The effects of re-homogenization on plagioclase hosted melt inclusions



Melt inclusions trapped in phenocrysts provide a unique picture of magma systems prior to modification by crustal processes. However, post-entrapment crystallization complicates their interpretation. Re-heating the phenocryst to the temperature of entrapment is a commonly applied method to recover the original melt composition. To understand the effects of re-homogenization, we compared the composition of re-heated and naturally quenched melt inclusions and inclusion compositions that had been subjected to over-heating and under-heating to examine the degree to which anomalous compositions were produced. Our results on plagioclase hosted inclusions from ocean floor basalts indicate that the general patterns represented by naturally quenched inclusions are the same as observed for rehomogenized inclusions. Most important, the range of minor elements described for plagioclase hosted inclusions from basalts is found in naturally quenched inclusions, and is therefore not a consequence of the re-homogenization process.

1. Introduction

Historically, most research in petrology has relied on bulk rock, glass and phenocryst compositional data. With the advent of high resolution micro-analytical techniques over the past couple of decades, information from more spatially restricted phenomena has become one of the most important growth areas in petrology/geochemistry. One of these areas includes the use of trapped melt inclusions (see review by Kent [2008]). Such inclusions are arguably a unique source of information on the array of magmatic compositions that existed prior to being “processed” by shallow level mixing and fractionation.

Interpretation of melt inclusion data requires that we make several assumptions about the degree to which the inclusion has been effected by entrapment processes, including the trapping of a melt boundary layer [Kuzmin and Sobolev, 2004; Faure and Schiano, 2005; Baker, 2008] or by post entrapment processes [Danyushevsky et al., 2002; Cottrell et al., 2002]. The most obvious of the post entrapment processes is crystallization of the melt inclusion. Unlike diffusional re-equilibration [Roedder, 1979; Qin et al., 1992; Danyushevsky et al., 2000; Gaetani and Watson, 2002; Hauri, 2002; Danyushevsky et al., 2002; Massare et al., 2002; Cottrell et al., 2002; Portnyagin et al., 2008; Gaetani et al., 2009; Chen et al., 2011] or breaching [Nielsen et al., 1995a], the affects of post entrapment crystallization are apparent under even the most cursory visual examination.

Taken together, all of the known entrapment and post entrapment phenomena, and our uncertainty with regards to how homogenization effects the composition of inclusions has lead to an understandable preference for naturally quenched inclusions [Kress and Ghiorso, 2004; Kent, 2008]. Nevertheless, in many cases, naturally quenched inclusions are not available. To recover the information trapped in “crystallized” inclusion, we must consider re-homogenization. Re-heating of inclusions has been used by a number of petrologists, and there has been some significant work done to examine the potential effects of re-melting [Nielsen et al., 1995a, 1998; Danyushevsky et al., 2000, 2002, 2003]. However, a number of open questions remain. Specifically, what is the magnitude of the effects of re-homogenization? How sensitive is the melt composition to over or under heating (re-heating to temperatures above or below the entrapment conditions)?

The goal of this work is to present a data set of both re-homogenized and naturally quenched plagioclase hosted inclusions from the same samples. We present data from under-heated and over-heated phenocrysts as well as those heated to a temperature close to that of entrapment, and describe the specific characteristics generated by partial homogenization (in the case of under heating) and dissolution of the host plagioclase (in the case of over-heating). Our focus is on the chemical effects of homogenization. The physical effects, particularly the effects of breaching, degassing and cracking have been examined in previous investigations [Nielsen et al., 1995a, 1998]. In addition, we will evaluate the degree to which the cited compositional characteristics of plagioclase hosted inclusions (e.g., anomalous Ti, HFSE, Fe) can be attributed to the re-homogenization process. We focus here on plagioclase hosted inclusions for several reasons. First, they are extremely abundant, providing the opportunity to examine numbers of analyses that would be difficult to obtain from other systems. Second, there are several important issues identified from plagioclase hosted melt inclusion data; one of which is that noted above (anomalous Ti, HFSE contents). Finally, the issues related to diffusive re-equilibration involve different components than do olivine or pyroxene hosted inclusions. For example, the Fe diffusion phenomenon described by Danyushevsky et al. [2000, 2002] for olivine hosted inclusions are not present in plagioclase (or at least are significantly different).

2. Sample Selection

Samples were chosen based on their plagioclase phenocryst contents, degree of quenching and location. The samples (Table 1) include axial lavas (Southeast Indian Ridge, Juan de Fuca, Gorda, etc.), off axis seamounts (Lamont Seamounts), and pull-apart basins (Blanco Fracture Zone). Phenocryst content ranges from 15 to 40 volume %. Most represent pillow rinds, and are characterized by variable glass contents, from glassy quenched matrix and glassy melt inclusions to microcrystalline matrix. Therefore, within each sample there is a range of post entrapment crystallization.

Table 1. Selected Ocean Floor Basalt Samples Used in This Investigation
Juan de Fuca Ridge
HU81-017-11-E-1Karsten et al. [1990]
TT175-39-10-O-2Karsten et al. [1990]
D22-3: West ValleyCousens et al. [1995]
East Blanco Depression
1R1Dziak et al. [2000]
Lamont Seamounts: Sasha
F2-1Allan et al. [1989]
Southeast Indian Ridge
D69Douglas-Priebe [1998]
Gorda Ridge
D9-2-1Davis and Clague [1987]
Galapagos Platform
PL13-2Sinton et al. [1993]
Chile Ridge
19-1Sherman et al. [1997]

3. Experimental and Analytical Methods

3.1. Rehomogenization Technique

The crystals used for this study were hand-picked from the sample after coarse (0.5–3 mm) crushing. As a preliminary step to re-homogenization, a set of unheated crystals from each sample were mounted, polished and examined using backscattered electron imaging. Analysis of the phenocrysts, and the glassy inclusions were performed by electron microprobe. Post- entrapment crystallization (present to some degree in all samples) of melt inclusions took the form of rims of quench crystals in the inclusions and as daughter crystals in the glassy matrix (example in Figure 1). This texture was described in detail by Nielsen et al. [1995a], as were the analytical complications that these textures caused. Analyses described as “unheated” represent the glass in the center of the inclusion (Figure 1).

Figure 1.

(top) Backscattered electron images of post entrapment crystallization in plagioclase hosted melt inclusions. Note the growth of daughter crystals on the inner wall of the inclusion (light gray) in the plagioclase host (darker gray). (bottom) Partially re-homogenized (under heated) melt inclusion with remnant olivine daughter crystal (dark gray spot in light gray melt inclusion).

Re-homogenization of the effects of post entrapment crystallization was performed using a one atmosphere quench furnace. Crystals were suspended by 0.003-in.-thick Pt wire in a Pt “boat” and heated in a furnace atmosphere set near the QFM-buffer. The host crystals were held at 1000°C for 15 min, and then heated to the initial approximate rehomogenization temperature for ∼45 min. Evaluation of the accuracy of the temperature of entrapment and the degree to which the composition of the melt at the time of entrapment was recreated depends on a number of assumptions. First, for these ocean floor basalt systems, we have assumed that the system was saturated with both olivine and plagioclase at the time of entrapment. Second, we assume that all crystals we run are similar in composition and the melt inclusions have the same entrapment temperature. Third, we assume that diffusive transport within the host crystal (the post entrapment daughter crystals should have re-melted) during the re-heating process is minimal for most elements of interest (but allow for the possibility of diffusional re-equilibration during magma storage and transport). The validity of each of these assumptions must be evaluated for each sample we run.

The accuracy of the re-homogenization temperature is determined by running a set of incremental heating experiments at 10°C intervals covering the estimated range of entrapment temperatures. Evaluation of the appropriate rehomogenization temperature can be achieved by calculating the phase equilibria of the re-heated melt inclusions, together with the assumption of olivine plus plagioclase saturation. Specifically, if the calculated liquidus temperature and phase compositions match the experimental results, we can assume that the rehomogenization temperature was close to the entrapment temperature [Sours-Page et al., 1999; Nielsen et al., 2000].

3.2. Electron Microprobe

Major element analyses of melt inclusions and host lavas were performed using the CAMECA SX-100 Electron Microprobe at Oregon State University. Analyses were performed using a 30-nA beam current, 15-kV accelerating voltage and defocused (5 μm) beam. Standards, including USNM 113498/1 (Makaopui Lava) for Si, Al, Fe, Ca and Ti, USNM 133868 (Kakanui An orthoclase) for Na, USNM 143966 (Microcline) for K and USNM 122142 (Kakanui Augite) for Mg were used for glass calibrations. Na was counted first due to its susceptibility to beam damage. Major elements were counted for 10–20 s, while elements in low concentrations, particularly K, P, Cl and S, required counting times of 100, 100, 500 and 50 s, respectively. Due to low concentrations in the glasses, the primary standard for Cl, Tugtapite (7 wt. % Cl) was cross standardized with Scapolite (USNHMR6600–1). Error on individual analyses (based on counting statistics) is ∼1% relative for Na, Mg, Al, Si, Ca, and Fe oxide. Error for K, P, Cl and S are on the order of 5% for S, 15% for K, 20% for P and Cl.

A complete list of analyses is reported in Table S1 in the auxiliary material.

4. Results

4.1. Effects of Over- and Under-Heating During Rehomogenization

Evaluation of melt inclusion data depends on our understanding of what can go wrong during the process of rehomogenization. For example, if there are a range of melt inclusion and phenocryst compositions in a specific sample, it is likely that there is a range of entrapment temperatures represented. In addition, larger inclusions will require longer times to re-homogenize [Danyushevsky et al., 2002].

One of the reasons we as petrologists are interested in melt inclusions is the existence of melt inclusions in single crystals that were trapped from a range of liquid compositions. Re-heating such a phenocryst will result in over heating some inclusions and under heating others. Over- or under-heating of a host mineral during rehomogenization can have a significant effect on the composition of the melt inclusions [Nielsen et al., 1998; Danyushevsky et al., 2002]. To a first approximation, under-heating results when the crystal and inclusion have not been heated to a temperature high enough to melt all post-entrapment crystals. Over-heating occurs when the inclusion is heated past the temperature at which the mineral host begins to melt, causing dilution of the trapped melt composition.

Elements in high concentration in the host are particularly susceptible to the effects of over- (or under-) heating. For plagioclase, these include Al, Eu and Sr. However, much of the controversy regarding the composition of melt inclusions relates to their minor and trace element diversity [Nielsen et al., 1995b, 2000; Sours-Page et al., 2002; Danyushevsky et al., 2002], and the possibility of disequilibrium melting and diffusive re-equilibration with the host [Cottrell et al., 2002]. Therefore, to understand the possible role of the re-homogenization process, separate from the entrapment process in controlling the composition of melt inclusions, we must examine the composition of inclusions that have undergone significant over- and under heating within the context of unheated melt inclusions from the same samples and the range of composition of the host lava suite. In the section below, we will examine a number of cases where we can document both the areas of consistency among all plagioclase hosted inclusions, and where we can document the distinctive characteristics of each sample. We will also examine the distinctive characteristics of naturally quenched inclusions versus those that have been re-homogenized.

4.2. Observations

The first case we will examine is from a plagioclase ultraphyric lava from the Lamont Seamounts (F2–1). The results of the rehomogenization experiments are presented as analyses of individual melt inclusions (re-heated and naturally quenched), together with the host glass and the associate lava suite (Figure 2). Examination of the major element results documents a number of characteristics that are shared by inclusion populations from other plagioclase ultraphyric lavas. (1) The major element composition of the host lava suite is consistent with a trend generated by olivine and plagioclase fractionation. (2) The trends represented by the unheated inclusions describe a trend consistent with post entrapment crystallization of the host. (3) There are multiple populations of re-homogenized inclusions. In this case (Figure 2), one of the inclusion populations lies near the olivine plagioclase cotectic, while others trend away from the array defined by the host lava suite.

Figure 2.

Comparison of melt inclusions that have undergone re-heating, with naturally quenched inclusions for sample F2–1 and the host lava suite for Sasha Seamount (Lamont Seamount group [Allan et al., 1989]). Arrows represent post entrapment plagioclase crystallization, plagioclase addition and the calculated olivine plus plagioclase cotectic (calculated using COMAGMAT [Ariskin et al., 1993], version 3.59). Model parameters used were, 0.1 GPa pressure, 0% water. Plagioclase and olivine liquidus temperatures were adjusted to match the observed phase equilibria in the rehomogenization experiments and those of Kohut and Nielsen [2003].

These results suggest that the entrapment temperature range for the inclusion population straddled the re-heating temperature applied for this set of phenocrysts.

Minor element data from this sample document some of the common characteristics of plagioclase hosted inclusions. Most inclusions have K contents that are near those one would predict based on the host lava composition (e.g., low K for this low K host). A small minority of inclusions (∼1–3%) are characterized by high K. The pattern exhibited by the Ti data is dramatically different. Most inclusions are lower in Ti than one would predict based on the Ti content of the host lava. It is important to note here that the range in Ti contents is similar for the naturally quenched and homogenized inclusions.

The second case differs in both tectonic setting and in host lava composition. Sample 1R1 is from the East Blanco depression [Dziak et al., 2000], a pull-apart basin along the transform system linking the Gorda and Juan de Fuca Ridges. The host lava composition is more enriched than the Sasha Seamount sample, and contains a higher proportion of phenocrysts, most of which are significantly less anorthitic.

The major element results suggest that some of the naturally quenched inclusions lay on the cotectic, as do a proportion of the inclusions heated at 1200–1230 C (Figure 3). However, this data set is different from that for the Sasha Seamount PUB in that the inclusions have lower Mg (and Mg#) than their host glass.

Figure 3.

Comparison of melt inclusions that have undergone re-heating, with naturally quenched inclusions for East Blanco lava 1R1 and the host lava suite. Arrows represent post entrapment plagioclase crystallization, plagioclase addition and the calculated olivine plus plagioclase cotectic.

The minor and trace element data exhibit for the host lava and the associate suite of lavas are distinctly different from one another (Figure 3). Values for K are distinctly higher than the suite (but without the “excursion” of values to higher K as seen in the Sasha lavas), and values for Ti are lower (for both naturally quenched and re-heated). Taken together, this data suggests that the plagioclase phenocrysts in this lava are xenocrystic. However, this interpretation requires that we better understand the potential effects of entrapment and post entrapment crystallization.

The possible effects of any of these processes should be a function of inclusion diameter [Qin et al., 1992; Cottrell et al., 2002; Danyushevsky et al., 2002]. For example, small inclusions will be more susceptible to diffusive re-equilibration with the host, sidewall crystallization and the entrapment of a boundary layer. Therefore, compositional effects should exhibit trends as a function of inclusion size. Examination of the melt inclusion data demonstrates that composition is a function of re-homogenization temperature – however in no case did we find that inclusion composition was a function of inclusion size (Figure 4).

Figure 4.

Correlation of melt inclusion composition with inclusion size (minimum dimension) for heated and unheated melt inclusions from East Blanco lava 1R1.

The next case, HU81-017-11-E-1 from the Juan de Fuca Ridge [Karsten et al., 1990] represents conditions where the re-homogenization temperature was set higher than the entrapment temperature. This is evident in the major element data, where the rehomogenized inclusions trend away from the host lava (Figure 5). In addition, at least two distinct populations of inclusions are evident in the data set. The naturally quenched inclusions trend away from the host lava on plagioclase control line in the opposite direction. Therefore, we can see from this major element data that the naturally quenched and re-homogenized inclusions are separated from the host lava suite by a combination of post entrapment, and re-homogenization processes.

Figure 5.

Comparison of melt inclusions that have undergone re-heating, with naturally quenched inclusions for Juan de Fuca lava HU81-017-11-E-1 and the host lava. Arrows represent post entrapment plagioclase crystallization, plagioclase addition and the calculated olivine plus plagioclase cotectic.

The results for Ti and K suggest that the effects of phase equilibria (addition or removal of plagioclase) has a significant effect on the major element content of the melt inclusion populations – however, they do not change the overall trends for the minor elements. Specifically, the data exhibits two distinct trends for K and to a lesser degree Ti. In addition, the wide range of Ti values in HU-81 is similar to that observed in other PUB lavas. Again, the range of Ti values for the naturally quenched samples is as great as that seen in the re-homogenized samples. The K and Ti data from the naturally quenched inclusions appear to lie on an olivine + plagioclase control line with the host lava. However, as we can see from the Mg-Al data, such a model is not supported by the major element data, which is on a much shallower plagioclase only control line that trends down to values similar to those of the re-homogenized inclusions. It is important to note that the re-homogenization process did not result in a narrower range of Ti or K – as one might predict if re-heating cause diffusive re-equilibration.

A similar data set was generated from (TT175-39-10- O-2), a depleted MORB from the Juan de Fuca Ridge [Karsten et al., 1990]. Note that the re-heated melt inclusion data exhibits multiple populations (Figure 6). The trends within those data intersect an olivine + plagioclase control line that connects with the host lava suite. The naturally quenched data trend off from that same line on a plagioclase only control line. The majority of the inclusions exhibit K contents that are generally similar to the host lava. However, as with the Sasha Seamount lava and the other, more enriched Juan de Fuca lava, some inclusions exhibit significantly higher K. This is not seen in the Ti data for the same inclusions. The heated and naturally quenched inclusions exhibit similar ranges in Ti at any individual Mg content.

Figure 6.

Comparison of melt inclusions that have undergone re-heating, with naturally quenched inclusions for Juan de Fuca lava TT175-39-10- O-2 and the host lava. Arrows represent post entrapment plagioclase crystallization, plagioclase addition and the calculated olivine plus plagioclase cotectic.

Examination of the dependence of inclusion composition as a function of inclusion size illustrates the low proportion of large (>50 micron) inclusions. Such inclusions should be much more sensitive to entrapment and diffusional re-equilibration. However, there is no evidence of a relationship between size and composition for any component (Figure 7).

Figure 7.

Correlation of melt inclusion size with composition for Juan de Fuca lava TT175-39-10- O-2. Each point represents a single inclusion analysis and measured minimum inclusion dimension.

The data from a depleted MORB from the SEIR (D69–1) illustrates that each lava exhibits somewhat different characteristics with respect to the population of inclusion compositions (Figure 8). In this case, the major element trends are similar to those seen in TT175. Naturally quenched inclusions describe a trend on a plagioclase control line attributable to post entrapment crystallization. In this case, the temperature applied in the re-homogenization process resulted in a narrow distribution of major element data that is similar to the host lava. In addition, K contents of re-heated inclusions are similar to the host lava, and do not contain the sub-set of high K inclusions seen in TT175. In contrast, the Ti contents of both the re-heated and naturally quenched inclusions exhibit a significant range. Therefore, the re-homogenization process cannot be the cause of the observed range in minor element composition.

Figure 8.

Comparison of melt inclusions that have undergone re-heating, with naturally quenched inclusions for Southeast Indian Ridge lava D69–1 [Douglas-Priebe, 1998] and with the host lava. Arrows represent post entrapment plagioclase crystallization, plagioclase addition and the calculated olivine plus plagioclase cotectic.

One of the other parameters cited as being sensitive to the re-homogenization process is volatile content, specifically S and its potential effect on Fe content [Nielsen et al., 1998; Danyushevsky et al., 2002]. Exsolution of sulfide may be driven by changes in the Fe content of the inclusion, or by changes in oxygen fugacity (e.g., reduction of sulfate to sulfide). Breaching of the inclusion will result in sulfur concentrations that are significantly below the Fe-S sulfide saturation line. Our data (Figure 9) indicate that plagioclase hosted inclusions from ocean floor basalts fall on the sulfide saturation line near their host lava [Mathez, 1976; Wallace and Carmichael, 1992].

Figure 9.

Comparison of S-Fe for rehomogenized melt inclusions for Southeast Indian Ridge lava D69–1 that have undergone re-homogenization, with its host glass composition and the SEIR lava suite [Douglas-Priebe, 1998]. Sulfide saturation line is as reported by Wallace and Carmichael [1992].

In the cases above, most post entrapment crystallization was limited to growth of plagioclase on the wall of the inclusions. Sample PL12–2 from the Galapagos Platform [Sinton et al., 1993] represents a sample where the inclusions are not quenched. This is typical of lavas that cooled more slowly, resulting in microcrystalline daughter crystals of olivine and plagioclase.

Rehomogenization of these more crystalline inclusions (Figure 10) results in a slightly different pattern for partially rehomogenized inclusions compared to those that contain only plagioclase daughter crystals. Specifically, under heating produces a pattern characterized by a wide range of major element composition, and liquids that are more evolved than the host lava. Heating to increasingly high temperatures results in a narrower range of major element compositions (e.g., Mg versus Al), but not for the incompatible minor elements – particularly Ti.

Figure 10.

Comparison of melt inclusions that have undergone re-heating, with naturally quenched inclusions for Galapagos Platform lava PL13–2 and the host lava [Sinton et al., 1993]. Arrows represent post entrapment plagioclase crystallization, plagioclase addition and the calculated olivine plus plagioclase cotectic.

5. Summary and Conclusions

Interpretation of melt inclusion data requires that we have a clear understanding of the potential effects of entrapment, post entrapment processes, as well as the effects of rehomogenization. We have examined of melt inclusion populations from a number of different plagioclase phyric and ultraphyric ocean floor basalts. The results suggest that over heating or under heating of plagioclase hosted melt inclusions has a significant effect on the major element composition of the inclusions. Equally important, each plagioclase phyric or ultraphyric sample has its own unique set of characteristics attributable to the processes related to the a) differentiation or accumulation of the primary melts and b) the provenance of the plagioclase phenocrysts.


We wish to acknowledge the generosity of all our colleagues who freely gave us samples of plagioclase phyric basalts over the years. This work could not have been done without access to those samples, and the advice of many on where to look in the dim and dark recesses of our sample repositories. This work was supported by NSF grant 0927773.