The role of H₂O in the deformation and microstructural evolution of the upper mantle

Mantle deformation governs the most important large-scale geological processes, such as plate tectonics (Fig. 1), on Earth and in other terrestrial planets because the rheological properties of the mantle have a first-order influence on how this deformation is accommodated at great depth (e.g.Karato, 2010). It is widely accepted that the rheology of subsolidus peridotite is essentially controlled by the crystal plasticity of olivine, which is the most abundant and the weakest mineral phase in all prevalent lithologies in the upper mantle (Karato and Wu, 1993; Bürgmann and Dresen, 2008). However, despite decades of intense research, processes allowing for ductile deformation of the upper mantle remain poorly understood, yet the role of volatiles has been of interest since the earliest experimental works. The reason is that even the smallest quantities of fluids potentially weaken the rheology (see Kohlstedt and Hansen, 2015 for an overview), and – given that a completely dry and refractory upper mantle can be considered unlikely – the response of the shallow mantle to deformation is probably influenced by the distribution of aqueous fluids and hydrous melts¹ in a fundamental way.

The dominant volatiles of the Earth’s upper mantle are species in the C-O-H system, mostly in the form of H₂O and CO₂ molecular species. However, minor CO, CH₄, H₂, N₂, H₂S and SO₂, electrolytic species, as well as halogens and noble gasses can be present locally (Wyllie and Ryabchikov, 2000; Zhang et al., 2009; Klemme and Stalder, 2018; Mysen, 2022). In this chapter, the attention will be focused on phases containing the H₂O component, and we therefore refer to H₂O as a thermodynamic component of the system. The main pathways to supply this hydrous component to the upper mantle are: (1) the direct crystallization of hydrous minerals, (2) the incorporation of hydrogen (more specifically compensated protons) through point defects in the crystal structure of nominally anhydrous minerals (NAMs), and (3) the percolation of aqueous fluids or hydrous melts at mantle depths (Fig. 1). Ultimately, the source and distribution of this H₂O component is largely controlled by subduction zones and dehydration processes in the slab.

The exact role of each of the above processes in controlling the rheology of the upper mantle is, however, controversial. Some of these controversies stem from an unclear identification of the H₂O-bearing phases that might actually control the deformation processes, both in nature and in experiments. From an experimental perspective (e.g.Hirth and Kohlstedt, 2003) the effect of H₂O on the strain rate (⁠$∊․$⁠) for a given stress (σ) in a ‘wet’ rheology is usually accounted for by the water fugacity $fH2O$ term. For example (disregarding the effect of partial melts or oxygen fugacity, Kohlstedt and Hansen, 2015),

where A_w is a constant for the ‘wet’ rheology, n is the exponent of the stress term, d is the grain size with a p exponent, r is the exponent of the H₂O fugacity term, $Ew*$ and $Vw*$ are the activation energy and volume terms under ‘wet’ conditions, respectively (Mei and Kohlstedt 2000a,b) and R is the universal gas constant. While the $fH2O$ term offers a quantitative means to constrain the influence of H₂O on deformation (cf.Eq. 1) – which is highly appreciated, e.g. in geodynamic modelling – it may obscure the underlying physical processes that govern weakening. Moreover, the widespread use of ambiguous terms in the literature – such as water-saturated and water-undersaturated conditions, water activity, ‘hydrous’ deformation, ‘water’ or hydrolytic weakening, and notably, the direct substitution of the $fH2O$ term in Eq. 1 by ‘water’ content in olivine (C_OH, as discussed by Karato and Jung 2003) – contributes to the overall lack of clarity regarding the actual weakening mechanism. This results in a limited significance of its use as a quantitative parameter in flow laws. The recognition that the term $fH2O$ in a fluid-rock system – arbitrarily defined from an intensive variable, the H₂O chemical potential (μH₂O) – does not depend solely on pressure and temperature (as implicitly stated in Eq. 1), but fundamentally relies on the relative proportion of H₂O content in the system under consideration, serves as a cautionary note when applying Eq. 1. In an attempt to clarify these ambiguities, a set of definitions illustrated in simple binary and ternary systems including H₂O as a component is presented below.

The thermodynamic description of a system where chemical reactions – in the present case, hydration, dehydration and hydrous melting – are feasible, requires the definition of a ‘compositional space’. This compositional space constitutes the minimum set of components necessary to adequately describe (or, more accurately, ‘model’) the chemical variability of the system and its constituent phases (Spear, 1995). The definition of the system components is arbitrary, with the only constraint that they need to be linearly independent (i.e. linear combinations among components are not allowed). For example, in pure molecular fluids (i.e. those consisting of species without charge), the choice of including the two components H₂ and O₂ or, alternatively, the single component H₂O as part of the compositional space depends on whether one intends to model or not, respectively, redox reactions among fluid/melt species and solids. Assuming electrical neutrality and ignoring redox reactions, H₂O serves as a convenient thermodynamic component to describe phase relations among H₂O-bearing melts, aqueous fluids, hydrous minerals, and even NAMs as discussed below. The common use of the term ‘water’ to refer to the H₂O component, although colloquial, is not appropriate because (1) ‘thermodynamic components’ are just entities of the compositional space and lack of any physical meaning, and (2) ‘water’ is a phase with a well-defined and limited pressure and temperature stability, which falls well below the conditions commonly encountered in the upper mantle. Therefore, in this book chapter, we prefer to apply the H₂O tag, which should be understood as the colloquial ‘water’ term, widely – and perhaps unfortunately – used in the mantle literature.

There is no consensus in the literature regarding the terminology for expressing the amount of the H₂O component in hydrated minerals like NAMs, where protons typically balance vacancies and associate with oxygen to form hydroxyl groups (OH^–). However, adopting a consistent approach – as employed for nominally hydrous phases such as amphibole and serpentine that also possess structural hydroxyl groups – justifies the use of H₂O as a thermodynamic component for NAMs too. This becomes more evident when considering the representation of hypothetical hydrous endmembers in NAMs, such as Mg₂H₄☐O₄ (defined in terms of MgO and H₂O components) in the case of forsterite, where the empty square indicates a vacancy in the tetrahedral site that satisfied charge balance (Jollands et al., 2023).

The computation of equilibrium phase diagrams enables the definition of the chemical potential of all components in the system (e.g.Powell et al., 2019), including H₂O. From a computational perspective, a phase diagram can be constructed by changing either the ‘quantity’ of the H₂O component (as an extensive variable²), or the H₂O chemical potential (as an intensive variable³). When the amount of the H₂O component is selected as an independent variable, the H₂O chemical potential becomes a dependent variable⁴ with a numerical value that is a function of pressure, temperature and, importantly, phase assemblage. Therefore, from that very moment, the H₂O chemical potential is intrinsically linked to the bulk chemical composition and, in our particular case, the amount of H₂O present in the system in question. From an experimental perspective, the H₂O chemical potential is also a dependent variable unless the experimenter deliberately imposes H₂O fluid saturation, or induces saturation with a fluid containing another non-reacting component in the solution (e.g. Ar; for a comprehensive discussion see the seminal work of Greenwood, 1961). In this latter case, the non-reacting component results in a fixed chemical potential of the H₂O component at the run conditions, which would be lower than the chemical potential of a pure fluid containing only H₂O at the same conditions (i.e. imposing an H₂O activity lower than one). Thus, the H₂O chemical potential is equal to that of pure H₂O – or, in the buffered experiments, the chosen fixed chemical potential of the H₂O component – exclusively in this situation. It is also experimentally feasible to establish a numerical value for the H₂O chemical potential by introducing hydrous solid phases as H₂O-buffers, as discussed by Otsuka and Karato (2011). However, this approach is constrained to discontinuous values, dictated by the buffering effect observed in specific H₂O-undersaturated mineral assemblages. Notably, Otsuka and Karato (2011) investigated such buffering effects in mineral combinations like clinohumite–periclase–forsterite and brucite–periclase (see the following section).

Figure 2 shows a simple MgO-H₂O system computed at constant pressure and temperature conditions to illustrate: (1) the geometric definition of the μH₂O (i.e. the H₂O chemical potential); (2) derive corresponding values for $aH2O$ (i.e. the H₂O activity) and $fH2O$ (i.e. the H₂O fugacity); and (3) show their dependence on X(H₂O) (i.e. the relative amount of H₂O). In this system, there are no solid solutions, and the determination of stable phases involves the computation of their Gibbs energies per mol of components in the system (number of mols of H₂O and MgO in each phase) under the specified pressure and temperature (P-T) conditions. Subsequently, the identification of stable phases is achieved by locating the minimum surface defined by two phases: periclase-brucite and brucite-H₂O in Fig. 2a, and periclase-H₂O in Fig. 2b. Notably, in the latter case, the appearance of brucite above the line connecting periclase-H₂O indicates that the univariant reaction brucite = periclase + H₂O has been crossed at the corresponding P-T conditions, and brucite is no longer stable. Once the stable phases are found, it is possible to compute the two chemical potentials by extending the line that joins the two phases until its intersection with the vertical axis of the two system components (i.e. MgO-H₂O) at X(H₂O) = 0 for the MgO component, and at X(H₂O) = 1 for the H₂O component. By definition, both chemical potentials (μMgO and μH₂O) are equal for each of the individual two-phase assemblages that are stable (e.g. per-br and br-H₂O in Fig. 2a define a total of two different pairs of μH₂O and μMgO). Each two-phase assemblage defines, therefore, a unique μH₂O. In the illustrated example in Fig. 2a, considering the periclase-brucite assemblage, the line joining these two stable phases (solid black line) is extended until it intersects with the vertical axis at X(H₂O) = 1 (dashed violet line). This intersection point occurs at –268,844 J/mol (depicted as a violet dot in Fig. 2a), and it defines the H₂O chemical potential of periclase-brucite $(μH2Oper−br)$ at 2 GPa and 500°C. Note that $μH2Oper−br$ is lower than the chemical potential of pure H₂O (⁠$μH2O○$⁠) at the same conditions, which is situated at –252,003 J/mol and is depicted as a red dot at X(H₂O) = 1 in Fig. 2a. Consequently, all compositions ranging from X(H₂O) = 0 to 0.5 (corresponding to the compositional periclase-brucite joint) are fluid undersaturated and exhibit identical μH₂O. The difference in energy between the pure H₂O at the given P-T conditions – denoted as $μH2O○(P,T)$ and $μH2Oper−br(P,T)$ – defines the H₂O activity (⁠$aH2O$⁠), as given by Eq. 2.

In the specific case depicted in Fig. 2a,

Figure 2b shows the same MgO-H₂O system at lower P (0.01 GPa) and higher T (850°C) conditions, where H₂O saturation is achieved. As evident in the figure, at H₂O saturation, both H₂O chemical potentials – i.e.$μH2O○(P,T)$ and $μH2Oper−H2O(P,T)$⁠, depicted as a red dot at X(H₂O) = 1 – are identical, hence the H₂O activity equals to one (⁠$aH2O=1$⁠), according to Eq. 2.

In contrast to the reference used in the definition of H₂O activity (⁠$aH2O$⁠) at the P-T conditions of interest, the H₂O fugacity (⁠$fH2O$⁠) is defined relative to the H₂O chemical potential at the temperature of interest, yet at a reference pressure of 1 bar, denoted as $μH2O○$ (1 bar, T). In the studied MgO-H₂O system at 2 GPa and 500°C (Fig. 2a), the $fH2O$ can be calculated as shown in Eq. 3.

where $fH2O○$ is the standard state fugacity at the reference pressure and equals 1 bar. Thus, for the H₂O-undersaturated conditions shown in Fig. 2a (at X(H₂O) = 0.0–0.5), we have:

At the H₂O saturation condition shown in Fig. 2b, the $fH2O$ can be computed directly by determining the difference between $μH2O○(P,T)$ and $μH2O○(1 bar,T)$⁠. Importantly, this calculation is solely dependent on the pressure and temperature of the system, and it remains unaffected by X(H₂O) (i.e. for X(H₂O) > 0.5 in Fig. 2a, and for all X(H₂O) values in Fig. 2b). It is crucial to realize that without knowledge of the stability conditions of brucite – specifically without Gibbs energy minimization of the full system – establishing a priori whether fluid saturation conditions are met is not possible. Consequently, $μH2O$⁠, $fH2O$ and $aH2O$ cannot be computed solely based on a chosen P-T condition if this information is lacking. In experimental works, employing Eq. 1 directly, using $fH2O$ from the equation of state (EoS) for pure H₂O at the running pressure and temperature conditions assumes that hydrous phases, hydrous melt or H₂O in NAMs are equilibrated with excess H₂O, and such an approach implicitly considers H₂O saturation.

The aforementioned approach gains greater relevance for a mantle composition when extended to the MgO-SiO₂-H₂O system. This extension enables the exploration of the implications of incorporating solid solutions, such as hydrous melts and NAMs with variable compositions. In this context and for illustrative purposes, the continuous Gibbs energy is piecewise linearized using pseudocomponents (i.e. discrete compositions along the solid solution that are considered as discrete phases), as outlined by Connolly and Kerrick (1987). Solid solution models for NAMs are still under development (Padrón-Navarta, 2019).

For didactic purposes, a simplified example for forsterite is presented in a sandbox environment, considering only protons associated with oxygen in silicon vacancies (Mg₂H₄☐O₄, where the empty square indicates a vacancy; see Jollands et al., 2023 for further details), and referred here to as hydrous forsterite (hfo), a hypothetical endmember. To facilitate a more straightforward analysis in 2D plots, the system can be projected conveniently from enstatite⁵, as illustrated in Fig. 3 (purple arrows in the ternary diagrams showing the direction of projection from enstatite). Despite its simplicity, this system allows the definition of three crucial types of phase relations. At extremely low H₂O concentrations, the only stable phase is hydrous forsterite (and enstatite) at all temperatures, as depicted by the blue region in the compatibility diagrams in Fig. 3. It is noteworthy that in this single-phase field (hfo), there is a continuous increase in H₂O content of hydrous forsterite with increasing X(H₂O) (see inset in Fig. 3a for an enlarged view; the same process applies to Fig. 3b–c), with tangents to its energy ω-surface approximated by two pseudocomponents. In the example, the tangent line shown in Fig. 3a is defined by hfo₁ and hfo₂ (pale blue circles labeled accordingly in the inset of the figure). In any other two-phase assemblage involving hydrous forsterite – i.e. hfo + atg in Fig. 3a, hfo + H₂O in Fig. 3b and hfo + melt in Fig. 3c –, the H₂O content in forsterite does not increase with increasing X(H₂O), and the composition of hfo is fixed. At low temperatures, the hydrous phase antigorite is stable. Antigorite has a lower Gibbs energy than pure H₂O and one of the pseudocomponents from hydrous forsterite (i.e. hfo₃ in Fig. 3a). Thus, it plots in a lower position in the diagram than a metastable line joining them (Fig. 3a). As temperatures increase, antigorite is no longer stable, resulting in a simplified compatibility diagram (Fig. 3b) with only two fields (+enstatite): hydrous forsterite, and hydrous forsterite+H₂O, depicting H₂O-undersaturated and H₂O-saturated conditions, respectively. These regions are divided by one pseudocomponent (equivalent to hfo₃⁶). This scenario is important, as only a relatively small amount of H₂O – greater than the solubility of H₂O in the NAMs – would be enough to reach H₂O saturation, which has significant implications for the interpretation of hydrous deformation experiments.

At the highest temperatures, the hydrous melt surface intersects the H₂O-hfo joint, resulting in four regions. As a consequence, hydrous forsterite is no longer stable with pure H₂O (Fig. 3c). In the hfo + melt region, the H₂O content in olivine is buffered by the hydrous melt and does not increase with the H₂O content of the system, because the system is divariant. This situation may also be relevant for deformation experiments involving a small degree of hydrous partial melting. In such cases, the $fH2O$ cannot be directly computed from the P-T conditions, but the inclusion of a solubility model for H₂O in the melt – dictating the H₂O chemical potential – is needed (cf.Hirth and Kohlstedt, 1996; Asimow et al., 2004).

Figure 4 illustrates a T-X(H₂O) diagram for the entire MgO-SiO₂-H₂O system, maintaining a constant MgO/SiO₂ ratio at a fixed pressure of 2 GPa, within a small range of bulk H₂O contents (0.0–0.4 mol.%; equivalent to 0–1470 ppm wt. H₂O), representing conditions typical of the upper mantle. These simple diagrams bring a novel perspective by examining the impact of H₂O in NAMs on the chemical potential of H₂O (panel 2) and other derived variables under H₂O-undersaturated conditions in a thermodynamically consistent manner, as opposed to iterative calculations (cf.Hirth and Kohlstedt, 1996; Asimow et al., 2004). In the H₂O-undersaturated region, where hydrous phases or hydrous melts are not stable (shown as a semi-transparent field in Fig. 4), the chemical potential depends on the incorporation of H₂O in into the hydrous forsterite, which thus controls all dependent quantities (i.e. H₂O activity and H₂O fugacity). Suffice to say that experiments aimed at investigating the rheological effect of H₂O incorporated in the structure of olivine should be conducted specifically within this T(P) and X(H₂O) window. Unfortunately, this window is poorly constrained in natural systems, primarily due to the dependence of H₂O incorporation into NAMs on various factors, such as trace element concentration and oxygen fugacity (e.g.Tommasi et al., 2017). It has been suggested that the correlation of the strain rate with H₂O fugacity can be used to support the equivalent effect of H₂O concentration in NAMs on strain rate (e.g.Hirth and Kohlstedt, 2003; Karato and Jung, 2003; Kohlstedt and Hansen, 2015). The $fH2O$ is very sensitive to changes in P and T under H₂O-saturated conditions; therefore, the fact that this dependency is identified when considering data at different P-T conditions does not, by itself, demonstrate that the dependency of $fH2O$ (or H₂O content in olivine) can be quantified (but see Faul et al., 2016).

In the following example, we disregard the effect of solid solution for H₂O in NAMs (as these models are not yet fully available; Padrón-Navarta, 2019), and we focus on the Na₂O-CaO-FeO-MgO-Al₂O₃-SiO₂-H₂O-Cr₂O₃ (NCFMASHCr) system to explore processes that take place in a more realistic mantle composition. Among these components, Na₂O is relevant for the stability of pargasitic amphiboles that – together with NCFMASH melts – exert a primary influence on the H₂O chemical potential. Note that the K₂O component – although not considered in the system discussed in this section – is also important in other settings (see Fig. 1) for the stability of phlogopite and K-richerite amphibole, which are stable at depths >180 km (e.g.Konzett and Ulmer, 1999). Figure 5 illustrates that regions saturated in a free H₂O fluid phase (fields with water pattern, where the H₂O activity is = 1 as shown in Fig. 5b,e) become wider at lower pressures. Only in these regions do the H₂O fugacity, the modal content of hydrous phases, and the melt fraction remain independent of the amount of H₂O in the system, as evidenced by the solid horizontal lines in Fig. 5. For the low bulk H₂O contents (0.0–1.0 wt.% H₂O) considered in Fig. 5, which are feasible in various settings in the shallow upper mantle (Fig. 1), most phase assemblages are H₂O-undersaturated with either hydrous melts at high temperature or hydrous phases – such as amphibole or chlorite – at low temperatures. In complex multicomponent systems, these H₂O-undersaturated fields containing hydrous phases or hydrous melts also constrain the values of the H₂O activity and H₂O fugacity, as is done in simpler systems (Fig. 3a, c).

However, due to the existence of many possible continuous multivariant reactions between all phases (e.g. tremolite content in amphibole, anorthite content in plagioclase, chromium in spinel, etc.), the values of H₂O activity/fugacity do not vary linearly with the relative amount of H₂O in the system, as opposed to simple systems (cf.Fig. 4) because they depend on the amount of hydrous phases or hydrous melts at H₂O-undersaturated conditions (i.e. lower half or upper half of the diagrams, respectively). We emphasize that under these H₂O undersaturated conditions, introducing the H₂O fugacity term in rheological constitutive laws (Eq. 1) is unlikely to have any physical meaning without considering all the variables that may influence it, particularly the total H₂O content in the system at the relevant P-T conditions of deformation. The H₂O fugacity can only be computed independently from the chemical complexities of the rock in situations of H₂O saturation, which can only be achieved for realistic low H₂O contents when neither hydrous phases nor hydrous melts are stable, and the H₂O content in the system exceeds that required for H₂O saturation of NAMs (Fig. 4). In any case, it is highly probable that under those circumstances, the rheological behaviour would be more influenced by the presence of a stable interstitial free fluid phase than by the numerical value of the H₂O fugacity, or even the actual H₂O content in the NAMs.

These observations suggest the need for further research focused on a better understanding of the conditions in the upper mantle that control the stability of hydrous minerals, as well as the distribution and composition of aqueous fluids and hydrous melts. Additionally, the solubility of H₂O in NAMs (particularly for pyroxenes with mantle composition) and its dependency on other factors, such as trace elements and eventually fO₂, need to be further constrained. This is important because under natural conditions, their impact on rheology may be major and, perhaps, different than previously thought. Moreover, in future constitutive rheological laws accounting for the effect of hydrous components, it is essential to differentiate clearly between direct measurements of extensive variables (i.e.X(H₂O) in the experimental system, as illustrated in Figs 2–5), and dependent intensive parameters (such as H₂O fugacity), the latter being a complex function of the reactive deforming system.

In recent decades, many efforts have been made to obtain quantitative experimental data on the rheological behaviour of the upper mantle. Below we review the major achievements that – at least marginally – addressed the potential role of H₂O components presented in Fig. 1, and we summarize what the numerous recent natural case studies demonstrate when these components are inferred to be at play during ductile deformation of the upper mantle.

In contrast to anhydrous melts (e.g.Hirth and Kohlstedt, 1995a,b; Kohlstedt and Zimmerman, 1996; Bai et al., 1997; Zimmerman et al., 1999; Holtzman et al., 2003a,b; Zimmerman and Kohlstedt, 2004; Scott and Kohlstedt, 2006; Takei and Holtzman, 2009a,b,c; King et al., 2010; Kohlstedt et al., 2010; Soustelle et al., 2014), there are very few experiments on the mechanical behaviour of hydrous melts during deformation, but available data indicate that the presence of H₂O enhances strain rates by approximately two orders of magnitude compared to equivalent anhydrous systems (Mei et al., 2002). Of particular importance in hydrous melts is that they are common metasomatic agents in the upper mantle, which contribute to the generation of chemical and mineralogical heterogeneities (O’Reilly and Griffin, 2013 and references therein). Amphibole and mica are well-known products of such melt percolation in a wide depth range and in various tectonic settings (e.g.Dawson and Smith, 1982; Vannucci et al., 1995; Zanetti et al., 1996; Coltorti et al., 2007; Shaw, 2009; Fumagalli and Klemme, 2015; Mandler and Grove, 2016; Bonadiman et al., 2021; Wang et al., 2021). Moreover, amphibole is the most important H₂O-storing mineral in the upper mantle down to ∼3 GPa, ∼90–100 km depth in an oceanic environment (Green et al., 2010; Green, 2015), while phlogopite – subject to sufficient potassium – plays a similar role to somewhat deeper levels (Mengel and Green, 1989; Wyllie and Ryabchikov, 2000). The breakdown of amphibole causes a remarkable drop in the solidus temperature of fertile lherzolites and, at the same time, reduces the H₂O-storage capacity of the mantle, which is then controlled only by the NAMs under subsolidus conditions at greater depths (Green et al., 2010; Kovács et al., 2021). Although these results demonstrate that the presence of hydrous minerals controls the H₂O-storage capacity of the mantle at lithospheric mantle depths, to date, natural observations are scarce and no experimental data are available to constrain their implication for the rheology while they are present as stable phases in an otherwise nominally anhydrous, four-phase peridotite. Tommasi et al. (2017) provided one of the first microstructural observations from coarse-grained pargasite- and phlogopite-bearing peridotites, which record pervasive ductile deformation in the dislocation creep regime coeval to percolation of hydrous Si-rich melts at mantle depths (980–1080°C and <2 GPa) in the Finero massif (southern Alps, Italy). Their study concurs with the available experimental data (Mei et al., 2002) and shows that the synkinematic presence of interstitial hydrous melts leads to significant rheological weakening of mantle rocks by accommodating large amounts of deformation due to a switch of the deformation mechanisms to stress-controlled dissolution-precipitation creep. However, the lack of correlation of olivine fabric strength and the symmetry of crystallographic-preferred orientation (CPO) – or lattice-preferred orientation (LPO) – of olivine to the occurrence of hydrous minerals (pargasite and/or phlogopite) in the peridotites implies little or no direct effect of these phases on the microstructural evolution of mantle rocks (Fig. 6). Even when hydrous minerals are present in peridotites at modal abundances of up to 25 vol.%, the fabric strength remains relatively weak in the peridotites (Fig. 6). This observation indicates that the mere presence of stable hydrous minerals in peridotites during deformation does not exert a significant influence on the rheological behaviour of the upper mantle (Tommasi et al., 2017). It is conceivable that, as long as hydrous minerals are stable, rather than simply present, it is the saturation of rocks in H₂O and/or hydrous melts that controls the rheology of the mantle. Moreover, it is important to note that hydrous metasomatism does not consistently influence the H₂O content of the coexisting NAMs. In cases where hydrous minerals are present, the NAMs – particularly olivine – tend to remain relatively dry or only moderately saturated in H₂O (e.g.Denis et al., 2015; Hao et al., 2016; Aradi et al., 2017; Tommasi et al., 2017; Demouchy and Tommasi, 2021).

The weakening effect of the hydroxyl concentration in NAMs has been known since the earliest experiments on quartz crystals in the 1960s (Griggs and Blacic, 1965; Griggs, 1967; Blacic, 1975), where the decrease in strength of the mineral was observed with increasing hydrogen (proton) concentration and explained by easier movement of dislocations in the wet crystalline lattice. Subsequently, the potential impact of H₂O-linked weakening on mantle rheology has been proposed through experiments on olivine (Avé Lallemant and Carter, 1970) and pyroxene (Avé Lallemant, 1978). Later studies have been focused on the overview of the weakening mechanisms in major constituent phases of mantle peridotites and pyroxenites (e.g.Kohlstedt et al., 1995; Mei and Kohlstedt, 2000a,b; Karato and Jung, 2003; Chen et al., 2006; Tielke et al., 2017) and resulted in the widespread acceptance that, in continental environments, the rheology of the lithospheric mantle, in general, and the crust-mantle coupling, in particular, may be controlled by the behaviour of olivine, with variable concentrations of hydrous defects. In the last decade, however, low-T (800–1100°C) dry deformation experiments on olivine challenged this paradigm (Demouchy et al., 2013, 2014). These experiments demonstrated that the inferred important weakening effect of hydrous defects is partly due to the large error in the extrapolation of high-T (.1200°C) deformation experiments to natural conditions prevalent in the shallow lithospheric mantle (∼900–1000°C), and so the rheology of dry olivine is indeed significantly weaker at these low-T conditions than previously thought. Thus, numerical modelling of lithospheric strength profiles may not need to operate with wet olivine rheology to weaken the lithospheric mantle (Boioli et al., 2015). Currently, there is a general consensus within the scientific community that hydrous defects incorporated into the olivine structure have a discernible weakening effect, at least up to ∼18 ppm wt. H₂O concentrations (Girard et al., 2013; Faul et al., 2016). However, it is also acknowledged that the significance of this weakening effect may have been somewhat overestimated in both diffusion (Fei et al., 2013) and dislocation creep regimes (Demouchy et al., 2012; Girard et al., 2013).

As for the microstructural record of deformation at upper mantle conditions, CPO in olivine is often, although not exclusively, induced by deformation flow that activates different dislocation slip systems (e.g.Karato, 2008). The resulting CPO patterns are potentially important tracers in seismogenic modelling of the Earth’s interior (Tommasi et al., 1999; Mainprice et al., 2000; Karato et al., 2008; Hansen et al., 2021). Natural record shows that the most common CPO symmetries in mantle rocks – developed by simple shear deformation in the dislocation creep regime – are characterized by the [100] olivine crystallographic axes distributed subparallel to the deformation flow direction and the {0kl} planes around the shear planes, with preferential alignment of (010) planes in the plane of the foliation (e.g.Tommasi et al., 2000; Tommasi and Vauchez, 2015; Demouchy et al., 2023). These CPO symmetries are explained by the dominant activation of {0kl}[100] and (010)[100] dislocation slip systems, and are often referred to as D-type (also known as axial-[100], [100]-fibre, or olivine ‘pencil-glide’) and A-type olivine fabric in the literature, respectively. Deformation experiments have suggested many microscopic and macroscopic factors to control the olivine CPO patterns in the dislocation creep regime but, among the intrinsic processes acting at the intracrystalline level, the biggest emphasis was put on the role of temperature, confining pressure, differential stress and structurally bonded hydrous defects (e.g.Carter and Avé Lallemant, 1970; Durham and Goetze, 1977; Bai et al., 1991; Jung and Karato, 2001; Couvy et al., 2004; Katayama et al., 2004; Raterron et al., 2009, 2011, 2012; Tasaka et al., 2016). In the experimental works of the early 2000s, uncommon olivine fabric types were revealed, which indicated [001] slip direction and/or switches in the common slip plane to (001) or (100). The resulting CPO symmetries were consistent with the dominance of (001) [100], (010)[001] or (100)[001] slip (E-, B- and C-type olivine fabric, respectively), and were suggested to be controlled by the differential stress and the concentration of structurally bonded hydrous defects during deformation (Fig. 7a) (Jung and Karato, 2001; Katayama et al., 2004; Jung et al., 2006; Karato, 2008; Karato et al., 2008). These observations have been used widely to interpret a seismic anisotropy signal in the mantle wedge (e.g.Mizukami et al., 2004; Kneller et al., 2005; Katayama and Karato, 2006; Lassak et al., 2006; Tasaka et al., 2008) as well as to render deformation conditions, particularly H₂O content and differential stress, to peridotites displaying a specific olivine CPO symmetry (e.g.Karato, 2004). However, considering the saturation of olivine in H₂O at the P-T conditions of experiments (Fig. 7b), large amounts of aqueous fluids were probably present during deformation, which is also corroborated by the published Fourier-Transform Infrared (FTIR) spectra displaying broad bands instead of sharp OH-stretching related peaks (e.g.Jung et al., 2006). This suggests that olivine crystals contained the possible maximum of structurally bonded hydrous defects during most of the deformation experiments (Fig. 7b). Hence, apart from a few experiments where olivine may not have reached the saturation (e.g. data points in the yellow-shaded field in Fig. 7b), the dependence of olivine CPO types on the concentration of structurally bonded hydrous defects remains ambiguous.

The lack of correlation between structurally bonded H₂O and olivine CPO symmetry is consistent with the results of a recent comprehensive study of olivine CPO types as a function of H₂O content and stresses in natural mantle samples (Bernard et al., 2019). Those authors found that all olivine CPO types were able to form at very low H₂O contents and stresses, and there was no clear relationship between hydrous components in the NAMs and CPO types in natural rocks. The only agreement with experimental results was that A-type olivine CPO symmetry did not occur at H₂O contents above the experimental A-to-E type boundary, and C-type fabric recorded the highest H₂O concentrations (Bernard et al., 2019). However, in the case of the C-type fabric, the effect of pressure in the activation of [001]-directed olivine slip could not be excluded because the corresponding peridotites were garnet-bearing, hence originated at higher-pressure conditions (Bernard et al., 2019). Although structurally bonded hydrous defects in the olivine crystalline lattice possibly have an impact on slip system activity (e.g.Tasaka et al., 2016; Faul et al., 2016; Wallis et al., 2019) by altering dislocation self-energy in an anisotropic manner (e.g.Karato et al., 2008; Karato, 2013), the natural record indicates that the role of other factors (e.g. deformation history, interstitial fluids/melts) in the development of uncommon olivine CPO symmetries in mantle peridotites is significantly larger than that of H₂O concentration and differential stress (e.g.Bernard and Behr, 2017; Bernard et al., 2019; Kumamoto et al., 2019). Alternatively, instead of the absolute concentration of hydrous defects bonded in the crystalline structure, it is the incorporation site that may control the microstructural evolution and weakening of H₂O-bearing olivine (Padrón-Navarta et al., 2014, Padrón-Navarta and Hermann, 2017). Recent experimental data are consistent with this hypothesis, suggesting that the type of hydrous defects in olivine and the incorporation site of protons in the crystalline structure coupled to certain trace elements – such as Ti – may influence the hydrolytic weakening of olivine. This influence is pivotal as it ultimately controls the absolute concentration of H₂O under lithospheric conditions when hydrous phases or hydrous melts are not stable (Berry et al., 2005, 2007a,b; Faul et al., 2016). The future direction of research in the field of hydrous weakening of mantle NAMs and the variation of olivine CPO types is thus expected to explore the dependence of rheological behaviour on the incorporation site of structurally bonded protons under truly H₂O-undersaturated conditions.

Experimental data addressing the effect of synkinematic interstitial aqueous fluids on the rheological behaviour of mantle rocks are scarce. Torsion experiments at 300 MPa and 1200°C on wet olivine polycrystals showed a limited strength decrease by a factor of 2–3 relative to dry samples, which is explained by the presence of both structurally bonded hydrous defects in the olivine structure and importantly by H₂O-derived species in grain boundaries or pores (Demouchy et al., 2012). On the distribution of fluids during ductile deformation, recent high-pressure experiments conducted at 1.2 GPa and 900°C on olivine-pyroxene aggregates have shown that large amounts of aqueous fluids can be concentrated in domains experiencing the highest strain (Précigout et al., 2019). This observation suggests that deformation may enhance fluid pumping into ductile shear zones in the mantle (Précigout et al., 2019), which is primarily driven by dynamically changing pore-fluid pressure at the grain scale during creep cavitation (e.g.Fusseis et al., 2009). This potentially has an important feedback on generating shear instabilities, because if melts/fluids can form continuous films at grain boundaries, effective pressure is probably reduced significantly (e.g.Karato, 2008). However, unlike interstitial melts, at pressures relevant to the shallow subcontinental upper mantle, H₂O-rich fluids have a large dihedral angle (≥60°) at triple junctions between mineral grains (e.g.Watson and Brenan, 1987; Mibe et al., 1999), which impedes efficient wetting of grain boundaries and promotes the formation of fluid pockets instead of interconnected layers, except for very large pore-fluid volumes. Nevertheless, both deformation (e.g.Hier-Majumder and Kohlstedt, 2006) and impurities dissolved in the aqueous fluids (e.g.Huang et al., 2019; Mysen, 2022) tend to decrease the size of dihedral angles, leading to the formation of interconnected networks of aqueous fluids so that they may reach the base of the crust. If that is the case, the aqueous fluid-saturated grain boundaries – similar to those wetted by interstitial melts (e.g.Bai et al., 1997 and references therein) – are expected to serve as a fast intergranular diffusion path. Therefore, free interstitial aqueous fluids may foster the shift of deformation from grain interiors to grain boundaries, favouring fluid-assisted grain boundary migration (e.g.Urai et al., 1986) and/or dissolution-precipitation (pressure-solution) creep, which involves advection of elements by the fluid in response to applied stresses (e.g.Rutter, 1976; Spiers et al., 2004). This latter deformation mechanism is common in the crust (e.g.Wintsch and Yi, 2002; Bestmann et al., 2004) and has also been described in various rock types exhumed from subduction zones (e.g.Wassmann and Stöckhert, 2013), but is usually considered to be of minor importance in the deformation of peridotites due to the limited availability of free aqueous fluids in the low-porosity mantle. Nonetheless, experimental data on the solubility of enstatite and forsterite in aqueous fluids at deep lithospheric conditions (700–900°C and 0.4–1.5 GPa; Newton and Manning, 2002) show that silica solubility is sufficient for dissolution-precipitation creep to be efficient during disequilibrium conditions if time-integrated fluid-rock ratios are high. Thus, if free, impure aqueous fluids are present, it is conceivable that dissolution-precipitation creep could equally be a potential deformation mechanism in the shallow lithospheric mantle.

The microstructural record of aqueous fluid-assisted deformation is rare in fresh (i.e. non-serpentinized) mantle peridotites. In a plagioclase-facies subcontinental lithospheric mantle peridotite shear zone from the Ronda massif (Spain), Hidas et al. (2016) interpreted the progressive strain localization in the presence of synkinematic aqueous fluids during deformation. The shear zone microstructures (e.g. crystallization of secondary orthopyroxene in the pressure shadow of porphyroclasts, Fig. 8a) and major element geochemical compositions (e.g. no compositional difference between shear-zone rocks and host peridotite) were consistent with dissolution-precipitation creep in a melt-free system. This process is inferred to be responsible for the formation of olivine-rich mylonitic domains alternating with orthopyroxene-rich mylonitic-ultramylonitic bands (Fig. 8a), leaving the modal composition essentially intact at the bulk rock scale, compared to country host peridotites. Moreover, the olivine CPO symmetry recorded a switch from the country host peridotite (axial-[100] or D-type olivine fabric) to the H₂O-impregnated mylonites (so-called E-type olivine fabric) coupled to secondary orthopyroxenes mimicking the olivine CPO within the shear zone (Fig. 8b). This latter observation is inconsistent with dislocation creep deformation of orthopyroxene and indicates oriented crystallization controlled by olivine CPO (Hidas et al., 2016). The origin of the unusual E-type olivine fabric in the plagioclase facies peridotites, however, remained unclear due to the low solubility of hydrous defects in olivine at shallow mantle conditions (<1 GPa; Padrón-Navarta and Hermann, 2017). Nevertheless, from a fossil mantle wedge environment exhumed in the Semail Ophiolite, Oman (Prigent et al., 2018), and from an oceanic transform fault setting (Bickert et al., 2023), similar observations were revealed between porphyroclastic tectonites and intermediate-T (700– 900°C) mylonitic-ultramylonitic peridotites, where olivine displayed a remarkable switch to (001)[100] CPO symmetry (i.e. E-type olivine fabric) associated with synkinematic oriented crystallization of secondary pyroxenes, mimicking the olivine CPO. Thus, the natural record shows that ductile shear zones can channelize fluid flow in the upper mantle (e.g.Précigout et al., 2017, 2019), eventually promoting dissolution-precipitation creep assisted by siliceous fluids (Hidas et al., 2016; Bickert et al., 2023).

The conditions at which dissolution-precipitation may be active in the lithospheric mantle, however, are very limited. This process requires the presence of significant amounts of impure aqueous pore fluids in unaltered mantle, which could be achieved locally in a domain constrained by the upper limit of serpentine and chlorite stability (∼750–820°C; e.g.Vissers et al., 1995; Padrón-Navarta et al., 2010 and dashed line #2 in Fig. 5a,d in this chapter) and the onset of hydrous melting (<1000–1100°C; Green, 1973 and lower limit of free melt domain in Fig. 5a,d in this chapter). So far, the natural record documenting dissolution-precipitation creep in non-altered peridotites is scarce yet these data show that it can be a viable deformation mechanism in the mantle wedge of subduction zones (∼700–900°C; Prigent et al., 2018), at oceanic transform faults (∼735–945°C; Bickert et al., 2023), and in the shallow subcontinental lithospheric mantle (∼750–1000°C; Hidas et al., 2016) (cf.Fig. 1), allowing for deformation of peridotites at lower stresses than dislocation creep. If active, dissolution-precipitation creep would imply feedback between deformation and fluid transport and, hence, would promote highly heterogeneous deformation in the shallow upper mantle in diverse tectonic settings. Association of highly localized deformation and the very weak olivine CPO produced by dissolution-precipitation creep may also provide an explanation for the very weak seismic anisotropy usually recorded in the fore-arc domains of subduction zones (delay times <0.5 s; e.g.Di Leo et al., 2012; Long and Silver, 2008). Altogether, fluid-assisted dissolution-precipitation creep could also be a major rheological process controlling strain localization and faulting at plate boundaries in general (Bickert et al., 2023).

At typical tectonic strain rates (10^–12 to 10^–15 s^–1), unweathered mantle rocks transition progressively from ductile to semi-brittle behaviour as temperature decreases below ∼850–900°C, and they primarily deform by brittle mechanisms below 600°C (e.g.Boettcher et al., 2007; Kohli and Warren, 2019; Prigent et al., 2020). In oceanic peridotites, extensive shearing at these low-T conditions – yet occurring beyond the serpentinite stability – often produces ultramylonitic shear zones with remarkably fine-grained (<10 µm) minerals, in which strain is predominantly accommodated by grain-size sensitive deformation mechanisms (e.g.Jaroslow et al., 1996; Warren and Hirth, 2006). Nevertheless, recent microstructural, petrological and geochemical data indicate that at mid-ocean ridges, the hydrologic cycle might extend past the brittle-ductile transition, and substantial amounts of fluids can reach deep levels of the mantle lithosphere along transform faults (e.g.Harigane et al., 2019; Kohli and Warren, 2020; Prigent et al., 2020; Vieira-Duarte et al., 2020; Patterson et al., 2021; Kakihata et al., 2022). Besides ample microstructural evidence for brittle deformation, such as crosscutting intra- and transgranular fractures, as well as pulled apart and/or displaced microboudinage structures, these hydrated low-T, non-serpentinized ultramylonites are accompanied by the synkinematic formation of hydrous phases, such as amphibole ± chlorite (e.g.Prigent et al., 2020; Vieira-Duarte et al., 2020; Kakihata et al., 2022), and occasionally they may exhibit weak to moderately developed CPO patterns. In such cases, olivine is observed with weak bimodal (010)[001] and (001)[100] CPO symmetries, potentially indicating that low-temperature plasticity was at play under hydrous conditions at high-stress (Kakihata et al., 2022).

At even lower temperatures, provided that H₂O fluid supply is sufficient, and depending on the Si content of the ultramafic rocks (e.g.Früh-Green et al., 2004), infiltration along microfractures of the brittle peridotite matrix initiates serpentinization processes, which start with the development of typical mesh textures and culminate in the formation of pure serpentinites at the expense of peridotites (e.g.Boudier et al., 2010; Evans et al., 2013; Schwartz et al., 2013; Rouméjon and Cannat, 2014; Rouméjon et al., 2015; Escario et al., 2018). In general, the degree of serpentinization weakens peridotites significantly, and it has been shown that slightly serpentinized peridotites have a rheological behaviour similar to that of pure serpentinites (e.g.Escartín et al., 2001). Given the high strength of unweathered mantle minerals at low temperatures, serpentinization is expected to promote strain localization in the alteration products, facilitating the development of major (semi-)brittle faults in both subduction zones and oceanic settings. Furthermore, if interstitial fluids accompany such low-T deformation, the deformation is likely to be increasingly accommodated by dissolution and precipitation processes, particularly at the plate interface in subduction zones (e.g.Wassmann et al., 2011; Padrón-Navarta et al., 2012; Wassmann and Stöckhert, 2013).

In this contribution, we briefly reviewed the effect of different H₂O components on the rheology and microstructural evolution of the upper mantle by contrasting theoretical considerations with available experimental data and the natural record. We recalled that H₂O activity (⁠$aH2O$⁠), H₂O fugacity (⁠$fH2O$⁠), and H₂O chemical potential (μH₂O) are all related terms but that only μH₂O has a true thermodynamic meaning as it emerges once the stable phases are determined through Gibbs energy minimization. Through thermodynamic calculations in both simple and chemically complex systems, we emphasize that H₂O activity is equal to one only when the system is saturated in H₂O, and this is the only situation in which H₂O fugacity can be computed at any pressure and temperature conditions independently of the bulk amount of H₂O present in the deforming rock. However, under H₂O-saturated conditions, it is more likely that the rheological behaviour of any rock is governed by grain boundary processes rather than the potentially weakening effect of intracrystalline H₂O or the numerical value of $fH2O$⁠. Furthermore, available laboratory data and natural observations suggest that:

• The formation of hydrous minerals is very dependent on the whole-rock geochemical composition (e.g. availability of Na, K, etc.), as well as the prevailing equilibrium pressure and temperature conditions. Consequently, the presence of only small modal amounts (<1–2 vol.%) of these minerals in refractory peridotites at lithospheric mantle conditions is entirely compatible with the percolation of substantial quantities of interstitial aqueous fluids within the rock, as fluids in excess may not lead to an increased crystallization of hydrous phases if other physicochemical conditions in the system are unfavourable. Nevertheless, compared to the dry mantle, the presence of small modal amounts of amphibole indeed implies greater H₂O fugacity and, therefore, larger H₂O contents in NAMs. However, there are situations, such as during mantle exhumation or low-pressure mantle metasomatism, where this expectation is not met. This discrepancy is primarily due to the combination of the low-pressure conditions and, particularly in the case of olivine, the complex behaviour of trace elements in the presence of hydrous phases that limit the potential for the incorporation of extrinsic hydrous defects in the crystalline structure. Overall, small modal amounts of stable hydrous phases and small H₂O contents in coexisting NAMs in refractory peridotites of the lithospheric mantle cannot be taken as direct evidence of a dry mantle environment.

• Stable, coarse-grained hydrous minerals, such as amphibole and mica, control the H₂O storage capacity of the upper mantle, but their presence alone has little or no impact on the microstructural evolution of their host peridotite during deformation in the dislocation creep regime. Based on natural samples, this observation seems to apply to relatively large modal amounts of such hydrous phases, at least up to 25 vol.%.

• Structurally bonded hydrous defects might weaken rheology, but olivine may not need to be wet to effectively weaken the rheology of the shallow upper mantle.

• Predicted correlation of experimentally deformed olivine CPO-types with H₂O content and stress may not directly apply to nature, where other factors – such as temperature, pressure, deformation mechanism, history and geometry of deformation, strain magnitude and presence of interstitial fluids/melts – play a more significant role in controlling CPO symmetry.

• Synkinematic hydrous or anhydrous melts and aqueous fluids significantly weaken the rheology but their effect is transient, i.e. deformation probably remains active as long as a liquid phase is present and it may cease under liquid-absent conditions. This indicates that: (1) in certain tectonic environments (e.g. subduction zones, mid-ocean ridges, continental rifts), response of mantle rocks to deformation may be subject to the availability of melts/fluids and, as such, can rapidly change at the geological time-scale; and (2) in these settings the relatively short intermittent periods of active, fluid/melt-assisted deformation may be responsible for the permanent microstructural changes rather than the long periods of tectonic inactivity of the otherwise resistant mantle.

• Interstitial, free aqueous fluids promote potential switches in the dominant deformation slip systems and/or mechanisms, and dissolution-precipitation creep may be a viable deformation mechanism if specific physico-chemical conditions meet in the shallow upper mantle. The presence of synkinematic aqueous fluids, however, is difficult to recognize in most cases because such fluids may not leave an obvious trace in the rock and determining them requires detailed microstructural, petrological and geochemical analyses.

• At low temperatures, but above the stability field of serpentinites, deformation of the upper mantle in the presence of synkinematic H₂O components is progressively dominated by grainsize-sensitive deformation mechanisms, leading to the development of brittle-ductile shear zones that often contain hydrous phases (typically amphibole±chlorite). Reaching the temperatures of serpentinite stability, H₂O-assisted deformation is essentially controlled by serpentinite minerals, which develop within a resistant matrix composed of original mantle minerals, promoting (semi-)brittle faults and brittle-ductile strain localization at various scales.

We conclude that further experiments and carefully documented natural studies are needed for a better understanding of the effect of the H₂O component on upper mantle rheology. In the future, the role of hydrous volatile species on the microstructural evolution of the mantle must be explored through thermodynamically controlled models, which take into account the availability, speciation, concentration and incorporation into fluids (?) of a wide range of volatile species at different levels of the Earth’s deep interior.

The authors appreciate the constructive reviews by Marguerite Godard and Alberto Zanetti, as well as the kind assistance and careful editorial work of Costanza Bonadiman and Elisabetta Rampone throughout the writing of this manuscript. José Alberto Padrón- Navarta acknowledges a Ramón y Cajal fellowship (RYC2018-024363-I) funded by the Spanish Ministry of Science and Innovation (MICINN/AEI/10.13039/501100011033) and the FSE program ‘FSE invierte en tu futuro’, and support by the Junta de Andalucía research project ProyExcel 00757. Károly Hidas acknowledges funding by the Agencia Estatal de Investigación of the MICINN (PID2020-119651RB-I00/AEI/10.13039/501100011033).