Effects of sampling and mineral separation on accuracy of detrital zircon studies



[1] We investigated some of the sampling and mineral separation biases that affect the accuracy of detrital zircon provenance studies. The study has been carried on a natural catchment in the Scottish Highlands that represents a simple two-component source system and on samples of synthetic sediment prepared for this study to test the effects of heavy mineral separation on the resulting zircon age spectra. The results suggest that zircon fertility of the source rocks and physical properties of zircon represent the most important factors affecting the distribution of zircon age populations in the stream sediments. The sample preparation and selection of zircons for analysis may result in preferential loss of information from small zircon grains. Together with the preference for larger crystals during handpicking, it can result in several-fold difference compared to the real age distribution in the sediment sample. These factors appear to be more important for the reproducibility of zircon age spectra than is the number of zircon grains analyzed per sample.

1. Introduction

[2] Since the early studies that utilized U-Pb dating of detrital zircon in sedimentary provenance studies [e.g.,Ledent et al., 1964; Tatsumoto and Patterson, 1964], the main goal has been to identify different age populations in sedimentary rocks and to link them to particular geological events or to potential source areas (qualitative approach). The recent advancements in in situ isotopic dating techniques, namely the widespread use of ion probe (SIMS) and laser ablation (LA) ICP-MS in sedimentary provenance has allowed for larger quantities of data to be obtained per sample. This made it possible to apply a more quantitative approach to the analysis of distribution of detrital age populations in sedimentary rocks and to relate them to specific geological terrains in their source areas. Numerous factors can bias the results of detrital zircon age distribution in provenance studies. These include (1) natural processes, such as change of relief in the source area, mixing of sedimentary components, grain size sorting and quality of the grains, and (2) sampling and sample preparation biases such as heavy mineral separation, preference for size, shape and color during hand picking and analysis and the effects of number of analyzed grains per sample. While the natural processes and sample preparation usually result in random biases of the age distribution that are difficult to quantify, effects of data acquisition and processing (e.g., measurement protocol, use of calibration standards, mass discrimination and fractionation corrections, etc.) can skew the results in a systematic way and lead to inter-laboratory differences.

[3] Cawood et al. [2003] have studied the effects of catchment topography on the zircon distribution where the active (erosion) or passive (deposition) behavior of stream can strongly influence the uptake of zircon from the surrounding rocks. The bias caused by disproportion of zircon contents in rocks has been discussed by Moecher and Samson [2006]. These factors together with the local sourcing of sediments [Fedo et al., 2003; Sircombe et al., 2001] can cause a deviation of the age distribution in the sedimentary rock from the original age distribution in the source terrain. This bias is further propagated into the final zircon age spectra and cannot be simply removed by subsequent sample handling. Additional biases may be introduced during sample preparation. As an example, although Larsen and Poldervaart [1957] argued that sample crushing does not result in significant zircon loss when compared to acid sample dissolution, Hay and Dempster [2009]have recently demonstrated that sample crushing can cause significant loss of low-temperature zircon population. Magnetic separation of heavy mineral fraction on a Frantz isodynamic separator is also known to shift the age spectra in detrital samples with the preference for less magnetic and less discordant zircon grains [Heaman and Parrish, 1991; Sircombe and Stern, 2002]. Sieving and handling the finely crushed/sieved material sets a limit to the size of zircon grains that are eventually analyzed [Fedo et al., 2003]. The effects of Wilfley table and heavy liquid separation on detrital zircon age distribution have not yet been studied in detail [Fedo et al., 2003] but Hietpas et al. [2011]mention the preferential loss of the fine-grained component by Wilfley table separation.Moecher and Samson [2006] discussed biases caused by preference for larger zircon grains during sample preparation and analytical biases caused by more frequent analysis of grain cores compared to the rims. The effect of minimum number of zircon grains that must be analyzed to avoid loss of a zircon population and various strategies for grain selection have been discussed in a number of studies [Dodson et al., 1988; Vermeesch, 2004; Andersen, 2005; Link et al., 2005].

[4] This study investigates the effects of some of the natural and sample preparation biases that affect age distribution of detrital zircons extracted from sedimentary rocks. Modern sediment samples were used to explore the effects of mixing of two zircon components along the river stream on a detrital age spectrum. Synthetic samples with known zircon age distribution were analyzed to explore the reproducibility of the age spectra and the potential biases caused by sample preparation, especially the effects of grain size, heavy mineral separation and selection and number of grains analyzed.

2. Samples and Techniques

2.1. Modern Sediment Samples

[5] Total of 7 samples of modern stream sediments were collected along the river Dee in the Scottish Highlands (Figure 1). This river springs in the Cairngorm Mountains that mostly consist of the Cairngorm granitic intrusion of Caledonian age (404 ± 18 Ma [Oliver et al., 2008]) and it later cuts into the siliciclastic metasedimentary rocks of the Neoproterozoic Grampian Group (Dalradian) with detrital zircon ages between ca. 830 and 2000 Ma. Since the aim of this part of the study was to investigate the mixing of two age components along the stream, we focused on the lower part of the river where a sharp boundary between the Cairngorm granite and the Dalradian metasediments is well exposed in the gorge of the river Dee. The samples (ca. 1 kg of sand from the stream) were collected in ca. 1–2 km intervals from the margin of the Cairngorm granite intrusion (sample Dee-2,Figure 1) along the river down to the main valley where sedimentary material brought by other streams accumulates (sample Dee-7,Figure 1). The first sample (sample Dee-1,Figure 1) has been taken from the stream within the Cairngorm granite to monitor potential contamination by zircons other than those released from the granite. In this part the stream flows in relatively flat terrain and the average width of the river is ca. 5 m with depth of up to 0.5 m (variable due to seasonal changes). The samples Dee-2 and Dee-3 were taken from steeper parts of the stream below the granite-metapsammite boundary where high stream dynamics resulted in sediment-rich flatter parts alternating with numerous small rapids and waterfalls. In contrary, the samples Dee-4 to Dee-7 were taken from the less steep parts of the stream. In the collection site of sample Dee-7 the stream reaches an average width of ca. 10 m and the depth of ca. up to 1 m, with strong seasonal variations. Additional rock samples (ca. 1 kg) of the Cairngorm granite (Dee-1 granite) and metapsammite (Dee-4 Grampian metapsammite) of the Grampian Group from outcrops adjacent to the Dee-1 and Dee-4 sample locations (Figure 1), respectively were collected in order to characterize the zircon age signatures of the two respective source rocks.

Figure 1.

Simplified geological map of the studied natural catchment of the River Dee in Scottish Highlands. The arrow shows the location of the studied area on the map of Scotland. Marked with circles and labeled are the locations of collected samples. Zircon U-Pb ages for the Grampian group metasediments were obtained in this study. The concordia age of the Cairngorm granite was adopted fromOliver et al. [2008].

[6] The stream sediment samples were washed in water and sieved to obtain detrital fraction between ca. 50–500 microns in size. The samples were loaded to tetrabromoethane (TBE) and subsequently to diiodomethane (DIM) heavy liquids followed by removal of the magnetic minerals from the heavy fraction using the Frantz magnetic separator. Zircons were then transferred in ethanol from the pool of separated zircons to a double-sided tape using pipette (tweezers were not used in order to avoid bias for larger grains during the grain transfer), mounted to epoxy resin blocks and polished using SiC and diamond paste to expose and to smooth the inner grains sections suitable for laser ablation ICP-MS analysis. The samples of Cairngorm granite (Dee-1 granite) and Grampian metapsammite (Dee-4 Grampian metapsammite) were crushed, sieved and washed in water to obtain a size fraction between ca. 50–500 microns. The samples were carefully homogenized and 50 g of material from each sample were processed in heavy liquids and using the Frantz magnetic separator (see above) to obtain zircon concentrate for analysis. Zircon grain mounting and polishing followed the same procedure as for the stream samples.

2.2. Synthetic Sediment Samples

[7] Zircon-free quartz sand and known number of zircon grains of known age distribution were used to prepare four synthetic sediment samples (ART-1 to ART-4). The commercially available natural quartz sand (glass sand ST 08 from the Hrdoňovice quarry, Czech Republic) with high certified purity intended for glass production was further cleaned to remove potential impurities. The sand was first processed twice on the Wilfley shaking table and the final cleaning was then done by manual panning. Relatively large portion of the sample was discarded in each cleaning step in order to ensure the purity of the final sand fraction – only ca. 2 kg of zircon-free sand were prepared by processing ca. 10 kg of commercially available quartz sand. The purity of the final sand fraction was tested by processing two sand sub-samples (each weighing 100 g) in heavy liquids, followed by careful inspection of the heavy fraction that could potentially contain zircon under binocular microscope. No zircon grains were found in either of the two sand sub-samples.

[8] We used six natural zircons of known ID TIMS ages to produce the synthetic sediment sample. These were (listed here from the youngest to the oldest): VC # 1–2 (213 ± 1 Ma (J. Košler et al., manuscript in preparation, 2012)), Plešovice (337 ± 1 Ma [Sláma et al., 2008]), Seiland (531 ± 2 Ma [Pedersen et al., 1989]), FC-Z5 (1099.3 ± 0.3 Ma [Paces and Miller, 1993]), 9980 (1150 ± 2 Ma (Košler et al., manuscript in preparation, 2012)) and QGNG (1852 ± 1 Ma, [Black et al., 2003]). The larger zircon fragments were crushed down to small pieces (ca. 50–1000 μm). These were further processed using an air abrasion chamber [Krogh, 1982] in two steps – first in a diamond-lined chamber to produce round grains and then in a stainless steel chamber filled with pyrite powder to polish the grain surfaces to better mimic natural detrital zircons. After dissolving the pyrite remnants in 20% HNO3 and cleaning in deionized water the abraded round zircon grains were split by sieving into two size fractions that were smaller and larger than 100 microns, respectively. The zircon grains were mixed in proportions given in Table 1, photographed in binocular microscope for subsequent image analysis (see below) and mixed with the zircon-free quartz sand (Table 1). Synthetic detrital sediment samples ART-1 and ART-2 were prepared by mixing 2000 zircon grains with 500 g of zircon-free quartz sand; samples ART-3 and ART-4 were prepared by mixing 400 zircon grains with 100 g of zircon-free quartz sand.

Table 1. Composition of Zircons in the Synthetic Sediment Samples
SampleART-1, ART-2ART-3, ART-4
Volume of sand500 g100 g
Total number of grains2000400
Zircons proportions:number of grains (% [number])number of grains (% [number])
<100 μm>100 μm<100 μm>100 μm
VC # 1–2 (213 ± 1 Ma)5% [100]5% [20]
2.5% [50]2.5% [50]2.5% [10]2.5% [10]
Plešovice (337 ± 1 Ma)24.5% [490]24.5% [98]
12.25% [245]12.25% [245]12.25% [49]12.25% [49]
Seiland (531 ± 2 Ma)24.5% [490]24.5% [98]
24.5% [490]0% [0]24.5% [98]0% [0]
FC-Z5 (1099.3 ± 0.3 Ma)15% [300]15% [60]
7.5% [150]7.5% [150]7.5% [30]7.5% [30]
9980 (1150 ± 2 Ma)30% [600]30% [120]
15% [300]15% [300]15% [60]15% [60]
QGNG (1852 ± 1 Ma)1% [20]1% [4]
0.5% [10]0.5% [10]0.5% [2]0.5% [2]

[9] Heavy mineral fractions of the prepared synthetic sediment samples were then retrieved by two procedures: samples ART-1 and ART-2 were processed on Wilfley shaking table and the heavy fraction was subsequently separated in TBE and DIM heavy liquids; samples ART-3 and ART-4 were loaded directly into TBE and DIM without pre-concentration of the heavy mineral fraction on the Wilfley table. The aim of this procedure was to test the potential bias introduced by passing the samples through the Wilfley table, especially with respect to potential loss of the small size fraction and less abundant zircon populations. The zircon fraction retrieved from each synthetic sample was photographed under binocular microscope for subsequent image analysis. The digital images of samples before and after zircon separation were scaled and processed in ImageJ software (NIH, USA) and data tables with pixel area inμm2 of each individual zircon grain were exported. These data were then recalculated to diameter in μm assuming spherical shape of zircon grains. This appears to be a good approximation of grain size as most of zircon grains in the synthetic sample have spherical or near-spherical shape. The resulting grain size data were sorted into 10μm bins and used to calculate the grain-size distribution before and after the zircon separation.

[10] Following the zircon separation, approximately 200 grains from each sample were transferred in ethanol to a double-sided tape using a pipette, mounted in epoxy resin blocks and polished for laser ablation ICP-MS analysis. Transferring the zircons in ethanol using a pipette helped to achieve random grain selection without any size, color or shape preference. In order to asses the potential effects of handpicking on the selection of zircon grains, two of the sample mounts (ART-1 repl and ART-2 repl) were prepared in replicate by a different operator using an identical grain mounting procedure.

2.3. U-Pb Dating Technique

[11] Isotopic analysis of zircon in all samples by laser ablation ICP-MS followed the technique described inKošler et al. [2002] and Košler and Sylvester [2003]. A Thermo-Finnigan Element 2 sector field ICP-MS coupled to a 193 nm ArF excimer laser (Resonetics RESOlution M-50 LR) at Bergen University was used to measure Pb/U and Pb isotopic ratios in zircons. The sample introduction system was modified to enable simultaneous nebulisation of a tracer solution and laser ablation of the solid sample [Horn et al., 2000]. Natural Tl (205Tl/203Tl = 2.3871 [Dunstan et al., 1980]), 209Bi and enriched 233U and 237 Np (>99%) were used in the tracer solution, which was aspirated to the plasma in an argon - helium carrier gas mixture through an Apex desolvation nebuliser (Elemental Scientific) and a T-piece tube attached to the back end of the plasma torch. A helium gas line carrying the sample from the laser cell to the plasma was also attached to the T-piece tube.

[12] Laser was fired at repetition rate of 5 Hz and energy of 80 mJ. Linear laser rasters (30– 100 microns) were produced by repeated scanning of the leaser beam at a speed of 10 microns/second across the zircon surface. Typical acquisitions consisted of a 40 s measurement of blank followed by measurement of U and Pb signals from zircon for another 105 s. The data were acquired in time resolved - peak jumping - pulse counting mode with 1 point measured per peak for masses 202 (flyback), 203 and 205 (Tl), 206 and 207 (Pb), 209 (Bi), 233 (U), 237 (Np), 238 (U), 249 (233U oxide), 253 (237 Np oxide) and 254 (238U oxide). No common Pb correction was applied to the data due to variable 204Hg signal that interferes with 204Pb and that typically exceeded the low 204Pb signal from zircons by a factor of 5–10. Raw data were corrected for dead time of the electron multiplier and processed off line in a spreadsheet-based program (Lamdate [Košler et al., 2002]). Data reduction included correction for gas blank, laser-induced elemental fractionation of Pb and U and instrument mass bias. Minor formation of oxides of U and Np was corrected for by adding signal intensities at masses 249, 253 and 254 to the intensities at masses 233, 237 and 238, respectively. No common Pb correction was applied to the data. Residual elemental fractionation and instrumental mass bias were corrected by normalization to the Plešovice reference zircon (337 Ma [Sláma et al., 2008]). Zircon reference materials 91500 [Wiedenbeck et al., 1995] and GJ-1 [Jackson et al., 2004] were periodically analyzed during the measurement for quality control and the obtained mean values of 1066 ± 4 Ma for 91500 zircon (concordia age ± 2σ; n = 57 measurements) and 605 ± 5 Ma for GJ-1 zircon (concordia age ± 2σ; n = 40 measurements) correspond well with the published reference values (91500: 1065 Ma [Wiedenbeck et al., 1995]; GJ-1: 609 Ma [Jackson et al., 2004]). Discordance (in %) was calculated using equation of Sircombe [2000]. All U-Pb isotopic analyses of zircon are summarized in theauxiliary material. Zircon age distributions are presented as histograms combined with cumulative Gaussian probability (probability density) curves generated with the ISOPLOT program (v. 3.50 [Ludwig, 2003]; Figures 2 and 4).

Figure 2.

Distribution of zircon 206Pb/238U ages of the two main age components (source rocks) in samples from the River Dee in Scottish Highlands (see Figure 1 for the sample locations). The labels show the % fraction of zircons derived from the Grampian metapsammites, the estimated proportions of the Grampian metapsammites in the sediment are in brackets. The amount in percent of strongly discordant ages that has not been included into the plot is stated in the brackets with the abbreviation “disc.” The color coding corresponds to that in Figure 1.

3. Results

3.1. Samples of Modern Sediments and Their Source Rocks

3.1.1. Dee-1 Granite

[13] Heavy liquid separation of 50 g of the crushed sample material provided ca. 1400 zircon grains in the size range of 60–500 μm from which the majority (ca. 70%) were grains approximately 120 μm long. Separation of zircons from a second sample of the Cairngorm granite yielded ca. 850 zircon grains. Most of zircon grains from the Cairngorm granite showed evidence of metamictization; the grains were cracked and dark-colored and provided discordant ages (77% of all measurements;Figure 2). The U-Pb dating of 39 grains yielded only 9 reliable measurements (Figure 2) with the mean 206Pb/238U age of 409 ± 19 Ma, which is in agreement with the published TIMS age of 404 ± 18 Ma [Oliver et al., 2008].

3.1.2. Dee-4 Grampian Metapsammite

[14] Approximately 25 000 zircon grains (number estimated under binocular microscope by counting a subset of the whole zircon pool) in size range between 60–200 μm were recovered from 50 g of the sample Dee-4 Grampian metapsammite. Most of the grains (ca. 80%) were approximately 120μm long. Second sample of the Grampian metapsammite collected from a nearby location as the stream sediment sample Dee-2 (Figure 1) yielded approximately 20 000 zircon grains that ranged in size between 40–160 μm, from which approximately two thirds were crystals ca. 50–60 μm in length.

[15] The U-Pb dating of the sample Dee-4 Grampian metapsammite resulted in 128 ages in the range of ca. 830–2 000 Ma (Figure 2) which corresponds very well with the published detrital age data from the same sedimentary formation (Inverlair psammite [Banks et al., 2007]).

3.1.3. Stream Samples of Modern Sediments

[16] Stream sediment samples Dee-1 to Dee-7 contain detrital zircons derived from both source rock units - the Caledonian Cairngorm granite and the Grampian metapsammites (Figure 2). Significant amount (21–59%; Figure 2) of strongly discordant ages found within these samples and excluded from the data sets most probably represent metamict zircon grains derived from the Cairngorm granite. The proportions of these strongly discordant ages correlate with proportions of (metamict) zircons derived from Cairngorm granite (Figure 2).

3.2. Synthetic Detrital Sediment Sample

[17] Digital imaging of the zircon fraction prior it was added to the synthetic sediment samples, together with images of zircon grains recovered from the synthetic samples were used to calculate the zircon separation yield (Figure 3). The zircon yield from the synthetic samples ART-1, ART-3 and ART-4 was between ca. 77 to 80% (Figures 3e, 3g and 3h). Loss of part of the sample ART-2 during the separation on a Wilfley table resulted in a somewhat lower yield compared to the other three samples (64.9%;Figure 3f). The age distributions obtained from U-Pb dating of the synthetic samples are shown on probability distribution plots inFigures 4c–4h together with the expected age distributions (Figures 4a and 4b). The expected age distributions were calculated for 2000 zircon grains with age proportions identical to those in the synthetic sediment sample (Table 1) and assuming normal (Gaussian) size distribution in all age populations. The mean analytical uncertainties (standard error of the mean, SE) and reproducibilities (standard deviation, SD) obtained during measurements for each zircon component (Table 2) were used in calculation of the expected age distributions in Figure 4a. The expected age distributions in Figure 4bwere calculated using an arbitrary reproducibility of 5% (1 SD) for each age component in the sample and analytical uncertainties of individual measurements of 1% (1 SE) – these represent the best case scenario that can now be routinely achieved in laser ablation ICP-MS U-Pb dating of zircon [Horstwood, 2008].

Figure 3.

(a–d) Grain-size distribution (in green) of zircon used for preparation of the synthetic sedimentary samples, see theTable 1for the proportions of individual age fractions. Red curve shows grain-size distribution of zircon after mineral separation. Grey shading represents the loss of individual size fractions during separation. (e–h) Loss (in %) of individual zircon size fractions during mineral separation, total (cumulative) yield of zircon for all size fractions (in %) is shown in gray labels.

Figure 4.

(a) Expected age distribution for a hypothetical 100% zircon recovery from the synthetic samples; the plot is based on the measured uncertainties. (b) Expected age distribution for a hypothetical 100% zircon recovery; the plot is based on arbitrary analytical uncertainties (see text for details). (c–h) The measured age distributions obtained by U-Pb dating of the synthetic samples. Ages based on206Pb/238U and 207Pb/206Pb ratios are plotted for zircons younger and older than 1000 Ma, respectively. Labels show percentages of the major age components, which are differentiated by color coding. Color dotted reference lines represent the reference TIMS ages of individual zircon populations. Width of the bins is 24 Myr.

Table 2. Analytical Accuracy (1 Standard Deviation - SD;%) and Precision (1 Standard Error of the Mean - SE;%) Obtained for the Synthetic Sediment Samplesa
ZirconAccuracy 1 SD %Precision 1 SE %
  • a

    One SD represents the standard deviation of the whole set of data of given age population while 1 SE is the average analytical precision of the individual measurements.

  • b

    Values of the two overlapping zircon populations FC-Z5 and 9980 were calculated from data subsets divided at the arbitrary chosen value of 1125 Ma (approximate mean value of the actual ages of the two zircons).

VC # 1–2 (213 ± 1 Ma)11.57.4
Plešovice (337 ± 1 Ma)13.76.1
Seiland (531 ± 2 Ma)27.86.2
FC-Z5 (1099.3 ± 0.3 Ma)b20.71.7
9980 (1150 ± 2 Ma)b23.51.4
QGNG (1852 ± 1 Ma)23.80.7

4. Discussion

[18] Quantification of individual biases in sedimentary provenance is crucial for understanding the limits of the technique. Below we discuss some of these natural and sample handling factors that may bias the observed zircon age distribution.

4.1. Natural Mixing of Components

[19] Distribution of zircon populations in the two source rocks, i.e., zircons derived from the Cairngorm granite and zircons derived from the Grampian metapsammite along the studied section of the river Dee (Figure 1) shows an increase of the Grampian metapsammite population downstream from the granite-metapsammite boundary (Dee-2,Figure 1), although there is a significant variation between individual stream sediment samples (Figure 2). The presence of small number of Grampian detrital zircons in the stream sample Dee-1 collected within the Cairngorm granite suggests that some Grampian zircons may have been included within the granitic intrusion or that the granite – metapsammite boundary is not as sharp as it appears from the available outcrops and as documented by the geological map (cf.Figure 1). The observed contamination of the Cairngorm granite by zircons inherited from the Grampian metapsammites cannot, however, account for the scale of variation observed in the sediment samples down the stream (Figure 2). This variation is rather a result of the source rocks being located in the proximity of the sampling sites and of the highly dynamic character of the stream. The active (erosion) or passive (deposition) role of the substrate rocks and size sorting of the clasts, which is linked to the speed and strength of the stream, slope of the bed and depth of the collection site strongly influence the distribution of different zircon population in such sedimentary environment. Accordingly, it is difficult to infer the abundance of individual source rocks in the source terrain(s) from the age of zircon grains in the proximal stream sediments or sedimentary rocks. We expect that due to sediment mixing, the spatial variations in detrital age observed in proximal sediment samples became less pronounced in distal sediments further downstream. However, the complexity of the source terrain along the stream has to be taken into account. Recently Hietpas et al. [2011] demonstrated strong variability in age distributions of samples from distal part of the catchment that picks up material from variable sources. Moreover, depending on the erosion efficiency, i.e., the passive or active behavior of the river [Cawood et al., 2003] the detrital ages sourced from rocks in proximal part of the stream can be significantly suppressed in distal sediments in case the stream cuts into another rock unit with contrasting zircon ages. Since the geological link between the source rocks and the sediment is usually not preserved in older sedimentary sequences, these limits need to be considered during provenance studies and paleogeographic reconstructions [Cawood et al., 2003]. Recycling of older sedimentary rocks can further complicate the provenance studies since it is often difficult to distinguish between several magmatic rock sources and a sedimentary rocks with multiple zircon ages that were recycled into the new sediment.

4.2. Zircon Properties and Abundance in Source Rocks

[20] The natural processes of grain size sorting, break-up of zircon grains and change in the proportions of individual age populations along the stream result from the erosion and sediment transport and deposition. All these factors affect the zircon age distribution in sediments; however, they are secondary to the physical properties (size and resistance) of zircon grains and their abundance in the source rocks. Variable zircon fertility of the source rocks has been suggested byMoecher and Samson [2006] as one of the most critical parameters in detrital zircon provenance studies.

[21] In this study we found large differences in the zircon contents of the Grampian metapsammite and Cairngorm granite −50 g of granite provided approximately 1 000 grains of ca. 120 μm long zircon crystals compared to ca. 20 000 grains of the same size retrieved from the 50 g of metapsammite. Additionally, most zircons (ca. 75% of all zircon grains) in the Cairngorm granite were strongly metamict and did not yield reliable U-Pb ages. The Grampian metapsammite provides approximately 80 times more zircon grains to the stream compared to the corresponding amount of the Cairngorm granite. This estimate does not account for additional effects such as are differences in erosion rates of different rock types. Accordingly, the distribution of detrital zircons in the studied stream sediment samples does not reflect the real contributions of the source rocks to the sediment samples. For example, in case of the sample Dee-7 (Figure 2), the estimated contribution of only ca. 6% of the Grampian metapsammite to the stream sediment produced ca. 84% of the zircon grains present. This effect will be even more pronounced over larger stream distances, as the high-U metamict zircons tend to disintegrate during the transport of the grains [Hay and Dempster, 2009]. The topographic and climatic effects on erosion rates in the studied area were not part of this study. However, higher elevation and steeper slopes of the terrain built by the Cairngorm granite would suggest generally higher erosion rates for the granite compared to the metapsammite. On the other hand, the “softer” metapsammite disintegrates faster and should provide more material to the stream compared to the granite. Variations in the rates of these processes over time have no significant effect on the results of this study since the sampled stream sediments likely represent a short period in the latest history of the studied terrain. Moecher and Samson [2006] also pointed out that lack of certain zircon age population in detrital sediment does not necessarily mean that rocks of that age is not present in the source area. The source rocks may also not contain any zircons or there can be small zircon overgrowths that are rarely detected in the detrital sample.

4.3. Mineral Separation

[22] Distribution of detrital zircon ages can be significantly altered during the process of heavy mineral separation in the laboratory. The zircon yield from the sample, on the other hand, seems to be rather insensitive to the method used for mineral separation. This experiment with synthetic sediment sample shows almost constant zircon yield of 77–80% for most of the samples, except for sample ART-2 where material was accidentally lost during handling (Figure 3). In contrast, the yield of different size fractions from samples treated by different separation techniques was variable. While samples ART-1 and ART-2 (Wilfley table + heavy liquids) have lost uniformly ca. 20% of all zircon fractions between 50 and 330μm (Figures 3e and 3f), the samples ART-3 and ART-4 (heavy liquids only) have experienced almost complete loss of the smallest zircons (ca. 50–70μm) while zircons with diameter of ca. 150 μm and larger were almost fully recovered (Figures 3g and 3h). It appears that mineral separation using Wilfley shaking table in this study was responsible for almost uniform partial loss of zircon fraction below ca. 330 μm while recovery of larger zircon grains was not affected. Loss of the very small zircons (ca. 50–70 μm) during the heavy liquid separation of samples ART-3 and ART-4 (Figure 3) can possibly be attributed to electrostatic charging of the grains before the samples were loaded into the heavy liquids. This would also explain why samples ART-1 and ART-2 processed on Wilfley table prior to their separation in heavy liquids did not show a significant loss of small-size grains. This finding is in contrast toHietpas et al. [2011]who suggest that Wilfley table separation contributes most to the loss of fine-grained fraction during sample handling. Previous to this study, no experimental data that would quantify the effects of Wilfley table separation on the yield of zircon have been available [Fedo et al., 2003]. The side and forward tilt of the table, the sample grain-size, grain-shape and mineral contents, the speed of sample feeding, amount of water and volume of the sample cut taken off the table will all have some effect on the heavy mineral separation efficiency and grain size distribution. Since all operations in the sample handling are potentially prone to a partial sample loss or contamination, it is of advantage to reduce the number of separation steps. This includes avoiding the pre-concentration of heavy minerals using a Wilfley table because of the possible loss of the small zircon grain fraction. For the same reason, the use of plastic ware should be avoided to limit static charging and the potential loss of very small grains.

4.4. Grain-Size Distribution and Handpicking

[23] Results of the experiment with synthetic detrital sample suggest that zircon populations represented entirely by small zircon grains (e.g., <100 μm, Seiland zircon, Table 1) tend to be under-represented in the resulting age probability density plots (Figure 4). With the exception of sample ART-1, where the smallSeiland zircon grains are almost as abundant (21.3%; Figure 4c) as in the sample prior to the mineral separation (24.5%; Figure 4a), the remaining samples show two to threefold decrease in the abundance of this small-size zircon population (Figures 4d–4h). There is no significant correlation between the loss of the very small zircons (ca. 50–70 μm) in samples ART-3 and ART-4 (Figure 3) and the overall decrease of the small-size zircon population (<100μm; zircon Seiland, Table 1) in the age probability plots (Figure 4). This seems to reflect preference for larger zircon grains during handpicking despite the effort of both operators to select representative samples from the pool of zircon grains. The cumulative plots for the two equally represented zircon age populations (24.5%, Plešovice and Seilandzircons) that illustrate a systematic under-representing of sample consisting only of small grains (Seiland zircon) are shown in Figure 5. Overall preference for larger zircon grains (100–300 μm) during the preparation of zircon grain mount appears to be common in routine U-Pb isotopic analysis [Moecher and Samson, 2006]. While this study failed to reproduce the distribution of small grain-size population, the direct transfer of zircons grains that avoids their selective picking, e.g., by tweezers (see the Samples and technique chapter), is considered suitable for detrital zircon studies. Additional experiments will be necessary in order to evaluate whether selection of zircon grains according to their properties (shape, color, size, inclusions etc.) would provide more robust results for sedimentary provenance applications. The combination of random and nonrandom selection of zircon grains from each sample as suggested byAndersen [2005] also appears to be a good practice for achieving representative grain sampling.

Figure 5.

Cumulative plots showing appearance of the two equally represented age populations (24.5% of the total zircon load) during the U-Pb analysis of synthetic detrital sediment samples. Note the deficiency ofSeiland zircon (green dotted line), which contains only small zircon grains.

4.5. Number of Analyzed Grains

[24] Number of analyzed zircon grains needed to obtain accurate results in detrital provenance studies has been extensively discussed in the literature [Andersen, 2005; Dodson et al., 1988; Link et al., 2005; Vermeesch, 2004]. Large number of grain analyses per sample required by some of these studies can now be achieved with modern instrumentation. Two goals may be considered while analyzing detrital samples for sedimentary provenance – one is to detect the zircon age populations with low abundance in the sample and the other is to reproduce the age distribution of detrital zircons in the sediment. In order to achieve both goals at the same time, it is necessary to analyze as many as several hundreds of zircon grains per sample [Andersen, 2005; Link et al., 2005] - which is rarely the case in studies of natural sediment samples.

[25] The most often used constraint in sedimentary provenance studies is the minimum number of analyzed grains needed to avoid the loss of an age population of a given abundance at a given confidence level. In the experiment with synthetic sediment sample at least one zircon grain from the 5% population has been recovered and analyzed within the first 21 measurements of the six different synthetic sample mounts (Figure 6). It is much less compared to 60 grains suggested by Dodson et al. [1988] as a minimum number of grains that should be dated to gain 95% confidence that no age population of 5% abundance and higher is missed. In another study, Vermeesch [2004]suggested that for the same level of confidence at least 117 grains should be dated. However, this worst-case scenario of simultaneous detection of 20 equally large age fractions may not be typical of many common natural samples.Link et al. [2005] emphasize that having more than 60 analyses does not significantly increase the number of detected age populations and thus supports recommendation of Dodson et al. [1988]. Accordingly, analyzing 60 grains per sample appears to be a robust and sufficient routine for detrital zircon analysis. The effort by most authors to report 100 or more analyses per sample makes the results more robust but it may not be necessary for deciphering the sources of many natural detrital samples.

Figure 6.

Cumulative plots showing the appearance of 1% (zircon QGNG; blue solid line) and 5% (zircon VC #1–2; cyan dotted line) age populations during the U-Pb analysis of synthetic detrital sediment samples.

[26] The experiment further confirms the conclusion of Andersen [2005] that the probability of finding at least one grain of 1% population within 35 analyzed grains is 50%. Three out of the six analyzed synthetic samples (i.e., 50%) contain at least one grain of the 1% population within the first 35 measurements (Figure 6). The experiment also shows that increasing number of measured zircon grains over reasonable limit (i.e., 60 analyses) does not significantly improve the reproducibility of the age spectra. This is best documented in Figure 7, which shows deviation of individual age populations in synthetic samples from their expected (original) abundances relative to the increasing number of analyzed grains. Deviation of the Plešovice population (24.5%; yellow dashed bold line in Figure 7) from the expected abundance in all 6 synthetic samples is less than 50% after 101 analyses; in all samples except the sample ART-2 (Figure 7) the 50% limit of deviation is reached within 29 analyses. On the contrary, deviation of the Seiland population (24.5%; green dotted bold line in Figure 7) from the expected abundance in three synthetic samples (ART-1 repl, ART-2 and ART-2 repl) is larger than 50% after all grains in these samples have been analyzed. The uneven recovery of small-size zircon fractions was observed during preparation of all samples and it was the main cause of the overall deficiency of theSeiland population (all samples average deviations plot in Figure 7). The consequence of decrease of one zircon age population is inevitably linked to an increase of abundances of other populations as is for example the most abundant component in the samples - the mixedFC-Z5 + 9980 age group between ca 1.0–1.3 Ga (gray bold line in Figure 7) that makes up 45% of the zircon content. For this mixed age group the deviation in all samples but one (ART-1) stays above +15% even after all grains have been analyzed (Figure 7). This is an important limitation for studies that attempt to compare age spectra between samples. The smoothly decreasing abundances of the 1% QGNGpopulation in samples ART-2, ART-2 repl and ART-4 (Figure 7) result from a dilution effect of a single grain detected early in the sequence of analyses. Even for large number of analyses, such small population is likely to be underestimated as indicated by positively skewed binomial probability distribution calculated for low-abundance fractions [Andersen, 2005].

Figure 7.

The deviation (in %) of individual age populations in synthetic sediment samples from their expected abundance plotted as a function of the number of analyzed zircon grains. The expected abundance corresponds to zero % deviation (black reference line). The average deviations (bottom plot) were calculated using the deviations of individual samples. While the average deviations show well the general trends for abundant age populations/groups (45% and 24.5%), they can be misleading for low-abundance populations (5% and 1%) that are usually only represented by a few grains and show great variability in their abundance for individual samples. In order to smooth the curves, all plots were calculated as moving averages from subsets of 5 measurements; hence first 4 data-points are not shown. Color coding in the plots corresponds to that inFigures 46.

4.6. Separation of Adjacent Age Peaks

[27] The quality of the analyzed zircon grains (e.g., metamictization, inclusions) has an influence on the age spectra as it affects the accuracy and precision of individual U-Pb isotopic analyses. Correct data handling and calculation of analytical uncertainties [Horstwood, 2008] is crucial for obtaining reliable and reproducible results from the samples and leads to realistic estimates of age uncertainties. The underestimation of analytical uncertainties can lead to a complete separation of individual measurements that otherwise belong to one age group. On the other hand, an overestimation of analytical uncertainties can cause the adjacent age peaks to merge into one peak only [Sircombe and Hazelton, 2004].

[28] The experiment with the synthetic samples shows that the two age populations of FC-Z5 (1099.3 ± 0.3 Ma) and 9980 (1150.3 ± 2 Ma) zircon that are separated by ca. 50 myr, i.e., approximately by 4.5% cannot, with exception of sample ART-2 repl, be clearly resolved (gray shaded bars inFigures 4a and 4c–4h) by the analytical technique used and analytical uncertainties achieved in this study (Table 2). The age resolution could be further improved by improving the analytical precision but it is limited by the accuracy of LA ICP-MS U-Pb zircon dating which is presently ca. ±2% [Gehrels at al., 2008; Hanchar, 2009]. The other age groups that are temporally further apart form well separated peaks in the probability density plots (Figures 4a and 4c–4h). The mixture modeling approach of Sambridge and Compston [1994] can be employed to estimate the minimum number of age components in natural detrital samples. However, to avoid detection of spurious age probability density peaks, the sample must be representative of the sedimentary rock. Andersen [2005] suggests that even as many as 100 randomly selected grains are not sufficient for this spectrum deconvolution routine.

[29] The two diagrams (expected age distributions) in Figures 4a and 4b also show how the data quality affects the resulting age spectrum. While the two 24.5% populations (Plešovice in yellow and Seiland in green) are visually almost equal in the plot in Figure 4b (1SD = 5%, 1SE = 1% for all zircons), the peak of the Seiland zircon in Figure 4a is lower and wider compared to the Plešovice peak. This is an effect of lower precision in age determination of the 531 Ma (Seiland) population compared to the 337 Ma (Plešovice) population (Table 2) due to their different contents of radiogenic Pb. It highlights that the proportions of individual age groups must be taken into account rather then the height of the peaks when the age probability density plots are compared as the simple visual inspection of the age spectra based on the probability density distributions of ages may give a false impression of populations abundance [Sircombe and Hazelton, 2004].

5. Concluding Remarks

[30] Proportions of individual age populations in the sedimentary samples are considered to be diagnostic for the sediment source area and hence the accurate reproduction of the zircon age spectra from the sample is crucial. Natural variations in geological environments, however, preclude accurate assessment of the volume and ratio of source rocks that contribute to sediments. Besides other variables not investigated in this study (e.g., passive versus active behavior of the substrate rocks in the stream, change of relief in the source areas, size sorting during transport, variable rock resistance to weathering), the size, quality and quantity of zircon grains in the source rocks represent the most significant parameters that affect the age spectra obtained from detrital zircons. Diverse zircon fertility of the source rocks, together with variable susceptibility of detrital zircon to disintegrate during the transport, can account for several fold differences in the distribution of individual zircon age populations in the sediment compared to the distribution of corresponding rocks in the source area.

[31] While the complex natural processes affect redistribution of zircons from the source rocks into the sediments, sample handling in the laboratory, grain analysis and data processing can bias the accuracy of age distribution and has impact on how we reproduce the detrital zircon age pattern of the sediment. Partial loss of small zircon grains during mineral separation and apparent preference for larger zircon grains during handpicking can account for up to fourfold differences in the proportions of individual zircon populations compared to the zircon age distribution in the sediment. These factors affect the reproducibility of detrital zircon age spectra more significantly than the number of zircon grains analyzed per sample. Accordingly, it is advisable to reduce the number of operations during zircon separation, including the use of Wilfley shaking table.

[32] In this study, the 5% zircon age population in all samples has been identified within the first 21 measurements. In addition, the 50% probability of detecting 1% population within 35 measurements [Andersen, 2005] has been experimentally confirmed. In this study we were not able to reliably distinguish between two adjacent age peaks separated by only 4.5% but more data from different labs and techniques are needed to verify this observation. The analytical limitations, such as precision and accuracy of the dating technique, must be considered when evaluating potential overlap of zircon populations that are closely spaced in time.


[33] This study has been financed through the Statoil-funded Earth System Modeling project at Department of Earth Science, University of Bergen. Sklopísek Střeleč, a.s., Czech Republic is thanked for providing the glass sand ST 08 for the experimental part of the work. Tom Andersen and an anonymous reviewer are thanked for their helpful and constructive comments that significantly improved this manuscript.