INTRODUCTION

The Gulf of California hosts the trans-tensional domain of the Pacific-North America plate boundary (^{Fletcher et al., 2014}). This boundary is defined by lengthwise dominant transform faults that connect an en-echelon array of pull-apart basins at variable stages of crustal separation and sea-floor spreading (Figure 1). The maturity of the pull-apart basins and their interconnecting transform faults decreases north of the Guaymas basin where the Guaymas transform fault splays into several strike slip faults that shear continental crust (^{Lonsdale, 1989}). The geological transition from mature oceanic spreading centers and transform faults in the southern Gulf of California to typically much smaller basins in the northern Gulf of California coincides with major oceanographic and faunal changes in the region of the so-called Midriff Islands which include Isla Ángel de la Guarda, San Esteban, and Tiburón (Figure 1).

Figure 1 

Within the Ballenas channel, an ~80 km long submarine continental transform fault system separates the Baja California peninsula from Isla Ángel de la Guarda. The Ballenas Transform fault (BTF) juxtaposes continental crustal blocks of similar structure and composition along most of its length: on the east side, the continental block of Isla Ángel de la Guarda has become part of the North American plate, whereas the Baja California peninsula moves with the Pacific plate, with both blocks passing each other at a velocity of 47.3 ± 0.8 mm/yr (^{Plattner et al., 2015}). It further connects two young rift basins, the Lower Delfín basin to the north, and the more diffuse Ballenas and Salsipuedes basins in the south (Figure 1). The length of new oceanic crust is < 20 km for the Lower Delfín basin, where new crust is inferred to occupy a narrow rift with several submarine volcanoes located in the axial trough and over the faulted margins (^{Persaud et al., 2003}). The extent of new oceanic crust is likely smaller in the Salsipuedes basin (Figure 1), although multibeam bathymetry also depicts volcanoes in the southern Salsipuedes and the centrally located Ballenas basins (^{Plattner et al., 2015}). The seamounts in the Ballenas basin form a north-northeast alignment of small volcanoes that projects south onto Isla Coronado (ICO), locally also referred to as Isla Smith (Figure 2). This island is a 300 m-high dacite dome offshore the Main Gulf Escarpment where continental crustal rocks are down-faulted towards the basin. Another volcanic field occurs in the west central part of Isla Ángel de la Guarda, which shows a more pristine morphology compared to the regional volcanic deposits of Miocene age (Figure 2). This newly defined Lobera Volcanic Complex (LVC) is named after the eponymous seal colony (lobo marino) that permanently occupies the cliffy shore of the LVC.

Figure 2 Bathymetry and structural map of the Ballenas basin. Sample locations (red stars) and submarine volcanic mounts (red circles) reported by ^{Plattner et al (2015)} are shown. The Lobera volcanic complex (LVC) and Isla Coronado are Pleistocene volcanic events at the margins of the sheared continental crust. 

To better understand the timing and origin of magmas in sheared and thinned transitional continental crust, we investigate volcanic rocks from ICO and the LVC on Isla Ángel de la Guarda. Their youthful, but not quite pristine morphology (cf. Isla San Luis; ^{Paz-Moreno and Demant, 1999}), is consistent with Pleistocene activity, but no radiometric age constraints have been reported so far. In this paper we present U-Pb zircon and Ar-Ar whole-rock geochronologic data to constrain magmatic differentiation and eruption ages, and report reconnaissance major and trace element geochemistry as well as Nd-Sr isotopic compositions of lavas to evaluate the source and composition of parental magmas in this important transitional part of Gulf of California rift zone.

TECTONIC AND GEOLOGIC FRAMEWORK

The BTF accommodates nearly all of the Pacific-North America plate motion in the northern Gulf of California, with less than ~10 % of plate motion occurring along the Tosco-Abreojos fault in the Pacific margin of Baja California (^{Plattner et al., 2007}). It connects two proto-oceanic pull-apart basins, the Lower Delfín in the north and the Ballenas-Salsipuedes basins located in southeast end of the BTF (Figure 1). This transform fault developed in the late Pliocene when the Tiburón continental transform fault became inactive and the main plate boundary shifted west into the BTF, resulting in the capture of the continental block of Isla Ángel de la Guarda (^{Aragón-Arreola and Martín-Barajas, 2007}; Martín-Barajas et al., 2013).

Low-temperature thermochronology of crystalline rocks onshore the Ballenas transform margin in Baja California suggests that dextral shearing and uplift of the transform margin in Baja California occurred by ~1.8 Ma (^{Seiler et al., 2009}). Furthermore, correlation of distinctive volcanic deposits across the Ballenas Channel constrains dextral offset in the BTF between ~100 and 130 km (^{Stock et al., 2009}). This displacement is broadly consistent with the ~90 km length separation of continental crust in both the Upper and Lower Delfín basins based on interpretation of industry seismic profiles (Martín-Barajas et al., 2013; ^{Téllez-Velázquez, 2018}).

Extension in the Lower and Upper Delfín basins is transferred to dextral shear in the BTF and has detached the continental block of Isla Ángel de la Guarda from the Baja California peninsula. Although the BTF concentrates most of the North America - Baja California relative motion, the capture of Isla Ángel de la Guarda is probably incomplete as Quaternary deformation in the south part of Isla Ángel de la Guarda produces a north-northeast trend of normal and oblique faults (^{Higa et al., 2022}), and significant historic seismicity is reported east of the BTF (^{Castro et al., 2017}).

Magmatism in this young rift-transform system is ubiquitous, and submarine volcanoes are identified in high-resolution seismic profiles (^{Persaud et al., 2003}; ^{Hurtado-Brito, 2012}), and sea-beam bathymetry (^{Plattner et al., 2015}). Some of these submarine volcanoes are rhyolite and andesite pumice cones (Martín-Barajas et al., 2008; ^{Henyey and Bishop, 1973}), and a seamount reported by ^{Plattner et al. (2015)} in the Ballenas basin is composed of basaltic lava (P. Lonsdale written comm.) (Figure 2). The ICO and LVC volcanoes may represent magmatic activity at the intersection between rift basins and transform faults under the influence of adjacent continental crust that is being actively sheared along the transform boundaries.

GEOCHEMISTRY AND GEOCHRONOLOGY METHODS

RESULTS

Lobera Volcanic Complex (LVC)

We conducted field reconnaissance and satellite spectral analysis on EO-1 data and Google Earth images of the LVC in the western-central part of Isla Ángel de la Guarda (Figure 3). LVC is a composite lava dome complex with three main lava units, the lower lava unit (Qv1) is exposed at the northwest cliffy shore (sample IAG07-38; supplementary Figure S2), but its vent location is obscured. The eroded cliff indicates this lava flow is at least 20-30 m-thick and displays ramp structures and columnar jointing. The top of Qv1 is brecciated and overlain by whitish pyroclastic deposits ranging a few meters in thickness (Qt in Figure 3). Pumice lapilli in this deposit are reworked towards the top and grade upwards into a ~15-20 m-thick conglomerate covered by the uppermost lava flow unit (Qv3).

Figure 3 Google Earth image of the Lobera Volcanic Complex (LVC) with major map units and sample locations. Map units were defined from satellite spectral images (see Figure S1 in the Supplementary Material), and field reconnaissance. 

The two younger lava units Qv2 and Qv3 crop out along the S-SE and NE flanks of LVC (Figure 3). The largest and morphologically youngest flow is a ~2 km long coulee, displaying arcuate pressure-ridges that indicate E-SE directed transport (Figure 3). An isolated lobe in the NE corner of the volcanic complex displays a circular ridge crest nearly 200 m in diameter (Qv2), which suggests a local vent or alternatively, this lobe could be part of a larger lava flow of intermediate age between the lower and upper lava units.

All three LVC lava units have similar mineral composition although they display textural differences. The lower flow (Qv1) is composed of a dark grey, aphanitic glassy lava, with <1 % plagioclase and pyroxene microphenocryst <1.5 mm long (see photomicrographs in the Supplementary Material). The plagioclase microlites are commonly aligned within the glassy matrix, and glomerophyric plagioclase and pyroxene are common. Lava units Qv2 and Qv3 are dark grey to black with an aphanitic to microporphyritic texture. Plagioclase microlites within the glassy matrix define a hypohyaline flow texture. In this matrix, agglomerates of plagioclase (andesine)-orthopyroxene phenocrysts and opaque minerals along with zoned plagioclase microphenocrysts are common.

Isla Coronado (ICO)

The ICO comprises a ~300 meter-high dacite dome located in the north end of the island (Figure 4), situated north of Bahía de los Ángeles (BLA in Figure 1b) This lava dome overlies a low relief crystalline basement composed of Mesozoic granitic rocks and hornblende schist of presumably Paleozoic age. The lava dome includes two cooling units (ICO08-1 and ICO08-2) separated by a flow breccia, which also comprises most of the small islet near the northern tip of ICO (e.g., sample ICO08-3).

Figure 4 Google Earth image of Isla Coronado (Smith). The dacite dome is ±300 m high and erupted over granitic and metamorphic continental crust. 

The ICO lavas are dominantly hypohyaline to felsitic in texture with common glomerophyric plagioclase, pyroxene, and rare hornblende. Glomerophyric plagioclase is commonly euhedral and up to 1.5 mm in diameter; it displays oscillatory zonation and complex twinning. Plagioclase microlites are euhedral to subhedral and commonly associated with disseminated orthopyroxene and opaque minerals.

Major oxide and trace elements

Major oxide geochemistry in both ICO and LVC indicates dacitic composition of lavas (Figure 5a), although ICO samples are high-silica dacite compared to LVC lavas. The higher differentiation of ICO lavas is also reflected in lower content of MgO, TiO₂ and FeOt, and higher K₂O (Table 1). Rare Earth elements display similar patterns in both ICO and LVC samples (Figure 5b), with a slightly higher LREE fractionation and Eu depletion in ICO lavas (Lan/Ybn 3.75-3.91) compared to LVC dacite (Lan/Ybn 3.20-3.39).

Figure 5 a) Total alkalis vs silica plot (from ^{Le Bas et al., 1986}) with ICO and LVC samples analyzed for major oxide and trace elements geochemistry. b) Chondrite-normalized REE diagrams of selected samples from LVC (Isla Ángel de la Guarda) and Isla Coronado (ICO). 

Geochronology and zircon trace elements

Two U-Pb and one Ar-Ar ages were determined for lava flows from the LVC. The Ar-Ar whole-rock analysis obtained from one sample of Qv2 lava of LVC comprises two experiments that yielded overlapping ages for the high-temperature release steps which average 692 ± 164 ka (n = 10; MSWD = 1.0) (Figure 6).

Figure 6 Ar-Ar stepwise heating diagram for two runs of a whole-rock sample IAG07-40 from Qv2 in the northeastern lava of the LVC. Plateau age for the high temperature steps is indicated using steps 2-6 for 1st experiment. 

Due to large error obtained in the Ar-Ar whole rock analysis we processed a larger amount of rock in search of young zircon crystals. In spite of the low zircon yield, we obtained U-Pb crystallization ages from Qv1 and Qv3. The stratigraphically lowermost sample IAG07-39 (Qv1) yielded 9 zircon crystals with mostly mid-Miocene ages and one Cretaceous age (Figure 7a). Only two crystals (analyzed in triplicate) yielded a homogeneously Quaternary date with an error-weighted average of 1016 ± 60 ka (n = 6; MSWD = 1.50). The overlying lava flow Qv3 yielded 17 zircon crystals, but only one of Quaternary age (Figure 7b). This crystal was analyzed in triplicate, and a consistent error-weighted average of 358 ± 60 ka (n = 3; MSWD = 1.26) was obtained. Xenocrysts range in age from ca. 6.9 to 1013 Ma, with clusters in the Miocene (n = 3) and the Cretaceous (n = 6).

Figure 7 U-Pb and U-Th zircon geochronology results. U-Pb data are plotted uncorrected for common Pb with a linear regression fixed at common 207Pb/206Pb = 0.828 where the radiogenic component and age is determined from the lower Concordia intercept. U-Th data are shown in an equiline diagram where isotopic activities are indicated by parentheses. The age is determined from the slope of the isochron defined by zircon with young U-Pb ages. Open symbols are omitted from regression. 

About 20 zircon crystals were separated for ICO dacite lavas, with 10 crystals yielding Quaternary ages (Figure 7c). The concordia intercept age of 247 ± 14 ka (n = 10; MSWD = 0.67) suggested a sufficiently young age for disequilibrium dating (230Th)/(238U), and therefore the same crystals (along with xenocrysts for reference) were re-analyzed. An isochron age of 236 +156/-62 ka (n = 10; MSWD = 2.50) confirmed the 206Pb/238U concordia intercept age; because of the larger uncertainties of the U-Th age, however, we refer to the more precise U-Pb age in the subsequent discussion.

Trace elements in young zircons indicate a continental affinity (Figure 8). Quaternary zircon crystals from LVC have moderate U abundances (180-520 ppm) and U/Yb (0.19-0.48) vs. Hf (8090-9630 ppm) that fall within the field of continental zircon. Similarly, ICO zircon trace elements display elevated U abundances (810-2890 ppm) and U/Yb (0.94-2.1) at Hf abundances that mostly overlap those of LVC zircon (5180-9020 ppm) and that also fall within the continental field in zircon trace element discrimination diagrams (Figure 8).

Figure 8 Zircon trace element classification diagram after ^{Grimes et al. (2007)}. Data for northern Gulf zircons from ^{Schmitt et al. (2013)} shown for comparison. 

DISCUSSION

Radiometric ages for LVC and ICO reveal Pleistocene evolved magmatism along the onshore margins of the Ballenas basin. Activity in the LVC appears to be older, with zircon ages (representing maximum eruption ages) in the lower Pleistocene. A stratigraphically consistent younger age was determined for zircon from the upper lava flow (Qv3), which is concluded to be of mid Pleistocene age. The lava flow Qv2 located in the northeastern part of the complex is intermediate in age between both lava flows dated using zircon separates, and yielded a 40Ar/39Ar whole rock age of 692 ± 164 ky (Figure 6) that is consistent with its intermediate stratigraphic position in the LVC.

Zircon ages for ICO, again interpreted as maximum eruption ages, postdate LVC zircon, and suggest activity at or after ca. 250 ka (Figure 7c and 7d), also during the mid-Pleistocene.

To further elucidate the age significance of zircon in these lavas, zircon trace element geochemistry is compared to whole rock compositions via published zircon-melt partition coefficients (table 5 in ^{Sano et al., 2002}) (Figure 9). Because zircon saturates early in metaluminous melts, and the investigated rocks are crystal-poor, melt compositions may be reasonably well approximated by whole-rock compositional data. Thus, the comparison between predicted melt compositions (using measured zircon REE abundances divided by zircon-melt partition coefficients that range from 0.00046 for La to 325 for Lu, ^{Sano et al., 2002}) and actual whole rock data is a first-order assessment whether zircon is in equilibrium with its host rock. Although the calculated range of REE from LVC zircon is wider than observed whole rock compositions (Figure 9b), they broadly overlap, some deviations for La and Ce notwithstanding. Both elements, however, are difficult to model as La is only present in very small amount in zircon, and Ce partitioning is redox dependent so that the values from ^{Sano et al. (2002)} may not be adequate for LVC zircon-melt equilibria. Nonetheless, the REE models of Figure 9b suggest that zircon from LVC crystallized from a melt similar to the whole rock compositions. Zircon genetic classification diagrams (^{Grimes et al., 2007}) also indicate a larger crustal input into LVC zircon-crystallizing melts compared to those for Pleistocene volcanoes in the northern Gulf of California (Figure 8) (^{Schmitt et al., 2013}).

Figure 9 Chondrite-normalized rare earth element (REE) diagrams for (a) zircon and (b) whole rock along with calculated melt compositions that are expected to be in equilibrium with zircon. Calculations use literature zircon-melt partitioning coefficients (^{Sano et al., 2002}). With the exception of La (difficult to analyze due to extremely low abundances in zircon) and Ce (with a redox-sensitive partitioning coefficient), calculated melts agree with observed whole rock compositions for LVC, whereas ICO zircon trace elements reveal crystallization from much more evolved melts than indicated by whole rock compositions. 

In contrast to LVC zircon, trace element abundances in ICO zircon are higher, suggesting that they crystallized from a more evolved melt (Figure 9a). This is also underscored by a pronounced negative Eu anomaly displayed by ICO zircon. Calculated melt compositions in ICO samples (see above) are significantly higher in REE (except Eu) than REE measured in whole rock samples (Figure 9b), again disregarding La and Ce for the reasons stated above. The mismatch between modelled and actual whole rock (melt) compositions implies that ICO zircon crystallized from a significantly more evolved melt than the dacitic host lava. In the light of the higher U/Yb of ICO zircon compared to LVC (e.g. Figure 8), a stronger crustal input into the zircon-crystallizing melt is also invoked for ICO. This is supported by whole rock radiogenic isotope data (Table 2) discussed below.

Magmatism associated with the Ballenas continental transform faults may results from decompression melting of the upper mantle in regions where oblique rifting generates trans-tensional domains at right steepened strike-slip faults. Similarly, volcanic products in the western margin of the Cerro Prieto and Wagner pull-apart basin are commonly evolved, reflecting protracted ascent and residence times within continental crust and sediments, and differentiation of basaltic melts occurring via crystal fractionation and/or crustal assimilation (^{Schmitt et al., 2013}; ^{Herzig and Jacobs, 1994}). In this work assimilation-fractional crystallization (AFC) models were set up for a consistency check for a magma genetic scenario where mid ocean ridge basalt (MORB) as a parental magma interacted with continental basement (Figure 10). Calculations were based on simple binary mixing, the classical AFC algorithm in ^{De Paolo (1981)}, and energy-constrained (EC) AFC models (^{Spera and Bohrson, 2001}). More sophisticated models accounting for recharge and temperature-dependent changes in the crystallizing mineral assemblage and partitioning coefficients (cf. ^{Heinonen et al., 2021}) were not attempted as these would require additional constraints that are currently unavailable. The parental MORB composition was taken from submarine basalts of the Alarcón Basin in the southern Gulf of California (^{Castillo et al., 2002}). Crustal assimilate as indicated from regional geological mapping include Tertiary volcanic rocks that are abundant on the surface, Mesozoic tonalites belonging to the eastern Peninsular Ranges Batholith suite, as well as metamorphic rocks including hornblende schists.

Figure 10 Isotope variation diagram for whole rock 144Nd/143Nd (εNd) vs. 87Sr/86Sr compositions. Modelled mixing and assimilation-fractional crystallization (AFC) curves are indicated for parental MORB (Alarcón; εNd = +9.8; 87Sr/86Sr = 0.702476; ^{Castillo et al., 2002}) and crustal contaminant typical of eastern Peninsular Ranges Batholith (εNd = -8.5; 87Sr/86Sr = 0.709; ^{Herzig and Jacobs, 1994}). Classical AFC curves are calculated for r (rate assimilation/crystallization) = 0.5 and bulk mineral-melt partitioning coefficients D(Nd) = 0.28 (Nd magma = 6.74 ppm; Nd assimilate 25.6 ppm) and D(Sr) = 1.6. (Sr magma = 203 ppm; Sr assimilate 289 ppm). Energy-constrained (EC) AFC curves use parameters as above and thermal conditions similar to ^{Spera and Bohrson (2001)} with an initial (liquidus) magma temperature of 1100°C. The assimilant initial and liquidus temperature were set at 600°C and 1100°C, respectively. Magma and assimilant heat capacities (enthalpies of fusion) were 1484 J∙kg-1∙K-1 (390 kJ∙kg-1) and 1370 J∙kg-1∙K-1 (270 kJ∙kg-1), respectively. All wall rock melt was mixed into the magma. Data from Lower Delfín andesite and rhyolite pumice from Martín-Barajas et al. (2008), data from Roca Consag is from ^{Schmitt et al. (2013)}. 

Pre-Quaternary zircon encountered in LVC samples is dominated by Tertiary ages (48 %), with Cretaceous ages between 105 and 90 Ma being the second largest population (30 %). In ICO samples, Cretaceous ages between 113 and 98 Ma were also found, but the remainder is Proterozoic with the largest population of Proterozoic xenocryst zircon dated around ~1100 Ma (Figure 7c). In the absence of direct constraints on crustal compositions, we thus use eastern Peninsular Ranges Batholith Nd-Sr compositions as a crustal contaminant because (1) it matches a significant proportion of detrital ages, (2) such rocks presumably constitute a significant portion of the mid to lower continental crust of the Baja California peninsula, and they are more likely to become assimilated than upper crustal Tertiary volcanic rocks, and (3) the eastern Peninsular Ranges Batholith is itself a mantle-crustal hybrid which sampled and integrated pre-batholithic basement for which there are no outcrops in the direct vicinity.

Model curves for mixing and AFC scenarios in the åNd vs. 87Sr/86Sr space plot closely to the ICO data (Figure 10), and also match compositions for lava from Roca Consag, a volcanic island in the Wagner basin of the northernmost Gulf of California (^{Schmitt et al., 2013}). LVC data are identical in åNd to Roca Consag, but displaced to higher 87Sr/86Sr. Similar sub horizontal trends are displayed by submarine lavas from the Lower Delfín Basin (Martín-Barajas et al., 2008) and also known for volcanic rocks from geothermally active onshore basins of the northern Gulf of California rift (^{Schmitt et al., 2013}). They are interpreted as results of hydrothermal alteration, and therefore high-87Sr/86Sr data are considered unreliable as indicators of magmatic compositions. It is therefore concluded that binary mixing and/or AFC involving MORB magmas and tonalitic basement can explain the compositions of the Ballenas Channel evolved volcanic rocks. The amount of crustal material added to the erupted magma, as indicated by the different models, is at most ~10-20 %, within the range of other Pleistocene volcanic rocks from the northern Gulf of California magmatic province (^{Schmitt et al., 2013}; Figure 10). A slightly larger crustal input into the ICO source relative to LVC is consistent with findings from zircon and whole-rock geochemistry.

CONCLUSIONS

Two sets of volcanic vents have developed during the Pleistocene on opposite sides of the Ballenas transform fault within the Midriff section of the Gulf of California. These vents form Isla Coronado (ICO) at the southwestern terminus of the Ballenas basin, and the Lobera Volcanic Complex (LVC) on Isla Ángel de la Guarda on the northeastern end. Both volcanoes are located on the edge of sheared crystalline basement comprising tonalite and hornblende schists, which are overlain by Miocene volcanic rocks. This crustal segment has been previously thinned in dilatational domains of the transform fault system such as the Ballenas basin, which comprises several submarine volcanic knolls (^{Plattner et al., 2015}). Zircon crystallization ages of the ICO and LVC volcanoes range between ca. 250 and 1000 ka. Exact eruption ages are unknown, except for a ca. 700 ka 40Ar/39Ar age for a lava unit (Qv2) in the northeastern part of the LVC, but younger than the zircon crystallization ages. The young morphology of unit Qv3 in the southeastern LVC is consistent with the youngest zircon ages detected for the LVC of ca. 360 ka. The ICO eruption postdates ca. 250 ka as indicated by zircon crystallization ages. Radiogenic (Nd, Sr) isotopes suggest a MORB-type primary magma with significant (~10-20 %) crustal assimilation involving tonalitic basement of the eastern Peninsular Ranges Batholith.