Ancient Craton‐Wide Mid‐Lithosphere Discontinuity Controlled by Pargasite Channels

The mechanisms governing a commonly observed seismic velocity drop in the cratonic lithosphere, referred to as the mid‐lithospheric discontinuity (MLD), have been widely debated. To identify the composition and seismic structure of MLDs, we have analyzed Sp receiver functions (SRF) and mantle xenocrysts for six regions across Australia. We utilize locations where seismic stations and kimberlite‐hosted mantle xenocrysts are both available, allowing for comparison between seismological and petrological constraints. Our results show negative SRF phases indicative of the MLD coincide with clinopyroxene‐depleted zones at 60–140 km depth. Clinopyroxenes with different chemical compositions across the MLD define a litho‐chemical discontinuity. Modeling and experimental data show that MLDs may be explained by modified lherzolite with 10%–20% modal pargasite. Pargasite MLDs may form when rising H2O‐bearing melts cross the amphibole dehydration curve and react with clinopyroxene in lherzolite. Because the amphibole dehydration curve is isobaric at 80–120 km, pargasite will be precipitated as horizontal channels.


Supporting Information:
Supporting Information may be found in the online version of this article.Jaques et al., 2023).The lack of interdisciplinary constraints on the structure and composition of the lithospheric mantle has so-far failed to provide a consensus on the cause of MLDs.
To better understand MLDs we have conducted an interdisciplinary study on the seismological and petrological properties of the lithospheric mantle beneath Australia.At six locations (Figure 1), seismic discontinuities within the lithosphere are imaged with Sp receiver functions (SRF).We also use mantle xenocrysts from adjacent volcanic pipes (Figure 1) to constrain the geochemistry of the lithospheric mantle and interpret SRF profiles.Mantle xenocrysts come from kimberlites emplaced in craton interiors (Gawler Craton, Kimberley Craton, North Yilgarn Craton) and their Proterozoic margins (Delamerian Orogen, Adelaide Fold Belt, Arunta Inlier) (Figure 1) with emplacement ages from Miocene (∼20 Ma) to Paleoproterozoic (∼1,800 Ma) (Supporting Information S1).Our interdisciplinary study allows the seismological and petrological properties of the MLD to be simultaneously constrained on a local scale, thus providing new insights into the potential cause of MLDs.

Seismology of MLDs
The S-to-P converted waves at discontinuities beneath seismic stations are widely used to image upper mantle structures such as the MLD and LAB (e.g., Farra & Vinnik, 2000).We present SRFs calculated with the iterative time-domain deconvolution method (Ligorría & Ammon, 1999) using vertical and rotated radial-component recordings up to the end of 2023 for six permanent stations STKA, HTT, BBOO, WRKA, FITZ, and MEEK deployed in Archean-Proterozoic cratons and their margins throughout Australia (Figures 1 and 2) (Supporting Information S1).SRFs are migrated to depth using the interpolated Australian seismological reference model (AuSREM) (Kennett et al., 2013;Salmon et al., 2013) (Supporting Information S1).In this study, to prevent stacks from being dominated by structure sampled by a specific cluster of events, we average individual traces within 10°back azimuth and then average these binned traces for a final SRF stack (Figure 2).Unlike previous SRF studies where only the station stacks of SRFs are shown (Birkey et al., 2021;Ford et al., 2010), we present the high-quality SRFs over back azimuth along with the stacks and bootstrap-derived mean and standard deviations for each station (Supporting Information S1).For all stations, we observe positive Moho signals at 20- 45 km that generally agree with the Moho depths defined in the national reference AusMoho model (Kennett et al., 2023) and previous P-and S-wave receiver function results (Birkey et al., 2021;Ford et al., 2010) (Figure 2) (Supporting Information S1).Between 6 and 8 s (i.e., ∼60-120 km depth), negative Sp phases indicative of the seismic MLD, are clearly seen on both the individually binned traces and the stacks for most stations (Figure 2) (Supporting Information S1).In all cases there is generally a high amplitude negative phase directly beneath the positive Moho signal (Figure 2), however this may be caused by processing artifacts such as side-lobes.Beneath this first negative phase there is either distributed negative energy extending over 50 km or more (e.g., HTT, WRKA, STKA, BBOO, Figures 2a-2d), indicative of an extended velocity gradient, or distinct deeper negative phases (e.g., FITZ, MEEK, Figures 2e and 2f).For all stations, these deeper or depth-distributed negative phases match with the depths of a distinct lower sampling density of clinopyroxene, and strong chemical inflections in clinopyroxene composition constrained by analysis of mantle derived xenoliths and xenocrysts (Figure 2 and discussed in the next section).We interpret a depth for the MLD (as labeled on Figure 2), considering the presence of negative energy on the SRF and the base of the drop in clinopyroxene (Figure 2) (see below).Our interpretation of the depth of the MLD (60-120 km) is based on both the seismological and petrological observations.It is also generally consistent with a recent SRF study of MLDs by Birkey et al. (2021) on the same set of stations but using 2-year less data, a different rotation (rotated to P-SV-SH coordinate system) and deconvolution method (an extended-time multitaper cross-correlation method) for calculating SRFs, as well as different selection criteria.Several of the negative phases identified by Birkey et al. (2021) for the same set of stations are shown (dashed blue line) on Figure 2. The depth of negative phases reported in Birkey et al. (2021) generally agree well with our observations, albeit with slightly different interpretations for the MLD.For each location (a-f), we report the depth distribution of clinopyroxene xenocrysts from kimberlite and lamproite provinces (left panels) and Sp receiver functions (SRF) stacking profiles calculated for the nearest permanent seismic station (right panels).Clinopyroxene depth was determined using single-grain clinopyroxene geothermobarometry.KDE is kernel density estimate.Note the absence of clinopyroxene within the middle lithosphere (∼90-120 km) for each location (shown as gray band).Interpreted MLD is marked by solid black line based on both negative signal on SRF profile and the depletion in clinopyroxene.Dashed blue line labeled as B21 is the depth of MLD-like negative signals that were imaged by independent SRF study by Birkey et al. (2021).For several stations, Birkey et al. (2021) identified multiple negative phases, however we have only reported the phase nearest to our interpreted MLD.

Composition of MLDs
Mantle xenocrysts are the products of accidental sampling of the lithospheric mantle by rapidly ascending magmas such as kimberlite.Xenocrysts provide snapshots of the chemistry and lithology of the upper mantle at the time of eruption.To study the lithology and geochemistry of the MLD we have collected ∼3,000 clinopyroxene mantle xenocrysts from six volcanic provinces across Australia (Figure 1) (see Sudholz, 2024 for online repository of xenocryst compositions).We use recently published xenocryst data from the Gawler Craton and Adelaide Fold Belt (Sudholz et al., 2022), Kimberley Craton (Sudholz and Jaques et al., 2023) and Arunta Inlier (Sudholz and Reddicliffe et al., 2023) (Figure 1).Our data set includes new xenocryst data for the Yilgarn Craton and Delamarian Orogen (Figure 1).The emplacement age of the host volcanic rocks includes Miocene (Kimberley Craton), Jurassic (Gawler Craton and Adelaide Fold Belt), Permian (Delamarian Orogen), Neoproterozoic (Arunta Inlier) and Paleoproterozoic (Yilgarn Craton) (Supporting Information S1).The geological setting of kimberlites includes craton interiors (Gawler, Yilgarn and Kimberley Cratons) and craton margins (Delamarian Orogen, Arunta Inlier, Adelaide Fold Belt) based on surface geology and inferred subsurface distribution of cratonic lithosphere (Cawood & Korsch, 2008) (Figure 1).The equilibration pressure and temperature of clinopyroxenes within the lithospheric mantle were made using the calibrated Nimis and Taylor (2000) and Sudholz et al. (2021) geothermobarometers (Supporting Information S1).The uncertainty on pressure-temperature estimates is ±35-50°C and ±2.5-4.5 kbar (equivalent to <10 km).At each site, pressure estimates for clinopyroxenes define a bimodal distribution, with populations occurring within the shallow (60-80 km) and deep lithospheric mantle (>110 km) with the mid-lithosphere characterized by a lower sampling density of clinopyroxene (Figure 2 highlighted by gray band).The large sample size for each xenocryst suite (see Figure 2) suggests that the absence of clinopyroxenes within the mid-lithosphere is a robust feature that reflects a region of the lithosphere where clinopyroxene has been poorly sampled (i.e., clinopyroxene-free lithology).This layer occurs at similar depths to the negative SRF pulses that characterize the seismic MLD for adjacent seismic stations (Figure 2).The clinopyroxene-depleted zone occurs between 80 and 140 km for craton interiors (Gawler, Kimberley and Yilgarn Cratons) (Figures 2c, 2e, and 2f) and 60-100 km for craton margins (Delamerian Orogen, Arunta Inlier and Adelaide Fold Belt) (Figures 2a, 2b, and 2d).The layer also separates populations of clinopyroxenes from the shallow and deep lithosphere that have different chemical compositions (Supporting Information S1).Clinopyroxenes from the deep lithospheric mantle beneath the Gawler Craton (green circle) are more enriched in Na 2 O and Y and depleted in La (Supporting Information S1).Clinopyroxenes from the Kimberley Craton (red circle) record enrichments in La and K 2 O within the deep lithospheric mantle and lower concentrations of Y and Sc (Supporting Information S1).Clinopyroxenes from the Arunta Inlier (purple square) and Adelaide Fold Belt (blue diamond) record enrichments in TiO 2 and Cr 2 O 3 within the deep lithospheric mantle, and clinopyroxenes from the deep lithospheric mantle beneath the Yilgarn Craton (yellow triangle) have elevated concentrations of Cr 2 O 3 , FeO, and Na 2 O (Supporting Information S1).The decrease in the abundance of clinopyroxene and strong chemical inflections between the shallow and deep lithospheric mantle, and its association with negative SRF pulses, suggests that the clinopyroxene-depleted zone is an important petrological discontinuity that relates to the MLD.
The formation of pargasite and related amphiboles from a clinopyroxene precursor has been replicated experimentally and their modal abundances are approximately inversely correlated (Fumagalli et al., 2009;Mandler & Grove, 2016;Niida & Green, 1999;Saha et al., 2021) (Supporting Information S1).Assuming an upward percolation of H 2 O-bearing melt or fluid, the above reactions will promote the depletion in clinopyroxene and enrichment in pargasite once the melt or fluid crosses the isobaric section of the dehydration curve (Figure 3a).Such H 2 O-bearing melts or fluids may occur as kimberlites or proto-kimberlites, carbonatites, basanites, or other low-volume exotic melts.This process will lead to the formation of narrow channels or veins of pargasite-bearing clinopyroxene-depleted lherzolite at, or slightly above the dehydration curve at an approximately uniform depth.has been observed in xenolith suites from the cratonic lithosphere (Dawson & Smith, 1982).A limited compilation of published xenolith mineral compositions shows that amphibole-bearing xenoliths record equilibration pressures and temperatures between 22 and 48 kbar and 785-1093℃.The PT range of amphibole-bearing xenoliths corresponds to mid-lithospheric depths beneath most craton/craton margins settings and overlaps with our clinopyroxene-free layer and associated MLD seismic phases.The preservation of pargasite is hindered by its exceptionally low melting temperature (<1050℃) and susceptibility to weathering at the surface.The elevated  -c).For each lherzolite composition three modal abundances of pargasite were modeled: 5%, 10%, and 20%.Line color and width represent different pargasite modal abundances for each model (bold line is 20% modal pargasite, intermediate line is 10% modal pargasite, and thin line is 5% modal pargasite).Figure includes the modal abundances of clinopyroxene on the x-axis, calculated seismic Vs models, and synthetic Sp receiver functions.The compositions of each model and the Vs models output are reported in an online data repository (Sudholz, 2024).

10.1029/2024GL108433
temperatures of most kimberlitic melts (∼1050-1350℃) likely dissolve pargasite during transport.Pargasitebearing xenoliths may also be poorly represented in xenolith suites due to the narrow width of most MLDs (<20 km).Nonetheless, based on these observations we infer that pargasite-bearing lherzolite within the cratonic lithosphere causes lower modal abundances of clinopyroxene along the isobaric section of the dehydration curve (∼80-120 km).The elevated H 2 O contents associated with pargasite channels contribute toward a seismic velocity drop which characterizes the MLD.

Petrophysical Modeling
Based on the evidence presented above, we interpret the seismic velocity drop associated with the MLD to be caused by pargasite-bearing modified mantle lherzolite.To demonstrate this, we have conducted a series of simple 1D and 2D petrophysical simulations for Vs and SRF along a cratonic geotherm (∼40 mW/m 2 ) (Figure 3).Our models are performed at 400-1400℃ and 10-70 kbar at intervals of 25℃.We base our models on three conceptual lherzolite compositions: Lhz A, Lhz B, and Lhz C, which have Mg# values consistent with refractory (Mg # 93 olv ), intermediate (Mg # 90 olv ), and fertile (Mg # 88 olv ) mantle compositions respectively.The range in Mg# captures the compositions of most peridotites from the cratonic lithosphere (Pearson & Wittig, 2008).We include a pargasite-bearing layer at 90-120 km.For each lherzolite we have tested three different modal abundances of pargasite: 5%, 10%, and 20%, resulting in a total of nine models.We incorporate pargasite into the midlithosphere by depleting clinopyroxene and garnet at a ratio of 2:1 (2 pargasite = 1 clinopyroxene + 1 garnet).Future studies are needed to explore more precise enrichment and depletion systematics for pargasite.Notwithstanding, our approach is consistent with approximate phase relations in experimental lherzolites (Mandler & Grove, 2016;Saha et al., 2021) and depletion in clinopyroxene and garnet across the mid-lithosphere beneath Australia and other cratons (Figure 3) (Shaikh et al., 2020;Sudholz et al., 2022).Mineral abundances along the cratonic geotherm were used to construct 1D seismic Vs models and SRF synthetics.Values for Vs were calculated using the thermo-elastic properties reported in Abers and Hacker (2016).We have not used the updated parameters for pargasite reported in Saha et al. (2021), however, it is expected that they would produce a stronger signal for pargasite-bearing MLDs.The mineral abundances and Vs models are reported in an online data repository (Sudholz, 2024).Synthetic SRFs were calculated for each Vs model using the same rotation and deconvolution method and filters applied to real data, using associated reflectivity codes that compute 3component seismograms (Levin & Park, 1997;Park, 1996) (see Acknowledgments for source code).
The modelled modal abundances of clinopyroxene and pargasite and resulting Vs models and SRF synthetics for an intermediate (Mg # 90 olv ) and refractory (Mg # 93 olv ) lherzolite composition are reported in Figures 3b and 3c.Depletion in clinopyroxene occurs across the interval where pargasite is stable (90-120 km).The greatest depletion in clinopyroxene occurs for lherzolite with 20% modal pargasite.Conversely, the highest modal abundance of clinopyroxene occurs for lherzolites with 5% modal pargasite (left panels in Figures 3b and 3c).Intermediate and refractory lherzolite compositions record a strong decrease in Vs between 90 and 120 km (middle panels in Figures 3b and 3c).Both compositions have similar trends in Vs versus depth although the refractory lherzolite has larger values for Vs.As expected, the strongest decrease in Vs across the mid-lithosphere occurred for lherzolites with 20% modal pargasite and fertile (less depleted) whole-rock compositions.The smallest decrease in Vs occurred for models with 5% pargasite.For compositions with 20% modal pargasite, the decrease in Vs across the pargasite layer was approximately 2.5%-5%.The decrease in Vs for compositions with 10% and 5% modal pargasite were approximately 1.5%-2% and 1%-1.5% respectively (Figures 3b and 3c).Each modelled composition produced SRF negative signals across the pargasite bearing layer, with the strongest negative excursions produced for lherzolites with 20% modal pargasite (right panels in Figures 3b and 3c).Similar excursions occur for intermediate and refractory compositions.The models show that the abundance of pargasite has a much greater effect on the modelled Vs and SRF than whole rock Mg#.Because most MLDs observed globally have velocity decreases on the order of 2%-8% (i.e., Rader et al., 2015), our modelling suggests that >15% modal pargasite is likely required in order to produce strong enough decreases in seismic velocity (Figures 3b and 3c).The amount of pargasite may be slightly lower if very strong contrasts in Mg# also exist between the pargasite-bearing layer and the overlying lithospheric mantle.As suggested by Kovács et al. (2017Kovács et al. ( , 2021)), low volume (<1%) partial melts at the dehydration solidus (underlying the pargasite channels) may further reduce Vs.Such melts are probable given that Niida and Green (1999) found that the degree of melting in peridotite increases from 0% to 5% at more than 15℃ above the dehydration solidus.

Forming Pargasite MLDs
Petrological and seismological data has constrained the MLD beneath six seismic stations and kimberlite provinces across Australia.In all instances, the MLD is preserved between 60 and 140 km through a depletion of clinopyroxene and at slightly shallower depths (60-120 km) through a seismic discontinuity (Figure 2).The MLD is interpreted to have formed from bottom-up percolation of H 2 O-bearing melts from the deep lithospheric mantle or shallow asthenosphere (Figure 4a).H 2 O may have been transported by proto-or failed kimberlites, plumes, slab-triggered upwellings or other mechanisms (Freitas & Manthilake, 2019;Green, 2015;Kuritani et al., 2019;Sudholz et al., 2022;Sudholz and Reddicliffe et al., 2023;Yang et al., 2018).Given the temporal and spatial extent of the MLD in the global lithosphere, the supply of H 2 O to the mid-lithosphere is a universal, albeit poorly understood process that has persisted throughout geological time.The primary evidence supporting a subdehydration solidus origin for H 2 O (bottom-up hydration) comes from the horizontal nature of the MLD in geophysical and petrological data.In our preferred model, rising H 2 O-bearing melts/fluids from the deep lithospheric mantle or shallow asthenosphere crystallize horizontal channels of pargasite at the expense of clinopyroxene once they cross the isobaric section of the dehydration curve (Figure 4b).The isobaric section that controls pargasite stability at mid-lithospheric depths is the primary reason why most MLDs occur as horizontal layers across a narrow depth range in the cratonic lithosphere.Although phlogopite can cause strong velocity drops in Vs due to its high OH contents (see Saha et al., 2021 for discussion), the exceptionally large stability field of phlogopite provides no suitable explanation why a negative SRF phase would be concentrated exclusively at mid-lithospheric depths (60-140 km) within otherwise thermally stable cratons.The isobaric section that controls pargasite stability may have corresponded to a paleo-LAB that existed before cratonization (i.e., Kovács et al., 2017Kovács et al., , 2021;;Sudholz et al., 2022;Sudholz and Jaques et al., 2023).This interpretation is supported by the present-day seismic structure of un-cratonized continental lithosphere (i.e., Tasmanides of eastern Australia), which shows negative SRF phases between 80 and 120 km that correspond to the base of lithosphere (Birkey et al., 2021;Ford et al., 2010).Thin pargasite MLDs may occur when the H 2 O budget of the percolating melt/fluid is low, whereas thicker MLDs may originate from high-volume H 2 O-rich melt/fluid.Because of its low melting temperature, the pargasite MLD is likely a dynamic discontinuity that has evolved and reformed throughout Earth's history, explaining its semicontinuous nature within the cratonic lithosphere and depth misalignment between some stations.Variations in the depth of the MLD may also be due to small changes in the whole-rock composition of peridotite (i.e., Mg#, H 2 O, K 2 O, and TiO 2 ) which can greatly influence the depth of the isobaric section of the amphibole dehydration curve (Foley, 1991;Juriček & Keppler, 2023;Mandler & Grove, 2016;Niida & Green, 1999) (Figure 3).The occurrence of a pargasite MLD in all major cratons of Australia and for kimberlite provinces to Paleoproterozoic age suggests that the formation of pargasite MLDs may be preserved from craton formation.It is expected that the precipitation of pargasite will produce a marked change in the viscosity of the lithospheric mantle and will produce a "mid-lithospheric sponge" that can act as a reservoir for volatiles and incompatible elements.The layer will also have important implications for Earth geodynamics and the composition of small volume magmas erupted through the lithospheric mantle.The strong compositional and density contrasts across the MLD will result in a sharp change in rheology which may influence the stability and fate of the cratonic lithosphere.A pargasite MLD will be a weak layer in the lithosphere, providing a suitable "tear point" for lower lithosphere delamination (Kovács et al., 2017(Kovács et al., , 2021)).Similarly, the low melting point of pargasite-bearing lithologies may contribute to low-volume mafic cratonic magmas (Conceição & Green, 2004).MATLAB code for computing synthetic seismograms.PZ thanks Dr. Hongjian Fang from Sun Yat-sen University for insightful discussions on seismic data processing.All EPMA analyses were completed at the Centre for Advanced Microscopy, a precinct of Microscopy Australia.ZS acknowledges funding by Geoscience Australia (Exploring for the Future).PZ acknowledges funding support by the Australian Research Council (ARC) Linkage Proposal (LP180101118).We thank Linda Frewer (Diatech Heavy Mineral Services), Wayne Taylor, Steven Cooper (Orogenic Exploration), and Peter Downes (WA Museum) for providing samples used in this study.Two anonymous reviewers are thanked for providing constructive feedback that improved the mansucript.Quentin Williams is kindly thanked for editorial handling.David Headley Green is thanked for stimulating some of the ideas explored in this manuscript and for his 60+ year contribution to experimental studies of the lithospheric mantle.KC publishes with the permission of the CEO of Geoscience Australia.

•
Sp receiver functions and mantle xenocrysts are used to study midlithosphere discontinuities beneath Australian cratons • Mid-lithosphere discontinuity at 60-120 km depth corresponds with mantle xenocryst populations that are depleted in clinopyroxene

Figure 1 .
Figure 1.Topographic map of Australia with locations of permanent seismic stations (red inverted triangle) and kimberlitehosted mantle xenocryst and xenolith suites (yellow circle) in six different tectonic units considered in this study.Pink shading reflects the inferred outline of Archean-Proterozoic cratons.Orange shading reflects Proterozoic cratons, basins and orogens.

Figure 2 .
Figure2.Petrological and seismological results on MLDs from six regions across Australia.For each location (a-f), we report the depth distribution of clinopyroxene xenocrysts from kimberlite and lamproite provinces (left panels) and Sp receiver functions (SRF) stacking profiles calculated for the nearest permanent seismic station (right panels).Clinopyroxene depth was determined using single-grain clinopyroxene geothermobarometry.KDE is kernel density estimate.Note the absence of clinopyroxene within the middle lithosphere (∼90-120 km) for each location (shown as gray band).Interpreted MLD is marked by solid black line based on both negative signal on SRF profile and the depletion in clinopyroxene.Dashed blue line labeled as B21 is the depth of MLD-like negative signals that were imaged by independent SRF study byBirkey et al. (2021).For several stations,Birkey et al. (2021) identified multiple negative phases, however we have only reported the phase nearest to our interpreted MLD.

Figure 3 .
Figure 3. Approximate pressure-temperature stability range of pargasite in fertile and depleted peridotite and the dehydration curve and oxidized solidus of peridotite (depleted) (a).Model paleogeotherms taken from Sudholz et al. (2022) (Gawler), Sudholz and Reddicliffe et al. (2023) (Webb), and Sudholz and Jaques et al. (2023) (W.Kimberley). Figure modified after Green and Falloon (2005) and Mandler and Grove (2016).Petro-physical modeling of intermediate (Mg# 90 olv ) and depleted (Mg# 90 olv ) lherzolite compositions with variable modal abundances of pargasite (b-c).For each lherzolite composition three modal abundances of pargasite were modeled: 5%, 10%, and 20%.Line color and width represent different pargasite modal abundances for each model (bold line is 20% modal pargasite, intermediate line is 10% modal pargasite, and thin line is 5% modal pargasite).Figure includes the modal abundances of clinopyroxene on the x-axis, calculated seismic Vs models, and synthetic Sp receiver functions.The compositions of each model and the Vs models output are reported in an online data repository(Sudholz, 2024).

Figure 4 .
Figure 4. Schematic diagram of the formation of MLDs through the precipitation of pargasite and breakdown of clinopyroxene along the dehydration curve within the mid-lithosphere of cratons and craton margins.(a) Upward percolation of H 2 O-bearing fluids or melts from the deep sub-crust lithospheric mantle (>120 km) or shallow asthenosphere resulted from various mechanisms.(b) Pargasite is crystalized at mid-lithospheric depths when H 2 O-bearing melts or fluids cross the dehydration curve.The reaction between H 2 O-bearing melts or fluids with clinopyroxene in lherzolite at the dehydration curve will produce narrow channels of pargasite-bearing lherzolite because the dehydration curve controlling pargasite formation is isobaric at mid-lithospheric depths.(c) The pargasite MLD is preferentially sampled by kimberlites and related volcanics during rapid transport to the surface.The low melting temperature of pargasite may result in poor preservation and chemical weathering during transportation.