Proc. Natl. Acad. Sci. USA
Vol. 76, No. 9,pp.4192-4200, September 1979
Structural geology of the Earth's interior*
THOMAS H. JORDAN
Geological Research Division, Scripps Institution of Oceanography, La Jolla, California 92093
Communicated by Preston Cloud, June 14, 1979
picture of the Earth's internal structure that shouldlead to im-
proved models of mantle dynamics. Lateral variations in seismic
wave speeds have been documented in all majorlayersof the
Earth external to its core, with horizontal scale lengths ranging
from 10 to 104 km. These variations can be described in terms
of three types of heterogeneity: compositional, aeolotropic, and
thermobaric. All three types are represented in the lithosphere,
but the properties of the deeper inhomogeneities remain hy-
-ntheticaL It is argued that sublithospheric continental root
structures are likely to invoke compositional as well as ther-
mobaric heterogeneities. The high-velocity anomalies charac-
teristic of subduction zones-seismic evidence for detached and
sinking thermal boundary layers-in some areas appear to ex-
tend below the seismicity cutoff and into the lower mantle or
mesosphere. Mass exchangebetweentheupperandlowerman-
tles is implied, but the magnitude ofthe flux relative to the total
mass flux involved in plate circulations is as yet unknown. Other
observations, such as the vertical travel time anomalies seen in
the western Pacific, may yield additional constraints on the flow
geometries, but further documentation is necessary. Thermo-
6aric heterogeneities associated with a thermal boundary layer
at the base of the mantle could provide the explanation for some
of the observations of heterogeneities in the deep mantle. The
evidence for very small scale inhomogeneities (<50 km) in re-
gion D" and for topography on the core-mantle interface mo-
tivate the speculation that there is a chemical boundary layer
at this interface, as well as a thermal one.
Seismology is providing a more sharply focused
The colors are chocolate, purple, lavender, and magenta,
ofmany tones and shades. If it were not for this powerful
coloring, which discloses every band and layer with em-
phasis, and each with a habit peculiar to itself, we could
not venture to assert so much about them as we have
done. For we have been reading geologyfrom miles away
from our rocks. But what are miles in this Brobdingna-
-Capt. C. E. Dutton, describing the Paleozoic sequence at
the head of the Grand Canyon (1).
The U.S. Geological Survey was founded by pioneers whose
gaze was set on the grand scales of geology exposed in the
American West. Confronted by a record laid starkly bare in
these arid wastelands, they could not help but sense the movings
within our planet. They vocalized their perceptions with bold
new words: J. W. Powell gave us diastrophism to comprise the
various processes of crustal deformation, which G. K. Gilbert
(2) divided into orogenic (mountain-making) and epierogenic
(continent-making); Dutton (3) defined isostasy to connote the
tendency toward equality of total lithostatic loads, fully rec-
ognizing its control on vertical motions. The intervening century
has witnessed considerable development of these concepts.
From studies of Dutton's isostasy came the modern notions of
lithosphere and asthenosphere. We now interpret the litho-
sphere to be a mechanically strong boundary layer that slides
across the Earth's surface in huge rigid plates and the asthe-
nosphere to be an underlying layer of convective flow. The
great convulsions Gilbert called "orogenies" we now explain
as the deformations caused when two plates collide, and we
suspect that at least some epierogenic movements are isostatic
responses to convectively induced variations in mantle tem-
Yet, even in this bright era of the plate paradigm, very little
confidence can be placed in our understanding of the dynam-
ical controls on surface tectonics, for their mechanisms are still
only dimly perceived. Nearly every geophysicist postulates
some form of thermal convection operating within the mantle,
but many rudimentary constraints on the geometry of the flow
are lacking. We do not know, for example, to what depth the
mass circulation associated with plate motions persists. We
cannot predict with much certainty or completeness the geo-
graphical distribution of ascending and descending currents
nor, for that matter, even their statistics. We do know that the
Earth is a chemically differentiated planet, and from the vol-
canoes seen on its surface we must conclude that some of
the differentiation processes continue today, but we are only
vaguely aware of what roles these processes have played in the
planet's dynamical evolution. We are beginning to understand
the complex history of the continental crust, but we can say
almost nothing about the history of the mantle directly beneath
this crust. In fact, much controversy clouds the discussions of
continental deep structure.
There is little wonder, then, that our dynamical models are
weak. We need a more sharply focused picture of Earth's in-
ternal structure. To give the dynamical theories some teeth,
geophysicists will have to become structural geologists and map
the features in the mantle and core that are manifestations of
How shall we do the structural geology of the Earth's inte-
rior? It is no easy task, for the phenomena to be studied lie be-
neath hundreds or even thousands of kilometers of rock,
through which we have no direct access. The data must come,
of course, from indirect observations near the surface. Studies
of crustal structure and tectonics are providing critical facts
about the underlying mantle, especially the studies of very large
scale geological features; many of these reflect variations in
upper mantle temperatures and compositions. Observations of
the transient, noninertial responses of the surface to loading and
unloading by large masses such as ice sheets and volcanoes, or
by coseismic deformations, are yielding constraints on the
mantle's constitutive parameters; these parameters control the
balance of forces and are obviously critical to dynamical
modeling. Petrological and geochemical analyses of the rocks
brought up from the depths by volcanic action are demon-
strating the existence of systematic, although puzzling, varia-
tions in both the major and minor elemental compositions of
the mantle. Some of these chemical variations have evidently
persisted in the mantle for eons and, thus, can contribute in-
formation about the dynamics averaged over time periods
*Presented at the symposium "Earth Science and Earth Resources-A
Centenary Salute to the U.S. Geological Survey," 23April 1979, at
the Annual Meeting of the National Academy of Sciences of the
United States of America.
Proc. Natl. Acad. Sci. USA 76(1979)
comparable to the age of the Earth. Work in all of these fields
is proceeding at a vigorous pace, and hardly a day passes
without the report of some new results.
But to map the details of interior structure necessary for
understanding its dynamics, we must resort to geophysical
techniques with higher resolutions: to studies of the Earth's
gravity, electromagnetic, and elastic displacement fields. In this
address I shall concentrate on seismological investigations of
Earth structure-partially because it is the subject with which
I am most familiar and also because I believe it to be the most
promising source of information about the deep interior.
If the Earth were a perfect sphere whose pressure, temperature,
and composition, and hence whose elastic parameters, varied
only with distance from its center, then the travel times and
amplitudes of the various seismic waves would depend only on
the relative geometry of the source and receiver and not on their
absolute coordinates. In fact, however, significant geographical
differences in travel times and amplitudes have been observed,
and analysis of them is revealing various structures within the
mantle. For the surface and body waves sensitive to mantle
properties, the spatial fluctuations of travel times are typically
small, only rarely exceeding 10% in magnitude. Specific studies
indicate that the wave speed variations responsible for these
fluctuations are similarly small, at least for heterogeneities with
scale lengths greater than 100km or so. The weakness of lateral
variations simplifies the analysis of the seismic data, especially
travel time and eigenfrequency data, because variational
principles can be used to construct approximate, but accurate,
linear relationships between these particular data functionals
and models of the heterogeneity.
Lateral variations in seismic wave speeds have been docu-
mented in all layers of the Earth external to its core, with hor-
izontal scale lengths ranging from 10 to 104 km. In discussing
this heterogeneity I will adopt the radial regionalization illus-
trated in Fig. 1. Although most of the terms have been stan-
dardized by historical usage, some of the nomenclature requires
clarification. Following Daly (5), the silicate portion of the
Earth is divided into three shells: a lithosphere, an astheno-
sphere, and a mesosphere. The lithosphere-asthenosphere
boundary is taken to be the depth of compensation, here de-
fined to be the level above which significant deviatoric stresses
(say, >1 MPa or 10 bars) are sustained for geologically long
periods of time (say, >106 yr) and below which material be-
haves as a viscous fluid with no significant strength (i.e., <1
MPa). This dynamical definition is more or less the classical one
and does not necessarily identify the lithosphere with the region
occupied by the coherent kinematical entities we call plates (for
which I prefer and shall use Elsasser's term tectosphere), nor
Depth of compensation
- ______ JSeismicity cutoff
vp gradient change
Principal layers of the silicate Earth.
Proc. Natl. Acad. Sci. USA 76 (1979)
does it necessarily identify the asthenosphere with the Guten-
berg low-velocity zone for shear waves.
Below the asthenosphere is the mesosphere, evisaged by Daly
to be an inert, strong, lower mantle that does not participate in
present-day tectonic processes. Isacks et al. (6) incorporated this
concept into their model of lithospheric plate tectonics; they
postulated that the convective circulation associated with plate
motions is confined to a relatively thin asthenosphere whose
base is marked by the abrupt termination of Benioff zone
seismicity at depths of about 650-700 km. Most recent workers
have abandoned the idea of a mechanically rigid mesosphere
in favor of a convectively unstable lower mantle (7-14), but the
question of whether or not there is much convective mixing
between the upper and lower mantles remains the subject of
controversy (discussed further below). Therefore, I will follow
Isacks and others in defining the asthenosphere-mesosphere
boundary by the seismicity cutoff, without necessarily accepting
any inferences about the details of mantle circulation.
Fig. 1 also specifies the major seismological divisions of the
silicate Earth-crust, upper mantle, lower mantle-as well as
the finer regionalization of Bullen (15). The boundary between
the upper mantle and the lower mantle is placed at the dis-
continuity in wave speeds observed near 670 km depth, which
is also taken to be the base of Bullen's region C. This demar-
kation is somewhat shallower than Bullen's original definition
based on theJeffreys-Bullenvelocity distribution (984 km), but
it is concordant with more recent Earth models (16-18). Fur-
thermore, it has the advantage of making this boundary es-
sentially coincident with the seismicity cutoff (perhaps causally
related-to the existence of the 670-km discontinuity) and, hence,
with the asthenosphere-mesosphere boundary.
In contrast with the mesosphere, neither the lithosphere nor
the asthenosphere is specifically identified with any of the
seismologically defined regions or zones. In some (and perhaps
most) localities the subcrustal portions of the lithosphere appear
to be distinguished by an anisotropic velocity structure (19-22),
and the lithosphere-asthenosphere boundary may conform to
the top of the Gutenberg low-velocity zone (both corresponding
to constant, but possibly different, isotherms). At the present
time, however, it seems wise not to generalize these relationships
Region A Heterogeneity. Numerous local, regional, and
global studies have established that the crust is inhomogeneous
on essentially all scales sampled by seismic waves, from the
hemispheric asymmetry of the continents and oceans to the
small-scale heterogeneities responsible for high-frequency
scattering (23, 24). Of course, these observations only confirm
what is obvious to every structural geologist.
Region B Heterogeneity. All lines of evidence, seismological
and otherwise, indicate that the uppermost mantle is also in-
homogeneous across a wide spectrum of horizontal scales. As
in the crust, high-frequency seismic waves are incoherently
scattered by more-or-less randomly distributed bodies with
characteristic dimensions of less than 30 km (24-26), although
these scatterers could be more anisotropic than their crustal
counterparts (22). Presumably much (but probably not all) of
this inhomogeneity resides in the lithosphere, either frozen in
during its formation or subsequently introduced by various
magmatic and tectonic processes. Because their sizes are small
compared to their depths, the exact mapping of these bodies
will be difficult, and we may have to be contented with de-
scribing the gross statistical parameters of their distributions
(e.g., root mean square amplitudes and correlation lengths).
For scale lengths greater than a few tens of kilometers,
however, wave-speed maps of the near-surface heterogeneities
can be obtained by the systematic inversion of relative travel
time anomalies observed across dense seismic arrays (27-30).
These techniques are still under development, with much
current work aimed at increasing their three-dimensional re-
solving power. Preliminary reports indicate that the relative
amplitudes of the primary waves are tractable model discri-
minants (30) and that gravity measurements can be successfully
combined with travel time data in the formal inversion pro-
cedures (29). Already, important conclusions are being drawn.
Aki (28) has surveyed the results, and he reports: "Significant
everywhere to a depth of 100
estimate of root mean square fluctuation about 3%. The
lithosphere-asthenosphere boundary seems to manifest itself
as change in the roughness of anomaly pattern or in the trend
of anomaly." He notes that there is a good correlation between
the lithospheric anomalies and the surface geological features
in active areas such as California, Hawaii, and Yellowstone but
that the correlation is not as obvious in stable continental areas
such as eastern Montana and Norway.
The manner in which region B anomalies correlate with
seismology because its answer can provide diagnostic infor-
mation for interpretive modeling. Consider, forexample, that
tectospheric heterogeneities are defined to translate with the
crust during normal plate motions and are therefore available
to participate in long-term crustal evolution, whereas hetero-
geneities in the more mobile substrata arelikelytoparticipate
only as transient disturbances or not at all. Actually, at the in-
termediate and large scales (>300 km), the correlation between
long-term crustal history and the underlying wave speed
variations in region B is remarkably good, especially in the
continents. To illustrate some of these correlations I will use the
crude tectonic map of Fig. 2, in which the oceanic crust is
partitioned according to its age (or, more precisely, thesquare
root of its age) and the continents are dividedaccordingto their
Phanerozoic tectonic histories-i.e., by their generalized tec-
tonic behaviors since -600 Myr ago.t Representative Rayleigh
wave dispersion curves for the six tectonic regions are sum-
marized in Fig. 3, and several kinds of shear wave travel time
residuals are displayed as histograms in Figs. 4 and 5.
There is a general increase inRayleigh wavephasevelocities
with increasing age of the oceanic crust and acorresponding
decrease in shear wave travel times. This correlation is rea-
sonably well explained by the growth, byconductive thermal
decay, of a high-velocity low-attenuation boundary layer
identified as the oceanic lithosphere (20, 36, 37, 40-42). Evi-
dently this mantle layer, which seismologists term the "lid,"
attains a thickness of about 100 km in the oldest oceanbasins,
such as the western Pacific (41). Lid thicknesses derivedby
dispersion curve inversion (20, 36, 41) are in substantial
agreement with other geophysical estimates of subcrustal
lithospheric thicknesses, including the depth of Hawaiian
seismicity (43), estimates based onloading response (44),and
thermal models (45).
More puzzling, and more controversial in theirinterpretation,
are the observations related to the large-scale lateral hetero-
geneity of continents and oceans. Both regional (33, 34)and
; 50 km) lateral inhomogeneity is observed
150 km, with the minimum
is an important question for structural
tThis map, which has appeared elsewhere (31),wasoriginally designed
for the analysis of low-frequency eigenspectra and surface wave data
to replace the schemes previously used by seismologists le.g., plate
IV of Umbgrove (32)1. The reader iscautioned, however, thatmany
of the boundaries are approximate or interpretative, especially in
the southern oceans, in continental regions whose true tectonic history
has been obscured by Cenozoic platform cover, and inmy necessarily
subjective definition of what constitutes a "Phanerozoicorogeniczone
or mobile belt."
Proc. Natl. Acad. Sci. USA 76 (1979)
regions: A, 0-25 Myr (Il); B, 25-100 Myr (S); and C, >100 Myr (M). Subaerial continental crust is partitioned into three regions based on
Phanerozoic tectonic history: S, shields and platforms of exposed Archean and Proterozoic rocks with little or no Phanerozoic cover (J), P,
platforms with relatively flat-lying, undisturbed Phanerozoic cover (Ei); and Q, orogenic zones or mobile belts with significant deformation
or magmatic activity in the Phanerozoic (I). White areas are regions ofsubmerged continental or transitional crust, including continental margins,
island arcs, and oceanic plateaus adjoining continental crust.
Generalized tectonic map of the globe between 750N and 700S (Mercator projection). Oceanic crust is partitioned into three age
global (46, 47) surface wave studies have conclusively demon-
strated that region B velocities are higher beneath the continents
than beneath the oceans. The unanswered question is: How
deeply do these variations extend? On this point, the litho-
spheric plate hypothesis, which identifies tectosphere with
lithosphere, offers a specific, testable prediction: any structural
variations globally coherent with the parameters of long-term
(say, greater than 600 Myr) continental evolution should be
confined to the lithosphere (say, above 130 km), because plate
the six tectonic regions of Fig. 2. Letters identify regions; phase ve-
locities sample the following paths: S, Canadian Shield (33); P, Shiraz,
Iran, to Jerusalem (34); Q, Charters Towers to Adelaide, Australia
(35); C, Aleutian Is. to Afiamalu, Samoa (36); B, Gulfof California to
Atiamalu, Samoa (36); A, Rivera Fracture Zone to Galapagos Is. (36).
-, Continental; ---, oceanic.
Representative Rayleigh wave dispersion curves for
motions should continually rearrange sublithospheric hetero-
geneities with respect to surficial features. Furthermore, this
model-at least its simple version-predicts that region B
structural variations should be similar in old continents and old
ocean basins, where the lithospheric thicknesses are presumably
nearly equal (44).
Despite the merit of this model, its implications are difficult
to reconcile with the seismic data now available. Consider the
evidence in Fig. 4, where Sipkin and Jordan's (37) ScS2-ScS
travel time residualst are grouped according to the crustal re-
gionalization of Fig. 2. The median difference between one-
way travel times for old oceans (category C of Fig. 2) and stable
continents (categories S and P) is +3.0 s.§ About 1.5 s must be
added to correct this value for known differences in crustal
structures, so that, on the average, the actual one-way transit
times of shear waves through the mantle differ by more than
+4 s. The shear velocity variations required by this observation
t ScS is the shear phase reflected once from the core-mantle boundary;
ScS2 is reflected twice from the core and once from the surface. Be-
cause the near-source and near-receiver portions of their paths are
similar, their travel time differences are insensitive to upper mantle
velocity variations, except in the vicinity of the ScS2 surface reflection
points. Therefore, their differential travel times are useful for probing
upper mantle heterogeneity in regions lacking seismic stations, such
as ocean basins.
§Okal and Anderson (48) have claimed that there is no significant
difference in ScS2-ScS times for these two regions, but a recent
analysis (unpublished data) of a large set of multiple ScS phases
digitally recorded by the High Gain Long Period Network confirms
the results stated above.
Proc. Natl. Acad. Sci. USA 76 (1979)
One-way travel time residual, s
Sipkin and Jordan's (37) ScS2-ScS differential travel
times, grouped by tectonic classification (Fig. 2) ofthe crust sampled
at the ScS2 surface reflection point. Residuals are observed times
minus times computed by using the Jeffreys-Bullen tables (38); they
are corrected for source depths, ellipticity, station elevations, and
elevations of surface reflection points. All residuals have been divided
by 2 to normalize them to one-way travel times. Arrows indicate
medians for each category. Observations with surface reflection points
on continental margins are included in category Q. Data were collected
from the long-period instruments of the World Wide Standardized
Seismographic Network (WWSSN).
cannot be accommodated in the lithosphere without violating
surface wave constraints. Hence, the asthenosphere must be
characterized by large-scale continent-ocean heterogeneity.
One likely location for these variations is the Gutenberg low-
velocity zone,' which is weakly expressed or absent in most stable
continental areas (33, 34) but is well-developed in the oceans,
perhaps extending from the base of the lithosphere to depths
on the order of 200km (20, 36, 47). The surface wave data ap-
parently preclude the confinement of the heterogeneity above
this level, however, and some sort of deeper variations are im-
probably extends throughout Bullen's region B and perhaps
below it (49). This inference is concordant with Sacks and co-
workers' (50, 51) interpretation of continental structure in South
America and Alexander's (52) interpretation of the group and
phase velocities of surface waves, but it remains controversial
(48, 53, 54). Current work on higher mode dispersion (55-57)
should resolve these questions.
The surface wave and travel time data also demonstrate the
existence of intracontinental inhomogeneity in region B that
is globally coherent with surface properties and tectonic history
(34, 37). Continental areas that have acted as orogenic zones or
mobile belts during the Phanerozoic (category Q of Fig. 2) tend
to have lower shear velocities than do the more stable conti-
nental regions, and their velocity structures are more variable.
Much of the category-Q variations seen in Figs. 4 and 5 are
correlated with heat flow variations and magmatic activity,
implying a thermal control. The two most positive (>+5 s)
station anomalies in category Q of Fig. 5, forexample, are sites
in the magmatically active rift zones of Africa (Addis Ababa,
I have argued that continent-ocean heterogeneity
P, L, M
(> 100 Myr)
One-way travel time residual, s
Hatched bars are Sengupta's (39) S-wave station anom-
alies from deep-focus earthquakes; open bars are Duschenes and
Solomon's (40) S-wave source anomalies for oceanic earthquakes,
grouped by tectonic classification (Fig. 2) ofreceiver and source points,
respectively. Source anomalies are relative to the Jeffreys-Bullen base
line (38); station anomalies are relative to Sengupta's base line (39),
which differs from the former by an average of -0.9 s. Arrows indicate
medians for each category. Stations in island arcs are included in
Ethiopia, and Nairobi, Kenya); high values are also associated
with station anomalies or ScS2 surface reflection points in the
Basin and Range and in eastern Australia, other areas of recent
magmatism and high heat flow, whereas negative residuals
occur in magmatically quiescent regions of lower heat flow,
such as the southeastern United States and northeastern Siberia.
These correlations have been recognized by many authors (e.g.,
refs. 34, 37, and 46) and have been attributed primarily to
variations in or above the low-velocity zone, but the existence
of even more deeply seated heterogeneity has not been ex-
cluded. In the more stable continental regions, Phanerozoic
platforms (category P) and areas of exposed Precambrian rocks
(category S) are distinguished by the data; the former has lower
average shear velocities in region B. This fact is important for
theories of epierogenesis. Evidently, the large areas of persistent
upwarping-Suess's shields-.-have more mature root zones than
do areas that have accepted Phanerozoic cover. A more detailed
explanation requires a better theory of continental deep
structure, however, and such theories are only now taking shape
The constraints on small- to intermediate-scale (<3000 km)
inhomogeneities within the oceanic upper mantle are less good,
primarily because the oceans lack seismic stations, but pre-
liminary evidence suggests their existence. Strong azimuthal
variations in shear wave travel times have been documented
around the Hawaiian Islands (59, 60). In addition, differential
travel times between various multiple ScS phase pairs are
consistently less for paths from the Kuril-Kamchatka Arc to
Oahu than for other paths in the western Pacific (unpublished
data); these paths sample the anomalously high topographyof
the northwest Pacific, including the Hawaiian Swell. Our ob-
servations in the western Pacific also show considerable scatter,
amounting to several seconds in one-way travel time, which
indicates the existence of heterogeneities with scale lengths of
z1000 km. At present, however, we do not know if these vari-
Proc. Natl. Acad. Sci. USA 76 (1979)
ations are localized in region B or if they are distributed in other
Much better documented are the heterogeneities associated
with descending lithospheric slabs. In the well-studied Benioff
zones of the western Pacific, the travel times and amplitudes
of body waves confirm the model for plate subduction ad-
vanced by Isacks et al. (6): i.e., an n~100-km-thick slab pene-
trating an upper mantle which has lower velocities and higher
specific attenuation (6, 61-68). Detailed velocity and attenu-
ation models have been proposed for Japan by Sacks and his
coworkers (50, 66, 68); they estimate that the velocity contrasts
across the slab's upper boundary in region B are about 5 i 1%
for compressional waves and 7 ± 2% for shear waves, generally
consistent with other estimates (e.g., ref. 62). In South America,
on the other hand, the descending slab evidently interacts with
a thick, advancing, continental tectosphere, and the structure
is more complex (50, 66, 68, 69).
Region C Heterogeneity. Velocity anomalies associated with
descending slabs have been shown to extend at least as deeply
as the Benioff zone seismicity and therefore into region C (62,
64, 70, 71). The velocity contrasts between the slab and sur-
rounding mantle in some deep-focus zones appear to be com-
parable to those at shallower depths (70, 71), although the data
are not unanimous on this point (72).
Aside from Benioff zones, very little evidence has been put
forth for lateral variations which can be specifically assigned
to region C. Some workers, quoted previously, advocate the
extension of continent-ocean heterogeneity into region C, but
their arguments have not gained wide acceptance.
If large-scale region C heterogeneities do exist, they may be
most obviously expressed as regional variations in the depths
of the phase transitions bounding or internal to this zone. A
number of regional studies have produced velocity structures
of the upper mantle, and these indeed show variations (+25 km)
in thedepths of discontinuities. There are substantial trade-offs
between adiscontinuity's depth and the velocity structure above
it, however. Given the large variability of region B velocities,
any inferences must be made with considerable caution because
few studies have explored these trade-offs in detail. A notable
exception is the recent work by England et al. (73) who argue
plausibly for significant differences in the velocity structures
of western Europe and the Russian platform extending to 500
km in depth.
Region D Heterogeneity. Bullen's region D-the lower
mantle or mesosphere-was divided by him into two layers, D'
andD",on the basis of a decrease in the compressional velocity
gradient about 200km above the core-mantle boundary (Fig.
1; ref. 15). Until 10 years ago, region D', comprising the bulk
of the lower mantle, was regarded as a relatively homogeneous
layer, both radially and laterally (15, 74). The first clear indi-
cations of widespread heterogeneity came from the detailed
analysis of P-wave arrival angles, facilitated by the advent of
large-aperture seismic arrays (16, 75, 76). Evidence for inho-
mogeneity has grown steadily since then, although the models
are still quite primitive and qualitative.
The effects of lower mantle heterogeneity on traditional
seismic data are largely obscured by the relatively strong ve-
locity and attenuation variations in the crust and upper mantle.
Toinvestigate deep variations, experiments must be designed
to reduce this interference. One successful technique has been
to use the difference between the travel times of core-reflected
(PcP or ScS) and direct (P or S) body waves with similar upper
mantle paths. Hales and Roberts (77) demonstrated that ScS-S
differential travel times show anomalous scatter noteasilyat-
tributed touppermantle structure.Subsequentstudies of data
fromvery deep earthquakeshave confirmed that this scatter
is caused by heterogeneities residing in region D (12). As illus-
trated in Fig. 6, ScS-S residuals from deep-focus events fluctuate
by 5 s or more throughout the distance range of 30-80°.
It was first speculated that the inhomogeneities implied by
these fluctuations might lie in the vicinity of the core-mantle
boundary (77, 78), but further examination reveals that at least
part, and perhaps most, of the scatter results from near-source
heterogeneities. In particular, material characterized by
anomalously high velocities appears to underlie deep earth-
quake zones in both the eastern and western Pacific (70, 79, 80).
The results strongly suggest that, in some regions of rapid plate
subduction, descending slabs penetrate the mantle to depths
exceeding 800 km-that is, below the seismicity cutoff and into
the mesosphere (12, 70). If this hypothesis is correct, then ma-
terial is being exchanged between the asthenosphere and the
mesosphere, at least locally.
The extension of subduction zone anomalies into the lower
mantle has also been inferred from anomalies in P-wave arrival
angles from Benioff-zone earthquakes (76, 81-83), although
the relationship of these anomalies to the subduction process
is less clear. When plotted as "mislocation vectors" (the dif-
ference between an event's actual location and its apparent
location as seen by a seismic array), these data display a geo-
graphic coherence that implies near-source heterogeneity;
namely, changes in the patterns of mislocation vectors often
occur at changes in subduction zone geometry. The fact that
deep events show this behavior argues for variations at or below
700 km in depth. The nature of this heterogeneity has not been
precisely quantified, because the inverse problem for mislo-
cation vectors is not well understood, but Powell (83) has at-
tempted to model some aspects of the data. As she has demon-
strated, the data evidently require variations beneath the seis-
mic zones with horizontal scale lengths of 1000 km or more. In
particular, simple slab models cannot explain all of the obser-
vations, although deep slab penetration may be responsible for
certain features such as the splitting of mislocation vectors ob-
served at the Norwegian Seismic Array for events in the Middle
America Trench (82).
The evidence for inhomogeneities at deeper levels within
region D' is equally strong. Julian and Sengupta (84) have ob-
served that the scatter in P-wave travel times from deep-focus
events increases substantially beyond 700, implying a greater
quakes as a function of epicentral distance. Residuals are observed
times minus times computed from Jeffreys-Bullen tables (38), cor-
rected for source depth. Data were collected from the long-period
instruments of the WWSSN (78) and show a large scatter attributed
to region D heterogeneity (12).
ScS-S differential travel times from deep-focus earth-
Proc. Natl. Acad. Sci. USA 76 (1979)
Summary of characteristic horizontal scale lengths of wave-speed heterogeneity observed in silicate layers of Earth
Subcrustal lithospheric heterogeneity (19-30, 34)
Subduction zone anomalies (6,50,61-69); continental
roots (12, 31, 33, 34, 37, 46-59); thermal decay (20,
Subduction zone anomalies (62, 64, 70-72); transition
zone topography (73)
Subduction zone anomalies (12, 70, 76, 79-83)
Deep-mantle P-wave anomalies (84, 85)
Basal heterogeneity (78, 84-95)
lateral variability in the mantle below a depth of 2000 km. They
estimated that this heterogeneity is characterized by horizontal
scale lengths of approximately 1000 km and velocity fluctua-
tions of 1% or more. A much larger set of P-wave times (nearly
700,000) compiled by the International Seismological Centre
has been analysed by Dziewonski et al. (85), who inverted the
data for the lowest order (1 < 3) coefficients in a spherical
harmonic representation of deep-mantle velocity variations.
These authors resolved large-scale features below 1500 km in
depth and concluded that the perturbation amplitudes increase
near the base of the mantle, in agreement with Julian and
Sengupta's findings for smaller-scale heterogeneity. Most in-
terestingly, they obtained a significant negative correlation
between the satellite-derived gravity field and the gravity
anomalies computed by assuming isochemical proportionality
between velocity and density anomalies. The implications of
this important finding are not fully clear, but it appears to be
at odds with a simple model of local, convectively induced
Various techniques have demonstrated the existence of sig-
nificant lateral structures in region D" with a wide spectrum
of horizontal scales, supporting the notion that the lowermost
mantle is more heterogeneous than the region above it. The
observational techniques have generally made use of the special
properties of waves propagating tangentially or nearly tan-
gentially to the core-mantle boundary. For example, Alexander
and Phinney (86) found that the spectral characteristics of P
waves diffracted along this discontinuity are geographically
variable, and they inferred large-scale inhomogeneities in re-
gion D". In contrast, heterogeneities on a much finer scale are
implied by Haddon's (87) hypothesis that certain peculiar
high-frequency waves precursory to the core phase P'DF are
energy scattered at or near the core-mantle boundary. Detailed
studies have generally confirmed Haddon's idea (88-92), and
the horizontal scale lengths of the scatterers has been estimated
at 10-50 km (89, 92). The data are less specific on the exact
location of the heterogeneity; they can be explained by either
"lumps" in region D", with velocity variations of a few percent,
or "bumps" on the core-mantle interface, with amplitudes of
a few hundred meters. Recently, however, Chang and Cleary
(93) have observed precursors up to 65 s before the PKKP phase
at distances near 600, which they show are most simply ex-
plained by bumps on the interface. Sacks and his co-workers
(94, 95) have examined the amplitudes of the phase P'AB, which
intersects the core interface at an oblique angle, relative to P'DF,
which is almost normally incident. From the geographic co-
herence of these ratios, they argue for heterogeneity on a
somewhat larger scale (z150 km) and speculate that it maybe
a manifestation of convective activity in region D".
QUESTIONS OF STRUCTURAL
The salient observations of wave-speed lateral heterogeneity
are summarized in Table 1, where the horizontal dimensions
characterizing the heterogeneity spectrum are indicated for
each of the major layers.$ The table is certainly not complete:
observed features not yet assignable to specific layers have been
omitted (e.g., the mesoscale anomalies of the western Pacific
evident in the multiple ScS travel times), and many more must
surely lie undiscovered. But even in this early stage of structural
investigation, we can begin the geological interpretation of
To phrase our questions, it is convenient to define some
specific terms for the important types of heterogeneity. Com-
positional heterogeneities are those attributable to spatial
variations in chemical composition, including the variations in
minor elements such as volatiles. Field observations demonstrate
that crustal inhomogeneities are primarily of this type and that
compositional heterogeneities extend at least into the litho-
spheric mantle. Aeolotropic heterogeneities are those associated
with spatial variations in anisotropic properties. Mesoscale and
large-scale aeolotropic heterogeneities have been mapped in
the oceanic lithosphere (19-21), and, according to Fuchs (22),
some of the small-scale and microscale heterogeneities of the
lithosphere might be aeolotropic as well. Thermobaric heter-
ogeneities are those caused by the spatial variations in tem-
perature and pressure, including any variations due to isoche-
mical phase changes and electronic transitions. The wave-speed
variations in the lithosphere induced by conductive cooling are
thermobaric heterogeneities. Lithospheric heterogeneity is thus
a mixture of all three types: compositional, aeolotropic, and
What properties characterize the deeper variations? At this
time, the answers are speculative; the models are hypothetical.
The problem of continental deep structure, for example, is often
discussed in terms of purely thermobaric heterogeneity. Ac-
cording to the thermal boundary layer hypothesis, the sub-
continental upper mantle is strongly heated during major or-
ogenesis and subsequently cools by the conduction of heat to
the surface; the cratons are explained as areas that have been
cooling for a very long time and where the thermal boundary
layer has grown to extreme thickness. The time scales are about
right to explain the basic seismic observations, but one difficulty
with this model is the apparent conflict between the predicted
isostatic behavior of the cratons and the observations of epier-
ogenesis: stable continental areas simply do not show the ex-
¶In discussing the heterogeneity spectrum relevant to seismology,I
shall use the terms "microscale," "small-scale," "mesoscale," and
"large-scale" to designate horizontal dimensions of order 101, 102,
103, and 104 km, respectively.
Proc. Natl. Acad. Sci. USA 76 (1979)
tended history of subsidence and crustal thickening required
by the model. Indeed, the highest region-B shear velocities, and
presumably the lowest temperatures, are found beneath the
shields, whose recent history is one of uplift and erosion. Fur-
thermore, a thick, chemically homogeneous, thermal boundary
layer is probably convectively unstable (49); it would be quickly
disrupted and eventually destroyed rather than persist through
the eons as a stable continental root. One is forced to conclude
that continental deep structure cannot be purely thermobaric
unless it is maintained by advection, a possible but unattractive
alternative. In fact, continental deep structure is likely to involve
compositional as well as thermobaric heterogeneity (49, 58).
The specific model that I favor incorporates a type of hetero-
geneity best described as isopycnic, because it postulates that
the density effects of thermal and compositional variations in
the sublithospheric continental root zones are roughly equal in
magnitude but opposite in sign. Isopycnic heterogeneities ca-
pable of surviving for long periods could be generated by the
consolidation beneath the continents of mantle material pre-
viously depleted in basaltic constituents, an idea that receives
some support from studies of kimberlite xenoliths (58, 96). The
hypothesis implies that continental roots are part of the tec-
Heterogeneities that are likely to be primarily thermobaric
are the subduction zone anomalies, the seismic expressions of
detached and sinking thermal boundary layers. The question
here is: How deeply extends the circulation they imply? If one
accepts the evidence for high-velocity subduction zone
anomalies in the upper parts of region D, then it is difficult to
postulate no mass exchange between the asthenosphere and the
mesosphere. On the other hand, the mass flux across the 670-km
discontinuity might still be small compared to the total mass
flux involved in the plate tectonic circulation. Within the next
decade or so, structural seismology should provide the con-
straints on the flow geometry necessary to resolve these and
other problems of mantle dynamics. Could it be, for example,
that the mesoscale western Pacific anomalies are thermobaric
expressions of flows confined within the asthenosphere?
Even morechallenging are the deep-mantle anomalies and
basal (region D") heterogeneities of Table 1. Of these features
very little is known, but we can engage in some fanciful spec-
ulations. Significant mass flux across the core-mantle interface
is presumably inhibited by its large density contrast (t4300
kg/mi3), which substantially exceeds that at the Earth's surface.
If the lower mantle is convecting, due in part to heat flux from
thecore, then there will be a thermal boundary layer at its base,
probably occupying a significant portion of region D" (97).
Basal heterogeneities could thus be thermobaric variations
within this boundary layer, and at least some of the anomalies
inregion D' could be caused by its detachment and rise through
the mantle. We would expect anomalies of this sort to be
strongest near the boundary layer, which could explain the
apparent increase in heterogeneity below about 2000 km in
The analogy with upper mantle heterogeneity is obvious and
attractive, and perhaps it can be extended even further. At the
surface we find not only a thermal boundary layer but also a
chemical boundary layer-the crust-inhomogeneous on many
scales, including microscale variations that scatter seismic
waves. At the base of the mantle we again find a layer inhom-
ogeneous on many scales, where scattering by microscale
variations is also observed. Could these variations indicate the
existence of another chemicalboundary layerat the Earth's
othermajorchemical transition? Wemight postulatea solidslag
on the core'sliquid iron surface too dense to beeasily trans-
portedout ofregionD"byconvectiveaction, justas the con-
tinents remain on mantle's outer surface. Such a crust could
provide isostatic support for any subtle topography on the
core-mantle interface (a more dynamically feasible scenario
than uncompensated topography). Mountain ranges could be
piled up and huge rafts pushed about, perhaps in a tectonics not
totally strange to the surface geologist.
This research was supported by National Science Foundation Grant
Dutton, C. E. (1882) U.S. Geol. Surv. Monogr. 2.
Gilbert, G. K. (1890) U.S. Geol. Surv. Monogr. 1.
Dutton, C. E. (1889) Bull. Philos. Soc. Wash. 11,54-64, reprinted
in (1925) J. Wash. Acad. Sci. 15, 359-369.
Menard, H. W. (1973) Eos 54, 1244-1255.
Daly, R. A. (1940) Strength and Structure of the Earth (Pren-
tice-Hall, Englewood Cliffs, NJ).
Isacks, B., Oliver, J. & Sykes, L. R. (1968) J. Geophys. Res. 73,
Goldreich, P. & Toomre, A. (1969) J. Geophys. Res. 74, 2555-
Dicke, R. H. (1969) J. Geophys. Res. 74,5895-5902.
Morgan, W. J. (1972) Am. Assoc. Pet. Geol. Bull. 56,203-213.
Cathles, L. M. (1975) The Viscosity of the Earth's Mantle
(Princeton Univ. Press, Princeton, NJ).
McKenzie, D. & Weiss, N. (1975) Geophys. J. R. Astron. Soc. 42,
Jordan, T. H. (1975) Nature (London) 257,745-750.
O'Connell, R. J. (1977) Tectonophysics 38, 119-136.
Davies, G. F. (1977) Geophys. J. R. Astron. Soc. 49,459-486.
Bullen, K. E. (1963) Introduction to the TheoryofSeismology,
(Cambridge Univ. Press, New York), 3rd Ed.
Johnson, L. R. (1969) Bull. Seismol. Soc. Am. 59,973-1008.
Anderson, D. L., Sammis, C. & Jordan, T. H. (1971) Science 171,
Jordan, T. H. & Anderson, D. L. (1974) Geophys. J. R. Astron.
Raitt, R. W., Shor, G. G., Morris, G. B. & Kirk, H. K. (1971)
Tectonophysics 12, 173-186.
Forsyth, D. W. (1975) Geophys. J. R. Astron. Soc. 41, 103-
Shimamura, H. & Asada, T. (1978) Eos 59, 1133 (abstr.).
Fuchs, K. (1977) Geophys. J. R. Astron. Soc. 49, 167-179.
Aki, K. (1969) J. Geophys. Res. 74,615-631.
Cleary, J. R., King, D. W. & Haddon, R. A. W. (1975) Geophys.
J. R. Astron. Soc. 43, 861-872.
King, D. W., Haddon, R. A. W. & Husebye, E. S. (1975) Phys.
Earth Planet. Inter. 10, 103-127.
Haddon, R. A. W., Husebye, E. S. & King, D. W. (1977) Phys.
Earth Planet. Inter. 14, 41-70.
Aki, K., Christoffersson, A. & Husebye, E. S. (1977) J. Geophys.
Res. 82, 277-296.
Aki, K. (1977) J. Geophys. 43, 235-242.
Savino, J. M., Rodi, W. L., Goff, R. C., Jordan, T. H., Alexander,
J. H. & Lambert, D. G. (1977) Inversion of Combined Geo-
physical Data for Determination of Structure Beneath the
Imperial Valley Geothermal Region, Technical Report SSS-
R-78-3412 (Systems, Science and Software, La Jolla, CA).
Haddon, R. A. W. & Husebye, E. S. (1978) Geophys. J. R. Astron.
Soc. 55, 19-43.
Jordan, T. H. (1979) Sci. Am. 240 (1), 92-107.
Umbgrove, J. H. F. (1947) The Pulse of the Earth (Nijhoff, The
Brune, J. & Dorman, J. (1963) Bull. Seismol. Soc. Am. 53,
Knopoff, L. (1972) Tectonophysics 13, 497-519.
Landisman, M., Dziewonski, A. & Sat6, Y. (1969) Geophys. J. R.
Astron. Soc. 17,369-403.
Schlue, J. W. & Knopoff, L. (1977) Geophys.J. R. Astron. Soc.
Sipkin, S. A. & Jordan, T. H. (1976) J. Geophys. Res. 81,
Proc. Natl. Acad. Sci. USA 76 (1979)
Jeffreys, H. & Bullen, K. E. (1940) Seismological Tables (Br.
Assoc. Adv. Sci., London).
Sengupta, M. K. (1975) Dissertation (Massachusetts Inst. Tech-
nol., Cambridge, MA).
Duschenes, J. D. & Solomon, S. C. (1977) J. Geophys. Res. 82,
Leeds, A. R. (1975) Phys. Earth Planet. Inter. 11, 61-64.
Canas, J. A. & Mitchell, B. J. (1978) Bull. Seismol. Soc. Am. 68,
Eaton, J. P. (1962) in Crust of the Pacific Basin, Monograph 6,
eds. Macdonald, G. A. & Kuno, H. (Am. Geophys. Union,
Washington, DC), pp. 13-29.
Walcott, R. I. (1970) J. Geophys. Res. 75,3941-3954.
Parsons, B. & Sclater, J. G. (1977) J. Geophys. Res. 82, 803-
Toksoz, M. N. & Anderson, D. L. (1966) J. Geophys. Res. 71,
Dziewonski, A. M. (1971) J. Geophys. Res. 76,2587-2601.
Okal, E. A. & Anderson, D. L. (1975) Geophys. Res. Lett. 2,
Jordan, T. H. (1975) Rev. Geophys. Space Phys. 13 (3), 1-12.
Sacks, I. S. & Okada, H. (1974) Phys. Earth Planet. Inter. 9,
Sacks, I. S. & Snoke, J. A. (1977) J. Geophys. Res. 82, 2011-
Alexander, S. S. (1974) Eos 55, 358 (abstr.).
Okal, E. A. (1977) Geophys. J. R. Astron. Soc. 49,357-370.
England, P. C., Kennett, B. L. N. & Worthington, M. H. (1978)
Geophys. J. R. Astron. Soc. 54,575-585.
Nolet, G. (1975) Geophys. Res. Lett. 2,60-62.
Cara, M. (1978) Geophys. J. R. Astron. Soc. 54,439-460.
Chou, T. A. & Dziewonski, A. M. (1978) Eos 59, 1141 (abstr.).
Jordan, T. H. (1978) Nature (London) 247,544-548.
Sipkin, S. A. & Jordan, T. H. (1975) J. Geophys. Res. 80,
Hart, R. S. & Butler, R. (1978) Bull. Seismol. Soc. Am. 68,
Davies, D. & McKenzie, D. P. (1969) Geophys. J. R. Astron. Soc.
Mitronovas, W. & Isacks, B. L. (1971) J. Geophys. Res. 76,
Utsu, T. (1971) Rev. Geophys. Space Phys. 9,839-890.
Toksoz, M. N., Minear, J. W. & Julian, B. R. (1971) J. Geophys.
Res. 76, 1113-1138.
Sleep, N. H. (1973) Bull. Seismol. Soc. Am. 63, 1349-1373.
Sacks, I. S., Snoke, J. A. & Linde, A. T. (1976) Carnegie Inst.
Washington Yearb. 75,216-223.
Barazangi, M., Pennington, W. & Isacks, B. (1975) J. Geophys.
Res. 80, 1079-1092.
Suyehiro, K. & Sacks, I. S. (1978) Carnegie Inst. Washington
Yearb. 77, 494-504.
Stauder, W. (1975) J. Geophys. Res. 80, 1053-1064.
Jordan, T. H. (1977) J. Geophys. 43, 473-496.
Fitch, T. J. (1977) in Island Arcs, Deep Sea Trenches and
Back-Arc Basins, eds. Talwani, M. & Pitman, W. C. (Am. Geo-
phys. Union, Washington, DC), pp. 123-136.
Sondergeld, C. H., Isacks, B. L., Barazangi, M. & Billington, S.
(1977) Bull. Seismol. Soc. Am. 67,537-541.
England, P. C., Worthington, M. H. & King, D. W. (1977) Geo-
phys. J. R. Astron. Soc. 48, 71-79.
Jeffreys, H. (1962) Geophys. J. R. Astron. Soc. 7,212-219.
Chinnery, M. A. (1969) Phys. Earth Planet. Inter. 2, 1-10.
Davies, D. & Sheppard, R. Mv. (1972) Nature (London) 239,
Hales, A. L. & Roberts, J. L. (1970) Bull. Seismol. Soc. Am. 60,
Jordan, T. H. (1972) Dissertation (California Inst. Technol.,
Jordan, T. H. & Lynn, W. S. (1974) J. Geophys. Res. 79,2679-
Engdahl, E. R. (1975) Geophys. Res. Lett. 2,420-422.
Weichert, D. H. (1972) Earth Planet. Sci. Lett. 17, 181-188.
Sheppard, R. M. (1973) Seismic Discrimination Semi-Annual
Technical Summary, EDS-TR-73-175 (Lincoln Laboratory,
Powell, C. (1976) Dissertation (Princeton Univ., Princeton,
Julian, B. R. & Sengupta, M. K. (1973) Nature (London) 242,
Dziewonski, A. M., Hager, B. H. & O'Connell, R. J. (1977) J.
Geophys. Res. 82,239-255.
Alexander, S. S. & Phinney, R. J. (1966) J. Geophys. Res. 71,
Haddon, R. A. W. (1972) Eos. 53,600 (abstr.).
Cleary, J. R. & Haddon, R. A. W. (1972) Nature (London) 240,
Haddon, R. A. W. & Cleary, J. R. (1974) Phys. Earth Planet Inter.
Wright, C. (1975) Bull. Seismol. Soc. Am. 65,765-786.
Husebye, E. S., King, D. W. & Haddon, R. A. W. (1976) J. Geo-
phys. Res. 81, 1870-1882.
Doornbos, D. J. (1978) Geophys. J. R. Astron. Soc. 53, 643-
Chang, A. C. & Cleary, J. R. (1978) Bull. Seismol. Soc. Am. 68,
Sacks, I. S. & Beach, L. (1974) Carnegie Inst. Washington Yearb.
Snoke, J. A. & Sacks, I. S. (1977) Carnegie Inst. Washington
Yearb. 76, 233-239.
Jordan, T. H. (1979) in The Mantle Sample: Inclusions in
Kimberlites and Other Volcanics, Proceedings of the Second
International Kimberlite Conference, eds. Boyd, F. R. & Meyer,
H. 0. A. (Am. Geophys. Union, Washington, DC), Vol. 2, pp.
Elsasser, W. M., Olson, P. & Marsh, B. D. (1979)]. Geophys.Res.