Data from 90 permanent broad-band stations spread over central and eastern Europe were analysed using Ps receiver functions to study the crustal and upper-mantle structure down to the mantle transition zone. Receiver functions provide valuable information on structural features, which are important for the resolution of European lithospheric dynamics. Moho depths vary from less than 25 km in extensional areas in central Europe to more than 50 km at stations in eastern Europe (Craton) and beneath the Alpine–Carpathian belt. A very shallow Moho depth can be observed at stations in the Upper Rhine Graben area (ca. 25 km), whereas, for example, stations in the SW Bohemian Massif show a significantly deeper Moho interface at a depth of 38 km. Vp/Vs ratios vary between 1.60 and 1.96, and show no clear correlation to the major tectonic units, thus probably representing local variations in crustal composition. Delayed arrivals of converted phases from the mantle transition zone are observed at many stations in central Europe, whereas stations in the cratonic area show earlier arrivals compared with those calculated from the IASP91 Earth reference model. Differential delay times between the P410s and P660s phases indicate a thickened mantle transition zone beneath the eastern Alps, the Carpathians and the northern Balkan peninsula, whereas the transition zone thickness in eastern and central Europe agrees with the IASP91 value. The thickening of the mantle transition zone beneath the eastern Alps and the Carpathians could be caused by cold, deeply subducted oceanic slabs.
Knowledge of the deep seismic and thermal structure of the earth's crust, together with its composition as far down as the upper mantle and the mantle transition zone, is important in understanding the geodynamic history of Europe. In surface geology, there exist huge differences between the Pre-Cambrian East European Craton (EEC), the Palaeozoic platform of central Europe, and the recently active collision zone of the Alpine–Carpathian orogenic system (e.g. Banka et al. 2002; Fig. 1). However, it is still unclear how deeply the origins of these different tectonic units and related processes penetrate the upper mantle.
The EEC, finally consolidated in the late Precambrian, is divided from the rest of Europe by the Transeuropean Suture Zone (TESZ), which runs from the North Sea in the northwest to the Black Sea in the southeast. The TESZ is commonly split into the Tornquist—Teisseyre (TTZ) and Sorgenfrei–Tornquist zones (STZ). The Thor Suture is generally interpreted as a thrust of Caledonian nappes on Baltica (EEC). The basement of central Europe was mainly consolidated after the Caledonian (early Palaeozoic) and Variscan (Late Palaeozoic) orogenies. During the Mesozoic, central Europe was further consolidated and covered by platform sediments. Its southern periphery was influenced by the rifting and opening of the Tethys Ocean. During the late Mesozoic and Cainozoic, the study area southwest of the TESZ was influenced strongly by the Alpine orogeny and rifting in the North Atlantic. The Alpine–Carpathian belt is a result of the Tertiary African–European convergence and of the closure of the Tethys ocean, associated with the collision of several microplates with the European plate (e.g. Nemcok et al. 1998; Sperner et al. 2001; Seghedi et al. 2005). Within the central Alps (Austrian, Swiss), subduction of the central European lithosphere is observed towards the south (e.g. TRANSALP Working Group 2002; Lueschen et al. 2006). In the Carpathians, subduction is (was) mostly directed towards the Pannonian basin (e.g. Sperner et al. 2001).
The palaeotemperature distribution and composition of the deep lithosphere can be constrained by xenolith studies and exhumed slivers of mantle rocks. Unfortunately, outcrops of these rocks are scarce, both in time and space. They provide information only very locally, whereas geophysical methods can study the recent temperature distribution and compositional variations over broader areas. As identified in previous studies, there exist large differences in several geophysical fields between central and eastern Europe across the TESZ. Gravity studies first revealed the strong differences between the old core of Europe (EEC) and the Palaeozoic to recent mobile belts in the SW (Caledonian, Variscan and Alpine belts) (Tornquist 1908; Yegorowa & Starostenko 1999). Artemieva (2003) found that the uppermost mantle beneath the EEC is up to 1.6 per cent less dense than it is beneath central Europe. Strong anomalies in the distribution of seismic velocities were found by tomography and surface wave studies (e.g. Zielhuis & Nolet 1994; Meier et al. 1997; Ritzwoller et al. 2002; Shapiro & Ritzwoller 2002). Goes et al. (2000) calculated temperature differences of up to 300 K in the uppermost mantle using seismic velocities, under the assumption that these velocities depend more on temperature than on compositional variations. Using traveltime tomography, Wortel & Spakman (2000), Piromallo et al. (2001) and Piromallo & Facenna (2004) found evidence for high P-wave velocities in the mantle transition zone at depths from 500 to 600 km beneath the Alpine–Carpathian belt and the Mediterranean Sea. These high seismic velocities were interpreted as being caused by cold, deeply subducted material.
One possible method for studying temperature variations at the bottom of the upper mantle is to analyse the variations in the seismic discontinuities of the mantle transition zone at depths of about 410 and 660 km. It is now widely accepted that most of the discontinuities in velocity and density at these depths are the result of phase changes in olivine and other minerals (e.g. Vacher et al. 1998; Helffrich & Wood 2001; Lebedev et al. 2002, 2003). The transformation of the olivine component has a major impact on the seismic velocities observed. Olivine is transformed to a spinel structure at a depth of about 410 km and finally breaks down to perovskite and magnesiowüstite at about 660 km (e.g. Helffrich & Wood 2001). Both transformations are temperature-dependent but have opposing Clapeyron slopes (e.g. Fei et al. 2004; Katsura et al. 2004). If the temperature is increased (e.g. in a plume environment), then the transformation normally observed at 410 km should be found at a deeper level, and the one observed at 660 km should be found at a shallower level. The opposite holds true for a temperature decrease as it can be expected in subduction environments.
SS precursor and the Ps receiver function studies are the most common techniques used to study discontinuities of the mantle transition zone (Lawrence & Shearer 2006). Using a SS precursor study, Gossler & Kind (1996) found a transition zone that was 14 km thicker beneath continents than beneath oceans. According to Shearer (2000), the topography of the upper-mantle seismic discontinuities at 410 and the 660 km seem to be largely unrelated at a global scale, with the 660 topography exhibiting significantly larger variations in peak-to-peak amplitude than the 410 topography. However, Li et al. (2003a) and Chevrot et al. (2000) found that the delay times of Ps-converted 410 and 660 phases are generally quite well correlated, which implies that most of the variations in delay time can be attributed to variations in upper-mantle seismic velocity. Assuming a more or less homogenous transition zone, its thickness dTZ can be determined more reliably than the individual depths of the 410 and 660 km discontinuities. The sharpness of the 410 and 660 km discontinuities could be 5 km or less as estimated from short period P′P′ precursors (Benz & Vidale 1993). There is still some disagreement over the existence of a detectable seismic discontinuity (enhanced velocity gradient) at a depth of about 520 km (Shearer 1990).
In this paper, we apply the Ps receiver function method to study the upper mantle down to the mantle transition zone at the contact of the Precambrian lithosphere in eastern Europe and the Palaeozoic-to-recently consolidated central Europe, including large parts of the Alpine–Carpathian orogenic belt.
The Ps receiver function method is a widely accepted approach for studying lithospheric and upper-mantle seismic discontinuities. Our analysis of teleseismic broad-band recordings follows the work of Vinnik (1977), Kind et al. (1995, 2000) and Yuan et al. (1997). Rotation from the Z, N–S and E–W (ZNE) components into the P, SV and SH system (LQT components) is used to separate different wave types. For the rotation of the horizontal components, in most cases we used theoretical values of backazimuth calculated from the coordinates of the event and the station. The values thus obtained commonly differ by less than 10 degrees from the observed ones. However, at stations in orogens, such as the Alps and Carpathians, the backazimuth deviations are significantly larger. In such cases, the backazimuth was estimated from polarization analysis of the horizontal components. Generally, the angles of incidence were determined by minimizing the energy on the SV component (Q) at the time of the P signal. This was achieved by computing the eigenvalues of the covariance matrix of the P signal, following Kind et al. (1995). We used a time-domain deconvolution method to remove the source signal and source-side reverberations from the records, in order to allow records from different events to be stacked. Amplitudes of the SV and SH components were normalized in relation to the incident P wave. Arrival times were measured at the maximum of the deconvolved P-wave signal. For moveout correction, the IASP91 Earth reference model (Kennett 1991) was used to reduce the time scale of records at any distance to the fixed reference epicentral distance of 67° (slowness: 6.4 s deg−1). This means stretching (or compressing) waveforms for events at larger (or shorter) distances, respectively. In this way, primary P-to-S converted phases are aligned parallel to the P phase and multiple phases can be identified by their slowness difference (not parallel to the P phase). Moho depth and average crustal Vp/Vs ratios were analysed using crustal multiples (Zhu & Kanamori 2000).
In order to study the effect of lithospheric structure on the delay times of P410s and P660s, we analysed synthetic seismograms in the same way as the real data. The reflectivity method was used to compute the theoretical seismograms (Kind 1985).
We analysed teleseismic data from 90 permanent broad-band stations belonging to various seismic networks in central and eastern Europe (Fig. 2, Table 1). Most of the stations are equipped with STS-1 or STS-2 seismometers. Generally, we analysed hundreds of events at single stations, but for some stations less than 50 events could be used due to different reasons. At the stations of the Gräfenberg array (GRA1, GRB1 and GRC1) we analysed more than 1000 events from 25 yr. Most data were filtered using a third-order Butterworth high-pass filter with a corner period of 50 s. At some stations in coastal regions and in sedimentary basins, we filtered the data using a corner period of 30 s. Furthermore, we performed filter tests to enhance the signals from the mantle transition zone. Good results were obtained using a 3 s low-pass filter. Only stations where similar arrival times were obtained using different filters were assumed to show reliable Ps conversions from the mantle transition zone. For the final determination of Ps conversion delay times, we used the high-pass filtered data. We also tested P410s and P660s arrivals for subgroups of events with different levels of signal-to-noise ratios of the Moho signal relative to the noise in front of zero delay time (the P arrival time). Only arrivals with consistent signals were used.
Figs 3–5 show stacks of moveout-corrected receiver functions. Strong converted phases within the first 20–30 s delay time are caused by the sedimentary cover (especially in the North German–Polish Basin, beneath the Gräfenberg Array, or the East European Platform), discontinuities in the crystalline upper crust or the Moho and its multiple phases. The quality of the observations of the conversions from the mantle transition zone is influenced by higher order multiples and worse recording conditions in the sedimentary basins, as well as a complicated lithospheric structure beneath the active orogens. However, the stations on crystalline rocks in consolidated areas show excellent converted phases in most cases.
Generally, the most pronounced phase is the conversion at the Moho, which occurs at a delay time of between <3 and >6 s (Table 1, Figs 5, 6). Amplitudes of the Moho Ps conversions correspond to approximately 10 per cent of the amplitude of the primary P phase (see Table 1). At some stations (e.g. NRDL, BRNL and HAM), the strongest conversion occurs at the base of the sedimentary basin. This masks the converted energy from the crust–mantle boundary. The Moho Ps delay times (Moho depths) agree to a large extent with previous regional receiver function studies as well as with results of regional deep seismic transects. A very shallow Moho (less than 3 s Ps delay time) can be observed in areas under extension, such as the Upper Rhine Graben area (ECH, BFO, STU and TNS) or the Pannonian Basin (PKSM). Late Moho Ps arrivals (up to more than 6 s Ps delay time) can be observed in the Alps (e.g. BNI, WTTA), the Carpathians (MLR) and the EEC (e.g. SUW, TRTE/VSU and PUL); see Fig. 6. Apart from the direct conversion from the crust–mantle boundary, strong multiples can also be observed, which have been used to analyse Moho depths and crustal Vp/Vs ratios (Figs 17 and 18) using the approach of Zhu & Kanamori (2000). The uncertainties of the crustal parameters are about ±0.1 s for Moho Ps delay time, less than ±2 km for crustal thickness and ±0.05 for crustal Vp/Vs ratio (see Geissler et al. 2005).
Mantle transition zone
An approach widely used for studying the depth of the seismic discontinuities in the upper mantle is the common station stacking of the moveout-corrected single traces from all backazimuths. Since the stations analysed are not uniformly distributed within the study area, we used this approach rather than common conversion point stacking, because the latter approach is mainly used for densely spaced temporary networks. The converted phases from the 410 and 660 km discontinuities can be observed at most stations at delay times of about 42–46 and 66–70 s, respectively (Table 1). Their amplitudes are about four times lower than those from the Moho conversions (about 2–4 per cent of the incoming P wave). Generally, no direct dependence of the delay times of the mantle transition zone on the Moho Ps delay time is observed (see Fig. 5). Delayed arrival times of the 410 and 660 km discontinuities of up to 2 s (in comparison to theoretical delay times calculated from IASP91 Earth reference model) are observed in central Europe, for instance, at stations WLF, TNS, GRA1 and NOTT, as well as at several stations in the western Alps. In contrast, all stations in eastern Europe (on the EEC) show earlier arrivals, both from the 410 and 660 km discontinuities (Figs 9 and 10). The differential travel times for the transition zone (P660s–P410s) are about 23.8 ± 0.5 s for most of the stations in central and eastern Europe, as predicted by the IASP91 reference model (Figs 10 and 11). However, in the eastern Alps and the Carpathian–Pannonian region, the MTZ measured by receiver functions seems to be thicker than normal (>25 s differential delay time).
As observed in previous studies (e.g. Kind & Vinnik 1988), the conversion amplitudes and delay times of the 410 km discontinuity may vary with azimuth and frequency. The same is true for the 660 km discontinuity. However, in our stack traces, the 660 km discontinuity is clearer than that of the 410 at most of the stations. The most pronounced conversions from the MTZ can be observed at stations on the EEC.
Phases around 30 and 54 s Ps delay time
At many stations, additional converted phases can be observed at Ps delay times of around 30 and 54 s (for instance, stations MOX, RGN, KWP, MLR, BFO and ARSA). By investigating the distance (slowness) dependence of these phases, we found that in most cases, these phases are multiples of the lithospheric structure; hence, they most probably represent multipath crustal reverberations. This is also indicated by results obtained from analysis of the synthetic receiver functions (next section). These second-order multiples are an indication of a very strong velocity difference at the Moho and minor scattering/damping of the long-period waves within the crust. This agrees with the already very strong ‘primary’ multiples observed at these stations.
Synthetic receiver functions
In order to study the cause of the delayed or earlier arrivals from the MTZ converted phases at different stations, we computed receiver functions from synthetic seismograms (Figs 11 and 12) using the reflectivity method (Kind 1985). We used models of seismic velocity that reproduce the differences in crustal structure between central and eastern Europe and tried to vary the seismic velocities in the upper 200 km of the mantle, where significant differences have previously been shown to exist between eastern and central Europe. As can be seen in Fig. 10, the first 20–30 s are dominated by crustal primary conversions and reverberations. It is for the BM3A model (BM; Bohemian Massif), where sharp boundaries of a low velocity zone (LVZ) in the upper-mantle (asthenosphere) exist, that clear conversions can be observed from these discontinuities, at about 9 and 24 s delay time. Smoothing the LVZ boundaries (model BM3B, Fig. 12) makes the converted phases less significant. In noisy data or at stations with more complicated crustal structure, these phases might not be observable in Ps receiver function data. Clear arrivals can be modelled for both the 410 and 660 km discontinuities. As observed in several tests, the arrival time depends more on the Vp/Vs ratio than on the absolute velocities in the upper mantle. Model EEC4D (EEC) causes earlier Ps phases from the MTZ, which occur at 42.7 and 66.6 s (dt = 23.9 s). From the IASP91 reference model, we obtained arrival times of 44.2 and 68.0 s (dt = 23.8 s). From models BM3A and BM3B, we obtained values of 45.4/45.1 s and 69.2/68.9 s (dt = 23.8/23.8 s), as summarized in Table 2. Models with lower or higher absolute seismic velocities in the upper mantle but with the same Vp/Vs ratio as in the IASP91 model show almost no deviation of the MTZ arrival times from the predicted values (BM1A: 44.3/68.1 − dt = 23.8 s; EEC4C: 44.1/67.9 − dt = 23.8 s).
Crustal parameters: Moho depth and Vp/Vs ratio
Crustal thickness in central and eastern Europe has previously been investigated mainly by active seismic studies (both steep and wide-angle), the results of which show that the typical thickness beneath the Palaeozoic platform is about 30 km. This value can reach up to 50 km beneath the EEC, and an even deeper crust-mantle boundary has been observed beneath parts of the Alpine–Carpathian orogenic belt (e.g. Giese 1995; Grad et al. 2002; Dezes et al. 2004).
The Moho depths obtained in this study (Table 1, Fig. 8a) agree very well with previous results obtained from active-source studies and reflect the location of stations within different tectonic provinces (Pre-Cambrian craton, Palaeozoic massifs and Cainozoic areas under extension or compression). In addition, the Moho pseudo depths obtained from Ps delay times agree quite well with the Moho depths obtained by the Zhu and Kanamori method (compare Figs 6 and 8a). However, no clear dependence of the calculated Vp/Vs ratio on the major tectonic regime exists. Even if Vp/Vs ratios do not exhibit differences between the major tectonic units, there seem to be consistent areas within these units that show similar Vp/Vs ratios. Increased Vp/Vs ratios can be observed at stations in northern Germany, southern Sweden/eastern Denmark, in the Sudeten mountains at the Czech–Polish border and at stations in the Alpine–Carpathian orogenic belt. Stations with a lower than average Vp/Vs ratio (red in Fig. 8b) are more scattered, showing no consistent spatial distribution. Some stations show anomalous low or high Vp/Vs ratios, which might not be realistic. This might then also apply for the corresponding Moho depth values. It is probable that the very complicated low values observed are caused by complicated crustal structure (wave paths, for example, caused by a high-velocity lower crustal layer, as may be the case for stations MHV and OBN) and therefore, interference of intracrustal multiples with direct Moho conversions or Moho topography, as discussed by Geissler et al. (2005) and Zandt et al. (1995). From our data we conclude that the Vp/Vs ratio depends more on local crustal structure and composition than on the age of the tectonic unit.
Upper mantle: apparent topography and depths of the 410 and 660
Previous studies of the seismic discontinuities of the mantle transition zone beneath central Europe were carried out with data from permanent stations (Kind & Vinnik 1988; Stammler et al. 1992; Grunewald et al. 2001) as well as from temporary experiments (e.g. TOR & SVEKALAPKO –Alinaghi et al. 2003; EIFELPLUME –Budweg et al. 2006). Grunewald et al. (2001) investigated the topography of these discontinuities in Germany using data from the German Regional Seismic Network (GRSN) and the Gräfenberg Array (GRF). We extend this work by including more data from German stations and adding data from other permanent broad-band stations in central and eastern Europe (Fig. 2).
As mentioned above, apart from the Moho depths, the most striking difference between eastern and central Europe can be observed in the absolute Ps delay times of the mantle transition zone discontinuities. In eastern Europe, they arrive up to 2 s earlier than predicted by the IASP91 Earth reference model, whereas in central Europe they show a delay of up to 2 s (Figs 10a and b).
Although there appears to be sufficient data available to study the MTZ in the Alpine–Carpathian region, its complicated lithospheric structure hampers easy identification of the converted phases from the MTZ. At some stations, multiple phases seem to mask the MTZ conversions. However, conversions from the 660 km discontinuity can be identified clearly at most stations. These are delayed by up to 2 s in the eastern Alps/Carpathian region, whereas the P410s conversions identified do not show such dramatic delays. This results in an MTZ that is approximately 20 km thicker (up to 2 s differential delay time) beneath the Carpathian–Pannonian region (e.g. PKSM, VTS, VYHS and KWP) compared to that found in central and eastern Europe.
Fig. 13(b) shows the differential delay time plotted against the P410s delay time. The dashed line shows the expected correlation by Clapeyron slopes of +3 MPa K−1 for the 410 and −2 MPa K−1 for the 660 km discontinuity (see Li et al. 2003b). An increase in the 410 time for a constant differential time (the area to the right-hand side of the dashed line) can be explained by a decrease in the upper-mantle velocities, which is in most cases accompanied by an increase in the Vp/Vs ratio in the upper mantle. The converse is true for the area to the left-hand side of the dashed line. It is clear that stations on the EEC are clustered in a small area to the left-hand side of the expected line of correlation. Data from stations in central Europe (Palaeozoic) show a greater degree of scatter and almost all fall to the right-hand side of the correlation line in the plot. This could indicate higher Vp/Vs ratios on the travel path, because differences in delay times seem to depend to a large extent on the Vp/Vs ratio of the mantle and crust rather than on the absolute velocities as discussed previously. The general trend in the diagram (plotting along tdiff = 24 s) confirms results from the global study of Chevrot et al. (1999) and indicate that delays in the P410s (and P660s) are caused mainly by the velocity (Vp/Vs) structure in the upper mantle.
Significant variations exist between delay times of the 410 km (and 660 km) discontinuity recorded at different stations only some tens of kilometres apart, as for example in the case of the western Bohemian Massif (e.g. stations GRA1, GRB1, GRC1 and NOTT). We therefore, conclude that the main cause of the apparent topography of the 410 km discontinuity might be located in the uppermost mantle very close to the individual stations. Although data from stations in the Alpine–Carpathian belt are more complicated, it may be seen that some stations are plotted at differential delay times >25 s.
Mantle transition zone: thickened MTZ beneath SE Europe
As stated above, we observed significant variations in transition zone thickness across central and eastern Europe. Beneath central Europe north of the Alps and the EEC, the transition zone thickness is close to the value found in the IASP91 global reference model. Beneath the eastern Alps, the Pannonian Basin and the Carpathians, the transition zone seems to be more than 20 km (up to 2 s differential delay time) thicker than the IASP91 reference value, whereas beneath the western Alps, the extent of the thinning and thickening appears to be mixed. According to Bina & Helffrich (1994), variations in thickness of ±20 km could be explained as variations in temperature of ±150–200 K within the MTZ. This would imply at least a 150 K cooler mantle transition zone beneath the Carpathian–Pannonian region. The slightly reduced differential delay times (compared to the IASP91 model) of more than 1 s less at stations BNI, TUE, VDL and MUO in the western Alps would translate into an approximately 75 K hotter MTZ in that area. North of the Alpine–Carpathian belt, the temperature variations seem to be negligible. In particular, there seems to be no difference between the temperatures of the mantle transition zone beneath the Variscan consolidated central European platform and the Pre-Cambrian EEC. If these estimates of temperature variation are correct, this results in lateral temperature variations in the mantle transition zone beneath central and eastern Europe of about 225 K.
The observed thickening of the MTZ beneath the eastern Alps and the Carpathians may be explained by the model of, for example, Piromallo et al. (2001) and Piromallo & Facenna (2004), who postulate old subducted (cold) slabs in the MTZ beneath large parts of southeastern Europe. The strongest indication for thick accumulated cold material (former oceanic slabs) can be found in the area marked by Piromallo & Facenna (2004; Fig. 1) with ‘Carpathian’. In this region, very high P-wave velocities are mapped at depths of 650 km. The authors also find similarly high velocities beneath the (western) ‘Alps’, where we observe enlarged differential delay times only at stations at the western rim, with stations within the western Alps showing reduced values. This could point to a more complicated structure of the mantle transition zone in this area than previously described. The thickening of the MTZ beneath the eastern Alps, the Carpathians and the Pannonian Basin is comparable to SW Japan, where a thickening of the MTZ of about 30 km was observed by Tonegawa et al. (2006). It may well be the case that a large flat (cold) slab rests within the mantle transition zone, showing high seismic velocities and causing a thickening of the MTZ. Surprisingly, such a thickening seems not to exist beneath Tibet, where the thickness of the mantle transition zone is close to the IASP91 values (Yuan et al. 1997; Kind et al. 2002). It may be assumed that in the case of SW Japan and southeastern Europe, the subduction processes reach into the mantle transition zone, whereas in Tibet, processes related to continental collision seem to stop above the 410 km discontinuity.
According to Swieczak et al. (2004), the thickness of the mantle transition zone beneath eastern Europe varies from 240 to 280 km. Swieczak et al. assumed a thicker MTZ beneath the Pre-Cambrian platform (cold lithosphere), whereas the MTZ should be thinner beneath the Carpathians (hot lithosphere). Our results do not confirm the assumptions regarding the MTZ thickness made by Swieczak et al. (2004). They agree more with results from North America instead, which also fail to show any imprint of the cold lithosphere of the continental keel on the mantle transition zone thickness (Li et al. 1998) and thus indicate a decoupling of the EEC lithosphere (tectosphere) from the MTZ. Ramesh et al. (2002) observed a depression of the 410 km discontinuity by about 2 s in the western USA, which confirms a seismically slow upper mantle and probably has no imprint on the MTZ.
Apparent local deepening of MTZ discontinuities is also indicated in our results for specific backazimuth windows at single stations. We observe an apparent deepening of the 410 km discontinuity northwest of GRA1 (Fig. 3), as well as a local deepening of the 660 km discontinuity northeast of BFO. However, the data is presently too sparse (caused by the stations being too far apart) to analyse these structures in more detail. The observed anomalies may well be related to small-scale depth variations of the MTZ discontinuities close to the Eifel Plume (Budweg et al. 2006), but could also be caused by complex upper-mantle structures.
Influence of non-olivine component phase transformations
The preceding discussion was based on the assumption that all the seismic discontinuities of the mantle transition zone are dominated by the transformations of the olivine component. However, as shown by different authors (e.g. Revenaugh & Jordan 1991; Vacher et al. 1998; Deuss et al. 2006), the transformations of garnet-to-ilmenite and ilmenite-to-perovskite may also have an effect on the seismic velocity distribution, depending strongly on the temperature regime. As shown by Vacher et al. (1998), ilmenite is stable in cold regions (1000 K adiabat). Multiple phase transitions should, therefore, occur at depths of from 620 to 710 km, especially in garnet or MORB rich material, in a manner similar to some subducted oceanic slabs. According to Vacher et al. (1998), the gt-sp breakdown occurs in a very sharp (pressure/depth) interval (ca. 2 km thickness). The ilmenite transitions (gt-ilm, ilm-pv) are about 22 and 56 km thick, respectively, and may, therefore, be invisible in high-frequency seismic data. In hotter regimes, most of the seismic gradient is caused by the ‘olivine’ transitions; only a small gradient below the 660 km discontinuity might be the contribution from the ‘garnet’ transitions.
Using high-pressure mineralogy and physics, it can be shown that there appear to exist multiple transitions, especially at the base of the mantle transition zone. Unfortunately, the seismic data presented so far is unconvincing. Simmons & Gurrola (2000) and Niu & Kawakatsu (1996), for example, showed indications for multiple seismic discontinuities, although they did not show that the observed phases are in fact primary conversions. Moreover, Tonegawa et al. (2006) did not show any complexity of the 660 conversion in their study in an area close to that of Niu & Kawakatsu (1996). Our own observations of very late (complex and even doubled) phases with delay times of from 68 to 71 s, especially for stations in the eastern Alps and the Carpathians, may of course be related to these multiple phase transitions of the olivine and garnet (ilm-pv) components. In both cases, the late arrivals of the P660s point to a cold and, therefore, thickened transition zone in that region.
Sharpness of MTZ discontinuities
As previously mentioned, we also observed strong variations in the amplitudes of converted phases from the 410 km discontinuity beneath central Europe. As indicated in our study, the 410 phase might interfere with multipath reverberations from the crust and uppermost mantle at some stations (see Figs 4 and 11). Other possible causes of weak or complicated P410s phases are strong topography (van der Lee et al. 1994) or a broad transition layer (50 km wide) instead of the sharp discontinuity, as discussed by Chevrot et al. (1999). Broadening and frequency dependence of the P410s phase could be caused by the presence of water (Wood 1995; Helffrich & Wood 1996; van der Mejde et al. 2003). This may cause a ‘410 transition’ up to 40 km wide (Smyth & Frost 2002), in which case a conversion from the 410 km discontinuity would be barely detectable. Furthermore, broadening may occur due to temperature decrease (Bina & Helffrich 1994), which would contradict our observations beneath the EEC where we observed a particularly sharp 410 km discontinuity in all cases, in contrast to stations in central Europe. It is clear that the conversions from the MTZ at stations in the EEC are sharper than at stations in the rest of the study area. We may speculate that beneath the EEC, the discontinuities in the mantle transition zone are also sharper or at least show less small-scale topography. It may also be the case that less damping/scattering of higher frequencies of the converted phases in the upper mantle may be affecting our observations. If Bina & Helffrich (1994) are correct, then the MTZ below the EEC should not be colder than one might expect from the cool thick lithosphere above. In contrast, partly weak 410 conversions in the Carpathian region may be caused by lower temperatures.
Compositional differences in the upper mantle beneath central and eastern Europe
As shown by our synthetic seismogram analysis, most of the apparent deepening/up-warping of the 410 and 660 km discontinuities can be explained by reduced/enlarged seismic velocities/heterogeneity in the crust and uppermost mantle, which was already proposed by Kind & Vinnik (1988). High Vp/Vs ratios modelled in the upper mantle (71–210 km depth) of central Europe (models BM3A/B: 1.85; IASP91: 1.80) may be caused by lower seismic S (and P) velocities due to higher temperatures and the presence of small amounts of melt or water (asthenosphere). Different chemical composition cannot alone explain such high Vp/Vs ratios (Goes et al. 2000; Hacker & Abers 2004). Another possible source of high Vp/Vs ratios (>1.8) may be anisotropy, as stated by Kobussen et al. (2006). However, if we stack our data over all backazimuths, the effect of anisotropy should be less important (at least at stations with good data coverage).
Vp/Vs ratios as low as 1.75 were modelled for the EEC upper mantle down to 271 km (model EEC4D). Such low values can be partially explained by a cold olivine-orthopyroxene-rich mantle, according to our calculations, using the Excel worksheet of Hacker & Abers (2004). The lower Vp/Vs ratio might be caused mainly by higher shear wave velocities (3.5 per cent above IASP91 velocities), which were previously found by surface wave studies (e.g. Zielhuis & Nolet 1994; Cotte et al. 2002). According to Kopylova et al. (2004), the depleted cratonic mantle is characterized by lower Vp and higher Vs (leading to a lower Vp/Vs ratio) due to opx-enrichment (effect of up to 0.05 km s−1 for the Slave craton). Kopylova et al. (2004) state that many xenoliths from Archean cratons are characterized by extreme depletion, abundant opx and low ol/opx ratios. Xenoliths from the Baltic (Fennoscandian) Shield (Kukkonen & Peltonen 1999) are dominated by harzburgites (up to 77 per cent opx, ‘mean’: 26 per cent). A xenolith geotherm indicates a temperature of about 1300 °C at a depth of approximately 220 km. High Mg# harzburgites are characterized by lower densities, which support the study by Artemieva (2003), which showed that the upper mantle beneath the NE Baltic Shield is up to 1.4 per cent less dense than typical Phanerozoic mantle. As previously discussed by Priestley & McKenzie (2006), low temperatures should have a greater impact on the shear wave velocities than any variations in chemical composition.
A Vp/Vs ratio of 1.75 is likely to be the lower limit of what can be explained by isotropic velocities from chemical composition. It is possible that some amount of the observed Vp/Vs anomalies could also be explained by anisotropic effects, but we cannot explore such a possibility in this study. An opx-rich mantle could explain low Vp/Vs, but not high Vp velocities just beneath the Moho as observed in wide-angle seismic studies (e.g. Grad et al. 2002). It may be that the observed high P-wave velocities occur only within a thin layer just beneath the Moho.
This receiver-function study confirms previous estimates of crustal parameters in central and eastern Europe. There exist big differences in crustal thickness between Palaeozoic Europe, the EEC and the Alpine–Carpathian belt. Small differences exist between extended and non-extended regions in the Palaeozoic crust. Moho Ps delay times for stations on the Palaeozoic platform are generally <4 s (<33 km), but can reach more than 6 s (50–>55 km) in the EEC and beneath the Alps/Carpathians. The amplitudes of the converted phases are normally about 10 per cent of the incoming P-wave amplitude and in the EEC they are generally below 10 per cent. Average crustal Vp/Vs ratios show no correlation with tectonic age of the station basement. The mean value is 1.76, but the Vp/Vs ratio varies from about 1.6 to almost 2.0. However, the accuracy of extreme values is questionable.
The apparent topography of the 410 km discontinuity seems to depend mainly on the Vp/Vs ratio in the upper mantle (above the MTZ), so the delay of 410 conversions compared to those predicted in IASP91 could be caused mainly by a reduction in shear-wave velocity (Vp/Vs increase) within the asthenosphere, whereas the early arrivals from the 410 km (and in consequence also partly of the 660 km) discontinuity beneath the cratonic area are caused mainly by higher shear wave velocities, as is also indicated by studies on surface waves (e.g. Zielhuis & Nolet 1994; Cotte et al. 2002). High shear wave velocities beneath the craton could be explained by the absence of an asthenospheric layer, low temperatures, and a depleted chemical (opx-rich) composition. Goes et al. (2000) interpreted velocity variations in tomography models of Europe at a depth of 80 km as being caused mainly by temperature variations of up to 300 K across the TTZ. Given that we found no large-scale temperature variations in the MTZ across the TTZ, the difference must be a feature mainly of the upper mantle. Assuming that the MTZ is more or less flat below central and eastern Europe, Ps delay times of the MTZ conversions could be used to investigate the upper mantle by Ps tomography (for Vp/Vs variations).
According to our results, the MTZ below the eastern Alps and the Carpathians seems to be thicker by up to 20 km, which confirms the results obtained from tomography studies (e.g. Piromallo et al. 2001; Piromallo & Facenna 2004). It seems probable that the thickening is real and not the effect of an anomalous Vp/Vs ratio in the mantle transition zone, because such large velocity differences are not observed in tomography studies. The MTZ thickening beneath the eastern Alps, the Pannonian Basin and the Carpathians is, therefore, comparable to that observed in the active subduction in Japan (Tonegawa et al. 2006, ∼30km). It may be that large flat slabs rest in the transition zone, similar to the case reported in Japan. Surprisingly, similar effects were not observed beneath Tibet, where the mantle transition zone seems to be of normal thickness (Yuan et al. 1997; Kind et al. 2002). This might point to different depth extents of collisional tectonics.
We thank our colleagues from the SZGRF Erlangen, the GRSN, GEOFON, IG CAS Prague, ReNaSS, ZAMG Vienna, PAN Warsaw, Bratislava, Hungary, Zurich and MedNet for providing data. Most data were obtained from SZGRF, GEOFON and IRIS data centres. This research was supported by the Deutsche Forschungsgemeinschaft and the EU project NERIES. We used the software packages SeismicHandler (Stammler 1993) and GMT (Wessels & Smith 1998) for most of the data processing and plotting.
By analysing 25 yr of data from station GRA1 (1980–2004) for each year separately, variations in the Moho Ps delay times of 0.1 s could be observed (see Table A1). Therefore, the uncertainty of Moho Ps delay time measurements for temporary stations, which were operated for 1 yr, might also be in the range 0.1–0.2 s. Delay times tP410s and tP660s, observed in station stacks of GRA1 and OBN (no figure) for different years of observation, vary by about 0.5 (up to 1.0) s, showing no dependency on the filter used. It may therefore be assumed that the uncertainties for delay time measurements of conversions from the mantle transition zone (410 and 660) are also in the range 0.5–1.0 s for temporary stations. For many stations, absolute delay times for the 410 and 660 km discontinuities depend on the filter used (most obviously for the 410 conversion). This might be caused by higher-order crustal reverberations, which may interfere with the primary 410 conversion (see Fig. A2).
In Fig. A2 we show single moveout-corrected traces of station GRB1 sorted by distance and moveout, corrected for primary conversions Ps (a) and multiple phases PpPs (b), respectively. It is obvious from frequency content and remaining moveout that multiple phases also exist at later Ps delay times, disturbing the primary signals of the 410 and 660 conversions. Theoretical receiver functions for the IASP91 earth model show beside primary conversions from the middle crust, Moho and mantle transition zone also strong multiple phases at around 30 s (Fig. 11a). In the sum trace of moveout-corrected traces this multiple phase is suppressed. If we have a real dataset with excellent distance coverage, multiple phases will be diminished during stacking. If we stack traces over small bins of distance and backazimuth, we will get a mixed signal of primary and multiple energy. This fact should be considered when stacking data for a common conversion point or interpreting migrated sections in detail.