Water, Hydrous Melting, and Teleseismic Signature of the Mantle Transition Zone

Recent geophysical and petrological observations indicate the presence of water and hydrous melts in and around the mantle transition zone (MTZ), for example, prominent lowvelocity zones detected by seismological methods. Experimental data and computational predictions describe the influence of water on elastic properties of mantle minerals. Using thermodynamic relationships and published databases, we calculated seismic velocities and densities of mantle rocks in and around the MTZ in the presence of water for a plausible range of mantle potential temperatures. We then computed synthetic receiver functions to explore the influence of different water distribution patterns on the teleseismic signature. The results may improve our understanding and interpretation of seismic observations of the MTZ.


Introduction
Three major phase transitions define the mantle transition zone (MTZ) between the upper mantle and lower mantle: olivine transforms to wadsleyite at~410 km, wadsleyite to ringwoodite at~520 km, which then breaks down into bridgmanite, and periclase at~660 km (e.g., [1]). The exact depths of these transitions (in the following referred to as "410", "520", and "660", respectively) vary with temperature, pressure, chemical composition, and volatile content in rocks. Phase changes may cause strong seismic velocity discontinuities, which can be imaged using different seismological methods (e.g., [2][3][4][5]). The MTZ can act as a strong thermomechanical boundary between the upper and lower mantle, possibly impeding mass exchange between the upper and lower mantle (subduction and upwelling). Mapping of the actual depths of phase transitions is essential in order to understand the geodynamics of the Earth (e.g., [6,7]).
For decades, olivine with spinel structure has been recognised as a potential water-carrying mineral in the Earth's mantle [1,[8][9][10][11]. The ability of the MTZ to capture, store and release water may be a crucial aspect of the global water budget/cycle, mantle dynamics, and plate tectonics [12][13][14][15]. Reference [16] described the peridotite-water system and proposed hydration of the MTZ by dehydration of subducting slabs which has been supported by other studies (e.g., [17]). The finding of a water-rich ringwoodite inclusion in diamond was direct evidence for the presence of water in the mantle [18,19] and proposed the presence of water in the MTZ to explain the electrical conductivity of mantle rocks; however, this interpretation is debated (e.g., [20]). Reference [21] interpreted high seismic attenuation anomalies in the MTZ beneath the North Atlantic as evidence for water-rich rocks. A similar study proposed a 20-35 km thick layer with up to 700 ppm of water, close to the 410 [22,23], used water-bearing nominally anhydrous minerals (NAMs) for thermodynamic modelling to explain observed seismological data.

Water Saturation and Considered Mineral Phases
The model involved only nominally anhydrous minerals (NAMs) which is a common assumption for the mantle far away from subducting slabs, likely the main modern source of water influx to the mantle (e.g., [12]). The upper limits of water content in NAMs are well-studied, and these estimates are consistent. Olivine can carry a few thousands of ppm of water, garnet can hold hundreds of ppm (except hydrated species), pyroxene may contain up to 700 ppm of water, and omphacite up to 2000 ppm [64]. High-pressure olivine polymorphs can bear much more water, up to 1.5 wt.% for wadsleyite and 1.7 wt.% for ringwoodite [65].
The reduction of elastic moduli with increasing water content is usually assumed to be linear. The derivatives ∂K S /∂C H2O and ∂G/∂C H2O are studied for olivine and its polymorphs, but measurements for pyroxene and garnet are still sparse. Therefore, we considered the effects of water content only for olivine and its polymorphs which can also hold much more water than other NAMs. We included a linear correction for iron content [66]. Experiments compared to ab initio estimates (Table 1) predicted lower elastic moduli reduction for olivine and higher reduction for wadsleyite and ringwoodite.

Thermal Profile
The mantle thermal (and pressure) profile determines phase stability, density, and chemical composition of coexisting minerals. Higher mantle potential temperatures (T p ) lead to a reduction of seismic velocities due to the decreasing elastic moduli (V P : 0.75 ± 0.15%, V S : 1.3 ± 0.30% per 100 • C, [68]). Independent estimates of the T p were derived from petrological analyses of mantle-derived rocks and depend on the presence of volatiles and the primary melt composition. Fluid-bearing conditions result in a T p range of 1335-1355 • C and up to 1430 • C for dry depleted rocks [69]. However, newer studies propose a much wider temperature range of 1280-1400 • C for "ambient mantle" (e.g., [70,71]). Estimates of the adiabatic gradient in the mantle are very consistent at approximately 0.4-0.5 K/km for the upper mantle and 0.3 K/km in the lower mantle [62]. We computed models for a range of mantle potential temperatures (1250 • C "cold", 1350 • C "medium", 1450 • C "hot") and a temperature gradient of 0.4 • C/km.

Water Distribution Patterns
We tested two general patterns of water distribution. The first one ("linear") assumed a linearly increasing water content in the MTZ in a lateral direction. The second one ("Gaussian") comprised a 2D Gaussian water distribution centred at a 410 km depth and/or at a 575 km depth which resulted in two approximately ellipsoid hydration anomalies with lateral and vertical variations in water content. Three values of maximum water content in the different anomalies were defined at 0% ("dry"), 50% ("intermediate") or 100% ("wet") of the saturation level for a given phase. We tested different combinations of saturation levels in the upper and lower MTZ (wadsleyite and ringwoodite bearing layers): 0-0%, 0-50%, 50-0%, 50-50%, 100-0%, 0-100%, and 100-100% (each in percent with regards to the maximum saturation level) for each distribution pattern (linear and Gaussian).

Domain Calculation
The mantle phase diagram includes stability fields for olivine, wadsleyite, ringwoodite, and oxides. The reference Clapeyron slopes and pressures of phase transitions adjust to the given water content [1,25], temperature [72], and iron content [73]. The depth of each phase transition was estimated individually in each rock column. We used the third-order Birch-Murnaghan equation with the Mie-Grüneisen approach to compute the elastic moduli and densities under lithostatic pressure (derived from the weight of overlying rocks) and the Adams-Williamson equations to calculate seismic velocities [55].
We used a reference melting diagram from Reference [1]. Temperatures of liquidus and solidus decrease linearly from dry to wet conditions with increasing water content [26]. The melt fraction also increases linearly between solidus and liquidus [74]. We used parameters from Reference [31] for the equation of state, which allows to derive the bulk modulus of a melt-bearing rock by the Biot-Gassmann equation (e.g., [75]). Due to the high melt fractions that we obtained in our simulations (up to 30%), we also altered the shear modulus value using a relationship from Reference [76], in which we assumed a shear modulus of 0 for the melt fraction.

Seismic Velocity Reduction Due to the Attenuation
Seismic attenuation depends on the grain size, mechanism of dislocations, temperature, pressure, and rock properties (e.g., [77]). The exact mechanisms and magnitudes of seismic attenuation are still poorly known (e.g., [78]). However, a first-order estimate for the influence of seismic attenuation on the synthetic RFs is an important step to make comparisons to natural data.
Generally, seismic attenuation may be divided into three components: a "dry" value, a correction for the presence of water, and a correction for the presence of melt. We used a parameterisation summarised in Reference [78] for the "dry" component which resulted in an up to 0.5% decrease in the seismic wave amplitude (e.g., [14]). Reference [77] presented an approach to quantify the influence of water on seismic attenuation which we used to correct the value of "dry" seismic attenuation. If melt existed in the rock, we used attenuation values provided by Reference [79]. The decrease in seismic energy in this case reached values up to several percent depending on the actual melt fraction.

Receiver Function Modelling
P-receiver functions (RFs) are deconvolved waveforms of converted P-to-S phases from teleseismic earthquakes and a common tool to image seismic discontinuities, traditionally in the crust and upper mantle (e.g., [80][81][82][83]). The P-to-S conversions become visible as pulses at the delay time of the converted S-wave after the arrival of the primary P-wave, which is dependent on the overlying S and P wave velocity structure and ray parameter of the earthquake wave [84].
From the derived Vp, Vs, and density structure, we computed synthetic RFs using a 1D wavefield to calculate the surface response to an incoming plane wave [85]. Subsequent water-level deconvolution with a Gaussian factor of 1 isolates the RF waveforms (e.g., [80,86]). Although lithospheric layering was included in the thermal-petrological modelling, the evoked lithospheric seismic discontinuities were disregarded in this step to avoid multiples that may compromise the MTZ signals, although these can be assumed to only cause minimal interference [7]. Since the true velocity structure was known, we are able to perfectly back-migrate the signals to their true conversion depth.

Phase Equilibria
The depths of the MTZ-phase transitions computed for a dry system (405, 525 and 655 km, Figure 1 and Figure S1) were close to the reference levels of 410, 520, and 660 km. These depths changed considerably under wet conditions: the depth of the olivine-wadsleyite transition (410) shifts to a depth of 300 km, and the depth of wadsleyite-ringwoodite transition (520) moves to 450 km. This behaviour follows the general principle that the presence of a specific component (water) increases the stability field of a phase that effectively dissolves it. A lack of experimental data did not allow us to constrain the behaviour of the lower-upper mantle boundary.

Phase Equilibria
The depths of the MTZ-phase transitions computed for a dry system (405, 525 and 655 km, Figures 1, S1) were close to the reference levels of 410, 520, and 660 km. These depths changed considerably under wet conditions: the depth of the olivine-wadsleyite transition (410) shifts to a depth of 300 km, and the depth of wadsleyite-ringwoodite transition (520) moves to 450 km. This behaviour follows the general principle that the presence of a specific component (water) increases the stability field of a phase that effectively dissolves it. A lack of experimental data did not allow us to constrain the behaviour of the lower-upper mantle boundary.    (Figure 2a,e). For the medium 1350 • C geotherm melting is only observed in almost water-saturated rocks (more than 80% for wadsleyite and 90% for ringwoodite). The melt fraction reaches 14% and 7%, respectively. Melting starts at much lower water saturation levels in a hot 1450 • C geotherm (≥60% for wadsleyite and ≥70% for ringwoodite). The melt fraction reaches 30% for water-saturated wadsleyite and 21% for water-saturated ringwoodite. saturated (Figure 2a,e). For the medium 1350 °C geotherm melting is only observed in almost watersaturated rocks (more than 80% for wadsleyite and 90% for ringwoodite). The melt fraction reaches 14% and 7%, respectively. Melting starts at much lower water saturation levels in a hot 1450 °C geotherm (≥60% for wadsleyite and ≥70% for ringwoodite). The melt fraction reaches 30% for watersaturated wadsleyite and 21% for water-saturated ringwoodite.  . The effect of water and temperature on the melt fraction (%) (a,e), relative S-wave velocity change (b,f), relative P-wave velocity change (c,g), and relative V p /V s ratio change (d,h) for wadsleyite (a-d) and ringwoodite (e-h). The relative changes in V p , V s , and V p /V s display the difference between the maximum and minimum value, normalised by the maximum ((V max − V min )/V max ). A constant, low-gradient linear increase of this relative change was observed where no melt is present, which seems almost temperature independent. After the appearance of the first melt, the gradients increased significantly, while the gradients were still similar relative to the onset of melting.

Velocity Models
Variations in the MTZ thickness due to the water content and temperature did not allow to compare the different models directly in the spatial domain. We therefore computed a relative velocity change ((V max − V min )/V max ) for each MTZ layer and every model. This relative velocity reduction in dry systems was less than 2% for V s and 2.5% for V p , both in the lower and upper MTZ layers and for all the studied geotherms (Figure 2b,c,f,g).
Even small amounts of water are enough to substantially alter the velocity structure: 10% of water saturation reduces V s by 6%-10% and V p by 4.5%-7.5%. Further addition of water decreases seismic velocities linearly up to 10%-15% (V s ) and 7.5%-12% (V p ) dependent on mantle potential temperature. The reduction of seismic velocities is larger for wadsleyite-bearing rocks compared to ringwoodite-bearing rocks: 17%-20% larger for V s and 22%-32% for V p at 1250 • C, 21%-26% for V s and 30%-35% for V p at 1350 • C, and 35%-45% for V s and 40%-50% for V p at 1450 • C.
The presence of melt reduces the shear moduli of rocks drastically. Our parameterisation for the shear modulus [76] and bulk modulus (Biot-Gassmann) resulted in a non-linear velocity reduction with increasing melt fraction and temperature ( Figure 2). The V s was reduced by up to 80% and V p by up to 34% in the wadsleyite-bearing rocks (for a 1450 • C geotherm and 100% water saturation). In the same conditions, ringwoodite-bearing rocks showed a 65% reduction for V s and 26% for V p . The seismic attenuation parameterisation predicted an additional velocity reduction which reached~10% of V p and~7% V s under wet and hot conditions.

Receiver Functions
Receiver functions in the MTZ region are usually free of large-amplitude lithospheric multiples and, thus, observed peaks corresponded solely to velocity discontinuities within the MTZ itself. Here, we completely removed lithospheric multiples. As discussed above, both temperature and water content change the depth of the MTZ-phase transitions and both quantities can enhance or reduce existing seismic discontinuities or add new ones.
We performed a number of simulations to explore the influence of different water distribution profiles (linear and Gaussian), mantle potential temperature (1250 • C "cold", 1350 • C "medium", 1450 • C "hot"), and water content in MTZ olivine polymorphs ("dry", "intermediate", and "wet" cases). As a reference example illustrated in the main text, we focused on the Gaussian water anomaly distribution, both in the upper and lower MTZ reaching up to 100% water saturation. The corresponding figures of this example are shown in the main text (Figures 3-6). We also describe the corresponding linear model in the text, but figures are shown only in the Supplementary Materials. The other tested models are also documented in the Supplementary Materials investigating potential mantle temperatures.     In contrast to the Gaussian models, the linear water distribution models preserve the seismic discontinuities exactly at the depth of the actual phase transitions, as these also form the boundaries The results discussed are entirely based on this small set of defined models to illustrate the general problem and behaviour of water in the MTZ and its teleseismic signature. Any combination and geometry could, in theory, be tested and would result in a unique model and receiver function signal.

General Observations
The water-induced reduction of elastic moduli was proportional to the absolute amount of water. Therefore, only MTZ-lithologies were strongly affected by the addition of water, because the water-storing capacity of the upper mantle (above the MTZ) and lower mantle main rock-forming minerals is very low. The amplitude of the 410 was generally reduced if water was present in the upper MTZ and that of the 660 was increased if water was present in the lower MTZ. In some cases (high water content and temperature), we created pronounced low-velocity zones which was only the case when melts were formed (from approximately 70% water saturation for 1450 • C or and 90% for 1350 • C, Figure 2a,e). Therefore, no model with an intermediate water content showed occurrence of melt and visible low-velocity zones.

Gaussian Distribution Models
Models with Gaussian water anomalies (Figures 3 and 4, Figures S2-S7) exhibited much weaker discontinuities compared to the linear models presented hereafter, due to the smooth vertical gradient in the water content, an inherent feature of the Gaussian model parameterisation, compared to the abrupt changes exhibited in the linear models. There were no melt formed in the dry and intermediate cases as well as under wet conditions and cold geotherm.
In the presence of water in the upper MTZ (intermediate and wet setups), the 410 discontinuity appeared shallower and the corresponding RF-amplitude weaker (Figures 3 and 4, Figures S2-S5). High water content may reduce the MTZ seismic velocities to values similar to the overlying upper mantle, so that the seismic 410 discontinuity vanishes. The effects on the 520 discontinuity were usually much less pronounced, because the centre of the Gaussian water anomalies were placed closer to the 410, and, logically, the water anomaly at the 520 km depth was at the lower end of the Gaussian distribution and weaker.
In the case of a hydrous lower MTZ (Figures 3 and 4, Figures S2, S3, S6, and S7), which is generally thinner than the upper MTZ, the shallowing of the 520 appeared to be less affected by the addition of water; however, the weak, positive 520 peak also disappeared with increasing water content and an associated velocity reduction in the lower MTZ. Whilst no negative peaks were observed in the upper MTZ for cases without melt (intermediate water anomalies and/or cold geotherms), in the lower MTZ, weak but clearly negative conversions were produced just below the actual 520. This was due to the relatively sharp increase of water saturation, even in cases of a cold geotherm and intermediate water saturation.
As soon as the first melts were generated (see Figure 2 for an overview), V p and V s were extremely reduced and the V p -V s ratios increased, resulting in low-velocity "melt pockets" both in the upper and lower MTZ. These melt pockets were clearly visible in the RFs, marked by high-amplitude negative conversions at the top and high-amplitude positive conversions at the bottom of the area of partial melt ( Figure 5, Figures S2, S4, and S6). In the case of the lower MTZ, the 660 appeared as a double-peak, the shallower one coinciding with the lower limit of the melt pocket and the deeper with the actual lower-upper mantle boundary.

Linear Distribution Models
The other set of models included a homogeneous water anomaly across the entire vertical column of the upper and/or lower MTZ, i.e., simulating the case in which the entire vertical MTZ-layers were saturated with the same amount of water ( Figures S8-S13). The water content increased from zero at the left model bound to the maximum water content of 50% or 100% and the right model bound. In the linear model setup, it was possible to observe the impact of pressure versus temperature on melting more systematically. More melt was produced at shallower depths (lower pressure) than at greater depths (higher pressure) with the same water concentration, although the temperature was lower (Figures S8, S10, and S12). In this specific model setup, the distribution of melt appeared like an "inverted" (upward opening) triangle.
In contrast to the Gaussian models, the linear water distribution models preserve the seismic discontinuities exactly at the depth of the actual phase transitions, as these also form the boundaries between hydrated and non-hydrated layers. The signal of the 410 generally weakens due to the reduction of seismic velocities with increasing water contents. When only the lower MTZ carried water, we observed that the 520 amplitude decreased and could even become negative for high amounts of water, also for models without melt generation (Figures S12 and S13). This amplitude reversal was obviously amplified in the presence of melt ( Figure S12). The 660 amplitude increased with increasing water content in the lower MTZ, but the difference was not extremely clear, even when melt was generated. When both the upper and lower MTZ had the same level of water saturation, we observed that the 520 discontinuity remained largely unchanged.
A wet upper MTZ exhibited a small negative conversion under the cold conditions (1250 • C geotherm), corresponding to a weak low-velocity zone. Intensive melting under higher temperatures (1350 • C and 1450 • C geotherms) in the upper MTZ resulted in a prompt and abrupt high-amplitude negative conversion at the 410, which was now at shallower depth, clearly visible in the RFs ( Figures  S8 and S10). Similarly, high water content in the lower MTZ at high temperatures causes an immediate and high-amplitude negative conversion at the 520, which now almost reached the same depth as the 410. However, we deemed models with 100% water saturation throughout the entire layers of the vertical MTZ column as rather unrealistic.

Velocity Pull-Down and Depth Conversion
We observed that, in cases with high water content and hot geotherms, both for upper and/or lower MTZ layers, the RF conversion appeared to be interfering with spurious negative conversions that did not seem to belong to any primary phase ( Figures S8, S10, and S12). Conversions beneath a water/melt anomaly with reduced seismic velocities arrived later. This delay became particularly large when the melt was produced and V p /V s increased. For moderate low-velocity anomalies, the conversions arrived still within the shown delay time limit (80 s). For the more extreme (and perhaps less realistic) cases with melt in one of the entire MTZ rock columns (e.g., linear model, 1450 • C geotherm, 100% water saturation, Figure S8), the conversions arrived extremely late (up to 150 s compared to approximately 67 s for the dry case) and interacted with multiple conversions of MTZ discontinuities above the melt layers such as the 410. This theoretical situation is perhaps rather unlikely, as it requires fully water-saturated layers of the MTZ. However, the general problem remains that substantial water/melt and corresponding velocity anomalies may result in extremely delayed conversion times compared to the dry case. These large delay times also carry an opportunity for identifying water in the MTZ and hydrous melts, as the amount of water can be modelled in order to fit observed delay time anomalies in RFs if the data are good enough.
In this study, the RFs were perfectly depth-converted, as we know the exact velocity model, so that the true model depths are recovered. However, in reality, it requires an accurate a priori velocity model (V p and V s ) from other sources to place the conversion at the right depth and place. At MTZ depths, the size of such melt pockets may be easily below resolution of common tomographic techniques which are usually used to perform depth conversions. The resulting, usually over hundreds of km, smoothed models used for depth conversions may therefore result in large depth errors.

Discussion
Our model did not include full thermodynamic calculations (Gibbs minimisation) and was therefore limited to the phase diagrams from the published literature. However, it predicted all main MTZ phase transitions ( Figure 1) within 5 km from the reference depths at 410 km (olivine-wadsleyite), 520 km (wadsleyite-ringwoodite), and 660 km (ringwoodite-bridgmanite) at ambient mantle temperatures. This approach is simplistic, because it does not include zones where two mineral phases may coexist (e.g., olivine and wadsleyite). Experiments showed that phase transitions may occur over the intervals of 4-8 km [87] to 7-13 km [58], which is below the vertical seismic resolution at these depths (e.g., [7,88]). Transformations between olivine polymorphs are fast (tens of hours), so that metastability effects might be minor [89] and negligible for our simulations.
Seismic velocities computed with our code for dry conditions fit the reference Earth models (PREM [90] and ak135-F [91]) within 1%-3% and reproduce the most prominent features such as depths of the 410 and 660 discontinuities (Figure 1). We parameterised the shifts of phase transitions after [72], but we did not include water partitioning between liquid and solid phases nor between the two coexisting olivine polymorphs fields described in this paper. Therefore, our simple calculations tended to overestimate the depths shift of phase transitions in the wet systems.
We presented models of water and hydrous melt anomalies within the MTZ. Indeed, some seismological studies (RFs and precursors) did show signals indicative of low-velocity zones within the MTZ, usually in the lower MTZ (e.g., [37,[92][93][94][95][96][97][98][99]). These anomalies can, in principle, be "reproduced" by our code when placing a water-melt anomaly of the right shape and amplitude at the observed depths.
The influence of melt is a challenging problem since its amount reaches 30% for 100% water saturation for a 1450 • C geotherm. That is close to the minimum values (30%-50% of liquid) of the rheological transition from brittle to viscous behaviour in crystal mushes [100,101]. Therefore, shear moduli and seismic velocities should be sufficiently reduced. The values that we obtained were generally consistent with the results developed by other studies using the Gassmann approach (e.g., [102]). However, our simulations did not include migration and accumulation of melts. This missing element is likely to be a significant factor altering chemical compositions and the thermal and velocity structure in and around the MTZ.
The presence of water expands the stability regions of water-bearing phases such as wadsleyite in the upper mantle. If a relatively small and isolated water-bearing anomaly is located above the 410 discontinuity entirely surrounded by dry mantle pyrolite, a rock "enclave" or "lens" with wadsleyite-bearing lithology may form in the upper mantle. The developed algorithm predicts water-rich lenses well above the MTZ ( Figure 6). Such lenses can maintain wadsleyite-bearing lithologies at depths as shallow as 350 km (in the most extreme case) and may contain melt depending on the actual mantle temperature (Supplementary Materials Figure S14). They result in very prominent low-velocity RF signatures. This case may therefore be relevant to the nature of ultra-low zones shown in aforementioned papers. The long-term stability and formation mechanisms of such lenses in nature are beyond the scope of this paper and require a detailed self-consistent thermodynamic modelling.
The vertical seismic resolution in the MTZ region is typically of several tens of kilometres, the horizontal resolution is of the order of hundreds of kilometres (Fresnel zone); hence, imaging below this scale is not possible. Nevertheless, water-and melt-anomalies may occur at a much smaller scale than assumed in this study which will consequently not be detectable by many seismological techniques. Instead, we assumed large-scale anomalies varying over large wavelengths. Water and melts will more likely distribute in heterogeneous small-scale systems and not homogeneously within a large area. For example, using full waveform scattering imaging, reference [99] showed that the MTZ beneath the US is characterised by a small-scale (<100 km horizontal) low-velocity heterogeneity that possibly represents bodies of subducted, hydrated harzburgite, either in the form of isolated blobs or small-scale density instabilities. The important issue is that these anomalies are too small to be detected by tomographic methods but require high-resolution imaging techniques such as the one proposed by the authors. In many teleseismic images of the MTZ region, these would appear simply as smoothed large-scale bodies with approximately average properties distributed within the target area.

Conclusions
We provide a practical and computationally fast tool to produce seismic models of water-bearing lithologies and hydrous melts in and around the MTZ that can be used to investigate the resulting RF images. Our setup with pyrolithic lithology without water exhibits phase transitions close to the well-known 410, 520, and 660 discontinuities. Water expands stability fields of the water-carrying lithologies by several tens of kilometres.
The model predicts melting in regions with high mantle potential temperature and sufficient amounts of water. Melt-bearing anomalies placed at the depths of observed low-velocity zones produce synthetic RF signatures similar to observations. According to our model, hydrous melting is a necessity to explain clear low-velocity zones observed RFs and precursors. The presence of water only, without the occurrence of melts, merely results in amplitude changes of the 410 or 520 discontinuities, but the corresponding velocity reduction is not enough to produce noticeable negative conversions.
This paper may provide a reference for understanding the origin and nature of hydrous anomalies and interpreting seismological images of the MTZ. However, our tool relies on extremely simple parameterisations and approximations, so a more complex and detailed framework is needed for seismological studies and solution of inversion problems.  Figure S14: S14: Upper mantle "water lens" with maximum water-saturation of 100%.
Author Contributions: I.F. contributed more to the general framework of the developed script and its thermodynamic part, while C.S. worked mostly on its seismic part. The authors contributed nearly equally to the discussion of the results and writing of the paper.
Funding: This research received no external funding.