Unexpectedly High Magma Productivity Inferred From Crustal Roughness and Residual Bathymetry on the Eastern Part of the Ultra‐Slow Spreading Gakkel Ridge Since ∼45 Ma, Eurasian Basin, Arctic Ocean

The Gakkel Ridge in the Eurasian Basin has the slowest seafloor spreading worldwide. The western Gakkel Ridge (3°W–85°E; 14–11 mm/a) alternate between magmatic and sparsely magmatic zones, while the eastern Gakkel Ridge (85–126°E; 11–6 mm/a) appears to be dominated by magmatic zones despite ultraslow spreading. Little is known about the seafloor spreading conditions in the past along the entire ridge. Here, we exploit the residual bathymetry and basement roughness to assess the crustal accretion process of the Gakkel Ridge over time using 23 published regional multichannel seismic reflection profiles. Full seafloor spreading rates were faster (20–24 mm/a) up to ∼45 Ma, and residual bathymetry for the older crust is deeper than the world average in the entire Eurasian Basin. There is a sharp transition to 300–400 m shallower residual bathymetry for seafloor <45 Ma in the eastern Eurasian Basin. The crustal roughness versus spreading rate of the western Eurasian Basin is on the global trend, while that of the eastern is significantly below. Both low roughness and shallow residual bathymetry of the eastern Eurasian Basin is close to that of oceanic crust for spreading rates above 30 mm/a, demonstrating increased magmatic production of the eastern Gakkel Ridge since ∼45 Ma. A recent mantle tomography model predicts partial melting in the upper mantle based on the low Vs anomaly underneath. The sedimentary pattern toward the Lomonosov Ridge indicates that this hot mantle anomaly started to cause dynamic uplift of the area at ∼45 Ma.


Introduction
The Gakkel Ridge in the Eurasian Basin is the slowest spreading mid-ocean ridge on earth.Its current full spreading rate decreases from 14.6 mm/a to 6.3 mm/a from west to east (DeMets et al., 1994).The Eurasian Basin was created when the continental Lomonosov Ridge rifted off from the northern Barents and Kara seas at about 56 Ma (Karasik, 1968;Vogt et al., 1979).Initial seafloor spreading rates were more than twice the present, but fell to ultraslow in the 43-47 Ma interval (e.g., Glebovsky et al., 2006).The western part of the Gakkel Ridge (3°W-85°E), has been sampled, and proved more magmatically active than expected from its ultraslow seafloor spreading rate (Michael et al., 2003).However, magmatism is quite variable along the ridge, and Michael et al. (2003) divided the western Gakkel Ridge into a central sparsely magmatic zone (SMZ), flanked by the Western Volcanic Zone (WVZ) and the Eastern Volcanic Zone (EVZ) (Figure 1).Seismic results from within the western Gakkel Ridge rift valley indicate that the crust is thin (2-3.5 km) and without a well-developed (intrusive) layer 3 (Jokat et al., 2003;Jokat & Schmidt-Aursch, 2007) (Figure 2).3D gravity modeling show that the oceanic crustal thickness along the western Gakkel Ridge can be very variable, from 1 to 3 km at its thinnest, to 4-6 km under volcanic centers (Schmidt-Aursch & Jokat, 2016).The eastwards decrease in the spreading rate does not correlate with decreasing magmatism (Jokat et al., 2003;Michael et al., 2003), as seen globally (Bown & White, 1994;Dick et al., 2003), suggesting that magmatic productivity along the western part of the Gakkel Ridge is controlled by variations in mantle temperature and/or composition (e.g., Michael et al., 2003;O'Connor et al., 2021;Wanless et al., 2014).It remains unclear to what extent this influence continues along the eastern part of the Gakkel Ridge (85-126°E) at even lower spreading rates (11-6 mm/a).Nikishin et al. (2018) indicate that the eastern Gakkel Ridge is dominated by four magmatic segments, separated by quite narrow sparsely magmatic or amagmatic zones.Recent high-resolution multibeam bathymetric data show at least three volcanic centers from 85°E to 100°E (Ding et al., 2022).Preliminary results of the wide-angle seismic refraction data from the same study shows a crustal thickness of 6-7.5 km under the magmatic centers, decreasing to 4 km between centers (Ding et al., 2022) (Figure 2).At ∼120°E, the Gakkel Ridge Deep (GRD) on the spreading axis is surrounded by volcanic ridges.Dredging of a seamount at the GRD rim shows an enriched mantle source somewhat similar to that of the SMZ in the west (Jokat et al., 2019).
There are areas with focused magmatism on the spreading axis today, tied to basement ridges that continue off from the Gakkel Ridge.These can be followed out to seafloor of up to ∼25 Ma age in the west (Jokat et al., 2003), showing that they were stable over time.Four similar ridge-orthogonal volcanic ridges at the eastern Gakkel Ridge suggest that focused magmatism over time also occurred here (Jokat et al., 2019).Off-axis from the spreading ridge, several ice-floe based sonobuoy records primarily in the western Amundsen and Nansen basins indicate 2-6 km crustal thickness (Castro et al., 2023;Duckworth & Baggeroer, 1985;Døssing et al., 2014;Funck et al., 2022;Jackson et al., 1982Jackson et al., , 1984;;Kristoffersen et al., 1982;Poselov et al., 2011) (Figure 2).In summary, publications show a range of results, which could either indicate that the Gakkel Ridge seafloor accretion was variable, or that results are poorly constrained due to limitations imposed by the arctic environment on these studies.
The existing crustal thickness measurements are insufficient to map the magmatic development of the Gakkel Ridge since its separation from the Siberian shelf in Cenozoic times, and to compare between regions.However, there are other proxies for magma productivity that can be applied here.One approach is to calculate the rootmean-square (RMS) crustal roughness (e.g., Goff, 1991;Ma & Cochran, 1997;Malinverno, 1991).The crustal roughness depends on the axial morphology, which is created by the interplay between magmatic and tectonic processes (Ehlers & Jokat, 2009;Goff, 2020;Goff et al., 1997).A number of studies (e.g., Ehlers & Jokat, 2009;Goff, 2020;Malinverno, 1991;Sauter et al., 2011) show that there is a negative correlation between the spreading rate and crustal roughness globally: when the spreading rate decreases from slow to ultraslow, the roughness increases significantly.Slow spreading tends to reduce the magmatic crustal thickness due to a colder and thicker lithosphere (Lin & Morgan, 1992;Reid & Jackson, 1981).This strengthens the lithosphere, giving increased fault spacing and abyssal hills larger than those generated at intermediate and high spreading rates.However, the spreading rate is not the sole controlling factor of the crustal roughness; variations in the axial zone mantle temperature and compositions can also contribute due to its control of mantle melting (Goff, 2020;Goff & Jordan, 1989).Thus, deviations from the global data set can be interpreted in terms of magma productivity.
Seafloor subsidence is mainly controlled by passive thermal cooling, though variations can correlate with regional and local tectonics and volcanism (Crosby & McKenzie, 2009;Magde & Sparks, 1997).Under slow to ultraslow spreading, the expected reduction in magma generation in the axial zone will give a thinner oceanic crust.That will result in a deeper seafloor as an isostatic response both to the reduced crustal thickness, and to the increased cooling of the axial zone resulting in increased mantle density underneath (Lin & Morgan, 1992).Reduced crustal production and deeper basement with falling spreading rate is commonly observed at mid-ocean ridges (e.g., Klein & Langmuir, 1987;Lizarralde et al., 2004).Increasing crustal roughness should therefore correlate with deeper residual bathymetry (sediment load and thermal subsidence corrected bathymetry) if the seafloor spreading is magma-starved.
Five seismic reflection profiles have previously shown a high crustal roughness (450-584 m) of the western Eurasian Basin (Ehlers & Jokat, 2009;Weigelt & Jokat, 2001).We combine these profiles with 17 multichannel seismic profiles from Nikishin et al. (2018), and one additional seismic profile from the Alfred Wegener Institute (AWI) (Weigelt et al., 2020), together covering the axis and/or flanks of the Gakkel Ridge from 3°W to 126°E (Figure 1).These profiles cover a significant part of the Eurasian Basin to improve our understanding of the interaction between magmatic and tectonic processes along the entire Gakkel Ridge over much of its spreading history, which we explore here.

Materials and Methods
In this study, we derive the bathymetry and the depth of the basement along profiles from the published images of the seismic profiles (Jokat & Micksch, 2004;Nikishin et al., 2018;Weigelt et al., 2020).Of the 23 seismic reflection profiles, seven have depths in kilometers, while 16 have depths in two-way travel time.

Sediment Corrections
To convert the time sections to depth, we employ regional empirical velocity-depth relationships (Nansen Basin: V p = 1.644 + 0.423Z km/s; Amundsen Basin: V p = 1.7161 + 0.5459Z km/s [Z: sediment thickness in km]) for the sedimentary package based on regional seismic velocity models (Engen et al., 2009;Jokat & Micksch, 2004;Nikishin et al., 2018;Weigelt et al., 2020).The seafloor depth is estimated from the seabed reflection times using  (Jakobsson et al., 2020) showing regional features and locations of the multichannel seismic profiles.Profiles 11-003 to 11-035; 14-07 and 14-05 are from the Russian Federal projects Arktika-2011, 2012, and 2014(Nikishin et al., 2018), while Line 1, Line 2, AWI-20010100, AWI-20010300, AWI-20180300 are from the Alfred Wegener Institute (Jokat et al., 1995;Jokat & Micksch, 2004;Weigelt et al., 2020).The black lines mark profiles with depth in kilometers, while the white lines show profiles with two-way travel time depths.Red solid lines indicate the boundaries of the three axial regions identified by Michael et al. (2003).Red dashed line shows the Gakkel Ridge spreading axis.EVZ: eastern volcanic zone, GRD: Gakkel Ridge deep, SMZ: sparsely volcanic zone, WVZ: western volcanic zone.The Nansen Basin in the south and the Amundsen Basin in the north constitute the Eurasian Basin.a water velocity of 1,480 m/s.We adopted an Airy isostatic correction to remove the effect of the sediment loading (Le Douaran & Parsons, 1982): where B is the observed basement depth, B s is the basement depth corrected for the sediment loading, ρ a is the density of the asthenospheric mantle (3,200 kg/m 3 ), ρ w is the density of water (1,030 kg/m 3 ), and Z s is the sediment thickness.Additionally, ρ s is the average sediment density, which is calculated from: where ρ z is an empirical density-depth relationship of the sediments in the Arctic region (Engen et al., 2006): where ρ c =2,890 kg/m 3 is the crystalline crustal density, ρ 0 =1,900 kg/m 3 is the density of the uppermost part of the sediment column, α = 0.15 is the sedimentary compaction parameter, and z is the depth below seafloor in km.
In order to keep the morphology of the abyssal hills intact, we apply a Gaussian filter with a half-width of 50 km to the sediment load correction before applying it.

Subsidence Analysis
There should be a linear relationship between the sediment-load-corrected basement depth and the square root of the seafloor age for subsidence constrained only by passive thermal cooling (e.g., Crosby & McKenzie, 2009).The rate of subsidence can therefore be obtained by plotting the observed sediment load corrected basement depth against the square root of the seafloor age, followed by a linear regression.For this, we use a sampling interval of 1 km.The age of the oceanic crust was estimated from the magnetic anomalies identified along each profile (Jokat & Micksch, 2004;Nikishin et al., 2018;Weigelt et al., 2020), with ages from the geomagnetic polarity timescale of Ogg (2020).Since the spreading changes from slow to ultraslow during the 43-47 Ma interval (Brozena et al., 2003;Glebovsky et al., 2006;Karasik, 1968;Vogt et al., 1979), we estimate the subsidence before and after that time.To avoid a data-gap, we use 45 Ma to divide between the two.
We further calculate the residual bathymetry from the difference between the sediment-load-corrected basement depth and a global subsidence curve (subsidence rate: 325 m/Myr 1/2 and zero-age depth: 2,600 m) (Crosby & McKenzie, 2009).A positive residual bathymetry indicates an anomalously deep basement.

Crustal Roughness Analysis
The crustal roughness is based on the RMS deviation between the unloaded basement depth and a reference curve along a given profile length (Ehlers & Jokat, 2009;Malinverno, 1991).The reference curve of earlier investigations have either been based on a curve fit of basement depth for each profile (Ehlers & Jokat, 2009), or from a regional curve with a given subsidence rate and zero-age depth (e.g., Malinverno, 1991;Weigelt & Jokat, 2001).The correlation by Malinverno (1991) is derived from 101 profiles, 500-1,000 km in length, while the correlation by Ehlers and Jokat (2009) is based on five profiles from the northern North Atlantic and Arctic, approximately 200-500 km in length.Here, we calculate a local reference curve for each profile, based on linear regression between the sediment-load-corrected basement depth and the square root for seafloor age before and after 45 Ma, separately.That will prevent arbitrarily large roughness values if the local bathymetry deviates systematically from a regional or global model.The length for the roughness calculation was chosen to eliminate the effects of the window borders and self-affinity (typically 70 km, but ranging from 40 to 100 km) (Goff & Jordan, 1989;Malinverno, 1991).In addition, every chosen length excluded inside and outside corner reliefs.We also exclude the area close to the ridge (seafloor age younger than 10 Ma) from both the subsidence analysis and the crustal roughness calculations, since the morphology appears to still be developing there.
We calculate the regional spreading rates over time based on flow lines made from rotation poles between Eurasia and North America by Gaina et al. (2002), using the program backtracker, which is part of GMT 6.0 (Wessel et al., 2019).These are representative for the regional spreading-induced mantle upwelling and melting for plotting the roughness values.

Basement Subsidence
Figures 3b and 4b show the basement depths corrected for sediment loading for all profiles, where the red lines represent the local reference curve of each profile used to calculate roughness, and the green lines show the global subsidence trend (Crosby & McKenzie, 2009).We estimate subsidence and roughness separately before and after ∼45 Ma as described above.We also group the results separately for the western (Figure 3b) and eastern (Figure 4b) Eurasian Basin, with the boundary at 85°E on the Gakkel Ridge.
For most profiles of the western Eurasian Basin, the basement is 300-500 m deeper than the global trend (Figure 3b).The exceptions here are Line 1 in the westernmost part of the basin and Line 2 in the central western Amundsen Basin, where the basement depth for seafloor age younger than ∼45 Ma is close to the global trend.For Line 2 the oldest basement is deeper than the younger basement, while for Profile 11-005 it is shallower, and there are no systematic differences between the older and younger parts, often there is little change.This is in contrast to the eastern Eurasian Basin, where the oldest basement is 300-400 m deeper than the younger parts for most profiles, particularly for Profiles 11-026, 11-028, 11-029 and AWI20180300 in Figure 4b.The transition is sharp and confined to the 43-47 Ma interval.
In Figure 5, we plot the basement depths of all the profiles from Figures 3b to 4b against the square root of time, with linear regressions to estimate subsidence rate for the two age intervals in the western and eastern Eurasian Basin, respectively.For seafloor younger than ∼45 Ma, the subsidence rate of the west is 305 ± 20 m/Myr 1/2 , and 299 ± 10 m/Myr 1/2 in the east, but the zero-age depth for the west (3,120 ± 100 m) is 320 m deeper than that of the east (2,800 ± 100 m) (Figure 5), even though the west has a higher spreading rate.For the older seafloor, the subsidence rate of the west is 308 ± 10 m/Myr 1/2 , and 293 ± 10 m/Myr 1/2 in the east, while the zero-age depths for the west and east are the same (3,150 ± 50 m).

Residual Basement Depth
To explore these regional variations in basement depth further, we calculate the residual bathymetry for all profiles (Figures 6 and 7).Averages for crust older and younger than ∼45 Ma are shown for each profile.For the western Eurasian Basin, the residual bathymetry averages are positive (290-641 m) for seafloor age older than ∼45 Ma on both sides of the basin.However, for seafloor younger than ∼45 Ma, there is an apparent asymmetry between the Amundsen and Nansen basins (Figures 6 and 8).The depths range from 30 to 101 m in the Amundsen Basin, while the profiles (except Line 1) in the Nansen Basin have positive residual basement depths from 154 to 712 m (Figure 6).For the eastern Eurasian Basin, all the profiles except for profile 14-07, cover only the Amundsen Basin (Figures 7  and 8).For oceanic crust older than ∼45 Ma, the average residual bathymetry is between 206 and 831 m, comparable to that of the western Eurasian Basin.For seafloor younger than ∼45 Ma, the average residual bathymetry is much reduced, showing alternating positive and negative values ( 158-287 m) (Figures 7 and 8).It is also less variable than in the west.Profile 14-07 shows a similar asymmetry in residual bathymetry between the Amundsen and Nansen basins (Figures 7 and 8) as seen in the west: For the younger oceanic crust, the Amundsen Basin shows slightly negative residual bathymetry (average: 63 m), while the Nansen Basin has positive values  with an average of 287 m (Figure 7).For oceanic crust older than ∼45 Ma, both Amundsen (average: 206 m) and Nansen basins (average: 294 m) have positive residual bathymetry on this profile.

Crustal Roughness
The roughness for individual profiles is calculated from the difference between basement depth and the local reference curve of each profile (red lines) in Figures 3a and 4a. Figure 9a shows the correlation between the full spreading rate and the crustal roughness of the seafloor for each profile.The crustal roughness values are divided into the western (green circle) and eastern (blue circle) parts of the Eurasian Basin, with error bars showing standard deviations.The red and black curves show the global correlation between spreading rates and crustal roughness from Malinverno (1991), and Ehlers and Jokat (2009), respectively.The green curve represents the best fit between the spreading rates and crustal roughness based on our analysis from the western Eurasian Basin (Rc = 1,236(v) 0.5541 , where the Rc is the crustal roughness (m), and v is the full spreading rate).
The roughness values for the western and eastern parts are 215 ± 50 m and 260 ± 40 m, respectively, which is consistent with the curves from Ehlers and Jokat (2009) and Malinverno (1991).For basement younger than ∼45 Ma, the western part had spreading rates between 11 and 14 mm/a, corresponding to higher roughness values, but with large variations (300 ± 100 m).Our curve fit for the western Eurasian Basin is close to that of Figures 9b and 9c show the correlation between the residual bathymetry and the crustal roughness of the western and eastern Eurasian Basin before and after ∼45 Ma.For the western Eurasian Basin, the results show much scatter, and there are no clear trends.For the eastern Eurasian Basin, there is a clear trend from high crustal roughness and deep residual bathymetry for the older seafloor, to smaller crustal roughness and shallower residual bathymetry for the younger, despite the spreading rate decrease from slow to ultraslow.

Uncertainties
The primary error in 2D velocity-depth analysis comes from unconstrained velocity variability, typically ranging between 10% and 5% (Engen et al., 2009).Higher velocities lead to greater basement depth, increased sediment density and thickness, but these factors counteract during sediment loading corrections.To illustrate how this works, we use profiles 11-005 and 11-029, which have thick sediment covers in the Nansen and Amundsen basins, respectively.By changing the velocity from 10% to 5% results in basement depth and sediment thickness changes by 420 to 120 m, and if the density is coupled to the velocity following an empirical relationship (Ludwig et al., 1970), average sediment density varies by 0.15 to 0.05 g/cm 3 .Cumulatively, these changes only result in a small residual bathymetry uncertainty of up to ±50 m, and much less where the sediments are thin (Figure S1 in Supporting Information S1).
The sedimentary thickness of the Amundsen and Nansen basins differ significantly.During the opening of the Eurasian Basin, the Amundsen Basin received sediment from Barents-Kara and Laptev Sea shelves, as well as from the Lomonosov Ridge (Castro et al., 2018;Nikishin et al., 2018;Weigelt et al., 2020).From middle Eocene, the Amundsen Basin was increasingly isolated from these sediment sources, and became more dominated by pelagic sedimentation.In contrast, the Nansen Basin continued to receive sediments form the south, accumulating about twice as much as the Amundsen Basin (Engen et al., 2009;Jokat & Micksch, 2004;Nikishin et al., 2018).This results in greater velocities for sediments of the same age in the Nansen Basin due to the increased loading resulting in increased compaction, though the differences at the same burial depth is much less (Engen et al., 2009;Jokat & Micksch, 2004;Nikishin et al., 2018;Weigelt et al., 2020).The compaction parameter will describe this increase as the sedimentary package becomes thicker, and it is a reasonable way to predict densities in the absence of high-resolution velocity data for the sedimentary column.Still, lithological differences may result in density uncertainties (Ludwig et al., 1970).To simulate this, we test the sensitivity to the sedimentary compaction parameter (α) of profiles 11-005 and 11-029 by changing it from 0.05 to 0.25, resulting in density uncertainties of 0.13 to 0.1 g/cm 3 (Figure S2 in Supporting Information S1).For seafloor ages younger than ∼45 Ma, the basement depth of the profile 11-005 changed by 120-80 m, while for the profile 11-029 it remains almost unchanged due to the thin sediment cover.For ages older than ∼45 Ma, the basement depth of the profiles 11-029 and 11-005 was changed by 120-60 m and 160-100 m, respectively (Figure S2 in Supporting Information S1).
The shallowing of the basement over seafloor ages of ∼45 Ma is insensitive to this uncertainty due to its sharp character.We consider these estimates as maximum uncertainty, as the effect of depth conversion due to expected accompanying velocity variations are not accounted for.Thus, the residual bathymetry difference between the western and the eastern Eurasian Basin still largely remains.

Discussion
The discussion will cover three themes.First, we take a look at studies of other ultraslow spreading ridges, how any anomalies are explained, and how they compare to the Gakkel Ridge.Secondly, we take a closer look at a recent mantle tomography model of the area, which offer an explanation for the unexpectedly high magma productivity our results show for the eastern Gakkel Ridge.Lastly, we take a closer look at published stratigraphic interpretations and bathymetry for clues to how the change in magma productivity and the low subsidence rate found by our results, may have developed over time.

Comparison to Other Anomalous Ultra-Slow Spreading Ridges
There are other spreading ridges around the world that have very slow spreading, for example, the South-West Indian Ridge (SWIR).Sauter et al. (2011) analyzed the crustal roughness within areas on the flanks of the SWIR between 54°E and 67°E, based on multibeam bathymetric mapping.Their study showed that the roughness increases significantly from 220 ± 20 m to 300 ± 20 m when the full spreading rate decreased from slow (30 mm/ a) to ultraslow (15 mm/a).Furthermore, Sloan et al. (2012) reported a reduction in roughness from volcanic to nonvolcanic seafloor from 372 ± 23 m to 304 ± 20 m for the SWIR east of 61°E, even though both areas have a similar spreading rate of 15 mm/a.
An enriched mantle source promotes magmatic generation, causing a warmer and thinner lithosphere, and combined with increased crustal thickness, results in a reduced crustal roughness.This appears to be the case for the ultraslow spreading South Pandora (18 mm/a; 158 ± 16 m) and western Sheba ridges (20 mm/a; 103 m) (d'Acremont et al., 2010;Lagabrielle et al., 1996;Sauter et al., 2018).The South Pandora and Sheba Ridges have crustal thicknesses of 6 and 8 km, respectively (d'Acremont et al., 2010;Garel et al., 2003).This can be compared to the Reykjanes Ridge (full spreading rate: 20 mm/a), where the low crustal roughness (<100 m) and large crustal thickness (9 km) are attributed to the thermal influence of the Iceland plume (Shi et al., 2018;Small, 2013).
Our crustal roughness values (300 ± 100 m) in the western part younger than 45 Ma are comparable to those predicted from other ultraslow spreading ridges (Figure 9a).Our curve fit is slightly lower than the world average from Malinverno (1991), but significantly below that of Ehlers and Jokat (2009) for the western Eurasian Basin (Figure 9a).We include three profiles from the Ehlers and Jokat (2009) study, and with 8 additional profiles, the curve fit is based on 11 profiles.Differences could be explained by the larger data set.However, the roughness of individual lines from Ehlers and Jokat (2009) is lower in our analysis, primarily caused by differences in the reference lines.There is a large variability to the results, indicating that the spreading conditions varied in space and time.Roughness within the ultraslow spreading domain can vary between 100 m (Profile 11-006) and 532 m (Profile AWI20010300) (Figure 3a).Similarly, the residual bathymetry shows much variation, but is on average about 400 m deeper than the global model (Figures 6 and 9b).Uneven data coverage makes it difficult to see how systematic these variations are.The low correlation between residual bathymetry and crustal roughness in the western Eurasian Basin (Figure 9b) could indicate later dynamic uplift of some areas, likely caused by movement of warmer mantle material in the asthenosphere into the region (e.g., Hartley et al., 2011), in agreement with the lower than normal subsidence rate.The crustal roughness (232 ± 40 m) of the eastern Eurasian Basin younger than 45 Ma is ∼68 m lower than that of the western Eurasian Basin, and 200-300 m lower than predicted by Malinverno (1991) and Ehlers and Jokat (2009) (Figure 9a).The rapid decrease in basement depth from older to younger crust at ∼45 Ma, coeval with a substantial decrease in spreading rate, is also highly unexpected.It could possibly be due to a large increase in serpentinization.Sloan et al. (2012) observed a basement roughness reduction from 372 ± 23 m to 304 ± 20 m going from volcanic seafloor to one dominated by mantle serpentinization, for the SWIR east of 61°E.This roughness difference is similar to that observed between the western and eastern parts of the seafloor created by the Gakkel Ridge (Figures 9a and 9c).However, the crustal roughness (232 ± 40 m) in the eastern Eurasian Basin is much lower than that of the serpentinized seafloor at the SWIR (304 m) (Sloan et al., 2012), and thus does not support the onset of extensive mantle serpentinization as an explanation for the abrupt shallowing of the basement.Thus, the crustal accretion process of the eastern Gakkel Ridge cannot be explained by the seafloor-spreading rates, other factors must contribute.

Comparison to Mantle Tomographic Results
A number of mantle tomography models of the Arctic region have shown that there is an upper mantle low-shear velocity anomaly, indicating a hotter mantle, situated under parts of the Eurasian and Polar basins (e.g., Auer et al., 2014;Debayle et al., 2016;French et al., 2013;Jakovlev et al., 2012;Lebedev et al., 2018;Ritsema et al., 2011;Schaeffer & Lebedev, 2013).The Vs tomography model of Lebedev et al. (2018) is the most recent and explores some implications for the Arctic region in depth, which we use for the discussion.The Amundsen Basin is located directly above a low-velocity anomaly extending down to a depth of more than 200 km, while the mantle anomaly underlies only a minor part of the Nansen Basin (Lebedev et al., 2018) (Figure 8a).This anomaly extends across the central Arctic from the Canada Basin to the Eurasian Basin.The observed high heat flow in the Amundsen Basin also supports the presence of this mantle low velocity anomaly (Urlaub et al., 2009).Lebedev et al. (2018) estimate the likelihood of partial melting to presently occur in the mantle from their tomographic model by converting low-velocity shear-wave anomalies to temperature.Results indicate that partial melting should occur under the eastern Eurasia Basin, but to a lesser degree under the western basin, predicting higher melt production under eastern Gakkel Ridge today compared to the west.We know from the roughness analysis and residual bathymetry that the magma production increased at ∼45 Ma and remained robust up until present in the east.It is therefore reasonable to assume that warmer mantle material entered the melting zone of the spreading ridge at that time, increasing oceanic crustal thickness despite the ultraslow spreading rate.The western Eurasian Basin is less affected by this anomaly, having an oceanic crust of more variable thickness with time and location, but generally thinner.Dredged samples from the western Gakkel Ridge show mantle source heterogeneity, influencing its magmatic variations (Goldstein et al., 2008;Michael et al., 2003;O'Connor et al., 2021;Wanless et al., 2014).However, no dredged samples exist east of 85°E due to thick sediment cover, except at the GRD rim (120°E), revealing a mantle source similar to the SMZ (Jokat et al., 2019).Low S-wave mantle velocity may be caused by mantle heterogeneity (Goes & Van Der Lee, 2002).However, the velocity anomalies caused by mantle composition variations are unlikely to exceed 1% (Priestley & Mckenzie, 2006;Schutt & Lesher, 2006).Since S-wave anomalies commonly are 4% beneath the eastern Eurasian Basin (Figure 8a), it's likely that the eastern Gakkel Ridge magmatism is caused by increased mantle temperature with extra fertile mantle at most playing a minor role.
Low-velocity Vs anomalies with the associated higher upper mantle temperature will have two effects on the area.It will not only increase the melting in the melt-zone underneath the spreading ridge, but also cause dynamic topography due to the associated lower density (e.g., Gurnis, 1993;Gurnis et al., 2000;Hartley et al., 2011;Parnell-Turner et al., 2014;Rickers et al., 2013).The work of Hartley et al. (2011) in the NE Atlantic explores the effect of mantle plume movement on the uplift of landscape by studying its effect on the sedimentary stratigraphy, concluding that a 200-400 m landscape uplift could be attributed to a 30-60°C increase in mantle temperature within a plume channel thickness of 200 km.The present anomaly should have developed throughout the Cenozoic, combined with plate movements relative to its location at any time.The dynamic topography related to the anomaly should also develop as a response, leaving a trace not only in the low subsidence rate we observe, but also in the sedimentary depositional pattern and bathymetry, which we explore in the next sub-section.

Vertical Movements Recorded by Stratigraphy/Bathymetry
Low subsidence rates (293-299 m/Ma 1/2 ) are observed in the eastern part of the Amundsen Basin, which should be the result of dynamic topography uplift countering the thermal subsidence.Some of the profiles even indicate a shallowing of the residual bathymetry with age for the oldest oceanic crust adjacent to the Lomonosov Ridge, see for example, profiles 11-029 and 11-030 in Figure 4. Comparing subsidence curves based on a regional subsidence rate of 293 m/Ma 1/2 to a global subsidence rate of 325 m/Ma 1/2 , we estimate the uplift of oceanic crust near the Lomonosov Ridge, aged between 45 and 54 Ma, to be approximately 0.2-0.25 km.An uplift of this magnitude may be linked to mantle temperature anomalies of 20-30°C at depth of 50-200 km, based on Hartley et al. (2011), and is well within the range predicted by mantle tomography (Lebedev et al., 2018).
The seismic profiles derived from Nikishin et al. (2018) in the eastern Eurasian Basin show that sedimentary deposits older than ∼45 Ma generally thicken toward the Lomonosov Ridge, as would be expected from normal thermal subsidence.That should also apply to the ∼45 to ∼34 Ma sequence, but instead it thins in the same direction (Figure 10).This is not observed in the western Eurasian Basin (Jokat et al., 1995;Jokat & Micksch, 2004;Nikishin et al., 2018).The anomalously low oceanic basement subsidence adjacent to the ridge in east (Figures 4 and 5) corresponds with this, showing that this pattern is not a result of early high sedimentation rates.The influence of the mantle anomaly appears to have developed mostly through the ∼45 to ∼34 Ma interval, as the ∼34 to ∼20 Ma sedimentary sequence shows less thinning toward the Lomonosov Ridge (Figure 10).Furthermore, the 0-20 Ma sequence does not show any systematic thinning in that direction at all.The influence of the anomaly therefore appears to have developed mostly through the ∼45 to ∼34 Ma interval so that the mantle anomaly presently observed was well established under the Gakkel Ridge by ∼20 Ma (Figure 10).
There is an abrupt shallowing of the basement to the younger side over the 43-47 Ma interval observed at almost all the profiles of the eastern Amundsen Basin (Figures 7 and 8).A low subsidence rate requires heat input into the mantle underneath carried by convective movement.It would, however, not create a scarp, just a slight change in slope.Dynamic uplift is therefore unsuitable to explain the formation of the sharp transition.Oceanic lithosphere also rapidly gains flexural strength away from the spreading ridge (e.g., Watts, 1978), which would prevent the formation of a scarp due to uneven sedimentation.Rather, both the low crustal roughness and the shallowing basement depth indicates that there was an increase in crustal thickness due to increased mantle melting at the time, when the melting underneath the eastern Gakkel Ridge started to source from the hot mantle anomaly.We estimate the increase in crustal thickness (ΔH c ) from isostasy (Watts, 2001): where ΔH b is the basement depth change, ρ m is the mantle density at 3,300 kg/m 3 , ρ w is the water density at 1,030 kg/m 3 , ρ c is the crustal density at 2,850 kg/m 3 based on a regional gravity study from Døssing et al. (2014).
To explain a 300-400 m shallower basement, a crustal thickness increase of 1.5-2 km is required for isostatic balance.An earlier study shows that the seafloor age older than ∼45 Ma in the eastern Eurasian Basin has a crustal thickness of 4-5 km (Poselov et al., 2011) (Figure 2).If this result is reliable, it suggests an increase in crustal thickness to within the 5.5-7 km range since ∼45 Ma.This is in agreement with the findings from a recent wideangle seismic refraction study conducted along the Gakkel Ridge spreading axis from 80°E to 102°E, showing crustal thickness of 6-7.5 km beneath magmatic centers, gradually decreasing to 4 km between these centers (Ding et al., 2022) (Figure 2).
There is another feature of the Gakkel Ridge that should be noted, since it has several implications.The width of the axial valley as defined by the highest axial mountains in the International Bathymetric Chart of the Arctic Ocean bathymetric compilation (Jakobsson et al., 2020), seems to more than double in the east compared to the west (Figure 1).This implies that the crustal accretion zone must become wider before complete crustal development is reached at the lowest spreading rates.The ultraslow spreading will inevitably give the axial zone more time to cool and create a thicker lithosphere, with less shallow melting (e.g., Bown & White, 1994).This will affect the distribution of melting underneath the axial zone, and we propose that this probably increases the width of the zone for magma extraction.Crustal thickness may not be fully accreted before 40-50 km from the axis center, and at 3-6 mm/a half-spreading rates that takes 7-16 Ma.If this is the case, it should also show in the sedimentary record.2020)) (Figure 11c).That dates the event to occur when the oceanic crust was approximately 10-15 Ma old giving support to a wide crustal accretion zone in the eastern Eurasian Basin, possibly with intensifying magmatism for the last ∼15 Ma.

Conclusions
To estimate changes to the crustal accretion process of the Gakkel Ridge through time, we calculated thermal subsidence rate, residual bathymetry and the RMS crustal roughness of 23 published multichannel seismic profiles from Russian Federal projects and from the AWI.These profiles cross the axis and/or flanks of the Gakkel Ridge from 3°W to 126°E, covering large parts of the Eurasian Basin.The data were grouped into the western and eastern parts of the Eurasian Basin seafloor, divided at 85°E on the Gakkel Ridge.Full seafloor spreading rates were slow (west: 22-24 mm/a; east: 20-22 mm/a) from breakup at ∼56 Ma up to ∼45 Ma.
Seafloor created during this early interval has an average residual bathymetry 550 m below the world mean both in the east and in the west.During the 43-47 Ma interval, the spreading rate fell to ultraslow (west: 11-14 mm/a; east: 6-11 mm/a).The average residual bathymetry in the west remains largely unchanged over this interval, but for the eastern part there is a sharp shallowing of 300-400 m to the younger side.The residual bathymetry for seafloor younger than ∼45 Ma in the east is less than 200 m deeper than the world average.The RMS crustal roughness of seafloor created by the western Gakkel Ridge (300 ± 100 m) is comparable to those of other ultraslow-spreading ridges.However, the seafloor created by the eastern Gakkel Ridge has an RMS crustal roughness of 232 ± 40 m, 200-300 m lower than predicted for ultra-slow spreading by the global crustal roughness data set.Since the eastern Eurasian Basin has crustal roughness and residual bathymetry similar to that of normal oceanic crust at a spreading rate above 30 mm/a, we conclude that the seafloor spreading of the eastern Gakkel Ridge is magmatically robust since ∼45 Ma.The change to shallower basement and hence thicker crust over the 43-47 Ma interval, indicates that there was a change in the mantle melt source at the time, dominated by increased mantle temperature affecting the eastern Gakkel Ridge primarily.Anomalously hot mantle underneath the eastern Eurasian Basin presently is indicated by mantle tomography (Lebedev et al., 2018), and the arrival of this anomaly at ∼45 Ma is recorded in the sedimentary depositional pattern toward the Lomonosov Ridge in the eastern Eurasian Basin (Nikishin et al., 2018;Weigelt et al., 2020).The anomaly affected the western Eurasian Basin much less, but local discrepancies between crustal roughness and residual bathymetry, may indicate postspreading dynamic uplift of some areas.

Figure 2 .
Figure 2. Crustal thickness measurements for the Eurasian Basin.The dot and bars indicate the measurements from ice-floe based sonobuoy records and seismic refraction profiles.Thin green lines indicate seafloor ∼45 Ma old.

Figure 3 .
Figure 3. Basement depth after sediment load corrections of the seismic reflection profiles in the western Eurasian Bain (Figure 1).(a) The panel above each profile shows the difference between the observed basement depth and the local reference curve, which is used for root-mean-square crustal roughness calculation.The parts used and their corresponding crustal roughness values before and after ∼45 Ma are indicated by red and green bars, respectively.The oceanic ages are shown on top of each profile.(b) Red line: local reference curve of each profile before and after ∼45 Ma. Green line: Global seafloor subsidence curve from Crosby and McKenzie (2009).Black line: observed sediment-unloaded basement depth.

Figure 4 .
Figure 4. Basement depth after sediment load corrections along the seismic reflection profiles of the eastern Eurasian Basin (Figure 1).Markings are the same as for Figure 3.

Figure 5 .
Figure 5. Sediment-load-corrected basement depth along the profiles of the Eurasian Basin ((a) western part (3°W-85°E); (b) eastern part (85°E 126°E)) against the square root of the sea-floor age.The full spreading rates before and after ∼45 Ma are indicated at the top.The light green lines represent the profiles with the basement depth in kilometers, while the light gray lines show the basement depth calculated from two-way travel time.The average of the basement depth over the square root of time (sqrt(t)) interval of 0.1 is shown by circles with standard deviation error bars.The red solid lines represent the best-fit age-depth relationship for the seafloor.The dotted blue line shows the global subsidence (Crosby & McKenzie, 2009).

Figure 6 .
Figure 6.Residual bathymetry along each seismic reflection profile of the western Eurasian Basin.The residual is the difference between the unloaded basement depth (black solid line) and the global seafloor subsidence curve (green solid line) (Figures 3b and 4b).Positive values indicate anomalous deep basement.The oceanic ages, and the average of the residual bathymetry over seafloor older and younger than ∼45 Ma, are shown at the top of each profile.Gray lines correlate oceanic ages ∼45 Ma and ∼10 Ma between profiles.Line locations are shown on the inset bathymetry map.

Figure 7 .
Figure 7. Residual bathymetry along each seismic reflection profile of the eastern Eurasian Basin.Markings are the same as for Figure 6.

Figure 8 .
Figure 8.(a) The Vs tomography at 150 km depth (Lebedev et al., 2018).(b) Residual bathymetry and crustal roughness for the Eurasian Basin are indicated by color and grayscale bars on the bathymetry map, respectively.Gray lines correlate ∼45 Ma oceanic age between profiles.

Figure 9 .
Figure 9. (a) Top basement roughness versus full spreading rate.The western and eastern parts of the Eurasian Basin are represented by filled green and blue circles, respectively, with error bars showing standard deviations.The red curve represent the correlation from Malinverno (1991) based on 101 profiles, while the black curve is that of Ehlers and Jokat (2009), based on five high-latitude profiles.The green curve shows the best fit based on our crustal roughness calculations from western Eurasian Basin, based on 11 profiles.The results from previous Eurasian Basin studies are indicated by different symbols.(b, c) The correlation between the residual bathymetry and crustal roughness of the western and eastern Eurasian Basin before and after ∼45 Ma.Red arrows show the expected correlation trends due to changes in the magma budget of the spreading ridge.

Figure 10 .
Figure 10.Depth converted interpretations of select reflection seismic sections with horizons of 45, 34, and 20 Ma over the eastern Eurasian Basin derived from Nikishin et al. (2018) and Weigelt et al. (2020).(a) The colored dashed lines show the sediment thickness of each sequence; (b) The colored solid lines represent the seafloor, sequence boundaries and top acoustic basement.The vertical black dashed lines indicate ∼45 Ma oceanic age.

Figure 11 .
Figure 11.Line drawings of the stratigraphy of the selected seismic reflection profiles from Nikishin et al. (2018) to show the uplift of the strata with subsequent strata onlapping the 45 Ma (a) and 20 Ma horizon (c), and a small post 45 Ma deformation (b).