Figure 5. A Julia set associated with the first
cascade of period doubling bifurcations of the
logistic equation. Julia sets for quadratic maps
are intimately related to the Mandelbrot set.
Benoît was one of the first to use computers to
make pictures of mathematical objects:
computations which took hours to run on
expensive mainframes can now be performed in
seconds on handheld devices. This image and
Figures 1 and 13 were computed using free
software (Fractile Plus) on an iPad.
in a systematic fashion. It took many years be-
fore these ideas began to pay off, but that’s how
pioneering mathematics often goes.
The conjecture whose proof so pleased him
(see [3, contribution by Ian Stewart]) was the work
of Gregory Lawler, Oded Schramm, and Wendelin
Werner in their paper “The dimension of the
planar Brownian Frontier is 4/3” [10]. It is part
of the work for which Werner received a Fields
Medal, and it shows that fractals have given rise
to some very deep mathematics. I suspect that
only now are we beginning to see the true legacy
of Mandelbrot’s ideas, with a new generation of
researchers that has grown up to consider chaos
and fractals to be as reasonable and natural as
periodic motion and manifolds. Mandelbrot was
a true pioneer, one of the greatest mathematical
visionaries of the twentieth and early twenty-first
centuries.
David Mumford
Benoît Told Me: “Now You Can See These
Groups and See Teichmüller Space!”
Benoît Mandelbrot had two major iconoclastic
themes. First, that most of the naturally occurring
measurements of the world were best modeled
by nondifferentiable functions, and second, the
histograms of these measurements were best
modeled by heavy-tailed distributions. Even if
he did not bring a new unifying law like Newton’s
David Mumford is emeritus professor of mathematics at
Brown University. His email address is DavidMumford@
brown.edu.
F =
ma and even if he did not have the deep and
subtle theorems that make waves in the pure math
community, this vision was revolutionary. What
his lectures made clear was that fractal behavior
and outlier events were everywhere around us,
that we needed to take these not as exceptions but
as the norm. For example, my own work in vision
led me later on to express his ideas about out-
liers in this way: that the converse of the central
limit theorem is true, namely, the only naturally
occurring normal distributions are ones which are
averages of many independent effects.
Benoît’s immediate effect on my work was to
reopen my eyes to the pleasure and mathematical
insights derived from computation. I had played
with relay-based computers in high school and
with analog computer simulations of nuclear reac-
tors in two summer jobs. But at the time I thought
that only white-coated professionals could han-
dle the IBM mainframes and puzzled over what
in heaven’s name my colleague Garrett Birkhoff
meant when I read “x = x + 1” in some of his
discarded code. But Benoît told us that complex
iterations did amazing things that had to be seen
to be believed. These came in two types: the lim-
iting behavior of iterations of a single analytic
function and the limiting behavior of discrete
groups of Möbius transformations. The second of
these connected immediately to my interests. I
was always alert to whatever new tool might be
available for shedding any sort of light on moduli
spaces, whether it was algebro-geometric, topolog-
ical, characteristic p point counting, or complex
analytic. I had sat at the feet of Ahlfors and Bers
and learned about Kleinian groups and how they
led to Teichmüller spaces and hence to moduli
spaces. Benoît told me, “Now you can see these
groups and see Teichmüller space!”
I found an ally in Dave Wright, learned C, and
with Benoît’s encouragement, we were off and
running. When he returned to his position at the
IBM Watson Lab, he set up a joint project with us,
and we visited him and his team there. Later, Curt
McMullen, who also appreciated the power and
insight derived from these experiments, joined
us. It turned out that, in the early hours of the
morning, their mainframes had cycles to spare,
and we would stagger in each morning to see what
these behemoths had churned out. There was no
way to publish such experiments then, but Dave
and I astonished the summer school at Bowdoin
with a live demo on a very primitive machine of
a curvy twisting green line as it traced the limit
point set of a quasi-Fuchsian group. Ultimately,
we followed Benoît’s lead in his Fractal Geometry
of Nature [18] and, with Caroline Series, published
our images in a semipopular book, Indra’s Pearls
[27]. One anecdote: We liked to analyze our figures,
1212
Notices of the AMS
Volume
59, Number 9
estimating, for example, their Hausdorff dimen-
sion. We brought one figure we especially liked to
Watson Labs and, thinking to test Benoît, asked
him what he thought its Hausdorff dimension was.
If memory serves, he said, “About 1.8”, and indeed
we had found something like 1.82. He was indeed
an expert!
Hillel Furstenberg
He Changed Fundamentally the Paradigm
with Which Geometers Looked at Space
Let me begin with some words of encouragement
to you on this project, dedicated to memorializing
an outstanding scientist of our times and one we
can be proud of having known personally.
What do you see as Benoît’s most important con-
tributions to mathematics, mathematical sciences,
education, and mathematical culture?
Benoît Mandelbrot sold fractals to mathemati-
cians, changing fundamentally the paradigm with
which geometers looked at space. Incorporating
fractals into mainstream mathematics rather than
regarding them as freakish objects will certainly
continue to inspire the many-sided research that
has already come into being.
Kenneth Falconer
It Was Only on the Fourth or Fifth Occasion
That I Really Started to Appreciate What He
Was Saying
Benoît’s greatest achievement was that he changed
the way that scientists view objects and phenom-
ena, both in mathematics and in nature. His
extraordinary insight was fundamental to this,
but a large part of the battle was getting his ideas
accepted by the community. Once this barrier was
broken down, there was an explosion of activity,
with fractals identified and analyzed everywhere
across mathematics, the sciences (physical and
biological), and the social sciences.
Benoît realized that the conventional scien-
tific and mathematical approach was not fitted
to working with highly irregular phenomena. He
appreciated that some of the mathematics needed
was there—such as the tools introduced by Haus-
dorff, Minkowski, and Besicovitch—but was only
being used in an esoteric way to analyze spe-
cific pathological sets and functions, mainly as
Hillel Furstenberg is professor of mathematics at Bar-
Ilan University, Israel. His email address is harry@math.
huji.ac.il.
Kenneth Falconer is professor of pure mathematics at the
University of St. Andrews in Scotland. His email address is
kjf@st-andrews.ac.uk.
Figure 6. A self-similar fractal of Hausdorff
dimension (4ln2)/ln5 ≐ 1.72
(4
ln2
)/ln5 ≐ 1
.72
(4
ln2
)/ln5 ≐ 1
.72 associated with the
pinwheel tiling.
counterexamples that illustrated the importance
of smoothness in classical mathematics.
Benoît’s philosophy that such “fractal” objects
are typical rather than exceptional was revolu-
tionary when proposed. Moreover, he argued that
the mathematical and scientific method could
and should be adapted to study vast classes
of fractals in a unified manner. This was no
longer mathematics for its own sake, but math-
ematics appropriate for studying all kinds of
irregular phenomena—clouds, forests, surfaces,
share prices, etc.—that had been ignored to a
large extent because the tools of classical smooth
mathematics were inapplicable.
Benoît also realized that self-similarity, broadly
interpreted, was fundamental in the genesis, de-
scription, and analysis of fractals and fractal
phenomena. Given self-similarity, the notion of di-
mension is unavoidable, and “fractal dimension”
in various guises rapidly became the basic mea-
sure of fractality, fuelling a new interest in the
early mathematics of Hausdorff, Minkowski, and
others.
Benoît had many original ideas, but his presen-
tation of them did not always follow conventional
mathematical or scientific styles, and as a result it
often took time for his ideas to be understood and
sometimes even longer for them to be accepted.
A case in point is that of multifractal measures.
Multifractals are, in many ways, more fundamen-
tal than fractal sets. Many of the now standard
notions of multifractals may be found in his 1974
paper in the Journal of Fluid Dynamics [14], but
this is not an easy paper to fathom, and it was not
until the 1980s that the theory started to be ap-
preciated. Benoît suggested that “the community
was not yet ready for the concept,” but I think the
delay was partly because of the way the ideas were
presented. I heard Benoît’s talk on multifractals
many times in the 1980s; he was charismatic, but
his explanations were such that it was only on
the fourth or fifth occasion that I really started to
appreciate what he was saying.
October
2012
Notices of the AMS
1213