Distribution of Refractive Index Values in Sheet Glasses, by Koons and Buscaglia (Forensic Science Communications, January 2001)
January 2001 - Volume 3 - Number 1
Research and Technology
Distribution of Refractive Index Values in Sheet Glasses
Robert D. Koons
Forensic Science Research Unit
Federal Bureau of Investigation
After determining that a fragment of glass is indistinguishable from a window at a crime scene, the glass examiner is then faced with the task of assessing the significance of this result and providing that significance measure to the interested parties. There are two approaches that can be used for this. Some glass examiners will determine that two samples are indistinguishable based on refractive index (RI) and then use a database of RI values to obtain the frequency of occurrence of glass having the particular RI in question. Other glass examiners may use the likelihood ratio approach in which they use databases of RI values to determine the relative probabilities of competing hypotheses as to the origin of a questioned fragment of glass. In either approach, it is essential that the glass examiner have access to an accurate database containing RI values determined from appropriate, representative samples.
It has been noted that the distribution of RI values for sheet glass, beginning roughly in the 1960s, became narrower than that observed previously (Almirall 1996; Buscaglia 1994; Curran et al. 1997; Curran et al. 2000; Koons et al. 1991; Stoecklein 1996). Most of these authors used this observation to support their contention that more discriminating methods of comparison, such as elemental composition, should be used in glass examination. They based their statements on comparisons of recently acquired databases with older databases, usually those compiled by the FBI Laboratory in Washington, DC (Miller 1982), and the Forensic Science Service in the United Kingdom (Lambert and Evett 1984). It cannot be determined from these reports whether the purported changes in RI distributions reflect changes in glass manufacture or are an artifact of the limited source distribution in small geographical areas of individual studies.
Buscaglia and Kubic (1991) took a different approach, breaking down FBI glass case data into five-year intervals, and found that there is evidence of a narrowing of RI distributions about a central value. Because the sources and glass types in their sample population did not change drastically during the five-year periods, the observed narrowing of RI distributions was explained as a change in glass manufacturing rather than an artifact of changing sample selection. Unfortunately, their study was never widely disseminated.
Prior to the early 1960s, drawing and polishing processes were used to manufacture most sheet glass. Since the 1960s, the flat glass industry has converted almost entirely to the float process. As a result, recent glass collections contain a higher percentage of float glass than older collections. Curran and colleagues (2000) stated that an observed narrowing of the range of RI in modern sheet glass results from improved quality control in the manufacture of float glass. This is brought about by computerized delivery of raw materials and a standardization of methods and formulations among manufacturers throughout the world (Curran et al. 2000). Although these statements are possible explanations for observed changes in RI distributions, the authors provide no supporting data. Further, any narrowing of refractive indices of glass from a single manufacturing facility is offset somewhat in the overall distribution by a globalization of the glass products market.
The FBI Laboratory has maintained a database of refractive index values on flat glass control samples received as evidence in casework since the 1960s. In casework, examiners in laboratories throughout the United States use this database to determine, at least in approximate terms, the frequency of occurrence of glass having a given RI range. To assist in determining whether this frequency value can be applied to current glass evidence, an evaluation of the FBI Laboratory data has been made to assess whether the distribution of refractive indices has, indeed, changed appreciably since float glass products began to dominate evidentiary samples. This study is not designed to determine whether the database is appropriate for reliable frequency of occurrence calculations, such as those required for a likelihood ratio approach to the evaluation of evidence. A further study to address that question is still needed and would require collecting additional data on hypothesis-specific glass populations.
The FBI Laboratory glass database evaluated in this study consists of RI measurements on 2,337 samples of glass. Only those RI measurements made at the sodium D-line wavelength (nD) were considered in this study. The samples included in this study are flat glass samples, including architectural, automotive, mirror, and display case glasses submitted as exemplars in case evidence that were analyzed during the period from 1964 through 1997. The number of samples analyzed each year varies according to the number of specimens received in casework.
All data were obtained using the Emmons Double Variation Method and were normalized using a single in-house glass RI standard, which was analyzed along with every glass case over the 34-year period. The data through 1982 were discussed previously in a widely read textbook (Miller 1982). As a result, it is well known among glass examiners and has been introduced into the U.S. legal system many times. Since Miller's presentation of the database, it has approximately doubled in size.
The distribution of RI values for all flat glass from the period 1964 through 1997 is shown in histogram form in Figure 1. To determine whether trends with time exist in these data, the database was divided into glass cases received before 1980 and those received in 1980 and later. The histograms for these two time periods are shown in Figures 2 and 3. These two figures represent similarly sized glass populations: 1,275 samples before 1980 and 1,062 samples in and after 1980.
Several conclusions can be drawn from comparing the RI distributions in Figures 2 and 3. The range of measured RI values does not change appreciably, being approximately 1.512 to 1.533 in each of the time periods. The RI having the greatest frequency of occurrence, represented by the maxima in the two histograms, also does not change, that is, 1.5185 before 1980 and 1.5184 after 1980. However, the distribution of samples around these most frequently observed values is much narrower in the post-1980 samples than for those received prior to 1980. For example, the percentage has increased from 6.8 percent of samples with RIs between 1.5184 and 1.5186 before 1980 to 15.6 percent with RIs between 1.5183 and 1.5185 after 1980.
Another clear difference shown in the two figures is that, in the older glass population, there are a greater number of samples having RIs that are quite different than the most frequently occurring RI. In particular, the percentage of glasses with RIs in the 1.523 to 1.528 range drops from 34 percent in the pre-1980 samples to 9 percent in the post-1980 samples. These observations are consistent with the fact that the more recent sheet glass samples contain a greater proportion of glass made by the float process than the older glass and also provide some support to the belief of improved quality control by the manufacturers.
It is noted, however, that glasses from pre-float sources still make up a portion of the glass population seen in current case evidence. Because the FBI samples are received from a wide geographical area (United States and abroad), the figures indicate that concerns about a widening of RI distributions from an influx of foreign-made glass appear unfounded, at least at this time.
At this point, it is appropriate to comment on the use of these and other databases. The observed narrowing of the spread among RI values supports the contention of earlier authors that it is advantageous to seek additional, more discriminating measures that can be used to compare fragments of glass having indistinguishable RI values, particularly when those RI values are of high frequency of occurrence. On the other hand, the narrowing of RI distributions increases the evidentiary significance when glass specimens are found to have indistinguishable RIs of an uncommon value.
The FBI's and other databases may be used to get a general idea of the rarity of glass having a particular refractive index. None of the existing databases, however, is wholly appropriate for evaluating the likelihood that an innocent person would be associated with glass fragments of a particular RI. Reasons for this include biases toward sheet glass or scene-related glass, improper representation of a critical population in a given geographic area, and a variety of other databasing problems discussed in the literature (for examples, see Curran et al. 2000).
Even when considering large databases collected during long time periods, as shown in this study, variation over time occurs and renders these databases improper for statistical calculations. Therefore, the user is cautioned to be wary in attempting to use databases to determine exact frequency-of-occurrence values or to perform precise statistical calculations.
Almirall, J. R. Forensic Glass Analysis: Overview and New Developments. Presented at the International Symposium on the Forensic Examination of Trace Evidence in Transition, San Antonio, Texas, 1996.
Buscaglia, J. Elemental analysis of small glass fragments in forensic science, Analytica Chemica Acta (1994) 288:17–24.
Buscaglia, J. and Kubic, T. A. Forensic Glass Analysis in the 1900s. Presented at the Northeast Association of Forensic Sciences, Huntington, New York, 1991.
Curran, J. M., Triggs, C. M., Almirall, J. R., Buckleton, J. S., and Walsh, K. A. J. The interpretation of elemental composition measurements from forensic glass evidence, Science & Justice (1997) 37:241–244.
Curran, J. M., Hicks, T. N., and Buckleton, J. S. Forensic Interpretation of Glass Evidence. CRC Press, Boca Raton, Florida, 2000.
Koons, R. D., Peters, C.A., and Rebbert, P. S. Comparison of refractive index, energy dispersive x-ray fluorescence and inductively coupled plasma atomic emission spectrometry for forensic characterization of sheet glass fragments, Journal of Analytical Atomic Spectrometry (1991) 6:451–456.
Lambert, J. A. and Evett, I. W. The refractive index distribution of control glass samples examined by the forensic science laboratories in the United Kingdom, Forensic Science International (1984) 26:1–23.
Miller, E. Forensic glass comparisons. In: Forensic Science Handbook. Prentice-Hall, Englewood Cliffs, New Jersey, 1982, pp. 139–183.
Stoecklein, W. Determination of Source and Characterization of Glass of International Origin. Presented at the International Symposium on the Forensic Examination of Trace Evidence in Transition, San Antonio, Texas, 1996.