- © The Mineralogical Society Of America
The rock stratigraphic record exists as a rich, albeit complex and incomplete repository of Earth history, tracing the processes of biological evolution, climate change, oceanic and atmospheric chemistry, sea-level fluctuations, mountain-building and erosion, and basin subsidence. Without a detailed and precise temporal framework, however, the richness of the record, including global stratigraphic correlations, evaluation of cyclicity and rates of regional and global change, major sedimentary depositional hiatuses, and biological extinctions cannot be fully exploited.
In the past two decades major advances in U-Pb zircon geochronology have allowed us to evaluate the distribution of time in the rock record and rates of geological processes with unprecedented precision (see Parrish et al., this volume; Davis et al., this volume). While much work has focused on major transitions in Earth history, there is considerable promise for a highly calibrated time scale from the Neoproterozoic to Holocene that will permit increasingly more sophisticated questions to be addressed using the rock record. It has been long appreciated that there are dramatic events in the diversification and extinction of life such as the Cambrian radiation, the end-Permian extinction, and the end-Cretaceous extinction. However, important questions remain regarding the tempo and causes of evolutionary radiation and extinction. For example, what are the durations of mass extinctions? How long does ecological recovery take following a major extinction? Do evolutionary radiations correlate with changes in chemistry and temperature of the ocean-atmosphere system and global climate? Are there relationships between evolution and the aggregation and dispersal of supercontinents? Are apparently abrupt isotopic excursions in seawater chemistry globally synchronous and of similar duration? Although the biostratigraphic record has been historically used to address these questions, we are now moving into an era where an essential test of correlation and tempo will and must be high-precision geochronology of volcanic rocks interlayered with fossil-bearing rocks. This means moving beyond calibration of the time scale to understanding the detailed distribution of time in the rock record.
Volcanic rocks in the form of thin air-fall tuffs, regional ash-flow tuffs, and lavas are the most useful temporal markers in the geologic record. Volcanic rocks have long been appreciated as important markers even prior to the advent of high-precision dating. In many cases the physical volcanology and mineral chemistry of ash-beds have been used for regional correlation (e.g., Haynes 1994). Volcanic debris is often injected into the stratosphere by explosive eruption and can be transported hundreds to thousands of kilometers before being deposited. Individual ash layers in the rock record often range in thickness from a few millimeters to as much as several meters. Geochronological data obtained from such units can be used to estimate rates of sediment accumulation and biological evolution, calibrate excursions recorded in chemostratigraphic proxies, calibrate molecular clocks, constrain the timing of tectonic events, and confirm or deny problematic lithostratigraphic correlations (e.g., Ordovician bentonites). In the past decade the recognition and significance of volcanic ash-beds in the rock record has gained wider appreciation, although it is our opinion that many are still unrecognized.
In a broad sense the time scale is well-enough resolved through biostratigraphy and a coarse framework of absolute dates to predict the age of a volcanic rock interlayered with fossil-bearing rocks to within 5–10 Myr of its true age. However it is now possible to determine the age of Phanerozoic volcanic rocks to a precision approaching 0.1% or better, allowing a much finer resolution of time and process in the rock record. At the same time, by pushing the current analytical limits it has become necessary to examine in detail all possible sources of analytical uncertainty as well as geological complexity that can affect high-precision U-Pb geochronology, before we can determine the ultimate resolving power of the method. This paper will review the current state of the application of U-Pb geochronology to the stratigraphic record, evaluate the current limitations, and discuss the future.
THE GEOLOGIC TIME SCALE
High-precision geochronology is revolutionizing our understanding of how time is distributed in the Vendian to Mesozoic rock record (Fig. 1⇓). The unification of paleontological and geochronological records has allowed the development of a new field best described as quantitative biostratigraphy. For much of the Paleozoic rock record this field is in its infancy, and it is widely recognized that a comprehensive absolute time scale for much of the Paleozoic and Mesozoic, including some of the major extinctions, is far from complete. There have been a number of notable compilations of available geochronological constraints on the distribution of time in the rock record (e.g., Harland et al. 1990, Young and Laurie 1996). Unfortunately, in an attempt to include all available data, many of these time scales have typically averaged several dates obtained by different techniques and often of highly variable quality. The result, which is propagated into subsequent publications, is a time scale that is often poorly calibrated in absolute terms. However, the numbers of calibration studies and their geochronological resolution have increased in the past decade, which is in turn stimulating paleontological and paleoenvironmental research.
The geological time scale has two components: (1) a chronostratigraphic scale based on the lithostratigraphic record of geological and biological events established through the principles of superposition, which is calibrated in terms of (2) a geochronometric scale based on isotopic ages. The bounds of chronostratigraphic units, like the base of the Cambrian or the end-Permian and end-Cretaceous extinctions, are defined by the first appearance of diagnostic fossils in defined sections or GSSPs (Global Standard Stratotype-section and Point). Geochronometric calibration of a relative, chronostratigraphic time scale is conceptually straightforward. In the simplest case, a dateable volcanic rock occurs at or very close to a point in a stratigraphic section chosen as a global stratotype for the boundary between two geological intervals. Unfortunately, this situation is not common, and more often calibration requires the dating of rocks in sections other than the stratotype which are then correlated to the stratotype by means of biostratigraphy, chemostratigraphy, and magnetostratigraphy. In many cases the age of the boundary is estimated by interpolation.
It cannot be assumed that the first or last appearance of a fossil occurs at exactly the same time in different depositional settings. The terminal Neoproterozoic-Cambrian boundary is an excellent example. The boundary is formally defined as a point in rock in a stratigraphic section located in southeastern Newfoundland that coincides with the first appearance of a distinctive trace-fossil called Phycodes pedum, now known as Treptictnus pedum (Landing 1994). However, there are no volcanic rocks close to the boundary in Newfoundland. Consequently, the age of the boundary can only be calibrated through correlation with other sections that contain dateable rocks.
Proxies for radiometric geochronology: chemostratigraphy
The use of chemostratigraphy as a tool for global correlation assumes that carbonate rocks precipitated from seawater record a global signal of the geochemistry of seawater at a given time. Thus, the variation in chemical signatures of carbonates such as carbon, oxygen, and strontium isotopes with stratigraphic position can be used to correlate between sections (e.g., Kaufman et al. 1997, Brasier et al. 1996, Pelechaty et al. 1996, Bartley 1998). This is especially useful for late Neoproterozoic and earliest Cambrian rocks that contain a paucity of biostratigraphically useful fossils. For example, the Newfoundland sections lack the small shelly fossils and the well-defined carbon-isotopic record that is found in rocks of similar age elsewhere. Other sections in northwestern Canada (Narbonne et al. 1994), northeastern Siberia (Knoll et al. 1995), and Namibia (Grotzinger et al. 1995) contain carbonates, fossils, and/or ash-beds which have been used to support the view that the base of the Cambrian Period is approximately coincident in time (Landing 1994) on a global scale. However, given interbasinal differences in accumulation rates, diagenesis, tectonic setting, unconformities, preservation potential, and first appearance of diagnostic fossils, detailed correlation of chemostratigraphic signals requires precise geochronology. In particular, evaluating the potential of global diachroneity of a biostratigraphic boundary can only be done in the context of high-precision geochronology. A crucial issue in the evaluation of abrupt isotopic excursions such as are associated with the Neoproterozoic-Cambrian boundary and the Permo-Triassic boundary is their duration. Because accumulation rates can vary widely, the only way to test the global synchroneity and determine the duration of isotopic excursions is with high-precision geochronology.
The end-Permian extinction is characterized by an abrupt disappearance of Permian fossils followed by recovery to Triassic assemblages (Erwin 1993, 1994; Jin et al. 2000, Erwin et al. 2002). At the type section at Meishan, China, the boundary is marked by bentonites interlayered with fossil-bearing carbonate rocks. Chemostratigraphic studies from many different sections globally, both marine and terrestrial, indicate a shift in carbon isotopic values of at least −4‰ from the late Permian to the Early Triassic. However, whether the isotopic shift exactly coincides with the extinction and whether the extinction is globally synchronous are still open questions, resolvable only with the integration of high-precision geochronology, chemostratigraphy, and paleontology (Jin et al. 2000) from multiple stratigraphic sections.
Modern high-precision geochronological studies associated with calibration of the time scale rely on two principal geochronological methods: 40Ar/39Ar in the minerals biotite, hornblende, and sanidine, and U-Pb in zircons and less commonly baddeleyite, monazite and xenotime. Each system has strengths and limitations (above all the availability of appropriate minerals), but many silica-rich lavas and tuffs can be dated using either or both methods. The most profound differences between the two methods are that the U-Pb method exploits two independent chronometers (238U-206Pb, and 235U-207Pb) that can be compared to evaluate open-system behavior, and when measured by isotope dilution can be gravimetrically calibrated to high precision and accuracy, whereas the Ar-Ar method, based on the singular decay of 40K-40Ar, can partially evaluate open-system behavior through stepwise degassing, but relies on calibration with “standard” minerals. In the quest for ever more accurate and precise dates on volcanic rocks, it is becoming clear that both techniques can suffer from geological complexity such as post-depositional modification and pre- to syn-eruption inheritance of older minerals. Additionally, due to the low abundance of 235U relative to 238U, studies of Mesozoic and younger zircons often rely on the more precise measurement of the 238U-206Pb system, compromising the evaluation of closed-system behavior.
Until now, these two techniques have been compared in only a few studies that were devoted to resolving the problem of the absolute calibration of 40Ar/39Ar mineral ages (e.g., Renne et al. 1998a, Villeneuve et al. 2000, Min et al. 2001, Schmitz and Bowring 2001). Although the dates of individual zircons are inherently more robust in absolute terms, the assumption that zircon crystallization can be equated with eruption age is not always warranted given evidence from some studies for resolvable magma residence times in upper crustal reservoirs (Christensen and DePaolo 1993, Davies et al. 1994, Reid et al. 1997, Reid and Coath 2000). Even in rocks where magma residence times are minimized, there is preliminary evidence that 40Ar/39Ar dates could be up to 1% younger than U-Pb zircon dates from the same rock (e.g., Min et al. 2001) but it is only recently that errors calculated for each date have been precise enough to make this potential bias noteworthy, or establish its cause. However, as the time scale becomes better calibrated it may be necessary to distinguish between 40Ar/39Ar and U-Pb years until systematic biases are reconciled.
There are two techniques for the measurement of U-Pb decay that have been applied to calibrating the time scale: Secondary Ion Mass Spectrometry (SIMS)— specifically the Sensitive High Resolution Ion MicroProbe (SHRIMP)—and conventional Isotope Dilution analysis using Thermal Ionization Mass Spectrometry (IDTIMS). The two techniques are fundamentally different and each has advantages and disadvantages that help to elucidate many of the issues associated with time scale geochronology (see Ireland and Williams, this volume, for review of SIMS; Parrish et al., this volume, for a review of IDTIMS). Among their major differences is that the SIMS typically analyzes small, nanogram-sized domains within individual zircons, whereas state-of-the-art IDTIMS analyses are made on single grains, from less than one hundred to several thousands of nanograms in weight. While analyses performed with an ion-probe generally have very small amounts of common Pb, counting statistic limitations arise from the small quantity of zircon and radiogenic Pb being analyzed. Analyses of unknown zircons must also be referenced to a standard zircon of known age in order to calculate U-Pb ages and account for instrumental bias and drift. This calibration contributes a substantial amount of uncertainty to the calculated U-Pb dates. The appropriate selection and use of well characterized, concordant standards has been a recurring issue in SHRIMP U-Pb analysis (Compston 1999, 2000a, b; 2001)
With high spatial resolution comes a trade-off with precision, and the analytical errors for SHRIMP analyses are typically an order of magnitude greater than IDTIMS (Fig. 2⇓, Compston 1999). This imprecision makes the detection of subtle amounts of Pb-loss or inheritance difficult or impossible. For time scale work, often 10–50 spots from a variety of zircons are analyzed and a weighted mean date is calculated using the individual 238U-206Pb dates (e.g., Claoue-Long et al. 1991, 1995; Compston 1999). This approach can yield reduced statistical uncertainties for samplings of Paleozoic zircons that are on the order of ±2–3 million years, assuming a single-aged population. It has become commonplace in zircon studies using the SHRIMP to assess large data sets with gross measured dispersion using the mixture modeling approach of Sambridge and Compston (1994). This approach attempts to statistically deconvolve discrete age components from variably complex probability distributions. In some cases, components older and younger than the interpreted age may be resolved from multi-modal distributions, and are usually attributed to inheritance and Pb-loss respectively. While a very powerful technique for recognizing multiple components in complex data sets, it cannot overcome the basic limitations of the large uncertainties associated with individual ion probe analyses, nor systematic errors associated with standardization.
Recent application of high-precision IDTIMS U-Pb zircon geochronology to Phanerozoic volcanic ash beds has pushed the limits of current analytical capability, and emphasized the rigorous assessment of analytical and geological uncertainty in U-Pb zircon age determinations (Tucker and McKerrow 1995, Brack et al. 1996, Mundil et al. 1996, Bowring et al. 1998, Tucker et al. 1998, Ludwig et al. 1999, Davidek et al. 1998, Mundil et al. 2001). Recently, Schmitz and Bowring (2001) investigated the precision, accuracy, and suitability of modern isotope dilution U-Pb zircon geochronology for the Cenozoic, by collecting a large data set of single- and multi-grain zircon and titanite analyses from the well-studied Oligocene Fish Canyon Tuff (FCT) (Lipman et al. 1970, Steven and Lipman 1976, Bachmann et al. 2000). With these data, it was possible to quantify the various sources of uncertainty in U-Pb zircon analysis, including geological factors like Pb-loss, inheritance, accurate initial common Pb corrections, and intermediate daughter product disequilibria, as well as analytical factors including instrumental fractionation, tracer calibration, and blank estimation. The results serve as an example of the resolving power and limitations of a reasonably straightforward U-Pb zircon data set.
The major sources of uncertainty in IDTIMS U-Pb analyses are: (1) measurement error, including uncertainty in mass fractionation and the linearity of the ion detection system; (2) common Pb correction, including both analytical blank and initial common Pb present in the zircon; (3) U/Pb tracer calibration; (4) U decay constant uncertainties and (5) U-series disequilibrium effects.
With modern ion-counting systems and stable, high-yield (>1%) ionization, Pb isotope ratios can be measured precisely (e.g., 207Pb/206Pb, σμ ≤ 0.05%) for samples containing as little as 5 picograms of radiogenic Pb. Mass dependent fractionation of both U and Pb ions during analysis is a potentially important source of error. U fractionation is best monitored by using a double spike (233U, 235U, or 236U) and can be determined to 0.01%/amu or better. Lacking two constant abundance isotopes, Pb fractionation is most commonly estimated by repeated analysis and long-term reproducibility of the NIST SRM 981 Pb wire standard. Based on literature reports and our own experience, the magnitude of Pb isotope mass fractionation is typically in the range of 0.10–0.15%/amu on Faraday cups and 0.15–0.20%/amu for ion-counting Daly detection systems on VG-Micromass instruments. Our database of hundreds of NBS 981 runs illustrates the typical long-term reproducibility (±0.03–0.05%/amu) in fractionation with only a slight correlation with temperature. As many labs move to analyzing very small samples (<20 picograms of Pb), the question of whether fractionation changes with sample size must be evaluated (e.g., Roddick et al. 1987). At MIT we routinely analyze small (<100–200 picogram) NBS 981 standards for 6–8 hours and can show that fractionation tends to be somewhat higher at the lowest temperatures of a run (<1300°C), but is nearly constant until the load is taken to near exhaustion at very high temperatures (>1550°C) when it decreases. We view it essential that analysis protocols of samples resemble those of standards as closely as possible. In this regard, we note the intermittent trend in U-Pb geochronology to load a dissolved sample directly onto a Re filament without ion-exchange chromatography. While this clearly saves time and can reduce analytical blanks, the effect of additional matrix in the sample on Pb fractionation must be rigorously evaluated.
When analyzing small samples and standards with small ion-beams it is common to observe transient isobaric interferences across the Pb mass range, particularly at filament temperatures below approximately 1400°C (Parrish and Krogh 1987). In our experience these effects are minimized by the preparation of silica gel-phosphoric acid emitter solutions (cf. Gerstenberger and Haase 1997) that yield a stable ionization at relatively higher temperatures (1450–1500°C). Early transient interferences can be monitored and affected ratios discarded by examining the time series of ratio measurements, as well as the off-peak baselines. Another possible source of uncertainty that can arise is non-linearity of the ion-counting system. Typically, linearity is monitored by analysis of standard isotope ratios with a range of ion-beam intensities (Hayes et al. 1978) and Daly-type photomultiplier detectors can be shown to be linear to within ≤0.1% at count rates ≤1,500,000 cps.
Common Pb correction
All zircon analyses have a component of common Pb. Historically, there has been some controversy about how much common Pb is incorporated into the zircon crystal structure and how much occurs along fractures, in solid and fluid inclusions, and in the analytical blanks. Tom Krogh pioneered low blanks in zircon analyses in the 1970s and 1980s and first showed that most zircons contained little to no indigenous common Pb at the picogram level (Davis et al., this volume). Experimental work also determined the limited solubility of Pb in the crystalline zircon structure (Watson et al. 1997), and in situ SHRIMP analysis has further demonstrated the vanishingly small amounts of common Pb in the majority of zircons. Although there are documented exceptions including poorly understood incorporation into radiation damaged zircon (Mattinson 1994), most common Pb in IDTIMS analyses is apparently hosted by inclusions, present as surface contamination, or introduced during chemical processing.
As a result, minimization of laboratory blanks remains the single most important requirement for high-precision U-Pb zircon analysis, and in the last decade, most U-Pb laboratories have been able to reduce analytical blanks to below 5 picograms, and some to less than 1 picogram. With the growing appreciation that diamagnetic, clear, crack- and inclusion-free zircons separated from volcanic rocks have little to no indigenous common Pb, the assumption that all measured common Pb arises from laboratory blank is warranted for many samples with picogram levels of total common Pb. Most laboratories conservatively incorporate a substantial uncertainty in the magnitude of the laboratory blank (e.g., 50%) into error propagation, which is usually the dominant source of error in each analysis unless the blank is well constrained by numerous and consistent total procedural blank determinations (Ludwig 1980).
The isotopic composition of laboratory blank Pb is also of concern. Given that there are multiple sources of laboratory common Pb from airborne particulate, labware, and reagents, and that the contribution and isotopic composition of these sources change with time, it is challenging to determine an accurate isotopic composition of common Pb in a blank. This difficulty is only exacerbated when the measured ion beams constituting those blanks are small in magnitude. Nonetheless, most laboratories have characterized and adopted a reported isotopic composition with a realistically large uncertainty for the blank that includes temporal variability.
Even in the best geochronology labs the total amount of common Pb in an analysis can exceed estimated laboratory blanks, and the assignment of an isotopic composition to this supposed indigenous, or initial common Pb can have a discernable effect on the calculated date and error associated with an analysis, depending upon the contrast between the assumed blank and initial Pb isotopic compositions. Most geochronology labs use two approaches for estimating the composition of initial common Pb. The simplest is to use a model for Pb isotopic evolution such as that of Stacey and Kramers (1975), and assign a model composition corresponding to the nominal age of the zircon. It is ironic that while this model was proposed to describe the average evolution of Pb in a mantle reservoir, many silicic, zircon-bearing rocks are derived from melting older crust rather than mantle, thus calling into question the applicability of this model approach. Another approach is to use the isotopic composition of Pb in a comagmatic low U/Pb mineral such as alkali feldspar (Ludwig and Silver 1977, Housh and Bowring 1991). In general, the older the rocks the less likely it is that one can recover fresh feldspar and that its original isotopic composition can be determined. This problem is particularly acute in altered bentonitic ash layers without preserved feldspar.
Uncertainty in the composition of the initial common Pb is not usually propagated into the age error for many zircon geochronological applications, generally because of its minimal contribution for radiogenic samples. However, for high-precision geochronology and time scale calibration, where a complete description of uncertainties is vital to some interpretations, this systematic error is best propagated by calculating individual analysis dates and errors using reasonable bounds on initial common Pb compositions, deriving weighted means and errors for the resulting data sets, and appropriately supplementing the weighted mean error calculated from the assumed average initial common Pb composition according to the resultant dispersion introduced by varying the common Pb composition.
We must emphasize, however, that in any U-Pb zircon analysis the most crucial parameter is the ratio of radiogenic to common Pb, often indirectly expressed by the measured 206Pb/204Pb ratio. The higher this ratio the less sensitive the analysis is to the content and composition(s) of both blank and initial common Pb. Ultimately, analyses with low radiogenic to common Pb ratios are best rejected or given less weight in age calculation, as incorrect assignment of common Pb and uncertainty can have a discernable effect on the calculated age. This is especially true for the analysis of young, low-U zircons where the radiogenic to common Pb ratio can approach unity (Bowring et al. 1998, Mundil et al. 2001). One can see this effect qualitatively in Figure 2⇑, which shows representative errors for analyses of zircons from near the Permo-Triassic boundary at Meishan, South China. The error ellipse for a single SHRIMP analysis [A] is at one standard deviation (67% confidence interval) while those for two IDTIMS analyses [B and C] are two standard deviations (95% confidence interval). The IDTIMS analysis [B] represents a single grain with 26 picograms of total Pb with a blank of 2.6 picograms while [C] is two grains with a total of 12 picograms Pb and a blank of 2.1 picograms. The less favorable radiogenic Pb/common Pb ratio for [C] results in a proportionately larger error ellipse than for [B]. Also shown are error ellipses for hypothetical analyses that show a small amount of inheritance [D] and Pb-loss [E].
Tracer calibration and interlaboratory comparison
The uncertainty in U-Pb tracer calibration must be considered when comparing data between laboratories and can be among the larger sources of uncertainty (along with decay constant errors) in the absolute age of a population of zircons. The majority of laboratories have calibrated tracers against gravimetric standard solutions to arrive at an estimated uncertainty of approximately 0.1% (2σ). When this is propagated through error calculations, it introduces a proportionate uncertainty in the absolute age that is applicable to the intercomparison of U-Pb ages (but generally not Pb-Pb ages for clean 205Pb spikes) between laboratories. In the next decade as more and more laboratories become involved in very high-precision geochronology of the time scale, subtle interlaboratory biases should be rigorously evaluated by analysis of a suite of concordant zircon secondary “standards” (e.g., Wiedenbeck et al. 1995, Schmitz et al. 2003).
U decay constants
Renne et al. (1998b) referred to the U-Pb zircon method as the “gold standard” of geochronology because of its unique attributes of having two completely independent decay schemes, well-calibrated decay constants, and extreme resistance to resetting. The 235U and 238U decay constants were precisely determined more than 30 years ago with 2σ errors of 0.136 and 0.108%, respectively (Jaffey et al. 1971), thus Begemann et al. (2001) suggest using the 206Pb/238U system as the absolute standard for age normalization because dates derived from this system are least affected by decay constant uncertainties. However, there has been discussion of the possibility of systematic errors in the U decay constants based on comparing the degree of concordance in zircon analyses (Mattinson 1987, 2000) and the suggestion that the counting errors associated with the constants of Jaffey et al. (1971) underestimate the true error. Schmitz and Bowring (2001) suggested on the basis of a large number of concordant analyses (corrected for disequilibrium effects) of the Fish Canyon tuff zircons that the published decay constants and their uncertainties appear robust. Furthermore, analysis of the commonly used 1.1 Ga ion microprobe zircon standard (AS-3) also yields a cluster of concordant and equivalent analyses (Schmitz et al. 2003). Further work using large data sets of different aged zircons is still necessary, but at present there is no compelling reason to not accept the published experimentally determined decay constants and uncertainties. Mattinson (1987) and Ludwig (2000) have suggested propagating the decay constant uncertainties into calculated zircon dates, which can lead to very large errors in the 207Pb/206Pb date. While this is essential for intercomparison between U-Pb and 40Ar/39Ar, a time scale built entirely on the U-Pb method need not include it in error calculation if the dates that are being compared all use the same decay constants.
INTERMEDIATE DAUGHTER PRODUCT DISEQUILIBRIA
The U-Th-Pb systematics of young accessory minerals may be affected by disequilibrium partitioning of intermediate daughter nuclides within the decay chains during crystallization of the mineral-isotopic system under scrutiny. Sufficiently long-lived intermediate daughter products that may significantly perturb equilibrium U-Pb isotopic systematics are 230Th and 234U in the 238U-206Pb decay chain, and 231Pa in the 235U-207Pb decay chain (Mattinson 1973, Schärer 1984). In the former decay chain, 234U is not substantively fractionated from 238U during high-temperature magmatic processes, such that only 230Th disequilibrium need be considered. Because of the slightly greater ionic radius of Th relative to U and the octahedral site in zircon, intermediate daughter 230Th is normally preferentially excluded from zircon, resulting in 230Th deficiency and anomalously young 206Pb/238U dates for young zircon. An extremely rare exception to this deficiency has been noted in carbonatitic zircon (Amelin and Zaitsev 2002).
Disequilibrium between minerals and magma may be quantified and 230Th deficiencies corrected with a standard methodology utilizing 232Th/238U ratios (cf. Schärer 1984). Evaluation of the disequilibrium partitioning is mainly limited by our knowledge of the Th/U of the magma coexisting with the crystal during growth. Assessing this ratio at the time of eruption can be relatively simple for unaltered volcanic rocks, through analysis of pumice or glassy lava compositions.
The large zircon data set for the Fish Canyon Tuff presented by Schmitz and Bowring (2001) serves to illustrate the general problems of initial intermediate daughter disequilibria, and their mitigation. Their analysis suggests that 230Th disequilibrium corrected 206Pb/238U dates are generally shifted by only +0.08 Myr relative to measured 206Pb/238U dates, and that the magnitude of this correction is robust to within 0.02 Myr for large variations in magmatic Th/U. Thus uncertainties associated with 230Th disequilibrium do not appear to hamper our ability to resolve time at the ±0.1 Ma level using zircon 206Pb/238U ages in Oligocene and certainly older volcanic rocks.
In the 235U-207Pb decay chain, 231Pa-disequilibrium effects should be of lesser magnitude than those described above for 230Th-disequilibrium, because of the shorter half-life of 231Pa relative to 230Th—unless mineral-magma fractionation is significantly greater for Pa than for Th. While a quantitative knowledge of the partitioning of 231Pa between accessory minerals and melts is lacking, qualitative assessments of Pa partitioning relative to U and Th, based on relative ionic radii, suggest that Pa4+ depletion in zircon would be less than that of Th4+ (Shannon 1976, Mattinson 1973, Barth et al. 1989). However, Barth et al. (1989) emphasized the likelihood that in relatively oxidizing magmatic conditions, protactinium is present as Pa5+, in which case its minor enrichment relative to U4+ may occur in zircon, due to its closer similarity in ionic radius to Zr4+. Regardless of the details of 231Pa partitioning, it does not appear that its associated disequilibrium effects in zircon commonly exceed those of 230Th; more dramatic postulated 231Pa disequilibrium effects in zircon are extremely rare and are restricted to rocks with unusual alkaline, pegmatitic compositions (Mortensen et al. 1992, Anczkiewicz et al. 2001).
GEOLOGICAL COMPLEXITY AND OPEN SYSTEM BEHAVIOR
Crystal inheritance and Pb-loss
It is a common occurrence in airfall ash deposits to recognize zircon grains, often identical in appearance to the indigenous magmatic population, that are necessarily <1 to >>10 Myr older than the true age of the ash (e.g., Landing et al. 1998, Bowring et al. 1998, Mundil et al. 2001, Palfy et al. 2000). Since the lifespan of single strato-volcanoes to large magmatic arcs can be on the order of one to tens of millions of years respectively, the mechanisms and potential for this style of inheritance is clear. The two major mechanisms include magmatic inheritance (overgrowth of an older core by new zircon) and/or physical addition of older grains into the eruption column or during dispersal and deposition of the ash. In our experience, older zircons can make up as much as 50% of the zircons recovered from certain ashes. Restricting analyses to single grains or grain fragments is the best way to detect and overcome this problem. On the other hand large error ellipses related to either measurement uncertainty or unfavorable radiogenic to common Pb values in some single-grain analyses can mask subtle inheritance.
Zircon analyses that are too young and/or exhibit discordance attributable to Pb-loss are also common in the study of volcanic ash beds. Pb-loss has been inferred from large data sets obtained by both conventional and ion-probe analysis (e.g., Compston 2001a,b), indicating the various spatial scales of the phenomenon. Experimental measurements of Pb diffusivity in zircon indicate that Pb-loss is not likely dominated by volume diffusion through crystalline zircon (Lee et al. 1997, Cherniak and Watson 2000) but rather related to loss from radiation-damaged domains through crystal defects and fractures. While selection of the best quality, diamagnetic zircons for analysis and the aggressive abrasion of their outer portions can minimize the effects of Pb-loss, it is a problem that can be difficult to recognize without a large number of analyses. Discordance of analyses is obviously the best indicator of Pb-loss, but unfortunately, for Late Paleozoic and younger zircons, the limited curvature of concordia combined with imprecision in the measured 207Pb/235U date can limit the discernment of discordance as an obvious signal of Pb-loss.
Resolving a geological “age” from a large population of zircon dates
For Paleozoic zircons a typical propagated error in individual zircon dates is about 0.3–1.5%, or several hundred thousand years to several million years. If it can be shown that the range of single zircon ages exceeds analytical uncertainty it raises the problem of distinguishing between the competing phenomena of zircon inheritance and Pb-loss in order to deconvolve the true eruption and depositional age of a volcanic unit. On the other hand, in tightly clustered data sets where Pb loss and inheritance are apparently minimal, then the small standard error statistics of the inverse variance weighted mean age of a population of zircons can place tight constraints on the eruption age of a tuff. Either scenario requires a closer examination of both the distribution of single zircon ages, as well as the applicability of the applied sample statistics.
Ideally one should obtain a relatively large number of zircon analyses in order to evaluate these issues through probability density functions (e.g., Sambridge and Compston 1994, Schmitz and Bowring 2001, Schmitz et al. 2003). If all analyses fall within two standard deviations of a single inverse-variance weighted mean, one may confidently conclude that it is consistent with a single episode of zircon growth. A fair question is how many analyses are enough? The answer in part lies in how the data are being interpreted—the higher the resolution and confidence desired the more data per rock that is required. For time scale work, our experience suggests that 10 to 20 or even more high quality zircon analyses, including a large cluster of concordant and equivalent data, are often necessary to arrive at a single ash bed age, depending upon the complexity of the zircon population.
A given U-Pb zircon data set provides a multiplicity of dates that can be used to calculate an age of crystallization, including the weighted mean 206Pb/238U, 207Pb/235U, or 207Pb/206Pb dates, the Concordia age algorithm of Ludwig (1999) which combines all three chronometers, or the upper intercept date of an array of variably discordant data. With optimal data sets of concordant and equivalent zircons, these dates converge and the Concordia age algorithm provides the best age estimator, including the appropriately minimized decay constant errors.
In early Paleozoic and older rocks, zircon analyses often yield discordia arrays for which an upper intercept date can be calculated; this date is generally more robust when anchored by one or more concordant analyses. When the lower intercept is within error of present day, the upper intercept age approaches the weighted mean 207Pb/206Pb date, and both can be robust and precise estimates of the age of the rock (Davidek et al. 1998, Landing et al. 2000), although Ludwig (2000) has pointed out the magnification of decay constant errors associated with these ages.
In Late Paleozoic and younger rocks that have low U zircons, the lower abundances of 235U and radiogenic 207Pb result in a larger uncertainty in the 207Pb/235U date due to measurement errors and increased sensitivity to common Pb corrections. Small biases in 207Pb/235U are correspondingly magnified in the 207Pb/206Pb date because of the relatively reduced ingrowth of 207Pb over the past several hundred million years (reflected in the limited curvature of concordia). Because of these factors the 207Pb/206Pb date is generally not the preferred age for Late Paleozoic and younger rocks. Instead, the U/Pb dates are more robust, and the weighted mean 206Pb/238U date is most often used to calculate the age of a rock because of its better precision and insensitivity to systematic error relative to the 207Pb/235U date (Kamo et al. 1996, Bowring et al. 1998, Mundil et al. 2001).
It should be remembered that the power of the U-Pb system derives in part from the fact that two independent decay paths can be used to evaluate concordance and closed system behavior. IDTIMS analyses of very small (low radiogenic Pb) samples or ones with a relatively high proportion of common Pb do not allow this power to be fully appreciated. However, with low blanks and reasonably high radiogenic to common Pb ratios it is possible to evaluate concordance of the two decay systems even in Cenozoic zircons (e.g., Schmitz and Bowring 2001).
More complex data sets with significant dispersion must be rigorously examined in terms of distinguishing the eruption age from the effects of Pb-loss and inheritance. It should be stressed that this level of attention is essential for establishing a very high-precision record. In complex zircon data sets, it is clear that at some point the rejection of outliers can devolve into a subjective process, and statistically valid means of identifying and rejecting outliers and incorporating excess scatter into age errors remain areas of active research (Ludwig and Mundil 2002). Looking to the future, establishing other geochemical fingerprints for inheritance and Pb-loss could partially resolve problematic data distributions. Nonetheless, in some cases neither careful statistical treatment nor analysis of additional zircons may adequately resolve problematic distributions. In these cases it may be more productive to attempt recollection of the ash layer in nearby sections, or to date another volcanic layer from the same general stratigraphic interval.
Sample selection and analytical strategies
The best way to approach high-precision calibration of the time scale by dating individual ash beds is to accumulate a large amount of data for each bed so that one can evaluate the potential effects of inheritance and Pb-loss. This involves processing enough rock so that a substantial number of the best quality, preferably large zircons can be isolated for analysis. In our experience, successful heavy mineral separation from clay-rich bentonitic ashes incorporates high-energy ultrasonication during settling of increasingly diluted slurries to minimize loss of zircons through flocculation. Resulting crystalline concentrates are processed by standard heavy liquid and magnetic methods to isolate the highest quality, nonmagnetic to diamagnetic zircons.
Ideally single grains of zircon should be analyzed, although as previous discussions make obvious, a smaller amount of radiogenic Pb analyzed in a single zircon translates into larger errors, so that at some point there are trade-offs between the requirements of analyzing single grains and obtaining adequate precision. Mundil et al. (2001) have discussed the potential for averaging bias in ages calculated from multi-grain zircon fractions picked from complex zircon populations. However, while many zircon populations from volcanic rocks can be quite complex, others are not, and as was documented in the case of the Fish Canyon Tuff, analysis of multiple grains of zircon can allow for more precise analyses with no observable bias in calculated dates (Schmitz and Bowring, 2001).
Most U-Pb workers favor aggressive air abrasion of zircons recovered from volcanic ash-beds. Two positive effects are generally noted with air-abrasion: reduced common Pb and enhanced concordance. Following the technique of Krogh (1982), abrasion appears to remove the outer and/or damaged parts of the zircon crystals and only the best quality grain domains survive the process. The fine polishing of zircons resulting from abrasion with pyrite also minimizes the surface area of the grains, and corresponding sites for retention of surficial common Pb. Aggressive washing and ultrasonication with warm dilute HNO3 is a necessary and effective means of removing surface correlated common Pb from abraded and unabraded zircons without discernable U/Pb fractionation. Mundil et al. (2001), following Mattinson (1994), has also suggested hot HF leaching as an alternative to mechanical abrasion as a means of improving concordance, although the increased potential for laboratory fractionation of U and Pb during leaching makes the process somewhat unattractive.
In an ideal situation, one has a sequence of volcanic rocks interlayered with fossil-bearing rocks so that the stratigraphic succession is known. A first order test of the geochronology is that the ash-beds yield ages that either all overlap within analytical uncertainties or correspond to the correct stratigraphic order. If neither of these situations result, then it is difficult to know which dates to accept as depositional ages. As discussed, it is quite possible to remobilized precursor volcaniclastic material with the magmatic products of an eruption, and to have resulting volcanic deposits with few to no indigenous zircons but many incorporated during eruption and deposition.
The best way to illustrate both the power and more problematic aspects of U-Pb zircon geochronology as applied to the calibration of the time scale is to discuss case studies such as the Neoproterozoic-Cambrian radiation, the Triassic-Jurassic boundary and the Permo-Triassic extinction, where high-precision geochronology is a crucial tool.
THE NEOPROTEROZOIC-CAMBRIAN TRANSITION
It has long been recognized that all of the classes and orders of multicellular animals present today appear relatively suddenly in the fossil record in the lower Cambrian during the so-called “Cambrian explosion.” The age and nature of the Neoproterozoic-Cambrian boundary has thus been of considerable interest for many years. Estimates for the absolute age of the boundary and the definition of the boundary itself have varied widely (Fig. 1⇑), although 542 Ma is currently the best estimate (Grotzinger et al. 1995, Amthor et al. 2003). The possibility that a mass extinction coincides with the boundary (e.g., Knoll and Carroll 1999, Amthor et al. 2003) raises the important question whether it reflects a fundamental geological and/or biological event or whether the boundary is part of a protracted evolutionary history that culminates in a final pulse of increased fossil preservation. The past decade has seen dramatic new discoveries regarding the paleontology and chronostratigraphy of the transition between the late Neoproterozoic and the Cambrian (e.g., Grotzinger et al. 1995, Jensen et al. 1998, Gehling et al. 2001, Martin et al. 2000, Knoll and Carroll 1999, Erwin and Davidson 2001). This transition corresponds to a time of major tectonic and climatic change including the assembly and dispersal of a supercontinent, glaciation, and large fluctuations in the chemistry of the oceans and atmosphere (Hoffman et al. 1998, Kaufman et al. 1997, Knoll and Walter 1992). Chemostratigraphic and biostratigraphic studies are now focusing on evaluating the connections between tectonics, climate change and the evolution of life and the age, duration, and global synchroneity of dramatic isotopic excursions such as the one that coincides with the Neoproterozoic-Cambrian boundary (e.g., Bartley et al. 1998, Kaufman et al. 1997, Amthor et al. 2003). Further resolution of this question will depend on abundant high-precision geochronology of ash-beds integrated with the paleontological and chemostratigraphic records.
The dating of the Neoproterozoic-Cambrian boundary (see Fig. 1⇑) is a good example of pushing the limits of geochronology to solve a problem. A surprisingly young estimate of ca. 543 Ma was published in 1993 (Bowring et al. 1993). This estimate was based on the age of a pumice-rich volcanic breccia erupted immediately above the base of the Cambrian as defined by lithological and chemostratigraphic correlation. The breccia is overlain by siltstones and sandstones that contain both Cambrian small shelly fossils and Phycodes pedum (Bowring et al. 1993). This volcanic breccia contains abundant xenoliths of country rock and many of the separated zircons were distinctly older than the emplacement age. However, a population of grains yielded an upper intercept date of 543.8+5.1/−1.3 Ma that at the time was the first robust minimum age estimate for the boundary. Subsequent work in Namibia (Grotzinger et al. 1995) showed that an ash bed, overlain by Ediacaran fossils, had an age of 543.3±1 Ma and based on variations in carbon isotopes was assumed to be just below the Neoproterozoic-Cambrian boundary. The less precise age from the Siberian breccia is nominally older than the Namibian ash although they overlap within uncertainties. Most recently, Amthor et al. (2003) report the occurrence of the Neoproterozoic-Cambrian boundary from the subsurface of Oman with ash-beds on either side of the boundary. Chemostratigraphic and paleontological data are interpreted to indicate the simultaneous occurrence of an extinction of Precambrian shelly fossils (the lightly calcified Namacalathus and Cloudina) and a negative excursion in carbon isotopes at the boundary. The ash-beds on either side of the excursion have ages of 543 and 542 Ma consistent with the data from Namibia and provide the best estimate of the age of the boundary and a very short duration of the isotope excursion.
The questions that can now be asked are whether the Neoproterozoic-Cambrian boundary and associated carbon isotope excursion are globally synchronous, and whether there is a global -scale extinction across the boundary? The data that bear on the first question was obtained over a period of about ten years in the MIT geochronology lab where techniques and protocols have been and continue to be refined. Nonetheless there is a remarkable agreement for the age of the boundary around 542–543 Ma. The first determined and least precise date (Siberia) should now be revisited in an attempt to reduce errors and to evaluate any difference in age with Oman and Namibia. Clearly the data from Oman is now the most precise estimate of the age of the boundary and of the duration of the negative excursion in carbon isotopes. Future work could attempt to resolve small differences in time between the three sections. Ultimately this should be possible at the ±300–500 kyr level. An answer to the second question is more difficult but will involve study of the sections where Cambrian and Neoproterozoic fossils appear to coexist and establishment of a precise chronostratigraphy.
Figure 3⇓ shows existing timing constraints and important fossil occurrences for late Neoproterozoic and Cambrian time as well as generalized variations in carbon isotopic composition of carbonates. Of particular importance are the coexistence of distinctive small shelly (Cloudina-Namacalathus) and Ediacaran fauna below the boundary, the negative spike in δ13C close to the boundary, and the first appearance of Treptichnus pedum in the basal Cambrian. At present, the data suggest that late Neoproterozic time is best viewed as a period of biological diversification and preservation characterized by distinctive faunal assemblages and chemostratigraphic signals. Ediacaran fossils are present globally from >565 Ma to at least 543 Ma with the appearance of a mollusc-like animal (Kimberella) ca. 555 Ma (Martin et al. 2000). Ediacaran fossils disappear from the rock record ca. 542 Ma due either to closing of a taphonomic window (Gehling 1999) or because they suffered a mass extinction. As no known stratigraphic section across the transition has a combination of diagnostic fossils, chemostratigraphic constraints, and dated ash-beds, the nature and significance of the boundary is not fully understood. However, a number of studies indicate the occurrence of T. pedum or similar trace fossils below the putative boundary in the western US (Hagadorn and Waggoner 2000), Australia (Jenson et al. 1998), Namibia (Jenson et al. 2000) and even the type section in Newfoundland (Gehling et al. 2001) and Ediacaran fossils above the boundary (Hagadorn et al. 2000). Droser et al (1999) concluded that despite the uncertainties, T. pedum is still a reliable indicator for the approximate location of the boundary and represents an important evolutionary milestone. Thus, T. Pedum and other complex trace fossils became common at a time roughly coincident with the last occurrence of many Ediacaran soft-bodied taxa, the last occurrence of lightly calcified Cloudina and Namacalathus, and a major C-isotopic excursion (Grotzinger et al. 1995, Bartley et al. 1998, Amthor 2003). This raises the questions of whether or not the isotopic record of environmental change and the possible extinction are related (Bartley et al. 1998) and whether extinction may have been in part responsible for the Cambrian explosion?
What triggered the Cambrian radiation is still an open question. The relative roles of extrinsic (environment) and intrinsic (developmental) controls on this radiation are still not understood well. However, the idea that a major change occurred in the chemistry of the oceans and atmospheres, in particular a rise in oxygen concentrations, played a role in the dramatic increase in animal size and complexity is intriguing (Knoll 1996, Knoll and Carroll 1999). We believe that a well calibrated global chemostratigraphy and biostratigraphy will ultimately allow us to better understand this crucial transition in the history of life.
The Triassic-Jurassic (Tr-J) boundary is another excellent example of how U-Pb geochronology has been used to resolve the age and significance of a major extinction. The Tr-J boundary marks one of the five largest mass extinctions in Earth history. There are both well-studied terrestrial (Olsen et al. 1996) and marine (Palfy et al. 1999) sections. For many years the age of the boundary was poorly constrained with estimates as low as 190 Ma (Seidemann 1988). Harland et al. (1990) proposed an age of 208.0±7.5 Ma whereas more recently Gradstein et al. (1994) have proposed a refined age of 205.7±4 Ma.
The cause of the extinction is not known, although both bolide impact (Olsen et al. 2002) and flood basalt magmatism (Marzoli et al. 1999) have been implicated. Olsen et al. (2002) reported that a rapid rise in dinosaur diversity, preserved in terrestrial rift-basin rocks of eastern North America, immediately followed the Tr-J boundary and that the boundary is associated with an iridium spike and a fern spore spike consistent with the extinction being caused by a bolide impact. Reports of shocked quartz from the Tr-J boundary in Italy support this idea (Bice et al. 1992).
Another potential cause of the Tr-J extinction is the eruption of a massive amount (ca. 2 ×106 km3) of basaltic magma at ca. 200 Ma, remnants of which are now exposed in once-contiguous parts of North America, Europe, Africa, and South America (Marzoli et al. 1999). These magmatic rocks are referred to as the Central Atlantic Magmatic Province or CAMP (Marzoli et al. 1999). The basaltic and gabbroic rocks that occur on either side of the Tr-J boundary in the rift basins of the eastern U.S., as described below, are thought to be part of this event. A large number of Ar-Ar dates from a variety of intrusive and extrusive rocks from both the eastern U.S. and Brazil are consistent with emplacement of this magmatic province between 199 and 201 Ma and a possible link with the Tr-J extinction (Marzoli et al. 1999, Hames et al. 2000) that could involve high levels of atmospheric CO2 and destabilization of methane hydrates (Beerling and Berner 2002, Palfy et al. 2001).
For terrestrial sections, Dunning and Hodych (1990) reported U-Pb zircon and baddeleyite dates of 201±1 Ma for two intrusive sills thought to be correlative with basalt flows in the lowermost Jurassic, and concluded that the boundary was slightly older than 201 Ma. Subsequently Hodych and Dunning (1992) reported a U-Pb zircon date of 202±1 Ma for the North Mountain basalt of Nova Scotia, which is interpreted by Olsen et al. (1987) to be less than 0.2 Ma younger than the T-J.
A U-Pb age of 199.6±0.3 Ma for has been reported (Palfy et al. 2000) for a tuff just below the base of the Jurassic from a marine section on Kunga Island in western Canada. This age coupled with additional data from other marine sections (Palfy et al. 1999, 2000) suggests an age of approximately 200 Ma for the marine Tr-J boundary. Palfy et al. (2000) compared the terrestrial and marine records and concluded that the marine extinction did not occur before 199.9 Ma and the terrestrial extinction no later than 200.6 Ma (youngest possible age of the intrusive sills considering uncertainties) and therefore the terrestrial extinction preceded that in the marine realm by 0.7 Myr. This discrepancy argues against a catastrophic bolide impact but could be consistent with a more complex chain of events that led to both extinctions (Palfy et al. 2000).
The age and significance of the Tr-J boundary highlights many of the issues discussed in this paper and is an excellent example of a problem that could benefit from additional high-precision geochronology. While intriguing, the apparent 0.7 Myr difference between the terrestrial and marine extinctions needs additional scrutiny. While both estimates rely on U-Pb zircon geochronology, they were obtained by two different laboratories with differing analytical parameters. The question of interlaboratory bias can only be evaluated by routine analysis of concordant zircon standards by both labs. While the marine and terrestrial ages appear to be distinct, confirmation with further analysis by both labs is desirable. Also, the apparent coincidence of the CAMP magmatism and the extinction based on Ar-Ar geochronology is noteworthy but must be tempered by the observation that Ar-Ar dates may be approximately 1% too young when compared to U-Pb dates (Min et al. 2001). If true, the 199–201 dates would be shifted by about 2 Myr. Clearly more high-precision U-Pb dates on the basaltic rocks would help with this evaluation as would Ar-Ar dates on the same rocks. Ultimately, dates on the Tr-J boundary from other localities would serve as additional tests and serve as a model for evaluating the detailed timing and causes of other mass extinctions.
THE END-PERMIAN EXTINCTION
In south China, several marine stratigraphic sections preserve the Permo-Triassic boundary, including the type section at Meishan. Interlayered with the fossil-bearing rocks is a series of volcanic ash beds, which have been precisely dated using the U-Pb zircon technique (Bowring et 1998, Mundil et al. 2001). In these studies, ages of individual ash-beds were determined with uncer tainties of less than 0.5 Myr. However despite analysis of more than 100 zircons in both studies there is a lack of consensus on the age of the boundary and the duration of the extinction.
The age, duration, and cause of extinction at the Permo-Triassic boundary have been the sub ject of considerable debate (Erwin 1993). Claoue-Long et al. (1991) obtained a SHRIMP U-Pb on bed 25 at Meishan of 251.1±3.4 Ma. Subsequently, Campbell et al. (1992) dated a gabbro cuts the Siberian traps and drew attention to the fact that within errors, the two overlapped in and that the extinction could be due in part to the eruption. Renne et al. (1995) presented high precision 40Ar/39Ar data for sanidine and plagioclase feldspar separated from Bed 25 at Meishan calculated a weighted mean age of 250±1.5 Ma. In addition, they showed that this age was synchro nous with the intrusion of a gabbro that cuts the lower third of the Siberian traps and suggested casual link between the two events.
Bowring et al. (1998) published a series of ages for ash-beds interlayered with carbonate rocks from Meishan and two other localities in south China. The age of the boundary was estimated to be 251.4±0.3 at Meishan and just younger than 251.7±0.2 at Heshan. This boundary age compares with U-Pb zircon and baddeleyite age from the N’orlisk intrusion and lavas of 251.2±0.3 obtained by Kamo et al. (1996), again suggesting synchrony of the extinction and the eruption of the Siberian Traps.
The geochronological results from south China were used by Bowring et al. (1998) to two further conclusions. The first is that the age of the boundary is the same within errors for sections 1500 km apart (Meishan and Heshan). The second is that the extinction occurs in less 1 Myr. Three possible scenarios were suggested to explain the events at the P-T boundary. In first, eruption of the Siberian flood basalts in the latest Changhsingian released large amounts CO2 (and possibly sulfates, producing acid rain) and initiated a period of global warming. Warming of shallow seas lowered the lysocline sufficiently to release some 1200 Gt of oceanic methane hydrates (Bowring et al. 1998). Following the arguments of Renne et al. (1995), a short volcanic winter may have been triggered by volcanic aerosols and was followed by greenhouse conditions and warming. This cooling-warming cycle could have triggered convective overturn of the oceans, dumping deep CO2-rich bottom water onto the shelf regions (Knoll et al. 1996), leading to hypocap nia and a rise of atmospheric CO2 levels.
In the second scenario, volcanism caused collapse of primary productivity and cessation export of light carbon to the deep ocean, producing a transient isotopic shift after which the oceans returned to more normal δ13C values of around 0 during recovery. It is possible that the export sequestered light carbon and CO2 charged water by upwelling, combined with volcanic-induced extinction could explain the observations. The third possibility is that the latest Permian biota declining as a result of the above scenarios and that the collision of Earth with an icy object pushed the planet to the brink of total extinction (Bowring et al. 1998). All of the mechanisms discussed above have been operative at other times in Earth history and may have even caused other extinc tions. However, none before or since has been as dramatic as the Permo-Triassic extinction. single mechanism is apparently sufficient to explain all the geologic and paleontological data, the massive eruption of the Siberian traps may well have been the proximal cause for a cascade events leading to the apparent synchroneity of marine and terrestrial extinctions.
Mundil et al. (2001) challenged the conclusions of Bowring et al. (1998) with the publication of additional data for five ash beds from the Meishan section (Fig. 4⇓). The data for each ash-were used to either calculate a weighted mean date or to estimate a minimum age of the bed based on interpreting the data in terms of inheritance and Pb-loss. In general, the calculated dates violate stratigraphic order; for example, one of the most precise dates of 252.0±0.4 Ma is reported for Bed 15 which is 17.3 m below the boundary, yet Bed 28, 0.08 m above the boundary has a reported date of 252.5±0.3 Ma (Fig. 4⇓). Mundil et al. (2001) appear to hang their interpretation of the entire section on the latter date, yet it is difficult to understand how a similarly tight array of data indicating an age of 252.0 Ma for Bed 15 can be ignored.
One of the strongest criticisms of the Bowring et al. (1998) data set was that it included both single grain and multiple grain data. Mundil et al. argue that multiple grain analyses run the risk of averaging small differences in age and result in precise but inaccurate results. While this may be true, in single grains with small amounts of radiogenic Pb, analyses are much more susceptible to small differences in the isotopic composition and relative proportion of the blank, as well as measurement uncertainty. Thus it is not always clear whether small differences in dates of single grains are real or in part related to the problems of analyzing very small amounts of Pb. Thus, much as when comparing ion-microprobe analyses to conventional IDTIMS analyses, it is a question of whether taking the weighted mean of several large error ellipses associated with analyzing small amounts of zircon is better than combining enough grains together to yield a high sample/blank ratio and precise analysis.
If Siberian Traps volcanism was a driver for the Permo-Triassic extinction, precise geochronology should reveal a very close association. Kamo et al. (1996) obtained high precision, single-grain U-Pb geochronological data on both zircon and baddeleyite from the ore-bearing Noril’sk-1 intrusion that yielded an age of 251.2±0.3 Ma. In addition, Kamo et al. (2000) and Fedorenko et al. (2000) report and discuss respectively, U-Pb geochronological results from the Maymecha-Kotuy area that confirms the suggestion that the entire sequence was erupted in less than 1 Myr. From near the base of the sequence in the Maymecha-Kotuy area, perovskite from a melanephelinite has a date of 252.1±0.4 Ma and two lavas near the top yield U-Pb zircon dates of 251.1±0.5 Ma. At face value, the age and duration of the Siberian Traps overlaps the extinction based on the U-Pb dates obtained at Meishan (Bowring et. al. 1998), with the oldest dates at 252.1 Ma and the youngest at 251.1 Ma. In contrast, if the age of the extinction is as old or older than 253 Ma, as proposed by Mundil et al. (2001), then volcanism is distinctly younger. This is a matter that must be resolved with additional constraints on the age of the boundary both in China and in other localities. It is clear that we need to know the age and duration of both the Siberian Traps and the Permo-Triassic boundary to better than 200,000 years. This is surely a challenge for U-Pb geochronology.
The age of the Permo-Triassic boundary at Meishan is a good example of the issues associated with high-precision geochronology of critical intervals in earth history. It demonstrates the potential complexity of zircon populations in volcanic ash-beds and illustrates the need for dating multiple ash-beds in stratigraphic order. It also illustrates the unresolved analytical difficulties inherent in the analysis of picogram quantities of radiogenic Pb, and the current impasse between analysis of single zircon grains to isolate Pb-loss and inheritance issues and multiple grains to minimize measurement uncertainties. Hopefully, the zircon systematics of the Meishan ashes constitutes a worst-case scenario and other Phanerozoic sections will continue to yield more easily interpretable results. In fact the data that are emerging from the Triassic (Mundil et al. 1996, Brack et al. 1996, Palfy et al. 2003) are providing a consistent and highly resolved record for the post-extinction recovery and lending insight into the duration and periodicity of cycles preserved in carbonate platform rocks. We conclude by emphasizing the demonstrated ability of U-Pb zircon geochronology to resolve absolute Neoproterozoic to Phanerozoic time at the sub-million year level, and the important insights its application may reveal in the near future.
Until recently, the distribution of time in the geologic record has been viewed mostly in terms of biostratigraphy. However it has been shown that by integrating high-precision geochronology with paleontology one can begin to address fundamental issues such as rates of evolution, diachrony in the fossil record, and the affect of environmental change on the tempo of evolution. Only when the rock record is highly calibrated in an absolute sense will we fully appreciate the detailed history contained within. A highly calibrated rock record will lead to closer scrutiny of evolutionary change in index fossils such as conodonts and trilobites and could lead to major new insight into the causes of adaptations.
Modern U-Pb geochronological techniques can yield uncertainties approaching 0.1% for the ages of volcanic ash-beds interlayered with fossiliferous rocks. However there is a need for community-wide participation in the analysis of well-constrained ash-beds as well as “standard” zircons so that small interlaboratory biases can be included in propagated errors. The geochronological and paleontological communities can look forward to a highly resolved rock record that will permit a much more detailed understanding of earth history.
In the future, geochronology, paleontology, and molecular biology must be better integrated to determine rates of evolution, extinction, and recovery. Distinguishing between extrinsic and intrinsic controls on evolution will require a full understanding of the tempo of assembly and dispersal of continents and the related climatic changes. There is no doubt that a more comprehensive understanding of the distribution of time in the rock record will open vast new areas of study in the Earth Sciences.
MIT’s work on high-precision U-Pb geochronology is and has been supported by NASA’s Exobiology (NAG5-8814), The NASA Astrobiology Institute (NCC2-1053), and NSF (EAR 9725727 and EAR 9804988). We thank Dan Condon for discussion, review and a great deal of help with manuscript preparation. Randy Parrish provided a particularly thoughtful and thorough review, which greatly improved the manuscript. Chris Fedo and John Hanchar also reviewed the manuscript and their insightful comments are appreciated.