Neodymium Evidence for Increased Circumpolar Deep Water Flow to the North Pacific During the Middle Miocene Climate Transition

Low salinity surface water inhibits local deepwater formation in the modern North Pacific. Instead, southern‐sourced Circumpolar Deep Water (CDW) fills the basin, which is the product of water masses formed from cold sinking centers in the Southern Ocean and North Atlantic. This CDW is responsible for transporting a significant amount of global heat and dissolved carbon in the deep Pacific Ocean. The history of its flow and the broader overturning circulation are widely assumed to be sensitive to climate perturbations. However, insufficient records exist of CDW presence in the deep North Pacific with which to evaluate its evolution and role in major climate transitions of the past 23 Ma. Here we report sedimentary coatings and fish teeth neodymium isotope values—tracers for water‐mass mixing—from deepwater International Ocean Discovery Program Site U1438 (4.7 km water depth) in the Philippine Sea, northwest Pacific Ocean. Our results indicate the water mass shifted from a North Pacific source in the early Miocene to a southern source by ~14 Ma. Within the age model and temporal constraints, this major reorganization of North Pacific water mass structure may have coincided with ice sheet build up on Antarctica and is most consistent with an increased northward flux of CDW due to enhanced sinking of cold water forced by Antarctic cooling. The northward extent of this flux may have remained relatively constant during much of the past 14 Ma.


Introduction
Overturning circulation in the modern Pacific Ocean is driven by Circumpolar Deep Water (CDW), which is sourced from cold sinking centers around Antarctica and the North Atlantic (Ferrari et al., 2014;Talley, 2013;Talley et al., 2011). Low surface water density in the North Pacific precludes significant local convection, and Pacific meridional overturning circulation (PMOC) is sluggish compared with the Atlantic (Ferrari et al., 2014;Kawabe & Fujio, 2010). As CDW enters the Pacific at depth, it travels northward (Figure 1), gradually exchanging with intermediate water before flowing back to the Southern Ocean as North Pacific Deep Water (NPDW, sometimes referred to as Pacific Central Water) between 1 and 3 km depth (Kawabe & Fujio, 2010). Thus, CDW transports heat and dissolved carbon into the deep North Pacific, forming a major carbon reservoir separated from the atmosphere (Ferrari et al., 2014). Changes in PMOC have the potential to impact global temperature, moisture supply, and the carbon cycle, and there have been several attempts to reconstruct PMOC over the Cenozoic to understand its past roles in climate change and tectonic evolution (e.g., Butzin et al., 2011). In particular, it is not yet known how PMOC responded to major climate transitions such as the middle Miocene climate transition (MMCT, ~14 Ma).
Several neodymium (Nd) isotope and modeling studies have shown that the Pacific was characterized by local overturning in the north as well as the south during much of the warmest intervals in the Paleogene, with a major shift toward southern-sourced deep water in the north after~40 Ma (e.g., Thomas, 2004; exhibit an approximately unchanging isotopic offset from each other with depth throughout the Neogene, attesting that there were distinct water mass signatures somewhat similar to today (van de Flierdt et al., 2004). They show a long-term trend of gradually more radiogenic values at all depths from~20-3 Ma, suggested to be the result of the Isthmus of Panama closure (Bartoli et al., 2005;Duque-Caro, 1990) gradually cutting off the supply of nonradiogenic Atlantic water into the Pacific (Duque- Caro, 1990;Martin & Haley, 2000). Alternatively, gradual Pacific Plate movement transporting all sites toward more proximal volcanic centers in the West Pacific has been suggested by Le Houedec et al. (2016), but we note that western sites do not show more radiogenic values than eastern sites of a similar water depth (e.g., Sites D11-1 and CD29-2 have similar values; Ling et al., 1997).
To address this data set gap and gain insights into deep ocean circulation dynamics of the Neogene, we measured the Nd isotopic values of fossil fish teeth and ferromanganese oxyhydroxide coatings as a water-mass tracer from IODP Site U1438 (Figure 1) in the Philippine Sea, NW Pacific (Arculus, Ishizuka, Bogus, & Expedition 351 Scientists, 2015;Arculus, Ishizuka, Bogus, Gurnis, et al., 2015), over the past 23 Ma. The site is located at 4.7 km depth and is currently bathed in CDW.

Site U1438
In 2014, IODP Expedition 351 cored Site U1438 (Arculus, Ishizuka, Bogus, & Expedition 351 Scientists, 2015). The lithology of Site U1438 consists of pelagic noncalcareous abyssal clays with discrete tephra layers and dust in the top~160 m (Unit I, 26 Ma to Holocene), transitioning below into coarser grained tuffaceous muds and sands that continue down to~305 m (Unit II, 29 to 26 Ma). The age model is robust, with nearly all paleomagnetic reversals present, and tied to radiolarian biostratigraphy in the middle Miocene (~15 Ma) and nannofossil biostratigraphy in the late Oligocene (at 27 Ma; Arculus, Ishizuka, Bogus, & Expedition 351 Scientists, 2015; Figure 2). Site U1438, currently at 4.7 km water depth, has been situated below the carbonate compensation depth (CCD) for the past 26 Ma (modern CCD depth~4.5 km) based on the lack of carbonate throughout the  Kaneko et al., 2001), here referred to as Circumpolar Deep Water (CDW). Numbers refer to Sv (volume water flow per second). The dashed lines and question marks denote uncertainty. (b) Bathymetric map of the Pacific Ocean, with locations of sites discussed in the text, and the regional path of lower CDW influence centered around 4,500-m water depth (Kawabe & Fujio, 2010). whole of Unit I (<0.5 wt%; Arculus, Ishizuka, Bogus, & Expedition 351 Scientists, 2015). The CCD has been constrained to deeper than 4.5 km throughout the Oligocene and Miocene in the equatorial Pacific (Pälike et al., 2012), indicating that Site U1438 was monitoring >4.5 km water depth for the duration of our 23 Ma record. The Philippine Sea Plate, upon which Site U1438 is located, was in a position about 6°southward of its current location at~20 Ma (Richter & Ali, 2015;Wu et al., 2016) and has rotated steadily northward since then (Figure 3).

Use of Nd Isotopes as a Water Mass Tracer
Dissolved Nd is a robust tracer of water-mass composition owing in part to its short residence time (~0.3-1 ka, Tachikawa et al., 2003) relative to oceanic mixing (~1.5 ka). Seawater Nd isotope ratios ( 143 Nd/ 144 Nd normalized to bulk Earth, expressed as ε Nd , De Paolo & Wasserburg, 1976) are different in each of the modern ocean basins, as ε Nd varies depending on the source region of fluvial discharge, boundary exchange, and dust sources (Arsouze et al., 2007;Jones et al., 1994;Wilson et al., 2013). The North Pacific is surrounded by geologically radiogenic sources, and thus, shallow waters have highly radiogenic values (À2), while underlying water is less radiogenic because of the influence of deeper waters from the Southern Ocean (À8). Therefore, North Pacific Deep Water has a somewhat elevated signature (À5) due to gradual exchange with intermediate water, resulting in a reduced but robust ε Nd gradient that is clearly shown in water column measurements (Amakawa et al., 2009) but has proved challenging to model (Arsouze et al., 2007). One complication with reconstructing water masses is the radiogenic pore water flux detected in marginal settings, such as the Oregon Margin (Abbott et al., 2015) and Gulf of Alaska (Haley et al., 2014) up to~100 km from the shelf, where seawater ε Nd at least in the top 3.5 km water depth is elevated above more typical open ocean Pacific values.

Analytical Procedures
We reconstructed past water mass signatures at Site U1438 by measuring ε Nd of fossil fish teeth and ferromanganese oxyhydroxide coatings,  which record the isotopic value of bottom water at the sediment surface (Reynard et al., 1999). The fossil fish teeth and teeth debris collected from Site U1438 sediment samples were prepared using the general methods of Basak et al. (2011) and Xie et al. (2012). To isolate the teeth/debris, bulk sediment was disaggregated and washed over a 63-μm sieve. The retained material was dried overnight in an oven (50°C). For analysis, approximately 15-30 specimens per sample were handpicked using a binocular microscope and fine brush. The specimens were then washed 3 times with ultrapure water (Milli-Q).
For Fe-Mn oxyhydroxide coating analysis, bulk samples were dried overnight and were subsequently pulverized and homogenized with an agate mortar and pestle. These homogenized samples were decarbonated for 2 hr using sodium acetate buffered acetic acid solution that was precleaned in cation exchange resin. The remaining material was washed 3 times with Milli-Q. To reduce the Fe-Mn phases, the oxide fraction was  Figure 5 after repeating. It is assumed reworked or possibly affected by dispersed ash.

Paleoceanography and Paleoclimatology
leached in 14 ml of 0.02 M hydroxylamine hydrochloride (HH) in 20% acetic acid buffered to a pH of 4. The samples in HH solution were placed on a rotary shaker and left for 2 hr. Once the samples were leached, they were centrifuged, the supernatant decanted into a separate clean tube and centrifuged for an additional 1 hr, then decanted and dried. The samples were digested in concentrated HNO 3 overnight and then placed in 2 N HNO 3 for column chemistry.
Given the presence of disseminated ash throughout the sediment sequence, we also analyzed a few, targeted samples to characterize the detrital sediment (alumino-silicate fraction) Nd isotopic composition. The detrital silicate fraction was taken from the remaining material once the supernatant was decanted. These were placed back in HH for 1.5 hr and rinsed 3 times with Milli-Q before being left for 4 hr in HH. The samples were then rinsed 3 times with Milli-Q and dried overnight. The dried samples were homogenized and placed in 23 M HF for~5 days until completely digested. They were dried down and placed in concentrated HNO 3 -HCl-HNO 3 before finally being placed in 2 N HNO 3 in preparation for column chemistry.
All samples were then dissolved in 500 μL of 2 N HNO 3 . The rare Earth elements (REEs) were isolated using Tru Spec column chemistry (isolating the REE suite from the bulk sample) and the samples collected with 3 ml of 0.05 N HNO 3 in Teflon beakers and dried down. The samples were then dissolved in 200 μL of 0.18 N HCl and placed on a 100°C hotplate overnight. The Nd fraction was isolated from the bulk REE via Ln Spec column chemistry by sequentially separating out the bulk REE. Once dried down, the remaining Nd portion was loaded onto a degassed rhenium filament (0.76 mm width, 25 μm thickness) using 1 μL of 2 N HCl and analyzed on a Thermo Scientific Triton thermal ionization mass spectrometer as Nd+. The ion beams analyzed ranged from 1 to 10 × 10 À11 A, depending on the amount of Nd loaded and geometry of the double filament assembly, which was not tightly controlled (Pin et al., 2014). A run consisted of blocks of 16, each block cycled 8 times (collecting individual measurements), and a background measurement every 2 blocks; the gain was recalibrated after every 5 blocks. External precision was 15 ppm (2σ) with a value of 0.512104 based upon analysis of JNdi-1 standard (n = 34). Samples were only used if the absolute error was <10 À6 . Values higher than this were discarded, with the exception of 7HCC; the ε Nd errors of these samples were all within acceptable range (Tables 1-3). Several duplicates (from different samples) were run throughout to ensure reproducibility.

Calculation of ε Nd (t)
The ε Nd (t) (Taylor & McLennan, 1985). The fish teeth/debris and Fe-Mn coating values were plotted together when they were analyzed from the same sample and show similar values (R 2 = 0.92; Figure 4).

Isotope Values and Potential Ash Influence
The ε Nd values (Tables 1 and 2) at Site U1438 fluctuated from À3.4 to À4.2 between 15 and 23 Ma, before dropping by 1.2 ε units to À4.7 by 14 Ma (Figure 5). Over the next 9 Ma, ε Nd values exhibit modest  The active eruptive phase of the proximal and extinct proto-Izu-Bonin volcanic arc-the Kyushu-Palau Ridge (Figure 1a)-had ended by the late Oligocene (Arculus, Ishizuka, Bogus, Gurnis, et al., 2015;Ishizuka et al., 2011), as evidenced at Site U1438 by the lithological change from Unit I to II (Figure 2) and the drastic change in sedimentation rate. We regard it as unlikely that upward flow of pore waters from the underlying >1-km-thick volcaniclastic sediment package (Units II to IV; see Arculus, Ishizuka, Bogus, & Expedition 351 Scientists, 2015), deposited by the active proto-Izu-Bonin arc, could have been extensive. Pore water chemistry results show the boundary between Units I and II (at 160 mbsf; Figure 2) is a very low-permeability clay-rich layer (Arculus, Ishizuka, Bogus, & Expedition 351 Scientists, 2015). However, background Neogene volcanism within the Philippine Sea region had the potential to contribute sedimentary radiogenic Nd to Site U1438 through ash. Although we avoided sampling the discrete ash layers at our site (37 ash beds over 23 Ma; Figure 6), studies have shown that dispersed ash occurs in deepwater sites east of the Izu-Bonin Arc (Scudder et al., 2014; Figure 1a). Dispersed ash in Unit I could hypothetically have impacted the Nd signal of the fish debris and Fe-Mn coatings by either postdepositional authigenic growth in ash-rich pore waters (Martin & Haley, 2000), ash-rich pore water migration into oceanic bottom waters (Abbott et al., 2015;Martin & Haley, 2000), or through oceanic exchange with dissolving dust or volcanic sediments on the seafloor (Wilson et al., 2013). The water column ε Nd profile in the Oregon Basin, NE Pacific (black symbols, Figure 7a), has been argued as partially controlled by upward flux of pore fluids (Abbott et al., 2015), a process that could account for the positive offset of these (and samples from the Gulf of Alaska; white symbols, Figure 7a) from values more typical of the North Pacific (pink shading, Figure 7a). Although Site U1438 in the Philippine Sea is more distal from the continental shelf (~400 km) than these marginal sites, there is the possibility that pore water ash could impact bottom water ε Nd values. However, ε Nd values at Site U1438 over the last 24 million years are mostly lower than either the Gulf of Alaska or Oregon Basin (red symbols, Figure 7a) and more in line with typical water values at 4.7 km depth from a range of stations in the North Pacific. There is, however, a need for future pore water studies on NW Pacific margin sediments and water and at depths greater than 3 km.
To consider the Nd isotopic composition of the alumino-silicate fraction (and hence any potential ash that might contribute dissolved Nd to the pore waters or overlying bottom waters), we measured the aluminosilicate component of several targeted samples in the middle Miocene section of most significant change ( Table 3). The extracted alumino-silicate fractions from samples at 11.63, 14.85, and 20.4 Ma have ε Nd values of À11.1, À3.5, and À10.7, respectively (triangles in Figure 7b). These values generally reflect continental weathering inputs (possibly dust from East Asia) with the exception of À3.5, which indicates a proportion of andesitic-type sediment. While the alumino-silicate value of À3.5 is similar to the fish tooth analysis recorded at this interval, it is much lower than the sediment (and pore water) values at the Oregon margin (ranging from À2 to 0; Abbott et al., 2015) considered necessary to positively offset bottom waters by~1 ε unit from typical Pacific values at those depths (À3.5 to À3). Therefore, if the fish tooth sample at 14.85 Ma was controlled by radiogenic sediment and associated pore water flux, we would expect to see an alumino-silicate fraction value higher than the fish teeth value.
Overall, the ε Nd of sediment from ash-rich deposits proximal to volcanic island arc centers of the North Pacific is much higher (most commonly ranging from À2 to +8 ε units; Figure 7b) than our alumino-silicate samples. The range of ε Nd values at U1438 is more typical of Pacific deepwater compositions (box-and-whisker plot, Figure 7b), indicating an overriding water mass control on the fish teeth samples at U1438. This lack of evidence for a large ash influence at U1438 is also indicated by the broadly similar values and trends recorded at nearby Site 786 (Martin & Haley, 2000; Figures 1 and 6), which is also far lower than for typical Pacific volcanic margin sediments (Figure 7b) and Oregon Margin sediments and pore waters. The most radiogenic values occur in the Pliocene interval, which does coincide with the greatest number of discrete ash layers (Figure 6), and we cannot with our data discount a volcanic ash influence here. However, these elevated Pliocene values are also seen at distal Site 807 (Le Houedec et al., 2016;Figures 1 and 5) and so we suggest they may be reflecting changing seawater values in the West Pacific region. Although hot spot activity in the Philippine Sea (e.g., Benham Rise; Figure 3) had ended by 35 Ma (Ishizuka et al., 2013), we cannot rule out the influence of volcanic activity in that region.

Water Mass Changes During the Neogene
When compared with other available data from Pacific Ocean sites ( Figure 5), with no clear ash influence (from comparison with ash-rich zones, Figure 7), our results from U1438 signify likely changing deepwater sources to the Philippine Sea as the predominant control on ε Nd values. North Pacific sites from~10 to Figure 6. The ε Nd (t) of fish teeth and Fe-Mn oxyhydroxide coatings from Philippine Sea sites U1438 (this study) and nearby 786A (Martin & Haley, 2000), compared with ash layers from U1438. Note that none of the samples from U1438 (this study) was taken directly at ash layers.

10.1029/2017PA003309
Paleoceanography and Paleoclimatology 30°N (Ling et al., 1997;Martin & Haley, 2000) at intermediate NPDW depths (ranging from 1.8 to 3.1 km) exhibit generally higher radiogenic values over the past 25 Ma, with averages from À4 to À3 (Figure 5c). Equatorial Pacific Site 807 is monitoring a more southerly extent of NPDW today (2.8 km), and its greater variability (averages from À4.5 to À2.5) has been interpreted as due to alternating contributions of Upper CDW and NPIW (Le Houedec et al., 2016). We observe the majority of the Site 807 record from 25 Ma to Recent overlaps with NPDW end member values (Figure 5), although some of the less radiogenic values may not represent pure NPDW. Equatorial Pacific Site VA13-2 (4.8 km water depth; Ling et al., 1997), which does monitor CDW today (Figure 8a), exhibits relatively less radiogenic values with ε Nd ranging from À5.5 to À4 ( Figure 5). Finally, Equatorial Pacific Site DC137-01 (van de Flierdt et al., 2004), at an abyssal depth of 7.2 km and most impacted by CDW today (Figure 8a), shows the least radiogenic ε Nd values ranging from À6.5 to À5.5 since 25 Ma ( Figure 5).
With the Neogene evolution of Pacific intermediate and deepwater mass values thus constrained ( Figure 5), the ε Nd results from U1438 can be seen as indicating changes in the relative contribution of CDW to the deep Philippine Sea. Before~14.8 Ma (Figure 8c), Site U1438 was bathed in local intermediate NPDW (ε Nd values closer to intermediate water), and between~13.7 Ma and present (Figure 8b), CDW penetrated further northward into the deep North Pacific, lowering ε Nd values. Our results therefore indicate a major deep ocean reorganization occurred in the Pacific Ocean between 14.8 and 13.7 Ma. This ε Nd shift of >1 is comparable to the 1.5 ɛ unit shift of NPDW in the North Pacific during the Paleogene, interpreted as caused by local deepwater formation (Thomas, 2004). The Philippine Sea Plate modest 6°of latitude gradual movement northward away from the CDW source over the last 20 Ma (Figure 3) could not explain the negative ε Nd changes at Site U1438, as this movement would have distanced Site U1438 from the less radiogenic southern-sourced water mass.

Tectonic Influences
Between~15 and 23 Ma, our results show that, remarkably, the deep NW Pacific (>4.7 km) was bathed in NPDW (Figure 8c), not CDW as it was after 14 Ma ( Figure 8b) and as it is today (Figure 8a). Ocean circulation modeling by Butzin et al. (2011) andvon der Heydt andDijkstra (2006) shows that a continental configuration change from the Oligocene to Miocene, which included a more restricted Isthmus of Panama ( Figure 5a) and less redistribution through the Tethys circumequatorial flow, caused North Pacific overturning to become more limited and southern component deepwater to penetrate further northward into the Pacific. Thus, if this modeling is correct, it is possible that the enhanced CDW flux since 14 Ma was the result of restricted Atlantic high salinity inflow, and lower North Pacific salinity causing stifled North Pacific overturning. However, if this process does explain the gradual increase in both deep and intermediate Pacific Ocean water ε Nd (e.g., Martin & Haley, 2000), it is then less able to explain the abrupt evolution of Site U1438 ε Nd values as they show a different trend to other sites ( Figure 5). This leads us to favor nontectonic processes for the increased CDW flux to the Philippine Sea in the middle Miocene.
Another significant tectonic change impacting the Pacific, the closure of the Indonesian Gateway by 14 Ma (Gourlan et al., 2008), prevented the flow of less radiogenic water (ε Nd below À7) into the Pacific and could thus have resulted in a trend toward more radiogenic values in the Pacific. Therefore, this gateway closure could not have directly caused the negative ε Nd shift at U1438. We note, however, that it is possible the Figure 7. The ε Nd (t) of fish teeth and Fe-Mn oxyhydroxide coatings from Site U1438 (this study) compared with various modern North Pacific Ocean water and sediment samples. (a) Modern water sample ε Nd data (black symbols, with 7-point black smoothing spline and ±1σ gray area) from the marginal Gulf of Alaska (Haley et al., 2014) and Oregon margin (Abbott et al., 2015). Data show a significant positive offset from NE Pacific modeled seawater (dark pink zone; Abbott et al., 2015) and from measured NW and NE Pacific seawater (light pink zone; sites BO-3, BO-5, LM-2, LM-6/11, TPS 24 271-1, and TPS 24 76-1;Amakawa et al., 2009). Site U1438 data, spanning the last 24 Ma, largely sit within the expected range of seawater. (b) Modern ε Nd values for sediment collected close to island arcs (Jones et al., 1994), considered influenced by radiogenic volcanic material. The ε Nd (t) of fish teeth and Fe-Mn oxyhydroxide coatings from Site U1438 (red box-and-whisker), and detrital alumino-silicates (black triangles), do not show a strong radiogenic signal, but rather a typical range for intermediate and deep North Pacific seawater (shown as pink bar). diversion of Indian Ocean deepwater southward around Australia (Gourlan et al., 2008) could have had some unconstrained influence on source waters for CDW. In the West Pacific, Le Houedec et al. (2016) hypothesize irregular seafloor tectonic reorganization to explain the pseudo cyclic 7-11 Ma variations in ε Nd at Site 807 from 35 Ma onward. If tectonic variations did occur, they may have controlled the rate of margin volcanism in the West Pacific. Although our record is of lower resolution than 807, there are similar changes to U1438 from 14 Ma onward (Figure 5), in particular a positive excursion in the Pliocene, which may be indicative of greater margin activity around the Philippine and Japan margins (Le Houedec et al., 2016). However, as the relatively abrupt negative ε Nd shift to Site U1438 at~14 Ma is counter to the more positive NPDW records (including Site 807), regional margin activity does not appear to be a plausible cause of the less radiogenic shift at~14 Ma.

Antarctic Glaciation at 13.8 Ma
We conclude that the rather abrupt appearance of CDW in the Philippine Sea between 14.8 and 13.7 Ma could not have been caused by gradual tectonic closure of the Isthmus of Panama or the closure of the Indonesian Gateway (which would have elevated West Pacific ε Nd values). An alternative explanation is Antarctic glaciation at 13.8 Ma, increased deepwater formation around Antarctica, and increased export of CDW further north into the Pacific. The approximate temporal coincidence of increased CDW flux to the Philippine Sea with the rapid (<100 ka) ice sheet build up on East Antarctica and sea level fall (Lear et al., 2010) at the MMCT ( Figure 5) indicates that greater production of CDW may have resulted from Antarctic processes, although we acknowledge that higher resolution records are needed to test this temporal link. Greater production of CDW in the Southern Ocean during the MMCT was interpreted from increased glacial offsets in δ 18 O with depth at various sites east and west of New Zealand (Flower & Kennett, 1995), and significantly strengthened overturning circulation in the Southern Ocean during the MMCT was modeled from sea level fall, ice sheet growth, and CO 2 reduction (Huang et al., 2017). In that study, Antarctic ice sheet growth, declining temperatures, and a greater extent of sea ice particularly in the Ross and Weddell Seas promoted AABW formation during the MMCT (Huang et al., 2017). Despite no change in ε Nd at equatorial Pacific intermediate sites over the MMCT, and a modest positive shift in anomalously unradiogenic intermediate water in the South China Sea (Ma et al., 2018), Holbourn et al. (2013) suggested that enhanced benthic δ 13 C gradients between Pacific intermediate and deep sites were caused by an invigorated PMOC at 13.8 Ma and that increased benthic δ 18 O offsets were caused by cooling NPDW. In the context of our new ε Nd data, the rapidity of these δ 13 C and δ 18 O changes and approximate coincidence with the MMCT (Holbourn et al., 2013) suggests that the CDW expansion detected at Site U1438 was possibly caused by enhanced CDW formation and northward flux in response to Antarctic glaciation since the MMCT, the largest Antarctic climate switch of the past 25 Ma.

Conclusions
New Nd isotopic data from IODP Site U1438 in the Philippine Sea (4.7 km water depth) provide the first constraints on the evolution of deep water in the abyssal North Pacific over the Neogene. Although there is a need for future studies of NW Pacific marginal marine sediments and pore waters near volcanic centers, comparisons with available ε Nd data from these settings, and from open Pacific sites, suggest that changing ocean water values rather than pore waters were the principal driver of the data set. Our results indicate that the water source for the deep Philippine Sea remained relatively stable for the majority of the last 23 Ma, as ε Nd values follow the gradual trends that are present in both intermediate and deepwater sites in the Pacific Ocean attesting to its long term stratification. The gradual increase toward more radiogenic values in all Pacific intermediate and deep sites from~15 to~5 Ma may be due to a gradual decline in Atlantic Ocean inflow through a closing Isthmus of Panama (Martin & Haley, 2000). However, a unique change in ε Nd at Site U1438, expressed as a secular drop of 1 ε unit compared to other Pacific sites, occurred between 14.8 and 13.7 Ma which signifies a likely northward expansion of CDW into the abyssal Philippine Sea. It is possible that a stifled inflow of saline Atlantic water through a restricting Isthmus of Panama reduced North Pacific surface water salinity and inhibited local deepwater formation. However, the coincidence of this increased northward flux of CDW with global cooling and ice sheet expansion at the MMCT (13.8 Ma) could be explained by enhanced formation of CDW source waters in a cooling Southern Ocean (Huang et al., 2017) causing a greater flow of CDW into the deep North Pacific. Higher temporal resolution records from other North Pacific sites are now required to constrain the timing and extent of this expanded CDW flow. A smaller change in our records occurred at 5 Ma, with a 0.5 ε unit increase that is mirrored in West Pacific Site 807, and could be a regional radiogenic water mass signal or possibly the coincidental influence of dispersed ash. Our records suggest that PMOC responded to the most significant long-term climate change event of the Neogene. By contrast, outside of this interval, PMOC appears to have been relatively stable, within the resolution and sensitivity of the records, for much of the Neogene. Our results have implications for future studies aiming to constrain global carbon cycle changes over the MMCT, as an increased carbon-rich CDW flux to the deep North Pacific may have replaced local NPDW ventilation before 14 Ma.