Journal of Occupational Health
Online ISSN : 1348-9585
Print ISSN : 1341-9145
ISSN-L : 1341-9145
Originals
Comparison of sensor characteristics of three real-time monitors for organic vapors
Hajime Hori Sumiyo IshimatsuYukiko FuetaMitsuo HinoueToru Ishidao
Author information
JOURNAL FREE ACCESS FULL-TEXT HTML

2015 Volume 57 Issue 1 Pages 13-19

Details
Abstract

Objective: Sensor characteristics and performance of three real-time monitors for volatile organic compounds (VOC monitor) equipped with a photo ionization detector (PID), a sensor using the interference enhanced reflection (IER) method and a semiconductor gas sensor were investigated for 52 organic solvent vapors designated as class 1 and class 2 of organic solvents by the Ordinance of Organic Solvent Poisoning Prevention in Japan. Methods: Test vapors were prepared by injecting each liquid solvent into a 50 l Tedlar® bag and perfectly vaporizing it. The vapor concentration was from one-tenth to twice the administrative control level for all solvents. The vapor concentration was measured with the monitors and a gas chromatograph equipped with a flame ionization detector simultaneously, and the values were compared. Results: The monitor with the PID sensor could measure many organic vapors, but it could not detect some vapors with high ionization potential. The IER sensor could also detect many vapors, but a linear response was not obtained for some vapors. A semiconductor sensor could detect methanol that could not be detected by PID and IER sensors. Conclusions: Working environment measurement of organic vapors by real-time monitors may be possible, but sensor characteristics and their limitations should be known.

(J Occup Health 2015; 57: 13–19)

Introduction

Real-time monitors, that is, direct reading instruments, for organic vapors equipped with various chemical sensors have been developed1). Although real-time monitors have not been approved for use in working environment measurement for organic solvents in Japan, they are helpful for understanding work environment conditions, especially when the vapor concentrations are fluctuating. Because most real-time monitors have data loggers, the temporal change in the vapor concentration can be analyzed after measurement. This shows that the real-time monitors are suitable for not only working environment measurement for conducting technical measures of work environments but also for measuring the personal exposure level for performance of risk assessments2).

We have been studying the sensor characteristics of several real-time monitors for organic vapors or volatile organic compounds36). First, we tested a monitor with a photo ionization detector (PID)4) for 52 organic solvents. Although this monitor had high sensitivity and a linear response was obtained for many solvents, it showed low or no response for some solvents with a high ionization potential. Next, we also tested a monitor with the interference enhanced reflection (IER) method6) under the same experimental conditions. The IER sensor could also detect many organic vapors. A linear response were obtained, but the regression lines between the concentrations measured by the monitors and those of the actual ones did not pass through the origin for 14 of the 52 solvents.

In these studies, we selected almost the same concentration range for the test vapors (around 50–300 ppm) for all organic solvents in order to obtain fundamental sensor characteristics of the monitors. However, the occupational exposure limits (OELs) or the administrative control levels (ACLs) of organic solvents are different from solvent to solvent. In particular, the ACLs have been lowered in association with revision of the OELs, and currently, solvents with ACLs of 50 ppm or less comprise about 70% of the designated organic solvents7). In addition, because the Japanese working environment evaluation standards requires a quantification limit at least one-tenth of the ACL, the sensor characteristics up to 1/10 of the ACLs should be important. In particular, for solvents in which the regression lines did not pass through the origin when measured with the IER sensor, the concentration dependence might be different between low and high concentration ranges. Therefore, we should discuss whether the sensors are practical or not under such concentration range.

In this study, we investigated the sensor characteristics at the concentration range between one-tenth to two times of the ACLs. A real-time monitor equipped with a semiconductor gas sensor was used in addition to the monitors with PID and IER sensors, and the characteristics and limitations of the three monitors for 52 organic vapors were discussed.

Materials and Methods

Organic solvents

We used all of the organic solvents designated as class 1 and class 2 organic solvents and required to be subjected to are obliged the working environment measurement by the Ordinance of Organic Solvent Poisoning Prevention in Japan. The ordinance designates 47 solvents, but we tested 52 solvents including isomers. All solvents used were of reagent grade (Wako Pure Chemical Industries, Osaka, Japan and Kanto Chemical, Tokyo, Japan).

Real-time monitors

Three different real-time monitors were used, that is, a VOC monitor equipped with a semiconductor sensor (Handy TVOC Monitor FTVR-02, Figaro Engineering, Osaka, Japan), a monitor equipped with the interference enhanced reflection (IER) technology (the Handy VOC Sensor VOC-121H, O.S.P. Inc, Sayama, Japan) and a portable VOC monitor equipped with a photo ionization detector (PID, ionization potential: 10.6 eV) (MiniRAE 3000, RAE Systems, San Jose, U.S.A.). Specifications of these monitors are shown in Table 1. In all measurements, zero calibration was carried out using filtrated air passed through a charcoal bed in advance to test the monitors.

Table 1. Specifications of real-time monitors used in this work
MiniRAE 3000 (RAE Systems) VOC-121H (O.S.P.) FTVR-02 (Figaro Eng.)
Sensor PID IER Semiconductor
Range 0.1–15,000 ppm 1–100 ppm or 25–2,500 ppm 1–3,000 ppm
Resolution 0.1 ppm 0.1 ppm 1 ppm
Accuracy ± 3% ± 3% of full span ± 1%
Sampling flow rate 450–550 ml/min 200–300 ml/min 800 ml/min
Data logging time 16 h 6–8 h 8 h

Methods

A schematic diagram of the experimental system is shown in Fig. 1. Fifty liters of clean air was introduced into a plastic bag (Tedlar® bag, volume 50 l) by using a dry gas test meter (Model DC-2A, Shinagawa Corporation, Tokyo, Japan). Test vapor was prepared by spiking liquid solvent into the bag and vaporizing it perfectly. The vapor concentration was one-tenth, or one or two times the ACL of each solvent.

Fig. 1.

Schematic diagram of the experimental system.

(1) 50 l Tedlar bag, (2) sampling cell, (3) MiniRAE 3000, (4) VOC-121H, (5) FTVR-02, (6) FID-gas chromatograph.

After being left for about 20 to 30 minutes, air containing organic vapor in the bag was sampled with a gastight syringe, and the vapor concentration was determined with a gas chromatograph (GC) (GC-17A, Shimadzu, Kyoto, Japan) equipped with a flame ionization detector. Then, a sampling cell made of Teflon® was connected to the Tedlar® bag, and the sensors of the monitors were inserted into the cell. Air in the bag was continuously introduced into the sampling cell, and the vapor concentration was monitored and recorded when the concentration became stable for 5 minutes or more.

Results

Figure 2 shows typical examples of the relationship between the vapor concentrations displayed by the monitors and the values measured by the GC. Regression lines and the coefficients of determinations are also included in the figure. A linear relationship was obtained between the values measured by all the monitors and those measured by the GC. If the slope of the regression line is unity, the values measured by the monitor and those measured by the GC were equal. In such a case, the monitor can be used without any corrections. In Fig. 2, the slope of the regression line of toluene was 2.54 for the MiniRAE 3000, 0.91 for the VOC-121H and 0.12 for the FTVR-02, respectively. This result indicates that, for the MiniRAE 3000, the values displayed on the monitor were about 2.5 times greater than the actual values, but for the FTVR-02, the values were almost one-tenth lower.

Fig. 2.

Comparison of toluene concentrations measured with the three monitors and those measured with the GC.

Figure 3 and Fig. 4 show results for 1,4-dioxane and chloroform, respectively. In Fig. 3, although a linear relationship was obtained between the monitors and the GC, the regression line for the VOC-121H did not pass through the origin. For the MiniRAE 3000 and FTVR-02, the regression lines passed through the origin. The slope of the regression line for the MiniRAE 3000 was close to unity, indicating that the values measured by the monitor were almost the same as those measured by the GC. On the other hand, the slope of the regression line for the FTVR-02 was almost one-tenth lower than that of the GC. Figure 4 shows the case of chloroform. Only the VOC-121H responded to the vapor, but the regression lines did not pass through the origin. The MiniRAE 3000 and FTVR-02 did not respond in the concentration range used in this study.

Fig. 3.

Comparison of 1,4-dioxane concentrations measured with the three monitors and those measured with the GC.

Fig. 4.

Comparison of chloroform concentration measured with three monitors and those with the GC.

As shown in Fig. 2 to 4, the sensor characteristics and sensitivities were largely different by sensors and vapors. Table 2 summarizes the slopes and intercepts (IC) of the regression lines between the monitors and the GC for 52 vapors. If the regression line passes through the origin and the slope of the regression line is larger than unity, the value measured by the monitor is larger than that measured by the GC. For the MiniRAE 3000, the lines for 47 of the 52 vapors passed through the origin, and the slopes for 22 vapors were greater than unity. But chloroform, carbon tetrachloride, 1,2-dichloroethane, 1,1,2,2-tetrachloroethane and methanol were not detected by this monitor. The VOC-121 could detect all the vapors except for methanol, and the slope was greater than unity for 20 of the 52 vapors, but it did not pass through the origin for 18 vapors.

Table 2. Slopes and intercepts (IC) of regression lines between the monitors and GC
Solvents Administrative control level (ppm) Regression line with GC
MiniRAE 3000 VOC-121H FTVR-02
Slope Slope IC Slope IC
Class 1
Chloroform   3 NR 0.32 2.58 * NR
Carbon tetrachloride   5 NR 0.06 1.01 * NR
1,2-Dichloroethane  10 NR 0.12 1.73 * NR
cis-1,2-Dichloroethylene 150 1.03 0.12 0.01 *
trans-1,2-Dichloroethylene 150 2.73 0.07 4.47 0.01 *
1,1,2,2-Tetrachloroethane   1 NR 3.87 1.98 * NR
Trichloroethylene  10 2.08 0.27 2.98 * NR
Carbon disulfide   1 0.95 * 0.32 * NR
Class 2
Acetone 500 0.92 0.04 0.08 *
Isobutyl alcohol  50 0.22 0.26 2.42 0.09 *
Isopropyl alcohol 200 0.15 0.09 0.07 *
Isopentyl alcohol 100 0.31 1.06 0.09 *
Ethyl ether 400 0.95 0.04 0.08 *
Ethylene glycol monoethyl ether   5 0.47 * 0.84 2.93 * NR
Ethylene glycol monoethyl ether acetate   5 1.16 3.61 NR
Ethylene glycol monobutyl ether  25 0.59 3.91 0.06 0.74 *
Ethylene glycol monomethyl ether 0.1 0.48 * 0.39 * NR
o-Dichlorobenzene  25 2.13 7.03 NR
o-Xylene  50 2.72 3.88 0.11 *
m-Xylene  50 2.23 1.46 0.09 *
p-Xylene  50 1.71 1.28 0.07 *
o-Cresol   5 1.09 5.27 NR
m-Cresol   5 1.03 5.10 NR
p-Cresol   5 1.03 3.90 NR
Chlorobenzene  10 2.89 1.51 1.79 * NR
Isobutyl acetate 150 0.36 0.65 0.12 *
Isopropyl acetate 100 0.38 0.16 5.01 0.06 *
Isopentyl acetate  50 0.47 1.90 0.15 *
Ethyl acetate 200 0.27 0.16 0.12 *
Butyl acetate 150 0.40 1.15 0.16 *
Propyl acetate 200 0.27 0.38 0.12 *
Pentyl acetate  50 0.50 2.47 0.14 *
Methyl acetate 200 0.16 0.05 3.04 0.05 3.35 *
Cyclohexanol  25 0.93 0.84 23.9 0.11 -0.84 *
Cyclohexanone  20 1.13 1.21 6.14 0.15 *
1,4-Dioxane  10 1.05 0.38 2.41 * 0.11 *
Dichloromethane  50 0.02 * 0.03 4.20 * NR
N,N-Dimethylformamide  10 2.16 1.38 0.32 -0.96 *
Styrene  20 3.52 3.22 0.15 *
Tetrachloroethylene  50 2.11 0.80 NR
Tetrahydrofurane  50 0.52 0.08 0.11 *
1,1,1-Trichloroethane 200 0.03 0.11 0.01 *
Toluene  20 2.54 0.91 0.12 *
n-Hexane  40 0.21 0.04 3.29 * 0.02 *
1-Butanol  25 0.26 0.77 0.26 -1.28 *
2-Butanol 100 0.23 0.12 6.09 0.07 *
Methanol 200 NR NR 0.11 *
Methyl isobutyl ketone  20 1.16 0.68 0.19 *
Methyl ethyl ketone 200 1.20 0.12 0.25
Methyl cyclohexanol  50 0.99 3.37 0.12 *
Methyl cyclohexanone  50 1.33 2.98 NR
Methyl butyl ketone   5 1.24 0.39 3.14 * NR
*  Difficult to determine 1/10 of the administrative control level.

The values measured by the FTVR-02 were generally low compared with those measured by the MiniRAE 3000 and VOC-121. A linear response was obtained for 31 of the 52 vapors, but the slope was generally small. This monitor did not respond to 18 vapors, but methanol, which could not be detected by the MiniRAE 3000 and VOC-121H, was detected, and a linear response was obtained.

Discussion

We tested the performance of three real-time monitors equipped with a PID sensor (MiniRAE 3000), a sensor using IER technology (VOC-121H) or a semiconductor sensor (FTVR-02) for 52 organic solvent vapors.

The MiniRAE 3000 responded to many organic vapors with a linear response, but it did not respond to five vapors, that is, chloroform, carbon tetrachloride, 1,2-dichloroethane, 1,1,2,2-tetrachloroethane and methanol. The reason for this must be that the ionization potentials of these vapors are larger than that of the built-in PID lamp (10.6 eV). Because these vapors could not be ionized by the PID lamp due to its low ionization power, these vapors could not be ionized, and consequently, they could not be detected by this monitor. The relationship between the ionization potential of organic vapors and the slope of the regression lines for the MiniRAE 3000 is shown in Fig. 5. In general, the slope of the regression lines decreased with increasing ionization potential. This suggests that the relative sensitivity of the PID sensor for vapors can be roughly estimated by knowing the ionization potential. Theoretically, if the ionization potential of a vapor is greater than 10.6 eV, the slope of the regression line must be zero, that is, the sensor cannot determine the vapor concentration. However, dichloromethane had a small slope (0.02), although its ionization potential was 11.32 eV, which was greater than the ionization potential of the PID (10.6 eV) (Table 2). The reason for this is unclear, but the effects of stabilized regents or unpurified contaminants may be considerable. We should confirm this by using a more purified solvent.

Fig. 5.

Relationship between ionization potential of organic vapors and slope of regression line.

The VOC-121H could detect all the vapors except for methanol. For 18 vapors, however, the regression lines did not pass through the origin. The reason for this might be related to the sensor characteristics. The IER sensor consists of a thin polymer film. When the polymer film adsorbs organic vapors, the thickness of the film changes depending on the adsorbed amount of adsorbed vapors. The difference in the film thickness before and after adsorption is detected by measuring the intensity of reflection light that is obliquely emitted from a laser diode (LD) or light emitting diode (LED) to the polymer film, and the concentration is determined from this difference. The enhancement of the polymer membrane may depend on the vapors, and because the enhancements for the above 18 vapors may be small, the polymer membrane would have been almost completely saturated with vapors at low concentration ranges; therefore, the film thickness would not swell in proportion to the vapor concentration in the concentration range used in this study8).

As shown in Table 2, the values measured by FTVR-02 were generally low. In principle, semiconductor sensors respond to many organic vapors nonselectively. However, the slope of the regression line for toluene was only 0.12, although this monitor was calibrated with toluene. This suggests that the sensor of this monitor had deteriorated. Many types of semiconductor sensors are available, and in general, semiconductor sensors are so sensitive to alcohols that they are widely used as alcohol sensors. Therefore, it is reasonable that methanol, which could not be detected by the other two monitors, can be detected by this monitor with a linear response. In addition to semiconductor sensors, we also previously confirmed that the PID sensor also exhibited temporary deterioration when we used another monitor4). Therefore, when we use such real-time monitors for determination of organic vapors, calibration should be conducted in advance of measurement.

The work environment evaluation standards describe that one-tenth of the administrative control level can be used as the minimum measured value9, 10). Therefore, when we use real-time monitors for working environment measurement, the following condition should be satisfied to measure one-tenth of the administrative control level with a precision of ± 20%, which is the average allowable range for gas detector tubes (± 15% when the stain length is 1/3 or more of full scale and ± 25% when the stain length is 1/3 or less)11).

  

where, R is the minimum resolution of the monitor (ppm), S is the slope of the regression line (-), and E is the administrative control level (ppm).

The left side of Eq. (1) indicates the actual minimum resolution value (ppm), and the right side indicates 20% of 1/10 of the administrative control level.

For example, the resolution of the MiniRAE 3000 is 0.1 ppm (Table 1). Therefore, when the slope of the regression line is unity, the minimum resolution of the monitor is also 0.1 ppm. However, when the slope is 0.1, the minimum resolution should be 1 ppm even when the displayed resolution is 0.1 ppm because the displayed value is one-tenth of the actual value. Eq. (1) shows that, in such a case, the concentrations of vapors with an administrative control level of 5 ppm need to be determined with a precision of ± 20%. If the slope of the regression line does not satisfy Eq. (1), the monitor cannot be used for the working environment measurement because it cannot determine 1/10 of the administrative control level. Considering Eq. (1), the MiniRAE 3000 responded to 47 of the 52 vapors, but the resolution was not enough for 4 vapors, that is, carbon disulfide, ethylene glycol monoethyl ether, ethylene glycol monomethyl ether and dichloromethane.

The VOC-121H responded to all vapors except for methanol. However, it could not measure 1/10 of the administrative control level for 12 vapors. The FTVR-02 responded to 34 vapors, but 1/10 of the administrative control level could only be determined for only methyl ethyl ketone with a resolution of 1 ppm. However, the validity of this monitor should be reconfirmed using a new sensor because the sensitivity of the sensor was much lower than the original one, as described above. Considering that the slope of the regression line for toluene should be 1.0, the slope in Table 2 should be multiplied by 8.3 (=1/0.12). If this is done, 1/10 of the administrative control level may be able to be determined for many vapors with this monitor.

The real-time monitors used in this study cannot detect individual components of multicomponent organic vapors separately. This is a limitation of the monitors for practical use in workplaces using mixed solvents. However, we previously confirmed that a monitor with a PID sensor showed an additive response for two-component organic vapors by correcting sensitivity of the sensor3). Therefore, evaluation of work environments using multicomponent organic solvents may be possible if the kinds of vapors and their compositions are known because the sensitivity of the monitors for each vapor can be determined from Table 2. This should be confirmed in future investigations.

Acknowledgments: The authors thank Ms. Ayaka Nakayama for technical help. This work was partly supported by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (JSPS).

References
 
2015 by the Japan Society for Occupational Health
feedback
Top