The diapirism suite includes true diapirs (pierced-up-folds), migrating spines(commonlyof magma halite, or clay), diapiric ridges (of magma or halite), more-or-less vertical sheets (igneous dykes), volcanoes, lava flows, salt extrusions, salt domes, salt-spine lakes, sandstone dykes, and some aspects of boudinage. Plastic spines may migrate upwards, or downwards; they move in response to either a massdensitydifferential in a gravityfield, or a pressure differential across a plasticity gradient; the latter drives most or all halite spines, and may drive many if not all magma spines.

The normal fault suite includes normal (or gravity) faults, grabens, half-grabens, antithetic faults (isolated graben fragments) graben-edge roll-overs, upside-down grabens (necking), sag basins, some aspects of boudinage, and others. Halite ridges occupy the space made by necking (upside-down grabens).

Both suites are specifically the product of simple tension (that is, primary tension, rather than tension combined more or less equally with shear). That tension is ordinarily developed in a horizontal sense. Primary compression produces features from these suites only in a higher-order sense (that is, first order compression may give rise to second order or third order, tensional features). The sizes of the individual members of the suite serve as a reasonably good index to the lowest order of the system. Thus major deep-sea trenches, sag basins and volcano fields - taken together - indicate primary simple first-order horizontal tension, with or without a secondary shear component.

Because both suites develop in a tension field, a great deal of overlap should be expected. Grabens, for example, may or may not be accompanied by volcanism. The relationship is determined by the nature of the structural connection between the graben, on the one hand, and the potential magma source, on the other.

A wide variety of scaled dynamic models has praduced much evidence to help clarify the production of members of both suites a great deal of this evidence can be confirmed by examination of the prototypes. The model work has yielded, despite considerable effort, not a single example of anything that could be called a descending slab, but has provided many examples of the various features in the two suites, all of them the result of horizontal tension.

A diapir is an upfold which has been pierced, or ruptured, by the upward movement of a plastic core. The word is derived from a compound Greek work having a stem peiro meaning "to pierce", but carrying the idea dia of "through" or "completely" rather than merely "to indent". It is, therefore, a more narrowly defined word than is "piercement structure", in view of the fact that "to pierce" means either to punch a hole into, or to punch a hole through.

The diapir is the ruptured fold, not the plastic core, and not the spine or needle- shaped body which one may observe after several successive layers have been ruptured. Spines may be composed of various materials, such as halite, clay, sand, or magma. Upward-moving halite produces salt spines, associated with well-known salt domes; but spine, diapir, and dome are three different concepts. Upward- migrating magma may or may not reach the earth's surface; if it does reach the surface , it produces a volcano, a line of volcanoes, or a sheet flow. A volcano can be defined as the eruption which results when a magmatic spine reaches the earth's surface. The extent to which this spine pierces a nested set of up- folds is far from clear, however.

A salt spine which reaches the surface may give rise to a lake (humid climate) or to a partly-breached topographic dome, or to a salt hill (dry climate). The more-or-less circular map pattern which we coromonly associate with salt domes is not the only shape available. Many piercing masses are roughly sheet-shaped; if composed of magma, they chill to form (generally) igneous dykes, and if composed of sand (alternatively, sandstone) they may be seen in the field as one of two classes of sandstone dyke (the other is produced by sedi- mentation, under water, from above).

Furthermore, some plastic masses form neither spines nor sheets, but wedge-shaped bodies (ridges) or lunate fold fills. One known example is the halite, or salt, ridge observed in the subsurface in the Gulf Coast area of the United States, where the halite mass is shaped something like a very long army, or campaign, tent, having a fairly clearly defined crest line and two oppositely-sloping sides (Atwater and Forman, 1959). Such a ridge, in certain cases, may give rise to individual spines, which grow from the crest.

The phenomenon of diapirism is associated with a geometric system which includes breached upfolds (diapirs), which may or may not be nested; spines and sheets; halite ridges; and certain surface and near-surface manifestations (volcanoes, lava flows, salt-dome lakes, salt extrusions, salt domes, igneous dykes, sand- stone dykes). Not all of these can be seen at any given locality, surface or subsurface. In many places, the list of evidence types may be small.

Therefore when we discuss diapirism, we necessarily treat a wide variety of subjects other than diapirs.

The present author has been carrying on a program of studies of diapirism for years. This has involved some field work, some subsurface analysis, and a good deal of dynamic mechanical modelling, using principles which are thought to produce dynamically scaled results (Hubbert, 1951; Tanner, 1973c).

The purpose of this work has been to elucidate, if possible, the interrelationships between various styles of deformation (such as compression, shear, and tension), on the one hand, and sedimentation, on the other hand. This purpose fits well with the larger goal of clarifying the interactions between tectonism, in general, and sedimentation. The overall program has produced no single underlying "law" or algorithm, ror was such expected; however, valuable results have been obtained (for example, the use of sequences of transport vec- tors, such as cross-bedding, to provide details of deforroational history; Tanner, 1971a).

An early part of the model work was designed merely to see if the field geologist should believe the old saying that compression, shear and tension are merely three aspects of the same deformational system, and cannot be distinguished from each other in hind-sight. In the models, at least, it is commonly a fairly sirople roatter to distinguish them from each other after the fact (but with no know- ledge of the stress system used). As the model designs evolved, one of the intermediate goals became the matter of fitting diapirism into a deformational pattern of some kind.

With this goal, a long series of models was tested in order to compare tensional, com- pressional and shear effects (Tanner and Williams, 1968; Tanner, 1973a). Each of these models incorporated at least two different materials, one of which (pitch) had the proper- ties necessary for it to model halite spines, if such were possible within the operating rules. Many models were built and operated for each of the three primary deformational cate- gories (compression, shear, tension). The results of this diapirism series were quite clear: compressional models produced no visible diapirism (spines, sheets, ridges, or diapirs); shear models produced very little in that category; and tension models yielded a rich variety. This variety included ridges and spines having the correct geometries (including spacing) for matching with Gulf Coast sub- surface ridges and spines; also included were many grabens and single faults. One model series included a thermal gradient, and pro- duced "magma" accumulations in low pressure areas, but was not designed to create "volcanoes". From the model work, one should conclude that extensive diapirism indicates a pervasive tensional stress field (although not necessarily of first order).

Diapirs (ruptured folds) were not observed in this series, at least in part because (a) the scale was not favourable for that pur- pose, and (b) the background material (dry, powdered drilling mud) was not suitable. How- ever, other models, having a distinct strati- graphy and a more favourable scale, have shown diapirs (breached folds).

The models require a great deal of time, patience and care, hence were built and oper- ated by a team of investigators rather than by one person. One of the team cherished the idea that diapirism is forced (in a geometric- ally highly limited sense, it is), and there- fore that salt spines should be produced in compression models. He was, accordingly, en- couraged to operate a separate series of experiments, with permission to modify the model in any way he wished in an effort to obtain diapirism. No model grabens or salt ridges or spines developed in that series.

In the case of magmatic diapirism, one might argue that molten magma either melts or dissolves its way upward. In the case of halite, clay, or sand, these arguments cannot be used. From the geometry of the diapirism models, one should reject ideas of melting, dissolution, or forcible injection.

Instead, the ridges and spines were pro- duced at those points where combined pressure (both geostatic and dynamic) was least; at these locations, the most plastic material available (pitch or other suitable material) flowed toward the lowest pressure. That is, highly plastic materials (such as halite, clay or magma) move down the pressure gradient in a less plastic medium (but perhaps upward, against gravity), exploiting fractures and faults if and where they exist. Initially, plastic matter in a flat-lying but thick bed (e.g. halite) tends to be swept into ridges; these are commonly located specifically under- neath surface grabens, and result in a measur- able thinning (between graben floor and ridge crest) of the substance between them. Their geometry appears to indicate the presence of the graben above. Spines appear either after ridges have begun to form, or in cases where ridges do not form; they are located along or close to graben boundary faults, or other faults, or in positions where presumably they are exploiting minor fracture intersections.

Various other model experiments have been run (e.g. Tanner, 1962) in an effort to clarify some of the problems. Out of all of the model work, certain conclusions can be drawn:

1. The size of the feature is an iroportant consideration. For example, boudinage, itself, indicates a higher order stress field (in the Moody-Hill sense) than does, for example, major grabens or salt domes.

2. Graben-and-horst fields indicate simple tension, whereas en echeZon grabens (with or without associated horsts and upwarps) reflect shear having minor or incidental tension. Iso- lated grabens of higher order may develop in a compressional field (e.g. on top of an anti- cline).

3. Sand dykes are likely to be third order, or higher; mud lumps and clay plugs are prob- ably third order or higher but may be second order locally; a well-developed set of halite spines is likely to be first order or perhaps second order; and magma spines indicate a first order tensional stress field.

Many tensional features do not fit into any of the categories given above. For example, sag basins occur at various scales in many parts of the world, but their occurrence has not been systematized properly, and they are commonly ignored in structural treatises. The largest basins which can be put, with assurance, into the sag group are what the present writer has called "middle-sized basins" (that is, somewhat smaller than major ones such as the Pacific and Atlantic; Tanner, 1968a). An out- standing example is the Gulf of Mexico, but others are known along the eastern edge of Asia (Karig, 1971; Tanner, 1968b, 1974), and else- where.

As one goes down the size scale, sag basins go through a marked change in geometry. Middle-sized basins such as the Gulf of Mexico (Tanner, 1971b) are so large, in terms of rock properties, that they cannot be bounded by one or two sharply-defined faults, and instead may not have any obvious structural boundaries at all. Sag basins that are much smaller, however, such as Reelfoot Lake, in Tennessee (Hodgson, 1964) and Lake Okeechobee, in Florida (Bond, Tanner and Smith, 1981) may be outlined by one or two faults, and also may have sides where only minor tilting of the ground surface can be seen. In the case of Lake Okeechobee, formed within the 60o angle between two faults, the orientation of the tensional axis can be drawn reasonably accurately. The small sag basins which dot the valley floor of the lower Magdalena River, in Colombia (Tanner, 1973b) were forroed in alluvium having properties such that fault traces are apparently not preserved, and stress orientations cannot be deduced readily.

The deep-sea trenches of the world look, at first glance, like large isolated grabens. There have been, over the years, two main hypo- theses to account for them. The first, popular during the 1940's and 1950's, was the tectogene hypothesis; according to this, each trench is the sea-floor expression of a tightly-closed syncline caused by compression at right angles to the trench. The second, popular during the 1970's and 1980's, is an essential part of the sea-floor spreading scheme; in this version, compression at right angles to the trench axis is attributed to some aspect of sea-floor translation from spreading centres (mid-ocean ridges) to areas of crustal destruction (sub- duction zones).

Precisely how it is that the trenches look so much like grabens, when they are supposed to be directly due to lateral compression, might be "explained" in several different ways. One of these (a) envisages the crust as rolling over, assuming a vertical or steep dip, and then falling into the mantle under its own weight, tearing away from that portion of the crust still remaining at the earth's surface. In this version, the subducted crust is falling, and the trench is the pull-apart structure left behind. Another, (b) explains the trench as the true graben that is formed on the crest of the flexure, or tilted anticline, where the crust bends so that it can drop into the mantle as it is being destroyed or consumed. The second of these two makes the trench an "isolated graben of higher order" located "on top of an anticline", as was specified in an earlier paragraph, thereby presumably reconcil- ing it with the assumed compression field.

A third (c), not stated anywhere explicit- ly as far as I know, recognizes a series of normal faults, parallel with each other, staggered in such a fashion that the youngest is topographically highest and also farthest seaward, whereas the oldest is topographically lowest, farthest landward, and buried beneath trench-floor sediment or sub-trench crust. The oldest is therefore structurally (but not topo- graphically) higher than the youngest. In this scheme, the set of normal faults make up a com- posite fault zone, the latter having a dip which is significantly less than that of the average single fault. Presumably fault iden- tity would be lost as any given slice of crust moved downward into the mantle.

The first of these three skirts the question of how we can have a permanent, con- tinuous gap in the "conveyor belt" that the crust is supposed to be, from spreading centre to, and into, the subduction zone; that is, how it is that the subducted crust can pull away from the horizontal crust, leaving a gap which no part of the crust is crossing. One possible way around this dilemma is to postu- late that we are now studying the system in an atypical time, and that "tomorrow" some more horizontal crust must be rammed into the trench, obliterating it temporarily. This, in turn, introduces several ogres whom we may not wish to face.

The second poses geometrical questions which will be dealt with subsequently. The third might be applied in areas where the sub- duction zone is thought to have rather low dips (e.g. about 45a or lower), but does not seem applicable elsewhere (the zone must dip less steeply than normal faults). However, normal fault planes are typically concave- upward, except where parallel or subparallel faults merge at depth. The "venetian blind" system visualized here appears to be more use- ful in a very narrow or near-surface setting than at any great depth in the crust or mantle.

The present author (Tanner, 1973a] review- ed seismological and other evidence which shows that deep-sea trenches are basically tensional, in a horizontal sense (not a vertical sense), and that they fit into the normal-fault suite of middle-sized sag basins, volcanic activity, smaller grabens, and high marginal mountains (Tanner, 1972) which characterize many conti- nental margins. This kind of reasoning is rejected by advocates of sea-floor spreading, who find that subduction is the only imaginable way to dispose of all the tremendous amount of crust that they are generating along mid-ocean ridges.

A basic requirement of the subduction idea can be stated in the following form, where the sentence has been designed to bring out some of the contradictions which are involved: "Hori- zontal compression lets a lighter crust fall freely through a heavier mantle." Some advo- cates of subduction will argue that the crust is indeed heavier than the mantle; but no evi- dence has been collected in support of this point of view, and in fact the evidence is vol- uminous that the crust is lighter; the only support for a heavier crust comes from the hypothesis itself, an inadmissable source. Other advocates may choose to argue that the crust does not fall; rather, it is shoved down (somewhat after the manner of the tectogene, except that the latter buckled, whereas the subducted slab is not supposed to buckle).

This horizontal compression, combined with falling (or shoving) is supposed to leave a trench, or trough, at the surface. However, the following possibilities roust be considered:

a. If the trench is a true first-order graben, standing low, then the first-order stress field must be dominated by horizontal tension (not by vertical compression, which is not the same thing as horizontal tension, despite many statements to the contrary).

b. If the trench is a second-order graben, standing high because it lies on or close to the axis of an arch, then the scales are un- acceptable. A little bit of trigonometry shows that, for a strip between 400 and 500 km wide, 2 km of up-arching (above the earth's surface) produces only about 100 m of vertical offset in the graben, and approximately 40 km of up-arching is needed to get 10 km of drop within the graben. These numbers show that the grabens in Japan, for example, may well be tle result of arching, but the trench, offshore, cannot be.

c. If the trench is a graben-like feature, due to flexing, it is a more-or-less symmetrical structure having a vertical syrometry axis, developed from a fold having a 45o (more-or- less) axial plane, and located to one side of, rather than on, the presumed flexed material. This geometry is, as far as is now known, impossible.

d. If the trench is a secondary graben, located between a primary normal fault and an anti- thetic fault, then one of the first three restrictions applies.

e. If the trench is the result of a roll-over, such as is commonly found along the edges of many graben floors, giving them a whale-back shape, then a restriction from among the first three must apply.

The geometry of deep-sea trenches indi- cates that they are priroary and first-order grabens; that is, that they are due to first- order horizontal tension. The detailed work of Bowin (1968) on the Cayman trench, later repeated, amplified, or confirmed by other workers (e.g. Erickson et aZ., 1972), shows that' this is true. This trench is not an anomaly, as soroe have claimed, but is precisely what should be found (Tanner, 1973a), is represent- ative of model results, and matches well other deep-sea trenches, which also conform with model results. There is at present no rational alternative model; and present conceptual res- trictions and model limitations do not permit a radical alternative.

Peter, Elvers and Yellin (1965) produced a geophysical cross-section of the Aleutian trench, showing several interesting features. They showed a crust, under the trench, only 4 km thick (e.g. the thinning known in metal- lurgy as "necking") and a ridge of igneous rock underlying the north slope. This ridge is supposed to have risen to fill a "fissure" in the crust. This is the kind of structure observed in the tensional models. Instead of halite, however, igneous material played the same role, rising from below (not necessarily as a fully-developed magma) to fill a low pressure area above.

In the many series of models which have been operated at Florida State University, one set was designed to test the idea that a slightly heavier surface layer would, either with or without shoving from the side, sink into the lighter underlying material, and duplicate a subducted slab. It should be obvious that a lighter surface layer will not sink into the lower layer, nor can it be shoved downward in a thin sheet. Therefore the model contained a built-in error: the upper layer was heavier than the lower one. It is custom- ary, when working with an inherently unstable design such as this, to construct the model in inverted position, to freeze it, to turn it back right-side up, and to let it thaw at room temperature. This is the procedure that has been followed with many models designed to ex- plore the diapirism phenomenon. In an effort to duplicate the down-going slab of the subduc- tion hypothesis, models of this kind were built and operated. No slab of any kind was pro- duced. Instead, upside-down diapirs developed, with the spines composed of thinner layer material (the surface heavy layer); they migrated downward (because of the mass density differential, in this case). (Ordinary diapir- ism develops from a thin lower layer, which migrates upward either because of a density differential, or because a plastic medium is squeezed into low-pressure strips or spots. This is, of course, what one should expect.) It is quite instructive, after doing the necessary algebra for model design, to watch the actual operation of the system, and to realize that, whatever else may happen, a thin down-going slab is mechanically the least probable of all models proposed to date. A row, or field, of inverted diapirs, is more realis- tic, by far.

It is also helpful to consider some other aspects of the model work. A mass density differential of about 1.6 (c.g.s. units; 0.9 subtracted from 2.5) in a compression field did not produce any diapirism at all. The sub- duction idea reyuires that a vanishingly small mass density differential (probably running the wrong direction) drive massive subduction through a horizontal compression field, a proposition that is ludicrous.

BENIOFF ZONE DIAPIRISM

On the other hand, if the deep-sea trench is indeed a graben, surely there are plastic materials at depth, and diapirs should be found in the same region. From the global standpoint, the piercing spines should not be halite or clay (which do not occur universally), but should be magma (which either occurs in, or is generated in, a tension field). The depth of generation of magma is on the order of 50-100 km, or more, and perhaps much more. With Benioff Zone dips of 30o to 60o, the horizontal distance from the trench to the volcanic belt (magmatic diapirs) should be as follows:
 

The width typically encountered in island-arc areas is between 75 and 200 km, indicating, from the table, a depth of generation of 50 to 200 km.

The sea-floor spreading hypothesis is supposed to require destruction of crust at deep-sea trenches, and this destruction is supposed to take place by means of consumption (at great depths) of a down-going slab. But extensive model work on diapirism and graben formation has shown that all of the gross sur- face geometry, all of the known major internal structures, and all of the diapirism is produced by important (first order) horizontal tension. This result is obtained without violating mass density relationships, amounts of fault throw or heave, stress directions, or pertinent angles. The subduction concept stands in opposition at all essential points, and therefore must be in error.

The expanding earth hypothesis yields general first-order horizontal tension, but in the wrong places and patterns, and therefore cannot be correct, either. Hence it is necess- ary to continue to look for an improved model, having greater explanatory power.

The diapirism suite of structures and the normal-fault suite of structures are evidence of horizontal tension at the appropriate scale and order. The two suites overlap, but do not occupy precisely the same field of rock and deformational parameters. Scaled dynamic model work has contributed a great deal of insight into how some of these structures come into existence and how they evolve, but has not produced any information to support the concept of subduction.

Geological, geophysical and model work, taken together, require that deep-sea trenches, island arcs, and their associated diapiric and normal-fault structures, be the result of first-order horizontal tension.

Atwater, G.,and Forman, M.J., 1959: Nature of growth of southern Louisiana salt domes and its effect on petroleum accumulations. Am. Assoc. Petr. GeoZ., 8u22., 43: 2592- 2622.

Bond, P., Tanner, W.F., and Smith, L., 1981: Structural patterns in South Florida. GuZf Coast Assoc. GuZf Soc., Trans., 31: in press.

Bowin, C.O., 1968: Geophysical study of the Cayman trough. Journ. Geophs. Res., 73: 5159-5173.

Erickson, A.J., Helsley, C.E.,and Simmons, G., 1972: Heat flows and continuous seismic profiles in the Cayman trough and Yucatan basin. GeoZ. Soc. Am., BuZZ., 83: 1241- 1260.

Hodgson, J.H., 1964: Earthquakes cmd Earth Structure. Prentice-Hall, New York: 166 p.

Hubbert, M.K., 1951: Mechanical basis for certain familiar geologic structures. GeoZ. Soc. Am., BuZ2., 62: 355.

Karig, D.E., 1971: Origin and development of marginal basins in the western Pacific. Journ. Geophs. Res., 76(11): 2542-2561.

Peter, G., Elvers, D., and Yellin, M., 1965: Geological structures of the Aleutian trench southwest of Kodiak Island. Jourrc. GeophzJs. Res., 70: 353-366.

Tanner, W.F., 1962: Surface structural patterns obtained from strike-slip models. Journ. GeoZ., 70: 101-107.

Tanner, W.F., 1968a: Tensional basins on the east edge of Asia. GeoZ. Soc. Japan, Jourm., 74: 583-588.

Tanner, W.F., 1968b: Origin of the Gulf of Mexico, II; additional data. GuZf Coast Assoc. GeoZ. Socs., Trans., 18: 98-107.

Tanner, W.F., 1971a: Poleward drift of North America. GeoZ. Rundsch., 60: 848-867.

Tanner, W.F., 1971b: Sediment-vector sequences and tectonism. The Mountain GeoZogist, 8(3): 151-153.

Tanner, W.F., 1972: High marginal mountains: evidence for tension. Pacific GeoZogg, 5: 1-5.

Tanner, W.F., 1973a: Deep-sea trenches and the compression assumption. Am. Assoc. Petr. GeoZ.,.BuZ2., 57(11): 2195-2206.

Tanner, W.F., 1973b: The incomplete floodplain. GeoZog, 2(2): 105-106. Tanner, W.F., 1973c: Canons of tectonism. Geo2. Rundsch., 62: 969-988.

Tanner, W.F., 1974: The Mexican "island" arc. GuZf Coast Assoc. GeoZ. Soc., Trans., 24: 81-88.

Tanner, W.F,, and Williams, G.K., 1968: Model diapirs, plasticity and tension. In: J. Braunstein and G.D. 0'Brien (Ed.), D2apirism and Diapirs, A Symposium. Am. Assoc. Petr. Geol., Mem., 8: 10-15.