elsewhere,
we enter the variable of interest, the one whose expression in terms of other
variables we seek, in this case the gas pressure p, with an exponent of 1 in
the first position of the product . Then we find the three
exponents, α, β, and γ, that render this product dimensionless. Thus,

(1.11)

and, therefore, the exponents must be solutions of

M: 1 + β = 0,
(1.12a)
L: −1 − 3α + γ = 0,
(1.12b)
and
 T: −2 − γ = 0.
(1.12c)
From equations (1.12) we find that α = − 1, β = − 1, and

γ = − 2. Therefore, the dimensionless product is , and so

(1.13)

where C is an undetermined dimensionless number.

While interesting, this result is not quite what is known as Boyle’s law.
The latter claims that the pressure p of a gas in thermal equilibrium with a
constant-temperature environment is inversely proportional to its volume V,
that is, p ∝ 1/V. Equation (1.13) is equivalent to Boyle’s law only if is
in one-to-one correspondence with the temperature – which, in the ideal gas
model, it is.
If we had allowed the volume V to enter the calculation independently
of the number density N/V, then the number of variables, p, N, V, m, and ,
would increase to 5 (NV = 5) while the number of dimensions, M, L, and T,
would remain at 3 (ND = 3). According to the rule of thumb NP = NV − ND,
the analysis will produce 2(=NP) independent, dimensionless products.
Since and N are two such independent dimensionless
“products,” we need look no further. The result is where
f(N) is an undetermined function – a much less informative result than is
. In general, if our model assumptions allow us to reduce
the number of dimensional variables and constants, we should do so.
 1.9 A Mistake to Avoid
A model of a state or process incorporates certain idealizations and
simplifications. Skill and judgment are required to decide which quantities
are needed to describe the state or process and what idealizations and
simplifications should be incorporated. Similar skill and judgment are
required in dimensional analysis, for the analysis in dimensional analysis is
the analysis of a model. And the model we adopt in a dimensional analysis
is determined by the dimensional variables and constants we adopt and the
dimensions in terms of which they are expressed.
Does the object change shape or does it remain rigid? Does the fluid
become turbulent or does it remain laminar? Does the hot object radiate or
does it merely conduct? While a certain part of dimensional analysis
reduces to the algorithmic, no algorithm helps us answer these questions.
Rather, our answers define the state or process we describe and the model
we adopt.
We will, on occasion, make mistakes. And the mistakes we make in
adopting and dimensionally analyzing a model are of several kinds. The
best way to identify these mistakes is to identify them in particular
examples.
Suppose, for instance, I am a relatively inexperienced modeler. And I
try to find an expression for the frequency of oscillation ω of a mass m of
an extended body of density ρ and volume V that hangs from the end of a
spring characterized by spring constant k – as illustrated in Figure 1.2.
Since the weight of the mass pulls on the spring, I judge that the
acceleration of gravity g is also relevant. The symbols, descriptions, and
dimensional formulae in Table 1.2 summarize this, rather naïve, model.

Figure 1.2. Spring-mass system.

Table 1.2
ω Frequency T−1

m Mass M

k Spring constant MT−2

ρ Mass density ML−3

V Volume L3

g Acceleration of gravity LT−2

Since there are 6(=NV) dimensional variables and constants and

3(=ND) dimensions, the dimensional analysis should, according to the rule
of thumb NP = NV − ND, produce 3(=NP) dimensionless products.
Accordingly, we seek exponents α, β, γ, δ, and ε that make the product
ωmαkβργVδgε dimensionless. We find that

(1.14)
and, therefore, that the exponents are constrained by

T: −1 − 2β − 2ε = 0,
(1.15a)
M: α + β + γ = 0,
(1.15b)
and

L: −3γ + 3δ + ε = 0.
(1.15c)
Equations (1.15) are solved by α = 1/2 − δ + 4ε/3,
β = − 1/2 − ε, and γ = δ − ε/3. Consequently,

(1.16)

for arbitrary δ and ε. As expected the analysis produces three dimensionless

products: ωm1/2/k1/2, ρV/m, and m4/3g/(kρ1/3). The result

(1.17)

where f(x, y), is an undetermined function of two variables, formally

expresses the frequency of oscillation ω in terms of the other dimensional
variables and constants. Even so, (1.17) by itself tells us little or nothing.
We expected something more informative. What went wrong?
 The Mistake of Not Minimizing the Number of Dimensional
Variables and Constants
One mistake has been to ignore a relation among the variables and
constants implied by the model. Clearly, the mass m of the object on the end
of the spring is related to its (presumed constant) mass density ρ and its
volume V by m = ρV. If so, then (1.17) reduces to

(1.18)

where f(1, x) is now an undetermined function of only one variable.

Equation (1.18) also follows from recognizing that since m = ρV, the
set of model variables and constants need include only two of the following
three variables: m, ρ, and V. This reduces NV from 6 to 5. Since the number
of dimensions ND remains the same at 3, the number of dimensionless
products NP(=NV − ND) shrinks to 2(=5 − 3): ωm1/2/k1/2 and m4/3g/(kρ1/3).
These produce (1.18). Yet equation (1.18), which may seem an
improvement on (1.17), still tells us very little.
Neither (1.17) nor (1.18) are incorrect. Rather, neither tells us what we
want to know. A more experienced modeler and analyst might observe that
we have not yet sufficiently minimized the number of dimensional variables
and constants. Should we not know, for instance, that the acceleration of
gravity g merely shifts the position of equilibrium without changing the
frequency of oscillation ω around that equilibrium? With such knowledge
we would leave g out of the list of relevant dimensional variables and
constants. Then we would have 4 dimensional variables and constants and 3
dimensions, and the analysis would produce 1 dimensionless product
ωm1/2k−1/2. Consequently,

(1.19)

where C is an undetermined number.

At this point one might worry that dimensional analysis works well
only when the analyst already knows the solution. Not true! Of course,
more knowledge always helps. But we need not know everything in order to
learn something from dimensional analysis. Dimensional analysis has other
ways to protect us from monstrosities such as m4/3g/(kρ1/3) and, thus, other
ways to move us beyond the unformed and uninformative (1.18) to the
optimally informed and informative (1.19). We will learn more of these
“other ways” in the chapters that follow.
 Essential Ideas
According to the π theorem, every dimensionally homogeneous equation
can be cast as a relation among dimensionless products. Furthermore, the
number of these dimensionless products NP is equal to the number of
dimensional variables and constants NV that describe the state or process
minus the minimum number of dimensions ND in terms of which these
dimensional variables and constants can be expressed, that is,
NP = NV − ND. The Rayleigh algorithm is an analytical machine for
determining the dimensionless products. A dimensional model consists of
the dimensional variables and constants and their dimensional formulae.
 Problems
The problems illustrate developments in the text. Answers, except those
given in the problem statement, are listed in the Appendix.

1.1 Dimensionless products. The radiant energy per volume per

differential frequency interval in a cavity surrounded by walls at
temperature T is ρ where

This spectral energy density ρ has the dimension of an energy divided by a

volume and a frequency. The symbols c, h, and kB are, respectively, the
speed of light, Plank’s constant, and Boltzmann’s constant. Assume this
relation is dimensionally homogeneous and that [h] = ML2T−1,
[kB] = ML2T−2Θ−1, and [T] = Θ. What are the two dimensionless products
π1 and π2 in terms of which the spectral energy density may be expressed?
answer
1.2 Ice skaters. Two skaters each of mass m and speed υ approach each
other along straight, parallel lines. A distance r separates their lines of
approach. At closest approach they catch each other’s hands and begin
rotating around their common center of mass. Use the Rayleigh algorithm
to determine how their frequency f of rotation depends upon these variables.
answer
1.3 Centripetal acceleration. An object of mass m moves in a circle of
radius r with speed υ. Use the Rayleigh algorithm to determine an
expression for its acceleration a.
answer
1.4 Walking Froude number. Consider the assumption that the normal
walking speed υ of a biped or quadruped is a function of the length l of its
leg (considered as a physical pendulum) and the acceleration of gravity g.

(a) Show that the mass of the animal’s leg cannot enter into a dimensionless
product with υ, l, and g.

(b) Show that the one dimensionless product composed of these variables is
the walking Froude number υ2/gl.

(c) How does the walking speed υ depend on the leg length l?
answer
(d) Given that many humans prefer to walk at a speed of approximately
1.4 ⋅ m/s, estimate the walking Froude number of a typical human being.
(You will have to estimate the human leg length l.)
answer
(e) Estimate your own walking Froude number by measuring your normal
walking speed and your leg length.

1.5 Vibrating wire. A wire of length l and mass per unit length λ pulls with
a force (or tension) τ on two posts between which the wire is stretched tight
and to which each end is attached. Suppose the middle of the wire is
plucked. Use the Rayleigh algorithm to determine how the period Δt of the
wire’s oscillation depends upon l, λ, and τ.
answer
1.6 Falling through the center of the Earth. Suppose you drill a tunnel
through the Earth’s center as illustrated in Figure 1.3. Then you drop an
object of mass m from rest into the tunnel. What is the time required Δt for
the object’s passage from one side of the Earth to the other? This duration
should depend on the gravitational constant G, the density of the Earth ρ,
and the Earth’s radius R. It might also depend on the mass m of the object
dropped. These dimensional variables and constants, their descriptions, and
their dimensional formulae are collected in Table 1.3.

(a) Since there are 5 dimensional variables and constants expressed in terms
of 3 dimensions, there should be, according to the rule of thumb
NP = NV − ND, 2 dimensionless products. Use the Rayleigh algorithm to
find the two dimensionless products π1 and π2.
answer
(b) There is a good reason why the mass of the object m should be
eliminated from the list of variables and constants on which the duration Δt
depends. Again use the Rayleigh algorithm to find the single dimensionless
product implied by the variables and constants neglecting m. Derive an
expression for the duration Δt.
answer

Figure 1.3. Object falling through center of Earth.

Table 1.3
Δt Duration of passage T

G Gravitational constant M−1L3T−2

ρ Density of Earth ML−3

R Radius of Earth L

m Mass of object M

a A complete set of independent products plus all dimensionless products

that can be formed from them is itself a group because these products: (a)
are closed under multiplication, (b) contain an identity element 1, and (c)
each product πi has an inverse .
 2
Mechanics
◈
 2.1 Kinematics and Dynamics
The science of mechanics is divided into kinematics, the description of
motion, and dynamics, the study of forces and their consequences.
Dynamics, in turn, is divided into equilibrium dynamics (also known as
statics), concerned with forces that balance each other, and non-equilibrium
dynamics, in which forces cause bodies to accelerate. If we wish merely to
describe an object in free fall, our task is kinematical. If we wish to know
what determines the drag force that balances the object’s weight when
falling at terminal speed, our task is one of equilibrium dynamics. And
finally, if we wish to know how forces cause an object to accelerate to
terminal speed, our task is one of non-equilibrium dynamics. Knowing what
kind of problem we face is an important part of discerning which
dimensional variables and constants enter into its solution. The dimensional
formulae of mechanical variables are usually expressed in terms of M, L,
and T. Thus, the dimensional formula of speed is LT−1 and of acceleration is
LT−2.
Using “the dimension of” operator […] on both sides of an equation is,
in general, a good way to discover a dimensional formula. For instance, the
dimensional formula of the dynamical quantity force follows, in non-
equilibrium processes, from Newton’s second law of motion F = ma. Thus,
[F] = [ma], [ma] = [m][a], and [m][a] = MLT−2, and, therefore,
[F] = MLT−2. Likewise, the dimensional formula of a spring constant k
follows from the linear restoring force F = − kx applied by a spring that is
extended or compressed a distance x. Therefore, [k] = [F]/[x] and so
[k] = MT−2.
 2.2 Effective Dimensions
Consider a gas enclosed within a container fitted with a frictionless,
massless piston of area A on which a weight of mass m rests – as shown in
Figure 2.1. Given that we ignore atmospheric pressure, how does the gas
pressure p depend on the mass m? The answer, p = mg/A, is not in question.
Rather, we are here concerned with the peculiar way dimensional analysis
produces this result.

Figure 2.1. Gas in container.

We could use 4 variables and constants, p, m, g, and A, in our

dimensional analysis. But suppose we are aware that the mass m and the
acceleration of gravity g enter into the result only as a single product we
call the weight w(=mg). Then we have 3 variables: p, w, and A. These
symbols, their descriptions, and their dimensional formulae are collected in
Table 2.1.

Table 2.1
p Pressure ML−1T−2

w Weight MLT−2
 A Area L2

Since the number of dimensional variables NV = 3 is the same as the

number of dimensions ND = 3, the Rayleigh algorithm should produce
0(=3 − 3), that is, no dimensionless products – at least according to the rule
of thumb NP = NV − ND. Yet, because we know there is, indeed, one
dimensionless product pA/w, we continue with the analysis.
According to the Rayleigh algorithm, the dimensionless product is
found by choosing exponents α and β so that the product pwαAβ is
dimensionless, that is, so that

(2.1)

Therefore, we require that

M: 1 + α = 0,
(2.2a)
L: −1 + α + 2β = 0,
(2.2b)
and

T: −2 − 2α = 0.
(2.2c)
Since equations (2.2a) and (2.2c) are algebraically
equivalent, only one of them is needed. Thus, we find that α = − 1 and
β = 1. As expected, this solution produces the dimensionless product pA/w
or, equivalently, pA/mg and the result
 (2.3)

where the dimensionless number C is undetermined by this analysis.

Apparently, the Rayleigh algorithm works even when the rule of thumb
NP = NV − ND does not.
The dimensional formulae ML−1T−2, MLT−2, and L2 listed in the
rightmost column of the table reveal an interesting structure. The
dimensions M and T always appear together in the combination MT−2.
Operationally, it is as if this coupling of M and T reduces the number of
dimensions from 3 (M, L, and T) to 2 (MT−2 and L). Then NV = 3 and
ND = 2, and so NP = NV − ND yields NP = 1. This tactic saves the rule of
thumb NP = NV − ND. But why? And, does it always work?
 The Number of Effective Dimensions
The rule NP = NV − ND is always observed if ND is defined as the minimum
number of dimensions or combinations of dimensions needed to express the
dimensional formulae of all the dimensional variables and constants. [8]
We see, for instance, that we need only 2 dimensions or combinations of
dimensions, MT−2 and L, to express the dimensional formulae of p, w, and
A. Alternatively, we need 3 dimensions or combinations of dimensions, M,
L, and T, to express the dimensional formulae of the 4 dimensional
variables and constants p, m, g, and A. In each case, the rule of thumb
NP = NV − ND is observed.
For convenient reference I call “the minimum number of dimensions
or combinations of dimensions needed to express the dimensional formulae
of all the dimensional variables and constants” the number of effective
dimensions. Observe that the form of the first of the two effective
dimensions in this example, MT−2 and L, encodes some physics – and the
physics encoded is not hard to identify. The mass m resting upon the piston
that encloses the ideal gas is in equilibrium because the forces on it balance
one another. Neither this mass nor any part of the gas is accelerating.
Nowhere is Newton’s second law F = ma needed or invoked. Therefore, the
dimensional formula of F is, in this case, disconnected from the
dimensional formula of ma. What, in a more general context, has a
dimensional formula of MLT−2 is here itself an effective dimension that we
denote as the dimensional combination MLT−2 or more concisely as F. We
depend on context to distinguish between the variable F, denoting force,
and its dimension F.
 Clearly only 2 dimensions, L and MLT−2, that is, L and F, need appear
in the dimensional formulae of p, w, and A. These symbols, their
descriptions, and their dimensional formulae in terms of these effective
dimensions are collected in Table 2.2. Note that 3 dimensional variables and
constants and 2 effective dimensions produce, according to NP = NV − ND, 1
dimensionless product, which we know is pA/w.

Table 2.2
p Pressure (MLT−2)L−2 FL−2

w Weight (MLT−2) F

A Area L2 L2

Because effective dimensions, used in this way, recognize and encode

relationships among the dimensional variables and constants, their number
is always less than or equal to the number of dimensions in which these
variables and constants are otherwise expressed, in this case, M, L, and T.
The concept of effective dimension is usually unremarked, because the
Rayleigh algorithm automatically, if implicitly, identifies the number ND of
effective dimensions in such a way that NP = NV − ND is always observed.
In particular, the number of effective dimensions is the number of
independent constraints that determine the exponents defining the
dimensionless products. [9] However, once identified, effective dimensions
may be explicitly adopted – as we did in the last column of Table 2.2.
 2.3 Imposed Dimensions
Modeling a physical state or process is more art than algorithm. One aspect
of the dimensional analyst’s art is to choose dimensional variables and
constants that appropriately describe a physical process or state. Another is
to choose the dimensions in terms of which the dimensional formulae of
these variables and constants are expressed. Often, in mechanical problems
we choose the dimensions M, L, and T, but in the last column of Table 2.2
we chose, instead, the dimensions L and F.
I call the dimensions in terms of which one chooses to denominate the
dimensional variables and constants the imposed dimensions because the
dimensional analyst consciously adopts or imposes them. [10] Imposed
dimensions no less than effective dimensions define a dimensional model.
As we have seen and shall see, the dimensions M, L, and T are not the only
dimensions that can be imposed on a mechanical state or process.
Throughout this book, we will invariably use Rayleigh’s algorithm
and, therefore, will automatically take advantage of the effective
dimensions implied by the list of dimensional variables and constants and
the dimensions in which their formulae are expressed. On occasion,
especially here in Chapter 2 and in Chapters 3 and 4, we will impose
dimensions that explicitly encode relationships that further define a
dimensional model even before we execute the Rayleigh algorithm.
Consciously imposed dimensions may increase ND and consequently
decrease the number of dimensionless products NP(=NV − ND). Comparing
the following analysis with the related one of Section 1.9 makes this point.
 2.4 Example: Hanging Spring-Mass
System
We first encountered the hanging spring-mass system in Section 1.9. As the
presumably rigid mass m at the end of the spring bounces up and down, its
volume V is not related to its position or to the extension or compression of
the spring. Therefore, the dimensional formula of volume [V] is
independent of the dimension L that describes the position of the mass and
the extension or compression of the spring. In particular, [V] is not L3.
Rather, [V] = V where, in this case, V is the symbol of an imposed
dimension as well as the symbol of a dimensional variable denoting
volume. Consequently, we reconfigure the table in Section 1.9 in terms of
the imposed dimensions M, L, T, and V. Then the number of dimensions ND
increases from 3 to 4 and, consequently, the number of dimensionless
products NP(=NV − ND) decreases from 3 to 2. The symbols that
characterize the hanging spring-mass system, their descriptions, and their
dimensional formulae in terms of these imposed dimensions are collected in
Table 2.3.

Table 2.3
ω Frequency T−1

m Mass M

k Spring constant MT−2

ρ Mass density MV−1

 V Volume V

g Acceleration of gravity LT−2

The product ωmαkβργVδgε becomes dimensionless when

(2.4)
The constraints

T: −1 − 2β − 2ε = 0,
(2.5a)
M: α + β + γ = 0,
(2.5b)
V: −γ + δ = 0,
(2.5c)
and

L: ε = 0
(2.5d)
are solved by β = − 1/2, γ = 1/2 − α, δ = 1/2 − α, and
ε = 0. Therefore, ωmαkβργVδgε =(ωρ1/2V1/2/k1/2)(m/ρV)α where α is arbitrary,
and so the two dimensionless products are and ρV/m.
Because these two products are complete, we may form all others from
them. In particular, if we multiply the first by the inverse square root of the
second, we produce the convenient pair and ρV/m. These are
related by

(2.6)
where h(x) is an undetermined function. Given m = ρV, this result reduces
to

(2.7)

where C = h(1) is an undetermined, dimensionless number. The

“monstrous” dimensionless product m4/3g/(kρ1/3) that plagued the analysis
of Section 1.9 never emerges.
When used uncritically, as they are in Section 1.9, the dimensions M,
L, and T are absolute because they allow, however implicitly, any and every
kind of mechanical state or process. In contrast, imposed dimensions are
relative because they restrict the particular state or process to the model
adopted – with its idealizations and simplifications – rather than to the
whole universe of physical possibilities.
 2.5 Example: Hanging, Stretched Cable
While an idealized spring pushes or pulls at a point, materials in bulk, such
as steel, glass, concrete, bone, water, or air, push or pull across an area. The
equivalent “spring constant” for these materials is called the bulk modulus.
The bulk modulus K of an isotropic material is defined as the
proportionality constant between the increment in material pressure Δp and
the relative increment in the material volume ΔV/V so that
Δp = − K(ΔV/V). The negative sign indicates that a decrease in the volume
V of the material causes an increase in the pressure p it exerts. Therefore,
the dimensional formula of the bulk modulus [K] is the same as that of
pressure [p]. The bulk moduli of common materials in terms of their SI unit,
the Pascal or Pa, are collected in Table 2.4.a

Table 2.4
Material Bulk moduli (in Pascals)

Steel 160 ⋅ 109

Glass (35 − 55) ⋅ 109

Concrete 30 ⋅ 109

Bone ∼9 ⋅ 109

Water 2.2 ⋅ 109

Air 1.01 ⋅ 105

 Consider a uniform, massive cable that hangs from one end. How
much does the cable stretch s as a fraction of its relaxed length l? We
believe that this stretch must depend not only on l but also on λ the mass per
unit length of the relaxed cable, g the acceleration of gravity, and K the bulk
modulus of the cable material. These symbols, their descriptions, and their
dimensional formulae are collected in Table 2.5. The third column lists the
dimensional formulae in terms of M, L, and T while the fourth column lists
the dimensional formulae in terms of the imposed dimensions F, M, and L.
This is an equilibrium problem, but introducing force F as an imposed
dimension does not, in this case, increase the number of dimensions ND. In
each case, NV = 5, ND = 3, and, according to the rule of thumb
NP = NV − ND, the analysis should produce 2(5 − 3) dimensionless products.

Table 2.5
s Stretch L L

l Length L L

λ Mass per ML−1 ML−1

length

g Acceleration of LT−2 FM−1

gravity

K Bulk modulus ML−1T−2 FL−2

Another tactic would be to combine mass per length λ with the

acceleration of gravity g to form a weight per length λg where [λg] = FL−1.
In this case the table is reduced to four rows as shown in Table 2.6. This
tactic reduces both NV and ND each by 1 and, therefore, preserves the
number NP(=2) of dimensionless products. Any of these three approaches
serves. Since it is easier to work with fewer variables, constants, and
dimensions, we adopt Table 2.6, in which F and L are the only imposed
dimensions.

Table 2.6
s Stretch L

l Length L

λg Weight per length FL−1

K Bulk modulus FL−2

In this case we do need the Rayleigh algorithm to discover the two

independent and complete dimensionless products. A simple inspection of
the table will do. The pair of dimensionless products s/l and λg/lK is
particularly convenient because the variable of interest s appears in only
one place. Therefore, these dimensionless products are related by

(2.8)

where the function f(x) is undetermined. This is as far as dimensional

analysis per se takes us.
However opaque (2.8) may seem, it has reduced an unknown relation
among 5 dimensional variables and constants to an unknown function
between 2 dimensionless products. Furthermore, (2.8) is a good starting
point for a different kind of analysis that, when justified, transforms an
unknown function into a power of a dimensionless product.
 2.6 Asymptotic Behavior
We are sometimes confronted with an undetermined function of a single
variable f(x) where x is either very small or very large compared to one.
When x ≪ 1 or x ≫ 1 there are three possibilities [11]: (1) f(x) approaches
a non-vanishing dimensionless constant C, (2) f(x) approaches a power law
C ⋅ xn where n is a non-zero exponent, and (3) neither of these.
Possibility (3) includes, for instance, the function f(x) oscillating
indefinitely or changing exponentially or logarithmically. Possibilities (1)
and (2) are summarized in the power law approximation

f(x) ≈ C ⋅ xn
(2.9)
valid when either x < < 1 or x > > 1 and where both C
and n are undetermined and n may vanish.
Of course, dimensional analysis alone does not reveal the asymptotic
behavior of f(x). But our physics sense might. If x < < 1 or x > > 1, we
should ask ourselves, “How should f(x) behave as x → 0 or x → ∞? Should
f(x) approach a finite constant or should it become monotonically small or
large?” If our answer is “Yes” to either of these questions, the power law
limit f(x) → C ⋅ xn may be appropriate. And, if so, we may be able to
determine whether n = 0, n > 0, or n < 0.
To illustrate, we return to the stretched, hanging cable of Section 2.5.
Suppose we are most concerned with cables of, say, a few meters, more or
less, a mass density of less than one kilogram per meter, and common, solid
materials with a bulk modulus on the order of 109 ⋅ Pa or greater. In this
case the dimensionless ratio λg/lK ≤ 10−8. Since there is no reason f(λg/lK)
should oscillate or change exponentially or logarithmically as λg/lK
decreases, a power law asymptotic form makes sense. Therefore, we
approximate (2.8) with

(2.10)

where the number C and the exponent n are undetermined.

Our expectations for the dependence of the cable stretch s on the
dimensional variables and constants, l, λ, g, and K, further constrain n. First,
n > 0. Otherwise s would not increase with the chain’s mass per unit length
λ. Furthermore, because longer cables of the same composition should
stretch more than shorter ones, it must be that n < 1. Therefore, 0 < n < 1. A
guess of n = 1/2 produces the scaling . While consistent
with dimensional homogeneity and our expectations, this result goes
beyond what we know.
 2.7 Example: Speed of Sound
Sound travels in air, water, and metals. All these are elastic materials whose
parts, when displaced a small distance, are pushed or pulled back into
position. The best expression of the internal forces that keep these materials
close to equilibrium is their bulk modulus K.
This displacement from and return to equilibrium of one part of a
material results in an oscillation whose effect is transmitted to its
neighboring parts at a characteristic speed called the sound speed.
Therefore, the sound speed cS of an elastic material should depend on its
bulk modulus K and on its mass density ρ, since the latter describes the
material’s inertia. What else? The internal kinetic energy of its parts as
determined by a particle mass m and speed ? Possibly.
The 4 dimensional variables and constants cS, K, ρ, and describe a
non-equilibrium, mechanical model. For this reason we impose the 3
dimensions M, L, and T as indicated in Table 2.7.

Table 2.7
cS Sound speed LT−1

K Bulk modulus ML−1T−2

ρ Mass density ML−3

Internal kinetic ML2T−2

energy
 If the imposed dimensions M, L, and T are the effective dimensions,
the Rayleigh algorithm will produce 1(4 − 3) dimensionless product. The
dimensional variables and constants form a dimensionless product
where cS occupies the place reserved for the variable of
interest. The dimensional formula

(2.11)
implies that when the exponents α, β, and γ satisfy

L: 1 − α − 3β + 2γ = 0,
(2.12a)
T: −1 − 2α − 2γ,
(2.12b)
and

M: α + β + γ = 0
(2.12c)
the form will be dimensionless. The
solution of equations (2.12) is α = − 1/2, β = 1/2, and γ = 0. Therefore,

(2.13)

where C is an undetermined dimensionless number. Since the bulk modulus

K of an ideal gas is its pressure p, the speed of sound in an ideal gas is given
by .
 2.8 Example: Side Window Buffeting
Many of us are familiar with the “whuppa, whuppa, whuppa,” sound made
when a rear window of a rapidly moving car is opened. This sound can be
quite annoying – even painful. Usually the only remedy is to open more
windows or to close them all. Sometimes speeding up or slowing down
helps. This side window buffeting is more pronounced in newer, tighter,
more aerodynamic cars than in older, less tight, less aerodynamic ones.
A car with a single open window is actually a large version of what is
known as a Helmholtz resonator. When Hermann von Helmholtz (1821–
1894) described their properties in 1885, he had no idea that someday
human beings would climb inside large, self-propelled versions of these
resonators. Helmholtz’s own resonators were handheld glass or metallic
cavities of volume V, usually spheroidal (but the shape is not crucial), with
a small opening of area A. Helmholtz chose values of A and V so that the air
inside the cavity resonated with a particular frequency ω. On the side
opposite the opening of area A he added a protruding orifice with an even
smaller opening that could be inserted into his ear. Helmholtz used his set
of resonators to isolate and listen to particular frequencies in the sounds
made by “the whistling of the wind, the rattling of carriage wheels, the
splashing of water.” [12] When one blows across the mouth of an empty
beer bottle, one hears the low pitch to which the bottle, an accidental
Helmholtz resonator, is tuned.
The resonate frequency ω of a car with an open window of area A is
determined by the volume of the passenger compartment V, the mass
density ρ of air, and the pressure p of the air. After all, it is the air in the car
that vibrates, and the state of the air is completely determined by its volume
V, its mass density ρ, and its pressure P. These symbols, their descriptions,
and their dimensional formulae are collected in Table 2.8.

Table 2.8
ω Frequency T−1

A Area of opening L2

V Resonator volume L3

ρ Density of air ML−3

p Air pressure ML−1T−2

Because there are 5 dimensional variables, ω, A, V, ρ, and p, and 3

imposed dimensions, M, L, and T, we expect to find 2 dimensionless
products related to one another by an undetermined function. These
dimensionless products assume a form ωAαVβργpδ whose dimensional
formula is given by

(2.14)

The 4 exponents α, β, γ, and δ that render ωAαVβργpδ dimensionless are

constrained by the three equations:

T: −1 − 2δ = 0,
(2.15a)
L: 2α + 3β − 3γ − δ = 0,
(2.15b)
and

M: γ + δ = 0.
(2.15c)
Their solution, β = 1/3 − 2α/3, γ = 1/2, and δ = − 1/2,
reduces the dimensionless product to ωV1/3(ρ/p)1/2(A/V2/3)α where α is
arbitrary. Therefore,

(2.16)

where f(x) is an undetermined function.

According to (2.16), the resonant frequency ω is directly proportional
to , which itself is proportional to the speed of sound in air. Also
resonators with the same shape yet of different sizes, that is, having the
same ratio A/V2/3 but with different volumes V, are such that ω ∝ 1/V1/3.
However, we also know that A/V2/3 is very small in a typical
Helmholtz resonator. Certainly the linear size of a car’s passenger
compartment V1/3 is much larger than that of an open window A1/2. When
A/V2/3 < < 1 (2.16) is approximated by the power law

(2.17)

where C and n are undetermined numbers. Helmholtz determined that

n = 1/4 in which case (2.17) reduces to

(2.18)
 2.9 Example: Two-Body Orbits
The moon orbits the earth, the earth orbits the sun, and the stars in a double
star system orbit each other. Each of these systems contains two bodies, one
of mass m1 and the other of mass m2, separated by a characteristic distance
r. Their gravitational attraction causes each body in a system to orbit their
mutual center of mass. This motion has a period Δt after which time the
massive bodies return to their initial positions and velocities. Thus, 4
dimensional variables, Δt, r, m1 and m2, and 1 dimensional constant, G,
completely characterize this two-body, gravitationally bound system. Their
symbols, descriptions, and dimensional formulae are collected in Table 2.9.

Table 2.9
Δt Period T

r Characteristic L
separation

m1 Mass 1 M

m2 Mass 2 M

G Gravitational constant M−1L3T−2

Since there are 5 dimensional variables and constants and 3 imposed

dimensions, two independent and complete dimensionless products should
characterize this non-equilibrium dynamical system. With practice we will
develop patterns of thought that allow us to forgo the Rayleigh algorithm
and quickly identify these dimensionless products. However, the Rayleigh
algorithm always works and has the advantage of methodological clarity if
not calculational efficiency.
Accordingly, we require that be dimensionless.
Therefore,

(2.19)

and

T: 1 − 2δ = 0,
(2.20a)
L: α + 3δ = 0,
(2.20b)
and

M: β + γ − δ = 0
(2.20c)
renders dimensionless. The solution to these
constraints is α = − 3/2, γ = 1/2 − β, and δ = 1/2, and so

(2.21)

where β is arbitrary. Therefore, the two independent dimensionless products

are Δt2m2G/r3 and m1/m2. These are related to one another by

(2.22)
where the function f(x) is undetermined. This result expresses Kepler’s third
law, according to which Δt2 ∝ r3 where r is directly proportional to the
semi-major axis of a planet’s elliptical orbit around the sun. This is as far as
dimensional analysis takes us.
Symmetry considerations take us a little further. Replacing m1 with m2
and m2 with m1 should leave (2.22) unchanged. Therefore,

(2.23)

or, equivalently,

(2.24)

where x is any positive dimensionless number. Equation (2.24) is a

functional equation one solution of which on the domain x > 0 is
f(x) = C/(1 + x) where C is an undetermined dimensionless number. An
analysis based on Newton’s second law leads to f(x) = 4π2/(1 + x) and
therefore to

(2.25)
 Essential Ideas
The dimensions in terms of which the dimensional formulae of the
dimensional variables and constants are expressed are not absolute, but
rather relative to the model adopted. These relative or imposed dimensions
encode idealizations and simplifications that define the model. The number
of effective dimensions is the minimum number of imposed dimensions or
combinations of imposed dimensions in terms of which the dimensional
variables and constants are expressed. The rule of thumb NP = NV − ND is
always observed when ND is the number of effective dimensions.
 Problems
2.1 Gravitational constant. Use Newton’s law of gravitation
F = Gm1m2/r2 to determine the dimensional formula of the gravitational
constant G in terms of M, L, and T .
answer
2.2 Interior pressure. A non-rotating, self-gravitating, spherical body has a
mass density ρ and radius R. [In parts (a) and (b) below use the result of
Problem 2.1.]

(a) Use the Rayleigh algorithm to show that po = C ⋅ Gρ2R2 where po is the
pressure at the center of the sphere and C is an undetermined dimensionless
constant.

(b) Also determine an expression for the pressure p(r) at a distance r from
the center of the sphere.
answer
2.3 Gravitational instability. A self-gravitating, spherical body of mass
density ρ and radius R rotates with at an angular frequency ω. Find how the
critical frequency ω* at which the sphere begins to break up depends on ρ,
R, and G. [13]
answer
2.4 Escape velocity. Determine how the speed υ required for any object to
escape from a planet depends on the planet’s mass M and radius R. See
Figure 2.2.
answer
 Figure 2.2. Escape velocity.

2.5 Transverse waveform. A long wire with mass per unit length ρ is
tightly strung between two posts as illustrated in Figure 2.3. Its tension τ is
essentially uniform throughout. The wire is plucked near one end and a
transverse waveform propagates as shown in Figure 2.3. Determine an
expression for the waveform’s speed of propagation υ in terms of τ and ρ.
answer

Figure 2.3. Transverse waveform.

2.6 Drag force in an ideal fluid. A sphere of radius r moves through a

fluid of mass density ρ at a speed υ. Assume the fluid is merely pushed out
of the way by the sphere without, in any way, adhering to it.

(a) How does the drag force FD that resists this motion depend on these
variables?
answer
(b) Suppose that the sphere is falling at terminal speed through the fluid.
Given that the drag force FD perfectly balances the sphere’s weight mg, find
how the terminal speed υ depends on mg, ρ, and r.
answer
2.7 Free vibrations. A body of mass m, size l, and bulk modulus K vibrates
with a natural (unforced) frequency ω. Find an expression for ω in terms of
m, l, and K.
answer
2.8 Soap bubble. The air pressure within a spherical soap bubble of radius
r and surface tension σ exceeds that outside the bubble by Δp. Since the
soap bubble is in an equilibrium state, force is an imposed dimension F
independent of M, L, and T.

(a) How is Δp related to r and the surface tension σ of the bubble? One
should find the, possibly, counterintuitive result that Δp ∝ r−1.

(b) Determine NV, ND, and NP. Show that NP = NV − ND.

2.9 Blast wave. Energy E is quickly released at a point near the surface of
the Earth. The resulting blast wave propagates through air of density ρ and
creates a hemispherical shock of radius R at time t. How does R depend on
time t, ρ, and E? [14]
answer

aThese bulk moduli are at standard temperature, 0o Celsius, and standard

pressure,1.01 ⋅ 105 Pascals or 1 atmosphere.
 3
Hydrodynamics
◈
 3.1 Fluid Variables
A fluid is a continuously distributed material characterized by a mass
density ρ, a speed υ, and, when isotropic or non-directional, a single
pressure p at a point. In general, these basic kinematical descriptors ρ, υ,
and p may vary from place to place. Here we consider only incompressible
fluids whose mass density ρ is constant throughout. Under normal
conditions water is the most familiar realization of an incompressible fluid.
The adjacent parts of a fluid can shift with respect to one another and
push, pull, and accelerate each other with forces originating in non-uniform
pressure p, in viscosity μ, and in surface tension σ – as explained later in
this chapter. Gravity, which tends to straighten out the free surface of a fluid
by pulling its parts down toward the center of the earth, also operates.
These terms (mass density ρ, speed υ, pressure p, viscosity μ, surface
tension σ, and the acceleration of gravity g) are related to one another by
equations of motion that, because of their many applications, have been
carefully studied. Our concern is with the constraints that dimensional
homogeneity imposes on the dimensional variables and constants that enter
into these equations.
 3.2 Example: Water Waves
Water is a relatively dense fluid whose free surface, say, on a pond or lake,
is flat – at least when that surface is in equilibrium. When the surface is
disturbed, gravity pulls the water back toward flatness, the water overshoots
its equilibrium position, disturbs neighboring regions, and the disturbance
propagates. Figure 3.1 illustrates the variables of a water wave disturbance:
wavelength λ, wave amplitude a, depth of the water d, and propagation
speed υ. Since the restoring force is gravitational, the fluid mass density ρ
and the acceleration of gravity g also play a role. How does the wave speed
υ depend on λ, a, d, ρ, and g?

Figure 3.1. The variables of a wave disturbance on the free surface of an

incompressible fluid.

These 6 dimensional variables and constants, their descriptions, and

their dimensional formulae are collected in Table 3.1. Because we express
their dimensional formulae in terms of 3 imposed dimensions, M, L, and T,
the Rayleigh algorithm should produce 3 dimensionless products. An
inspection of Table 3.1 reveals them to be , λ/d, and a/d or, if one
prefers, , , and . Note that the mass density ρ cannot
and does not enter into these dimensionless products. Dimensional analysis
takes us no further.

Table 3.1
υ Wave speed LT−1

λ Wavelength L

a Wave amplitude L

d Fluid depth L

ρ Mass density ML−3

g Acceleration of gravity LT−2

To go further we appeal to the approximations of wave theory and to

the properties of water waves. In particular, we limit our analysis to small
amplitude waves whose amplitude a is negligibly small compared to both
the wavelength λ and the water depth d. This approximation is, of course,
not valid for waves that crash on the beach, but otherwise works well. With
wave amplitude a eliminated from the analysis, we have only 5 dimensional
variables and constants υ, λ, g, ρ, and d expressed in terms of 3 imposed
dimensions M, L, and T. The 2 remaining dimensionless products,
and λ/d, are related to one another by

(3.1)

where f(x) is undetermined. However, (3.1) tells us only that the speed υ of
a small-amplitude water wave is proportional to the square root of the
acceleration of gravity g.
The limiting cases of deep-water waves for which d > > λ and for
shallow-water waves for which d < < λ yield to further analysis. In both
cases (3.1) reduces to an asymptotic form. Assuming this asymptotic form
is that of a power law, (3.1) becomes

(3.2)

where C and n are undetermined and may be different in the two cases
λ/d < < 1 and λ/d > > 1.
Because deep-water waves on the surface of the ocean seem to be
independent of ocean depth, it must be that in this limit n = 0 and therefore
the speed of a deep-water wave

(3.3)

Thus, longer-wavelength, deep-water ocean waves propagate more quickly

than do shorter-wavelength ones. We may have observed this phenomenon
firsthand or in a film of a long-wavelength swell passing under and
proceeding beyond shorter-wavelength waves on the swell’s surface.
Further analysis indicates that, in this case, .
Shallow-water ocean waves for which d < < λ are called tsunamis
after the Japanese word for harbor waves. Tsunamis, which are initiated
when a large volume of ocean water is quickly displaced from equilibrium –
say, by earthquake, underwater landslide, or asteroid impact – can have
wavelengths as long as 1,000 kilometers. A different kind of analysis, as
well as observation, indicates that tsunami speeds are independent of
wavelength. In this case n = − 1/2 and so

(3.4)

It has been shown that . The wave speed of the devastating

tsunami of 2004 was as high as 1,000 km/hour.
 3.3 Surface Tension
The force that restores the surface of a liquid to equilibrium on a small scale
is surface tension rather than gravity. Surface tension originates in the
attraction of the individual molecules of liquid to nearby molecules. Within
the body of a liquid a single molecule is pulled indifferently in all directions
and the net attraction vanishes. However, these attractions pull a molecule
on the free surface of a liquid inward – toward the body of the fluid. It is for
this reason that while large bodies of liquid, under the dominant influence
of gravity, have a flat surface, small drops of liquid, under the dominant
influence of surface tension, are spherical.
The surface tension of a liquid is characterized by a dimensional
quantity σ defined as the net force f that the molecules on one side of a short
line on the surface of the liquid exert on the adjacent molecules on the
opposite side of that line, divided by the length l of the line. Thus, σ = f/l
and [σ] = FL−1 or [σ] = MT−2 depending on whether these forces are
balanced or not.
 Capillary Waves
Capillary waves are small surface waves or ripples whose dominant
restoring force is surface tension. Suppose a liquid is indefinitely deep and
the amplitude of a capillary wave is so small as to play no role in its
behavior. Then its wave speed υ can depend only on wavelength λ, surface
tension σ, acceleration of gravity g, and, possibly, the mass density ρ of the
liquid. These 5 symbols, their descriptions, and their dimensional formulae
are collected in Table 3.2.

Table 3.2
υ Wave speed LT−1

λ Wavelength L

σ Surface tension MT−2

g Acceleration of gravity LT−2

ρ Mass density ML−3

Because the dimensional formulae of these 5 dimensional variables

and constants are expressed in terms of 3 imposed dimensions, M, L, and T,
the variables and constants should combine into 2 independent,
dimensionless products. These take a form whose dimensional formula,

(3.5)

implies that the exponents α, β, γ, and δ are constrained by

 L : 1 + α + γ − 3δ = 0,
(3.6a)
T : − 1 − 2β − 2γ = 0,
(3.6b)
and

M : β + δ = 0.
(3.6c)
The solution to equations (3.6) is β = − 1/4 − α/2,
γ = − 1/4 + α/2, and δ = 1/4 + α/2. Consequently, the dimensionless
products are υ[ρ/(σg)]1/4 and λ(gρ/σ)1/2. By multiplying the first by the
square root of the second, we produce υ(λρ/σ)1/2. This dimensionless
product and λ(gρ/σ)1/2 compose an independent pair. In this way we find
that the wave speed

(3.7)

where f(x) is undetermined by our analysis.

The size of the dimensionless argument of the function f(x)
determines whether it is gravity or surface tension that dominates. When
gravity dominates , it must be that

in order to recover (3.3), , the

latter derived under the assumption that σ = 0. When surface tension

dominates , it must be that in order

that the acceleration of gravity g vanish from the result. In this case,
where C′ = f(0). When , both gravity and
surface tension play a role.
 The distance that separates these two regimes is easily
calculated for water. Water at 25 °C and atmospheric pressure has a surface
tension σ = 0.0720 ⋅ N/m and density ρ = 0.997 ⋅ 103 ⋅ kg/m3. Given
g = 9.8 ⋅ m/s2, we find that or 2.7 ⋅ mm. In general,
pure capillary waves have wavelengths so small (λ < < 2.7 ⋅ mm) as to be
invisible to the naked eye. Visible waves with wavelengths of a few
millimeters up to a centimeter are mixed capillary and gravity waves.
 3.4 Example: Largest Water Drop
The distance (=2.7 ⋅ mm) also determines the size of the largest drop
of water. Imagine a drop that hangs from a water faucet and is very slowly
filled. Eventually the drop becomes so heavy it detaches from the faucet
and falls. What is the volume V of this largest drop? Certainly V depends on
σ, for without surface tension no drop would form. The mass density of
water ρ and the acceleration of gravity g must also play a role, because it is
only because ρ and g are non-vanishing that the drop detaches and falls.
These dimensional variables and constants, their descriptions, and their
dimensional formulae are collected in Table 3.3.

Table 3.3
V Volume L3

σ Surface tension MT−2

ρ Mass density ML−3

g Acceleration of gravity LT−2

Since this is an equilibrium problem, we are free to impose force F as

a dimension. In this case, the variables, descriptions, and formulae are given
in Table 3.4. Note that the dimension F enters into those dimensional
formulae that express counterbalancing forces.

Table 3.4
V Volume L3
 σ Surface tension FL−1

w Weight density FL−3

Both sets of variables and formulae produce the same dimensionless

product. Since it is slightly easier to work with 3 rather than with 4
dimensional variables and constants, we do so here. We need no special
algorithm to discover, from the second table, the single dimensionless
product produced: V(w/σ)3/2. Therefore, given w = ρg, we find that

(3.8)

where C is an undetermined dimensionless number. Here

(σ/gρ)3/2 = 2.0 ⋅ 10−8m3 or 20 ⋅ mm3.
Dimensional analysis does not determine the number C in (3.8). But C
can be empirically determined. Catching and counting the water drops
falling slowly from my kitchen faucet and measuring their total volume
yields C ≈ 9 so that

(3.9)

Therefore, each of these drops has a mass of about 180 milligrams, a

volume of about 180 ⋅ mm3, and a radius of about 5.6 ⋅ mm. These are the
largest drops that can be formed under the circumstance described. Smaller
drops can, of course, be formed, say, by a nozzle that mixes air and water,
and larger drops could be formed in a weightless environment.
 3.5 Viscosity
Among the variables and constants that describe an incompressible fluid,
the viscosity μ is likely to be the least familiar. The viscosity of a fluid
quantifies its tendency, through internal friction, to reduce the relative
motion of its adjacent parts. Without viscosity, eddies in water would
circulate forever. All fluids are, to a degree, viscous, but viscosity is not
always important in all phenomena.
Imagine a fluid trapped between two large, flat, horizontal plates. One
plate is fixed in place and the other moves horizontally at constant speed.
The fluid resists the motion of the moving plate with a drag force FD per
unit area A that is proportional to its speed υ and inversely proportional to
the separation y of the two plates. Symbolically,

(3.10)

The dynamic or shear viscosity μ is the proportionality constant that turns

(3.10) into the equality

(3.11)

Therefore, [μ] = ML−1T−1 or, in equilibrium problems, [μ] = FL−2T. In

practice, measurements of viscosity are often carried out on a thin layer of
fluid trapped between two closely spaced, concentric cylinders, one of
which can rotate while the other is attached to a torsion meter.
 Terminal Speed
Since viscosity quantifies the ability of a fluid to resist the motion of an
object passing through it, viscosity contributes to the phenomenon of
terminal speed. Objects falling in a fluid, such as air, will approach a
terminal speed that depends on the weight and shape of the object and the
properties of the fluid – particularly its viscosity.
Because objects fall relatively quickly in air, the Greek philosopher
Aristotle (384–322 BCE) may have developed his ideas about falling
objects by extrapolating from a study of the descent of stony pebbles in a
clear pool of water. [15] According to Aristotle, all space is full, that is,
filled with a fluid, whether water or air, or something less dense but akin to
these fluids. Thus, Aristotle imagined that descent in water is structurally
similar to descent in air.
The dimensional analysis of a spherical pebble falling at terminal
speed through a viscous fluid employs the symbols, descriptions, and
dimensional formulae found in Table 3.5. Since this is an equilibrium
problem, the fourth column includes the force as an imposed dimension F.
In either case there are 4 dimensional variables and constants and 3
imposed dimensions. Therefore, the analysis should produce 1
dimensionless product.

Table 3.5
υ Terminal speed LT−1 LT−1

μ Viscosity ML−1T−1 FL−2T

mg Weight MLT−2 F
 r Radius L L

Adopting the imposed dimensions M, L, and T, we find that

(3.12)

where the exponents must satisfy

L : 1 − α + β + γ = 0,
(3.13a)
T : − 1 − α − 2β = 0,
(3.13b)
and

M:α+β=0
(3.13c)
in order to make υμα(mg)βrγ dimensionless. The solution
to these constraints is α = 1, β = − 1, and γ = 1. Therefore, the single
dimensionless product is υμr/mg, and so

(3.14)

where C is an undetermined, dimensionless number. The terminal speed of

an object falling in a viscous fluid is, indeed, directly proportional to the
object’s weight mg and inversely proportional to the fluid’s resistance, that
is, to its viscosity μ – just as Aristotle claimed for all falling objects. [16]
 Stokes’ Law
At terminal speed the drag force FD the fluid exerts on a spherical object by
virtue of the fluid’s viscosity balances the object’s weight mg. Given (3.14),
this drag force

FD = C′ ⋅ υrμ
(3.15)
where C′ = 1/C. Relation (3.15) with C′ = 6π is known as
Stokes’ law after George Stokes (1819–1903), the British mathematician,
scientist, and engineer who discovered it in 1851.
 3.6 Example: Hydraulic Jump
Anyone with a kitchen sink has observed the hydraulic jump. As water falls
from the faucet and strikes the relatively flat bottom of the sink, it spreads
out in a smooth circular pattern that is surrounded by a stationary wave
form or jump. Beyond the jump, the water flows more deeply and roughly.
This jump develops in order to maintain a constant rate of flow given that
viscosity continually slows the water flowing along the sink bottom. Figure
3.2 illustrates the situation.

Figure 3.2. Hydraulic jump.

The radial position R at which the hydraulic jump forms depends on

the mass density ρ of water and its viscosity μ. Because the water’s
volumetric flow rate πr2υ is also important, the radius r of the bottom of the
cylindrical column of falling water and the speed υ with which the water at
the bottom of the column strikes the sink are also determinants. (Because
πr2υ is constant along the column, it narrows as the falling water picks up
speed.) These symbols, their descriptions, and their dimensional formulae
are collected in Table 3.6.

Table 3.6
R Position of jump L

ρ Mass density ML−3

μ Viscosity ML−1T−1

υ Speed LT−1

r Radius of column L

Since there are 5 dimensional variables and constants and 3 imposed

dimensions, we expect, according to the rule of thumb NP = NV − ND, to
find 2 dimensionless products. These have the form Rραμβυγrδ. Thus, their
dimensional formulae are given by

(3.16)

where the exponents α, β, γ, and δ are constrained by

L : 1 − 3α − β + γ + δ = 0,
(3.17a)
M : α + β = 0,
(3.17b)
and

T : − β − γ = 0.
(3.17c)
The solution of (3.17) is β = − α, γ = α, and δ = − 1 + α. These determine
two dimensionless products, R/r and rρυ/μ, that are related by

(3.18)

where f(x) is undetermined.

 Reynolds Number
The dimensionless ratios rρυ/μ and Rρυ/μ are examples of Reynolds
numbers, so called after Osborne Reynolds (1842–1912) who promoted
their use. In general, the size of a Reynolds number in a given geometry
determines whether a fluid in that geometry is dominated by viscosity or
not. When the Reynolds number is small, viscosity dominates and the flow
is smooth, streamline, or laminar. When the Reynolds number is large,
viscosity is not so important and the flow is rough or turbulent. The critical
Reynolds number separating these two regimes is not necessarily 1, but
rather depends on the geometry of the flow. When water flows over a flat
surface, as it does along the sink bottom, the relevant, critical Reynolds
number Rρυ/μ is, in fact, about 5 ⋅ 105.
Numbers that characterize the flow of water in my sink are
R = 0.20 ⋅ m, r = 0.02 ⋅ m, and υ = 2 ⋅ m/s. Given that for water
ρ = 103 ⋅ kg/m3 and μ = 8.9 ⋅ 10−4 ⋅ kg/(m ⋅ s), I find that Rρυ/μ = 4.5 ⋅ 105 –
a value close to the critical Reynolds number for this geometry. Because
rρυ/μ is also large compared to 1, we replace f(rρυ/μ) on the right-hand side
of (3.18) with its power law approximation. Thus,

(3.19)

where C and n are undetermined dimensionless numbers.

Experience helps us determine the possible values of n. First, we
rewrite (3.19) as
 (3.20)

I observe that a larger flow rate ρr2υ causes a larger jump position R.
Therefore, n > 0. I also observe that increasing r while preserving ρr2υ, say,
by raising a cookie sheet under a steady stream of water, weakly increases
R. Therefore, n < 1 and so 0 < n < 1. According to one model for which
there is experimental support, n = 1/3. [17] In this case (3.20) becomes

(3.21)

Given the above data, I find that, for my sink and faucet, C ≈ 0.25.
 3.7 Example: Equilibrium Pipe Flow
Consider the equilibrium flow of an incompressible fluid at a volumetric
rate Q from left to right through a straight, horizontal pipe of length l and
uniform cross-sectional area A as illustrated in Figure 3.3. The quantity Q/A
is the average fluid speed υ over a cross-section. Thus, Q = υA. By
equilibrium flow I mean one for which the net force on each part of the
fluid, that is, on each fluid element, vanishes. Therefore, the velocity of
each fluid element is constant.

Figure 3.3. Pipe flow variables.

Given the interaction with the walls of the pipe and the retarding effect
of the fluid’s viscosity μ, a force must push the fluid from left to right in
order to keep it moving at constant speed. This force is maintained by a
pressure gradient ∇p pointing in the direction opposite to the flow, that is,
from right to left. Therefore, as the fluid moves from left to right over
length l, its pressure drops by an increment l∇p. How does the flow rate Q
depend on the cross-sectional area A, the pressure gradient ∇p, the fluid
viscosity μ, and the fluid mass density ρ?
 Since this is an equilibrium problem, the dimension F is an appropriate
imposed dimension along with and independent of M, L, and T. As usual,
the dimension F enters only into those dimensional formulae that represent
counterbalancing forces. In particular, [Δp] = FL−3 and [μ] = FL−2T.
Note that imposing the dimension F, as we do here, is mathematically
distinct from using the relationship [F] = [m][a] and its consequence
F = MLT−2 to eliminate one of the dimensions M, L, or T in favor of F. The
former restricts the solution to one in which the net force on every fluid
element vanishes. In this case, forces are independent of mass times
acceleration. The latter applies Newton’s second law F = ma by adopting
the equivalence F = MLT−2.
The dimensional variables and constants, their descriptions, and their
dimensional formulae are collected in Table 3.7. Since there are 5
dimensional variables and constants (Q,∇p,μ,ρ, and A) and 4 imposed
dimensions (M, L, T, and F), there should be 1 dimensionless product. Table
3.7 reveals that, since the dimension M appears only once, the mass density
ρ cannot enter into this dimensionless product.

Table 3.7
Q Flow rate L3T−1

∇p Pressure gradient FL−3

μ Viscosity FL−2T

ρ Mass density ML−3

A Cross-sectional area L2
 Therefore, the dimensionless product assumes a form Q(∇P)αμβAγ
where the exponents, α, β, and γ, make

(3.22)

equal to 1. Thus,

L : 3 − 3α − 2β + 2δ = 0,
(3.23a)
T : − 1 + β = 0,
(3.23b)
and

F : α + β = 0.
(3.23c)
The solution to equations (3.23) is α = − 1, β = 1, and
γ = − 2. Therefore, the single dimensionless product is Qμ/(A2∇p) and so

(3.24)

where C is undetermined. As seems reasonable, the volumetric flow rate Q

increases with increased cross-sectional area A, increases with increased
pressure gradient ∇p, and decreases with increased viscosity μ.
 3.8 Example: Non-Equilibrium Pipe Flow
If we allow parts of the fluid to accelerate, then F is no longer an
appropriate imposed dimension and should, in the above table in Section
3.7, be replaced with MLT−2. The result is shown in Table 3.8. Since there
are now only 3 imposed dimensions in terms of which the original 5
dimensional variables and constants are expressed, the Rayleigh algorithm
should produce 2 dimensionless products. One of these, Qμ/(A2∇p), has
already been determined.

Table 3.8
Q Flow rate L3T−1

∇p Pressure gradient ML−2T−2

μ Viscosity ML−1T−1

ρ Mass density ML−3

A Cross-sectional area L2

The dimensionless products assume a form Q(∇p)αμβργAδ whose

dimensional formula

(3.25)
equals 1. Therefore,
 3 − 2α − β − 3γ + 2δ = 0,
(3.26a)
−1 − 2α − β = 0,
(3.26b)
and

α + β + γ = 0.
(3.26c)
The solution to (3.26), β = − 1 − 2α, γ = 1 + α, and
δ = − 1/2 + 3α/2, produces two dimensionless products: Qρ/(μA1/2) and
ρA3/2∇p/μ2. Multiplying the first by the inverse of the second yields the
dimensionless product Qμ/(A2∇p) that, according to Section 3.7,
characterizes equilibrium flow in a straight pipe. The relation between this
dimensionless product and Qρ/(μA1/2),

(3.27)

where the function f(x) is undetermined, characterizes non-equilibrium flow

in a straight pipe.
Since Q = Aυ, the argument Qρ/(μA1/2) is the Reynolds number υρA1/2/
μ of this flow. We recover equilibrium flow (3.24) from non-equilibrium
flow (3.27) by making the fluid mass density ρ negligibly small compared
with μA1/2/Q or, equivalently, by making the viscosity μ very large
compared to Qρ/A1/2. These moves are equivalent to replacing the function
f(Qρ/μA1/2) with its low Reynolds number asymptotic value f(0) which
produces (3.24) with C = f(0). Apparently, a vanishing Reynolds number, in
a straight pipe, means the fluid moves down the pipe at constant velocity.
The general result (3.27) also has a high Reynolds number, asymptotic
limit, describing fully developed turbulent flow, in which viscosity μ
becomes negligible compared to Qρ/A1/2. In order that viscosity be
completely eliminated from (3.27) it must be that f(x) → C′/x when x > > 1.
Therefore, in this fully developed turbulent flow limit Q → C′ ⋅ A5/2∇p/(Qρ)
or, equivalently,

(3.28)

where .
 3.9 Scale Models
The result of a dimensional analysis is often that one dimensionless product
equals an unknown function of another. Relation (3.27) among the variables
describing pipe flow for arbitrary Reynolds number,

(3.29)

is one example.
Suppose our job is to design a large pipe of cross-sectional area A and
length l that transports motor oil. We want to test our design on a smaller-
scale version of that pipe. But pumping motor oil through a smaller pipe
with the same aspect ratio A1/2/l is not adequate. While proportionally
decreasing A1/2 and l preserves the geometry of the pipe, doing so does not
preserve the Reynolds number Qρ/(μA1/2) of the flow, and it is the Reynolds
number that determines whether the flow is laminar or turbulent. In fact,
preserving geometric similitude is not necessary at all. Rather, what needs
to be preserved, in this case, is dynamic similitude.
One preserves dynamic similitude by making the values of the two
dimensionless products Qμ/(A2∇p) and Qρ/(μA1/2) realized in the scale
model identical to those in the fully designed pipe. We do this by varying
the pressure gradient ∇p in the scale model and measuring the resulting
flow rate Q – the other quantities being constant – and so determining the
relationship between Qμ/(A2∇p) and in the scale model, that is,
by determining the function f(x). The pressure gradient ∇p in the scale
model is then chosen so that the two dimensionless products are the same in
the scale model and the fully designed pipe. Cars, ships, and planes, as well
as pipes, have been designed in this way.
However, scale models should not depart from the physics
incorporated in the full design. If, for instance, the scale model of a pipe is
too small, surface tension becomes significant and the scaling implied by
(3.29) breaks down.
 Essential Ideas
Dimensional models of incompressible fluids introduce the dimensional
variables mass density ρ, flow speed υ, and pressure p as well as surface
tension σ and viscosity μ. The Reynolds number is a dimensionless product
that when relatively small produces laminar flow and when relatively high
produces turbulent flow.
 Problems
3.1 Capillary waves. The context is Sections 3.2 and 3.3. Use the Rayleigh
algorithm to show that the speed υ of a capillary wave with wavelength λ in
an compressible fluid with mass density ρ and surface tension σ is given by
where C is an undetermined dimensionless number.

3.2 Water drop oscillation. Use either the Rayleigh algorithm or a less
formal dimensional argument to determine how the oscillation frequency ω
of a small drop of incompressible fluid in free fall depends on the volume V
of the drop, its mass density ρ, and its surface tension σ.
answer
3.3 Capillary effect. When a small-diameter tube open at both ends is
inserted into water, as shown in Figure 3.4, the water is drawn a certain
distance h up the tube. This distance depends on the diameter d of the tube,
the surface tension of water σ, the mass density ρ of the water, and the
acceleration of gravity g. Find the 2 dimensionless products implied by
these dimensional quantities and cast them into an expression for h.
answer

Figure 3.4. Capillary effect.

3.4 Terminal speed. The context is the discussion of viscosity dominated
terminal speed in Section 3.6. Find a dimensionally correct expression for
the time t required for an object with weight mg and size r falling in a fluid
of viscosity μ to reach terminal speed.
answer
3.5 Viscous rebound. When a spoon is depressed into and removed from
the surface of a pool of molasses, an interval Δt is required for the molasses
to return to its original, equilibrium shape. (a) Find the two dimensionless
products π1 and π2. (b) Express Δt in terms of the other variables and
constants.
answer
3.6 Lift. Determine a dimensionally correct expression for the mass m of a
rock with mass density ρ that can be lifted from the bottom of a streambed
and moved downstream. The stream is flowing with a speed υ.
answer
3.7 Dimensionless products. According to Section 3.1, mass density ρ,
speed υ, pressure p, viscosity μ, surface tension σ, and the acceleration of
gravity g are the basic dimensional descriptors of hydrodynamics. Given
these 6 dimensional variables and constants expressed in terms of the 3
dimensions M, L, and T, find 3 independent, dimensionless products, π1, π2,
and π2.
answer
 4
Temperature and Heat
◈
 4.1 Heat Transfer
Joseph Black’s early work (circa 1760) on heat capacity and latent heat
effectively disentangled the concepts of temperature and heat. Black
(1728–1799) noticed, for instance, that boiling water absorbs heat without
changing its temperature. The role of temperature is, apparently, to
determine the direction and rate of heat transfer when two bodies, with
unequal temperatures, are brought into thermal contact. Furthermore, heat
moves from place to place while keeping its quantity conserved – or so it
was thought.
This understanding of heat as an inherently conserved quantity
remained possible until the 1840s. Then James Joule (1818–1889), in a
series of increasingly precise experiments, began generating, through the
dissipation of work, predictable amounts of this supposedly conserved heat.
In this way Joule forged a new understanding of heat as heating, that is, as
simply one way of transferring energy to or from a system. The other
category of energy transfer is work done on or by a system. In 1850,
Rudolph Clausius (1822–1888) harmonized this understanding of heating
and working and Sadi Carnot’s earlier (1826) articulation of a universal law
of heat engine performance into the first and second laws of the new
science of thermodynamics.
Thermodynamics is a theory that, while providing a new foundation
for known facts, preserves the validity of older explanations in particular
limits. One of these limits is that in which no work is done on or by the
system in question and heat moves from place to place as if it were a
conserved quantity. In this limit it is permissible and often convenient to
use the word heat as short hand for “that energy transferred when no work
is done.” This limit is easily recognized and rich in phenomena.
The limit in which heat retains its status as an independent, conserved
quantity is the regime of heat transfer. Calorimetry experiments are done
and heat capacities are sometimes measured in this regime. [The word
calorimetry refers to the traditional unit of heat – the calorie.] Interestingly,
Joseph Fourier (1768–1830) first worked out the mathematics of heat
transfer in the same treatise, The Analytical Theory of Heat (1822), in
which he articulated the principle of dimensional homogeneity.
 4.2 The Dimensions of Temperature and
Heat
The physics of heat transfer requires two dimensions, temperature Θ and
heat H, beyond those that characterize non-equilibrium dynamics. Thus, for
instance,

[cp] = HΘ−1M−1
(4.1)
where cp is a heat capacity per unit mass at constant
pressure. We depend on context to distinguish between the traditional
symbols for temperature T and the dimension time T. Thus, [T] = Θ while,
if υ is a velocity, [υ] = LT−1.
When work is dissipated into or produced out of the internal energy of
a thermodynamic system, the first law of thermodynamics is necessarily
invoked. Then the dimension heat H loses its independence and should be
replaced with the dimension of energy ML2T−2. Yet when no work is done
on or by a system, heat H is an appropriate imposed dimension just as when
no part of a mechanical system is accelerated force F is an appropriate
imposed dimension.
Occasionally, one reads that the variable temperature does not need its
own dimension Θ because temperature can always be measured in energy
units with dimensional formula ML2T−2. The three examples in this chapter,
in Sections 4.4, 4.7, and 4.8, are counterexamples to this claim. Those in
Sections 5.7, 6.2, 6.3, and 6.6 illustrate states and processes in which
temperature can be measured in energy units.
 4.3 Conduction and Convection
Heat conduction is accomplished when the energy of some atoms and
molecules is imparted to other nearby atoms and molecules. Solids, liquids,
gases, and plasmas all conduct heat. Thermal conductivity is a dimensional
constant that characterizes a particular material’s ability to conduct heat. In
general, thermal conductivity may vary from point to point and even change
in time as the material properties change.
The extent to which a material conducts heat at a point is proportional
to the temperature gradient at that point. A one-dimensional quantification
of this statement is

(4.2)

where q is the heat flux or rate at which heat flows past a point per unit
area, T is the temperature, and ∂T/∂x is the x component of the temperature
gradient. The negative sign in (4.2) ensures that the heat flux is always in a
direction opposed to the temperature gradient, that is, always from hot to
cold. The thermal conductivity k is the proportionality constant that turns
(4.2) into an equality. Therefore,

(4.3)

governs heat conduction.

Convection occurs when a fluid, by virtue of its bulk motion, carries its
internal energy from one place to another. Convection may occur either by
design, in heating and cooling systems, or naturally, as in weather patterns.
 4.4 Example: Cooking a Turkey
Each year on their Thanksgiving holiday Americans consume some 50
million turkeys. Unlike chickens, which have been bred to uniform size,
turkey weights vary over an order of magnitude. Wild turkey hens weigh as
little as 3.0 kilograms (6.6 pounds) while domestic toms weigh as much as
36 kilograms (79 pounds). Turkeys found in grocery stores are usually
between 4.5 and 11 kilograms (10–24 pounds). So the question arises,
“How long should we cook our turkey? Last year our 9-kilogram turkey
was perfect after 4 hours. This year we have a 6-kilogram turkey.”
What are the physical variables and constants that contribute to a
turkey’s cooking time Δt? Certainly its mass m and possibly the difference
ΔT(=Toven − To) between the turkey’s initial temperature To and the
temperature Toven(>To) at which the oven is set are relevant. Since the
turkey’s thermal conductivity k determines how a temperature difference
leads to a heat flux, k should also contribute to the cooking time. So should
the turkey’s heat capacity per unit mass at constant pressure cp and its mass
density ρ. These parameters seem sufficient. Their symbols, descriptions,
and dimensional formulae are collected in Table 4.1. Note that the imposed
dimension heat H appears in the dimensional formulae of the dimensional
quantities that quantify heat conduction.

Table 4.1
Δt Cooking time T

m Turkey mass M
 ΔT Temperature difference Θ

k Thermal conductivity HT−1L−1Θ−1

cp Specific heat at constant HM−1Θ−1

pressure

ρ Mass density ML−3

Since there are 6 dimensional variables and constants, Δt, m, ΔT, k, cp,
and ρ and 5 imposed dimensions, M, L, T, H, and Θ, our analysis should
produce 1 dimensionless product. This product will assume a form
with exponents, α, β, γ, δ, and ε, that render the form
dimensionless. Therefore,

(4.4)
where the 5 exponents α, β, γ, δ, and ε satisfy the 5
constraints

T: 1 − γ = 0,
(4.5a)
M: α − δ + ε = 0,
(4.5b)
Θ: β − γ − δ = 0,
(4.5c)
H: γ + δ = 0,
(4.5d)
and

L: −γ − 3ε = 0.
(4.5e)
Their solution, α = − 2/3, β = 0, γ = 1, δ = − 1, and ε = − 1/3, produces

(4.6)

where C is a dimensionless number.

Note that the difference ΔT between the turkey’s initial temperature
and the oven temperature does not enter into this result (4.6). If we had
instead of ΔT(=Toven − To) separately included the turkey’s initial
temperature To, the oven temperature Toven, and the turkey’s desired interior
temperature Tinside among the dimensional variables and constants, our
analysis would have yielded

(4.7)

where g(x, y) is an undetermined function of two arguments.

Even so, (4.7) leaves out other possible variables – in particular those
that determine shape. After all, a pancake-shaped “turkey,” with relatively
large surface area per volume, should cook more quickly than a spherical
one. But suppose all turkeys are similarly shaped, their initial temperatures
are identical, all ovens are set identically, and each turkey is cooked until its
interior reaches that at which bacteria are killed (74 Celsius or 165
Fahrenheit). Then scaling (4.6), Δt = C ⋅ m2/3ρ1/2cp/k, suffices and only one
datum is needed to determine the constant C.
It takes me about 4.25 hours to cook a 9.0-kilogram (20 pound) turkey
in a 163 Celsius (325 Fahrenheit) oven. Since different turkeys have
approximately the same thermal conductivity, specific heat, and mass
density, these numbers produce the handy formula
 (4.8)

While the proportionality constant in (4.8), based as it is on my experience,

may not suit everyone, the scaling Δt ∝ m2/3 should apply to all cooks and
to all turkeys.
The United States Department of Agriculture (USDA) publishes a
table of cooking time versus turkey mass that seems to incorporate this
scaling. Figure 4.1 shows the USDA data (translated into SI units) with
error bars representing ranges given, for example, in the instruction “cook a
5.5–6.6 kilogram turkey for 3–3.75 hours.” I have plotted Δt = 1.0 ⋅ m2/3 on
a graph of the USDA data where m is in kilograms and Δt is in hours.

Figure 4.1. Data: USDA turkey cooking time recommendations. Solid

line: Δt = 1.0 ⋅ m2/3.
 4.5 Diffusion
Recall that a temperature gradient ∂T/∂x generates a heat flux q, that is, a
rate of heat transfer per unit facing area, determined by (4.3), according to
which

(4.9)

where k is the thermal conductivity of the material through which the heat
conducts. However, (4.9) tells only half the story of heat conduction. For a
non-uniform heat flux q, that is a non-vanishing ∂q/∂x, inevitably leads to
regions into which more heat enters than leaves. Furthermore, the
temperature T of a region into which more heat enters than leaves increases
at a rate ∂T/∂t. Thus, we have

(4.10)

where the negative sign indicates that if more heat enters a region than
leaves, then ∂q/∂x < 0 and the temperature increases, that is, ∂T/∂t > 0.
Combining (4.9) and (4.10) we find, providing the thermal conductivity k is
uniform in space, that

(4.11a)

or, equivalently,
 (4.11b)

where

(4.12)

DT is called the thermal diffusivity. The mechanism by which heat conducts

through an object with constant thermal diffusivity DT is called diffusion.
Note that [DT] = T−1L2.
The diffusion equation (4.11) is one of the basic partial differential
equations of applied physics. Its single parameter DT, the thermal
diffusivity, determines the rate at which heat diffuses or conducts. When we
are confident that a process is pure conduction, as in “Cooking a Turkey,”
we may construct our dimensional model with the thermal diffusivity DT
rather than with its separate components k, ρ, and cp.
 4.6 Making the Equations Dimensionless
We can also extract the scaling Δt = C ⋅ m2/3ρ1/3cp/k that describes thermal
diffusion directly from the diffusion equation (4.11b). Accordingly, we
recast (4.11b), by multiplying and dividing equals by equals, into a form

(4.13)

where To, Δt, and l2 are dimensional constants that characterize the initial
temperature To and the time Δt required for a temperature T to penetrate a
distance l. Therefore, the dimensionless product DTΔt/l2, which appears in
the dimensionless equation (4.13), is equal to a dimensionless number C.
Thus, given Δt = C ⋅ l2/DT and DT = k/ρcp, we find that Δt = C ⋅ l2ρcp/k.
Also since l2 ∝ V2/3 and V = m/ρ we find that Δt = C′ ⋅ m2/3ρ1/3cp/k – which
reproduces (4.6).
Making the equations that govern a process or define a state
dimensionless and then extracting the dimensionless product or products
that parameterize these equations is a distinct method of dimensional
analysis. According to this method we should list all the relevant equations
and identify the dimensionless products entering into them. If there is only
one dimensionless product, that product is equal to a dimensionless number.
If there is more than one, the products are related to each other by an
unknown function.
The process of making the equations dimensionless takes the place of
the Rayleigh algorithm. Let’s call this procedure making the equations
dimensionless rather than the unattractive, if more common, non-
dimensionalization. Of course, making the equations dimensionless implies
that we know what those equations are. And sometimes we do not.
 4.7 Example: Growth of Arctic Ice
For the purpose of being the first to reach the North Pole, the Norwegian
explorer Fridtjof Nansen (1861–1930) allowed his ship, the Fram (in
Norwegian “Forward”), to become frozen in the ice of the Laptev Sea (north of
Siberia). Nansen believed that the natural drift of polar ice would carry the
Fram across the North Pole. As it happened, Nansen made it only to 86 degrees
north latitude (a record), but he and his men survived. Nansen later wrote an
excellent account of his journey entitled Farthest North.
One of Nansen’s occupations that first winter of his expedition (1884–
1885) was to measure the thickness of the ice that surrounded the Fram. He
noted that,

From measurements that were constantly being made, it appeared that the
ice which was formed in the autumn in October or November continued to
increase in size during the whole of the winter and out into the spring, but
more slowly the thicker it became.

[18]

That the ice grew “more slowly the thicker it became” is confirmed by
dimensional analysis. Figure 4.2 illustrates the geometry.

Figure 4.2. Layers of sub-freezing air, ice, and freezing water.

 The water below the layer of ice is close to the freezing temperature of
seawater, −2.0 C (=To), while the temperature T(<To) of the air above the layer
of ice is, during the Arctic winter, sub-freezing. Under these conditions the ice
thickness λ grows in time t. As the liquid water just under the layer of ice
freezes, the heat expelled from that liquid conducts through the ice up to the
cooler air. Therefore, this process, although one of heat transfer, is not one of
pure diffusion but also involves the fusion of liquid water into ice. The heat of
fusion of water h, the thermal conductivity of ice k, the mass density of ice ρ,
and the mass-specific heat capacity of ice at constant pressure cp are all
important. These symbols, their descriptions, and their dimensional formulae
are collected in Table 4.2.

Table 4.2
λ Ice thickness L

t Time T

ΔT (=To − T) Temperature difference Θ

k Thermal conductivity HT−1L−1Θ−1

h Heat of fusion HM−1

ρ Mass density of water ML−3

cp Mass specific heat of ice HΘ−1M−1

Since there are 7 dimensional variables and constants and 5 imposed

dimensions, the analysis should produce 2 dimensionless products of the form
. Thus,
 (4.14)
implies that

L: 1 − γ − 3ε = 0,
(4.15a)
T: α − γ = 0,
(4.15b)
Θ: β − γ − ϕ = 0,
(4.15c)
H: γ + δ + ϕ = 0,
(4.15d)
and

M: −δ + ε − ϕ = 0.
(4.15e)
Constraints (4.15) are solved by α = γ = − 1/2, δ = − β,
ε = 1/2, and ϕ = β + 1/2. Therefore, the dimensionless products are
λ(hρ/tΔTk)1/2 and h/cpΔT, and the ice thickness λ is described by

(4.16)

where f(x) is an undetermined function. Since the ice thickness λ increases as

t1/2, the sea ice surrounding the Fram, indeed, grew “more slowly the thicker it
became.”
If we had replaced the imposed dimension heat H with its energy
equivalent ML2T−2, an extraneous dimensionless product th2ρ/kΔT would have
been produced – one that represents what we do not intend our model to allow:
the dissipation of work into or the production of work out of internal energy.
Josef Stefan (1835–1893) was the first in 1891 to construct and solve a
complete model of the ice-thickening process [19]. Since Stefan’s time a
mathematical problem in which a solution is required in a region whose
boundary moves, as does the lower boundary of the ice in Figure 4.2, is called
a Stefan problem.
Stefan’s analysis was probably motivated by news generated from earlier
Arctic explorations including that of the ill-fated Franklin expedition of 1845
in which all perished searching for the Northwest Passage, and the disastrous
USS Jeanette expedition of 1879–1882 whose captain, George Washington De
Long, and many of his crew died while searching for the mythical “open polar
sea.” Nansen’s relatively successful expedition occupied three years, 1893–
1896. Later (1910–1912), the Fram carried Roald Amundsen to Antarctica for
his trek to the South Pole. To this day the Norwegian people lovingly preserve
the Fram in an Oslo museum.
 4.8 Example: Stack Effect
I once worked in a fine old building constructed in the 1920s. Its design
included small passageways or chimneys that connected its upper floors to
roof vents. By encouraging a natural flow of air – as illustrated in Figure
4.3 – these chimneys kept the upper floors of the building from becoming
too hot in the summer.

Figure 4.3. Natural convection through a chimney.

Air of density ρ and mass-specific heat capacity cp enters the room

through an open window and flows out through a chimney of height h. If
the inside temperature TH becomes higher than the outside temperature
TC(<TH), a buoyant force lifts the air up the chimney at speed υ – a
tendency called the stack effect. How does υ depend on the dimensional
variables and constants whose dimensional formulae are collected in Table
4.3? Note that we do not include the heat H among the imposed dimensions.
After all, the buoyant air does work on its environment as it expands and
ascends the chimney.
Table 4.3
υ Speed of air flow LT−1

g Acceleration of gravity LT−2

h Height of chimney L

cp Mass specific heat L2T−2Θ−1

TH − TC Temperature increment Θ

TH Inside temperature Θ

ρ Mass density ML−3

From this list we may immediately eliminate the mass density ρ of the
air since its dimensional formula ML−3 cannot be combined with the others
in a dimensionless product. This leaves 6 dimensional variables and
constants, υ, g, h, cp, TH − TC, and TH, and 3 imposed dimensions, L, T, and
Θ. Therefore, the analysis should produce 3 dimensionless products. An
inspection of the table reveals these to be , cp(TH − TC)/gh, and
(TH − TC)/TH. Therefore, the speed υ of the airflow may be expressed as

(4.17)

Dimensional analysis takes us no further.

Note, however, that the quantities gh and cp(TH − TC) are energies per
unit mass and the ratio (TH − TC)/TH, that is, 1 − TC/TH, is the Carnot
efficiency of an ideal heat engine operating between the two temperatures
TH and TC. The stack effect, essentially, sets up a heat engine that generates
work from a temperature difference. Because an engine whose Carnot
efficiency vanishes produces no work, we know that f(0, 0) = 0. For this
reason, the leading order term in an asymptotic expansion of f(x, y) in terms
of small x and y is given by f(x, y) = C ⋅ xnym where n and m do not both
vanish. Another kind of analysis shows that n = 0 and m = 1/2 in which case
(4.17) becomes

(4.18)

where C is an undetermined dimensionless number.

 Essential Ideas
The dimensional variable temperature T and the two dimensions Θ,
denominating the dimension temperature, and H, denominating the imposed
dimension heat, are introduced. Heat is “that energy transferred when no
work is done.” When no work is done on or by the system, as in the
examples “Cooking a turkey” and “Growth of Arctic ice,” the imposed
dimension heat H increases the number of dimensions, decreases the
number of dimensionless products generated, and produces a more
informed result. When work is done, as in the example “Stack effect,” the
dimension H must be replaced by its energy equivalent ML2T−2.
 Problems
4.1 Ideal gas. The ideal gas pressure equation of state can be expressed as
pV = nRT where p is the gas pressure, n is the number of moles in volume
V, R is the “ideal gas constant,” and T is the absolute temperature. Find the
dimensional formula of R, that is, find [R].
answer
4.2 Pure conduction. A periodic change of temperature of frequency ω is
impressed on the face of a half-infinite solid of diffusivity DT. A wave of
temperature variation conducts into the half-infinite solid. Find (a) how the
wavelength λ of this variation and (b) its speed of propagation υ depend on
ω and DR. [20]
answer
4.3 Conduction and convection. A fluid of speed υ flowing around a hot
object conducts and convects heat away from the object. The temperature
difference between the hot object and the cooler fluid is ΔT. The size of the
object is l where [l] = L, the mass-specific heat capacity of the fluid is cp,
the mass density of the fluid is ρ, and the thermal conductivity of the fluid is
k. Determine how the heat flux q depends on these dimensional variables
and constants, that is, on υ, ΔT, l, cp, ρ, and k. Since this is a heat transfer
problem, the heat H is an effective dimension along with M, L, T, and Θ.
[21]
answer
 5
Electrodynamics and Plasma
Physics
◈
 5.1 Maxwell’s Equations
An electric charge density ρ generates an electric field while moving
charges compose a current density that generates a magnetic field .
James Clerk Maxwell (1831–1879) devised equations that relate these
fields, and , to their sources, ρ and . These are, in differential vector
form,

(5.1a)

(5.1b)

(5.1c)

and

(5.1d)

The dimensional constant εo is known as the vacuum permittivity or, more

prosaically, the electric constant, and μo as the vacuum permeability or the
magnetic constant. Once created, electric and magnetic fields exert a
force on a charge q as described by the Lorentz force law

(5.2)

where is the velocity of the charge. Equations (5.1) and (5.2) govern all
electrodynamic phenomena.a
 Maxwell first composed his eponymous equations (5.1) in 1865 from
contributions made during the previous 80 years. He was especially
indebted to Michael Faraday (1791–1867) and urged those who would
understand his own work first to read Faraday’s 1,100-page Experimental
Researches in Electricity. Faraday pioneered the concept of field as opposed
to the action at a distance view that prevailed in his time. According to
action at a distance, charged particles exert forces directly on each other.
Faraday believed, rather, that fields are real objects that mediate the
interaction of charged particles. Maxwell’s effort to mathematize Faraday’s
pictorial understanding of electric and magnetic fields led him to formulate
equations (5.1). The second term on the right-hand side of the Ampere-
Maxwell law (5.1d) introduces Maxwell’s own contribution: the
displacement current term .
Maxwell showed that the Ampere-Maxwell law (5.1d) and Faraday’s
law (5.1b) together allow for self-supporting electromagnetic field
structures that detach from their sources and propagate through space at the
speed of light. In this way Maxwell predicted that light waves are part of
the spectrum of electromagnetic waves – a prediction Heinrich Hertz
(1857–1897) verified experimentally, if accidentally, in 1886.
Maxwell’s equations (5.1) require one new dimension, the electric
charge, whose SI unit is the coulomb and whose symbol is Q. Therefore, the
dimension of charge density ρ is QL−3 and of current density J is QT−1L−2.
From [ρ] = QL−3, [J] = QT−1L−2, and (5.2) we find that

[E] = MLT−2Q−1,
(5.3a)
and
 [B] = MT−1Q−1.
(5.3b)
From Gauss’s law (5.1a) and from (5.3a) we find that

[εo] = M−1L−3T2Q2
(5.4a)
and from the Ampere-Maxwell law (5.1d) and from (5.3b)
we find that

[μo] = MLQ−2.
(5.4b)
Note that since [μoεo] = L−2T2, the dimensional constant
has the dimensional formula of speed LT−1. Furthermore, the
magnitude of , to three places 3.00 ⋅ 108 ⋅ m/s, is the speed of light
in vacuo. We will also make use of the potential difference or voltage V
defined by . Therefore, given (5.3a), [V] = ML2T−2Q−1 .
A useful approximation of Maxwell’s equations is one in which both
Maxwell’s displacement current term in (5.1d) and the right-hand side of
Faraday’s law (5.1b) are unimportant. In this approximation, electric and
magnetic fields decouple and each is generated separately from its own kind
of source, electric fields from charges and magnetic fields from currents.
This decoupling creates the separate realms of electrostatics and
magnetostatics. The vacuum permittivity εo enters into the first and the
vacuum permeability μo into the second.
 5.2 Example: Oscillations of a Compass
Needle
By the end of the eighteenth century, small magnetized pieces of iron had
been used for six or seven centuries as compass needles to help guide
seafarers. Also by the end of the eighteenth century, the direction of
magnetic north relative to celestial north (magnetic declination) and the
deflection of a compass needle in the vertical direction (magnetic dip) had
been mapped over much of the Atlantic. Magnetic declination and dip
together describe the direction of what was eventually called the
geomagnetic field. But how can the magnitude of this field be measured?
In 1776, Jean Charles Borda (1733–1799), a French naval engineer
who would shortly be commanding French warships aiding the American
revolutionaries, had a clever idea. Just as the oscillation frequency ω of a
pendulum is directly proportional to the square root of the magnitude g of
the local acceleration of gravity so that , the oscillation frequency
ω of a compass needle should be proportional to the square root of the
magnitude B of the local geomagnetic field so that . By measuring
the oscillation frequency ω of a compass needle Borda hoped to measure B.
[22]
Because [ω] = T−1 and [B] = MT−1Q−1 the frequency ω must depend
on more than the magnitude of B. Certainly ω should also depend on how
strongly the compass needle responds to a magnetic field and on the
compass needle’s rotational inertia. The magnetic dipole moment μ of the
compass needle (whose symbol should not be confused with that of the
vacuum permeability μo) parameterizes the first and the needle’s moment of
inertia I the second. These symbols, their descriptions, and their
dimensional formulae are found in Table 5.1.

Table 5.1
ω Frequency T−1

B Magnetic field strength MT−1Q−1

μ Magnetic dipole QT−1L2

moment

I Moment of inertia ML2

Since there are 4 dimensional variables and constants and 4 imposed

dimensions, these should not, according to the rule of thumb NP = NV − ND,
form a dimensionless product – unless the number of effective dimensions
is less than 4. In the latter case the dimensionless product will have the form
ωBεμβIγ where the exponents make

(5.5)

equal to 1. Therefore,

T: −1 − α − β = 0,
(5.6a)
M: α + γ = 0,
(5.6b)
Q: −α + β = 0,
(5.6c)
and
 L: 2β + 2γ = 0.
(5.6d)
Equations (5.6) are solved by α = − 1/2, β = − 1/2, and
γ = 1/2. In this way, we find that

(5.7)

where C is an undetermined dimensionless number.

Note that only two of the three equations (5.6b), (5.6c), and (5.6d) are
linearly independent – a sure sign that only 3 of the 4 imposed dimensions
are effective ones. A brief inspection of the table reveals that the three
dimensions M, L, Q appear only in two combinations: ML2 and QL2T−1.
These latter are, respectively, dimensional formulae of the moment of
inertia and the magnetic dipole moment. Therefore, there are only 3
effective dimensions, T, ML2, and QL2T−1, rather than 4, M, T, L, and Q.
Scaling (5.7) confirms Borda’s speculation that . In principle,
Borda could have compared the magnitude B of the magnetic field at
different places by measuring the oscillation frequency ω at different places
with the same compass or with identically constructed compasses. Indeed,
the French government commissioned a two-ship expedition in 1785, under
the direction of the Comte de La Perouse, in part to measure the magnetic
declination, dip, and magnitude at various places in the Pacific.
Unfortunately, in 1788, the expedition’s ships were wrecked on coral reefs
off the coast of Vanikoro, one of the Solomon Islands. Evidence suggests
that some of La Perouse’s sailors survived for some years, even for some
decades. A French search and rescue mission sighted Vanikoro in 1793 but,
not realizing the expedition had foundered there, failed to land. In 1827, an
Irish captain, Peter Dillon, commanding a British vessel, landed on
Vanikoro where he bought “a ship’s bell, a plank with a fleur-de-lys and
guns with the maker’s identification mark still visible” – all items of
European manufacture, the remains of La Perouse’s ships – but found no
survivors. [23]
 5.3 Example: Radiation from an
Accelerating Charge
A charged particle moving with constant velocity is attended by electric and
magnetic fields that move along with it. These fields do not detach from
and propagate away from their source, that is, they do not compose
electromagnetic waves. Only accelerating charges radiate electromagnetic
waves that carry energy, momentum, and, when designed to do so,
information away from their source. Cell phone and radio transmission
towers, for instance, support antennas along whose arms electrons move
back and forth. Because these electrons continually change their speed and
direction of motion, they continually accelerate. And because they
continually accelerate, they continually radiate electromagnetic waves.
Consider a charged particle with constant acceleration a, say, one
moving with constant speed υ in a circle of radius r so that its centripetal
acceleration a = υ2/r. At what rate P does this charged particle radiate
energy? Since radiation is an electromagnetic phenomenon, the power
radiated P must depend on both of the dimensional constants that enter into
Maxwell’s equations, εo and μo, as well as on the charge q and its
acceleration a. These dimensional variables and constants, their
descriptions, and their dimensional formulae are collected in Table 5.2.

Table 5.2
P Power radiated ML2T−3

q Charge Q
 a Acceleration LT−2

εo Vacuum permittivity M−1L−3T2Q2

μo Vacuum permeability MLQ−2

Since there are 5 dimensional variables and constants and 4 imposed

dimensions, we expect 1 dimensionless product of the form . We
choose the exponents in order to make

(5.8)
equal to 1. Therefore,

M: 1 − γ + δ = 0,
(5.9a)
L: 2 + β − 3γ + δ = 0,
(5.9b)
T: −3 − 2β + 2γ = 0,
(5.9c)
and

Q: α + 2γ − 2δ = 0.
(5.9d)
The solution of (5.9) is: α = − 2, β = − 2, γ = − 1/2, and
δ = − 3/2. Therefore, is dimensionless and so

(5.10)
where C is an undetermined dimensionless number. Note that in (5.10) we
have replaced μo with c via . J. J. Larmor first derived (5.10)
with C = (6π)−1 in 1897.
Larmor’s result indicates that the planetary model of the hydrogen
atom, for a short time in vogue after Rutherford’s 1911 discovery of the
atomic nucleus, is untenable. For the electron, in this case, would
continually radiate energy and, therefore, continually descend into the
nucleus. Niels Bohr was aware of this problem when, in 1913, he devised
quantum conditions that not only saved the stability of the hydrogen atom
but also predicted its emission and absorption spectra.
 5.4 Example: Child’s Law
The terms anode and cathode denote two parts of a vacuum tube, the simplest
geometry of which is illustrated in Figure 5.1. Two parallel, metallic plates
with surface area A are separated by a distance s . One of the metallic
plates is charged negative (the cathode) and the other positive (the anode) with
a potential difference ΔV between them. A flow of electrons of current density
J is emitted from the cathode and accelerated toward the anode.

Figure 5.1. Parallel-plate geometry of an electron tube.

C. D. Child (1868–1933) discovered, in 1911, a relationship, now bearing

his name, that determines the maximum current density Jmax an electron tube
with this geometry may produce. Because like charges repel, the larger the
current density J, the less accelerating force on an electron just emitted from
the cathode. Indeed, if J is large enough, an electron just emitted from the
cathode will be so repelled by the electrons in the space between the two
plates, it will experience no acceleration. This maximum current density Jmax
is called the space charge limited current density.
The space charge limited current density Jmax is a function of the plate
separation s, the potential difference ΔV, the vacuum permittivity εo, the
electron charge e, and the electron mass me. These symbols, their descriptions,
and their dimensional formulae are collected in Table 5.3.

Table 5.3
Jmax Maximum current QT−1L−2
density

s Plate separation L

ΔV Potential difference ML2T−2Q−1

εo Vacuum permittivity Q2L−3M−1T−2

e Electron charge Q

me Electron mass M

The dimensionless products produced by these dimensional variables and

constants assume a form where the exponents are
constrained by the requirement that

(5.11)
be equal to 1. Thus,
 C: 1 − β + 2γ + δ,
(5.12a)
T: −1 − 2β − 2γ = 0,
(5.12b)
L: −2 + α + 2β − 3γ = 0,
(5.12c)
and

M: β − γ + ε = 0.
(5.12d)
These constraints are solved by β = − 7/2 + α, γ = − 3 + α,
δ = 3/2 − α, and ε = 1/2. The two dimensionless products produced,
and sΔVεo/e, are related by

(5.13)

where f(x) is an undetermined function. This is as far as dimensional analysis

takes us.
However, if one could be assured, as is here the case, that the electron
charge e and mass me enter into (5.13) only in the charge to mass ratio e/me,
then it must be that f(x) = C ⋅ x2 and so

(5.14)

A more detailed analysis reveals that this is so and that . The

cloud of electrons in front of the cathode that maintains the space charge
limited current is called a virtual cathode.
 5.5 Plasma Physics
A plasma is a collection of disassociated electrons and positive ions.
Powerful electrostatic forces usually keep plasmas globally charge neutral.
Even so, plasmas can support waves that create local regions of net charge
and current density and, therefore, local regions of non-vanishing electric
and magnetic fields. Most natural and artificially produced plasmas are
gases that have been to some degree ionized. Yet the free electrons and
immobile ions within a metallic conductor and the charges within an
electrolyte can also be considered plasmas.
Plasmas are produced naturally on the surface of the earth in lightning
strikes and in the upper atmosphere or ionosphere by ionizing solar
radiation. The arc of a welder’s torch is plasma, as is the hottest part of a
flame. Plasmas are used to etch integrated circuits. While our nearby
environment is largely plasma-free, most of the matter of the universe is hot
enough and tenuous enough to maintain significant ionization and is,
therefore, in the plasma state. The sun, for instance, is plasma, as is most of
the interstellar and intergalactic gas.
 5.6 Example: Plasma Oscillations
Consider a plasma composed of electrons of density ne and positive ions of
density ni and charge state Z such that Zni = ne. If the plasma electrons are
shifted from a position of equilibrium, they will be pulled back and
accelerate toward their equilibrium position, overshoot that position
because of their inertia, and continue to oscillate at a frequency ωp called
the plasma frequency. Since the electrons are much less massive than the
ions, most of this oscillatory motion will be in the electrons.
How does the plasma frequency ωp depend on the electron density ne,
the electron charge e, the electron mass me, and the vacuum permittivity εo?
These 5 dimensional variables and constants are denominated in 4 imposed
dimensions M, L, T, and Q. Therefore, according to the rule of thumb
NP = NV − ND, the analysis should produce 1 dimensionless product. These
symbols, their descriptions, and their dimensional formulae are collected in
Table 5.4.

Table 5.4
ωp Plasma frequency T−1

ne Electron density L−3

e Electron charge Q

me Electron mass M

εo Vacuum permittivity M−1L−3T2Q2

 In this case we can easily construct the dimensionless product without
employing the Rayleigh algorithm. One starts with εo and divides or
multiplies by powers of the other quantities in each case so that the division
or multiplication eliminates one of the imposed dimensions M, L, T, or Q
from the dimensional formula of the product. The single dimensionless
product produced in this way is . Thus

(5.15)

where C is a dimensionless number. As one might expect, the larger the

electron mass me, the smaller the oscillation frequency ωp. Also, the larger
the restoring force as characterized by the parameter nee2/εo, the larger the
oscillation frequency ωp. According to a more detailed analysis, C = 1.
 5.7 Example: Pinch Effect
The electrostatic attraction of unlike charges and the electrostatic repulsion
of like charges are not the only kinds of electromagnetic interaction.
Whenever like charges, say electrons, all move or drift in the same
direction, they compose a current whose different streams attract each other.
Oppositely directed streams of current repel each other. These interactions,
mediated as they are by a magnetic field, are normally much weaker than
electrostatic attraction and repulsion. But when a dense cloud of electrons
drifts through an equally dense cloud or array of relatively stationary ions,
the electrostatic forces are neutralized and leave the magnetic force
dominant. The resulting attraction of different streams of current is called
the pinch effect.
When generated by the high current of a lightning bolt, the pinch effect
may be so large as to implode a rigid metallic lightning rod carrying that
current. In 1905, two Australian engineers, J. A. Pollock and S.
Barraclough, examined the crushed and distorted remains of a lightning rod,
originally a copper tube that had been erected above a refinery in the
province of New South Wales (Australia). Their conclusion was that the
copper tube had, indeed, been struck by lightning and imploded by the
pinch effect.
In mid-twentieth century the pinch effect was thought capable of
producing plasma hot enough and dense enough to support the nuclear
fusion of light nuclei. The idea was that the initial part of a current pulse
directed along a long straight wire would vaporize and ionize the material
of the wire and the gas around it while the pinch effect associated with the
remaining part of the pulse would compress and heat the plasma thus
produced. Eventually, the outwardly directed pressure exerted by the
plasma would balance the inwardly directed pinch force. The relation
between the plasma current I, the plasma radius R, the plasma density per
unit length NL, and the plasma absolute temperature T when these forces are
balanced can be found with dimensional analysis. Figure 5.2 illustrates the
“z-pinch” geometry of this cylindrical plasma.

Figure 5.2. Z-Pinch Geometry.

Because this is not a heat transfer problem but rather one in which the
magnetic field compresses and heats the plasma, the dimension H is not
independent of M, L, and T. Boltzmann’s constant kB and the vacuum
permittivity μo should be included among the dimensional variables and
constants. Furthermore, since we are concerned only with the fully pinched
and balanced configuration, we need include neither a compression time nor
the inertia of the plasma particles. (But see Problem 5.5.) Table 5.5 collects
these variables and constants, their descriptions, and their dimensional
formulae.

Table 5.5
I Current QT−1

R Radius L
 NL Density per unit length L−1

T Temperature Θ

kB Boltzmann’s constant ML2T−2Θ−1

μo Vacuum permittivity MLQ−2

The table clearly indicates that Boltzmann’s constant kB must multiply

the absolute temperature T – otherwise the dimension of temperature Θ
would not be eliminated from the dimensionless product. For this reason we
save ourselves a little effort by including these two dimensional variables
and constants only in the product kBT. Then we have 5 dimensionless
variables and constants I, R, NL, kBT, and μo denominated in 4 imposed
dimensions M, L, T, and Q.
The dimensionless products are of the form where the
exponents make this form dimensionless. Thus,

(5.16)
implies that

Q: 1 − 2δ = 0,
(5.17a)
T: −1 − 2γ = 0,
(5.17b)
L: α − β + 2γ + δ = 0,
(5.17c)
and
 M: γ + δ = 0.
(5.17d)
The solution of (5.17) is β = α − 1/2, γ = − 1/2, and
δ = 1/2. Note that only two of the three equations (5.17a), (5.17b), and
(5.17d) are linearly independent. Evidently, only 3 dimensions are effective.
An inspection of the table reveals these to be QT−1, MT−2, and L or
alternatively and equivalently QT−1, MLT−2, and L. These latter identify the
total current QT−1 and the force MLT−2, that is, F, as effective dimensions.
With enough foresight we could we have adopted these as imposed
dimensions.
The dimensionless products assume the forms and
RNL. Therefore,

(5.18)

where f(x) is an undetermined function. This is as far as dimensional

analysis takes us. In most applications RNL > > 1. A more detailed analysis
indicates that as RNL → ∞, f(RNL) → C where C = 8π.
 Essential Ideas
Electrodynamics introduces three new dimensional fundamental constants:
the vacuum permittivity εo, the vacuum permeability μo, and the speed of
light c. Because εo, μo, and c are related by , no more than two
of these three are needed in any one description. One new dimension, the
charge Q, is also introduced. To model an electrostatic state or process, one
needs εo, and to model a magnetostatic state or process, one needs μo.
Electromagnetic processes in which electric and magnetic fields are coupled
require both εo and μo.
 Problems
5.1 Electret oscillations. An electret is a material, possibly quartz or
polymer, that can, for an extended period, hold a charge separation and in
this way create a semi-permanent electric dipole moment p for which
[p] = QL. Imagine a needle-like electret with moment of inertia I mounted
so that it can freely rotate and align itself with an applied electric field of
magnitude E. Use dimensional analysis to determine its frequency ω of
oscillation around its equilibrium position. See Section 5.2.
answer
5.2 Fields. A particle with charge q and constant velocity of magnitude υ
passes within a distance d of an observer. (a) How does the maximum
electric field strength E detected by the observer depend on these variables
and εo and μo? (b) How does the maximum magnetic field strength B
detected by the observer depend on these variables and εo and μo?
answer
5.3 Cylindrical Child’s law. The context is Section 5.4 on Child’s law.
Suppose the cathode is a long straight wire and the anode a concentric
cylinder of radius R and length L as shown in Figure 5.3. A potential
difference ΔV between the cathode and anode causes a maximum current
Imax to flow between the two. Find an expression for the maximum current
per unit length Imax/L between the cathode and anode that is parallel in
structure to (5.13) and also to (5.14).
answer
 Figure 5.3. Geometry of a wire cathode and a concentric cylinder anode.

5.4 Plasma oscillation. The context is Section 5.6. Use the Rayleigh
algorithm to show that only one dimensionless product is
produced from the dimensional variables and constants ωp, me, εo, ne, and e.

5.5 Pinch effect compression time. The context is Section 5.7. Use the
Rayleigh algorithm to determine an expression for the time Δt required for a
current I to cause a “cold,” that is, a zero-pressure, plasma of initial radius
Ro composed of ions of mass m and density per unit length NL to become
fully compressed. (Include the vacuum permeability μo among the
dimensional variables and constants.)
answer
5.6 Pinch effect. Again, the context is Section 5.7 on the pinch effect.
Reformulate the table listing the symbols, names, and dimensional formula
of 5 dimensional variables and constants, I, R, NL, kBT, and μo. In doing so
express the dimensional formulae in terms of 3 effective dimensions: a
current QT−1, a pressure per unit length ML−2T−2, and a length L.
answer

a Maxwell’s equations can be cast into a form, favored by theorists, in

terms of only one dimensional constant: the speed of light c. This is
possible because the vacuum permeability μo is, in SI units, a
conventional number 4π ⋅ 10−7 and, for this reason, can be absorbed into
the definition of the electric field and the charge density.
 6
Quantum Physics
◈
 6.1 Planck’s Constant
Max Planck’s professional goal in the period 1894–1900 was to model the
properties of electromagnetic radiation in thermal equilibrium with
blackbody materials, so called because they absorb (and emit)
electromagnetic radiation at all frequencies. Thus, the radiation in
equilibrium with a blackbody is called blackbody radiation. At first, Planck
(1858–1947) made little progress – classical descriptions of blackbody
radiation failed to reproduce even its most basic properties. And before
1900, classical physics was the whole of physics.
Planck eventually looked for a simple postulate that could
mathematically reproduce the properties of blackbody radiation even if that
postulate went beyond what was then known. What he discovered in 1900
was the bold assertion that a blackbody can absorb and emit radiation of
frequency ν only in countable bundles or quanta each with energy E
proportional to ν. Thus, E = hν where h is a constant now called Planck’s
constant. Planck’s postulate led to expressions that fit the available data
well when h assumed a value close to

h = 6.63 ⋅ 10−34 ⋅ kg ⋅ m2/s

(6.1)
where (6.1) gives, to three places, the current value of
Planck’s constant in SI units. Note that [h] = ML2T−1.
While Planck’s postulate was a great success, it took some time for
physicists to fathom its consequences. In 1905, Albert Einstein (1879–
1955) showed that Planck’s quanta, later called photons, also explained the
photoelectric effect, while Arthur Holly Compton, in 1922, showed that a
photon and an electron preserve their total momentum and energy as they
interact with one another. Eventually, it became clear that Planck had
discovered a new realm of physics – the realm of the quantum. Today we
recognize Planck’s constant h as a sure sign of quantum physics.
 6.2 Example: Blackbody Radiation
The equation of state of radiation in a cavity surrounded by blackbody
material can be explored dimensionally. Because blackbody radiation is a
quantum phenomenon, its energy density E/V depends on Planck’s constant
h as well as on the speed of light c and on the product kBT where kB is
Boltzmann’s constant and T is the absolute temperature of the radiation.
These symbols, their descriptions, and their dimensional formulae are
collected in Table 6.1.

Table 6.1
E/V Energy density ML−1T−2

h Planck’s constant ML2T−1

kBT Boltzmann’s constant ML2T−2

times absolute
temperature

c Speed of light LT−1

Because there are 4 dimensional variables and constants, E/V, h, c, and

kBT, and 3 imposed dimensions, M, L, and T, the dimensional analysis
should, according to the rule of thumb NP = NV − ND, produce 1
dimensionless product. We find this product by making (E/V)hα(kBT)βcγ
dimensionless. Thus,
 (6.2)
implies that

M: 1 + α + β = 0,
(6.3a)
L: −1 + 2α + 2β = 0,
(6.3b)
and

T: −2 − α − 2β − γ = 0.
(6.3c)
Equations (6.3) are solved by α = 3, β = − 4, and γ = 3.
Therefore, the product (E/V)(hc)3(kBT)−4 is dimensionless and

(6.4)

where C is a dimensionless number. Equation (6.4) with C = 8π5/15 is

known as the Stefan-Boltzmann law.
 6.3 Example: Spectral Energy Density of
Blackbody Radiation
The energy density E/V of blackbody radiation is composed of electromagnetic
waves of frequency ν associated with photons having energy hν whose density
is non-uniform over the spectrum of electromagnetic waves from ν = 0 to
ν = ∞. Thus, we may ask, “How does the energy density of blackbody radiation
per differential frequency, that is, V−1(dE/dν), depend upon frequency ν?” This
quantity V−1(dE/dν), known as the spectral energy density of blackbody
radiation, depends on Planck’s constant h, on the normalized absolute
temperature kBT, and on the speed of light c. These symbols, their descriptions,
and their dimensional formulae are collected in Table 6.2.

Table 6.2
(dE/dν)V−1 Spectral energy ML−1T−1
density

ν Frequency T−1

kBT Boltzmann’s ML2T−2

constant times
absolute temperature

h Planck’s constant ML2T−1

c Speed of light LT−1

Since there are 5 dimensional variables and constants and 3 imposed

dimensions, we should find 2 dimensionless products. These assume a form
(dE/dν)V−1να(kBT)βhγcδ that the exponents α, β, γ, and δ render dimensionless.
Thus

(6.5)
implies

M: 1 + β + γ = 0,
(6.6a)
L: −1 + 2β + 2γ + δ = 0,
(6.6b)
and

T: −1 − α − 2β − γ − δ = 0.
(6.6c)
These are solved by β = − 3 − α, γ = 2 + α, Δ = 3. Therefore,
(dE/dν)V−1h2c3(kBT)−3 and hν/kBT are dimensionless and so

(6.7)

where the function f(x) is undetermined by dimensional analysis.

However, the form of the undetermined function f(x) is constrained by the
requirement that the integral of the spectral energy density V−1(dE/dν) over all
frequencies is the average energy density E/V. Symbolically,

(6.8)

and so given (6.7),

 (6.9)

Shifting the integration variable from ν to x where x = hν/kBT transforms (6.9)

into

(6.10)

Given the equation of state (6.4) equation (6.10) becomes

(6.11)

This result relates the function f(x) and the value of the number C, which
otherwise remain undetermined. A more detailed analysis finds that
f(x) = x3/(ex − 1) and so, given (6.11), C = 8π5/15.
 6.4 Example: Bohr Model
Niels Bohr (1885–1962) used Planck’s constant h to forge a new
understanding of the hydrogen atom. According to the model Bohr
proposed in 1913, the hydrogen atom is composed of an electron, with mass
me and charge e, orbiting, in specially defined non-radiating orbits, a proton
with mass mp(=1836me). Since the electron and proton are bound together
electrostatically, the vacuum permittivity εo helps determine these orbits.
Because Bohr’s orbits are, by hypothesis, non-radiating and we assume the
electron speed in a hydrogen atom is non-relativistic, the speed of light c is
not relevant. Therefore, the dimensional constants me, e, εo, and h complete
the list of those that describe Bohr’s hydrogen atom.
Bohr’s model determines, among other quantities, a ground state
orbital radius r1 that characterizes the hydrogen atom’s spatial extent.
Dimensional analysis determines how r1 depends on me, e, εo, and h. Table
6.3 collects these symbols, their descriptions, and their dimensional
formulae.

Table 6.3
r1 Bohr radius L

me Electron mass M

e Electron charge Q

εo Vacuum permittivity M−1L−3T2Q2

h Planck’s constant ML2T−1

 Since this table contains 5 fundamental, dimensional constants
denominated in 4 imposed dimensions, there should be only 1
dimensionless product having the form . The exponents α, β, γ,
and δ that determine this product are those that make
dimensionless. Therefore,

(6.12)

implies that

L: 1 − 3γ + 2δ = 0,
(6.13a)
M: α − γ + δ = 0,
(6.13b)
Q: β + 2γ = 0,
(6.13c)
and

T: 2γ − δ = 0.
(6.13d)
The solutions to equations (6.13) are α = 1, β = 2, γ = − 1,
and δ = − 2. Therefore, r1mee2/(εoh2) is dimensionless and

(6.14)

where C is a dimensionless number. Bohr’s own, simple analysis

determines that C = 1/π. The quantity r1 with C = 1/π in (6.14) is called the
Bohr radius.
 The above dimensional analysis is often done in a slightly different
way, in which the vacuum permittivity εo is replaced with 4πεo and Planck’s
constant h with its “reduced value” ℏ(=h/2π). Then the dimensionless
product produced is r1mee2/(4πεoℏ2) and the resulting scaling is

(6.15)

with C′ = 1. While this result suggests that we should use 4πεo in place of εo
and use ℏ in place of h, we have no reason to believe that this tactic will
always produce proportionality constants equal to 1.
 6.5 Atomic Units
The dimensional constants me, e, εo, and h enter into all models in which
mechanics, electrostatics, and quantum physics play a role. Because atomic
physics draws on these subjects, εoh2/e2me (proportional to the Bohr radius)
characterizes lengths in atomic physics. Similarly, the time (see
Problem 6.2) and, of course, the mass me and charge e similarly
characterize atomic physics.
The rule of thumb NP = NV − ND suggests and the Rayleigh algorithm
confirms that only one length can be fashioned out of the 4 dimensional
constants, me, e, εo, and h, whose dimensional formulae are expressed in
terms of the 4 imposed dimensions, M, L, T, and Q. Similarly, only one time
follows from these same 4 dimensional constants. In all, the 4 dimensional
fundamental constants, me, e, εo, and h, produce 4 unique fundamental,
characteristic quantities: a length εoh2/e2me, a time , a mass me,
and a charge e. These compose what we call a characterizing system of
units or a scale – a scale of atomic-sized units. The elements of this scale,
their formulae, and their values to three places in SI units are listed in Table
6.4.

Table 6.4
atomic scale

Mass me 9.11 ⋅ 10−31 kg Electron mass

 atomic scale

Length 1.67 ⋅ 10−10 m Bohr radius

Time 3.82 ⋅ 10−17 s Period of

electron
orbiting at
Bohr radius

Charge e 1.60 ⋅ 10−19 C Electron

charge

Because atomic physicists exert much effort calculating with me, e, εo,
and h, they sometimes take the shortcut of setting each of these to unity,
that is, so that me = 1, e = 1, εo = 1, and h = 1. The effect of this shortcut is
that all masses produced in this way are in units of me. For instance,
mp = 1836 means mp = 1836me. Furthermore, all lengths are in units of
εoh2/e2me, all times are in units of , and all charges are in units of
e.
 6.6 Example: Quantum Ideal Gas
A gas is ideal if its particles do not interact with one another at a distance or
through a field. We already know that the equation of state of a classical
ideal gas p = (N/V)kBT relates its pressure p, its particle density N/V, and its
absolute temperature T where kB is Boltzmann’s constant. How do quantum
effects modify this equation of state?
Certainly Planck’s constant h should enter into any quantum version of
p = (N/V)kBT. But the Rayleigh algorithm applied to the 4 dimensional
variables and constants p, N/V, kBT, and h that are denominated in terms of
the 3 imposed dimensions M, L, and T generates exactly one dimensionless
product pV/(NkBT).
Possibly we have ignored some dimensional variable or constant that
effectively brings Planck’s constant h into play. Since, by definition, no
force fields play a role, the dimensional constants G, εo, μo, and e are not
involved. Because we limit our model particles to non-relativistic speeds,
neither is the speed of light c. It seems that only the mass m of the ideal gas
particles remains to be considered. In retrospect, it may surprise us that m
does not enter directly into the classical equation of state p = (N/V)kBT. (But
review Section 1.7.)
Therefore, we allow the pressure p of a quantum ideal gas to depend
on N/V, kBT, h, and m. These 5 symbols, their descriptions, and their
dimensional formulae are collected in Table 6.5.

Table 6.5
p Pressure ML−1T−2
 N/V Number density L−3

kBT Boltzmann’s constant ML2T−2

times absolute
temperature

h Planck’s constant ML2T−1

m Particle mass M

Accordingly, NV = 5 and ND = 3. Therefore, we expect, according to

the rule of thumb NP = NV − ND, that the Rayleigh algorithm will produce 2
dimensionless products, one of which we already know is pV/(NkBT). The
dimensionless products formed are those that make p(N/V)α(kBT)βhγmδ
dimensionless. Given that

(6.16)
we have

M: 1 + β + γ + δ = 0,
(6.17a)
L: −1 − 3α + 2β + 2γ = 0,
(6.17b)
and

T: −2 − 2β − γ = 0.
(6.17c)
The solution to (6.17) is β = − 5/2 − 3α/2, γ = 3 + 3α, and
δ = − 3/2 − 3α/2. Therefore, ph3/[m3/2(kBT)5/2] and (N/V)h3/(mkBT)3/2 are
the 2 dimensionless products. Multiplying the first by the inverse of the
second produces pV/(NkBT). Therefore, we may write

(6.18)

where f(x) is undetermined. This is as far as dimensional analysis takes us.

However, we do know that p = (N/V)kBT must be recovered from (6.18) in
the limit of vanishing Planck’s constant h. Therefore, as
(N/V)h3/(mkBT)3/2 → 0, f(0) → 1.
Quantum effects dominate in the high-density, low-temperature, small-
mass regime for which (N/V)h3/(mkBT)3/2 > > 1. If in this (quantum) regime
the function f(x) is replaced with the power law f(x) ≈ C ⋅ xn, then (6.18)
becomes

(6.19)

where C and n are undetermined. Our physics sense takes us a little further.
Because we expect that as T → 0, the pressure p should not diverge, it must
be that n ≤ 2/3.
In the special case n = 2/3 (6.19) becomes

(6.20)

This result (6.20) with C = (3/π)2/3/20 describes the degeneracy pressure

exerted by a gas of particles that observe Fermi-Dirac statistics according to
which no two particles in the same region can occupy the same single-
particle quantum state. Equation (6.20) also expresses the dependence of
the critical pressure at which a gas of bosons begins to transition to a
condensed state. Bosons are particles that observe Bose-Einstein statistics
according to which any number of particles may occupy one single-particle
quantum state.
Of course, dimensional analysis does not answer or even pose crucial
questions such as “Why is there a degeneracy pressure?” or “Why does a
Bose-Einstein condensate exist?” Only a more complete theory can frame,
much less answer, these questions.
 6.7 Example: Quantized Radiation from
an Accelerating Charge
According to the result of Section 5.4, the rate P at which a charge q moving
with speed υ in a circle of radius r radiates electromagnetic waves that carry
away energy is given by

(6.21)

where a = υ2/r and C is an undetermined dimensionless number. But we now

know that this electromagnetic energy is emitted in quanta or photons each
with energy hν for various frequencies ν. How does this knowledge modify
(6.21)?
Clearly, Planck’s constant h must enter into any quantum modification
of (6.21). Therefore, this quantum modification must be a relation among the
following dimensional variables and constants: the power radiated P, the
charge q, its acceleration a, the vacuum permittivity εo, the speed of light c,
and Planck’s constant h. These symbols, descriptive names, and dimensional
formulae are listed in Table 6.6.

Table 6.6
P Power radiated ML2T−3

q Charge Q

a Acceleration LT−2

εo Vacuum permittivity M−1L−3T2Q2

 c Speed of light LT−1

h Planck’s constant ML2T−1

Since there are 6 dimensional variables and constants and 4 imposed

dimensions, we expect 2 dimensionless product of the form . We
have already identified, in Section 5.4, one of these: Pεoc3/(q2a2). The other
we find in the usual way, that is, by choosing exponents α, β, γ, δ, and ε that
render the form dimensionless. Therefore,

(6.22)
implies that

M: 1 − γ + ε = 0,
(6.23a)
L: 2 + β − 3γ + δ + 2ε = 0,
(6.23b)
T: −3 − 2β + 2γ − δ − ε = 0,
(6.23c)
and

Q: α + 2γ = 0.
(6.23d)
The solution of (6.23) is: α = − 2, β = − 2, γ = − α/2,
δ = 2 − α/2, and ε = − 1 − α/2. In this way we find the two dimensionless
products to be Pc2/a2h and . By multiplying the first by the inverse
square of the second, we eliminate h and produce, as expected, Pεoc3/(q2a2).
The square of the second independent, dimensionless product is q2/εoch.
Thus, we may write
 (6.24)

where the function f(x) remains undetermined.

The dimensionless product q2/(2εohc) with the electron charge e
replacing q is called the fine structure constant – of which more will be said
in Chapter 7. The undetermined function in (6.24) is not easily discovered or
expressed. One approach to determining f(x) is to expand the semi-classical
(and thus semi-quantum) expression for P in terms of powers of the fine
structure constant e2/(2εoch)[≈0.007]. [24]
 Essential Ideas
The dimensional model of a quantized state or process must include the
dimensional constant known as Planck’s constant h.
 Problems
6.1 Ground state energy. The context is Section 6.4. Determine how the
ground state energy E1 of a hydrogen atom depends on relevant
dimensional constants.
answer
6.2 Atomic time unit. (a) Use the Rayleigh algorithm to show that only one
quantity with the dimension of time, , is fashioned out of the
fundamental constants me, e, εo, and h. (b) Show that this time is equal to
Planck’s constant divided by the energy found in Problem 6.1.

6.3 Ideal gas. Show that, as claimed in Section 6.6, the Rayleigh algorithm
applied to the dimensional variables and constants P, N/V, kBT, and h
generates one dimensionless product PV/(NkBT).
 7
Dimensional Cosmology
◈
 7.1 The Fundamental Dimensional
Constants
Each subject we have explored thus far, mechanics and the mechanics of
fluids, thermal physics, electrodynamics, and quantum physics, has
introduced one or more fundamental dimensional constants: the
gravitational constant G, Boltzmann’s constant kB, the vacuum permittivity
εo, the speed of light c, the electron charge e, the electron mass me, and
Planck’s constant h. These subjects, considered expansively, encompass
most of what we know of the physical world, and these fundamental
constants quantify that knowledge. These seven constants, their descriptive
names and symbols, their numerical values to three places in SI units, and
their dimensional formulae are collected in Table 7.1.a

Table 7.1
Name Symbol Value in SI Dimensional
units Formula

Gravitational G 6.67 ⋅ 10−11 M−1L3T−2

constant

Boltzmann’s kB 1.38 ⋅ 10−23 ML2T−2Θ−1

constant

Vacuum εo 8.85 ⋅ 10−12 M−1L−3T2Q2

permittivity

Speed of light c 3.00 ⋅ 108 LT−1

 Name Symbol Value in SI Dimensional
units Formula

Electron charge e 1.60 ⋅ 10−19 Q

Electron mass me 9.11 ⋅ 10−31 M

Planck’s h 6.63 ⋅ 10−34 ML2T−1

constant

What I find remarkable is that these constants are related to one

another through their dimensional formulae in ways that compose a
dimensional structure or cosmology. Here I use the word cosmology
broadly to denote the general structure of the universe, its elements, and its
laws. This chapter seeks to uncover that cosmology. The results, partial and
preliminary though they may be, are those that can be obtained with the
methods of dimensional analysis. Other methods may take us further, but
dimensional analysis brings us close to the boundary separating the known
from the unknown.
 7.2 Eddington-Dirac Number
We simplify our initial investigation in two ways, first by eliminating
Boltzmann’s constant kB from consideration and second by concentrating
on the dimensional structure of classical physics. A brief inspection of the
above table reveals that the dimension of temperature Θ appears only in
Boltzmann’s constant kB. Consequently, kB cannot enter into a
dimensionless product with the other constants – and finding these
dimensionless products is our first task.
Of course, Boltzmann’s constant kB has its uses. Besides converting
the dimension of temperature Θ into that of energy ML2T−2 and vice versa,
Boltzmann’s constant kB provides entropy with its dimensional formula.
After all, the statistical entropy S of an isolated system is given by
S = kBlnΩ where Ω is the number of microstates accessible to that system.
Therefore, [S] = [kB] = ML2T−2Θ−1.
Once we eliminate Boltzmann’s constant kB and the non-classical
Planck’s constant h from the list of fundamental constants, the remaining 5
classical ones, G, ε0, c, e, and me, characterize gravitational, electrostatic,
and electromagnetic interactions, and the properties of an electron. Because
these 5 constants are denominated in 4 imposed dimensions, M, L, T, and Q,
they produce among themselves, according to the rule of thumb, 1
dimensionless product. The Rayleigh algorithm easily produces this
dimensionless product: .
Two electrons separated by a distance r illumine its meaning. These
two electrons repel each other with an electrostatic force e2/(4πεor2) and
attract each other with a gravitational force . Therefore, the ratio
, or more simply , characterizes the relative
strength of these two forces.
The value of the ratio,

(7.1)

fascinated Arthur Eddington (1882–1944) and Paul Dirac (1902–1984).

Eddington noticed that 1042 is approximately the square root of the number
of atoms in the observable universe, while Dirac found that 1042 is
approximately the age of the universe in units of a classical unit of time
e2/(εomec3) ≈ 10−23s. Eddington and Dirac used these correlations as the
basis of speculations that, while interesting, have failed to bear fruit. Even
so, we honor their efforts by calling the Eddington-Dirac
ratio.
The large size of the Eddington-Dirac ratio means that
the electrostatic force is much stronger than the gravitational force. Gravity
dominates the large-scale structure of the universe only because large
numbers of positive and negative charges tend to neutralize each other.
Electrostatics dominates the small-scale structure of the universe. Together
these two forces make possible a universe with an amazing variety of
structures from atoms, to living organisms, to galaxies.
One consequence of the Eddington-Dirac ratio is the dimensional
equivalence

(7.2)
or or . We may, for instance, substitute e2/
εo for in any term and preserve that term’s dimensional formula. Of
course, in doing so we shift its numerical value by a factor of 1042.
 7.3 Example: Gravitational-Electrostatic
Mass
What quantity with the dimension of mass can be composed out of the three
fundamental constants e, εo, and G? The dimensional equivalence (7.2), that
is, , contains the answer. For if , then
, [e2/εo] = [G][me]2, and so . Therefore,
the quantity has the dimension of mass – the mass of a spherical
particle with charge e whose gravitational self-attraction is large enough to
balance its electrostatic self-repulsion. I call the gravitational-
electrostatic mass.
 7.4 Example: Classical Electron Radius
What quantity having the dimension of length can be composed out of the
fundamental constants me, e, εo, and c? Since the Eddington-Dirac ratio
does not contain the speed of light c, this question cannot be answered by
using the dimensional equivalence (7.2) based on that dimensionless
product.
Instead, denote the length sought with l. Given that the 5 dimensional
quantities, l, me, e, εo, and c, are expressed in terms of 4 imposed
dimensions, M, L, T, and Q, as shown in Table 7.2, the Raleigh algorithm
applied to these quantities should produce 1 dimensionless product.

Table 7.2
l Characteristic length L

me Electron mass M

e Electron charge Q

εo Permittivity of free M−1L−1T2Q2

space

c Speed of light LT−1

In this case, we do not need the Rayleigh algorithm to uncover the

dimensionless product. We simply multiply and divide εo by powers of the
other constants until the result becomes dimensionless. The dimensionless
product produced in this way is lmeεoc2/e2. Therefore, e2/(εomec2) has the
dimension of length. This length is, apart from a dimensionless factor, the
radius r of a spherical electron whose potential electrostatic energy, roughly
e2/(4πεor), is equal to its rest energy mec2. The length e2/(4πεomec2)
produced in this way is equal to 2.81 ⋅ 10−15 ⋅ m and is sometimes called the
classical electron radius.
 7.5 Classical Scales
From among the five classical, fundamental dimensional constants G, εo, c, e, and me we chose the
following four εo, c, e, and me to form a quantity e2/(εomec2), the classical electron length, with the
dimension of length. We can also divide this length e2/(εomec2) by the speed of light c to form a time
e2/(εomec3). The Rayleigh algorithm indicates that these are the only quantities with the dimensions of
length and time that are combinations of εo, c, e, and me.
This length and this time along with the electron mass me and charge e form a set of four
quantities with dimensions M, L, T, and Q. It is customary to call this set of quantities a scale, in this
case, the classical scale.b,c The numerical values of these quantities, to three places in SI units, are
collected in Table 7.3.

Table 7.3
classical scale

Mass me 9.11 ⋅ 10−31 kg Electron mass

Length 3.53 ⋅ 10−14 m Classical electron

radius

Time 1.18 ⋅ 10−22 s Time for light to

cross a classical
electron radius

Charge e 1.60 ⋅ 10−19 C Electron charge

Other classical masses, lengths, durations, and charges can be formed in similar fashion, that is,
by applying the Rayleigh algorithm to other sets of 4 dimensional constants taken from the 5 classical
ones: G, εo, c, e, and me. There are in all 5 ! /4 ! 1! or 5 such classical scales. These are collected in the
first five rows of Table 7.4 in Section 7.7.

Table 7.4
Constants Mass Length (meters) Time (seconds) Charge
(kilograms) (coulombs)
Constants Mass Length (meters) Time (seconds) Charge
(kilograms) (coulombs)

G , me , e , εo me ∼ 10−30 e ∼ 10−19

G , me , εo , c me ∼ 10−30

G , me , e , c me ∼ 10−30 e ∼ 10−19

G , e , εo , c e ∼ 10−19

me , e , εo , c me ∼ 10−30 e ∼ 10−19
classical
scale

G , me , e , h me ∼ 10−30 e ∼ 10−19

G , me , εo , h me ∼ 10−30

G , me , c , h me ∼ 10−30

G , e , εo , h e ∼ 10−19

me , e , εo , h me ∼ 10−30 e ∼ 10−19
atomic
scale

G,e,c,h e ∼ 10−19
me,e,c,h
Constants me ∼ 10−30
Mass Length (meters) Time (seconds) e ∼ 10−19
Charge
(kilograms) (coulombs)

G , εo , c , h
Planck scale

me , εo , c , h me ∼ 10−30

e , εo , c , h e ∼ 10−19
 7.6 Fine Structure Constant
Restoring Planck’s constant h to the classical, dimensional constants (and
again excluding Boltzmann’s constant kB) composes a set of 6 fundamental,
dimensional constants: G, εo, c, e, me, and h. Since their dimensional
formulae are expressed in terms of 4 imposed dimensions, M, L, T, and Q,
they can be combined into 2 dimensionless products. One of these, we
know, is proportional to the Eddington-Dirac ratio . The
other, which we met for the first time in Section 6.7, is e2/(εohc). The
particular form e2/(2εohc) is called the fine structure constant.
Arnold Sommerfeld (1868–1951) observed that the ratio of the
velocity of an electron in the first Bohr orbit to the speed of light is the fine
structure constant. But the fine structure constant also carries with it another
meaning. Because it is proportional to the ratio of the potential energy of
two electrons separated by a distance d, that is, e2/(4πεod), to the energy of
a photon with wavelength d, that is, hc/d, the fine structure constant
regulates the interaction of electrons and photons. Indeed, e2/(2εohc)
characterizes a generalization of Maxwell’s equations, called quantum
electrodynamics or QED, in which electrons and photons interact.
The value of the fine structure constant was first thought to be exactly
1/137. Consequently, a number of physicists, including Wolfgang Pauli
(1900–1958), sought the meaning of this apparently special integer 137. As
it happened, Pauli was unsuccessful in this effort and balefully pointed out
the number of his hospital room, #137, to a visitor shortly before his death.
We now know that the fine structure constant is only approximately 1/137.
The current value of its inverse is, to 12 places, 137.035999139.
Nevertheless, since the equation

(7.5)

is accurate to three places, the number 137 is a handy way to remember the
value of the fine structure constant.
No one has been able to derive numerical values of either the
Eddington-Dirac ratio or the fine structure constant. Instead, these
dimensionless products or their constituent dimensional constants must be
measured. Yet there are physicists who believe the task of physics will not
be complete until the value of all the fundamental dimensional constants
can be derived from some very deep principle – now unknown to us.
The fine structure constant e2/(2εohc) provides a way of generating
dimensional quantities with the dimensions of M, L, T, and Q that contain
Planck’s constant h from quantities that do not. One simply uses
dimensional equivalences like

[e2/εo] = [hc]
(7.6)
or [e2] = [εohc] or [c] = [e2/hεo] to transform classical
quantities into quantum ones that contain Planck’s constant h. See, for
example, Section 7.9.
 7.7 The 15 Scales
From the set of 6 fundamental dimensional constants, G, εo, c, e, me, and h,
4 can be taken, without regard to order, in 6 ! /4 ! 2! or 15 different ways.
Employing the Rayleigh algorithm on each of these 15 sets of 4 constants,
we derive quantities with the dimensions of M, L, T, and Q.
Of course, the Rayleigh algorithm produces a particular quantity only
when that quantity exists. And occasionally it does not. For instance,
because the dimensional formulae of the 4 constants G, me, c, and h do not
contain the dimension Q, the Rayleigh algorithm cannot generate, from
these constants, a quantity with the dimension Q. Therefore, this particular
set of 4 possible dimensional quantities remains incomplete.
Each of the 15 sets of dimensional quantities derived from 4
fundamental dimensional constants composes a scale. The 15 scales,
classical and quantum, complete and incomplete, are listed in Table 7.4.
Three of them have been given special names: the classical scale, based on
me, e, εo, and c, the atomic scale based on me, e, εo, and h, and the Planck
scale based on G, εo, c, and h. Attending each dimensional combination is
its value in SI units to the nearest order of magnitude.d,e,f
 7.8 Planck Scale
Max Planck (1858–1947) was quite taken with the “naturalness” of what
later were called Planck scale units. [25]g Unlike SI or Gaussian units or
other units in widespread use, Planck units do not depend on the properties
of arbitrarily chosen, if universally recognized, measures like the kilogram,
nor do they depend on the properties of any one particle. The dimensional
formulae and the values of quantities that make up the Planck scale are
listed in Table 7.5. The Planck scale is the only one of the 15 possible ones
in which Planck’s constant h appears in all four characteristic quantities.

Table 7.5
Planck scale

Mass 2.18 ⋅ 10−8 kg

Length 1.62 ⋅ 10−35 m

Time 5.39 ⋅ 10−44 s

Charge 1.33 ⋅ 10−18 C

Planck scale quantities are those at which gravitational forces and

quantum effects interact and, consequently, can be interpreted in these
terms. For instance, the Planck length is the geometrical mean of
the Schwarzchild radius mG/c2 and the Compton length h/mc of an object
with mass m. The Schwarzchild radius mG/c2 is that radius of a sphere of
mass m whose escape velocity becomes the speed of light c, while the
Compton wavelength h/mc is the de Broglie wavelength of an object with
momentum mc where here and elsewhere m refers to an arbitrary rest mass.
The Planck time is that required for light to cross a Planck
distance. Furthermore, the Planck mass is that mass whose
Compton wavelength h/mc is equal to the Planck length . And the
Planck charge is that of oppositely charged particles separated by
distance d equal to the wavelength of a photon with energy, hc/d, sufficient
to disassociate the particles, that is, one for which hc/d = e2/εod.
 7.9 Example: Planck Mass
The Planck mass is easily derived from the electron mass me by
using the dimensional equivalences built into the Eddington-Dirac ratio
and the fine structure constant e2/(2εohc). In particular, we

use the equivalence , derived from the Eddington-Dirac

ratio, to identify a mass . Then we use the equivalence [e2/
ε0] = [hc], derived from the fine structure constant, to generate the Planck
mass . Interestingly, the Planck mass (∼10−8 ⋅ kg) is relatively large
on a human scale considering that the Planck length (∼10−35 ⋅ m) and
Planck time (∼10−43 ⋅ s) are so small. The best digital balances can resolve
masses to within a Planck mass.
 7.10 The 22 Quantities
The 60(=15 × 4) positions in the table of Section 7.6 are filled with 58
quantities. Of these 22 are unique. Each of these 22 unique quantities can be
interpreted in terms of certain kinds of interactions (gravitational,
electrostatic, and electromagnetic), certain effects (quantum mechanical),
and certain particle properties (electron charge and mass).
The 22 quantities separated into categories with the dimensions of M,
L, T, and Q create sequences of magnitudes. The masses in kilograms, from
largest to smallest, are

10−7 , 10−8 , and 10−30.

The lengths in meters, from largest to smallest, are

1034 , 10−10 , 10−12 , 10−14 , 10−32 , 10−35 , 10−36 , and 10−57.

The durations in seconds, from largest to smallest, are

1077 , 10−17 , 10−20 , 10−22 , 10−38 , 10−43 , 10−44 , and 10−66.

And, finally, the charges in coulombs, from largest to smallest, are

10−18 , 10−19 , and 10−41.

One is drawn to especially large and small magnitudes. For instance, the

charge (∼10−41 ⋅ C) is a factor of 1021 times smaller than the

charge of an electron. This charge is that of a spherical particle of

mass me whose repulsive self-electrostatic force and attractive self-
gravitational force are equal – a gravitational-electrostatic charge.
The length (∼1034 ⋅ m) is a factor of 108 times larger than the
size of the observable universe (∼1026 ⋅ m). This distance is the separation
of two uncharged masses me each pulled by gravity from rest at “infinity”
until equal to their de Broglie wavelengths. In contrast, the very small
length Gme/c2(∼10−57 ⋅ m) is the Schwarzchild radius of an object with the
mass of an electron.
Even more startling is the duration (∼1071 ⋅ s) – at least a
factor of 1056 times longer than the age of the universe (∼4 ⋅ 1017 ⋅ s). This
duration is, according to the Heisenberg uncertainty principle, the
quantum lifetime of a gravitating sphere of mass me and radius . By
“quantum lifetime” I mean the maximum lifetime Δt of an object
having energy ΔE that may, according to the Heisenberg
uncertainty principle ΔtΔE ≤ h, fluctuate into existence.
Other quantities are less fancifully interpreted. For instance,
h/mec(∼10−12 ⋅ m) is the Compton wavelength of an electron, and the
quantity h/mec2 (∼10−20 ⋅ s) is the time required for light to traverse this
wavelength.
We have in this chapter uncovered dimensionless products and
dimensional scales and quantities using only the methods of dimensional
analysis. As we have come to expect, dimensional analysis takes us only so
far and no further. Consequently, these dimensionless products and
dimensional scales and quantities may seem more the input than the
substance of a complete dimensional cosmology.
 Essential Ideas
The 6 fundamental constants G, εo, c, e, me, and h generate two
dimensionless products, the Eddington-Dirac number and
the fine structure constant e2/(2εohc). These 6 constants also generate
quantities with dimensions of mass, length, time, and charge that compose
scales at which various forces (gravity and electromagnetism), quantum
effects, and the properties of fundamental particle interact.
 Problems
7.1 Eddington-Dirac number. (a) Derive the dimensionless product
by applying the Rayleigh algorithm to the set of fundamental,
classical, dimensional constants: G, ε0, μ0, e, and me. (b) Check to see that
is dimensionless by using the dimensional formulae in Table
7.1 of Section 7.1.

7.2 Fine structure constant. Show that the fine structure constant
e2/(2ε0hc) is dimensionless by using the dimensional formulae in Table 7.1
of Section 7.1.

7.3 Planck scale. Derive the Planck length, time, and charge from the
classical electron length h/mec, time h/mec2, and charge e by substituting
dimensional equivalences built into the Eddington-Dirac number
and the fine structure constant e2/(2hc) as illustrated in
Section 7.9.

7.4 Electron scales? Speed of light scale? According to Section 7.8, “The
Planck scale is the only one of the 15 in which Planck’s constant h appears
in all four characteristic quantities.” But there other scales in which a
particular fundamental constant appears in all four characteristic quantities.

(a) Identify two scales that could be called an “electron mass scale” because
me appears in all four characteristic quantities.
answer
(b) Identify two scales that could be called an “electron charge scale.”
answer
(c) Identify one scale that could be called a “speed of light scale.”
answer

a I have included the vacuum permittivity εo and the speed of light c in

place of the electromagnetic constants εo and μo. Given that
only two of the three constants εo, μo, and c are
independent. Therefore, only two are needed in any one description.

b The name classical scale is something of a misnomer, since, as we shall

soon see, there are five scales composed solely of classical, fundamental,
dimensional constants.

cOne could always add to the classical or any other scale consisting of a
mass, length, and time, a characteristic temperature defined by a quantity
with the dimensional formula of energy ML2T−2 divided by Boltzmann’s
constant kB.

d The quantities in the table proportional to the classical electron radius,

the classical duration, and the Planck mass, length, time, and charge omit
factors of 4π and 2 that are included in their customary definitions.

e Of course, from any one complete scale of characterizing quantities

many other quantities can be derived by composition. For instance, given
a characterizing mass, length, time, and charge, one can derive a
characterizing speed, momentum, kinetic energy, angular momentum,
current, magnetic moment, and so on.

fNote that of these 15 scales 3 are incomplete. The fundamental

constants foundational to each of these 3 incomplete scales combine, in
turn, to make the Eddington-Dirac number from G, me, e,
and εo, the fine structure constant e2/(εohc) from e, εo, c, and h, and their
quotient from G, me, c, and h. The existence of each of these
dimensionless products reduces the number of characterizing quantities
comprising each of the incomplete scales.

gPlanck’s constant h is often replaced by the reduced Planck’s constant

ℏ(=h/2π) and εo with 4πεo in definitions of the Planck mass, length, time,
and charge.
 Appendix
Answers to Problems
1.1 π1 = ρc3/hν3and π2 = hν/kBT.

1.2 f = C ⋅ υ/r.

1.3 a = C ⋅ υ2/r.

1.4 (c) . (d) For a leg length of 1 meter υ2/gl = 0.2.

1.5 .

1.6 (a) π1 = GρΔt2 and π2 = ρR3/m. (b) π = ρGΔt2 and .

2.1 [G] = M−1L3T−2.

2.2 (b) p = Gρ2R2 ⋅ f(r/R).

2.3 .

2.4 .

2.5 .

2.6 (a) D = C ⋅ ρr2υ2. (b) .

2.7 .

2.9 R = C ⋅ E1/5t2/5/ρ1/5.

3.2 .

3.3 h/d and σ/(ρgd2) or, equivalently, h = d ⋅ f(σ/ρgd2) where f(x) is

undetermined.

3.4 t = C ⋅ μr2/mg.

3.5 (a) π1 = Δt(ρg2/μ)1/3 and π2 = lρg1/2/μ. (b) Δt = (μ/ρg2)1/3 ⋅ f(lρg1/2/μ).

3.6 m = C ⋅ ρυ6/g3.

3.7 π1 = ρσ3/(μ4g), π2 = υμ/g, and π3 = pσ/(μ2g) or other complete sets of

3 independent dimensionless products that can be formed from these.

4.1 [R] = ML2T−2Θ−1.

4.2 (a) . (b) .

4.3 The dimensionless products are ql/κΔT and υlcp/κ. Therefore,

q = (κΔT/l) ⋅ f(υlcp/κ) where f(x) is undetermined.

5.1 .

5.2 (a) The dimensionless products are Ed2εo/q and υ/c. Therefore,
E = (q/d2εo) ⋅ f(υ/c) where f(x) is undetermined. (b) The dimensionless
products are Bd2εoc/q and υ/c. Therefore, B = (q/(d2εoc)) ⋅ g(υ/c) where
g(x) is undetermined.
5.3 and
.

5.5 (a) The dimensionless products are mΔt2/(μoI2) and NLR. (b)
. In typical pinches NLR > > 1 and
.

5.6 (a)

I Current (QT−1)

R Radius L

NL Density per unit length L−1

kBT Boltzmann’s constant times absolute (ML−2T−2)L4

temperature

μo Vacuum permittivity (ML−2T−2)L3(T2Q−2)

6.1 .

7.4 (a) “Electron mass scales” are based on G, me, εo, and c, and on G,
me, εo, and h. (b) “Electron charge scales” are based on G, e, εo, and c,
and on G, e, εo, and h. (c) A “speed of light scale” is based on G, εo, c,
and h. The “speed of light scale” is identical to the Planck scale.
 References

1. Galileo Galilei, Two New Sciences, translated by Henry Crew and

Alfonso de Salvio, (Chicago, IL: Encyclopedia Britannica Inc., 1952), p.
187.

2. H. E. Huntley, Dimensional Analysis (Mineola, NY: Dover, 1967), p. 33.

3. Joseph Fourier, The Analytical Theory of Heat, translated by Alexander

Freeman (Mineola, NY: Dover, 1878), Book II, Section IX, Articles 157–
162. Fourier closes his discussion of dimensional analysis with the
comment that, “On applying the preceding rule to the different equations
and their transformations, it will be found that they are homogeneous with
respect to each kind of unit, and that the dimension of every angular or
exponential quantity is nothing. If this were not the case some error must
have been committed in the analysis … .”

4. Lord Rayleigh (John William Strutt), The Principle of Similitude. Nature,

March 18, (1915).

5. Edgar Buckingham, On Physically Similar Systems; Ilustrations of the

Use of Dimensional Equations. Physical Review, Vol. IV, no. 4, (1915),
345–376.
6. H. L. Langhaar emphasizes this aspect of the π theorem. See his
Dimensional Analysis and Theory of Models (Hoboken, NJ: Wiley, 1951),
p. 18.

7. Edgar Buckingham, On Physically Similar Systems; Ilustrations of the

Use of Dimensional Equations. Physical Review, Vol. IV, no. 4, (1915),
345–376.

8. This definition reformulates an equivalent one found in E. R. Van Driest,

On Dimensional Analysis and the Presentation of Data in Fluid Flow
Problems. J. Applied Mechanics, Vol. 13, no. 1, (1946) A–34. See also H.
L. Langhaar, Dimensional Analysis and Theory of Models (Hoboken, NJ:
Wiley, 1951), p. 29.

9. The number of effective dimensions ND is also the “rank of the

dimensional matrix” – a mathematical concept exploited in fluid mechanics
engineering texts. See, for instance, R. W. Fox, A. T. McDonald, and P. J.
Pritchard, Fluid Mechanics (Hoboken, NJ: Wiley, 2004), pp. 282–283.

10. The concept, although not the name, of imposed dimension originates
with Percy Bridgman’s highly recommended text Dimensional Analysis
(New Haven, CT: Yale University Press, 1922). See, in particular, pp. 9–11,
63–66, 67–69, and 77–78. Others, including H. E. Huntley, Dimensional
Analysis, (Mineola, NY: Dover, 1967), use the concept of imposed
dimensions.

11. G. I. Barenblatt, Scaling, Self-Similarity, and Intermediate Dynamics

(Cambridge, UK: Cambridge University Press, 1996), pp. 18 ff.
12. Hermann Helmholtz, On the Sensations of Tone 6th Edition (Gloucester,
MA: Peter Smith, 1948), pp. 43–44.

13. Percy Bridgman, Dimensional Analysis (New Haven, CT: Yale

University Press, 1922), p. 107.

14. G. Taylor, The Formation of a Blast Wave by a Very Intense Explosion.

II. The Atomic Explosion of 1945. Proceedings of the Royal Society of
London. Series A, Mathematical and Physical Sciences, Vol. 201, No. 1065,
(Mar. 22, 1950), 175–186.

15. David C. Lindberg, The Beginnings of Western Science (Chicago, IL:

University of Chicago Press, 1992), p. 305.

16. See Aristotle, The Basic Works of Aristotle editor Richard McKeon,
(New York, NY: Random House, 1966) Physics, Book IV, chapter 8, p.
216a, lines 14–17. See also David C. Lindberg, The Beginnings of Western
Science (Chicago, IL: University of Chicago Press, 1992), pp. 59–60.

17. R. P. Godwin, The Hydraulic Jump (‘Shocks’ and Viscous Flow in the
Kitchen Sink), American Journal of Physics 61 (9), (1993) 829–832.

18. Fridtjof Nansen, Farthest North (Edinburgh, UK: Birlinn, 2002), p. 186.

19. C. Vuik, Some Historical Notes on the Stefan Problem, Nieuw Archief
voor Wiskunde, 4e serie 11 (2), (1993) 157–167.

20. Percy Bridgman, Dimensional Analysis (New Haven, CT: Yale

University Press, 1922), problem 19, p. 108.
21. Traditionally the “Boussinesq problem.” See Percy Bridgman,
Dimensional Analysis (New Haven, CT: Yale University Press, 1922), pp.
9–11.

22. Lloyd W. Taylor, Physics: The Pioneer Science (Mineola, NY: Dover,
1941), pp. 592–593.

23. John Dunmore, Pacific Explorers: The Life of John Francois de La

Perouse 1741–1788 (Palmerston North, NZ: The Dunmore Press Limited,
1985), pp. 286–292.

24. L. Schiff, Quantum Mechanics 3rd Edition (New York, NY: McGraw
Hill, 1968) pp. 397 ff.

25. Max Planck, Theory of Heat Radiation (Mineola, NY: Dover, 2011), pp.
205–206. Number 164.
 Index
Ampere-Maxwell law, 62
Analytical Theory of Heat, The, 50
arctic ice, growth of, 55–57
Aristotle (384-322 BCE), 39
asymptotic expansion, 24–25, 59
atomic scale, 79–80, 92
atomic time, 84

Black, Joseph (1728–1799), 49

blackbody radiation, 3, 74–77
blast wave, 32
Bohr atom, 77–79
Bohr, Niels (1885-1962), 66, 77
Boltzmann’s constant, 3, 70, 85–86
Borda, Jean Charles (1733–1799), 63
Bose-Einstein statistics, 82
boson, 82
Boyle’s law, 11
Buckingham π theorem, 7–8
Buckingham, Edgar (1867–1940), 7
Buckingham’s rule of thumb, 9–12, 18–19, 30
bulk modulus, 23, 26
 of ideal gas, 27

calorimetry, 49
capillary waves, 36–37, 47
Carnot, Sadi (1796-1832), 49
centripetal acceleration, 15
charge density, 61–62
Child, C. D. (1868–1933), 66
Child’s law, 66–68, 73
classical electron radius, 87–88
classical scale(s), 88–89, 91
classical time, 86
Compton wavelength, 93–94
Compton, A. H. (1892–1962), 74
convection, 51, 60
cooking a turkey, 51–53
current density, 61–62, 66

deep-water wave, 35
degeneracy pressure, 82
diffusion equation, 54
dimensional constants
fundamental, 85
table of, 85
dimensional cosmology, 85
dimensional formulae, 4–5, 8, 17, 19–20
dimensional homogeneity, 1–3, 5
dimensional model, 11
dimensionless product, 3–4, 6, 48
 complete, 8
independent, 8
dimensionless products
number of, 9
dimensions
absolute vs. relative, 22
effective, 17–20
imposed, 21–22
number of, 8–9
Dirac, Paul (1902–1984), 86
displacement current, 62

Eddington, Arthur (1882–1944), 86

Eddington-Dirac number, 85–87, 90, 93, 95
effective dimension, 17–20, 26, 30, 60, 64, 72–73
number of, 19–20
Einstein, Albert (1879–1955), 74
elastic materials, 25
electret oscillations, 72
electric constant. See vacuum permittivity
electric dipole moment, 72
electric field, 61, 65, 68, 72
electromagnetic wave, 62, 65, 76, 82
electron tube, 66
escape velocity, 31
Experimental Researches in Electricity, 62

Faraday, Michael (1791–1867), 62

Faraday’s law, 62–63
Fermi-Dirac statistics, 82
fine structure constant, 83, 89–90, 93, 95
fluid element, 43
fluid variables, 33
Fourier, Joseph (1768–1830), 2, 50
free vibrations, 32
freeling falling object, 3
Froude number, 15

Galileo (1584–1642), 1
geomagnetic field, 63
gravitational constant, 15, 31, 85
gravitational instability, 31
gravitational-electrostatic charge, 87
gravitational-electrostatic mass, 87
gravity wave, 37

hanging stretched cable, 23–24

heat, 49–50, 59
heat capacity, 49–50, 56
heat conduction, 50–51, 60
heat flux, 51, 54
heat transfer, 49, 54
Heisenberg uncertainty principle, 94
Helmholtz resonator, 27–28
Helmholtz, Hermann von (1821–1894), 27
hydraulic jump, 41–43

ideal fluid, 32
ideal gas, 9–11, 19, 27, 59, 80, 84
quantized, 80–82
ideal gas constant, 59
imposed dimension, 20–21, 30
ionosphere, 68

Joule, James (1818–1889), 49

Kepler’s third law, 30

laminar flow, 42
latent heat, 49
light waves, 62
Lorentz force law, 61

magnetic constant. See vacuum permeability

magnetic dipole moment, 63–64
magnetic field, 61, 65, 68, 70, 72
Maxwell, James Clerk (1831–1879), 61
Maxwell’s equations, 61–63

Nansen, Fridtjof (1861–1930), 55

Newton, Isaac (1643–1727), 1
Newton’s law of gravitation, 31
Newton’s second law, 1, 20
non-dimensionalization, 55

orbits, two-body, 29–30

oscillations of a compass needle, 63–65

Pauli, Wolfgang (1900–1958), 90

pinch effect, 70–73
pipe flow
equilibrium, 43–45
fully developed turbulence, 46
non-equilibrium, 45–46
scale model, 46–47
Planck scale, 92–93, 95
charge, 93
length, 93
mass, 93
time, 93
Planck, Max (1858–1947), 74, 92
Planck’s constant, 76, 85
Planck’s constant, 3, 74, 77, 80
plasma frequency, 69
plasma oscillation, 69, 73
plasma physics, 68
power law approximation, 25
pressure, 9–11, 17–19, 31
gradient, 43, 47

quantum electrodynamics, 89
quantum physics, 74

radiation from accelerating charge, 65–66

quantized, 82–83
Rayleigh algorithm, 4–7, 47
modified, 7
Rayleigh, Lord, 5, 47
Reynolds number, 42–43, 46
Reynolds, Osborne (1842–1912), 42

scale model, 46–47

Schwarzchild radius, 93–94
shallow-water wave, 35
shear viscosity. See viscosity
side window buffeting, 27–28
similitude, principle of, 5
dynamic, 47
geometric, 47
small amplitude wave, 34
soap bubble, 32
Sommerfeld, Arnold (1868–1951), 89
space charge limited current density. See Child’s law
spectral energy density, 15, 76–77
speed of light, 62
speed of sound, 26
spring-mass system, 12–13, 21–22
stack effect, 58–59
Stefan problem, 57
Stefan, Josef (1835–1893), 57
Stefan-Boltzmann law, 3, 75
Stokes, George (1819–1903), 40
Stokes’ law, 40
Strutt, John William (1842–1919). See Rayleigh, Lord
surface tension, 35–37
symmetry, 2, 30
temperature, 49
gradient, 51, 54
terminal speed, 39–40, 48
thermal conductivity, 50–51, 54, 56
thermal diffusivity, 54
thermodynamics, 49
Thomson, William (1824–1907), 49
tsunami, 35
turbulent flow, 42

USDA data, 53

vacuum permeability, 61
vacuum permittivity, 61
vacuum tube, 66
variable of interest, 7, 10, 24, 26
virtual cathode, 68
viscosity, 39, 48

water drop oscillation, 47

water drop, size of, 37–38
water waves, 33–35
wave speed
deep-water, 35
of capillary waves, 37
of light, 62
of sound, 26
of tsunami, 35
shallow-water, 35
waveform, transverse, 31

z-pinch, 70