Vol.2 No.3 2009

Research paper : A marked improvement in the reliability of the measurement of trace moisture in gases (H. Abe)−213−Synthesiology - English edition Vol.2 No.3 (2009) was confirmed that flow rate control at a standard uncertainty of within 0.15 % was achievable[22]. Figure 10 shows the result of the comparison between the standard values of the trace moisture after introducing the critical-flow Venturi nozzle flow meters and the indications of the CRDS trace moisture analyzer. As in Fig. 8, the differences between the indications and standard values are shown as a relative value. The figure shows that the differences are within 6 % and are almost constant, in contrast to the result shown in Fig. 8 where the maximum difference reached 11 %. The remaining difference in Fig. 10 can be explained on the basis of the uncertainty due to the effect of the temperature on the absorption cross section and that due to the inaccurate value of the absorption cross section used in the CRDS trace moisture analyzer[22].3.3 Evaluation of uncertaintyThe standard uncertainty u(xw) of xw can be expressed as follows, assuming that there is no correlation between the physical quantities on the right side of Eq. (1): (3)u(A) on the right side represents the standard uncertainty of physical quantity A. Moreover, u(A) can be expressed as (4)where ci is the sensitivity coefficient and u(ai) represents the standard uncertainty of physical quantity ai. ci can be determined by theoretical consideration or by experiment. It is also important to identify what physical quantities should be included in Eq. (4) as u(ai). Minor uncertainties can be ignored. However, it often takes considerable time to determine which uncertainties can actually be ignored. There are also cases where an uncertainty component considered to be negligible was in fact not negligible. In one example, we observed variability in the evaporation rate when measurements were performed several times, but this variation could not be explained on the basis of the variabilities of the pressure and temperature in the generation chamber. After gathering long-term data, we found a correlation between the evaporation rate and the room temperature, and we performed an experiment where the temperature of the room was varied intentionally using an air conditioner. Figure 11 shows the temperatures of the room and generation chamber recorded during the experiment. Although the temperature of the chamber appeared to remain constant even if the room temperature was varied, the evaporation rate measured using the magnetic suspension balance depended on the room temperature, as shown in Fig. 12(a). In order to verify that this was not an effect of temperature on the indication of the magnetic suspension balance that biased the measurement value but the effect of temperature on the evaporation rate, we also examined the effect of temperature on the indication of the CRDS trace moisture analyzer shown in Fig. 12(b). The indication of the CRDS trace moisture analyzer clearly depended on the room temperature (the bias of the indication due to the temperature change was compensated), and we concluded that the temperature affected the evaporation rate. The sensitivity coefficients for the change in evaporation rate against the change in room temperature obtained from data shown in Figs. 12(a) and 12(b) were consistent. This phenomenon is probably attributable to unexpected heat transfer, for instance, through the dry gas introduced into the generation chamber. Furthermore, the temperature of the point monitored using a thermometer (controlled to be a set value) may differ from the temperature of the water in the diffusion cell (affected by the variability of room temperature). Further details of this temperature effect were discussed in Ref. [23]. Because it appeared that the temperature of the generation chamber was well controlled, we considered that the temperature of the water in the diffusion cell was also well controlled, and we did not initially recognize this uncertainty component. However, from the above experiment, we found that the uncertainty component due to the variability of room temperature should be included in u(N) as u(ai), and its Fig. 12 Room temperature dependency of evaporation rate.Reproduced with permission from Ref. [23]. Mass loss rate / (µg·h-1)xw / (nmol·mol-1)Temperature / ℃Temperature / ℃24262830242628309510011.012.0(b)(a)0.925 nmol / (mol℃)→0.132 µg / (h℃)0.142 µg / (h℃)13.0u(xw)FFF2=u(N)u(Nb)Nu(F)u2(xb)222+++u2(A)=c12u2(a1)+c22u2(a2)+…


page 40