What Can Body Waves Tell Us About The Mantle Transition Zone

1
What Can Body Waves Tell Us About The Mantle
Transition Zone?

• Presented by Jesse Fisher Lawrence
• Institute of Geophysics and Planetary Physics
• Scripps Institution of Oceanography
• University of California, San Diego
• Location Stanford University
• Date October 19th, 2006
• Web Site http//titan.ucsd.edu/
• In collaboration with Peter Shearer

2
General Structure of the Earth
From http//garnero.ucsd.edu/
Dziewonski and Anderson, 1981 PEPI
3
Why Study the Transition Zone?

• Phase changes associated with the discontinuities
  inhibit mantle flow to some extent.
• How much depends on the density contrasts, which
  can vary from place to place.
• The transition zone can also change as a result
  of convection.

4
Phase Transformations

• While most velocity density jumps are near 410
  660 km depth, the actual depth can vary
  depending on temperature.
• In seismology we can measure waves that reflect
  off of these discontinuities.

e.g., Bina Helffrich, 1994
5
Receiver Functions

• P-to-S converted waves (Pds)
• Recorded 30-90? from earthquakes.
• P-waves are recorded on the vertical component of
  a 3-component seismometer.
• P P-to-S converted waves are recorded on the
  radial horizontal component.

After Ammon, 1991 BSSA
6
Receiver Functions

• The receiver function is the vertical record
  deconvolved from the radial component.
• rf(t) dR(t)-1dZ(t)
• or
• rf(t) lR(t)-1lZ(t)noise

Vertical Record dZ(t)sZ(t)fZ(t)lZ(t)iZ(t)nZ(
t) Radial Record dR(t)sR(t)fR(t)lR(t)iR(t)
nR(t)
instrument
source
s
i
far field
f
l
local
n noise
7
Receiver Functions

• Spectral Division
• or
• In practice this is

8
Actual Receiver Functions

• Stacked Receiver functions isolate P660s P410s
  well.
• 3000 receiver functions can be calculated
  stacked in 10 minutes.
• Little interference from other waves like PP and
  PcP.
• 3317 traces added to this stack.

Lawrence Shearer, 2006 JGR
9
Look at 1D or 3D variations
Lawrence Shearer, 2006 JGR
10
SS-Precursors

• Our understanding of the transition zone was
  revolutionized by work on SS-Precursors in the
  1990s.
• Stack all available long-period records on the
  peak amplitude of the SS wave,
• Group by distance,
• Coherent signal constructively builds,
• Incoherent signal destructively interferes.

ScSScS
After Flanagan Shearer 1998 JGR
11
SS-Precursors

• Wave stripping
• S410S, S520S, S660S
• The 520-km discontinuity, while week, is a robust
  global feature.
• There is structure below the 660-km
  discontinuity.

Sub-660 gradient
Shearer 1996 JGR
12
PP-Precursors
PP

• There is a strong P410P.
• There is evidence of a P520P.
• But where is P660P?
• Is there no P660P?
• Or are other waves interfering with it?
• Estabrook Kind, 1996 Science

Lawrence Shearer, in press G3
13
PP-Precursors An Alternate Look
Lawrence Shearer, in press G3
14
PP-Precursors 1D Stack

• About 5-6 times the signal-to-noise ratio of the
  2D stacks because there are 20-40 times more
  waves in each stack.
• P660P and P520P do appear.
• The radial and vertical stacks are very similar!

P410P
P660P?
P520P?
Lawrence Shearer, in press G3
15
Topside Ppdp Reflections

• Pp660p is weaker than Pp410p, but it is much
  stronger than P660P.
• So why is the P660P so weak?
• What is different
• about the 660?

660
Lawrence Shearer, in press G3
16
Receiver Functions

• While P410s P660s are strong, where is P520s?
• If anything, P520s has a negative impulse.
• Why is the 520
• different from the
• 410 and 660?

Lawrence Shearer, in press G3
17
Modeling Method

• A linear inversion is problematic when fitting
  just one type of data.
• We use the Niching Genetic Algorithm.
• Solve for changes in
• P-velocity ?VP - 0-10
• S-velocity ?VS - 0-10
• density ?? - 0-10
• Interface Depth ?z - ?30 km
• Interface Thickness ?H - ?30 km
• Synthetic calculations with generalized ray
  theory using a priori pulse heterogeneity
  constraints.

Lawrence Shearer, in press G3
18
Modeling Each Waveform
Lawrence Shearer, in press G3
19
The Most Optimal Model
Lawrence Shearer, in press G3
20
Comparing Models

• PREF (blue) is a suite of seismic models
  calculated from mineral physics properties of a
  pyrolitic composition mantle. Camarano et al.,
  2005

• No 520-km discontinuity.
• 660-km discontinuity depth is deeper.
• There are a lot of assumptions that go into PREF
  (both seismic mineral physics).
• More similar than AK135 PREM.

21
Big Picture Results

• The 410 is a lot like we thought it was.
• The 520 is a discontinuity in density and VP, not
  VS.
• The 660 is much less significant discontinuity,
  and more of a gradient.
• If the interfaces have some finite thickness,
  then the 410 is 3X thicker than the 660.

• Previously studies, lacking observations of a
  positive P520s pulse wrongly concluded that the
  absence of the 520-km discontinuity.
• While the 660 likely impedes convection (to some
  extent), this effect is less as a gradient rather
  than a discontinuity.

22
Transition Zone Thickness Topography

• Topography of the 410 660 are anti-correlated
• Average thickness
• ? 241 km.
• Topography
• ? 20 km

Flanagan Shearer, 1998 JGR
23
Models agree at long wavelengths
GD02

• Degree-6 there is good agreement
• But fine structures are harder to get.

GD02
Gu Dziewonski, 2002 JGR
24
SS-Precursors v. Receiver Functions?

• Chevrot et al., 1999
• Average Thickness 252 km
• Thickness Variation ? 15 km
• Low correlation with SS-precursor studies.

25
Receiver Functions

• Receiver function stacks for 118 stations
• Mean thickness 247 km.
• Median thickness 246 km.
• Strong P410s P660s
• Most lack a P520s

Lawrence Shearer, 2006 JGR
26
Correcting Biases

• Bias 1 P Pds actually follow slightly
  different paths through the Earth
• While Chevrot et al., 1999 accounted for this
  during stacking, they did not correct for this
  when calculating depth.
• They simply corrected to a particular distance.
• 2-4 km overestimation in Chevrot et al., 1999.

Lawrence Shearer, 2006 JGR
27
Correcting Biases

• Bias 2 Stations are predominantly on continents,
  not oceans, but the Earth is 70 ocean.
• When we look at the long wavelength (harmonic
  degrees l lt 6) the Pds is very similar to SdS.
• Average Thickness 242 km
• Thickness variation 20 km.
• Correlation at R20.5

Lawrence Shearer, 2006 JGR
28
Higher Resolution

• Current stacking method requires large bin sizes.
• Short wavelength is smoothed over
• Amplitudes are less than they should be

29
Structure Depends on Stack
30
SS-Precursor Sensitivity Kernels
31
Discretization

• With larger blocks the pattern gets smeared out.
• Less X-shaped.
• More circular.

32
Adaptive Stacking

• SdS has a very small amplitude (often below the
  noise).
• Stack gt 100 traces to increase signal to noise
• Provides more reliable travel time.
• I also stack the sensitivity kernels.
• I then invert the stacked travel times and
  stacked sensitivity kernels for the true
  structure.

33
Sensitivity of a Stack

• The second problem for a stable inversion, we
  must have lots of data.
• Stack many times with different geometries
• 100-1000 waveforms per stack
• Bootstrap method (25 X) ensures stack stability
• Stack with variable sized bins

dt SdS-SdSAK135 travel time residual Kij
Sensitivity of jth wave to topography in the ith
block. ?zi topography of the ith node
60,000 stacks 2X108 non-zero Kji values.
34
Inversion
Transition Zone Thickness
rk ?10? Stacks

• Similar but different.
• Larger amplitude shorter wavelength features.
• Some anomalies moved or disappeared.
• Others appeared or strengthened.

Inverted Structure
35
Test the Model (1)

• Given our model (?zi) and sensitivity (Kji), can
  we reproduce our stacks?

Yes!
36
Test the Model (2)

• Given a checkerboard pattern (?zi) and the
  sensitivity (Kji), can we reproduce the
  checkerboard from theoretical stacks?

Yes!
37
Topography v. Seismic Velocity
38
What does this model show?

• Slabs?
• Expected from plate motion tomography.

200 km
200 km
1000 km
200 km
250 km
250 km
Lithgow-Bertelloni Richards, 1998 R. Geophys.
39
What does this model show?

• Slabs?
• Expected from plate motion tomography.
• Hotspots?
• From tomography convection modeling.

200 km
200 km
250 km
250 km
40
Conclusions

• Our views of the transition zone are still in
  flux.
• Even the average structure is taking on shape,
  which can be used to infer the mineral physics,
  geodynamic, and geochemical environment.
• By improving upon old techniques, we gain insight
  into the nature of the transition zone.
• Constraining the scale of transition zone
  thickness anomalies is crucial for understanding
  understanding how the transition zone interacts
  with slabs and plumes.
• Thin slabs equate to through-going anomalies.
• Broad anomalies equate to stagnant anomalies.