A synthesis and meta-analysis of the Fe chemistry of serpentinites and serpentine minerals

Abstract

The iron chemistry of serpentinites and serpentine group minerals is often invoked as a record of the setting and conditions of serpentinization because Fe behaviour is influenced by reaction conditions. Iron can be partitioned into a variety of secondary mineral phases and undergo variable extents of oxidation and/or reduction during serpentinization. This behaviour influences geophysical, geochemical and biological aspects of serpentinizing systems and, more broadly, earth systems. Iron chemistry of serpentinites and serpentines is frequently analysed and reported for single systems. Interpretations of the controls on, and the implications of, Fe behaviour drawn from a single system are often widely extrapolated. There is a wealth of serpentinite/serpentine chemical composition data available in the literature. Consequently, compilation of a database including potential predictors of Fe behaviour and measures of Fe chemistry enables systematic investigation of trends in Fe behaviour across a variety of systems and conditions. The database presented here contains approximately 2000 individual data points including both bulk rock and serpentine mineral geochemical data which are paired whenever possible. Measures of total Fe and Fe oxidation state, which are more limited, are compiled with characteristics of the systems from which they were sampled. Observations of trends in Fe chemistry in serpentinites and serpentines across the variety of geologic systems and parameters will aid in verifying and strengthening interpretations made on the basis of Fe chemistry.

This article is part of a discussion meeting issue ‘Serpentinite in the Earth system’.

1. Introduction

The reaction of water with peridotites of the ocean crust and shallow mantle results in the alteration and hydration of those ultramafic rocks. Serpentinization of olivine and pyroxene produces secondary mineral phases with characteristics that differ from those of the primary minerals and by-products such as H₂ gas; thus affecting geochemical processes, such as elemental cycling, geophysical properties of the rocks and geobiological processes. Serpentinites occur in a variety of geologic and tectonic settings and the chemistry of serpentinites is thought to record the setting and conditions of serpentinization even as those change through geologic time [1].

The behaviour of iron in these systems is particularly dynamic as Fe is variably oxidized prior to incorporation into a variety of possible secondary mineral phases. Observations of the partitioning of Fe into secondary mineral phases, and its oxidation state in those phases, are interpreted to reflect reaction conditions and are often used to draw conclusions about the serpentinization process (e.g. [2–9]). Regardless of its mineral host, oxidation of iron can lead to the production of hydrogen gas which can fuel subsurface and surface chemolithotrophic biospheres (e.g. [10–12]). The generation of H₂ may also facilitate production of CH₄ and other abiotic organic compounds (e.g. [13–19]) thus affecting the geochemical cycling and sequestration of carbon. As such, it has been suggested that the origin of life on Earth may have occurred in a serpentinizing system (e.g. [20–22]). Signatures of Fe-bearing mineral phases detected on other planets are interpreted with these implications in mind (e.g. [23,24]). With such wide-ranging implications, it is important to understand how and why iron behaves the way it does in serpentinizing systems.

The formation of serpentine minerals is a prerequisite of the serpentinization process. Typically, the serpentinization process is represented as written in reaction (1.1) (e.g. [25]).

$$\underset{\text{Olivine}\hspace{0.17em}\text{F}{\text{o}}_{90}}{\text{M}{\text{g}}_{1.8}\text{F}{\text{e}}_{0.2}\text{Si}{\text{O}}_4}+\hspace{0.17em}1.37{\text{H}}_2\text{O}\to \underset{\text{Serpentine}}{0.5\text{M}{\text{g}}_3\text{S}{\text{i}}_2{\text{O}}_5{(\text{OH})}_4}+\underset{\text{Brucite}}{0.3\text{Mg}{(\text{OH})}_2}+\underset{\text{Magnetite}}{0.067\text{F}{\text{e}}_3{\text{O}}_4}+0.067{\text{H}}_2$$ 1.1

However, Fe can partition into serpentine, both as Fe(II) and Fe(III), and into brucite, a hydroxide mineral, as Fe(II). The substitution of Fe(III) into serpentine can be accomplished through a variety of possible mechanisms (e.g. [26,27]). Reaction (1.2) (adapted from [27,28]) depicts a ferri-tschermaks (or cronstedtite) substitution (coupled substitution of Fe(III) onto both the octahedral and tetrahedral sites), a common substitution mechanism:

$$\begin{array}{rl}& \underset{\text{olivine}}{{(\text{Mg},\text{F}{\text{e}}^{2+})}_2\text{Si}{\text{O}}_4}+a{\text{H}}_2\text{O}\to b{(\text{Mg},\text{F}{\text{e}}^{2+},{\text{Fe}}_x^{3+})}_3\underset{\text{serpentine}}{({\text{Fe}}_x^{3+}\text{S}{\text{i}}_{2-x})}{\text{O}}_5{(\text{OH})}_4\\ & +\underset{\text{brucite}}{c(\text{Mg},\text{ F}{\text{e}}^{2+}){(\text{OH})}_2}+\underset{\text{magnetite}}{y\text{F}{\text{e}}_3{\text{O}}_4}+\hspace{0.17em}d{\text{H}}_{2(\text{aq})}\end{array}$$ 1.2

In reaction (1.2), variable x indicates the extent of ferri-tschermaks substitution and variable y indicates the amount of magnetite produced per mole of olivine. The remaining coefficients can be expressed in terms of x and y to balance the reaction: $a=1+((1-x)/(2-x))-2y;$ $b=1/(2-x);$ $c=((1-3x)/(2-x))-3y;$ and $d=x/(2-x)+y.$

Serpentine possesses less Fe than magnetite and may be less reactive than Fe-bearing brucite but it is ever present, and the secondary phase in greatest abundance, in serpentinites. Furthermore, it is known that serpentine can undergo multiple episodes of recrystallization and replacement reactions as geologic and geochemical conditions change with time (e.g. [29]). As such, serpentine minerals play a particularly important role in the Fe chemistry of serpentinization. Yet, the timing and conditions under which Fe, and particularly Fe(III), is partitioned into serpentine minerals are still enigmatic (e.g. [1,4,26,30–34]).

It has been suggested that geologic setting, protolith composition, temperature, extent of serpentinization, water/rock ratios, fluid chemistry (e.g. aSiO₂), thermodynamic equilibrium between secondary mineral phases and kinetics, among other factors, may control Fe partitioning ([25,28], e.g. [29,35–39]). Since Fe may respond to any/all of these system parameters, its behaviour, as recorded in serpentinites is used to broadly define and interpret these same system parameters. To infer reaction histories and conditions from Fe chemistry, it is necessary to determine the chemical reactions that Fe underwent during serpentinization. Most often, we have only a single snapshot in time from which we try to reconstruct a series of many, often distinctly different, events. Yet, it is based on this extrapolation that we try to understand origin of the rocks, setting of serpentinization, number of episodes of alteration, water/rock ratio, temperature, and fluid chemistry, etc. As recently stated by McCollom et al. ([33] in revision), tests of the effects of many of these parameters on Fe partitioning are limited. Isolating single parameters for experimental investigations is a challenge and systematic, well-designed laboratory experiments to isolate and test possible controlling parameters will take a massive investment of effort, creativity and money; not to mention dedicated researchers and time. Yet, as geologically and geochemically complicated as natural serpentinizing systems are, there is a wealth of compositional data available in the literature.

We have conducted a survey of the existing literature on Fe in serpentinizing systems, collecting and compiling the results of analyses of Fe in serpentinites and serpentine minerals from many individual studies into a single database. The creation of this database enables a synthesized integration of the many results reported from a large number of individual studies. Earlier works have compiled chemical data from diverse serpentinizing studies (e.g. [40–43]) but none have focused on iron chemistry. Yet, many interpretations of both the controls on, and implications of, Fe behaviour in a single serpentinizing system are applied across systems. This database makes it possible to test Fe behaviour across a variety of parameters and systems, allowing visualization of the commonalities and differences between single systems or single parameters. This database will enable researchers to provide an increasingly broad and consistent context in which to interpret Fe data from unique serpentinizing systems. The insights gained from observations of this integrated dataset will aid in determining the robust controls and implications of Fe behaviour as well as aid in prioritizing and targeting future investigations of natural systems and experimental studies. Furthermore, this database is publicly available and can be added to and modified by other researchers who can tailor it to address open questions informed by their own expertise.

2. Methods

(a). Compilation of data from the literature

Data were initially gathered through literature searches using Google Scholar and GeoRef. Key words were first limited to ‘iron reduction oxidation serpentine’ to retrieve studies reporting quantitative Fe redox data from serpentine minerals. However, the majority of studies that publish Fe chemical and redox data are not actually focused on the Fe data. Thus, it was most effective to retrieve references that were cited within the papers originally identified via literature searches. When data was not available in tables, WebPlotDigitizer [44] was used to extract data from plots. The number of studies on Fe oxidation state in serpentine minerals is relatively limited. The database was expanded to include studies reporting Fe chemical and Fe redox data for bulk rocks, of which there are many. Not all studies could be included and work was prioritized for inclusion if it included (i) quantitative measures of Fe oxidation state and (ii) the data was collected from specimens of natural, geologic serpentinizing systems. Care was taken to include only the highest quality data in the database. Careful attention was paid to the methods employed in the original papers as well to the acknowledgement and assessment of assumptions and complications inherent to the methods. It was not feasible to compile the individual error for each measurement included in the database particularly because many of the source papers did not provide that information. Furthermore, it would be inaccurate to apply modern uncertainties to older datasets, since the uncertainties of many methods have often improved substantially over time. Data from experiments and thermodynamic models were not included.

In his 2008 Journal of Petrology paper [38], Evans presented bulk rock Fe(III)/Fe_T data from 43 literature sources and he generously shared this dataset with us. Any data for which the original reference could be obtained, and which met the other structural requirements of the database, were included in the current database (electronic supplementary material, table S1, citations indicated with asterisks). Reports of total Fe measurements in bulk rock serpentinites are common and while some studies reporting only bulk rock Fe chemical data (not Fe oxidation state data) were included in the database not all available studies could be included. Studies that made data available in a spreadsheet format (e.g. in electronic supplementary material) were prioritized for inclusion in the database. Similarly, studies that present contextual data for the geologic system and included paired serpentinite/serpentine data were also prioritized for inclusion. Seventy-six unique studies (citations provided in electronic supplementary material, table S1) were included in the final database.

(b). Description of the database

(i). Extraction and transformation of geochemical data

Major element geochemical data is traditionally reported in oxide weight per cent. Iron commonly occurs in two different oxidation states (2+ and 3+) within silicate rocks. The different valence states result in two different oxides of iron, FeO and Fe₂O₃. Depending on the specific study and whether data are reported for bulk rock specimens or serpentine minerals, iron contents may be reported in a variety of ways, including as weight per cent: (i) as FeO_T, which represents all of the iron in the rock, regardless of actual oxidation state, as Fe(II); (ii) as Fe₂O_3T, which represents all of the iron in the rock, regardless of actual oxidation state, as Fe(III); (iii) as independent measures of Fe(II)O and Fe(III)₂O₃; and (iv) as independent measures of Fe(II) and/or Fe(III). Chemical data may also be reported as atoms per formula unit (apfu) of a mineral normalized to the number of oxygens in that mineral. In a few cases, Fe(II) and Fe(III) were reported as per cent of the total Fe apfu. The ratio Fe(III)/Fe_T is also commonly reported when Fe redox data is included in a study. Some studies report a single value, the most common being FeO_T wt%, while others report a subset of the values and still others report all of the possible variations. All Fe data reported in the original studies were included in the database. In order to standardize the data format to enable comparisons between studies, if FeO_T wt% was not reported, it was calculated from the data that was provided. (Conversion of Fe₂O_3T to FeO_T was conducted according to FeO = Fe₂O₃ × 0.8998.) Similarly, Fe(III) wt% and Fe(III)/Fe_T were calculated whenever possible. Iron contents reported as apfu were converted to weight per cent values through a variety of calculations depending on the exact format of the available data. When a single Fe apfu value was reported in the original literature, the value was converted to FeO_T. When valence specific apfu values were reported, they were converted to Fe(II) and Fe(III) weight per cent as appropriate.

(ii). Contextual characteristics of the serpentinizing system, serpentinites and serpentine minerals

As addressed in §1, there are a number of different parameters that affect Fe behaviour in serpentinites and serpentine minerals. These parameters can be categorized as to whether they describe (i) the serpentinizing system, including geologic setting, temperature; (ii) the serpentinite rocks, including protolith type/composition, rock type, extent of serpentinization; or (iii) the serpentine minerals, including type of serpentine, texture of serpentine. Whenever available, quantitative data or qualitative descriptions of these system parameters were included in the database. Apart from the inherent characteristics of the serpentinite, the way in which the specimen was obtained (e.g. core versus outcrop) and analysed (e.g. wet chemistry versus electron microprobe or Mössbauer versus µXANES) may reflect differences in the Fe data. This information was also included in the database when available.

Specific modifications to some contextual data was necessary. For instance, qualitative descriptive information provided in the original sources was modified, when reasonable, to enable use of the information in assessing trends. Those modifications are stated here. The degree of serpentinization was often reported in the original sources as either a range of values or as a qualitative descriptor. In order to use these data, ranges were assigned a value equal to the middle of the range and qualitative descriptors were assigned a numeric value (e.g. ‘least’ = 10%; ‘partial’ = 40%; ‘high’ = 90%). The large variety of descriptions of serpentine texture were modified to be more consistent between studies according to the following: ‘after ol’ was transformed to ‘mesh’; any variety of ‘after pyroxene’ was transformed to ‘bastite’.

To characterize the geologic setting of serpentinites and serpentine minerals, three categories were included in the database. The first is the geologic setting as described in the original source of the data (‘published geologic setting’). The second is the geologic setting as it is currently understood (‘current geologic setting’), whenever it was possible to gather this information. Of particular note is that most mid-ocean ridge (MOR) samples (except for Hess Deep) were labelled as MOR oceanic core complexes (OCC) in this category to account for the current understanding that most MOR samples are from OCCs that were not previously recognized (M. Cannat 2019, personal communication; A. McCaig 2019, personal communication). For the current geologic setting, the term ‘ophiolite’ requires evidence for the entire oceanic crustal rock sequence (which was not always the case in earlier literature). The term ‘alpine’ is used as a general descriptor in this category signifying that the samples appear to have undergone prograde metamorphism. Thus, the term ‘alpine ophiolite’ describes ophiolites that underwent prograde metamorphism. This current geologic setting category is presented in the main body text, figures and tables as it should represent the most current science. The third category is the ‘inferred initial setting’, which is the setting or process that the samples are thought to have originated from or experienced. In this category, if a specific area of a subduction zone setting was inferred by the authors of the source data, it was included. In many cases the authors inferred that the samples had undergone subduction and then the term ‘subduction zone’ was used as the inferred geologic setting/process label. In many cases, an inferred initial setting or process was not provided in the source papers or easily gleaned from other sources, in which case that category was left blank. Data on the geologic setting as described in the original publications and inferred initial setting or process are presented in the electronic supplementary material (electronic supplementary material, figures S2 and S5).

(iii). Structure of the database

In many cases, multiple Fe chemical measurements from serpentine minerals were reported from the same bulk rock sample. Whenever possible, the serpentine-specific measurements are paired to the appropriate bulk rock chemical data within the database. This is done to enable direct comparison of bulk rock and mineral-specific Fe chemistry, as bulk rock Fe chemistry is assumed to be a predictor of serpentine mineral chemistry. However, this approach results in duplication of the bulk rock data; for example, serpentine analyses 1, 2, 3, 4 and 5 were all collected from rock A, thus the chemical data from rock A appears in the database in five separate instances. In order to avoid pseudoreplication when analysing the bulk rock data, the database was subdivided into two databases. The Bulk Rock Fe database includes only data generated from chemical analyses of bulk rock specimens, has no duplicate entries and is used for analyses of bulk rock data. The Serpentine Fe database includes serpentine mineral chemical analyses either paired with the appropriate bulk rock chemical data or independent of bulk rock data. This database is used for analyses of serpentine mineral chemical data and serpentine mineral/bulk rock chemical relationships.

(c). Data analysis

All data cleaning and analysis were conducted in RStudio (Version 3.5.0) [45]. The data were cleaned, organized and, when necessary, transformed before being analysed. The two databases are stored in the EarthChem Library data repository from where they can be freely accessed.

Exploratory data analysis revealed that the data often displayed non-normality, heterogeneous variance, uneven sample sizes and other characteristics that could interfere with many statistical analyses. Therefore, we used a non-parametric approach based on permutation analysis that avoids assumptions about the sample distributions (e.g. [46]). Permutation-based ANOVA (categorical independent variables) and linear regressions (continuous independent variables) were analysed using the lmPerm package [47] in R. When the ANOVA gave evidence of significant variation, permutation-based post hoc pairwise comparisons of means with control of the familywise error rate using the method of [48] were calculated with the coin [49] and StepwiseTest [50] packages in R. One million permutations were sampled for each permutation analysis, which was sufficient for convergence on a consistent decision to accept or reject the null hypothesis, although p-values do fluctuate due to the subset of permutations sampled. Permutation approaches can still be sensitive to heterogeneous variance and uneven sample sizes, and this could lead to an inflated type-1 error rate in spite of these efforts to control it [51,52]. To compensate we have used a conservative threshold for significance (α = 0.01), and we have included the results in order to facilitate interpretation of the dataset, while acknowledging that some findings of statistical significance could still be erroneous.

3. Testing predictors of Fe chemistry of serpentinites and serpentine minerals

In this section we describe the geochemistry of Fe in serpentinites and serpentine minerals included in the database from 76 unique studies. This description is subdivided into two sections, one focusing on the bulk rock chemistry (56 studies) and the other on the serpentine mineral chemistry (37 studies). The records of total Fe and Fe redox state are addressed for both bulk rock serpentinites and serpentine minerals. Contextual descriptive data of the serpentinizing systems and the rocks themselves are considered, as are sample and data collection methods, where they help to inform Fe behaviour. Many of the system-specific and rock-specific characteristics are themselves interrelated. Gaps in the availability of paired geochemical and contextual data that effect our ability to interpret Fe behaviour during serpentinization are identified and discussed in §5.

(a). Bulk rock Fe chemistry of serpentinites

(i). Total Fe content

The Bulk Rock Fe database includes 741 specimens of which 722 have data for total Fe content in serpentinites. A summary of these data is presented in table 1 and electronic supplementary material, table S2 and figure 1a. While there is a wide range of values (approx. 0–18 wt% FeO), most frequently the values range from approximately 6 to 8.5 wt% with an average of 7.22 wt% FeO.

Table 1.

Statistics for bulk rock Fe chemistry of serpentinites for all specimens and grouped by geologic setting and protolith.

	FeOT (wt%)				Fe(III) (wt%)				Fe(III)/FeT				LOI (wt%)
variable/statistic	mean	s.d.	median	n	mean	s.d.	median	n	mean wt%	s.d.	median	n	mean	s.d.	median	n
all specimens	7.22	1.49	7.14	722	3.27	1.35	3.16	588	0.58	0.20	0.60	601	11.92	4.64	12.34	667
by current geologic setting
alpine	7.42	0.62	7.58	47	2.31	1.32	1.95	47	0.40	0.22	0.34	47	6.24	4.47	5.02	26
alpine ophiolite	7.20	1.52	7.15	168	3.58	1.34	3.57	156	0.65	0.19	0.71	156	10.81	2.84	11.86	157
forearc	6.62	0.42	6.62	17	2.18	0.82	2.02	7	0.43	0.16	0.42	7	14.84	1.16	15.13	14
MOR OCC	7.35	1.38	7.10	212	3.50	1.33	3.52	172	0.61	0.18	0.62	183	12.11	2.88	12.73	206
MOR OCC hydrothermal field	7.64	2.64	6.80	21	3.78	2.75	2.63	4	0.54	0.07	0.53	4	13.68	0.96	13.50	21
ophiolite	6.99	1.72	7.39	120	2.56	0.97	2.57	76	0.45	0.19	0.45	78	11.94	8.60	9.51	110
passive margin	6.47	0.59	6.48	56	2.74	0.50	2.75	46	0.54	0.08	0.55	46	16.19	1.40	16.10	52
by protolith
dunite	7.61	1.18	7.36	86	3.45	1.32	3.59	68	0.58	0.17	0.64	69	11.72	3.48	12.40	86
harzburgite	6.92	1.31	6.99	215	3.00	1.36	3.01	185	0.57	0.22	0.59	190	11.00	3.63	12.38	194
lherzolite	8.74	1.27	8.79	14	4.37	0.69	4.65	14	0.65	0.08	0.62	14	10.89	2.14	11.41	14

Figure 1.

Histograms of bulk rock Fe chemistry of serpentinites. (a) Total FeO abundance (wt%) of all specimens in the database. (b) Total FeO wt% by geologic setting as it is currently understood. (c) Total FeO wt% by protolith. (d) Total Fe(III) abundance (wt%) of all specimens in the database. (e) Total Fe(III) wt% by geologic setting as it is currently understood. (f) Total Fe(III) wt% by protolith. (g) Fe(III)/Fe_T of all specimens in the database. (h) Fe(III)/Fe_T by geologic setting as it is currently understood. (i) Fe(III)/Fe_T by protolith. Diamonds indicate mean values and filled circles indicate median values. F statistic and p-value results of ANOVA tests are shown at the top of each plot. For relationships with two or more categories where ANOVA was significant (p < 0.01), letters indicate the results of pairwise comparison tests, such that categories sharing at least one common letter are not statistically significantly different from one another. (Online version in colour.)

Across geologic settings. The geodynamic or tectonic setting of serpentinization is not always equivalent to the geologic setting from which a specimen is collected and serpentinites may reflect reaction in multiple distinct geodynamic settings. In order to conduct a first order assessment of the relationship between geologic setting and Fe chemistry using this database, the current understanding of the geologic setting of the specimens was interrogated and included here (the geologic setting as defined in the original publication from which the data was obtained and the inferred initial setting/process are also included in the database and data are shown in electronic supplementary material, table S2 and figure S2). Whenever possible the most specific and current definition of the geologic/tectonic setting was used even if it was included in another publication on the same site (see §2b(ii) for definition of some specific categories). Total Fe content are from specimens that represent nine unique geologic settings, of which seven possess n ≥ 10 specimens and were included in subsequent data analyses (figure 1b). The mean FeO_T values of serpentinites from different geologic settings vary from approximately 6.5 to 7.6 wt% (table 1). There is little difference between the mean and median values for any one geologic setting, varying no more than 0.4 wt%. Serpentinites from hydrothermal fields situated on oceanic core complexes have the highest mean FeO_T, while those from passive margins have the lowest (figure 1b). While the ANOVA test indicated statistically significant variation among categories, post hoc pairwise comparisons did not identify any statistically distinct categories (figure 1b).

Across protolith type. Forty per cent of the specimens included in the database were assigned a protolith. Six protolith types were identified, three of which have n ≥ 10 specimens (table 1). Serpentinites with harzburgitic protoliths dominate the data, accounting for approximately 50% of the rocks with known protoliths. Specimens with dunitic or lherzolitic protolith compositions have significantly higher FeO contents than specimens with harzburgitic protoliths (figure 1c).

Across rock type. Examination of Fe chemistry as it varies by rock type allows greater distinction between serpentinites that have unique features but which may have had the same protolith (electronic supplementary material, figure S1a). For example, the rock type ‘antigorite serpentinite’ is more descriptive of the rock than ‘serpentinized harzburgite’ and could have formed from a variety of original protoliths. Half of the specimens were assigned a rock type, representing 22 unique types of serpentinites, six of which have n ≥ 10 specimens (electronic supplementary material, table S2). Of course, type of serpentinite is often linked to type of protolith and similar patterns in Fe content hold true between these categories. However, by distinguishing between types of serpentinites, it becomes clear that carbonated serpentinites are shifted to much lower FeO contents, statistically distinct from all other rock types except serpentinized harzburgite (electronic supplementary material, figure S1a). Antigorite serpentinites have FeO_T between that of serpentinized harzburgites and serpentinized dunites, and are statistically similar to both of these rock types, possibly because either harzburgites or dunites may be the protolith of antigorite serpentinites.

With degree of alteration. Approximately 90% of the specimens have data for loss on ignition (LOI wt%) which can be used as a proxy for degree of serpentinization (table 1). Values of FeO_T for LOI wt% up to approximately 15% are relatively stable (figure 2a). When specimens with LOI > 15% are included in the analysis, a negative relationship between FeO_T and LOI is apparent (electronic supplementary material, figure S3). Many of the specimens with LOI > ∼15 wt% are carbonated serpentinites thus it seems likely that the presence of Fe-poor carbonate minerals reduces the overall FeO abundance of the whole rock.

Figure 2.

Bulk rock Fe chemistry as a function of the extent of serpentinization. (a) Total FeO abundance (wt%) versus LOI wt% for non-carbonated serpentinites. (b) Total Fe(III) abundance (wt%) and moles H₂ per kg of rock versus LOI wt%. The calculation of moles H₂ per kg of rock assumes that all Fe(III) is associated with H₂ production. (c) Fe(III)/Fe_T versus LOI wt%. (Online version in colour.)

(ii). Redox state of Fe

Iron redox data is available for approximately 80% of the specimens included in the Bulk Rock Fe database (table 1, figure 1d). The mean abundance of Fe(III) is 3.27 wt%, though the standard deviation is large (1.35 wt%) relative to the total abundance and the most frequent values range from 2 to 4.5 wt%. The mean Fe(III)/Fe_T is greater than 0.50 with most values ranging from 0.4 to 0.8 (table 1, figure 1g). It is important to acknowledge that bulk rock measurements of Fe redox integrate the signal from all Fe-bearing phases included in the sample. The mineral assemblage of serpentinites is complex and frequently includes Fe-(hydr)oxides such as magnetite and Fe-bearing brucite. In more silica-rich systems Fe-bearing clays other than serpentine phases may form. Where alteration may have occurred under more oxidizing conditions, haematite and/or goethite may be observed. Formation of many of these minerals will lead to an increase in the abundance of Fe(III) and Fe(III)/Fe_T as measured by bulk analytical techniques, which is consistent with the interpretation that serpentinites bearing these mineral assemblages have experienced greater degrees of alteration.

Across geologic setting. Five geologic settings have Fe(III) wt% and Fe(III)/Fe_T values for n ≥ 10 specimens. Mid-ocean ridge OCC and alpine ophiolites have significantly higher Fe(III) wt% values than all other settings (figure 1e). The Fe(III)/Fe_T of specimens from alpine ophiolite settings are statistically distinct from all settings other than MOR OCC (figure 1h).

Across protolith type. The bulk rock Fe(III) content varies with type of protolith such that Fe(III) wt% in lherzolite > dunite > harzburgite, with only the difference between lherzolite and harzburgite being statistically significant (figure 1f). This is consistent with the variation of FeO_T with protolith (figure 1c). The differences between Fe(III)/Fe_T by protolith are not statistically significant (table 1, figure 1h). The mean Fe(III)/Fe_T for dunites and harzburgites is essentially the same; while the absolute amount of Fe(III) is greater in specimens with dunitic protoliths, the total FeO content is also higher so the ratios of Fe(III)/Fe_T are similar. The same trends in Fe(III) wt% and Fe(III)/Fe_T exist when sorted by rock type (electronic supplementary material, figure S1d,g). If we assume that serpentinization is isochemical with regards to Fe, it is possible to investigate the variation in Fe(III) content with the bulk rock FeO_T of the serpentinites. There is a statistically significant increase in Fe(III) content with increasing FeO_T though there is a wide range in Fe(III) wt% represented by the average FeO_T content (figure 3).

Figure 3.

Bulk rock Fe(III) abundance (wt%) as a function of the bulk rock total FeO content. (Online version in colour.)

With extent of alteration. Both Fe(III) wt% and Fe(III)/Fe_T increase with increasing extent of serpentinization when LOI is used as a proxy for the degree of serpentinization (figure 2b,c). Linear models indicate that the correlations between these variables are statistically significant. The moles of H₂ produced per kg of rock was calculated from Fe(III) wt% and is depicted on the right-hand y-axis of figure 2b. This calculation assumes no starting Fe(III) in the rock as the actual value in the primary igneous minerals that compose peridotites varies between phases and is thought to be quite small.

Across sample type and measurement method. Serpentinite rocks can be sampled from the surface or subsurface of the Earth and from marine or subaerial settings. These attributes of the sample may affect the measured Fe redox state of the bulk rock. Serpentinites collected from subaerial outcrops are the most common and there are nearly as many specimens obtained from cores, particularly from the ocean crust. Specimens dredged from the seafloor have statistically significantly higher Fe(III) content than other sample types (electronic supplementary material, figure S1e). Cored and outcrop specimens are not significantly different from one another. The same trends are apparent when comparing values of Fe(III)/Fe_T (electronic supplementary material, figure S1 h). Techniques for measuring Fe redox state also vary between studies, and while there are approximately 30× as many samples measured using wet chemical techniques versus Mössbauer, there were no significant systematic differences between the techniques for either Fe(III) wt% or Fe(III)/Fe_T. The averages differ by only approximately 0.1 wt% Fe(III) and 0.01 Fe(III)/Fe_T (electronic supplementary material, figure S1f,i).

(b). Fe chemistry of serpentine minerals

(i). Total Fe content

The Serpentine Fe database includes 1318 entries of which 1253 have total Fe content in serpentine minerals (table 2). The following summary excludes the single data points for greenalite, cronstedtite, polyhedral and nanotubular serpentine, which were not included in subsequent grouped analyses. Values of FeO_Tserpentine range from approximately 0 to 20 wt% with the most frequent values ranging from approximately 2.5 to 6.5 wt%. The mean is approximately 5 wt% though the standard deviation is relatively large (approx. 2.5%) (table 2, figure 4a).

Table 2.

Statistics for serpentine mineral Fe chemistry for all specimens and grouped by geologic setting, protolith, serpentine polymorph and serpentine texture.

	FeOT (wt%)				Fe(III) (wt%)				Fe(III)/FeT
variable/statistic	mean	s.d.	median	n	mean	s.d.	median	n	mean	s.d.	median	n
all data	4.85	2.68	4.43	1252	1.01	0.65	0.93	126	0.51	0.26	0.52	154
by current geologic setting
alpine ophiolite	3.57	2.10	3.00	201	1.18	0.62	1.08	64	0.60	0.25	0.65	71
forearc	6.83	3.67	5.68	159	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
layered mafic intrusion	8.87	4.13	9.60	23	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
MOR	4.86	2.39	3.98	8	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
MOR OCC	4.48	2.04	4.11	524	1.78	0.72	1.63	10	0.57	0.17	0.61	24
ophiolite	5.92	1.83	6.00	66	n.a.	n.a.	n.a.	0	0.45	0.18	0.50	10
passive margin	4.97	2.33	4.72	161	0.63	n.a.	0.63	1	0.79	n.a.	0.79	1
by protolith
dunite	4.24	2.76	4.00	160	0.59	0.42	0.46	15	0.47	0.23	0.41	15
harzburgite	4.95	2.58	4.53	865	1.22	0.61	1.09	72	0.55	0.23	0.60	77
websterite	5.73	3.61	4.55	33	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
wehrlite	3.09	0.46	3.10	14	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
by serpentine type
antigorite	4.08	2.21	3.92	123	0.96	0.68	0.75	37	0.38	0.26	0.26	42
chrysotile	2.73	1.65	2.40	67	0.69	0.31	0.70	36	0.45	0.19	0.38	29
lizardite	4.80	3.16	3.75	140	1.33	0.61	1.36	29	0.70	0.17	0.78	32
by serpentine texture
banded vein	4.21	1.96	3.26	28	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
bastite	5.36	2.44	5.30	295	1.33	0.39	1.44	11	0.52	0.26	0.47	4
blades	4.16	2.95	2.75	28	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
fibrous vein	3.42	1.51	3.44	41	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
granular vein	4.65	1.96	4.39	34	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
interlocking	2.57	0.56	2.38	12	1.19	0.29	1.21	6	0.62	0.08	0.62	2
isotropic vein	4.21	3.09	2.84	13	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.
mesh	5.39	2.76	4.97	109	1.39	0.43	1.23	5	0.50	0.14	0.50	5
mesh core	6.78	4.38	5.20	129	n.a.	n.a.	n.a.	0	0.66	0.10	0.66	4
mesh rim	4.58	1.93	4.42	244	n.a.	n.a.	n.a.	0	0.54	0.20	0.55	17
microfracture	5.09	1.49	4.52	15	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
serp2	4.77	2.16	4.08	34	2.90	0.18	2.90	2	0.65	0.06	0.65	2
serp3	3.59	1.40	3.00	17	n.a.	n.a.	n.a.	0	n.a.	n.a.	n.a.	0
vein	3.83	2.53	3.80	35	0.75	n.a.	0.75	1	0.59	0.36	0.61	3

Figure 4.

Histograms of the Fe chemistry of serpentine minerals. (a) Total FeO abundance (wt%) of all specimens in the database. (b) Total FeO wt% by geologic setting as it is currently understood. (c) Total FeO wt% by protolith. (d) Total FeO wt% by serpentine polymorph. (e) Total FeO wt% by serpentine texture. F statistic and p-value results of ANOVA tests are shown at the top of each plot. Diamonds indicate mean values and filled circles indicate median values. For relationships with two or more categories where ANOVA was significant (p < 0.01), letters indicate the results of pairwise comparison tests, such that categories sharing at least one common letter are not statistically significantly different from one another. (Online version in colour.)

Comparison to bulk rock characteristics: Across geologic setting. Iron contents were reported for seven unique geologic settings. There are only eight MOR samples from Hess Deep but because they represent a unique type of marine sample they were included in subsequent analyses. The FeO_T of serpentine minerals varies widely with geologic settings with means ranging from approximately 3.5 to 9 wt%. Serpentine minerals from layered mafic intrusions have the highest FeO_T content. Mid-ocean ridge samples from Hess Deep have significantly lower FeO_T than most other groups including samples from MOR OCC. Forearc specimens are unique from all other settings. As seen in the bulk rock chemistry, specimens from MOR OCC and alpine ophiolite settings are statistically similar to one another (table 2, figure 4b). Data for the other geologic setting categories are shown in the electronic supplementary material (figure S4).

Across protolith type. Protoliths were assigned to 1083 of the serpentine mineral specimens with four unique protoliths that have n ≥ 10 specimens. A comparison of FeO_T values by protolith, reveals that serpentine formed from alteration of websterite, a pyroxenite, is enriched in Fe relative to serpentines formed from peridotites (figure 4c). However, while the p-value of the ANOVA test indicated statistically significant variation with protolith, post hoc pairwise comparisons did not identify any statistically significant pairwise differences.

There are 479 serpentine mineral specimens for which there exists paired bulk rock Fe chemical data. Using bulk rock FeO content as a proxy for FeO content of the protolith, we observe a general trend of increasing FeO in serpentine with increasing FeO content of the bulk rock (electronic supplementary material, figure S6a). However, there is considerable scatter in the data and relatively few data at low and high values of bulk rock FeO_T.

Across rock type. The database includes 15 unique types of serpentinites. Six of these have greater than or equal to 10 specimens accounting for 1015 specimens with a rock type more specific than ‘serpentinite’ (electronic supplementary material, table S3). Approximately 80% of these are serpentinized harzburgites. The trends identified when categorizing the data by protolith hold true when categorizing by rock type. Serpentine from serpentinized harzburgite is frequently enriched in FeO relative to serpentine in other rock types yet only serpentinized wherlites and serpentinized lherzolites are significantly different from one another (electronic supplementary material, figure S5a). Serpentine in carbonated serpentinites has approximately the same FeO content as serpentines from serpentinized harzburgites despite significantly lower iron in the bulk carbonated serpentinites. This might suggest that during carbonation, the composition of the serpentine minerals do not change significantly, at least with respect to iron.

With extent of alteration. Studies investigating the composition of serpentine minerals tend to assess and report degree of serpentinization versus LOI as a measure of extent of alteration. Increasing extent of serpentinization is generally correlated with decreased FeO content of serpentine. However at high degrees of serpentinization there is a wide range in FeO_Tserpentine values (electronic supplementary material, figure S6b).

Comparison to serpentine mineral characteristics. The data have demonstrated that characteristics of the bulk rocks from which serpentine specific measurements are made are related to the total FeO content of those serpentines. There are also characteristics specific to the serpentine minerals themselves which can affect FeO_T.

Across serpentine type. Fifteen unique categories of serpentine types, including mixtures of 2 or more different polymorphs of serpentine or serpentine mixed with other silicate or (hydr)oxide phases, were identified in the publications included in the database. Of these, only the three main polymorphs of serpentine have n≥10 individual specimens. The mean FeO_T of lizardite is significantly higher than that of chrysotile due to frequent occurrences of Fe-rich lizardite (figure 4d). However, there is no statistical difference between antigorite and either lizardite or chrysotile (figure 4d).

Across serpentine texture. For most of the specimens within the dataset for which Fe data is reported, the type of serpentine has not been determined (approx. 75% of samples) and thus the majority of the FeO_T data cannot be used to discern trends in Fe content with type of serpentine. Texture is a commonly assessed and published attribute of serpentine minerals. However, the relationship between FeO_Tserpentine and texture is complicated by the variation in terminology; in 37 studies, 55 different terms were used to describe serpentine texture. Similar and repetitive categories were combined for a total of 38 different textures, 13 of which have n ≥ 10 individual specimens. Most textures have a wide range of FeO_T values and the mean is frequently greater than the median value (table 2, figure 4e). Mesh cores are significantly enriched in Fe relative to all other textures except bastite. Bastite and the different mesh textures are generally enriched in Fe relative to veins, though the difference is not significant (figure 4e). The category ‘mesh’ has intermediate mean and median values and likely consists of data from all different areas of the mesh texture. The mean and median values within each textural category vary more than observed in other categories (e.g. serpentine type, geologic setting). This may be due to the subjective manner in which serpentine texture is described and may possibly be magnified by combining textural groupings, although the mean/median variation is apparent even in textures that were not combined. It may also be due to the many different reaction conditions that potentially affect the serpentine texture (e.g. [29]).

(ii). Redox state of Fe

There are significantly fewer data available on Fe redox of serpentine minerals. Of the 1318 entries in the database, 126 of them include data for Fe(III) wt% and 156 include values for Fe(III)/Fe_T. The vast majority of these have 0–1.5 Fe(III) wt% with a mean value of 1 wt% (table 2, figure 5a). The whole range of values for Fe(III)/Fe_T are represented by the data though values from 0.2 to 0.85 are the most frequent with an average of approximately 0.5 (table 2, figure 5e).

Figure 5.

Histograms of the Fe redox chemistry of serpentine minerals. (a) Total Fe(III) abundance (wt%) of all specimens in the database. (b) Total Fe(III) wt% by geologic setting as it is currently understood. (c) Total Fe(III) wt% by protolith. (d) Total Fe(III) wt% by serpentine polymorph. (e) Fe(III)/Fe_T of all specimens in the database. White circle represents the Fe(III)/Fe_T of magnetite for reference. (f) Fe(III)/Fe_T by geologic setting as it is currently understood. (g) Fe(III)/Fe_T by protolith. (h) Fe(III)/Fe_T by serpentine polymorph. Diamonds indicate mean values and filled circles indicate median values. F statistic and p-value results of ANOVA tests are shown at the top of each plot. For relationships with two or more categories where ANOVA was significant (p < 0.01), letters indicate the results of pairwise comparison tests, such that categories sharing at least one common letter are not statistically significantly different from one another.

Comparison to bulk rock characteristics: Across geologic setting. Only two geologic setting categories have n ≥ 10 individual specimens with Fe(III) wt% data (table 2; figure 5b). There is a statistically significant difference in Fe(III) values between serpentine from MOR OCC and alpine ophiolite settings (figure 5b). Fe(III)/Fe_T data is similarly limited (table 2). There is no statistical difference between the three categories for which there is enough data to interrogate (figure 5f).

Across protolith type. The data are heavily dominated by rocks with harzburgitic protoliths which have higher Fe(III) wt% and Fe(III)/Fe_T values for serpentine minerals than rocks with dunitic protoliths (figure 5c,g). There is a statistically significant difference in Fe(III) values between serpentine from harzburgites and dunites (figure 5c). There are not enough specimens for which bulk rock Fe(III) wt% and serpentine mineral Fe(III) wt% are provided to investigate a relationship between the two. There is, however, a statistically significant trend of increasing Fe(III)/Fe_T of serpentine minerals with increasing Fe(III)/Fe_T of the bulk rock (figure 6a).

Figure 6.

Serpentine mineral Fe chemistry as a function of serpentinite bulk rock Fe chemistry and extent of serpentinization. (a) Fe(III)/Fe_T of serpentine versus Fe(III)/Fe_T of the bulk rock. (b) Fe(III)/Fe_T of serpentine versus degree of serpentinization. (c) Fe(III)/Fe_T of serpentine versus bulk rock LOI wt% as a proxy for extent of serpentinization. (Online version in colour.)

With degree of serpentinization. Though there are relatively few data, the Fe(III)/Fe_T of serpentine minerals increases significantly with increasing extent of serpentinization represented both as degree of serpentinization and LOI (figure 6b,c).

Across sample type and measurement method. Differences in Fe redox chemistry between sample types are not significant (electronic supplementary material, table S3 and figure S5c,d). Mössbauer and µXANES analyses are the two most common ways to measure the oxidation state of Fe in minerals, and the results vary significantly between the two methods. The average Fe(III)/Fe_T as assessed by µXANES is higher than that measured by Mössbauer spectroscopy (electronic supplementary material, table S3 and figure S5e). Ratios of Fe(III)/Fe_T in serpentine minerals determined by µXANES may be overestimated by 0.05 (or underestimated by 0.1) due to crystal orientation effects (e.g. [53]). Beam oxidation can also result in an overestimation of Fe(III)/Fe_T by 0.05–0.1 depending on serpentine phase, though for the µXANES data included in this database, the data collection parameters were designed to minimize this effect [6,54,55]. The Mössbauer and µXANES data included in this database are taken from a limited number of studies by a small number of authors and focused on only a small number of serpentinizing systems. Thus, real differences between the suites of samples analysed by the different techniques may also contribute to the observed bias.

Comparison to serpentine mineral characteristics: Across serpentine type and texture. Lizardite is significantly enriched in Fe(III) relative to both antigorite and chrysotile (figure 5d). The mean and median proportions of oxidized iron in lizardite (0.70/0.78) are much greater than those in antigorite (0.38/0.26) and chrysotile (0.45/0.38). However, they are only significantly different from antigorite (figure 5h). The number of samples for which both texture and Fe redox were assessed is too small to provide any valuable insights to the relationship between serpentine texture on Fe redox state.

4. Relationships between Fe chemistry and system and reaction parameters

A main goal of studying the chemistry of serpentinites and serpentine minerals is to shed light on the alteration history they record. The complex and dynamic behaviour of iron in these systems is of particular interest because of its responsiveness to the characteristics and conditions of the reactions. But, the question remains, can we reliably infer reaction histories and conditions from Fe chemistry? To address this question, we must determine if there are robust relationships between Fe chemistry and characteristics of the serpentinizing system and reaction parameters. With this database, it is possible to interrogate a much larger dataset to see where observations and inferences/interpretations from single systems and parameters are consistent across/between many systems and many parameters.

(a). System and rock specific parameters

(i). Across geologic setting

The geodynamic setting of serpentinization can be difficult to accurately assign on the basis of the current geologic setting of the sample and is inextricably linked to other reaction parameters including fluid composition and temperature. Thus, teasing apart the relationship between Fe chemistry and geodynamic setting of serpentinization is complex. However, it has been demonstrated that the extent of Fe oxidation is related to the geologic setting of the serpentinite (e.g. [6,55,56]). Recent work on the effect of geodynamic setting on serpentinites and serpentine minerals has focused on understanding the effect of serpentinizing fluids on the redox state of subduction zones (e.g. [41,54,56–59]). Many serpentinites derived from subduction zones that have undergone prograde metamorphism after serpentinization are likely to have experienced Fe reduction during those episodes (e.g. [55]). However, the bulk rock data presented here (at least when grouped by current geologic setting) illustrate a similarity between specimens from MOR OCC, which presumably experienced uplift and exhumation, and alpine ophiolites, which presumably experienced prograde metamorphism during orogenesis. Whereas these two categories had no significant difference between them, they both have significantly higher values of Fe(III) wt% and Fe(III)/Fe_T than the other categories. When viewed from this perspective, these data might suggest that the overarching signal recorded by the Fe in these rocks is extent of serpentinization. Specimens from both MOR OCC and alpine ophiolites may have experienced a greater extent of water/rock reaction than samples from other settings. Rocks from MOR OCC settings may have experienced serpentinization where they originated on the seafloor, as they were uplifted and exhumed during the formation of the oceanic core complex, and serpentinization may continue well after OCC formation (e.g. Lost City). Rocks from alpine ophiolites possibly serpentinized during original formation, obduction and orogenesis. However, ophiolites may have experienced serpentinization on the seafloor and during obduction but did not undergo orogenesis. Alpine serpentinites may have only been altered during orogenesis and serpentinites from passive margins have not experienced obduction or orogenesis. Additionally, iron in serpentinites from MOR and ophiolite settings is more oxidized than iron in basalts and gabbros from these same settings [60], likely to be reflective of the serpentinites having experienced a greater degree of alteration.

(ii). With protolith

Inherently, the process of serpentinization requires a protolith, i.e. a primary igneous peridotite, that possesses its own (unique) chemical composition and which affects the resulting mineral assemblage of the serpentinite (e.g. [7,41,61,62]). Peridotites (greater than or equal to 60 wt% olivine) are the dominant protolith types included in this analysis and include dunites (typical composition approx. 100% olivine), harzburgites (approx. 70–80 vol% olivine, 10–20 vol% orthopyroxene, 1–5 vol% clinopyroxene), and a few examples of lherzolites (approx. 60–70 vol% olivine, 10–30 vol% orthopyroxene, 5–15 vol% clinopyroxene, 2–5 vol% spinel) (e.g. [7,63,64]) and wehrlites (wide ranges in olivine and clinopyroxene content and minor (less than 5%) amounts of orthopyroxene). A few of the serpentinites in the database have pyroxenite protoliths (≤60 wt% olivine, websterites). Olivine in pyroxenites has been demonstrated to be more Fe-rich than olivine in peridotites (e.g. Fo_82–86 versus Fo_88–92; [65] and references therein). Studies on the oxidation state of Fe in the primary minerals show Fe(III)/Fe_T in clinopyroxene > orthopyroxene > olivine though orthopyroxene often contributes more Fe(III) overall because it is more Fe rich than clinopyroxene (e.g. [66–68]). The data presented here show that serpentinites with harzburgitic protoliths have significantly less FeO than those with dunitic and lherzolitic protoliths (figure 1c). That serpentinites with dunitic protoliths have higher FeO content than those with harzburgitic protoliths is consistent with a study of peridotites from the Oman ophiolite, where it was shown that dunites have higher FeO contents then harzburgites (7–10 versus 7–8.5 wt%), generally due to their higher olivine content [69]. Generally, the whole rock composition of pyroxene-rich peridotites includes slightly more Fe than peridotites with less pyroxene [40]. Thus, serpentinites with the more pyroxene-rich lherzolitic protoliths would be predicted to have greater FeO contents than serpentinites with harzburgitic protoliths. Serpentinites with lherzolitic protoliths have significantly more Fe(III) than serpentinites with harzburgitic protoliths (figure 1f) but there is no significant variation in Fe(III)/Fe_T between types of protoliths (figure 1i).

Protolith characteristics are frequently discussed as a possible control on secondary mineral compositions (e.g. [29]) and particularly on the incorporation of Fe into serpentine minerals (e.g. [28,70,71]; McCollom et al. [33]). Here, the Fe chemistry of the serpentinite is used as a proxy for the Fe chemistry of the protolith because attempts to derive the quantitative composition of the protolith from the composition of altered serpentinites are not reliable for rocks that have experienced greater than approximately 40% serpentinization (e.g. [72]). This assumes that serpentinization is close to isochemical with regards to FeO (though the isochemical nature of serpentinization has been debated through time) (e.g. [59,73–80]). It might reasonably be expected that when there is more Fe available in the starting material, more will be incorporated into the secondary serpentine minerals (e.g. [28]). However, the data presented here do not show a statistically significant relationship between bulk rock Fe content and serpentine mineral Fe content (electronic supplementary material, figure S6a). Models indicate that serpentinites with pyroxenite protoliths have more Fe-rich serpentine than those with peridotite protoliths [28]. The data here show that while the mean FeO content of serpentine minerals with websteritic protoliths is higher than that of serpentine minerals with peridotite protoliths, the difference is not statistically significant (figure 4c).

(iii). With degree of alteration

The extent to which a rock is serpentinized, or altered from its original peridotitic composition, is known to influence the resulting chemical composition. While iron is not commonly thought to be particularly mobile during serpentinization (e.g. [2,29,55,59,81,82]), observations of Fe mobility during serpentinization vary depending on the characteristics of the system (e.g. natural versus experimental; high temperature versus low temperature) or with the scale being considered (e.g. grain versus rock versus system) [55,61,64,70,83–93]. The data here indicate a minor, but statistically significant, negative correlation between bulk rock FeO and LOI up to 15 wt% (figure 2a). Though the data is sparse, the negative correlation is stronger at higher values of LOI (up to 50 wt%; electronic supplementary material, figure S3). This may suggest that some Fe is lost from the system, at least at this scale, during serpentinization.

An increase in Fe(III)/Fe_T of serpentinites with increasing extent of serpentinization has long been observed and has been demonstrated previously for single systems and across systems (e.g. [32,38,59,94]). This database expands the number of specimens investigated at a single time and demonstrates that the increase in both Fe(III) wt% and Fe(III)/Fe_T with greater degrees of serpentinization is statistically significant (figure 2b,c). While there are distinct differences in the Fe redox state when comparing serpentinites with the lowest and highest values of LOI, for the typical serpentinite (LOI ∼ 12 wt%) there is a wide range of Fe(III) and Fe(III)/Fe_T values (figure 2b,c). Thus, while the extent of serpentinization is certainly a predictor of Fe redox state, there are clearly other influential parameters.

Extent of serpentinization has also been invoked as a predictor of the Fe content of serpentine minerals. The overall Fe content of serpentine minerals is predicted to decrease with extent of serpentinization such that less altered rocks tend to have more Fe rich serpentine than highly altered rocks (e.g. [3,4,26,37,38,95]). One explanation for this is the partitioning of Fe into magnetite as it precipitates due to an increase in oxygen fugacity during later stages of reaction (e.g. [9]). The data here do not show a statistically significant relationship between the Fe content of serpentine minerals and degree of alteration (electronic supplementary material, figure S6b). The proportion of Fe occurring as Fe(III) has been predicted to increase [4,28,38] with increasing serpentinization as, for example, at the MARK system where Fe(II)-rich serpentine was observed at early stages (0–40% serpentinization) whereas at greater extents of serpentinization the serpentine had less Fe but it was greater than 50% Fe(III) [4]. In this compilation of data from across systems, there is a statistically significant trend of increasing Fe(III)/Fe_Tserpentine with increasing degree of serpentinization (figure 6b,c).

(iv). Across sample type

The method by which a specimen was collected is another parameter that has been noted to affect the redox state of Fe in serpentinites. In Evans's 2008 [38] analysis of bulk rock Fe(III)/Fe_T versus extent of serpentinization, specimens collected by dredging the seafloor were omitted. The assumption was that these samples would reflect alteration that occurred on the seafloor, not just alteration as a result of serpentinization. Specimens obtained from drill core material, which has not spent long periods of time exposed to seawater or subaerial conditions, might better preserve the in situ Fe redox state versus specimens obtained from dredging or outcrops. We included specimens from dredging and outcrops in the current database because they are a significant number of the samples for which Fe redox was measured. In fact, Fe in bulk rock dredge samples is significantly oxidized relative to other sample types (electronic supplementary material, figure S1h). Core samples, which might be expected to be the least oxidized, appear more oxidized than outcrop samples; however, the difference is not statistically significant (electronic supplementary material, figures S4d). However, bulk rock outcrop samples do have a greater range of Fe(III)/Fe_T values extending to fully oxidized while values from core samples do not exceed 80% oxidized Fe (electronic supplementary material, figure S1h). It is important to bear in mind that these differences in Fe redox state are also likely to be affected by the geologic setting. Specimens from mid-ocean ridges are most likely to be those that are derived from drill core material whereas specimens from ophiolites are more likely to be from outcrops. The inter-relationship between these two specimen characteristics (geologic setting and sample type) makes it difficult to further determine which has more influence on the redox state of iron.

(b). Mineral specific parameters

(i). Across serpentine type

Lizardite is commonly thought to accommodate more Fe in its structure than the other polymorphs of serpentine [62,96]. Interrogating the 330 specimens for which a type of serpentine was defined, it is apparent that lizardite is significantly enriched in Fe relative to chrysotile (figure 4d).

The type of serpentine that forms may be related to the Fe(III) content of serpentine (e.g. [29,30]). It has been observed that lizardite can accommodate significant amounts of Fe(III) with lesser amounts in antigorite [1,29,32,62,97–99]. Lizardite may even be stabilized to higher temperatures (relative to antigorite) by increasing substitution of Fe(III) into its mineral structure [29]. The data included here generally support these earlier observations. The abundance of Fe(III) in lizardite is significantly higher than in chrysotile (figure 5d) while the Fe(III)/Fe_T of lizardite is significantly higher than the Fe(III)/Fe_T of antigorite (figure 5h).

(ii). Across serpentine texture

Serpentine texture is related to recrystallization and replacement, which can occur from changes in temperature and fluid composition, and may even be related to characteristics of the protolith (e.g. texture of the protolith) [29]. Thus, serpentine textures may reflect some of these characteristics of the serpentinizing system. Decades ago it was noted that serpentines formed from pyroxenes, often described as having a bastite texture, have higher Fe contents than mesh texture serpentines (formed after olivine) (e.g. [92,98,100]). More recently, models of pyroxenite serpentinization produced serpentine that was more Fe-rich than serpentine resulting from the serpentinization of dunite [28]. Observations from 295 specimens identified as bastite appear to have an average FeO content that is higher than the other serpentine textures except for mesh cores though there are no statistically significant relationships between bastite and the other textures (figure 4e). However, mesh cores do have significantly higher FeO content than all textures other than bastite. This may be in part due to the absence of other Fe-bearing secondary phases (e.g. magnetite) in either mesh cores or bastite. Serpentine in veins tends to have lower FeO contents, which in many cases may be due to co-precipitation with magnetite which may preferentially accommodate the iron.

5. Knowledge gaps and open questions

In total, the Serpentine Fe database contains 1318 specimens. However, the number of serpentine specimens for which there is paired bulk rock data is much lower (479). The number of specimens for which Fe redox data is available for both bulk serpentinites and serpentine minerals is even more limited (58). Of course, the database presented here does not include all published data but it does demonstrate the frequency with which paired bulk rock and mineral Fe redox data are presented together and the ease of obtaining that data. The lack of complimentary data limits our ability to interrogate how bulk serpentinite composition and, by proxy, protolith composition might be related to serpentine mineral composition. Coordinated efforts to conduct both bulk and in situ analyses of Fe redox on the same specimens in future investigations of serpentinizing systems will hopefully result in the production of more paired bulk and mineral Fe redox data.

Sample type (e.g. core versus outcrop) has been demonstrated to affect the bulk rock redox state of Fe and it seems likely that serpentine minerals in cored material might better preserve the in situ Fe redox state versus serpentine minerals in samples obtained from dredging or outcrops. To date, measurements of Fe redox state in serpentines has been primarily focused on samples collected from outcrops so it is not possible to quantitatively assess the effect of sampling procedure on the Fe redox of serpentine minerals. However, some recent serpentinite drilling projects have dedicated core material for assessment of Fe redox state (e.g. IODP Exp. 357; Oman Drilling Project) and, in some cases, samples are being stored and prepared anaerobically to further ensure that the original Fe redox state is preserved (e.g. Oman Drilling Project). This work may begin to shed some light on the potential for a sampling procedure bias in Fe redox data.

Temperature is thought to be a main driver of variation in the Fe content and redox state of serpentine minerals. Specifically, the Fe content of serpentine is thought to decrease with increasing temperature (e.g. [28,62]) and it has been observed that there is a greater proportion of Fe(III) in serpentine when it is formed at lower temperatures (e.g. [62]). While this should be a testable hypothesis, it is one that is difficult to pin down. Results from experimental studies on the effect of temperature on the Fe chemistry of serpentine are widely variable ([64], e.g. [71,101]) (McCollom [71] and references therein; McCollom et al. [33] in revision and references within) and thermodynamic models often suggest that Fe(II)/Fe(III) of serpentine does not vary widely with temperature (e.g. [25,28,71]). It is not often possible to measure the temperature of the reactions occurring in modern serpentinizing systems. Complicating the study of naturally serpentinized rocks is the observation that many serpentinites have undergone multiple episodes of water–rock interaction resulting in multiple variations and/or generations of serpentine minerals. Bulk rock proxies for temperature, such as oxygen isotopic measurements, have greatly informed our understanding of the serpentinization process. Ideally, in order to understand the effect of temperature on serpentine mineral Fe chemistry and redox state, these measurements would be conducted on the same scale as the serpentine minerals. This necessitates in situ measurements, which are just now being more widely applied to serpentinizing systems (e.g. [102–104]).

Water/rock ratio is a key system parameter that is difficult to assess but plays an important role in the Fe chemistry of serpentinites and serpentine minerals. Generally, under low water/rock ratios (e.g. closed system conditions) it is predicted that the Fe content of serpentine is greater than under high water/rock ratios (e.g. open system conditions) [7,26]. In models of low temperature systems, water/rock has been implicated as a key control on the Fe(III)/Fe_T of serpentine where Fe(III)/Fe_Tserpentine increases with increasing water/rock [7,26,28,38,105]. In natural modern systems and in preserved ancient systems, the W/R cannot be measured so it must be inferred from observations. In a study of the MARK system, observations of increased Fe(III)/Fe_T of serpentine were used as evidence of increasing W/R with time [4]. The design of flow-through water/rock reaction experiments where the W/R can be varied may lend more insights into the effect of W/R on serpentine mineral chemistry.

Fluid composition is inextricably linked to both rock composition and water/rock ratio such that it influences mineral precipitation and is, in turn, influenced by reactions between minerals (e.g. [71]). Activity of SiO₂ is thought to be particularly influential in regards to both the precipitation and composition of serpentine (e.g. [29,37,71]). Silica mobility between mineralogically distinct regions of the rock also affects Fe behaviour [9,37,106,107]. It has been suggested that the ferric iron component of lizardite may increase with decreasing aSiO₂ [29]. Similar to W/R, in ancient systems the fluid composition cannot be measured, and because of the feedback between fluid and mineral composition it is difficult to isolate and test the effects of varying fluid compositions. Investigation of modern serpentinizing systems where fluid composition can be measured and coupled to solid phase analyses (e.g. Oman Drilling Project) may help unravel the relationship between fluid and mineral chemistry.

Primary and secondary mineral assemblages also affect serpentine mineral chemistry. The secondary mineral assemblage is itself dependent on the fluid composition and water/rock ratio. Magnetite and brucite particularly influence serpentine chemistry because of their ability to accommodate Fe. In studies of serpentinites, the abundance of magnetite is often reported in terms of magnetic susceptibility. An inverse relationship between magnetite abundance and Fe content of serpentine has been observed [108–110]. This relationship is important to understand because of the implications for H₂ production as Fe(III) incorporation into serpentine may play a significant role in H₂ production at low temperatures (e.g. [4,28]). Brucite is difficult to detect, let alone quantify, because it is often intimately intergrown with serpentine (e.g. [3,6,7,9,95]). New microscale techniques that enable identification and chemical assessment of brucite versus serpentine will help to elucidate the relationship between these phases.

Finally, there is a great need to standardize both the measurements conducted and the reporting of data when describing the Fe content and redox state of serpentine minerals. When Fe content and redox are measured, they may be reported in any of a myriad of different formats (see §2b(i)). In a few cases, the annotation of the data tables and the method section of papers lack key details making it difficult to determine what form of Fe was actually measured and reported. Iron plays an important role in responding to reaction conditions precisely because it is redox active and it is relied upon as an interpretable record of these conditions. Thus, characterization of both the Fe content and redox state in serpentinites and serpentine minerals should be standard measurements when trying to unravel the history of serpentinization.

The compilation of data such as presented here, but generated from experiments and models, will be helpful to shed light on some of these open questions. Reaction parameters such as temperature, fluid chemical composition and water/rock are more easily measured and, in many cases, constrained in experiments and models. To date, the results of experimental investigations aiming to understand the role of these parameters in influencing Fe chemistry are considerably variable. A statistical treatment of a larger dataset might help to illuminate some of the more robust trends.

Data accessibility

Additional tables and figures as well as references for the literature data that was included in the databases presented in this work are available in the electronic supplementary material. The Serpentinite Bulk Rock Fe chemical database and the Serpentine Mineral Fe chemical database are freely available through the EarthChem Library data repository. Serpentinite Bulk Rock Fe chemistry: https://doi.org/10.1594/IEDA/111420. Serpentine Mineral Fe chemistry: https://doi.org/10.1594/IEDA/111419.

Authors' contributions

L.E.M. conceived of the idea and compiled the data for the databases, analysed and interpreted data, and led authorship of manuscript. E.T.E. contributed to the statistical analysis of the data and provided critical revision of the manuscript.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by funding from the Rock-Powered Life NASA Astrobiology Institute (Cooperative Agreement NNA15BB02A).