Ozone Cross-Section Measurement by Gas Phase ... - ACS Publications

Article pubs.acs.org/ac

Ozone Cross-Section Measurement by Gas Phase Titration Joel̈ e Viallon,* Philippe Moussay, Edgar Flores, and Robert I. Wielgosz Bureau International des Poids et Mesures (BIPM), Pavillon de Breteuil, F-92312 Sèvres Cedex, France ABSTRACT: Elevated values of ground-level ozone damage health, vegetation, and building materials and are the subject of air quality regulations. Levels are monitored by networks using mostly ultraviolet (UV) absorption instruments, with traceability to standard reference photometers, relying on the UV absorption of ozone at the 253.65 nm line of mercury. We have redetermined the ozone cross-section at this wavelength based on gas phase titration (GPT) measurements. This is a well-known chemical method using the reaction of ozone (O3) with nitrogen monoxide (NO) resulting in nitrogen dioxide (NO2) and oxygen (O2). The BIPM GPT facility uses state-of-the-art flow measurement, chemiluminescence for NO concentration measurements, a cavity phase shift analyzer (CAPS) for NO2 measurements, and a UV ozone analyzer. The titration experiment is performed over the concentration range 100−500 nmol/mol, with NO and NO2 reactants/calibrants diluted down from standards with nominal mole fractions of 50 μmol/mol. Accurate measurements of NO, NO2, and O3 mole fractions allow the calculation of ozone absorption cross section values at 253.65 nm, and we report a value of 11.24 × 10−18 cm2 molecule−1 with a relative expanded uncertainty of 1.8% (coverage factor k = 2) based on nitrogen monoxide titration values and a value of 11.22 × 10−18 cm2 molecule−1 with a relative expanded uncertainty of 1.4% (coverage factor k = 2) based on nitrogen dioxide titration values. The excellent agreement between these values and recently published absorption cross-section measurements directly on pure ozone provide strong evidence for revising the conventionally accepted value of ozone cross section at 253.65 nm.

A

ambient conditions to the virtual exclusion of any potentially competing gas phase chemistry involving NO, O3, NO2, and O2. In addition, the reaction has a simple 1:1 stoichiometry, and in a well-designed flow reactor, the reaction may be driven essentially to completion (>99.9%) in a relatively short residence time, less than 30 s. It is therefore recognized as an alternative secondary measurement principle and method to calibrate ambient ozone analyzers in the International Standard 15337:2009.6 This method can also be seen as a metrological triangle in which three different traceability chains are being compared: NO, NO2, and O3. NO reacted and NO2 gain are traceable to reference materials which are independently prepared by gravimetry. The O3 reacted, measured by an ambient ozone analyzer, is traceable to a standard reference photometer (SRP) for ozone which is considered as a primary standard for numerous national and international ozone monitoring networks.7 This instrument relies on the measurement of the amount of UV light emitted at 253.65 nm by a mercury lamp absorbed by the sample of ozone in dry air in the instrument’s gas cells. The ozone concentration is therefore anchored to the value of the ozone absorption cross-section at the mercury line wavelength, which has been conventionally adopted to be the value measured by Hearn in 1961,8 equal to 11.476 × 10−18 cm2 molecule−1 with a relative expanded uncertainty of 2.12%.9

tmospheric ozone levels are of worldwide concern at ground level where increases in concentration have detrimental effects on public health and the environment, and international effort is focused on monitoring and reducing surface ozone concentrations. There is a strong appreciation for the requirement of comparable measurements in this field. The principal method of ozone concentration determination is ultraviolet photometry, for which there is an international standard1 and where calibration against primary UV photometers is required. This method was introduced in the 1970s by the U.S. National Bureau of Standards (now National Institute for Standards and Technologies, NIST) and the U.S. Environmental Protection Agency after a series of studies demonstrated its simplicity of application with a reasonable relative uncertainty of 2.5%.2 Those studies were motivated by the need to replace the former recommended method, which used a neutral buffered 1% KI solution (“1% NBKI” method).3 Another alternative method that was also considered at that time was the gas phase titration (GPT) of ozone with nitrogen monoxide (NO), already proposed by Saltzman et al. as the “nitrogen dioxide equivalent method,” and further studied for its chemiluminescent emission by Clyne et al.4 The titration of ozone with nitrogen monoxide is a bimolecular reaction leading to the formation of nitrogen dioxide and molecular oxygen: NO + O3 → NO2 + O2

with a rate constant k = 1.9 × 10−14 cm3 molecule−1 s−1 at 298 K.5 This reaction readily lends itself to exploitation as a means of measuring ozone, as the reaction proceeds very rapidly under © XXXX American Chemical Society

Received: August 23, 2016 Accepted: October 11, 2016

A

DOI: 10.1021/acs.analchem.6b03299 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 1. Measurement setup. EPCi, pressure controllers; Fi, mass flow controllers and flow measurement devices (molbloc/molblox); NV, needle valve; O3-GEN, ozone generator; RV, reaction vessel; Vi, valves.

sieve and a methane burner in a commercial air cleaning system (AADCO 737-R). When no O3 was produced, the same air was directed to the reaction vessel bypassing the ozone generator to be mixed with the mixture of NO and/or NO2 in nitrogen. The reaction vessel was a custom-made glass tube of 8 mm diameter and an internal volume close to 1.2 L. The internal volume of the RV had been chosen to have a residence time of 30 s, sufficient to reach reaction completion. The RV could be bypassed with a short length of electropolished 316L stainless steel tubing with a variable restriction set to balance the backpressure generated by the RV as measured inside the ozone generator. Potential losses of each gas inside the reaction vessel were estimated by comparison of two paths for the flow of gas: via the reaction vessel and bypassing it (between valves V10 and V11), at various concentrations. No difference could be observed when flowing NO or NO2, and when flowing ozone, maximum losses of 1 nmol mol−1 have been observed at 500 nmol mol−1. At every step of the titration, all flow rates were adjusted so as to maintain a constant flow rate in the reaction vessel and a constant oxygen level in the mixture, between 19.93% and 20.02%. The system was fully automatic, making use of pneumatically actuated twin valves (V1 to V14). The facility was designed to reduce dead volumes as far as possible. All gas lines were 1/4″ or 1/8″ electropolished stainless steel. Further details of the ozone generator and the gas analyzers are provided below. O3 Generator. Ozone in air was generated with a modified Environics S-6100 Ozone Generator, producing ozone from air by photolysis of molecular oxygen after absorption of the radiation at 185 nm emitted by a mercury lamp. The modification consisted in changing all Teflon lines inside the instrument with stainless steel lines, in order to avoid as much as possible permeation of water inside the gas lines. The stability of the ozone concentration was tested during measurements performed over 20 min or more at every titration values. A maximum variation of 0.5 nmol mol−1 of O3 was observed.

In this paper, after the description of the experimental setup in the first section, the metrological traceability of the three measurable compounds of the GPT is described in the second section. The excellent agreement of changes in NO and NO2 concentrations during titration are demonstrated in the third section, allowing the ozone absorption cross sections in the UV to be calculated based on the GPT measurements anchored to NO and NO2.

■

MEASUREMENT SETUP AND PROCESS In the setup presented here, GPT is executed by reacting the ozone with an excess of NO, under kinetic conditions such that the reaction proceeds to completion, so that the reacted NO and O3 can be considered equal to the gain of NO2: −Δx NO = −ΔxO3 = +Δx NO2

(1) −1

where Δx = xfinal − xinitial in nmol mol and xfinal and xinitial are the final and initial mole fractions of the compound, in nmol mol−1. Figure 1 shows a schematic of the setup. Input gases to the titration are represented in the left part of the scheme, the reaction vessel (RV) in the middle, and outputs gases and corresponding analyzers on the right part. The system aimed at titrating between 100 nmol mol−1 and 600 nmol mol−1 of O3 with NO in excess. As NO and NO2 in nitrogen mixtures are not stable enough at those low levels, standard mixtures at the micromole per mole level were selected and diluted down with nitrogen, then mixed with O3/ air or air only just before the reaction vessel, and finally analyzed simultaneously with the NO, NO2, and O3 analyzers. NO/nitrogen was introduced in line 1, NO2/nitrogen in line 2, O3/air or air only in line 3, and nitrogen only in line 4 (line number i can be identified by the mass control number Fi on Figure 1). Flows in each of the four lines were measured by commercial laminar flow measurement devices named molbloc/molbox in order to get simultaneous measurements on all lines, and controlled with mass flow controllers (Fi). O3 was generated with a commercial ozone generator (O3-GEN) from compressed air-dried by adsorption and cleaned by molecular B


Article

Analytical Chemistry Analyzers. NO was analyzed with a chemiluminescence analyzer, Thermo Environmental Instruments Inc. model 42C. This well-known technique involves a gas-phase ozone-induced chemiluminescence detection of NO.4 NO2 was analyzed with a cavity attenuated phase shift (CAPS) analyzer, Environment S.A. model AS32M. The instrument detects NO2 in its cell from its absorption of light emitted by a broadband incoherent LED at 430 nm.10 It was selected for its selectivity of NO2, in particular to avoid interferences with nitric acid (HNO3), a common impurity in NO2 in air standard mixtures as discussed later. O3 was analyzed with a dual cell UV absorption analyzer, Thermo Environmental Instruments Inc. model 49C, described for example by Zucco et al.11 This analyzer’s aim was to provide the link between the ozone titration setup and the SRPs maintained at the BIPM. Measurement Process. A full experiment of ozone titration included the reaction between NO and O3 at 10 different mole fractions to cover the range 150−600 nmol mol−1. Each of these titrations formed one sequence, itself comprised of six phases to perform the necessary checks, calibrations, and corrections. Figure 2 displays the mole

(t6) O3 alone: This retains the same configuration as in phase t4, to estimate the stability of the amount of ozone injected in the titration step.

■

TRACEABILITY AND UNCERTAINTIES The measurement equation is very simple, reflecting the stoichiometric reaction eq 1. Calculations performed to deduce each of the three terms are detailed in this section, together with the traceability of the measurements and the uncertainties. For each term, a first section details which quantities were measured and which calculations were performed to obtain final results. All uncertainties displayed in tables were calculated using the law of propagation of uncertainties, as recommended in the Guide to the Expression of Uncertainties in the Measurement.12 O3 Reacted. O3 reacted during a titration step is deduced from the O3 mole fraction introduced in the reaction vessel from the ozone generator and corrected for small amounts of unreacted O3 measured during titration steps. As expected from the reaction kinetics, the amount of O3 left is inversely proportional to the amount of O3 introduced in the reaction vessel. Typically, less than 0.5 nmol mol−1 were measured for input values higher than 500 nmol mol−1, up to 4 nmol mol−1 at an input value of 100 nmol mol−1. Considering possible variations in the ozone amounts produced by the generator, the input values were themselves deduced from a mean between the values measured just before a titration step, in phase t4, and just after the step in phase t6. Finally, it was noticed that NO2 could produce small interferences in the ozone analyzer, estimated as 9 × 10−4 times the NO2 concentration. This was also accounted for by correcting the ozone analyzer response, to obtain the uncalibrated ozone reacted ΔyO3. Calibration factors of the ozone analyzers obtained during a precomparison with the BIPM SRPs were then applied. Traceability of O3 Measurements. Ozone mole fractions measured by the ozone analyzer are traceable to SRP27 maintained by the BIPM, one of the SRPs invented by Bass and Paul,13 produced by NIST in 2002. SRP27 is the reference of the international comparison BIPM.QM-K1 and has been compared with 21 national standards for ozone since 2007.14 O3 Analyzer Calibration. The ozone analyzer had been regularly calibrated against SRP27 and was again prior to this exercise, following the process decsribed in Viallon et al.15 Measurement data were fitted using a linear regression model with uncertainties in both axes, following the generalized leastsquares approach as described for example by Riu et al.16 Calculations were performed with software adapted from a Matlab package developed by the National Physical Laboratory (NPL) and implementing the algorithm described in the ISO standard 28037:2010.17 One possible drawback of such a method is an underestimation of the slope uncertainty due to the averaging effect of the regression analysis, which by default considers all points as independent. To deal with the case of correlated measurement results, as in the case of two measurements of SRP27 which are considered highly correlated, an alternative regression model can be used as previously developed in Bremser et al.18 In particular, the standard uncertainty on the ozone cross-section was included in the covariance term between two points. The effect of this can be observed in the relative uncertainty associated with the slope of the regression (next section), equal to 1%, where a typical

Figure 2. (Uncalibrated) Mole fraction of NO, NO2, and O3 during a typical measurement sequence. Vertical lines indicate the six phases of one sequence, happening at time ti.

fractions of NO, NO2, and O3 as measured by each corresponding analyzer during one typical titration sequence, in this example at about 600 nmol mol−1 of ozone. Each phase is listed below: (t1) Zero: During this phase, only nitrogen was flowed through each line in order to verify and compensate the zero drift of the NOx analyzers. (t2) NO span: A mixture of NO in air at 900 nmol mol−1 was injected in line 1 to calculate the response factor drift of the NO analyzer and correct for it. (t3) NOx analyzers calibration: NO from line 1 and NO2 from line 2, at mole fractions chosen to match the titration step, were mixed in the reaction vessel to calibrate the NOx analyzers. (t4) O3 alone: By flowing air in line 3 inside the ozone generator, O3 in air at the target mole fraction of the titration step was sent to the reaction vessel to get a first ozone value. (t5) Titration step: NO in excess was added to the O3 reactant; both mixed during the 30 s residence time in the reaction vessel and reacted to form NO2 and O2. C


Article


values. The calibration uncertainty was between 0.07% and 0.1% relative for flow rates above 50 mL min−1 and between 1.4% and 0.3% below. It was further combined with a type A uncertainty derived from the standard deviation of the mean of the last three readings, taken at 1 min intervals. Typical observed values were lower than 0.02 mL min−1 for flow rates below 100 mL min−1 and up to 0.7 mL min−1 for the highest flow rate of 2.2 L min−1 in line 3. The mole fraction of NO left after the reaction vessel, xfinal(NO), was deduced from measurements with the NO analyzer in phase t5, the titration phase, after proper calibration and correction for drifts as described later. Traceability of NO Standard Gas. The standard mixture of NO in N2 was a primary reference material produced by gravimetry by the NPL, certified at a mole fraction of 59.928 μmol mol−1 with a standard uncertainty of 0.051 μmol mol−1, considering only the gravimetric uncertainty. The confidence in the certified value of the standard was reinforced by comparison with a larger set of standard mixtures maintained by the BIPM using two other analyzers (based on chemiluminescence and UV absorption), following the procedure developed for NO/N2 standard comparisons by the BIPM in 2006.20 The last repeat of this comparison demonstrated an agreement of better than 0.05 μmol mol−1 between a set of 10 standards. In addition, analysis by a Fourier transformed infrared spectrometer (FTIR) equipped with a 45 m optical path length gas cell showed no detectable impurity. The levels of N2O and NO2, two compounds which had been detected in some standards during the 2006 comparison, were found to be below 15 nmol mol−1 and 5 nmol mol−1, respectively, in the standard mixture used for this exercise. NO Analyzer Calibration and Corrections. The NO analyzer was calibrated in line with the titration measurements. This method automatically compensates for small linear drifts in the instrument. Phase t3 of each titration step was used to provide one calibration point, consisting of the diluted NO mole fraction as a reference (eq 2), and the analyzer measured value corrected for the offset and span drifts. The offset drift was linearly interpolated at time t3 from two successive zero measurements (phase t1) of two sequences. Similarly, the span drift was interpolated from two successive span measurements (phase t2) of two sequences. These calculations were performed systematically, although the drifts were often negligible, with typical values of −4 × 10−5 nmol mol−1 min−1 for the offset drift and below −0.06 nmol mol−1 min−1 for the span drift at 900 nmol mol−1. The response of the NO analyzer was further averaged over the last three measurement points to provide the value ycal, and the associated standard deviation was used as standard uncertainty, with typical relative values of 0.1%. Each set of 10 calibration values (xr(NO), ycal) was fitted using a linear regression model with uncertainties in both axes, with the same software mentioned earlier. In that case, as each concentration point was obtained by dilution from the same standard mixture, the covariance between two calibration points i and j was described as follows:

value of that uncertainty without the cross-section would be 0.3%. Uncertainty Budget. Choosing one titration sequence as an example, all quantities and associated uncertainties impacting the O3 reacted were summarized in Table 1, with the Table 1. Typical Uncertainties on the O3 Reacted quantity y m

analyzer response (phase t4) analyzer responsem (phase t6) unreacted ozonem (phase t5) NO2 interferencesm uncalibrated O3 reacted calibration slope a1 calibration intercept a0 O3 reacted ΔxO3

yva

u(yv)b

598.87 598.40 0.49 0.56 598.70 1.0039 0.04 601.09

0.15 0.17 0.15 0.003 0.19 0.0109 0.28 6.46

a

yv typical value. bu(yv) standard uncertainty. All mole fractions are in nmol mol−1 and all flow rates in mL min−1. The exponent m indicates measured quantities.

distinction between measured quantities and results of calculations. The uncalibrated O3 reacted was calculated following the process explained in the previous section. The O3 reacted was obtained after application of the calibration parameters. Its combined standard uncertainty results from the law of propagation of uncertainties, including the regression parameters uncertainty. This table provides a good representation of the most important sources of uncertainty on the O3 reacted, namely the calibration slope, itself dominated by the uncertainty on the ozone absorption cross-section in the SRP, equal to 1.12%. The standard uncertainty associated with the analyzer response is the standard deviation calculated on the last 5 min of measurements. A comprehensive uncertainty analysis of the SRP was published in Viallon et al.9 NO Reacted. With NO being in excess with a factor closed to 2 compared to ozone, the NO reacted in the titration is the difference between NO introduced in the reaction vessel and NO left after the reaction vessel, as measured with the NO analyzer. The mole fraction of NO introduced in the reaction vessel, xr(NO), was deduced from the mole fraction of NO in the standard mixture and the dilution factor, following dilution with nitrogen from lines 1, 2, and 4 and with air from line 3: qv xr(NO) = xs(NO) 4 1 ∑i = 1 qvi (2) where xs(NO) is the mole fraction of NO in the standard mixture in nanomoles per mole and qvi is the flow rate in line i in milliliters per minute. The relative standard uncertainty associated with xr(NO) is a combination of the relative uncertainties of the standard mixture and of the dilution factor expressed as a ratio of flow rates in the right part of the equation, as described below. Flow Measurements. Accurate measurement of gas flow rates is crucial in this setup, as they represent a major source of uncertainty. All flow measurement devices were calibrated in situ versus a piston prover Bios ML-800, itself regularly calibrated by the Swiss Federal Office of Metrology and Accreditation (METAS). Molblocs are known to be matrix dependent.19 Therefore, they were calibrated with the same matrix gas as used during the titration experiment, i.e., air or nitrogen. Additional measurements using NO and NO2 in nitrogen mixtures demonstrated negligible impact on the

u(xi , xj) =

qv1, j qv1, i

u 2(xi)

(3)

where u(xi) is the standard uncertainty on the mole fraction of the more concentrated gas mixture in nanomoles per mole and qv1,i is the flow rate in line 1 during the calibration point i in milliliters per minute. D


Article

Analytical Chemistry Finally, the NO reacted −ΔxNO was calculated as the difference between the mole fraction of NO introduced in the reaction vessel during the titration step, xr(NO), and the mole fraction left after the reaction vessel xfinal(NO) deduced from the calibration line calculated above. The associated combined standard uncertainty was between 2.5 nmol mol−1 and 4.4 nmol mol−1, dominated by the uncertainty on the dilution factor in eq 2. Uncertainty Budget. Choosing one titration sequence as an example, all quantities and associated uncertainties impacting the final result were summarized in Table 2. This

spectrometer, equipped with a mercury cadmium telluride (MCT) high D* liquid N2-cooled mid-infrared detector, and set to 1 cm−1 resolution. The spectrometer was configured with a 10 m multipass White-type cell (Gemini Scientific Instruments, USA). The gas cell had wetted surfaces of electropolished stainless steel treated with SilcoNert 2000 and gold (mirror coatings) to minimize surface interactions with reactive gas phase species. The interferometer was scanned at 360 scans per minute, and spectra were coadded for 5 min. The optical path length of the gas cell was calculated using a standard gas mixture of dry air certified by the NIST to contain methane at 1842.74 nmol mol−1 with an uncertainty of 0.71 nmol mol−1 and carbon dioxide at 400.4 μmol mol−1 with an uncertainty of 0.16 μmol mol−1. Knowing the mole fractions of these two compounds while quantifying them with the FTIR using the method described below allowed the cell path length to be deduced, based on the value needed for measured and certified mole fractions to agree. From those calculations, a length of 10.01 m with a relative uncertainty of 5% was deduced. This level of uncertainty is conservative and dominated by the uncertainty in the molecular parameters required, as explained below. Only HNO3 could be observed at significant levels in the NO2 in nitrogen mixtures. It was quantified using a synthetic calibration method published in 201322 and using the HNO3 ν2 band at 1709 cm−1. Synthetic spectra were generated and used to fit experimental spectra with the software MALT version 5.0, as described by Griffith et al.,23 based on line strengths and other molecular parameters tabulated in HITRAN 2012.24 Compared to the 2013 method, the inner surface treatment of the FTIR gas cell with SilcoNert allowed a noticeable improvement of the HNO3 stability during measurements, reducing this uncertainty component. An overall uncertainty of 10% was considered on HNO3 values. This is very conservative but acceptable as its impact on the final NO2 uncertainty is negligible. All three NO2/N2 mixtures were analyzed twice, first within a few days before the titration measurement, then about four months later. In all of them, HNO3 was present, at mole fractions ranging from 500 nmol mol−1 to 1010 nmol mol−1, and with an observed drift of 2 nmol mol−1 month−1 in the worst case. These values were at least twice as large as the certified values provided by the suppliers. After discussion with the suppliers, it was concluded that HNO3 resulted from the reaction of NO2 with trace water inside the cylinders, with this reaction taking place quickly initially, after introduction of the gases into the cylinders, leading to a first period of rapid growth followed by a stabilization of HNO3 content. The gravimetric mole fractions of NO2 required correction for the measured HNO3 mole fractions, and this bias was corrected to obtain the unbiased values xs(NO2), with unchanged associated analytical uncertainties. NO2 Analyzer Calibration and Corrections. Following a similar process as for NO, the NO2 instrument was calibrated in line with the titration measurements, during the same phase t3. As previous validation studies had demonstrated a negligible span drift, the analyzer measured values were only corrected for the offset drift, which was typically of −1 nmol mol−1 during the entire sequence. This drift was again linearly interpolated at time t3 from two successive zero measurements (phase t1) of two sequences. The response of the NO2 analyzer was further averaged over the last three measurement points to provide the value ycal(NO2), and the associated standard deviation was used

Table 2. Typical Uncertainties on the NO Reacted quantity y

yva

u(y)b

NO in standard xs flow in line 1 m qv1 (phase t3) flow in line 2m qv2 (phase t3) flow in line 3m qv3 (phase t3) flow in line 4m qv4 (phase t3) NO reactant xr (phase t3) analyzer responsem ycal (phase t3) analyzer responsem ytit (phase t3) flow in line 1m qv1 (phase t5) total flow qvtot (phase t5) NO reactant xr (phase t5) NO final xfinal (phase t5) NO reacted ΔxNO

59928.20 24.47 49.43 2200.59 50.00 630.86 624.00 639.66 48.91 2325.60 1260.41 639.66 616.15

50.80 0.11 0.13 1.55 0.06 2.82 0.32 0.51 0.11 1.57 3.05 1.80 3.55

a


table provides a good representation of the most important sources of uncertainty on the NO reacted, namely the measurement of flow rates lower than 100 mL min−1. Final values displayed in this table are for information only, as they were actually calculated using a least-squares regression on the entire measurement range, rather than a single-point calibration presented here. NO2 Gain. The NO2 gain was measured with the NO2 analyzer calibrated with NO2 in air mixtures during the phase t3. Adopting a similar approach as for NO standards, each NO2 in the air calibration mixture was obtained from the standard mixture properly diluted with nitrogen. The mole fraction of NO2 at the analyzer was calculated using an equation similar to eq 2 for NO. Traceability of NO2 Standard Gas. Three different standard mixtures of NO2 in N2 were used during the measurements, one produced by the NPL and two by the Dutch Metrology Institute (VSL), with the following certified values and associated standard uncertainties indicated in brackets: 35.0 μmol mol−1(0.175 μmol mol−1), 29.66 μmol mol−1 (0.15 μmol mol−1), and 39.88 μmol mol−1 (0.2 μmol mol−1). When anchoring the ozone titration to NO2 standards, a difficulty arises from the stability and purity of the mixture when NO2 is at the micromole per mole level. As observed during the international comparison CCQM-K74, HNO3 at nanomole per mole levels is very commonly observed as an impurity in such mixtures, which would result from a reaction between NO2 and traces of water inside cylinders.21 All three standards were therefore analyzed by FTIR as explained below. Purity Analysis of NO2 Standards by FTIR. NO2 in nitrogen standards were analyzed with a Bruker Vertex 70v FTIR E


Article


ozone absorption cross-section. Showing relative differences leads to easily interpretable data, and these are displayed with all measurement points in Figure 3, using the reacted ozone mole fraction as a reference. Uncertainties in the figure are combined expanded uncertainties (95% confidence interval).

as standard uncertainty, with typical absolute values of 0.3 nmol mol−1. The same process already described for the NO analyzer calibration was applied to fit each set of 10 calibration values, including the calculation of covariances between two calibration points. It was further noted that the NO2 analyzer calibration line was slightly different when calibrating directly with NO2 in nitrogen standards or indirectly with NO in nitrogen standards after titration with ozone. The two calibration lines slopes were in relative agreement of better than 10−3, but the two intercepts showed a 5 nmol mol−1 difference, not covered by a standard uncertainty of 1.5 nmol mol−1. As there was no reason to prefer one calibration process rather than the other, it was decided to use both procedures and incorporate the difference into the uncertainty of the calibration procedure. The NO2 analyzer calibration intercept was further corrected by half of the difference, 2.5 nmol mol−1. Its uncertainty was also combined with an additional component equal to 1.44 nmol mol−1 as estimated from a rectangular distribution between the two intercepts, resulting in an intercept standard uncertainty of 2 nmol mol−1. Finally, the NO2 gain ΔxNO2 was simply deduced from the calibration line calculated above. The associated combined standard uncertainty was between 1.8 nmol mol−1 and 3.7 nmol mol−1, also dominated by the uncertainty on the dilution factor. Uncertainty Budget. As for NO, all quantities and associated uncertainties impacting the final NO2 gain were summarized in Table 3. Unlike the NO reacted uncertainty budget, the

Figure 3. Relative difference between NO and O3 reacted (diamonds) and NO2 gain and O3 reacted (circles). O3 mole fractions were all measured by standard UV photometry.

Figure 3 highlights the constant relative bias of about 2.1% between NO and O3 reacted, over the entire range of measurements, which can be attributed to the impact of the ozone cross-section value used. This bias is just covered by the expanded uncertainties, mainly dominated by the ozone absorption cross-section uncertainty as in this figure. A similar trend is observed between NO2 gain compared to O3 reacted, with a relative bias of 2.6% on average. The present results are in line with most recent published work on ozone titration, believed to be biased compared to UV photometry when measured with sufficiently low measurement uncertainties. Results of such comparisons appeared first in 1976 in the study of DeMore and Patapoff.2b Using an ozone absorption cross-section 1.5% higher than the conventional value adopted later on, they found a good agreement between both methods, however within experimental uncertainties of 5%. In 1982, Fried and Hodgeson25 successfully introduced laser photoacoustic detection of NO2 during GPT experiments, in order to avoid previously observed issues with the simultaneous detection of NO and NO2 with a chemiluminescence detector. They performed a comprehensive comparison between measurements of NO, NO2, and O3, leading to a good agreement between NO and NO2 but lower O3 values by 3.6%. Analysis of possible concurrent reactions leading to the oxidation of NO to NO2 failed to find a chemical explanation of this bias. More recently, a NIST SRP was compared to ozone titration with excess NO by Tanimoto et al.,26 with a reduced 0.4% uncertainty on GPT, and an observed 2% bias between both methods. This was further confirmed by Walden,27 who observed a bias of 2.8% in a GPT system designed to allow reaction at the micromole per mole level followed by measurements of diluted NO2.

Table 3. Typical Uncertainties on the NO2 Gain quantity y

yva

u(y)b

NO2 in standard xs (corrected for HNO3) flow in line 1m qv1 (phase t3) flow in line 2m qv2 (phase t3) flow in line 3m qv3 (phase t3) flow in line 4m qv4 (phase t3) NO2 in mixture xr (phase t3) analyzer responsem ycal (phase t3) analyzer responsem ytit (phase t5) NO2 gain ΔxNO2

29050 24.47 49.43 2200.59 50.00 617.76 697.05 698.69 619.22

150 0.11 0.13 1.55 0.06 3.62 0.3 0.3 3.64

a


uncertainty on the standard mixture is a major component. This fully reflects the difficulty in producing stable mixtures of NO2 in nitrogen in cylinders. The NO2 mole fraction indicated in this example takes into account the amount of HNO3 measured by FTIR.

■

COMPARABILITY BETWEEN REACTANTS AND PRODUCTS Three series of measurements have been performed with each of the three NO2 standards, resulting in a total of nine series. Each series included 10 points at different mole fractions covering the range 100−600 nmol mol−1. Each measurement point was formed by the values −ΔxO3, −ΔxNO, and ΔxNO2. It is interesting to first analyze those results by looking at the level of agreement between reactants/products, when ozone is measured by UV photometry with the conventional value of the F


Article


■

OZONE ABSORPTION CROSS-SECTION DEDUCED FROM TITRATION WITH NO AND NO2 Comparing the GPT results to those from UV photometry allows an absorption cross section of ozone in the UV to be calculated. The ozone absorption cross-section being a constant, the two comparisons performed during a titration experiment, namely one between NO and O3 reacted and the second between NO2 gain and O3 reacted, were used to calculate two new values of the ozone absorption cross-section: σGPT = σhearn/a1, where a1 is the slope of the linear regression between ozone mole fractions deduced by GPT anchored to either NO or NO2, versus mole fractions traceable to the SRP. To calculate values from all measurements, the two sets of nine series of measurements were fitted with a straight line, taking into account all uncertainties and correlations between measurements performed with the ozone analyzer, except the ozone cross-section uncertainty. Results are plotted on Figure 4, showing the two parameters of each of the nine regressions (intercept a0 and slope a1), together with their expanded uncertainty.

measurement of the ozone absorption cross-section performed by spectroscopy in pure gaseous ozone,28 equal to 11.27 × 10−18 cm2 molecule−1, as well as most recent results recently summarized by the Absorption Cross-Section of Ozone (ACSO) committe.29

■

CONCLUSION An experimental setup was developed to perform the titration of O3 with excess NO in the gas phase at nanomole per mole levels, with traceability of O3, NO, and NO2 measurements to standards of the highest metrological level. Both the NO reacted and the NO2 gain were compared to the O3 reacted in the experiment, demonstrating a constant relative bias of 2.1% to 2.3% between O3 and the two other compounds when O3 was measured by UV photometry anchored to the conventional value of the ozone absorption cross-section. Using nine series of repeated measurements to obtain a good statistical representation, the GPT results within the range 100−600 nmol mol−1 were used to calculate two GPT-based ozone absorption cross-section values, one with traceability to NO standards, the other to NO2. They provided very close values of 11.24 × 10−18 cm2 molecule−1 and 11.22 × 10−18 cm2 molecule−1, respectively, in agreement with most recent measurements of this constant performed by absorption spectroscopy on pure ozone samples.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +33 1 45076270. Author Contributions

All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors would like to acknowledge the contributions of Michael Esler (BIPM until 2006) in the development of the first GPT system.

Figure 4. Slope a1 and intercept a0 of the nine linear regressions between NO and O3 reacted (blue diamonds) and NO2 gain and O3 reacted (pink triangles).

■

This figure shows a good reproducibility of measurements among the nine series, with standard deviation of the mean well below the typical standard uncertainty of each parameter. All intercepts are also consistent with zero within the uncertainty, as expected from the titration principle. It can be noted that those values reflect the results discussed in the previous section, with a better agreement between NO and NO2 than with O3. The mean slope of each series was computed. All series are strongly correlated with a domination of type B components in the uncertainty budget. Therefore, the overall uncertainty associated with the mean slope was taken as a mean of the nine uncertainty values as a conservative value. This resulted in the following values and standard uncertainties in parentheses: 1.0207 (0.0091) for NO and 1.0227 (0.0074) for NO2, interpreted as constant relative biases of 2.1% and 2.3%. Two GPT-based cross-section values were deduced from the ratio between the Hearn value and the mean slopes, resulting in the values and expanded uncertainties of 11.24 × 10−18 (0.20) cm2 molecule−1 and 11.22 × 10−18 (0.16) cm2 molecule−1, respectively. These two results are in agreement with our

REFERENCES

(1) Ambient Air - Determination of Ozone - Ultraviolet Photometric Method; International Organization for Standardization: Geneva, Switzerland, 1996; Vol. 13964. (2) (a) Paur, R. J.; McElroy, F. Technical Assistance Document for the Calibration of Ambient Ozone Monitors; Department E (MD 77): Research Triangle Park, NC, 1979; Vol. 27711. (b) DeMore, W. B.; Patapoff, M. Environ. Sci. Technol. 1976, 10 (9), 897−899. (3) (a) Saltzman, B. E.; Gilbert, N. Anal. Chem. 1959, 31 (11), 1914−1920. (b) Saltzman, B. E.; Gilbert, N. Am. Ind. Hyg. Assoc. J. 1959, 20, 379−386. (4) Clyne, M. A. A.; Thrush, B. A.; Wayne, R. P. Trans. Faraday Soc. 1964, 60 (0), 359−370. (5) Atkinson, R.; Baulch, D. L.; Cox, R. A.; Crowley, J. N.; Hampson, R. F.; Hynes, R. G.; Jenkin, M. E.; Rossi, M. J.; Troe, J. In Atmospheric Chemistry and Physics; IUPAC Task Group on Atmospheric Chemical Kinetic Data Evaluation: Research Triangle Park, NC, 2004; Vol. 4, p 1461. DOI: 10.5194/acp-4-1461-2004. (6) Ambient Air -- Gas Phase Titration -- Calibration of Analysers for Ozone; International Organization for Standardization: Geneva, Switzerland, 2009; Vol. 15337:2009. (7) Galbally, I. E.; Schultz, M. G.; Buchmann, B.; Gilge, S.; Guenther, F. R.; Koide, H.; Oltmans, S.; Patrick, L.; Sheel, H. E.; Smit, H.; Steinbacher, M.; Steinbrecht, W.; Tarasowa, O.; Viallon, J.; Voltz-

G


Article

Analytical Chemistry Thomas, A.; Weber, M.; Wielgosz, R. I.; Zellweger, C. Guidelines for Continuous Measurements of Ozone in the Troposphere; World Meteorological Organization: Geneva, 2013. (8) Hearn, A. G. Proc. Phys. Soc., London 1961, 78, 932−940. (9) Viallon, J.; Moussay, P.; Norris, J. E.; Guenther, F. R.; Wielgosz, R. I. Metrologia 2006, 43, 441−450. (10) Kebabian, P. L.; Herndon, S. C.; Freedman, A. Anal. Chem. 2005, 77 (2), 724−728. (11) Zucco, M.; Curci, S.; Castrofino, G.; Sassi, M. P. Meas. Sci. Technol. 2003, 14, 1683−1689. (12) BIPM; CEI; FICC; ISO; OIML; UICPA; UIPPA Guide to the Expression of Uncertainty in Measurement; International Organization for Standardization: Geneva, Switzerland, 1995; p 101. (13) Paur, R. J.; Bass, A. M.; Norris, J. E.; Buckley, T. J. Standard Reference Photometer for the Assay of Ozone in Calibration Atmospheres, 6963; National Institute of Standards and Technology: Gaitherburg, 2003; p 65. (14) BIPM International comparisons of ozone standards (BIPM.QM-K1). http://www.bipm.org/en/bipm/chemistry/gasmetrology/ozone-comparisons.html. (15) Viallon, J.; Moussay, P.; Idrees, F.; Wielgosz, R.; Zhou, Z. Final report on ongoing key comparison BIPM.QM-K1, ozone at ambient level, comparison with NIM (July 2014). Metrologia 2015, 52 (1A), Tech. Suppl. 08012. (16) Riu, J.; Rius, F. X. J. Chemom. 1995, 9 (5), 343−362. (17) Determination and Use of Straight-Line Calibration Functions; International Organization for Standardization: Geneva, Switzerland, 2010; Vol. 28037. (18) Bremser, W.; Viallon, J.; Wielgosz, R. I. Metrologia 2007, 44, 495−504. (19) Niederhauser, B.; Barbe, J. Metrologia 2002, 39 (6), 573. (20) Wielgosz, R. I.; Esler, M.; Viallon, J.; Moussay, P.; Oh, s. H.; Kim, B. M.; Tshilongo, J.; Mokgoro, I. S.; Maruyama, M.; Mace, T.; Sutour, C.; Stovcik, V.; Musil, S.; Castorena, A. P.; Murillo, F. R.; Kustikov, Y. A.; Pankratov, V. V.; Gromova, E. V.; Thorn, W., Jr.; Guenther, F. R.; Smeulders, D.; Baptista, G.; Dias, F.; Wessel, R. M.; Nieuwenkamp, G.; Van der Veen, A. M. H.; Botha, A.; Valkova, M.; Caballero, V. S.; Konopelko, L. A. Metrologia 2008, 45 (Tech. Supl.), 08002. (21) Flores, E.; Idrees, F.; Moussay, P.; Viallon, J.; Wielgosz, R.; Fernández, T.; Ramírez, S.; Rojo, A.; Shinji, U.; Waldén, J.; et al. Metrologia 2012, 49 (1A), 08005. (22) Flores, E.; Viallon, J. l.; Moussay, P.; Wielgosz, R. I. Appl. Spectrosc. 2013, 67 (10), 1171−1178. (23) Griffith, D. W. T. Appl. Spectrosc. 1996, 50 (1), 59−70. (24) Rothman, L.; Gordon, I.; Babikov, Y.; Barbe, A.; Benner, D. C.; Bernath, P.; Birk, M.; Bizzocchi, L.; Boudon, V.; Brown, L.; et al. J. Quant. Spectrosc. Radiat. Transfer 2013, 130, 4−50. (25) Fried, A.; Hodgeson, J. A. Anal. Chem. 1982, 54, 278−282. (26) Tanimoto, H.; Mukai, H.; Hashimoto, S.; Norris, J. E. J. Geophys. Res. 2006, 111, D16313. (27) Waldén, J. Metrology of Gaseous Air Pollutants; University of Helsinki: Helsinki, Finland, 2009. (28) Viallon, J.; Lee, S.; Moussay, P.; Tworek, K.; Petersen, M.; Wielgosz, R. I. Atmos. Meas. Tech. 2015, 8 (3), 1245−1257. (29) Orphal, J.; Staehelin, J.; Tamminen, J.; Braathen, G.; De Backer, M.-R.; Bais, A.; Balis, D.; Barbe, A.; Bhartia, P. K.; Birk, M.; Burkholder, J. B.; Chance, K.; von Clarmann, T.; Cox, A.; Degenstein, D.; Evans, R.; Flaud, J.-M.; Flittner, D.; Godin-Beekmann, S.; Gorshelev, V.; Gratien, A.; Hare, E.; Janssen, C.; Kyrö lä, E.; McElroy, T.; McPeters, R.; Pastel, M.; Petersen, M.; Petropavlovskikh, I.; Picquet-Varrault, B.; Pitts, M.; Labow, G.; Rotger-Languereau, M.; Leblanc, T.; Lerot, C.; Liu, X.; Moussay, P.; Redondas, A.; Van Roozendael, M.; Sander, S. P.; Schneider, M.; Serdyuchenko, A.; Veefkind, P.; Viallon, J.; Viatte, C.; Wagner, G.; Weber, M.; Wielgosz, R. I.; Zehner, C. J. Mol. Spectrosc. 2016, 327, 105.

H


Ozone Cross-Section Measurement by Gas Phase ... - ACS Publications

Recommend Documents