Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Critical porosity: A key to relating physical

properties to porosity in rocks


AMOS NUR, GARY MAVKO, JACK DVORKIN, and DORON GALMUDI, Stanford Rock Physics Laboratory,
Downloaded 05/21/13 to 147.188.128.74. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Stanford, California

Discovering accurate relations tively) and density ρ as M = ρ Vp2 (between an upper and a lower
between porosity and seismic wave and G = ρ Vs2. The bulk modulus K, bound) possible at a given porosity.
velocities in porous rocks has been the inverse of compressibility, is K = The best known bounds are the
an important topic of rock physics M - 4/3G. Reuss lower bound and the Voigt
for decades. Such relations, if they Approaches to establishing poros- upper bound. Consider a saturated
actually exist, are critical when seis- ity-velocity or porosity-modulus rela- rock of porosity ø. Then, for exam-
mic and/or sonic data are to be tions can be categorized as empiri- ple, for its bulk modulus K, we have
used to infer porosity and porosity- cal relations, theoretical bounds, and K R ≤K≤K V , K V = (1-ø)K s + øK f and
related properties (e.g., permeabili- deterministic models. 1/KR = (1-ø/Ks + ø/Kf where KR is
ty, strength, saturation) in situ. Empirical relations are obtained the Reuss bound, KV is the Voigt
These inferences became especially statistically for a given data set. They bound, K s is the modulus of the
important in recent years, with the are useful for summarizing and solid phase, and Kf is the modulus
rapid development of seismic and parameterizing empirical observa- of the pore fluid. An example of
sonic methods which are applied to tions but often fail when applied out- using these bounds is given in Fig-
the characterization of heteroge- side the range (porosity and rock ure 1 for a quartz rock filled with
neous reservoirs and recovery mon- type) covered by the data. They sel- water.
itoring. dom carry physical insight and thus Bounding the moduli may be
Often the elastic moduli of rock resist generalization. extremely useful if the bounds are
are used instead of velocities. Bounds, in contrast, are rigor- narrow. However, this is not the case
Among them are the compressional- ously derived from basic physical when the moduli of the two con-
wave modulus (or the M-modulus) principles and thus are widely stituents differ from each other as
and the shear modulus (G). They can applicable. However, they do not much as those of quartz and water.
be expressed through the P- and S- give specific velocity estimates but In this case, even more accurate
wave velocities (Vp and Vs respec- provide a range of velocity values bounds, such as those of Hashin and

Figure 1. (a) Bounds for a


quartz rock with water.
1 = Voigt, 2 = Reuss, 3 =
upper Hashin-Shtrikman,
4 = lower Hashin-Shtrik-
man. The lower Hashin-
Shtrikman bound is iden-
tical to the Reuss bound,
because the shear modu-
lus of water is zero. (b)
Deterministic models for
the same rock. 1 = Wyllie,
2 = Raymer, 3 = Kuster
and Toksöz model where
pores have aspect ratio of
0.5, 4 = the latter model
with porosity scaled by
critical porosity. (c)
“Dirty” sandstones; veloc-
ity predicted by Raymer
versus measured velocity.
(d) Unconsolidated high-
porosity sands from the
Troll Field in the North
Sea. RHS = rescaled lower
Hashin-Shtrikman bound.
Bars on the porosity axes
are at critical porosity.

MARCH 1998 THE LEADING EDGE 357


Shtrikman (“A variational approach
to the elastic behavior of multiphase
materials” in Journal of Mechanics and
Physical Solids, 1963), fail to narrow
the possible span of moduli at a
given porosity (Figure 1a).
“Naive” deterministic models
attempt to find a meaningful expla-
Downloaded 05/21/13 to 147.188.128.74. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

nation to experimental observa-


tions. The best known example is
Wyllie's time average (“Elastic wave
velocities in heterogeneous and
porous media,” GEOPHYSICS, 1956):
1/Vp = (1 - ø)/Vps + ø/Vpf, where Vps Figure 2. Physical meaning of critical porosity.
is the velocity in the mineral phase
and Vpf is that in the fluid phase.
The popularity of this formula is
due to its simplicity and also due to
the fact that by “fudging” the Vps
and Vpf values for various rocks and
fluids, one can match practically
any data point.
Raymer et al. (“An improved
sonic transit time-to-porosity trans-
form,” SPWLA 21 Annual Logging
Symposium, 1980) modified this for-
mula by suggesting different laws
for different porosity ranges. In a
sense, these authors recognized the Figure 3. (a) Critical porosity in clean sandstones separates the consolidat-
need for two porosity domains: The ed-rock domain from the suspension domain. Mechanical behavior drasti-
consolidated-rock domain and the cally differs between these two domains. (b) Bulk modulus versus porosity
suspension domain. It is remarkable for air-filled glass foam. Critical porosity is about 90%. 1 = Voigt bound,
how well this model describes exper- 2 = upper Hashin-Shtrikman bound, 3 = upper Hashin-Shtrikman bound
imental data for clay-free sandstones scaled with critical porosity.
(Figure 1b), as well as for “dirty”
sandstones with clay (Figure 1c).
However, these models fail to
describe unconsolidated sandstone
reservoirs such as found, for exam-
ple, in the North Sea (Figure 1d). We
can quickly fix this problem by
applying the critical porosity princi-
ple. Once we rescale the lower
Hashin-Shtrikman bound to include
the critical porosity point, it starts
acting as a fairly accurate lower
bound for the collected data (see
details in “Elasticity of high-porosi-
ty sandstones: Theory for two North
Sea data sets,” by Dvorkin and Nur, Figure 4. M-modulus (left) and shear modulus (right) versus porosity for
GEOPHYSICS, 1996). sandstones. Bar on the porosity axis is critical porosity.
It is clear that the above “naive”
deterministic models cannot be used diagnose a rock; i.e., to understand to that used in the model. Resulting
to estimate Vs since the shear-wave its internal topology from velocity are avenues to estimating perme-
velocity in fluids is zero. Still, we are and porosity measurements. Such ability, strength, and other important
aware of attempts to assign certain diagnostics may be important for parameters.
shear-wave velocity values to brine, estimating, e.g., permeability and Rigorous models that assign
oil, or gas in order to force Wyllie's strength from porosity and velocity some idealized shape to the grains or
time average to “predict” Vs. Of measurements. pores typically relate the elastic
course, such attempts are nonphysi- Rigorous deterministic models are properties of rocks to porosity, over
cal, but this is only part of the story. based on first principles. They treat the entire porosity range — from
More important is that they are not rock as a composite and thus require zero to 100%. An example is given in
based on first principles and thus specific geometric information about Figure 1b. Here, the Kuster and Tok-
may force experimental values into the constituents. If such a model fits söz (“Velocity and attenuation of
an inappropriate functional form. the data, it is valid to conclude that seismic waves in two-phase media,”
Also, these models cannot be used to the rock's internal topology is close GEOPHYSICS, 1974) model's predic-

358 THE LEADING EDGE MARCH 1998


tions are not only far from the exper- lated velocity-porosity function from we are able to drastically improve
imental data points but also fail to ø to ø/øc, where øc is the porosity at the performance of a rigorous deter-
recognize the apparent break in the which the break in the velocity- ministic model.
velocity-porosity behavior occurring porosity trend occurs (Figure 1b). In This result encourages us to pay
at porosities of about 40%. It is inter- other words, let us plot function Vp close attention to the porosity value
esting that the “naive” Raymer et al. (ø/øc) where Vp (ø) is obtained from at which a break in the velocity-
model does recognize this break — the Kuster and Toksöz model. The porosity trends occurs. Probably this
because it is empirical, and the data result is striking: the rescaled model critical porosity is a key to relating
Downloaded 05/21/13 to 147.188.128.74. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

require the break. closely matches the data for porosi- physical properties to porosity in
Let us ad hoc rescale the results ties between zero and øc. Thus, by rocks.
of the Kuster and Toksöz model by recognizing that rocks behave differ-
changing the argument in the calcu- ently in different porosity domains, A primer on critical porosity. To elu-
cidate the concept of critical porosi-
ty, let us consider relatively clay-free
sandstones. It is valid to assume that
the starting point of formation of
such rocks is quartz sand. A typical
porosity of such sand is about 40%.
It is remarkably close to a funda-
mental value of 36% which is the
porosity of a close random pack of
identical spheres. As diagenesis pro-
gresses, sand grains become com-
pacted and cemented, and porosity
decreases down to zero, at which
point we have solid quartz (Figure
2). In such sandstones, cemented
Figure 5. M-modulus versus porosity for brine-saturated basalts. Young grains form a continuous frame
oceanic crusts on to the left, and deep-sea basalts to the right. Bar on the which is responsible for transmitting
porosity axis is critical porosity. any applied load. This rock is frame-
supported.
Is it possible to have sands with
porosities exceeding 40%? Yes, but
then the grains have to be separated.
They do not (or barely) touch each
other and thus cannot transmit
applied load. Now pore fluid bears
the load, and the material is a fluid-
supported suspension. If the rock is
vacuum-dry, then, in the suspension
state, the elastic moduli are zero.
In this suspension case the sand-
stone data fall close to the Reuss
bound (Figure 3). This is because
the Reuss bound is derived by
assuming that stress is uniform
Figure 6. Vp (left) and M-modulus (right) versus porosity for chalks. Bar throughout the composite — the sit-
on the porosity axis is at critical porosity. uation typical for suspensions
where solid grains are totally sur-
rounded by fluid (an isostress sys-
tem). Isostress systems include sand
grains and foraminiferal skeletons
on the ocean floor, liquefied sedi-
ments, slurries, and rocks with suf-
ficient crack concentration to cause
them to fall apart. Relevant geolog-
ic situations and processes in which
such suspensions may be involved
include sedimentation, liquefaction
due to earthquakes, sanding in
boreholes, sand volcanoes in accre-
tionary wedges, crack percolation
in breaking rocks, and fault zones.
The Reuss bound is identical to the
Figure 7. Vp (left) and M-modulus (right) versus porosity for cracked lower Hashin-Shtrikman bound if
igneous rocks. Bar on the porosity axis is at critical porosity. one of the phases is fluid.

MARCH 1998 THE LEADING EDGE 359


Therefore, the critical porosity, mica. In spite of this diversity, there explanation for this large value.
which is in this case about 40%, sep- is a general trend of moduli decreas- Chalks have a granular structure
arates two fundamentally different ing with increasing porosity and similar to that of sandstones. There-
domains — one of consolidated, becoming zero for porosities higher fore, the transition from frame-sup-
frame-supported rocks, and another than 40%. This value is the critical ported state to fluid-supported state
of fluid-supported suspensions. porosity value for sandstones. in chalks should occur at grain con-
These speculations are supported The M-modulus versus porosity centration of about 60%, or void frac-
by the data where the critical porosi- for basalts from the young oceanic tion øv of 40% (the porosity of a ran-
Downloaded 05/21/13 to 147.188.128.74. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

ty value separates the two domains of crust and from the deep sea is given dom pack of identical spheres). At
the mechanical behavior of sand- in Figure 5. The critical porosity the same time, chalk drains have
stones (Figure 3a). These data cannot value for these rocks is apparently intrinsic porosity øg. Let us assume
be matched by an effective medium 15-20%. that this porosity is about 40%. Then
model that does not use the critical The apparent critical porosity critical porosity can be found from
porosity concept. However, once the value for chalks, young and old, is 1-øc = (1-øv)(1-øg), which gives øc =
model is modified to recognize the about 65% (Figure 6). There is an 0.64.
critical porosity phenomenon, it
matches the data perfectly (Figure 1b).
Measurements of the bulk mod-
ulus of air-filled glass foam are sum-
marized in Figure 3b. Glass foam has
a honeycomb structure where air
bubbles are separated by thin glass
membranes. These membranes form
a continuous frame that can support
load up to almost 100% porosity. In
this case, the material is frame-sup-
ported practically over the whole
porosity domain. This is probably
the stiffest possible arrangement of
glass and air. Indeed the data are
very close to the upper Hashin- Figure 8. Vp (left) and M-modulus (right) versus porosity for dolomites. Bar
Shtrikman bound. The match be- on the porosity axis is at critical porosity.
tween the data and the theoretical
predictions becomes perfect if we
rescale the upper Hashin-Shtrikman
curve by assuming that the critical
porosity is 90%.
Close analogs for glass foam
among geomaterials are pumice (see
below), and mudcake accumulated
on the walls of a well during drilling.
Mudcake may have porosity up to
80% and remain mechanically stable
and practically impermeable.
Based on the evidence presented
for clean sandstones and glass foam,
we formulate the following working
hypothesis: A critical porosity value Figure 9. Vp (left) and M-modulus (right) versus porosity for limestones.
exists which is typical for a given class Bar on the porosity axis is at critical porosity.
of porous materials. Each class is defined
on the basis of common mineralogy
and/or diagenetic porosity reduction
process. In order to validate this
hypothesis and support the above-
formulated critical porosity concept,
we present data collected by various
authors for different rock types.

Application to various rocks. Dry


sandstone elastic moduli at high
pressure (derived from ultrasonic
velocity measurements) are given in
Figure 4. Here clean consolidated
sandstones are mixed with clay-rich
sandstones and with unconsolidated Figure 10. Vp (left) and M-modulus (right) versus porosity for rock salt. Bar
high-porosity sandstones containing on the porosity axis is at critical porosity.

360 THE LEADING EDGE MARCH 1998


Critical porosity for cracked identical spheres. This is why critical Chen and Nur, Critical concentra-
igneous rocks is about 5% (Figure 7). porosity here is, similar to sand- tion models for porous materials in
The reason for such a low critical stones, about 40% (Figure 11). Advances in Porous Media, 1994).
porosity value is that these rocks are Data for pumice indicate that the Below we give a simple example of
fractured. Therefore, they cease to critical porosity is as high as 80% a critical porosity effective medium
exist as intact structures at very small (Figure 12). The reason is the foam- theory — the modified Voigt
porosities. like structure of this porous volcanic bound.
For dolomites, limestones, and rock. The original Voigt bound is a lin-
Downloaded 05/21/13 to 147.188.128.74. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

rock salt (Figures 8-10), critical ear interpolation between the elastic
porosity is about 40%. Modified Voigt bound. Several ef- moduli of two constituents of a two-
Sintered glass bead pack is an fective medium theories have e- phase composite. Consider a mixture
artificial material whose starting merged that are based on the criti- of quartz and water. At zero porosi-
point is a dense random pack of cal porosity concept (see chapter by ty, the volumetric concentration of
quartz is one, and the modulus of the
composite is that of quartz. At 100%
porosity, the volumetric concentra-
tion of water is one, and the modu-
lus of the composite is that of water.
For other porosities, the bound is a
straight line connecting the two end
members (Figure 13a).
Let us modify the original Voigt
bound by replacing the water end
member (at 100% porosity) with the
suspension end member (at critical
porosity). The latter can be calculat-
ed by using the Reuss bound (e.g.,
for bulk modulus): Kc = [(1-øc)Kq-1 + øc
Figure 11. Vp (left) and M-modulus (right) versus porosity for sintered Kw-1]-1 where Kc is the bulk modulus
glass. Bar on the porosity axis is at critical porosity. of the suspension at critical porosity
øc , and Kq and Kw are the bulk mod-
uli of quartz and water, respectively.
Let us linearly interpolate between
the quartz bulk modulus at zero
porosity and the suspension bulk
modulus at critical porosity: Keff = (1-
ø/øc)Kq+ ø/øc Kc. The effective shear
modulus Geff of a composite can be
calculated in a similar way: Geff = (1-
ø/øc )Gq where Gq is the shear modu-
lus of solid quartz, and the shear
modulus of suspension is zero. This
modified Voigt bound provides,
apparently, an excellent estimate for

Figure 12. Vp (left) and M-modulus (right) versus porosity for pumice. Bar
on the porosity axis is at critical porosity. Table 1. Critical porosity for vari-
ous natural and artificial rocks

Material Critical porosity

Natural rocks
Sandstones 40%
Limestones 40%
Dolomites 40%
Pumice 80%
Chalks 65%
Rock salt 40%
Cracked igneous rocks 5%
Oceanic basalts 20%
Artificial rocks
Figure 13. (a) Voigt, Reuss, and modified Voigt bounds for bulk modulus ver- Sintered glass beads 40%
sus porosity. (b) Clean sandstone data and modified Voigt bound for P-wave Glass foam 90%
velocity versus porosity.

MARCH 1998 THE LEADING EDGE 361


Figure 14. Elastic modulus (nor-
malized by the mineral modulus)
versus porosity (normalized by crit-
ical porosity).

Vp in consolidated clean sandstones


(Figure 13b).
Downloaded 05/21/13 to 147.188.128.74. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Conclusion. The critical porosity


values implied by the above obser-
vations on the velocity (moduli)
versus porosity trends in natural
and artificial rocks are summarized
in Table 1. The generality of the crit-
ical porosity concept becomes evi-
dent if data for different rock types
are consolidated in a single plot. To
do so, we normalize the elastic
modulus of a rock by that of the
appropriate solid mineral, and
porosity by the appropriate critical
porosity value. As a result, data for
various classes of natural and arti-
ficial materials have a simple linear
trend (Figure 14).
The critical porosity value
divides the entire porosity range into dated, frame-supporting state. If the Acknowledgment: This work was sup-
two domains. Most reservoir rocks traditional effective medium theories ported by the Stanford Rock Physics and Bore-
fall in the first consolidated-rock are modified to recognize the effec- hole Geophysics industrial consortium (SRB).
domain. In this domain, where tive porosity divider, they can give
porosity varies from zero to the crit- improved predictions for the Corresponding author: J. Dvorkin, phone,
ical porosity, rock is in the consoli- acoustic properties of rocks. LE 605-725-9296

362 THE LEADING EDGE MARCH 1998

You might also like