Ill sx GALAXIES AND EDDIES
9 » Fractal View of Galaxy Clusters
In Chapters 6 and 7, the Kock and Peano
fractals ae Inrodued via-geomorpholoey,
but the mou sigaican tes of fal ae
rooted elsewhere, Inching toward the tal
Stream af tiene, tht chapter and the nxt
two tackle two ise exceplonal antiguy,
Importance and ificaliy
"he disbuon afte tar, the gli
the chister e ealanics, and o on fasinates
the amateur ab wel the speci, otc
asuophgic "a whol. The base reason Is
{hat no one has yt expsined why the dt
arety at Tessin certain range af
Seues While thre ae allusions to scing
Gal development hasten to sweep tuner the
rag iming that on sel beyond tome
tnformly dsibued, nt
Lest fundamentally, the Bsitation in eal
ing with the irregular arises from the absence
of tools to describe it mathematically. State
shed 10 decide between two assump
tions, only one of which is thoroughly ex
plored (asymptotic uniformity). Is it surpie
{ng thatthe results are inconclusive?
The questions, however, refuse to be set
ie. In parallel with efforts to explain, |
think it indispensable to describe clustering,
and to mimic reality by purely geometric
means. The fractal treatment of this subject,
Scattered over several chapters of this Esay
proposes to show by explicitly constructed
model thatthe evidence is compatible with 3
degree of clustering that extends far beyond
the limits suggested by existing models
‘The present introductory chapter describes
an influemial theory of the formation of stars
and galaxies, due to Hoyle, the principal de
scriptive model of their distribution, due te
Fournier d'Albe (also known a8 the Charlier
model), and, most important, sketches some{9 5m FRACTAL VIEW OF GALAXY CLUSTERS.
empirical data. It is shown that both theories
fand data can be interpreted in terms of a sca:
ing fractal dust. T argue that the distribution
Of galaxies and of stars includes zone of
self-similarity in which the fractal dimension
Satisfies O€D<3, Theoretieal reasons for ex
posting D=1 are sketched, raising the ques
tion of why the observed D is “1.23.
Preview. Chapter 22 uses fractal tools to
improve our understanding of what the cosmo
logical principle means, how itean and should
‘be modified, and why the modification de
‘mands randomness. A’ discussion of improved
‘odel clusters is withheld until Chapters 22,
23, and 32 0 3.
Is THERE A GLOBAL
DENSITY OF MATTER?
Let us begin with a clase examination of the
concept of global density of matte As with
the concept of the length ofa coastline, things
scom simple, but in fact go awry very quickly
And most interestingly. To define and measure
ensity, one starts with the mass M(R) in a
sphere of radius R centered on Earth, The
approximate density, defined as
M(R)/((4/3)2R),
is evaluated. After thai, the value of R is
made to tend toward infinity, and the global
ensity js defined a6 the limit toward which
the approximate density converges
But need the global density converge to a
ws
positive and finite limit? If 50, the speed of
‘comergence leaves a great deal to be desired,
Furthermore, the estimates of the limit dens:
ty had behaved very oddly in the past. As the
depth of the world perceived by telescopes
Increased, the approximate density diminished
ina surprisingly systematic manner, Accord=
ing to de Vaucoulevrs 1970, it has remained
RO? The observed exponent D is much
Salle than 3, the best estimate, on the basis
of indirect evidence, being D= 123,
‘The thesis of de Vaucouleurs is that the
behavior of the approximate density reflects
reality, meaning that M(R)=R®. This Formula
‘recalls the classical result that a ball of radius
R ina Euclidean space of dimension E has a
volume RE. In Chapter 6 we encounter the
same formula for the Koch curve, with the
major difference that the exponent is ot the
Euclidean dimension E=2 but a fraction-
valued fractal dimension D. And Chapter
erives M(R)=RP for the Cantor dust on the
time axis (For which Em)
‘Al these provedents suggest very strongly
that the de Vaueouleurs exponent D isa feac-
tal dimension
ARE STARS IN THE SCALING RANGE?
‘Obviously, the sealing range in which D satis:
fies O And the smallest ball including the
basic 7 balls is to be called a “stellar aggre-
gate of order 1.” An aggregate of order 2 is
achioved by enlarging an aggregate of order 1
Jn the ratio 1/rm7 and by replacing each of
the resulting Balls of radius 7 by a replica of
the aggregate of order 1. In the same way, an
aggregate of order 3 is achieved by enlarging
an aggregate of order 2 in the ratio 1/
and by replacing each ball by a replica of the
agerepate of order 2. And so on,
Ta sum, Between two succesive orders of
Aggregation, the number of points and the
radius are enlarged in the ratio 1/r=7. Con
sequently, whenever R is the radius of some
aggregate, the function Mo(R) expressing the
number of points contained in a ball of radius
Riis Mo(R)=R, For intermediate values of Ry
a7
M(R) is smaller (reaching down to R/7), but
the overall tend is Mo(RJER,
Starting trom aggregates of order 0, itis
also possible to interpolate by successive
stages to aggregates of orders ~1, 2, and so
on, The fist stage replaces each aggregate of
order O with an image of the aggregate of
order 1, reduced in the ratio 1/7, and so
forth. If one does so, the validity of the rela-
tionship Mo(R)eR is extended to ever smaller
values of R. After infinite extra- and interpo
Tation, we have a self-similar set with
Dalog 7 /log 7=1
We may also note that an object in
espace for which D=1 need not be a straight
Tine nor any other rectfiable curve. Tt need
rot even be connected. Each D is compatible
with any lesser oF equal value of the topologi-
cal dimension. In particular, since the doubly
infinite Fournier universe is a totally discon
nected “dus,” its topological dimension is 0
DISTRIBUTION OF MASS:
FRACTAL HOMOGENEITY
The step from geometry tothe distribution of
mass is obvious. If each stellar aggregate of
order O is Toaded with a unit mass the mass
MGR) within 4 ball of radive R> is identical
to Mo(R), hence «R. Furthermore, o generate
aggregates of order —1 from agerepates of
frder 0 amounts to breaking up a fall that
had been viewed as unifora, and finding it to
‘be made of seven smaller ones, The stage ex:
tends the rule M(R)=R below R=.
‘When viewed over the whole 3:space, the
resulting mass distribution is grossly inom
geneous, but over the Fournier fractal itis as
homogeneous a= can be, (Recall Plate 80.) In
particular, any two geometrically identical
portions of the Fournier universe eaery ident
fal masses I propose that such a distribution
fof mass be called fractally homogeneous
“= The preceding definition is phrased in
terms of sealing fractals, but the concept of
Fractal homogeneity is more general. It ap-
plies to any fractal for which the Hausdorft
measure for the dimension D is positive and
Finite, Fractal homogeneity requires the mass
carted by a set to be proportional tothe ses
Hiausdorif measure. =
FOURNIER UNIVERSE VIEWED AS
CANTOR DUST. EXTENSION TO Dyt
1 wust the reader was not distracted by the
sasual use of fractal terminology in the open
ing seotions ofthis chapter. ti abvious that,
without being aware of the Tact, Fournier was
traveling. along a track paraliel to that of
Cantor, his contemporary. The main differ
nce is that the Fournier construction is im-
bedded in space instead of the line. To further
improve the resemblance, it suffices to change
Fournier's aggregates from being balls to be-
ing bricks (filled-in cubes). Now, each aggre
tate of order O isa brick of side 1, and it i
sludes 7 aggregates of side 1/7; one of them
has the same center as the inital eube, and
the other six touch the central subscuares of
the faces of the original cube
Later we will examine how Fournier ob
tains the value DI from basic physical phe
nomena, and how Hoyle obtains this same
value. Geometrically, however, = is a spe
cial case, even if one preserves the overall oc
tahedron and the value N=7. Since the balls
do not overlap, 1/r ean take any value be-
tween 3 and infinity, yielding M(R)=RP, with
iog 7/log (1/r) anywhere betwosn O and
Jog 7 /log 3=1.7712.
Purther, given any D satisfying D3, it is
easy fy changing N to construct variants of
Fournier's model having this dimension
‘THE CHARLIER MODEL AND
OTHER FRACTAL UNIVERSES
‘The above constructs share every one of the
characteristic defects of first fractal models
‘Most conspicuously, just like the Koch cusve
‘model in Chapter 6 and the Cantor dust mod
fl in Chapter 8, the Fournier model isso ree-
ular as to be grotesque. AS corcectve, Char
lier 1908, 1922 suggests that one allow N and
+ to vary Feom one hierarchical level 1 anoth-
‘er, taking On the values Nip and Fry
‘The scientific eminence of Charlier was
such that, despite the praise he lavished on
Fournier, writing inthe leading scientific lan
‘guages of the day, even the simple model soon
became credited to its mous expositor in
stead ofits unknown author. It vas much die
‘cussed in its time, in particular in Selety
1922, 1923a, 19238, 1924, Furthermore, themodel attracted the attention of the very in
Ayential Emile Borel, whose comments in Bo
rel 1922, while dry, were perceptive. But from
then on, aside from fitfl revival, the model
Tall into neglect (for not very convincing rea
sons noted in North 1965, pp. 20-22 and
4408-409), Nevertheless, it refuses to die. The
basi idea was independently reinvented many
times to this day, notably in Lévy 1930. (See
the LEvY entry in Chapter 40.) Most impor-
tant, the Fractal core notion of the Fournier
universe is implicit in the considerations about
turbulence and galaxies in von Weizsicker
1950 (see Chapter 10), and jn the model of
the genesis of the galaxies due to Hoyle 1953,
which will be disused momentarily
‘The basic fractal ingredient is also present
in my models, Chapters 32 10 35.
Th this light, the question arises of whether
8 model of galaxy distribution ean fail to be a
Fractal with one or two cutoffs I think aot. IP
tone agrees that the distribution must be scal-
ing (for reasoas to be elaborated in Chapter
11) and that the set on which matter concen-
trates isnot a standaed sealing st, it must Be
a fractal set
Granted the importance of sealing,
Charlier nonsealing. generalization of the
Fournier mode! is ill-inspired. <2 Incidentally,
i allows 10g Nm /log (L/t) to vary with m
between two bounds; Dipin?O and Omox <3.
We have here yet another theme; effective
dimension need not Have a single value, and
may drift betwoen an upper and a lower limit
This theme ie picked up senin in Chapter
FOURNIER'S REASON TO EXPECT D=1
We now describe the impressive argument
that leads Fournier 1907, 103, to conclude
that O must be eausl wo 1. This argument is 3
strong reason for nat forgeting its author.
Consider a galactic aggrezate of arbitrary
‘order with mast Mand radius R. Using with:
fut misgivings « formula applicable to objects
with spherical symmetry, assume that the gra
Vitational potential on the surface is GM/R
G being the gravitational constant). A star
falling on this universe impacts with the ve
locity V equal to (2G8/R)"
‘To paraphrase Fournier, an important con-
clusion may be drawn from the abservation
that no stellar velocity exceeds 1/300 of the
velocity of light. tis thatthe mass comprised
Within a world Ball increases as its radius, and
not as its volume, or in other words, that the
Abensity within » world ball varies inversely as
the surface ofthe ball. To make this clearer,
the potentil at the surface would be always
the same, being proportional to the mass and
inversely proportional to the distance. And as
2 eonsequence, stellar velocities approaching
the veloity of light would not prevail in any
part of the universe,
HOYLE CURDLING: THE JEANS
CRITERION ALSO YIELDS D=1
A hierarchical disteibution also arises in a
theory advanced in Hoyle 1983, according 0
Which galaxies and stars form by a cascade
process starting witha uniform gas.
‘Consider a gas cloud of temperature T and
mass Mo, distributed with a uniform deasity
Throughout a bal of radius R. As shown by
Jeans. a “critical” situation prevails when
Mo/RoJKRT/G. (Here, k isthe Boltzmann
constant ané J a numerical coefficient.) Ta
this critical case, the primordial gaseous cloud
ic unstable and must inevitably eontzact.
Hoyle postulates (2) that Mo/Ro takes on
iis critical value at some initial stage, (8)
that the resulting contraction stops when the
volume of the gas cloud drops to 1/25:th,
tnd (@) that each cloud then splits into five
louds of equal size, mass Mj-=-Mo/S, and
qual radius Ry=Ro/5. Thus the process ends
a it started: in an unstable situation followed
by a sevond stage of contraction and subd
sion, then a thied, and so on. But cusdling
Stops as clouds become so opaque that the
heat due to gas collapse ean no longer escape
‘As in the diverse other fields where the
same cascade process is encountered, 1 pro
pose thatthe five clouds be called curds, and
that the cascade process be ealled carding
|AS said when T introduced this last term, 1
sould not resist its juxtaposition with palacte
Fournier injects N=? to facilitate the
fsraphical illustration, but Hoyle claims that
Ne has & physical basis. In another otrast
with Fournier, whose geometrical illustretion
is detailed beyond wha is reasonable oF need
fd, Hoyle i vague about the curds" spatial
scatter. An explicit implementation as 10
wait until we describe random curdling in
Chapter 23. But these discrepancies do. not
matter; the main fact is that r=1/N, 30 that
Dal must be part of the design if curdling is
tw end as it begen in Jeans instability
Further, ifthe duration of the frst stage is
taken 10 be 1, gas dynamics shows that the
mth stage’ duration is 5°". I follows that
the same process could continue to infinity
‘within # total ime of 1.2500.
EQUIVALENCE OF THE FOURNIER
AND HOYLE DERIVATIONS OF D=1
AA the edge of an unstable gas cloud satisfy
ing the Jeans eriterion, the velocity and the
temperaiuze are linked by VE/2aJT, be
‘case GM/R is equal to V2/2 (Fournier) and
to JKT (Jeans), Now recall that in statistical
thermodynamics the temperature of & gus is
proportional tothe mean square velocity ofits
Molecules. Hence the combination of the
Fournier and Jeans ctiteria suggests that at
the edge of a cloud the velocity ofthe fall of a
macroscopic object s proportional to the aver
age velocity ofits molecules. A careful analy
Sis of the role of temperature in the Jeans cri-
terion is Bound to show the two criteria to be
‘equivalent. <2 Most likely, the analogy ex
tends to the M(R)=R relationship izhin gal-
nies, reported in Wallenguist 987. =
WHY D=1.23 AND NOT D=1?
The’ disagreement between the empirical
1.23 and the Fournier and Hovle theret90 FRACTAL VIEW OF GALAXY CLUSTERS
cal Del raises on important issue, P. J. E
Peebles tacked it in 1974 by relativity theory.
See Peebles 1980, a full treatment of the
physict and of the statistics (Out not of the
foometty af this topic,
‘THE SKY'S FRACTAL DIMENSION
The sky is projection ofa universe, in which
every point is first described by ite spherical
svordinates p, 8, and @ and then replaced by
the point of spherical coordinates 1, , and g
‘When the universe is a fractal of dimension D,
and the origin of the frame of references be
Tongs to the universe (see Chapter 22), the
structure of this projection is “typically”
ruled by the following alternative: O>2 im-
plies that the projection covers a nonzero pro
Portion of the sky, while D<2 implies tht the
Projection is self of dimension D. =a As ex
femplified in Plates 95 and 96, txpcal allows
for exceptions, due to the structure of the
fractal and/or the choice of origin. Tt often
means “true with probability 1." =
ASIDE ON THE BLAZING SKY EFFECT
(WRONGLY CALLED OLBERS PARADOX)
‘The rule in the preceding section bears dre
Ty upon the motivation that led diverse writers
(including Fournier) to variants of @ fractal
Universe. They recognized that such vniverses
“exorcise” geometrically the ‘Blazing. Sky
Epfect, often (but ¥ronsly) called Olbers
1
paradox. Under the assumption that the dis
tribution of eslestial Boties is uniform, mean
ing that D=3 for all sale, the sky i lit near
uniformly, during the night and during the
day, to the brighness of the solar dive
‘This paradox is no longer of interest to
physicists, having been eliminated by telat
ty theory and the theory of the expansion of
the Universe, and other arguments, But its
demise left a peculiar by-product: numerous
commentators invoke thei preferred explana
tion of the Blazing Sky Effect as an excuse
for neglecting clustering, and even an argu
ment for denying its reality. This is truly
‘odd viewpoint: even if galaxies meed not be
clustered to avoid the Blaring Sky Effect,
they are clustered, and this characteristic de-
‘mands careful study. Furthermore, as seen in
Chapter 32, the expaasion of the Universe is
compatible not only with standard homogenei=
ty but also with fractal homogeneity
‘The Blazing Sky argument is simplicity
itself. When the light emitted by a star is pro-
portional to its surface area, the amount of
Tight reaching an observer at a distance of
ReL/R2, but the stars apparent surface is
itsel’ <1/R?. Thus, the apparent ratio of
Tight to spherical angle is independent of R.
‘Also, when the distribution of stars in the
Universe is uniform, almost any direction in
the sky sooner or later intersects some stat
‘Therefore, the sky is uniformly bright, and
seems ablaze. (The Moon's dis would form
fn exceptional dark domain, at least, in the
Absence of atmospheric diffsion.)
‘On the other hand, the assumption thatthe
cy
ives is eactal with D<2 resolves the paras
doe In tha case, the aniverses projection on
the sky ba fava withthe same D, ence a
{ref Yro aren. Even i the sar a6 given &
nonzero radius, «large proportion of direc
Tom go to infty without encountering any
Star, Ang these tirestion, the might sky
Black. When the range in which D3 i fl
lowed by 2 range in which D=3, the aks
backround isnot stristiyBlack ut lumina
ed entremel funy.
“he Blaving Sky Bifect was noted by
Kepler shorty ater Galles Sideral
Message bas fommented favorably on the
notion thatthe Universe unbounded. In his
T6l0" Comverseton with the. Sidereal
Mestenger Kleene "You do nat hs
tate so. declare that there ace vibe over
10,00 sas. IF this tee ad theses,
tavel the sume nature as out son, hy donot
these sins eollcctvelyouidinance our sum in
brilliance? But maybe the intervening ether
dbscirs then? Not inthe feast. Its gute
dlear the. hs world of our doc nt long
toan uniferentited swarm of countless oe
sin" (Rosen 1968, pp. 38-35)
This conchsion remained sontovesal,
but the argument was not forgotten, witness
the comment by Eamind Halley. in 1720,
that “Another Argument | have heard urged
that if the number of Fint Stare mere more
than finite, the whole seers of thir ap
patent Sphere woul be lino" Late ts
Zonclsin wap discussed by De Chsetey and
1H. Lambert, ut came to be sree fo
Gauss reat fond, Obes. Tae term
“Olbers paradox” that became attached toi!
fs scandalous but symptomatic. Observations
that had been rejected into the “unclassified
residuum" (page 28) become all 100 ofter
Credited to the first Establishment figure whe
decorates them by 2 classifiable wrapping
however transient. Historical discussions are
found in Gamow 1934, Munitz 1957, North
1965, Dickson 1968, Wilson 1963, Jaki 1969,
Clayton 1975, and Harrison 1981
ASIDE ON NEWTONIAN GRAVITATION
‘The Rev. Bentley kept pestering Newton with
an observation closely related to the Blazing
Sky Effect if the star” distribution is homo
geneous, the force they exert on one Among
thom is infinite. One may add that their gra
‘tational potential is infiaite, And that aay
Aistribution wherein M(R)=R® for large R
Yields an infinite potential unless D1. The
Imodesn theory of potentials (Frostman ‘theo:
£5) confirms that thers is a privileged link
between Newton's gravitation and the value
1. The Fourier and Hoyle derivations of
Del cannot fail to be seated to ths link
“Fournier's theme of “the gravitational po-
Cental at the surface being always the same”
is contal to modern potential theory. =
ASIDE ON RELATIVITY THEORY
Fo paraphrase de Vaucouleurs 1970:
Relativity theory led us co believe that to bein! FRACTAL VIEW OF GALAXY CLUSTERS
optically observable, no stationary. material
ball can have a radius R less than the
Schwarzchild limit Ry=2GM/c2, where ¢ is
the velocity of light. Tn a plot of the mean
density p and the characteristic radius of
‘various Cosmical systems, pyy = 3¢*/BeGRy?
{efines an upper limit. The ratio o/ong may be
called the Sebwarzehild filing Tactor. For
Inost common astronomical bodies (stars) or
systems (galaxies), the Filling factor is very
sinall, on the ofder of 10-* to 10°®." The
Square of the velocity ratio postulated by
Fournice is (800)"*~10", precisely in the
range middle of the above. =
AN AGGLUTINATED FRACTAL UNIVERSE?
Many authors think one may explain the gen
cis of stars and other celestial abjects by an
ascending cascade (i.e, the agglutination of
areatly dispersed dust particles into inereas-
ingly bigger pieces) rather’ than by
descending cascade & Ta Hoyle (ie, the feng
mentation of very large and diffuse masses
‘nto smaller and smaller piece).
‘An analogous alternative arses in connec:
ion with the cascades postulated in the study
of turbulence, Chapter 10. Richardson's cas-
‘ade descends toward ever smaller ees, but
ascending cascades may also be present; soe
Chapter 40, under RICHARDSON, Thus it may
be hoped that the interrelations between de
sconding and ascending cascades will be eari-
Fied soon,
FRACTAL TELESCOPE ARRAYS
To wind up this discussion, nothing can be
‘more appropriate than a comment about the
tools used to observe the galaxies. Dyson 1977
suggests that it may be advantageous to re-
place one piece telescopes by arzays of small
telescopes. The diameter of each would be
bout 0.1m, equal to the patch size of the
smallest optically significane atmospheric dis-
turbance, thei centers would form a fractally
Iierarchical pattern, and they would be linked
by Currie interferometers. A rough analysis
leads to the conclusion that a suitable value
for the dimension would be 25. Dyson's con-
clusion: “A Skilometer array of 1024 ten-
centimeter telescopes connected by 1023
interferometers is not a practical proposition
today. [Ie is offered} as a theoretical ideal,
show what can be done in principle.”
‘SURVEY OF RANDOM FRACTAL
MODELS OF GALAXY CLUSTERS
‘one grants the claim thatthe distribution of
eslanies is described usefully by unknowingly
fractal models of limited subtlety and versatl
ity, one should not be surprised that knowing
|y fractal random models provide even more
tveful desriptions. To begin with, ovr under
Sanding of Hoyle curdling improves when it
is-set in its proper context: random fractals
(Chapter 23), OF greater significance, I think,
are the random models T developed and dis
fuss in Chapters 32 to 38. One reason for
8
Aielling on several models is hat improve
trent in the gualty of description i “paid
for" by inereased complication A second ea
fon i that cach model valves a fractal dust
that deserves attention, Let me survey these
inode het, out of logical order.
‘Around 1965, my ambition wes to imple
ment the relationship M(R}«R° with O<3
with a model in which thee is no esntr of
the universe” irs achieved this goal by the
fandom walk model described in Chaplet 32
Then, aan alternative, { developed a tema
Imodsh, which consist in cating out from
Space a collection of mutually independent
randomly placed tremas of random rad,
aning up to an upper cutoff L that may be
ier finite or infin.
Since both models had been selected solely
on the basis of formal simpli, it was der
Tightftty surprising to diveover they hve pre
dBctve valu My theorsticaleorciatonfne-
tions (Mandelbrot 19754) agree withthe
urvestited nes reporid in Peebles 1980
{Gee pp. 243-249). More precisely my two
Approaches agree on the 2pointeorelation,
ty random walk yields good 3-poim corre=
lation and a bad 4-pintcorslaton, and my
spherical temas model is very good for all
known correlations, m= ve
Unfortunately, the appearance of samples
generated by ether mole sue unveaistic.
Using « notion tht developed for thi very
purpose and describe in Caper 35, they have
unaceepiable lacunaity properties. For the
trema mode! this tet is coreted By into
dling more elaborate ema shapes For the
random walk model, T use a less lacunae
lbordinator.”
Thus, the study of galaxy clusters hat
greatly stimulated the development of Fractal
seometry. And today the uses of fractal geom
elty in the study of galaxy’ clusters go well
beyond the tasks of streamlining and house:
leaning accomplished in the present chapter
CUT DIAMONDS LOOK LIKE STARS
‘And the distribution of raw diamonds in the
Earth's crust resembles the distribution of
Sars and galaxies in the sky. Consider a large
‘world map on whieh each diamond mine or
diamond rich site—past or present-is repre
sented by pin. Where examined from fat
away, these pins" density i extraordinarily
lancuen. A feo are lated here and there, but
most concentrate in a few Blessed (or ac
cursed) areas, However, the Earth surface in
there areas is not uniformly paved with dae
rmonds. When examined more closely, any of
these areas turns out itself to be mostly blank,
With scattered subreas of much greater di:
‘mond concentration, The process continues
ter several orders of magnitude,
Is it not irsesistible to inject curdling in
this context? Indeed, an unknowingly Trectal
model has been advanced by de Wij, as seen
Under NONLACUMAR FRACTALS in Chapter 39,Plate 95 « PROJECTION OF FOURNIER'S MULTIUNIVERSE (DIMENSION D~0.8270)
This plate represents to seale both the projec-
tion and the “equatorial” section of & Uni-
verse of dimension DI described in the tex.
Sce alo Plate 96,
‘To paraphrase the caption in Fournier
1907: “A. multiuniverse constructed upon a
cruciform or octahedral principle is not the
plan of the world but is useful in showing that
{an infinite series of similar successive univer
5 may exist without producing’ “blazing
Sky." The matter ih each world sphere is pro
portional to its radius. Ths is the condition
Fequired for fulfiling the laws of gravitation
4nd radiation, In some dizections the sky
would appear quite black, although there is an
Infinite succession of universes. The “world
ratio’ inthis ease is Ne? instead of 1022 a¢
in reality.”
Tn the seme described in Chapter 34, a
universe with Dal and Na1022 is of very
Tow lacunarity, but extraordinarily stratified
-
Plate 96 « A FLAT FOURNIER UNIVERSE WITH D=1
Plate 95, being drawn to exact sale, is not
dnly har 10 print and to ses bet potenti
tiseading. Indeed, ite no universe Of ee
inenson D1 but js plana projeton, whose
dimension Delog'5/iag 7=0.8570<1
Therefore, inorder to avoid fesvng he wrong
Impression, we hasten to exhibit a regular
Fournierlike planar pattern of dimensior
Del. The construction, which involves
1/r25 instead of 1/e=7, is carted one step
further than is possibie in Plate 95. ma