1502
ANALYTICAL CHEMISTRY, VOL. 51, NO. 9, AUGUST 1979
2. Specific combinations and relative concentrations of nitrosatable species and coupling reagents. 3. Reaction of nitrite with ring substituents other than the amino group. 4. Reaction of nitrite with the coupling reagent. 5 . Formation of more than one pigment. 6. Oxidation of the diazonium ion intermediate. 7. Oxidation of the pigment. 8. Oxides of nitrogen in the air. 9. Reduction of the diazonium ion by residual reductants. 10. Formation of semistable nitroso-reductant intermediates. 11. Pre-reaction of NS and nitrite. These factors, in addition to pH and temperature, are important in some degree to all reagents studied and, it may be assumed, to any other compounds that have been or may be proposed for the purpose of determining nitrite. We know of no papers on nitrite determination where new or different reagents have been proposed in which more than one of these factors has been investigated. Since the operation of many of these factors results in significantly different total amounts of pigment formed, no new reagents or combinations thereof should be recommended without an assessment of the effects of these factors. This study also provides information that could be useful in improving the method of analysis. In a study of a number of sample preparation procedures ( 4 ) ,we found that the different procedures yielded different amounts of nitrite. I t is axiomatic that measured nitrite is a function of both sample preparation and colorimetric development, with interactions occurring between the two. The next step in improving the determination of nitrite by Griess reagents would be to study these interactions using the observations made in this study to determine those sample components affected by a given preparation procedure. For example, the difference in the effects of reductants on different colorimetric reagent combinations could be used to define the nature of residual reductants in cured meats and the effectiveness of sample preparation procedures in removing them. The foregoing factors are not the entirety of extraneous factors or reactions that are of concern in nitrite determi-
nation. Nitrite itself is highly reactive in acid solutions. In addition to the bimolecular reaction to form the nitrosating compound, Nz03,nitrous acid undergoes a termolecular reaction to form NO and NO Fo,05,20,94 = 1.7). The relative precision can be expected to degrade as the boron concentration approaches the detection limit (0.01 pg/mL) which imposes an absolute limiting precision of *0.003% B203. The plasma precision can therefore be stated as *0.8% RSD or *0.003% B203,whichever is greater. The exception is zirconium-aluminosilicates as stated. T h e accuracy of the plasma method for boron in alkali aluminosilicates (zirconium absent) is within the limits imposed by the precision of the comparisons made in Table IV-A. There is no overall bias in the average results for the 16 glasses considered. Conversely, alkali aluminosilicates containing zirconium (IV-B) produce an average plasma result which is
1.4% relative lower than the average wet result for these three samples. The 95% CL for a single comparison of the average of eight plasma determinations (1.2% RSD, t = 2.1) and one wet determination (70RSD = (0.02/16.1) (100) = 0.1%) is 0.9% relative ( = 2.1(1.22/8 + O.lz/l)l’z). For three such comparisons the limit is 0.5% relative. One of methods is therefore biased for zirconium aluminosilicates. Figure 4 shows that the effect of zirconium on plasma boron response is negligible below 100 pg/mL (equivalent to 27% ZrO, in the original sample). But synthetic studies of the effect of zirconium on the wet method have indeed indicated a sporadic and positive bias ( 4 ) . For high lead glasses in the range of 4-770 B203(Table V-A), the plasma results average 1.2% relative lower than the average of the wet results. Again taking the plasma precision as 0.8% RSD and the wet precision as 0.4% RSD (0.02 SD a t the 5.5% level), then a single comparison of the average of three plasma determinations with one wet determination has a 95% CL of f1.2% relative ( = 2.0(0.82/3 + 0.42/1)1/2). For nine such comparisons the limit is f0.4% relative. Thus, the 1.2% bias between the two methods is significant. Even if the zirconium-containing samples are omitted, a significant bias of 0.8% remains. Note that the three samples in Table V-B were not considered in the calculations to confine the B203region of interest. The first of these is in a range where the wet method produces large relative errors (more than 6%) and should be eliminated from consideration. The last two might be included in the calculations, but they would not significantly change the results. The explanation and further verification of the bias is beyond the scope of this work, but it is small enough that the agreement between the methods is still good and acceptable in most analytical situations. The comparison of wet and plasma results for fluoride glasses (Table VI) produces an average plasma result which is higher than the average wet result by 3.2% relative a t the 1.5% B203 level (Table VI-A) and a t the 15% B203 level (Table VI-B). If the plasma precision is again taken as 0.8%
1508
ANALYTICAL CHEMISTRY, VOL. 51, NO. 9, AUGUST 1979
relative and the wet precision as 1.3% relative (0.02 SD a t the 1.5% level), then the 95% CL for a single comparison of the average of three plasma determinations and one wet determination is 2.8% relative (= 2.0(0.8*/3 1.3'/1)'''). For seven such comparisons the limit is 1.1% relative. The two glasses in Table VI-B could be included in this calculation without changing the results. Thus the bias of 3.2% relative is significant. Its explanation probably lies in the known volatility of boron trifluoride which could be formed and released during the fusion treatment employed by the wet titrimetric method. This explanation is also consistent with the low analytical totd results (
