Figure 10. A self-affine fractal provides a simple
model for the geometry of a fern.
of Nature and landed a consulting job on stock
options pricing. I concluded, as Mandelbrot had
surmised decades earlier, that the stock price is
not a “random walk”, that complexity and noise
are often indistinguishable, and that complexity
may be modeled by fractals. Market pricing is
essentially deterministic, not random. At that
time, on a daily basis, traders would run their
Black-Scholes models, which assume pricing is a
random diffusion process, and bring the results
to the floor each morning like racing forms at
the horsetrack. They trusted these cheat-sheets to
tell them when to buy and sell. But I was able to
exploit the limitations of the Black-Scholes model
using fractals and made a decent little fortune for
someone who had recently been a poverty-stricken
student.
The notion of “fractals as antennas” occurred
to me in 1987 while attending a lecture by Man-
delbrot. I went home and explored this curious
idea, which has subsequently become a major
theme of my efforts and a field in its own right.
Some years later I saw Benoît again at a fractals-
in-engineering conference. This was finally the
opportunity to converse with him and the first
of several lunch meetings and subsequent phone
conversations in the last dozen years of his life.
No one who had such conversations can forget the
brilliant, witty joy of Benoît the polymath. In par-
ticular, they helped me to realize that Maxwell’s
equations require self-similarity for frequency in-
variance, a fundamental and what should have
been obvious result. Now I see many problems
that benefit from fractals: metamaterials, a new
form of radiative transport, optimization, and
fluid mechanics and drag reduction. I only regret
that I can’t share these with Benoît anymore.
Figure 11. Various pictures constructed from the
orbit of a leaf picture under a system of three
affine transformations. The limit set of the
semigroup is illustrated in red and yellow.
Figure from [2].
Stéphane Jaffard
Parts of Mathematics Are Totally Bathing in
the Ideas That Benoît Introduced
Benoît was one of the first to apply computer
graphics to mathematical objects. He used them to
develop intuitions and to make either discoveries
or deep conjectures.
He also put forward particular entities such
as Mandelbrot cascades, the Mandelbrot set, Lévy
dusts, and so on as beautiful objects, worthy of
study in their own right. At that time, this was
orthogonal to the main direction of mathematics
towards generalizations and abstract structures. I
believe that Benoît’s influence on the mathemat-
ical community was very helpful in that respect:
mathematics was able to admit a down-to-earth
component. Some parts of mathematics are now
totally bathing in the ideas that Benoît introduced.
For example, the idea of scale invariance is every-
where present in the mathematics of signal
processing, my area.
More broadly, the notion of fractal prob-
ability has been one of the most important
unifying concepts in science introduced in the
last fifty years. It has allowed scientists with di-
verse specializations to draw connections between
seemingly unrelated subjects and has created un-
expected cross-fertilizations. This was driven by
the mesmerizing and enthusiastic personality of
Benoît.
Note that fractals are one of the few parts of
mathematics that can be “shown” to the general
Stéphane Jaffard is professor of mathematics at Univer-
sité Paris Est (Créteil Val-de-Marne). His email address is
stephane.jaffard@u-pec.fr.
1216
Notices of the AMS
Volume
59, Number 9
Figure 12. “…[E]ighty students in my fractal geometry course learn in a single class how to generate
the fractals pictured here.…”
public. As a teenager, I was influenced by Benoît’s
fascinating books. They explained a part of math-
ematics that was under construction yet could be
readily understood.
My thesis was on the then-new topic of
“wavelets”. I worked at École Polytechnique under
the supervision of Yves Meyer. Once Benoît visited
École Polytechnique, and he heard that a Ph.D.
student was working on systems of functions that
could be decomposed into elementary blocks,
related inter alia by dilations and translations. He
came to my office, and we had long conversations
about new possibilities offered by wavelet anal-
ysis. For me, this was the start of interactions
which influenced me considerably; it certainly
pushed me towards specializing in multifractal
analysis, a part of fractals where Mandelbrot’s
ideas are prevalent. Our interactions resulted in
two joint papers on Polya’s function, whose graph
is space-filling and multifractal (its Lipschitz
regularity index jumps everywhere). The interest
that Benoît showed in this example, which was
quite forgotten at that time, was typical of his
fascination for beautiful mathematical objects
and the art with which he managed to draw
a correspondence between their mathematical
beauty and their graphical beauty. In all the con-
versations that we shared, I was always amazed
by the uninterrupted flow of original and brilliant
ideas that he very generously shared.
Sir Michael Berry
How to Model…a Surface With No Separation
of Scales
In the early 1970s, I was studying radio-wave
echoes from the land beneath the ice in Antarc-
tica. Existing theories separated the “geography”,
supposedly measured by the start of the echo,
from the “roughness”, indicated by the disorderly
Sir Michael Berry is professor of physics at the Univer-
sity of Bristol, UK. His email address is asymptotico@
physics.bristol.ac.uk.
echo trail. The separation was modelled by a flat
surface (“geography”) superimposed on what was
single-scale randomness (“roughness”), typically
gaussian. I found this not only unappealing but
also scientifically absurd: in a natural landscape,
any apparent dichotomy must be an illusion, an
artifact of the wavelength used to interrogate it.
But how to model, or even describe, a surface with
no separation of scales? I had no idea until I read
Philip Morrison’s review of the English edition of
Mandelbrot’s Fractals: Form, Chance and Dimen-
sion [16]. I cannot remember being so excited by a
book review. It was immediately clear that fractal
dimension was the key idea I needed, and this was
confirmed by the book itself.
Quickly came the identification of a new class of
wave phenomena: “diffractals”, that is, waves inter-
acting with fractal objects. In the echo-sounding
of landscapes, the interaction is mainly reflec-
tion. Later, a grim consequence of an absorption
interaction emerged: we realized that the pro-
longed winter predicted to occur after a nuclear
war, because of the absorption of sunlight by
smoke, would be significantly intensified by the
fact that smoke particles are fractal (it would also
be prolonged, because smoke’s fractality slows
the particles’ fall). From the development of quan-
tum chaology in the late 1970s came a conjecture
about the spectra of enclosures (“drums”) with
fractal boundaries: the “surface” correction to the
“bulk” Weyl eigenvalue counting formula would
scale differently with frequency and depend on
the fractal dimension. This generated considerable
mathematical activity.
In diffractals it is the objects interacting with
the waves, not the waves themselves, that are
fractal. But in some phenomena the wave intensity
is fractal on a wide range of scales down to the
wavelength. One such, unexpected in one hundred
fifty years, is the Talbot effect, associated with light
beyond diffraction gratings whose rulings have
sharp edges: the fractal dimensions of the wave
across and along the beam direction are different.
All this sprang from Benoît Mandelbrot’s insight,
October
2012
Notices of the AMS
1217