A prerequisite for the reconstruction of transport pathways is the identification of specific source areas on the Antarctic Peninsula and the adjacent islands. However, the heterogenous petrology of magmatic arc rocks and associated units as well as the extensive ice cover make this identification difficult. The investigation of the clay mineral distribution in surface sediments of the Atlantic sector of the Southern Ocean reveals important clues for mapping specific clay mineralogical source areas. Clays supplied by the rocks exposed on the Antarctic Peninsula represent a significant source for chlorite and smectite, which are transported to the shelf break by surface currents and glaciomarine processes. There, chlorite and smectite are injected into the eastward-flowing ACC and then exported to the Atlantic sector (Petschick et al., 1996; Diekmann et al., 1996).
A detailed analysis of the clay mineralogical assemblages deposited on the shelf northwest of the Antarctic Peninsula reveals a varied pattern of clay mineralogical provenance. Smectite concentrations are highest (>60%) in surface sediments in the far northeast (Fig. F4). This area is referred to as the "smectite province." Detrital and/or authigenic smectite is probably supplied by Mesozoic volcanic arc rocks and by associated volcano-sedimentary forearc basin strata exposed in Graham Land, on the adjacent islands, and possibly within the shelf (Pirrie, 1991; Petschick et al., 1996). Because these rock units, however, occur along the whole axis of the Antarctic Peninsula, the smectite maximum off northern Graham Land requires an additional source.
Potential further sources are the Cenozoic volcanic, pyroclastic, and volcaniclastic deposits outcropping mainly on the South Shetland Islands, Brabant Island, and Anvers Island (Petschick et al., 1996; Hillenbrand, 2000). This is supported by a smectite content of 95% in sediments near the active volcano on Deception Island (Fig. F4). In the South Shetland Islands area, submarine weathering of Cenozoic tephra and volcanic rocks producing smectite may play an important role (Petschick et al., 1996). Furthermore, some of the alkaline volcanics exposed in northern Graham Land and on the adjacent islands were extruded subglacially (e.g., Smellie and Skilling, 1994). This process is known to represent a very effective mechanism for smectite formation. In contrast, authigenic smectite found in volcanogenic soils on the South Shetlands was shown to be produced by hydrothermal alteration rather than by chemical weathering (Blümel et al., 1985). Therefore, soils in the South Shetland Islands area only represent a secondary source for smectite.
We consider that the highest smectite concentrations within the smectite province are probably restricted to the shelf northwest and southwest of the South Shetland Islands, as indicated by clay mineral studies of Shuitu (1990). Surface sediments taken from the inner parts of the Bransfield Strait, in contrast, contain more illite and chlorite (Shuitu, 1990; Yoon et al., 1992). Two samples in the smectite province are characterized by relatively low smectite but higher chlorite concentrations (Fig. F4). These samples were taken from the deeper Bransfield Basin, below 850 m water depth. An increase in chlorite near Smith Island was reported by Yoon et al. (1992), who assume chlorite supply from Paleozoic metamorphic rocks exposed on that island. We speculate that detrital supply from similar basement rocks known to crop out on the steep margins of deep basins within the Bransfield Strait (Jeffers and Anderson, 1990) cause the relative chlorite increase observed in our samples.
The shelf sediments south of 64°S contain less smectite and more chlorite and illite, forming a "chlorite-illite province" (Fig. F4). Whereas chlorite is the most abundant clay mineral offshore southern Graham Land, illite is dominant northwest of Alexander Island. Chlorite and illite are predominantly supplied by physical weathering under the polar climatic conditions of basement metamorphics and igneous rocks of the Andean Orogen. Altered calc-alkaline volcanics might represent the most important source for chlorite, as indicated by petrographic analyses (West, 1974; Hoecker and Amstutz, 1987). Lower smectite concentrations in the chlorite-illite province with respect to the smectite province possibly point to the rare occurrences of Cenozoic volcanics in the southern Antarctic Peninsula area.
Within the chlorite-illite province, illite concentration increases toward the southwest but only two surface sediment samples were investigated from the shelf northwest of Alexander Island. Both have illite contents around 60%. Together with illite derived from gneissic rocks, which are widespread in northern Palmer Land (Davies, 1984; Smith, 1987), considerable amounts of illite may be supplied from a Tertiary batholith exposed in the Rouen Mountains at the northern tip of Alexander Island (Care, 1983). This suggestion is confirmed by nearby recovered basal tills, which are virtually monolithologic and comprise quartz-mica schistose/gneissic pebbles interpreted to be derived from the Rouen Mountains (Kennedy and Anderson, 1989). Therefore, we speculate that the illite dominance in the shelf sediments northwest of Alexander Island might represent a restricted local signal.
Unfortunately, no clay mineral data for surface sediments from the Marguerite Bay area were available. To get an idea of the representative clay mineral assemblages of this shelf region, we investigated samples from ODP Site 1097 located at the northern end of Marguerite Trough (Figs. F1, F4). Marguerite Trough was eroded by ice streams, which had advanced from Marguerite Bay northward during glacial periods, draining both northern Palmer Land and eastern Alexander Island (e.g., Pope and Anderson, 1992). The mean clay mineral assemblage of 20 samples from Site 1097 consists of 54% chlorite and 36% illite, indicating that the bedrock in northern Palmer Land/eastern Alexander Island represents a source for detritus enriched in these clay minerals. Nevertheless, illite content at Site 1097 is generally higher than in the northern part of the chlorite-illite province. The higher illite concentrations are probably supplied by the gneissic rocks exposed in northern Palmer Land.
The clay mineral assemblages in surface sediments on the continental rise offshore from the chlorite-illite province exhibit a clear enrichment in smectite (Fig. F4). At present, glaciogenic detritus derived from the Antarctic Peninsula and the adjacent islands is released into the sea near the coast and transported farther offshore by tidal and wind-driven circulation (e.g., Domack and Ishman, 1993; Ashley and Smith, 2000). We consider terrigenous particles released in the Bransfield Strait not relevant to sedimentation on the slope and rise because they are probably carried to the northeast by a cyclonic current located within the strait (Hofmann et al., 1996). Fine-grained components of the glaciogenic detritus derived from the northwestern coasts of the South Shetlands as well as from the coasts in the chlorite-illite province, however, may travel across the shelf driven by currents. Thereby, circulation patterns on the shelf suggest meandering rather than straight transport pathways to the shelf edge (Smith et al., 1999).
We expect clay-sized particles settling down through the water column above the slope and rise offshore from the smectite province to be transported southwestward by the current regime measured there (Nowlin and Zenk, 1988). In contrast, fine-grained particles settling down above the continental slope west of 63°W, including those derived from the chlorite-illite province, may be deposited more or less in situ because of the lack of along-slope currents (Whitworth III et al., 1998). Offshore from Graham Land, intense erosional features reported from the 16° steep upper continental slope by Vanneste and Larter (1995) may point to a still-recent activity of turbidity currents there. We assume, however, that the possible supply of terrigenous detritus to the rise by turbidity currents only plays a minor role at present.
Bottom-current regime on the upper continental rise is dominated by general southwestward flow in the whole study area (Nowlin and Zenk, 1988; Camerlenghi et al., 1997). Therefore, fine-grained particles settling down to the upper rise offshore from the smectite province and from the chlorite-illite province are entrained into this bottom current. We expect at least some of the particles to be transported farther to the southwest before deposition. Consequently, the clay mineral assemblage in the rise sediments offshore from the chlorite-illite province comprises higher amounts of smectite, which was injected into the bottom current at the continental rise offshore from the smectite province (Hillenbrand, 2000; Pudsey, 2000).
Late Quaternary variations of clay mineral assemblages deposited at the Antarctic Peninsula's continental rise show a remarkable glacial-interglacial cyclicity with enhanced smectite concentrations during interglacial periods and high chlorite contents during glacials (Hillenbrand, 2000). At gravity core site PS1565, which is located offshore from the chlorite-illite province (Fig. F4), these variations are depicted for the last 130 k.y. (Fig. F5). Smectite contents increase at the transition between oxygen isotope Stages 6 and 5, are high during the early as well as late Stage 5, and decrease during Stages 4-2. Smectite increases again at the transition between Stages 2 and 1. Chlorite concentrations mirror the smectite variations inversely, thereby reaching maximum values during the late period of Stages 4-2, which possibly represents the last glacial maximum. In contrast to the variations of smectite and chlorite, no clear correlation of the illite content with global climatic cycles is visible. The observed cyclicity in smectite-chlorite fluctuations seems representative for the whole continental rise area between Drifts 3 and 5 (Pudsey, 2000).
During interglacial periods, smectite is delivered from the smectite province to site PS1565 by bottom-current transport along the continental rise. The decrease of smectite during glacial periods may be explained by two different mechanisms. The first is a consequence of previously suggested climate-induced changes in depositional processes on the margin (e.g., Pudsey and Camerlenghi, 1998; Rebesco et al., 1998; Hillenbrand, 2000; Pudsey, 2000). During glacial periods, grounded ice streams advanced across the shelf of the chlorite-illite province, eroded predominantly chlorite-rich detritus, and transported the debris seaward. The supply of eroded detritus to the outer shelf and the shelf edge increased, particularly during glacial maxima, when grounded ice streams had reached the shelf break. Rapid accumulation of unsorted debris near the shelf edge caused gravitational instabilities. These instabilities triggered slumps and slides on the upper slope, which turned into turbidity currents at the base of the slope. The turbidity currents cut channels into the upper continental rise and transported most terrigenous detritus down to the lower rise and to the Bellingshausen abyssal plain. Fine-grained particles within the suspension cloud of the turbidity currents, however, were captured by the southwestward-flowing bottom current and were deposited at the drifts as well as in the adjacent rise areas. Consequently, the higher input of detritus derived from the chlorite-illite province into the bottom current during glacial periods resulted in a higher proportion of chlorite within the clay mineral assemblage deposited at site PS1565. During the glacial-interglacial transitions, grounded ice streams retreated and thus the supply of chlorite-enriched debris to the slope and rise decreased.
On the other hand, a weakening of bottom-current intensity during glacial periods might have restricted the supply of smectite-rich detritus to the continental rise off southern Graham Land. The hypothesis of a glacial reduction in bottom-current strength may be supported by reconstructions of paleo-grounding lines in the Weddell Sea, indicating that no large ice shelves existed there during glacial maxima. Instead, grounded ice had advanced at least to the outer shelf (e.g., Bentley and Anderson, 1998). At the present time, dense bottom-water masses in the Weddell Sea, which crucially contribute to the production of Weddell Sea Deep Water (WSDW) via mixing processes, are derived from super-cooled shelf waters forming beneath the large ice shelves (e.g., Orsi et al., 1993). Widely grounded ice on the Weddell Sea shelf during glacial times would have limited the formation of WSDW and weakened bottom-water flow on the continental rise west of the Antarctic Peninsula. Any glacial reduction of bottom-water flow should not only have reduced southwestward transport of smectite but also should have caused a significant increase in deposition of clay-sized particles.
In core PS1565, there is only a weak correlation between changes in clay contents and variations of chlorite (Fig. F5). In principle, the proportion of the clay fraction does not reflect glacial-interglacial cyclicity. More sophisticated grain-size analyses, which were carried out on the upper Quaternary sequences recovered at Sites 1095 and 1096 by Pudsey (Chap 12, Chap 25, this volume) show a slight decrease of the clay fraction during interglacials. This decrease, however, may at least be partly caused by an increase of silt-sized diatoms diluting the proportions of terrigenous clay. The diatoms were primarily supplied by vertical settling during interglacial times, independent of changes in bottom-current strength. Therefore, we suggest that glacial-interglacial cyclicity observed in the clay mineral assemblages deposited on the continental rise off southern Graham Land was predominantly controlled by changes in the supply of chlorite.
According to the existing age model, the stratigraphic sequence of Site 1095 reaches back to ~9.3 Ma and that of Site 1096 to ~4.7 Ma. Throughout these intervals, the clay mineral composition is characterized by major fluctuations of the individual clay minerals, particularly smectite and chlorite. The clay mineral assemblages alternate between two end-member assemblages (Figs. F6, F7). The first assemblage is characterized by smectite concentrations <20% and chlorite concentrations >40%. The second assemblage has smectite concentrations >20% and chlorite concentrations <40%. The illite concentrations in the sediments recovered at Sites 1095 and 1096 are generally higher than in gravity core PS1565 because the ODP sites are located offshore from the southern part of the chlorite-illite province where illite is more abundant (Fig. F4). As in core PS1565 sediments, the illite fluctuations cannot be easily correlated with fluctuations in the other clay minerals.
In general, a very similar pattern in the clay mineral distribution to that described in core PS1565 can be observed in the drill cores from Sites 1095 and 1096 if one takes into account the different sample spacing. The short-term cyclic changes in the clay mineral assemblages deposited at Sites 1095 and 1096 occur throughout the late Neogene and Quaternary (Fig. F6) independent of changes in the clay contents. The smectite-poor assemblages are interpreted to indicate more glacial conditions, whereas smectite-rich assemblages may indicate more interglacial conditions (Fig. F7). We suppose that the short-term fluctuations of clay mineral assemblages at Sites 1095 and 1096 were caused by the repeated advance and retreat of grounded ice masses across the shelf of the Antarctic Peninsula, resulting from changes of Antarctic ice volume.
Antarctic ice volume changes are thought to have influenced the oxygen isotope record in benthic foraminiferal tests on Milankovitch timescales since the early Pliocene (Shackleton, 1995; Barker et al., 1999). There is some debate, however, whether ice volume in Antarctica already fluctuated on orbital timescales during the late Miocene (Barker et al., 1999). For the time interval between 7 and 5 Ma, benthic 
18O fluctuations with the 41-k.y. period of orbital obliquity were reported both from Morocco (Hodell et al., 1994) and from the equatorial West Atlantic (Shackleton and Hall, 1997). Whereas those from Morocco were interpreted to be controlled, at least in part, by Antarctic ice volume changes (Hodell et al., 1994), those from the Atlantic were assumed to exclusively reflect changes in deep-water temperatures (Shackleton and Crowhurst, 1997).
The pronounced short-term fluctuations in the clay mineral record at Site 1095 strongly suggest that Antarctic ice volume changes already occurred during the late Miocene. This might support the interpretation of Hodell et al. (1994). We consider, however, that our sample set from Site 1095 yields too low a stratigraphic resolution for time-series analyses in the obliquity band. Moreover, the assumption of periodical expansion and contraction of the Antarctic Peninsula ice cap may imply that the Antarctic ice sheet had reached a size to make it sensitive to sea-level changes, which might have been internally generated (Barker et al., 1999).
In contrast to other sedimentological parameters (e.g., the opal record) (Hillenbrand and Fütterer, Chap 23, this volume), we observe only weak long-term changes in the clay mineral assemblages deposited on the continental rise west of the Antarctic Peninsula. At Site 1095, smectite contents slightly increase at ~7.0 Ma and decrease at ~5.4 Ma, fluctuating afterward around a higher mean value than before ~7.0 Ma (Fig. F6A). Chlorite variations reflect this pattern inversely. The increase of smectite roughly coincides with an intensification of Neogene volcanism in the northern Antarctic Peninsula region (Smellie, 1990), which ultimately may have resulted in a greater supply of smectite-enriched clay to Site 1095. A more pronounced intensification of volcanism in the northern Antarctic Peninsula region, however, took place during the Pliocene and Pleistocene (Smellie, 1990). Therefore, we conclude that it is improbable that intensified late Neogene volcanism in the smectite province represents the only cause for the slightly enhanced smectite concentrations between ~7.0 and ~5.4 Ma.
Because grain-size distribution at Site 1095 gives no clear evidence for a significant change in paleobottom-current strength during the late Miocene (Pudsey, Chap 12, this volume), we speculate that local changes in glacial erosion linked to tectonic processes have contributed to the increase of smectite at Site 1095. The occurence of redeposited neritic diatoms in the drift sediments at Site 1095 points to a shallowing of the Antarctic Peninsula's shelf between ~7.8 and ~6.2 Ma (M. Iwai, pers. comm., 1999). No global sea-level fall is known from that time interval (e.g., Abreu and Anderson, 1998). Therefore, the sea-level lowstand recorded at Site 1095 must have been a consequence of local uplift, which probably resulted in an intensification of glacial erosion in the corresponding hinterland. South of 64°S, the Cenozoic subduction of oceanic crustal segments below the Pacific margin of the Antarctic Peninsula had ceased at ~14.1 Ma, at the latest (Larter and Barker, 1991; Larter et al., 1997). Uplift related to the ridge crest-trench collision, which was concentrated on a relatively narrow zone on the shelf, lasted only for 1-4 m.y. (Larter and Barker, 1991; Larter et al., 1997). Thus, the chlorite-illite province should have been in subsidence since 10 Ma and glacial erosion should have remained rather constant during the late Miocene.
In contrast, ridge-crest segments collided with the paleo-trench offshore from Anvers Island and Brabant Island from 11.2 to 3.3 Ma and subduction in the South Shetland Trench may still continue recently (Larter and Barker, 1991; Larter et al., 1997). Most collision ages offshore from the southern part of the smectite province date at ~6.6 Ma. We assume that in the southern smectite province, uplift processes related to ridge crest-trench collision started already at ~7.8 Ma, causing a local sea-level lowstand. As a further consequence, glacial erosion of uplifted smectite-bearing rocks intensified at 7.0 Ma. The strengthening of erosion resulted in an increased supply of detritus enriched in smectite to the continental rise, where bottom currents carried the fine-grained debris southwestward to Site 1095. Subsidence in the southern part of the smectite province started only 1 Ma after the ridge crest-trench collision. At ~6.2 Ma, continued subsidence and repeated advance of grounded ice masses during cold periods had deepened the shelf to a water depth in which benthic diatoms could not survive. Glacial erosion in the southern smectite province had weakened to the pre-uplift level at ~5.4 Ma, but intensified volcanism in the northern Antarctic Peninsula region during the late Neogene produced more smectite, so that smectite supply to Site 1095 did not decrease to the mean value recorded before ~7.0 Ma.
At Site 1096, illite contents slightly increase between ~1.5 and ~0.2 Ma, whereas illite concentrations at Site 1095 remain rather constant (Fig. F6). This suggests that during most of the Quaternary, northern Alexander Island, which represents the direct hinterland of Drift 7, was a more important source for fine-grained detritus supplied to Site 1096 than northern Palmer Land/eastern Alexander Island. Indications for changing contributions of local source areas within the southern chlorite-illite province may be derived from reconstructions of paleo ice-stream pathways on the shelf. Rebesco et al. (1998) inferred ice flow in Marguerite Trough during glacial periods exclusively in a northerly direction, indicating that glacial detritus from northern Palmer Land/eastern Alexander Island was predominantly supplied to the slope adjacent to Drift 5. Bentley and Anderson (1998), in contrast, concluded from grounding zone wedges, which they found on the shelf northwest of Alexander Island, that ice streams also had spread out from Marguerite Trough to the outer shelf break adjacent to Drifts 6 and 7.
We suppose that throughout the late Neogene and Quaternary, glacial debris was delivered from the southern chlorite-illite province to the outer shelf and the shelf edge via the ice stream pathways suggested by Bentley and Anderson (1998). Between ~1.5 and ~0.2 Ma, however, the ice streams passing northern Alexander Island may have reached the shelf break adjacent to Drift 7 more often. This assumption is supported by the relative age relationships between the grounding zone wedges found on the shelf. They indicate that grounded ice streams advanced to the outer shelf during the last glacial maximum, whereas they had reached the shelf edge during older glacial periods (Bentley and Anderson, 1998). Moreover, detailed grain-size data show silty layers to be more abundant at Site 1096 between ~1.7 Ma and ~100 ka (Pudsey, Chap 25, this volume). These silty layers are interpreted as distal turbidites (Barker, Camerlenghi, Acton, et al., 1999), which were probably triggered by more frequent advances of grounded ice streams to the shelf break adjacent to Drift 7. The increased activity of turbidity currents at the continenal margin should have enriched the bottom current on the rise with glacial debris enriched in illite. While the illite-rich detritus was predominantly deposited at proximal Site 1096, the bottom current supplied more fine-grained detritus derived from the northern part of the chlorite-illite province to the distal Site 1095.
In the late Neogene to Quaternary variations of clay mineral assemblages at Sites 1095 and 1096, we do not recognize pronounced long-term changes caused by major glaciological or climatic changes. This strongly suggests that the onset of vast glaciation on the Antarctic Peninsula predates the late Miocene and that polar weathering conditions prevailed on the Antarctic Peninsula throughout the last 9 m.y. Our suggestion is supported by the occurrence of IRD in upper Miocene sediments from the Bellingshausen Sea recovered both at Site 1095 and Site 325, Deep Sea Drilling Project Leg 35 (Hollister, Craddock, et al., 1976) and by mineralogical and sedimentological evidence from continental margin deposits recovered in the Weddell Sea during ODP Leg 113 (Kennett and Barker, 1990). These findings also indicate that major cooling and glaciation had affected the Antarctic Peninsula region at least since the late Miocene. Geochemical investigations carried out on sedimentary rocks point to a prevalence of glacial climatic conditions on the Antarctic Peninsula even since the early Oligocene (Dingle and Lavelle, 1998). The buildup of large ice sheets in West Antarctica, including the Antarctic Peninsula ice cap, may have taken place at ~15 Ma and contributed to a major eustatic sea-level drop (Abreu and Anderson, 1998), which is known to be caused by a significant increase in global ice volume (Lear et al., 2000). The clay mineral records in the drift sediments on the Antarctic continental rise give no evidence for major deglaciation events during the late Neogene (e.g., Webb and Harwood, 1991), supporting the assumption that the Antarctic ice sheet as a whole had been quite stable since that time (Barker et al., 1999).
