Geophysical Prospecting 32, 1-17, 1984.

SHEAR WAVES BY A N EXPLOSIVE POINT-SOURCE: THE EARTH SURFACE AS A GENERATOR OF CONVERTED P-S WAVES*
J. FERTIG**

ABSTRACT
FERTIG, 1984, Shear Waves by an Explosive Point-Source: The Earth Surface as a GenerJ. ator of Converted P-S Waves, Geophysical Prospecting 32, 1-17.
The most common source of seismic energy is an explosion at some depth in a borehole. The radiated waves are reflected not only at the subsurface layers but also at the free surface. The earth's surface acts as a generator of both P- and S-waves. If the source depth is much less than the dominant wavelength the reflected waves resemble closely the waves generated by a single force. Theoretical seismograms were computed with different methods to look for the relevance of the surface-reflected waves. The numerical experiments show reflected shear waves even for small shotpoint-receiver distances. Due to their polarization these waves can be detected most easily on in-line horizontal geophones. The existence of these waves was examined during a conventional survey in Northern Germany. Conventional data analysis shows a large variability in the up/u, ratio. The method used here produced a shear-wave section with a rather good signal-to-noise ratio down to 4 s S-wave reflection time.

, I

1. I N T R O D U C T I O N
The general formulation for the displacements in an elastic medium leads to very complicated formulae. The general solution of these equations is unknown. Only simplifications of the problem or approximations of the medium lead to equations which we can solve analytically, or at least approximately. We have the most complete knowledge for waves propagating in a homogeneous, infinite, isotropic, and ideally elastic medium. In a medium with so many contraints we have a wavepropagation of two independent kinds of waves. The most important one is the P-wave which is faster than the S-wave. Both propagation velocities depend on the elastic parameters of the medium. These velocities describe the kinematical aspects of wave-propagations. The causes of these waves can be described by single forces.

* Paper read at the 44th meeting of the European Association of Exploration Geophysicists, Cannes, June 1982, revision received May 1983. ** Preussag AG, Erdol und Erdgas, PO Box 4829,3000 Hannover 1, German F.R.
1

J. FERTIG

In relation to a specified coordinate system we may speak of vertical and/or horizontal forces. An explosive point source can be described by an equivalent bodyforce system. An actual description of the representation of seismic sources can be found in many monographs, e.g., Ben-Menahem and Singh (1981), Aki and Richards (1980) and Pilant (1979). Many point-source systems produce shear waves-at least in the nearfield. These waves decrease rapidly with increasing distance from source to geophone. Therefore special sources are used to generate shear waves. Current technology of shear-wave generation can be found in Helbig and Mesdag (1982). Most of these sources are confined to new processes for the generation of shear waves either by a single force or with a combination of single forces. Another method to find information about the shear-wave velocity is the use of converted waves. For some years we tried to use conventional shots in a deep borehole which produced mainly compressional waves (P-waves). We used the reflected S-waves from the deep layered structure as a secondary source of S-waves (Dohr and Janle 1980). The energy of these waves depends strongly on the angle of incidence at the deep interface. Therefore one needs large offset to observe these waves. These waves belong to rays which are asymmetrical even for a plane and horizontally layered subsurface. Therefore standard processing techniques like the CMP-technique cannot be used for processing these data. Although recording of these waves is rather easy-one only needs horizontal in-line geophones to record these wavesspecial identification methods are necessary to obtain information about the shearwave velocity U , . The identification was done with motion-product seismograms, the stacking velocities were found by a searching algorithm over variable v,/u,-ratio following an optimum stack for compressional waves as was shown in a paper presented in Venice 1981 by Fertig and Hentschke. In offshore surveys this method is the only one to generate and to observe S-waves: the S-wave is produced here by transmission through-and not by reflection at-the seabottom. On land, shear waves are generated not only at deep interfaces but also by reflection of compressional waves at the earths surface. The source considered below is an explosive point-source at some depth below the free surface. This surface acts as a strong interface and a perfect reflector. To satisfy the boundary conditions at the surface we must have reflected P-waves and S-waves. Due to the polarization of these S-waves it is sufficient to plant horizontal geophones parallel to the normal recording procedure. These geophones have to be sensitive to the in-line component of the shearwave movement as these waves have SV-polarization. This is in contrast to specific sources for S-waves that generally produce SH-waves polarized at right angles to the source-geophone line. The existence of secondary S-waves is studied with simple theoretical considerations and by synthetic seismograms computed with the reflectivity method. The efficiency of S-wave generation by reflection at the free surface is demonstrated with a field example from Northern Germany. The processing of these data is simple as common processing packages require at most a slight modification.

SHEAR WAVES

2. K I N E M A T IA N D D Y N A M I C C ASPECTS F REFLECTED A N D O PS - W A V E S R O M T H E FREE U R F A C EO R A P O I N T F S F SOURCE AT DEPTH

To get some insight into the reflection process of a compressional wave at the free surface it is useful to look at the rays, wavefronts, and amplitudes of the reflected waves. The kinematic properties of wave-propagation can be derived from the eikonal equation. The dynamic properties are related to displacement amplitudes. Thus, the existence of particular wavefronts and ray paths is part of the kinematic aspect of the problem. The dynamic aspect may solve the question whether a certain model of the subsurface is adequate for the displacement amplitudes observed at a given source-receiver distance for the specified source type.

To

i ,
I------ /
Q = point or line source

1
r

= movement of particles

Fig. 1. Wavefronts and rays of direct and reflected waves for a point source Q at depth h. Direct wave P, reflected P-wave pP, reflected S-wave pS.

J . FERTIG

2.1. Wavefronts and rays for a source in a homogeneous half-space

The main assumption throughout the paper is the existence of a point-source which produces only compressional waves. The results shown below are correct only in a homogeneous medium. This medium can be described by three elastic parameters : the velocity for compressional waves up, the velocity for shear-waves U,, and the density d. If the half-space consists of a smoothly varying medium we can still use the distinction between compressional and shear waves, at least for sufficiently high frequencies. The idealized circumstances are shown in fig. 1. The characteristics of the waves are described by methods of standard raytheory; Snells law for reflection and refraction is used. In addition to the direct P-wave we have to take into account reflected P-waves. Due to the assumption of an elastic medium these also exist reflected S-waves. The upper part of fig. 1 shows the wavefronts of the direct and reflected compressional waves. The direct P-wave and the reflected compressional wave belong to a spherical wavefront, whereas the reflected S-wave has a more ellipsoidal wavefront. At a point on the free surface ( P o ) all three wavefronts interfere. A detailed study of individual phases can be carried out for a point inside the half-space (Pl).The ray diagram gives a simple picture of the wave process at any point P , in the half-space. The movement of the particles can be separated into a vertical (V) and a horizontal (H) component. The particle movements are shown by arrows in fig. 1. At the free surface we have a coincidence of the H-component and the x-coordinate axis; therefore the H-component is sometimes called the X-component. At least in the far field of the source we may have longitudinally polarized P-waves, and S-waves which show a transversal polarization. By virtue of Snells law we can compute traveltimes for every ray in fig. 1. This step solves the kinematic part of our task. The rays and traveltimes of these waves belong to rays which follow Fermats principle of least time. The dynamical aspect-that is, the determination of amplitudes-still remains to be solved.

2.2. T h e dynamic behavior of rejlected compressional and shear waves from f the free surface o a source at depth
The amplitudes of the direct and reflected waves are determined by two factors: 1. the sperical spreading for circular or cylindrical wavefronts ; 2. the reflection of curved wavefronts at interfaces. The first part can be determined from the knowledge of the kinematic part of our problem. It mainly takes care of different ray paths. The calculation of the second part is more complicated. The physical behavior of a discontinuity is prescribed by boundary conditions. These conditions must be established by physical reasoning. In the case of a stress-free boundary it is sufficient to set all stresses to zero. For a curved wavefront the solution of this problem is not trivial. Since the beginning of this century we know representations of spherical waves as a superposition of plane or cylindrical waves. The Sommerfeld integral gives a superposition by cylindrical waves. A change of variables in the Sommerfeld integral

SHEAR WAVES

FREE URFQCE RPP

Fig. 2. Reflection coefficient (displacement)for an incident P-wave and a reflected P-wave. The angle of incidence varies between O" and 90". The numbers indicate different up/u, ratios.

allows one to see this integral also as a superposition of scaled homogeneous and inhomogeneous plane waves (Aki and Richards 1980). Therefore it is worth studying the reflection process for our special problem for plane waves and different angles of incidence in more detail. 2.2.1. The rejection of plane P-waves at a free surface. The conditions for a stress-free surface lead to a system of equations which can be solved easily. The reflection coefficients for an incident P-wave are denoted by R,, and Rps (for the reflected P-wave and for the reflected S-wave, respectively). These coefficients are displayed in figs 2 and 3 for different angles of incidence. The parameter in these figures is the velocity ratio vp/v,. The coefficients are shown for displacements and not for potentials as usual. The analytical expressions are:

R,, R
ps

= -

(1 - 2a' sin'

e,)2 - 4a3 sin'

A

8, cos e,(i - a2 sin'

ep)i/2
2

=--

1 4a2 sin 8, cos ,(i - 2a' sin' A

e,)
7

J . FERTIG

D
FREE URFRCE RP

Fig. 3. Same as fig. 2, but for the reflected S-wave.

where
a = v,/v,,

= (1

- 2a sin 8,)

+ 4a3 sin

8, cos ,(l - a sin ,)l/,

8, = angle of incidence for the P-wave.

The reflected S-wave shows a simple amplitude behavior for all vp/v, ratios. For some ratios and ranges of angles of incidence nearly all P-wave energy is transformed into S-wave energy. The reflection coefficient R,, always has the same sign and is real; therefore the reflected signal changes only its amplitude but not its phase. The coefficient R,, shows a much more complicated behavior but a detailed discussion of this coefficient will be omitted here. For more information see Pilant (1979). 2.2.2 The waveJield in the interior of the half-space. To evaluate the wavefield inside the half-space one could use numerical or analytical methods. An analytical method would be the exact ray-theory where we have to solve integrals for different

SHEAR WAVES

paths of integrations, taking into account saddle points, poles and branch cuts (Cerveny 1956). We took the numerical approach for solving this problem. The finite-difference technique was used to solve the problem for propagation of elastic waves generated by a line-source in inhomogeneous media. Special considerations were made for the approximation of the boundary conditions (Kummer 1983). These approximations lead to an accuracy of second order. The main advantage of this method is that we can compute seismograms at any place within the medium. The position of the line-source is arbitrary too. Disadvantages of this procedure are long computing times and exhaustive computer storage requirements. The parameters of the calculations by this method were as follows: P-wave velocity S-wave velocity Density Dominating wavelength Time increment Grid spacing Depth of line-source Depth of profile Geophone spacing
up = 2000 m/s, U, = 1155 m/s, d = 1 g/cm3, A = 120 m, At = 1.2 ms, Az = Ax = 3.5 m, h = 0.125 WL, Zd = 3 WL, Ag = 42 m.
V$U, =

a,

The profile is situated three wavelengths below the free surface. The first receiver lies directly below the source. The computations were done for vertical and horizontal components (figs 4 and 5). We used this configuration here to study the wavefield in the lower half-space and in order to be free of the influence of surface waves (Rayleigh waves). The direct P-wave cannot be separated in the time domain, but the superposition can be recognized by the change of the puise shape. The undistorted pulse can be seen at geophone position 11. In the neighborhood of the shotpoint we can see this pulse again but now differentiated. The identification of all individual arrivals is greatly facilitated by the use of geometrical ray theory (fig. 1), which predicts three body waves P, p P and pS for a buried point- or line-source. In addition to the expected wave types another wave denoted S* can be clearly seen. The subvertical S-arrival is also not expected by the standard ray theory. The S*-wave as well as the subvertical pS-wave were first shown at the EAEG Meeting in Venice 1981 by Hron and Mikhailenko. These waves were computed by another numerical method and are analyzed in a paper by Hron and Mikhailenko (1981). Both wave types belong to nongeometrical rays. The existence of these waves is strongly dependent on the depth of the point- or line-source. They can only be seen for source depths which are much less than one wavelength (A). The S*-wave shows a linear polarization which distinguishes it well from the Rayleigh wave. Both nongeometrical wave types are linked to the propagation of energy along the wavefronts. Mathematically this is equivalent to the concept of inhomogeneous waves characterized by the exponential amplitude decay along the wavefronts. These inhomogeneous waves are present in the mathematical representation of spherical waves in the form of the Sommerfeld integral. The S*-wave may be explained by the

J. F E R T I G

C !

200

'

2-COMP.

400

'

600
(

'

MSEC 1

BOO

'

1000

'

1200

Fig. 4. Synthetic seismograms (vertical component) computed by a finite-difference technique. The source is situated 0.125 wavelengths (A = 120 m) below the free surface. The seismograms are computed for a depth of 3 A and a geophone spacing of 42 m. The direct P-wave and the surface reflected pP-wave interfere. For increasing distance the pS-wave separates into pSand S*-wave (for other parameters see text).

excitation of an ordinary S-wave at the point O above the source (fig. 1). It occurs when the energy reaches the free surface after decaying along the wavefronts of inhomogeneous plane waves radiated by the source. This explains the observed exponential dependence of the S*-amplitude on the depth of the source. The numerical method used includes all homogeneous and inhomogeneous waves radiated by the source automatically. The reflectivity method used in the next section does not include all inhomogeneous waves.

SHEAR WAVES

Hor izontai component

3
L

5
6

7
8 9
10
11

1 2 13

11

l6
l7

7 8 l9 1
~

2o

21
22

Fig. 5. Same as for fig. 4, but for the horizontal component.

Figures 4 and 5 indicate that a separation of pS- and S*-waves is possible only at sufficiently large offsets.

2.3. The meaning of pS-waves in reflection seismology

Up to now we described the pS-reflection in its behavior in the half-space. Usually observations are made at the free surface or near it. Therefore it is important to study the behavior of this wave in reflection at deeper layers. To study these waves, synthetic seismograms were computed with the reflectivity method (Fuchs and

10

J . FERTIG

Mller 197 1). This method automatically includes all multiple reflections and wave conversions between P- and S-waves for a point-source at the free surface. Kind (1978) extended this method with regard to a buried point-force or an explosive point-source at any depth. The method produces seismograms in real amplitudes, takes into account spherical spreading and head waves. By a suitable choice of the parameters, surface waves can also be simulated. The method is based essentially on Sommerfeld's integral in combination with a modified Haskell matrix multiplication of " layer-matrices ". The velocity-depth function is a first-order discontinuity at a depth of 1000 m with a high velocity contrast. The explosive point-source lies at a depth of 18 m (fig. 6); the shear-wave velocity is U, = vJ(,,h). To keep the necessary integrations stable, an artificial absorption with a Q of 200 throughout the whole model was used. Figures 7 and 8 show the results for the vertical (2)and horizontal component ( X ) . The most important phases are marked. The reflection from the interface is marked by Pl P , , which includes the pP, P,-wave. These two waves interfere and are only detectable through a change in the shape of the pulse. The assumed velocity model allows the generation of different head waves which also cannot be explained by plane-wave theory. In the chosen distance range we only observe the P-head-wave
h = 18m

vP= v*w S

o.

o.
r\

P "

YlnO
v

1.5

Fig. 6. Simple velocity-depth function. The source lies at a depth of 18 m. The up/us ratio is

fi.

SHEAR WAVES

11

o.

-+

--

I , ,
( <

4n
<

5. O

I -

0:o

,'O 1.'5 2.'0 DI TQNCE ( KM 1

2.' 5

Fig. 7. Synthetic seismograms (vertical component) for the velocity depth function of fig. 6, computed by the reflectivity method. The most important phases are: direct P-wave P, Rayleigh-wave R, reflected P-wave P, P, , converted P, S I wave, and the surface-generated pS, SI-wave.

from the lower half-space ( P , P , Pl). The PS-reflection from the interface is marked as Pl S, and is accompanied by the P , P , S,-head-wave. The following wavegroup pS, SI can be recognized clearly on the vertical as well as the horizontal component panel. The clear appearance of this wave on the vertical component is mainly due to the range of angle of incidence at the free surface. The surface multiple of the P-wave can be seen at a reflection time of 2 s. The X-component is nearly undisturbed in this time interval.

3.

PRACTICAL

R E S U L T SI N

THE O B S E R V A T I O N S O F P S - W A V E

3.1. Description offield parameters

Parallel to the numerical computations we have run a field experiment in Northern Germany in a structurally simple area. The vertical components were recorded with 120 traces in a split-spread configuration with a geophone spacing of 50 m. The maximum coverage was 30-fold. Close by a second cable with 60 horizontal geophones was laid out. The spread configuration was an in-line offset spread with an

12

J . FERTIG

4n01
50+ .-

o. O

O, 5

D I STRNCE

1. O

1, 3

2.0 KM 1

Fig. 8. Same as fig. 7, but for the horizontal component.

initial offset of 100 m. The coverage was 15-fold at maximum. Each geophone group consisted of six single 7 Hz geophones with a spacing of 1 m. Experience with earlier experiments indicated that a large pattern would cause difficulties due to large infra-group static corrections caused by lenses of sand and mud near the surface. Shots were fired at a depth of 18 m and consisted of a small pattern with three holes and a total charge of less than 1.5 kg. We believe that this configuration models a buried point-source rather well.

3.2. Results of conventional data processing of the vertical and horizontal component
Figure 9 gives an impression of the data quality of the individual recordings. For comparison a corresponding record of the vertical component (2)is also shown. The horizontal in-line component ( X ) shows some groups of reflection which are characterized by a relatively low mean frequency and a large moveout. These are optical indicators of another wave type. The large moveout means low velocity and the low mean-frequency might be the influence of anelasticity. The strong surface waves are not suppressed due to our small geophone pattern.

SHEAR WAVES am
0.200

13

a~)

am
a loo
a200

axa
040 .0

a m a0.500

a500 a600

a m
0.830

a0 60

a m
am

am

1 . m
1 im .

a=
1 .

1 . m
loo
1.200
1 xa . 1.400 I. 500 1.600 1 . 1.830 1.900

1.200
1.10 1 . 1.500 1.600 1 . 1 830 . 1m .

2 z
2.200

t z
20 20 2.400
2.500

axa
2.500

a 4m

axa
a0 60 2.700 a830
2.900

2.600 2m . 2m .

3z ;
3,200

am

3 . m
3 loo . 3,200

3,xa
30 40 3,500 3.800
3,700

3.xa
3,400

3500 3,600
3m . P m 3 . -

cz
4.1200 4.30

380 .3 3,900

E :?

4m .

4.m
4.500 4 . m

4.300
4a .

4.500 4.600
4m . 4 . m

4.m
4m .

am

4.900

am

4.900

Fig. 9. Field records for the vertical ( Z ) and horizontal ( X ) component at the same position.

The vertical component was processed in a conventional way without any difficulties. To confirm our observations again we computed synthetic seismograms for a simplified velocity model for this region. We used the reflectivity method again. The velocity-depth model was :
Layer 1 2 3

Thickness (m) 354 520 450

up (ml4

1600 2000 3200 4100

The explosive point-source was assumed and constant density maintained at a depth of 18 m. For simplicity we used a v p / v , ratio of 2 throughout the model. Figure 10 gives a comparison between theoretical and practical results. The individual pS-reflections are marked by A, B and C. The strong reflections in between are converted waves (PS) and multiple reflections. The practical and theoretical records show nearly the same amplitude behavior over the distance range used. A detailed

14

J. FERTIG

PRAXIS

THEORIE
4 .

0.ana "
a m

Q " ana
a m a m
a m
i.im
am am am am
L2m L m
I03 I5m

a m a03
am am am am

1 . m Iim
IXa
I .

. 1.o00

IQ0

Im
1.600 I. 7m
1 . m

L m
I7m

Im

%? ana
a m a03
L W

. 2o00
aim am

IBm Im

a m
a m

a 0
Zm i

am

am am am

a m
L m

am

3;"

. LOO0
3 im 3200 3m i

3na

n m

303

303
3 . m

3m
3600

30 60

37m

a m
Lao

am
3m

%?
4x0

. Go00 im
4.

3 m

40x1

43m
4 . a

4m .
4.-

4.553 4.600

4m .
4 . m

4.m
4m
4m

4m
4m

Loo0

- !%O00
2.' O 2.'5

4m .

2.5

0.5

0.'5

DI TQNCI

1' .0

Fig. 10. Comparison between theoretical and real seismograms for a simple three-layer model (see text). The figure shows the horizontal component. The surface-generated S-waves which are reflected by the individual boundaries are marked by A, B and C.

comparison shows that our assumption of a up/u, ratio of 2 is not correct. The recorded section indicates longer traveltimes for the S-wave which means a smaller velocity U,. The processing of the horizontal component was also done in a conventional way. Due to the low stacking velocities for the upper part of the seismograms, strong muting was necessary. This reduced our stacking factor considerably. Therefore the stacking results are not quite so reliable in the first second. The static corrections-the most difficult part of our processing routine-were taken from the vertical component and multiplied by a factor of 3. This is a very crude procedure and should only be seen as a first attempt to overcome this problem. The results of our processing of the individual components are shown in fig. 11. This part of the profile covers a range of 5km. To make comparison easier we halved the time-scale for the horizontal component. We believe that there is a fairly good correlation between the two components. The easiest comparison is possible

SHEAR WAVES

15

Vertical component

Fig. 11. Comparison of the vertical and horizontal sections. The time-scale is halved for the horizontal component. Letters indicate a correlation of individual reflectors.

16

J. F E R T I G

for the reflectors B, C and K. The tertiary layers do not compare well; this might be caused by the poor stacking quality we have in this time range. The topmost reflections might also be influenced by surface waves. Nevertheless, we believe that we can improve our results when we get better static corrections.

4. CONCLUSION
Supported by simple theoretical considerations and theoretical seismograms the relevance of a shear wave generated by the reflection of a compressional wave at the free surface is supported. The existence of such a wave is a simple means to generate S-waves. These waves can be observed as a by-product with normal recordings-no extra source is needed. We only need a second cable with in-line-horizontal geophones. The greatest benefit is to be expected from direct combination of shearwave observations with compressional wave observations. Many physical phenomena which affect the vp/v, ratio can be studied. With better static corrections we believe that we can improve our results. This will also lead to more reliable determinations of the vp/v, ratio derived from traveltime measurements. The question whether this wave is the so-called S*-wave cannot be answered. The distance range used might not be sufficient to separate pS- and S*-waves reflected from the subsurface.

ACKNOWLEDGMENTS
This research was sponsored by the Bundesministerium fr Forschung und Technologie of the FRG under contracts ET 3062-A and ET 3104-A. I would like to thank Dr Dohr for his advice and support and C . Lange for his expert processing. The manuscript was typed by Mrs R. Mller.

REFERENCES
AKI, K. and RICHARDS, P.G. 1980, Quantitative Seismology, Theory and Methods, Vol. 1, W.H. Freeman, San Francisco. BEN-MENAHEM,and SINGH, 1981, Seismic Waves and Sources, Springer-Verlag, Berlin, A. S.J. Heidelberg, New York. CERVENY, 1956, The reflection of spherical elastic waves at a plane boundary, Travaux de V. l'Institut Geophysique del'Academie Tchecoslovaque des Sciences 44,343-365. DOHR, and JANLE, 1980, Improvements in the observation of shear waves, Geophysical G. H. Prospecting 28,208-220. FUCHS, and MLLER, 1971, Computation of synthetic seismograms with the reflectivity K. G. method and comparison with observations, Geophysical Journal of the Royal Astronomical Society 23,417433. HELBIG, and MESDAG, K. C.S. 1982, The potential of shear-wave observations, Geophysical Prospecting 30,413-431. HRON,F. and MIKHAILENKO,1981, Numerical modeling of nongeometrical effects by the B. Alekseev-Mikhailenko method, Bulletin of the Seismological Society of America 71, 1011-1029.

SHEAR WAVES

17

KIND, R. 1978, The reflectivity method for a buried source, Journal of Geophysics 44, 603612. KUMMER, 1983, Differenzenapproximationen des elastischen Wellenfeldes in isotropen, linB. earen Medien mit geneigten Grenzflachen, Dissertation, Universitat Hamburg. PILANT, 1979, Elastic Waves in the Earth, Elsevier, Amsterdam. W.L.