Commentary on Dr. Ken Rahn's Work on the JFK Assassination Investigation
by Patrick M. Grant
Normally, when a concerned scientist has issues with a published work, he or she will submit a letter of protest, criticism, or alternate viewpoint to the journal in which publication occurred. That letter will then be evaluated by the journal editor, perhaps given to other technical staff for opinions of appropriateness, and, if published, often answered or rebutted in the same issue by an accompanying letter from the original authors. The protest letter is usually constrained by the journal to be succinct, focused, and deal with valid technical concerns, and such published communications are considered to be on a reasonable scientific footing. The original authors may or may not sway the interested reader with their subsequent response.
The narrative by Dr. Rahn in his recent Internet posting does not fall into such a category of reasonable technical merit. Consequently, Rick Randich is not inclined to participate in such rambling opinions, and I do agree with Rick in the traditional sense of accepted scientific protocol. However, I will play this game to some limited extent because I have serious technical reservations about Dr. Rahn’s published work that may preclude his contributions to the assassination debate. It is likely time that someone pointed them out to interested readers, despite the completely unconstrained informality of the Web ballpark that Dr. Rahn has chosen to play in.
First, some factual, but scientifically unimportant, errors in this latest posting: Dr. Rahn claims that all of the metallurgy in our July 2006 Journal of Forensic Sciences (JFS) paper was exclusively Rick’s, while the statistical analyses were exclusively my own. He must have fabricated this demarcation within his own mind, as neither Rick nor I have ever claimed this. Indeed, the paper is jointly coauthored, and responsibility for its entire content consequently lies with both authors by definition. Were Dr. Rahn’s supposition correct, two separate papers with a single author each would have reasonably resulted. In fact, Rick is quite knowledgeable about statistical analysis, uses it routinely as a professional analyst, and it was a necessary foundation for his work opposite the FBI in the interpretation of bullet-lead compositional analyses and comparative databases. As for myself, I have been an analytical separations scientist for nearly 40 years, and the macro- and microsegration behaviors of metallurgy are very analogous to the technique of fractional crystallization or zone refining, used historically by chemists to separate different compositional phases from mixtures or solutions [see, e.g., E. W. Berg, Physical and Chemical Methods of Separation, McGraw-Hill, 1963, chapter 9]. The two disciplines merely use different terminology for the same phenomena.
Secondly, I was never in Vince Guinn’s research group, nor was I ever at the University of Maryland, where Vince moved after retiring from UC Irvine. Rather, I worked with Vince at UCI during the late 60s-early 70s. He was a member of my graduate oversight committee, and his was one of three authorization signatures on my PhD thesis, but my research advisor was always Professor F.S. Rowland. However, I have always regarded Vince as an esteemed mentor in NAA and forensic science, and we did collaborate together on one technical article [Science 175: 1121 (1972)].
At UCI, I helped build the TRIGA nuclear reactor that Vince used for his work in the JFK investigation, as well as in other projects, and was an AEC-licensed senior operator for that reactor. George Miller and I set up the Ge(Li) spectrometer system that Vince used for his NAA work, and I performed the detector energy and efficiency calibrations necessary for accurate results. I understand first-hand how Vince did his JFK bullet analyses, the apparatus and reactor irradiation ports that he used, and the relative errors inherent in those various protocols. My PhD thesis with Sherry Rowland was on a novel combination of NAA with hot-atom chemistry to obtain molecular, not just elemental, information. Yet, Dr. Rahn would label me a latter-day NAA revisionist? Exactly who is the poseur here?
Response to Internet Posting
But now let’s get on with the real deal, and I’ll address the major issues presented in Dr. Rahn’s posting. The first part of his offering alludes to words of extremely questionable wisdom from a bandleader and spends several pages on his subjective assessment of “strong,” “absolute,” and “weak” characteristics of various sentences written by Rick and me in the JFS article. Now, I must admit that I’ve been around for some time, but this is the first time I’ve ever seen a completely arbitrary semantic development used as a technical argument. I thought it reasonably entertaining actually, and kudos to Dr. Rahn for forensic novelty (only in the debating sense, however), but it would not pass muster as legitimate commentary in any valid scientific forum that I know of. That’s the beauty of the Web as a communication medium, of course, and likely why Dr. Rahn chooses it in lieu of a substantive letter to JFS. But don’t kid yourself that personal opinion necessarily has any semblance of merit in the absence of formal review -- thus the well-established refereeing system for legitimate technical publication. So, for example, while Wikipedia is a convenient and free on-line encyclopedia, it is not guaranteed accurate, and there are errors of content, some deliberately posted as such. Although the refereed-journal system is not perfect either, it at least has more scientific checks and formal technical inputs than an Internet monologue.
Dr. Rahn claims that his counting of the number of grains in the thin cross-sections of the MC round that we presented in Figure 5 of the JFS article falls above what he would expect by factors of about 2-3. He calculates this by considering that the diameter of a MC bullet is 6.5 mm and that our paper gives the typical grain sizes in the MC rounds we inspected as 500-1000 µm in linear dimension (0.5-1 mm for those who might otherwise have a hard time). However, his dilemma is actually worse than that, as he either does not know, or has overlooked the fact that 6.5 mm is the diameter of the entire bullet, including the gilding-metal jacket. The diameter of the lead core is actually 4.3 mm for most of the length of the bullet and, of course, is a little less at the nose. This lack of fundamental understanding by Dr. Rahn would otherwise serve to increase his factors of 2-3 to discrepancies in the range of factors of 3-7. Now, the only way that Dr. Rahn’s calculation works is if grains of dimension < 500 µm exist, which could give him a higher count across a given transversal. The grains have irregular shapes and are randomly oriented in three dimensions within the bullet structure. If a given bullet cross-section happened to cut across tapering edges of some grains, they could register as apparent, < 500-µm grain sizes. Empirical cross-sections will invariably intersect such irregular grain boundaries, at times producing higher grain counts than would result were all grains transected only at their widest dimension. In fact, however, what we actually wrote in the JFS article was:
It was observed that the individual grain sizes spanned a large range, from less than 100 µm to greater than 1,000 µm. The average grain dimension was approximately 500-1,000 µm, and several grains were > 1,200 µm in length.
The difference factor between < 100 µm and 500 µm is greater than 5×, which would account for Dr. Rahn’s (corrected) factors of 3-7 quite easily. Dr. Rahn unwittingly makes our point nicely here, in that we have observed, and say, that the grain size is highly variable. There will therefore be areas with more, and some with less, than the average size. This is nature at work, not formal statistics. Statistics would be valid if taking into account many thousands of grains on average, but we did not. We considered reality as we observed it, not through or via some artificial, constructed model. I also note that the samples for the NAA analyses were not, and could not, be chosen to obtain an exactly “average” number of grains.
However, Dr. Rahn primarily takes us to task for not providing numeric data on elevated antimony (Sb) concentrations at lead (Pb) grain boundaries: a “gaping hole” in the supporting data. It is true that we “only” showed two indicative photomicrographs of the grain-boundary concentration effect in our JFS paper, thinking that each picture might be more worthy and worth, say, 1000 words (and even more data points) to an unobsessed reader. The referees and journal editor obviously agreed with us. This information could certainly be quantitated given the time and funding to do so. What we did provide, however, were nine classic and historic references (numbers 27 through 35 in the JFS paper) to this fundamental and often taken-for-granted precept of basic metallurgy. These references also provide their own citations to further related work. We are confident that Dr. Rahn can find all of the supporting data he might need in these published studies. However, it may take some effort since, as he opines later in his posting, you really don’t need to make any effort to understand metallurgy for any of this bullet-casting or ammunition-manufacturing stuff anyway.
Dr. Rahn also thinks that, for the metallurgy we discussed in the JFS article to be valid, there must exist a direct correlation between the concentrations of Sb and copper (Cu) in the measurements of bullet-lead compositions. This proposition is quite wrong and reveals a lack of understanding, not only of metallurgy, but also of basic principles of physical chemistry. The reason that Sb and Cu separate to some degree from cooling Pb in the first place is that, chemically and thermally, Sb and Cu are not Pb. That is also why different phases concentrate and separate during slow cooling of a host matrix in fractional crystallization.
For Dr. Rahn’s absolute correspondence between Sb and Cu to be true, these two different elements would have to have the same atomic size and be chemically and thermally identical, or at least very, very similar. However, not only is Sb not Pb, it is not Cu either, so their properties cannot be identical. Now, there are different groupings of elements in nature that are chemically very similar, and these are designated as members of various chemical series. They can be found summarized in an efficient manner on Mendeleev’s Periodic Table of the Chemical Elements. Those elements with similar chemistries are generally found in vertical columns of the Periodic Table (e.g., the Group IA alkali metals or the Group VIIA halogens), although there are some instances where they are depicted as horizontal rows (e.g., the lanthanides and actinides). Thus, sodium (Na) is very much like potassium (K), and they have similar material properties, undergo similar chemical reactions, and form similar salts in chemical combinations.
However, Sb and Cu are not co-members of any chemical series, nor do they have the same atomic radius. Indeed, Sb is a Group VA metalloid, while Cu is a Group IB transition metal. Antimony is regarded an appreciably volatile element, alone and in various molecular combinations (e.g., SbH3, SbX5). It primarily forms compounds such as halides, hydrides, oxides, sulfides, and nitrides, and finds uses in semiconductors, flameproof formulations, and (of course) as an alloying element for hardening Pb. Copper, on the other hand, is a much more refractory element, with both melting point and boiling point many hundreds of °C higher than those of Sb, and it has many more metallic properties due to its excellent electrical and thermal conductivity. Copper also forms various halides, oxides, etc., but very different from those of Sb. Moreover, Cu is additionally found in myriad chemical complexes of diverse nature, such as blood hemoglobin, industrial catalysts, and antibacterial medications. Copper finds wide use as the major component of bronzes and brasses, as an electrical conductor, in coins, and for many other functions.
In short, Sb is not Cu, nor are they at all similar in their physical chemical or thermal properties by the stretch of anyone’s imagination except, it appears, Dr. Rahn’s. The chemical methods used to remove Sb and Cu in lead refining are quite different, since no single process removes both equally well. I also note that the room-temperature solubility of Cu in Pb is less than 70 ppmw, while that of Sb in Pb is approximately 4000 ppmw. Thus, Sb could be preferentially lost through volatility, diffuse at a different rate than Cu, undergo very different interactions along the way to a grain boundary, etc. Copper and Sb therefore need not absolutely concentrate to the same extent at exactly the same grain boundaries in a cooling Pb matrix, especially if the mixture has not been perfectly homogenized. Indeed, two direct quotes from our JFS article are:
Not all elements segregate in the same way since the size of the impurity atom, the diffusion rate of that element, and the crystal structure of any new phases all affect how segregation occurs.
Copper may also segregate via a different route. When a kettle of molten soft lead is refined, it is stirred forcefully up to the point of casting to keep it as homogeneous as possible. The temperature of the melt will be in the range of 350-500°C, and the density of the melt will be essentially that of lead, or approximately 13.2 g/cm3. The melting point of copper is 1083°C, and its density at the melt temperature, where it is solid, is about 8.9 g/cm3. Any small particles of copper remaining in the melt will tend to float. This phenomenon of layering occurs for several low-solubility elements and impurity compounds commonly found in lead alloys. Intermetallics, sulfides, oxides, and silicate stringers are routinely found entrapped in most metal alloys (27, 30). During solidification, the solid particles of copper are reinserted into the liquid phase by the advancing (solid phase) dendrites. The copper will be ultimately trapped in the last liquid to solidify, i.e. at the grain boundaries. These particles of copper can be relatively large compared to those that precipitate during the eutectic transformation.
Personally, I would consider this language somewhat “strong,” using Dr. Rahn’s semantic guidelines, but he must have judged it “ignorable” in crafting his critique. It is actually a well-known fact that solid Cu floats on molten Pb, as it is often routinely removed in slag during Pb refining. Certainly, it is most obvious from Vince’s accurate NAA measurements that the Sb and Cu levels are not necessarily equally concentrated at the same grain boundaries. Were it so, all of his Cu measurements would have been on the same order as his “big outliers,” or all on a “low side,” or all grouped somewhere in between (with a suitable statistical distribution).
I therefore suggest that all of Dr. Rahn’s presented Cu-Sb (and other) element-correlation plots, whether linear or log-log, are of no relevance whatsoever for any applicable consideration of the analyses we presented in the JFS paper. Dr. Rahn’s tutorial on generic log-log graphs, showing what to expect from direct, no, or inverse correlations between variables, is actually pretty good. However, to assess the JFK bullet evidence from an isolated mathematical perspective, while ignoring the physicochemical principles that underlie and define that evidence, comprises serious blinders to reality. As Dr. Rahn states toward the end of his posting, “Notice how far we have come without having to know a thing about metallurgy.” Indeed. (However, going so far in the wrong direction accomplishes little.)
Fatal Flaws in Dr. Rahn’s Assessment of the JFK Assassination Evidence
However, the real purpose of my response to the Internet criticism is less the defense of Rick’s and my JFS article than it is to point out my misgivings in Dr. Rahn’s analysis of the JFK bullet composition data. In my opinion, technical errors, and the faulty assessment that derives from them, are such as to make that work fatally flawed. Dr. Rahn was correct in the semantic efforts of his posting that the word, “proper” in our JFS title did allude to previous improper evaluations. Although we made no serious or specific allegations in that paper, he was well positioned to recognize that particular nuance of our work.
Rick and I performed our assessment of the assassination bullet-lead evidence independent of any previous developments by Dr. Rahn or anyone else. Our involvement in the JFK investigation was stimulated by Rick’s prior work on the interpretation of bullet composition in contemporary jurisprudence, and by my previous association with Vince Guinn and an enduring interest in this singular case. Somewhere along the way, however, someone (perhaps Gary Aguilar?) sent us the two papers that Dr. Rahn published in 2004 in the Journal of Radioanalytical and Nuclear Chemistry with Larry Sturdivan. We did read them and were able to form (hopefully objective) scientific opinions about their effort. I believe that their assessment is based on serious errors that consequently invalidate the entire work.
The Rahn and Sturdivan (R&S) fundamental argument goes basically like this (see the figure after the Cu-Ag log-log plot in the posting, as well as Figs. 2 and 3 in their first JR&NC paper): The inhomogeneity of MC ammunition, as given by the relative standard deviation (%RSD) of Sb concentration measurements, decreases from 90% for whole or individual bullets, to 24% for quarter-bullets, to 3-5% for small fragments and the assassination evidence. Hence, the homogeneity of the bullet Pb increases, with commensurate decreases in the variability of Sb concentration measurements (i.e., smaller compositional RSD errors), as you progress to smaller and smaller sample sizes. Consequently, you should believe that the reported precisions of the NAA measurements of the assassination evidence are indicative of good specimen homogeneities, that they also reflect good accuracies of those data, and that their original interpretation by Vince should be accepted as irrefutable fact.
It appears that R&S believe that it is the size of the original or parent specimen that is important for consideration, rather than the size of the analytic subsample that is actually measured. However, this is completely untrue for an assessment of inhomogeneity of the parent specimen, unless that parent is absolutely homogeneous in its composition. Only then is that subsample representative of the whole. But we already know that bullet Pb can be inherently inhomogeneous in various aspects of its elemental composition and microstructure. Even in Dr. Rahn’s posting:
“. . . . complete chemical homogenization was not necessary to meet the terms of the contract. Thus the final product was ‘clumpy’ in antimony, and to a lesser extent in the other elements associated with the scrap material.”
Consequently, the determination of the Sb concentration in a subsample of any Pb specimen is only a measure of the composition of that subsample, and it cannot be extrapolated to the constitution of any neighboring fragments or to the parent specimen from which it came. It is certainly not necessarily representative of the Sb in an inhomogeneous progenitor.
The weight of an intact, 6.5-mm MC cartridge is approximately 10.3 g. The Pb core of this round weighs 7.13 g, with the remainder being supplied by the metal jacket. Now, "whole bullets" to me implies an analysis (i.e., measurement of Sb concentration) of complete 7.13-g masses of MC bullets. That is, it requires this large mass to establish the R&S highly variable, 90% heterogeneity region of the MC Sb variation in the Pb. In fact, however, the data that R&S averaged for these statistical calculations were Vince Guinn's measurements, as he reported in Table II-A, Appendix D, p. 547, of his HSCA testimony. So did Vince irradiate and γ-count entire MC bullet specimens for these experiments? Or equivalently, perhaps he dissolved whole bullets in individual acid solutions and then removed homogeneous aliquots of each for analysis? The answer is “no” for both of these only possibilities to measure Sb (or any other) concentration integrated over an entire, individual bullet. The following was Vince’s protocol for this study, taken directly from his written testimony (HSCA, Appendix C, p. 540):
“Samples ranging in weight from 44 to 58 milligrams (each weighed to within ± 0.1 milligram) were drilled out from the open base end of each bullet with a cleaned small steel drill, after first scraping the surface of the bullet face free of oxidized layer with a clean stainless-steel scalpel.”
Thus, these were not 7-g specimens, but rather, they ranged in weight from 44-58 mg. These sample sizes were therefore nominal 50-mg drillings from whole bullets, or 0.7% of an individual bullet, but certainly not whole bullets. This protocol was followed for all of Vince’s NAA samples taken from the purchased MC bullets in lots 6000 through 6003, except for the subsamples described next.
Similarly, do the R&S-advertised "quarter-bullets" represent entire bullets cut into fourths, and then suitably measured for Sb content in those 25% subsamples? The answer is “no” again: they were not 7.13 / 4 = 1.8-g specimens analyzed in their entirety. They were actually Vince's above-discussed 50-mg specimens that he subsequently divided into fourths for his homogeneity study. The experimental protocol that Vince testified to is (HSCA, Appendix C, p. 545):
“To study the degree of homogeneity of individual Mannlicher-Carcano bullets, four samples of bullet lead were analyzed from each of three individual bullets (bullets 6001 C, 6002 A, and 6003 A). The larger drillings obtained from each of these three bullets (which were made approximately down the longitudinal axis of each bullet) were cut into four pieces – one of which was the specimen analyzed earlier.”
These subsample sizes were consequently only approximately 50 mg / 4 = 10-15 mg each. They were not 25% of entire MC bullets, but actually less than 0.2% of a whole bullet.
The undisputable result is that the 10-58 mg sizes of the R&S “individual” and "quarter-bullets" are actually in the same range as their designated "smaller fragments" and the assassination evidence (5-50 mg specimens). Thus, the assigned errors for all of their assessed "smaller specimens" should really range from 3-90% for any ensuing analysis and conclusions. Put another way, a ¼-fraction of a 50-mg subsample of Pb does not a quarter-bullet make. And size is the crux of the R&S argument: it is strictly that size (i.e., the mass or weight) of the MC specimens that lies at the foundation of the R&S development. To quote Dr. Rahn in this posting:
“Thus the size scale for heterogeneities in antimony extends to at least the size of whole bullets.”
“Thus the bigger the sample of WCC/MC lead, the more it will differ in antimony from other samples . . . .” (The rest of this sentence reads, “of the same size,” which does not make sense; I presume that Dr. Rahn really meant to conclude this sentence another way.)
It is the size of a sample that would determine, e.g., the number of grains, grain boundaries, and random inhomogeneities relevant to this discussion within an individual sample, completely independent of whether that mass came from drilling a pristine bullet in the lab or whether they were the shattered fragments of a fired round at a crime scene.
By the way, the 24% error datum from Vince's homogeneity measurements is entirely consistent with Rick's and my analysis in the JFS paper. Our corresponding number was 13% and arose from the use of weighted statistics. The 24% number derived from Vince's original calculations of unweighted means and standard deviations. At this point, any reader who has persisted this long with my response should note that all of the %RSDs discussed above have been one standard deviation (1σ).
But now let’s ignore the underlying science in this closing section and look at the bottom line of what “proper” math and error-propagation techniques with the extant data tell us. And, ok, you won’t really need any metallurgy or separation science here. First of all, I think that it's pretty tough to make the case for two distinct populations of assassination bullet evidence when the data are that close and n = 2 and n = 4 for the two purported groups. Nevertheless, using all of Vince’s MC measurements reported in his HSCA testimony, we showed in the JFS article that, if you do a better job of propagating actual accuracies (rather than just precisions and minimum errors) and expand the resulting error bars to 2σ, all of the measurements of the assassination specimens overlap, and no evidence for two separate populations remains. To illustrate this fact, see the graphical representation of the resultant data in the following figure.
Minimum error bars of 2σ (95% confidence) are the modern scientific standard for the cut-off criterion for distinct groups when doing comparative compositional analyses. This detail has been most recently reiterated by J. R. Almirall and T. Trejos, “Advances in the Forensic Analysis of Glass Fragments with a Focus on Refractive Index and Elemental Analysis,” Forensic Science Review 18: 73-96 (2006):
“If the mean value of the recovered sample is within the mean value of the known, ± 2 or 3 standard deviations, they are considered to “match,” suggesting that the samples have a common source of origin.” (p. 88)
To conclude, Dr. Rahn has opined that the metallurgic foundation of our JFS article failed miserably, despite him having no credibility whatsoever in the discipline. His main rebuttal to our thesis, as quoted from his Web posting, is:
“The problem is that actual measurements on WCC/MC lead show exactly the opposite – the variability of antimony decreases as fragments progress from larger to smaller.”
As discussed above, however, this supposed variability is based on vastly inflated (erroneous) representations of 50-mg (“individual bullets”) and 10-15 mg (“quarter-bullets”) specimen sizes. Now, Rick and I are generally happy to discuss our JFK bullet-evidence analysis, as published in JFS, to address the questions or concerns of any credible investigator. In this instance, though, it might be more productive for Dr. Rahn to reconsider the foundation of his own assessment, rather than throw more rocks at ours. His criticism is predicated on that foundation, but it is a mere a house of cards.