Okki Verdiansyah MT

STTNAS Yogyakarta

Evolving fracture patterns: columnar joints, mud cracks and polygonal terrain
Lucas Goehring
Published 4 November 2013.DOI: 10.1098/rsta.2012.0353
Cracks in cooling or drying media can form captivating patterns of connected networks,
Two types of patterns are common.
1) The first is a rectilinear pattern, such as one that may form when a homogeneous
slurry is dried or cooled uniformly [1–3]. The vertices of these patterns typically
contain one crack path intersecting another straight crack at right angles, a T-
junction.
2) Second are crack networks with a more regular hexagonal pattern .These are
dominated by cracks intersecting at 120° or Y-junctions. Ex : mud cracks.
• Hexagonally ordered patterns may form as the result of either desiccation or thermal
contraction cracks, and there is a deep physical symmetry between these two
mechanisms [8,9].
• Furthermore, they may be classified into crack patterns that have evolved in both time
and space (such as a network of crack tips that delimit the prismatic forms of
columnar joints as they advance [10]), and those that have recurred in the same place
but evolved over time as the pattern repeatedly healed and recracked [3,11].
Hexagonal crack patterns are found naturally in many settings. They can be driven by thermal or
desiccation contraction and appear to require the opportunity to evolve, in either space or time.
Shown here are examples of columnar joints (a) in lava , (b) in desiccated starch; (c) desiccation
cracks in the cemented sands (d) ‘young’ thermal contraction cracks, or polygonal terrain, (e)
desiccation-induced mud-crack patterns in the laboratory after (f) one drying and (g) 25 cycles of
wetting, drying and cracking [3], and (h) as found naturally in soil
• Cracks in a homogeneous film will tend to pattern the surface
into roughly equal-sized rectilinear pieces, bounded with right-
angled T-junctions. However, many natural crack patterns, such
as columnar joints, show hexagonal planforms and equi-angled
Y-junctions. In both cases, Euler’s formula implies that the
average number of neighbours for any polygon is six, as long
as cracks do not cross each other, or vertices form from more
than three cracks. It is possible to continuously deform a
rectilinear crack network into a hexagonal one (allowing
vertices to pass each other, if necessary [24]), if the pattern is
allowed to evolve
In lava, columnar joints form by the gradual extension of thermal contraction
cracks into a cooling body, with the cracks growing perpendicular to the
isothermal surfaces at the time of cracking [Mallet R, 1875 On the origin and
mechanism of production of the prismatic (or columnar) structure of basalt.].
At any particular time, the crack tips represent a network in a thin region near
the surface defined by the glass transition temperature isotherm [Ryan MP &
Sammis CG1981 The glass transition in basalt] and the columnar forms can be
interpreted as a record of this network as it evolved in time, while propagating
through space.
Lava may cool, and thus columns may grow, in any direction; in many cases,
both an upper and lower colonnade are present in a single flow unit, as the
lava cools from both its top and bottom surfaces [Long PE & Wood BJ, 1986
Structures, textures and cooling histories of Columbia River basalt flows] [DeGraff
JM, Long PE, Aydin A, 1989 Use of joint-growth directions and rock textures to infer
thermal regimes during solidification of basaltic lava flows].
In starch, columns grow normal to the surfaces of constant moisture content
and track the position of the so-called funicular–pendular transition, where the
water transport mechanisms of capillary flow and vapour diffusion cross over
in efficiency [Goehring L, 2009 Drying and cracking mechanisms in a starch slurry].
Columnar joints in lava. (a) A cross-section of a colonnade near Fingal’s Cave, Staffa, UK, appears
hexagonal, but (b–d) cracks very near the Y-junctions curve slightly, reflecting interactions between
cracks during any particular growth increment. (e) A single column near Whistler, Canada, showing
lateral striae bands. These can be used to infer the crack front growth velocity [56], which is faster
(the striae are smaller) near the base of this flow. (f) The crack front velocity in basaltic (black)
and other lavas (green) is inversely correlated with the size of the columns, as it is also (g) in starch
for constant (green) and slowly varying (black) front velocities ((f,g) adapted from [25]). Curves
show the expected correlation for constant Péclet numbers. (h) Plumose, or hair-like, marks on
striae, here sketched over with chalk, indicate the direction of crack propagation of each crack
increment. (i) The crack growth direction recorded on striae can be towards (white) or away (grey)
from any vertex. Shown here are the direction of crack propagation for the sequence of striae on
the faces of three columns, near different vertices.
Elastic stresses during columnar joint formation are essentially confined to a thin layer. (a) In
cooling lava, water infiltration within cracks maintains most of a colonnade at 100°C [61]. A
region of steep thermal gradients connects the water-cooled zone with the still-liquid melt, where
viscous dissipation relieves stress [61]. The network of crack tips that define the polygons lies
within a temperature window of a few degrees, near the glass transition temperature [56]. Each
crack face advances intermittently: some point breaks first (marked by a small open circle),
causing a crack to advance locally into the warmer lava, where it rapidly halts. This advance
propagates laterally, until it reaches a vertex. Each such advance leaves behind a single stria
[62,63], as seen in figure 8e,h. (b) In starch columns grow from the drying surfaces inwards,
following (c) a steep transition in water content that generates stress [25]. The cooling/drying
fronts advance at a rate v, which sets a natural length scale D/v to the cracking problem, in place
of the film thickness h. (Online version in colour.)
Schematic model for the evolution of a rectilinear crack pattern into a hexagonal pattern. (a)
Initially, the order of crack opening is captured by the shape of the T-junctions. (b) If the cracks
heal and recur, a new crack can be guided by the previous cycle’s cracks, but deflected by the
asymmetric stress concentration it feels near a vertex, in which case (c) the vertex is shifted and the
crack paths are curved. (d) This process may repeat and evolve the crack pattern. Over many
cycles (e), with cracks equally likely to approach a vertex from any branch, symmetry will lead to
a Y-junction. (f) Close to the vertex, however, the order of cracking in any particular cycle will still
curve cracks slightly. (g) If the evolution of the crack network is stacked in space, it provides a
model for extended patterns such as columnar joints.
• The crack patterns : are commonly seen in dried mud, frozen soils or old lava flows
• A hexagonal network of equi-angled Y-junctions or sometimes an intermediate state.
• A genetic model was sketched which could account for these observations, among others,
as the result of simple interactions.
• A crack curves to maximize the difference between the strain energy release rate, at the
moment of cracking, and the cost of creating the new crack surfaces—an established
principle of fracture mechanics
• This is a local energy consideration: the crack tip only senses its immediate environment
and responds accordingly.
• When a layer repeatedly cracks and heals, the positions of the previous cracks may be
weaker, or more compliant, than the surrounding material.
• Similarly, in a crack network like that of columnar joints, which is growing through space,
the stress state in an uncracked layer is strongly influenced by the prior pattern.
• In this way, cracks are guided along the paths taken in previous cracking cycles.
• At vertices, however, this argument introduces asymmetries and a growing crack can be
deflected. If the order in which cracks arrive at a vertex changes with time, in different
cracking cycles, these deflections can accumulate and will distort the vertex, moving it in
a direction to equalize the angles between all cracks, creating a Y-junction.
Crack vertices move during repeated drying cycles. To measure this, (a) raw images are (b)
thresholded, (c) skeletonized, and then (d) vertex positions are found as the branch-points of the
skeletons. Shown here are the positions of three vertices over 10 drying cycles, graded in colour
from the first (light; red online) to the tenth (dark; blue online) cycle. (e) The average vertex
displacement increases with the number of drying cycles, and scales with the layer thickness.
Representative error bars are given for h=3.6 mm. The data from a previous study [3] where
h=3.1 mm are also shown. Vertex motion is most apparent in the first four cycles, and then slows
down to a gradual drift. The arrow shows the measured vertex displacement when cracks round a
vertex, from figure 4f, for comparison. (f) The vertex motion is directed towards the crack that
originally approached a vertex at two right angles. The angle θ here is the difference between
the direction in which a vertex has been displaced between the first and tenth cracking cycles (for
h=2.6 mm) and the angle of approach of this crack, as demonstrated by the inset
Polygonal terrain in the Dry Valleys, Antarctica. (a) Oblique/aerial view of cracks near New
Harbour delta, of age 1–2 ka [7]. Crack positions are highlighted by white arrows along the
perimeter of the image. The vertices shown are distorted and appear to be deflected in the
direction indicated by the black arrows from what would otherwise be T-junctions. (b) A close-up
image of a Y-junction shows a small vertex ‘core’ where cracks in any particular year can interact
to approach each other at right angles. (c) Black & Berg [34] installed pairs of rods here
bracketing the crack pattern in 1963. By 2007 this crack (highlighted by white arrows) has shifted
to lie outside the rod pair. (d) At Victoria Valley, where the terrain is 10–12 ka [7], the pattern is
more hexagonal.
Okki Verdiansyah MT
STTNAS Yogyakarta
• Inclined columnar joints in flow interiors, spiracles that rise from
the bottoms of flows, and primary foreset bedding in
palagonite breccias are the most reliable features from which
to determine the direction of flow of ancient basaltic lavas.
 Inclined columns in flow centers
Columns grow' at right angles to isothermal surfaces within the cooling lava,
but if the lava flow is thin, or if cooling is rapid and irregular, columns may not
form. Instead the lava breaks into hackly fragments or joints irregularly.
Many joints are vertical, however, and so simulate columnar structure if
observed only in cross section.

Most of the thick flows of Columbia River basalt show excellent two-tier
columnar jointing: a low'er tier of coarse but generally well shaped columns
growls upward from the base, while a longer tier of thin and more irregular
columns sprouts downward from beneath the flow top. The twTo meet along a
prominent horizontal parting about two-thirds of the distance from the top of the
flowr.
Under normal conditions of static crystallization (fig. 1. left) isothermal surfaces
are nearly horizontal, and so the two tiers stand vertical.
Thus in figure 1. left, lava at points A and B has reached the same temperature,
columnar joints are growing in it. and they continue to grow vertically because
the isothermal surfaces between A and B are essentially horizontal.

But if forward movement of the lava is resumed after columnar joints have
grown some distance from the top and bottom of the flow the isothermal
surfaces will be tilted because of transport of hotter lava from nearer the vent
over the already solidified lower tier of columns. For example, in figure 1, right,
creep of the still liquid interior has carried A to A'. Now two similarly related
points of the same temperature (A' and B) are no longer in the same vertical
line, and in consequence the isothermal surfaces between them are inclined.
Continued growth of both the basal and the top tier of columns, of course,
occurs at right angles to these tilted isothermal surfaces. Hence new extensions
on both the upper and lower tier of columns will be tiked (fig. 1, lower right,
and plate 1, fig. 1). The columns appear to have been dragged in the direction
of flow, but of course the ‘‘drag” is not caused by actual bending or distortion
of already solidified columns: it is only the result of continued growth of the
joints in conformity with the tilted isothermal surfaces.
The upper ends of most gas chimneys
and tubules bend in the direction of
lava motion, but some break up into
irregular cavities and clusters of
vesicles that trail out horizontally
downstream. Steam explosions in water
saturated sediments beneath the flow
blast mud and other debris into the
chimneys; this, too, is scattered to
leeward by lava movement.

The large gas chimneys are called

spiracles (Fuller. 1931, p. 292-293.
299). In the Columbia River basalt—
where they are common—spiracles
range from a few' inches to scores of
feet in diameter. Most widen slightly or
even mushroom after rising a few feet
into the flow. After rising vertically for
as little as a foot or as much as forty
feet most of them bend abruptly or
else trail out as vesicle clusters in the
direction of flow.
often propagate inward and divide the rock into prismatic columns.
It has been shown that:
(1) the initial crack pattern that forms at the surface of the body is not as regular as the pseudo-hexagonal pattern
which evolves later;
(2) columnar structures develop by repeated, step-wise crack advances, forming cm to dm transverse bands on the
column faces;
(3) even within well-developed colonnades, the polygonal outlines continue to shift slightly from one crack advance
to the next.
We propose that each new crack should propagate parallel to the highest thermal gradient ahead of the current
crack tip. When adjacent columns are of unequal size, the local asymmetry of the isotherms drives the new crack
toward the biggest and hottest column.
We present a geometrical model and algorithm that follows this rule and mimics the successive crack advances as
the cooling front migrates into the lava sheet, or dyke.
The model polygonal pattern obtained consists of convex, irregular polygons with a variable number of sides, but
in which pentagons and hexagons predominate.
The algorithm successfully reproduces the rapid evolution from an initial, immature crack pattern to a pseudo-
hexagonal, mature one in which the average number of sides approaches six. It predicts also the predominance of
Y-type crack junctions, the rapid evolution toward columns of approximately equal sizes, and the persistent changes
in length and diverging orientation of growth steps observed along the sides of columns.
The model polygonal patterns exhibit geometrical properties which are very similar to those observed in well-
developed columnar jointed flows such as the Giant's Causeway.

The columns may form sets. Straight, regular
columns are called colonnade. Irregular, fractures
columns are called entablature.
From Spry (1962).
The columns form due to stress as the lava
cools (Mallet, 1875; Iddings, 1886, 1909;
Spry, 1962). The lava contracts as it cools,
forming cracks. Once the crack develops it
continues to grow. The growth is
perpendicular to the surface of the flow.
Entablature is probably the result of cooling
caused by fresh lava being covered by water.
The flood basalts probably damned rivers.
When the rivers returned the water seeped
down the cracks in the cooling lava and
caused rapid cooling from the surface
downward (Long and Wood, 1986). The
division of colonnade and entablature is the
result of slow cooling from the base upward
and rapid cooling from the top downward.
Possible mechanism for the formation of
columnar jointing at Devils Tower. Isotherms are
layers with the same temperature. Joints formed
perpendicular to the isotherms as the rock
cooled. From Spry (1962).
Stria and plumose features. (a) shows chalked demarkations of individual striae, and some
plumose structure. (b) shows the curvature of several striae that imply that this column cooled
from the top of the picture downwards. (c) shows alternating rough and smooth regions that
can be used to measure stria heights (d) shows large striae and plumose. (e) shows a
schematic representation of these features, with the crack initiation points shown as black
dots, plumose shown as thin lines, stria edges as thick lines, and the alternation of smooth
and rough surfaces as surface shading. A dashed circle in the upper left corner emphasizes
that subsequent striae often curve slightly in order to intersect at right angles. Labels indicate
the growth stages of a single stria. In (a),(d), arrows show the inferred direction of crack
propagation within individual striae.
Observation
Columnar ordering
Stria height vs. column width
Stria heights vs. position relative to the flow margin
Statistical distribution of stria heights
Undulations of the column width

Analysis
Conductive cooling regime
Convective cooling regime

Examples of wavy columns from the (a) Banks Lake II. and (b)
Discussion Bingen II sites. This undulation of the olumn faces can also be
Striae observed at the Devil’s Past-pile (c). in California. In (a), the
inner joint between two wavy columns, marked by an arrow, can
Convection be seen to undulate. This indicates that wavy columns are not an
Column scaling erosional pattern. The region indicated by a rectangle in (a) is
shown close up, with enhanced contrast, in (d), in order that the
Wavy column striae can be seen. The wavelength of the undulations is much
Residual disorder larger than the width of the striae, (e) shows a sketch of the
surface of a wavy column, as it would appear in cross-section
Model extinction (cut perpendicular to the column surface, and parallel to the
direction of crack propagation), and shows how the wavelength
ç and angle Ø of wavy columns are defined