Detection of petroleum hydrocarbons at low ppb levels using quartz resonator sensors and instrumentation of a smart environmental monitoring system Iwao Sugimoto,* Michiko Seyama and Masayuki Nakamura NTT Lifestyle and Environmental Technology Laboratories, Midori-cho, Musashino-shi, Tokyo 180–8585, Japan Received 27th November 1998, Accepted 8th February 1999 Petroleum hydrocarbon vapors at low ppb levels can be detected using a thickness shear mode resonator (TSMR) coated with a chemical-sensing overlayer, prepared by radiofrequency sputtering of porous sintered-polyethylene (PS-PE).The sensing capabilities of PS-PE sensors were profoundly aVected by the sputtering methods; they were enhanced by the photo-excitation eVect, and were reduced by carbonization and water treatment.The photoassisted PS-PE sensor was extremely sensitive and could detect linear hydrocarbon (>C12) vapors below the ppb level. The time constant of the sorption curve, however, was large, indicating a slow sensing speed. Toward creating instrumentation for a smart environmental monitoring system, the TSMR sensors were arrayed on a circuit board equipped with a serial interface and signal processing chips of the oscillation drive and frequency counter. Co-sorption with water vapor at a relative humidity of about 10% has almost no eVect on the sensing ability of PS-PE sensors for 1,2,4-trimethylbenzene.Conversely, it enhances the sensitivity of the TSMR sensor coated with a D-phenylalanine film. Upward shifts in the baseline are evident with elapsed time.However, a rigorous ten-cycle iteration test for 100 ppm toluene vapor demonstrated good reproducibility of the sensor’s signals. The continuing rise in petroleum hydrocarbon pollution is structural advantages in their chemical-sensing layers. The plasma-produced molecular networks are uniquely charac- driving the development of instrumentation that can rapidly detect spills and contamination in the ground and in water.terized by the radical sites generated by the beam-induced bond cleavage in plasmas.3 Some radicals form cross-linking Some of the largest sources of environmental pollution are the storage, distribution, and transportation facilities of petroleum bonds, yielding high atomic densities.4 Others may remain as trapped radicals, allowing flexible molecular networks. Both fuels such as gasoline, diesel oil, kerosene, and fuel oil.The demand is increasing for field-remote multi-point sensing of these features achieve high dissolution capacities for any type of solute molecules. Of the plasma processing methods, systems that can provide immediate detection of petroleum pollution. sputtering techniques have great potential for preparing highly sensitive layers because this approach is likely to produce high Both in the atmosphere and in water, optical and piezoelectric sensing systems are promising instrumentation for atomic density films with high gas-sorption capacities due to the energetic character of the sputtered species.These sputtered monitoring the level of environmental hydrocarbon pollutants.The optical type has been commercialized by a US company.1 species are likely to enhance the adhesiveness to the transducing film-substrate suitable for chemical-sensing overlayers. The piezoelectric type has been extensively exploited for practical interest due to its capability of measuring extremely small From the chemical sensing aspect, we focused particularly on the fact that PPF networks produced by radiofrequency mass changes below the nanogram level.Among several piezoelectric devices, the thickness shear mode resonator (TSMR) sputtering contain unsaturated carbons (dangling bonds and multiple bonds). These can interact with solute molecules by has been widely used as a mass transducer owing to its high oscillation stability and high signal-to-noise (S/N) ratio.2 specific spin-related interaction5 and p-electron-derived interactions: p–p,6 cation–p,7 CH–p8 and Cl–p9,10 interactions.The The gas-sorption chemistry in chemical-sensing overlayers of TSMRs governs their sensitivities and selectivities. To concentrations of these unsaturated carbons can be increased by photo-assisted sputtering using ultraviolet (UV) light exci- enhance the monitoring ability of the TSMR sensors for petroleum pollution, the overlayers should be interactive with tation.Thus, they play important roles in characterizing the gas-sorption behavior of PPFs.11 hydrocarbons, not with water. In monitoring the gas phase in the field, water vapor is considered the strongest competitor In this paper, we report on PPF-coated TSMR sensors for volatile petroleum hydrocarbons with detection limits at the to the target materials.In the environmental atmosphere, temperature, humidity, and wind are ever-fluctuating and sub-ppb level. The extremely sensitive films can be prepared by a photo-assisted sputtering technique. Our approach uses aVect the concentration of water vapor as well as the target materials.Fortunately, the lipophilic sensing layers are hydro- a porous sintered polyethylene (PE) disk as a sputtering target and UV light irradiation as a supplementary plasma excitation phobic due to their non-polar character. For satisfactory sensing ability, the detection limit for petroleum hydrocarbons source. It has recently been reported that PPFs composed of hydrocarbon networks can be obtained by sputtering a porous in the gas phase must be below the low ppb level.Plasma processing to produce plasma-polymer films (PPFs) sintered PE disk, not the commonly applied low-density PE disk, which is widely manufactured by compression molding.12 is conducted for the most part in a vacuum; this approach is inherently superior to wet processing, because it allows cleanli- We have developed an environmental gas monitoring system using an array of these PPF-coated TSMR sensors.The array ness without the need for organic solvents. In addition, polymeric films produced by plasma processing have salient was installed on a circuit board equipped with a serial interface J. Environ. Monit., 1999, 1, 135–142 135for multi-point sensing and signal processing chips of the compression-molded PE disk.The PE disk coated with the amino acid was then used as a sputtering target after the oscillation drive and frequency counter. Using the dynamic sensor response, the pattern matching technique can classify ethanol had fully evaporated. Helium was introduced into the sputtering chamber at 4 mL min-1, and the rf power was the gases based on a discrimination algorithm.The system can provide on-line monitoring of atmospheric conditions and can applied to the lower target electrode at a density of 0.42 Wcm-2. The detailed procedure of the rf sputtering of classify the gas species via telecommunications networks. amino acids has been described previously.15 Experimental Instrumentation of smart gas-sensing system for identifying environmental vapors Preparation of plasma-polymer-film-coated TSMR sensors We have developed a smart environmental gas-sensing system, The TSMR was fabricated from an AT-cut quartz crystal which can detect and distinguish vapor species.It consists of plate, which was 8.5 mm in diameter and 0.1 mm thick, with a sensor module (Fig. 1), a power supply, and a signal a 5 mm diameter gold electrode.The fundamental resonance processing computer. The sensor module is composed of an frequency of this TSMR was 9 MHz and it was manufactured array of eight PPF-coated TSMR sensors installed on a circuit by Nihon Dempa Kogyo, Tokyo, Japan. board equipped with a custom LSI operating the oscillation PPFs were produced by a diode-type radiofrequency (rf ) and measuring resonance frequency and a multi-port serial sputtering apparatus using the sputter-up configuration.The interface (RS-485) for multi-point sensing. The computer interelectrode space was capable of excitation by irradiation display can show the time-dependent dynamic sensor response of UV light through a CaF2 window using a 300 W low- and gas classification map for discrimination of vapor species. pressure mercury lamp.The detailed structure of the photo- The discrimination algorithm for identifying the vapor species assisted sputtering apparatus has been described previously.13 is based on a pattern matching technique that uses learning Sputtered PPFs were deposited on both sides of the quartz vector quantization (LVQ).16 Two featured parameters were crystal plates, which were placed on the grounded upper extracted: the time constant and the maximum frequency shift electrode.The thickness of the film on each side of the quartz from the dynamic sensor response.17 This procedure based on crystal plate was controlled by the sputtering time to about LVQ is advantageous for fast and accurate discrimination of 0.5 mm.several types of vapor species. For six kinds of vapors, the Polyethylene (PE) films were prepared by using, as a sputter- processing time for determining the winning vector in the LVQ ing target, a porous PE disk (135 mm in diameter) produced map, which can visualize the results of LVQ processing, is less by sintering PE granules (with diameters ranging from 200 to than 1 s.The system can monitor the atmospheric condition, 300 mm). PE disks of two thicknesses (5 and 10 mm) were identify the vapor species, and make this information immediused. We will refer to the films produced by the supplementary ately available for on-line environmental surveillance via telephoto- excitation method as photo films. Those produced using communications networks.We have shown the validity of the a 5 mm thick disk will be termed 5 mm films, to diVerentiate sensing system for discriminating 14 chemical vapors.18 them from the major films obtained from 10 mm thick disks. The PE disk was severely blackened after about 12 sputtering Measurements of sensing capability cycles over about 300 h. The sputtered PE film obtained using this blackish PE target was carbonaceous.We will refer to this Fig. 2 shows schematically the set-up of the gas flow system film as a carbon film. The properties and structures of the for measuring the detection capability of the TSMR sensors. sputtered films are also aVected by the background pressure, The organic vapors were generated by the diVusion tube which can be considered to be a measure of the amounts of method19,20 using a PD-1B standard gas generator from residual gases in the vacuum chamber.These amounts increase Gastec, Kanagawa, Japan. A glass tube preserving the hydrowith background pressure; water is one of the main residual carbon solvent (vapor source) was kept in a temperaturegases attributable to the characteristics of plasma products.In controlled bath using synthetic air (99.9999% pure) as a carrier most cases, the background pressure was 9×10-5 Pa. In some gas and also as a standard or cleaning gas for establishing the cases, however, the atmosphere in the chamber was kept humid initial state of the TSMR sensors. Commercial hydrocarbon with water-saturated cotton and the chamber was evacuated solvents were used without further purification (>98% pure).to be 4×10-3 Pa, which was used as the background level. They are regarded as environmental-residing components of We will refer to the film prepared under humid conditions as gasoline and fuel oil and are classified by the World Health a water film. Organization as very volatile organic compounds (VVOCs), In all cases for PE films, krypton was introduced into the sputtering chamber at 6 mL min-1, and the rf power was applied to the lower electrode at a power density of 0.389 Wcm-2.A fluoropolymer-coated TSMR sensor was also fabricated. The film was produced by rf sputtering of a poly(chlorotrifluoroethylene) (PCTFE) disk, which was manufactured by Daikin Kogyo, Osaka, Japan, using ordinary compression molding.Argon was introduced into the sputtering chamber at 5 mL min-1, and the rf power was applied to the lower target electrode at a density of 1.12 Wcm-2. This fluoropolymer film, which has a solvophobic character14 except for the fluorocarbons, is considered to be a good reference for PE films. For comparison purposes, polar chemical-sensing overlayers of TSMR were also prepared, using a-amino acids.Film synthesis was carried out by rf sputtering of D-phenylalanine and DL-histidine (>98% pure, purchased from Kanto Chemicals, Tokyo, Japan). Each amino acid was suspended Fig. 1 Magnified photograph of sensor module with an embedded array of TSMR sensors. in ethanol and the dispersed solution was spread on the 136 J. Environ. Monit., 1999, 1, 135–142Fig. 2 Schematic diagram of set-up for measuring sorption capacities of TSMR sensors for organic vapors. volatile organic compounds (VOCs), and semi-volatile organic Quadrex MS (50 m x 0.53 mm x 0.25 mm) was used as the capillary column and helium was used as the carrier gas. compounds (SVOCs).21 A naphtha reference standard (Supelco, Bellefonte, PA, USA, Cat. No. 4–7488) was used for the gasoline sample.Also, special-A-grade fuel oil produced Results and discussion commercially in Japan was used for the fuel oil sample without purification. The vapor concentration was controlled by the To determine gas concentrations accurately, a cross-check shape of the diVusion tube and by its temperature, which was between the diVusion tube method and GC analysis was either 30, 40, or 50 °C.The flow rate of the carrier gas was set performed. Based on the diVusion tube method as shown in at 0.2 L min-1, using a mass flow controller. It was possible Fig. 2, the standard gases can be generated, controlling their to install eight TSMR sensors in a 30 mL sensor cell, which concentrations by changing such factors as the shape of the can be used as the sensor module in the smart gas-sensing diVusion tube, the temperature, and the flow rate of the carrier system.The sensor cell was connected to the gas flow line and gas. The gas concentration [C (ppm)] can be estimated as was placed in a temperature-controlled chamber kept at 25 °C. follows: The hydrocarbon stream generated from the gas-generation cell was introduced into the sensor cell after the baseline C= K·Dr·103 F (3) (background) fluctuation had been brought down to below 1 Hz for 10 min.Switching between the baseline and measuring modes was eVected rapidly using a four-way valve, which K= 22.4 M · 273+T 273 · P 760 (4) produced a step response and kept the gas pressure constant before and after switching. The sorption capacities were deter- where Dr (mg min-1) is the diVusion rate, F (mL min-1) the flow rate of the carrier gas, M the molecular weight of the mined by the maximum frequency shifts for 3 h sorption measurements.Sauerbrey’s equation22 was used to relate the generated gas, T (°C) the temperature of the diVusion tube, and P (mmHg) the pressure of the generating gas. frequency shift (Df ) to the change of mass loading (Dm): Table 1 summarizes the vapor concentrations, which were Dm=-[A(rqmq)1/2/2F02]Df (1) calculated using eqn.(3) and measured by GC analysis, under various conditions of diVusion tube shape and temperature. where F0 is the fundamental resonant frequency of the unloaded TSMR (9 MHz), A is the electrode area (0.39 cm2), To calculate the gas concentration precisely, the diVusion rates were determined by the weight loss of a diVusion tube, rq is the density of the quartz (2.65 g cm-3), and mq is the shear modulus of the quartz (2.95×1011 dyn cm-2). With preserving the hydrocarbon solvent at a constant temperature over a period of 1 week.These are some diVerences between these constants, we obtain Cc and Cm. In general, the values for calculated concentrations Dm(ng)=-1.05Df (Hz) (2) (Cc) over 0.1 ppm tend to be lower than the corresponding measured concentrations (Cm).Conversely, values for Cc below A frequency counter (Advantest, R5361A) was used for measuring the resonant frequencies of TSMR sensors to an 0.1 ppm tend to be higher than the corresponding Cm. Under these low-concentration conditions, the contaminants accuracy of 0.1 Hz.It is becoming increasingly important to confirm the accu- in the flow line are clearly detectable by GC analysis. In most cases, the main GC signal of the contaminants can be identified racy of the vapor concentration estimated using the diVusion tube method for low concentrations at the sub-ppm level. The as toluene by its retention time. The other clearly observable contaminants are acetone and n-decane, which are reported as concentrations of hydrocarbon vapors were measured by using gas chromatography (GC).The exhaust gas eluted from a being among the most ubiquitous pollutants in the indoor environment.23 These contaminants are probably generated sensor cell was collected in a gas sorption tube packed with Tenax TA. This sampling tube was set in a thermal desorption from the surface of the inner wall of the gas-lines and the gasgeneration cell, which was coated with fluoropolymer resin injection system (Perkin-Elmer, ATD 400) and the desorbed gas was analyzed by using an HP-5890 II gas chromatograph paint containing typical organic solvents.Even though the gas flow line was cleaned by a flow of pure synthetic air for 4 d equipped with a hydrogen flame ionization detector.A J. Environ. Monit., 1999, 1, 135–142 137Table 1 Gas concentration generated by diVusion tube method Measured vapor DiVusion DiVusion rate Calculated vapor concentration by GC Gas source tubea Temperature/°C (Dr)/mg min-1 concentration (Cc) (ppm) (Cm) (ppm) None D-1 30 — — 0.002b D-5 50 — — 0.0065b 2-Methylpentane D-1 30 17.750 25.210 28.000 2,2,4-Trimethylpentane D-1 30 5.550 5.890 2.100 n-Decane D-1 30 0.121 0.100 0.140 D-1 40 0.233 0.200 0.290 0.007c D-3 30 0.721 0.620 0.850 0.083c n-Dodecane D-2 30 0.040 0.030 0.031 D-3 30 0.114 0.080 0.039 0.005c D-4 30 0.184 0.130 0.092 0.028c n-Tetradecane D-2 50 0.017 0.010 0.005 0.012c D-3 50 0.045 0.030 0.014 0.001c n-Hexadecane D-2 50 0.005 0.003 N.D.d 0.048c D-3 50 0.012 0.007 N.D.d 0.009c D-5 50 0.040 0.022 0.003 0.002c 1,2,4-Trimethylbenzene D-1 30 0.195 0.200 0.210 0.007c D-2 30 0.466 0.470 0.650 3-Ethyltoluene D-2 30 0.729 0.740 0.920 Gasoline D-1 30 — — 1.000b D-2 30 — — 2.900b Fuel oil D-1 30 — — 0.100b D-2 30 — — 0.230b aShape of diVusion tube [cross section (mm2), length (mm)]: D-1 (1.65, 50), D-2 (4.91, 52), D-3 (12.56, 50), D-4 (18.85, 40), D-5 (27.33, 31).bConcentration estimated as n-octane concentration using the total GC signals. cConcentration of toluene, which is detected as the main co-existing compound. dNot detected; lower than the detection limit of 0.0002 ppm. while heating the pipe to 50 °C, the flow line was not completely by a gradual decrease in the frequency shift after taking the maximum positive values, which are as large as a few clean for measuring the sensor ability at the ppb level.In some cases, the concentrations of contaminants were higher than Hertz. Most of the cases marked with an asterisk belong to polar amino acid films. These negative shifts may reflect that of the vapor species to be generated, as shown in Table 1. In particular, for the semi-volatile hydrocarbon n-hexadecane re-vaporization of the sorbed (adsorbed and/or absorbed) gas molecules from the solvent PPF owing to their weak intermol- with a low vapor pressure, it is noteworthy that GC measurements can detect only the contaminants, not the n-hexadecane, ecular interactions.To clarify the background levels of the sensor responses, which was introduced into the gas-generation cell.Table 2 summarizes the downward shifts in the resonant blank tests (as indicated by ‘none’ in Table 2) were performed as follows. The frequency shifts were measured by changing frequencies, over 1 h, of the PPF-coated TSMR sensors exposed to the vapor flow shown in Table 1. In Table 2, the the pure air stream through the cleaning cell to the other pure air stream through the gas-generation cell without introducing asterisks denote negative values of the downward shift caused Table 2 Frequency shifts of PPF-coated TSMR sensors over 1 h.The asterisks denote negative values of the downward shift caused by a gradual decrease in the frequency shift after taking the maximum positive values Condition Sensing film DiVusion PE (water)/ PE (carbon)/ PE/ PE (photo)/ Phe/ His/ PCTFE/ Gas source tube Temperature/°C Hz Hz Hz Hz Hz Hz Hz None D-1 30 2 3 4 5 0 — 0 D-5 50 4 5 10 16 0 — 0 2-Methylpentane D-1 30 7 5 35 39 6 — 2 2,2,4-Trimethylpentane D-1 30 3 2 8 11 * — 1 n-Decane D-1 30 6 4 23 28 6 2 1 D-1 40 35 8 150 186 5 2 0 D-3 30 62 12 235 280 5 1 0 n-Dodecane D-2 30 10 4 22 32 9 6 0 D-3 30 56 13 83 124 7 5 0 D-4 30 84 * 169 229 6 * 0 n-Tetradecane D-2 50 17 12 25 36 9 8 0 D-3 50 18 9 29 39 16 7 0 n-Hexadecane D-2 50 8 7 27 34 5 2 0 D-3 50 5 7 16 23 21 3 0 D-5 50 18 9 31 41 11 4 0 1,2,4-Trimethylbenzene D-2 30 12 7 55 57 1 * 0 3-Ethyltoluene D-2 30 14 7 82 79 * * 0 Gasoline D-1 30 45 17 169 261 1 * 0 D-2 30 29 8 118 135 4 2 0 Fuel oil D-1 30 17 24 38 49 8 * 0 D-2 30 18 8 46 97 10 3 0 138 J.Environ. Monit., 1999, 1, 135–142a gas source. Two ‘none’ conditions correspond to the situations in which contaminant vapors can be generated at maximum (D-5, 50 °C) and minimum (D-1, 30 °C) concentrations. The total amounts of contaminants correspond to noctane concentrations of 6.5 ppb (maximum) and 2 ppb (minimum) (see Table 1). Taking into account this information on the background level of the sensors and the contamination level of the gas flow line, we will now interpret the results of the gas-sorption measurements listed in Tables 1 and 2.Compared with the amino acids and PCTFE films, PE-class films induce larger frequency shifts of the TSMR sensors for all of the hydrocarbon vapors, as shown in Table 2. In particular, PE and PE Fig. 3 Response curves of TSMR sensors for 14 ppb n-tetradecane. (photo) films are extremely sensitive and can detect the linearshaped large hydrocarbons at the sub-ppb level. The hydrocarbon networks of PE-class films are attributable to the lipophilic behavior of the TSMR sensors, applicable to environmental hydrocarbon sensing at the ppb level. The inferior sorption properties of amino acid and PCTFE films are probably due to the low values of the dispersion factor, which is one of the most important molecular descriptors in the well-established linear solvation energy relationship (LSER) equation24–26 quantifying the gas-sorption characteristics of polymeric films. In the LSER equation, the sorption capacity [corresponding to the partition coeYcient (K)] of a solvent polymeric film for the solute vapor can be expressed by a linear combination of the factors of intermolecular interactions: Fig. 4 Response curves of TSMR sensors for 650 ppb log K=C0+r(R2)+s(p2)+a(b2)+b(a2)+l(logL 16) (5) 1,2,4-trimethylbenzene. where C0 is a regression constant; r is the polarizability of the solvent polymer and R2 is that of the solute vapor; s is the 1,2,4-trimethylbenzene, which is representative of the alkylated benzenes. All the sensor response curves can be expressed by polarity of the solvent polymer and p2 is that of the solute vapor; a is the hydrogen-bonding acidity of the solvent polymer a time-dependent equation, based on the sorption/desorption kinetics presented by Langmuir,29 as follows: and b2 is the hydrogen-bonding basicity of the solute vapor; b is the hydrogen-bonding basicity of the solvent polymer and DF(t)=a[1- exp(-t/t)] (6) a2 is the hydrogen-bonding acidity of the solute vapor; and l and logL 16 are the dispersion/cavity factors of the solvent where the gas sorption starts at t=0, DF(t) is the sensor response value (frequency shifts of TSMR) at time t, a is the polymer and solute vapor. The dispersion factor is closely correlated with the lipophilicity evaluated by solubility in n- maximum value of the sensor response at the saturated state equilibrated between sorption and desorption, and t is the hexadecane at 25 °C (known as the Ostwald solubility coeYcient27).time constant. The validity of this kinetic equation has been shown elsewhere.30 In Fig. 3, all the sensors responding to Despite the high vapor concentrations (>1 ppm) of the VVOCs 2-methylpentane and 2,2,4-trimethylpentane, the fre- linear-shaped hydrocarbons have large time constants in their saturation curves, while the time constants in Fig. 4 are smaller. quency shifts of all the TSMR sensors are small for both VVOCs. These low sensitivities are probably caused by their This behavior indicates the slow diVusion of linear-shaped hydrocarbon molecules into the bulk of the film concomitant low dispersion force values, due in turn mainly to their low molecular weight and their branch-shaped molecular struc- with multilayered adsorption at the film surface.Despite the slow responses, the linear-shaped hydrocarbons can eventually tures. The latter structural factor can be explained by the thermodynamic ‘entropy eVects’28 in solvation.In general, the induce large frequency shifts due to their extensive dissolution into the solvent film and self-cohesive accumulation at the film dissolution capabilities of branch-shaped hydrocarbons are inferior to those of linear-shaped hydrocarbons. The linear- surface. The sensitive PE (photo) films are expected to induce detectable response signals for linear-shaped hydrocarbons shaped hydrocarbons are likely to be solvated without disturbing the highly ordered molecular networks of the solvent at the sub-ppb level.Conversely, the alkylated benzene 1,2,4-trimethylbenzene is not likely to diVuse into the bulk of molecules because of their high capability for taking any shape of molecule adaptable for solvent molecular networks.the film and exhibits fast saturation due to self-association occurring in the vicinity of the film surface, because it has a Compared with the background level detected by blank tests, the PE- and PE (photo)-TSMR sensors are suitable for planar and rigid molecular structure. GC analysis reveals that the eluent vapors from gasoline detecting those branch-shaped hydrocarbons classified as VVOCs.and fuel oil have no more than 46 and 39 components, respectively. The vapor concentrations of the sum of all of the The test vapors belonging to the VOCs and SVOCs can be classified as linear-shaped hydrocarbons and alkylated ben- GC signals for gasoline and fuel oil correspond to n-octane concentrations of 1.0 and 0.1 ppm, respectively.As shown in zenes. Compared with the amino acid and PCTFE films, all of the PE films induce satisfactorily large frequency shifts of Table 1, even though the vaporization conditions for both petroleum hydrocarbons are the same, the values for the the TSMR sensors, applicable for environmental monitoring, as shown in Table 2. Fig. 3 and 4 show the TSMR sensor number of components and vapor concentration of gasoline eluents are larger than those of fuel oil, because gasoline is response curves for 14 ppb (Cc) n-tetradecane, which is a typical linear-shaped hydrocarbon, and for 650 ppb (Cc) composed mainly of hydrocarbons (>C10) with higher vapor J.Environ. Monit., 1999, 1, 135–142 139Fig. 7 Response curves of TSMR sensors for 100 ppm toluene vapor Fig. 5 Response curves of TSMR sensors for gasoline vapor (corre- induced by the iterative sorption–desorption cycle. sponding to a 1.0 ppm concentration of n-octane vapor). To evaluate the sensing ability in the presence of humidity, the sensor responses of PPF-coated TSMR sensors exposed to a flow of a binary gaseous mixture of water and a petroleum hydrocarbon, e.g., 1,2,4-trimethylbenzene or n-decane, were measured.Using a humidified vapor measuring apparatus composed of the flow lines of water and petroleum hydrocarbon (as shown in Fig. 8), the background levels at which each TSMR sensor was exposed to a flow of pure synthetic air were first established by setting two sets of four-way valves to feed pure synthetic air. After the sorption equilibrium had been attained, the gas flow into the measuring cell was changed to pure water vapor by turning the four-way valve 2, leaving the four-way Fig. 6 Response curves of TSMR sensors for fuel oil vapor (corre- valve 1 in the pure synthetic air mode. All the response curves sponding to a 230 ppb concentration of n-octane vapor). of the TSMR sensors for sorption of water vapor, in which the relative humidity is at about 10%, are shown in Fig. 9. After the TSMR sensors had been exposed to pure water pressures than those of the components of fuel oil. The sensor response curves for 1.0 ppm gasoline vapor and 0.1 ppm fuel vapor for 3 h, a petroleum hydrocarbon stream was introduced into the humidified sensor cell, replacing the pure synthetic oil vapor are shown in Fig. 5 and 6, respectively. The volatile characteristics of gasoline are attributable to air, by turning the four-way valve 1. The sensor response curves for 1,2,4-trimethylbenzene and n-decane are shown in alkylated benzenes, which are present in gasoline in large amounts but are rare in fuel oil. Therefore, the sensor response Fig. 10 and 11, respectively. Fig. 10 and 11 show the frequency shifts from pure water vapor (background) to a mixture of curve of gasoline vapor (Fig. 5) is similar to that of 1,2,4-trimethylbenzene (Fig. 4).While one of the main compo- water and petroleum hydrocarbon vapors. In all of the water vapor sorption curves (as shown in nents of fuel oil are the linear-shaped hydrocarbons with their high molecular weight, the sensor response curve of fuel oil Fig. 9), the D-phenylalanine film showed a remarkably high sorption capacity for water, reflecting its polar molecular vapor (Fig. 6) resembles that of n-tetradecane (Fig. 3). Hence, the sensor response curve of petroleum products reflects that structure; this was reported previously.15 Conversely, PE-class films with a non-polar character12 had low water sorption of their main components, which can characterize the vaporization behavior of petroleum products.capacities. The PCTFE film in particular shows a negligible aYnity for water vapor, as expected from the hydrophobic To confirm the reproducibility of the sensor’s responses, iterative testing of the sorption/desorption of the typical character of the fluoropolymer. After the water vapor attained suYcient sorption at the hydrocarbon toluene was carried out.Severe conditions were set for the chemical-sensing overlayers of the TSMR sensors, quasi-equilibrium state, petroleum hydrocarbon vapor at a concentration of 200 ppb was introduced into the sensor cell to the extremely high concentration of 100 ppm, to accelerate the solvation damage induced by the restructuring of the to measure the capability for sensing petroleum hydrocarbons in the presence of humidity (Fig. 10 and 11). The responses molecular networks. Ten cycles of the sorption/desorption tests were performed varied widely between the PE-film-coated TSMR sensors. However, regardless of the type of petroleum hydrocarbon, as shown in Fig. 7. The duration of each sorption or desorption period was 30 min.All of the sensor curves responding to the response level was comparable. The frequency shifts of PE-film-coated TSMR sensors are basically the same as the sorption and desorption were then superimposed on their time-dependent drift curves. Good reproducibility was values for pure 1,2,4-trimethylbenzene, as shown in Table 2. Compared with the sensitivity for pure n-decane vapor, the observed for all of the PPF-coated TSMR sensors, taking into account the oVset due to sensor drift.The changes in frequency frequency shifts of these sensors are reduced by the presorption of water vapor. This may be caused by the lipophobic shifts (strength of sensor response) induced by gas-sorption/ desorption had a profound correlation with the degree of character of the water-sorbed PE film, preventing the incorporation of n-decane molecules with their hydrophobic character.oVset induced by individual drift. Toluene vapor induced larger frequency shifts for the TSMR sensors coated with the In all cases, the PE (photo) film shows pronounced sensing ability coincident with its high sensitivity for pure petroleum PE (photo, 5 mm), PE (photo), and PE films, which are composed of lipophilic hydrocarbon networks.12 hydrocarbons. The sensing ability of the D-phenylalanine film for petroleum In practical situations, TSMR sensors are profoundly aVected by the concentration of water vapor, which is usually hydrocarbons was abruptly increased by the co-sorption with water vapor, compared with the negligible sensing ability for higher than that of the petroleum hydrocarbon to be detected. 140 J. Environ. Monit., 1999, 1, 135–142Fig. 8 Schematic diagram of set-up for measuring sorption capacities of TSMR sensors for binary mixtures of water and organic vapors. Fig. 9 Response curves of TSMR sensors for water vapor (RH: Fig. 11 Response curves of TSMR sensors for binary mixture of 8.7% 10.0%). RH water and 200 ppb n-decane.References 1 D. P. Saini, R. Leclerc, S. Coulter and H. Li, SPIE, 1995, 2504, 334. 2 K. Bodenho� fer, A. Hierlemann, G. Noetzel, U. Weimar and W. Go� pel, Anal. Chem., 1996, 68, 2210. 3 Techniques and Applications of Plasma Chemistry, ed. J. R. Hollahan and A. T. Bell, Wiley, New York, USA, 1974. 4 H. B. Olayan, H. S. Hamid and E. D. Owen, J. Macromol. Sci.- Rev.Macromol. Chem. Phys., 1996, C36, 671. 5 J. S. Miller and A. J. Epstein, Angew. Chem. Int. Ed. Engl., 1994, 33, 385. Fig. 10 Response curves of TSMR sensors for binary mixture of 6 C. A. Hunter and J. K. M. Sanders, J. Am. Chem. Soc., 1990, 10.0% RH water and 200 ppb 1,2,4-trimethylbenzene. 112, 5525. 7 J. C. Ma and D. A. Dougherty, Chem. Rev., 1997, 97, 1303. 8 M.-F. Fan, Z.Lin, J. E. McGrey and D. M. P. Mingos, J. Chem. these petroleum hydrocarbons, as shown in Table 2. The Soc., Perkin Trans. 2, 1996, 563. hydrophobic interaction of the water-sorbed film molecules 9 S. W. Benson, J. Am. Chem. Soc., 1993, 115, 6969. and hydrocarbons may play an important role in enhancing 10 I. Sugimoto, M. Nakamura and H. Kuwano, Sens. Actuators, B, 1996, 37, 163. the sensitivity of the sensor. J. Environ. Monit., 1999, 1, 135–142 14111 I. Sugimoto, M. Nakamura, H. Kuwano and R. Shimade, Thin 22 G. Z. Sauerbrey, Z. Phys., 1959, 155, 206. Solid Films, 1997, 310, 303. 23 I. L., Gebeluegi and F. Korte, in Proceedings of the Fifth 12 I. Sugimoto, Analyst, 1998, 123, 1849. International Conference on Indoor Air Quality and Climate, 13 I. Sugimoto and S. Miyake, J. Appl. Phys., 1988, 64, 2700. INDOOR AIR ’90, ISIA@, Toronto, Canada, 1990, vol. 2, 14 Y. Ishikawa, H. Kuwahara and T. Kunitake, J. Am. Chem. Soc., pp. 701–706. 1989, 111, 8530. 24 J. W. Grate and M. H. Abraham, Sens. Actuators, B, 1991, 3, 85. 15 I. Sugimoto, M. Nakamura and H. Kuwano, Anal. Chem., 1994, 25 M. H. Abraham, Pure Appl. Chem., 1993, 65, 2503. 66, 4316. 26 R. A. McGill, M. H. Abraham and J. W. Grate, CHEMTECH, 16 T. Kohonen, Self-Organizing Maps, Springer-Verlag, Berlin, 1994, 27. Germany, 1995. 27 M. H. Abraham, P. L. Grellier and R. A. McGill, J. Chem. Soc., 17 M. Nakamura, I. Sugimoto, H. Kuwano and R. Lemos, Sens. Perkin Trans. 2, 1987, 797. Actuators, B, 1994, 20, 231. 28 J. N. Israelachvili, Intermolecular and Surface Forces, Academic 18 M. Nakamura, I. Sugimoto and H. Kuwano, in Proceedings of the Press, London, UK, 1985. 1997 IEEE International Conference on System, Man, and 29 I. Langmuir, J. Am. Chem. Soc., 1916, 38, 2221. Cybernetics, SMC ’97, IEEE, Orlando, FL, USA, 1997, vol. 4, 30 M. Nakamura, I. Sugimoto and H. Kuwano, J. Intell. Mater. Syst. pp. 3036–3041. Struct., 1994, 5, 315. 19 J. M. H. Fortuin, Anal. Chim. Acta, 1956, 15, 521. 20 G. A. Lugg, Anal. Chem., 1968, 40, 1072. 21 WHO Regional OYce for Europe, Indoor Air Quality: Organic Pollutants, EURO Reports and Studies No. 111, Copenhagen, Paper 8/09298B Denmark, 1989. 142 J. Environ. Monit., 1999, 1, 135&ndash