have my original copy of his book, signed by
Benoît, on the one occasion at Princeton that
we met.
—John Hutchinson
Iterated function systems (IFSs) are now a
standard framework for handling deterministic
fractals, self-similar sets and measures. They were
named by this author and Stephen Demko [1],
though Benoît thought we should have called
them “map bags”. He was fascinated by models of
leaves with veinlike internal structures made by
invariant measures of IFSs.
Hutchinson’s paper and the work of many oth-
ers influenced by Mandelbrot ended a long period
where geometry and the use of pictures played
little role in mathematics. Mandelbrot believed
passionately in pictorial thinking to aid in the de-
velopment of conjectures and formal proofs. His
advocacy has enabled it to be okay once again for
mathematicians to do experimental mathematics
using pictures.
Mandelbrot’s ideas have inspired a huge amount
of research, from pure mathematics to engineer-
ing, and have resulted in deep theorems; a new
acceptance of geometry and pictures as having a
role to play in experimental mathematics; and var-
ious applications, including image compression
and antenna design. The notion of a fractal now
forms part of good preuniversity mathematics ed-
ucation, while the mathematical study of fractals
has its own specialist areas, including, for exam-
ple, analysis on fractals [8] and noncommutative
fractal geometry [9].
One important idea of Mandelbrot was that
various random phenomena, such as stock market
prices, are governed by probability distributions
with “fat tails”. This led him to warn in 2004 that
“Financial risks are much underestimated. I think
we should take a strongly conservative attitude
towards evaluating risks.” The subsequent global
financial crisis underlined his point.
Prior to editing both this article and [3], we
emailed colleagues to ask for memories and com-
ments on Benoît’s contributions to mathematics,
influence, and personal recollections. We received
replies from many: not only mathematicians but
artists, physicists, biologists, engineers, and so on.
Using these replies we have produced two articles:
this one and [3], which is more focused on recol-
lections of the man himself. Our goal has been to
put together something special using the words
of everyone who wrote but, in general, editing and
shortening to avoid repetition of themes.
From early on, Mandelbrot was driven by a
desire to do something totally original, to look at
problems that others found too messy to consider,
and to find some deep unifying principles. As the
Figure 3. Superposition of the attractors, colored
using fractal transformations (see [2]) of two
simple bi-affine iterated function systems.
words in the following contributions show, he
succeeded.
Roger Howe
Participating in a Conversation That Takes
Place over Long Spans of Time
One pleasure of doing mathematics is the sense
of participating in a conversation that takes place
over long spans of time with some of the smartest
people who ever lived. Benoît’s work on fractals
provides a good example of this kind of long-term
dialogue.
A significant factor in the invention of calculus
was the idea of representing a curve by the graph
of a function and, reciprocally, of representing the
time variation of a quantity by a curve. This back-
and-forth identification allowed one to connect
the drawing of tangent lines with finding the rate
of change of quantities that vary in time.
When calculus was invented in the seventeenth
century, the concept of function was not very
precise. Work during the eighteenth century on
solving the wave equation using sums of sine and
cosine functions led to a sharpening of under-
standing of the essential properties of functions
and of their behavior. This led in the first half of
the nineteenth century to the isolation by Cauchy
of the notion of continuity, which made clear
for the first time the distinction between con-
tinuity and differentiability. During the rest of
the nineteenth century, mathematicians explored
this difference, which contributed to the general
unease and insecurity about the foundations of
mathematics. Hermite is quoted as “recoiling in
horror from functions with no derivatives.” The
early twentieth century saw the production of
Roger Howe is professor of mathematics at Yale University.
His email address is roger.e.howe@yale.edu.
1210
Notices of the AMS
Volume
59, Number 9
Figure 4. Two illustrations of IFS semigroup tilings. The triangle on the left is tiled with the orbit of a
six-sided figure under a system of two affine transformations. The limit set of the set of triangular
tiles on the right is the attractor of a system of three affine transformations. A theme of Benoît’s work
was that the iteration of simple rules (e.g., elementary geometrical transformations) can produce
nondifferentiable (rough) objects. Figure from [2].
a menagerie of striking examples (the Cantor
set, the Koch snowflake, the Sierpinski carpet,
etc.) illustrating the difference between continuity
and differentiability. However, for several decades
these examples were regarded as exotica, mon-
sters with no relation to the physical world. They
were objects only a mathematician would inves-
tigate. They were liberated from this marginal
status by Mandelbrot, who said, “Wait a minute.
A lot of things in the world—clouds, river sys-
tems, coastlines, our lungs—are well described by
these monsters.” Thus started the use of these
mathematical objects to study complicated, messy
nature.
Ian Stewart
No Lily-White Hands
I first learned about fractals from Martin Gar-
dener’s Scientific American column. I promptly
bought a copy of Fractals: Form, Chance, and Di-
mension [16]. Despite, or possibly because of, its
unorthodoxy and scope, it seemed to me that
Benoît Mandelbrot had put his finger on a brilliant
idea.
I’m pleased that, towards the end of his life, he
received due recognition, because it took a long
time for the mathematical community to under-
stand something that must have been obvious to
him: fractals were important. They were a game
changer, opening up completely new ways to think
about many aspects of the natural world. But for
a long time it was not difficult to find professional
Ian Stewart is emeritus professor of mathematics at the
University of Warwick, UK. His email address is I.N.
Stewart@warwick.ac.uk.
research mathematicians who stoutly maintained
that fractals and chaos were completely useless
and that all of the interest in them was pure hype.
This attitude persisted into the current century,
when fractals had been around for at least twenty-
five years and chaos for forty. That this attitude
was narrow-minded and unimaginative is easy to
establish, because by that time both areas were
being routinely used in branches of science rang-
ing from astrophysics to zoology. It was clear that
the critics hadn’t deigned to sully their lily-white
hands by picking up a random copy of Nature or
Science and finding out what was in it.
To be sure, Mandelbrot was not a conventional
academic mathematician, and his vision often
carried him into realms of speculation. And it was
easy to maintain that he didn’t really do much
that was truly novel—fractal dimension had been
invented by Hausdorff, the snowflake curve was
a century old, and so on. Mathematicians would
have cheerfully gone on employing Hausdorff-
Besicovitch dimension to consider such questions
as finding a set of zero dimension that covers
every polygon, but they would not have figured
out that quantifying roughness would make it
possible to apply that kind of geometry to clouds,
river basins, or how trees damp down the energy
of a hurricane.
Mandelbrot’s greatest strength was his instinct
for unification. He was the first person to real-
ize that, scattered around the research literature,
often in obscure sources, were the germs of a
coherent framework that would allow mathemat-
ical models to go beyond the smooth geometry
of manifolds, a reflex assumption in most areas,
and tackle the irregularities of the natural world
October
2012
Notices of the AMS
1211