Sample sites and collection methods:
Seawater samples were collected from cruises R/V Roger Revelle RR1804 and RR1805 from April to May 2018 and R/V Kilo Moana 1919 and 1920 in September 2019. Samples above 750 meters from both cruises were collected in 5-liter (L) acid-cleaned Teflon-coated external-spring "Niskin-type" bottles (Ocean Test Equipment) on a powder-coated trace metal clean rosette (Sea-Bird Electronics). After the recovery of the rosette, the Niskin-like bottles were carried into a laminar-flow clean van, pressurized with nitrogen, and filtered through either 0.2 µm AcroPak® capsule filters or 0.4µm Nuclepore® membrane filters. Samples deeper than 1000 m were collected by regular internal spring closure Niskin bottles, and filtered through 0.2 µm AcroPak® capsule filters. Samples between 300 m to 500 m were collected in both sampling systems and demonstrated no contamination from regular Niskin bottles on both cruises.

The locations of the stations are summarized in Table S1 and Fig. 1 of Huang et al, 2021. Station 23 is off the west coast of Baja California Mexico at the northern fringe of the ODZ, where its oxygen level is never below 2 μM. Station P3 is located at the mouth of the Gulf of California which is controlled by complex and dynamic currents and mesoscale eddies (i.e. Lavin et al., 2003). Station P1 is an onshore station near the west coast of Mexico. The "coring station" is on the continental margin slope, with fully anoxic bottom waters down to the bottom at 460 m. Station 15 is an offshore station near the center of the ETNP ODZ.

Cr Purification and Analysis:
The total dissolved chromium (Cr) and Cr(III) isotope analysis procedures were described in detail in Moos and Boyle (2019) and Huang, et al. (2021), respectively. Briefly, seawater samples were acidified with concentrated HCl to pH ~1.9 after being shipped back to MIT. Lowering the pH reduces Cr(VI) to Cr(III). To accelerate this process, samples were put in a 60°C oven for 7 to 14 days. After that, a 50Cr-54Cr double spike (DS) was added to samples (0.5~1L) in 1 L separatory funnels with a roughly 1:1 double spike/sample Cr ratio. This mixture was put on a shaker table overnight for sample Cr(III) and double spike Cr(III) to reach equilibrium. Cr was then pre-concentrated using a Mg(OH)2 co-precipitation method by adding concentrated NH3⋅H2O. Mg(OH)2 pellets were dissolved in HCl, and the pH of the solution was adjusted to reach a [H+] of 0.02M for column chromatography.

Cr(III) isotope analysis was conducted on frozen samples. The frozen samples were stored in either -80°C or -20°C freezers on the ship and in the lab to preserve Cr redox species. They were taken out of the freezers to thaw at room temperature on a shaker table, minimizing the time for thawing. Immediately after complete thawing, 2 nanomoles (nmol) of the double spike were added into the sample bottles and allowed for sample-double spike equilibrium before transferring the samples to separatory funnels and Mg(OH)2 co-precipitation. Spiked samples were equilibrated for 0.5 - 12 hours. We have shown that this range of equilibration time gives identical [Cr(III)] and δ53Cr(III) (Huang et al., 2021). The procedure afterwards was the same as the total dissolved Cr analysis method above.

Cr(VI) isotope ratios were measured from seawater samples where their Cr(III) had been extracted by Mg(OH)2 co-precipitation. The Cr(III)-extracted seawater samples were filtered through clean 0.4μm Nuclepore® filter membranes to remove any Mg(OH)2 precipitate. They were then acidified to pH 1.9 and put in a 60°C oven for 14 days to accelerate the reduction of Cr(VI) to Cr(III). They can then be treated as acidified seawater following the same procedure for the total dissolved Cr analysis.

Three-step ion chromatography was conducted to remove all matrix and isobaric ions (Moos and Boyle, 2019). All three chromatography columns use AG1X8 anion exchange resin. The first column is designed to remove major seawater matrix cations. This is done by oxidizing sample Cr(III) to Cr(VI) with ammonium persulfate (APS, (NH4)2S2O8) (1 hour, 110°C) before loading it onto the first column, and following elutation of matrix cations, reducing it back to Cr(III) while retained on the resin by eluting with 2M HNO3 + 2% (v/v) H2O2. This process releases the Cr(III) from the anion resin. The second column is aimed at removing polyatomic sulfur interferences, which come from both the seawater itself and the degradation from the APS in the previous step. The third mini-column was to remove traces of Fe. This is achieved by dissolving the dried-down samples in 6M HCl. In the strong acidic medium, Fe would complex with Cl- to form FeCl4-, which has a high affinity to the anion exchange resin, whereas Cr(III) would pass through. The second and the third columns were each followed by digestion with 100µL aqua regia to decompose resin residues. The final samples were dissolved in 2% (v/v) HNO3 to be analyzed on a multi-collector inductively-coupled-plasma mass spectrometer (MC-ICP-MS).

Blanks and reproducibility:
One quality control seawater sample from the surface Atlantic Ocean has been measured repeatedly with each batch of samples since November 2017. The external reproducibility of δ53Cr and [Cr] are 1.02 ± 0.10‰ (2SD, n=29) and 3.19 ± 0.12 (SD, n=29), respectively. The external reproducibility of low-Cr (i.e. Cr(III) or Cr(VI)) analysis was also determined using the same quality control seawater with a sample Cr to double spike ratio of 0.3. The resulting reproducibility of the δ53Cr and [Cr] of low-Cr samples are 1.02 ± 0.15‰ (2SD, n=9) and 3.24 ± 0.06 nmol/kg (SD, n=9), respectively.

Procedural Cr blanks were on the order of 0.02 nmol, which is comparable to that in Moos and Boyle (2019). Considering a typical Cr amount for total dissolved Cr analysis (2 nmol), the blank is negligible. Therefore, total dissolved Cr data was not corrected for the blank. As for Cr(III) and Cr(VI) analyses, sample Cr varies from 0.44 to 1.66 nmol. The blank accounts for less than 5% of the sample Cr. No correction was made on Cr(III) or Cr(VI) isotope data, either.

Instruments:
All Cr isotope measurements were made on an IsoProbe MC-ICP-MS. A ‘peak-jump’ mode was applied in the Cr isotope analysis on IsoProbe MC-ICP-MS. The plasma mass spectrometer was tuned on an 100 mM NH4S2O8solution with 1 µM Cr to minimize polyatomic sulfur interferences (i.e. 32S16O1H+ on mass 49, 34S16O+ on mass 50 and 34S16O1H+ on mass 51) by lowering the signal of mass 49 (polyatomic) relative to mass 52 (Cr). Each sample was bracketed by two SRM-DS mixtures with the same sample Cr to DS ratio and similar signal level (within 10%). The δ53Cr data was calculated by iteration on each isotope correction and instrumental mass fractionation (β) in an Excel spreadsheet (Moos and Boyle, 2019). The [Cr] concentration was calculated by averaging the results from the single isotope dilution formula using corrected 50Cr and 54Cr.