Roadblocks and speed limits: Mantle-to-surface volatile flux through the lithospheric-scale Denali fault, Alaska

Active strike-slip faults have three-dimensional geometries that vary along strike and can facilitate deep fluid flow and mantle-to-crust connections, with implications for the seismic cycle (e.g., Vauchez et al., 2012; Norris and Toy, 2014). Conversely, ancient “dead” strike-slip faults have been seismically imaged as crustal-scale features offsetting the Moho (Diaconescu et al., 1998), but it is unclear whether these maintain mantle fluid connections. Furthermore, strike-slip faults that penetrate to mantle depths can facilitate lithosphere-scale deformation and act as transform plate boundaries, but not all continental strike-slip faults share these characteristics (Andrew et al., 2023).

Isotopic signatures of gases in springs (e.g., ³He/⁴He, δ¹³C) along active faults are ideal for quantifying mantle-to-crust connections and testing geophysical and structural models for strike-slip fault systems (e.g., Kennedy et al., 1997; de Leeuw et al., 2010; Klemperer et al., 2013). Documenting the presence of persistent mantle-to-surface volatile flux has implications for how fluid pressure gradients may impact fault strength and seismicity along strike-slip faults (Kennedy et al., 1997).

The Denali fault (northwestern North America) is a long-lived (>10⁷ yr), >2000-km-long dextral strike-slip system (e.g., Benowitz et al., 2022). Here we present ³He/⁴He and δ¹³C data from previously (Richter et al., 1973) and newly identified thermal springs along the Denali fault system in Alaska (USA) (Fig. 1). We sampled springs located along and outside of the rupture associated with the 2002 M7.9 Denali fault earthquake (Eberhart-Phillips et al., 2003). Results clarify questions about the lithospherescale nature of the Denali fault, hydrologic connectivity of fault splays and the main strand, and possible hydrologic impacts from the 2002 earthquake. We suggest that similarities in the ³He/⁴He ratio of springs at analogous tectonic settings place constraints on the mantle fluid flux into continental strike-slip faults.

Southern Alaska, composed primarily of the Wrangellia composite terrane (Trop et al., 2020), is delineated by the Denali fault system to the north and east and by the Aleutian Trench to the south (Fig. 1). The Denali fault is active, demonstrated by a M7.9 earthquake in 2002 that ruptured ~340 km of the Denali and Totschunda faults (Fig. 2) (Eberhart-Phillips et al., 2003). Geodetic data indicate that southern Alaska is moving westward and independent of North America (e.g., Elliott and Freymueller, 2020). Hence, further knowledge of the Denali fault system’s lithospheric-scale nature is critical to discerning its role in the northeastern Pacific plate transform system and whether southern Alaska comprises an independent tectonic plate.

The Denali fault system can be divided into the primarily strike-slip Totschunda fault (e.g., Trop et al., 2022), the main strand of the Denali fault which becomes more transpressive east to west (Regan et al., 2021), and the ancient strikeslip and now dominantly dip-slip Hines Creek fault (Benowitz et al., 2022). Based on receiver functions, Brennan et al. (2011) concluded the active central Denali fault does not extend into the mantle nor offset the Moho but the Hines Creek fault offsets the Moho by ~10 km. Conversely, Allam et al. (2017), using double difference tomography and receiver functions, concluded the active strand of the Denali and the Hines Creek and Totschunda faults all extend into the mantle with ~10 km Moho offsets.

Pleistocene slip rates decrease east to west from ~13 to ~5 mm yr^–1 along the Denali fault (Fig. 1) (Haeussler et al., 2017), which are compatible with Oligocene reconstructions of longterm slip decreasing east (~300 km) to west (~150 km) (e.g., Regan et al., 2021). Brocher et al. (2004), based on aftershocks from the 2002 earthquake, inferred splay faults like the Granite Mountain and McCallum Creek thrusts connect to the Denali fault at depth (Fig. 1), but it is not clear whether these faults are also hydrologically connected. Given that deep crustal fluid circulation, including mantle contributions, affects the strength of faults (Kennedy et al., 1997; Lee et al., 2019), any variation in fluid flow along strike of the Denali fault and along splay faults would influence how the faults behaved during earthquakes.

Helium isotope ratios (³He/⁴He) are sensitive for identifying mantle volatiles in crustal fluids and are reported relative to the air ratio (R_A, 1.4 × 10^-6) (Graham, 2002). The convecting upper mantle has values of 8 ± 1 R_A, and crustal fluids have values <0.1 R_A (e.g., Graham, 2002). Mantle fluids introduced to the crust evolve toward lower ³He/⁴He due to mixing with crustal fluids and ⁴He ingrowth (Kennedy et al., 1997). Helium is a trace gas that migrates with other volatiles such as water and CO₂, and integrated ³He/⁴He and carbon stable isotopes (δ¹³C values) can track the provenance of volatiles in fluids in continental settings (Sano and Marty, 1995; Hiett et al., 2021).

We sampled 12 springs along the Denali, Hines Creek, McCallum Creek, and Slate Creek, and Totschunda faults (Fig. 2; Table S1 in the Supplemental Material¹). Detailed methods are in the Supplemental Material. Travertine deposits were observed at most locations, and Windy Creek, Railroad, and Slate Creek springs were bubbling. Springs range from 2.2 to 12.4 °C, with pH of 6.7 ± 0.6 (one exception at 2.4), Na – HCO₃, Ca + Mg – HCO₃, and Ca + Mg – SO₄ compositions, variable dissolved inorganic carbon (DIC) of as much as 7700 mg L^–1, and total dissolved solids of 410–11,600 mg L^–1 (Table S2). Most spring δ¹⁸O and δ²H values (averaging -20‰ and -157‰ relative to Vienna Standard Mean Ocean Water, respectively) are consistent with meteoric recharge (Table S4; Fig. S1). Values from Soda Creek springs on the Totschunda fault are shifted to the right of the meteoric water line (Fig. S1) and indicate a history of waterrock interaction at elevated temperatures along groundwater flow paths (e.g., Hiett et al., 2022).

Spring ³He/⁴He ratios range from 0.65 to 2.65 R_C/R_A (measured ³He/⁴He with air corrections, relative to R_A; Table S3). Eight springs have values indicating ~8%–30% mantle-derived helium (Fig. 2), assuming an 8 R_A source. The highest values are located west of the 2002 rupture at Railroad and Windy Creek springs (Fig. 2). Modest (8%–12%) mantle He contributions are observed at Soda Creek springs (0.65–0.99 R_C/R_A), located near the eastern terminus of the 2002 rupture on the Totschunda fault. A spring along the north-dipping (85°) Slate Creek splay normal fault (Rose, 1967), which did not rupture in 2002, yields ~16% mantle-derived helium (1.30 R_C/R_A). A spring along the northdipping (<45°) McCallum Creek splay thrust fault (Waldien et al., 2018), which also did not rupture during the 2002 earthquake, has air-like ³He/⁴He of ~1 R/R_A. The three springs sampled along the main strand rupture zone also have He/Ne and ³He/⁴He ratios consistent with airsaturated meteoric recharge (Fig. 3).

The δ¹³C values of DIC in non-bubbling springs range from -9.2‰ to +0.1‰ (Fig. 2B), and at bubbling springs, the δ¹³C values for DIC and CO₂ range from -4.1‰ to +1.7‰ and -10.8‰ to -9.0‰, respectively (Table S4). The δ¹³C values from the bubbling springs are used to calculate “initial values” prior to near-surface phase separation and degassing of -10.8‰ to -7.9‰ (Supplemental Material; see ^{footnote 1}). The concentrations and isotopic values of “external carbon” (C_ext and δ¹³C_ext; Fig S2; Table S4) are calculated by removing contributions from water-rock interaction with marine carbonates (C_carb) (Supplemental Material; Chiodini et al., 2000; Hiett et al., 2022). C_carb in our analysis accounts for ~20%–50% of the dissolved carbon in the springs. External carbon is derived from sedimentary organic matter oxidation and endogenic sources (e.g., mantle or metamorphic fluids). Fingerprinting the endogenic carbon sources can be aided by considering ³He/⁴He with the δ¹³C_ext (Hiett et al., 2022). Springs along the main strand rupture and parallel splays likely derive their C_ext from sedimentary organic matter (e.g., shales) with additions of endogenic carbon with δ¹³C values between -10‰ and 0‰, overlapping with carbonate and deep crustal sources, although mantle carbon cannot be completely discounted (Fig. S2). The C_ext at Windy Creek and Railroad springs appears dominated by an ~–9.5‰ endogenic source (Fig. S2), and considering the ~30% mantle He at these locations, we interpret that a proportion of the C_ext is mantle derived. The C_ext at Soda Creek springs is consistent with metamorphic devolatilization of marine carbonates, with negligeable mantle carbon.

Helium isotope data from these springs are not available for prior to the 2002 Denali fault earthquake, but data patterns suggest that the rupture altered flow paths along the fault zone. Mantle He at springs west and east of the main earthquake rupture segment indicates that deep fluid flow paths are present at these locations (Figs. 2 and 4). Soda Creek springs along the Totschunda fault, near the eastern termination of the rupture, have >50% lower mantle He contributions compared to springs west of the rupture (~12% versus ~30%). After the 2002 earthquake, Soda Creek springs had low flow and were no longer bubbling or depositing travertine as reported before the earthquake (Richter et al., 1973). These results suggest that deep conduits are still present but flow has changed. Springs along the 2002 rupture segment of the Denali fault have noble gas concentrations and ³He/⁴He consistent with shallowly circulated airsaturated meteoric water. However, these springs have fossil travertine accumulations suggestive of deeper connections with carbonic fluids in the recent past. Springs along splay faults parallel to the ruptured segment provide additional hydrological insight. The spring on the Slate Creek normal fault, connecting with the main Denali strand at ~18 km depth (Rose, 1967), has ~16% mantle-derived He, while the spring along the McCallum Creek thrust fault, intersecting the main Denali strand at <10 km depth (Waldien et al., 2018), has air-like noble gases (Figs. 2 and 3). Based on these observations, we infer the Denali earthquake disrupted the hydrological system by locally increasing the upper-crustal permeability (<10 km) and flooding the system with meteoric water and/or by sealing deep volatile connections due to along-fault damage and displacement. These types of seismicity-induced changes have been documented at other major fault systems, including reductions in mantle helium contributions correlated with increased hot spring discharge along the North Anatolian fault zone (Turkey) after the 1999 >M7 earthquakes (Güleç et al., 2002; Doğan et al., 2009). We observe no correlation between ³He/⁴He in springs and Pleistocene slip rates nor longerterm (10⁷ yr) along-strike variations in convergence on the Denali fault (e.g., Regan et al., 2021), suggesting that earthquake rupturing may be a stronger factor than strain rates in controlling mantle fluid connections.

The highest mantle He signatures along our transect are derived from springs in the Yakutat flat-slab segment and linked Denali volcanic gap (Fig. 1), where arc magmatism is absent due to limited melt generation (Chuang et al., 2017), discounting a magmatic ³He source. The presence of mantle-derived He in springs along the Denali, Hines Creek (Kansas Creek segment), and Totschunda faults, in the absence of nearby volcanism, indicates mantle-to-surface connections, consistent with present-day geophysical evidence that these structures penetrate the mantle (Allam et al., 2017) and are lithospheric boundaries (Gama et al., 2022). Without magmatism, the movement of mantle He into the crust in this setting requires volatile flux from the asthenosphere (Kennedy et al., 1997) or lithospheric mantle contributions via metamorphic reactions driving volatile releases (Hiett et al., 2021). This implies that the Denali fault taps actively devolatilizing mantle (Fig. 4), similar to other transform faults and major intracontinental strike slip faults like the San Andreas fault (California, USA), North Anatolian fault zone, Dead Sea transform, and Karakorum fault (Himalaya region) (Fig. 3) (e.g., Kennedy et al., 1997; de Leeuw et al., 2010; Klemperer et al., 2013; Torfstein et al., 2013).

Based on the (1) transcurrent nature of the Queen Charlotte–Fairweather–Connector–Totschunda–Denali fault system (Spotila and Berger, 2010), (2) lithospheric-scale nature of the Totschunda and Denali faults, (3) potential continuation of the Denali fault offshore toward the western Aleutian Trench (Vayavur, 2017), (4) difference in motion of south-central Alaska with respect to the North American plate (Elliott and Freymueller, 2020), and (5) extrusion of south-central Alaska toward the Aleutian Trench (Redfield et al., 2007), we suggest the Denali fault has characteristics of a continental transform. With the Aleutian Trench separating south-central Alaska from the Pacific plate, the tectonic setting is similar to the Anatolian plate delineated by the North Anatolian fault zone as the northern transform and Cyprean trench to the south.

The presence of mantle He in springs can be used to estimate the flow rate and flux of mantle volatiles along the Denali fault. One-dimensional estimates (Supplemental Material) based on radiogenic ⁴He ingrowth in the crust (after Kennedy et al., 1997; Kulongoski et al., 2013) suggest vertical mantle-derived fluid flow rates and ³He flux along the Denali fault may be as much as 160 mm yr^–1 and 7.4 × 10^-16 mol cm^–2 yr^–1. This requires mechanisms to move fluids in extremely low-permeability rocks via high fluid pressures through vertical conduits, with implications for fault strength and seismicity, similar to interpretations along the San Andreas fault (Kennedy et al., 1997). Denali fault mantle fluid flow rates and ³He flux fall in the range calculated for other strike-slip faults such as the North Anatolian fault zone (9–55 mm yr^–1; ~4 × 10^-15 mol cm^–2 yr^–1), San Andreas fault (2–147 mm yr^–1; ~2 × 10^-15 mol cm^–2 yr^–1), and Karakorum fault (19 mm yr^–1) (Kennedy et al., 1997; Doğan et al., 2009; de Leeuw et al., 2010; Klemperer et al., 2013; Kulongoski et al., 2013). Notably, Denali fault ³He/⁴He values and relative contribution between air, crust, and mantle volatiles are similar to those observed at other continental strike-slip and transform faults (Fig. 3). Efficient mantle volatile transit to the surface is evident at arc volcanoes, exemplified by as much as 100% mantle helium at Aleutian arc volcanoes. Apparently, at continental strike-slip fault settings, there is a maximum mantle He contribution of 30%–50%, suggesting a mantle-to-surface volatile “speed limit” in the absence of magmatism.

Funding was provided by U.S. National Science Foundation Early-Concept Grants for Exploratory Research (EAGER) to Newell and Regan. We thank Denali and Wrangell–St. Elias National Parks (permit numbers DENA-2021-SCI-0010; WRST-00242) and Ahtna Inc. for access, and Alaska Land Explorations for helicopter support. We thank Karen Fischer and an anonymous reviewer for constructive feedback.