A method for resolving changes in atmospheric He /N2 as an indicator of fossil fuel extraction and stratospheric circulation

The atmospheric He/N2 ratio is expected to increase due to the emission of He associated with fossil fuels and is expected to also vary in both space and time due to gravitational separation in the stratosphere. These signals may be useful indicators of fossil fuel exploitation and variability in stratospheric circulation, but direct measurements of He/N2 ratio are lacking on all timescales. Here we present a high-precision custom inlet system for mass spectrometers that continuously stabilizes the flow of gas during sample– standard comparison and removes all non-noble gases from the gas stream. This enables unprecedented accuracy in measurement of relative changes in the helium mole fraction, which can be directly related to the He/N2 ratio using supplementary measurements of O2/N2, Ar/N2 and CO2. Repeat measurements of the same combination of high-pressure tanks using our inlet system achieves a He/N2 reproducibility of ∼ 10 per meg (i.e., 0.001 %) in 6–8 h analyses. This compares to interannual changes of gravitational enrichment at ∼ 35 km in the midlatitude stratosphere of order 300–400 per meg and an annual tropospheric increase from human fossil fuel activity of less than ∼ 30 per meg yr−1 (bounded by previous work on helium isotopes). The gettering and flow-stabilizing inlet may also be used for the analysis of other noble-gas isotopes and could resolve previously unobserved seasonal cycles in Kr/N2 and Xe/N2.


Introduction
The atmospheric mole fraction of helium in dry air is typical ~5.24 ppm (Glückauf, 1944) with an isotopic abundance of 4 He 20 about 10 6 times greater than 3 He. On geological time scales, the natural concentration of 4 He in the atmosphere is set by a balance of 4 He loss to space and 4 He release from the Earth's crust, where it is produced by radioactive decay of uranium and thorium (Kockarts, 1973;Pierson-Wickmann et al., 2001;Sano et al., 2013;Torgersen, 1989;Zartman et al., 1961). Over the past century, human exploitation of fossil fuels likely has accelerated the release of crustal He (Boucher et al., 2018c;Evans, 2013, 2004;Mabry et al., 2015;Oliver et al., 1984;Pierson-Wickmann et al., 2001;Sano et al., 1989), but direct 25 observations of a secular increase of atmospheric 4 He are lacking. Additionally, recent measurements and model simulations reveal a small depletion of heavy gas argon in the stratosphere by gravitational separation (Belikov et al., 2019;Birner et al., 2020;Ishidoya et al., 2020Ishidoya et al., , 2018Ishidoya et al., , 2013Ishidoya et al., , 2008Sugawara et al., 2018) suggesting a corresponding enrichment of the light gas helium. Gravitational separation is only partially counteracted by the large-scale stratospheric circulation and mixing, which tends to homogenize the atmosphere. Variability in stratospheric circulation and stratosphere-troposphere exchange (STE) 30 could therefore impact the degree of fractionation and cause additional interannual changes in the stratospheric and, to a much lesser extent, the tropospheric abundance of 4 He.
These gases all have significant sources or sinks in the stratosphere that cause strong stratosphere-troposphere concentration differences. Global circulation models consistently predict an acceleration of the stratospheric Brewer-Dobson Circulation (BDC;Brewer, 1949;Dobson, 1956) under global warming (Butchart, 2014). Stratospheric circulation is also naturally modulated on a range of shorter time scales from synoptic-scale events to decadal variations (e.g., Holton et al., 1995;Li et 40 al., 2012;Flury et al., 2013;Butchart, 2014;Ray et al., 2014). Circulation changes have typically been observed using measurements of numerous different trace gases in the stratosphere (e.g., CO2, SF6, H2O, O3, CO, or N2O) (e.g., Engel et al., 2009Engel et al., , 2017Ray et al., 2010;Haenel et al., 2015). However, interpretation of these tracers of stratospheric circulation is complicated by complex chemical sources, sinks and tropospheric histories, whereas gravitational fractionation of He/N2 is governed by comparatively simple physics and expected to increase smoothly in the troposphere. 45 Atmospheric He/N2 measurements may also provide an indication of the history of fossil-fuel usage. Previous attempts to measure the fossil-fuel signal in He have centered on measurements of changes in the atmospheric 3 He/ 4 He isotope ratio typically using multicollector, static vacuum mass spectrometers (Boucher et al., 2018c;Evans, 2013, 2004;Mabry et al., 2015;Oliver et al., 1984;Sano et al., 1989). However, measurements of the 3 He/ 4 He ratio are fundamentally limited by the extremely low abundance of 3 He (e.g., Mabry et al., 2015;Boucher et al., 2018b), with only about 1 in 730,000 He atoms 50 being 3 He. Therefore, the precision on individual 3 He/ 4 He analyses is currently limited to ~±0.2% (2σ). This is insufficient for the detection of the stratospheric and anthropogenic signals we are interested in, and which we estimate to cause variations in the atmospheric 4 He abundance on the order of 0.0030 to 0.04% y -1 (see section 2.1 & 2.2.). Moreover, small changes in 3 He from radioactive decay of tritium in nuclear warheads may complicate the interpretation of 3 He/ 4 He results (e.g., Boucher et al., 2018c;Lupton and Evans, 2004). 55 Thus far, the most promising direct measurements of the atmospheric 4 He mixing ratio were produced by Holland and Emerson (1987). Holland and Emerson repeatedly introduced sample and standard air into a mass spectrometer through a charcoal trap to concentrate helium. However, their method also only achieved an instrument precision of 0.22% (2σ) and is thus not suitable for the science discussed above.
Here we describe a method to measure relative differences in 4 He mole fraction ( 4 He/M) between two large samples of air 60 using a custom mass spectrometer inlet system. The helium mole fraction can later be mathematically translated to our target ratio, 4 He/N2, given supplementary measurements of O2, Ar, and CO2 (see discussion). This is advantageous because N2 is near-constant in the atmosphere making 4 He/N2 more readily interpretable than 4 He/M. The 4 He/M method depends on stabilization of the air flow to the ion source between a sample and standard gas to achieve high precision differencing. Novel elements in our setup include continuous flow removal of reactive gases via titanium gettering immediately upstream of the 65 mass spectrometer inlet, and the use of an actively-controlled open split (Henneberg et al., 1975) for balancing pressures upstream of a shared capillary directed towards the mass spectrometer. Gas handling techniques, the inlet system and the continuous-flow getter oven are described in detail below.

Gravitational fractionation of He/N2 in the stratosphere
The notion that the stratospheric and tropospheric He/N2 ratio must vary in response to fluctuations in stratospheric circulation 70 is based on studies of the atmospheric Ar/N2 ratio (Birner et al., 2020;Ishidoya et al., 2020). Relative changes in the Ar/N2 ratio (or He/N2) are commonly expressed in delta notation: where subscripts SA and ST refer to the ratio in a sample and a standard gas mixture, respectively. δ(Ar/N2) is multiplied by 10 6 and expressed in "per meg" units.
Sensitivity tests with the 2-D chemical-dynamical-radiative model of the atmosphere SOCRATES by Ishidoya et al. (2020) 75 indicate that significant temporal changes in stratospheric Ar/N2 should occur in response to an acceleration or deceleration of the BDC. The simulations also suggest a weak stratospheric influence on tropospheric Ar/N2. Ishidoya et al find that imposing a gradual acceleration of the BDC of 4% dec -1 leads to a 40 per meg dec -1 increase in δ(Ar/N2) at ~35 km altitude in northern mid-latitudes, and a corresponding 1.3 per meg dec -1 decrease of δ(Ar/N2) in the troposphere. Furthermore, they find that imposing 3-year periodic changes of 10% in BDC yields anomalies of ±25 and ±0.4 per meg in stratospheric and tropospheric 80 δ(Ar/N2), respectively. Tropospheric observations of δ(Ar/N2) by Ishidoya et al. (2020) would be consistent with larger STEinduced interannual changes of tropospheric Ar/N2. Variability of the BDC on the order of 10% or more on seasonal to decadal time scales is consistent with published estimates (Flury et al., 2013;Ray et al., 2014;Salby and Callaghan, 2006).
The atmospheric He/N2 ratio must be more strongly impacted by gravitational fractionation than Ar/N2 due to the larger mass difference and higher diffusivity of He than Ar, which brings He closer to gravitational equilibrium. The gravitational 85 fractionation effect on He/N2 can be scaled from Ar/N2 (Birner et al., 2020) using the molar mass difference to air Δ (Δ = − 0.02896 kg mol -1 ) and the molecular diffusivity of gas in air as: Using the Fuller method (Reid et al., 1987), is 3.6 (3.5) times greater than ( 2 ), and Δ is more than twice as large and opposite sign than Δ (Δ = −0.02496, Δ = 0.01102, Δ 2 = −0.0009466). This makes δ(He/N2) ~-7.5 times more strongly fractionated by gravity than δ(Ar/N2) in the stratosphere but in opposite direction. 90

Other controls on tropospheric He/N2
A variety of known natural processes influence tropospheric 4 He/N2 is summarized in Figure 1 and Table 1. Natural 4 He release from the Earth's crust is mediated by volcanism, ground water discharge and diffusive leakage. At the same time, helium is lost to space by thermal and non-thermal escape (Kockarts, 1973;Oliver et al., 1984;Pierson-Wickmann et al., 2001;Sano et al., 2013;Torgersen, 1989). Based on these natural fluxes and the total atmospheric burden, the atmospheric residence time of 95 4 He is estimated to be ~1 million years.
Over the few last centuries, He release from fossil-fuel extraction has dwarfed the natural release rates of 4 He by several orders of magnitude. Based on knowledge of fossil fuel usage and He content of the material (Table 1) the additional 4 He release rate is estimated to be of order 3 to 30×10 10 mol yr -1 (e.g., Oliver et al., 1984;Sano et al., 1989Sano et al., , 2013Pierson-Wickmann et al., 2001) implying that 3 He/ 4 He should be decreasing at rates between 35 and 350 per meg y -1 . However, in contrast to these 100 predictions and some earlier observations (Oliver et al., 1984;Pierson-Wickmann et al., 2001;Sano et al., 2010Sano et al., , 1989, no significant trend in atmospheric 3 He/ 4 He has been observed using archived air samples spanning from the beginning of the 20 th century to today. These more recent observations bound any trend in 3 He/ 4 He to within roughly ±30 per meg per year (2σ), suggesting similarly small increase rates in δ(He/N2) (Boucher et al., 2018c;Evans, 2013, 2004;Mabry et al., 2015). 105 He release from fossil-fuel extraction is also expected to impose an interhemispheric gradient in δ(He/N2). A rough upper bound can be estimated by assuming all fossil-fuel derived He emissions occur in the Northern Hemisphere and interhemispheric mixing of the atmosphere has a time scale of about one year. This would yield a north-south difference of 30 per meg, equal to the expected annual rise in δ(He/N2).
Seasonal and long-term ocean warming can cause small changes in He/N2, mainly due to the impact on N2. From observations 110 of δ(Ar/N2) (Keeling et al., 2004) and solubility data of Ar, He and N2 (Hamme and Emerson, 2004;Weiss, 1971), we estimate that the impact on He/N2 of air-sea exchanges is on the order of 0.16 per meg y -1 for the secular ocean warming trend and 3-9 per meg for seasonal heat exchanges. Therefore, the ratio of stratospheric signals to ocean warming is ~12 times greater for He/N2 than Ar/N2 and the effect of slow ocean warming is over two orders of magnitude smaller than the influence of fossil fuel exploitation. 115 The He/N2 ratio could also be impacted by processes changing atmospheric N2. However, the annual removal of 7.5x10 12 mol N2 y -1 by anthropogenic nitrogen fixation in agriculture, combustion, and industry is clearly negligible compared to the ~1.4x10 20 moles of N2 in the whole atmosphere (Fowler et al., 2013). Volcanic emissions of N2 are likewise negligible, on the order of 10 9 mol y -1 .

Methods 120
Our He/N2 analysis method relies on measuring the helium mole fraction relative difference between an air sample and a standard gas using a single collector for 4 He + on a magnetic sector mass spectrometer (MS). Crucially, whole dry air is pressurestabilized to a high level prior to gettering, so that the beam intensity ratio being measured is effectively the 4 He to air ratio.
Measurements of the He mole fraction difference can also be expressed similarly to Eq. (1) as δ( 4 He/M) where M is total moles. By applying small corrections for variations in O2/N2, Ar/N2, and CO2, the quantity δ( 4 He/M) is easily related to 125 δ( 4 He/N2).
The MS is interfaced to a custom inlet system with on-line gettering and active flow stabilization using an actively pressurecontrolled open split (Henneberg et al., 1975), as shown in Figure 2.

The Inlet system
The design of the inlet system incorporates elements of an open split (Henneberg et al., 1975) but further stabilizes the pressure 130 using active control elements and allows active switching between a sample (SA) and reference standard gas stream (ST). Fig. 2) alternately slide 0.3-mm tubes exhausting sample or standard gas close to a shared intake capillary which is placed at the end of the stabilization chamber (C in Fig. 2) and connects the chamber to the getter oven and MS. Air actuation of the pistons is controlled by the MS through an electronic valve assembly (Clippard, model:

Pneumatically-actuated pistons (B in
EMS-08). The flexible 0.3mm tubes are mounted leak-tight inside sturdier 1/16" OD tubing which is fixed to the piston and 135 moved with a stroke length of 7 cm. A sliding seal is made between 1/8" OD outer tubing and the 1/16" OD tubing using a compressed O-ring lubricated with TorrLube vacuum grease. This setup creates a movable feedthrough port for the 0.3-mm tubes containing sample and standard gas and the pressure stabilization chamber, thus allowing the chamber to be operated at a selected pressure above or below ambient. The default setting is 14 psia (96.5 kPa).The chamber is shaped as a funnel to guide the sliding tubing into a reproducible resting position. Variations in chamber pressure are measured with a 0.2-Torr MKS 140 223B differential pressure gauge and are limited to better than 1 part in 10 6 by opening an MKS Type 248 Control Valve, which allows most of the gas in the stabilization chamber to be pumped away by a vacuum pump. The valve is controlled via an MKS 250E Control Module. The shared outlet capillary from the pressure-stabilization chamber is crimped and thermally insulated to avoid changes in conductance and thus air flow caused by room temperature fluctuations. The pressure in the getter oven (D in Fig. 2) is about 2 mTorr (0.3 Pa) because the getter material effectively acts as a vacuum pump. 145
This increases the concentration of He in the gas mixture by a factor of about 100, boosting precision. The getter oven has an inner diameter of 0.94 cm and a length of 22.5 cm. It is manufactured from heat resistant stainless steel (SS310) and equipped 6 with VCR face seals for easy maintenance. The temperature of the getter oven is determined by manually adjusting the power provided to two OMEGA CRWS semi-cylindrical heaters surrounding the getter. The heaters are additionally equipped with an independent limit controller for safety.
The gettering efficiency depends on the heaters' temperature and must be balanced against material tolerance and increased evolution of H2 gas from the metal in the getter. H2 forms a solid solution in Ti and is continuously released to the gas stream 155 when Ti is heated. The solution process is reversible and H2 is absorbed if the Ti is cooled down. H2 could interact with He + in the source or combine with ionized gas to form hydride compounds such as ArH + (Fig. 3) However, since the H2 flux into the gas stream varies slowly compared to the 30-second timescale of switching between sample and standard gas, the impact of H + cancels during sample-standard comparison. In its current size (~10-12 g Ti), the getter can be used for 70-80h before the Ti must be replaced to prevent N2 breakthrough. This requires breaking vacuum in the inlet approximately once every four 160 weeks depending on usage. After replacement, fresh titanium is gradually heated to 900°C over ~12h in isolation from the MS to allow degassing without contaminating the MS. A coarse mesh of metal wire and 2μm SWAGELOK filters on both sides of the getter prevent getter-derived dust from entering the vacuum system and MS.

Inlet operation
We have developed customized scripts using the software ISODAT provided with any MAT253 mass spectrometer to control 165 the inlet system and operate the pneumatic actuators for He/M analysis (Fig. 3). In a typical run, the instrument performs sample-standard gas switching with a ~30 second switching time (~60 sec full cycle), using a conservative 18-second idle time to allowing the MS signal to stabilize before integration . As customary in dynamic MS noble gas application, we group each analysis into blocks consisting of (i) adjusting the accelerating voltage to find the center of the 4 He peak followed by (ii) 20 sample-standard comparisons. Background concentrations of 4 He in the MS are determined by closing the inlet upstream of 170 the getter oven and subtracted daily. Data are quality controlled and anomalous cycles are rejected when delta values deviate by more than 3 standard deviations from the smoothed time series or when there are abrupt changes detected in the ion beam associated with instability in the MS source (not shown). ISODAT also monitors the MS source pressure and closes the external change-over-valve (Fig. 2) to protect the MS in case of a pressure control failure.

Gas handling and sample delivery systems 175
Air is delivered to the inlet system from a pair of high-pressure gas cylinders (A in Fig. 2). For He/N2 standard gas, we rely on compressed dry air stored in high-pressure cylinders, as is conventional for atmospheric measurements of O2/N2, CO2, and Ar/N2 (Keeling et al., 2007). All cylinders are stored horizontally for 2 days in a thermal enclosure at ambient temperature before analysis to minimize the risk of thermal fractionation. The pressure is dropped to slightly above ambient directly at the head valve of high-pressure cylinders using capillaries rather than regulators. The use of capillaries ensures that all wetted 180 parts are exclusively metal, which is impermeable to He, and eliminates problems we encountered using regulators during initial tests. Due to the use of capillaries, the gas delivery system cannot be evacuated efficiently and instead must be purged for several hours ahead of analysis until the signal stabilizes. The flow rates in the lines are monitored using 0.1 liter per minute Omron DF6-P flow meters and are manually balanced at around 27-28 cm 3 min -1 before every analysis by adjusting the crimping of both 316 stainless steel capillaries. Sample and standard gas streams are both dried before entering the pressure 185 stabilization chamber (C in Fig. 2) by flowing through U-shaped cold traps made from about 25 cm of 1/4" stainless steel tubing. The traps are held at about -80°C by submerging the metal tubing in a dry ice and ethanol mixture for the duration of the analysis.

Gettering performance
A mass scan of air introduced through the gettering and flow-stabilizing inlet system revealed that N2 and O2 are almost completely removed from the air by the on-line getter (Fig. 4). He is effectively preconcentrated in the gas mixture. 40 Ar ions with one or more charges yield the largest beams in the scan followed by 36 Ar, and H2 evolving from the hot metal in the getter oven. 205

Response time
Our setup demonstrates the ability to transition between sample and standard gas with a e-folding time scale of ~4 seconds (Fig. 5). The e-folding time is primarily controlled by the volume of the getter and the total flow of gas through the getter.
Regions of the inlet system upstream of the getter experience ~100x faster flushing than downstream of the getter because the gas upstream still contains N2 and O2 and hence flows much faster. An associated large drop in pressure however ensures that 210 all parts of the inlet system are flushed out similarly quickly. The e-folding time does not change substantially over the life span of the getter.

Analytical precision
Using the default 60 sec sample-standard cycle, the gettering and flow-stabilizing inlet system achieves an internal precision in δ(He/M) of approximately ±15 per meg over 1.5h and ±8 per meg for samples run 6h or longer (1σ, standard error of ~90 215 and ~360 cycles respectively). The (external) reproducibility of repeated 6-8h measurements of the same sample and standard gas cylinder combination is comparable and essentially as expected from the shot-noise on the 4 He ion current (Figs. 6&7).
The zero enrichment, i.e., the delta value observed when introducing the same gas through sample and standard side of the inlet, is generally small and stable over time. It is tested by mounting the crimped delivery capillaries (A in Fig. 2) to a tee fitting, which splits the gas stream at high pressure from a single tank of air. This tee minimizes thermal fractionation by 220 dividing the flow at a junction machined into the center of a large brass block (Keeling, 1988). Identical delta values (within error) obtained after reversing the outlet from the tee demonstrate that no measurable fractionation occurs within the tee and therefore that the zero enrichment reflects a persistent asymmetry somewhere downstream, most likely within the pressure stabilization chamber. The typical zero enrichment varies slightly with the mean flow of gas into the stabilization chamber (F), the pressure in the chamber (P), and the flow offset between SA and ST side (ΔF) before entering the stabilization chamber 225 (Fig. 7). Weighted multiple linear regression analysis using 3 different pressure levels (9 psi, 14 psi, and 16 psi, i.e, 62.1, 96.5, and 110.3 kPa) and predictors F, P, and ΔF reveals that the zero enrichment value decreases by 2.8±0.9 per meg per 1 cm 3 min -1 change in mean flow away from 27.5 cm 3 min -1 and increases by 17.2±4.8 per meg per 1 cm 3 min -1 flow imbalance between SA and ST (1σ). The analysis also finds that the dependence of δ(He/M) on F and ΔF is significant at the 5% level.
For a balanced flow of 27.5 cm 3 min -1 and a pressure of 62.1, 96.5 and 110.3 kPa in the stabilization chamber, the mean zero enrichment is -9.61±7.2, 1±3.7, and -15.7±4.7 per meg, respectively. For actual measurements, P is held constant at 96.5 kPa with very high precision. F and ΔF are stable over 8h to within ±0.2 cm 3 min -1 . This typically yields a correction for mean gas flow and flow imbalance of less than 10 per meg with an uncertainty smaller than 6 per meg, which increases the overall analytical uncertainty in repeat tank analysis from 8 to 10 per meg.

Discussion 235
The gettering and flow-stabilizing inlet system has demonstrated the ability to determine the helium mole fraction difference between a sample and standard gas, δ(He/M), to about 10 per meg in a single 6-8h analysis and has a range of possible applications.
Our primary targets are (i) to use stratospheric δ(He/N2) as a tracer of the large-scale stratospheric circulation and (ii) to evaluate tropospheric δ(He/N2) trends as a possible indicator of anthropogenic fossil fuel exploitation. 240 We expect an excellent signal-to-noise ratio for the detection of stratospheric changes in δ(He/N2). Interannual variability in stratospheric δ(He/N2) is likely on the order 300-400 per meg (Table 1). Repeat 6-8h measurements of a high-pressure cylinder currently achieve a precision of ~10 per meg, or about 40 times better than the stratospheric signal. Associated changes in tropospheric δ(He/N2), in contrast, are likely much smaller at around 6 per meg and therefore at or below the current limit of detection even after averaging of multiple samples. 245 He/N2 measurements can help quantify the anthropogenic 4 He release over time due to fossil fuel extraction (Boucher et al., 2018c;Evans, 2013, 2004;Mabry et al., 2015;Oliver et al., 1984;Sano et al., 2010Sano et al., , 1989. Although theoretical predictions clearly support an anthropogenic 4 He increase, past observational studies produced conflicting evidence. Recent improvements in analytical methods and sampling have narrowed the uncertainty in 3 He/ 4 He trend estimates to <30 per meg y -1 with a mean statistically indistinguishable from zero (Table 1). However, with a precision of ~10 per meg on single samples, 250 measurements of δ( 4 He/N2) on decades-old archived air may allow trend detection to ~1 per meg y -1 or better, while also avoiding possible complications from 3 He emissions that could bias estimates of the 4 He source from 3 He/ 4 He.
Another possible application is the investigation of spatial gradients in atmospheric δ(He/N2) caused by the distribution of local volcanic or anthropogenic sources (e.g., Sano et al., 2010;Boucher et al., 2018c). High precision δ(He/N2) may allow the detection of diffuse helium release in regions of volcanic activity (Boucher et al., 2018b). Furthermore, global north-south 255 δ(He/N2) gradients from anthropogenic emission sources concentrated in the Northern Hemisphere are likely on the order of 10s of per meg and thus may also be detectable directly. Alternatively, studies could target more local gradients around oil or natural gas facilities that are likely even greater.
The method developed here is potentially applicable to measure the abundance of any noble gas in air. The intensity of the ion beam and thus the precision for different noble gases depends on their natural abundance and ionization efficiency in the MS 260 source. 20 Ne and 22 Ne have isobaric interferences from doubly charged Ar and CO2, but Kr and Xe yield usable ion beams (Table 2). We estimate a precision of ~5 and ~19 per meg for repeat 6-8h analyses of δ( 84 Kr/ 28 N2) and δ( 129 Xe/ 28 N2) respectively, by assuming that precision scales with the square root of the total ions counted as expected from shot-noise behavior. This estimate compares favorably to the precision currently reported in conventional dual inlet mass spectrometry studies (Baggenstos et al., 2019;Bereiter et al., 2018). For example, Baggenstos et al. (2019) achieved a precision of 88 per 265 meg and 203 per meg for repeat ~2h analyses of δ( 84 Kr/ 40 Ar) and δ( 132 Xe/ 40 Ar) in ambient air, respectively. The improved precision enabled by our inlet system should be sufficient to resolve the previously unobserved annual cycle of Kr and Xe caused by the seasonal release and uptake of both gases by the ocean as it warms and cools. The seasonal cycle of δ( 40 Ar/ 28 N2) has an amplitude of 5-15 per meg in the extratropics (Keeling et al., 2004). δ( 84 Kr/ 28 N2) and δ( 132 Xe/ 28 N2) however are ~3.4 and ~8.9 times more sensitive than δ( 40 Ar/ 28 N2) to changes in ocean temperature owing to the different temperature-270 dependences of Ar, Kr and Xe solubility in seawater (Hamme and Emerson, 2004;Jenkins et al., 2019). This implies that seasonal variations in δ( 84 Kr/ 28 N2) and δ( 132 Xe/ 28 N2) have a magnitude of 17-51 and 45-134 per meg respectively, which would be readily resolved if precision of our system scales as expected with signal strength.
The gettering inlet and MS system was applied here only for single ion (He + ) detection, but alternately could be applied for multi-ion collection. The acquisition of Kr and Xe isotope ratios for example would provide valuable additional information 275 for detecting artefactual fractionation during sampling and allow further improvements in precision by increasing the total number of ions collected.
The need for only a single ion detector also allows the gettering and flow-stabilizing inlet to be interfaced with simpler and more affordable mass spectrometers, such as quadrupole systems. The performance of the system will depend on the stability of the 4 He + -ion beam over the time scale of switching and will need to be evaluated critically, but any variability on time scales 280 longer than the switching time is canceled by sample-standard differencing.
Additional work is needed to further improve calibration methods and to establish standard procedures for collecting air samples while avoiding artifacts in He/N2 at the 10 per meg level. We currently need samples of ~16-20l for a full 8h analysis because long purging and analysis times are necessary to achieve a precision of 10 per meg. If reduced precision is acceptable, analyses time can be shortened but purging of the inlet system for at least one hour is needed before each analysis even for 285 lower precision work. Furthermore, air samples must currently be provided at pressure greater than 3 atm to allow sufficient flow through the narrow tubing into the pressure-stabilization chamber. The reproducibility of measurements also depends on adequate calibration strategies. The short-term reproducibility of high-pressure cylinders shown in Figure 6 and the long-term stability established for O2/N2, CO2, and Ar/N2 standard gases in previous work (Keeling et al., 2007) suggest that long-term stability in δ(He/N2) is achievable but needs further evaluation. 290 into the MS from an actively pressure-controlled open-split (Henneberg et al., 1975), such that variability of the 4 He + ion beam 295 directly reflects differences in the helium mole fraction of the gas mixtures. Measurements of the helium mole fraction can easily be converted to δ(He/N2) if O2/N2, Ar/N2, and CO2 concentrations of the sample are determined as well. An online getter preconcentrates He and other noble gases before entry into the MS by chemically removing >99.99% of all N2 and O2 in a reaction with titanium sponge. Our method thereby avoids the need for peak jumping and a multi-collector mass spectrometer, while achieving a precision of ~10 per meg (1σ) on repeat analysis of δ(He/N2) in high-pressure tanks. 300 In future work, the gettering and flow-stabilizing inlet system could be used to investigate possible interannual to decadal changes in stratospheric δ(He/N2) linked to variability in stratospheric circulation and stratosphere-troposphere exchange processes. Additional applications could include the search for a signal of anthropogenic helium release during fossil fuel extraction and burning, or measurements of spatial gradients resulting from localized human or natural sources of helium. The setup is also suitable for the analysis of other noble gases and could therefore be used to study seasonal ocean warming 305 associated with degassing or uptake of Kr and Xe from the ocean (Baggenstos et al., 2019;Bereiter et al., 2018).

Acknowledgements
We thank Ross Beaudette, Alan Seltzer, Sarah Shackleton, Jacob Morgen, Jessica Ng, and Eric Morgan for laboratory support and insightful discussions during the development of the He/N2 analysis system. We are grateful to Shane Clark, Savannah Hatley, Adam Cox, and Timothy Lueker for providing high-pressure cylinders used during testing. We also thank them for 310 maintaining and operating the Ar/N2, O2/N2 and CO2 analysis systems in the Keeling laboratory. This work was supported by the National Science Foundation grants MRI-1920369 and AGS-1940361.

Data availability
Data presented in this manuscript are available as an electronic supplement to this paper from the journal website.

Competing interests 315
The authors declare that they have no conflict of interest.

Author contribution
BB designed and build the inlet system with important design expertise from WP, JS and RK. BB performed all testing and prepared the manuscript with contributions from all co-authors.  pathways and letters A-E in blue circles label the main sections of the inlet system. Red double arrows indicate manual switching option in the inlet system. Sample or standard gas is delivered from high-pressure cylinders (A). The flow can be measured by two Omron flow meters before entry into the pressure stabilization chamber (C). Pistons (B) alternately move fine metal tubing in the pressure stabilization chamber, pushing either the sample or standard gas stream deeper into the stabilization chamber where the gas will be picked up by a single capillary leading to the MS. A sliding seal is made using 470 lubricated O-rings between 1/8" and 1/16" OD tubing at the entry to the pressure-stabilized chamber. This ensures sufficient rigidity and protects the fine metal tubing inside The chamber is exhausted to a vacuum system and the pressure is monitored and controlled by a differential pressure gauge combined with an automatic MKS flow control valve. The stainless-steel getter oven (D) has an inner diameter of 1/2" and is filled with 10-12 g of titanium sponge. 2μm-filters prevent particles from contaminating the MS and gas delivery system. In case of an anomalous pressure change in the MS or when venting the getter 475 oven, the getter oven can be isolated from the pressure-stabilization chamber with a change-over-valve controlled directly by the MS software. The entire inlet vacuum system is backed by a diaphragm vacuum pump and a turbomolecular pump (E). A manual shutoff-valve can isolate the getter oven from the MS.    respectively. Analysis 6 for cylinder A was shorter resulting in a larger uncertainty for that measurement. Data are not corrected for zero enrichment effects discussed in the text. 500 Figure 7. Difference in δ(He/M) between two identical gas streams (i.e., the zero enrichment) measured repeatedly under different conditions over 1.5-3h. Error bars show 1σ uncertainty (internal precision). Measurements were made at different pressure levels (a), with slightly varying gas flows to the stabilization chamber (b), and imbalances in flow between SA and ST side (c). The same shared capillary was used for all analysis. Therefore, the pressure in the stabilization chamber controls 505 the intensity of the ion beam and the internal precision of the analysis, illustrated by the greater scatter of observations at 9 psi (62.1 kPa). Delta values shown in (d) were corrected for the influence of pressure, mean flow, and flow imbalance according to coefficients found by multiple linear regression (see text). For a pressure of 14 psi (96.5 kPa), corrected delta values generally show scatter as expected from shot-noise behavior and corrected delta values are stable over time.  Figure 4. Relative ion beam intensities on MAT253 are calculated from the scan 515 with identical source tuning. Xe isotope beams were not observed but scaled from previous observations in the lab.