You are on page 1of 20

Journal of Volcanology and Geothermal Research 116 (2002) 279^298

www.elsevier.com/locate/jvolgeores

Emplacement and arrest of sheets and dykes in


central volcanoes
Agust Gudmundsson 
Geological Institute, University of Bergen, Allegt. 41, N-5007 Bergen, Norway
Received 18 December 2001; accepted 30 January 2002

Abstract

Sheet intrusions are of two main types: local inclined (cone) sheets and regional dykes. In Iceland, the inclined
sheets form dense swarms of (mostly) basaltic, 0.5^1 m thick sheets, dipping either at 20^50‡ or at 75^90‡ towards the
central volcano to which they belong. The regional dykes are (mostly) basaltic, 4^6 m thick, subvertical, subparallel
and form swarms, less dense than those of the sheets but tens of kilometres long, in the parts of the volcanic systems
that are outside the central volcanoes. In both types of swarms, the intrusion intensity decreases with altitude in the
lava pile. Theoretical models generally indicate very high crack-tip stresses for propagating dykes and sheets.
Nevertheless, most of these intrusions become arrested at various crustal depths and never reach the surface to supply
magma to volcanic eruptions. Two principal mechanisms are proposed to explain arrest of dykes and sheets. One is
the generation of stress barriers, that is, layers with local stresses unfavourable for the intrusion propagation. The
other is mechanical anisotropy whereby sheet intrusions become arrested at discontinuities. Stress barriers may
develop in several ways. First, analytical solutions for a homogeneous and isotropic crust show that the intensity of
the tensile stress associated with a pressured magma chamber falls off rapidly with distance from the chamber. Thus,
while dyke and sheet injection in the vicinity of a chamber may be favoured, dyke and sheet arrest is encouraged in
layers (stress barriers) at a certain distance from the chamber. Second, boundary-element models for magma chambers
in a mechanically layered crust indicate abrupt changes in tensile stresses between layers of contrasting Young’s
moduli (stiffnesses). Thus, where soft pyroclastic layers alternate with stiff lava flows, as in many volcanoes, sheet and
dyke arrest is encouraged. Abrupt changes in stiffness between layers are commonly associated with weak and partly
open contacts and other discontinuities. It follows that stress barriers and discontinuities commonly operate together
as mechanisms of dyke and sheet arrest in central volcanoes. : 2002 Elsevier Science B.V. All rights reserved.

Keywords: dyke injection; dyke arrest; magma chamber; stress ¢eld; layered crust; Young’s modulus

1. Introduction have now been carried out for more than a cen-
tury (Geikie, 1897; Harker, 1904, 1909). These
Systematic studies of dykes and inclined sheets have included extensive ¢eld studies as well as
theoretical proposals as to the mechanics of sheet
and dyke emplacement. While ¢eld observations
* Tel.: +47-5558-3521; Fax: +47-5558-9416.
dominated in the early part of the last century,
E-mail address: agust.gudmundsson@geol.uib.no the theoretical studies of Anderson (1936, 1951)
(A. Gudmundsson). formed a basis for understanding the stress trajec-

0377-0273 / 02 / $ ^ see front matter : 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 7 - 0 2 7 3 ( 0 2 ) 0 0 2 2 6 - 3

VOLGEO 2462 12-7-02


280 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

tories along which ideal sheets and dykes propa- model of the stress ¢elds of magma chambers and
gate. emplacement of dykes and inclined sheets that
There has been a renewed interest in the me- combines results from ¢eld observations, experi-
chanics of sheet and dyke emplacement in the past ments (through petroleum engineering studies),
30 years (Pollard, 1973; Pollard and Muller, and theoretical analysis. For this purpose, I ¢rst
1976; Halls and Fahrig, 1987; Parker et al., present generalised results on sheet and dyke ge-
1990; Baer and Heimann, 1995). The early work ometry based on studies of several thousand sheet
of this period focused on the static stress ¢elds intrusions in Iceland and Tenerife (Canary Is-
surrounding magma sources of various geometries lands). The emphasis is on the overall geometry
(Muller and Pollard, 1977; Delaney and Pollard, of swarms of dykes and inclined sheets, as well as
1982; Pollard and Segall, 1987), but later work those geometrical parameters of individual intru-
has explicitly dealt with the £uid dynamics of sions that are useful in developing and testing
magma £ow in ideal dykes (Wilson and Head, quantitative models on their emplacement. Sec-
1981; Spence and Turcotte, 1985; Spence et al., ond, I present analytical and numerical models
1987; Lister and Kerr, 1991; Rubin, 1995; Fialko of the stresses associated with magma chambers
and Rubin, 1999; Bolchover and Lister, 1999). and tips of sheets and dykes, and discuss how
Theoretical work on the propagation of, and these stresses may contribute to the arrest of the
£uid £ow in, dykes must be tested on, and sup- sheets and dykes or, alternatively, their propaga-
ported or refuted by, quantitative ¢eld studies of tion to the surface.
actual dykes. Most models to date assume a ho-
mogeneous and isotropic crust with either a con-
stant or linearly varying stress ¢eld. Thus, little 2. Sheet swarms and dyke swarms
attention has been paid to e¡ects of variations
in the elastic properties and tensile strength of 2.1. General
the crust, its layering, and e¡ects of discontinu-
ities and abrupt stress changes on the propagation In regions with central volcanoes, there are
and arrest of dykes and sheets. For example, it is commonly two types of swarms of sheet intru-
commonly assumed that their rate of propagation sions : swarms of regional, subvertical and thick
is primarily controlled by the viscosity of the mag- dykes, and swarms of local, inclined and thin
ma and that fracture toughness (or, alternatively, sheets. Regional dykes have been studied for a
the tensile strength) of the host rock can be long time in the British Isles (Geikie, 1897), but
ignored (Spence and Turcotte, 1985). While this the identi¢cation of inclined (cone) sheets was ¢rst
may be correct for homogeneous, isotropic rocks, made in Skye by Harker (1904). The regional
it is certainly not so for layered rocks. Not only is dykes normally occur in elongate swarms, some
it common to see the e¡ects of layering in that of which may reach 1000^2000 km in continental
sheets and dykes are diverted out of their path- areas (Ernst et al., 1995), and at least hundreds of
ways at discontinuities, and have to expand on kilometres in the oceanic areas, as inferred from
entering layers with high Young’s moduli or ten- ophiolites and active ocean-ridge segments. A
sile strengths, but it is also one of the basic con- large part of the oceanic crust at divergent plate
clusions of thousands of hydraulic fracture experi- boundaries is generated by dykes, and their me-
ments in petroleum engineering that sharp chanics of emplacement and arrest is thus of great
contrasts in Young’s modulus, tensile strength importance for understanding crustal formation.
(or toughness), and fracture-normal compressive By contrast, swarms of inclined sheets are nor-
stress are very e¡ective in arresting the tips of mally circular or slightly elliptical in plan view,
hydraulic fractures (Valko and Economides, and are usually con¢ned to central volcanoes
1995; Charlez, 1997; Yew, 1997) and, by implica- (Walker, 1992; Gudmundsson, 1998). Dyke com-
tion, dykes and sheets. plexes of many oceanic islands have characteris-
The principal aim of this paper is to provide a tics similar to those of swarms of inclined sheets

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 281

(Walker, 1992, 1999; Marinoni and Gudmunds- contain many thousands of sheets (Fig. 1), are
son, 2000). Large fractions of many central volca- almost exclusively con¢ned to central volcanoes,
noes consist of inclined sheets. Their mechanics of and are commonly associated with large gabbro
emplacement and arrest or, alternatively, propa- plutons. These plutons, exposed at depths reach-
gation to the surface, is important for the under- ing 2 km beneath the initial top of the lava pile,
standing of magma transport and volcanic hazard are the uppermost parts of shallow, crustal mag-
in central volcanoes. ma chambers that supplied magma to the sheets.
The following description of the geometry of Most sheets are basaltic. For example, only a few
inclined sheets and regional dykes is primarily acid and intermediate sheets were found in a
based on the author’s studies of sheets and dykes study of more than 500 sheets in the Hafnarfjall
in Iceland and Tenerife (Canary Islands). The fo- Central Volcano in West Iceland (Gautneb et al.,
cus is on the geometrical aspects that are consid- 1989; Gudmundsson, 1995a). A sheet swarm is
ered important for the propagation and arrest of normally circular or slightly elliptical in plan
sheets and dykes. The overall structures of local view and several kilometres in radius, its margins
sheet swarms and regional dyke swarms are di¡er- essentially coinciding with those of the central
ent, and so these are treated in separate subsec- volcano within which the swarm develops.
tions. However, both types of sheet intrusions are The attitude of the sheets is very variable within
hydrofractures, meaning that their propagation is each swarm, but becomes less so at greater dis-
driven by £uid (magmatic) pressure that is greater tances from the centre. Generally, the sheet strike
than the minimum compressive principal stress in has a roughly circular distribution (Fig. 2A).
the host rock. Most sheets dip towards the source chamber,
the common mean dips in swarms being between
2.2. Sheet swarms 45‡ and 65‡. Many swarms, however, show two
peaks in the dip distribution (Fig. 2B), at 75^90‡
Swarms of inclined sheets in Iceland normally and at 20^50‡ (Gudmundsson, 1998; Klausen,

Fig. 1. Dense swarm of inclined (cone) sheets in the valley Kalfafellsdalur in the Tertiary lava pile of Southeast Iceland (Gud-
mundsson, 1995a; Klausen, 1999). Most of the sheets dip towards the extinct shallow magma chamber of an extinct central vol-
cano. View northwest: essentially the entire wall consists of inclined, basaltic sheets, most of which have a gentle, apparent dip
to the southwest. The person standing to the right of the river provides a scale.

VOLGEO 2462 12-7-02


282 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

from the central part of the swarm to its marginal


parts at the rate of V16‡/km (Klausen, 1999).
The average sheet dip in Icelandic swarms is
steeper than that in swarms in the British Isles,
where the average dip is V45‡ (Billings, 1972;
Hills, 1972), and Gran Canaria (Canary Islands),
where the average dip is V41‡ (Schirnick et al.,
1999). This di¡erence in average dip is presum-
Fig. 2. Strike distribution (A) and dip distribution (B) of 515 ably related to di¡erent loading conditions. The
inclined (mostly basaltic) sheets in the Tertiary Hafnarfjall source chambers of the Icelandic sheets were sub-
Central Volcano in West Iceland. The frequency distribution
of sheet strike (A) peaks at N10^20‡E, which coincides
ject to internal £uid excess pressure, but also to
roughly with the NNE trend of the rift zone within which external absolute or relative tensile stresses related
this sheet swarm developed. The frequency distribution of to plate pull (Gudmundsson et al., 1997; Gud-
sheet dip (B) has two peaks: one at 75^85‡, the other at 35^ mundsson, 1998). By contrast, the source cham-
45‡ (Gautneb et al., 1989). bers of sheets in the British Isles and Gran Cana-
ria were presumably subject to internal £uid
1999). The steep dipping sheets are mostly in- excess pressure as the main, or the only, loading.
jected from the central part of the source magma For a given geometry of the source chamber, the
chamber (Figs. 3 and 4), whereas the shallow dip- trajectories of c1 , along which most sheets prop-
ping sheets are mostly injected from the marginal agate, are, on average, more shallow when the
parts of the chamber (Gudmundsson, 1995a, only loading is internal £uid excess pressure
1998; Klausen, 1999). A study of 1128 sheets in (Fig. 4) than when there is, in addition, external
the extinct Therartindur Swarm in Southeast Ice- horizontal tensile loading (Fig. 3).
land indicates that the average sheet dip decreases Sheet thickness ranges from a few centimetres

Fig. 3. Stress ¢eld around a magma chamber having a circu- Fig. 4. Stress ¢eld around a magma chamber having a circu-
lar vertical cross-section subject to a remote tensile stress of lar vertical cross-section subject to internal magmatic excess
5 MPa (excess pressure pe = 0). The ticks represent the direc- pressure pe = 5 MPa. The ticks represent the direction (trajec-
tion (trajectories) of the maximum principal compressive tories) of the maximum principal compressive stress, c1 ,
stress, c1 , along which inclined sheets injected from the along which inclined sheets injected from the chamber would
chamber would propagate. The stress ¢eld was calculated us- propagate. The stress ¢eld was calculated using the BEASY
ing the BEASY (1991) boundary-element program. The en- (1991) boundary-element program. The whole model is
tire model is shown in the inset where the crosses indicate shown in the inset where the crosses indicate the boundary
the boundary conditions of no displacement. Modi¢ed from conditions of no displacement. Modi¢ed from Gudmundsson
Gudmundsson (1998). (1998).

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 283

verses, consists of sheets (Gautneb et al., 1989).


Similarly, in 1000^1500-m-long pro¢les in the ex-
tinct Reykjadalur Central Volcano in West Ice-
land, 5.8^7.9% of the rock, and as much as 90%
in shorter traverses, is composed of sheets (Gaut-
neb and Gudmundsson, 1992). Next to extinct
magma chambers (mainly plutons of gabbro)
commonly the whole rock consists of sheets (Gud-
mundsson, 1998, Fig. 7). De¢ning intensity as
number of sheets (or dykes) per unit length of
pro¢le, the high-intensity parts of the Icelandic
swarms are very similar to parts of ophiolites as
regards sheet intensity and thickness range (Pallis-
ter, 1981; Lippard et al., 1986). The number of
sheets falls o¡ very rapidly with distance from the
margin of the extinct chamber, indicating that
most of the sheets never reached the surface to
feed volcanic eruptions but rather became ar-
rested at various depths.
Fig. 5. Changes in thickness (A) and dip (B) with distance Most sheets occupy essentially pure extension
from the centre of the Tertiary Reykjadalur Central Volcano fractures (mode I cracks). For most cross-cutting
in West Iceland. At a distance of about 9 km the mean
thickness of the sheets increases abruptly from around 1 m
relationships the only displacement on the earlier
to 3^4 m (A) and the mean dip of the sheets increases from sheet is opening perpendicular to the trend of the
30^50‡ to 70^80‡ (B). These changes occur where the region- later sheet (Fig. 6). The sheets commonly appear
al dykes take over from the local inclined sheets. Modi¢ed somewhat irregular in shape, partly because they
from Gautneb and Gudmundsson (1992). form their pathways by linking up of o¡set col-
umnar joints and other discontinuities in the host
to 14 m, but is most commonly 0.5^1 m. For rock. The sheet pathway is normally not a shear
example, the average thickness of 1128 inclined fracture formed by tectonic forces prior to the
sheets in the Thverartindur Swarm in Southeast sheet emplacement, although sheets occupying
Iceland is 0.86 m (Klausen, 1999). Compared with shear fractures have been reported. Field observa-
the regional dykes, sheets are, on average, much tions (Gudmundsson, 1995a,b) indicate that most
thinner and also have a much more restricted sheets form their own pathways when propagat-
thickness range. The thickness distribution is nor- ing into the host rock as magma-driven extension
mally a law with a negative exponent (Gud- fractures.
mundsson, 1995a; Klausen, 1999). The sheet
thickness range is generally similar to that ob- 2.3. Dyke swarms
served in many ophiolites (Pallister, 1981) and
oceanic islands (Walker, 1992, 1999; Marinoni In contrast to the sheet swarms, which are con-
and Gudmundsson, 2000). The thickness and ¢ned to central volcanoes, regional dyke swarms
dip of sheet intrusions increase abruptly at that extend for tens of kilometres from the central
distance from the centre of the associated volcano volcanoes. Some dykes may be fed, partly at least,
where the regional dyke swarm takes over from by the shallow chambers, but it is likely that most
the local sheet swarm (Fig. 5). regional dykes are fed by large magma reservoirs
The proportion of sheets in the swarms varies in the deeper crust or at the contact between the
widely. In 540-m-long pro¢les in the extinct Haf- crust and upper mantle (Gudmundsson, 1998,
narfjall Central Volcano in West Iceland, 6.1^ 2000). While both regional dykes and local sheets
10.9% of the rock, and up to 60% in shorter tra- propagate as hydrofractures and are mostly com-

VOLGEO 2462 12-7-02


284 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

Fig. 6. Cross-cutting, inclined sheets and subvertical dykes in the Laxardalur Central Volcano, of Plio^Pleistocene age, in South
Iceland. The cross-cutting relationships indicate that all these sheet intrusions are primarily mode I (extension) fractures. The
0.3-m-long hammer provides a scale.

posed of basalt, swarms of dykes and sheets have distribution of regional dykes is because their at-
several structural di¡erences, brie£y summarised titude is controlled directly by the essentially sta-
below. ble regional stress ¢eld (plate pull) associated with
Dyke swarms are typically 5^10 km wide and the rift zone within which they form.
around 50 km long, similar in dimensions to Ho- The regional dykes show a much smaller range
locene ¢ssure swarms (Gudmundsson, 2000). At in dip than the local sheets (Figs. 2 and 7). Shal-
the depth of exposure, commonly between 500 low dips of 20^50‡ (Fig. 2) are generally absent
and 1500 m below the initial top of the non- from the regional swarms where most dykes are
eroded lava pile, each dyke swarm contains hun- subvertical (Figs. 7 and 8). For example, the aver-
dreds of dykes. However, the dyke intensity in- age dip of 1547 regional dykes in East and South-
creases with depth in the lava pile (Walker,
1960). The majority, and presumably most, of
the dykes are non-feeders (Gudmundsson et al.,
1999; this paper). The geometry of the swarms
supports their being derived from elongate, ellip-
soidal, regional magma reservoirs located beneath
large parts of each volcanic system (Gudmunds-
son, 1998, 2000).
The strike distribution of most regional dyke
swarms follows a narrow normal curve, with a
standard deviation normally much less than that Fig. 7. Strike distribution (A) and dip distribution (B) of 700
of most local sheet swarms (Figs. 2 and 7). Re- regional basaltic dykes in East Iceland. Dyke strike (A)
gional dykes are mostly subvertical and subparal- peaks in the northeast sector, the arithmetic average strike
being N20‡E and coinciding roughly with the NNE trend of
lel and rarely cross-cut (Fig. 8), whereas inclined the rift zone within which these dykes developed. Dyke dip
sheets commonly cross-cut (Fig. 6) and dip from (B) indicates that most of these regional dykes are subverti-
horizontal to vertical (Fig. 2). The narrow strike cal, dipping at 75^90‡ (Gudmundsson, 1995a).

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 285

Fig. 8. Segmented, subvertical regional dykes, up to 13 m thick, in the Tertiary lava pile in Vidbordsdalur, Southeast Iceland,
eroded to nearly 2 km beneath the initial top of the pile. View northeast, the upper part of the thick dyke to the left (northwest)
on the photograph is segmented. From the top to bottom of the 400-m-thick pile seen here, the basaltic lava £ows increase their
dips by 2^3‡.

east Iceland is around 79‡, with 58% of the dykes may be as thick as 60 m, but also as thin as a few
dipping in excess of 80‡. Most dykes cut the lava centimetres. The mode thickness in individual re-
£ows at right angles, so that dyke dip is partly a gional swarms is commonly 1^2 m, while the aver-
function of the dip of the lava pile which increases age thickness is 4^6 m in the Tertiary swarms and
with depth of exposure (Fig. 8). In East Iceland, 1^2 m in the Pleistocene swarms. The thickness
for instance, where the regional dykes are gener- distributions in most swarms are log-normal
ally observed at crustal depths of 1^1.3 km in a laws, power laws (Gudmundsson, 1995a,b,
lava pile dipping at 6^9‡, only 62% of the dykes 1998), or other laws with negative exponents
dip in excess of 80‡. Conversely, in Northwest (Klausen, 1999), so that the arithmetic average
Iceland, where most regional dykes are observed dyke thickness in a swarm does not present its
at crustal depths of 0.5^1 km in a lava pile dip- most common (mode) thickness. Whereas sheets
ping at 3^7‡, 81% of the dykes dip in excess of may constitute as much as 100% of the rock
80‡. The di¡erences in dip between the local sheet (Fig. 1), the regional dykes (in pro¢les 5^10 km
swarms and the regional dyke swarms are readily long) most commonly constitute only 5^6%
explained in terms of the di¡erent sources of these (Fig. 8) of the rock, the maximum being 15^28%
types of swarms. The sheets follow local stress (Gudmundsson, 1995a). Because the intensity
trajectories that dip from near zero to vertical of regional dyke swarms increases with depth of
(Figs. 3 and 4), whereas the regional dykes follow exposure, dykes may, however, constitute a
subvertical trajectories, perpendicular to the mini- higher percentage of the rock at deeper crustal
mum principal stress which, in the relatively £at levels.
rift zones, is normally subhorizontal.
The regional dykes are on average much thick-
er, and have also a greater range in thickness, 3. Geometry of individual dykes and sheets
than the local sheets. While the thickness of a
local sheet rarely exceeds 14 m, a regional dyke As regards the mechanics of sheet emplacement

VOLGEO 2462 12-7-02


286 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

Fig. 9. O¡set dyke segments (outlined) in the Tertiary lava pile in Tenerife (Canary Islands). The dyke segments are basaltic and
slightly overlapping. The 0.6-m-high rucksack provides a scale.

and arrest, important geometrical parameters in- tion front. Commonly, the ¢nal geometry of a
clude segmentation, general shape, and tips. Be- dyke results from many magma injections, where-
cause these geometrical characteristics are mostly by the dyke may expand and grow over a consid-
identical for dykes and sheets, in this section the erable period of time (Gudmundsson, 1995b). As
term dyke is used as a generic term for both prop- the dyke grows, its segments propagate both lat-
er dykes and inclined sheets. erally and vertically and may link up into gradu-
ally larger segments. Thin dykes normally consist
3.1. Segmentation

All major dykes are segmented (Figs. 8 and 9).


In vertical sections, o¡sets are particularly com-
mon at sharp contacts between di¡erent layers
and other discontinuities. Some segments are con-
nected by thin, igneous veins (Fig. 10); other seg-
ments have no visible connections (Gudmunds-
son, 1995b). Some o¡set segments overlap (Fig.
9) and may form hook-shaped junctions. All ma-
jor dykes are also o¡set in lateral sections. These
o¡sets are seen when following well-exposed,
eroded dykes along their length, but are easiest
to demonstrate for feeder dykes, exposed at the
surface of active volcanic systems as volcanic ¢s-
sures (Gudmundsson, 1995b, 2000).
Dykes, like other fractures in heterogeneous
Fig. 10. Dyke segments connected by thin, inclined igneous
rocks, are unlikely to propagate as single units veins (black) across a scoria layer located between two basal-
but, as observations indicate, rather as sets of tic lava £ows. The dyke segments are vertically and laterally
loosely connected segments, each with a propaga- o¡set in a manner very commonly observed for dykes.

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 287

of short segments, thick dykes of long segments, to form in a direction that is perpendicular to the
illustrating the self-similar or self-a⁄ne growth of minimum principal compressive stress.
many dykes (Gudmundsson, 1995b).
Although dyke segments are commonly seen as 3.3. Tips
disconnected in lateral and vertical sections (Figs.
8^10), it is likely that there are some connections Many tens of dyke tips (ends) have been ob-
between many, if not most, segments that consti- served in Iceland and Tenerife, Canary Islands
tute a dyke. These connections, however, are com- (Gudmundsson, 1995a,b; Gudmundsson et al.,
monly very small in comparison with the sizes of 1999). Many dykes and sheets observed to termi-
the segments (Fig. 10). The probability of a sec- nate in vertical sections, for example in Iceland
tion dissecting such small connections between (Gudmundsson, 1995a,b) and in Tenerife (Gud-
segments is much less than that of the section mundsson et al., 1999), simply taper away in an
dissecting the segments themselves, so that most essentially homogeneous and isotropic rock (Fig.
dyke segments are disconnected in outcrops. 11). Then the most likely explanation for the ar-

3.2. Shape

Dykes and their individual segments are com-


monly roughly elliptical in plan view (Pollard and
Segall, 1987), although many segments deviate
considerably from ellipses (Peacock and Marrett,
2000). In regional dyke swarms of the continental
areas, dyke lengths reach 500^1000 km and dyke
thicknesses 100^200 m (Ernst et al., 1995). In Ice-
land, 16 measured regional dykes are 4^22 km
long, the mean values being 8 km for length,
9.5 m for thickness, and 926 for the aspect
(length/thickness) ratio (Gudmundsson, 1995a,b).
Thus, apart from consisting of segments, in lateral
sections all these dykes are, to a ¢rst approxima-
tion, £at ellipses with aspect ratios of 900^1000.
In vertical sections, many dyke segments are
also elliptical although some, particularly thick
ones, are rather straight (Gudmundsson, 2000,
Fig. 10). Other segments are either sinuous or
appear irregular in shape. This apparent irregu-
larity is partly attributable to the frequent o¡sets
of dyke parts at weak contacts and other discon-
tinuities. Partly, however, the irregularity is due to
the dyke overpressure distribution which, in turn,
depends on the anisotropy of the local stress ¢eld
and the variations in the mechanical properties of
the host rock. Regardless of shape, dykes follow
pathways that, on average, are perpendicular to Fig. 11. Arrested dyke tip of a subvertical dyke in the rocks
the minimum compressive principal stress, in pile of the Laxardalur Central Volcano, of Plio^Pleistocene
age, in South Iceland. The tip is rounded, with a large radius
agreement with the observation that most dykes of curvature (Eq. 6), indicating that the host rock was rela-
are extension fractures (mode I cracks). In the tively soft (with a low Young’s modulus) at the time of dyke
analysis below, dykes and sheets are thus assumed emplacement. The 0.3-m-long hammer provides a scale.

VOLGEO 2462 12-7-02


288 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

rest of the intrusion is the stress change with dis-


tance from the source magma chamber.
An example of a dyke tip being arrested at a
contact between contrasting rock layers is from
Tenerife (Fig. 12). There, the volcanic pile is com-
posed of alternating layers of basaltic lava and
pyroclastic rocks with contrasting mechanical
properties and sharp contacts. Commonly, such
contacts contribute to the arrest of vertically
propagating dykes.
The ¢eld observations above indicate that a
vertically propagating dyke may become arrested
at a weak contact. Occasionally, however, the
dyke propagates laterally along the contact as a
sill. Examples from Iceland exist where a vertical
dyke meets with a contact between lava £ows,
follows the contact horizontally for many metres,
and then propagates again vertically from that
contact and upwards. Dykes that are not arrested
normally become o¡set at weak contacts, some
with igneous veins connecting the o¡set segments
(Fig. 10) but others with no such connections.

4. Initiation and tip stresses of dykes and sheets

If magma is to be transported to the surface of Fig. 12. Basaltic dyke (outlined) with a blunt tip, seen in a
a volcano, a sheet or a dyke must be initiated vertical section in Tenerife (Canary Islands). The subvertical
from the source chamber and be able to propa- dyke is arrested at a sharp contact between a basaltic lava
gate to the surface. In this section, I shall examine £ow (upper part of the section) and a layer of pyroclastic
rock (lower part of the section). There is no evidence of ero-
the stress conditions for sheet and dyke initiation sion contributing to the blunt tip shape. At the bottom of
and tip stresses. A magma-¢lled chamber ruptures the section (a road cut), the dyke is 0.38 m thick, but it is
and initiates a sheet or dyke when: 0.28 m thick where its tip is arrested. The notebook provides
a scale.
pl þ pe ¼ c 3 þ T 0 ð1Þ
tion of Eq. 1 are reached at any point on the
Here pl is the lithostatic stress (or pressure) at the chamber walls, irrespective of the shape of the
depth of the chamber; pe = Pt 3pl , the excess mag- chamber or its depth below the surface. It follows
matic pressure, is the di¡erence between the total that the local c3 and T0 are those that are referred
magma pressure, Pt , in the chamber at the time of to in Eq. 1. Also, it is irrelevant whether the con-
its rupture and the lithostatic stress; and c3 and dition of Eq. 1 is reached when pe increases as a
T0 are the minimum principal stress and the in result of adding £uid to the chamber, or when c3
situ tensile strength, respectively, in the roof of decreases because of increased tensile stress con-
the chamber. Compressive stress is considered centration (Gudmundsson, 1988, 1998). Thus,
positive ; thus, when there is an absolute tension, stress concentration e¡ects due to the shape of
c3 is negative, whereas the maximum compressive the magma chamber, its irregular boundaries,
principal stress, c1 , is always positive. and other such factors are included in the local
Sheet or dyke injection occurs when the condi- magnitude of c3 . A magma chamber is normally a

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 289

long-lived £uid-¢lled structure which is presum- tips (ends), so that a knowledge of the tip stresses
ably in lithostatic equilibrium with its surround- is necessary to understand their propagation and
ings for most of its lifetime. Thus, in Eq. 1, pl = c3 arrest. To calculate the tip stresses, consider ¢rst a
( = c1 ) is the usual situation everywhere along the static sheet intrusion modelled as an ellipse of a
contact between the host rock and the chamber. It major axis 2a and a minor axis 2b subject to a
is only during short-term events when pe s 0, ei- constant magma overpressure P0 . Then the mini-
ther because of added volume of magma (absolute mum compressive (maximum tensile) principal
increase in £uid pressure) or reduction in c3 , that stress c3 at the crack tip is:
the condition of Eq. 1 may be satis¢ed, resulting  
in rupture and sheet or dyke initiation. 2a
c 3 ¼ 3P0 31 ð5Þ
It is sometimes more convenient to refer to the b
total £uid pressure Pt in the chamber, in which
case Eq. 1 may be written as: When the tip is exposed and measurable, its radi-
us of curvature, rc = b2 /a, may be used by rewrit-
Pt ¼ c 3 þ T 0 ð2Þ ing Eq. 5 in the form:
h i
When the conditions of Eqs. 1 and 2 are met, a c 3 ¼ 3P0 2ða=rc Þ1=2 31 ð6Þ
dyke (or a sheet) initiates. The dyke may be ar-
rested at, or soon after, its initiation ; alterna- Using the typical lateral aspect (a/b) ratios of re-
tively, it may be able to propagate away from gional dykes in Iceland, 900^1000, and, from Eqs.
the source magma chamber. If the chamber is 3 and 4, their estimated static overpressures of
located in a host rock that behaves as elastic, several tens of MPa (Gudmundsson, 1990,
and if the magma £ow is vertical and the vertical 2000), Eq. 5 yields 3c3 at a lateral dyke tip as
co-ordinate z is positive upwards, the pressure P0 V104 MPa. The radius of curvature of dyke tips
available to drive the propagation of the dyke is in vertical sections are, except for blunt-ended
given by (Gudmundsson, 1990): dykes, typically a few centimetres (Figs. 9 and
11). When 2a, which for a vertical section is the
P0 ¼ 3ð b r 3 b m Þgz þ pe ð3Þ dip dimension of the dyke, is many kilometres,
Eq. 6 yields 3c3 at a vertical dyke tip as
where br is the density of the host rock, bm is the V103 ^104 MPa. If the dip dimension was taken
density of the magma, g is the acceleration due to as that appropriate for individual dyke segments,
gravity, pe is the excess pressure, de¢ned in Eq. 1, rather than whole dykes, 2a in Eq. 4 would be
and the minus sign is because g is positive down- 101 ^102 m and Eq. 6 would give 3c3 at a dyke
wards but the co-ordinate z is positive upwards. tip as V102 ^103 MPa.
The driving pressure P0 is also commonly referred Because in situ tensile strengths of typical host
to as net pressure or overpressure; here the terms rocks of dykes and sheets are only 0.5^6 MPa
driving pressure and overpressure will be used. (Haimson and Rummel, 1982; Schultz, 1995;
Eq. 3 applies to the case where the state of stress Amadei and Stephansson, 1997), these theoretical
in the host rock is isotropic, so that all the prin- crack-tip values of 3c3 are from 100 to 10 000
cipal stresses are equal. If the state of stress is not times greater than the tensile strength. But these
isotropic, in which case there is di¡erence between results apply only when the dykes (and sheets), or
the principal stresses, then, for the two-dimen- their segments, are modelled as ellipses. A di¡er-
sional case examined here, the term cd = c1 3c3 ent, and perhaps more commonly used, model for
must be added to Eq. 3, which thereby becomes : dykes and sheets is that of a mathematical crack.
Consider a two-dimensional crack located
P0 ¼ 3ð b r 3 b m Þgz þ pe þ c d ð4Þ along the x-axis. The other coordinate is y so
that the crack length 2a can be de¢ned by y = 0,
Dykes and sheets propagate by expanding their 3a 9 x 9 a. Inside the crack tips, the stress c3 (x,

VOLGEO 2462 12-7-02


290 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

Fig. 13. Cylindrical (stock-like) magma chamber and the state of stress around it. (A) Horizontal cross-section (in the xy-plane)
through a cylindrical magma chamber, of radius R1 , subject to total magmatic pressure Pt (Eq. 2) and external total horizontal
and isotropic compressive stress cH at R2 . Radius vector r measures the radial distance from the chamber centre. (B) Decrease in
radial cr and circumferential ca stresses (made dimensionless through dividing by the magmatic excess pressure, pe (Eq. 1)) with
dimensionless distance (r/R1 ). At the margin of the chamber 3ca = cr = pe (tensile stress is negative, compressive stress positive).

0) = 3p(x), for 0 9 x 9 a. Outside the crack tips, buoyant intrusion should continue its propaga-
for x s a, the stress c3 is given by (Maugis, tion to the surface.
2000): Nevertheless, the majority of sheets and dykes
Z Z t do not reach the surface to feed volcanic erup-
2x a tdt pðxÞdx tions but become arrested at various crustal
c3 ¼ 3 2 2 3=2 2 2 1=2
ð7Þ
Z 0 ðx 3t Þ 0 ðt 3x Þ depths (Gudmundsson et al., 1999). This apparent
paradox can be understood partly in terms of the
where 0 6 t 6 a. We shall consider only the case of stress ¢eld associated with typical magma cham-
a constant overpressure, p(x) = P0 . Substituting bers, and partly in terms of stress barriers, con-
this in the second integral in Eq. 7, we obtain tacts and other discontinuities that transverse the
P0 Z/2. Using this and solving the ¢rst integral, potential pathways of dykes and sheets.
Eq. 7 gives:
 
1 1 5. Sheet arrest in a non-layered crust
c 3 ¼ 3P0 x 3 ð8Þ
ðx2 3a2 Þ1=2 x
Many central volcanoes, particularly those that
Eq. 8 shows that when xCa, so that the tip of the generate radial dykes, are thought to have shallow
dyke (or sheet) is approached (from outside the magma chambers, the shape of which may be ap-
tip), the theoretical tensile stress c3 becomes in- proximated by that of a cylinder. Exposed, extinct
¢nite, or c3 C3r. Because microcracking and magma chambers of this shape are referred to as
plastic deformation at the tip of the sheet intru- plugs or necks, when they are small, and as stocks
sion relax part of the stress, in¢nite stress is never when they are larger. Plugs range in diameter
reached in the host rock. However, these results from metres up to 1^2 km, whereas stocks may
indicated that, for constant magmatic overpres- be as much as 10^11 km in diameter, some of
sure, regardless of whether the sheet intrusion is which may connect upwards with volcanic plugs.
modelled as a mathematical crack or an elliptical Consider a cylinder of an inner radius R1 and
hole, the tip tensile stresses are so high that a outer radius R2 (Fig. 13A), where R2 ER1 (e¡ec-

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 291

Fig. 14. Spherical magma chamber and the state of stress around it. (A) The chamber is located at the origin of the spherical co-
ordinate system. At a point P(r, a, P) ^ where r is the radius vector, a is the angle between the radius vector and a ¢xed axis,
and P is the angle measured around this axis ^ the state of stress is de¢ned by the principal stresses cr , ca and cP , where cr is
the radial stress (compressive) and the tangential stresses ca and cP are tensile and equal. (B) Decrease (in the vertical xz-plane)
in stresses cr and ca (made dimensionless through dividing by the magmatic excess pressure (Eq. 1)) with the dimensionless dis-
tance (r/R1 ), where R1 is the radius of the spherical magma chamber. Regionally, at the margin of the chamber cr = pe (Eq. 15)
but 3ca = 0.5pe (Eq. 16). The stresses due to pe fall o¡ as the cube of the distance and thus a¡ect sheet attitude only in the im-
mediate vicinity of the chamber.

tively assuming that R2 Cr). We use polar coor- is equal to the lithostatic stress in the crust outside
dinates, with radius vector r and polar angle a, the chamber, then cH = ch = pl . For this stress
and let the cylinder be subject to total internal state, we can substitute the excess pressure pe
£uid pressure Pt (de¢ned in Eq. 2) at R1 and ex- for Pt and use cH = 0 in Eqs. 9 and 10, which
ternal total horizontal compressive stress cH at thereby become :
R2 . We assume that the state of stress is isotropic  2
so that the maximum, cH , and minimum, ch , hor- R1
c r ¼ pe ð11Þ
izontal compressive stresses are equal. The radial r
stress cr is then given by (Saada, 1983):
 2
 2   2  R1
R1 R1 c a ¼ 3pe ð12Þ
c r ¼ Pt þ c H 13 ð9Þ r
r r
These equations show that the intensity of the
and the tangential or circumferential stress, ca , stress ¢eld generated by magmatic excess pressure
by: pe in a vertical, cylindrical magma chamber falls
 2   2  o¡ with the square of the distance, r (Fig. 13B). If
R1 R1 the stress in the host rock of the chamber is not
c a ¼ 3Pt þ cH 1 þ ð10Þ
r r lithostatic, the magnitude of the overpressure in a
propagating sheet intrusion follows Eq. 4.
These equations can be simpli¢ed when, instead Many shallow chambers are modelled ap-
of using the total £uid pressure in the cylinder, we proximately as spherical (Mogi, 1958; McTigue,
use the excess pressure pe , here the di¡erence be- 1987; Gudmundsson, 1988, 1998; Delaney and
tween the total magmatic pressure in the cylindri- McTigue, 1994). Estimated volumes of active
cal chamber and the regional horizontal compres- magma chambers are commonly between 5 and
sive stress, cH . From Eq. 1 it follows that rupture 500 km3 (Chester, 1993); the corresponding radi-
of a cylindrical magma chamber occurs when the us of a spherical chamber would be V1^5 km.
local pe = T0 , where T0 is the tensile strength of Consider a hollow sphere of inner radius R1
the host rock at the margin of the chamber. If cH and outer radius R2 , where R2 ER1 (again, e¡ec-

VOLGEO 2462 12-7-02


292 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

tively assuming that R2 Cr so that free-surface


e¡ects are not considered), subject to internal to-
tal £uid pressure Pt at its inner radius and litho-
static stress pl at its outer radius (Fig. 14A). We
use spherical polar coordinates (r, a, P), where r is
the radius vector, a is the angle between the radi-
us vector and a ¢xed axis, and P is the angle mea-
sured around this axis (Fig. 14A). Then the radial
stress cr becomes (Saada, 1983):
 3   3 
R1 R1
c r ¼ Pt þ pl 13 ð13Þ
r r

Because of spherical symmetry, the stresses ca


and cP are equal and given by:
    
Pt R1 3 pl R1 3
ca ¼ cP ¼ 3 þ þ2 ð14Þ
2 r 2 r
Fig. 15. Boundary-element model (BEASY, 1991) showing
the contours, in MPa, of the maximum principal tensile
Using the excess magma pressure pe , as de¢ned
stress c3 around a magma chamber in a layered crust. The
above, rather than the total pressure Pt , Eqs. 13 chamber is subject to an excess pressure of 10 MPa (twice
and 14 simplify to: that in Figs. 3 and 4). The sti¡nesses of the layers increase
 3 from 5 GPa for the surface layer A to 50 GPa for the layer
R1 hosting the chamber. In this model, and the models in Figs.
c r ¼ pe ð15Þ
r 16 and 17, the height is a unit, the vertical diameter of the
chamber is 0.3 units, the thickness of each of the layers A, B
  and C is 0.1 units, and the depth to the top of the chamber
pe R1 3 from the free surface is 0.35 units. These models (Figs. 15^
ca ¼ cP ¼ 3 ð16Þ
2 r 17), like the earlier ones (Figs. 3 and 4), are two-dimension-
al, as is generally justi¢ed (Gudmundsson, 1998). The maxi-
mum tensile stress is truncated at 10 MPa (Figs. 15^17)
If the magma chamber is very small in relation to
which is equal to the tensile strength used.
its depth (Fig. 14A), so that R1 C0, but pe R31 is
¢nite, the intensity or strength S (point pressure)
of the chamber is given by S = pe R31 , with the units 0.5^6 MPa (Haimson and Rummel, 1982;
of Nm or work (energy). This type of point pres- Schultz, 1995; Amadei and Stephansson, 1997).
sure forms the basis of the well-known ‘Mogi It follows that the tensile stress due to excess pres-
model’ (Mogi, 1958), used to account for surface sure in the chamber before rupture and initiation
deformation due to in£ation and de£ation of of a sheet or dyke is very small and decreases
magma chambers. Eqs. 15 and 16 show that the rapidly with distance along the potential pathway
intensity of the stress ¢eld associated with a of the propagating intrusion.
spherical magma chamber falls o¡ inversely as These results apply primarily if the regional
the cube of the distance or radius vector, r, stress ¢eld around the chamber is isotropic and
from the chamber (Fig. 14B). the host rock itself homogeneous and isotropic.
One general conclusion from Eqs. 9 to 16 is Normally, however, host rocks of magma cham-
that the stress intensity due to the excess pressure bers are heterogeneous and highly anisotropic.
pe falls o¡ rapidly with distance from a magma Therefore, even if the regional loading is uniform
chamber. The excess pressure pe is normally equal there may be signi¢cant variations in the stress
to the local tensile strength at the time of chamber ¢eld that can contribute to the arrest of sheet
rupture and sheet intrusion (Eq. 1), and is thus intrusions. In the following sections, I analyse

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 293

these e¡ects using two main sets of numerical decreases with increasing fracture frequency of a
models. In the ¢rst set, the e¡ects of crustal layer- rock mass (Priest, 1993). Because all rock masses
ing on the stress ¢eld around a magma chamber contain fractures and other discontinuities, the in
are analysed (Figs. 15^17). In the second set, the situ Young’s moduli of rocks are generally lower
general e¡ects of abrupt changes in Young’s mod- than those of small-scale laboratory samples of
uli, as well as those of discontinuities, on the the same rock types (Goodman, 1989).
crack-tip stress ¢elds of dykes and sheets them- Here boundary-element models (BEASY, 1991)
selves are considered (Figs. 18 and 19). are used to study the e¡ects of a large variation in
sti¡ness between layers on the stress ¢eld inside a
central volcano. The sti¡nesses of crustal layers
6. Sheet arrest in a layered crust tend to increase with depth, a variation that is
used in the ¢rst model (Fig. 15). An active central
Crustal layering is common in volcanic zones in volcano and its surroundings, however, often con-
general and in central volcanoes in particular. tain soft pyroclastic layers between sti¡ lava
Many volcanic zones, such as in Iceland, consist £ows, in which case there is not a simple relation-
primarily of (mostly) basaltic lava £ows with ship between crustal depth and sti¡ness. Thus, in
layers of scoria and sediments in between (Figs. the next two models (Figs. 16 and 17) a Young’s
8 and 12). Inside central volcanoes, however, acid modulus of 1 GPa is used for the soft pyroclastic
and intermediate lava £ows and pyroclastic rocks layers and 100 GPa for the sti¡ lava £ows.
are normally more common. In the uppermost Although such extreme sti¡ness variations may
parts of many active volcanic zones and central not be very common, they highlight the poten-
volcanoes, layers of soft sediments and pyroclastic tially large e¡ects that Young’s modulus contrasts
rocks o¡er strong mechanical contrasts to the can have on volcano stresses and associated sheet
sti¡er lava £ows. and dyke arrest. Because injection and arrest of
Laboratory measurements indicate that dykes and sheets occurs primarily in the layers
Young’s moduli (sti¡nesses) of common solid above the source chamber of the volcano, the fo-
rocks are mostly within the range of 1^100 GPa cus is on sti¡ness variation between those layers,
(Johnson, 1970; Jumikis, 1979; Afrouz, 1994). and the rock surrounding the chamber itself is
Some rocks, however, exceed these limits. For ex- modelled with a uniform sti¡ness (Figs. 15^17).
ample, some igneous rocks, in particular basalt In all these models (Figs. 15^17) the only load-
and gabbro, reach sti¡nesses of 110^130 GPa ing is a magmatic excess pressure of 10 MPa in
(Bell, 2000), whereas some clays have sti¡nesses the chamber (cf. Eq. 1). The value 10 MPa is
of 0.1 GPa (Waltham, 1994) and others reach similar to values of small-scale laboratory tensile
values as small as 0.01 GPa (Johnson, 1970). Mer- strengths of common solid rocks (Jumikis, 1979;
cia Mudstone has a sti¡ness down to 0.002 GPa, Afrouz, 1994; Waltham, 1994). In situ tensile
that is, 2 MPa (Bell, 2000). Similarly, some vol- strengths, however, are normally lower, com-
canic tu¡s have sti¡nesses of 0.1^0.4 GPa while monly ranging from 0.5 to 6 MPa (Haimson
very soft ones have a sti¡ness of only 0.05 GPa and Rummel, 1982; Schultz, 1995; Amadei and
(Bell, 2000). Poisson’s ratios of solid rocks show a Stephansson, 1997). Here the laboratory tensile
much more limited range than the Young’s mod- strengths and sti¡nesses are used in all the mod-
uli. For many rocks, Poisson’s ratio is typically els.
around 0.25 (Johnson, 1970; Jumikis, 1979). For In the ¢rst model (Fig. 15) the chamber itself is
example, 0.25 is a common Poisson’s ratio of ba- located in a sti¡ layer D (50 GPa), whereas the
salts and also of weak volcanic tu¡s (Bell, 2000). layers above gradually decrease in sti¡ness from C
These values are based on small-sample labora- (30 GPa) through B (20 GPa) to the surface layer
tory measurements and may be signi¢cantly dif- A (5 GPa). No signi¢cant chamber-generated ten-
ferent from those of in situ rock masses. In par- sile stress is transferred to the surface layer A.
ticular, the in situ Young’s modulus normally This layer would thus be likely to contribute to

VOLGEO 2462 12-7-02


294 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

the arrest of many dykes and sheets and make it stress associated with the tip of a dyke or sheet
impossible for them to feed volcanic eruptions. intrusion so much as to make it impossible for the
The next two models (Figs. 16 and 17) focus on intrusion to continue its extension. Sheets and
the stress-¢eld e¡ects of an extreme variation in dykes propagating into soft layers commonly de-
sti¡ness between layers. The magma chamber it- velop blunt or rounded tips (Fig. 11), thereby
self is located in a low-sti¡ness layer D, for exam- making the crack-tip tensile stress concentration
ple sedimentary or pyroclastic rocks, with a smaller than in the case of a sharp and narrow
Young’s modulus of 10 GPa. The crust above tip.
the magma chamber is composed of layers with Many layered rocks, such as in active central
large and abrupt contrasts in sti¡ness. The very volcanoes and volcanic zones, contain sharp con-
soft layers (1 GPa) are considered to represent tacts between layers, and these contacts are often
weak pyroclastic layers, whereas the very sti¡ mechanically weak. When a horizontal disconti-
layers (100 GPa) represent basaltic lava £ows. nuity, such as a contact, a joint or a fault, is
In the model in Fig. 16, layers C and A are very approached by the tip of a vertically propagating
sti¡, whereas layer B is very soft. The high tensile dyke, the discontinuity may open up and contrib-
stress concentration in layer C indicates that it ute to arresting the dyke. This mechanism of frac-
would normally be easy for dykes and sheets to ture arrest is referred to as the Cook^Gordon
propagate through that layer. Similarly, tensile debounding mechanism (Atkins and Mai, 1985).
fractures and, depending on the thickness of the A transverse discontinuity such as a weak contact
layer, normal faults might develop in the surface is a very e¡ective stopper of sheets and dykes
layer A during periods of signi¢cant excess pres- (Fig. 19). Horizontal contacts, joints, or thin
sure and in£ation of the magma chamber. The layers of di¡erent mechanical properties (for ex-
very soft layer B, however, receives no signi¢cant ample, scoria between lava £ows) are also a com-
tensile stress during these same periods, and mon reason for dykes being o¡set in vertical sec-
would thus normally contribute to the arrest of tions (Fig. 10).
dykes and sheets injected form the magma cham-
ber. Thus, layer B would act as a stress barrier
where many sheets and dykes could become ar- 8. Discussion
rested.
In the model in Fig. 17, by contrast, layers C The model proposed in this paper recognises
and A are very soft, whereas layer B is very sti¡. two principal mechanisms by which sheets and
In this model, considerable tensile stress concen- dykes become arrested. One is generation of stress
trates in layer B, so that this layer might develop barriers whereby sheet intrusions meet layers with
normal faulting. By contrast the surface layer A local stress ¢elds that are unfavourable to their
and, in particular, layer C just above the magma propagation and become arrested. The other is
chamber, would act as stress barriers and contrib- rock anisotropy, whereby sheet intrusions become
ute to the arrest of sheets and dykes from the arrested at contacts, joints and other discontinu-
magma chamber. ities. These mechanisms are related, particularly
through changes in Young’s moduli which are
common at contacts between di¡erent rock types.
7. Sheet arrest due to soft layers and This model can account for the following impor-
discontinuities tant ¢eld observations of arrested dykes and
sheets.
A soft layer suppresses not only the tensile First, in vertical sections the tips of many dykes
stresses associated with pressured magma cham- and sheets become arrested at weak contacts,
bers (Figs. 15^17), but also the stresses associated joints or other discontinuities that transverse the
with the tips of propagating sheets and dykes potential pathway of the intrusion. Commonly,
(Fig. 18). A soft layer may suppress the tensile the intrusions that terminate at such discontinu-

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 295

sile strengths, and weak contacts are commonly


e¡ective in arresting sheet intrusions. For exam-
ple, studies of man-made hydraulic fractures in-
dicate that many contacts, particularly when com-
bined with stress contrasts, are very e¡ective in
arresting vertical propagation of the fractures
(Valko and Economides, 1995; Charlez, 1997;
Yew, 1997). Many dykes and sheets that become
arrested at bedding planes in sedimentary rocks,
or at contacts between lava £ows, change into
concordant intrusions such as laccoliths or sills
(Mudge, 1968).
Of other models on the arrest of sheet intru-
sions, the principal one is that of dykes becoming
arrested because of cooling of magma at the tip,
that is, when £ow of magma into the tip is
blocked by solidi¢cation so that its propagation
Fig. 16. Boundary-element model (BEASY, 1991) showing
velocity decreases below a critical level (Hoek,
the contours, in MPa, of the maximum principal tensile 1994; Fialko and Rubin, 1999; Bolchover and
stress c3 around a magma chamber. This chamber is located Lister, 1999). As a general model of arrest of
in a layered crust with a very strong contrast in sti¡ness be- sheet intrusions, however, this model fails to agree
tween a soft (1 GPa) layer B and the sti¡ (100 GPa) layers with several results, including the following. First,
A and C (cf. Fig. 15).
many dyke and sheet tips are blunt and associated

ities have blunt tips (Fig. 12). There is no doubt


that abrupt changes in sti¡ness (Figs. 15^18), and
discontinuities at contacts between layers (Fig.
19), commonly alter the local stress ¢elds so as
to encourage the arrest of sheets and dykes.
Second, in vertical and horizontal sections,
many dykes and sheets taper away within essen-
tially homogeneous rock layers (Fig. 11, Gud-
mundsson, 1995a,b; Gudmundsson et al., 1999).
The general shape of these intrusions on ap-
proaching their tips indicates that they entered
layers where the compressive stress was so high
as to arrest the dyke tip. The observed tips are
generally identical whether they are just the ends
of dyke segments or the true ends of a whole
dyke. This type of sheet arrest can often be attrib-
uted to the fall o¡ in intensity of the (relative or
absolute) tensile stress with distance from the
source chamber (Figs. 13 and 14).
The existence of stress barriers is supported by Fig. 17. Boundary-element model showing the contours, in
abundant empirical data for sudden stress MPa, of the maximum principal tensile stress c3 around a
magma chamber. This chamber is located in a layered crust
changes (Haimson and Rummel, 1982; Amadei with a very strong contrast in sti¡ness between the soft
and Stephansson, 1997). Stress barriers that coin- (1 GPa) layers A and C and the sti¡ (100 GPa) layer B
cide with abrupt changes in Young’s moduli, ten- (cf. Fig. 15).

VOLGEO 2462 12-7-02


296 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

Fig. 18. Dyke arrest due to contrast in Young’s modulus be- Fig. 19. E¡ects of discontinuities on dyke arrest. This model
tween layers A (10 GPa) and B (40 GPa), both with the has all the same boundary conditions and elastic properties
same Poisson’s ratio, 0.25 (Gudmundsson, 1988). This as in Fig. 18 except that the contact between layers A and B
boundary-element model (BEASY, 1991) has a unit height; is here a discontinuity modelled as an internal spring with
the dyke height (dip dimension) is 0.6 units, the thickness of sti¡ness 6 MPa/m. The maximum tensile stress at the upper
the surface layer A is 0.05 units, and the upper tip of the tip of the dyke is again 50 MPa (the maximum shown is 10
dyke is at 0.1 units below the free surface. The dyke mag- MPa) but, in contrast to the model in Fig. 18, here there is
matic overpressure varies smoothly from zero at its tips to no tensile stress transferred to layer A. Thus, a combination
6 MPa in its centre. The maximum tensile stress (indicated of a contrast in Young’s modulus and a sharp discontinuity
by the contours of 3c3 ) is 50 MPa at the upper dyke tip is probably one of the most e¡ective mechanisms for dyke
but is truncated at 10 MPa (cf. Fig. 15). The maximum ten- arrest.
sile stress in layer A is only 0.5 MPa and this, together with
the £attening of the contours at their contact with layer B,
suggests that the dyke becomes arrested. sheet intrusion and propagation may, therefore,
be satis¢ed in a shell of a certain thickness sur-
with discontinuities, indicating arrest primarily by rounding the chamber, but not outside that shell
mechanical, rather than thermal, e¡ects. Second, (Fig. 20). Thus, once the sheet or dyke propagates
so long as there is any signi¢cant magma over- to the outer surface of that shell, the location of
pressure in a sheet intrusion, it follows from Eqs. which is a function of time, the intrusion becomes
5^8 that the tensile stresses at the tip would nor- arrested.
mally be many times greater than the tensile
strength of any solidi¢ed part of the magma.
Thus, the propagation should continue. Third,
simple calculations indicate that hardly any of
the inclined sheets in Iceland, many being injected
from depths of only 2 km below the surface,
could be thermally arrested (Gudmundsson,
1995a,b, 1998). This conclusion applies even if
the sheets were injected into geothermal systems:
many thin sheet intrusions do not become ar-
rested but rather reach the surface at the ocean
bottom or under lakes where the upper crust is Fig. 20. During a stress cycle (the average time between suc-
saturated with water. cessive eruptions in a central volcano), most sheets injected
The analytical and numerical results presented from the source chamber become arrested in various crustal
here have implications for the development of a shells or levels (1^5). These injections, however, gradually ho-
sheet swarm in a central volcano. Eqs. 9 to 16 mogenise the stress ¢eld to the surface (6), at which stage
one or more eruptions may occur. Central volcanoes are
show that, for a pressured chamber, the stress commonly supplied with magma through a double chamber
intensity falls o¡ rapidly from the margin of the where a deep-seated reservoir supplies primitive magma to a
chamber into the host rock. The condition for shallow chamber. Modi¢ed from Gudmundsson (2000).

VOLGEO 2462 12-7-02


A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298 297

The numerical models (Figs. 15^17) support Atkins, A.G., Mai, Y.W., 1985. Elastic and Plastic Fracture.
Ellis Horwood, Chichester.
this general conclusion, but indicate that for mag-
Baer, G., Heimann, A. (Eds), 1995. Physics and Chemistry of
ma chambers hosted in rocks of contrasting me- Dykes. Balkema, Rotterdam.
chanical properties the geometries and locations BEASY, 1991. The Boundary Element Analysis System User
of the shells may be a complex function of time. Guide. Computational Mechanics, Boston, MA.
Irrespective of the geometries of these shells, it is Bell, F.G., 2000. Engineering Properties of Soils and Rocks,
likely that the sheets and dykes gradually homog- 4th edn. Blackwell, London.
Billings, M.P., 1972. Structural Geology, 3rd edn. Prentice-
enise the stress ¢eld to the surface of the central Hall, Englewood Cli¡s, NJ.
volcano (Fig. 20). Thus, the models presented Bolchover, P., Lister, J.R., 1999. The e¡ect of solidi¢cation on
here make two main predictions as to the evolu- £uid-driven fracture, with application to bladed dykes. Proc.
tion of a swarm of sheets and dykes in a central R. Soc. London Ser. A, Math. Phys. 455, 2389^2409.
volcano. One prediction is that most sheets and Charlez, P.A., 1997. Rock Mechanics, Vol. 2. Petroleum Ap-
plications. Editions Technip, Paris.
dykes become arrested, primarily because of dis- Chester, D., 1993. Volcanoes and Society. Edward Arnold,
continuities and stress barriers that transverse London.
their potential propagation pathways. The second Delaney, P.T., McTigue, D.F., 1994. Volume of magma accu-
prediction is that as the sheets and dykes gradu- mulation or withdrawal estimated from surface uplift or
subsidence, with application to the 1960 collapse of Kilauea
ally propagate for greater distances from their
Volcano. Bull. Volcanol. 56, 417^424.
source chamber, they homogenise the stress ¢eld Delaney, P.T., Pollard, D.D., 1982. Solidi¢cation of basaltic
to the surface (Fig. 20). Thus, stress-¢eld homog- magma during £ow in a dike. Am. J. Sci. 282, 856^885.
enisation of the crust above the source chamber Ernst, R.E., Buchan, K.L., Palmer, H.C., 1995. Giant dyke
may be regarded as a necessary condition for swarms: characteristics, distribution and geotectonic appli-
eruptions fed by sheets and dykes to occur in a cations. In: Baer, G., Heimann, A. (Eds.), Physics and
Chemistry of Dykes. Balkema, Rotterdam, pp. 3^21.
central volcano. Fialko, Y.A., Rubin, A.M., 1999. Thermal and mechanical
aspects of magma emplacement in giant dike swarms.
J. Geophys. Res. 104, 23033^23049.
Acknowledgements Gautneb, H., Gudmundsson, A., 1992. E¡ect of local and
regional stress ¢elds on sheet emplacement in west Iceland.
J. Volcanol. Geotherm. Res. 51, 339^356.
I thank Sonja L. Brenner for running the mod- Gautneb, H., Gudmundsson, A., Oskarsson, N., 1989. Struc-
els presented in Figures 15^17, and Yosuke Aoki ture, petrochemistry, and evolution of a sheet swarm in an
and an anonymous referee for helpful comments. Icelandic central volcano. Geol. Mag. 126, 659^673.
My work on dykes in Tenerife was made in pleas- Geikie, A., 1897. The Ancient Volcanoes of Great Britain, Vol.
2. Macmillan, London.
ant collaboration with Laura B. Marinoni. This
Goodman, R.E., 1989. Introduction to Rock Mechanics, 2nd
work was supported by grants from the Iceland edn. Wiley, New York.
Science Foundation, the Norway Research Coun- Gudmundsson, A., 1988. E¡ect of tensile stress concentration
cil, and the European Commission. around magma chambers on intrusion and extrusion fre-
quencies. J. Volcanol. Geotherm. Res. 35, 179^194.
Gudmundsson, A., 1990. Emplacement of dikes, sills and crus-
tal magma chambers at divergent plate boundaries. Tecto-
nophysics 176, 257^275.
References Gudmundsson, A., 1995a. Infrastructure and mechanics of
volcanic systems in Iceland. J. Volcanol. Geotherm. Res.
Afrouz, A.A., 1994. Rock Mass Classi¢cation Systems and 64, 1^22.
Modes of Ground Failure. CRC Press, London. Gudmundsson, A., 1995b. The geometry and growth of dykes.
Amadei, B., Stephansson, O., 1997. Rock Stress and its Mea- In: Baer, G., Heimann, A. (Eds.), Physics and Chemistry of
surement. Chapman and Hall, London. Dykes. Balkema, Rotterdam, pp. 23^34.
Anderson, E.M., 1936. The dynamics of formation of cone Gudmundsson, A., 1998. Magma chambers modeled as cav-
sheets, ring dykes and cauldron subsidence. Proc. R. Soc. ities explain the formation of rift zone central volcanoes and
Edinburgh 56, 128^163. their eruption and intrusion statistics. J. Geophys. Res. 103,
Anderson, E.M., 1951. Dynamics of Faulting and Dyke For- 7401^7412.
mation, 2nd edn. Olivier and Boyd, Edinburgh. Gudmundsson, A., 2000. Dynamics of volcanic systems in Ice-

VOLGEO 2462 12-7-02


298 A. Gudmundsson / Journal of Volcanology and Geothermal Research 116 (2002) 279^298

land: example of tectonism and volcanism at juxtaposed hot Parker, A.J., Rickwood, P.C., Tucker, D.H. (Eds.), 1990.
spot and mid-ocean ridge system. Annu. Rev. Earth Planet. Ma¢c Dykes and Emplacement Mechanisms. Balkema, Rot-
Sci. 28, 107^140. terdam.
Gudmundsson, A., Marti, J., Turon, E., 1997. Stress ¢elds Peacock, D.C.P., Marrett, R., 2000. Strain and stress: reply.
generating ring faults in volcanoes. Geophys. Res. Lett. J. Struct. Geol. 22, 1369^1378.
24, 1559^1562. Pollard, D.D., 1973. Derivation and evaluation of a mechan-
Gudmundsson, A., Marinoni, L.B., Marti, J., 1999. Injection ical model for sheet emplacement. Tectonophysics 19, 233^
and arrest of dykes: implications for volcanic hazards. 269.
J. Volcanol. Geotherm. Res. 88, 1^13. Pollard, D.D., Muller, O.H., 1976. The e¡ect of gradients in
Haimson, B.C., Rummel, F., 1982. Hydrofracturing stress regional stress and magma pressure on the form of sheet
measurements in the Iceland research drilling project drill intrusions in cross section. J. Geophys. Res. 81, 975^984.
hole at Reydarfjordur, Iceland. J. Geophys. Res. 87, 6631^ Pollard, D.D., Segall, P., 1987. Theoretical displacements and
6649. stresses near fractures in rocks: with application to faults,
Halls, H.C., Fahrig, W.F. (Eds.), 1987. Ma¢c Dyke Swarms. joints, veins, dikes and solution surfaces. In: Atkinson, B.
Geol. Assoc. Can. Spec. Pap. 34. (Ed.), Rock Fracture Mechanics. Academic Press, London,
Harker, A., 1904. The Tertiary Igneous Rocks of Skye. UK pp. 247^349.
Geol. Surv. Mem. Priest, S.D., 1993. Discontinuity Analysis for Rock Engineer-
Harker, A., 1909. The Natural History of Igneous Rocks. ing. Chapman and Hall, London.
Hafner, New York. Rubin, A.M., 1995. Propagation of magma-¢lled cracks.
Hills, E.S., 1972. Elements of Structural Geology, 2nd edn. Annu. Rev. Earth Planet. Sci. 23, 287^336.
Chapman and Hall, London. Saada, A.S., 1983. Elasticity: Theory and Applications.
Hoek, J.D., 1994. Ma¢c Dykes of the Vestfold Hills, East Krieger, Malabar, FL.
Antarctica. PhD Thesis, Utrecht University, Utrecht. Schirnick, C., van den Bogaard, P., Schmincke, H.U., 1999.
Johnson, A.M., 1970. Physical Processes in Geology. Freeman Cone sheet formation and intrusive growth of an oceanic
and Cooper, San Francisco, CA. island ^ The Miocene Tejeda complex on Gran Canaria
Jumikis, A.R., 1979. Rock Mechanics. Trans Tech Publica- (Canary Islands). Geology 27, 207^210.
tions, Clausthal. Schultz, R.A., 1995. Limits of strength and deformation prop-
Klausen, M.B., 1999. Structure of Rift-Related Igneous Sys- erties of jointed basaltic rock masses. Rock Mech. Rock
tems and Associated Crustal Flexures. PhD Thesis, Univer- Eng. 28, 1^15.
sity of Copenhagen, Copenhagen. Spence, D.A., Turcotte, D.L., 1985. Magma-driven propaga-
Lippard, S.J., Shelton, A.W., Gass, I. (Eds.), 1986. The Ophio- tion of cracks. J. Geophys. Res. 90, 575^580.
lite of Northern Oman. Blackwell, Oxford. Spence, D.A., Sharp, P.W., Turcotte, D.L., 1987. Buoyancy-
Lister, J.R., Kerr, R.C., 1991. Fluid-mechanical models of driven crack propagation: a mechanism for magma migra-
crack propagation and their application to magma transport tion. J. Fluid Mech. 174, 135^153.
in dikes. J. Geophys. Res. 96, 10049^10077. Valko, P., Economides, M.J., 1995. Hydraulic Fracture Me-
Marinoni, L.B., Gudmundsson, A., 2000. Dykes, faults and chanics. Wiley, New York.
palaeostresses in the Teno and Anaga massifs of Tenerife Walker, G.P.L., 1960. Zeolite zones and dike distribution in
(Canary Islands). J. Volcanol. Geotherm. Res. 103, 83^103. relation to the structure of the basalts of eastern Iceland.
Maugis, D., 2000. Contact, Adhesion and Rupture of Elastic J. Geol. 68, 515^528.
Solids. Springer, Berlin. Walker, G.P.L., 1992. ‘Coherent intrusive complexes’ in large
McTigue, D.F., 1987. Elastic stress and deformation near a basaltic volcanoes ^ a new structural model. J. Volcanol.
¢nite spherical magma body: resolution of the point source Geotherm. Res. 50, 41^54.
paradox. J. Geophys. Res. 92, 12931^12940. Walker, G.P.L., 1999. Volcanic rift zones and their intrusion
Mogi, K., 1958. Relations between eruptions of various volca- swarms. J. Volcanol. Geotherm. Res. 94, 21^34.
noes and the deformations of the ground surfaces around Waltham, A.C., 1994. Foundations of Engineering Geology.
them. Bull. Earthq. Res. Inst. 36, 99^134. Spon, London.
Mudge, M.R., 1968. Depth control of some concordant intru- Wilson, L., Head, J.W., 1981. Ascent and eruption of basaltic
sions. Bull. Geol. Soc. Am. 79, 315^332. magma on the Earth and Moon. J. Geophys. Res. 86, 2971^
Muller, O.H., Pollard, D.D., 1977. The stress state near Span- 3001.
ish Peaks, Colorado, determined from dike pattern. Pure Yew, C.H., 1997. Mechanics of Hydraulic Fracturing. Gulf
Appl. Geophys. 115, 69^86. Publishing, Houston, TX.
Pallister, J.S., 1981. Structure of the sheeted dike complex of
the Samail ophiolite near Ibra, Oman. J. Geophys. Res. 86,
2661^2672.

VOLGEO 2462 12-7-02

You might also like