Article Cite This: Anal. Chem. XXXX, XXX, XXX−XXX
pubs.acs.org/ac
Compact, Automated, Inexpensive, and Field-Deployable VacuumOutlet Gas Chromatograph for Trace-Concentration Gas-Phase Organic Compounds Kate M. Skog,† Fulizi Xiong,† Hitoshi Kawashima,† Evan Doyle,† Ricardo Soto,† and Drew R. Gentner*,†,‡
Downloaded via LA TROBE UNIV on January 4, 2019 at 13:14:40 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
†
Department of Chemical & Environmental Engineering, School of Engineering and Applied Science, Yale University, New Haven, Connecticut 06511, United States ‡ SEARCH (Solutions for Energy, Air, Climate and Health) Center, Yale University, New Haven, Connecticut 06511, United States S Supporting Information *
ABSTRACT: The identification and quantification of gasphase organic compounds, such as volatile organic compounds (VOCs), frequently use gas chromatography (GC), which typically requires high-purity compressed gases. We have developed a new instrument for trace-concentration measurements of VOCs and intermediate-volatility compounds of up to 14 carbon atoms in a fully automated (computer-free), independent, low-cost, compact GC-based system for the quantitative analysis of complex mixtures without the need for compressed, high-purity gases or expensive detectors. Through adsorptive analyte preconcentration, vacuum GC, photoionization detectors, and need-based water-vapor control, we enable sensitive and selective measurements with picogram-level limits of detection (i.e., under 15 ppt in a 4 L sample for most compounds). We validate performance against a commercial pressurized GC, including resolving challenging isomers of similar volatility, such as ethylbenzene and m/p-xylene. We employ vacuum GC across the whole column with filtered air as a carrier gas, producing longterm system stability and performance over a wide range of analytes. Through theory and experiments, we present variations in analyte diffusivities in the mobile phase, analyte elution temperatures, optimal linear velocities, and separation-plate heights with vacuum GC in air at different pressures, and we optimize our instrument to exploit these differences. At 2−6 psia, the molecular diffusion coefficients are 6.4−2.1 times larger and the elution temperatures are 39−92 °C lower than with pressurized GC with helium (at 30 psig) depending on the molecular structure, and we find a wide range of optimal linear velocities (up to 60 cm s−1) that are faster with broader tolerances than with pressurized-N2 GC.
G
power micro-GCs for single or multidimensional GC, including GC with etched chips.6,8,9,13−20 Existing devices have been typically employed for VOC measurements in workplaces or indoor environments with higher concentrations (i.e., ppb or ppm) or for focused detection of targeted analytes in the parts per trillion range (e.g., explosives and TCE).12,18,21 However, Zhong et al. detected most of their 31 VOCs down to sub-parts-per-billion levels.22 One limitation for many of these and many commercially available portable GC-MS instruments is the necessity of on-board carrier gas cylinders (e.g., PerkinElmer Torion T-9 and those in refs8, 11, and 23). In addition, standalone low-cost VOC sensors are commercially available, but they are limited to ppb or ppm concentrations and often report lumped measurements without detailed chemical speciation.
as-phase organic compounds, including volatile organic compounds (VOCs), play central roles in the composition and chemistry of the atmosphere and indoor air as well as in a diverse range of biological, industrial, and commercial processes. Chemically speciated measurements of VOCs and intermediate-volatility organic compounds (IVOCs; e.g., C12− C18 volatility range) are also important because of their direct or indirect effects on health, the environment, and climate.1,2 The cost or size of accurate, commercially available instrumentation, especially for working with trace concentrations, such as online mass spectrometers (MS) or gas chromatographs (GCs) with relevant detectors (e.g., PerkinElmer Clarus or PAMS Auto-GC), frequently limits accessibility to measurements. GCs and their detectors are also limited by necessary high-purity, compressed gases and other infrastructure (e.g., low-pressure vacuum pumps). Nonetheless, customized research GCs have become smaller and more portable with new heating techniques;3−9 microscale inlets and preconcentrators;10−12 and microfabrication of low© XXXX American Chemical Society
Received: July 11, 2018 Accepted: November 15, 2018
A
DOI: 10.1021/acs.analchem.8b03095 Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
■
A common approach to trace (i.e., ppt to low ppb) measurements is the preconcentration or “trapping” of analytes from several liters of sample air onto a sometimes cooled adsorbent material via physical or chemical adsorption or via similar nonadsorbent-based cryofocusing.4,11,12,24−26 A range of adsorbents are commercially available, but they require application-dependent selection, as previous work with a subset of adsorbents reported that some analytes can decompose in the presence of O2 at high temperatures.27 Adsorbents should be selected to minimize water adsorption, or other water-vapor mitigation (e.g., trapping or purging) should be used. High-purity pressurized carrier gases remain a major consumable for GC operations. In conventional GC, the use of pressurized N2 as a carrier gas requires slower optimal linear velocities because of lower diffusion coefficients of analytes in N2 versus in He or H2.28,29 However, vacuum GC separations operated at subambient column pressures have been shown to have faster analysis times.30−32 In vacuum GC, orifices or short microbore column segments at the inlet create a large pressure drop before the GC column, produce vacuum conditions across the column, and require short or megabore analytical columns to reduce pressure differentials across the column.14,27−38 These lower pressures result in higher diffusion coefficients within the column,28,29,32 which produce faster optimal carrier-gas velocities and therefore higher-speed separations.39−41 Additionally, the use of air (i.e., N2, O2, and Ar) as a carrier gas has shown promise but requires careful selection of the column active phase to prevent degradation with O2 at elevated temperatures.12,22,29,31 A variety of detectors exist for GC applications, but some require supply gases (e.g., makeup gas or fuel). Whereas, photoionization detectors (PIDs) represent a good option for small, affordable, independent VOC detectors. Although they have been integrated with larger GC systems in the past,42,43 sufficient, leak-proof housings designed for vacuum GC and targeted analyte delivery are not available. PIDs operate using 9.6, 10.0, 10.6, or 11.7 eV ionization lamps containing a noble gas that targets analytes with equal or lower ionization energies. Developments in micro-electromechanical-system (MEMS) sensor arrays22,37,44−50 and miniature mass spectrometers present more selective options for VOC detection independently (i.e., ambient MS) or after GC separation.45,51−64 Such sensor arrays and mini-MS could be integrated with our system in future work. We present a new instrument that performs similarly to more expensive, conventional GC-based instruments, to make parts-per-trillion-level measurements of separated VOCs and IVOCs in a fully portable, independent, and automated setup that is completely free of gas-supply cylinders. This is accomplished via the successful integration of a VOC− IVOC-preconcentration system, purified air as the carrier gas, vacuum GC with temperature control, and sensitive PID quantification. The objectives of this paper are to (a) present the design and operation of our instrumentation; (b) characterize and quantify the theoretical operation and advantages of vacuum GC, including with air as a carrier gas; (c) evaluate the optimized performance of our instrumentation and compare it with that of conventional pressurized GC; and (d) discuss future directions for research and applications.
Article
EXPERIMENTAL SECTION
Instrument Design. The whole instrument measured 14′′ × 12′′ × 8′′ and was supplied with 115 VAC power. The flow path of analytes proceeded according to Figure 1. All tubing that encountered analytes was constantly heated to 55 °C (Thermon 120 VAC heating cable, 10-FLX-1) and made of PEEK or passivated stainless steel (AMCX, Inertium). Flows were directed via two- and three-way valves (Parker Series 25) and pulled by a microdiaphragm pump (Parker CTS-12 Series) with a minimum pressure of 4.0 psia. The pump voltage was electronically controlled to keep a consistent pump speed to maintain stable pressure and flow through the instrument. The inlet was equipped with a PM filter (PTFE, 1 μm pore size, Tisch) to remove aerosols or a sodium thiosulfate treated quartz filter (Tisch) to remove aerosols and ozone (when necessary), which could have reacted with the collected analytes.65 During sample collection, analytes were collected over 5−40 min at flows of 100−300 sccm (determined by a downstream constriction) onto a thermoelectrically cooled (5−15 °C) preconcentration trap filled with hydrophobic adsorbents (Sigma-Aldrich quartz-wool plugs around a 75 mg, Restek ResSil-B in a 0.065′′ i.d. tube).
Figure 1. Simplified instrument flow diagram with “collection” and “analysis” modes shown.
During analysis, air from the inlet was pulled through an activated-carbon trap (Sigma-Aldrich, 29204, 10 g), to remove background VOCs, and a microbore (0.007′′ i.d.) constriction, which set flow at 0.8−2.2 sccm (application dependent). All flow paths from the trap to the detector were carefully designed to minimize deadspace and reduce volume upstream of the column. For analysis, flow was reversed through the adsorbent trap, which was then heated to 180 °C at 150 °C min−1 or greater (McMaster 120 VAC cartridge heaters, 8376T27) to desorb analytes in as narrow a plug as possible, and this was held for several minutes to ensure no carryover of heavy analytes. All but the lightest compounds were then cryofocused at the head of a megabore GC column (MXT-1, -5, or -200; 0.53 mm i.d.; 15 m length; phase thickness of 1−3 μm; Restek), which was thermoelectrically cooled (TE Tech, VT-127-1.4-1.5-72, wired in parallel) to subambient temperatures (5−15 °C) in a low-thermal mass oven. The oven was cooled and heated (Omega 120 VAC cartridge heaters, B
DOI: 10.1021/acs.analchem.8b03095 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
proportional to 1/P1.286 instead.69 Here, the diffusion coefficients in air under vacuum are conservatively calculated as 1/P according to eq 1, but they could be higher, which would result in more theoretical plates and better separations. Calculations for the theoretical plate height were made using the extended Golay equation:70 ÄÅ ÉÑ ÅÅ dc 2 Ñ ÑÑ 2 ÅÅ (11 k 6 k 1) + + ÑÑ ÅÅ 2Dm 2 Å ÑÑÑ H = ÅÅ u + ÑÑ 24Dm(k + 1)2 ÅÅÅ u ÑÑ ÅÅ ÑÑ ÅÇ ÑÖ 4 2 2kdf2 9(P − 1)(P − 1) 3(P 2 − 1) u × + 8(P 3 − 1)2 3Ds(k + 1)2 2(P 3 − 1)
HDC00022) via a temperature program set according to application needs. In the presented data, an MXT-1 column in the oven was held at 7.5 °C for 5.35 min and heated at 10 °C min−1 to 185 °C (5−20 min hold). When necessary on the basis of relative-humidity (RH)-sensor data, adsorbed moisture was purged from the system at high flows via filtered and desiccated air (activated carbon and 3 Å molecular sieves, 12 g, Sigma-Aldrich, 208574). Dry-purging occurred at the end of the oven-temperature ramp after the analytes of interest were eluted, while the oven and trap were still at elevated temperatures. The effluent from the GC column was measured via a PID with a 10.6, 10.0, or 9.6 eV lamp (Alphasense, PID-AH) depending on the application and analytes of interest. Column effluent was plumbed directly into the ionization region using a custom PID housing that also measured pressure, RH, and temperature via microsensors (STMicroelectronics, Sensirion), which were all used for system diagnostics and PID-signal standardization. Chromatogram analysis and peak integration was handled using the TERN (TAG ExploreR and iNtegration) software package in IgorPro.66 A PID was used because it does not require supply gases, it operates well at vacuum GC pressures without specialized pumping, and it is sensitive with fairly straightforward electronics for signal capture. PIDs are known to be sensitive to pressure, temperature, and relative humidity, which were tracked immediately after the detector electrodes and were used for signal correction. High precision required a stable supply voltage to the PID and could have affected linearity and sensitivity if not properly set and regulated.67 The noble-gas composition of the PID lamp determined the range of efficiently ionizable analytes (Figures S1 and S2). This presented some limitations for measuring certain analytes (e.g., C2−C4 alkanes and many halogenated species) but also opportunities for partially selective measurements of analytes that were still responsive to low-energy 9.6 or 10.0 eV lamps (e.g., aromatics). Response factors will vary because of differences in analyte ionization energies and can span 2 orders of magnitude.67,68 All system operations and independent temperature control for the oven and trap were managed by custom electronics with multiple Cypress PSoC ARM microprocessors; no computer was required. All electronics were designed to be ultralow noise, with I2C digital communications to maintain accuracy and precision. Data was recorded locally on an SD card and can be transmitted via a cellular network. Heat dissipation was achieved through the use of three fans, one each for the oven, trap, and electronics. Instrument supplies cost less than $2,500. Methods. Diffusion coefficients in the mobile phase were calculated using the Fuller-Schettler and Giddings equation:69
(
(0.00143)T1.75 DAB =
1 MA
+
1 MB
( )
(2)
where H is the plate height, Dm is the diffusion coefficient in the mobile phase, u is the linear velocity, k is the partition ratio, dc is the inner diameter of the column, P is the ratio of columninlet pressure to column-outlet pressure, df is the film thickness of the stationary phase, and Ds is the diffusion coefficient in the stationary phase. Extra-column band broadening is considered in the SI. To better represent the conditions across the whole column, the diffusion coefficients in the mobile phase were calculated using the average pressure within the column.40 Diffusion coefficients of n-alkanes in dimethyl-polysiloxane were calculated using the following equation:71 ln(Ds) = −8.870 + 0.3836cn +
where Pvp is the pure vapor pressure (bar), Pc is the vapor− liquid critical pressure (bar), Tc is the vapor−liquid critical temperature (K), and a, b, and c are constants. The difference in effective elution temperature was calculated as the difference between the equivalent vapor-liquid equilibrium partitioning ratios (Pvp(T)/PGC) for vacuum GC versus pressurized GC, and it was confirmed via the Clausius-Clapeyron equation (see the SI). Throughout the study, 45 and 30 psia (30 and 15 psig) were chosen for the comparison with pressurized GC as examples within the diverse range of typical operating pressures (the 30 psia results are shown in the SI). Authentic liquid and gas VOC and IVOC standards (Accustandard, Apel-Riemer, and Sigma-Aldrich) were used for calibration and testing. Liquid standards with 1−13 compounds were all diluted in methanol, which does not produce significant signal on a PID. Gas standards were diluted from a multicomponent standard cylinder (Table S1) with concentrations spanning an order
0.5
2
(3)
where cn is the number of carbon atoms in the molecule, and R is the ideal-gas constant (1.987 cal/K/mol). The optimal linear velocity (uopt) was then determined as the minimum of H, and to examine optimal range, we calculated the linear velocities for H values within 5−10% of the minimum. Temperature-dependent vapor pressures for C5−C18 nalkanes were calculated using parameters from Poling et al.72 via: ÄÅ 1.5 Å ij TC ÅÅÅ ijj T yzz T yzz j ln(Pvp) = ln(PC) + z + bjj1 − z ÅÅaj1 − j T ÅÅÅÅ jjk TC zz{ TC zz{ k Ç É 2.5 5Ñ ij yz ij yz ÑÑÑÑ T T zz + djj1 − zz ÑÑ + c jjj1 − jj j TC zz{ TC zz{ ÑÑÑÑ k k (4) Ö
)
1/3 PGC(sumv1/3 A + sumv B )
(44.61 − 498.1cn) RT
(1)
where DAB is the diffusion coefficient of compound A in gas B (cm2/s), T is the temperature (K), PGC is the GC pressure (atm), MA and MB are the molar masses, and sumvA and sumvB are the sums of the atomic-diffusion volumes for A and B. The relationship between the diffusion coefficients and pressure has been treated as being inversely proportional (D ∝ 1/P) in many studies, but other evidence suggests that it is C
DOI: 10.1021/acs.analchem.8b03095 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Analyte GC Elution Temperatures. In addition to its effect on the molecular diffusivity, the pressure of the GC mobile phase also modifies the vapor−liquid equilibrium of analytes in the GC column and thus the retention factor (k) of a given analyte (at a fixed temperature). Two studies have mentioned differences ranging from 30 to 70 °C with the application of vacuum GC, but these differences in elution temperature had not been theoretically explored or modeled, which limited fundamental understanding.28,30 To more closely examine the effect of pressure, we calculated the temperatures necessary to achieve similar vapor-liquid equilibrium partitioning at a range of vacuum pressures compared with those at 30 and 45 psia (Figure 3). At 4 psia, the temperature difference versus pressurized GC at 45 psia is −45 to −75 °C, depending on the retention index, and −36 to −60 °C when compared with pressurized GC at 30 psia (Figure S5).
of magnitude (17.9−301 ppb), producing concentrations ranging from 6 ppt to 7 ppb. Breakthrough tests used the same standard concentrations from 50 ppt to 14 ppb or mass loadings of 3 to 150 ng (see the SI). The breakthrough testing setup and 1/4′′ o.d. back-up adsorbent tubes are described in detail in Sheu et al.73 Indoor sampling took place in New Haven, CT. All dilution and flow measurements were collected via temporary (external or internal), independent mass-flow meters (Alicat). PID sensitivities to T, RH, and P were assessed within the instrument and on a similar benchtop setup in which a toluene gas standard was introduced directly into the PID housing.
■
RESULTS AND DISCUSSION
Theoretical Operation and Advantages. Analyte Diffusivity in Vacuum GC with Air. Molecular diffusivity is dependent not only on the analyte and carrier gas but also on the temperature and pressure in the GC column. N2 carrier gas traditionally requires a slower separation because of the low diffusivities shown in Figure 2, but dropping the pressure to vacuum conditions greatly improves performance. Figure 2 shows that decreasing the pressure greatly increases the diffusivity of analytes within a GC column, which ultimately greatly increases the number of “plates” in a GC column and its resolving capability. Diffusion coefficients at ∼12.7 psia in air are roughly equivalent to those in helium at 45 psia, and at lower pressures, the diffusivity of analytes is higher than with 45 psia helium. The diffusion coefficients at 6, 4, and 2 psia are 2.1, 3.2, and 6.4 times greater than that of helium at 45 psia. However, the number of potential plates in the column with vacuum GC is limited by the megabore column diameter (0.53 mm), and shorter column lengths (95% up to 8 L sampling volumes. All tests at flow rates of 200 sccm and below performed similarly, and 300 sccm testing is shown with all the data in Figure S19. Alkanes and alkenes with fewer than eight carbon atoms had limited collection, but aromatic and oxygenated species of similar volatilities did not. For these compounds, smaller samples (2−4 L) or colder collection temperatures (5 °C) improved collection efficiencies. Future work will target trapping and thermal-desorption solutions for small alkanes and alkenes. The carbonyls, alcohols, and other oxygenated species in our standard (e.g., nonanal, aromatic aldehydes, citronellol) were transmitted through and measured by the instrument, including multioxygen compounds (e.g., furural and butyl acetate). It is unlikely that the losses of the high-molecularweight alkanes in Figure 8 are due to breakthrough but rather to potential upstream losses during these tests. With 5 min desorption times, we observe no carryover of analytes in our standards or samples. Thermal desorption of temperatures tested at 185 °C did not show any evidence of decomposition (i.e., peak losses or positive artifacts) when compared with desorption in high-purity N2 (Figure S20). By monitoring RH and purging when necessary, water vapor did not present problems for trapping, separation, or detection. This performance and handling is consistent with that of the offline adsorbent traps containing quartz wool and ResSil-B in Sheu et al.73 Photoionization Detector (PID): Sensitivity and Selectivity. With our ultralow-noise circuitry, we observed precision of 30 μV with the PID, which corresponds to concentrations of 0.69 ppb inside the detector for toluene, but this will be compound-dependent. The LOD with respect to the indetector concentration was 2.5 ppb for toluene. The stability, or drift, of the PID at constant toluene concentration was 4.7% over 13 h. However, the time scale of this drift in absolute signal will have minimal effects on GC-peak integrations of relative signals. The temperature, pressure, and RH sensitivity of the PID was tested with zero air and with 0−17 ppm toluene introduced constantly to the PID housing. The PID signal increased with increasing pressure and decreasing temperature, and the effect was concentration-dependent for RH. However, calibration with toluene across RH (0−80%), temperature (21−38 °C), and pressure (270−400 hPa) ranges showed G
DOI: 10.1021/acs.analchem.8b03095 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
or regenerate our activated-carbon filter over the year of use. Water-filter replacement is sample-RH-dependent, and the water filter did not need to be changed in the lab; the current capacity allows for up to 47 days of purge flow at 200 sccm, which is only used for a few minutes at a time. In high-RH environments, additional water-management solutions upstream of the inlet could be used to reduce the frequency of replacement.
■
CONCLUSIONS
■
ASSOCIATED CONTENT
This work advances the theory and application of vacuum GC using air as a carrier gas and achieves low 1−15 ppt LODs. It is applicable and flexible across a variety of applications with gasphase VOCs and IVOCs. The fully automated, cylinder-free, low-cost (
