Physics. -
Convection~currents
in the Earth *). By F. A.
VENINO
MEINESZ.
(Communicated at the meeting of February 22, 1947.)
In a previous paper the writer has drawn attention to the fact that the
hypothesis of convection~currents in the subcrustallayer under the eastern
half of the Indian Archipelago may give a good explanation of the deep~
focus and the intermediate earthquakes in this area. He emitted this
hypothesis already in 1932 1) for explaining the sinking down of the deep
basins which are c1early connected with the great folding processes in the
neighbouring tectonic beIts but which probably have originated with a
great time~lag of many millions of years after the folding. In the Banda
arc the last great folding period is put by UMBOROVE 2) in tertiary [2,
i.e. some twenty million years ago, while the sinking of the Banda basin,
although difficult to determine exactly by direct evidence, is probably
much more recent. Direct evidence that this area has provided erosion
products to the surrounding tectonic belt and that it, therefore, must have
been above sealevel, dates further back but, as MOLENORAAFF already
pointed out and as UMBOROVE also is inc1ined to assume, it seems likely
that the principal part of the sinking of the basin is simultaneous with the
rising of the adjacent tectonic belt which took place in the pleistocene, i.e.
only some 1 or 2 million years ago. MOLENORAAFF explained, for example,
that in Timor, the miocene folding is intersected by the present coast~line
of the deep basins and that it must, therefore, be anterior to the sinking.
His opinion about the simuItaneity of the rlsing and sinking movements
is widely accepted. It agrees with the explanation of both movements by
the hypothesis of a convection~current in the subcrustallayer which at the
same time must bring about a rising above the rising current and a sinking
above the sinking one.
This hypothesis mayalso explain the evident connection of these crustal
movements with the folding phenomenon and the great time~lag between
both. During the folding~period the Earth's crust may be supposed to have
down~buckled along the tectonic belt, thus forming a considerable crustal
bulge at the lower boundary of the crust, which according to the negative
anomalies and the topography in this belt and assuming a density difference
with the substratum of 0,6 must have a cross~section of 1500-2000 km 2 •
It must have pushed away the subcrustal material. As this last material is
poorer in radio~active constituents than the crust, we may expect a slow
heating up of this area by the excess of radio~active radiation caused by
the concentration of crustal material along the belt.
* ) Lecture on December 21, 1946.
1) F. A. VENING MEINESZ, J. H. F. UMSGROVE, PH. H. KUENEN, Gravity Expeditions
at Sea, Vol. 11, Pub!. Netherl. Geod. Co=., Waltman, Delft, p. 135.
2) Id. p. 140 e.s.
238
This heating up must have disturbed the equilibrium in the substratum.
If we assume with most geophysicists that the Earth is cooling n?twithstanding the amount of radio-active minerals present in the outer layers.
the sub stratum must have shown a downward temperature-gradient tending
to bring about instability as it causes layers of lower temperature to overlie
higher temperature layers. The writer. however. ag rees with JEFFREYS in
supposing that these layers have some strength or. in other words. that
below a certain limit. stresses only bring about elastical deformations; the
stresses have. therefore. to exceed this limit before flow can take place.
According to the smallness of the deviations from isostatic equilibrium of
10-15 mgal the writer supposes this limit to lie between 25 and 50 kgjcm 2 •
lf only the vertical temperature-gradient caused by the cooling of the Earth
i.<; present. this strength must prevent any convection-current to originate
and it has. therefore. a stabilising effect. H. however. at the same time. a
temperature gradient in a horizontal sense is present. the normal density
equilibrium in horizontal layers is disturbed. H this gradient is sufficient
the resulting stresses will overcome the streng th-limit of the substratum
and a current must set in which. because of the vertical temperaturegradient. takes the character of a convection-current.
We must ex peet this heating of the subcrustal layer below the tectonic
belt to a temperature sufficient to start the above phenomenon. to be a
slow process and so it does not seem unlikely that we can thus explain the
time-lag of many millions of years between the folding in the tectonic belt
and the coming about of the convection-current. It is important to investigate
this numerically to see whether this is possible.
We shall begin by assuming that the whole crustal root consists of
granite. This leads to a density difference from the substratum of about
0.6 (peridotite
3.23. dunite
3.29) as we assumed above when
mentioning the cross-section of 1500-2000 km 2 • For our deductions we
shall adopt a cross-section of 1700 km 2 •
For the radio-active heat-production we shall use the figures for different
types of rocks given by GUTENBERO in table 22 on page 155 of "Internal
Constitution of the Earth" and the figures of Table 18-2 on page 270
of "Handbook of physical constants" by BIRCH. SCHAIRER and SPICER. The
mean figure for granite given by GUTENBERO is 7.8 X 10- 13 caljcm 3 jsec
and by the second table 5.6 X 10- 6 caljgramjyear which gives 4.8 X 10- 13
caljcm 3 jsec. The mean figure is 6.3 X 10- 13 caljcm 3 jsec. Taking the mean
for peridotite and dunite in GUTENBERO's table we find 1.5 X 10- 13
caljcm 3 jsec and the Handbook gives for ultra-basic rocks 0.9 X 10- 6
caljgramjyear
0.9 X 10- 13 caljcm 3 jsec; the mean is 1.2 X 10- 13
caljcm 3 jsec. The excess heat developed in the root is the difference and so
we obtain 5.1 X 10- 1 3 caljcm 3 jsec. For the whole root of 1700 km 2 crosssection this gives 9 caljsecjcm (the cm dimension at right angles to the
cross-section) •
This figure has been derived for a root consisting entirelY of granite.
=
=
=
239
This is, however, unlikely. If the root has been formed by the downward
buckling of the crust, as it has been supposed, we must expect deep crustal
layers to be present and it is usually assumed that these layers are not
granitic but more basic. As, according to the assumed mode of originating
of the root, these layers may be expected to form the outer shell of it and
as we may suppose the root to have partially melted away because of the
higher temperature these materials must have been subjected to when they
were pushed downwards, it would of course be possible that a great part
of this outer shell has disappeared. The explanation of the more acid type
of volcanism in the nearby inner Banda~arc by the flowing oH of the
molten material of the root, as it has e.g. been advocated by ANDERS ON 3),
would seem to point to the granitic central part of the root being now
exposed to the melting and this would appear to confirm our last sup~
position.
We cannot, however, co me to any certainty about this point and so it is
no doubt possible that the root partially consists of deeper rocks than
granite. It is simple to see that if for these rocks the diHerence of the
heat~production from that of the sub~crustal material would be proportional
to the diHerence of the density from that of this same subcrustal layer
our result for the excess of the hèat~production of the root would remain
the same. The figures for the heat~production given by GUTENBERG and
others, however, deviate from this proportionality in the sense that the
heat~production in the crust diminishes probably quicker with depth.
Resuming we must recognize that the above mentioned figure of 9 cal/sec/cm
for the root may be too high and so we shall reduce it to 7.5 cal/sec/cm.
Adopting this figure the problem has to be solved what temperature
distribution in the subcrustal material is caused by this sou ree of heat
af ter a lapse of time of some 18 million years. For this solution we have
to apply the formula of heat~conduction for the two~dimensional case
represented by our problem. This is given by the diHerential equation for
the temperature (J
(1 A)
where
!:::,(J
expressed in polar coordinates r,
6 () = 0
2
(j
or
and
2
qJ
is
+ .!. 0 () + .!. 0
2
r Or
()
r 2 OqJ2
À
(1 C)
a=-
ce
A
coeff. of thermal conductivity,
c
heat~capacity.
e
density.
The quantities A, c and
e all refer to
(1 B)
the subcrustal material.
3) W. Q. KENN:EDY and E. M. ANDERSON, Crustallayers and the origin of rnagrnas,
Bull. Volcan. S. 11, Torne lIl, 1938.
240
We shall simplify our problem by the suppositions th at the root has a
circular cross~section with a radius ro and that the heat~production is
concentrated in the centre. We shall furthermore assume that the pheno~
menon is cylindrieally symmetrie. As the fundamental law of heat~con~
duction given by 1 A is linear in (J our phenomenon is not affected by the
norm al cooling of the Earth and so from this point of view th ere is no
objection to adopting this assumption.
The presence of the surface of the Earth at a distance of about 49 km
(thickness of the crust = 30 km + distance of centre below the lower
crustal boundary = -+- 19 km) must, however, affect it and so we have
to prove that this effect is negligible. It is simple to do this. The effect of
the boundary can be taken account of by assuming the presence of a sink
of heat of the same amount as the heat~source in the root and at the same
distance of 49 km above the Earth' s surface. This sink is, therefore, at a
distance of 98 km from the souree. We shaIl, however, find th at the effect
of the sink at this distance is negligible and so we can maintain our sup~
position of cylindrieal symmetry.
According to this supposition formulas 1 A and 1 B become
a
rpo + ~ (0)
( àr 2
r 0r
= 00.
àt
. (2)
The solution of this equation is an exponental integraI. whieh we shall as
usual indicate by the symbol
Ei
0= ~[-Ei(-~)J
inA.
(3)
'fat
where q is the heat~production in cal/cm 3 /sec.
We introduce
q = 7.5 cal/cm 3 /sec.
À.
0.01 (see "Handbook" p. 254).
c
0.20 (see "Handbook" p. 235 e.s.).
(}
3.27.
a
0.015.
This leads to
We compute (J for a time t = 18 million years = 5.67 X 10 14 sec.
Using for
(-x) the table in JAHNKE u. EMDE "Funktionentafeln"
p. 21. 22, we find the following va lues for (J. For the radius ro of the root
we introduce 23 km whieh corresponds to the supposed cross~section of
the root of 1700 km 2 •
Ei
r
km
23
40
60
80
100
r2
x=-
'fat
0.1555
0.470
1.058
1.880
2.940
-Ei(-x)
1.432
0.599
0.200
0.0506
0.0141
0
+ 85°.4
+ 35°.8
+ 12°.0
+ 3°.0
+ 0°.8
241
It is remarkable to see that a temperature of 85° at the surface of the
root requires such a long lapse of time to come into being. It is also
interesting to find that the heat af ter this long interval practieally did not
come beyond some 100 km. We see here our supposition confirmed that
the effect of a heat~sink at this distance is negligible.
Por making a rough estimate of the magnitude of the stresses caused
by this temperature distribution we compute the rise of the surface of the
substratum brought about by the expansion while neglecting -the elastieal
deformations caused by the resulting stresses. Por this rough estimate we
adopt a rectangular cross~section for the root of a height of 37 km and a
breadth of 46 km, and we as su me in the column of the substratum bordering
Fig. 1.
the root a temperature of 85°.4 over 37 km height and below this a falling
oH of the temperature as given by the above table i.e. at 17 km below the
point C a temp. of 35° .8, at 37 km below it a temp. of 12°, etc. We
adopt a volumetrie thermal expansion of the substratum of 3 X 10- 5
and we assume that the adjustment of the hydrostatie equilibrium of these
layers leads to the entire expansion appearing at the surface. We then
find a rise th ere of 146 meters whieh for a density of 3.27 represents an
anomaly of 14 mgal and an excess pressure of 45 kgfcm 2 • This is exactly
the order of magnitude we may presume for the strength~limit of the sub~
crustal layers and so we see that the great time~lag of 18.000.000 years
before the starting of the convection~current could weIl be explained.
When the strength~limit is exceeded by further heating, the sub cru stal
matter would flow off from the neighbourhood of the root and this would
disturb the equilibrium in the column below it; a rising current would set
in here while a sinking one would originate in the area whieh had become
loaded by this flow while the pressure diHerence in the deeper layer would
17
242
bring about a flow contrary to the one at the surface. The convection~
current would thus be started and as it begins by bringing about an
increase of the temperature~difference between the rising and sinking
columns it would accelerate till after some time a maximum difference
would be reached; this would probably occur af ter about a quarter of a
complete turn. The speed will th en decrease again till the current has about
made half of a complete turn, i.e. when it has brought the higher tempera~
ture matter on top and the lower temperature matter below; the rising and
sinking columns will then have assumed the same mean temperature and
the equilibrium is restored. The heat conduction will more or less alter this
picture and the amount of movement needed for restoring the equilibrium
but it does not change the principle of the phenomenon.
If we as su me that up to now about a quarter of a complete revolution
has been made and supposing in accordance with the dimensions of the
Banda basin and the depth of about 400 km of the deep earthquakes in
this area th at the current takes place inside a cross~section of about 500 km
height and 500 km breadth, we find that in 2.000.000 years it must have
travelled a distance of about 200 km and so the mean velocity must have
been about 10 cm per year. During this interval which is only one ninth
of the interval involved in the above problem, the cooling cannot have got
much further than to a dep th of about 40 km, i.e. to a fifth of the depth
of the upper horizontal part of the current. It does not seem likely that
this can have seriously affected the course of events.
For obtaining an estimate of the sinking of the surface above the sinking
column we have to as su me the temperature distribution in the subcrustal
layer before the current started and outside the area heated by the root.
Referring to the curve given by GUTENBERG in "Internal Constitution of
the Earth" page 162, we think that fig. 2 gives an acceptable estimate.
We may probably adopt the temperature at the lower boundary of the
rigid crust at about 700° which would mean a gradient in the crust varying
from 30 0 /km at the surface to lOo/km at the bottom.
If we as su me that during the history of the Earth the subcrustal layer
has been turned over several times by convection~currents, the temperature~
curve in this layer has to show, as GUTENBERG remarks, a much smaller
gradient of, for instance, 1°/km; from 100 km - 500 km depth he assumes
a temperature of 1500 - 1800°. In the upper layer, however, where it
touches the crust a cooling curve must have formed since the last con~
vection~current took place, which might have occurred af ter the Eocene
folding period in the tectonic belt. This has been represented by the curve
of fig. 2.
From this curve we may estimate the mean temperature of the upper
200 km of the subcrustallayer at some 1200° and af ter a quarter revolution
cf the current this mean temperature must have reigned in the greatest
part of the sinking column, increased, however, by a certain amount because
ot the heat~conduction from the surrounding matter. As a result we may
243
make the rough estimate that over the lower half of this column the mean
temperature will be about 500 0 below the normal one. Adopting again a
thermal expansion coefficient of 3 X 10- 5 this amounts to a shortening of
the column of 3.8 km.
2oao~
500'!
(>
35
3IJO
Fig. 2.
$O()
Ir",.
Temperature curve in the upper 500 km.
We may assume that about half of this shortening will appear at the
surface and the other half at the bottom of the column, thus causing the
pressure gradient needed for the horizontal parts of the currents at the
surface and at the bottom, the first being directed towards the sinking
column and the second away from it. This rough assumption is in harmony
with the formulas for a steady éonvection~current; in another paper the
writer hopes to enlarge on these deductions. We come to the conclusion
that we may expect a sinking of the surf ace over the sinking column of
about 2 km.
The reverse must be true for the rising column, and so we may estimate
the rising of the surface here at this same value. We neglect here the
rising at the start caused by the radio~active effect of the root as derived
above. The result would thus be a difference between the two are as of
about 4 km. This difference is in good harmony with the actual topography
of the Banda arc area and, in fact, of the whole eastern half of the archi~
pe1ago but there is an obvious disagreement in the fact that the mean level
of the whole reg ion has not remained the same when the deep basins
originated but that it has been lowered over 1.5 or 2 km. This point has
already been raised by MAC GILLAVRY in 1930; he drew attention to the
difficulty of explaining this lowering of the mean level, while we might
244
rather expect that the thickening of the sialic crust by the orogenic
phenomena would bring about a rising of this mean level.
Our hypo thesis of convection~currents in the subcrustal .layer can give
a simple explanation of this difficulty and it can at the same time make
clear that this area as a whole must show an excess of gravity as the
anomaly~field of the East Indies shows it to be the case. The explanation
is that the presence of convection~currents brings about an increased
cooling of the Earth in this area. This is clear as the sinking current brings
matter of lower temperature downwards which absorbs heat from the
surrounding matter while the excess heat brought to the surface by the
rising current is for a large part lost to the Earth by an increased radiation
at the surface. Admitting this excess of cooling up to a depth slightly
larger than that to which the currents go. it follows that the density is
correspondently larger than normal while the shrinking must lower the
surface. If the whole area consisted of rising and sinking columns and if
the excess heat of the rising column was entirely radiated into space we
should thus obtain a mean lowering of the surf ace of half the shrinking of
3.8 km derived before for the sinking column. i.e. about 2 km. This figure
may be too large but it is certainly possible that it would attain the amount
of 1.5 km which is present. It is also possible th at part of this amount has
been caused by older current~systems. In that case this part of the lowering
would. however. already have been present during the last folding in
tertiary f 2.
The discussion of the gravity field may be postponed to a future paper.
The writer wishes here only to add a short remark which follows from
the last deductions. If we are right in attributing the low level of the mean
topography in the eastern half of the East Indies and. in facto in the whole
area east of Asia wh ere we find tectonic basins to an abnormal cooling
here of the Earth up to a dep th of some 600-800 km. this must mean a
disturbance of the therm al equilibrium of the Earth of a much greater size
than that caused directly by the presence of the crustal root as derived
before. The question then arises whether this would not be likely to start
a correspondently greater type of convection~current. made possible by the
presence to a much greater depth of a slight temperature gradient of the
order of 10 /km or less caused by the cooling.
This current~system rising under a broad borderzone of the continent
and sinking under the adjoining oceanic zone would thus have continental
dimensions and might even be supposed to involve the whole thickness of
the mande up to a depth of 2900 km. It might be imagined to bring about
the major cycles of orogeny weIl known in the geological history of the
Earth 4). During each cycle the current would only make about one half of
a revolution in the same way as it has been explained for the smaller type
4)
See also: DAVID GRIGOS. A Theory of Mountain-building. Amer. Journ. o. Sc. 237.
611-650 (1939).
245
of current dealt with above. Af ter each cycle the temperature gradient
must have disappeared and it takes a long period of rest before the cooling
has brought it again into being.
There is much to be said in favour of this explanation of the orogenic
cycles. It explains their occurring more or less periodically with a variabie
period. It likewise explains the regressions in the beginning of an orogenic
period and the slow change to transgressions in the end of this period
and during the long period of rest between the cycles; these regressions
are connected with the rise of the surface over the borderzones of the
continents when a rising current originates there, while 'the transgressions
come into being wh en the current is slowly dying out.
If these speculations may be allowed we should come to the conclusion
that the initial disturbance of the thermal equilibrium needed for bringing
about these large cycles would be caused by the smaller types of current~
system here discussed which themselves are started as a direct conse~
quence of the folding together of the Earth' s crust. We thus would see
here a chain of events of steadily increasing size started in the beg inning
by al small surface effect of the Earth and leading eventually to a big
current~system having its main ca~se in the cooling of the Earth which
provides it with the energy required.