Anda di halaman 1dari 3

imental

[H+]

IsWvAN LENGYEL AND IRVING R. EPSTEIN


Recent experiments on the chlorite-iodide-malonic acid-starch reaction in a gel
reactor give the first evidence of the existence of the symmetry breaing, reactiondiffusion structures predicted by Turing in 1952. A five-variable model that describes
the temporal behavior of the system is reduced to a two-variable model, and its spatial
behavior is analyzed. Structures have been found with wavelengths that are in good
agreement with those observed experimentaily. The gel plays a key role by binding key
iodine species, thereby creating the necessary difference in the effective diffision
coefficients of the activator and inhibitor species, iodide and chlorite ions, respectively.

and

M,

10-5

M,

M,

that, with the kinetic parameters given in


Table 1, the first term in that rate law is
negligible compared with the second.
Numerical integrations, which reproduce
both the number of oscillations and the
concentration dependences found in the experiments, show that, although the concentrations of C102- and I- vary by several
orders of magnitude during an oscillation,
the concentrations of C1O2, I2, and MA are
slowly varying functions of time; they are
well approximated as constants during an
oscillatory period. By taking [C102], [I2],
and [MA] as time-independent parameters
and making other chemically reasonable
simplifications, we find that the five-variable
model can be reduced to a two-variable
model (7) (X I-, Y C102, A = I2):
=

A-PX rMl kl
kia [MA]0 [I2]I/(klb+[I2]0)
=

for all events taking place after the initial fast


(4)
k,'
consumption of C102- and I- to produce
that eliminates chemical gradients C102 and I2. The C102-I2-MA reaction can
X- Y
rM2 = k2 [X]
and leads to uniform spatial distributions. In be described with a five-variable model con(5)
k2P k2 [C102]0
a remarkable paper entitled 'The chemical sisting of three reactions, whose rate laws
basis of morphogenesis," Turing (1) pre- are known (6):
k3' [X][Y]/(u+[X]2)
4X + Y-e P
rM3
dicted in 1952 that the coupling of diffusion
+ + H+
+ I2IMA
IMA
kinetics
reaction
chemical
with suitable
(6)
k3' k3b [I2]0
could destabilize a homogeneous stationary
d [2]/dt = kla [MA] [I2]/(klb+ [I2]) (1)
A homogeneous system governed by Eqs.
state and result in the formation of tempo6 has only one steady state, which
4
through
rally stationary, spatially inhomogeneous
C102 + I-* C102 + Y2I2
for
certain
values of the parameters loses
structures. He suggested that this process
d[C102]/dt = k2 [CGb2] [I] (2) stability through a supercritical Hopf bifurmight lead to a variety of morphogenetic
cation. The two-variable model is amenable
phenomena. The possibility of such diffusion-induced instabilities has been widely C102- + 4F 4H -+ C + 2I2 2H20 to simple analytical investigation of the necessary and sufficient conditions for Turing
investigated in the chemical and biological
= k3a [C102-] [I-] [H+]
dICl02-]Idt
instability. Similar studies have been carried
literature (2-4).
+ k3b [C102-] [I2] [I-]/(u+[I-]2) (3) out on two similar inhibitory models, the
Only very recently, however, was the first
Degn-Harrison model (9) by Velarde and
dear experimental evidence of stationary
co-workers
in
the
(10) and the Thomas model (11)
reactant
Ithe
Turing structures reported (5). A strip of gel The inhibition (8) by
is
of
12).
Murray
(2,
3
by
for
law
rate
the
term
of
Eq.
second
with
was
fed
indicator
loaded with starch
Our system of reaction-diffision equaconstant concentrations of malonic acid particular significance. Under typical exper(MA) at one end and of chlorite (C102-)
and iodide (I-) ions at the other end. The
pattems, which are nonuniform both paral- Table 1. Model parameters.
lel and perpendicular to the direction along
EA
Value at
Rate or diffusion
Reference
Value at 7C
which the reactants diffuse toward each otheeec
Vlea '
250C
constant
(kj
mol-')
er (the starch is immobile on the gel), were
9 x 10-4
This work
observed in the color of the starch-I3 com- kia (s-1)
81.5
7.5 X 10-3
5 x 10-5
5 x 10-5
plex.
klb (M)
This work
62.0
1 x 103
s-'
6 x 103
Lengyel et al. have recently demonstrated k2 (M-'
120
51.4
)
460
(8)
(M-2
Sk3a
(6, 7) that in a closed, homogeneous reac- k3b (s')
1.5 x 104
(8)
110.0
2.65 x 10O3
1 X 10-'4
tion mixture temporal oscillations in the U (M2)
1 X 10-'4
DA

AILY EXPERIENCE SUGGESTS THAT


diffusion always causes a smoothing

C102--I--MA reaction arise from the sim-

chlorine dioxide (C102)-I2-MA system. The reaction of C102-I2-MA accounts


pler

of Chemistry, Brandeis Univer02254, and Institute of Physical


Lajos University, H-4010 Debre-

I. Lengyci, Department

sity, Walham,

MA

Chemistry, Kossuth

cen,

Hungary.

I. R. Epstein, Department of Chemistry and Center for


Brandeis University, Waltham, MA

Complex

02R5.
650

I05 X DCxh- (cm2 s-')

1.5

0.75

Estimated from
mobility data

105 X D,-* (cm2 s-)


105 x Dc102 (cm2 s-')
105 X DI2 (cm2 S-')

1.4
1.5
1.2

0.70
0.75
0.60

(25)
As C102

Estimated from
values for Iand I3- (25)
Estimated from
0.4
0.8
105 x DMA (Cm2 S-')
mobility data
*In the callations D,- was taken as 5 x 10-7cm2 s-' because of the binding ofl2 and I3- to the gel (see text).
SCIENCE, VOL. 251

Downloaded from on April 2, 2016

Modeling of Turing Structures in the


ChlorodideMalomc Acid-Starch Reaction System

[I-]
[I2]

conditions,

O1-4

1IUU
but, by removing X, it serves as an inhibitor
for
the
Turing
instability.
Stable steady state
k,'
80
The
necessary
condition,
Eq.
13,
requires
, 25/c,
Domain with
+ Dx (a2x/az2)
(7)
that the diffusion coefficients ofthe reactants
60
be different and that the inhibitor (C102-)
ay/at = k2'x - k3'xy/(u + x2)
40
diffuse more rapidly than the activator (I-).
>
1 *-,
!
+ Dy (a2y/az2)
(8) With plausible values for the other paramep ' 3a5-2/a
20
where z is the spatial variable, and Dx and ters, we find that the ratio Dy/DX must be at
Oscillatory state
DY are the diffusion coefficients of x and y, least 10 for Eq. 13 to hold. In aqueous
Uv
'
respectively. It is useful to define the follow- solution, diffusion coefficients of typical
20
40
60
80
0'
100
a
ing dimensionless constants:
small ions are of the order of -` CM2 S-1,
and the diffusion coefficient of C102 can- Fig. 1. Stability diagram of the two-variable
a = kl[MA]W/(k2ulu2 [CG02]0)
(9a) not be ten times greater than that
and the domain of diffiusion-induced instaof I-. In model
bility with DY/DX = 15 in the (a, 1) plane.
general,
diffusion-induced
with
instability
1 = k3b [I2]Wl(k2uu"2 [C102]o
(9b) nearly equal diffusion coefficients (16) can
When applied to Eqs. 7 through 9, the occur only under very special conditions,
necessary and sufficient conditions (13, 14) which are unlikely to be realized in practice.
4.0O
for diffusion-induced instability in a twoIn the Turing structure experiments, the
variable system can be summarized as fol- gel is not an inert medium like the solvent in
lows, where the aii are the elements of the the corresponding aqueous experiment. The ,hx
2.0
_
Jacobian matrix, and ss signifies steady state starch cannot diffiuse in the gel, nor can the
=
{as with a12 [a(ax/at)/ay]}J:
starch-3- complex; their diffusion coeffi1) The homogeneous system must have a cients are zero (17). More importantly, the
steady state {xs, = u1/2a/5, y,, = (u"1/21) [1 polyacrylamide gel is typically 10 to 30% by
1.0
2.0
3.0
+ (aL5)2]} that is stable to homogeneous weight (2 to 6 mol/dm3) in the acrylamide
Distance (mm)
perturbations. This condition is fiulfilled if monomer, which provides a hospitable environment for both I2 and 13-. Specific Fig. 2. Stationary structures in the five-variable
(10) interactions of I2 and 13- with organic model with parameters given in Table 1. Initial
all + a22 < 0
concentrations in the gel were zero for all compolymers such as polyvinylacetate (18) and ponents. Fixed boundary
and
conditions: all concen(in M) are zero at the gel ends, except
(11) starch (19) are well known. Such interaction trations
alla22- a12a21 > 0
with the -CONH2 side groups of the poly- [MA] = 1.0 x 102 at the left and [I2]=
In our case, inequality 11 always holds, and mer is highly likely and would reduce the 1.0 X 103 [CG02] = 6.0 x 10-4 at the right.
inequality 10 becomes
effective diffusion constant of I-. In particular, the relatively low electron density of constants were determined at 25C, but the
(12) the NH2 moiety would facilitate Lewis
p > 3a/5 - 25/a
acid- Turing structures were found at 7C. Ener2) The stability of the homogeneous base interactions with 137. Simple experi- gies of activation (EA) of the component
steady state against inhomogeneous pertur- ments in which C102- and 12 solutions are reaction are available, so rate constants at
bations can be investigated with either Neu- poured through a column of polyacrylamide 7C can be estimated. We were unable to
mann or Dirichlet boundary conditions. gel show that the latter do have significantly find a direct determination of the diffusion
The following inequality is a necessary con- longer retention times.
coefficient of C102-, but from ion mobility
dition for diffusive instability (14):
We may picture the I- ions moving not in data (22) it must be similar to those of
a simple gel but in a medium with many chlorate and perchlorate ions, that is
alIDY + a22Dx >
traps (gel-bound 12, starch) that capture I- =1.5 x 10-5 cm2 s-1 at 25C. With the
for a period of time before relasing it, there- Stokes-Einstein relationship (diffusion coef2[DxDy (a11a22 - a12a21)]"2 > 0 (13) by
causing I- to diffuse much more slowly ficient D = kT/6ir'a, where j is viscosity, a
Comparing inequalities 10 and 13, we ob- than in the absence of the traps. The effec- is the particle radius, and k is the Boltzmann
serve that a,1 and a22 must have opposite tive diffusible concentration and concentra- constant), the temperature (T) dependence
signs. In our case a22 < 0 at any parameter tion gradient of I- are lower than the cor- of the diffusion coefficients can be calculatvalues, whereas a1, > 0 if
responding stoichiometric quantities. Thus ed. The absolute temperature changes only
the
rate of I- diffusion decreases as if its slightly, but the viscosity of water approxia > 5(5/3)1/2
(14) diffusion
coefficient were lower. The re- mately doubles as the temperature drops
In the standard terminology used to dis- quired tenfold difference between the diffu- from 250 to 7TC.
cuss Turing structures (12), the activator has sion coefficients of I- and C102- is easily
The wavelength of the structures can be
a positive sign (a1l > 0) and the inhibitor achieved by this mechanism. In fact, poly- calculated from:
has a negative sign (a22 < 0) in the Jaco- acrylamide gels have been used for the anad = 2r/q
bian; consequently, I- is the activator and lytical separation of halogen-containing spe- where
C102- is the inhibitor. In the chemical cies (20). For reaching a qualitative picture
sense, X (I-) is a self-inhibitor (15) in the (21), we estimated DY/DX= 15. Figure 1
(15)
q = 4(all/Dx + a221DY)112
crucial process (3); that is, the higher the shows the phase diagram in the a-P phase
concentration of I-, the slower the reac- plane obtained from the inequalities 12 Using the parameters in Table 1, we find
tion. However, X indirectly stimulates its through 14. The domain of diffusive insta- that the wavelength is 50.15 mm, which is
own buildup by inhibiting its own destruc- bility broadens as DY/DX is increased.
close to the experimental result (0.2 mm).
tion. Chemically, Y (C102-) acts directly
The parameters used to calculate the Numerical integration of Eqs. 4 through 6
neither as an activator nor as an inhibitor, structures are given in Table 1. The rate confirmed the existence of Turing struc-

tions is

ax/at =

k2'x - 4k3'xy/(u + x2)

8 FEBRUARY 1991

REPORTS 65 1

tures, but the amplitude was only a few


percent of the total concentration of I-,
too small to be observed visually as in the
original

experiment.

To describe the system more realistically,


we must use the five-variable model, Eqs. 1
through 3, because in the experiments the
concentrations of the initial reactants at the
gel boundaries are a key factor in determining the nature (amplitude and wavelength)
of the solution. The relevant case is that in
which C102 and 12 diffuse from one end and
MA from the other. We used a finite difference method with 500 mesh points in the
3-mm gel length. With much coarser meshes
the inhomogeneity (23) (related to the number of peaks) was much larger than with 500
points, but above 400 mesh points the structures were independent of the resolution
and were stable in space and time after 7
hours of reaction time. The wavelength,
location, and number of peaks are much
more sensitive to the diffusion coefficients of
I- and C102- than to those of C102, I2, and
MA. An example of a calculated stationary
structure is shown in Fig. 2.
The experiments of Castets et al. (5) were
conducted on the Cl02--F-MA reaction,
whereas the present model treats the simpler
C102-I2-MA system. As Lengyel et al. have
shown (6), C102 and I2 are formed in the
C102-F- reaction and then dominate the
reaction with MA. However, the generation
of C102 and I2 takes place at some distance
into the gel. Unfortunately, there is not yet
a detailed mechanism that accounts for the
formation of C102 in the C102-F- reaction
at high C102- excess. Recent experiments
(24) show evidence of very similar Turing
structures in a C102-1--MA mixture in the
same reactor, lending support to the contention that the model presented here adequately describes the original system.
In retrospect, the discovery of Turing
structures in the C102-V--MA-starch system appears somewhat fortuitous, in that it
is the "inert medium," the gel, by binding
key iodine species, that makes it possible to
establish the crucial difference in the effective diffusion coefficients of activator and
inhibitor. Such circumstances may be rare in
inorganic chemical systems. In biological
systems, on the other hand, membranebound species play key roles, and substrate
inhibition, rather than the autocatalysis
found in most chemical oscillators, is a
common means of dynamical regulation. In
such systems, as Turing recognized, diffulsion-induced instability is likely to be a
major mechanism for pattern formation.
REFERENCES AND NOTES
1. A. M. Turing, Philos. Trans. R. Soc. London Ser. B
237, 37 (1952).

652

2. J. D. Murray, J. Theor. Biol. 98, 143 (1982).


3. A. R. Peacocke, The Physical Chemistry of Biological
Organization (Clarendon, Oxford, 1983).
4. J. D. Murray, Mathematical Biology (Springer-Verlag, Berlin, 1989).
5. V. Castets, E. Dulos, J. Boissonade, P. De Kepper,
Phys. Rev. Lett. 64, 2953 (1990).
6. I. Lengyel, Gy. RBbai, I. R. Epstein, J. Am. Chem.
Soc. 112, 4606 (1990).
7.
ibid., p. 9104.
8. D. M. Kern and C.-H. Kim, ibid. 87, 5309 (1965).
9. H. Degn and D. E. F. Harrison, J. Theor. Biol. 22,
238 (1969).
10. J. L. Ibafiez, V. Fairen, M. G. Velarde, Phys. Lett. A
59, 335 (1976); V. Fairdn and M. G. Velarde,
Progr. Theor. Phys. 61, 801 (1979).
11. J. P. Kemevez, Enzyme Mathematics (North-Holland, Amsterdam, 1980).
12. J. D. Murray, J. Theor. Biol. 88, 161 (1981).
13. G. Nicolis and I. Prigogine, Self-Organization in
Nonequilibrium Systems (Wiley, New York, 1977);
L. A. Segel, Modeling Dynamic Phenomena in Molecular and Cellular Biology (Cambridge Univ. Press,
New York, 1984); A. B. Rovinsky, J. Phys. Chem.
91, 4606 (1987).
14. L. Edelstein-Keshet, Mathematical Models in Biology
(Random House, New York, 1988).
15. Gy. Bazsa and M. T. Beck, Acta Chim. Hung. 73
425 (1972).
16. J. Pearson and W. Horsthemke, J. Chem. Phys. 90,
__,

1588 (1989).
17. Nevertheless, there is evidence that similar structures
arise in starch-free experiments: Q. Ouyang, K. Lee,
H. L. Swinney, unpublished results.
18. S. Hayashi, T. Hirai, S. Shimomichi, N. Hojo, J.
Polym. Sci. Part A 20, 839 (1982).
19. A. Cesdro, J. Benegas, D. Ripoli, J. Phys. Chem. 90,
2787 (1986).
20. D. Saunders and R. L. Pecsok, Anal. Chem. 40, 44
(1968).
21. Strictly speaking, the effective reactivity of I- should
also be reduced by this interaction, but we have
neglected this much smaller effect in the present
treatment.

22. R. Parsons, Handbook of Electrochemical Constants


(Butterworths, London, 1959).
23. C. Berding, Bull. Math. Biol. 49, 233 (1987).
24. P. De Kepper and I. Nagypl, unpublished results.
25. I. Ruff, V. J. Friedrich, K. Csillag, J. Phys. Chem.
76, 162 (1972).
26. We thank P. De Kepper and J. Boissonade for
discussions of Turing structures and for encouraging us to probe more deeply into the mechanism of
this system. This work was supported by grants from
the National Science Foundation (CHE-8800169)
and by a U.S.-Hungarian cooperative grant from
the National Science Foundation and the Hungarian
Academy of Sciences.
31 July 1990; accepted 8 November 1990

Potassium in Clinopyroxene Indusions from Diamonds


GEORGE E. HARLow AND DAVID R. VEBLEN
Analytical transmission electron microscopy, electron microprobe analyses, and singlecrystal x-ray diffraction data support the conclusion that high potassium contents, up
to 1.5 weight percent K20, of some diopside and omphacite indusions from diamonds
represent valid clinopyroxene compositions with K in solid solution. This condusion
contradicts the traditional view of pyroxene crystal chemistry, which holds that K is
too large to be incorporated in the pyroxene structure. These diopside and omphacite
inclusions have a high degree of crystal perfection and anomalously large unit-cell
volumes, and a defect-free structure is observed in K-bearing regions when imaged by
transmission electron microscopy. These observations imply that clinopyroxene can be
a significant host for K in the mantle and that some clinopyroxene inclusions and their
diamond hosts may have grown in a highly K-enriched environment.
peridotite xenoliths in kimberlites
and other extrusive rocks are too small to
account for the K levels in basalts produced
by partial melting (1). Eclogites are closer to
basalts in composition, and their K-bearing
phases may include sanidine as well as omphacite (a cpx with a nominal composition
of Na0.5Ca05A10.5Mg0.5Si2O6). Although
eclogites may be important reservoirs for K,
they are not considered to be the source of
most mantle-derived basalts. Moreover, the
relatively high K contents of omphacite
from Group I eclogite xenoliths (0.10 to
0.33% K20 by weight) have not been demonstrated to be in the cpx crystal structure
(2). In order to evaluate accurately the potential for cpx as a K reservoir in the mantle
and as a possible source for K in the genesis
of mantle-derived igneous rocks, the crystal
AmeriMineral
of
Sciences,
G. E. Harlow, Department
can Museum of Natural History, New York, NY 10024. chemistry of K in pyroxene must be underD. R. Veblen, Department of Earth and Planetary Scireconciled with measureences, The Johns Hopkins University, Baltimore, MD stood better and
ments on xenolith suites.
21218.

HE QUESTION OF WHERE SUBSTAN-

garnet

tial K resides in the mantle is fundamental for understanding the source


for K in basalts that originate from partial
melting of the mantle, the distribution of K
in the mantle, the role of K in mantle
metasomatism and diamond formation, and
the mantle's heat budget (because of radioactive decay of 'K). Garnet peridotites are
thought to have compositions typical of the
upper mantle, and among the nominally anhydrous solid phases in these rocks,
diopsidic clinopyroxene (approximately
CaMgSi206) has the highest K level, is
stable over a large range of pressures, and is
reasonably abundant. However, measured
levels of K in clinopyroxene (cpx) from

SCIENCE, VOL. 251

Anda mungkin juga menyukai