C o m m u n i c a t i o n s

Effect of relative humidity on the determination offormaldehyde with the NIOSH 3500 method(chromatropic acid method)

C. L. Paul Thomas*a, Frank Meuniera, Cathy A. Veaseya and Charlotte D. McGillba Department of Instrumentation and Analytical Science, UMIST, P.O. Box 88, Sackville Street,Manchester, UK M60 1QD b Department of Chemistry, University of Reading, Whiteknights, P.O. Box 224, Reading,UK RG6 6AD

The relative humidity of a formaldehyde test atmospherewas found to be a factor in formaldehyde determinationusing the chromatropic acid (CTA) method; also knownas the NIOSH 3500 method.1 Formaldehyde recoveryefficiencies of 100% were obtained when the relativehumidity was below 0.1% at temperatures across therange 22–27 °C. Relative humidity levels of 50% and then> 95% gave reduced efficiencies of 66 and 36%,respectively, across the temperature range 22–27 °C.Multiple linear regression yielded a correction algorithmfor the CTA method that accounted for the influence ofwater at all the concentrations, masses and relativehumidities studied:

MM M

SF W=

+ ×( ) +0 004 1 776

1 025

. .

.(1)

where MS is the mass of formaldehyde sampled, MF is themass of formaldehyde recovered by the CTA method andMW is the mass of water sampled. The correlationcoefficient was 0.997. Preliminary studies with2-hydroxymethylpiperidine [CAS registry number3433-37-2] sampling systems have revealed the existenceof similar dependence on relative humidity.

Formaldehyde is a ubiquitous airborne pollutant emitted from avariety of anthropogenic sources and materials.2,3 It is asuspected carcinogen in humans4 and has been reported to causea range of other symptoms in cases of prolonged exposure.5Individuals are generally not affected until levels reachconcentrations above 5 mg m23.4,5 However, for some who aresensitised, levels of 0.12 mg m23 can be detrimental;4 shortterm and long term exposure limits have been set at 2.5 mgm23.6 It is interesting to note that indoor levels have beenreported to reach concentrations of 1.23 mg m23.

The chromatropic acid (CTA) method for the determinationof formaldehyde is a standard method with a reported limit ofdetection of 0.5 mg per sample. Reported interferences to themethod are oxidizable organics, phenols (when in an eight-foldexcess over formaldehyde), ethanol and higher alcohols,olefins, aromatic hydrocarbons and cyclohexanone.

A recent paper looking into sampling intercomparisons ofaldehydes8 reported no relationship between the level of relativehumidity and the degree of uncertainty between laboratories.Water has not been reported previously as an interferent in theanalysis of formaldehyde vapour. However, during the develop-ment of a thermal desorption technique for formaldehydedetermination the results yielded by the CTA method wereobserved to be inversely correlated to the relative humidity.9This communication is a quantitative description of the effect.

Experimental

A standard formaldehyde atmosphere was generated from aparaformaldehyde (BDH Chemicals, Poole, UK) permeationtube source maintained at 80 °C with an airflow of 90cm3 min21 passing directly over it. Two diluent lines were used,to prevent repolymerisation of the formaldehyde and to vary therelative humidity of the system, see Fig. 1. Full details of thevapour generator used along with gravimetric validation studieshave been reported previously.9 All experimental work wasundertaken within the temperature range 22–27 °C.

The CTA method involves active sampling of air throughthree midget impinger bottles; the first two contained 10 cm3 ofa 1% m/m aqueous solution of sodium bisulfite (AnalaR grade,Aldrich, Gillingham, UK) in distilled deionised water, 18 MWgrade. The third impinger was left empty to protect the samplingpump from liquid carry over. The recommended sample flowrate for the method is between 0.2 and 1 dm3 min21, collectingno less than 1 dm3 of air for concentrations above 1 mg m23,and no more than 100 dm3 for lower levels.1

After sampling, the total volume of solution in each of theimpingers was noted and 4 cm3 aliquots removed from each oneand placed in separate 25 cm3 calibrated flasks. To each flask0.1 cm3 of of 1% m/m solution of CTA (BDH) was added,followed by 6 cm3 concentrated sulfuric acid (AnalaR grade,Aldrich). The flasks were carefully shaken and left for thereaction to proceed. Although it has been recommended that theflasks are heated to 95 °C for 15 min to allow the reaction to goto completion, other studies have shown that this is notnecessary as the reaction is exothermic and sufficient heat isgenerated in situ.10,11 We have also investigated the effect ofdifferent heating times and found that there was a slight increasein the absorbance of the solutions with increased heating time.However, the effect was not significant compared to betweenrun experimental variability.12 Thus, the solutions were left tostand and cool down to room temperature. A UV/VIS

Fig. 1 Schematic diagram of the test atmosphere generator. 1, Purified airsupply; 2, flow controllers; 3, flow meters; 4, permeation tube holder; 5,permeation tube; 6, dilution chamber; 7, Dreschel bottles; 8, samplingmanifold; 9, sampling port; 10, charcoal trap; and 11, exhaust.

Analytical Communications, March 1998, Vol. 35 (103–105) 103

Publ

ishe

d on

01

Janu

ary

1998

. Dow

nloa

ded

on 0

1/07

/201

4 05

:50:

41.

View Article Online / Journal Homepage / Table of Contents for this issue

spectrophotometer (Perkin-Elmer Lambda 5, Norwalk, CT,USA) was used to analyse the solutions with the absorbancemonitored at 580 nm. Calibration solutions in the concentrationrange 0.0125–5.0 mg cm23 were prepared using a formalinsolution diluted with a 1% sodium bisulfite solution. Thesestandards were then treated with CTA and analysed in the sameway and alongside the samples in a random order.

Three concentrations of formaldehyde were generated by thetest atmosphere generator (5.92, 3.35 and 1.79 mg m23) bychanging the flow rate of the diluent gas passed through thesystem. Each test atmosphere concentration was sampled at arate of 200 cm3 min21. Different sample volumes were taken togive sampled masses of formaldehyde in the range 10–140 mg,which were analysed as above. A fourth midget impingercontaining 1% m/m sodium bisulfite solution was used as acontrol blank.

The humidity of the system was determined, in the absence offormaldehyde vapour, with a relative humidity and temperatureprobe (Vaisala HMP 124B, Newmarket, UK). With thehumidification line shut off, the relative humidity was deter-mined to be < 0.1%.

Air in the humidification line was wetted by passing itthrough the Dreschel bottles containing distilled deionisedwater, 18 MW. Adjusting the ratios of the flows through thediluent humidification lines varied the relative humidity. Usingthe same formaldehyde concentrations as above, identicalmasses of formaldehyde were sampled with the relativehumidity set at 50 and > 95%. The masses of formaldehyderecovered were determined using the following equation:1

MF = ([H2CO]13 V1 + [H2CO]23 V2)2 2([H2CO]B3 VB) (2)

where [H2CO] is the concentration of formaldehyde in thesolution retained within the impinger, V is the volume ofsolution left in the impinger after sampling and the subscripts 1,2 and B refer to the first, second and blank impingers,respectively. If the mass trapped in the second impingerexceeded 1/3 of that trapped in the first, the sample wasdiscarded;1 breakthrough was deemed to have occurred.

Results and discussion

The masses of formaldehyde recovered by the CTA methodwere correlated to the mass loss measured gravimetrically fromthe permeation tube for each level of relative humidity. Theresults, which can be seen in Table 1, show that an increase inthe percentage relative humidity of the sampled atmosphere wasaccompanied by a decrease in the amount of formaldehyderecovered by the CTA method. This effect was reproducible andindependent of the concentration of formaldehyde that was

sampled. With the relative humidity set at 50%, only 66% of theanalyte was recovered and with it set at > 95% only 36% of theanalyte was recovered.

The regression equations obtained for the CTA method withatmospheres at different levels of relative humidity againstreleased masses of formaldehyde are shown in Fig. 2. It can beseen that linear behaviour is observed at each level of relativehumidity.

The possibility that the observed effect was an artefact arisingfrom different mixing patterns within the sampling manifoldproduced by the humidification line was tested. The CTAmethod was used to analyse samples of formaldehyde at each ofthe three concentrations. Initially all of the diluent flow wassupplied by flow 2, the humidification flow was then introducedwithout the Dreschel bottles, and the two flows combined tocreate the same concentration of analyte. No statisticaldifference between the two sets of data were found and the sametrends in reduced recovery efficiency were not observed, seeTable 2.

In aqueous solution, formaldehyde is in equilibrium withmethylene glycol, at low concentrations of formaldehyde; thiswill affect the kinetics of formaldehyde reactions as theapparent concentration will be reduced. In the gaseousphase, however, little is known about such behaviour.Formaldehyde–water complexes have been investigated usingmicrowave spectroscopy and a strongly bonded dimer with aring-like structure with hydrogen bonds linking a formaldehydeoxygen atom to a water hydrogen (2.68 Å) and a formaldehydehydrogen to a water oxygen (2.01 Å) has been postulated.13

However it is not readily apparent if the observed loss ofreactivity in forming such a complex is sufficient to account forthe observed effect in this study.

Multiple linear regression was applied to the CTA methoddata using the relationship:

MF = A0 + A1MS + A2MW (3)

where A0, A1 and A2 were the unknown parameters and thefollowing relationship was obtained:

MM M

SF W=

+ ×( ) +0 004 1 776

1 025

. .

.(4)

Previous data was selected at random to validate this correctionalgorithm and it was found that formaldehyde concentrationscould be predicted to an accuracy of 3.3%. The linearcorrelation between the concentrations predicted from gravi-metric data (X) and the concentrations obtained from eqn. (4) forthe sampled masses, concentrations and relative humidities

Table 1 Summary of the linear regression analysis of the recovered massesof sampled formaldehyde against released mass at different relativehumidities. All the experiments were undertaken in the temperature range22–27 °C

Relative[H2CO]/ humidity Intercept/ Efficiencymg m23 (%) Gradient mg r2 (%)

5.92 < 0.1 1.223 23.51 0.960 1223.35 < 0.1 0.974 28.7 0.988 971.79 < 0.1 1.011 22 0.991 101

5.92 50 0.6511 21.8 0.975 653.35 50 0.6625 22.5 0.989 661.79 50 0.6697 0.7 0.998 67

5.92 > 95 0.3629 0.21 0.916 363.35 > 95 0.3674 20.45 0.983 371.79 > 95 0.3509 1.28 0.926 35

Fig. 2 Summary of the linear regression analysis showing the effect ofrelative humidity (RH) on the recovery of formaldehyde. (2), RH > 95%,y = 0.34x + 1.46, r2 = 0.99; (8), RH = 50%, y = 0.56x + 3.85, r2 = 0.98;(.), RH < 0.1%, y = 0.92x + 3.02, r2 = 0.95.

104 Analytical Communications, March 1998, Vol. 35

Publ

ishe

d on

01

Janu

ary

1998

. Dow

nloa

ded

on 0

1/07

/201

4 05

:50:

41.

View Article Online

studied gave the equation MS = X + 0.011 with an r2 value of0.997.

Similar susceptibility to relative humidity has been observedwith a 2-hydroxymethylpiperidine-based method, NIOSH2541.14 2-Hydroxymethylpiperidine-coated denuder tubes wereused as a reactive inlet to derivatise formaldehyde. The resultantderivative was trapped on Tenax-TA which was subsequentlyanalysed by thermal desorption and gas chromatography.8Formaldehyde masses in the range 1–6 mg were sampled at arate of 15 cm3 min21 at the levels of relative humidity usedabove.

The results obtained followed the same trends as thoseobserved with the CTA method, with the peak areas obtainedfrom the relative humidities of 50 and > 95% over thetemperature range 22–27 °C reduced to 66 and 39% of the dryair responses, respectively.

The effect of relative humidity on the collection efficienciesof formaldehyde when a 2,4-dinitrophenyhydrazine (DNPH)derivatisation step is used has been reported.15 Glass fibre filtersimpregnated with DNPH were used to sample formaldehydeatmospheres at 20 and 80% relative humidity and 23 °C. In thedry air it was found that the amount of formaldehyde recoveredwas reduced by up to 20%. Grosjean et al.16 suggest that thiseffect may be due to incomplete reaction between theformaldehyde and the DNPH as the pH of the collectionmedium is lower in dry air than humid air. Overall thesefindings indicate an opposite trend to the results reported inTable 1.

These data and observations were surprising in that an effectof such magnitude has not been reported previously. Discus-sions with fellow researchers at NIOSH (National Institute forOccupational Safety and Health, 4676 Columbia Parkway,Cincinnati, OH 45226) suggest that no formal studies looking atpercentage relative humidity effects are known. At this stagetwo possible explanations for these data suggest themselves: theCTA and 2-hydroxymethylpiperidine methods may be suscepti-ble to relative humidity interference because water–formal-dehyde complexes are forming in the gas phase and these inhibitthe derivatisation process; or adsorbed water on the surfaces ofthe sampling train is removing formaldehyde via a denuder tubetype mechanism.17 Whatever the reason, both considerationsneed to be tested in follow up studies.

These preliminary data indicate that relative humidity needsto be accounted for in formaldehyde monitoring activity. Alarger follow up study is under way involving both of themethodologies reported above as well as DNPH and fuel celltechniques. This follow up work will seek to: repeat and confirmthe findings reported here using alternative designs of sampling

system and sampling manifold; and establish if the interferenceobserved so far is present in other methods of formaldehydedetermination.

The work undertaken in this paper has been supported byEPSRC and Jones Chromatography through the CASE awardprogramme and UMIST through a Graduate School ResearchScholarship. The authors would like to thank the staff atNIOSH, and in particular Dr. E. R. Kennedy, for helpfuldiscussions.

References

1 National Institution of Occupational Safety and Health, NIOSH,Manual of Analytical Methods, Method 3500, National Institution ofOccupational Safety and Health, Washington, DC, 1989.

2 Graelel, T. E., in Chemical Compounds in the Atmosphere, AcademicPress, London, 1978, p. 158.

3 Carlier, P., Hannachi, H., and Mouvier, G., Atmos. Environ., 1986,20, 2079.

4 Fielder, R. J., Toxicity Review 2, Formaldehyde, Health and SafetyExecutive, HMSO, London, 1981.

5 Light, E. N., Formaldehyde, Analytical Chemistry and Toxicology,ed. Turoski, V., Advances in Chemistry Series 210, AmericanChemical Society, Washington, DC, 1985.

6 Health and Safety Executive, HSE EH40/91 Occupational ExposureLimits 1991, Health and Safety Executive, London, 1991.

7 Bardana, E. J., and Montanaro, A., Resp. Cut. Innum. Annul. Allergy,1991, 66, 441.

8 Goelen, E., Lambrechts, M., and Geyskens, F., Analyst, 1997, 122,411.

9 Thomas, C. L. P., McGill, C. D., and Towill, R., Analyst, 1997, 122,1471.

10 West, P. W., and Sen, B., Z. Anal. Chem., 1956, 153, 177.11 Altshuller, A. P., Miller, D. L., and Sleva, S. F., Anal. Chem., 1961,

33, 621.12 Meunier, F., M.Sc. Dissertation, 1996.13 Lovas, F. J., and Lugez, C. L., J. Mol. Spectrosc., 1996, 179, 320.14 National Institution of Occupational Safety and Health, NIOSH,

Manual of Analytical Methods, Method 2541, National Institution ofOccupational Safety and Health, Washington, DC, 1989.

15 Levin, J., Andersson, K., Lindahl, R., and Nilsson, C., Anal. Chem.,1985, 57, 1532.

16 Grosjean, E., and Grosjean, D., J. Am. Chem. Soc., 1996, 30, 859.17 Thomas, C. L. P., and Alder, J. F., Anal. Chim. Acta, 1993, 27,

179.

Paper 7/08807HReceived December 8, 1997Accepted January 22, 1998

Table 2 Summary of the investigation to determine if the observed behaviour could be attributable to experimental artefacts. Experiment A was run usingthe third flow line, but with no humidification taking place and experiment B did not include the third flow line. Critical ‘t’ values are at the 95% confidencelimit

Mean Meanmass loss Standard mass loss Standard

Theoretical recorded deviation recorded deviation Criticalmass loss/ experiment experiment experiment experiment Experimental ‘t’mg A/mg A B/mg B ‘t’ value value

2 3 2.5 1 0.8 1.62 2.364 4 1.9 3 1.2 0.42 2.23

11 12 6.6 11 4.6 0.42 2.23

Analytical Communications, March 1998, Vol. 35 105

Publ

ishe

d on

01

Janu

ary

1998

. Dow

nloa

ded

on 0

1/07

/201

4 05

:50:

41.

View Article Online