Uranium-Lead Dating Fraught With Discordance

Behind the confidence of uranium-lead dates hides worry about numbers that don’t match up.

Uranium-lead (U-Pb) dating is a staple of the billions-of-years claims about igneous and metamorphic rocks and meteorites – giving rise to the consensus age of the solar system at 4.55 billion years.  The technique has been refined over the last century, but “discordances” (mismatches) remain.  That’s because there are two “isotopically distinct yet chemically identical” decay chains, F. Corfu described in the Geological Society of America Bulletin this month: the ²³⁵U to ²⁰⁷Pb pathway, and the ²³⁸U to ²⁰⁶Pb pathway.  (The ratios of the products, ²⁰⁷Pb/²⁰⁶Pb, can also be measured.)  “These twin decay systems, running at different speeds, allow an immediate verification of the validity of their ages, which must be concordant to be considered valid, although under favorable circumstances, discordant data can be extrapolated to the correct age.”  Since uranium is often locked in tightly-bound minerals called zircons, from which parent and daughter product are locked in, the method is thought by modern geologists to provide a reliable clock.

In “A century of U-Pb geochronology: The long quest towards concordance,” Corfu, a geologist at the University of Oslo, described geologists’ frustration at persistent discordance between the two methods.  The discordances are small; around 1% or less.  Still, he finds it troubling that there should be any discordance if the same physical processes are operating for the same amount of time.  At the beginning of the article, he quoted T. E. Krogh: “Only one answer is the right one, and only that one is good enough.”

The progress in developments of the U-Pb method is both a history of technical discoveries and advances, as well as a history of a long struggle toward concordance, toward an understanding of the causes of discordance, and toward ways to eliminate it. Despite the enormous progress achieved in this field, the problems of U-Pb discordance have not yet been completely resolved and will be one of the main hurdles to overcome in the future.

The paper is technical, so detailed evaluation of the contents should be left to geophysicists.  The following quotes give a sense of the kinds of problems Corfu discussed, and their implications.  Alert readers will want to watch for instances of card stacking, circular reasoning, extrapolation and sidestepping.  Philosophers of science will want to watch for data manipulation, unargued interpretation and possible self-deception.

• Now, 60 yr later, we see two contrasting trends, two philosophies: one that continues the battle to eliminate discordance, but also a second one that bypasses the problem and solves the Gordian knot by essentially eliminating the concept, declaring U-Pb systems concordant by definition (see section “Administrative Concordancy”).
• [By 1956] There was a gradually growing consensus that the dominant cause of discordance was loss of Pb, and different attempts were made to rationalize the effect and extract reliable ages.
• While intellectually stimulating, these models and discussions could not provide unique and fully reliable solutions, and the problems of discordant U-Pb data remained a challenge….
• The rebound from these rather discouraging early phases in the development of U-Pb can be credited largely to the work and imagination of Leon Silver, at Caltech in Pasadena. He realized that the weak point in the analytical procedure was “the absence of a suitable approach in selecting samples from which data could be successfully systematized and interpreted” (Silver, 1963a, 1963b, p. 281). He understood that zircon populations in rocks can be extremely variable in terms of composition, crystallinity, and other parameters. He experimented in separating fractions according to size and magnetic variability, and his measurements of the U-Pb ratios of the different fractions revealed extreme variation in the degree of discordance, which correlated with U content and degree of radioactivity. The U-Pb results defined collinear arrays for which intercepts could be interpreted in terms of primary crystallization and secondary Pb loss (Silver, 1963a, 1963b; Silver and Deutsch, 1963).
• Silver’s developments set up analytical approaches that were to become standard procedures in the following two decades. Given that discordance was next to unavoidable in most samples, the strategy became that of maximizing the spread in order to establish more robust discordia lines and increase the reliability of their intercept ages. A major hurdle, however, remained the immense analytical effort needed to produce individual analyses from zircon fractions of 100–500 mg (Silver and Deutsch, 1963), which severely limited the utility of the method.
• The change of focus by Tom Krogh, then at the Carnegie Institution of Washington, to U-Pb geochronology from his previous involvement with Rb-Sr dating (Kamo et al., 2011) represents the next major step in the evolution of the method, and the one that really enabled the more widespread application and testing of the principles and analytical approaches defined by Leon Silver’s work…. Although this flurry of activities and new applications helped to solve many geological questions, many of the studies involving zircon had to rely on extrapolating the age from discordant arrays, which in complex cases had detrimental effects on the precision and accuracy of the ages.
• Since model solutions to discordance did not work very well, and since extrapolating the ages from discordant arrays could be problematic in all those cases that had undergone a Pb evolution involving more than just two stages, Tom Krogh concluded that the only reliable strategy had to be that of finding, isolating, and analyzing concordant domains of zircon.
• Because altered domains are easily soluble and could be separated from the unaltered zircon by partial dissolution, several workers experimented with partial dissolution techniques. The results, however, were not encouraging because the technique introduced secondary effects, fractionating U from Pb in the remaining parts of the zircons (e.g., Pidgeon and Hopgood, 1975; Todt and Büsch, 1981; Turek et al., 1982).
• Krogh (1982a) chose instead a different track. To avoid the side effects of the chemical attacks, he opted to remove the outside of the grains, or of fragments of grains, by spinning them around in a steel chamber with a jet of compressed air (air abrasion method). His tests showed that, in many cases, the technique could indeed strip most of the discordant zircon matter… In general, to achieve the best results, these methods had to be combined with careful inspection and selection of the grains under a binocular microscope, and in specific cases with imaging (cf. Corfu et al., 2003). [From here, Corfu discusses other method refinements and cross-correlation techniques leading up to the ion microprobe.]
• When compared to ID-TIMS [isotope dilution—thermal ionization mass spectrometry], the main disadvantage of the ion probe was a considerable decrease in precision, and hence loss of temporal resolution. This stemmed mainly from the much smaller amount of material available for measurement. Another critical step was the determination of the U-Pb ratios, which in SIMS [secondary ionization mass spectrometry] must be calibrated against the corresponding ratios of an external standard, in contrast to ID-TIMS, where the ratios are obtained by mixing a tracer of known composition with the sample. Because the sputtering of U and Pb is controlled to some degree by the composition and/or structure of the sample (McLaren et al., 1994), discrepancies can arise between sample and standard when they have quite distinct basic properties. The anomalous behavior is evident in some zircons very rich in U, such as the cases reported by Harrison et al. (1987) and Wiedenbeck (1995). This potential problem is now largely recognized in the SIMS community, and modern analytical protocols are generally concerned with proper matching of the reference zircons.
• The latest addition to the arsenal of U-Pb dating methods is the laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) technique. The first experiments with U-Pb dating by LA-ICP-MS were done in the early 1990s (Fryer et al., 1993), and from there the technique rapidly improved. In less than a decade, many laboratories worldwide had adopted it, contributing to many improvements and refinements. The method is attractive mainly because of the high throughput capability, which makes it ideal for the study of large populations, such as in detrital zircon studies…. The main initial challenges for LA-ICP-MS were achieving control of the analytical fractionation between U and Pb, and correcting for common Pb….
• In my experience, it is generally difficult to apply the [chemical abrasion] method to U-rich and metamict grains because they dissolve too easily, even in cases where the grains, though rich in U, are free of alteration and yield concordant data when treated with air abrasion (e.g., Svensen et al., 2012).
• The Pb-Pb method is a special branch of U-Pb geochronology, but it has been used mostly to pursue a different type of problem, namely, employing the Pb composition of U-poor or U-free minerals and rocks to study the genesis of rocks rather than to date them. Occasionally, however, the two tracks overlap and complement each other.
• One of the most prominent applications of Pb isotopic analysis to dating was that of Patterson (1956), who determined the first (essentially) correct age of 4.55 Ga for meteorites and, by extension, the planet Earth.
• THE ROAD TOWARD CONCORDANCE: A UTOPIAN GOAL?
• At present, U-Pb dating is the method of choice in many studies of magmatic systems, metamorphism, and provenance. Thousands of studies have examined the behavior of the decay system, the behavior of U-Pb in different minerals, and the response in different geological settings.  The preoccupation with discordance, which laid a cloud of skepticism over the initial enthusiasm, has been largely dispersed by the subsequent developments. Today, we have reasonably good control of the behavior of U-Pb systems. Yet, do we really understand discordance; do we really know how to navigate around it?…  Although the basic definition of concordance is clear-cut, as soon as the analytical error is considered, the term becomes somewhat fuzzy, occasionally progressing into extreme fogginess.
• However, even the use of the better matching decay constant does not always succeed in bringing data sets into concordance.
• Another factor that can affect the degree of discordance is the U isotopic composition.
• A second set of causes that has been widely discussed, and convincingly proven in some cases, is related to initial disequilibrium in the decay chains from U to Pb.
• A related phenomenon is the excess of ²³¹Pa [protactinium], which produces an excess of ²⁰⁷Pb (Fig. 2A-I). The phenomenon has been demonstrated convincingly in two cases (Parrish and Noble, 2003; Anczkiewicz et al., 2001) where the effect is very strong, but it has been discussed as a possible cause in many other cases where the discordance toward too high ²⁰⁷Pb/²³⁵U ratios for uniform ²⁰⁶Pb/²³⁸U is distinct, but not extreme.
• In practice, subtle disequilibrium effects can be difficult to distinguish from effects caused by other factors, such as small amounts of inheritance of xenocrystic material, Pb loss, incorrect common Pb corrections, or simply analytical biases from improper blank corrections, fractionation, and instrumental nonlinearities.
• More severely discordant data will generally have been produced by either Pb loss or mixing.
• By contrast, Pb loss by recrystallization and/or redistribution often means that there may not be any parts of a mineral left with preserved close-system relationships and thus that are capable of providing concordant ages.
• SOME MODERN DILEMMAS
• The main difficulty in interpreting such U-Pb data is in identifying correctly the reasons for the specific distribution of discrete data patterns.
• One critical question affecting modern studies done with chemical abrasion ID-TIMS is whether Pb loss effects can be completely eliminated by the chemical abrasion procedure. Some examples show that this assumption is not always valid (Schoene et al., 2010a). A similar uncertainty, but in the opposite direction, is introduced by the difficulty of evaluating whether some slightly too old grains might indicate the presence of antecrysts or traces of xenocrystic material.
• Finally, the not always easily controllable role of disequilibrium in ²³⁰Th, ²³¹Pa, and ²²²Rn, discussed previously, can further complicate the interpretation of high-resolution data.
• Ultimately, the overriding criterion for the validity of an interpretation is the mutual coherence of the data, their consistency with the various properties of the analyzed zircons, the consistency with coexisting minerals, and the consistency within the geological framework….
• A somewhat related dilemma, which becomes more acute in interpreting Paleozoic and older U-Pb data sets, concerns the interpretation of data that are nearly, but not fully, concordant (Fig. 2A-III). One common solution is to extrapolate to the concordia curve….
• In some cases, the choice is strongly supported by the combined available evidence, whereas in other cases, it is very much open to debate and often decided on the basis of the interpreter’s bias.
• The question as to whether discordant data should be interpreted in terms of modern or ancient disturbances can also have a large impact on the use of Hf [hafnium] isotopic compositions in zircon.
• …the examples stress the importance of a proper consideration of the reality, and the causes and the effects of U-Pb discordance on the results and their interpretations…. Evidently simply ignoring discordance distorts the reality, because, as the present examples show, discordance matters.
• [Regarding a graph in Figure 5]  The initial and the final peaks correspond to the two real events, but the intermediate ones are spurious and geologically meaningless. The example shows that an awareness of discordance and precision are important for accurate interpretations.
• However, in a modern detrital zircon study, such a pattern would normally be interpreted as indicating real ages, and the possibility that individual subconcordant data may be recording partial resetting or mixing is rarely, if ever, considered. This may not be a very important factor in sedimentary rocks produced from relatively simple sources, such as an island arc, but it is likely to have serious repercussions on the apparent age distribution produced in zircon derived from more complex high-grade terranes.
• Administrative Concordancy [i.e., user choice to consider the data concordant by fiat]
  
• Some contamination by common Pb (and U) is essentially unavoidable in all dating techniques…. The ID-TIMS method introduces Pb and U contamination during the dissolution, the chemical separation, and the loading on a filament in preparation for the mass spectrometry. In SIMS and LA-ICP-MS measurements, contaminant Pb (and U) can be introduced mainly during the mounting and polishing process. Practitioners of all three methods are concerned with keeping the common Pb levels as low as possible, and have devised specific techniques to achieve this goal. Nevertheless, it is likely that some contamination will affect every analysis.
• Questions affecting the precision are usually related to the proportion of blank versus initial Pb, and to how well one knows, or can estimate, their compositions and uncertainties. These factors propagate into the total uncertainty, and for very small samples they become its dominant element.
• In LA-ICP-MS analyses, the main problem is caused by the interference on ²⁰⁴Pb by ²⁰⁴Hg introduced with the Ar gas. It is, in principle, possible to minimize the amount of Hg present and correct for it by measuring the abundance of ²⁰²Hg (Gerdes and Zeh, 2009), but most users consider that this correction introduces too much uncertainty, and choose instead to either make no correction at all, based on the assumption that zircon has no common Pb and that the blank is negligible, or they correct the common Pb with a model calculation that assumes a coherent behavior of Th/Pb and U/Pb and estimates the time of the isotopic disturbance (Andersen, 2002). Both approaches can have problematic implications for the accuracy of the data and the geological deductions.
• Factors that are important for a proper interpretation of U-Pb data include the geological and mineralogical context of the studied minerals, the internal textures of the minerals, the age relationships between different components of a mineral and between different coexisting minerals, the type and spatial arrangement of inclusions, and the mineral’s chemical composition.
• Progress in the development of U-Pb dating techniques has been driven forward by the spirit of technical innovation, but at the same time, it has been moderated by some inherent limitations. Many imaginative approaches had to be developed to make the system work.
• Discordance, however, remains a general problem, the minimization of which still requires work and creative solutions, unless, out of convenience, the historical consciousness of discordance simply disappears from the collective awareness.

In that last sentence, Corfu warns that researchers could simply ignore discordance out of “convenience.”  They could pretend that the problems don’t exist or aren’t important, because it’s too much work to get to the one right answer that Tom Krogh said is the only one that’s good enough.  If ignoring the problem becomes the choice of the “collective awareness,” credibility of the U-Pb dating method—or any other method—will suffer as a result.

For anyone taking the time to wade through all those sample quotations from the paper, or even to scan through the bold parts, it should become evident that a great deal of guesswork, human interference and manipulation precedes those confident announcements that such-and-such a rock is xyz million years old.  None of these techniques are simple; the potential for investigator interference is large.  The rock doesn’t just talk on its own.  Under enough torture, it can be made to squeal out any age the investigator wants.

Sometimes in science, the discordant data is the most significant part.  Remember when some 19th-century physicists downplayed the “ultraviolet catastrophe” in blackbody radiation theory?  Most of the theory worked just fine.  Those darned outliers were seen as a problem that would soon be solved within the consensus paradigm.  When a bold maverick, Max Planck, decided to focus on the outlier data, he gave birth to quantum mechanics – a revolutionary new paradigm that changed everything.

As a human with biases, a scientist can be prone to cover up the outlier data points when showing a graph to their colleagues that mostly shows a perfect fit.  In this paper, Corfu has done a good job identifying the known factors that can cause discordance.  But what about the unknowns?  What about the unknowable factors?  Since some of the known causes of discordance came to light over decades, we have no confidence now that new unknowns yet to be identified will not color or revolutionize the interpretations of age accepted today by the consensus.

One big factor is built into the interaction between geologist and lab.  It’s typical for the lab to ask the geologist how old he thinks the rock is based on its provenance in the geologic column.  How many discordant dates are simply tossed and never published because they are “obviously wrong”?  There’s a huge space there for circular reasoning.  How often, if ever, do labs perform true blind tests, publishing results with no prior knowledge of the expected date, letting the sample speak for itself?  Creation geologists have submitted blind samples (i.e., not telling the lab what date to expect), and obtained wildly discordant results that profoundly contradict evolutionary expectations.  See geologist Andrew Snelling’s Earth’s Catastrophic Past, Vol. 2, ch. 103 for examples.  This is a huge topic that deserves more “collective awareness.”

No human was present when the rocks were laid down.  The apparently minor discordances Corfu identified should be re-evaluated in light of what he said about assumptions and interpretations.  They could be much more important than even he, as an evolutionist and believer in long ages, would admit.  His statements waffle between “not a problem” and “big problem!”  For instance, he claimed that Patterson’s estimate of the age of meteorites (“and by extension, planet earth”) was “(essentially) correct”.  But a little later he said, “The main handicap of the method has been the necessity to rely solely on the Pb composition, without assistance from U-Pb, thus limiting the ability to evaluate the state of concordance of the analyses.”  So what is it, correct or handicapped?  Could it be a case of “extreme fogginess” he warned about?  Maybe the consciousness of discordance simply disappeared from his individual awareness, for the sake of convenience, in that instance.

The role of prior belief in millions of years in interpreting the rocks to be millions of years old would make a good study.  We hope our spotlight on Corfu’s paper will stimulate thinking along those lines.