You are on page 1of 23

J. theor. Biol.

(1982) 94555-577

Analytical Description of Growth?


R. SKALAKS, G. DASGUPTA AND M. Moss
Department of Civil Engineering and Engineering Mechanics,
Bioengineering Institute and Department of Anatomy, Columbia
University, New York, U.S.A.

E. OTTJZN AND P. DULLJZMEIJER


Zoological Laboratory, State University of Leiden, Netherlands
AND

H. VILMANN
Department of Anatomy, Royal Dental College, Copenhagen,
Denmark
(Received 10 March, 1980, and in revised form 7 August, 1981)
Methods for the mathematicaldescription of growth are reviewed and
extended. The main modeof descriptionis that of identifying the material
point paths(in space-time)of the masspointswhich makeup the animalor
plant. The spatial location and massof the elementary particles of the
animalare consideredto be continuousin space-time.Discontinuitiesare
allowed in displacementfunctions to allow for slip, as in a knee joint;
discontinuities are also allowed in material properties in the senseof
suddenadhesionor non-adhesionof surfacesand in the topology, suchas
the completion or opening of a ring. When cells proliferate or atrophy
throughout the tissuegrowth isconsideredasdistributedsourcesor sinksin
space-time.Depositionor resorptionon the surfaceof a boneisconsidered
to be a surface distribution of masssourcesor sinksin space-time.The
theoretical framework isillustrated by idealizedexamplesof mathematical
descriptionsinvolved.

1. Introduction
Mathematical methods for the description of growth have developed
steadily since the 1940’s but have accelerated particularly in the last decade.
The purposes of present paper are to review developments to date and to
t Thisresearchwassupportedby National Institutesof Health Grants HD-14371-01 and
DE-0154502 (M. L. Moss) and HD-14371-01 (R. Skalak).
$ Send correspondence to: Professor Richard Skalak, Director, Bioengineering Institute,
Room 610, SW. Mudd Building, Columbia University, New York, NY 10027, U.S.A.
555
0022-5193/82/030555+23$02.00/O @ 1982 Academic Press Inc. (London) Ltd.
556 R. SKALAK ET AL.

suggest some extensions, particularly for the description of surface growth,


and the use of allometric networks in conjunction with finite element
methods.
The basic view pursued in the present paper is that in principle, any finite
growth or change of form may be regarded as the integrated result of
differential growth, i.e. growth of the infinitesimal elements making up the
animal or plant. The notion of distributed growth has been previously
developed by many authors, often in the literature of plant growth as well as
animal growth. One of the earliest papers relating growth rates and changing
form was illustrated by the tobacco leaf in the work of Richards & Kavanagh
(1943, 1945). These authors introduced the ideas of a velocity field and a
distributed growth field as means to discuss change of form. These ideas of
distributed field variables were not much farther developed until the 1970’s
although relative growth rates were extensively discussed by such authors as
Teissier (1960) and Huxley (1932).
Distributed growth can be described conveniently by the use of the
vocabulary of continuum mechanics, both of fluid mechanics and solid
mechanics, Continuum analysis of growth by mapping functions such as used
in solid mechanics was discussed by Archer & Lockhart (1970) and
developed further in a treatment of growth stresses in trees (Archer &
Bymes, 1974; Archer, 1976). The analogy to fluid mechanics has been most
extensively developed in relation to plant growth by Erickson (1976a, b)
and Silk & Erickson (1978, 1979).
The finite changes of dimensions and alterations of form which take place
during ontogeny as well as phylogenetic differences have been analyzed for
planar cross-sections of skulls in the work of Bookstein (1978). He computes
elongations or stretch ratios which must be variable and distributed
continuously over the cross-section to produce the observed growth. The
present paper suggests the appropriate vocabulary to extend such studies to
three dimensions.
The mathematical description of distributed growth in three dimensions is
summarized in section 2. Most of the ideas and correlations to continuum
mechanics have been noted in the articles cited above. They are summarized
in section 2 by way of background and they are cast in the vocabularies of
both solid and fluid mechanics to bring out the relations to the continuum
mechanics literature.
In section 3, the kind of description used in section 2 is extended to
appropriately describe surface growth, as usually occurs in bone growth. The
particular method of description suggested is new and its use is illustrated by
examples which are not actual biological structures, but illustrate the
analysis involved.
ANALYTICAL DESCRIPTION OF GROWTH 557
Section 4 discusses possible discontinuities which may arise and section 5
gives some idealized examples which illustrate the method of description of
surface growth.
Section 6 discusses allometric growth and network models. This dis-
cussion is intended to help bridge the gap between the continuum approach
of sections (2) and (3) and the finite dimensions or lengths which are often
described and correlated in the biological literature. It is further suggested
that the gap may most readily be bridged by adaptation of finite element
methods which are currently in wide use in engineering and continuum
mechanics. This suggestion has already been introduced in the growth
literature by Niklas (1975) and Mauseth & Niklas (1979). In the present
discussion, these ideas are tied into the concepts discussed in the earlier
sections.

2. Volumetrically Distributed Growth


Suppose that at a time t1 some region R1 of space (see Fig. 1) is occupied
by the animal or part of an animal. It is assumed that every point in the
animal can be identified and located at the time ti with respect to Cartesian
orthogonal axes, ai (i = 1,2,3). If there is growth so that the animal occupies
another region R2 at a later time, tz, a complete description of the growth is
assumed known if all points of the region Rz are located relative to another
set of fixed Cartesian, orthogonal coordinates xi (i = 1,2,3). The xi axes may
be coincident with al, az, a3 but this is not necessary. One possible descrip-
tion of the growth is attained if the co-ordinates Xi of any point P2 at t2 are
known in terms of the co-ordinates Ui of the same material point PI at time

FIG. 1.
558 R. SKALAK ET AL.

tl. Let the transformation from RI to Rz be described by giving the final


co-ordinates xi in terms of the initial co-ordinates ai and r. Thus

Xi = Xi(Uj, t). (1)

The particular region described by equation (1) when (a i, u2, uj) traverse R 1
is the region R2 at fZ. If we have the transformation equation (1) for all f, we
have a complete description of the growth R 1. It is assumed in the concept of
distributed continuous growth, that the functions of equation (1) are
generally continuous.
It should be noted that to have a complete model, the mass density p at
every point of space and time must also be known. Sufficient information is
available if the density p is known as a function p(ai, t) or p(x,, t). If the basis
of equation (1) is accepted as a suitable description of growth, then a large
body of ideas, definitions, proofs and methods of the mechanics of finite
deformation theory of continuum mechanics can be adapted to the present
purposes. In the following, for brevity, only the results of interest will be
cited. Proofs are available in texts such as Fung (1965) or Green & Zerna
(1968) on large deformation theory.
A suitable measure of the extensions at any point due to growth is given
by what is usually called the material strain tensor in solid mechanics.
Analogously, in the present case, we define a growth extension tensor eij by

(ds)2 - (dSo)2 = e,. da, da,


(2)
(dS)2 ” ’ ’

where dSo and dS are the initial and final lengths of a line element dUi. The
definition (2) leads to (cf. Fung, 1965)

1 &Q dXk
eijz- ---6, . (3)
2 ( aUi dUj >

In equations (2) and (3) the usual summation and range conventions are
used. Indices which occur only once in each term, like i, j in equation (3) are
assumed to have the range 1,2,3. A repeated index like k in equation (3)
implies summation of the terms with k = 1,2, 3. The 6, is the Kronecker
delta; 6, = 1 for i = j and Sij = 0 for i # j.
Since eij is a symmetric second order tensor with nine real components (by
definition (3)), there are at any point, three principal axes of the growth
extension tensor and the formulas for finding them are given (for the strain
tensor) in Fung (1965). In principal axes, the growth extension tensor takes
ANALYTICAL DESCRIPTION OF GROWTH 559
the form
el 0 0
eij = 0 e2 0 (4)
I 0 0 e3I

where el, ez, e3 are the principal values. The directions of the co-ordinate
axes which give the form of equation (4) can be established at every point of
the space R2 and a system of orthogonal curves can be constructed, which
completely fill the space (three times) by defining each curve to be tangent to
one of the principal axes at each point along its lengths. These curves will be
called the principal growth trajectories, similar to the biorthogonal grids
introduced by Bookstein (1978). Along growth trajectories, it is convenient
to define growth extension ratios which are defined as the ratio of the final
length of a line element dS to its initial length dSo. Thus extension ratios Ai
are
Ai = (2ei + l)l” (i = 1,2,3). (5)
An alternate description can be given in terms of the velocities of the
material points. When velocities are used, the extensive vocabulary of fluid
mechanics including streamlines, sources, sinks and other useful ideas may
be directly adapted. This viewpoint has been quite fully developed by
Erickson (1976a,b), Silk & Erickson (1978, 1979) and by Cox & Peacock
(1978, 1979). Velocity fields are more attractive for some purposes because
they embody the instantaneous growth rate rather than a finite change as
implied by use of the growth tensor.
The velocity field associated with the growth expressed in equation (1) is
given by
dXi
lJ,i =z= Vi(Uj, t)

where the Ui are held fixed during the differentiation. Equation (1) can be
solved, at least in principle, for the xi in terms of the Q. Substituting for the ai
in equation (6) will yield the velocities in terms of the current co-ordinates xi.
This gives the spatial description usually used in fluid mechanics.
In fluid flow, a common method of depicting a motion is by the use of
streamlines which are defined as curves tangent to the instantaneous velocity
vector at each point (Batchelor, 1967). An example of streamlines in growth
is given by Cox & Peacock (1978).
The gradient of the velocity field forms a second order tensor which
contains the full information of the local rates of growth and of the rate of
560 R. SKALAK ET AL.

rotation experienced locally (cf. Silk &Erickson, 1978, 1979). The gradient
of velocity is as/ax,. Its symmetric part is the rate of deformation tensor dij
and the antisymmetric part is the vorticity slij defined by

(7)

The rate of deformation tensor dij contains complete information on the


growth rates at a point. The rate of extension of any line element is (seeAris,
1962, or Richards & Kavanagh, 1943)

&z= &j&n,. (9)

The principal growth rates di represent the rates of growth in three perpen-
dicular directions and include the maximum (d,) and minimum (d,) growth
rates at the point.
An important invariant of the rate of deformation tensor d, is the trace dii.
It can be shown that this is the rate of change of volume per unit volume at a
point. This will be called the rate of volumetric growth or rate of dilatation,
dii. ThUS

where V. v stands for the grad v in vector notation and V is the volume of a
small element of the tissue at the point under consideration. The intro-
duction of the rate of dilatation equation (10) leads naturally to considera-
tion of the growth of massof the tissue (Silk &Erickson, 1978, 1979). In the
general case in which a tissue may be growing in size and at the same time
changing its density the massrate of growth per unit volume is

ap a
g=,+-&(Pvi).
I
It can be recognized that when the specific growth rate g is set equal to zero,
then equation (11) becomes the usual equation of conservation of massin
fluid mechanics.
In the above discussion,it was implied that the massrate of growth, g, was
positive. It may also take on negative values which would correspond to
atrophy or resorption. All of the equations and interpretations given above
hold true in this case also.
ANALYTICAL DESCRIPTION OF GROWTH 561
A recent example of three-dimensional growth in which both linear
growth rates and volumetric growth rates are computed is given by
Hejnowicz & Nakielski (1979). This study of the shoot apical dome assumes
axial symmetry and illustrates the equivalent of equations (7) and (11) in
cylindrical co-ordinates.

3. Growth by Deposition or Resorption on a Surface


Consider a region RO which contains a surface SO(as shown in Fig. 2). At
time t,, the region R,, is divided into two parts, RI and R2, by the surface SO.

(0) (b)

so nl

“2
01
02

%
(cl

FIG. 2.

Suppose that at time to a surface growth on both sides of So starts. The


surface So may be regarded as a suture in a skull and may have different rates
of growth on each side. If the regions R1 and R2 are assumed not to have any
growth in them, they move as rigid bodies. The growth on So gives rise to a
relative velocity of RI and Rz and the growth creates two new regions R3
and Rd.
562 R. SKALAK ET AL.

In the general case shown in Fig. 2c, growth on the curved surface SO at
any time cl is specified by a vector growth rate G1 on one side and G2 on the
other. The Gi and GZ may be functions of position on SO and of time. The
vectors G1 and GZ allow for growth which takes place at any angle to the
normal n to SO as shown in Fig. 2c. The directions of G1 and G2 may be
regarded as those of the scleroblasts producing the growth. The magnitudes
G1 and GZ have units of mass rate per unit of surface area So (gr/cm’ set).
The surface SO itself may move and change size and shape with time.
Further, regions RI, Ra, R3, R4 may have distributed growth in them.
The situation of growth from two sides of a surface also occurs in trees.
The cambial growth in trees has been so modeled by Wilson & Howard
(1968) with xylem cells formed on the inner side and phloem cells to the
outer (bark) side of the sheet of “initials.”
In seeking a mathematical description of the above situation, an unusual
circumstance arises in that the regions R3 and R4 added by the growth on SO
do not have initial co-ordinates at time cl since the mass in them did not exist
at time cl. This difficulty of assigning initial co-ordinates may be remedied by
use of a new kind of initial or material co-ordinate system. In principle, it is
always possible to assign surface co-ordinates which will be called ([,, &) to
locate points on So (cf. Aris, 1962). Since S, and SZ originated off So, the
same co-ordinates may be used to identify points on S, and SZ. In fact,
material co-ordinates may be defined throughout R3 and R4 as ([,, &, 7)
where T stands for the time at which the particular portion of the tissue was
created. The surfaces 7 = constant are loci of all material points which were
created at the same previous time 7. The range of T is t, < T < t. The
descriptions of the particle positions in R3 and R4 at any time t are of the
form
x1 =x1(51,52, 7. t) (12)
x2 = x2(51, 52, 7, t) (13)
x3 = x3(511 52, 7, f). (14)
Each region, R3 and R4, will have a separate set of functions like equations
(12)-( 14) to describe its growth. The regions R 1 and R2 are described by sets
of functions of the form of equation (1). The surface S, may itself grow and
move. Its location at any time t is included in equations (12)-( 14) as the
surface obtained by setting 7 = t. Let these co-ordinates be x,~, x2O, x3,,.
Then S,, is given by

x10 = Qd51, 62, t) (15)


x20 = x20(5*, 52, ?I (16)
ANALYTICAL DESCRIPTION OF GROWTH 563
x30 = x30@-*, 22, d. (17)
In equations (15)-(17) the &, 6) are surface co-ordinates which may be
regarded as the co-ordinates of the cells producing the growth or resorption
on So.
The velocities of the material points in R3 and R4 are found by differen-
tiating equations (12)-(14) with respect to time, t:
fJ1= &(51,52, 7, c) (18)

02 = SZ(Sl, 52, 7, t) (19)


u3 = a,(&, 52, 7, 0 (20)
where the dot represents differentiation with respect to time holding
(er, &, t3, T) constant. Assuming equations (12)-(14) can be inverted to
solve for (&, t2, 7) in terms of (x1, x2, x3, t) the velocity equations (18)-(20)
can then be expressed in terms of (x1, x2, x3, t) as usual.
The velocity of the points on So which generate the new growth may be
found by differentiating equations (15)-(17) with respect to time. These
velocities Vi of the surface may be different from equations (18)-(20) which
are the material particle velocities. The surface velocities are
v1= ho(h, 52, t) (21)
v2 = ~20(51,~2, d (22)
v3 = ~30(Zlr 52, t) (23)
where the dot represents differentiation of equations (15)-(17) with respect
to time holding &, 5 2) constant. In general, the material points move away
from the surface of growth by the velocity V. given by G/( p cos 0). where 8
is the angle between Vo and n. When So itself is moving, Vo is the relative
velocity of the material points with respect to the surface So. Hence, at every
point of and on each side of So it is required that
(v-V)=VG=G/(pcos8) (24)
where v is the material point velocity given by equations (18)-(20).
Let II be a normal vector taken positive pointing away from So on each side
of So. The normal component of G may vary on each side of So indepen-
dently as functions of [i and f. The tangential components of (v-V),
equation (24), may be continuous, but could be discontinuous, i.e. take on
different values on each side of So to allow for the slip of a joint or for slip of
two tissues relative to each other during the growth process. Such slip may
occur in the skull between the bone plate and the brain tissue during growth.
564 R. SKALAK ET AL.

The velocity field Vi of material points in regions RS and R4 in Fig. 2 may


be used to find the streamlines of the motion and will be related to the
distributed rate of growth tensor dij and mass growth rate g as discussed for
the more usual case in section 2.
If there is volumetrically distributed growth in the regions R3 and R, after
their initiation by growth on So, there may be some growth extension within
these tissues of the type that was considered in section 2. To investigate the
growth extensions at a point, say P with co-ordinates (g-i, &, r), it is first
required to find the initial co-ordinates of this material. To do this, the time t
in equations (12)-(14) is set equal to 7p. These equations then give the
material particle locations when the particle of interest was first formed.
Hence for investigation of particle P, take the initial co-ordinates ai to be
a1 =x1(4+, 52, 7, T&J (2%
a2 = x2(&, 52, 7, TJ (26)
a3 = X3(51, 52, 7, T,,). (27)
Now, in principle, it is possible to eliminate (ti, t2, T) between equations
(12)-(14) and (25)-(27) to arrive at the material point description of the
form.
Xl = X1(&, a27 a39 t, 7p) cw
x2 = X2(&, a29 a39 t, Tp) (29)
x3 = x3(4, a2r a39 t, Tp). (30)
The transformation equations (28)-(30) are the co-ordinates xi at time t that
result when the co-ordinates Ui at time t = TV are used for initial co-
ordinates. The transformation equations (28)-(30) are of the form of
equation (1) when we regard TV as a fixed parameter. Then all the results in
section 2 apply here also. The growth extensions derived will be those
occurring since t = TV
The above discussion assumed, as shown in Fig. 2c, that accretion (or
resorption) takes place on both sides of a surface S,, with growth rates Gi and
G2. For most bones, the great preponderance of the surface area is not in
contact with another bone, but is in contact with some other tissue which
probably grows due to sources distributed in space. In such cases, the above
analysis holds with one growth rate on So, say, Gz equal to zero. The bone is
then located on the Gi side of So and the soft tissue in contact with the bone
must at least match the normal velocity of S,,, i.e.:
(v-V).n=O (31)
ANALYTICAL DESCRIPTION OF GROWTH 565
where v is the material particle velocity of the soft tissue in contact with S,,
and V is the velocity of SO. If the soft tissue and bone are not permitted to
slide over each in any tangential direction then equation (3 1) is replaced by
the stronger condition
v-v=o. (32)

4. Discontinuitles
The above formulations assume that location of any mass particle is a
continuous function of time and space. The later locations of initially
neighboring particles however, may be discontinuous. The sliding of the
surface produces a discontinuous tangential component of displacement
across the slip surface. A similar situation may occur in growth where one
tissue grows at a different rate or in a different manner than its neighboring
tissue. Another kind of discontinuity arises when a tissue is divided by the
intrusion of another ingrowing tissue.
If a tissue is a closed figure such as a sphere or a closed ring, then any
growth which results in opening the ring or making a hole through the sphere
also leads to discontinuous displacements of points which were formerly in
touch. The topology is also discontinuous in this case.
The discontinuities of displacement mentioned above are readily
incorporated in the descriptions discussed by assigning different location
functions in each region of interest.
It should be noted that although displacement may be discontinuous, no
holes or empty spaces are anticipated. In this sense all results are continuous.
A water or air-filled cavity provides a continuous mass distribution if the
fluids are considered part of the total system. Even air spaces such as the
external ear canal or the nasal passages and lungs which are connected to the
surrounding air may be considered part of the internal space of the animal.
This may be useful for purposes of describing the growth of these passages
and spaces.
In the early stages of embryonic growth it often occurs that when two
tissues first touch or sometimes when two parts of the same tissue touch each
other, there is a fusion or adhesion and a sudden change of growth rate. To
allow for this kind of sudden change, discontinuities of properties and
functions must be introduced in the theory.
A common variety of discontinuities are those of the growth rate tensor or
the mass rates of growth g and G. In uiuo, these rates probably never change
instantaneously, but may change abruptly, in a matter of minutes or hours
when the entire growth takes weeks or years. Then analytically, it may be
convenient to introduce discontinuous growth,rates.
566 R. SKALAK ET AL.

A common type of functional discontinuity is the fusion of two surfaces


which were previously not attached. Thus in an embryo, a muscle and a bone
may form and grow separately. At a certain instant they are joined, after
which the organism has more capability of motion. It is the capability or
organization which is discontinuous. Modeling may proceed with two
separate descriptions, before and after the discontinuity. The disjoining or
separation of a tissue along a certain internal surface is the reverse of
adhesion and may be similarly modeled as a discontinuity of properties in
time.
Topological changes during growth will generally be of two kinds, First, a
rearrangement of existing parts by fusion or separation may lead to a new
topology without requiring any new growth. A second kind of change
induced when a new tissue first forms and must be recognized as a new
region, like R3 and R, in Fig. 2b. To have a mathematical description of a
change in topology when no new growth is involved, it will generally be
sufficient to relabel the surfaces and volumes involved and to proceed with
the new system taking into account the continuity of location and mass of the
material points involved.
Changes in topology which are initiated by a new tissue starting to grow on
an interior surface, like S,, at t = ti in Fig. 2b, have already been analyzed in
section 3. In each case, the initial region R. must be considered as divided
and treated separately after the time r 1. The description is topologically
discontinuous in time.
In terms of the vocabulary of differential topology, the situation may be
readily summarized by noting that continuum theory is valid when cor-
responding regions are homeomorphic, but singular points or surfaces in the
co-ordinate transformations will generally produce one or more of the kind
of discontinuities discussed above.

5. Surface Growth Examples


The following examples are intended to illustrate simple cases of surface
growth. They are not biologically realistic in detail, but show the nature of
functions involved. It is to be expected that some combination of analytic
methods described here, approximations, and numerical, computerized
methods will be necessary to describe growth of actual living forms
realistically.

(A) GROWTH OF A TWO-BONE APPENDAGE

Consider two straight bones of the same cross-section having lengths I,


and l2 of the regions RI and R2 at t = 0, Fig. 3a. Assume growth starts on S,,
ANALYTICAL DESCRIPTION OF GROWTH 567
02, x2

(a) (bl
FIG. 3.

at t = 0 with mass rate Gi on the left and G1 on the right of So. The new
growth is shown shaded in Fig. 3b, regions R3 and Rd. For all four regions,
the x2, x3 co-ordinates remain fixed so x2 = u2, x3 = a3 throughout. The x,
material point paths are given by:
x1 = a1 in R, (33)
x1 = 11+ VG1T in R3 (34)
xl=I,+VG,t+VG2(t-7) in R, (35)
xl = %+(VGl+ vG2b in R2. (36)
The left end of RI is assumed to be fixed. The al and T in equations (33)-(36)
are material co-ordinates; Ii, V,, = G,/I and vo2 = G,/p are constants. It
follows from equations (33) and (34) that the particle velocity is zero in RI
and R3. Differentiating equations (35) and (36) with respect to t shows that
the particle velocity is (V,, + VG2) for material points in R2 and R.,. The
location of the boundary S, is derived by setting T = t in equations (34) or
(35). Either of these then gives the location of So as
x10= I,+ VGlt and hence VI = VG,. (37) and (38)
The growth extensions eij are everywhere zero, but the length of the first
bone Lr and the length of the second bone L2 increase with time:
L, = II + v,,t, L2 = 1,+ vG2t. (39)
This example serves to demonstrate a case in which the velocity of the
anatomical point (So) given by equation (38) differs from that of the material
points in any of the four regions, RI-Rd.

(B) HORNS AND SPIRAL SHELLS

Consider initial regions RI and Rz which are cones with bases of radius ho
in the (ai, u2) plane and cone heights lo, Fig. 4a. Suppose that at t = 0,
568 R. SKALAK ET AL.

(0) (b)

FIG. 4.

deposition begins on So with a constant angle a of the surface growth


velocity VG to the (a,, u2) plane and parallel to the (ai, u3) plane. The
regions R3 and R, are mirror images and only R, will be considered.
Assume the growth velocity has magnitude V, given by

v =G+G G,-G,

---6x (40)
G 2p sin ff 2p sin LY
where (Y is the angle of Vc to S,,, G, and G, are functions of time, and e1 is a
material co-ordinate equal to (al/h,) at t = 0. If G, and G, are suitably
assigned and the surface So is expanded to a larger circle in a proportionate
manner, a horn or sea-shell shape of logarithmic-spiral form can be
generated as shown in Fig. 4b. The following steps lead to consistent growth
rates and spiral shapes.
Suppose at a general time t, the center of the logarithmic spirals is at C and
a total angle of 8 has been produced as shown in Fig. 4b. This angle 0 is a
function of t. A material co-ordinate 7 is introduced as in section 3. The
surfaces T = constant are given by 4 = constant (see Fig. 4b). Thus 4 = (b(7)
and serves to mark the time of deposition of a given mass point. The angle
between the radius vector and a tangent to the spiral is a and we set
ANALYTICAL DESCRIPTION OF GROWTH 569
k = cot a. Finally, for convenience, assume the initial radius r. is given. In
terms of these variables, the co-ordinates X& &,7, t) are
Xl = roe Ire-(ro-&ho)eke cos (e-4) (41)
x2 = &hoe” (42)
~~=(r~-~~h~)e~‘sin(0-~). (43)
For any fixed 8 and C#B, the co-ordinates &, & range over *l (over a unit
circle). getting T = t in equations (41)-(43) gives 4 =8 and this yields the
surface So as indicated in section 3. Thus So is given by
x1 = &hoeke (44)
k0
~2 = &hoe (45)
x3=0. (46)
The magnitude of the growth velocities G1 and G2 are
G1 = (ro - ho)pe Irei (47)
G2 = (r. + ho)pe keti (48)
It is thus seen that growth rate of the circle So given by equations (44)-(46) is
proportional to the growth rates Gi and G2 at the inner and outer edges of
the horn. The distribution Vc in equation (40) is linear in 6 as it must be
since it measures the difference of velocity of two plane surfaces.
The horn generated by equations (44)-(46) will be such that every curve
on which & and t2 are constant will have a projection on the (x1, xg) plane
which is a logarithmic-spiral. This includes the bounding curves shown.
The description given above has been given, in effect, by the analysis and
discussion of D’Arcy Thompson (1942) and Huxley (1932). The form of the
description given here shows that specifying the shape of So and the
distribution of the growth vector G controls the form generated. This
viewpoint allows certain general statements to be made in a concise form.
For example, if plano-spirals and helical-spirals are considered, and o is a
vector along the axis of the spiral, then plano-spirals require G . o = 0, but
turbo-spirals are generated when G . w # 0.
When So is plane and R4 (Fig. 4) is rigid, the magnitude of G must vary
linearly over So. It is to be expected that this distribution can be affected by
the physiological distribution of nutrients and by stresses applied. The
surface So need not be plane, but can be of any shape in space. To generate
spirals, the growth on So need only be directed in the correct spiral direction
and vary appropriately in magnitude.
570 R. SKALAK ET AL

A narwhal’s tusk can be generated on a plane circular So with the G


vectors directed along axial spirals. Then G is along the axis on the axis, but
slopes increasingly with radius. On a circle of given radius, the G vectors are
tangent to a cone of this radius and at equal angles to So. The circle So must
also grow in the correct proportion. At any time the magnitude G is constant
over S,,. This keeps the tusk growing straight.

6. Allometric Growth and Network Models


If Ii and l2 are two different dimensions of a growing body, or correspond-
ing lengths in different bodies, the growth is said to be allometric if
l2 = bl: (49)
where b and k are constants. The relation (49) may be expressed
parametrically as the result of exponential growths of the form
1, = d,ekgl (50)
l2 = dze kZ’ (51)
where d,, dz, kI, k2 are constants and t is the time or pseudo time. (A pseudo
time is any monotonically increasing function of time to be selected such that
equations (50) and (5 1) hold.) It follows that k = k,/k,.
The possibilities for allometric growth are limited. For example, suppose
that in Fig. 5 the lengths 1, and l2 of two sides of the triangle ABC grow
allometrically with the angle 8 constant. Then it is readily seen that the third
side BC has length 13:
1, = [I: + 1: - 21,1, cos ey2. (52)
If 1, and 1, are of the form of equations (50) and (51) then 1, cannot be
represented by a similar form with constant k3. Equation (52) will give an
allometric growth only if k, = k, in which case k = 1 and the triangle ABC in
Fig. 5 grows gnomonically, that is, into strictly similar triangles.

FIG. 5
ANALYTICAL DESCRIPTION OF GROWTH 571
It is possible to have allometric growth in three dimensions in three
different directions which are not in one plane. An example of this follows.
Consider the motion given by
x* = u,ek” (53)
x2 = aZek2’ (54)

where k,, kZ, k3 are constants and r is the time or pseudo time. For this
motion, the principal axes of strain and strain rate are parallel to the
co-ordinate axes and the strains and strain rates are the same at every point.
If the density is assumed constant, the growth rate is found to be
g = dkl + kz + kd. (56)
Since there are, in general, a set of principal axes at each point, it may
be said that locally, in some neighborhood of space and time equations (53)-
(55) represents the most general growth possible at any given point and time.
It is of interest to see to what extent allometric growth can be generalized.
One possibility is that the principal axes of strain trajectories remain fixed
curves in ai space and the rates of deformation along each trajectory is a
constant. Then the lengths measured along the corresponding curves in xi
space are allometric in the sense equation (49). The value of k may vary for
neighboring trajectories so that this system would permit a variety of shape
changes.

(A) AN ALLOMETRIC NETWORK MODEL

Another way to achieve some degree of allometry is to assume there exists


a network of points (in a plane or in space) such that the length of each line
between points obeys equation (50) with its own k value. In relation to Fig.
5, it was pointed out that 1,, 12, 1, cannot all grow allometrically if 8 is fixed.
But if the line elements are allowed to rotate, then each line may grow
allometrically. The future geometry of the triangle is determined if the initial
size and shape is given and the growth rates of the three sides are known.
In Fig. 6a is shown a plane network of points and lines which divide the
plane up into triangles. If each element is assumed to grow allometrically,
then the future geometry is determined or may even be overdetermined. It
can be shown that the criterion for arbitrary growth rates to be permissible
for every line segment of a plane network is

m=2j-3 (57)
572 R. SKALAK ET AL.

m=ll j=7 m=l2 /=7


zj-3111 2/-S-11
m=Zj-3 ma2/-3
Determinate indeterminate

(0) (b)

m=l5 j=7 m=16 j=7


3j-6=15 3/-6115
m=3j-6 m>3j-6
Determinate Indeterminate

(c) Cd)

FIG. 6.

where m = number of line segments and j = number of nodes (joints).


Equation (57) is the same criterion as required for a framework or truss to be
statically determinate and can be derived by similar methods (see DiMaggio,
1963; Timoshenko & Young, 1965; Calladine, 1978). If m > 2j - 3 (Fig.
6b), the network is indeterminate. Growth rates of each member cannot be
arbitrary and must to some extent be related. If m < 2j - 3, the network is
unstable. Specifying the growth rate of each member will not yield a unique
future shape.
In Fig. 6c, a space network is illustrated. It can be shown that the growth
rates of each line segment can be arbitrary if
m =3j-6. (58)
The rule equation (58) is the same as for the statical determinacy of space
trusses and was given by Maxwell in 1864. Cauchy (1813) studied the
ANALYTICAL DESCRIPTION OF GROWTH 573
stability of networks of triangles like Fig. 6b. Euler (1758) also derived some
interesting results on geometry of polyhedra. Caspar & Klug (1962)
consider a virus coat problem of similar nature. Some networks that might
be used to simulate biological structures in 20 and 30 are shown in Figs 7
and 8 respectively.
The use of allometric networks such as shown in Figs 5-8 is attractive
because they would lead naturally to the adaptation of finite element
methods to the study of growth patterns. For example, if finite displacements
of points A, B, C are given, interpolations to derive the displacements of
interior points are provided by finite element approximations. For example,
a curvilinear quadralateral (in a plane) or a curvilinear parallelepiped (in
space) may be used to good advantage. If each side grows allometrically and
the intersections of the edges are at right angles, these grids become similar

I27 (a)
m=25
j=14
2j-3=25
m=2j-3
Determinate
w(b)
m=27
j=15
2j-3=2?
m=Zj-3
Determinate

m=21

Determinate

r 2j-3=37
m=2j-3
Determinote
(d)

FIG. 7.
574 R. SKALAK ET AL.

m= 3j-6
Determinate

FIG. 8.

to the biorthogonal grids of Bookstein (1978). Examples of the use of the


finite element method in large deformations due to stressmay be found in
Zienkiewicz (1977). This methodology may be adapted more or lessto the
study of growth. Examples of the use of finite elements in the description of
growth have been discussedby Niklas (1977) and Mauseth & Niklas (1979).
The rule equation may be expressed in terms of the degrees of freedom of
a linked set (chain) of rigid bars as defined in the kinematic theory of
mechanisms (seeHunt, 1978). The degrees of freedom, f, are given (in 30)
by

F?6(m-j-l)+ f fi (59)
i=l

where m and j are numbers of bars and joints as above, and fj is the number
of degrees of freedom of the ith joint. When f = 0, a system of rigid bars
could not move, but independent growth rates of the length of each bar are
kinematically permissible. The condition f = 0 equation (59) is equivalent to
equation (58).
In a network representing moving parts of an animal (like a fish’s head and
jaws), several types of joints besidespoint connections may be appropriate;
ANALYTICAL DESCRIPTION OF GROWTH 575
there may be more or less degrees of freedom. Further, stresses and
deformations applied by surrounding tissues may influence the observed
geometry in Go. During growth new connections or changes in degrees of
freedom may occur.

(B) A CENTERED ALLOMETRIC MODEL

A pattern of growth which may be readily visualized is that allometric


growth occurs along every radius emanating from one point of space. This is
possible analytically with all points moving on their fixed directions from the
center 0. Equation (50) would apply to every radial line segment, but the
value of ki would be a function of the direction i.e.
r = ree k(n,.n,.n,)r (60)
where r. is an initial radius and nl, nz, n3 are its direction cosines. Any
continuous function k of (its, n2, n3) may be used in equation (60). This
model will give allometric growth along any ray from 0, but not exactly
allometric growth for other line segments unless k is constant. If k does not
vary much, other line segments will grow very nearly allometrically. The
possibilities of a centered allometric model have been recently discussed
with respect to cranial growth by Moss et al. (1981). Another recent
exploration of the use of a centered growth model has been discussed by
Pittenger, Shaw & Mark (1979) and Todd et al. (1980). These authors have
shown characteristic perceptual information for age level of faces can be
embodied in an approximate distortion of outlines by a particular radial
growth function.

7. Discussion, Future Directions


The basic methods of description in sections 2 and 3 form a framework
within which growth and deformation may be discussed in regard to the
kinematics involved. These schemes are useful to investigate what the actual
growth patterns are in terms of distributed mass sources in space and on
surfaces.
After the basic patterns of growth are known it is of interest to investigate
what simplified patterns (such as discussed in section 6) give close approxi-
mations of real growth patterns. When this phase is in hand, questions
concerning such items as the relative rotation or displacement of one tissue
due to the growth of another can be quantitatively addressed and described.
The systems proposed will be useful to describe interactions of growth and
stress systems such as remodeling of bone or growth under stress as in the
branch of a tree (cf. Wilson & Archer, 1979; Cowin & Van Buskirk, 1979).
576 R. SKALAK ET AL.

When the kinematic description outlined above is complete it is reason-


able to look next for the physiological control mechanisms which determine
the kinematic growth observed. The growth response itself must be part of
the entire system which regulates growth as well as being the end product.
This larger system must involve mechanical stresses and many non-
mechanical influences such as chemical, neural, electrical, hormonal, nutri-
tional factors (see Goss, 1979).
An important aspect of broader studies must be the interaction of forms
and functional components in growth. Discontinuities and changes of inter-
linkages are to be understood from a functional viewpoint as well.
Allometric growth rate changes require functional explanations. The
kinematic modeling described in the present paper should be considered as a
prelude to quantitative functional and causative study.

REFERENCES
ARCHER. R. R. & LOCKHART, J. A. (1970). Continuum Analysis of Growth By Mapping
Functions. Proc. 23rd ACEMB, p. 81.
ARCHER, R. R. & BYRNES, F. E. (1974). Wood Sci. Technol. 8, 184.
ARCHER, R. R. (1976). Wood Sci. Technol. 10,293.
ARIS, R. (1962). Vectors, Tensors, and the Basic Equations of Fluid Mechanics. Englewood
Cliffs, New Jersey: Prentice-Hall.
BATCHELOR, G. D. (1967). An Introduction to Fluid Dynamics. Cambridge: Cambridge
University Press.
BOOKSTEIN, F. L. (1978). The Measurement of Biological Shape and Shape Change. (Lecture
Notes in Biomathematics, vol. 24). New York: Springer-Verlag.
CALLADINE, C. R. (1978). Jnt. J. Solids Structures 14, 161.
CASPAR, D. L. & KLUG, A. (1962). Cold Spring Harbor Symp. Guam. Biol. 27, 1.
CAUCHY, A. L. (1913). J. Fat. Polyt. 9, 87.
COWIN, S. C. & VAN BUSKIRK, W. C. (1979). J. Biomechanics 12,269.
Cox, R. W. & PEACOCK, M. A. (1978). J. Anaf. 126,555.
Cox, R. W. & PEACOCK, M. A. (1979). J. Anat. 128,207.
DASGUPTA, G., SKALAK, R., Moss, M. L. & MOSS-SALENTIJN, L. (1979). A Model for
Allometric Growth. Proc. Third Engineering Mechanics Division Specialty Conference
A.S.C.E., 17-19 Sept. 1979, University of Texas, Austin, Texas, pp. 497-500.
DI MAGGIO, F. L. (1963). Proc. Am. Sot. Civil Engineers 89-ST3, 63.
DULLEMEIJER, P. (1974). Concepts and Approaches in Animal Morphology. Assen, Nether-
lands: Van Gorcum.
ERICKSON, R. 0. (1976a). Ann. Rev. Plant Physiol. 27, 407.
ERICKSON, R. 0. (1976b). In: Descriptive and Theoretical Studies in Automata, Languages,
Deuelopment (P. Lindenmeyer and G. Rosenberg eds). Amsterdam: North-Holland.
EULER, L. (1758). Novi Comment. Acad. Sci. Imp. Petropol. IV, 109 seq. (ad annos 1752 et
1753).
FIROOZBAKHSH, K. & COWIN, S. C. (1980). J. Riomechanical Engrg 102,287.
FUNG, Y. C. (1965). Foundations of Solid Mechanics. Englewood Cliffs, New Jersey: Prentice-
Hall.
Goss, R. J. (1979). ‘Ihe Physiology of Growth. New York: Academic Press.
GREEN, A. E. & ZERNA, W. (1968). Theoretical Elasticity. Oxford: Oxford University Press.
ANALYTICAL DESCRIPTION OF GROWTH 577
H~NOWICZ, Z. & NAKIELSKI, J. (1979). Modeling of Growth in Shoot Apical Dome. Acta
Socieratis Boranicorum Poloniae. 58, 423.
HUNT, K. H. (1978). Kinematic Geometry ofM&hanisms. Oxford: Oxford University Press.
HUXLEY, J. S. (1932). Roblems of Relative Growth. London: MacVeagh.
MAUSETH, J. D. & NIKLAS, K. J. (1979). Am. J. Bat. 68,933.
MAXWELL, J. C. (1864). Phil. Msg. 27,294 (Paper XXVI in Collected Papers, Cambridge,
1890).
Moss, M. L., VILMANN, H., DASGUFTA, G. & SHAL.AK, R. (1981). In: CraniofacialBiology
(D. S. Carlson ed.). Monograph 10, Craniofacial Growth Series, University of Michigan, Ann
Arbor, pp. 61-81.
Moss, M. L., SKALAK, R., MOSS-SALENTIJN, L., DASGUFTA, G. M., VILMANN, H. &
MEHTA, P. (198 1). Proc. Finnish Dent. Sot. 77,119.
NIKLAS, K. J. (1977). Ann. Bat. 41, 133.
PIT~ENGER, J. B., SHAW, R. E. &MARK, L. S. (1979). J. exp. Psychol. (Hum. Percept.) 5,478.
RICHARDS, 0. W. & KAVANAGH, A. J. (1943). Am. Nat 77,385.
RICHARDS, 0. W. & KAVANAGH, A. J. (1945). In: Essays in Growth andForm (W. E. 1eGros
Clark and P. B. Medawar eds). Oxford: Clarendon Press.
SILK, W. K. & ERTCKSON, R. 0. (1978). Am. J. Bol. 65,310.
SILK, W. K. & ERICKSON, R. 0. (1979). J. theor. Biol. 76,481.
TEISSIER, G. (1960). In: The Physiology of Crustacea (T. H. Waterman ed.). New York:
Academic Press.
THOMPSON, D. W. (1942). On Growth and Form. 2nd edition. Cambridge: Cambridge
University Press.
TIMOSHENKO, S. P. & YOUNG, D. (1965). Theory of Structures. 2nd edition. New York:
McGraw-Hill.
TODD, J. T., MARK, L. S., SHAW, R. E. & PII-IXNGER, J. B. (1980). Sci. Am. 242, 106.
WILSON, B. F. & HOWARD, R. A. (1968). For. Sci. 14,77.
WILSON, B. F. & ARCHER, R. R. (1979). Bio. Sci. 29,293.
ZIENKIEWICZ, 0. C. (1977). The Finite Element Method. 3rd edition. New York: McGraw-
Hill.

You might also like