Seeds of Festuca arundinacea(Schreb.), also commonly called tall fescue, were first sown in a commercial potting soil (Terreau universel, Botanic, France; composition: black and blond peat, wood fibre, green compost, and vermicompost manure, organic and organo-mineral fertilisers and micronutrient fertilisers). During 15 to 20 d, they were then placed in a growth chamber of the Microcosms experimental platform of the European Ecotron of Montpellier (https://www.ecotron.cnrs.fr, last access: 9 March 2023) under diurnal light–dark cycles (Table S1 in the Supplement), air temperature set at 20 ∘C (Tair), air relative humidity (RH) at 80 %, and CO2 atmospheric concentration close to ambient air (concentration of CO2=400 ppm).

Twelve pots (8 cm × 8 cm × 12 cm with 180 to 200 g of dry soil) containing approximately 25 to 30 mature fescue plants were used for each experimental run. All plants were placed in a plastic tray filled with tap water inside an airtight transparent chamber manufactured from welded polycarbonate (10 mm wall thickness and 120 L volume) similar to the chambers used by Milcu et al. (2013) (Fig. 1). The sealing of the closed chamber was checked before each experiment using helium.

To control temperature and light intensity inside the closed chamber, this smaller chamber was placed in a larger controlled-environment growth chamber. Light was provided by two plasma lamps (GAVITA Pro 300 LEP02; GAVITA) with PAR = 200 µmol m-2 s-1, and air temperature inside the closed chamber was regulated at 19 ± 1 ∘C by adjusting the growth chamber temperature.

The closed chamber (Fig. 1) was used as a closed gas-exchange system with controlled, and continuously monitored, environmental parameters. Air and soil temperature (CTN 35, Carel), air relative humidity (PFmini72, Michell instrument, USA), and CO2 atmospheric concentration (GMP343, Vaisala, Finland) were measured and recorded using the growth chamber datalogger (sampling rate = 1 min). O2 concentration was continuously monitored using an optical sensor (Oxy1-SMA, Presens, Germany). Because precise O2 concentrations are determined in our samples by mass spectrometry (see next section), the measurements of the Oxy1-SMA were only used as a control during the experiment. The measured O2 value for atmospheric air was adjusted to 20.9 % before each sequence of experiments, and the same adjustment (offset) was then applied to the O2 record during the following sequence.

Air relative humidity was regulated between 80 % and 90 % using a heat exchanger (acting as a condenser) connected to a closed-cycle water cooling system. The condenser was positioned in a way to create a closed water cycle in the biological chamber (water vapour from evapotranspiration was condensed back into irrigation water). In order to keep the CO2 mixing ratio close to 400 ppm during the light periods, photosynthetic CO2 uptake was compensated with injections of pure CO2 using a mass flow controller (F200CV, Bronkhorst, the Netherlands). During the dark periods, a soda lime trap connected to a micro-pump (NMS 020B, KNF, Germany) was used to remove the excess CO2 coming from respiration. CO2 atmospheric concentration during the night was kept below 200 ppm.

To ensure atmospheric pressure stability in the closed chamber, a pressure compensation system, made of two connected 10 L gas tight bags (Restek multi-layer polyvinyl fluoride foil gas sampling bag, USA), was installed. Each bag was half full of atmospheric air; the first one was installed in the closed chamber while the second one was outside the chamber. This way, each bag inflated or deflated in response to pressure variations caused either by O2 or CO2 uptake or release. The pressure difference between the closed chamber and the atmosphere was regularly measured using a differential sensor (FD A602-S1K Almemo, Ahlborn, Germany).

Finally, the enclosed air was mixed using seven brushless fans and considered homogeneous.

To measure the isotopic composition along the experiment, small samples of gas were collected in 5 mL glass flasks, made of two Louwers HV glass valves (1-way bore 9 mm Ref. LH10402008, Louwers Hanique, the Netherlands) welded together. Those flasks, previously evacuated, were mounted on PFA tubing (1/4th) using two 1/4th UltraTorr fittings (SS-4-UT-9, Swagelok, USA). Two manual valves (SS-4H, Swagelok, USA) were also installed on the PFA tubes to open or close the circuit. A micro-pump (NMS 20B, KNF, Germany) was finally turned on during air sampling to ensure closed chamber atmosphere circulation through the flask. The flow rate was equal to 1.6 L min-1.

After each experiment, the plant leaves were collected, placed in airtight flasks, and immediately frozen at -20 ∘C for at least 24 h to make sure there was minimal loss of water through vaporisation when the vial was opened later. The extraction of water from leaves was done according to the procedure detailed in Alexandre et al. (2018). The vial was fixed onto a cryogenic extraction line and was first immersed in a liquid nitrogen Dewar to prevent any sublimation of the water. The water extraction line was emptied of most of its air (<10-5 Pa). Once this pressure was reached, the pump was turned off and a valve was closed in order to keep a constant static void within the system. The “reception” vial was then immerged in a liquid nitrogen Dewar which will act as a water trap, whilst the sample vial for the water was then transferred to a water bath maintained at 75 ∘C. The system was kept in these conditions for no less than 6 h, so that all the water present in the leaf and stems was extracted. Afterwards, in order to remove all of the organic compounds of the extracted water, an active charcoal was placed in the extracted water and left under agitation for the night.

For analysis of δ17O and δ18O of water, leaf water was converted to O2 using a fluorination line for reaction of H2O with CoF3 heated to 370 ∘C at LSCE. The isotopic composition of the dioxygen was measured with an IRMS equipped with a dual inlet (Thermo Scientific MAT253 mass spectrometer). The standard that was chosen was an O2 standard calibrated against VSMOW. The precision was 0.015 ‰ for δ17O, 0.010 ‰ for δ18O, and 6 ppm for Δ17O (Eq. 1); for more details, refer to Landais et al. (2006).

The values of δ18O and δ17O of leaf water measured with respect to VSMOW are then expressed with respect to the isotopic composition of dioxygen in atmospheric air (classical standard for δ18O and δ17O of O2 measurements). No consensus has been reached for the values of δ18O and δ17O of O2 in atmospheric air with respect to δ17O and δ18O of H2O of VSMOW. These differences are most probably to be attributed to the different analytical techniques used for preparing and measuring the samples (Yeung et al., 2018; Wostbrock et al., 2021). In our case, because we use a similar set-up to the one developed by Barkan and Luz (2003) for the analyses of the triple isotopic composition of O2 in the air (cf. next section), we have chosen to base our calculation on their estimates. In this study, we have thus chosen the value of 23.88 ‰ for δ18O of O2 values with respect to VSMOW following Barkan and Luz, 2005. As for the δ17O of O2 value with respect to the VSMOW value, we use two different possible estimates from these authors, either 12.03 ‰ (Luz and Barkan, 2011) or 12.08 ‰ (Barkan and Luz, 2005). We acknowledge that because of the absence of consensus, slightly different values could be obtained for the fractionation factors determined in this study if a different choice is made for the reference values of δ18O and δ17O of O2 in atmospheric air with respect to δ17O and δ18O of H2O of VSMOW.

The air samples collected in the closed chambers were transported to LSCE for analyses of the isotopic composition of O2. The flasks were connected on a semi-automatic separation line inspired from Barkan and Luz (2003) which was made up of eight ports in which two standards (outside air) and six samples were analysed daily (Brandon et al., 2020). After pumping the whole line, the air was circulated through a water trap (ethanol at -100 ∘C) and then through a carbon dioxide trap immersed in liquid nitrogen at -196 ∘C. After collection of the gas samples on a molecular sieve trap cooled at -196 ∘C, a helium flow carried it through a chromatographic column which was immersed in a water reservoir at 0 ∘C to separate the dioxygen and the argon from the dinitrogen. After separation of the dioxygen and argon from helium, the gas was collected in a stainless steel manifold immersed in liquid helium at -269 ∘C.

After collection, the samples were analysed by the IRMS, previously mentioned for leaf water analyses. The following ratios were measured: 18O/16O, 17O/16O, and O2/Ar (as an indicator of the O2 concentration because Ar is an inert gas). δ17O and δ18O of O2 each sample were obtained through three series of 24 dual inlet measurements against a standard made of O2 and Ar. This sequence was followed by two peak jumping analyses of the O2/Ar ratio, including separate measurements of the O2 and Ar signals for both the standard and the sample. The uncertainty associated with each measurement was obtained from the standard deviation of the three runs and from the repeated peak jumping measurement for δO2/Ar which was defined by n(O2)n(Ar)samplen(O2)n(Ar)standard-1×1000, and n(O2) is the number of moles of O2 and n(Ar) the number of moles of Ar. The uncertainty values for Δ17O, δ18O, and δO2/Ar were, respectively, 10 ppm, 0.05 ‰, and 0.5 ‰.

Each day, we performed measurements of the dioxygen isotopic composition and O2/Ar ratio on two samples of outside air, which is the standard for the isotopic composition of O2 (Hillaire-Marcel et al., 2021), so that the calibrated δ18O value for our sample was calculated as in Eq. (2). 2δ18Ocalibrated=δ18Omeasured/1000+1δ18Ooutsideair/1000+1-1×1000

Our goal was to calculate the fractionation factor associated with δ17O and δ18O for soil respiration, dark leaf respiration, and photosynthesis using the microcosm described above. In order to quantify the fractionation factors, we needed to work in closed and controlled conditions. Given the volume of the closed chamber (120 dm3, hence about 1.12 mol of O2) and the order of magnitude of dark respiration (order of magnitude of 0.08 µmol O2 s-1 for soil respiration) and net photosynthetic fluxes (order of magnitude of 0.45 µmol O2 s-1) inside the chamber, we calculated that experiments should last from 3 d to more than 2 weeks so that more than 110 of the O2 in the chamber can be recycled by the plant and soil. This recycling allows the creation of sufficiently large isotopic signals (especially Δ17O of O2) to be detected and measured. We set up two different experiments in the closed chamber, each experiment being repeated three or four times to characterise the experimental repeatability of the system.

The first experiment (repeated four times, i.e. in four sequences) aimed at studying the fractionation factors during soil respiration. The second experiment (repeated three times, i.e. in three sequences, each sequence being divided into several periods with or without light) aimed at studying the fractionation factors during dark respiration and photosynthesis of plants.

Prior to the aforementioned experiments, measurements were carried out on a closed empty chamber to check the absence of leaks as well as the absence of isotopic fractionation (Table S2).

To conduct the soil respiration experiment, 2.6 kg of soil (Terreau universel, Botanic) were placed in 12 different pots. The light was turned off during this experimental run (Table S1). We decided not to apply any diurnal cycles during dark respiration experimentations for two reasons. First, we wanted to prevent the development of algae, mosses, or any photosynthetic organisms in the chamber. Secondly, it was easier to optimise temperature control as the light radiation could increase the temperature inside the closed chamber. During this dark period, CO2 from soil respiration accumulates in the biological closed chamber. To have a stable concentration of CO2 during the whole dark period, the CO2 was trapped using soda lime. Four sequences were performed with respective durations of 53, 51, 43, and 36 d.

We used the same soil with plants (Festuca arundinacea) grown before the start of the three sequences of the photosynthesis and dark respiration experiment. In order to obtain a significant change of the Δ17O of O2 signal in our closed 120 dm3 chambers, the three experiments were run for 1 to 2 months. CO2 level was controlled to 400 ppm by a CO2 trap and CO2 injections. This was done to ensure that the CO2 in the chamber did not reach levels too far from the atmospheric composition as this could have affected the physiology of the plant. The light cycle was controlled to alternate between day (photosynthesis and respiration) and night conditions (respiration) (Table S1).

The values of the leaf water measurements are presented in Table S3. Because the experiments had to be carried in a closed chamber, we could not sample leaves during the experiment and only got a value at the end of each sequence. Nevertheless, we could compare the isotopic composition of the irrigation and soil water at the start and at the end of the experiment.

Respiration is associated with isotopic fractionation. The light isotopes, 16O, are more easily integrated by microorganisms than the heavy isotopes, 18O, which hence remain in the atmosphere. We express the fractionation factor for soil respiration as 718αsoil_respi=18Rrespired18Rair. In our experiment, the respiratory process took place in a closed reservoir so that we could calculate the fractionation factors from the evolution of the concentration and isotopic composition of dioxygen in the chamber. The number of molecules of dioxygen in the air of the closed chamber, nO2, between time t and time t+dt can be written as 8nO2t+dt=nO2t-dnO2, with dn(O2) the number of dioxygen molecules respired during the time period dt. A similar equation can be written for the number of dioxygen molecules containing 18O remaining in the air of the chamber: 918Rt+dt×nO2t+dt=18Rt×nO2t-18Rt×18αsoil_respi×dnO2. The evolution of the isotopic ratio of oxygen, 18R, between time t and time t+dt can be written as: 1018Rt+dt=18Rt+d18R Combining equations Eqs. (8), (9), and (10); neglecting the second order term d18Rt×dnO2t; and integrating from t0 (starting time of the experiment when the chamber is closed) to t leads to 1118εsoil_respi=18αsoil_respi-1=ln⁡δ18Ot1000+1δ18Ot01000+1ln⁡n(O2)tn(O2)t0. Because argon is an inert gas, we can link nO2tnO2t0 to δO2Ar, so that 12nO2tnO2t0=δO2Art1000+1δO2Art01000+1.

In order to calculate the isotopic fractionation associated with soil and plant respiration during the dark period, we followed the same calculation as for the soil respiration (Sect. 2.4.1). In this case, we selected only night periods from each sequence of the photosynthesis and dark respiration experiment.

During photosynthesis, the oxygen atoms in the dioxygen produced by the plant come from the oxygen atom of water consumed by photosynthesis in the leaves, so that the fractionation factor during photosynthesis can be expressed as 1318αphotosynthesis=18RproducedO218Rlw, where lw stands for leaf water.

For our study of Festuca arundinacea, we consider that the water in the mesophyll layer can be represented by bulk leaf water.

Photosynthesis occurs during the light periods. However, it should be noted that dark respiration, photorespiration, and the Mehler reaction occur at the same time. In a first approach, we made the assumption that respiration rates remain the same during the light and dark periods. This assumption is probably true for soil respiration since flux of heterotrophic dark respiration is not expected to change for different light conditions if the other environmental drivers (e.g. humidity, temperature, soil organic matter) are constant. However, autotrophic dark respiration is expected to decrease during light periods compared to dark periods. As a consequence, we present sensitivity tests to the dependence of a vanishing dark respiration of leaves during the dark period in Table S4.

Thus, at each stage, dioxygen is both produced by photosynthesis and consumed by the aforementioned O2 uptake processes (hereafter total_respi) by the plant according to the mass conservation equation 14nO2t+dt=nO2t-dntotal_respi+dnphotosynthesis, where dntotal_respi is the number of molecules of O2 consumed by dark respiration, photorespiration, and the Mehler reaction between time t and t+dt, and dnphotosynthesis is the number of molecules of O2 produced by photosynthesis between t and t+dt.

The budget for 18O of O2 can be written as 1518Rt+dt×nO2t+dtnO2t0=18Rt×nO2tnO2t0-18Rt×18αtotal_respi×dntotal_respinO2t0+18Rlw×18αphotosynthesis×dnphotosynthesisnO2t0, where 18αtotal_respi is the fractionation factors associated with each O2 consuming process period throughout the whole experiment.

We introduced the normalised fluxes of photosynthesis and total respiration as 16Fphotosynthesis=dnphotosynthesisnO2t0×dt,17Ftotal_respi=dntotal_respinO2t0×dt,18a18R=d18Rdt. This led to the following expression of 18αphotosynthesis: 1918αphotosynthesis=nO2t/nO2t0×a18R+18Rt×(Fphotosynthesis-Ftotal_respi+18αtotal_respi×Ftotal_respi)18Rlw×Fphotosynthesis. This equation can be simplified at t=0 for 18Rt=18Rt0=1 and nO2t=nO2t0: 18αphotosynthesis depends on the values of 18αtotal_respi and of Ftotal_respi, themselves dependent on the values of 18αMehler (fractionation factor associated with Mehler reaction), FMehler (flux of oxygen related to Mehler reaction), 18αdark_respi, Fdark_respi, 18αphotorespi (fractionation factor associated with photorespiration), and Fphotorespi (photorespiration flux of oxygen). These last four parameters could not be determined in our global experiment. Our determination of 18αphotosynthesis will thus rely on assumptions for the estimations of 18αMehler, FMehler, 18αphotorespi, and Fphotorespi.

To separate the 18αdark_respi from the other fractionation factors, we defined 2018αtotal_respi=18αphotorespi×fphotorespi+18αMehler×fMehler+18αdark_respi×fdark_respi with 21Ftotal_respi=Fdark_respi+Fphotorespi+FMehler. f indicates the fraction of the total oxygen uptake flux corresponding to each process (dark respiration, photorespiration, and Mehler reaction) so that 22fdark_respi+fphotorespi+fMehler=123Fdark_respi=fdark_respi×Ftotal_respi24Fphotorespi=fphotorespi×Ftotal_respi25FMehler=fMehler×Ftotal_respi. In the absence of further constraints, we used here as first approximation the global values from Landais et al. (2007) for fdark_respi (0.6), fphotorespi (0.3) and fMehler (0.1). Values for αphotorespi and αMehler were based on the most recent estimates of Helman et al. (2005).

During the four sequences, the respiration activity led to a decreasing level of the O2 concentration measured by the optical sensor or through the δO2/Ar evolution from IRMS measurements (Fig. S1 in the Supplement). The comparison of the evolution of the O2 concentration during the different sequences showed that respiratory fluxes were different with a maximum factor of 4 between the different sequences (Fig. S1). In parallel to the decrease in O2 concentration, the δ18O increased as expected because respiration preferentially consumes the lightest isotopes: over the 51 d of the second soil respiration sequence, we observed a linear decrease of oxygen concentration by more than 5 % while δ18O increased by 8 ‰ (Fig. 2). A Mann–Kendall trend test showed that the Δ17O of O2 does not show any statistically significant trend over the four sequences (Fig. S2) (p values were equal to 0.40, 0.08, 0.58, and 0.47, respectively).

We used the 15 to 20 samples obtained during each sequence of the soil respiration experiment to draw the relative evolution of ln⁡(18Rt/18Rt0) vs. ln⁡((δO2Art/1000+1)/(δO2Art0/1000+1)) following Eq. (11) (Fig. 3). The slope of the corresponding regression line provided the isotopic discrimination 18εsoil_respi and hence the fractionation factor 18αsoil_respi for each sequence (Table S5). It could be observed that despite differences in respiratory fluxes for the different sequences (the standard deviation is equal to 50 % of the average flux across sequences; see Table S5), the relationship between δ18O of O2 and O2 concentration (or δO2/Ar), and hence the calculated fractionation factor associated with respiration, is not much affected.

Using the results of the four sequences, we determined the values for the mean isotopic discrimination 18εsoil_respi (-12.3±1.7 ‰), the mean isotopic discrimination 17εsoil_respi (-6.4±0.9 ‰), and the average θsoil_respi (0.5164 ± 0.0005).

During the night periods, when only respiration occurred, we observed a decrease in O2 concentration by 1 % within 3 d and a δ18O increase by 1 ‰ during the same period (Fig. 4). The evolution was qualitatively similar to that of soil respiration experiments with higher fluxes. We observed the same trends for the evolution of δO2/Ar during the night periods as for the respiration experiment. During light periods, there was a marked decrease in δ18O (2 ‰) and a marked increase in the flux of oxygen released (1 %) during 1 d. We observed the same trends for the evolution of δO2/Ar during the night periods as for the respiration experiment.

The Mann–Kendall test (95 %) showed a significative increasing trend of the Δ17O of O2 over sequences 1 and 2 (Fig. S3) (≃ 100 ppm in 31 d for sequence 1, ≃ 100 ppm in 40 d for sequence 2), while no significant increase of Δ17O of O2 is observed over sequence 3 (Fig. S3).

The average of the isotopic discrimination for dark respiration 18εdark_respi and 17εdark_respi were calculated over the nine night periods, and we obtained values of, respectively, -17.0±2.0 ‰ and -8.5±0.8 ‰. The average of θdark_respi during the experiment was equal to 0.5124 ± 0.0084 (details in Table S6).

The dark respiration of this experiment includes respiration of both soil and leaves. Because soil respiration fractionation factor has been determined above, it is possible to estimate here the fractionation factor for the dark leaf respiration, and we consider that respiration rate during dark and light periods do not vary: 26Fdark_respi=Fsoil_respi+Fdark_leaf_respi,2718αdark_respi=fsoil_respi×18αsoil_respi+fdark_leaf_respi×18αdark_leaf_respi, with Fdark_leaf_respi the flux of leaf respiration during the night, fsoil_respi the fraction of soil respiration during night periods (Fsoil_respi/Fdark_respi), and fdark_leaf_respi the fraction of dark leaf respiration during night periods (Fdark_leaf_respi/Fdark_respi). 2818αdark_leaf_respi=18αdark_respi-fsoil_respi×18αsoil_respifdark_leaf_respi The isotopic discriminations 18εdark_leaf_respi and 17εdark_leaf_respi were, respectively, equal to -19.1±2.4 ‰ and -9.7±0.9 ‰. The average of θdark_leaf_respi was equal to 0.5089±0.0777. The standard deviations (1σ) were calculated by a Monte Carlo method from the individual uncertainties of the 18αdark_respi, 18αsoil_respi, Fsoil_respi, and Fdark_respi.

In order to calculate an average value for the fractionation factor associated with photosynthesis from Eq. (19), we first calculated the averages of the flux of the O2 consuming processes and of the fractionation factors associated with each sequence: 〈Ftotal_respi〉 and 〈18αtotal_respi〉. We also calculated the net O2 flux during light periods, aN=Fphotosynthesis-Ftotal_respi, as the linear regression, aN, of nO2tn(O2)t0 with time. a18R is also obtained as a linear regression of 18R with time over each light period. Our data support our assumption that the regime was stationary over time and nO2t/nO2t0 evolved linearly over time, which is why we were able to do linear regressions. 2918αphotosynthesis=a18R+aN+〈18αtotal_respi〉×〈Ftotal_respi〉18Rlw×Fphotosynthesis The results of the eight individual αphotosynthesis values are given in Table S10. The value of isotopic fractionation associated with the light period of period 1 of sequence 1 appeared clearly out of range. Following the Dixon's outlier detection test (Dixon, 1960), this value was considered an anomaly (likelihood >99 %) and was removed from further analysis.

We finally estimated the values of 18εphotosynthesis and 17εphotosynthesis as +3.7±1.3 ‰ and +1.9±0.6 ‰, respectively. The average of θphotosynthesis was equal to 0.5207±0.0537, a value which depends on the value taken for the δ17O value of atmospheric O2 vs. VSMOW (Sharp and Wostbrock, 2021), see Table 2.

We performed different sensitivity tests (Sects. S1 and S2 in the Supplement). Sensitivity test 1 (Table S4) quantifies the influence of vanishing flux of dark leaf respiration during the day. This test shows that the assumption of similar flux of dark leaf respiration during the night, and light periods did not influence much the values of photosynthesis fractionation factors. It results in an additional uncertainty of 0.0006 and 0.0005 for the values of 18αphotosynthesis and 17αphotosynthesis.

Sensitivity tests 2 (Tables S7, S8 and S9) were performed on values of the O2 flux and associated fractionation factors for photorespiration and the Mehler reaction. They resulted in additional uncertainties of 0.0007 and 0.0005 for the values of 18αphotosynthesis and 17αphotosynthesis (Table S10).

Sensitivity tests 3 concerned the possible evolution of the isotopic composition of leaf water over the course of an experiment. The comparison of the δ18O of irrigation water and soil water at the end of the experiment shows a possible increase up to 2 ‰ (Table S3). We thus estimate that our values of leaf water δ18O measured at the end of the experiment may be overestimated by 1 ‰ compared to the mean value of leaf water δ18O during the course of the experiment. Taking this possible effect into account would lead to a fractionation factor for photosynthesis higher by 1 ‰ compared to the presented one of 3.7±1.3 ‰, hence a higher isotopic discrimination associated with photosynthesis.

Finally, we evaluated by a Monte Carlo calculation how the different uncertainties listed in the three sensitivity tests described above influence the final uncertainty on the photosynthesis isotopic discrimination. We found a final standard deviations (1σ) equal to 0.3 ‰ for 18εphotosynthesisand 0.15 ‰ for 17εphotosynthesis.