Isotopic homogenization and scrambling associated with oxygen isotopic exchange on hot platinum: studies on gas pairs (O₂, CO₂) and (CO, CO₂)

Abstract

The catalytic exchange between O₂ and CO₂ on hot platinum leads to isotope scrambling in CO₂ and homogenization of the oxygen isotopes in the two phases (O₂ and CO₂) even though the two gases could be widely different in isotope ratios from each other (e.g., about 45‰ in δ¹⁸O) before the exchange. After the exchange, the δ¹⁸O values of the gases become close to each other (within 2‰) depending on the time and temperature. The clumped-isotope analysis of the post-exchange CO₂ samples shows that the Δ₄₇ values (relative to the pre-exchange values) decrease with increase in temperature from near zero (at ∼25 °C) to about −0.8 ± 0.1‰ (at ∼700 °C). The low value of −0.8‰ equals the values typically obtained for pure CO₂ heated to 1000 °C inside quartz tubes suggesting a scrambled state. If the catalytic exchange between CO₂ and O₂ is associated with isotopic scrambling (or random mixing) a microscopic mechanism of isotopic mixing can be inferred. It seems that CO₂ and O₂ dissociate in the adsorbed state on platinum (or quartz surfaces) and the product CO molecules and O atoms mix uniformly while doing a random walk on the hot surface and produce two reservoirs (CO and O) where the three oxygen isotopes are randomly distributed. This can produce the observed internal isotopic scrambling in CO₂ when CO and O recombine to form CO₂ molecules on the platinum surface and desorb. Experiments using CO–CO₂ gas pairs over hot platinum also show homogenization showing exchange of CO as a molecular entity and supporting the suggested mechanism.

---

A. Introduction

There are twelve stable isotopologues of CO₂ comprised of different combinations of ¹²C, ¹³C, ¹⁶O, ¹⁷O and ¹⁸O like ¹⁶O¹²C¹⁶O, ¹⁶O¹³C¹⁶O, etc. The equilibrium thermodynamic model predicts that heavy isotopes ¹³C and ¹⁸O tend to accumulate in the molecular species ¹⁶O¹³C¹⁸O and consequently its abundance is more than expected from the stochastic distribution.¹ This is due to the lower zero point energy of ¹⁶O¹³C¹⁸O compared to the isotopologue ¹⁶O¹³C¹⁶O or ¹⁶O¹²C¹⁸O (the two dominant minor species). This tendency is popularly known as ‘clumping’ which means enhanced association of two heavy isotopes. Clumping depends on the temperature of equilibration, and this dependence forms the basis of Clumped Isotope Thermometry which requires analysis of a given sample of CO₂ and its stochastic equivalent.¹ Theoretically, the stochastic distribution in a particular sample of CO₂ can be calculated if the abundances of isotopic atoms (¹²C, ¹³C, ¹⁶O, ¹⁷O and ¹⁸O) are known.² Experimentally, however, one cannot conclusively decide when a particular sample of CO₂ achieves stochastic distribution since there are issues of isotopic fractionation during sample preparation and mass spectrometric analysis. For analytical purpose one supposes that the random distribution is attained when a sample of CO₂ is heated to a high temperature (1000 °C) in a quartz tube for a few hours. It is thought that interaction of the gas molecules among themselves and with the wall should bring about the expected randomness. Such a gas sample is assumed to define the ‘zero’ (within ±0.03‰) in the scale of “clumpiness”.³

For clumped isotope thermometry, one first analyzes a sample gas for five isotopologue ratios (like ratio R⁴⁷ = n(47)/n(44) where n is number of ions of mass 47 and 44) and then heats up the same gas sample (after retrieving it from the mass spectrometer) to 1000 °C and reanalyzes it for the same ratios. Dividing the first ratio with the second one, after proper calibration, one obtains the clumpiness of the sample gas relative to the heated gas (in Δ₄₇ scale defined later). However, the question remains whether the sample gas heated in this way really attains the complete randomness. Wang et al.² have shown by a theoretical model that CO₂–CO₂ interaction and exchange should bring about randomness in the distribution if equilibrium is attained at high temperature ∼1000 °C but possible wall effect and finite time of heating was not considered. It is of course expected thermodynamically that at high enough temperature and long enough time any sample of CO₂ will attain stochastic distribution.

A recent analytical method^4,5 involving exchange of CO₂ and O₂ in presence of hot platinum (∼680 °C) for two hours provided a useful technique for determination of ¹⁷O/¹⁶O of CO₂. It was noted that this particular exchange reaction mixes the O-isotopes among the two gases so efficiently that the post-exchange isotopic ratios are nearly identical. The mixing was used to obtain the ¹⁷O/¹⁶O ratio in CO₂ which cannot, otherwise, be derived from the measured ion mass ratio of 45/44 due to interference from ¹³C/¹²C ratio. Despite the success of this method to achieve extremely high precision (about 7 parts per million) the mechanism of exchange was not clear. It was thought to be driven by a catalyzed break down of the adsorbed molecular species (O₂ and CO₂) on the surface of the platinum followed by isotopic exchange.⁵ But whether it was thermodynamic exchange process or simple mechanical mixing remained a puzzle.

In this study we aimed to understand the mechanism of this exchange by using a novel approach which combines both isotopic analysis of O₂ and CO₂ as well as analysis of Δ₄₇ (defined later) of the same CO₂ which has undergone exchange. The latter analysis gives us an idea about the degree of scrambling of the CO₂ gas in its isotope distribution before and after exchange. Isotopic homogenization and scrambling should be intimately related if the breakdown of CO₂ (to CO + O) and O₂ (to O + O) takes place followed by mixing of O-atoms from the two sources. It would imply that the process is not a Urey-type thermodynamic exchange where the zero point energy governs the reaction. It would also imply that the twelve CO₂ isotopic species (like ¹⁶O¹²C¹⁶O, ¹⁶O¹³C¹⁶O… etc.) would not obey intra-molecular thermodynamic distribution and the post-exchange CO₂ gas would show signature of scrambling by reduced amount of clumpiness.

While developing the platinum exchange technique for high precision reported in our recent work⁵ we analyzed a few CO₂ samples for clumped isotope ratio Δ₄₇ but the data were preliminary with large uncertainty. The present work is a more intensive study aimed to address exclusively the issue of mechanism of exchange with higher precision and accuracy. New experiments on temperature variation of Δ₄₇ show how the catalysis induced exchange follows a smooth variation (probably Sabatier effect). We also carried out blank experiments to isolate the effect of the quartz exchange on individual gases. In addition to the exchange of O₂ and CO₂ we investigated exchange of gas pair CO–CO₂. If CO₂ indeed breaks down as seen in case of nickel catalyst⁶ then CO molecules from this source would also be able to exchange with a second adsorbed reservoir of CO and one would have homogenization of both carbon and oxygen isotopes. Indeed the results obtained in this work justify our preliminary conclusion on a firmer basis. This study may have interesting applications in isotope geochemistry where surface reactions mediated by catalyst play an important role. For example, in case of leafs there is rapid exchange of H₂O and CO₂ mediated by carbonic anhydrase catalyst. This study may throw light in understanding such cases of isotope exchange.

B. Experimental method

A gas extraction line (made of glass and shown in Fig. 1) was fabricated in the Centre of Earth Sciences, Indian Institute of Science, Bangalore, to study the exchange of oxygen isotopes between O₂ and CO₂ on hot platinum. The reaction section consists of a 1/4 inch diameter quartz tube containing 3–4 small platinum strips (2 mm × 10 mm); the platinum containing part of the tube was placed inside a furnace. This section was connected to the main line by ultra-Torr connector and could be easily disconnected when required. Following the protocol of Mahata et al.,^4,5 the reaction time was kept fixed at 2 hours and the temperature maintained at ∼700 °C by using a controller where the thermocouple sensor was kept inside the cylindrical heater close to the platinum containing part. The temperature could be kept constant to within ±10 °C.

Fig. 1 Schematic diagram of the gas extraction line for CO₂ and O₂ isotope exchange in presence of heated platinum.

For O₂–CO₂ exchange nearly equal amount of the two gases (to ensure equal number of oxygen atoms in the two gas phases) are taken in the heated quartz tube section containing the platinum strips. The protocol is as follows. First, a sample of CO₂ is taken from an attached flask in the line. After taking an aliquot from this gas (to be used as the initial CO₂ sample) a section of the line (with known volume V) is filled with this gas and its amount (n mole) is calculated using pressure (P) read by a Baratron gauge at the room temperature (T) and assuming ideal gas law: PV = nRT. This aliquot of CO₂ is transferred completely to the quartz segment by cryogenically trapping in the cold finger and is isolated by closing the lower stop cock. Next, a sample of O₂ is taken from a second flask into the line including a transfer volume (between two stop cocks shown in Fig. 1) connected to the reactor. The pressure of O₂ is adjusted such that its amount in the transfer volume is close to that of the CO₂. After waiting for about 5 minutes to ensure uniformity the upper stop-cock is closed and the lower stop cock is opened to commence the reaction. The rest of the gas in the line is collected as the initial O₂ sample. Only the platinum containing part of the quartz tube is heated to the desired temperature and the rest of the reaction section remains at room temperature. We have earlier checked that in about two hours the exchange is complete and a steady state is attained (at a temperature of 680 °C) despite the temperature gradient in the reactor.⁴ For CO–CO₂ exchange exactly similar procedure was adopted except that the amount of CO was usually taken to be twice that of CO₂ to ensure equimolar O-atom content of the reacting gases.

After the exchange the CO₂ fraction is extracted from either O₂ or CO by passing the gas mixture through a cold finger (kept at −196 °C) while collecting the O₂ or CO in a sample tube (kept at the sampling port) containing a few pellets of molecular sieve (MS13X) and cooled to liquid nitrogen temperature.^7,8 Later on, the CO₂ fraction is also collected in a sample tube by liquid nitrogen. The gases (O₂, CO₂ and CO) are analyzed for isotope ratios expressed in the standard δ-notation defined below:

δ¹⁸O = [R¹⁸_sam/R¹⁸_std − 1] × 1000 in per mil (‰)

where R¹⁸ = ¹⁸O/¹⁶O isotope ratio and the subscripts ‘sam’ and ‘std’ refer to sample and standard respectively. Similar definitions apply for δ¹⁷O and δ¹³C where R equals ¹⁷O/¹⁶O and ¹³C/¹²C respectively. The standard for oxygen isotope ratios is Vienna Standard Mean Ocean Water (VSMOW) and for carbon isotope ratio it is Vienna Pee Dee Belemnite (VPDB) a carbonate fossil powder.⁹ Both these standards are distributed by International Atomic Energy Agency, Vienna with advice regarding their handling.⁹ If the sample oxygen has R¹⁸ value more than that of VSMOW (0.002004) the δ¹⁸O value is positive. Thus, VSMOW provides a common scale for comparison of oxygen isotope ratios. Clumped isotope ratio is expressed in terms of isotopologue ratio Δ₄₇ defined by

Δ₄₇ = [R⁴⁷_sam/R⁴⁷_stochastic − 1] × 1000

where R⁴⁷ refers to the ratio ¹⁶O¹³C¹⁸O/¹⁶O¹²C¹⁶O i.e., ionic mass ratio 47/44 in analysed CO₂. As mentioned before, stochastic distribution is practically achieved by heating CO₂ at 1000 °C for more than two hours and serves as internal standard.¹ For example, if the sample CO₂ is equilibrated at low temperature of 25 °C its R⁴⁷ value would be higher than the stochastic value.

Following the exchange, the CO₂ gas is analyzed for δ¹³C and δ¹⁸O and the O₂ gas is analyzed for δ¹⁷O and δ¹⁸O using a MAT 253 mass spectrometer (MS) operated in the dual inlet mode. We thus obtain the isotopic composition of the two reacting gases (O₂–CO₂ pair or CO–CO₂ pair) as well as the final composition of the product gases after the exchange. The MS working gases are calibrated relative to the international standards VPDB and VSMOW by standard techniques^9,10 and also through inter-laboratory (Academia Sinica and Utrecht University) cross-calibration. The δ-values (in ‰) of our local standards are as follows: Linde CO₂ with δ¹³C = −4.411 (VPDB), δ¹⁸O = 24.592 (VSMOW); OASIS CO with δ¹³C = −492.7 (VPDB), δ¹⁸O = −297.5 (VSMOW) and OASIS O₂ with δ¹⁷O = −11.504 (VSMOW), δ¹⁸O = −21.720 (VSMOW). Using these values of the three MS working standards all raw δ-values are converted to VPDB and VSMOW scale. The analytical precisions in the mass spectrometric analysis are typically 0.011, 0.023 and 0.011 for δ¹³C, δ¹⁷O and δ¹⁸O, respectively. However, reproducibility of analysis of samples (taking into account the processing protocols) based on repeat analysis of 12 samples have typical standard deviation (1σ) values: 0.17, 0.11 and 0.17‰ for δ¹³C, δ¹⁷O and δ¹⁸O respectively. Various uncontrolled kinetic fractionations taking place in the reactor and those associated with gas separation and absorption–desorption in molecular sieves add up to give higher dispersion. This is evident from the standard deviation of the oxygen isotope anomaly values (defined as Δ¹⁷O = δ¹⁷O − 0.516 × δ¹⁸O) which is about 0.028‰ since the ¹⁷O/¹⁶O and ¹⁸O/¹⁶O fractionations cancel out in the calculation of anomaly. This is the reason for getting a high precision (±0.007‰) in the estimation of Δ¹⁷O of CO₂ by Mahata et al.⁵ It is to be noted that any leakage in the vacuum system or presence of minute quantities of water would have deleterious effects on the results since that would mean extra sources of oxygen with widely different isotope ratios that can exchange with CO₂. However, as discussed in Mahata et al.⁵ the blank experiments did not reveal any such exchange. CO₂ samples were heated in the same set up with platinum but without oxygen and the changes in the δ¹⁸O values were less than 0.04‰ indicating insignificant contribution from any contaminant.

In addition to the IISc experiments, a second of set of experiments (using O₂–CO₂ pair) was performed in Academia Sinica using the system described in Mahata et al.⁴ and following essentially the same protocol as given above. However, there is a notable difference between the two setups. The relative δ-values of the O₂ and CO₂ in the two laboratories were quite different. In case of IISc experiments, the δ¹⁸O values (in VSMOW scale) are, O₂: −21.54 ± 0.18; CO₂: 25.16 ± 0.17‰. But for Academia Sinica the respective values are (for three different pairs): pair 1: [O₂: 32.75 ± 0.20; CO₂: 15.41 ± 0.05‰]; pair 2: [O₂: 32.79 ± 0.20; CO₂: 16.38 ± 0.05‰]; and pair 3: [O₂: 32.61 ± 0.20; CO₂: 20.82 ± 0.05‰]. These choices with different sets of values allowed us to check if the initial compositions have any influence on the degree of homogenization. As the results will show, it does not matter whether the δ¹⁸O of initial O₂ is less or more than the δ¹⁸O of initial CO₂ or how different the values are to the final homogenization state. The data obtained in the Academia Sinica system are added to the IISc data set and are given in ESI† tables and figures while discussing the results of Δ₄₇ analysis. It is evident that there is fair degree of consistency in the two sets of results.

C. Results

The analytical data related to the catalysis induced exchange (for IISc experiments only) are given in ESI Table 1† for O₂–CO₂ and ESI Table 2† (for δ¹⁸O) and ESI Table 3† (for δ¹³C) in case of CO–CO₂ exchange. Note that in these tables the simple d-value has been replaced by logarithmic definition, following Miller (2002)¹¹ who suggests that in the oxygen three isotope plot if we use the logarithmic definition of δ-values i.e., 1000 × ln(1 + δ/1000) the slope of the regression line is invariant to the choice of the reference material and is especially preferred when large δ-differences occur like in Fig. 2. The δ¹³C values for the O₂–CO₂ exchange case are not shown because, as expected, the initial and final values remain the same within ±0.005‰. This also proves that there is no mass loss or gain in the system as that would surely alter the isotope ratio of carbon. ESI Table 1† shows that as a result of exchange between O₂ and CO₂ (taken in nearly equimolar proportions and at temperatures ranging between 600 and 800 °C) the δ¹⁸O of O₂ phase increases by an amount close to the amount by which δ¹⁸O of CO₂ phase decreases (schematically shown in Fig. 2). Since the number of atoms of O₂ and CO₂ are nearly equal, mass balance requires that decrease in δ¹⁸O value of CO₂ must equal the increase in δ¹⁸O value of O₂. The δ¹⁸O value of O₂ at the start of the process is −21.8‰ (VSMOW) and that of CO₂ is +24.8‰ (VSMOW). The exchange brings their δ¹⁸O values close to each other within a few ‰ depending on the time and the temperature. Given a rather long time (∼25 h) and temperature of 700 °C the two δ¹⁸O values become almost identical to within 0.3‰ (sample 7 and Fig. 2). This demonstrates that the exchange leads to homogenization of isotope ratios at a particular time–temperature combination. It is obvious that the post-exchange δ¹⁸O values of the two gases are not what one expects in equilibrium isotope exchange reaction between O₂ and CO₂. In that case, at 700 °C CO₂ would have δ¹⁸O value higher by about 6.4‰ as can be derived from the ratio of β-values (see the table of reduced partition function ratios in Richet et al.¹²). We therefore characterize the process as isotopic mixing or homogenization rather than isotopic exchange in the sense of equilibrium thermodynamics.¹³ Platinum catalyst plays an important role in bringing about the homogenization. This is clearly evident from the results of the samples 20 and 21 where no platinum was used and the changes in δ¹⁸O values were minimal. We also note that the magnitude of exchange reduces with temperature and at ∼300 °C the change in δ¹⁸O value is close to zero (Fig. 3).

Fig. 2 Isotopic exchange between O₂ and CO₂ with μmole ratio of 47/49 in presence of platinum catalyst at 700 °C. The starting composition of O₂ (−21.8, −11.5) increases to (2.4, 1.0) after exchange while the composition of CO₂ decreases correspondingly from (24.8, 13.2) to (2.1, 1.4). At the final stage the two delta values should be identical but we find that O₂ δ¹⁸O is little higher (∼0.3) than that of CO₂. This is due to the interfering exchange with quartz (δ¹⁸O ∼ 10) which increases the δ¹⁸O of final O₂ while reducing the δ¹⁸O of final CO₂ (see Fig. 3). Numbers are in ‰ relative to VSMOW.

Fig. 3 Changes in δ¹⁸O of O₂ and CO₂ due to exchange in presence of platinum. In the absence of platinum the quartz tube at high temperature (700 to 800 °C) itself interacts with both O₂ and CO₂ and increases the δ¹⁸O of O₂ and decreases the δ¹⁸O of CO₂. This should be due to their relative difference; the δ¹⁸O of our quartz tube is about 10‰ (VSMOW) higher by about 12‰ than δ¹⁸O of O₂ and lower by about 15 than δ¹⁸O of CO₂.

As the temperature increases the δ¹⁸O value of each gas changes and reaches a plateau at about 600 °C. Additionally, with increasing temperature (for ∼2 hours) the difference between the two δ¹⁸O values reduces (Fig. 4). These two observations seem to display the well-known Sabatier effect showing temperature variation in the catalytic efficiency of the platinum.¹⁴ Fig. 5 shows that for both IISc and Academia Sinica experiments, the gas compositions approach each other and the final values are close (within 0.3‰) when the time given is large enough for the exchange reaction (∼20 hours for sample ID 15).

Fig. 4 The difference in δ¹⁸O values of CO₂ and O₂. The initial difference is large (24.9 + 21.9 = 46.8) but the difference reduces after exchange and the reduction is a function of temperature. At 300 °C the difference does not change but at 800 °C the difference reduces to almost zero. This could be due to the temperature variation of catalytic efficiency of platinum.

Fig. 5 The figure shows that for both IISc and Sinica experiments, the gas compositions approach each other and the final values are very close (within 0.3‰) when the time given is large (∼20 hours for sample ID 15). Sample ID's are nominal sequential numbers just to show the δ-values of gas pairs approaching each other.

In case of CO–CO₂ exchange the isotopes of both carbon and oxygen can change (Fig. 6) unlike O₂–CO₂ case where carbon isotopes remain invariant. However, in case of CO/CO₂ close to two, equal number of O-atoms from CO and CO₂ are available whereas C-atoms from CO were twice that from CO₂. Therefore, it is expected that the magnitude of exchange in the two isotope ratios would not be the same. Our CO tank happens to be extremely depleted in both ¹⁸O and ¹³C. We calibrated it by converting CO to CO₂ by discharge and measuring the product CO₂ against our tank CO₂ [δ¹⁸O = +24.8 (VSMOW) and δ¹³C = −4.411 (VPDB) in ‰]. The values for CO found by this method are δ¹⁸O = −280.2‰ (VSMOW) and δ¹³C = −495.0‰ (VPDB). We should mention that the discharge of CO does not convert 100% of the carbon to CO₂ and there is a large carbon deposit due to the disproportionation reaction. This can potentially change the δ¹³C value of the CO₂ product from that of the CO. Therefore, we checked the assignment of the δ¹⁸O and δ¹³C values by cross-calibration with the Utrecht University laboratory of Prof. Thomas Rockmann and found them to be consistent. However, the extremely negative δ¹³C value of the tank CO introduces non-linearity issue and the assigned value may have large uncertainty.

Fig. 6 Changes in δ¹³C of CO and CO₂ due to exchange in presence of platinum with time. The time required to reach a steady state (about 20 h) is much higher compared to O₂–CO₂ exchange time (about 2 h) on platinum (based on data in ESI Table 2†).

Despite the above-mentioned problem of calibration we find that the negative changes of δ¹⁸O and δ¹³C values in CO₂ match the corresponding increases in CO reasonably well when corrected for the atom ratio. For blank runs in both cases (only CO or only CO₂ with platinum in quartz tube) the changes are negligible. However, when both CO and CO₂ are present in the quartz tube, even without platinum there is a large change (in both δ¹⁸O and δ¹³C; see ESI Tables 2 and 3†) unlike the case of O₂–CO₂ (see ESI Table 1†).

This change is somewhat less than that obtained with platinum (compare samples 3, 4 and 5) but is still appreciably large. It seems that the heated quartz surface alone can act as catalyst for the CO–CO₂ exchange at 700 °C. However, this result has to be investigated further with a more positive (in δ¹⁸O and δ¹³C) CO tank. For oxygen isotope ratio the changes in the two gases are nearly equal as expected (see the 1 : 1 line in Fig. 7) but for carbon isotope ratio the change in CO is less than that in CO₂ (see Fig. 8) even when the latter is corrected for the atom ratio (i.e., multiplied by CO₂/CO molar ratio to normalize the change in CO₂ to the mole number of CO). A comparison of the numbers in column 6 and 9 (ESI Table 3†) show that the atom ratio corrections to the CO₂ δ¹³C change cannot compensate for the difference fully and apparently the carbon isotope balance is not satisfied within 10‰. We suspect that this could be due to the problem of calibration since the δ¹³C value of CO is −495‰ as against that of CO₂ being −4.4‰. A change of 10‰ in CO₂ may not show up as equivalent change in CO (corrected for mass ratio) unless the two scales are consistent to a few per mil level. We conclude that the CO–CO₂ exchange results are only indicative at this stage due to the limitation imposed by the hugely different isotope ratios of the CO tank that we have.

Fig. 7 The change (in absolute magnitude) in δ¹⁸O of CO is nearly equal to the change in δ¹⁸O of CO₂ leading to a 1 : 1 slope for the two changes.

Fig. 8 The δ¹³C change in CO is less than that in CO₂ even when corrected for the amount difference (since less the amount more is the change). In the present experiments CO amount was always more than CO and it seems that for δ¹³C changes the amount correction cannot compensate for the difference fully and we do not get a 1 : 1 slope.

D. Discussion

A model that can account for the above observations is one that involves catalyzed breakdown and exchange of molecules and atoms on the platinum surface. For CO₂–O₂ exchange we visualize the exchange reaction to occur as:

CO₂ → CO_2ads, CO_2abs → CO_ads + O_ads, O₂ → O_2ads, O_2ads → O_ads + O_ads

The catalytic reaction starts by adsorption of the reactants CO₂ and O₂ to the platinum catalyst spontaneously. For most metal surfaces the O₂ molecules dissociate upon adsorption. The oxygen isotopic exchange studies by several researchers (Bedrane et al., 2005; Sandler and Durigon, 1968; Chin et al., 2011; Boreskov, 1964)^15–18 provide ample proof of dissociative adsorption of O₂ by metal particles. The reason is two-fold: firstly, the free electrons from the metal fill the anti-bonding orbital of O₂ molecule and secondly, the bonding of the two atoms to the surface is energetically more favourable than that of one molecule.¹⁹ The CO₂* molecules (a small fraction of the total) can also dissociate to CO* and O* though with a much lower rate (energy barrier ∼ 125 kJ mol⁻¹) than O₂ dissociation (which does not require activation energy). In contrast, the reverse reaction of CO* + O* → CO₂* has a low energy barrier 50 to 100 kJ mol⁻¹ and can take place faster.¹⁴ The O*-atoms (and CO* molecules) move around on the platinum surface in a two-dimensional Brownian motion with a mean residence time of about 60 msec.^20,21 Consequently, a reservoir of adsorbed O*-atoms builds upon the platinum surface by contribution from both CO₂* and O₂* and they are mixed up due to migration of the O*-atoms from their original site of adsorption, thus losing their source identity (or isotopic composition). Therefore, when the molecules form again by reverse reactions, the O*-atoms are derived from a completely homogenized reservoir (with mass balance composition) and the oxygen isotopic compositions of the resultant O₂ and CO₂ are almost equal. In a comparable case, a homogeneous isotope mixing in the product CO₂ formed by catalytic CO oxidation on platinum surface was inferred by earlier studies.^22,23 Finally, the product molecule CO₂ being relatively unreactive with the platinum surface desorb from the catalyst surface in an endothermic step. The O₂ and CO₂ molecules in the gas phase soon attain an isotopic steady state with the adsorbed species of the same.

Due to this set of processes there is a reservoir of CO and O along with CO₂ and O₂ (all in the adsorbed state on the hot platinum surface) which can interact with each other. Depending on the temperature and time and the rates of the various processes a steady state is reached where the oxygen isotope ratios become indistinguishable in the two phases (CO₂ and O₂). From the results it is clear that in the present experimental configuration it takes about 2 hours at 700 °C to achieve this steady state. The isotope ratios are identical in the two gas phases (CO₂ and O₂) in this case and that is possible only if a random distribution is present where the zero point energy difference of the various species of molecules does not play any role in controlling the steady state composition. The clumped isotope results should show this clearly since the ¹³C¹⁸O¹⁶O abundance (or clumpiness) would be higher than dictated by random distribution only if the zero point energy of this isotopologue plays a role. For the purpose of present discussion the clumpiness Δ′₄₇ is defined here with reference to the composition of the gas before any exchange, that is as: Δ′₄₇ = Δ₄₇ (post-exchange) − Δ₄₇ (pre-exchange). With this definition, we note that in the present case of O₂–CO₂ exchange the clumpiness decreases with increase in temperature (Fig. 9) and even at a low temperature of ∼600 °C it attains the composition of the CO₂ gas heated to 1000 °C.

Fig. 9 Clumpiness of CO₂ [defined as the change: Δ₄₇ (post-exchange) − Δ₄₇ (pre-exchange) in ‰ VSMOW] as a function of temperature (ESI Table S1†). Increased negative values indicate more stochasticity. Catalytic exchange of CO₂ with O₂ at temperatures above 600 °C (for 2 h or more) induces clumpiness close to that of the CO₂ gas heated to 1000 °C without O₂ and platinum. At 300 °C the clumpiness is close to zero (less stochasticity). Symbols: IISc – Indian Institute of Science, SINICA – Academia Sinica, Taiwan, HG – Heated Gas CO₂ at 1000 °C. Pt – Platinum. Uncertainty in the values is typically about ±0.1‰.

The plot also suggests that for a range of temperature (500 to 1100 °C) the randomness obtained for the heated gas (without platinum and without O₂) has a broad peak (at ∼800 °C, in a resolution of 100 °C) after which it decreases slightly. This Gaussian type behaviour is the feature of catalytic Sabatier effect¹⁰ and the values provide a measure of scrambling brought about in CO₂ by contact with heated quartz inducing a catalytic reorganization of the isotopes. However, there is a large uncertainty in the values (∼±0.1‰) and this issue should be investigated further. If confirmed by future experiments, this could mean that heated gas (as obtained experimentally at 1000 °C) still has some amount of residual clumpiness. However, that does not diminish the usefulness of the heated gas as a convenient means of defining the zero of the clumpiness scale and the conventional clumped isotope thermometry based on this scale is perfectly reasonable. Sample numbers 20 and 21 refer to two cases of O₂–CO₂ reaction when there was no platinum in the reaction chamber. These data (ESI Table 1†) show that without platinum the two gases do not exchange their isotopes. However, a sample of CO₂ heated without O₂ and without platinum attains the randomness as before. This (ESI Table 1†) result probably shows that CO₂ in presence of heated quartz alone may achieve randomness by a process which is quite different from the homogenization process due to catalyzed exchange with O₂ on heated platinum. But both lead to randomness in the CO₂ molecules in the reactor. This could be the reason that in case of CO–CO₂ experiments there is reasonable exchange even when they are in heated quartz tube without any platinum.

The proposed mechanism is further supported by the results of CO–CO₂ exchange carried out in a similar way using platinum. In this case the adsorbed species are CO, CO₂ and O and the mixing takes place between CO coming from CO₂ breakdown and the original CO. The rate constants are presumably much slower here and therefore the O-isotope ratios in the post-exchange gases are not equal in two hours of exchange. For 25 hours exchange at T = 700 °C (sample CO_CO₂_5) the δ¹⁸O values (in ‰ VSMOW) obtained are: −94.3 (CO₂) and −179.3 (CO). These values differ widely (by about 40‰) from the number −132.6‰, the estimated mean δ¹⁸O value (weighted by the respective mole amounts) if the two gases were mixed uniformly. For O-isotopes the mixing is mediated by the two sources of oxygen, CO molecules and O-atoms, but that is not so in case of C-isotopes (mediated only by CO). Therefore, the change (brought about in a given time and temperature) in case of δ¹³C value is much smaller. For Sample 5 the δ¹³C change in CO₂ is about 40‰ which is lower than the δ¹⁸O change by more than a factor of two. The clumpiness of the CO₂ samples from CO–CO₂ exchange could not be measured due to slight contamination of CO (high δ⁴⁹ value). In future, we plan to clean the CO₂ by GC and analyze it for Δ₄₇. We note that there is another means of checking the randomness in the exchange of CO₂–O₂. If the suggested mechanism is correct the isotopes in the O₂ phase are also expected to be randomly distributed which can be checked by measuring the ¹⁸O¹⁸O abundance and comparing that to the one expected for random distribution.

E. Conclusions

O₂ and CO₂ do not react in simple gas phase but in presence of hot platinum they exchange their oxygen isotopes rapidly. It is known that this exchange carried out at ∼700 °C in a platinum containing quartz tube reaches a steady state in about two hours when the oxygen isotope ratios of the two gases become equal to each other within a few ‰. It is still not known how this exchange occurs in the microscopic scale. To get an insight, we carried out clumped isotope analysis of the post-exchange CO₂ gas (following standard procedure outlined by the CALTECH group) which measures the abundance of ¹³C¹⁸O¹⁶O isotopologue (mass 47) in a sample of CO₂ and calculates the deficiency of this species from the equilibrium thermodynamic prediction in terms of isotope anomaly Δ₄₇. The magnitude of Δ₄₇ gives the degree of randomness in the isotopic distribution of ¹³C, ¹⁸O and ¹⁶O to form the mass 47 species and has a value of nearly −1‰ (in the CALTECH scale) for a completely scrambled distribution. We find that the post-exchange CO₂ has Δ₄₇ value close to that expected from a random distribution. This suggests that two processes (exchange on platinum and scrambling due to heating) occur in tandem and can be explained simply as due to mixing of O-atoms and CO molecules formed by breakdown of the O₂ and CO₂ molecules to generate two homogeneous reservoirs on the platinum surface. We suspect that a similar process occurs on the quartz surface also but without any O₂ adsorption because O₂ and CO₂ do not exchange in a quartz tube alone. We also found that CO and CO₂ gas pairs exchange their oxygen and carbon isotopes which is only possible if CO₂ breakdown to CO + O occurs in the adsorbed state and the CO gas also gets adsorbed and mixes with the CO from CO₂ on the platinum surface, albeit with less efficiency.

Acknowledgements

This work was supported by Divecha Centre for Climate Change. This manuscript was improved by thoughtful feedback from Prof. John Eiler. SKB acknowledges Academia Sinica, Taiwan and Centre for Earth Sciences, Indian Institute of Science, Bangalore, India for support and facilities.

Notes and references

1. J. M. Eiler and E. Schauble, Geochim. Cosmochim. Acta, 2004, 68, 4767–4777.
2. Z. Wang, E. A. Schauble and J. M. Eiler, Geochim. Cosmochim. Acta, 2004, 68, 4779–4797.
3. K. J. Dennis, H. P. Affek, B. H. Passey, D. P. Schrag and J. M. Eiler, Geochim. Cosmochim. Acta, 2011, 75, 7117–7131.
4. S. Mahata, S. K. Bhattacharya, C.-H. Wang and M.-C. Liang, Anal. Chem., 2013, 85, 6894–6901.
5. S. Mahata, S. K. Bhattacharya and M.-C. Liang, Rapid Commun. Mass Spectrom., 2016, 30, 119–131.
6. T. Osaki and T. Mori, React. Kinet. Catal. Lett., 2006, 89, 333–339.
7. O. Abe, Rapid Commun. Mass Spectrom., 2008, 22, 2510–2514.
8. I. Ahn, M. Kusakabe and J. I. Lee, Rapid Commun. Mass Spectrom., 2014, 28, 1321–1328.
9. R. Gonfiantini, W. Stichler and K. Rozanski, IAEA-TECDOC, 1995, 825, 13–29.
10. W. A. Brand, J. Mass Spectrom., 1996, 31, 225–235.
11. M. F. Miller, Geochim. Cosmochim. Acta, 2002, 66, 1881–1889.
12. P. Richet, a. Y. Bottinga and M. Javoy, Annu. Rev. Earth Planet. Sci., 1977, 5, 65–110.
13. H. C. Urey, J. Chem. Soc., 1947, 562–581.
14. I. Chorkendorff and J. W. Niemantsverdriet, Concepts of modern catalysis and kinetics, John Wiley & Sons, 2006.
15. S. Bedrane, C. Descorme and D. Duprez, Appl. Catal., A, 2005, 289, 90–96.
16. Y. L. Sandler and D. Durigon, J. Phys. Chem., 1968, 72, 1051–1057.
17. Y.-H. Chin, C. Buda, M. Neurock and E. Iglesia, J. Am. Chem. Soc., 2011, 133, 15958–15978.
18. G. Boreskov, Adv. Catal., 1964, 15, 285–339.
19. N. Niemantsverdriet and B. E. Nieuwenhuys, Catalysis and Surface Science, UNESCO-EOLSS, Paris, France, 2014.
20. G. Ertl, Angew. Chem., Int. Ed., 2008, 47, 3524–3535.
21. S. Renisch, R. Schuster, J. Wintterlin and G. Ertl, Phys. Rev. Lett., 1999, 82, 3839.
22. S. Akhter and J. White, Surf. Sci., 1986, 171, 527–542.
23. J. Wintterlin, S. Völkening, T. Janssens, T. Zambelli and G. Ertl, Science, 1997, 278, 1931–1934.

---

Footnote

† Electronic supplementary information (ESI) available: ESI Tables 1–3. See DOI: 10.1039/c6ra08286f

---