Cordierite-bearing migmatitic gneiss. Two of the three MNHN samples have a very similar macroscopic aspect. These are Le Guillou's sample (8F636), labelled “gneiss à gros grains gris rougeâtre” (coarse-grained grey-reddish gneiss), and one of Hombron's samples (8G177: Fig. 2a) described as “gneiss à gros grains, rougeâtre, feldspath rouge et mica brun” (coarse-grained gneiss, reddish, with red feldspar and brown mica). They are coarse-grained and foliated gneiss, with abundant cm-sized crystals of pink K-feldspar, which give the rock a pinkish leucocratic aspect. Quartz, cordierite and biotite are also visible with the naked eye. The preferred concentration of biotite and quartz + feldspars in alternate layers indicates an incipient migmatisation, although a planar deformation is still clearly visible, with individual crystals flattened parallel to the foliation. No conspicuous lineation is visible. Le Guillou's sample (8F636) is slightly richer in biotite and less coarse-grained than Hombron's sample (8G177).

Under the optical and scanning electron microscopes, the two samples are almost undistinguishable. Biotite appears subhedral; anhedral quartz crystals display slight undulose extinction, and cordierite is largely unaltered. K-feldspar in the form of coarse microcline crystals is more abundant than plagioclase, and exhibits perthitic exsolution lamellae (Fig. 3a), but no Carlsbad twin. Crystals are commonly replaced by quartz + albite intergrowths (i.e., myrmekite) at their rims. Sillimanite, apatite, zircon and magnetite are minor phases (Fig. 3a). Under the scanning electron microscope (SEM), magnetite crystals show 10-μm-sized inclusions of spinel and an irregular corona of muscovite, mainly at the contact with quartz.

The mineral compositions were obtained at the electron microprobe (Table 2; see Appendix SM1 for the analytical conditions). Microcline composition (sample 8F636: Or₈₈Ab₁₁An₀₀Ce₀₁; 8G177: Or₉₁Ab₀₈An₀₀Ce₀₁) as well as that of plagioclase (8F636: Or₀₁Ab₆₀An₃₉; 8G177: Or₀₁Ab₆₅An₃₄) are very similar in both samples (Table 2; Fig. 4). Cordierite also shows similar X_Mg ratios in the two samples (8F636: X_Mg = Mg/(Mg + Fe + Mn) = 0.71; 8G177: X_Mg = 0.70; Table 2), as well as biotite (8F636: X_Mg = 0.60; 8G177: X_Mg = 0.65). Biotite shows Tschermak substitution (8F636: Phl₂₉Ann₃₄Mn-Ann₄East₂₈Prw₄; 8G177: Phl₃₅Ann₃₀Mn–Ann₃East₂₆Prw₆) and a noticeable F-content (8F636: X_F = F/(F + Cl + OH) = 0.124 ± 0.027; 8G177: X_F = 0.098 ± 0.025; Table 2).

Bulk-rock compositions are presented in Table 3 (major elements) and 5 (trace elements; Appendix SM2; see Appendix SM1 for the method). They were used to calculate the abundance of the minerals in both samples (in vol.%; see Godard, 2009, and Appendix SM1 for the method). The results show that the rocks are cordierite-bearing gneiss, which contain, by decreasing order of abundance (in vol.% of 8G177 and 8F636, respectively): Qtz (35.3, 30.2); Crd (23.0, 23.1); microcline (15.6, 24.7); Pl (19.3, 12.3); Bt (3.3, 5.9); Mag (2.7, 3.3); Ilm (0.6, 0.5); Sill (0.2, 0.0); Ap (0.1, 0.0).

The chemistry of the major elements is compatible with either a granitic or pelitic protolith for both samples, as shown by various binary diagrams (Fig. 7 in Appendix SM2). The two samples also show an almost identical REE pattern (Fig. 5a), with similar total REE contents (227–274), high LREE fractionation (La/Sm = 4.6–4.8), negative Eu anomaly (Eu/Eu* = 0.5) and moderate HREE fractionation (Gd/Lu = 2.5–2.8). The REE (Fig. 5a) and trace-element (Fig. 5b) patterns for 8G177 and 8F636 are also very similar to those of the Post-Archaean Average Australian Shale (PAAS), a geochemical standard proposed by Taylor and McLennan (1985) from metapelitic rocks of the Australian part of the same craton.

Mesocratic amphibolites: The third MNHN specimen (8G178: Fig. 2b), donated by Hombron, is macroscopically very different from the above gneiss samples. Described as “gneiss leptinoïde à grains fins presque compacte, gris verdâtre avec micas bronzé” by Grange (1848-1854), it is a fine-grained, almost equigranular, greenish and mesocratic rock, where only a few mm-sized biotite flakes are distinguishable with the naked eye.

Under the optical and electron microscopes, the rock appears made up of quartz crystals, mm-sized biotite flakes with included prehnite lamellae, clinoamphibole and numerous microcrystalline aggregates of muscovite + pumpellyite ± calcite ± albite ± quartz (Fig. 3c). Cummingtonite, with its distinctive twinning, magnetite, titanite, apatite and epidote are minor phases; calcite fills some late veinlets up to 10 μm in thickness.

The composition of the mafic minerals, obtained at the electron microprobe (Table 4), shows a higher X_Mg ratio than those in the cordierite gneiss (biotite: X_Mg = 0.84; clinoamphibole: X_Mg = 0.74; cummingtonite: X_Mg = 0.70; Table 4). Biotite is a phlogopite (Phl₇₃Ann₁₆Mn-Ann₁East₀Prw₁₁) rich in F (X_F = F/(F + Cl + OH) = 0.28); muscovite is close to the end-member, with little phengite substitution (Ms₈₁Pg₅Mrg₃[Mg,Fe,Mn]-Cel₁₁), and clinoamphibole is magnesio-hornblende, following the classification of Hawthorne et al. (2012). The average of 140 analyses from the muscovite + pumpellyite + albite + quartz aggregate of Fig. 3b gives a composition (Si_2.25Al_1.81Ti_0.00Fe_0.07Mn_0.01Mg_0.16Ca_0.24Na_0.05K_0.46O₈) roughly compatible with that of some feldspar (Si_2+xAl_2−xCa_1−x(Na, K)_xO₈), except for the alkalis, which were possibly lost during the pseudomorphing process.

The bulk-rock composition of sample 8G178 is distinct from that of the cordierite gneiss (Table 3; Table 5 in Appendix SM2). The same is true for the calculated abundance of the minerals (in vol.% of 8G178): 28.1 Cam; 26.9 Qtz; 32.2 Ms + 5.2 Pmp + 0.6 Ab (i.e., ca. 38% of Ms-bearing pseudomorphs); 2.6 Bt; 2.6 Mag; 1.8 Cum; 0.0 Ap. The major-element chemistry is intermediate between basic and pelitic compositions (see Appendix SM2). With respect to the cordierite-bearing gneiss, sample 8G178 shows a lower REE total content (103.9), lower LREE and HREE fractionations (La/Sm = 2.75; Gd/Lu = 1.92) and a less marked Eu anomaly (Eu/Eu* = 0.7) (Fig. 5a).

Other samples: The sample stored at Le Mans museum (2008.2.934: see Section 2, Table 1) consists of a mesocratic medium-grained granular rock. Berthois (1935) studied a thin section of this rock and described it as made up of oligoclase (An₃₄, determined by Michel-Levy's Carlsbad-albite method), hornblende with relics of uralitised pyroxene, minor biotite and quartz. It is thus a quartz diorite, very different from the above samples. Equally dissimilar are the five samples stored at Toulouse museum (MIN 2013 0 1-5; https://commons.wikimedia.org/wiki/File: Roquemaurel_Roches_Terre_Adélie_MHNT_MIN_2013_0_1-5.jpg). Three of these (1, 2, 5) are apparently volcanic rocks, showing feldspar and mafic phenocrysts in an aphanitic or microcrystalline greyish matrix. Two samples (3, 4) display subhedral mafic crystals in a fine-grained felsic matrix. None of these samples contain the cm-sized reddish crystals of K-feldspar that are so abundant at the “Rocher du Débarquement.”

Two samples from the “Rocher du Débarquement”, stored in the Antarctic collections of the Jean-Monnet University (Saint-Étienne, France), were chosen for a comparative study with the above historical samples.

Cordierite-bearing migmatitic gneiss: The first sample (RPM93-49B) was collected on the “Rocher du Débarquement” by R.-P. Ménot, in 1993 and is representative of the main rock present on the islet. It is a migmatitic coarse-grained gneiss with cm-sized reddish K-feldspar crystals. The other minerals, namely biotite, cordierite, quartz, plagioclase, magnetite, apatite and sillimanite, are clearly visible under the optical microscope. The macroscopic aspect of this sample, and its petrological features observed under the optical and electron microscopes are very similar to the above 8F636 and 8G177 historical sample. Perthitic exsolution lamellae, myrmekite and even the small muscovite coronas between quartz and magnetite are also present. The mineral compositions are very similar too (see RPM93-49B vs. 8F636 and 8G177 in Table 2).

Mesocratic amphibolites: The second sample (GD04-74) was collected on the “Rocher” by Guillaume Duclaux in 2004; it is representative of metre-sized enclaves of fine-grained mesocratic amphibolite boudinaged within the migmatitic gneiss. The petrological features of this sample are identical to those of Hombron's 8G178 sample: the rock is almost equigranular, with quartz, clinoamphibole, numerous microcrystalline muscovite-bearing aggregates and a few biotite flakes with included prehnite lamellae, although, relative to 8G178, cummingtonite is apparently absent and pumpellyite is less abundant in the muscovite aggregates. Mineral compositions are also similar (see GD04-74 vs. 8G178 in Table 4).

Based on its position, size and configuration, the “Rocher du Débarquement” has been identified with the northernmost islet of “Pointe Géologie” Archipelago (e.g., Perroud, 1955). According to Vincendon-Dumoulin's map (Fig. 1; Dumont d’Urville, 1840), the coordinates of the landing site are 66°34′29″ S, 137°50′25″ E relative to the Paris Meridian, and thus 140°10′39″ E of Greenwich (green circle in Fig. 1), whereas the actual latitude and longitude of the islet apex are 66°36′18.82″ S and 140°4′17.04″ E (Perroud, 1955; red circle in Fig. 1). The errors in latitude (3.4 km), longitude (4.7 km) and total distance (5.8 km) appear acceptable, if one considers the inaccuracies of the navigation at the time and the difficulties in using a compass in a region close to the southern magnetic pole.

Hombron's and Le Guillou's historical specimens stored at the MNHN in Paris are very similar to the cordierite-bearing gneiss (8G177, 8F636) and mesocratic amphibolite (8G178) recently sampled on the “Rocher du Débarquement” (see Section 3.2). This indeed confirms that the “Rocher” is the real place of the 1840 landing, which, at worst, could have occurred on a neighbouring islet of the same archipelago, made up of the same rocks.

On the other hand, the rocks preserved at Le Mans and Toulouse and reportedly sampled in “Terre Adélie” (see Sections 2 and 3.1, Table 1) are completely dissimilar to the rocks from the “Rocher du Débarquement”. Since Dumont d’Urville did not land on other sites of the Adélie Land coast, they clearly come from other places. The quartz diorite from “île Adelida” (Le Mans: 2008.2.934) could have been sampled in the Magellan Strait at the “Adelaida” Archipelago, where Grange (1848–1854, vol. 1, p. 199), while publishing the geological results of the mission, mentioned diorite dykes. As for the five samples stored at the Toulouse Museum (MIN 2013 0 1–5), their old labels with a numbering up to 47 suggest that they have belonged to a larger collection, which may have included rocks from Adélie Land. However, the five remaining specimens, volcanic for the most part, are totally different from the study rocks (Section 3.1); they possibly come from Patagonia, the Magellan Straight, Pacific Ocean, Indonesia or South Shetland, visited by the expedition. Samples 8F636, 8G177 and 8G178 of the MNHN in Paris are thus the only surviving pieces of the first rocks ever taken on the Antarctic Continent.

Vincendon-Dumoulin carried out his topographical survey from the ship's desk, over 250 km between longitudes 136°E and 142°E. On the resulting map, of which our Fig. 1 is extracted, the icecap edge lies northwards relative to the present coastline. The apparent gap between them is of 7.3 km on average, with a maximum of 22 km at Port-Martin (see fig. 2 in Perroud, 1955). In January 1840, Charles Wilkes surveyed a much larger portion of the same coast, between 100°E and 150°E, and his map, although less accurate, also shows a coastline shifted northward of the present coast by a few kilometres (Knox and Adélie coasts) up to ca. 100 km (Sabrina, Banzare and Clarie coasts) (see fig. 7 in Bertrand, 1972).

Indeed, most of these discrepancies resulted from underestimating the distance to the coast (see Dumont d’Urville, 1846–1853, vol. 1, ch. 8) and from mistaking ice shield for land. However, the relative position of the “Rocher du Débarquement”, reported at 1.7 km from the ice-capped land in 1840 (Fig. 1), 11.4 km in 1952 (Perroud, 1955) and 11.4 km today (Google map), suggests that, in this area, the icecap retracted southwards of ca. 9 km between 1840 and 1952, at the end of the Little Ice Age. The comparison of the coastlines reported in 1840 and 1952 (see red line in Fig. 1; Perroud, 1955) corroborates this evolution.

The “Rocher du Débarquement” consists of Crd-bearing migmatitic gneiss, whose composition in major, trace and rare-earth elements is very close to that of the PAAS standard (Fig. 5), indicating a pelitic protolith. It is however slightly depleted in HREE relative to the PAAS standard (Fig. 5a), which we attribute here to the presence in the same formation of HREE-enriched restitic garnet-bearing paragneiss. The Crd migmatitic gneiss encloses boudins of a fine-grained, mesocratic and almost equigranular rock, whose chemistry is intermediate between basic and pelitic compositions, and probably representative of a volcano-clastic or calc-silicate origin. It was boudinaged and folded close to the metamorphic peak. Several micro-structures (prehnite lamellae in partially chloritised biotite, Ms + Pmp + Ab pseudomorphs after feldspar, calcite veinlets…) indicate static alteration during retrogression under prehnite–pumpellyite-facies conditions.

A theoretical P–T diagram (or “pseudosection”) was modelled for the composition of sample 8G177 (Fig. 6), using the Thermocalc software of Powell and Holland (1988) with an upgraded version of the thermodynamic data set of Holland and Powell (1990, 1998). The modelled molar composition is: (SiO₂)_72.04(Al₂O₃)_10.01(CaO)_1.39(MgO)_4.17(FeO)_7.03(K₂O)_1.77(Na₂O)_1.67(TiO₂)_0.47O_1.60, with a_H2O = 1.0 (i.e. saturation in water), O being deduced from the modal abundance of magnetite (see Appendix SM3 for further details). The result shows that the main paragenesis (Qtz + Crd + Pl + Ksp + Mag + Bt + Ilm + Sil) is stable in a narrow P–T field extending from ca. 2.5 kbar & 620 °C to 3.2 kbar & 670 °C (Fig. 6a). The boundary of this field towards higher values of P is well constrained by the Crd-in curve, whereas the Sil-out curve that bounds the same field at lower P (Fig. 6b) is not so relevant, since it is not clear whether the minor sillimanite observed in the sample was relict or still stable at the metamorphic peak. Migmatisation indicates that the solidus curve was overstepped and the rock reached the field Melt + Qtz + Crd + Pl + Ksp + Mag + Bt + Ilm ± Sil, close to 3.2 kbar and 665 °C (green field in Fig. 6a and b). For these P–T conditions, the model provides mineral abundances, which, converted from mol.% to vol.%, are close to those calculated for the real 8G177 rock (see Section 3.1): Qtz, 35.8 vs. 35.3 vol.%; Crd, 21.5 vs. 23.0; Pl, 17.2 vs. 19.3; Ksp, 13.8 vs. 15.6; Bt, 7.2 vs. 3.3; Mag, 2.0 vs. 2.7; Ilm, 1.5 vs. 0.6; Sil, 0.9 vs. 0.2. The peak P–T conditions obtained for similar cordierite migmatites by Pelletier et al. (2005) at “Pointe Géologie” (4–6 kbar, 700 ± 50 °C for aH2O=0.9) are slightly higher than those obtained here (ca. 3.2 kbar, 665 °C for aH2O=1.0). This disparity can be partly explained by different hypotheses on water activity (0.9 vs. 1.0), since the solidus curve shifts towards lower T when aH2O increases.

The isopleths for cordierite (X_Mg = 0.70) and plagioclase (X_An = 0.34) cross in this narrow P–T field (green area in Fig. 6b), confirming that these minerals equilibrated close to the peak of metamorphism. This is not the case, however, for biotite and microcline, whose real compositions (X_Mg = 0.65, X_Ab = 0.08, respectively) do not match exactly those provided by the model (X_Mg = 0.51–0.60, X_Ab = 0.21–0.18: Fig. 6b). This seems due to the effect of retrogression, during which biotite was partly chloritised and microcline exsolved albite-rich perthitic lamellae. The difference in composition between the modelled K-feldspar at the peak (X_Ab ≈ 0.2) and the real perthitic microcline (X_Ab = 0.08) can well be explained by the exsolution of albite lamellae (ca. 10 vol.%) in the latter (Fig. 3a). Apart from these observations, the retrograde P–T path is poorly documented. We suggest a clockwise path that allows cordierite to remain stable for as long as possible (red dotted line in Fig. 6a), reducing the prospect of alteration of this mineral, which is well preserved in the rock. According to the model, the evolution from the peak towards the muscovite-bearing field at 570 °C and 1.9 kbar produces a diminution in the spinel content of magnetite (X_Al from 0.046 down to 0.025); at contact with quartz, i.e. in excess of SiO₂, such release of Al could generate the muscovite coronas observed around magnetite (see Section 3.1).

This study indicates low-P amphibolite-facies conditions at the metamorphic peak for this region of the “Terre Adélie” craton, corresponding to an abnormally high geothermal gradient (ca. 60 °C/km), even greater than that evidenced by Pelletier et al. (2005). No trace of a high-P overprint has been recorded here, contrary to what is suspected 120 km further east, at the edge of the “Terre Adélie” craton (see review in Godard and Palmeri, 2013).