Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Velocity Dispersion and

Wave Attenuation (Q)

189
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

(1) Seismic velocities almost always increase with

frequency, and

(2) Seismic waves are always attenuated as they

travel through rocks.

These two observations are usually intimately related.

Both usually increase from dry to fluid saturated
conditions, and both usually decrease with increasing
effective pressure.

These effects complicate the comparison of laboratory

and field data, but they also reveal details about the
pore space and the pore fluids it contains.

190
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Kramers-Kronig Relations

∞
−1 |ω | MR ( α) − M R (0) dα
Q (ω ) =
πM R (ω ) −∫∞ α α −ω
∞
−ω Q−1(α )MR (α ) dα
MR (ω ) − M R (0) = ∫− ∞
π |α | α −ω

Causality leads to a very specific relation

required between Q and modulus dispersion

191
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

I.1

In most rocks and sediments, the velocity tends to

increase with frequency. This is accompanied by
attenuation. Attenuation tends to be highest in
frequency range where velocity is increasing most
rapidly.

192
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Fluid “squirt”
Biot

Thermoelastic

In real materials, we expect that multiple mechanisms

of attenuation are present, each having its own
characteristic frequency and magnitude.

In fact, we might expect a fairly constant level of

attenuation over wide frequency bands.

193
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

I.3

The difference between dry and saturated velocities and the disagreement
with the low frequency Gassmann theory often increases with fluid
viscosity. Again the differences are greatest at low pressures.
Data from Winkler (1985).

194
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Measured
ultrasonic
Velocity

saturated
dispersion

Gassmann

dry rock Measured

ultrasonic

3-D Seismic ultrasonic

lab
frequency
The most common recipe for applying ultrasonic core
data to field conditions is to use velocities measured
on (nearly) dry cores and then use Gassmann to add
the fluids. The basic assumption is that velocity
dispersion is smaller for dry or nearly dry rocks, so
that the ultrasonic dry velocities are good estimates
of the low frequency dry velocities.

195
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

I.4

Failure of Gassmann's theory to predict saturated ultrasonic velocities

relative to dry velocities. Navajo sandstone data from Coyner (1984).

196
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Viscoelasticity
We have been talking about elastic materials where
stress is proportional to strain.
σ11 + σ 22 + σ 33
= K(ε11 + ε 22 + ε 33) volumetric
3
σ ij = 2µε ij shear

σ ij = λδ ijεαα + 2µε ij general

Viscoelastic materials also depend on rate or history.
σ˙ ij σ ij
ε˙ij = + Maxwell model
2µ 2η
σ ij = 2η ε˙ij + 2µε ij Voigt model
(
ησ˙ ij + ( E1 + E 2 )σ ij = E 2 ηε˙ij + E1ε ij ) Standard linear
solid
E1
E2

η
197
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Wave Propagating in a Viscoelastic

Solid

u(x,t) = u0 exp[−α (ω ) x ]exp[i (ωt − kx)]

At any point, the stress and strain are out of phase:

σ = σ 0 exp[i (ωt − kx)]

ε = ε 0 exp[i (ωt − kx − ϕ )]

The ratio of stress to strain is the complex modulus.

198
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Quality Factor “Q”

low Q: large dissipation

high Q: small dissipation

Different views of Q:
energy dissipated per wave
1 ΔW cycle
=
Q 2πW peak strain energy of the wave

1 αV velocity
=
Q πf
frequency

1 1  u(t)  amplitude loss per cycle

≈ ln 
Q π  u(t + τ )
1
= tan(ϕ )
Q phase delay

199
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Spectral Ratio Method

We can think of Q-1 as the fractional loss per
wavelength or per oscillation. Therefore over a fixed
distance there is a tendency for shorter wavelengths
to attenuate more: 1 αV
≈
Q πf
or
πf
α≈
If we propagate the wave VQ
u = u0 exp[−αx ]
Then we can compare the amplitudes at two different
distances: u 
ln x 2  = −α ( x 2 − x1 )
 ux1 
 ux 2  πf
ln  = −
 ux1 
( x − x1)
QV 2

200
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Standard Linear Solid

If we assume sinusoidal motion
ε = ε 0e iωt
σ = σ 0e iωt
Then we can write:
σ 0 = M(ω )ε 0
with the complex, frequency-dependent modulus
 ω 
M∞  M 0 + i M0 M∞ 
E 2 ( E1 + iωη)  ωr 
M(ω ) = =
E1 + E 2 + iωη ω
M∞ + i M0 M∞
ωr
In the limits of low frequency and high frequency
E 2 E1
M0 = , ω→0
E1 + E 2

M ∞ = E 2, ω→∞
  ω  2
M0 M∞ 1+   
  ω r  
Re{M (ω )} = 2
ω 
M ∞ +   M0
 ωr 
201
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Standard Linear Solid

Similarly, we can write Q as a function of frequency:
ω
M0 M∞ ( M∞ − M0 )
1 M I (ω ) ωr
= =
Q MR (ω )   ω 2
M 0 M ∞ 1 +   
  ωr  
ω
 
1 E2  ωr 
= 2
Q E1 ( E1 + E 2 ) ω
1+ 
 ωr 
where
E1 (E1 + E 2 )
ωr =
η
The maximum attenuation
 1 1 E2
  =
 Q max 2 E1(E1 + E 2 )
 1 1 M ∞ − M 0 1 ΔM
  = ≈
 Q max 2 M0 M∞ 2 M
occurs at

ω = ωr
202
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Standard Linear Solid Model

I.5

Attenuation and velocity dispersion tend to be most

localized in frequency. Attenuation is largest where
velocity is changing most rapidly with frequency.
Peak attenuation and modulus dispersion are related
by:
1 1 ΔM
  =
 Q  max 2 M
203
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Nearly Constant Q Model

ω1 ω 2

€ €

I.6

Attenuation is nearly constant over a finite range of

frequencies. It is sometimes interpreted as a super-
position of individual (Standard Linear Solid)
attenuation peaks. The broadening of the attenuation
peak is accompanied by a broadening of the range of
frequency where velocity increases.
 1  1 ΔM π
  ≈
 Q  2 M ln(ω 2 /ω1 )

204
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Nearly Constant Q Model

Liu, et al. (1976) considered a model in which simple
attenuation mechanisms are combined such that the
attenuation is nearly a constant over a finite range of
frequencies.
We can then write
V (ω 2 ) 1
= 1+ ln(ω 2 /ω1 )
V (ω1 ) πQ
which relates the velocity dispersion within the band
of constant Q, to the value of Q and the frequency.
We can express V (ω 2 ) /V (ω1 ) as:
€
V (ω 2 ) ΔM
≈ 1+
V (ω1 ) M1
€
Expanding for small ΔM /M and substituting in:
1 ΔM 1
≈ ln(ω 2 /ω1 )
€ 2 M1 πQ
Finally  1  1 ΔM π
  ≈
 Q  2 M1 ln(ω 2 /ω1 )
€ 205
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Constant Q Model

1/Q

log(ω) I.7

Attenuation is constant for all frequencies, and

velocity always increases with frequency.

206
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Constant Q Model

Kjartansson (1979) considered a model in which Q is

strictly constant. In this case the complex modulus
and Q are related by:
2γ
 iω 
M(ω ) = M1 
 ω1 
where

1  1
γ = arctan 
€ π  Q

As with the nearly constant Q model, we can simplify

this for large Q, giving:

 1  1 ΔM π
  ≈
 Q  2 M1 ln(ω 2 /ω1 )

207
€
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Fontainebleau Sandstone
Velocity Attenuation
5000 200
BR-E
4500 BR-S
US-P
US-S
4000 150
Velocity (m/s)

3500

1000/Q
3000 100

2500

2000 50
BR-E
BR-S
1500 US-P
US-S
1000 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

Pressure (MPa) Pressure (MPa)

Velocity Dispersion
1.2
Extension
Torsion
1.1 CQ Model
NCQ Model

1
Vbr/Vus

0.9

0.8

0.7

0.6

0.5
0 50 100 150 200 250 300
1000/Qbr I.8

Just as velocity increases with effective pressure, so does

Q. The strong pressure dependence is a clue that cracks
are important for the physical mechanism of attenuation.
From Nathalie Lucet, 1989, Ph.D. dissertation, Univ. of
Paris/IFP.

208
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Location Type of rock Depth (m) Measurement

frequency (Hz)

Limon (Colorado) Pierre shale 0-225 50-450 32 McDonald et al.

(1958)
Gulf Coast (30 km Loam/sand/clay 0-3 50-400 2
south of Houston) Sands and shales 3-30 50-400 181 Tullos and Reid
Sandy clay 30-150 50-400 75 (1969)
Clay/sand 150-300 50-400 136
Offshore- Clay/sand 1170-1770 ≤125 67
Lousiana Sands and shales 1770-2070 ≤125 >273
(Pleistocene) Same but more sandy 2070-2850 ≤125 28
Southeast Texas Sandbanks, silty shale 900-1560 ≤80 52
Southeast Texas Mostly shale 1560-1800 ≤80 >273
Southeast Texas Sand (23%) and clay 1800-2100 ≤80 30 Hauge (1981)
Sand (20%) and clay 600-1560 ≤80 41
Limestone and chalk 1590-1755 ≤80 >273
Sand (45%) and clay 660-1320 15-40 28
Sand (24%) and clay >1020 40-70 55
Beaufort Sea 549-1193 125 Ganley, Kansewich
(Canada) 945-1311 425 (1980)
Offshore Siliceous chalk 278-442 5000-15000 68 on
Baltimore ave.
Siliceous chalk with 442-582 5000-15000 287 Golberg (1958)
porcellanite joints on
ave.
S an d b an k s,ilty sh ale

from Carmichael (1984) and Goldberg (1985)

Some values of Q in rocks and sediments, summarized

by Bourbié, Coussy, and Zinszner, 1987, Acoustics of
Porous Media, Gulf Publishing Co.
209
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Dominant fluid mechanisms affecting velocity and Q

Gassmann Model
• static, zero frequency limit
• no viscous/inertial effects
• uniform pore pressure

Biot "global flow" Model

• viscous/inertial effects
• average flow only
• average relative motion of fluid
and solid lead to dispersion
and attenuation
Squirt "local flow" Model
Hi Pp • viscous/inertial effects
• grain-scale fluid motion leads to
Hi Pp
Hi Pp dispersion and attenuation
• effectively stiffens the soft porosity
• superimposed on Biot/Gassmann
Patchy Saturation Model
• large scale patches of saturation
• patch-scale diffusion leads to
dispersion and attenuation
• microscale squirt can be
superimposed

210
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Wave-induced fluid motion in the rock

appears to be the dominant source of
attenuation and dispersion:

• In fully-saturated rock, squirt flow between

the stiff and soft parts of the pore space
• In partially saturated rock, gas gives the
viscous liquids more mobility

SQUIRT
FLOW

211
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Q and Gas Saturation

Lab data and field observations indicate
that Q may be used to detect gas-
saturated reservoirs

Murphy's (1982) experiments show that

(a) attenuation in gas saturated rocks is larger
than in dry rocks,
(b) attenuation peaks at low gas saturation.
1000/Qe

70
811 - 846 Hz
60
1000/Qe

50

40
571 - 647 Hz

30
MASSILON SANDSTONE
EXTENSIONAL LOSS
20

0 0.5 Sw 1

212
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Q and Gas Saturation

Lab data and field observations suggest
that Q might help to detect gas

Klimentos (1995) reports, based on well log data,

that P-wave attenuation in gas-saturated sandstone
is larger than in liquid-saturated sandstone.
20 20
Gas +
Condensate
Klimentos '95
(dB/m)

(dB/m)
(dB/cm)

(dB/cm)

Well Logs
Mediium Porosity SS
15 15 ~ 2500 m
Attenuation
Attenuation
P-Attenuation

S-Attenuation
S-Wave
P-Wave

10 10
Oil +
Klimentos '95
Well Logs Water
Gas +
Mediium Porosity SS Oil +
~ 2500 m Condensate Water
5 5
1.5 1.6 1.7 1.8 1.5 1.6 1.7 1.8
Vp/Vs Vp/Vs

213
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Biot Theory
Biot developed a macroscopic theory to attempt to model
the behavior of fluid-saturated poroelastic systems.
His generalized form of Hooke’s law:
σ ij = λδ ijεαα + 2µε ij − βPδ ij
1
ξ= P + βεαα
M
where λ and µ are the dry rock moduli, and the fluid
pressure P is linearly related to the normal stresses (and
not the shears) by a new constant β. Similarly, the
increment of fluid content ξ in an elementary cell of solid is
linearly related to the pore pressure and the solid
volumetric strain εαα. These describe essentially the same
mechanical problem as the Gassmann theory for coupling
the fluid and solid.
The equations of motion are:
2 2
∂σ ij ∂ ui ∂ wi
= ρ 2 + ρuw 2
∂x j ∂t ∂t
∂P ∂ 2 ui ∂ 2 wi 1 ∂wi
= ρ uw 2 + ρw 2 +
∂x i ∂t ∂t χ ∂t
where ρuw describes an inertial coupling between the solid
and the fluid, and χ is a dissipation term.

214
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Biot Theory
The high frequency limiting velocities VP∞ and VS∞ , are
given by [in Geertsma and Smit’s (1961) notation]:
1
  ρ −1  K fr   K fr   2
  φ a + 1−   1− − 2φa−1  
 1  K + 4 µ  + ρ fl  K0   K0  
VP∞ = −1  fr
3 fr 
 ρ 0 (1− φ) + φρ fl (1− a )   K fr  1 φ

  1− − φ +
   K0  K0 K fl 
1
 µ fr  2

VS∞ =  −1 
ρ 0 (1− φ ) + φρfl (1− a ) 

where
Kfr, µfr bulk and shear moduli of dry rock frame
K0 bulk modulus of mineral
Kfl effective bulk modulus of pore fluid
φ porosity
ρ0 mineral density
ρfl fluid density
ρ low frequency density of saturated composite:
ρ = 1 – φ ρ0 + φρ fl
a-1 tortuosity
The low frequency limiting velocities are the same as
predicted by Gassmann’s relations.

215
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Biot Theory

I.9

Biot and later Stoll (1977) considered adding frame

attenuation on top of fluid effects. This plot by Stoll
shows attenuation vs. frequency for two extreme
cases and for a typical sand showing how frame
losses and fluid losses combine to control the overall
response.
216
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Squirt Flow

When a rock is compressed by the stress of a passing

wave, increments of pore pressure are induced in the pore
fluid. At very low frequencies there is time for the pore
pressure to equilibrate throughout the pore space, and the
fluid effect is described by the Gassmann theory.
However, at high frequencies we expect that unequal pore
pressures are induced on the microscale of individual
pores--larger increments in the soft, crack-like porosity and
smaller increments in the stiffer, equi-dimensional pores. If
these do not equilibrate, the rock will be stiffer, and the
velocities will be faster, than at low frequencies when they
do equilibrate.
This frequency-dependent distribution of pore pressure
leads to velocity dispersion, and the tendency for the fluid
to flow and adjust leads to attenuation.

217
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Estimating the High Frequency

Squirt Flow Modulus
We compute the high frequency bulk modulus in 2 steps:
1. the unrelaxed bulk modulus of the “wetted” frame where
liquid is trapped in the thinnest cracks and the remaining
space is dry, is given by
1 1  1 1  high P ∂Pi
− = − ∑P φi ∂σ
K high f Kmineral  K fluid K mineral  low

P
4

P P P
2 3
1

pore pressure in the ith thin crack:

∂Pi 1
≈
∂σ  1 1 
1+  −  φ i / (∂φ i / ∂σ )dry
 Kfluid K mineral 
Combining gives:
1  1   1 1 
≈
K high f  K dry 
+
K
−
K  φ soft (σ ) + K
highP
 fluid mineral 
So trapping water in the thinnest cracks is approximately the
same as closing the cracks under high pressure.
2. Finally the remaining pore space is saturated using
Gassmann with Khigh f used as the “dry” rock modulus.

218
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Squirt Flow

The shear modulus:

1 1 4 highP (∂φi / ∂σ )dry

µhigh f
≈ − ∑
µdry 15 lowP  1 1 
1+  −  φ / (∂φ i / ∂σ )dry
 Kfluid Kmineral  i

Comparing with the bulk modulus:

 1 1  4  1 1 
µ −  ≈ 15  K − 
 high f µ dry   high f Kdry 

219
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Constructing the Unrelaxed Moduli

I.10

220
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

I.11

In these plots, the dry data are taken as inputs. The ultrasonic
water-saturated data are compared with predictions by Gassmann,
the high frequency Biot limit, and the high frequency squirt limit.

221
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

I.12

222
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Most physical mechanisms of dispersion and attenuation

have a characteristic frequency where attenuation is
largest and velocity is changing most rapidly with
frequency. It also separates the low frequency “relaxed”
behavior from the high frequency “unrelaxed” behavior.

It is very difficult to predict the characteristic frequency

very accurately, because it depends on idealized model
assumptions, and details of the rock microstructure that
are not well known.

Nevertheless, here are some rough estimates:

Biot:
φη
f Biot =
2πρ f κ
squirt: K 0α 3
f squirt =
η
patchy saturation:
κK f
f patchy = 2
Lη
viscous shear in crack:
αµ
f visc.crack =
2πη
223
 Stanford Rock Physics Laboratory - Gary Mavko

Velocity Dispersion and Q

Biot Theory

Parameter Porosity Permeability Characteristic frequency

(%) (mD)
Sample Water Normal oil Heavy oil
(h = 1cP) (h = 10-50 (h = 100-
(4) cP) (4) 500 cP) (4)
Fontainebleau 5 0.1 80 MHz 800-4000 8-40 GHa
sandstone (1) MHz
Fontainebleau 20 1000 30 kHz 300-1500 3-15 MHz
sandstone (1) kHz
Tight sand (2) 8 0.02 1 GHz 10-50 GHz 100-500
GHz
Cordova 24.5 9 4.5 MHz 45-230 450-300
Cream MHz MHz
limestone (2)
Sintered glass 28.3 1000 42 kHz 420-2100 4.2-21 MHz
kHz
(1) Bourbié and Zinszner (1985)
(2) Carmichael (1982)
(3) Plona and Johnson (1980)

(4) Viscosity η is expressed in centipoises (1 cP = 1 mPa. s).

Compiled in Bourbié, Coussy, and Zinszner,

1987, Acoustics of Porous Media, Gulf Publishing Co.

224