Measurement of gas phase hydrogen peroxide ... - Wiley Online Library

JOURNAL

OF GEOPHYSICAL

RESEARCH,

VOL. 91, NO. D5, PAGES 5371-5378, APRIL 30, 1986

Measurement of Gas Phase Hydrogen Peroxide in Air by Tunable Diode Laser Absorption Spectroscopy F. SI•EMR, • G. W. HARRIS, 2 D. R. HASTIE, 3 G.I. M^cI(^¾,2 ANDH. I. SCHIFF 2'3 Tunablediode laserabsorptionspectroscopy has beenappliedto the determinationof gas phase hydrogenperoxidein ambientair with sub-ppbv(parts per billion by volume)detectionlimits for measurement timesof the orderof minutes.The methodsfor calibratingthe instrumentand for assuring the absenceof spectroscopic and samplinginterferences at the level of our presentdetectionlimits are described. Ambientair monitoringwith our systemindicatesthat the hydrogenperoxidemixingratio is often < 0.3 ppbv. Five-minuteaveragemixing ratios of up to 2.9 ppbv have beenmeasuredat sitesin southwestern Ontario, Canada,during the summermonthsof 1984 and 1985,while the highest1-hour averagevalue observedwas 2.1 ppbv.

INTRODUCTION

cation, 1985), the results of earlier studies must be treated with

Hydrogen peroxide is believed to be the most important oxidant in the aqueous phase conversion of S(IV) to S(VI) [Penkett et al., 1979; Middleton et al., 1980] and is therefore directly involved in the chemical transformationsleading to the acidificationof precipitation.Although formed mainly in the gas phase, H20 2 is highly soluble. Reliable measurements of H202 in both the gas and the aqueousphasesin the troposphereare thereforeessentialfor the understandingof the processesinvolved in acid deposition. Atmospherichydrogenperoxide is mostly formed by the

caution.

Measurementsof H202 mixingratiosin troposphericair are thereforealsoimportant for understandingthe behaviorof the odd hydrogenradicals[Loganet al., 1980]. Despite its importance,relatively few measurementsof this speciesin the gas phase have been reported.Bufalini et al.

occur in the aqueous trapping methods. In addition, the method has other advantages for monitoring atmospheric in H202; this paper shows that it can provide sensitivemeasurements (sub-ppbv detection limits) with averaging times of a few minutes, and a further advantage is its very high specificity towards the target molecule by virtue of the high spectral resolution employed. Unlike methods which depend on the chemical properties of the molecule, no special procedures are required to avoid interferencesfrom, for example, organic per-

Direct measurementof H202 in the gas phase avoids these artifacts effects.Kley and Stone [1978] proposed a method based on the VUV photolysis of H202 to yield excited OH radicals which could be sensitivelydetected by their fluorescence,but so far this method has not been implemented.The feasibility of H202 measurementsby kilometer path length Fourier transform infrared spectroscopywas assessedby Tuazon et al. [1980] and by Hanst et al. [1982], with resulting estimatesof detectionlimits between40 and 100 ppbv. recombinationof hydroperoxy,HO2, radicals. During the The tunable diode laser absorption spectrometer(TDLAS) daytime, photoiysisoI H2t) 2 producesthe Ht) radical, which method fulfils the requirement of detecting H20 2 directly in is recognizedto be a key speciesin atmosphericchemistry. the gas phase, removing the possibility of such artifacts as may

[1972] report40-180ppbv(partsperbillionby volume)H202 during moderateand severesmogepisodesat Riverside,California. Kok et al [1978a, b] found midafternoon values of 10-30 ppbv at two locations in the south coast air basin of

southernCalifornia and mixing ratios of 100 ppbv in the plume of wildfires.Kelly et al. [1979] report much lower mixingratiosof 0.3-3 ppbv at a rural sitenear Boulder,Colorado.Their valuesare closeto the 0.2- to 2-ppbvlevelspredicted by modelsfor summermid-latitudes[Logan et al.,

oxides.

EXPERIMENTAL

The TDLAS method takes advantage of the high monochromaticityand rapid tunability of Pb salt diode lasersto Thesepreviousstudieshavereliedon bubblingambientair measureabsorptionsarisingfrom singlerotational-vibrational throughan aqueous solutioncontaining sometrappingagent, lines in the mid infrared spectrumof a molecule.In order to with subsequent analysisof the resultingsolution.The forma- facilitate the measurement of very low optical densities tion of H202 as an artifact of this samplingprocedure,most (< 10-5) at line centerand to reducethe chancesof overlap probably by reactionsof ozone and its decompositionprod- between absorption lines, reduced pressures(•25 torr) are uctswith water,hasnow beendemonstrated[Zika and $altz- used to minimize pressurebroadening of the rotational lines. man, 1982; Heikes et al., 1982; Heikes, 1984]. In addition, the The atmospheric sample is pumped rapidly at the reduced aqueous trap methods may suffer from SO2 interferences pressurethrough a White cell, which also providesthe neces[Heikeset al., 1982] and from interference by organicperox- sary long optical path lengths (about 40 m in our present ides,whichare alsoexpectedto be presentin tropospheric air instrument). A description of our original system has been [Logan et al., 1980]. Although progresshas been made in published previously [Hastie et al., 1983], and details of our overcomingtheseartifacts(A. L. Lazrus, private communi- improved TDLAS systemsfor trace atmosphericgas monitor1980].

• Max PlanckInstitute forChemistry, Mainz,FederalRepublic of Germany.

2 UnisearchAssociates Incorporated, Concord,Ontario,Canada. 3 York University,Downsview, Ontario,Canada. Copyright1986by the AmericanGeophysicalUnion. Paper number 6D0014. 0148-0227/86/006D-0014505.00 5371

ing will be provided elsewhere(G. W. Harris et al., unpublishedmanuscript, 1986). Application of the TDLAS method to the quantitative measurement of a particular speciesrequires effort in three areas: (1) the selectionand characterizationof a suitable diode laser and of laser operating conditions; (2) the development and validation of a calibration method, and (3) the establishment of sampling procedures.

5372

SLEMRET AL.' GAS PHASEHYDROGENPEROXIDEMEASUREMENT

14--

13--

LINE

A

II-

I0--

9--

8-

3 -'

2 -'

! .-

ot

I

I

!

I

I

I

2

3

4

5

I

I

I

I

I

!

I

I

1

I

I

I

I

2

3

4

5

6

7

8

9

I0

(60

SEC

MEASUREMENT

NUMBER

I

I

I

I

I

I

I

1

I

2

3

4

5

6

AVERAGES)

Fig. 1. Measurements of an estimated11 ppbv H202 addedto scrubbedambientair usingthreedifferentH20 2 absorptionlines. Each data point is obtained from a 60-s averagespectrum.The horizontal lines show the mean of each data set. Line A was usedfor ambient air monitoring.

Selectionof Laser Diode and Operating Conditions

beam and the tunable diode laser (TDL) beam, was used to ascertainthe TDL output wavelengthto an accuracyof about

_ 0.5 cm-x. The linesare then assigned by comparison of the The most prominentvibrationalabsorptionband of H202

is the v6bandcenterednear 1260cm-x. Accordingto the Air Force GeophxsicsLaboratory (AFGL) absorptionline compilation [Rothrnanet al. 1983a,b], the strongestgroup of rotational lines in the band are near the center of the R branch

around 1285 cm -•.

The selection of an individual line for

recordedspectrumof H202 in this region with the line listings in the AFGL compilation [Rothrnanet al., 1982a, b'l. Simultaneouslyrecordedspectraof N20 and H20 aid in the assignment process.The grating spectrometeris usedsimplyto identify the spectral region of the absorption band under study. When actual measurementsare made, the grating spectrometer is removed, and the highly monochromatic laser beam is tuned over a region correspondingto a few absorp-

air-monitoring purposes involves a number of compromises tion line widths(about 10-2 cm-•) centeredat the target between the properties of the available lasers, the strengths of the absorption lines of the target molecule,and the avoidance absorption line. Most of the work reported here was done with a second of absorptionlinesfrom other atmosphericspecieswhich may interferespectroscopically. The important lasercharacteristics laser supplied by the Fraunhofer Institute for Physical for this applicationrelate to output wavelengthrange,output MeasurementTechniques(Freiburg, Federal Republicof Germode structure,frequencyand amplitude stability, and beam many),operatedat the H202 line at 1284.2044cm-•, which profile intensity distribution. hasa linestrength of 3.6 x 10- 20cm2 molecule-• cm- •. Only Five lasers were obtained from two manufacturers for the line at 1283.740cm-•, of the severalhundredlinesin the evaluation in this work, and two laserswere judged suitable v6 band of HeOe, has a significantlygreaterintensity(s = 5.04 for complete characterization.The laserswhich were rejected x 10-eø cme molecule-; cm-;). That line is, however,overeither provided insufficientoutput power, did not operatein a laid by an interfering absorption in ambient air and therefore single-frequency mode near the target region,were inherently cannot be usedfor air monitoring. It should be noted that neither the assignmentof a selected noisy(poor power-to-noiseratio), or had a poor (severelynonGaussian)beam intensityprofile. line to a particulartransitionin the moleculenor knowledge

Of the two lasersselectedfor furtherwork, onewassupplied by Laser Analytics(Bedford,Massachusetts) and after detailed evaluationwas operatedon the H202 linesat 1264.590cm-•

of its integratedline strengthis strictlynecessary for the present purposes. However, if the selectedline can be identified

and if its integratedline strengthis known, it is possibleto

and 1264.622cm-•. Theselineshaveintegrated linestrengths ascertainwhetherother lines which are significantlystronger of 0.4 and 1.6 x 10- 20cm2 molecule-• cm- •, respectively. To than the one selectedexist elsewherein the spectrumof the assigntheseand other lines in the H202 band, a grating spectrometer, calibrated by superimposing a helium-neon laser

molecule

and hence whether

further

work

on laser and ab-

sorption line selectionproceduresare warranted.

$LEMR ET AL.' GAS PHASE HYDROGEN PEROXIDE MEASUREMENT

5373

deviation. These data were obtained using background subtraction,as describedbelow,and 1-min signalaveraging.The means and standard deviations are 10.82 ___0.18(line A), Line assignmentand line compilations are not used for 10.54+_0.89 (line B) and 10.86 _ 0.84 (line C), further confirmscreeningthe selectedregion for spectroscopicinterferences. ing the absenceof interferenceson the lines. The absenceof interferencesin the vicinity of the wavelength Calibration Procedures of the selected line is checked experimentally. The investigation of interferences, the selectionof suitablelaseroperating The procedure for calibrating the TDLAS for ambient conditions, and the determination of system detection limits HeOe measurementsis addition of a known flow of the target are not independenttasks but are carried out iteratively. Degas to a known flow of air at the ambient air-samplinginlet. pendingon the moleculeand on the laser or lasersavailable, In this way the sensitivityof the completesystemis calibrated severalto severalhundred separaterotational-vibrational lines directly in terms of ambient H2Oe mixing ratio and exact may be candidatesfor usein a measurementsystemaimed at knowledge of instrumental factors, such as the relationships that molecule. between the magnitude of the 2f signal (signal detected at Duringan atmospheric measurement the laserfrequency is twice the modulation frequency,f) and optical density in the sweptacrossa narrow (about 10-2 cm-x) regioncenteredon White cell or of the factors governing the line width of the the target absorption line. Absorption featureslying outsideof absorption line and of the 2f line, is not required so long as lnvestiqation of Possible Spectroscopic Interferences

this narrow

window

do not interfere

with

the measurement.

such factors are held constant.

One possibleexceptionis the caseof very intenselines arising from specieswhich have large and variable mixing r.atios (such as water vapor) which may causea degradation in detection limit even if the interfering line is centeredoutsidethe window. This is becausethe Lorentzian "tail" of the intense line may have significant optical density in the window and as the mixing ratio of the interfering speciesvaries the shape of the backgroundspectrumwill alter. If this occurs,the precisionof background subtraction will be degraded and will result in poorer detection limits. This effect will not, however, lead to a

The calibration method will also compensate for any firstorder surface processesaffecting the throughput efficiency of the sampling system.Higher-order effects(if present)would be best accountedfor when the mixing ratio of the added calibration "spike" is comparable to the mixing ratio of the target gas in ambient air. For H202 the ambient mixing ratio is low, and this requirement must be compromisedby the need to maintain calibration precision. Calibration is performed on mix-

ecule has been removed ("scrubbed").Any interferenceswill be reflected in a lack of linearity of the instrument response toward added H202. If, in any of theseprocedures,interfering absorption lines are detected,that particular absorption line is abandonedand another absorptionline is selected.ß An unequivocaltest for the absenceof interferencesis af-

pretreated with a 50/50 mixture of HeOe (90%) and concentrated HeSO,•. The parts were immersed in this mixture, which was then cautiously warmed (inside a fume hood) until steady gas evolution was observed.This procedure appeared to prevent subsequentdecomposition of H202 by impurities on the surfaces.By suitable choice of coil length, solution

forded by comparisonof the mixing ratios calculatedfrom severalseparateabsorptionlines.There is a vanishinglysmall probability that any interferinggas(or gases)could have exact overlap and identical ratios of line strengthsas the target gas at each wavelength. In the case of H202 we have identified

concentration, andbathtemperature, permeation ratesin the

turescontaining5-30 ppbv H•Oe. The calibration gas is added at controlled flow rates in the range 10-50 standardmL min-x to an airstreamflowingat false identification of the target molecule. Chance interferences are readily identified. Since the laser is about 5 standardL min-x. Calibrationin the low-ppbvrange scanned across a wavelength region contained the target abthereforerequiresa stable sourceof HeO2 with mixing ratios sorption line, line shape information, and the exact line center of a few ppmv (parts per million by volume).The sourcedevelposition are determined. Interference resulting from partial oped for this work consistsof a 4-m-long coil of low-density overlap of an absorption line with the target line is immedipolyethylenetubing, 0.32-mm OD and 0.16-mm ID (Imperial ately apparent from' the line shape. Such interferences,and Eastman type P-22-1/8) which is immersed in either 30 or even the rare casesof exact overlap (line centers identical to 50% stabilizedsolutionsof HeOe thermostatedto ___0.2 K in +_10-3 cm-x), can be.studiedby changingthe concentration the range 300-320 K. The H202 permeating into the tube is of the target gas in the sampled air as well as by inspection of carried away by the dry nitrogen carrier gas. spectraarisingfrom ambientair from whichthe targetmolAll parts of the device in contact with H202 solution were

five different lines which pass all the above tests for freedom from interference. As pointed out above, it was not necessary to spectroscopicallyassigneach of theselines for this test, nor to determine their exact wavelengths. Figure 1 shows mixing ratios measured on three lines, labeled A, B, and C, which lie close to each other in the H:O:

spectrumand which werejudged to be interferencefree to the level of the detection limits on the basis of the above tests. The

instrument was calibrated on each line by addition of a 35ppbv spike to scrubbed ambient air, and then a second lower spike of •11 ppbv added to ambient air was measured on each line.

Lines B and C gave poorer detection limits than line A, as can be judged from the scatter in the data in Figure 1, and were not usedfor ambient monitoring. However, the meansof the three measurement

sets are identical

within

one standard

range5-500 ng min-x whichprovidethe desiredmixingratios of calibrating gas may be obtained with thesedevices. The permeation rate of the calibration source was determined by the colorimetric TiCI,• method of Pilz and Johann [1974]. A stock solution is prepared by dissolving11.6 mL of TiCI,• (Alfa Products, Danvers, Massachusetts)in 50 mL of concentratedHC1 (37%. 12 M). This solution is bright yellow and keeps indefinitely. Prior to use the stock solution is diluted tenfold with distilled water to produce a colorlesssolution which is somewhat

unstable

and must be discarded

after

2-3 days. One mL of this working solution and 10 mL of 1.2 M HC1 are placed in an impinger, and the calibration gas stream is bubbled through the impinger until a slight yellow coloration is observed (1-10 hours depending on the permeation rate). The solution is then brought up to 50 mL total volume by addition of distilled water, and its absorbance at 415 nm is compared to that of a blank solution which has not

been exposed to H202. Matched 10-cm-long cuvettes are used.The yellow coloration is due to a Ti-H202 complex of 1'! stoichiometry[Pilz and Johann, 1974] with an absorption

5374

SLEMRET AL.' GAS PHASEHYDROGENPEROXIDEMEASUREMENT I0

1

6

o

I

o

I

2

I

I

I

4

ADDED

I

I

I

6

H202

I

6

(ppbv)

Fig. 2. Response linearitytowards1.7-8.3ppbv H202 addedto scrubbedambientair. The leastsquaresline has slope 1.12 and intercept -0.15 ppbv.

coefficientat 415 nm equal to 735 M -• cm-• [Pilz and Johann,1974;Kok et al., 1978b].No changein absorbance at 415 nm wasobservedwhenthe nitrogencarriergasalonewas passedthroughthe impinger.The artifact H202 production which has causedproblems with impinger methodsfor ambient air-monitoring purposes does not occur when pure nitrogen is used as carrier gas [Heikes, 1984]. The H20: permeation rate may be calculatedfrom the absorbancemeasurement and the collection

time.

Repeatedmeasurementsof the permeation rate from individual permeation sourcesover a period of several months showed a decreaseof 10-20% per month. The deviceswere recalibratedat least weekly during measurementperiodsand were refilled with fresh solution if the permeation rate fell belowthe desiredrange. Materials Tests and SamplingProcedures Since H202 may readily decomposeon reactivesurfaces,all materialsto be usedin the parts of the instrumentcontacted

by the gasflow were testedfor inertnesstoward H20 2. Preliminaryscreening was performedby insertingthe materialin

the gasstreambetweenthe permeationdeviceand the TiCI,• impinger and noting any changesin the concentration.The H20 e mixingratioswerein the low-ppmvrangefor thesetests

and the gasflow rateswere10-50 standardmL min-•. Pyrex glass,perfluoroalkoxy(PFA) Teflon, and Care (usedfor IR windows)did not affect the H202 mixing ratio measuredat the impinger.One and a half metersof 6.4-mm OD polytetrafluoroethylene (PTFE) Teflon tubing initially removed

and a three-way alI-PTFE Teflon solenoidvalve (Nacom Industries, Tustin, California). Stainless steel tubing, metal needle valves,and aluminum tubing destroyedH•O• almost ,

completely, asdid severalepoxy-based glues(commonly used for affixing IR windows to Pyrex). Cells with Pyrex glass bodies and CaF• windows sealed with PTFE gasketspassed 95-100% of the incoming H202, and these cells were used for containing high concentrations of H20• for frequencycalibration (line locking) purposes.

Material testingin the 2-50 ppbv range was carried out by insertingsamplesof the material betweenthe ambient air inlet and the white cell. No HeOe losseswere observedon 30 m of 6.35-mm OD PTFE inlet tubing or on the Teflon solenoidand needle

valves

when

the airflow

was at its usual

value

of 5

standard L min- •.

Both the high- and low-concentrationmaterial testsshowed that FeSO,• and Hopcalite catalyst remove H•O2 quantitively from the gas stream. These materials were therefore used to scrub ambient air H•O• for acquiring background spectra. The scrubbers were glass tubes, 3.0-cm OD, packed with either 27 g of ferrous sulphate or with 35 g of 8-10 mesh Hopcalite (Mine Safety Appliances Company, Evans City,

Pennsylvania). The FeSO,•scrubberwaspreferredbecauseit is probably less perturbing toward other trace constituentsof ambient air (and the ideal is to use ambient air with only H•Oe removed for our backgroundspectra).FeSO,• reduces NOx in air, but interferences from NO2 and HNO 3 were explicitly ruled out. RESULTS

24% of theincomingHeOe,but afterovernight conditioning in the gas stream,no further HeO2 removaloccurred.Con- Linearity and ResponseTime Tests ditioning also preventeddestructionof H2Oe on a PFA The linearity of the instrumentresponsewas ascertainedby Teflonneedlevalve(GaltekCorporation,Chaska,Minnesota) addingH•O• in the mixingratio rangefrom 1.7 to 50 ppbv to

SLEMR ET AL.' GAS PHASE HYDROGEN PEROXIDE MEASUREMENT

ON

unity within one standard deviation (relative standard deviation 12%). This result shows that H:O: is not lost on the internal (Teflon coated) surface of the multiple-reflection cell and that the calibration system operates linearly over the flow ranges used. The responsetime of the instrument towards changesin the

OFF

ioo

./

8o

e/e..e•e....e-e -----e-e-

incoming H:O: mixing ratio was investigated by suddenly introducing or removing flows of calibration gas. The response time constant was approximately 0.6 min. Figure 3 shows the time responseof the signal to a pulse of 50 ppbv H202. This responsetime is about 6 times longer than the residence time of the gas in the multiple-reflectioncell and is

/

6o

4o

2o

o

5375

2 TIME

4 (M/N)

Fig. 3. Instrument responseto the sudden addition and subsequent removal of a spike of 50 ppbv H202 at the ambient air inlet.

intermediatebetweenvalues found for gasessuch as NO2, which appear not to interact with surfacesin the sampling train, and for strongly interacting gases such as HNO 3 [Hastie et al., 1983]. H:O: thereforeappearsto require some

time to cometo steadystatewith /he surfaces of the instruscrubbedambient air and measuringthe signal amplitude. ment. The response time is, however, considerably shorter

ppbv is less than the detection limit, while the r value for the linear regressionis 0.988.

than the averagingtime required to obtain sub-ppbv detection limits and so does not represent a constraint on instrument performance. Moreover, the fact that the response time was found to be independent of the mixing ratio over the range investigated, coupled with the responselinearity down to at

To investigatewhetherthe interiorsurfacesof the multiplereflectioncell affect the H202 mixing ratio, the signalsobservedwhen mixturesin the ppmv range flowed through a

least 1.7 ppbv, suggeststhat after the brief stabilizationperiod the gas inlet and white cell surfacesto not causeperturbations in the H:O: mixing ratios.

Figure 2 showsthe resultsof the linearity checksat the lower part of this range, 1.7-8 ppbv. The slope of the least squares line in Figure 2 is unity within 12% and the interceptof 0.15

9.5-cm-longcell were compared with those measuredin the 40-m path length White cell when the sameH20 2 flow was dilutedwith known flowsof scrubbedair to yield ppbv mixtures in the White cell. The ratios of the measuredoptical densitiesin the White cell to that calculatedfrom the path

Instrument

•

2.6

•

2.4

Limits

The detectionlimits are estimatedby either of two methods. The first is basedon the reproducibilityat the 95% confidence level of low-ppbv level measurementsof calibration gas mixtures [Parris et al., 1977]. Figure 4 showsthe responserepro-

length ratio and the optical densities in the short cell were

2.8

Detection

!

2 •

[.6

•

[.4

•

[.2

•

0.B

0.6

-'

0.4

-'

0.2

-'

0

'

I

I

$

I

I

MEASUREMENT

Fig. 4.

I

5

I

7

I

I

9

NUMBER

Responsereproducibilitytowards ,-•1.7ppbv H202. Each measurement is a 120-saverage.The horizontallines show the mean and _ 2 times the standard deviation of the data set.

5376

SLEMRET AL.: GAS PHASEHYDROGENPEROXIDEMEASUREMENT O.lõ

mV

northwest to northeast quadrant without encounteringlarge local sourcesof pollution and without crossingthe city. Suit-

ablewindconditions Obtainedon 5 sunnydaysbetweenFebruary 29 and June 3, I984,' and H202 measurementswere carriedout on eachof thesedays. Each ambient air measurementwas accompaniedby a detection limit determination,similar to that shownin Figure 5. In the measuring sequencea background spectrum(ambient air through the H202 scrubber)was first determined.An ambient spectrum was then accumulated, followed by a second backgroundmeasurement. With •n averagingtime of 180 s for each spectrum(correspondingto about 2000 individual scans) and a short waiting period after switchingvalvesto eliminate response time effects, four ambient air measurementswere madeeachhour accompaniedby four detectionlimit determi-

A

-o.

le5 mV

nations.

0.0õ

mV

H202 was deemed detected only if the hourly averagesof the ambient air measurementwere significantly(95% confidence level) higher than• the hourly averagesof the absolute values of the detection limits and if the ambient spectra were signficantly correlated, (again at the 95% confidence level) with the line shape information obtained from the calibration

0.05

-0.0õ

mV

mV

Fig. 5. (top) Calibration spectrumfrom 8.0 ppbv H20 2 (a) before and (b) after subtraction of a scrubbedambient air background spectrum. (Bottom) (a) Scrubbed ambient air background spectrum and (b) the residual resulting from subtraction of two such background spectra.The detection limit estimatedby least squaresfitting the calibration spectrum to the residual is 0.56 ppbv. Each spectrum was obtained with a 180-s averaging time. The spectra are centered at

1284.2044 cm-• andthescanrangeis about 10-2 cm-•.

ducibilitytowards • 1.7 ppbv H20 2 addedto scrubbedambient air. Twice the standard deviation correspondsto •0.3 ppbv.

-0.05

The secondmethodusesa computer-automated procedure.

mV

Two successivebackground spectra of scrubbedambient air are acquired, and one is subtracted from the other. The detec0.30

tion limit is thendetermined by theresultsof theleastsquares fit of a calibrationspectrum to the residual.This procedureis illustratedin Figure5. The top plot showsa calibrationspectrum dueto 8.0ppbvH202 bothbeforeandafterbackground subtraction.The bottomplot showsa backgroundspectrum and the resultof subtractingtwo suchbackgroundspectra acquiredsequentially. The fittingprocedureyieldsan estimate of the detectionlimit of 0.56 ppbv for the data shown in Figure5. Theestimates of thedetection limitsobtainedby the automaticmethodand by considering the reproducibility of

mV

the signal from known, constant calibration mixtures were very similar. The automatic estimation method was used most frequently. The detection limits obtained in this work for

H202 in ambientair are