WP4- NA4: Trace gases networking: Volatile organic carbon and nitrogen oxides Deliverable D4.9: Final SOPs for VOCs measurements Summary:

Transcription

1 WP4- NA4: Trace gases networking: Volatile organic carbon and nitrogen oxides Deliverable D4.9: Final SOPs for VOCs measurements Summary: This SOP provides a guideline for good measurement practice for the analysis of volatile organic compounds (VOCs) under the EU FP7 infrastructure project ACTRIS. Only active sampling is part of this SOP. For passive sampling respective guidelines from the EU should be used. The SOP contains the following topics: 1. General introduction Data Quality Objectives VOCS Measurement Setup Facility requirements Personnel requirements Occupational health and safety Instrumentation requirements Air inlet and sample lines Associated key measurements Environmental issues that affect ACTRIS stations and VOCs observations Sampling Location of the inlet Off-line sampling Sampling inlet lines (NMHCs and OVOCs) Adsorption tubes Stainless steel canisters OVOCs (DNPH) On-line sampling/quasi continuous observations Measurement techniques GC technique Removal of water/ozone/carbon dioxide/particles Water removal/management Ozone removal Carbon dioxide removal Particle filters Sample Preconcentration Capillary columns for GC analysis of VOCs and OVOCs Detection PTR-MS (provided by R. Holzinger, T. Petäjä, S. Dusanter)... 16

2 6. Reference materials Quality assurance Calibration measurements Method for measurements of Laboratory/Working standards and Target gases Zero Gas Method for Detecting Effects of Ozone on Reactive Compounds Audit procedures Measurement protocol Measurement uncertainties Calculation of mole fractions for linear detection systems Determination of Precision Determination of Uncertainty Determination of detection limit Data Management Data evaluation FID: effective carbon number Time series of calibration gas measurements Target gas measurements Results of standard addition measurements Data checks of final mole fraction data in time series QC in xy-plots (used at Rigi, Switzerland by Empa) QC in repeatability and reproducibility: Recommended QC and minimum QC, thresholds for flagging the data Metadata Ancillary data Data archiving at the station or laboratory Data submission References Appendices... 43

3 1. General introduction Volatile organic compounds (VOCs) consist of low-boiling non-methane hydrocarbons (alkanes, alkenes, alkynes, aromatics, terpenes) and oxygenated hydrocarbons, such as alcohols, ketones, or aldehydes. When released into the air VOCs play an important role in atmospheric chemistry and in the oxidizing power of the atmosphere, which then affects climate and air quality. VOCs are emitted by the biosphere and by anthropogenic activities, such as motor vehicle exhaust and solvent usage. A complex mixture of several hundred VOCs is emitted with half-lives ranging from several months in the case of ethane, to hours for the most reactive ones, such as isoprene or anthropogenic alkenes. VOCs are removed from the atmosphere primarily by their reaction with hydroxyl radicals a process which forms intermediate oxygenated organic compounds. The scientific background for the need of VOCs monitoring in global and regional networks has been extensively presented (for example in the GAW Reports 111 (WMO, 1995), 171 (WMO, 2007a), and 204 (WMO, 2012), Helmig et al., 2009). In populated areas VOCs and their degradation products are responsible, together with NOx, for the photochemical formation of ozone (O 3 ) and other photooxidant pollutants including secondary organic aerosol (SOA). Thus, they couple into photochemical ozone production, aerosol formation, and cloud processes and thus impact air quality and climate. Therefore, measurements of VOCs are essential and are among the long-term monitoring parameters in GAW (GAW Report 172 (WMO, 2007b)) and regional programs like EMEP. Only a few VOCs (e.g. formaldehyde) can be observed by satellite instruments. Therefore long-term in-situ measurements of these VOCs are essential for deriving trend estimates of globally representative mole fractions and source contributions. This standard operation procedure (SOP) covers exclusively ground-based in-situ measurements of VOCs. Table 1 The list of priority VOCs focused in the GAW report No. 171 (WMO, 2007a) Molecule Lifetime (OH= 1E6 cm-3) Importance to GAW Steel flask 1. Ethane 1.5 months source of methane natural sources biomass burning fossil fuel ocean production (S. hemisphere) trend in size of seasonal cycle indicator of halogen chemistry 2. Propane 11 days source of methane natural sources biomass burning fossil fuel ocean production (S. hemisphere) 3. Acetylene 15 days motor vehicle tracer biomass burning tracer ratios to other hydrocarbons trends Glass flask Analysis Method Network Type GC-FID global GC-FID global GC-FID global 4. Isoprene 3 hours biosphere product?? GC-FID Africa sensitive to temperature/land PTR-MS S. and N. use/climate change America O 3 precursor Europe oxidizing capacity precursor to formaldehyde 5. Formaldehyde 1 day indicator of isoprene - - DOAS Small 1

4 oxidation biomass burning comparison with satellites trends 6. Terpenes 1-5 hours precursors to organic aerosols 7. Acetonitrile year biomass burning indicator biofuel burning indicator 8. Methanol 12 days sources in the biosphere (methane oxidation) abundant oxidation product 9. Ethanol 4 days tracer of alternative fuel usage 10. Acetone 1.7 abundant oxidation product months free radical source in the upper troposphere 11. DMS 2 days major natural sulphur source sulphate aerosol precursor tracer of marine bioproductivity 12. Benzene 10 days tracer of combustion biomass burning indicator 13. Toluene 2 days ratio to benzene used for air mass age precursor to particulates 14. Iso/n-butane 5 days chemical processing indicator lifetime/ozone production 15. Iso/npentane NO 3 chemistry 3 days ratio provides impact on GC-FID is Gas Chromatography Flame Ionization detection GC-MS is Gas Chromatography - Mass Spectrometry DOAS is Differential Optical Absorption Spectroscopy PTR-MS is Proton Transfer Reaction Mass Spectrometry indicates state of current pracgce Deliverable WP4 / D GC-MS PTR-MS -? GC-MS PTR-MS -? GC-MS PTR-MS -? GC-MS PTR-MS?? GC-MS PTR-MS?? GC-MS PTR-MS? GC-FID GC-MS -? GC-FID GC-MS GC-FID GC-MS GC-FID GC-MS number of sites in Tropics for compari son with satellite s Selected sites in forested areas Global Global Global Global Global Global Global Global Global There are three groups of VOCs that are distinguished in current literature, GAW Reports and in this SOP: the non-methane hydrocarbons (NMHCs), the oxygenated volatile organic compounds (OVOCs) and biogenic VOCs (mainly isoprene and the group of monoterpenes, often summarized as BVOCs). Priority substances of these groups have been identified in GAW Report 171 (WMO, 2007a) and detailed guidelines for their measurements are provided in the respective report, following the general quality assurance (QA) recommendations and the strategic plan by GAW (GAW Report 172 (WMO, 2007b)). As analytical systems for the measurement of the different groups of VOCs are generally capable of analyzing not only the priority species identified in GAW Report 171 but also a list of chemically similar compounds, this SOP covers a broader range of compounds than specified in GAW Report 171 (Table 1). This is in line with the EMEP objectives and addressed in the ACTRIS Description of Work. The measurement of atmospheric VOCs is mostly done by gas chromatographic methods. However, for carbonyl OVOCs DNPH (dinitrophenylhydrazine) cartridge sampling and high-performance liquid chromatography (HPLC) exist for certain VOCs and the relatively new on-line method proton transfer 2

5 reaction mass spectrometry (PTR-MS). As gas chromatography (GC) is most widely used, we focus on this method in this SOP, but other techniques are discussed, too. The measurement of VOCs by GC is generally performed in a series of steps with (1) intake manifold and sampling line, (2) traps to remove water and ozone, (3) pre-concentration, (4) gas chromatographic separation, (5) analysis in detector, and (6) data processing and data delivery. A sample of atmospheric VOCs can be introduced to the analytical system directly from ambient air, a canister or a preconcentration tube. The sample is normally passed through a moisture and/or ozone removal system and then concentrated on an adsorbent medium that is cryogenically cooled using liquid nitrogen, liquid carbon dioxide, or thermoelectric closed-cycle coolers. The sample optionally can be refocussed cryogenically by a cooled secondary trap to narrow the band width for injection onto the capillary GC analytical column. The concentrated sample is then thermally desorbed into the analytical GC column and finally analysed by flame ionization detection (FID) or mass spectrometry (MS) (or any other suitable detectors). For the quantification of VOCs a calibration standard should be used, which either contains VOCs in ambient air, or contains artificially mixed VOCs in nitrogen (N 2 ). Mole fractions should be in the region of the expected ambient mole fractions. If needed the standard has to be diluted with zero air or N 2 to a mole fraction, which is in the mole fraction range measured at the specific station. 3

6 2. Data Quality Objectives Data quality objectives (DQOs) define qualitatively and quantitatively the type, quality, and quantity of primary data required and derived parameters to yield information that can be used to support decisions. In WMO/GAW DQOs were for example introduced in the strategic plan (WMO, 2001). For VOCS measurements the DQOs have the objectives to (a) detect long-term changes in background mole fractions and (b) quantify year-to-year variability. This section quantifies what are tolerable levels of uncertainty for reaching these goals. The following DQOs have been approved by the GAW-VOCS expert group: The enhanced ACTRIS DQOs defined in Table 2 are the objective for good performing stations. Within ACTRIS, generally the GAW DQOs shall be reached. These stated DQOs are valid for individual ambient air measurements. This is different from the interlaboratory comparability objectives used in the greenhouse gas community where the objectives refer to uncertainties in measurements of calibration standards comprising multiple measurements. During the ACTRIS non-methane hydrocarbon intercomparison most stations reached the GAW and ACTRIS DQOs when analysing NMHCs in nitrogen, but for compressed air (whole air) more scatter was observed and only a few stations reached the ACTRIS DQOs (Hoerger et al. 2014). The long-term objective is that all stations measure ambient air within the ACTRIS DQOs. Table 2 Data quality objectives for the measurements of VOCs GAW uncertainty GAW repeatability ACTRIS uncertainty ACTRIS repeatability alkanes 10% 5% 5% 2% alkenes incl. isoprene 20% 15% 5% 2% monoterpenes 20% 15% 10% 5% alkynes 15% 5% 5% 2% aromatics 15% 10% 5% 2% mole fraction <0.1 nmol/mol (ppb) (monoterpenes) 0.02 ppb ppb ppb (0.010 ppb) ppb (0.005 ppb) 4

7 3. VOCS Measurement Setup 3.1 Facility requirements Facility requirements include 24-hour available electricity and communications, a secure environmentally conditioned building suitable for the instruments and staff and ease of access. The facility and equipment should be suitable to sustain long-term observations with greater than 90% data capture (i.e. <10% missing data). The air sampling should be structured in a way to avoid local contamination sources. The laboratory building and inlet location on site should be set upwind of any other buildings, garages, parking lots, generators, other emission sources any nearby areas where fossil fuels or biomass may be combusted and where intensive agriculture is undertaken. Station personnel should also remain downwind of the sampling laboratory and refrain from smoking as necessary. Within the facility, temperature control and clean lab environment are required. Instrumentation should not be exposed to sunlight. 3.2 Personnel requirements Each set of measurements at an ACTRIS station should be conducted under the guidance of a designated Responsible Investigator (RI). For VOCs, it is recommended that the RI have training in atmospheric chemistry, meteorology and atmospheric composition monitoring. There are requirements for technicians with skills in (1) analytical chemistry, particularly atmospheric composition, (2) electrics and electronics, and (3) IT, particularly instrument control, data acquisition and data processing. It is recommended that station staff participate in the ACTRIS workshops, GAWTEC training programme and other GAW specialist activities where appropriate. Provision should be made for back up staff to cover the periods when regular staff is away at training, leave etc. 3.3 Occupational health and safety The VOCs program includes use of equipment that can cause the following occupational health and safety issues: High voltages; High-pressure gas lines (for example associated with the zero air generator); Noise; Heavy and awkward equipment. Other hazards may occur. Appropriate occupational health and safety information, protective equipment and training are required. 3.4 Instrumentation requirements The following instrumentation is required for a reliable long-term VOCs monitoring station in ACTRIS: VOCs system and suitable inlet as described in Sections 4 and 5. This system must be calibrated as recommended in Sections 7 and 8 of this SOP; Zero air supply that includes H 2 O, VOCs, O 3 and NO x removal (see Section 7); Inlet line and filter both inert to VOCs; Instrument control and data acquisition interface; 5

8 Computer; Internet connection/remote computer access; Deliverable WP4 / D4.9 Uninterruptable power supply. Equipment varies in specification and performance. The WCC, existing long-running ACTRIS and GAW stations and GAWTEC can provide advice on instrumentation that has performed successfully. Manufacturers instrument manuals should be available on site for all instruments used at the site Instrument replacement As long as an instrument performs within the specifications and the DQOs (Section 2), there is no necessity for replacement. If the instrument performance requires a replacement, the new and old system should run parallel for some time (6 month) if possible. Since IT equipment is subject to fast evolution, back-up equipment should be available and appropriate updates should be carried out depending on the availability of financial resources Instrument control and data acquisition software Instrument control and data acquisition usually depends on the available manufacturers software for the VOCS instrument. 3.5 Air inlet and sample lines The air inlet is an essential component of the ACTRIS monitoring system and any compromises made with regard to the inlet will affect all subsequent data. There are two key components of the inlet system, the location of the inlet and the materials of the inlet. In analytical chemistry terminology, the location of the inlet is an aspect of sampling and the passage of the air through the inlet corresponds to pre-treatment of the sample. See Section Associated key measurements Key measurements that will help in the interpretation of VOCs measurements include those used for processing the VOCs data, data selection and those related to VOCs chemistry. To understand the influence of nearby sources, to undertake data selection according to meteorological conditions and to quality control, the following additional parameters are useful: Wind speed and direction; Air temperature; Humidity; Particle number concentrations; Carbon monoxide mole fraction; Nitrogen oxides mole fractions; Radon concentration. To interpret the atmospheric chemistry processes affecting the observed ozone mole fractions, the following parameters are useful: Nitrogen oxides mole fractions; O 3 /CO/CH 4 /OH mole fractions; Water vapour concentration; Air temperature; 6

9 Spectral distribution of solar radiation (suitable for determining molecular photolysis rates)/solar radiation. Where VOCs measurements are undertaken at ACTRIS stations, consideration should be given to measurement of these additional parameters. The measurement techniques for these parameters are presented in GAW Report No.143 Global Atmosphere Watch Measurements Guide (WMO 2001b)) and in individual measurement guidelines (WMO, 2007a; WMO, 2010b; WMO, 2010c; WMO, 2011a). 3.7 Environmental issues that affect ACTRIS stations and VOCs observations The environmental conditions/hazards that affect VOCs observations include the following: Inlet blockage at polar and high-altitude sites, due to ice riming and blowing snow; Pollution events nearby roads, agriculture, biomass burning, industry etc.; Access limited by environmental conditions such as flooding, severe weather etc.; Lava flow for stations located on active volcanoes; Tourist activities. Consideration should be given to minimising the effect of the factors listed above where possible when setting up the station, while it is clear that the impact of natural hazards cannot be completely avoided. 7

10 4. Sampling The air from which VOCs are analysed can be sampled on-line at the measurement site or off-line, using either adsorption tubes or stainless steel canisters. Off-line samples are subsequently transported to the lab where they are analysed. The specific requirements of the different methods are described below. 4.1 Location of the inlet The height of the air sample inlet is critical to this sampling of representative air. The optimum inlet height depends on the surrounding area (type vegetation, orography, soil, water, snow). New stations should, if possible, for a trial period sample VOCs at 2-3 different heights to determine which inlet height is suitable. 4.2 Off-line sampling Sampling inlet lines (NMHCs and OVOCs) For the sampling of NMHCs the inlet line should be either Silco-treated steel or made of stainless steel. In the case of stainless steel, the line has to be heated up to 70 C to prevent condensation of VOCs on internal surfaces (Hopkins et al., 2011). Transfer lines for the analysis of OVOCs should be either silco-steel, PFA (perfluoralkoxy), PEEK (polyether ether ketone), or electro-polished heated stainless steel but not untreated stainless steel. Silco-treated steel should be humidified before first usage (e.g. by passing humid ambient air). The inlet line has to be as short as possible and the diameter should not be larger than 1/8 inch in order to minimize the dead volume of the sampling line unless it is permanently flushed. The residence time in the inlet should not exceed a few seconds. It is recommended to have a mesh with a maximum of 5 µm mesh size in the line in order to hold back particles. This has to be exchanged regularly. Experience from an urban site (Zurich) is that the grid has to be exchanged every 4 months Adsorption tubes Off-line sampling of NMHCs by adsorption tubes is a well-established method. Different providers have commercially available products. For this SOP it is advised to follow the specific procedures of the individual products. It has to be proven, however, that artefacts due to adsorption tube blanks (especially for aromatic compounds) are reproducibly low compared to the range of concentrations encountered at the sampling site. Due to the mentioned blank issues, it is generally not recommended to use the adsorption tube technique in very clean air. Moreover this method is not recommended within the EMEP or GAW programmes, and not yet tested in intercomparisons. However, for some compounds, like terpenes, adsorption tubes or canister measurements are useful to characterise these compound class, which are often not routinely analysed with on-line GC-systems Stainless steel canisters In the GAW Report No. 204 (WMO, 2012) a new Standard Operation Procedure (SOP) valid in the WMO GAW network was described. For the analysis of NMHCs in canister samples, this procedure can be applied. This SOP is based largely on the recommendations from the Accurate Measurements of Hydrocarbons in the Atmosphere project AMOHA (Plass-Dülmer et al., 2006) and from US-EPA (1998, 1999) on determination of VOCs in ambient air. Beside this SOP, many other 8

11 9 Deliverable WP4 / D4.9 methods exist for whole air sampling for VOCS analysis (e.g. glass flask air sampling in the US National Oceanic and Atmospheric Administration Cooperative Global Air Sampling Network (Pollmann et al., 2008)). Generally, the use of materials other than ultra-pure stainless steel, glass, silica coated stainless steel, PFA, and PTFE (polytetrafluorethylene) should be avoided for the measurement of NMHCs in air samples. Especially, plastics other than PFA and PTFE shall not be used to prevent memory and outgassing effects. The recommendations in the SOP constrain on electropolished stainless steel (ss) canisters. A variety of ss canisters with one or two valves may be purchased from several suppliers, e.g. Restek or TO-Can Air Monitoring Canisters. The use of two valve canisters allows more flexibility in air sampling and is recommended. The inner surface of canisters should be passivated, e.g. electro-polished and stainless steel valves (e.g. Swagelok) shall be used to seal the canisters inlets and outlets, respectively. The canister conditioning and sampling procedures are described in detail in GAW Report 204 (WMO, 2012). In this report problems of this method are discussed, too. Briefly, major problems usually arise from inappropriate conditioning, canister leaks, adsorptive losses of C 7 and higher boiling compounds, long storage times, and artefact formation of low boiling alkenes. These problems are linked to the canisters and depend strongly on the type of canisters used. Thus, this technique is recommended only for C 2 -C 6 alkanes, isoprene and benzene. However, when canisters of the new generation like Silonite (Entech) or SilcoCan (Restek) are used improvements for higher boiling compounds were found. Nevertheless, this has to be tested in detail for the used canister OVOCs (DNPH) Off-line sampling of OVOCs by DNPH-coated samplers with subsequent liquid chromatography (LC) is a well-established method within EMEP, therefore the chapter 3.8 of the EMEP manual for sampling and chemical analysis (Determination of aldehydes and ketones in ambient air, Revision Nov 2001) should be used. The SEP-PAK DNPH-Silica Cartridges (WATERS ) are recommended to measure individual aldehydes (from C 1 -C 6 ) and ketones (C 3 -C 5 ) over a concentration range (0,05-10 µg/m 3 ) encountered in background atmosphere. Since the blank can differ from a cartridge to another, it is strongly recommended to determine the blank by analyzing a set of minimum 7 cartridges per batch. For each measurement, the mean bank value representative of the batch is then systematically subtracted to the sample. Air flow through the DNPH cartridge may change during the sampling due to particles. Then the sampling device should include a flow controller and a pump efficient enough to maintain a constant flow rate of 1500 sccm over a minimum sampling time of 3 hours. The post sampling procedure (extraction) is fully described in the above mentioned EMEP Manual. The extract is analyzed by a high performance liquid chromatography (HPLC) equipped with a quaternary solvent supply system and a UV detector. The calibration procedure mainly consists in analysing a certified standard solution (e.g. SUPELCO Carbonyl-DNPH Mix). However, it is recommended to validated the whole method and determine accuracy and precision using a synthetic mixture obtained by diluting a certified gaseous standard (OVOCs standard) with humid air. 4.3 On-line sampling/quasi continuous observations On-line sampling avoids storage issues and minimizes leak issues, however, requires an analytical system at the sampling site and thus restricts the sampling intervals to the capabilities of the analytical system. The air sample is directly transferred via a sampling line into the VOCs instrument. Concerning sampling inlet lines see Section

12 5. Measurement techniques For on-line and off-line analyses of VOCs from ambient air GC systems are the method of choice. The advantages are medium cost, high sensitivity, excellent reproducibility, and, depending on the applied chromatographical details, large resolving power. Disadvantages are the need for well-trained and experienced operators, the restricted time resolution, and problems in analyzing more polar, surfacesticky compounds. An alternative discussed in section 5.2 are the PTR-MS systems capable of high time resolution measurements which are surface-contact free and allow analyses of the aforementioned problematic polar compounds, e.g. OVOCs. Disadvantages of PTR-MS systems are the high cost of the instrument, the high skills required to operate the systems and the fact that PTR-MS can practically only analyse OVOCs and certain unsaturated NMHCs. Furthermore, it cannot separate isobaric compounds like different monoterpenes. 5.1 GC technique As VOCs are only occurring in the atmosphere in the range of pmol/mol (ppt) up to some nmol/mol (ppb) they have to be preconcentrated before the analysis, using GC-FID or GC-MS. Preconcentration of VOCs is performed on a trap which contains enough of a suitable material or a combination of different adsorbents for fully retaining VOCs at a given temperature (section 5.1.2). The preconcentrated compounds are subsequently injected onto the analytical column where they are separated depending on the characteristics of the chosen column (Section 5.1.3). In the final step the compounds reach the detector (FID or MS, see Section 5.1.4) Removal of water/ozone/carbon dioxide/particles Prior to preconcentration, additional trapping devices may be required: Water (H 2 O) in ambient air affects the adsorption capacity of the preconcentration trap (see Appendix 2), the chromatography (peak shapes and retention times) and leads to ice formation in the preconcentration unit, when temperatures <0 C are applied. Ozone may react with unsaturated NMHCs such as alkenes (e.g. ethene, isoprene, monoterpenes) during the preconcentration step and form OVOCs. Furthermore, ozone could react with the adsorbent material itself (see Appendix 1). CO 2 can distort the chromatography or effect detector sensitivity in case of sample preconcentration at adsorption temperature <-78 C. Furthermore, particle filters should be used to avoid contamination of the system with particles Water removal/management Water management can be achieved by different methods such as Nafion dryer or a water trap (Table 3). Regardless, which water management system is chosen, its efficiency, potential artefacts (e.g. blind values) and the recovery of water soluble compounds (e.g. alcohols) needs to be tested (see Standard-addition measurements in Section 7.1.3). If hydrophobic adsorbents (see Appendix 2) at above ambient air temperature are used in the preconcentration trap, prior water removal can be neglected. In combination with a dry purging step - flushing of the preconcentration trap in the sample flow direction with dry gas (e.g. purified helium (He) 5.0 or He 6.0) subsequent to sampling residual water is further removed. This kind of sampling is applicable for C 4 and higher boiling compounds, and is regularly used for BVOCs sampling. Table 3 Methods to remove Water from the sample. Method Comments Recommended for 10

13 Nafion Dryer with a volumetric counter-flow of dry air or N 2, which is around 3 times higher than the flow of humid ambient air Water T < T ambient removes H 2 O effectively and substantial parts of the polar OVOCs and monoterpenes. Potential artefacts in C 2 -C 4 - alkenes may occur depending on the status of the Nafion Dryer. ((Gong and Demerjian, 1995; Plass-Dülmer et al., 2002) and references therein). H 2 O is adsorbed or frozen-out but not the analytes. The dew point should be measured and it should be at least 10 C lower than the trapping temperature of the preconcentration unit. For cryogenic trapping of VOCs, the dew point should be below -30 C. In NMHCs, BVOCs, and selected OVOCs analysis, freeze-out water traps are widely used (e.g. Cape Verde (Hopkins et al., 2003)). Deliverable WP4 / D4.9 NMHCs C 2 -C 7 ( sometimes C 8 ) NMHCs BVOCs OVOCs (risk of potential losses of highly water soluble compounds like alcohols (e.g. methanol*)) *When measuring highly water soluble compounds (e.g. methanol), the recovery of these compounds needs to be tested Ozone removal To avoid artefact formation from the reaction of unsaturated VOCs with ozone (O 3 ), several methods are available to eliminate ozone from the sample. Table 4 lists the most common methods. A more thorough compilation of available methods and their evaluation can be found in Appendix 1. Table 4 Ozone removal methods and recommendations. Method Comments Recommended for (e-polished) stainless Losses of OVOCs can occur (Hopkins et al., 2011; Englert et T > 70 C al., 2014) Titration with NO Slow reaction, alcohol losses were observed, poisonous NO (O 3 + NONO 2 ) reactant (Helmig, 1997;Komenda et al. 2003; Legreid 2006) into ambient air flow Cartridges filled or filters impregnated with sodium thiosulfate (Plass-Dülmer, 2002) (Na 2 S 2 O 3 ) potassium-iodide (KI) Has to be implemented after water removal**. Blank values for formaldehyde, acetaldehyde, alcohol losses (Helmig and Greenberg, 1994; Leibrock, 1996) sodium sulfate (Na 2 SO 3 ) Removes methyl vinyl ketone (MVK) and macrolein (??) (MAC), efficiency depends on H 2 O vapour content of air stream, humidity increases efficiency (Helmig, 1997). Manganese-Oxide Work in progress *Recovery has to be tested. NMHCS, OVOCs* NMHCs, BVOCs, OVOCs* NMHCs, BVOCs NMHCs,? NMHCs,? ** the ozone scrubber is efficient with a minimum moisture in the gas stream (Kliendienst et al, 1995). However special cautions have to be taken when the scrubber is used for extensive periods of time at high RH. The preparation procedure includes drying the tube properly Carbon dioxide removal 11

14 Methods to remove carbon dioxide (CO 2 ) from the ambient air flow are listed in Table 5. Table 5 CO 2 removal Deliverable WP4 / D4.9 Method Comments Recommended for Cartridge with Ascarite substance is hygroscopic, trap should be installed behind a water trap to avoid liquefaction; artefact might be possible Preconcentration trap slow heating of preconcentration trap to a temperature high enough for the CO 2 to be released but not for the analytes. To be determined NMHCs BVOCs Particle filters In order to avoid contamination of the system with particles, filters (Table 6) can be used in the analysis of VOCs. Table 6 Particle filters used in GC systems. Method Comments Recommended for PTFE membrane filter Pore size: µm, Metron Technology, Aschheim, Germany (used at Hohenpeißenberg) No artefacts are detected for recommended compounds (see Section 6.6.). Not suitable for OVOCs. NMHCs (C 2 -C 14 ) BVOCs Sample Preconcentration A compilation of different trapping adsorbents and their usage is provided in Appendix 2. Either cryogenic adsorption on glass beads, a combination of week adsorbents with low sub-ambient temperature or stronger adsorbent with higher, up to ambient temperature can be chosen, often also multi-bed adsorbents with increasing adsorbent strength in sampling flow direction are used. For each system break-through volumes have to be tested, using either increasing amounts of humidified synthetic standards or of ambient air spiked with standards (Section 7.6). For the trapping procedure a pump should be used downstream of the trap connected to a critical orifice or a mass flow controller (or any other suitable instrument) to regulate the flow through the trap. It is essential to determine the sampling volume with low uncertainty either by regularly calibrated mass flow controllers or by pressure rise measurement in a defined reference volume. If the pump is used before the trap it has to be ensured that no additional contamination is produced by the pump. After trapping, the trap should be flushed in forward mode at the same temperature for an adequate amount of time to allow the purge out of remaining water and potentially adsorbed gases (e.g. CO 2, noble gases) from the trap. Release of the analytes from the trap is normally done by heating the trap (either by ohmic resistance or by other means of heating) in counter-flow. The final temperature should be reached as fast as possible and should be high enough to release all analytes. Analytes are transferred to the gas chromatography system by carrier gas flow. After transfer of the analytes the trap should be 12

15 reconditioned (e.g. by flushing it further with carrier gas and heating it to a higher temperature than needed to release the analytes, the flow of cleaning gas is vented to the environment). In case that analyte injection is not rapid enough to obtain sharp peaks which may be due to large trap volumes or slow heating rate of the trap, a second focusing trap should be installed between preconcentration and column. This again may be adsorptive or cryogenic but needs to have a substantially smaller internal volume than the preconcentration trap. Following systems (Table 7) have taken part successfully in an ACTRIS intercomparison experiment (Hoerger et al. 2014): Table 7 List of successfully employed GC systems. Adsorbents Temperature and flows Sample custom made thermo desorption systems Glass beads in 1/8 Ads C and 50 ml/min Silcosteel tubing (LN2 cooling) Des. 340 C and 5ml/min Fused Silica beads, Caboxene 1003,Carboxene 1016, Carboseive S-III in stainless steel tube Carbopack BHT Tenax TA/ CarbopackX/Carboxene569 in fritted glass tube Ads. -45 C Des. 235 C Ads C Des. 200 C Ads. 30 C, 80 ml/min Des. 200 C, 20 ml/min** Commercial thermo desorption systems Markes UNITY TD Ads. -20 C, Carbopack B, Des. 350 C, Carboxen 1000 Volume Systems Recommended for 750ml Hohenpeißenberg NMHCs (C 2 -C 8 ) (Plass-Dülmer et al., 2002)* 600ml Rigi, EMPA NMHCs (C 2 -C 8 ) 400ml WCC-VOC, KIT NMHCs (C 2 -C 6 ) Garmisch 1500ml Hohenpeißenberg BVOCs (sabinene depletion on TenaxTA**), NMHCs (C 4 -C 14 ) 1000ml Cape Verde, (Hopkins et al., 2003) EMPA NMHCs (C 2 -C 8 ), OVOCs, ENTECH TD Glass beads Ads C, Des. 70 C,**** 360ml EMD NMHCs (C 2 -C 8 ) * Reference systems during ACTRIS intercomparison ** Refocussing on Methyl Silicone Capillary, XX -180 C 20ml/min, des. 60 C, 2.5ml/min *** needs to be tested regularly, depletion process increases with age of td tube **** Refocussing on glass beads, Tenax, Ads. -50 C, Des. 220 C Split injection During the ACTRIS NMHCS intercomparison experiment systems with split injection seemed to have a poorer performance than systems without (Hoerger et al., 2014). Reasons (e.g. variable split flows) are not understood, yet. Currently, it is recommended to inject directly onto the column, without split injection Capillary columns for GC analysis of VOCs and OVOCs Capillary columns exhibit better separation efficiencies and higher inertness compared to packed columns. Despite their lower capacity they are suitable for most applications in trace gas analysis. 13

16 There are two types of capillary columns that are most widely used for the analysis: PLOT (Porous Layer Open Tubular) and WCOT (Wall Coated Open Tubular) columns. Several possible analytical columns are listed in the Appendix 3. Table 8 lists a number of columns which have been successfully employed in VOCs analysis. Table 8 List of VOCS columns. VOCS Column T range NMHCs C 2 -C 8 AL 2 O 3 /KCL PLOT OVOCS BVOCS, NMHCS C5 and higher -100 C 200 C CP-LOWOX* 0 C 350 C CP-Porabond-U DB-1** DB-5** -100 C 300 C -60 C 350 C * or similar columns as listed in table 1 in Appendix 3 ** or similar columns as listed in table 2, in Appendix 3 Typ. Dim 50m x 0.53mm 20m x 0.53mm 25m x 0.32mm Comments strong selectivity but H 2 O dependent retention time shifts Long lifetime, retention times stable but co-elutions with aliphatic NMHCs (e.g.) 50m x 0.32mm Co-elution with 50m x OVOCs 0.22mm Citation Plass-Dülmer et al., 2002 Hopkins et al., 2003 Legreid, 2006 and Englert et al. (to be submitted) Riemer et al., Detection Two detection systems are mostly used in VOCs analysis: Flame Ionization Detection (FID) and/or Mass Spectrometry (MS). Advantages and disadvantages of these detectors are listed in Table 9. Table 9 Advantages and disadvantages of detection systems. FID Advantages + sensitive, robust, simple in design and easy to use + very stable performance with typically less than 2% sensitivity drift over one month + response is proportional to the mass or carbon number and allow easy quantification + with the effective carbon number (ECN) concept of the response, they allow for effective QA (see Section 8.1.1) + not sensitive to traces of water, N 2 and O 2, and noble gases from the sample gas + less expensive Disadvantages - not substance-specific Co-elution of peaks GC MS + compound identifying capabilities + substance-specific: overlaying peaks are detectable by compound specific mass tracks - variable sensitivity requires more frequent calibration measurements - instruments needs regular tuning - expensive A FID is the favourable detection system whenever identification can be achieved simply based on the retention times. If the resolution of the chromatographic system does not allow unambiguous 14

17 identification of different compounds based on retention time alone, a mass spectrometer is recommended as detector for its compound identifying capabilities Operating conditions: Flame Ionization Detector (FID) The operation principle is based on the ionization of organics in a hydrogen (H 2 ) flame. A FID needs thus air and H 2 to produce the flame and a make-up gas for proper operation; the flow rates should be well controlled to achieve stable operation of the detector. Essential is to have low VOCs levels or at least low fluctuation in VOCs levels in the operating gases. Table 10 lists the suitable operating condition for FIDs: Table 10 Operating conditions for FID GAS Supply Flow rate* Temperature Air Synthetic air (quality 5.0) or ambient air catalytically cleaned ml H 2 (Pd or Pt catalyst at 350 C-450 C) T FID **>= T column,max to avoid or minimize Cylinder (H 2 quality 5.0) or H 2 30 ml deposition of column residues generator Make Up Gas (e.g. N 2 ) Cylinders, grade 5.0 or higher 30 ml *The suitable flows might vary depending on the FID used; it is important to check the total flows of the individual gases, including the carrier gas, and stay within the specified margins by the FID manufacturer. ** but within the specification of the manufacturer The sensitivity of an FID is generally sufficient to do analysis in background atmosphere at pmol/mol levels (ppt), e.g. detection limits of GC-FID systems for analyzing 1 litre of air are typically better than 3 pmol/mol (e.g. Plass-Dülmer et al., 2002). For GC-FID systems, it is recommended to perform a calibration every 2 weeks in order to secure high data coverage, however, at least every 2 months Operating conditions Mass Spectrometer (MS) In a MS the analytes are ionized in the ion source either by chemical (CI) or electron ionization (EI). In VOCs analysis, usually EI is used. The resulting gas-phase ions are measured depending on their specific mass-to-charge ratio. Thus, even overlying peaks can thus be separated by analyzing different, compound specific mass tracks. Operating temperatures of a MS system are listed in table 11: Table 11 Operating temperatures of a MS system. MS System T transfer line T ion source T quadrupol AGILENT inert XL 150 C C 230 C C 150 C C High temperatures minimize the residence time and adsorption effects of compounds in the source Other MS systems To be determined To be determined To be determined The sensitivity of a MS is not stable and the signal depends on a set of tunable parameters (e.g. repeller voltage, lenses, multiplier voltage), which influence ionization and ion transmission process as well as the detection of the charged ions at an electron multiplier. Usually a decrease of MS 15

18 sensitivity is observed over time which results in a decrease of peak area. Three measures are thus required: i) Tracking the sensitivity with frequent working standard measurements. Frequency of the working standard measurements should ensure that the decline in sensitivity is accurately tracked over time (e.g. if continuous measurements are performed it is recommended to perform a working standard measurement every 2-4 sample; at least daily close in time to the ambient air sample). ii) Regular auto-tuning of the MS: weekly to monthly, depending on the drift strength observed in the individual systems but at least every second month. iii) If boundary conditions of the source (repeller voltage, lenses) do not allow a proper tuning of the source anymore it has to be cleaned using the procedure specified by the manufacturer. 5.2 PTR-MS (provided by R. Holzinger, T. Petäjä, S. Dusanter) PTR-MS techniques minimize potential losses of VOCs since ambient air is directly analysed without any preconcentration as in gas chromatography. Ambient air passes a drift tube where VOCs are ionized by proton transfer from hydronium ions (H 3 O + ), providing that the VOCS proton affinity is higher than that of water. Product ions are then detected and quantified by mass spectrometry at the targeted VOCS masses plus 1 amu. For more detailed description of PTR-MS techniques see e.g. Blake et al. (2009), De Gouw and Warneke (2007) or Wisthaler et al. (2006). The following recommendations are preliminary, further work is in needed and in progress. A focus on PTR-MS measurements is planned for ACTRIS-2. Inlet (recommendation): The length of the inlet line depends on the measurement place and height. However, it should be as short as possible to minimize the residence time in the inlet line. Recommended materials are PFA, PEEK and PTFE. Typical inlet diameters (ID) are 4-8 mm. Inlet flow varies from few l/min to some tens of l/min, depending on the inlet diameter and length. The inlet flow should be turbulent. The sampling line should also be heated at C to minimize wall-gas interactions. I think it is critical for sticky compounds such as carboxylic acids. In addition, it is important to make sure that there is no cold region between the sampling line and the PTRMS. PTR-MS sampling from this inlet line should be short and have a low volume (1/8 or 1/16 ) line. The PTR-MS sample flow should be ca. 100 ml/min or more. Background determination: The instrumental background is determined by measuring VOCS free air (zero air), which is produced by pumping ambient air through a catalytic converter (Pt catalyst with heating). The background should be measured every hour, but at least once a day. The background signals are subtracted from the 16

19 measured signals. Background signals should be measured at the same RH than ambient air because signals measured at some masses are RH-dependent. The quality of the zero gas is essential as residual mole fraction of a few pmol/mol to a few tens of pmol/mol can lead to negative offsets of the same magnitude for ambient measurements. Critical instrument settings (recommendation): Drift tube pressure: mbar Drift tube voltage: V Drift tube temperature: C Voltage between last drift ring and exit lense: 30 V (instrument dependent, optimization procedures have to be defined) The ratio of O 2 + to H 3 O + should be below 0.03 Table 12 Further settings to be discussed System PTR TOF-MS Q-PTR-MS IONICON SV valve setting (describe optimization procedure, this is applicable for the instruments with the switchable reagent ion (SRI)) Peak shape standards MCP voltage Other To be determined To be determined Tuning of mass scale and resolution SEM voltage Working standard (recommendation) A recommended calibration standard includes: methanol (m33); acetonitrile (m42); acetaldehyde (m45); acetone (m59); isoprene (m69); methyl vinyl ketone (MVK; m71); methyl ethyl ketone (MEK; m73); benzene (m79); toluene (m93); xylene (m107); trimethylbenzene (TMB; m121); α-pinene (m137, m81), trifluorobenzene (m133); trichlorobenzene (m181, m183, m185) A dilution of ca. 1/50 should be used. Calibration in field operation: When and how often: background measurements, at least once a day calibration (working standard) Full mass scan Calculation of VMR recommended rate constants Typically experimentally obtained values provided by Zhao and Zhang (2004) are used. For unknown compounds the usually recommended value is m 3 s -1. transmission recommended procedure 17

20 6. Reference materials The calibration scale is kept by the Central Calibration Laboratories (CCLs) by means of a system of standards (see below). The calibration scale is transferred to the stations and labs by tertiary standards. In case a station does not use a tertiary standard from the CCL, it has to demonstrate that the laboratory standard is linked to the calibration scale by regular and direct comparison. It is recommended that each station or laboratory holds the following calibration gases: 1. A laboratory standard which should be a multi-component laboratory standard (synthetic mixture) that covers the main components and should presumably be produced by the CCL (NPL) or another (NMI) linked to the CCL. 2. A certified multi-component working standard (synthetic mixture with certified mole fractions) with similar components as the laboratory standard. 3. Multi-component working standards that cover all components measured and which are calibrated versus laboratory standard, travelling standards, or other methods (carbon response FID, permeation/diffusion source, mixtures). One of these working standards should be of high mole fractions (upper nmol/mol range) for standard addition measurements. 4. A target gas which preferably is a whole-air working standard calibrated versus laboratory standard, other standards or by other means, but may also be a synthetic mixture. Minimum requirements for a station that need to be fulfilled: 1. A laboratory standard to define the calibration scale for each component measured at the station 2. One working standard (which may be custom-made) to check for drifts in the scale. In case of a GC-FID where calibration can in many cases be reasonably transferred from the laboratory standard to other compounds not present in the laboratory standard by means of the carbon response concept, a well-documented procedure to assign calibration factors and uncertainties to these compounds is needed. 3. A target gas which is presumably whole air but could also be a synthetic mixture The CCLs are for NMHCs, the National Physical Laboratory (NPL, for terpenes, the National Institute for Standards and Technology (NIST, and for OVOCs, a CCL is not determined, yet. 18

21 7. Quality assurance Quality assurance (QA) follows the principles of the GAW QA system ( i) Network-wide use of only one reference standard or scale (primary standard). In consequence, there is only one institution that is responsible for this standard (CCL). ii) Full traceability to the primary standard of all measurements made by Global, Regional and Contributing GAW stations. iii) The definition of data quality objectives (DQOs). iv) Establishment of guidelines on how to meet these quality targets, i.e., harmonized measurement techniques based on Measurement Guidelines (MGs) and Standard Operating Procedures (SOPs). v) Establishment of MGs or SOPs for these measurements. vi) Use of detailed log books for each parameter containing comprehensive meta information related to the measurements, maintenance, and 'internal' calibrations. vii) Regular independent assessments (system and performance audits, Performance audit: check measurements versus DQOs and traceability System audit: overall conformity of a station with the principles of GAW). viii) Timely submission of data and associated metadata to the responsible World Data Centre as a means of permitting independent review of data by a wider community. 7.1 Calibration measurements Frequent calibration measurements are essential for performing good measurements. Furthermore and as first QC measure, target tank measurements should be made. If results of target gas measurements are not in the ACTRIS DQO, the instrument and quality assurance system have to be optimized in order to achieve better results with potential consequences on more frequent calibration, blank and target gas measurements. In the following table 13 recommended calibration frequencies are listed. Table 13 Recommended calibration frequencies System Laboratory Standard GC-FID 2/year (1/year) GC-MS 2/year (1/year) Working Standard blank Target 2/month (1/month) 1/week (1/month) 1/month Every 2-4 th sample (1/day) 1/week (1/month) 1/month To stay within the DQOs, the sensitivity of a GC system should not drift by more than 3% between calibrations. Similarly, blank values (see below) and reproducibility should not change such that they introduce more than 3% effects on the measured data. As both, calibration and target gas measurements enable to detect drifts in the system it is up to the operators to decide the share of these measurements. Another issue is the reproducibility of such standard measurements. Often, 19

22 first measurements are off in a set of measurements due to insufficient equilibration of internal surfaces of sampling lines. For such conditions, series of standard measurements are to be performed containing at least one appropriate measurement. If a drift in the laboratory standard is observed as determined by a drift/inconsistency with a working standard or a discrepancy with a new laboratory standard beyond the combined uncertainties, the discrepancy has to be resolved as soon as possible. Options in such a situation are: send the laboratory standard for recalibration to the CCL or WCC ask other stations for a high level standard for an independent check check available results from past intercomparisons. Anyway, station operators should try to identify where the drift occurred and apply a correction for those periods in which the drift can be well described. If this is not possible, the uncertainty during this period needs to cover the range of unexplained drift. In case stations use working standards/target gases not comprising all components measured, it is justified to determine the sensitivity drift of the instrument by this reduced compound mix if it comprises major constituents of the various groups of VOCs and it covers the range of volatility and polarity encountered in the samples. Calibration factors of compounds not present in the working standard may then be scaled by calibration factors of physically similar behaving compounds present in the standard. OVOCs calibration and target gases might indicate lower repeatability and reproducibility as surface equilibria need more time to be established and slight changes in pressure may affect these equilibria. Accordingly, it might be necessary to apply temperature control to the cylinder valve, pressure regulator and transfer line. Also, frequently used dynamic dilution systems might require substantial warm-up times and it is recommended to keep them running all time Method for measurements of Laboratory/Working standards and Target gases Generally it is recommended to leave pressure regulators and transfer lines attached to the working standard/target gas cylinders in order to minimize the risk of contamination and reduce equilibration times. Laboratory gloves (i.e. powder-free latex) should be worn whenever working with parts in contact with test gases in order to avoid contamination. In order to set the stage for good calibration measurements several issues should be considered: Transfer line and ferrule material Silco steel or Sulfinert or other stainless steel tubing with a passivated internal surface. The use of Vespel/Graphite (VG) ferrules is recommended as these provide a tight sealing while not damaging the tubing. They can be used several times and should only be replaced in case that sealing or contamination problems are present (follow the mounting instructions of the manufacturer). Installation of a new standard gas cylinder 20

23 Pressure regulator and the transfer line with capped fitting on the GC connection side should be mounted at least 24 hours before the measurement. After installation, the regulator and transfer line need to be flushed at least 3 times with the calibration gas. Initial leak check: After flushing, pressurize the pressure regulator (cylinder pressure) and the plugged transfer line (at level pressure needed for the measurement set up). With the cylinder valve closed, check the pressure for a few minutes; if not constant, check all connections, tighten gently, and repeat the check. It is strongly recommended to use no liquid leak tester solutions as they might contaminate the system. Equilibration For equilibration keep the pressure regulator and the transfer line (plugged at the end) pressurized with the standard gas for at least 24 hours. During this equilibration time, the cylinder valve is closed to avoid back diffusion of potential contaminants into the cylinder and to avoid losing sample through possible leakages. This setup also serves as a static leak test as the upstream regulator pressure should not change during the 24 h equilibration period. Connection to the instrument Connect the test gas cylinder to an appropriate instrument inlet port. Then flush the whole inlet line for at least another 3 times and leave the gas cylinder connected to your instrument. It is recommended to open the standard cylinder valve only during the sampling periods unless you use an automated measurements sequence in unattended operation. It is recommenced to leave the standard cylinder permanently connected to the GC system. If this is not possible: 1. Leave the pressure regulator mounted on the cylinder, keep it pressurized and repeat the connection to the instrument method every time you connect the cylinder to the standard port. 2. If you have to dismount the pressure regulator, it is recommended to follow the complete installation of a new gas cylinder method every time. Measurement procedure The standard gas measurement should follow your typical measurement procedure. However, the measurement of the standard gas should be performed after an initial flushing period through the GC valve system which is sufficiently long to achieve equilibration in the lines (typically 10 min with 30 ml/min are sufficient for NMHCs) Zero Gas In this context, Zero gas is a hydrocarbon free gas. The routine measurement of zero gas is part of the QA program to be followed at all stations. It yields information about artefacts due to release of adsorbed hydrocarbons or leaks in the sample path. Blank values should be as low as possible. As zero gas you can use 21

24 catalytically cleaned ambient air (Pt or Pd catalyst at 400 C), which is preferred as this is identical to the sample gas matrix. or synthetic gas (e.g. He or N 2 ) of at least 5.0 or higher quality. This method is not as good but easier to handle. In N quality, often methanol can be observed. To reduce impurities in synthetic gas a post-cleaning is recommended (e.g. cooled charcoal and molecular sieve cartridges). For offline sampling humidified zero gas is needed. High quality water has to be used. Often, trace amounts of hydrocarbons in the pmol/mol range are present as impurities in the zero gas. This creates an inherent problem: blank values caused by impurities cannot easily be separated from blank artefacts as mentioned before. Accordingly, care has to be taken to identify the origin of blanks found in zero gas measurements. Stations have to test zero-gases by comparing the blank values obtained in measurements of different hydrocarbon free gases aiming at the lowest levels. As blank values might vary over time, it is recommended to conduct weekly zero gas measurements. Figure 1 represents the behaviour of blank mole fractions over time; in the here shown example He was used as zero gas in weekly measurements. Blank mole fractions were determined by applying the same calibration factors as for ambient air samples. Shown blank measurements were performed with the set up as depicted in Figure 2. Except for some single events and benzene, most blank values are observed at a rather constant level below 5 pmol/mol. The observed benzene variability is captured by the frequent measurements. Single events (e.g. as observed in June when propane and propene increased drastically) are recorded as well and yield valuable information about the system status. Some occasionally observed blank substances are listed in table 14 below. Table 14 Occasionally observed blank substances Compound Cause various column peaks, column bleeding leakages contamination benzene are occasionally observed and are generally associated to some kind of overheating of traps but the nature of this contamination is not really understood C 2 -C 4 alkenes often observed in systems using Nafion Dryers ((Gong and Demerjian, 1995; Plass-Dülmer et al., 2002, Hoerger et al., 2014) and references therein) acetaldehyde KI ozone filter (Helmig and Greenberg, 1994; Leibrock, 1996) formaldhyde KI ozone filter (Helmig and Greenberg, 1994; Leibrock, 1996) 22

25 30 He-blank, ppt Ethane Ethene Propane Propene Acetylene 1-Butene Benzene i-butene year 2009 Figure 1 Blank mole fractions from zero gas measurements performed with He and the configuration described below in Figure 2 (Hohenpeissenberg Observatory) Method for Measurement of zero gas (blanks) For blank measurements, a zero gas is sampled via the usual air sample path as depicted in Figure 3. Thus, the zero gas passes the ozone and particle filter (if present), the water trap, and sampling unit just like ambient air samples. The sample volume for zero gas should be the same as for ambient air samples. Often a high flow inlet manifold is used which cannot be easily flushed by zero gas. In this case, zero gas has to be introduced after the sampling split to the instrument. However, checks for this set-up should be performed with independent sampling. Figure 3 describes a possible set-up to perform a zero gas measurement. The zero gas is applied at an open T into the sampling line at a flow rate sufficiently higher than the sample flow; e.g. with a sampling flow of about 80 ml/min towards the GC, a zero gas flow of 100ml/min yields a 20 ml/min overflow towards the ambient air sample manifold. It cannot be excluded though, that small amounts of ambient air diffuse into the He against the He flow (at the open T -position). 23

26 Figure 2 Example for a zero gas measurement set-up (Hohenpeissenberg Observatory) Method for Detecting Effects of Ozone on Reactive Compounds In order to check for interferences with ozone and other reactive constituent of the ambient air sample gas, it is suggested to perform standard addition measurements. Briefly, a high concentrated standard gas mixture (e.g. VOCS at 100 nmol/mol level) is added into the ambient air stream such way that the ambient air peak areas are negligible, while the gas matrix itself is dominated by the O 3 rich ambient air (>90%). The standard mixture should contain ozone reactive compounds (e.g. alkenes). If the O 3 rich ambient air matrix does not have an effect on the sample, the peak area ratio of the standard addition and a pure standard measurement is defined only by the dilution factor and the C-response factor is constant (see Sections 8.1 and 8.3). The set up shown in Figure 3 can be used for the standard addition measurements. Figure 3 Set-up for standard addition measurements as used by Hohenpeissenberg. In this specific set-up a quartz capillary (red arrow) is used to add the low standard gas flow into the ambient air stream. It is recommended to perform standard additions on a monthly basis but at least 4 times per year. Strongest effects are expected in high ozone periods (e.g. summer). 24

27 7.2 Audit procedures Audits are performed by the WCC-VOC (KIT Garmisch). 7.3 Measurement protocol It is required that each station has the following log sheets/book either in electronic or paper-based form: 1. Instrument logbook with all operation parameters, significant changes, characterizations, tests results, etc. 2. Measurement logbook with all measurements including the type of measurement, the time of measurement in UTC (start sampling, end sampling, start GC run), sampled volume (dry volume), and comments (anything unusual). 3. A Log of the used calibration factors and blank value determinations from zero gas measurements. 4. A Log of all working standard and target gas measurements. 5. An Error Log with ascribed uncertainty contributions to compound measurements due to peak-overlap, scatter of blank values, unusual low reproducibility, instable sensitivity and so on as well as all other unexplained deviations from normal instrument performance. 6. Meteorological data Log (temp, humidity, wind velocity and direction) 7.4 Measurement uncertainties This section describes the routine assessment of measurement precision and uncertainty. While the precision reflects random errors in the measurement process, the uncertainty includes also possible systematic errors in the measurement. In the following it is illustrated which factors influence precision and uncertainty and how they are derived following the concept of the Guide to the Expression of Uncertainty in Measurement (GUM, 2008). Derived uncertainty values are 1σ errors, for the expanded uncertainty the values have to be multiplied by the coverage factor k=2 (representing the 2σ error). The error calculation is based on the calculation of mole fractions for linear detection systems as presented in section For data submission the precision (1σ) will be reported as well as the total expanded uncertainty (2σ) Calculation of mole fractions for linear detection systems For substances quantifiable via a standard gas mixture, the mole fraction χ sample,i of a compound i in a sample is calculated via: Asample, i Ablank, i χ sample, i = * fcal, i (F1) V sample With the calibration factor 25

28 f cal, i = A V cal cal, i * χ A cal, i blank, i = C i num 1 * C resp, i (F2) A sample,i = peak area of sample measurement of compound i A cal,i = peak area of calibration gas measurement of compound i A blank,i =possible blank value of compound i determined in zero gas measurements χ cal,i = certified mole fraction of calibration gas standard V cal = sample volume of calibration gas V sample = sample volume of sample C i num = Number of C atoms in the molecule i (e.g. for i = n-pentane, C num = 5) C resp,i = mean C-response factor of compound i In case of substances not quantifiable by the standard gas mixture, the mole fraction is calculated via the mean C-Response factor C resp, which is determined from selected compounds in the standard gas measurements averaging the C resp factors for those substances. The mole fraction of a substance is then Asample, i χ sample, i = (F3) V * C * C resp sample, i i num Determination of Precision The precision can either be derived from the target gas or working standard (whole air) measurements or series of air samples taken in similar, stable air mass conditions (meteorological and chemical) in series of 5 or more measurements. It covers the random errors of peak integration, volume determination and blank variation. In case of canister or adsorption tube sampling it includes the reproducibility of the sampling system as well. The precision δχ prec is determined as the standard deviation of a series of measurements of a sample σχ sample: χ prec = σχ sample The above given value represents the instruments precision only at the concentration level and complexity of the sample gas. When working with highly variable concentrations, hence variable peak areas, a more general description of the precision has to be applied for routine measurements: 26

29 χ * Where 1 rel prec = 3 DL + χ σ χ sample DL = detection limit of the system (as described below) C = mole fraction of considered peak rel σ χ sample = relative standard deviation of the sample (e.g. working standard) Thus, the DL is the dominant factor for rather small peaks, while the reproducibility of the instrument becomes the important term for larger peaks Determination of Uncertainty The total uncertainty χunc of a measurement does not only include the random errors described by the precision but also the systematic errors χsystematic of the measurement. χ 2 unc = χ 2 prec + χ 2 systematic Possible systematic errors are: - uncertainty in the standard gas mole fraction δχ cal - systematic integration errors (due to peak overlay or bad peak separation) δ Aint - systematic errors in sample volume determination δ V - further instrumental problems (e.g. sampling line artefacts, possible non-linearity of the detector (MS), changes of split flow rates) δχ instrument - offline sampling errors Following Gaussian error propagation, the overall systematic error is then described as χ 2 systematic = χ 2 cal + δ 2 int + χ 2 vol + χ 2 instrument I Referring to equations F1 to F3, the single error contributions are determined for each analysed compound as A * V =, sample cal χ cal *δχcal Vsample * Acal where δχ cal includes the certified relative uncertainty of the standard gas (or the working standard) and possible drifts of the standard. 27

30 2 2 * * f Asample Vcal χ cal cal χ int = * δa int, + * sample δa 2 int, cal, * Vsample Vsample Acal where δa int, cal measurement and 2 Deliverable WP4 / D4.9 represents the relative error in peak area due to integration of the calibration δa int, samplethe integration error of the sample measurement, respectively., the systematic error of the sample volume, can be neglected, since δ Vsample = δvcal, and thus the χ vol error cancels in equation F1. The statistic volume error is covered by the measurement precision. χ instrument, the error in mole fraction due to specific instrumental problems has to be evaluated for each site individually. These errors can be derived by tests or intercomparison measurements Determination of detection limit Due to impurities or analytical problems or limits the baseline of your gas chromatographic system is usually to a certain degree noisy. Thus, the lowest quantifiable quantity of a substance - the detection limit of the measurement system is different from zero. A simple way to calculate the detection limit is, to integrate a baseline signal over a time interval similar to the average peak width. This integration is performed for a statistically significant number of times (min 10 times). The derived standard deviation of the integrated area multiplied by a factor of 3 represents the detection limit. 28

31 8. Data Management 8.1 Data evaluation FID: effective carbon number The effective carbon number concept (ECN) (Sternberg et al., 1962, Dietz et al., 1967) states that the response (peak area) of the FID is proportional to the number of molecules times the effective number of carbon atoms per analyte molecule, e.g. 2 nmol/mol of ethane have the same integrated response as 1 nmol/mol of butane (when comparing identical sample volumes). If other than hydrogen or carbon bonds occur, the response of the respective carbon atom is adjusted to yield an effective carbon number. For example, an alcohol-group (-O-H) at a terminal C, like in ethanol, results in an effective response of 0.5 for this carbon or a total of 1.5. ECN are listed in original literature (see above), e.g. 1.0 for carbon in aliphatic and aromatic bonds, 0.95 per C in olefinic bonds, 1.3 in acetylenic, 0 for carbonyl, and 0.5 for primary, 0.25 for secondary and 0.75 for tertiary alcohol (Sternberg et al., 1962). Using the ECN-concept, reliable calibration factors for compounds not present in the calibration gas mixture can be estimated. A GC-FID system can be easily characterized for losses or artefacts by making use of the known carbon response C resp = With the peak area A, sample volume V, substance mole fraction and the number of carbon atoms C num. When the carbon responses for the various organic compounds are calculated, they should agree within a few percent. Deviations are often due to bad peak separation, adsorptive losses in the system, or artificial changes at active sites. Efforts should be taken to optimize the system. Especially standard addition measurements are of high value as they characterize interferences with other constituents of ambient air like water vapour and ozone Time series of calibration gas measurements Time series of the calibration factor, peak area or especially for GC-FID systems the C-response factor are valuable tools to monitor the system status over time. As mentioned above, the C-response factor should agree for different compounds within a few percent. And as long as the FID conditions do not change, the C-response factor is expected to be constant. Since a MS is more variable these time series are expected to show drifts and steps due to sensitivity changes. But as for the C-response factor, a similar behaviour is expected for similar compounds. In Figure 4, a time series of C-response factors for a number of NMHCs is shown. Several features can be observed in this example: With the exception ethyne, all shown substances agree within 3% and resemble the same behavior over time. A reason for the drift (~ -4%) might be e.g. a slow change of FID characteristics which are, however, captured by the frequent standard measurements. Ethyne reveals a different C-response behavior; it is not only more variable but shoes a sharp increase. The latter was connected to a change of working standard, which affected only the ethyne response. As a consequence 29

32 for ethyne, the average C-response factor is used, as it cannot be determined properly in the calibration measurements working standard C-response GC Hohenpeissenberg, 2013 ethane, AS_NPL ethene, AS_NPL propane, AS_NPL propene, AS_NPL 2-methylpropane, AS_NPL ethyne, AS_NPL n-butane, AS_NPL t-2-butene, AS_NPL 1-butene, AS_NPL c-2-butene, AS_NPL Figure 4 Time series of C-response factors of bi-weekly working standard measurements for several NMHCs with the C2-C8 GC-FID system of Hohenpeissenberg (Plass-Duelmer et al., 2003) during the year Target gas measurements In Figure 5, a series of target gas measurements at Hohenpeissenberg (DWD, Germany) is shown. Here, the determined mole fraction for all analysed compounds is plotted over time in a log scale. Relative changes are detectable as linear deviations from constant values. The first plot shows compounds with mole fraction of more than 50 pmol/mol, the second one C 3 -C 6 hydrocarbons with mole fractions below 50 pmol/mol, and the third one C 6 -C 9 hydrocarbons with mole fractions below 50 pmol/mol. Except for 2-methylpropene (due to blank values), it should be pointed out that repeatability gets poorer for compounds with higher molecular weight and towards lower mole fractions. However, the repeatability is still mostly within 2 pmol/mol or a few percent. This plot shows monthly repeatability of a series of 5 replicates, and monthly reproducibility throughout the year for ambient air mole fraction levels and ambient air matrix. 30

33 10000 mole fraction [pmol/mol] Target Gas ethane ethene propane propene 2-methylpropane Ethyne n-butane n-pentane 3-M-Pentan 2-methylpropene 2-methylbutane 2-methylpentane methylcyclopentane+ 2-2-dimethylbutane. toluene Figure 5 Measurements of compressed whole air from cylinder ( Referenzluft-04 or Ref-04 ) through 2009; generally 5 replicates are measured once a month Results of standard addition measurements The standard addition measurement is compared to a pure standard measurement of the same standard gas mixture. If the O 3 rich ambient air matrix does not have an effect on the sample, the calibration factor (for FID systems the C-response factor) should be the same for both measurements and thus 1= With = being the average peak area ratio for non O 3 -reactive compound with!"!# low mole fraction in the ambient air (e.g. alkanes like). This concept is applicable for all linear GCsystems, without knowledge of the exact flows or mole fractions in the standard gas mixture. In Figure 6, results of a standard addition measurement performed in 2009 at Hohenpeissenberg are shown. Plotted is the normalized peak ratio as described above. Some of them have fairly high contributions from ambient air (ratios up to about 3), but the important alkenes are clearly dominated by the added standard. If ozone interferences (losses) exist, these reactive alkenes should show lower ratios than 1. None of the alkenes shows any significant deviations from 1 and thus no indication of reactive losses with O 3. In case alkene measurements exhibit a normalized peak area ratio r norm < 1, the GC system is further checked and if necessary the O 3 filter is replaced. 31

34 NPL-Standard-Addition ethane ethene propane propene 2-methylpropane acetylene n-butane t-2-butene 1-butene 2-methylpropene c-2-buten2 propyne 1,3-butadiene t-2-pentene n-hexane Figure 6 Results of NPL standard addition measurements from 2009 performed once per month. Typically, 2-10% of sample volume is added to an ambient air Data checks of final mole fraction data in time series VOCs should be grouped in a convenient number (typically 3 or 4) of functionally similar compounds, e.g. alkanes or alkenes, in a log plot over a time interval of half a year or a year. The procedure is illustrated in Figures 7 and 8, where quality checks were performed for Hohenpeissenberg data in year Generally, it is expected that the variability of the data should increase with higher reactivity (variability-lifetime-relation) and changes should be more pronounced for shorter lived compounds (lower background). Spikes in positive direction may be attributed to plumes with local/regional pollution and should be checked for consistency with other compounds from similar sources, if not consistent, the raw data should be rechecked. Spikes in negative direction stand for clean air and should again be checked for consistency with other compounds from similar sources. If they are found to be not consistent with other compounds, the peak integration, breakthrough in trap or other potential loss problems should be checked once again. (e.g. the negative spike in ethane at the beginning of august (Fig. 8) was due to breakthrough; several positive spikes in November are found in all time-series and are thus most likely an atmospheric feature). 32

35 10000 ethane, ppt 2-methylpropane, ppt propane, ppt n-butane, ppt Figure 7 Time series (annual cycle) of C 2 -C 4 alkanes measured at Hohenpeissenberg, generally, measurements from 1:00 and 13:00 CET are shown. Due to the orographic situation on the top of a hill, Hohenpeissenberg is frequently decoupled from the boundary layer at night time and accumulations of trace gases as found at flat terrain sites are usually not observed. For compounds with similar relative annual cycles but with different mole fraction levels the ratio plots as shown below can be used (Fig. 9). In such plots, either structural similar compounds (as shown above for pairs of alkanes, alkynes, and aromatics) are compared or compounds originating from similar sources or compounds having similar lifetimes. For each distinct spike/outlier or deviations from the ratio, following checks are performed: i) logbook entries to identify irregular operation conditions ii) peak integration iii) other compounds deviating in these individual measurements and try to identify the reason for the spike For example, the n-pentane/2-methylbutane spikes are due to occasional n-pentane plumes (observed in two independent systems but the origin of the plume is unknown), the broad minimum in propyne/ethyne ratios in spring is due to the longer lifetime of ethyne and the correspondingly relaxed reaction to the OH annual cycle, and of the aromatics only benzene has a well-developed summer minimum, all others are flattened in summer but consistently as seen in the xylene/toluene ratios. 33

36 10 n-pentane/2-methylbutane toluene/benzene propyne/ethyne p-m-xylene/toluene Figure 8 Time series of ratios between pairs of hydrocarbons with similar structure; data from Hohenpeißenberg, QC in xy-plots (used at Rigi, Switzerland by Empa) A similar approach as used at Hohenpeissenberg is applied at the Rigi site by Empa. Beside time series, many xy-plots are generated to check for consistency with former years and within the year. In the correlation plots, compounds which are structural similar, or compounds originating from similar sources, or compounds having similar lifetimes are plotted against each other. In Figures 9 and 10 examples are shown of correlation plots using toluene vs. benzene and benzene vs. ethyne. Table 15 xy-plots used at the Rigi (Switzerland) site by Empa propane/ethane butenes/ethene isoheptane/isohexane n-butane/ethane pentenes/ethene 1,3-butadiene/isoprene propane/n-butane ethenes/butenes toluene/benzene 2-methylpropane/n-butane methylpropane/2-methylbutane m/p-xylene/benzene 2-methylbutane/n-pentane 1,3-butadiene/ethene ethylbenzene/mp-xylene propene/ethene n-hexane/n-pentane o-xylene/mp-xylene ethyne/ethene isohexane/n-hexane o-xylene/ethylbenzene ethyne/benzene 34

37 Deliverable WP4 / D Chrom. ok C5/C2-Benzene also higher toluene y = x R² = benzene Toluol * Toluol (11) Linear (Toluol (11)) Figure 9 Toluene vs benzene at Rigi (Switzerland). Blue 2011 data, brown data benzene y = x R² = ethyne Benzol * Benzol (11) Linear (Benzol (11)) Figure 10 Ethyne vs benzene at Rigi (Switzerland). Blue 2011 data, brown data QC in repeatability and reproducibility: One more check for the plausibility of the measurements is related to the repeatability (sets of 3-5 measurements each) and reproducibility (monthly or bimonthly) of the measurements of calibration gases or target gases. In Figure 11, the relative (%) and absolute (pmol/mol) standard deviations are plotted for 3 different target gases on a log scale versus the mixing ratio for all identified compounds. Such a presentation helps to identify problems with individual compounds. In the lower figure there are two curves 35

38 added that represent a variability by 1 pmol/mol + 1% (red line) and 3 pmol/mol + 3 % (broken red line). For a number of compounds the first line represents a good fit. However, some of the compounds exhibit higher scatter but are generally within the 3 pmol/mol + 3% line which still is a quite good level for the reproducibility of measurements of cylinder gases and within the ACTRIS DQOs as shown in Section 2. Compounds with higher deviations should be checked for the reason of the deviation, often peak-overlap or integration problems are associated with worse reproducibility. Also, heavier compounds tend to be less reproducible due to adsorption/desorption problems. 1000% 100% 10% reference gases in 2007 rel. standard dev. versus mixing ratio Ref-04 Target 1 SUPELCO Target 2 74-NMHC Target 3 1% 0% abs. standard dev. versus mixing ratio Ref-04 Target 1 1ppt+1% Target 2 SUPELCO Target 3 74-NMHC 3ppt+3% Figure 11 Standard deviations obtained from all measurements of target gases performed at Hohenpeissenberg Observatory in 2007 versus compounds mole fractions in pmol/mol in the respective standard; the upper panel shows the relative standard deviations, and the lower panel the pmol/mol standard deviations versus mole fractions (pmol/mol); two lines in the lower part show parameterizations of the standard deviations for good results (1 pmol/mol + 1%) and as an upper limit for most of the not so good compounds 3 pmol/mol + 3%. 36

39 8.1.8 Recommended QC and minimum QC, thresholds for flagging the data Deliverable WP4 / D4.9 As a minimum requirement of the QC we suggest to visually control time series of calibration gas (7.2), target gas (7.3) and ambient air measurements (7.5) and to generate xy-plots for the above mentioned compounds (7.6) and to compare the data to previous years Template is available from EMPA. Table 16 ACTRIS data flags Flag Data Valid (V) or invalid (I) Description 000 V Valid measurement 147 V Below theoretical detection limit or formal Q/A limit, but a value has been measured and reported and is considered valid 420 V Preliminary data 457 V Extremely low value, outside four times standard deviation in a lognormal distribution 458 V Extremely high value, outside four times standard deviation in a lognormal distribution 651 V Extremely high value, outside four times standard deviation in a lognormal distribution 652 V Construction/activity nearby 999 I Missing measurement, unspecified reason 8.2 Metadata Link to EBAS Meta data ( Ancillary data EBAS 8.4 Data archiving at the station or laboratory It is recommended to perform daily backups of the raw data. 8.5 Data submission All ACTRIS trace gases measurements are reported to, and stored in the EBAS atmospheric database The EBAS database, originally designed for the European Monitoring and Evaluation Programme (EMEP), today archives data on atmospheric composition from ground stations around the globe, as well as aircraft and ship platforms. All datasets in EBAS are associated to one or more projects/frameworks, having individual rules for data disclosure. Most data stored in EBAS are originating from programs encouraging an unlimited and open data policy for non-commercial use. Offer of co-authorship is made through personal contact with the data providers or owners whenever considerate use is made of their data. In all cases, an acknowledgment must be made to the data 37

40 providers or owners and to the project name when these data are used within a publication. The data policy for ACTRIS near-surface data is harmonized with the data policy for GAW, and available from EBAS, and from the ACTRIS data portal and also web: GB/ProjectResults/Dataconcept.aspx. The ACTRIS data portal links EBAS data, together with data from the two other ACTRIS databases, EARLINET DB and CloudNet DB, into one common data portal. The portal facilitates the combined analysis of all ACTRIS data, offering advanced tools for plotting and combining ACTRIS data from the three topic databases, and mapping tools for user defined visualization of distribution atmospheric sites and variables across networks and projects. The following section provides a summary of the data submission procedures for trace gas data to EBAS. The text below only address the main points as defined by August 2014, for a complete and, at any time, updated document please reference Trace gas near-surface data are qualified as ACTRIS data only if the measurement data are reported to EBAS by using the templates and formats recommended by the ACTRIS trace gas community, and following the procedures described in the current document. The data providers are responsible for the quality of the data submitted and the templates ensure proper and sufficient documentation of the data. ACTRIS partners shall label their contribution two EBAS with project/framework "ACTRIS". The data can also be associated to other programs and frameworks like GAW-WDCGG-node, EMEP, InGOS etc. Data submitted to EBAS need to be formatted in the EBAS NASA-Ames format by the data provider. The EBAS NASA-Ames format is based on the ASCII text NASA-Ames 1001 format, but contains additional metadata specifications ensuring proper documentation of the setup and procedures for each measurement principle. The term VOCs in ACTRIS consists of three subgroups: NMHCs (C 2 -C 9 hydrocarbons), OVOCs (oxygenated volatile organic compounds), and terpenes (biogenic hydrocarbons with a terpene-structure), which can be measured by several different measurement principles; on-line and off-line traps and off-line canisters. An overview of all ACTRIS variables and the associated recommended methodologies are available at the ACTRIS web in the document ACTRIS Data, concept, and variables : Specific templates for each of these are available from under the tab Submit Data -> Regular Annual Data Reporting -> VOCS. An EBAS NASA Ames file consists of two parts; a metadata header and a column formatted data part. The header section contains a number of important metadata items describing the measurement site, data variable, instrument, measurement principle and operating procedure. If nothing changes in the measurement set up, the header will remain the same from year to year, and the measurement data will be visible as one continuous dataset in the database. The data section of an EBAS NASA Ames file consists of a fixed column number format ASCII table, including time stamp, data value and flag for each single measurement point or data average point. The data formatting templates give the user a detailed line-by-line explanation of what metadata that should be included on which line of the header, in terms of correct procedure and wording: 38

41 Further information is available by clicking on the respective line number from the template. Flagging of data should be done according to the ACTRIS trace gas guidelines. For time being only flags from the tables at the format template pages are recommended, but a complete list of flags available in EBAS is located at The data centre recommends first to create the data table and then add the header. Name the file overusing the filename stated in the header. The data submission deadline for the ACTRIS project is following the EMEP submission deadline, this is July 31 for data from the year before. Example: 31. July 2014 is reporting deadline for all 2013 data. The files containing the data submissions must be uploaded to the EBAS anonymous FTP site, accessible at: ftp://ebas-submissions.nilu.no/incoming using the submitters as password. This site is for security reasons a blind drop page, so the submitter will not be able to see the data after submission. An automail from the system will be sent to the data submitter if the submission was successful. The submitted data will be collected, checked and inserted to EBAS. The data submitter will be notified in case of needs for correction in the submitted data. Status report to ACTRIS activity leaders will be made by the end of the year. 39

42 9. References Apel, E.C. et al. (2003): A fast-gc/ms system to measure C 2 to C 4 carbonyls and methanol aboard aircraft. Journal of Geophysical Research 108 (D20): Barkley, C.S. et al. (2005): Development of a Cryogen-Free Concentration System for Measurements of Volatile Organic Compounds. Analytical Chemistry 77 (21): Blake, R. S., Monks, P. S., Ellis, A. M. (2009): Proton-Transfer Reaction Mass Spectrometr. Chem. Rev. 109: De Gouw, J., Warneke, C. (2007): Measurements of volatile organic compounds in the earth s atmosphere using proton-transfer-reaction mass spectrometry. Mass Spectrom. Rev. 26 (2): Graus, M., Müller, M., Hansel, A. (2010): High Resolution PTR-TOF: Quantification and Formula Confirmation of VOCS in Real Time. J. Am. Soc. Mass Spectrom. 21 (6): Dietz W.A. (1967): Response factors for gas chromatographic analyses, J. of Gas Chromatography 5, 68. Folkers, A. (2002): Oxygenated volatile organic compounds in the troposphere: Development and employment of a gas chromatographic detection method. Report of the Research Centre Jülich Dissertation University of Köln, Jülich. Greenberg, J.P., Zimmermann, P.R., Pollock, W.F., Lueb, R.A., Heidt, L.E. (1992): Diurnal variability of atmospheric methane, nonmethane hydrocarbons, and carbon monoxide at Mauna Loa. Journal of Geophysical Research 97: 10,395-10,413. Helmig, D. (1997): Ozone removal techniques in the sampling of atmospheric volatile organic trace gases. Atmospheric Environment 31 (21): Helmig, D., Greenberg, J.P. (1994): Automated in situ gas chromatographic-mass spectrometric analysis of ppt level volatile organic trace gases using multistage solid-adsorbent trapping. Journal of Chromatography A 677: Hoerger, C.C., S. Reimann, A. Werner, C. Plass-Duelmer, E. Weiss, R. Steinbrecher, S. Sauvage, J.R. Hopkins, J. Aalto, J. Arduini, N. Bonnaire, A. Borowiak, J.N. Cape, A. Colomb, R. Connolly, J. Diskova, P. Dumitrean, C. Ehlers, V. Gros, H. Hakola, M. Hill, M.K. Kajos, J. Jäger, R. Junek, M. Leuchner, A.C. Lewis, M. Maione, D.Martin, E. Nemitz, S. O'Doherty, P. Pérez Ballesta, T. Petäjä, J-P. Putaud, N. Schmidbauer, G. Spain, E. Straube, M. Vana, M.K. Vollmer, R. Wegener, A. Wenger, Volatile organic compounds (VOCs) intercomparison experiment in Europe (within ACTRIS), (2014): submitted to Atmos. Meas. Tech. Hopkins J.R., Lewis A.C., Read K.A. (2003): A two-column method for long-term monitoring of nonmethane hydrocarbons (NMHCs) and oxygenated volatile organic compounds (o-vocs). Journal of Environmental Monitoring, volume 5, issue 1. Hopkins, J.R., Jones, C.E., Lewis, A.C. (2011): A dual channel gas chromatograph for atmospheric analysis of volatile organic compounds including oxygenated and monoterpene compounds. Journal of Environmental Monitoring DOI: /c1em10050e. Koppmann, R., Johnen, F.J., Khedim, A., Rudolph, J., Wedel, A., Wiards, B. (1995): The influence of ozone on light nonmethane hydrocarbons during cryogenic preconcentration. Journal of Geophysical Research 100: 11,383-11,391. Kuster, W.C, Goldan, P.D., Albritton, D.L. (1986): Ozone interferences with ambient dimethyl sulfide measurements: The problem and a solution. Eos 67: 887. Lee, J.H., Batterman, S.A., Jia, C., Chernyak, S. (2006): Ozone artifacts and carbonyl measurements using Tenax GR, Tenax TA, Carbopack B, and Carbopack X adsorbents. Journal of the Air & Waste Management Association 56:

43 Leibrock E. (1996): Entwicklung eines gaschromatographischen Verfahrens zur Spurenanalytik von oxidierten Kohlenwasserstoffen in Luft, Wissenschafts-Verlag Maraun, Frankfurt a.m., McClenny, W.A., Pleil, J.D., Evans, G.F., Oliver, K.D., Holdren, M.W., Winberrry, W.T. (1991): Canister-based method for monitoring toxic V0Cs in ambient air, J. Air Waste Manage. Assoc. 41, McClenny W.A., Pleil J.D., Evans G.F., Oliver K.D., Holdren M.W., Winberry W.T. (1991): Canister-based method for monitoring toxic VOCs in ambient air J. Air Waste Man. Ass., 41, Palluau F., Ph Mirabel, M Millet (2007): Influence of ozone on the sampling and storage of volatile organic compounds in canisters, Environ Chem. Lett. 5, Plass-Dülmer, C.; Michl, K.; Ruf, R.; Berresheim, H. (2002), C 2 - C 8 hydrocarbon measurement and quality control procedures at the Global Atmosphere Watch Observatory Hohenpeissenberg. J. Chromatogr. 953, Plass-Dülmer C., N. Schmidbauer J. Slemr, F. Slemr, H. D'Souza (2006): European hydrocarbon intercomparison experiment AMOHA part 4: Canister sampling of ambient air, J. Geophys. Res., 111, D04306, doi: /2005jd Pollmann J., D. Helmig, J. Hueber, Ch. Plass-Dülmer, P. Tans (2008): Sampling, storage, and analysis of C 2 C 7 non-methane hydrocarbons from the US National Oceanic and Atmospheric Administration Cooperative Air Sampling Network glass flasks, J. Chromatogr., A 1188, Sternberg J.C., W.S. Gallaway and D.T.L. Jones, in N. Brenner (1962), The mechanism of response of Flame Ionization Detectors, in J. Callen and M.D. Weiss (Editors) Gas-Chromatography, Academic Press, New York, p Taipale, R., Ruuskanen, T. M., Rinne, J., Kajos, M. K., Hakola, H., Pohja, T., Kulmala, M. (2008): Technical Note: Quantitative long-term measurements of VOCS concentrations by PTR-MS measurement, calibration, and volume mixing ratio calculation methods. Atmospheric Chemistry and Physics 8, US-EPA (1998): Technical assistance document for analysis of ozone precursors, US-Environmental Protection Agency, EPA/600-R-98/161. US-EPA TO-14A (1999): Compendium of methods for the determination of toxic organic compounds in ambient air: determination of volatile organic compounds (VOCs) in ambient air using specially prepared canisters with subsequent analysis by gas chromatography, US-Environmental Protection Agency, Method TO-14A, 2nd ed., EPA/625/R-96/010b. Wisthaler, A. et al. (2006): Recent developments in proton-transfer-reaction mass spectrometry. Photonic, Electronic and Atomic Collisions, 24th International Conference on Photonic, Electronic and Atomic Collisions, Rosario, Argentina: Doi: / _0060. WMO (1995), WMO-BMBF Workshop on VOCs - Establishment of a World Calibration/Instrument Intercomparison Facility for VOCS to serve the WMO Global Atmosphere Watch (GAW) Programme, WMO Report, 111. WMO (2001): Strategy for the Implementation of the Global Atmosphere Watch Programme ( ), GAW Report No. 142 (WMO TD No. 1077), 62 pp., World Meteorological Organization, Geneva, Switzerland. WMO (2007a): A WMO/GAW Expert Workshop on Global Long-Term Measurements of Volatile Organic Compounds, Geneva, Switzerland. WMO (2007b): WMO Global Atmosphere Watch (GAW) Strategic Plan: , Geneva, Switzerland. 41

44 WMO (2012): (GAW Report 204: Standard Operating Procedures (SOPs) for Air Sampling in Stainless Steel Canisters for Non-Methane Hydrocarbons Analysis (prepared by R. Steinbrecher and E. Weiss), 28 pp., Geneva, Switzerland. Zhao, J. and R. Zhang, 2004: Proton transfer reaction rate constants between hydronium ion (H3O+) and volatile organic compounds. Atmospheric Environment, 38, Blake, R. S., Monks, P. S., Ellis, A. M. (2009): Proton-Transfer Reaction Mass Spectrometr. Chem. Rev. 109:

45 10. Appendices Appendix 1: Ozone removal techniques for GC analysis of OVOCS Appendix 2: Adsorbents for sorbent-based enrichment of VOCs and OVOCs (oxygenated volatile organic compounds) in ambient air samples Appendix 3: Chromatographic separation 43

46 APPENDIX 1: OZONE REMOVAL TECHNIQUES FOR GC ANALYSIS OF OVOCS (OXYGENATED VOLATILE ORGANIC COMPOUNDS) IN AMBIENT AIR SAMPLES (COMPILED FROM LITERATURE J. ENGLERT) Reactions of concentrated VOCs with ozone during sampling process may alter the quantities of the target analytes and also contribute to the formation of artefacts which may mistakenly be interpreted as atmospheric constituents. Ozone reactions during cryogenic enrichment of VOCS: Ozone melting and boiling points (at atmospheric pressure) are at C and C. During cryogenic freeze-out of VOCs from ambient air samples ozone is concentrated together with the target analytes, whereas the main constituents of air nitrogen and oxygen do not condense under these conditions (boiling point of liquid nitrogen -196 C). Reactions of VOCs with ozone occur when heating the cryogenic trap to transfer the analytes to the GC system. Alkenes, such as isoprene and monoterpenes can be depleted in this reactions leading to artefacts like methacrolein and methylvinylketone. By collecting ambient air into stainless steel canisters prior to the analysis with cryogenic freeze-out techniques this effect is reduced because of the short lifetime of ozone in these canisters (Helmig, 1997; Greenberg et al., 1992). But this method is not suitable for oxygenated VOCs because of the high reactivity of these compounds on unheated stainless steel surfaces. Ozone reactions during solid adsorbent sampling of VOCs: Ozone artefacts are formed on and with some sorbents (e.g. graphitised carbon sorbents and Tenax TA) leading to both VOCs losses and formation (Lee et al., 2006; McClenny et al., 2001). Adsorbed unsaturated hydrocarbons might for example undergo reaction with ozone during ambient sampling leading to diminished alkene concentrations and the formation of oxygenated reaction products e.g. acetaldehyde and formaldehyde. Products from ozone - Tenax reactions include benzaldehyde, phenol, acetophenone and n-aldehydes (Helmig, 1997). Reactions with ozone can be reduced by selectively removing the oxidant in the sample flow prior to the concentrating of the analytes of interest. The ozone removing system should be easy to use, inexpensive, efficient in the ozone removal rate and have a high scrubbing capacity, long lifetime and eliminate the effects of ozone without interfering with the analytes of the target compounds and without introducing contaminants. Furthermore it should be universally applicable to allow the analysis of a wide range of compounds. Commonly reported techniques for ozone scrubbers include impregnated filters, impregnated glass wool, coated tubes and coated annular denuders. Catalytic destruction of ozone on metal surfaces: Aluminum, copper, lead and tin have low ozone depletion efficiency whereas silver, iron, zinc, gold, nickel, mercury and platinum have high ozone destruction capacities. The ozone removal acquirement of some metals is used e.g. by nickel tubing, which reduces ozone levels to less than 20 44

47 % of ambient air level (Helmig, 1997). Koppmann et al. (1995) found up to 50% destruction of ambient ozone by pulling the sample air through stainless steel inlet lines kept at 67 C. Hopkins et al. (2011): All gas transfer lines within the system are made from stainless steel and heated to 70 C to reduce ozone mixing ratios. Disadvantage: Loss of OVOCs on the surface of stainless steel even at high temperatures (150 C). Ozone deletion by nitric oxide (NO) titration: Titration of the ambient air sample with a few ppm of NO prior to the concentration step is a very efficient method to remove ozone. Ozone (O 3 ) deletion performance depends on sufficient reaction time and NO concentration in the mixing chamber. An example is the titration of the ambient air sample for 20 seconds in a 1 litre glass reaction vessel with a small flow of 200 ppm NO in nitrogen resulting in a NO concentration of 2 ppm. NO reacts with O 3 to nitrogen dioxide (NO 2 ) and oxygen (O 2 ) (Helmig, 1997). The reaction is: O 3 + NO O 2 + NO 2. Disadvantage: slow reaction, alcohol losses (but constant) Ozone deletion by potassium iodide (KI): In many cases KI is used for O 3 removal. This technique is very effective at ambient humidity levels while capacity is reduced in dry air respected in following equation (Helmig, 1997): O 3 + 2KI + H 2 O O 2 + I 2 + 2KOH. KI reacts with O 3 to potassium oxide (K 2 O) and elemental iodine. Example: PTFE-lined stainless steel or Silco steel capillary, OD 1/4, 5 cm filled with KI-coated glass wool. Disadvantage: formaldehyde and acetaldehyde blank values, alcohol losses (Helmig and Greenberg, 1994; Leibrock, 1996) Sodium sulphite (Na 2 SO 3 ): Most efficient in the presence of atmospheric water vapour and hence has to be positioned upstream of a water trap was found to remove 99% of the O 3 in a humid ambient air stream but inconsistent removal efficiencies from different suppliers and from different batches testing of individual O 3 traps is required (Helmig, 1997) Example: ¼ glass tube filled with 1 g of Na 2 SO 3 anhydrous crystals held in place by glass wool plugs and maintained at 100 C to prevent clumping of the Na 2 SO 3 Disadvantage: removal of methyl vinyl ketone (MVK) and methacrolein. Sodium thiosulphate (Na 2 S 2 O 3 ): 45

48 The reaction between thiosulfate and O 3 produces tetrathionate oxygen and water depends on the ph level: 2S 2 O O 3 + 2H + S 4 O O 2 + H 2 O Example: O 3 filters were prepared by flowing a 10% solution of aqueous Na 2 S 2 O 3 through commercial glass fiber filters followed by dry purge with nitrogen and had capacities in excess of 1 m 3 air at ambient O 3 levels (Helmig, 1997) Advantage: this glass fiber filters also reduce sampling artefacts from reactions with halogens Other O 3 removal agents are copper oxide (CuO), magnesium sulphate (MgSO 4 ), manganese dioxide (MnO 2 ), potassium carbonate (K 2 CO 3 ) and TPDDC (see Table 1). In-line O 3 scrubbers like granular KCl and crystalline Na2SO4 are prone to artefacts and require regular maintenance so that they are not suited to long-term instrument deployments (Hopkins et al., 2011). Table 1: Ozone removal techniques for VOCS monitoring and their characteristics. Technique Agent Characteristics Coated annular denuder Potassium iodide (KI) Very efficient Cellulose filter KI Improved formaldehyde and acetaldehyde recovery Packed Teflon tubing Crystalline KI Quantitative transmission of formaldehyde and acetaldehyde, partial loss of methacrolein and methyl vinyl ketone (MVK) Impinger KI 2% aqueous, buffered KI solution Impregnated glass wool KI Quantitative O 3 removal, iodated artefacts Coated tubing KI in copper tubing Commercial scrubber KI in polyethylene cartridge Low capacity at 5% RH Impregnated glass fiber filter Sodium thiosulphate (Na 2 S 2 O 3 ) High capacity, also reduces sampling artifacts from reactions with halogens Coated copper screen Manganese dioxide (MnO 2 ) High capacity, possible losses of terpenes (e.g. camphor, linalool), loss of formaldehyde Packed copper tubes Anhydrous mesh potassium 100% transmittance of light carbonate (K 2 CO 3 ), crystalline hydrocarbons Packed Teflon tubing K 2 CO 3 ozone and water removal, 100% transmission of light hydrocarbons Loss of unsaturated compounds Packed glass tube Crystalline sodium sulphite prevented, most efficient in the (Na 2 SO 3 ) presence of atmospheric water vapour Cartridge Copper oxide (CuO), crystalline No losses of carbonyl compounds Trap Removal of at least 100 ppb, loss Crystalline magnesium sulphate of O (MgSO 4 ) 3 removal efficiency with sampling length Gas-phase ozone titration Nitric oxide (NO) Very efficient, quantitative 46

49 Metal tubing Spiked cartridge Spiked cartridge Nickel (Ni) TPDDC (Tetramethyl-1,4- phenylenediamine dihydrochloride) 5% Na 2 S 2 O 3 aqueous solution on Tenax Deliverable WP4 / D4.9 recovery of formaldehyde, formation of artifacts on Tenax exposed to elevated NO x levels, possible chromatographic interferences of NO and NO 2 with NMHCS (Kuster et al., 1986), losses of alcohols, slow reaction O 3 reduced to less than 20% of ambient level Sampling of carbonyl compounds on microcartridges containing porous glass particles impregnated with dansylhydrazine (DNSH), agent added to the reagent solution at the time of cartridge preparation to serve as an O 3 scavenger Direct pretreatment of the adsorbent, improved monoterpene recovery Spiked cartridge Na 2 S 2 O 3 Interferences eliminated 47

50 APPENDIX 2: ADSORBENTS FOR SORBENT-BASED ENRICHMENT OF VOCS AND OVOCS (OXYGENATED VOLATILE ORGANIC COMPOUNDS) IN AMBIENT AIR SAMPLES (COMPILED FROM LITERATURE BY J. ENGLERT) Sampling of ambient air with sorbent tubes or traps and subsequent thermal desorption to transfer the sampled compounds to a GC system is widely-used for trace gas analysis of VOCs because of the high sensitivity of this method. There are two different sorbent-based sampling strategies: (1) on-line sampling of ambient air directly into (cooled) sorbent focusing traps or transfer of air samples from containers (stainless steel canisters or PTFE bags) into these (cooled) traps; and (2) off-line pumped (active) or diffusive (passive) sampling onto adsorbent tubes or cartridges held at ambient temperature. In the case of off-line sampling VOCs are transferred in a second step into a cooled focusing device (e.g. sorbent trap). For oxygenated VOCs a method with short transfer from sampling device to the analysis system is important because of the high losses of these analytes on surfaces, especially on unheated and not inert ones like untreated surfaces of stainless steel. When selecting a suitable sorbent or sorbent combination for VOCs and OVOCs several factors have to be considered including sorbent strength, artefacts, hydrophobicity, inertness, thermal stability and friability. It has to be verified that there is no breakthrough (most critical compounds methanol and acetaldehyde), getting stuck or back-diffusion of target compounds. Some special, low volatile analytes may also be lost through aerosol formation. The sorbents must be strong enough to retain target analytes from a specific sample volume but must also be weak enough to release them during thermal desorption. Sorbent strength is measured in terms of breakthrough volumes that are defined as the litres gas per gram adsorbent required to elute a VOCs off 1.0 gram adsorbent at an indicated temperature. This capacity of solid sorbents depends on temperature and is typically specified at 20 C. It approximately halves for every 10 C rise. Therefore, cooling the traps during sampling increases adsorbent performance. The lowest possible temperature is limited by the dew point of the sampled air (Brown and Shirey, 2001; Helmig and Greenberg, 1994; Woolfenden, 2010b). When using hydrophilic sorbents (molecular sieves) or temperatures below the dew point for ambient air samples some kind of water trap has to be installed in the sampling line. Otherwise there would be a reduction of sorbent performance that might reach a factor of 10 at high humidity conditions (90% RH) and after desorption of the trapped water moisture might interfere with the following chromatographic analysis. Weak and medium strength sorbents (porous polymers and graphitised carbon blacks) are hydrophobic and so they prevent trapping of excess water. Some sorbents especially carbon blacks contain chemically active materials (trace metals) and are unsuitable for labile (reactive) species. Most porous polymers except from Tenax TA have high inherent artefacts with blank peaks at 5-10 ng levels (Woolfenden, 2010b). Ozone (O 3 ) artefacts are formed on and with some sorbents (e.g. graphitised carbon sorbents and Tenax TA) leading to both OVOCs losses and buildings (Lee et al., 2006; McClenny et al., 2001). So the aspect of O 3 removal has to be considered in sorbent-based ambient air sampling. 48

51 Quartz wool or silica beads are not able to retain most of the compounds. They are usually used in multi-bed traps to prevent very high boilers to come in contact with a stronger adsorbent. Porous polymers are weak or medium strength sorbents. None of them could retain the very volatile analytes. In multi-bed traps they are often the first sorbent in sampling direction for the mid and higher boiling point analytes beginning from benzene. Porous polymers are hydrophobic and so are adequate for humid ambient air samples. CarbopackTM, CarbotrapTM and CarbographTM are graphitised carbon blacks. The three different types differ in mesh sizes. They are suitable for most of the VOCs depending on their different sorbent strength. The strongest CarbopackTM X should have a weaker adsorbent in front of it when sampling very high boiling point analytes. All graphitised carbon blacks are hydrophobic like porous polymers and so are adequate for humid ambient air samples (Brown and Shirey, 2001). CarboxenTM and CarbosieveTM adsorbents are very strong and not appropriate for analytes with boiling points higher than benzene because they have very small pores. They should always be used with a weaker adsorbent (porous polymer or graphitised carbon black) placed in front. Pore shape of the CarbosievesTM is different from the CarboxensTM. Pores of CarbosievesTM may be blocked by analytes with high boiling points. Both CarboxensTM and CarbosievesTM are not hydrophobic and so do need water removal for sampling humid ambient air samples. Charcoals are not suitable for thermal desorption because they are too strong to release most of the analytes with only heat. However, they are sometimes used in multi-adsorbent traps for very volatile analytes e.g. Halocarbon 12 and Chloromethane. Charcoals are hydrophilic (Brown and Shirey, 2001). Multi-adsorbent traps with up to four different sorbents allow a wide range of volatile compounds to be enriched simultaneously. Sorbents are arranged in order of increasing sorbent strength from the sampling end. Thermal desorption is in reverse direction to sampling flow so that low-volatile compounds do not come in contact with the stronger adsorbent for highly volatile analytes. Care should be taken when choosing sorbents for multi-adsorbent traps or tubes. The temperature required for conditioning the most thermally-stable sorbent must not exceed the maximum temperature of any other. Migration of loosely bound analytes from weak to strong adsorbent (e.g. from Tenax TA to a carbon molecular sieve) has to be inhibited by extending the bed length of the weaker sorbent or inserting a medium strength sorbent between (Woolfenden, 2010b). Multiadsorbent traps applied for oxygenated VOCs are for example CarbopackTM B : CarboxenTM 1000, 90 mg in total (Hopkins et al., 2003), 75 mg CarbopackTM B : 5 mg CarbopackTM X (Roukos et al., 2009) or 70 mg Tenax TA : 110 mg CarbotrapTM : 250 mg CarbosieveTM SIII (Folkers, 2002). There are different adsorbent bed sizes and densities depending on application and analytes. To allow high sampling flow rates coarse sorbent grain sizes (20/40 mesh) have to be used (Helmig and Greenberg, 1994). Important characteristics of the most common sorbents and their adequacy for OVOCs analysis are summarized in table 1. 49

52 Table 17 Sorbent Quartz wool/silica beads Carbograph TM 2TD Carbopack TM C Carbotrap TM C Tenax TA Carbograph TM 1TD Carbograph TM B Carbopack TM B Carbotrap TM Chromosorb 102 PoraPak TM Q Chromosorb 106 PoraPak TM N HayeSep TM D Carbograph TM 5TD Class Fused silica Graphitised carbon black Porous polymer Graphitised carbon black Porous polymer Porous polymer Porous polymer Porous polymer Porous polymer Graphitised carbon black Strengt h Very weak Max. Temp. [ C] Relative analyte size to n- alkanes >450 C 30 -C 40 Weak >450 C 8 -C 20 Weak 350 C 6 -C 30 Weak/ mediu m Mediu m Mediu m Mediu m Mediu m Mediu m Mediu m/stro ng >450 C 5/6 -C C 5 -C 12 Yes 250 C 5 -C 12 Yes 225 C 5 -C 12 Yes 180 C 5 -C 8 Yes 290 >450 C 3/4 -C 8 Adequacy for OVOCS (e.g. methanol, ethanol, ketones, aldehydes) No, too weak but suitable for cryogen enrichment No, too weak No, too weak (Leibrock, 1996; Woolfenden, 2010b) Yes (Hopkinset al., 2003; Roukos et al., 2009) Yes (Legreid, 2006) Characteristics Very inert, non-water retentive, hydrophobic, minimal inherent artefacts, friable, 40/60 mesh recommended to minimise back pressure Very inert, hydrophobic, minimal inherent artefacts, friable, 40/60 mesh recommended to minimise back pressure, O 3 artefacts Too weak for acetone and n-pentane, high benzene blank value, inert, hydrophobic, low inherent artefacts (e.g. aldehydes - Helmig and Greenberg, 1994), NMHCS, aldehyde and ketone artefacts in combination with O 3 (Lee et al., 2006), prone to chemical degradation and aging effects (Helmig and Greenberg, 1994) Hydrophobic, minimal inherent artefacts, friable, formation of fines, 40/60 mesh recommended to minimise back pressure, aldehyde and ketone artefacts in combination with O 3 (Lee et al., 2006) Inert, hydrophobic, high inherent artefact levels Inert, hydrophobic, high inherent artefact levels Inert, hydrophobic, high inherent artefact levels Inert, hydrophobic, high inherent artefact levels Inert, hydrophobic, high inherent artefact levels Hydrophobic, minimal inherent artefacts, friable, formation of fines, 40/60 50

53 Carbopack TM X Carboxen TM 569 Unicarb TM Carboxen TM 1003 Carbosieve TM SIII Molecular sieve 5Å Molecular sieve 13x Graphitised carbon black Carbonised molecular sieve Carbonised molecular sieve Carbonised molecular sieve Carbonised molecular sieve Molecular sieve Molecular sieve Mediu m/stro ng >450 C 3 -C 9 Strong >450 C 2 -C 5 Yes Strong >450 C 3 -C 8 Yes Very strong Very strong Very strong Very strong >450 C 2 -C 5 Yes >450 C 2 -C 5 Yes >400 C 2 -C 5 >400 C 2 -C 5 Yes (Roukos et al., 2009) No (not inert) No (not intert) mesh recommended to minimise back pressure, retention of very volatile compounds e.g. 1,3- butadiene Hydrophobic, minimal inherent artefacts, friable, formation of fines, 40/60 mesh recommended to minimise back pressure, retention of very volatile compounds e.g. 1,3- butadiene, no O 3 artefacts (Lee et al., 2006) Inert, less hydrophilic than most carbonised molecular sieves, minimal inherent artefacts Inert, hydrophilic, performance weakened in humid conditions, individual inherent artefacts, must be conditioned slowly, requires extensive purge to remove permanent gases Inert, hydrophilic, performance weakened in humid conditions, individual inherent artefacts, must be conditioned slowly, requires extensive purge to remove permanent gases Inert, minimal inherent artefacts, significantly water and CO 2 retentive, performance weakened in humid conditions, cold trap not lower than 0 C, easily and irreversibly contaminated by higher boiling components protect with front bed of weaker sorbent High inherent artefacts, significantly hydrophilic, not suitable in humid conditions, easily and irreversibly contaminated by higher boiling components High inherent artefacts, significantly hydrophilic, not suitable in humid 51

54 Charcoal Activated carbon Very strong >400 C 2 -C 4 No (Woolfende n, 2010a) conditions, easily and irreversibly contaminated by higher boiling components Limited to solvent extraction (too strong and reactive for thermal desorption metal content), hydrophilic, poor sensitivity only for ppm level concentrations, analytical interference when using MS detection Trademarks: Tenax TA - Buchem bv, Netherlands; Chromosorb - Celite Corporation, USA; PoraPak TM Waters Corporation, USA; Carbograph TM LARA s.r.l., Italy; UniCarb TM Markes International Ltd., UK, USA; HayeSep TM Hayes Separations Inc., USA; Carbotrap TM, Carbopack TM, Carboxen TM and Carbosieve TM Sigma-Aldrich, USA 52

55 APPENDIX 3: CHROMATOGRAPHIC SEPARATION (COMPILED FROM LITERATURE BY J. ENGLERT) There are two types of capillary columns that are most widely used for the analysis of volatile organic compounds (VOCs): PLOT (Porous Layer Open Tubular) and WCOT (Wall Coated Open Tubular) columns. PLOT columns feature a solid stationary phase consisting of a thin layer of small and porous particles (adsorbent) adhered to the surface of the tubing. Chromatographic results are achieved by adsorption of the analytes on the surface of the stationary phase by either surface charge interactions or shape selectivity and size exclusion interactions. PLOT columns in contrast to weaker retaining dimethylpolysiloxane columns are able to separate VOCs at ambient and above ambient oven temperatures which reduces liquid nitrogen consumption that is necessary in case of WCOT columns. Special highly polar OVOCs PLOT columns do not essentially retain most NMHCs as they have little or limited interactions with the surface of the stationary phase. By this way OVOCs are isolated and generally no co-elutions with NMHCs will appear. Thus, in principle a non-specific detector (flame ionisation detector FID) can be used as single detector. The disadvantage of PLOT columns is the need for water removal from the sample gas, otherwise sharp water peaks will co-elute with OVOCs, e.g. with propanal and acrolein on GS-OxyPLOT (Agilent). Furthermore, most PLOT columns are sensitive to water with respect to shifts in retention times depending on the moisture content of the ambient air sample. Another issue of PLOT columns may be occasionally occurring mobilisation of particles from the stationary phase (problem especially for MS), but this effect has decreased due to better bonding of the porous polymer layer. WCOT columns have a liquid stationary phase. They separate the solutes with different polarities and solubility depending on the physical properties of the stationary phase, e.g. in non-polar films the analytes dissolve according to the boiling points. The polar/non-polar interactions are much weaker than the adsorptive interactions in PLOT columns. There a two types of films: non polar dimethylpolysiloxane or polar polyethylene glycol. Dimethylpolysiloxane columns are versatile, very stable and can be operated at very low temperatures. But there are co-elution problems of OVOCS with NMHCs and so there is the need for a specific detector (MS). Another disadvantage is the low retention of alcohols on dimethylpolysiloxane columns. On the contrary on polyethylene glycol columns alcohols have high retention. Concurrently NMHCs have lower retention so that there are less co-elutions with OVOCs. But a drawback is the fact that aldehydes have also low retention. Furthermore, polyethylene glycol columns have shorter lifetimes, are susceptible to damage upon overheating or exposure to oxygen and they cannot be operated at sub-ambient oven temperatures. 53

56 1. PLOT columns Table 1: PLOT columns PLOT column equivalents GS-OxyPLOT (Agilent), CP-LowOx (Varian) CP-PoraBOND U (Agilent resp. Varian) AlO 3 PLOT (Agilent resp. Varian) Polarity High polar Midpolar High polar Composition Proprietary, salt deactivated Styrene-glycol methacrylate copolymer Proprietary, salt deactivated Operable temperature range 0 C to 350 C -100 C to 300 C -100 C to 200 C Analysis of alcohols Analysis of aldehydes Analysis of ketones Analysis of ethers Analysis of esters Analysis of aromatics Analysis of alkanes Analysis of terpenes +/- + Analysis of nitriles + + Methanol+n-butane, butanal+benzene+ n-butane and ethyne Expected co-elution problems Ethyl acetate+mvk+mek (2-butanone), water peak+propanal and acrolein ethylacetate+mvk, 2- butanol+mek, butylacetate+ ethylbenzene+m+pxylene+n-hexanal, pentanal+toluene isohexanes isoheptanes m/p-xylene Advantage Strong selectivity to OVOCs, high retention of OVOCs even at above ambient oven temperatures, no retention of saturated aliphatic NMHCS and so no co-elutions with Water resistance, retention times not influenced by water, long lifetime Strong selectivity on light hydrocarbons 54

57 OVOCs, long lifetime Disadvantage Need for humidity management, retention of water, tailing of unsaturated OVOCs due to reactions with the polar column, unsaturated NMHCs and aromatics both with carbon atom numbers higher than eleven stick in the column Co-elutions of OVOCs with aliphatic NMHCs, retention of water Not useful for OVOCs 55

58 Examples of ambient air chromatograms Deliverable WP4 / D4.9 1A) Al 2 O 3 (KCl) (from Rigi, Switzerland, Empa) 56

59 57 Deliverable WP4 / D4.9

60 Fig.1: Al 2 O 3 (KCl): a typical chromatogram at Rigi (Switzerland). 58

61 1B) LowOx Fig.2: CP-LowOx (Varian), 10 m x 0.53 mm x 10.0 µm (Hopkins et al., 2003). 59

62 Fig.3: CP-LowOx (Varian), 30 m x 0.53 mm x 10.0 µm (Roukos et al., 2009). 60

63 Fig. 4: CP-LowOx (Varian), 30 m x 0.53 mm x 10.0 µm (measurements École des Mines de Douai, Environmental & Chemistry Department, site: Paris suburban, 2010). 1C) PoraBOND U Fig. 5: CP-PoraBOND U (Varian), 25 m x 0.32 mm x 7.0 µm (measurements Empa 2012; system description: Ledreid, 2006): 4.0 min Methylether, 5.0 min Methanol, 5.1 min n-butane, 5.5 min. 1,3-Butadiene, 5.9 min Acetaldehyde, 7.9 min Ethanol, 9.2 min Isoprene, 9.9 min Acrolein, 10.0 min Propanal, 10.6 min Methylacetate, 10.8 min Isopropanol, 11.1 min Acetone, 13.0 min MTBE, 13.3 Methacrolein, 12.6 n- Propanol, 14.8 Ethylacetate, 14.9 Butanal + Benzene, 15.1 MVK, Butanol, 15.6 MEK, Methyl-3-butene-2-ol, 17.8 n-butanol, 19.8 Pentanal + Toluene, 24.1 Butylacetate + Ethylbenzene + m+p-xylene + n-hexanal, 24.8 o-xylene, 29.0 Benzaldehyde. 61

64 2. Dimethylpolysiloxane columns Deliverable WP4 / D4.9 Table 2: Dimethylpolysiloxane columns WCOT column equivalents DB-1 (Agilent), CP-Sil 5 CB (Varian), Rtx-1 (Restek), BP-1 (SGE), SPB-1 (Supelco) HP-5ms resp. DB-5 (Agilent), CP-Sil 8 CB (Varian), Rtx-5ms (Restek), BPX-5 (SGE), SPB-5 (Supelco) DB-624 (Agilent resp. Varian), Rtx-624 (Restek) Polarity Non-polar Non-polar Midpolar Composition 100% Dimethylpolysiloxane 5%-Phenyl-95%- methylpolysiloxane 6% Cyanopropylphenyl- 94%- dimethylpolysiloxane Operable temperature range -60 C to 350 C -60 C to 350 C -20 C to 260 C Analysis of alcohols Tailing Tailing + Analysis of aldehydes Analysis of ketones Analysis of ethers Analysis of esters Analysis of aromatics Analysis of alkanes Analysis of terpenes Analysis of nitriles Expected co-elution problems Advantage Disadvantage Propanal+acetone, ethanol+acetone, n-pentane+acetone, n-butane+ acetaldehyde, OVOCs+ NMHCs High thermal stability Low selectivity, tailing of alcohols and ketones, coelutions of OVOCs with NMHCs n-butane+acetaldehyde+ methanol, isobutene+ methanol, ethanol+isopentane, acetone+propanal+ isopropanol, butanal+mek, OVOCs+NMHCs More selective than DB- 1, high thermal stability Tailing of alcohols and ketones, co-elutions of OVOCswith NMHCs Propanal+acetone, OVOCs+NMHCs Good retention of alcohols, good selectivity, good thermal stability Co-elutions of OVOCs with NMHCs 62

65 Deliverable WP4 / D4.9 2A) DB-1 Fig. 6: DB-1 (Agilent J&W), 100 m x 0.25 mm x 0.5 µm (Riemer et al., 1998). 63