Integrated Ocean Drilling Program (IODP) Expedition 302 (Arctic Coring Expedition, ACEX) recovered a unique sediment record from the central Arctic Ocean, revealing that this region underwent major environmental fluctuations since the Late Cretaceous. Major and trace element composition of 1,300 samples were determined using X-ray fluorescence (XRF). The results show significant compositional variability of the sediments with depth that can be attributed to changes in (a) provenance and pathways of detrital material, (b) paleoenvironmental conditions and depositional processes, and (c) diagenetic overprint of the primary record. In addition to existing lithological units, we introduce new geochemical units for a more process-related approach interpreting the ACEX record. In detail, via the geochemical signature of Siberian flood basalts we are able to reconstruct the discontinuous rifting and deepening of the central Lomonosov Ridge during the Paleogene, accompanied by changing current regimes and the onset of sea ice. Eocene biosiliceous sedimentation took place in a relatively shallow setting under predominantly anoxic bottom water conditions, causing a positive anoxia-productivity feedback, although water column stratification was repeatedly interrupted by ventilation events. Anoxic to sulfidic conditions were even more extreme after biosilica production ceased, and significant amounts of pyrite were deposited on the Lomonosov Ridge. Especially in organic matter-rich Paleogene deposits, diagenetic processes obscured the paleoenvironmental signals. Fundamental environmental changes occurred in the Middle Eocene, but geochemical and micropaleontological proxies point not to the identical sediment depth. After approximately 26 Ma of non-deposition or erosion, the Middle Miocene record shows the transition to dominantly oxic bottom water conditions, although suboxic diagenesis seemingly affected these deposits.
 The Arctic is one of the key regions on Earth regarding climate change. Not only is it affected most directly and rapidly by present-day global warming, but it also has a great potential for recording climate fluctuation of the geological past. The climatological and environmental development of the Arctic Ocean (in particular its marginal parts) during the late Quaternary is comparatively well studied. However, much of the Cenozoic and Mesozoic history of this region remains poorly documented because there are relatively few sedimentary archives extending to beyond the Quaternary.
 Our understanding of Arctic long-term history improved considerably since Integrated Ocean Drilling Program (IODP) Expedition 302 (the Arctic Coring Expedition, ACEX) drilled the central Lomonosov Ridge (CLR) (Figure 1) at 88° N latitude [Backman et al., 2006; Moran et al., 2006]. Despite coring gaps and several hiatuses, the ∼428 m sediment record spans long intervals of time from the Late Cretaceous to the Holocene. Many important findings from this unique climate archive have already been published [e.g., Backman and Moran, 2008, and references therein; Stickley et al., 2009]. As part of initial post-cruise research, around 200 bulk sediment samples from the ACEX cores were analyzed by X-ray fluorescence (XRF) at the Geosciences Department, University of Bremen. The ensuing records showed remarkable variations in the elemental composition over depth and time, which may reflect major changes in oceanographic conditions, sediment provenance, and post-depositional processes [Backman et al., 2006]. While some basic interpretations were published, this first data set had certain limitations in terms of sample resolution, selection of analyzed elements, and data quality (as also noted by Martinez et al. ). Consequently, there has been a desire to augment and to refine the sedimentary chemistry records in these cores, and to broaden the interpretational scope.
Spofforth et al.  presented XRF scans of the middle Eocene part of the record. However, comparison of scanning data with XRF data on single samples indicates that the former method does not provide accurate measurements [see also Sluijs et al., 2008]. Martinez et al.  combined parts of the original post-cruise XRF data set with new measurements by Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES) and Mass Spectrometry (ICP-MS) as far as the quality of the analyses allowed. They presented elemental records down most of the ACEX sediment column and applied multivariate statistics, but with a focus on the terrigenous component and with some lithological and chemical key intervals left poorly examined.
 Here we present new high-resolution major and trace elemental data of the ACEX record obtained by single-sample quantitative XRF analysis. This data set covers most of the sedimentary record at higher resolution than previous work [Backman et al., 2006; Martinez et al., 2009], with the exception of the 190-52 mcd interval. Parts of this XRF record combined with ICP-MS data are published by Knies et al. , O'Regan et al. [2008a], Sangiorgi et al. [2008a], Sluijs et al. , and Stickley et al. . Building upon previous ideas and concepts developed in the aforementioned publications, our aim is an even more thorough inorganic-geochemical characterization of the ACEX deposits, to trace sediment provenance, paleoenvironmental changes, and diagenesis during and after sediment deposition. In addition to existing lithology-based units, we introduce a new geochemical subdivision to stimulate a process-related discussion of the ACEX record and Arctic paleoenvironment in general.
 The lithological characteristics of the ACEX record [Backman et al., 2006] provided a basis for the original subdivisions into lithological Units 4 to 1/1, the most recent age model was published by Backman et al. . In a nutshell, the ACEX record consists of an upper, Neogene part (middle Miocene to Pleistocene, 18.2 to 0 Ma, 199 to 0 mcd) poor in total organic carbon (TOC) and dominated by siliciclastic material, while the lower, “Paleogene” part (Campanian to middle Eocene, ∼80 to 44.4 Ma, 428 to 199 mcd) is overall rich in TOC and biogenic opal. Upper and lower part roughly represent the “Icehouse” and “Greenhouse” states of the Arctic Ocean, respectively, and are separated by a hiatus of ∼26 Ma duration. The Campanian-Maastrichtian Unit 4 is mainly siliciclastic (silty clay, silty sand), and unconformably overlain by late Paleocene to early Eocene Unit 3 that documents a very different depositional environment as indicated by finer grain size (mostly clay), pyrite, elevated TOC contents, and (bio-)siliceous material. Middle Eocene Units 2 and 1/6 are rather poor in siliciclastic material (mostly clay), but high in biosilica, pyrite and TOC. The upper boundary of Unit 1/6 is a major hiatus, covering ∼26 Ma and thus the upper Eocene, the Oligocene, and the lowest Miocene. The overlying Unit 1/5 is subdivided into a lower part that shows lithological similarities to the older deposits with high TOC and pyrite contents, and an upper part that marks the termination of biogenic TOC-rich sedimentation on CLR. All of the overlying deposits (Units 1/4 to 1/1) are similar to Late Quaternary Arctic sediments, poor in TOC, dominated by terrigenous clay and silt, with sand lenses and isolated pebbles.
2. Material and Methods
 Sediment samples (1,300 in total) were obtained from cooperating institutes (AWI Bremerhaven, Universities of Utrecht, Bremen, Boston). Samples were frozen, freeze-dried and ground in an agate ball mill for subsequent chemical analysis by XRF (Philips PW 2400). Around 600 mg of each sample were mixed with 3600 mg of a 1 + 1 dilithiumtetraborate + lithiummetaborate (Li2B4O7 + LiBO2) mixture, pre-oxidized at 500°C with NH4NO3 (p.a.), and fused to homogenous glass beads. In case of small sample amounts, sediment was mixed 1+1 with in-house Loess standard and corrected for additional element content. Glass beads were analyzed for major (Al, Ca, Fe, K, Mg, P, Si, Ti) and minor (As, Ba, Co, Cu, Mn, Mo, Ni, V, Zn, Zr) elements (Data Set S1). Analytical precision and accuracy were better than 5%, as checked by in-house and international standard materials. Notably, in the upper half and the lowermost part of the record, Mo was below the XRF quantification limit, thus only qualitative, but still informative results are given (Data Set S2).
 To correct for variable dilution by biogenic (e.g., carbonate, opal) or diagenetic (e.g., pyrite, barite, apatite, siderite) components, elemental records were normalized to Al, as done by Backman et al.  and Martinez et al. . Normalization to Al is displayed as %/% for major elements, and as ppm/% for minor elements. Element/Al ratios were further compared to average shale values (AS) [after Wedepohl, 1971, 1991] (Data Set S3). We recognize that normalization to AS composition bears a certain degree of inaccuracy because the drainage basins surrounding the Arctic Ocean have different lithologies, and because the contribution from these basins to the total sediment budget on CLR may have changed over time [e.g., Krylov et al., 2008; Martinez et al., 2009]. Minor deviations of element/Al ratios from AS composition may reflect this variable lithology in the hinterland. However, when deviations of element/Al ratios from AS values are pronounced, its application as a normalization parameter seems unproblematic. As an overview and to better illustrate geochemical trends with sediment depth, the average element/Al ratios for each lithological unit and for the whole ACEX record are given (Data Set S4). Notably, trace element/Al* ratios are semiquantitative if respective trace element contents are below quantification limit (Data Set S2).
 In this section, we will focus on long-term trends and geochemically peculiar intervals. Details of the elemental depth profiles are depicted in Figures 2, 3, 4, 5, 6, and 7, geochemical data are provided in Data Sets S1–S4. The description of our results and related depth plots (Figures 2, 4, 6, and 7) follow the onboard lithological classification (Units 4 to 1/1), while scatterplots (Figures 3 and 5) are based on geochemical units (F-A) defined below.
3.1. Elements Representing Detrital Input (Al, K, Mg, Ti, Zr)
 The average Al2O3 contents and K/Al ratios for samples from Unit 4 are close to their mean values for the entire ACEX record (Data Set S4). By contrast, the Mg/Al ratio is much lower, while the average Ti/Al and Zr/Al ratios are much higher than the mean values.
 In Unit 3, all detrital element records change significantly upcore, in particular display significant variability. The Al2O3 contents on average increase to values slightly higher than the ACEX mean, but within a range of ∼15–21 wt%. Mean values of K/Al in Unit 3 are similar to the ACEX mean, but range between ∼0.19 and ∼0.30. The Mg/Al ratios increase and reach ACEX average values, but vary between ∼0.13 and 0.31. There is a clear trend toward lower K/Al and higher Mg/Al ratios in the upper part of Unit 3. The mean Ti/Al and Zr/Al ratios are much lower than in Unit 4. They are particularly low in the upper part of Unit 3 and across the PETM and ETM2 intervals, while between these events the Ti/Al and Zr/Al ratios nearly reach 0.12 and 25, respectively.
 Unit 2 has the lowest Al2O3 contents of the whole ACEX records, ranging between ∼3 and ∼9 wt%. The mean K/Al ratio is also lowest in Unit 2, but increases from its minimum (∼0.16) to ∼0.26 at the top of Unit 2, In contrast, the mean Mg/Al in Unit 2 is the highest of the ACEX record, but decreases from its maximum (∼0.38) to 0.20 at the top of Unit 2. Thus, the K/Al and Mg/Al ratios mirror each other in this part of the ACEX record. The mean Ti/Al ratio is slightly lower than the ACEX mean, but gradually increases upcore from ∼0.04 to ∼0.06. The Zr/Al ratio is highly variable in Unit 2, with low values from ∼2 to ∼17 in its lower, and values from ∼15 to ∼30 in its upper part.
 Unit 1 is subdivided into Subunits 1/6 to 1/1 [Backman et al., 2006]. In Subunit 1/6, detrital element records display partly very abrupt changes. Al2O3 contents decrease upcore within a range of ∼16 to ∼4 wt%. The K/Al ratio follows the Al2O3 trend in the lower part of Subunit 1/6, decreasing from ∼0.25 o ∼0.18, but then almost doubles to ∼0.35 within an interval of 30 cm. The Mg/Al ratio shows a rapid increase (from ∼0.18 to ∼0.27) slightly below, and a rapid decrease (from ∼0.27 to ∼0.14) parallel to the marked K/Al increase. The Ti/Al and Zr/Al ratios are close to their mean ACEX values and display little variability in Subunit 1/6.
 Through Subunit 1/5 (including the “Zebra Layer” [Backman et al., 2006]), high variability persists for the Al2O3, Mg/Al and Zr/Al records, while K/Al and Ti/Al are rather stable. The Zr/Al ratio displays the most extreme geochemical pattern, reaching maximum values for the Cenozoic ACEX record (>40 ppm/%). For the overlying Subunits 1/4, 1/3 and 1/2, the Al2O3 content displays a slight upcore decrease, but mean values for these Subunits are still slightly higher than the ACEX mean values. The Zr/Al ratio is rather uniform and close to the ACEX average. The K/Al, Mg/Al and Ti/Al ratios slightly but steadily increase upcore, reaching mean element/Al ratios close to (Mg/Al) or slightly higher than (K/Al, Ti/Al) the respective ACEX mean values. In Subunit 1/1, the general geochemical trends of the underlying Subunits 1/4 to 1/2 continue. Mean values of Al2O3 decrease, while K/Al and Ti/Al increase. However, in Subunit 1/1 all detrital element records are more variable than in the Subunits below, and especially Zr/Al ratios cover a broad range of ∼18 to ∼40.
3.2. Elements Representing Nutrient Availability and Primary Productivity (Si, P, Ca, Ba)
 Interpreting the ACEX Si/Al ratio record (Figure 4a) in terms of paleoproductivity is not simple. Silica can be introduced into Arctic sediment both as detritus (quartz, alumosilicates) and as biogenic opal (e.g., diatoms, chrysophyte cysts, ebridians [Stickley et al., 2008]). Scatterplots of Al2O3 versus SiO2, and Ti/Al versus Si/Al (Figures 5a and 5b) show considerable variability throughout the ACEX record, confirming different pathways of Si input. In Unit 4 and lowermost Unit 3, Si/Al values range between 2 and 5 (Figure 4a), with mean values close to those of the whole ACEX record. Most dramatic changes in upper Unit 3 and Unit 2 are due to the occurrence of biogenic silica [Backman et al., 2006; Stickley et al., 2008]: The Si/Al ratios strongly increase to values > 5 (AS = 3.11), with maxima of 25 and mean Si/Al ratios that double the ACEX average. The Si-rich interval terminates at the lower boundary of Subunit 1/6, and the mean Si/Al values of all overlying Subunits are below the ACEX average.
 The variability of the P/Al record in ACEX sediments (Figure 4b) is highest in upper Unit 3 and Unit 2, with extreme values up to 3,500 ppm/% and mean values of ∼250 ppm/%. In the other parts of the ACEX record, P/Al mean ratios are 150 ppm/% or below (AS = 80), with most samples ranging between 60 and 100 ppm/%.
 Another element often related to marine productivity is Ca (Figure 4c), which is mostly bound to carbonate tests of foraminifera and coccolithophorids. However, Arctic sediments are very poor in biogenic carbonate. Figure 4c shows that compared to AS, most of the ACEX record is strongly Ca-depleted. The upper part of Unit 3 and Unit 2 are the only intervals where Ca/Al enrichments are common, reflected by highest mean Ca/Al ratios of the ACEX record. Comparison of the Ca/Al and P/Al records reveals a close correlation (R2 = 0.89), corroborating earlier geochemical and mineralogical studies that have described apatite in these horizons [Backman et al., 2006; Knies et al., 2008], which in fact could be detrital, authigenic, or biogenic. Thus, the use of Ca and P as paleoproductivity and nutrient proxies is limited in the ACEX record, especially within Unit 2.
 Barium in Quaternary Arctic sediments is mostly incorporated into detrital minerals like feldspar [Nürnberg, 1996; Pirrung et al., 2008], and not applicable as productivity proxy. In all lithological units including the biosilica- and TOC-rich Unit 2, mean Ba/Al ratios are between ∼85 and ∼40 ppm/% (Figure 4d). Only Subunits 1/6 and 1/5 have mean Ba/Al ratios that are 2–3 times higher than the ACEX average Ba/Al ratio, with several peaks reaching 600 to 1,400 ppm/%. However, the use of Ba/Al as productivity proxy is problematic even in these Ba-enriched Subunits due to potential barite diagenesis (see below [e.g., Brumsack, 1986; Bishop, 1988; Brumsack, 1989; Torres et al., 1996; McManus et al., 1998]).
 A number of elements in marine waters and sediments undergo redox changes due to different oxidation states under oxic, suboxic, and anoxic conditions and/or co-precipitate with pyrite, or form sulfide minerals under sulfidic conditions. Most elements related to primary productivity and OM export (Si, P, Ca, Ba) are prone to redox-related diagenetic re-distribution as well. For detailed reviews of redox-sensitive/sulfide-forming elements, we refer to Brumsack  and Tribovillard et al. [2006, and references therein]. In general, As, Co, Cu, Mo, V and Zn (Figures 6 and 7) are enriched in OM-rich sediments and/or under sub- to anoxic conditions, while Mn is depleted. The enrichment patterns of Fe and S in the ACEX record [Backman et al., 2006; Martinez et al., 2009] generally seem to be related to diagenetic pyrite formation, but Fe might also be affected by variable detrital material. Notably, Mo was below the XRF quantification limit in Units 1, 4 and lower Unit 3 of the ACEX record.
 A general observation from redox-sensitive/sulfide-forming element records is the strong enrichment (moderate for Mn and Co) relative to AS in the Si-rich upper Unit 3 and Unit 2, as well as around the 26 Ma-long hiatus between Subunits 1/6 and 1/5 (Figures 6 and 7). This is also reflected in highest average As/Al, Co/Al, Cu/Al, Mo/Al, Ni/Al, U/Al and Zn/Al ratios in the respective lithological units. Notably, the V/Al record has a less pronounced increase in the Si-rich deposits, and the Zn/Al record shows much higher enrichment in the lower than in the upper part of Unit 3. Around the 26 Ma-long hiatus (Subunits 1/6 and 1/5), the redox-sensitive/sulfide-forming elements show distinct variations. While As/Al, Co/Al and Fe/Al display partly extreme enrichments in Subunit 1/6, they abruptly decrease to AS values within a few centimeters at the boundary to Subunit 1/5. Within Subunit 1/5, Fe/Al values stay close to AS, while Co/Al and As/Al still exhibit considerable enrichments. In deposits of lower Unit 3, Unit 4 and Unit 1/4 to 1/1, the elements Co, Cu, Fe, Mo, V and Zn mostly exhibit element/Al ratios close to AS.
 In the upper part of the ACEX record, Mn/Al differs strongly from other trace elements, as evident from mean Mn/Al ratios in Subunits 1/4 to 1/1 that are consistently higher than the ACEX average, and from Mn/Al peaks of 1,000 to 1,800 ppm/% (Figure 6). The Mn/Al record is roughly paralleled by the Co/Al in Subunits 1/2 and 1/1 with numerous peaks up to 20 ppm/%.
 The interpretation of the geochemical data set presented here can be facilitated by a clear differentiation of the whole record into geochemical rather than lithological units. As already noted by Backman et al. , geochemical characteristics of certain intervals are not always paralleled by lithological contrasts, giving rise to differences between lithological units defined by visual inspection of the drilled core material, and geochemical characteristics. Here we introduce geochemical units F to A according to common features regarding sediment provenance, paleoenvironment, and diagenesis.
4.1. Late Cretaceous Geochemical Unit F
 Geochemical unit F is equivalent to lithological Unit 4 of Campanian-Maastrichtian age. Due to drilling disturbance and core loss [Backman et al., 2006], this unit is interpreted with care. Lithologically, it is composed of clayey mud, silty mud and silty sand [Backman et al., 2006]. It is characterized by slightly enriched Si contents, Al2O3 contents lower than AS, and high values of Ti/Al and Zr/Al (Figure 2), indicating enrichments of detrital quartz sand and heavy minerals. These elements are normally enriched under higher energetic conditions (current or wave activity) or at the base of turbidites. Deposition of Cretaceous sediments in a shallow neritic environment [Backman et al., 2006] as well as reworked OM [Stein, 2007] are compatible with the enrichment of Si, Ti and Zr by sediment reworking and winnowing of low density particles. This could also explain P enrichments (Figure 7), potentially related to detrital apatite. Overall, our geochemical data indicate deposition in relatively shallow agitated waters, proximal to the Late Cretaceous paleo-coastline. This is basically consistent with earlier interpretations suggesting that the CLR was attached to the Siberian shelf at this time [e.g., Karasik, 1968; Jokat et al., 1992]. Regarding redox conditions during the Late Cretaceous, despite the finding of pyrite grains in the sediments [Backman et al., 2006], Fe/Al ratios close to AS (Figure 6a) contradict extremely reducing depositional conditions. Also As, Mo, V and Zn records (Figures 6 and 7) suggest slightly reducing conditions at most, supporting a well-mixed shallow coastal environment.
4.2. Upper Paleocene to Lowest Eocene Geochemical Unit E
 The major compositional changes within geochemical unit E - comprising the lower part of lithological Unit 3 - are higher Al contents, but lower Ti/Al, Zr/Al and Si/Al ratios compared to unit F (Figures 2 and 4). This indicates less agitated waters, probably due to deepening of the depositional setting. Some intervals are laminated, and there are higher contents of redox-sensitive metals (Figures 6 and 7) and OM (1–3 wt% [Backman et al., 2006]). The OM is thermally immature and dominantly terrigenous [Stein et al., 2006; Weller and Stein, 2008]. However, during two rather short-lived intervals within geochemical unit E, aquatic OM clearly dominated over the terrigenous fraction: The Paleocene-Eocene Thermal Maximum (PETM) and Eocene Thermal Maximum 2 (ETM2) [Lourens et al., 2005; Stein et al., 2006; Sluijs et al., 2009]. These periods are characterized in the ACEX record by TOC peaks, pronounced negative δ13C excursions [Pagani et al., 2006; Stein et al., 2006], and biomarker evidence for subtropical sea surface temperatures [Sluijs et al., 2006, 2008, 2009].
 The heavy mineral fraction (Ti/Al, Zr/Al) shows significant variability in Paleocene-Eocene deposits (Figure 2) [see also Sluijs et al., 2008]. During the earliest Eocene (between PETM and ETM2 events), shallow water depths associated with higher energetic and well-oxygenated depositional conditions probably prevailed at the drill site, supported by high Ti/Al and Zr/Al ratios, the predominance of terrestrial OM, and low pyrite abundance. In contrast, periods of higher OM burial (PETM, ETM2) not only show low Ti/Al and Zr/Al ratios (Figures 2d and 2e), but also (slight) enrichments of redox-sensitive/sulfide-forming elements (Co, Cu, Fe, Mo, Zn; Figures 6 and 7), suggesting less agitated and oxygen-deficient (bottom) water masses. This is in line with earlier ideas of the persistence of at least periodic “Black Sea-type” conditions in the Paleocene-Early Eocene Arctic Ocean [Sluijs et al., 2006, 2008; Stein et al., 2006; Stein, 2007; Knies et al., 2008]. Elevated primary surface productivity (compared to present-day values) was paralleled by enhanced OM preservation, resulting from stratification of the water column, deep-water oxygen depletion, and even euxinic conditions that sometimes reached the photic zone [Sluijs et al., 2006; Weller and Stein, 2008; Sluijs et al., 2009]. However, paleoenvironmental conditions favorable for high OM burial during the Paleocene-earliest Eocene were repeatedly interrupted by periods of water column mixing [Weller and Stein, 2008]. In terms of sediment provenance, the data reveal a gradual increase in Mg/Al, and variable K/Al values (Figure 2). These changes are most probably related to a new configuration of current systems that delivered detrital sediment from different source areas to the CLR. Such changes in marine currents could have been related to the Paleocene initiation of rifting at the Gakkel Ridge [e.g., Karasik, 1968; Jokat et al., 1992], not only detaching CLR from its Cretaceous detrital source areas, but also opening the Eurasian Basin. The trends in detrital element composition will be explained and interpreted in more detail below.
4.3. Lower to Middle Eocene Geochemical Unit D
 This geochemical unit comprises upper lithological Unit 3 and all of lithological Unit 2. Its presence indicates drastic changes in paleoenvironmental conditions on CLR. In contrast to the older, mainly siliciclastic deposits, this section of the ACEX record is dominated by biogenic silica. In the upper part of unit D (equivalent to lithological Unit 2), the silica comprises mainly marine-brackish diatoms and ebridians as well as freshwater chrysophyte cysts, some silicoflagellates [Stickley et al., 2008; Onodera et al., 2008] and high abundances of sea ice diatoms above ∼260 mcd [Stickley et al., 2009]. However, in the lower part of unit D (equivalent to lithological Unit 3), the biogenic silica has been converted to cristobalite, tridymite and zeolites [Backman et al., 2006; C. Vogt, personal communication, 2009]. The deposits are mostly of dark color [Backman et al., 2006], finely laminated, rich in pyrite and in TOC (1–6 wt% [Stein et al., 2006]). Freshwater chrysophyte cysts [Stickley et al., 2008], large amounts of microspore clusters (massulae) from the free-floating freshwater fern Azolla [Backman et al., 2006; Brinkhuis et al., 2006; Speelman et al., 2009], and isotope data from fish bones [Waddell and Moore, 2008; Gleason et al., 2009] indicate an episodically brackish to fresh surface water layer in the Middle Eocene Arctic Ocean. This is in line with the occurrence of stable seasonal sea ice by approximately 47 Ma (260 mcd), as recently proven by Stickley et al. . The stable water column stratification obviously caused deep-water oxygen depletion, anoxic conditions, and enhanced OM preservation, and thus environmental conditions that might be interpreted as “Black Sea-type” [Stein et al., 2006; Stein, 2007; Weller and Stein, 2008; Boucsein and Stein, 2009]. The most remarkable trace of paleoenvironmental changes in geochemical unit D is a strongly increased Si/Al ratio with values of 5–25, or SiO2 contents of 50–80 wt% (Figures 4a and 5). The geochemistry strongly suggests that the upper part of lithological Unit 3 and all of lithological Unit 2 were deposited under the same general paleoceanographic conditions.
 Despite elevated Si contents related to higher productivity, Ba/Al is not enriched. Following the classic ideas of enhanced biogenic barite deposition in relation to OM export from the photic zone, the Ba/Al record should display an enrichment relative to AS [Church, 1979; Bishop, 1988; Dymond et al., 1992; McManus et al., 1998]. We suggest two possible explanations for this geochemical discrepancy: a water depth and/or a diagenetic effect [Torres et al., 1996; McManus et al., 1998]. Following the generally accepted idea of biogenic barite formation, a minimum water depth of 1000 m is required for the establishment of the sedimentary biogenic barite signal. Probably the Middle Eocene CLR was situated at shallower water depths [Moore and the Expedition 302 Scientists, 2006; Gleason et al., 2009], and biogenic barite simply did not form. Alternatively, biogenic barite was deposited, but diagenetically dissolved later on under sulfate-depleted conditions. Indeed, in the lower section, pore water data indicate active OM degradation via microbial sulfate reduction [Backman et al., 2006]. Long-lasting sulfate-depleted conditions within this geochemical unit could have caused almost complete remobilization of biogenic barite and re-precipitation in diagenetic fronts higher in the sedimentary column. However, in OM-rich, anoxic Cretaceous black shales Ba is still overall enriched because it was diagenetically redistributed within, rather than depleted from the sediments [e.g., Brumsack, 1986; Arndt et al., 2009]. If similar barite redistribution took place in the deposits on CLR, we would still expect some Ba enrichment in geochemical unit D, As this is not the case, we suggest that biogenic barite either did not form in shallow waters, or biogenic barite remobilization started shortly after its deposition. Dissolved pore water Ba concentrations [Backman et al., 2006] close to detection limit throughout the ACEX record indicate that little barite diagenesis is taking place at present, which is consistent with the latter idea.
 The general enrichment of Co, Cu, Fe, Mo, V and Zn (Figures 6 and 7) in unit D points to anoxic to sulfidic bottom waters during sediment deposition of geochemical unit D, as expected from TOC, N, pyrite and biosilica records [Backman et al., 2006; Stein et al., 2006; Stein, 2007; Knies et al., 2008; Stickley et al., 2008]. However, an enrichment of Mn is still observed throughout the biosilica-rich unit (Figures 6c and 6d). It was stated by Stickley et al.  that the Mn pattern in geochemical unit D was caused by precipitation of Mn carbonates in porous biosiliceous sediments. However, the co-enrichment of Mn and several trace metals that are usually enriched under anoxic-sulfidic conditions is contradictory in terms of redox conditions. Usually Mn is depleted from the sediment under sub- to anoxic conditions, especially in non-restricted marine environments like modern upwelling zones off Peru or Namibia, but also in restricted marine basins with carbonate-free sediments like the Black Sea [Froelich et al., 1979; Brumsack, 1986; Calvert and Pedersen, 1996; Burdige, 1993; Brumsack, 2006; Stickley et al., 2008]. Thus, while several geochemical parameters support the “Black Sea-type” interpretation, the Mn/Al enrichment contradicts it. In contrast, early diagenetic formation of Mn carbonate layers is known from the Baltic Sea, where deep subbasins (e.g., Gotland Deep, Landsort Deep) with anoxic-sulfidic and Mn2+-rich bottom waters were repeatedly flushed with well-oxygenated saline waters. This induced Mn oxide formation, followed by transformation to Mn carbonates in otherwise anoxic sediments with significant enrichments of trace metals indicative for anoxic to euxinic bottom waters [e.g., Huckriede and Meischner, 1996; Emeis et al., 1998; Sternbeck et al., 2000; Sohlenius et al., 2001; Neumann et al., 2002]. Episodic, but short-termed flushing events of the deep Arctic Ocean during Eocene times with saline marine waters from the Atlantic are supported by Nd isotope data from fish bones [Waddell and Moore, 2008; Gleason et al., 2009] and by silicofossil preservation [Stickley et al., 2008]. Also a more detailed look at the records of generally enriched trace elements of Co, Cu, Fe, Mo, Ni, V and Zn reveals considerable variability (Figures 6 and 7) that is probably related to rapid changes in bottom water redox conditions. The highly variable P/Al record (Figure 4b) also indicates redox fluctuations at the seafloor, as P is regenerated from sediments under anoxic conditions, while oxic conditions induce its preservation in the sediment as authigenic apatite [e.g., Filippelli, 1997; Delaney, 1998; Slomp and Van Cappellen, 2007]. Microfossil associations in ACEX unit D provide evidence for brackish to freshwater conditions, and the proposed shallow water depth on CLR [Moore and the Expedition 302 Scientists, 2006; Gleason et al., 2009] was probably in the same order of magnitude as the deep basins of the Baltic Sea (maximum of ∼450 m) In summary, geochemical data as well as previous paleoenvironmental considerations suggest a “Baltic Sea-type” rather than a “Black Sea-type” environment on CLR in the lower to middle Eocene.
 In terms of nutrient dynamics under variable redox conditions, ACEX unit D deposits share geochemical similarities with Cretaceous black shales. For these TOC-rich deposits, a positive feedback loop between P regeneration from the sediment, enhanced burial of organic carbon, and anoxic bottom waters was postulated [e.g., Ingall et al., 1993; Mort et al., 2007; März et al., 2008; Meyer and Kump, 2008]: Formation of P-enriched lower water masses and occasional upwelling of nutrient-rich deep waters fuelled primary productivity pulses. More OM was exported to the seafloor, and its degradation both provided “new” organic P and kept the deep water masses anoxic. Extreme sedimentary P enrichments suggest that the P- and probably also Fe-rich anoxic deepwater in the middle Eocene Arctic Ocean was periodically oxidized, leading to massive Fe oxide precipitation, adsorption of the dissolved P, and pulsed P deposition. Upon sediment burial, Fe-bound P was transformed into authigenic apatite in highly porous opal-rich sediments, leading to parallel maxima in Ca/Al and P/Al occurring in this interval (Figures 4b and 4c) [e.g., Jahnke et al., 1983; Filippelli, 1997; Delaney, 1998; Slomp and Van Cappellen, 2007].
 Another striking feature of geochemical unit D is the enrichment of Fe/Al relative to AS (Figure 6a), dominantly bound to pyrite [Backman et al., 2006]. This surplus of Fe could have been causally related to high diatom production and sedimentary Si enrichment, as suggested by Stickley et al. . We suggest that most pyrite-bound excess Fe was provided through a “Fe shuttle” mechanism like it is active in the modern Black Sea [Canfield et al., 1996; Wijsman et al., 2001; Anderson and Raiswell, 2004; Lyons and Severmann, 2006]. This mechanism implies a stratified Arctic Ocean with a suboxic water layer separating oxic surface from euxinic deeper waters. As the suboxic zone impinged on the continental shelf/slope, Fe was leached from the sediments, kept in solution as Fe2+ within the suboxic water layer and shuttled toward the open ocean by diffusion and currents. At the suboxic-sulfidic interface, Fe2+ precipitated as pyrite and was eventually deposited as excess Fe on the seafloor. In addition, the middle Eocene circum-Arctic rivers - comparable to the modern situation - probably drained extensive peat bogs, and were rich in dissolved Fe2+ and Mn2+ [e.g., Telang et al., 1991; Ponter et al., 1992; Hölemann et al., 1999; Gordeev et al., 2004; Hölemann et al., 2005]. Thus, a combination of Fe2+-rich river waters and the suboxic “Fe shuttle” best explains the high Fe/Al values in geochemical unit D.
 Apart from remarkable redox conditions, there were also changes in detrital sediment provenance during geochemical unit D. The trend of increasing Mg/Al values that commenced in geochemical unit 5 proceeds and is mirrored by decreasing K/Al values. A K/Al minimum and Mg/Al maximum is reached at about 300 mcd (Figure 2). Above about 300 mcd (lithological Unit 2), this geochemical trend is reversed, leading to gradually higher K/Al and lower Mg/Al values until 220 mcd. The K and Mg records throughout the lower and middle Eocene sections thus indicate a gradually increasing, then decreasing signature of the (relative to AS) Mg-rich and K-poor Siberian flood basalts forming the Putoran Massif [Rachold, 1999; Schoster et al., 2000; Schoster, 2005; Martinez et al., 2009]. These basalts and associated ejecta are drained by the Khatanga and Yenisei rivers, which deposit their suspended load on the Kara and western Laptev Sea shelf. The geochemical and mineralogical fingerprint of the Putoran Massive, in particular the K and Mg signals, is also translated to the central Arctic Ocean via the Siberian branch of the Transpolar Drift [e.g., Vogt, 1997; Schoster et al., 2000; Schoster, 2005; Darby, 2008; Krylov et al., 2008; Martinez et al., 2009], and in the ACEX record is most prominent around 300 mcd. In contrast, the suspended load of other Siberian rivers (e.g., Olenek, Lena, Yana) is compositionally similar to AS [Rachold, 1999; Schoster et al., 2000], as documented in geochemical unit E, and might rather imply delivery of terrigenous material via the Polar branch of the Transpolar Drift.
4.4. Middle Eocene Geochemical Unit C
 This geochemical unit is equivalent to lithological Unit 1/6, and partly comparable to geochemical unit D due to its high pyrite and TOC contents [Backman et al., 2006; Stein et al., 2006; Krylov et al., 2008]. A major difference is the merely sporadic presence of siliceous fossils in geochemical unit C [Stickley et al., 2008]. Only around 202.5–203.5 mcd (44.6 Ma) is there again a biosilica-rich interval [Stickley et al., 2008] with TOC > 5 wt% [Stein et al., 2006]. As in geochemical unit D, this biosilica-rich interval is a locus of diagenetic processes [Backman et al., 2006; Sangiorgi et al., 2008b; Stickley et al., 2008], where higher porosity supported formation of a diagenetic front [Stickley et al., 2008]. This front is most strongly enriched in Ca, P, Ba (Figure 4) and Mn (Figure 6d), but also in Mg (Figure 2c), Cu, Mo and V (Figure 7), consisting of minerals such as barite, apatite and mixed Ca, Mg and Mn carbonates [Krylov et al., 2008; Vogt, 2009]. Formation of such a distinct diagenetic front most probably resulted from a period of nonsteady state diagenesis and continuous supply of dissolved pore water constituents (e.g., dissolved Ba, Ca, Mn, sulfate and phosphate). One possible precondition for establishing such diagenetic conditions is the fixation of the sediment surface, e.g., during a prolonged period of reduced to non-sedimentation. Otherwise, the sedimentary diagenetic zones would migrate upwards with the sediment surface, leading to rather dispersed precipitation of authigenic minerals. The formation of the diagenetic front might be related to the 26 Ma-long hiatus at the top of Subunit 1/6 [O'Regan et al., 2008a, 2008b; Sangiorgi et al., 2008b], which has been interpreted as period of non-sedimentation or even erosion [Backman et al., 2006; O'Regan et al., 2008b; Sangiorgi et al., 2008b]. This is obvious from the pronounced Ba/Al peaks of up to 1400 ppm/% (Figure 4d), which most probably represent diagenetic Ba fronts precipitated within the sediment after deposition. While the diagenetic Ba enrichments found in geochemical unit C cannot be taken as proof for high paleoproductivity, precursor-barites of biogenic origin must still have existed below the diagenetic front.
 Also outside the silica-rich interval (202.5–203.5 mcd), redox-sensitive/sulfide-forming trace elements are enriched in geochemical unit C (As, Ba, Co, Cu, Mo, V, Zn; Figures 4, 6, and 7). In combination with dark sediment color, high TOC and pyrite contents [Backman et al., 2006; Stein et al., 2006], the geochemical data confirm at least periodically anoxic to euxinic bottom waters in the Middle Eocene Arctic Ocean. Micropaleontological data [Sangiorgi et al., 2008b; Stickley et al., 2008] indicate a still brackish to fresh surface water layer, inducing stratification of the water masses. However, in contrast to geochemical unit D, d13C data of the OM indicate that the source of OM shifted from dominantly marine to terrestrial, with the exception of the biosiliceous horizon (202.5–203.5 mcd [Stein et al., 2006]).
 Interestingly, the hiatus at the upper boundary of geochemical unit C, which displays no change in detrital provenance, shows a drastic decrease in Fe/Al values from around 3 down to 0.5 (Figure 6a). As Fe depletion already starts around 30 cm below the hiatus, it is probably not related to this gap in sedimentation alone. The very high Fe/Al values are bound to large amounts of pyrite [Backman et al., 2006], as are the associated enrichments of As and Co. We assume that the extreme Fe enrichment below the hiatus (Fe/Al > 3.5, Fe2O3 > 25 wt%) results from the same combination of boreal river input and “Fe shuttle” mechanism as described for the underlying deposits [Canfield et al., 1996; Wijsman et al., 2001; Anderson and Raiswell, 2004; Lyons and Severmann, 2006]. However, Fe/Al ratios in geochemical unit C are more than twice as high as those of unit D, and - referring to a modern analog - around three times as high as Fe/Al maxima found in Black Sea sediments (Fe/Al ∼1.2, Fe2O3 ∼5 wt% [Brumsack, 2006; Lyons and Severmann, 2006]). The higher Fe/Al ratio in unit C compared to D could be related to a) by a significantly decreased delivery of detrital Al to CLR, b) a much higher Fe/Al ratio of the delivered detrital material, c) diffusive input of Fe diagenetically mobilized deeper in the sediment column, and/or d) an even higher contribution of Fe delivered by Arctic river waters and suboxic shelf sediments. Regarding the latter consideration, an important factor could be the specific morphology of the Arctic Ocean basin with its extensive shelf areas and a very high potential to release Fe2+ under suboxic conditions. The Fe/Al increase from geochemical unit D to C may have been caused by sea level fluctuation and/or changing redox conditions in the water column. As a result, the suboxic zone impinged the vast Arctic shelves, huge amounts of dissolved Fe were released to the water column, and eventually deposited at the seafloor. However, based on our data alone this mechanism cannot be verified.
 In contrast to Fe, most redox-related trace metals display inconsistent patterns across the hiatus. While decreasing Zn/Al ratio follows the Fe record (Figure 7d), there are no observable trends for As, Cu, Mo, Ni, and V (Figures 6 and 7). Redox-related elements may document a marked change in bottom water, but not sediment redox conditions prior to the 26 Ma-long hiatus.
 Geochemical unit C is distinct from the units discussed so far also in terms of sediment provenance. Another short-lived increase in the K-poor, Mg-rich component related to the Siberian flood basalts is depicted from the records (Figures 2b and 2c). Throughout this unit, the periodic occurrence of IRD [St. John, 2008] and specific diatoms between 202.5 and 203.5 mcd [Stickley et al., 2009] indicate the presence of seasonal sea ice. As main source of the sea ice and the entrained sediments, Krylov et al.  suggested the western Laptev Sea due to heavy mineral associations related to the Siberian flood basalts of the Putoran Massif. Like in lower geochemical unit D, the K/Al and Mg/Al trends (Figures 2b and 2c) confirm that between 45.5 and 44.6 Ma the Putoran fingerprint on CLR was stronger again.
 The most dramatic provenance change of the whole ACEX record is located at 202.15 mcd, where the K/Al ratio nearly doubles within a few centimeters. Parallel to that, the Mg/Al ratio is reduced, resulting in a K/Mg ratio that increases from 0.15 to 0.34 (Figures 3b and 3c). It is noteworthy that this dramatic change does not coincide with the boundary between lithological Subunits 1/6 and 1/5, but predates the palynologically determined hiatus. The change in provenance must have occurred on a relatively short time scale, shifting from the western Laptev and Kara Seas to a region with K-rich and Mg-poor lithologies. The palynologically defined hiatus itself is characterized by little provenance variation, as K/Al and Mg/Al do not show significant shifts (Figure 2). A possible explanation could be the existence of more than one hiatus. This hypothesis requires further investigation, and what appears as a “multistep” hiatus might be related to problems with the age model in this part of the core and/or significant sediment reworking (C. E. Stickley, personal communication, 2010).
4.5. Middle Miocene Geochemical Unit B
 The interval directly above the hiatus, also termed “Zebra Layer,” is defined as geochemical unit B and equivalent to the lower lithological Unit 1/5. A detailed study of the sediment geochemistry around the 26 Ma-long hiatus will be presented in a forthcoming publication. Here we will mostly describe general compositional trends that characterize the interval above the hiatus as geochemical unit B. The sediments consist of silty clay and exhibit very characteristic, dark gray to black, slightly tilted bands of 0.5–3 cm thickness [Backman et al., 2006]. Within this light-dark banded interval, highest TOC values of the ACEX record are found (14.5 wt% [Backman et al., 2006; Stein et al., 2006]). Siliceous microfossils are rare, the lithology is mainly siliciclastic [Backman et al., 2006; Sangiorgi et al., 2008a; Stickley et al., 2008]. Mostly based on these previous findings, Jakobsson et al.  subdivided the sediment interval around the 26 Ma-long hiatus into an early “lake,” a transitional “estuarine,” and finally a fully ventilated “ocean” phase in the Arctic Ocean. This paleoenvironmental change was caused due to the establishment of the Fram Strait deep-water connection around 17.5 Ma ago. Geochemical results from the “Zebra Layer” are partly published by Sangiorgi et al. [2008b] and indicate that it marks the last period of anoxia/euxinia in the central Arctic Ocean. Thus, the interpretation by Jakobsson et al.  as a transitional interval is supported by geochemical data. Specifically, redox-related element records (As, Cu, Mo, Ni, U, Zn) are highly variable between values found in the underlying TOC-rich and the overlying TOC-lean deposits, documenting significant changes in bottom water oxygenation during deposition of unit B. Detailed examinations published by Sangiorgi et al. [2008a] reveal compositional differences of the dark and light layers, with more reducing conditions related to dark, OM-rich relative to light, OM-poor intervals. Interestingly, these fluctuations are not seen in the Fe/Al record, which is close to AS values (Figure 6a). Indeed, among the redox-related elements the Fe record shows the most drastic geochemical change related to the hiatus. Thus, the mechanism that led to strong Fe enrichment on CLR before the hiatus was not active anymore.
 The detrital element record allows for two main interpretations. First, the peaks in the Zr/Al record and also reworked sand grains [St. John, 2008] imply sediment sorting by high current or wave activity on CLR, hinting to very shallow water depths [Sangiorgi et al., 2008a]. Second, the Al2O3, K/Al, Mg/Al and Ti/Al records (Figures 2a–2d) show a gradual trend across the 26 Ma-long hiatus, which is unexpected given the long time on non-deposition or erosion. Our geochemical data reveal no major change in the detrital source area to CLR, but rather a continuous development. This indicates that either the transport mechanisms and Arctic Ocean current systems did not significantly change, or that both before and after the hiatus the detrital material was derived from a local source area (e.g., a subaerial part of the CLR). The aspect of sediment reworking is potentially important not only for the inorganic geochemical composition of the sediment in unit B, but might also have implications for the interpretation of other paleoceanographic proxies from the “Zebra layer.”
4.6. Middle Miocene to Pleistocene Geochemical Unit A
 Above the “Zebra Layer,” these sediments mark the onset of paleoenvironmental conditions in the central Arctic Ocean that are probably comparable to the modern ones. These deposits are largely barren of palynomorphs and biosiliceous fossils [Sangiorgi et al., 2008a; Stickley et al., 2008], which is also reflected in the Si/Al ratio. In geochemical unit B, the Si/Al ratio fluctuated between 3 and 4, while it is constantly around 3 in unit A (Figure 4a). The TOC content drops to < 0.5 wt% [Stein et al., 2006; Stein, 2007], marking the most pronounced shift in TOC of the ACEX record. Geochemical unit A includes the upper lithological Unit 1/5 and reaches up to the top of the ACEX record, thus is of Middle Miocene to Pleistocene age [Backman et al., 2008]. After water column anoxia defined the Cretaceous-Paleogene part of the ACEX record, this geochemical unit for the first time represents oxygenated bottom water conditions in the central Arctic Ocean. This is demonstrated by repetitive extreme Mn/Al enrichments starting above the “Zebra layer” (up to 1,800 ppm/%; Figure 6d). The dominant sources for these high amounts of Mn most probably were circum-Arctic rivers. As they drain extensive acidic peat bogs, they are particularly rich in dissolved Mn2+ [e.g., Telang et al., 1991; Ponter et al., 1992; Hölemann et al., 1999; Gordeev et al., 2004; Hölemann et al., 2005]. In addition, some redox-related elements like Cu, Mo, Ni and Zn (Figure 7) show slight enrichments relative to AS. However, these are not indicative for anoxic conditions as they are orders of magnitude lower than in the TOC-rich units and most probably bound to Mn oxides by adsorption.
 Abrupt and high Mn peaks frequently found in Quaternary Arctic sediments are sometimes interpreted as changes in redox conditions of the Arctic Ocean bottom waters and/or variable Mn input through Siberian rivers, and applied as chemostratigraphic marker horizons on glacial/interglacial time scales [Jakobsson et al., 2000, 2001; Polyak et al., 2004; Löwemark et al., 2008]. However, early diagenesis due to fluctuating redox conditions at the seafloor, variable input of reactive OM and/or changes in sedimentation rate might have overprinted the primary cyclicity [Li et al., 1969; Gobeil et al., 1999, 2001; Katsev et al., 2006]. Pore water data [Backman et al., 2006; Dickens et al., 2007] show that diagenetic redistribution of Mn is occurring within Miocene to Pleistocene ACEX deposits. Hence, the Mn peaks might partly represent authigenic Mn (oxyhyr)oxides, but also rhodochrosite has clearly been proven in the middle part of this unit [Backman et al., 2006].
 In terms of detrital provenance, the base of geochemical unit A coincides with the hiatus separating lithological Units 1/6 and 1/5. It documents intermediate K/Al and low Mg/Al ratios relative to the older ACEX deposits, which consistently increase from lithological Unit 1/5. The detrital record indicates that after the sedimentological hiatus of around 26 Ma duration, the provenance of Middle Miocene to Pleistocene sediments was not dominated by the Siberian flood basalts any more, but by a rather shale-like composition [Martinez et al., 2009]. The gradual compositional trend to both higher K and Mg, and lower Al2O3 values further upcore (Figure 2) cannot be interpreted in terms of this two-component system (shale-like versus flood basalt signature), as K and Mg trends parallel each other. We suppose the long-term trends reflect a changing provenance responding to the onset of perennial sea ice formation [Krylov et al., 2008; Martinez et al., 2009] or to tectonic processes related to uplift/subsidence of the CLR [O'Regan et al., 2008b]. Small-scale fluctuations during geochemical unit A rather document cyclic and short-termed events that may be related to climatic changes on glacial-interglacial timescales [e.g., Backman et al., 2008; O'Regan et al., 2008a; Pälike et al., 2008]. In this context, during geochemical unit A the CLR seemed to be periodically influenced by the Beauford Gyre as the dominant current system of the Canadian Basin. In this way, new detrital source areas as the Canadian Archipelago and the East Siberian shelf may have left their geochemical traces at the ACEX site, providing e.g., more K-feldspar and dolomite. In the uppermost part of geochemical unit A, an increase in the Mg/Al ratio (while K/Al is rather stable) clearly indicates that the detrital material arriving at CLR was to a large part derived from dolomite rocks of the Canadian Archipelago, transported by icebergs or sea ice via the Beauford Gyre [Vogt, 1997].
 Quantitative XRF analyses of discrete bulk sediment samples taken from the length of the entire ACEX record indicate depth intervals with significantly different elemental compositions. This allows for a subdivision of the sedimentary record into six geochemical units. These units likely represent times of common provenance, paleoenvironmental and depositional conditions, and diagenetic overprint (Figure 8). Moreover, these geochemical units only partly correspond to the lithological units defined onboard the ACEX expedition, as anticipated by Backman et al. . The geochemical units augment the lithological subdivision and offer a more process-oriented description of the sedimentary succession.
 Details and characteristics of the geochemical units include:
 Unit F is mainly siliciclastic and was deposited under oxic and high-energy conditions.
 Unit E is dominated by siliciclastic material as well, but with a slightly different provenance. Redox conditions were mostly oxic, and depositional energies were generally lower than in unit F, but displayed some variable. During the warm intervals of the PETM and ETM2, however, deposition took place in a little agitated environment with probably anoxic bottom water conditions.
 Unit D documents a period of strongly enhanced biosilica deposition and overall reducing bottom water conditions. However, the variable records of redox-related trace metals and the enrichment of Mn point to repeated bottom water ventilation events, and thus a “Baltic Sea-type” depositional environment. Strong diagenesis overprinted some paleoceanographic proxies (e.g., Ba, P), complicating paleoproductivity estimates in this strongly biogenic unit. Detrital element records reveal a first increasing, then decreasing influence of the Siberian flood basalts during unit D.
 Unit C marks the return from biosiliceous to siliciclastic sedimentation. However, bottom water redox conditions remained strongly reducing without evidence for ventilation events. Intense diagenesis caused the precipitation of authigenic minerals, especially barite. Extreme Fe enrichments were probably related to dissolved riverine input and/or Fe recycling from reducing shelf sediments. In the upper part of unit C, an abrupt increase in K/Al implies a significant provenance change, and raises the question of a hiatus prior to the 26 Ma-long one that defines the top of unit C.
 Unit B is equivalent to the “Zebra layer” and was deposited after the 26 Ma-long hiatus, but most detrital elements do not indicate a significant provenance change over this period. This unit is transitional between the older, mainly anoxic, and the younger, mainly oxic deposits, and thus characterized by strong variations in both redox conditions and depositional energy.
 Unit A is characterized by a steady and gradual trend in detrital sediment provenance. Redox conditions were dominantly oxic, but suboxic diagenesis caused precipitation of various authigenic minerals in distinct horizons. Strong and cyclic enrichments in Mn throughout this unit are either related to variable river discharge or variable bottom water redox conditions, but are clearly overprinted by diagenesis in the deeper part of unit A.
 Samples were provided by the Integrated Ocean Drilling Program (IODP). We are highly indebted to the masters, crews and shipboard scientific party of IODP Expedition 302. We are hugely indebted to R. Stein (AWI Bremerhaven), H. Brinkhuis and F. Sangiorgi (University of Utrecht), C. Vogt and M. Kölling (University of Bremen), and N. Martinez and R. Murray (Boston University) for sharing their sample sets. We thank E. Gründken, C. Lehners and M. Wagner for sample preparation and chemical analyses. We are grateful to C. Vogt and J. Matthiessen for most constructive criticism, and to R. Stein, H. Brinkhuis, F. Sangiorgi and A. Krylov for inspiring discussions. Insightful comments by an anonymous reviewer and C. Stickley improved the manuscript significantly. We especially thank the Editor G. Dickens for his very helpful and far more than editorial comments and suggestions. Funding by the Deutsche Forschungsgemeinschaft through grant BR 774/20 within the SPP 527 “IODP/ODP” is gratefully acknowledged.