0% found this document useful (0 votes)
159 views11 pages

A better way to construct the sunflower head

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
159 views11 pages

A better way to construct the sunflower head

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

A Better Way to Construct the Sunflower Head

HELMUT VOGEL
Technische Unicersitiit Mikehen, Lehrstuhl fiir
Physik,
D-8050 Freising- Weihemtephan, Federal Repiblic of Germany

Receiwd 23 October 1978

ABSTBACI

The flower or seed pattern in a sunflower head can be approximately constructed by


the method of Davis and Mathai [I], which has inspired the cover of this journal. This
construction reflects, besides the well-known spiral arcs, two kinds of irregularities, one of
which is also observed in natural heads, the other not. The latter kind of irregularities is
shown to be due to the semiempirical construction employed and can be avoided by a
more pertinent, wholly mathematical construction program, which predicts the whole
pattern, including the irregularities that are seen in nature, and only these. In particular,
concise formulae are given for the points where visible arcs of different order begin and
end, and for the concomitant locations of irregularly shaped junction seeds.

INTRODUCTION
The flower heads (capitula) of most compositae consist of a great
number of individual flowers, whose arrangement has intrigued observers at
least from 1754 on [2]. The most conspicuous feature of this pattern is two
sets of arcs, one running clockwise, the other counterclockwise, and com-
posed of particularly tightly packed seeds. These arcs are often (wrongly)
described as logarithmic spirals [3-51. Their number in each set is invariably
a member of the Fibonacci sequence 1, 1,2,3,5,8,13,21,. . . ; e.g. 21 and 34
for a small sunflower capitulum, up to 89 and 144 or even 144 and 233 for
big ones.
The conspicuous arcs are not the basic structural elements, but merely
consequences of a single primary growth spiral on which, in the developing
capitulum, new individual flowers are added from the center (the
meristema), each issuing at an angle of 137.5” with respect to the preceding
one. This 137.5” angle occurs also in the arrangement of leaves or side
branches around the main stem of many plants, especially compositae,
where it leads to a near-periodicity in Fibonacci distances: The 5th branch
is nearly above the 0th; the 8th and 13th are even nearer to it. Sometimes

MATHEMATICAL BIOSCIENCES 44: 119-189 (1979) 179


QElsevier North Holland, Inc., 1979 0025-5564/79/040179+ 1lf02.25
180 HELMUT VOGEL

one can pursue this sequence up to the 55th branch, e.g. with the steppe and
dump plant mugwort (Artemisia). Thus, the growth program does not have
to change from the stem to the capitulum as far as angles are concerned,
but the longitudinal growth is relieved by a lateral one.
For convenience of description, I will count the flowers or seeds outward
from the center, i.e., contrary to their seniority. In polar coordinates r, cp,
flower number k is at q+ = k8 (6 = 137.5”) and at r, = r(rpk). What the lateral
growth program [i.e., the function r(q)] is and how it produces the flower or
seed pattern seems still to be controversial. Almost any r(q) explains the
existence of the arcs. Finer features are more discriminating, e.g. the facts
pointed out by Richards [6, 71 and Davis, Mathai [I]: The arcs do not run
all the way to the center, but are successively relieved by other systems of
arcs, fewer, but always Fibonaccian, in number; at the junctions of different
arc systems, irregular seed shapes arise. Davis and Mathai [1] proposed a
graphical construction that well describes these features, and a computa-
tional procedure that claims to do so. The former is based on the assump-
tion that the area provided for individual flowers or seeds is the same all
over the capitulum, the latter on the assumption that the primary spiral is
logarithmic (r = r,e”v) and because of its flatness (small a) approaches an
Archimedes spiral rmr,( 1 + acp). Confirmation is derived from the construc-
tion, not from the calculation, which I will show to contradict that construc-
tion and the facts.
I will provide a growth program that is biologically reasonable and
predicts in a quantitative and concise form the whole pattern of the
sunflower head down to the most minute detail.

WHY 137.5”?
The angle of 137.5”, common to stem and capitulum, can be considered
as the consequence of a very simple growth program. We look at the stem in
its axial direction and mark the side branches as points on a circle. Between
them, in general, there remain gaps of different sizes. The program then
prescribes:

(1) Each new branch (leaf, flower) bears the same spatial relationship to
the existing stem and accordingly issues at a fixed angle 6 with respect to
the preceding branch.
(2) Each new branch fits into the largest still-existing gap between older
branches, cutting a constant fraction off that gap.

Condition 2 tends to uniformize the distribution of branches over the


circle (more precisely, it tends to a big-scale uniformity; microscopically,
there may always remain different gap sizes). The question is, what angle 6
satisfies condition 2.
BETTER WAY TO 181

We express S in turns, i.e., S=z2a. A rational z would lead to a


repetition, after which persisting gaps could not be filled. We establish a
series of rational z,, converging to the real value of z, starting with zO= f
(aa= 120”) and refining successively as follows: At a given z, =p,/q, (p,,, qn
relatively prime), there is a period of q, branches, which form exactly pn
complete turns, the distance Z~ between temporally consecutive branches
being divided into pn equal parts. Branch qn coincides with branch 0. We
now correct this by slightly increasing z to zn+ i, such that branch number qn
now cuts a fraction zn+, off the old gap between branch 0 and its nearest
angular neighbor (old length of this gap: z,/p,, = l/q,,). After this correc-
tion, the distance between consecutive branches is z~+ ,; the q,,th one lies not
at pn complete turns as before, but at p, +zn+,/qn =q,,zn+l. From this,

Pn+l W”
zn+,=-=
%+I q,z-1 ’
(1)

which is generally true, starting from z,,= f, if and only if pn and qn are
consecutive Fibonacci numbers of even order:

P,, = F2m qn= &u+2-

(1) then becomes equivalent to F2v+4F2v= Fiu+2- I, which follows from


F2u+3 F2v+1 =F2 2v+2+ I, and this in turn can be easily confirmed by complete
induction with respect to v.
At any refinement stage n, the branches fill the whole circle uniformly,
and each new one fits into one of the largest gaps remaining between the
older ones. This remains valid in the limit n+cc, zn+g2=i[(fi - 1)12=
0.382, i.e., i&+137.5”, the Fibonacci angle.

THE LATERAL GROWTH PROGRAM r(cp)


The condition of equal areas for all seeds fixes the function r(cp) and,
together with the Fibonacci angular distances, defines the whole pattern of
the capitulum. It is easy to see that an Archimedes or, a fortiori, a
logarithmic spiral would not provide equal areas. A big capitulum with K
seeds has radius r, and area nri. In order to provide equal spaces for K
seeds, this area must be proportional to K-i.e., since cp= KS, proportional
to ‘p:

r(cp)=a*. (4

This spiral is essentially identical to the orbit of a nonrelativistic charged


particle in a cyclotron, in which the energy E = imv2 increases by equal
steps at each passage through the electric field, i.e., E is proportional to cp,
182 HELMUT VOGEL

FIG. 1 The primary cyclotron spiral connecting consecutive seeds that have a
Fibonacci angular distance of 137.5”.

whereas the angular velocity w in the magnetic field is constant: mv2/r=


evB, i.e., w= v/r = eB/m, and hence ra vcc V% a &$ . For lack of a
better term, we will call the spiral (2) a cyclotron spiral (see Fig. 1).
With an Archimedes or a logarithmic spiral, the seed areas would
diminish absurdly towards the center. On the other hand, the decelerating
r(q) in the cyclotron growth program seems anatomically more reasonable
than the Archimedes or the logarithmic one, which are bound to terminate
much more abruptly when the final capitulum size is attained.

THE ARC STRUCTURE


We now discuss the secondary arcs, which are the most prominent
feature in a capitulum. They are composed of nearest neighboring flowers
BETTER WAY TO CONSTRUCT THE SUNFLOWER HEAD 183

or seeds. From (2) for the distance dk, between the centers of seeds k and I
we have, according to the cosine theorem,

dk:
aZ=k+l-2flcos(k-I)S. (3)

This is small whenever (k - 1)6zrn*2ir (with integer m) or

x=k-lxs: 2’
&27T (4)
g ET2 .

Since x = k - I is integer, this holds best if x and m are Fibonacci numbers

x=F,, m= Fn_2.

This explains the fact that in an arc, neighbors have always numbers
differing by a Fibonacci x. Of course, (4) becomes an exact equality only in
the limit, when F,_,/ F, attains g2. The difference x&m-277 is

1.073
Gg2- F,-z= 2,rFn_2. (5)

In the interesting region, one can thus expand the cosine around rna2r:

d2
‘ak+l-2m(l-$$),
a2

or after expanding also m mk(l-x/2k-x2/8k2-x3/16k3) and re-


arranging,

dk:
-= x2 p2k x3 p2 1.073
4k+-+s-p P-- - =2.809 (6)
a2 x2 g2

This has a minimum of

$ = fl = 1.677 at x,,,=m =2.37fi.

Obviously, x cannot always keep this optimal value, since it has to be a


Fibonacci number. With increasing k, the x that dominates the visible arcs
thus runs through the Fibonacci series. The transition from x = F,- , to
x= F, occurs when both provide the same distance d, i.e.,

d2(Fn) = d2(F,- I), whence F,'_ , F,’ = 4/I 2k2,


184 HELMUT VOGEL

i.e., the F, arc arrives at the first rank at seed number

This event is not as conspicuous as the appearance of F, at the second rank


or its disappearance from it, since at any time the two best F, are visible.
The F, arc appears at

because Fn_2Fn= F,‘, , + 1. At the same time, F,_, constitutes an optimum


and provides the minimal distance d, = I .677a. In the neighborhood of k;,
the transition between different spiral systems deforms the seeds most.
In the “life” of any arc, characterized by a number and a seed distance
F,, there are thus four phases, separated by the following events shown in
Table 1. The d entries are obtained by inserting the k limits into (6). These
are averages between the proximal and the distal neighbors of the seed
under consideration. The distal neighbor (the one that leads away from the
optimum) is somewhat farther off than the proximal one (typically 2.10,2.02
or 1.81,1.75). All these k, and k: are in general not Fibonacci numbers. The
arc F, runs through O.l78(F:+, - F:_,) places in the primary spiral and
itself contains s,=O.l78(Fj+,- F~_,)/F,,=O.l78(F,,+,- Fnp,) seeds. s, is
close to the Fibonacci number Fn_*, inasmuch as 0.178 = 1/2bxgz/2, and
F,_, + Fn_,= Fn~z+2F,~3z2Fn~z.
We prove that the arcs are not logarithmic spirals either, but they are of
the same type as the primary spiral, only very much steeper. The F, arc is

TABLE 1

Event Seed number Nearest-neighbor distance

Arc becomes k;=O.l78F,2_, d,&g2+g-q =2.0&l


visible

Arc arrives k,=O.l78F,_,F, d,&(g+g-‘) = 1.780


at 1st rank

Seeds have d, = 1.677~~


minimum distance

Arc declines kn+,=0.178FnF,,+, d,,,j/g = 1.78~


to 2nd rank

Arc disappears k;+,=O.l78F,f+, d,d/ =2.06a


BETTER WAY TO CONSTRUCT THE SUNFLOWER HEAD 185

FIG. 2 A reconstructed capitulum of 959 seeds, based on the cyclotron growth


program. All figures have been photographed from a PDP-12 screen.

composed by the seeds k = k; + mF,, m = 1,2,. . . , F, _ 2 [see (4)]. These seeds


have the coordinates r,,, = avm and ‘pm= ‘pOk rnF;\ [see (5)], i.e.,

One arc extends over an angular distance of about 1, i.e., 60”. Its inclination
with respect to the radius, arctan(rdq/dr), increases from 20” at the inner
end to 67” at the outer one. Thus, the angle between arc and radius changes,
which already suffices to show that the arcs are not logarithmic spirals, for
which this angle would be constant. (Compare the flight of nocturnal insects
around a street lamp: Constancy of angle between flight and light direction,
which normally leads to a straight course, here produces logarithmic
spirals.)
186 HELMUT VOGEL

All these details are confirmed equally well from actual capitula and
from constructions according to the cyclotron program (cf. Fig. 2). Davis
and Mathai’s [l] graphical (not their computational) procedure also fits
reasonably well, apart from inaccuracies which may be due to drawing. The
quasi-identity of the two constructions is not surprising, since Davis and
Mathai also use the Fibonacci angular condition and approximate the
equal-area condition by pushing circles or coins of equal size along their
Fibonacci radii until the new one touches at least one of the previous ones.

OTHER LATERAL GROWTH PROGRAMS


We consider more general primary spirals of the form

(pk = k& rk=aka.

Here,

d2
-$ = k2”+ 12”-2k”l”cos(k- /)a,

and instead of (6), we get

The minimal distance occurs at

namely d,,, = ak”-1/2m .

The cyclotron spiral (a = f) is the only one with d, independent of k; for


the Archimedes spiral (a = l), d, increases like X& towards the periphery.
The factor LYis carried through the ensuing results, so that all seed numbers
for the transitions between different arc systems and the arc lengths
(expressed in seeds as well as in radians) appear twice as big with the
Archimedes spiral as with the cyclotron one or with reality (see Fig. 3).
A logarithmic spiral

(pk= k& r, = r,e bk

yields as a condition for minimal distance

e -bx=l.---_,
P’ P2
bx’ 2x2
BETTER WAY TO CONSTRUCT THE SUNFLOWER HEAD 187

FIG. 3 A capitulum of 959 seeds, based on the Archimedes growth program instead
of a cyclotron one.

or, since always bx < 1,

dm=aebk(b-m)mabe”.

x,,, is independent of k, i.e., there is only one type of arc in either sense,
which runs all the way from the center to the periphery. The distance d,,,
heavily depends on k. Here and only here, the arcs are logarithmic.
Finally, we consider a program that ensures termination of lateral
growth, with still constant frequency of appearance of individual flowers:

qk = k& Iji=a(l-emck),

where k this time has to count in the order of seniority, i.e., small k at the
188 HELMUT VOGEL

rim, big k at the center. One obtains a near-neighbor distance of

d,~a~ee-2ckc2x2+~2x~2(1-e-ck)2 ,

which becomes minimal at

xql~, d,,,za $@~-~~(l -e-“) .

For ck<<l, one obtains of course the Archimedes-spiral results, and for
ck>>l

Seeds with ck>l (i.e., at less than 30% of the final radius a of the
capitulum) would become smaller and smaller, and the arc structure would
become indistinct, since v- runs through the Fibonacci sequence
very quickly.

DISCUSSION
From a teleological point of view, the problem of the sunflower head can
be considered settled. Two simple principles-most uniform angular dis-
tribution of branches and flowers (hence the Fibonacci angle), and most
uniform area distribution of flowers and seeds (hence the cyclotron primary
spiral)-explain the pattern down to the most minute apparent irregulari-
ties. The Fibonacci law permits the angular growth program to be the same
for stem and flower head; the cyclotron law prefigures the transition from a
growing flower head to the final state of this head. What has still to be
provided is a causal mechanism for both growth programs.
Agronomists use the arc structure as an aid in counting the seeds on a
capitulum, e.g. in resolving the crop yield into its components (number of
plants/acre, number of flowers/capitulum, percentage of flowers that bear
seeds, weight of seed, oil content), and in studying the he&abilities of these
components [8]. Irregularities in the arc structure, especially the fact that
arcs do not run all the way through, slightly falsify simple counts according
to the schema (number of peripheral arcs)x(number of seeds per arc).
More important, it is conceivable that regions of abortive flowers (produc-
ing no seeds or empty seeds), which one observes on actual capitula and
which can reduce the yield to 60% or less, are related to the regions where
one arc system is relieved by another [9]. The nature of this relationship,
and whether it is simply due to deformed flowers in the transition regions
being avoided by bees, are under study.

P thank Drs. G. Piquemal, G. Heim and J. Haiech for stimulating discus-


sions.
BETTER WAY TO CONSTRUCT THE SUNFLOWER HEAD 189

REFERENCES

Constructing the sunflower head, Math. Biosci.


20: 117-133 (1974).
C. Bonnet, Recherches sur I’usage des feuilies akns les p/antes, Gottingen and Leiden,
1754.
M. Gardner, The multiple fascination of the Fibonacci sequence, Sci. Am. 220: 116-
120 (1969).
K. Menninger, Zwischen Zuhl und Raum, Ullstein, 1960, pp. 116-121.
E. P. Northrop, Riddles in Mathematics, Pelican, 1960, p. 59.
F. J. Richards, The geometry of phyllotaxis and its origin, Sot. Exper. Biol. Symp.
11:217-245 (1948).
F. J. Richards, Phyllotaxis: its quantitative expression and relation to growth in the
apex, Philos. Trans. Roy. Sot. London Ser. B 235:509-564 (1950).
G. Piquemal, Recherches sur la structure du rendement en grains du toumesol, Ann.
Am&or. Pluntes 18(4):423-446 (1968).
G. Piquemal, personal communication.

You might also like