The Foundation Seamounts are a volcanic chain formed during the last 21 m.y. by the action of a hot spot, presently located near the axis of the Pacific-Antarctic Ridge. The western part of the chain is formed by volcanoes with ages ranging from 21 to 16 m.y., spatially distributed along two parallel lines roughly 200 km apart. The shape and distribution of the edifices are strongly suggestive of an emplacement along preexisting fissures or fractures, resulting from lithospheric deformation during the major kinematic reorganization of 25 m.y. Their chemical and isotopic composition becomes enriched toward young volcano ages, resulting from the progressive contamination of the upper mantle by the arrival of a plume pulse. The eastern part of the chain is formed by volcanoes younger than 5 m.y., also distributed along two subparallel lines. The distance between these two lines diminishes toward the Pacific-Antarctic Ridge axis. A geoid high is located very near the Pacific-Antarctic Ridge axis, over the volcanoes of the north line, which lies in the prolongation of the older Foundation Seamounts, and possibly marks the location of the Foundation hot spot. Both the geoid anomaly and the morphology of the edifices show that the north line is the main locus of the volcanism. The southern line was probably formed on top of the flexural arch resulting from the emplacement of the north line. The chemical and isotopic compositions of the basalts reveal a growing influence of the ridge on the off-axis plume magmatism. The pattern is coherent with a mixing between two sources, occurring when the two melting zones merge and overlap.
 Studies of the interaction processes between mid-oceanic ridges and hot spots have advanced significantly since the early works of Schilling , Vogt , and Morgan and Rodriguez . Vogt , Vogt and Johnson [1975,1976, and Schilling  discussed the possibilities and implications of a flow along the axis of a mid-oceanic ridge under the influence of an underlying hot spot, while Morgan  suggested that the proximity between a ridge and an off-axis hot spot could generate an asthenospheric flow between the two melting zones. Based on morphological and geochemical (isotopes and trace elements) observations, J.-G. Schilling and coworkers [e.g., Schilling, 1985; Schilling et al., 1985; Hanan et al., 1986] developed the so-called mantle plume source-migrating ridge sink (MPS-MRS) model, where the mid-oceanic ridge acts as a line sink, draining the material of the plume along a sublithospheric channel. For a ridge-centered plume the preferential flow direction is along the axis. When it migrates away from the plume, the ridge continues to be fed by a nonradial flow, directed toward it. According to this model the along-axis extension of the geochemical and topographic anomalies is related to the distance between the hot spot and the spreading center, and the connection could still be effective at distances larger than 1700 km [Schilling, 1985]. Off-axis, volcanism located between the spreading ridge and the plume and showing a linear trend of chemical mixing between the ridge mid-ocean-ridge-basalt-like (MORB-like) pole and the plume source is observed. This suggests that exchanges can occur along the asthenospheric channel linking the ridge and the hot spot [Kingsley and Schilling, 1998; Pan and Batiza, 1998]. The mixing trends were interpreted either as a two-way flux between the plume and the ridge [Haase et al., 1996] or as a mixing between the plume and the asthenospheric materials along the sublithospheric channel, the plume signature being diluted progressively as the distance from the hot spot increases [Kingsley and Schilling, 1998]. Numerical experiments [Ito et al., 1997; Ribe and Delattre, 1998] showed that the onset of an asthenospheric flow connecting plume and ridge is strongly controlled by the relative movement between the ridge and the hot spot. In a system where a spreading axis moves toward a mantle plume the additional drag created by the faster plate speed will hinder the connection. On the contrary, in systems where a ridge is moving away from a plume the slower plate speed will contribute to maintain the flow. As a consequence, ridge-hot spot interactions in the first case will be delayed, while in the second case, they will remain active at large distances.
 Linear features and small cones or lava flows grouped along narrow bands support the idea of narrow channels underlying weakened thermally thinned plates [Morgan, 1978; Rappaport et al., 1997]. However, all the studied cases considered either the situation of a ridge centered above a plume or of a ridge moving away from a plume. The case of a ridge approaching a hot spot was given much less attention because the only known examples are the Louisville hot spot-Pacific-Antarctic Ridge (PAR) and Foundation hot spot-PAR systems. Recently, the narrow and linear Hollister ridge, thought initially to represent the channeling flow between the Louisville hot spot and the PAR [Small, 1995], was shown to be of a more complex nature [Géli et al., 1999]. Therefore no morphological connection between the Louisville hot spot (the location of which is still controversial [Géli et al., 1999]) and the PAR can be observed on satellite gravity data. The Foundation-PAR system, where a series of narrow, elongated ridges nearly join the PAR axis [Mammerickx, 1992; Devey et al., 1997] and where the geochemical gradient between a plume source and a MORB-type source was clearly established [Devey et al., 1997; Hékinian et al., 1997], is the best example of a ridge approaching a hot spot.
 The Foundation Seamounts (Figure 1) form a volcanic chain located between 33°S, 131°W and 38°S, 111°W, roughly following the 100°N trend for the absolute motion of the Pacific plate [Mammerickx, 1992]. Ages decrease progressively, from 21 m.y. at the westernmost seamount to very recent ages near the PAR axis [O'Connor et al., 1998]. The main body of the chain (Figure 1), between 33°S, 125°W and 36°S, 115°W is made of well-individualized, elevated volcanoes, sometimes grouped into clusters of two or three edifices [Devey et al., 1997]. Rocks from these volcanoes display a chemical composition typical of intraplate hot spot samples [Hémond and Devey, 1996; Devey et al., 1997; Hékinian et al., 1997]. At its eastern end the Foundation chain joins the PAR. Lava compositions show that 400 km away from the PAR axis, near 115°W, hot-spot- and ridge-derived materials start to mix [Devey et al., 1997; Hémond and Devey, 1996] in what we call the “ridge-hot spot interaction zone.”
 A paleomicroplate is located between 125° and 121°W [Mammerickx, 1992]. It probably formed between 23.4 and 20 m.y. [Tebbens and Cande, 1997], when the southern Pacific-Farallon spreading axis propagated northward creating the eastern spreading ridge of the microplate. Its western boundary, the former northern Pacific-Farallon ridge axis, corresponds to a failed rift, located near 125°W, where the spreading ceased roughly 20 m.y. ago [Tebbens and Cande, 1997; Blais et al., 1999]. Isolated volcanoes and volcanic ridges are observed over a 300-km-wide area west of this failed axis in what will be called hereafter the “western province.”
 In this paper, we will focus on the analysis of the spatial distribution of trace element and lead isotopes and of geophysical and morphostructural data of both the “western province” and the “ridge-hot spot interaction zone” (boxes in Figure 1). The main body of the Foundation Seamount chain, located between these two areas and composed of the central volcanoes of the chain, will not be studied in this paper. These volcanoes represent the truly intraplate expression of the Foundation hot spot (see Figure 2 for the composition) and will be the subject of a specific paper. Trace element data will be discussed as (Nb/Zr)N ratios which characterize the nature (enriched or depleted) of the chemical composition of the lavas. 206Pb/204Pb isotopic ratios will be also used because they are an isotopic tool sensitive to magma mixing and source contamination when dealing with a high-μ-(HIMU)-like hot spot such as Foundation [Hémond and Devey, 1996]. This isotopic ratio will be used only to confirm that the variations observed in the (Nb/Zr)N ratio are essentially source features, rather than a magmatic differentiation effect (Figure 2). The objectives are twofold. First, to understand the temporal evolution of the interaction between the Foundation plume and the lithosphere by analyzing both the spatial distribution of the volcanism and the lithospheric control on the formation of the volcanoes. Second, to understand the temporal evolution of the interaction between the plume, the upper mantle, and the spreading ridge by analyzing the spatial distribution of the lava composition. Finally, the Foundation hot spot-PAR system will be compared with accepted models for ridge-hot spot interaction.
 Multibeam bathymetric data used in this paper were mainly acquired during the Foundation Hotline cruise of R/V L'Atalante (January to February 1997) that fully covered the present-day ridge-hot spot interaction zone and part of the oldest part of the chain. For most of the surveyed area the track spacing allowed a small amount of overlapping at the edges of the swaths. However, above the shallowest volcanic summits, data gaps were unavoidable. At these points, in order to improve the quality of the grids and reduce data gaps, data acquired during the SO-100 cruise of FS Sonne (January to February 1995; Devey et al. ) were occasionally included in our processing. The data set was acquired with Simrad EM-12 dual model during the Foundation Hotline cruise and with Hydrosweep during the SO-100 cruise. The final bathymetric grid has a 400-m step.
2.2. Altimetric Geoid
 In this work we used the gridded altimetric geoid derived from satellite data by Houry et al.  to constrain the present-day location of the Foundation hot spot. The marine geoid was filtered using a two-dimensional (2-D) band-pass filter [Gibert and Galdéano, 1985] in the 2000–500-km wave band in order to remove the long wavelengths that dominate the signal and the short wavelengths related to the topography. This filtering approach differs from the more current approach of removing the lower degrees of the geoid by means of a harmonic model [e.g., Monnereau and Cazenave, 1990], but it has been shown that such an approach may create some artifacts [Sandwell and Renkin, 1988]. The 2000-km cutoff limit was chosen on the basis of results showing that the geoid signature of hot spot swells has shorter wavelengths [Monnereau and Cazenave, 1990].
 From the limited set of major and trace element data published by Devey et al.  and Hékinian et al. [1997, 1999] only (Nb/Zr)N will be discussed. This ratio is useful because it reflects mostly the chemistry of the trace elements of the source when only primary melts are considered. These data are from samples dredged during both cruises (SO-100 and Foundation Hotline).
 The 206Pb/204Pb isotopic ratios will be used as a tracer for the mixing of the involved mantle domains. Pb isotopes were determined using Centre National de Recherche Scientifique (CNRS)-Université de Bretagne Occidentale-Institut Français de Recherche pour l'Exploitation de la Mer (Ifremer) facilities for isotope geochemistry. Thermal ionization mass spectrometry (TIMS) was used to determine Pb isotope compositions. The protocol as well as data are provided in Table 1. Since only the 206/204 ratio will be discussed in this paper, other Pb isotopic ratios are not included in this table. They will be the subject of a specific paper, where the isotopic characteristics of the Foundation plume are discussed in detail.
Table 1. Compilation of Names of Seamounts, Coordinates of the Dredges, Rock Types, and (Nb/Zr)N Ratios From the Literaturea
New 206Pb/204Pb ratios are presented. These data were obtained using a Finnigan MAT 261 mass spectrometer in static mode. The measurements were performed from Pb fractions obtained by digestion of sample chips and separation and purification of the Pb using a chemical procedure described by Manhès et al. . Pb mass fractionation during mass spectrometry analysis was corrected by 0.10%. This value was calculated by comparing the mean of 29 analyses of the NBS 981 standard done during the course of this study and loaded by the same operator with the value of the same standard given by Todt et al. . Our NBS 981 standard mean was 16.904 ± 0.014 (n = 29 and the error in the 2σ of the mean). Pb isotope data represent the average of duplicate or triplicate analyses. A complete isotope data set will be published and discussed in a paper where the isotopic characterization of the Foundation plume will be discussed in full detail.
 The western part of the Foundation Seamount chain (west of 125°W) is located on 33–20-m.y.-old lithosphere [Mammerickx, 1992; Tebbens and Cande, 1997]. The Resolution, Mocha, and Agassiz fracture zones display a 73°N direction and were formed prior to the major kinematic reorganization of 23.4 Ma (chron 6C), when the Pacific-Farallon ridge rotated ∼10° clockwise [Tebbens and Cande, 1997].
 The first volcanoes of the Foundation Seamounts are located south of the Resolution fracture zone (FZ) (Figures 1 and 3). Ages decrease from 21 m.y. at Ampère seamount to 16 m.y. at Becquerel seamount [O'Connor et al., 1998] and are consistent with the age regression toward the younger Foundation Seamounts to the east. Some of the volcanoes tend to form elongated ridges. These westernmost Foundation Seamounts appear to be the oldest volcanoes of the chain since no significant volcanic structures can be clearly identified from satellite maps between them and the Austral chain. A recent work [McNutt et al., 1997] reports ages for volcanic ridges at the southeastern end of the Austral chain which could be consistent with an ancient activity of the Foundation hot spot. Geochemical evidence, such as elevated Pb isotopes, is still required to confirm this. Nevertheless, whether the westernmost Foundation Seamounts are the product of an incipient hot spot or of the reactivation of an older one, they represent the beginning of an important magmatic pulse of the plume. South of these edifices of the main chain, volcanism is distributed along two different trends (Figures 1 and 3): the Del Cano ridge and a southern E–W line of elongated edifices.
 The Del Cano ridge (Figures 1 and 3) is an elongated high roughly parallel to the direction of the neighboring fracture zones [Mammerickx, 1992; Tebbens and Cande, 1997]. It was interpreted either as the wall of a fracture zone [Mammerickx, 1992] or as a chain of small off-axis volcanoes following the relative plate motion direction [Tebbens and Cande, 1997]. Our survey reveals that the structure of the Del Cano ridge is far more complex than previously suggested by satellite altimetry, where the ridge appears as a quasi-continuous feature, composed of two large segments (Figure 1). Only the eastern segment was surveyed (Figure 4). It is formed by seven en echelon massifs flanked by basins. Two main structural directions are observed, the first corresponding to the ridge general orientation, 80°N, and the second corresponding to the trend of the en echelon massifs themselves. The latter changes progressively from 60°N to 40°N from west to east. The massif morphology is asymmetric and suggestive of tilted blocks, with the steepest scarps generally face NW. At the western edge of the surveyed zone a flat-topped volcano shows an oblate form suggesting only slight deformation, but several small volcanic cones scattered all over the area show signs of more intense shearing. Minimum depths increase eastward from 1300 m (top of the flat volcano) to 2300 m (top of the last eastern massif). The easternmost massifs are isolated highs, separated by deep and wide basins, while the westernmost massifs tend to form a more continuous feature. This pattern results from an increase in the deformation toward the eastern edge of the ridge. The sense of the en echelon displacement is coherent with a lateral, senestral strike-slip shearing. A bottom video camera profile on the top of the flat volcano revealed sedimentary and manganese covering. A single altered basalt sample coated with a thin manganese crust was recovered by a dredge haul at the eastern edge of the ridge. The degree of alteration and the manganese crust suggest that the feature is quite old. Its plagioclase composition [Hékinian et al., 1999] is characteristic of depleted tholeiites. South of the Del Cano ridge, an E-W line is formed by elongated volcanoes. This line splits near Boltzmann seamount (126°W) into three lines (Figure 3). The first line, including Laplace seamount, intersects the failed rift of the Selkirk paleomicroplate just south of the Buffon seamount, the second line intersects the failed rift near a small discontinuity at 35°30′S, and the third line intersects the failed rift at the intersection with the Mocha fracture zone.
 Our cruise surveyed part of the fossil axis identified by Mammerickx  which corresponds to the western boundary of the Selkirk paleomicroplate. A dredge haul on the western wall of the fossil axis recovered altered basalts, doleritic blocks, and volcanic breccia, coated with manganese crust up to 7-cm thick [Hékinian et al., 1999]. Volcanism is present on the northern (Buffon seamount) and western (Laplace seamount) parts of the fossil axis. Identification of magnetic anomalies west of the failed axis is difficult because of the volcanic overprint. On the east flank, anomalies between 6A and 6B are identified. The ridge activity probably ceased shortly after chron 6A (20.5–20.7 Ma) [Blais et al., 1999]. Samples from the fossil rift are depleted tholeiitic lavas with respect to their trace element compositions [Hékinian et al., 1999], as shown in Figure 5a through the (Nb/Zr)N ratio, but the 206Pb/204Pb ratio (Figure 5b) indicates a transitional composition. This slightly higher isotopic ratio suggests that the mantle sampled by this ridge axis near the end of its activity was already being fertilized by the Foundation plume, although this fertilization is still reduced.
 The largest volcano of the area is Buffon seamount, a peculiar feature 8.8 m.y. old, a young age compared to the neighboring edifices [O'Connor et al., 1998]. It is a huge triangular-shaped volcano, emplaced at the intersection between the main chain and the northern part of the failed rift, and in the prolongation of the Del Cano ridge. Its elevation reaches 4000 m above the seafloor, its summit topping at −480 m. East of Buffon seamount, Celsius seamount, 13 m.y. old [O'Connor et al., 1998], was built on the lithosphere of the Selkirk paleomicroplate (Figure 3). East of this edifice the volcanoes of the main body of the Foundation Seamount chain begin.
 In terms of (Nb/Zr)N composition and 206Pb/204Pb isotopic ratios the volcanoes of this part of the chain can be divided into two groups. The first group is formed by the northern edifices from Ampère to Becquerel seamounts. The lavas are basaltic in composition and reveal low (Nb/Zr)N values, being among the least enriched in incompatible trace elements of the whole chain, except for the ridge-hot spot interaction zone, near the PAR [Hékinian et al., 1997, 1999]. Because of a moderate magmatic differentiation, Aristotelis seamount samples resulted in basaltic andesites [Hékinian et al., 1997]. The second group includes Boltzmann, Laplace, Buffon, and Celsius seamounts. A trend of increasing values of (Nb/Zr)N and 206Pb/204Pb from west to east (Figures 5a and 5b) is interpreted as an increase in a chemically and isotopically enriched component in the lava source toward Celsius seamount. Lavas from this last volcano are alkalic and differentiated [Hékinian et al., 1997] and show high (Nb/Zr)N and 206Pb/204Pb ratios (Figures 5a and 5b). Both ratios show that Celsius seamount is undoubtedly the westernmost edifice of the chain built mostly from hot-spot-derived material.
 In summary, in the western province (1) the volcanoes are younger than the underlying lithosphere [O'Connor et al., 1998] and display an intraplate morphology, (2) (Nb/Zr)N and 206Pb/204Pb ratios show a subtle but growing contribution of enriched material in the source (this is compatible with the increasing influence of plume-derived material and with the presence of a hot spot at depth), (3) the Del Cano ridge is likely to be formed by shearing, either contemporaneous or slightly ulterior to a volcanic episode, (4) the activity in the west axis of the Selkirk paleomicroplate stopped around 20 m.y., and (5) Celsius seamount represents the first edifice of the Foundation chain built mainly of hot-spot-derived material.
4. Present-Day Ridge-Hot Spot Interaction Zone
 The present-day ridge-hot spot interaction zone (Figures 1 and 6) extends 400 km away from the PAR axis and across a 200–90-km-wide band. The edifice morphologies vary, but average summit depths increase toward the PAR while the total volume of the edifices decreases. The largest volcanoes are grouped into oblique-striking ridges. Volcanism is organized along two roughly parallel lines, the distance between them diminishing toward the PAR from ∼130 km to ∼80 km. A rough estimation of volcanic output volume for the north line yields 2–3 times the volume of the south line, suggesting that the northern line, which lies in the prolongation of the older Foundation volcanoes, is the main locus of volcanism. Magnetic modeling of the volcanoes [Maia et al., 1999] shows that the emplacement of both lines was synchronous and that the edifices become progressively younger toward the PAR, in agreement with published radiometric ages [O'Connor et al., 1998].
4.1. Off-Axis Volcanism
 Most of the western edifices of the northern line are high, well-formed cones (Linnè, Mendeleiev, Mercator, Newton, and Ohm seamounts), some displaying a flat summit, capped with coral and sand [Devey et al., 1997], revealing an aerial or near-surface position in the past (Figure 6). These volcanoes are grouped into long ridges striking obliquely to the main trend of the Foundation chain. Some of the volcanoes of the northern line (Linnè Ridge and Mendel seamount) and all of the southern group (Mohorovicic, Wegener, Platon, Richter, and Planck seamounts) show a morphology suggestive of fissure volcanism, with narrow elongated shapes and linear crests. The bathymetry shows that these edifices are built by several hundreds of small coalesced volcanic cones.
 A rapid change in both the morphology and the spatial distribution of the volcanoes takes place near 112°W. Edifices become significantly smaller. The northern line becomes a cluster of flat volcanic domes with summit calderas. In the vicinity of the spreading axis, spacing between the cones diminishes, and they coalesce to form smooth, linear elongated ridges trending N110°, hereafter named North and South Ridges after Devey et al. . The eastern end of the southern line is essentially formed by the elongated Planck seamount.
 In terms of (Nb/Zr)N ratios the oblique ridges of the northern line, from Linnè Ridge to Ohm seamount, are heterogeneous (Figure 7a). The western part (Linnè Ridge and Mendel seamount) appears enriched and hot spot derived, with high (Nb/Zr)N ratios, while the eastern part is heterogeneous and reveals the irregularly growing influence of a depleted component in nearing the spreading axis. The 206Pb/204Pb isotopic ratios (Figure 7b) confirm this picture. The highest 206Pb/204Pb values are observed at the western end of the northern line, on samples from Linnè Ridge and Mendel seamount. These elevated values are conclusive evidence that the plume-enriched component is dominant in these lavas. The 206Pb/204Pb ratio decreases toward the east until Mercator seamount. Newton and Ohm seamounts show a sharp decrease in both (Nb/Zr)N and 206Pb/204Pb ratios with values that are typical of depleted tholeiites. Undoubtedly, these two seamounts were emplaced from a depleted tholeiite source while Linnè and Mendel seamounts are clearly hot spot type.
 The equivalent section of the southern line is more homogeneous (Figures 7a and 7b). Mohorovicic, Wegener, Platon, and Richter seamounts have comparable intermediate chemistry between alkali basalts and enriched tholeiites [Hékinian et al., 1999]. Their intermediate (Nb/Zr)N ratios reveal a slight but significant enrichment. This section of the southern line is also isotopically intermediate between the enriched and the depleted terms of the northern line: The 206Pb/204Pb ratios show limited variations. The eastern end of this line, Planck seamount, is clearly depleted [Hékinian et al., 1997], with low (Nb/Zr)N and 206Pb/204Pb ratios revealing an obviously large input of depleted material in the source.
 Approaching the PAR, the compositional pattern remains rather complex. The cluster of volcanoes of the northern line is heterogeneous and produced lavas which are either clearly alkalic (Watt and Pascal seamounts) or transitional [Hékinian et al., 1997, 1999], with highly variable (Nb/Zr)N and 206Pb/204Pb ratios. Some values are as high as those observed for Mendel seamount (Figures 7a and 7b). The North Ridge is moderately heterogeneous and built of both tholeiites and transitional basalts [Hékinian et al., 1997], with (Nb/Zr)N and 206Pb/204Pb ratios still showing some hot spot input into an abundant depleted component. The South Ridge is homogeneous both in terms of (Nb/Zr)N and 206Pb/204Pb ratios and exhibits intermediate values between depleted and enriched compositions.
4.2. Pacific-Antarctic Ridge Axis
 The PAR axis in the Foundation area is broader and shallower than in the neighboring ridge segments. The ridge morphology is smooth and flat, characteristic of fast spreading rates. On the east flank of the PAR the lithospheric fabric is undisturbed by volcanic features, with the exception of a small off-axis volcano, Rutherford seamount [Devey et al., 1997]. At the axis, crust, as inferred from gravity analysis, is thicker than what is generally observed for fast spreading ridges [Maia et al., 2000]. Both the morphology and the crustal structure of the axis show that the ridge is under the influence of the neighboring Foundation hot spot. However, the influence is significantly weaker than what is reported for systems where an accreting ridge is moving away from a hot spot. This probably reflects a less efficient connection in the case where a ridge is moving toward a hot spot [Maia et al., 2000].
 A feature of the spreading axis chemistry is the existence of intermediate lavas (andesites and dacites; Hékinian et al. ). Four dredges of a total of nine have collected intermediate rocks, of which three are located where off-axis volcanoes join the ridge axis. For these dredges, (Nb/Zr)N ratios are not a good indicator of source composition. Four other dredges have recovered “flat” tholeiites, and one dredge has recovered depleted tholeiites [Hékinian et al., 1997, 1999]. (Nb/Zr)N and 206Pb/204Pb ratios are not as heterogeneous as those from the off-axis volcanoes, remaining in the range of intermediate values between those characteristic of the Foundation hot spot and those characteristic of the depleted component usually seen in upper-mantle-derived normal MORB (NMORB). In summary, the spreading axis does not exhibit compositions as low as could be expected in magmas classically derived from the essentially depleted upper mantle. Enriched material must contribute to the axial magmatism, confirming a plume influence.
 In summary, in the ridge-hot spot interaction zone, the following can be said: (1) Volcanism is distributed along two lines. The distance between them diminishes toward the PAR from ∼130 to ∼80 km. (2) The morphology of the volcanoes evolves from tall cones to small, flat domes, with the change taking place near 112°W. (3) Some morphologies, mainly for the edifices of the southern line, are suggestive of fissure volcanism. (4) In terms of composition the two lines are highly contrasted. The northern line is very enriched in its western half and can be as depleted as NMORB in its eastern part. On the contrary, the southern line is more homogeneous, neither as enriched nor as depleted as the northern line. (5) The morphology [Maia et al., 2000] and the composition of the basalts [Devey et al., 1997; Hékinian et al., 1997, 1999] of the PAR axis show clearly the influence of the Foundation hot spot. This influence is, however, weak, considering the short distance between the ridge and the hot spot [Maia et al., 2000].
5. Position of the Foundation Hot Spot Through Time
 Extremely fresh lava flows were not found off axis, and the precise location of the hot spot is not known. Lack of volcanism on the Antarctic plate strongly suggests, however, that the hot spot is still located west of the PAR. A clue to the location of the hot spot can be gained from the study of the geoid. Hot spots are associated with broad positive geoid anomalies, centered over the plume and elongated in the direction of absolute plate motion [e.g., Crough, 1983; McNutt, 1988; Monnereau and Cazenave, 1990; Sleep, 1990]. A clear high of 15-cm amplitude and elongated to the west-northwest is visible on the filtered geoid (Figure 8). The maximum of the anomaly is centered over the north group of smooth elongated volcanic ridges. The amplitude is considerably smaller than those observed for midplate swells [e.g., Crough, 1978; Fischer et al., 1986; McNutt, 1988; Monnereau and Cazenave, 1990] but is consistent with the observation that near-ridge swells have low-amplitude geoid anomalies [Monnereau and Cazenave, 1990]. Such a geoid pattern is a strong argument favoring the location of the hot spot beneath the north volcanic system which would correspond to the main locus of the volcanic activity through time.
 If we consider a hot spot to be a small narrow conduit of some tens of kilometers in diameter, a reconstruction using the ages of the main body of the chain (13–5 m.y.; O'Connor et al. ) will bring the locus of volcanic activity north of the observed geoid anomaly. It could be therefore inferred that the trend of the edifices of the young part of the Foundation chain diverges from that predicted by rotation poles not including a recent change in absolute motion for the Pacific plate. However, if hot spots are considered to be a broad (∼200 km) thermal anomaly [McNutt et al., 1989; Wolfe et al., 1997], taking the location of the maximum of the geoid anomaly as being the present location of the hot spot and deriving its past positions using the rotation pole derived by Lonsdale  (or the pole derived by Yan and Kroenke ), all the observed volcanism of the chain falls within the hot spot radius without any need of change in plate motion (Figure 9). A slight change southward in the last 5 m.y., as suggested by Cox and Engerbretson , would better fit the general trend of the young part of the chain, but this improvement in the fit is in no way dramatic. However, a model considering a drastic change, as proposed by Wessel and Kroenke , is incompatible with our geoid data.
 In this section we will address two problems: first, the lithospheric control on the surface expression and distribution of the volcanism and, second, the mixing between the hot spot plume and the upper mantle components. We show that for both the older and the recent parts of the Foundation chain the emplacement of the volcanoes is mainly controlled by weak zones in the lithosphere. The spatial distribution of the chemical composition of the lavas of both areas is, however, explained by two different processes. For the old volcanoes of the western province the mixing between the plume and the upper mantle material is linked to the arrival of the plume head or of a major magmatic pulse. For the recent edifices the mixing is more conveniently explained by the approach of the plume and the Pacific-Antarctic Ridge and, consequently, the merging of two melting zones.
6.1. Origin of the Volcanic Ridges of the Western Province of the Foundation Chain
 The spatial distribution of the volcanism appears to be strongly controlled by lithospheric weak zones and/or preexisting structures. Kinematics, structural features, ages, and chemistry of the volcanoes show that the volcanic features were formed during two different periods. The earlier period is probably linked to the plate reorganization of 23.4 Ma [Goodwillie and Parsons, 1992; Tebbens and Cande, 1997]. During this reorganization a first short magmatic episode appears to be at least partly contemporaneous to the deformation of the plate and to the development of fractures in the different observed directions. The main part of the Del Cano ridge was built during this episode. A second magmatic episode is related to the beginning of the activity of the Foundation hot spot and is subsequent to the plate reorganization. Preexisting fractures were conduits for releasing magma to the surface.
6.1.1. Plate reorganization: First magmatic episode
 The Del Cano ridge probably corresponds to a small or zero-offset discontinuity or fracture zone that underwent an oblique slip normal displacement, thus both shearing and extension, most probably induced by the major plate reorganization at the time of chron 6C (23.4 Ma) [Goodwillie and Parsons, 1992; Tebbens and Cande, 1997]. Indeed, Goodwillie and Parsons  find a certain amount of extension on the old Pacific plate which could be a consequence of this reorganization. Several observations support this interpretation. First, the main trend of the Del Cano ridge is subparallel to the trend of the fracture zones in this region, its eastern end corresponding to crust formed at chron 6C. This suggests that the Del Cano ridge corresponds to an ancient fracture zone or to an axial discontinuity which ceased to exist around chron 6C time. Second, the structure of the Del Cano ridge clearly shows a deformation posterior to the active period of the transform, mainly shearing associated to a small amount of extension. The presence of both intact and truncated volcanoes suggests that the deformation is partly synchronous with a volcanic episode along the ridge. Both the height and the lateral dimension of the massifs reduce progressively from west to east. The amount of shearing is also variable along the Del Cano ridge, as can be seen by the morphology of the western edifice, showing little deformation, and of the easternmost massifs, showing intense deformation. The chron 6C clockwise rotation of the spreading direction and the formation of the Selkirk paleomicroplate could induce a certain amount of northwest-southeast extension as a response to the intraplate stresses generated by the new kinematic configuration. If the pole of the opening is located near the failed rift, the amount of extension would be greater at the western end of the Del Cano fracture zone, allowing more magma to reach the surface, therefore forming higher relief. Third, the sample obtained at the eastern tip of the Del Cano ridge [Hékinian et al., 1999] shows that at least this part of the feature did not undergo a strong influence of the Foundation hot spot. Also, the depleted tholeiite trace element and isotope compositions of the basalts from the failed rift of the Selkirk paleomicroplate argue in favor either of the absence of a hot spot influence or of a very incipient hot spot influence in this region at least at chron 6A (∼20 Ma), the time of the extinction of the western spreading axis of the paleomicroplate [Tebbens and Cande, 1997; Blais et al., 1999]. Since the Del Cano ridge was probably formed before the extinction of the Selkirk ridge axis, it is very unlikely that it could be linked to a later hot spot activity.
 Fracture zones have been known to provide preferential pathways for the magma and to respond to changes in relative plate motions by either tension or compression (e.g., the Marquesas FZ [McNutt et al., 1989; Jordahl et al., 1995]). Short oblique ridges observed between the West Ridge of Easter microplate and Pitcairn island are also related to fissure or leaky transform volcanism [Maia et al., 1994; Searle et al., 1995]. Extension with a shear component was also evoked as a mechanism for the formation of other Pacific intraplate en echelon ridges such as the Cross-grain [Winterer and Sandwell, 1987] and the Pukapuka ridges [Sandwell et al., 1995]. The Del Cano ridge is thus most probably formed by a similar process.
6.1.2. Plate-hot spot interaction: Second magmatic episode
 Reconstructions using different published rotation poles [Lonsdale, 1988; Yan and Kroenke, 1993; Géli et al., 1999] and ages published by O'Connor et al.  show that all the volcanic constructions of the western area could have been emplaced as the result of the movement of the Pacific plate over a stationary hot spot. Figure 10 shows a possible emplacement sketch of the edifices of the second magmatic phase, which began around 21 m.y. ago at ∼500 km to the west of the failed rift. Ampère, Archimidis, as well as Laplace seamounts lie in a trend coherent with the main body of the Foundation chain (Celsius-Linnè seamounts, 13–5 Ma age range, [O'Connor et al., 1998]), although Ampère and Archimidis are located slightly north of the hot spot center. Aristotelis, Becquerel, Buffon, and Celsius seamounts lie on the northern edge of the hot spot track (see Figure 9). Boltzmann seamount is located near its southern edge. The apparent draining of plume material by weak zones inherited from intraplate deformation can explain the irregularity of the trend as well as the widespread volcanism along this old part of the chain.
 From the spatial distribution of the (Nb/Zr)N and 206Pb/204Pb ratios (Figure 4) we can infer that the influence of the hot spot has progressively increased between 21 Ma (Ampère) and 13 Ma (Celsius). The compositions of the southern seamounts are more enriched than those of the northern ones and suggest that they contain more of the enriched component brought by the plume. This complex pattern cannot result from the decreasing influence on these melts of a neighboring spreading ridge axis as proposed by Devey et al.  and Hékinian et al. [1997, 1999]. The Selkirk paleomicroplate western axis was already extinct when most of these volcanoes formed, except for Ampère seamount, emplaced roughly 500 km away. The nearest active ridge axis, located north of the Resolution fracture zone (Figure 10) was located far away from the hot spot (∼300 km). Moreover, the distance between both did not vary significantly during the time of the building of these edifices, and an important variation in distance is required to explain the compositional trend.
 We therefore suggest that the whole area was influenced by the hot spot to various extents (Figure 11). Ampère seamount, which was located on the northern edge of the hot spot footprint, is derived from an asthenosphere-lithosphere which was moderately contaminated, probably by some of the first melts released by the plume pulse. Ampère lavas could then result from partial melting of this contaminated asthenosphere-lithosphere heated from below. This process may have been initiated through heat transfer from the rising plume to the lithosphere. The first melts derived from the plume are likely to have fertilized the asthenosphere-lithosphere in order to account for the chemical characteristics of the lavas. Aristotelis lavas could be generated by the same process, the seamount being located farther north and consequently less influenced by the hot spot. Becquerel's heterogeneous results may be related to the presence of a major fracture zone in its vicinity (Resolution FZ). The cold wall effect may account for the heterogeneity by allowing various unmixed partial melts to be erupted. The fact that this seamount was also built rather far away from the center of the plume supports the concept of small batches of partial melts erupting without being homogenized. Compared to Becquerel seamount data, the greater influence of the plume seen in Boltzmann and Laplace seamounts can be related to the following: (1) The plume has probably reached the mantle region involved in the partial melting, and (2) these seamounts were probably built closer to the center of the plume-contaminated area.
 Both the size and the anomalous age of Buffon seamount (9 Ma) may be related to its location on a very weak zone of the lithosphere. Its samples present ages which are too young compared to the neighboring edifices [O'Connor et al., 1998]. This is possible only if the samples represent late activity of a volcano whose construction started contemporaneously with the neighboring seamounts, i.e., around 16–13 Ma and/or during the final phase of extension of the failed rift (around 20 Ma). It is indeed possible that this seamount was built during three successive or near-continuous phases of volcanism. This long activity may be linked with the location of Buffon seamount, which sits right on the failed western axis of the Selkirk paleomicroplate. Its extended activity would have ceased only when hot-spot-derived material was no longer entrained toward the weak lithospheric zone underneath the failed axis, ∼9 m.y. ago (Figure 11). This is consistent with the existence of hot spot material channeling as proposed by Schilling [1985, 1991] when a ridge is moving away from a hot spot. Its chemistry undoubtedly shows input of hot-spot-derived material in its source. Magmas may have been drained by the slope of the lithosphere for ∼4 m.y. until the distance between plume and rift was large enough to end the connection. The age of Galilei seamount, 9 Ma [O'Connor et al., 1998], 300 km to the east (see Figure 1 for map location), tells us where the plume was located when Buffon's activity ceased.
6.2. Origin of the Volcanic Ridges of the Recent Parts of the Foundation Chain
 The spatial distribution and morphological evolution of the volcanic ridges in the youngest part of the Foundation chain show that heterogeneity in the strength of the lithosphere partly controls the surface expression of the hot spot volcanism. Several observations indicate that the emplacement of the southern volcanic line was controlled by the flexural response of the lithospheric plate to the loading of the northern line. First, the ridges form two subparallel systems, emplaced nearly simultaneously and separated by a distance that decreases toward the PAR. Both systems display the same morphological evolution from tall cones to small dome-shaped volcanoes. Edifices belonging to the south system are elongated, suggesting emplacement along fissures or cracks. Moreover, the volume of volcanic products differs between the two lines, the southern one being significantly smaller than the northern one. Second, the local maximum of the geoid (Figure 8) is located over the two smooth ridges of the north line, west of the PAR axis, which strongly implies a location of the Foundation hot spot beneath the north system. This is consistent with the observation that this north system lies in the prolongation of the older seamounts of the chain, while the south line formed only recently (<5 Ma). The observed differences in volume are also consistent with a northerly location of the hot spot, where the most vigorous melting regime, and therefore the larger lava production, would have occurred. Figure 12 shows a sketch representing the evolution of the ridge-hot spot interaction over the past 5 m.y. If a 200-km-wide thermal anomaly is considered, the south volcanic line could have formed on its southern edge as a secondary system of magmatic activity, erupted along fissures or cracks.
 The volcanic load of the north system deformed the plate. Extensional stresses associated with the upward bending of the elastic plate as a response to this loading may help to focus magma at the top of the flexural bulge [Sleep, 1996; McNutt et al., 1997; Hieronymus and Bercovicci, 2000], resulting in the formation of the south line. For this part of the Foundation chain the age of the lithosphere at the moment of the formation of the volcanoes decreases from ∼5–4 Ma for the westernmost edifices of this area (Linnè and Linnè Ridge) to less than 1 Ma for the dome-shaped volcanoes of the smooth ridges near the PAR [Maia et al., 1999]. This reduction in the age of the plate is consistent with the approach of the Pacific-Antarctic Ridge to the Foundation hot spot at a rate of 42 mm/yr, given by the HS2-NUVEL-1 model [Gripp and Gordon, 1990]. We may therefore expect the distance between both systems to reduce toward the PAR axis as a result of a decrease in the rigidity of the lithosphere at the time of formation of the volcanoes and, consequently, of the flexural wavelength [Walcott, 1970a, 1970b]. If the distance between the two systems can be considered to be the distance between the load (the north line) and the maximum amplitude of the flexural bulge (the south line), the flexural parameter corresponding to the observed values (∼130–80 km) would range from 40 to 23 km, yielding mechanical thickness between 12 and 6 km, which is consistent with the plate ages [Walcott, 1970a, 1970b; Turcotte and Schubert, 1982; McNutt, 1984]. These are only approximate figures, since recent models have shown that the interplay of tectonic and flexural stresses is very complex, especially for very young and thin plates [Hieronymus and Bercovicci, 2000]. The lack of volcanism north of the northern line could be explained by the presence of the Resolution-Challenger FZ, which could release at least part of the flexural stresses and prevent fracturing of the plate. The sketch presented in Figure 12 also shows that in the past the hot spot could have been located slightly to the south of the northern main volcanic line, therefore explaining the asymmetry of the flexural arch volcanism. In this case, the double system of the Foundation is analogous to the Taukina-Ngatemato ridges mapped by McNutt et al.  at the southeastern end of the Austral chain. According to these authors the Taukina ridge was emplaced on the flexural arch of the bending produced by the Ngatemato ridge. It is worth noting that the Taukina is considerably smaller than the Ngatemato ridge in both volume and size of the edifices, in a striking analogy to the differences between the north and the south lines of the Foundation.
 The existence of diffuse extension for the young part of the Pacific plate is still a matter of debate [Goodwillie and Parsons, 1992; Sandwell et al., 1995]. It is possible that a recent change in the absolute motion of the Pacific plate induced tensile stresses able to deform the young plate [Wessel and Kroenke, 1997]. There is some evidence, mostly derived from the eastern end of the Hawaiian chain, for such a change around 5–3 Ma [Cox and Engerbretson, 1985], but the question remains completely open [Wessel and Kroenke, 1997; Géli et al., 1999]. The age progression implies that the cracks allowing the Foundation volcanic ridges to form opened progressively. This differs from other Pacific ridges, such as Pukapuka, Cross-grain, or Rano Rahi, where no clear age progression was observed [Winterer and Sandwell, 1987; Sandwell et al., 1995; Scheirer et al., 1996]. Therefore extension, as evoked for the formation of other intraplate en echelon ridges, fails to account for all the observations in the eastern part of the Foundation chain.
 In summary, the flexural model explains the distribution of the volcanism in two systems. If the Pacific plate is undergoing a diffuse extension, this is likely to contribute to the local fracturing of the plate. Extension alone is, however, unable to explain all the observations for the young part of the Foundation chain.
6.3. Mixing Hot Spot and Spreading Axis Materials
 The spatial distribution of the chemical and isotopic composition of the northern line lavas shows that the dilution of the hot spot signature does not display a regular pattern. The hot spot signature peaks at Linnè Ridge and Mendel seamount. The variability in composition, which is highly heterogeneous, begins with Mercator seamount and continues until the spreading axis (Figure 7). It clearly means that a less enriched component gets involved in the source of these lavas and partly dilutes the hot spot signature. The pattern is very different from that observed for the Easter-Sala y Gomez seamount chain, where an almost linear trend of mixing between hot spot and upper mantle materials seems to exist [Kingsley and Schilling, 1998]. Compared to the northern line, the southern line is less heterogeneous, although there is a chemical variation from west to east (Figure 7). Such relative homogeneity implies a larger contribution of the surrounding depleted upper mantle. This is in agreement with the preferred geophysical model for the formation of the southern system. The opening of fissures due to the flexure of the plate has probably induced decompression melting in the mantle underneath the bulge, which was already contaminated by plume-derived material. This is consistent with a location near the edge of the hot-spot-fertilized area.
 The pattern observed in the northern line of the eastern Foundation apparently does not support the MPS-MRS model of Schilling [1985, 1991] of mixing along a sublithospheric channel connecting the plume to the spreading ridge axis. This model was based on studies of systems where a ridge is either located above or moving away from a hot spot. For these cases, the model conveniently explains both the off-axis and the on-axis observations (chemical compositions and ridge crustal structure and morphology) by a progressive mixing of hot-spot- and ridge-derived materials. However, the Foundation-PAR is a system where a spreading ridge approaches the hot spot. This difference in the relative movement strongly influences the dynamics of the interaction. According to recent models, when a spreading axis moves toward a hot spot, the relative movement, by entraining plume material away from the ridge, will hinder the connection, therefore reducing the distance at which both sources begin to interact [Ito et al., 1997; Ribe and Delattre, 1998]. In such a case, mixing probably happens in a rather inefficient way which preserves chemical heterogeneities in the samples at a rather small regional scale.
 The observed chemical pattern of the whole area shows that the interactions between the hot spot and the spreading axis started fairly recently. For example, a volcano as young as Mendel (4.5 Ma) does not show traces of mixing with ridge-derived products. From Mohorovicic seamount and eastward, lavas become heterogeneous and never as hot-spot-like as those at Mendel. This observation and the kinematic reconstruction (Figure 12) show that this mixing started only around 4 Ma. A drop of cumulated magmatic volume of the volcanoes, at ∼2 Ma age of loading, signifies the partial capture of the hot spot magmatic production by the spreading axis. This is supported by the observation that the oceanic crust is thicker by 1–1.5 km on the eastern flank of the PAR at least up to 1 Ma [Maia et al., 2000]. Both off and on axis the mixing taking place in the common melting zone is not very efficient, differently from the relatively efficient mixing considered in Schilling's MPS-MRS model (as observed between Easter hot spot and the East Pacific Rise [Kingsley and Schilling, 1998]). It preserves chemical heterogeneities as seen in the volcanic products erupted in the off-axis volcanoes. This could mean that this kind of mixing results from the introduction of partially molten blobs of hot-spot-derived material into the spreading axis melting zone. The modern lavas of the spreading axis demonstrate a major input of hot-spot-derived material into the ridge. This input seems to be larger for the northern system and particularly strong where the off-axis ridges join the axis. This is consistent with a more homogeneous composition for the southern line, with less input of hot spot material, and supports our favored model of lithospheric flexure for the origin of the double volcanic line.
 1. Lithospheric zones of weakness play a major role in the surface expression of the volcanism, at both the ancient and the young parts of the Foundation Seamount chain. Both areas were weakened, either by deformation linked to a kinematic change (old zone) and/or by deformation of the plate due to local flexural stresses (young zone). The kinematic reorganization of 25–20 Ma happened in the ancient area when the lithosphere later affected by the hot spot was young. The plate was therefore weak and easily deformed. The surface expression of the hot spot was therefore strongly governed by the distribution of fissures and cracks. A reasonably wide thermal anomaly could therefore be responsible for the volcanoes of the second phase. In the young part of the chain such a wide thermal anomaly can also explain both the north and the south volcanic systems. The young plate was deformed by the flexural stresses resulting from the emplacement of the northern volcanoes. The southern system formed along the cracks on top of a flexural bulge generated by the emplacement of the northern system.
 2. The comparison between the spatial distribution of the lava composition and the kinematic reconstructions shows that the processes acting in the ancient and in the recent parts of the Foundation chain are distinct. The composition of the basalts in the ancient part probably results from an interaction between a plume pulse and the surrounding mantle. The observations are not compatible with a model of interaction between a plume and a spreading axis. They rather support a growing influence of an upwelling mantle plume beneath the area. In the recent part the observations favor a model of interaction between a plume and a spreading ridge. However, they do not support an interaction of the kind suggested by the MPS-MRS model. The mixing between the two sources started only when both melting zones overlapped. This may be specific to the case when a spreading axis approaches a hot spot.
 The SO-100 and the Foundation Hotline cruises were part of a joint French-German program for the study of the South Pacific intraplate volcanism. The cruises were supported by the BMFT (Germany) and by CNRS-INSU (France). The authors would like to thank C. Bollinger, who helped with the isotope analysis. We also thank R. Hékinian for kindly providing unpublished data on trace elements to complete our data set and J.-Y. Royer for allowing us the use of his code for the kinematic reconstructions. This work benefited from discussions with J. Dyment and N. Ribe. J. Goslin, R. Maury, and A. Hofmann kindly read preliminary versions of the text and made helpful suggestions. C. Chauvel, W. White, P. Janney, B. Hanan, and M. McNutt made careful reviews. Their comments significantly improved the original manuscript. Several figures were drawn using the GMT software of Wessel and Smith . This study was funded by CNRS-INSU through the Intérieur de la Terre, PR70 “Océans” program, grant 97N51/0349.