BROMINE NUMBER SHOULD REPLACE FIA IN GASOLINE OLEFINS TESTING

Dec. 28, 1992
Axel J. Lubeck, Robert D. Cook Marathon Oil Co. Littleton, Colo . Fluorescent indicator adsorption (FIA) analysis, the ASTM test method proposed by the U.S. Environmental Protection Agency (EPA) for monitoring olefins in reformulated gasoline, is subject to significant bias and imprecisions.
Axel J. Lubeck, Robert D. Cook
Marathon Oil Co.
Littleton, Colo
.

Fluorescent indicator adsorption (FIA) analysis, the ASTM test method proposed by the U.S. Environmental Protection Agency (EPA) for monitoring olefins in reformulated gasoline, is subject to significant bias and imprecisions.

A more accurate, environmentally pertinent measure of olefin content in reformulated gasoline is bromine number, another ASTM method .2 Petroleum chemists should therefore work together with the EPA to select and optimize a bromine number procedure specifically designed for reformulated gasoline to replace FIA.

BACKGROUND

The EPA's proposed regulations covering reformulated gasoline require the measurement and reporting of various parameters which relate to potential ozone-forming emissions. 3 Among these parameters are olefins (vol %), as measured by FIA analysis (ASTM D1319).

While FIA analysis has been a mainstay of refinery laboratories for more than 30 years, its use for regulatory monitoring of reformulated gasoline is questionable for a number of reasons, including poor interlaboratory reproducibility, erroneous results (particularly at low olefin levels), and inappropriate units of measurement.

Marathon Oil Co. believes that a preferred technique for measuring olefins in reformulated gasoline is the determination of bromine number. This can be accomplished by using an existing standard method (ASTM D1159), or a suitable variation tailored to reformulated gasoline.

FIA ANALYSIS

The analysis of hydrocarbon types (saturates, olefins, and aromatics) by FlA was first published as an ASTM standard in 1954. Since that time, it has been successfully applied in refinery laboratories for process stream monitoring and in many locations for finished gasoline analysis.

The FIA method is based on separation of hydrocarbon components according to their adsorption affinities on activated silica get. A sample is introduced onto the top of a bed of silica gel and packed in a glass column of small, uniform bore, where it is adsorbed. Isopropyl alcohol is added to desorb the sample and force it down the column. Components with the least affinity for silica gel (saturates) move to the front of the traveling band of sample; those with progressively higher affinities (olefins and aromatics) lag behind correspondingly.

Fluorescent dyes, visible under ultraviolet (UV) light, mark the zonal boundaries of the saturates, olefins, and aromatics. The amount of each hydrocarbon type, in vol %, is calculated from the length of each zone in the column. The use of FIA has been favored because it is inexpensive to implement, and it determines the relative amounts of all three hydrocarbon types simultaneously. Based on its long history of use, it is not difficult to understand why the EPA has chosen to specify FIA analysis for monitoring olefins in reformulated gasoline.

FIA test results, however, are not precise; interlaboratory reproducibility is notoriously poor. Table 1 shows this poor precision (within-laboratory repeatability and between-laboratory reproducibility), as documented in the published ASTM standard.

Repeatability and reproducibility, as given in Tables 1 and 2, are calculated by statistical methods, as specified by ASTM.' The two parameters are approximately equal to three times the standard deviation of a given set of values.

If, as in Table 1, repeatability for a 1 vol % olefins sample is 40%, this means that the difference between two successive determinations of the same sample in the same laboratory will be less than or equal to 0.4 vol % (40% of 1 vol %), 95% of the time.

One of the sources of imprecision in the FIA method is the difficulty in determining the exact position of the boundaries between hydrocarbon types. The dyes marking the boundaries are observed as diffuse bands of fluorescence when exposed to UV "black" light.

Because the fluorescing bands have a measurable length, which can be significant relative to the zone length of a particular hydrocarbon type, differences in "reading" the position of the boundary add an undesirable subjectivity to the determination. Because the FIA procedure requires the presence of a human analyst, it does not lend itself to automation.

VOLATILITY ERROR

The presence in gasolines of significant amounts of volatile components (C4S and C5s) reportedly leads to errors in the direction of low olefin values. A note warning of the possibility of error from this source appears in the ASTM test procedure.

The error presumably arises from an anomalous swelling (lengthening) of the saturate zone because of the volatile saturates present. Erroneously low olefin and aromatic contents result as a consequence of the normalization procedure used to calculate the amount of each hydrocarbon type (i.e., dividing the length of each zone by the sum of the lengths of the three zones and multiplying by 100%).

While no data have been published, reports from reliable laboratories indicate that errors of 1-2 vol % (absolute) have been encountered. These errors were attributed to the presence of volatile components.

To circumvent this problem, ASTM Test Method D1319 states that samples containing more than 5% C4S or more than 10% C4S+C5S can be depentanized prior to analysis. Conflicting wording within the ASTM method creates ambiguity as to whether this corrective action is, in fact, mandated.

A recent informal poll of laboratories performing FIA analyses indicated that few of them depentanized gasoline samples on a routine basis. If the report of error from volatile components is correct, FIA analysis of nondepentanized gasolines is unacceptable.

For rigorous testing, all reformulated gasolines would need to be depentanized by distillation prior to FIA analysis. The overhead from the depentanization would have to be adequately collected and analyzed by gas chromatography, and the results added back to the FIA results on a weighted basis.

The poor reproducibility of the standard FIA method can only be further degraded by the addition of a distillation step and a second analysis. Such a complex multistep analysis is therefore equally unacceptable for routine regulatory monitoring.

NONLINEARITY

Marathon has observed additional errors in the FIA analysis not attributable to the presence of volatile components. In an effort to establish the internal consistency (linearity) of FIA olefin determinations, the authors analyzed a number of dilutions of a fluid catalytic cracking unit (FCCU) naphtha sample.

Eight simulated gasolines were prepared by diluting an FCCU naphtha with varying amounts of a mixture of reformate and alkylate. (This diluent mixture was purposely chosen to have the same density as the FCCU naphtha, allowing identical weight and volume dilutions.)

The eight "gasolines" were then analyzed by FIA in duplicate. The actual results were compared with expected olefin contents calculated (as weighted averages) from the FIA olefin values of the undiluted FCCU naphtha and the reformate/alkylate diluent.

A linear system would show little or no difference between the experimentally determined olefin contents and the expected values calculated from the weighted averages of the olefin contents of the undiluted FCCU naphtha and the diluent. As seen in Fig. 1, however, the experimental data show that the determined values are consistently lower than expected.

This bias is not attributable to the presence of volatile components in the gasoline because the expected olefin contents were based on the olefin content of the FCCU naphtha, which itself was determined by FIA. If the amount of volatile components were the same in the undiluted FCCU naphtha and each of the dilutions, the calculated expected values would account for any error due to volatile components.

The vapor pressure (and C4 + C5 content) of the reformate/alkylate mixture used, however, was less than that of the straight FCCU naphtha. Thus, the vapor pressure (and C4 + C5 content) of each of the dilutions was also less than the undiluted FCCU naphtha.

Assuming that the error associated with high C4 + C5 content (anomalous lengthening of the saturate band) would be reduced as the C4 + C5 content is reduced, the actual FIA results should actually show a slight positive bias. Instead, a significant negative bias was observed. It is evident that another error of greater magnitude is overshadowing the effect of the volatile components. While the cause of this additional error has not been identified, its presence is another strong argument against the use of FIA analysis. A better method is needed.

BROMINE NUMBER

The conventional approach to the chemical determination of olefins in gasoline is based on the addition of bromine to the double bond. With this approach, a sample of gasoline is dissolved in a specified solvent system and titrated with standard bromide-bromate solution. The result is expressed as the bromine number, defined as the number of grams of bromine that will react with 100 g of sample.

A number of procedures for determining bromine number have been published, differing primarily in three areas-temperature of the titration, use of a catalyst to increase reaction rates, and method for detecting the end point. Perhaps, the most widely used is ASTM D1159, first published in 1951.

The precision statement for the ASTM bromine number method (Table 2) shows the advantage of bromine number over FIA (Table 1). The importance of a precise method for regulatory analysis cannot be overemphasized. If test precision is poor, increased safety tolerances must be allowed to guarantee conformance to a specification or target value. This can result in an economic penalty (e.g., increased processing costs or reduced product volumes) for the manufacturer.

The bromine number determination is simple to perform and less time-consuming than FIA analysis. With modern instrumentation, bromine number titrations are automated and do not rely on the subjective judgment of the analyst. Because of its better precision and ease of use, Marathon favors the bromine number procedure.

LINEARITY

In the same manner as that for the FIA analysis, the authors examined the internal consistency (linearity) of bromine number results.

The same series of diluted FCCU naphtha samples was used, determining bromine numbers on each, in duplicate. The actual results were then compared with expected bromine numbers calculated (as weighted averages) from the bromine numbers of the undiluted FCCU naphtha and the reformate/alkylate diluent.

In contrast to the FIA results, the actual bromine numbers tracked the expected values almost perfectly (Fig. 2). The obvious linearity of the bromine number determinations lends credibility to the test method and its use for regulatory monitoring.

OXYGENATE EFFECTS

One of the defining parameters of reformulated gasoline is the presence of oxygenates. If Bromine number determination is to be used in reformulated gasoline monitoring, it must not be affected by the presence of these compounds.

It was not expected that bromine would react with aliphatic alcohols or ethers under conditions of the test procedure. However, to prove this, five oxygenates-methanol, ethanol, methyl tertiary butyl ether (MTBE), ethyl tertiary butyl ether (ETBE), and tertiary amyl methyl ether (TAME)-were each analyzed neat and in a gasoline blend prepared by mixing FCCU naphtha, reformate, and alkylate.

A summary of the results of this study is given in Table 3. For each of the oxygenate blends, the actual determined bromine number is in excellent agreement with the expected value (the weighted average bromine number of the neat oxygenate and the gasoline mixture). The greatest difference from expected values is 0.7% relative, which is well within the repeatability of the method. These typical oxygenates, at levels normally used in reformulated gasoline, do not interfere with the measurement of olefin content via bromine number determination.

BP NO. VS. FIA

A replacement analytical method is not readily accepted without challenge. It is desirable that the results provided by the proposed replacement method correlate with those obtained using the prevailing procedure. This is a reasonable expectation if both methods have comparable precision and accuracy.

In the case of FIA and bromine number, the authors have presented evidence that the methods do not have comparable precision and accuracy-there are errors associated with the FIA results.

The evidence suggests that one should expect a reasonably good correlation of values determined by the two methods, despite these errors, because the absolute magnitudes of the errors are small (less than the reproducibility of the FIA results). The evidence also suggests, however, that there should be a discernable negative bias in the FIA results for olefins.

To evaluate the above predictions, bromine numbers and olefin contents were determined by FIA analysis on 18 finished gasolines and six FCCU naphthas from five different refineries. For comparison on an equivalent basis, the bromine numbers were converted to a vol % olefin equivalent by dividing by the constant 2.2.

This simple conversion factor was derived using the calculations provided in the appendix to the bromine number method .2 The conversion factor applies to olefins derived from a typical full-range FCCU naphtha, which is the predominant source of olefins in most gasolines.

The calculated vol % olefin equivalents showed a good correlation (r 2 0.98) with vol % olefins determined directly by FIA (Fig. 3). Absolute differences between the vol % olefins calculated from the bromine number and those directly determined by FIA were, in all but one case, less than the between-laboratories reproducibility of the FIA procedure itself.

As predicted, Fig. 3 also shows that most of the FIA results, particularly at low olefin levels, lie below the solid line representing perfect correlation (assuming a 2.2 conversion factor). This is an example (with real, commercial gasolines) of the bias demonstrated in the FIA linearity study.

The authors also attempted to determine empirically a conversion factor from these same data. Ratios of bromine number to FIA vol % olefins (By no./FIA) were calculated for each of the 18 gasolines and six FCCU naphthas.

The variability in these ratios (Fig. 4) was great and led to the initial conclusion that there was no generally applicable constant to convert bromine number to vol % olefins. On closer examination, however, the ratios were seen to have an exaggerating effect. In other words, ratios are an expression of relative differences.

The variability in the calculated ratios was in large part caused by the transformation of small absolute variations into large relative variations. The authors also noted that there was both a systematic and a random component to the variability.

The systematic component, as shown by the sloped linear regression line, indicates a definite trend toward higher ratios as olefin content decreases. Both the magnitude and direction of this trend is consistent with the negative bias in FIA olefin results that has been demonstrated.

The random variability of the ratios, represented by the general scatter of the data, is likewise well within predictions based on the relative imprecision of the test methods. When FIA bias and the random variability are taken into consideration, Fig. 3 indicates that 2.2 is a reasonably good empirical estimate of Br no./FIA.

Inherent in the interconversion of FIA vol % olefins and bromine number are assumptions regarding the average molecular weight and average density of the olefins.

Although the predominant source of olefins in gasoline is FCCU naphtha, other streams (e.g., mixed butenes or propylene trimer) are at times blended into gasoline. The olefin molecular weights and densities for such streams may be significantly different from FCCU naphtha olefins.

The 2.2 factor thus would not be expected to be applicable to gasolines blended with substantial quantities of these streams (such gasolines are few in number).

UNITS OF MEASUREMENT

Any argument as to whether the 2.2 factor is applicable to all gasolines, however, should be a moot point. The explanation is as follows:

The purpose of measuring olefins in reformulated gasoline is to determine the potential contribution to atmospheric pollution-forming reactions. The accepted unit of measure for reactive species is the mole. Bromine number, expressed as g Br/100 g sample, is directly proportional to the number of moles of olefins per unit weight of sample, and is therefore an appropriate measure.

In contrast, FIA expresses olefin content in volume percent, allowing the possibility of misrepresenting the molar concentration.

As an example, two gasolines with identical bromine numbers could have as much as a 25% relative difference in vol % olefins, if one were blended with mixed butenes and the other with FCCU naphtha. Such inequity could be avoided if measurement units proportional to molar concentration were specified.

REFERENCES

  1. ASTM Standard Test Method D]319, Hydrocarbon Types in Liquid Petroleum Products by Fluorescent Indicator Adsorption (FIA).

  2. ASTM Standard Test Method D1159, Bromine Number of Petroleum Distillates and Commercial Aliphatic Olefins by Electrometric Titration.

  3. Federal Register, 57FR 1341613495, Apr. 16, 1992.

  4. Manual on Determining Precision Data for ASTM Methods on Petroleum Products and Lubricants, RR: D02-1007, Annual Book of ASTM Standards, Vol. 03.03.

Copyright 1992 Oil & Gas Journal. All Rights Reserved.