Rapid and accurate analyses of silicon and phosphorus in plants using a portable X-ray fluorescence spectrometer


Author for correspondence:
Stefan Reidinger
Tel: +44 1904 328590
Email: stefan.reidinger@york.ac.uk


  • The elemental analysis of plant material is a frequently employed tool across biological disciplines, yet accurate, convenient and economical methods for the determination of some important elements are currently lacking. For instance, digestion-based techniques are often hazardous and time-consuming and, particularly in the case of silicon (Si), can suffer from low accuracy due to incomplete solubilization and potential volatilization, whilst other methods may require large, expensive and specialised equipment.
  • Here, we present a rapid, safe and accurate procedure for the simultaneous, nonconsumptive analysis of Si and phosphorus (P) in as little as 0.1 g dried and ground plant material using a portable X-ray fluorescence spectrometer (P-XRF).
  • We used certified reference materials from different plant species to test the analytical performance of P-XRF and show that the analysis suffers from very little bias and that the repeatability precision of the measurements is as good as or better than that of other methods.
  • Using this technique we were able to process and analyse 200 ground samples a day, so P-XRF could provide a particularly valuable tool for plant biologists requiring the simultaneous nonconsumptive analysis of multiple elements, including those known to be difficult to measure such as Si, in large numbers of samples.


The elemental analysis of plants is an important tool for biologists in disciplines as diverse as ecology, physiology and agronomy. However, despite the routine application of digestion-based analytical techniques in many laboratories, the slow and often hazardous sample digestion process can create a bottleneck in the analysis of some elements, particularly where hundreds or even thousands of samples are to be analysed, as is the case for landscape-scale experiments in ecology or the rapid screening of new crop or biofuel varieties. Thus, the development of accurate and convenient high-throughput methods for assessing elemental concentrations in plants is of high importance. Here we describe a method for the rapid, safe and accurate elemental analysis of plant material using a portable X-ray fluorescence spectrometer (P-XRF). Although we concentrate on the measurement of phosphorus (P) and silicon (Si), both key elements for plant biologists and the latter notoriously difficult to analyse, P-XRF can potentially be applied to the simultaneous analysis of all elements from atomic number 12 (magnesium) up to atomic number 60 (neodymium).

Silicon typically constitutes between 0.1% and 5% of the dry weight of plants (Jones & Handreck, 1967). Despite being considered a nonessential element for the majority of higher plant species, Si can alter plant responses to a variety of environmental stresses, for instance by increasing drought and heavy metal tolerance (Neumann & zur Nieden, 2001; Hattori et al., 2005) or by acting as a defence against herbivores and fungal diseases (Fauteux et al., 2005; Massey & Hartley, 2006; Garbuzov et al., 2011). Soil Si application can boost crop health and yield, and its potential contribution to sustainable agriculture has recently been recognized (Datnoff et al., 2001). At the same time, an increasing global demand for biofuels requires the production of new plant varieties with low Si concentrations in their herbage, because Si particles that are dangerous to human health are emitted during the burning of the plant residuals (Blevins & Cauley, 2005), and Si forms sticky deposits on metal and refractory surfaces, thereby decreasing the burners’ performance (Miles et al., 1996). To date, advances in Si research are hindered by a lack of methods available for the economical, rapid, safe and accurate determination of Si in plant material.

In contrast to Si, the role of P in plant nutrition is, and has traditionally been, the focus of intense research. Phosphorus is an essential element for all life by being part of cell structural compounds such as nucleic acids and membranes, and by playing a key role in biochemical reactions such as photosynthesis and cell signalling. Soil P deficiencies frequently occur in both natural (Wardle et al., 2004) and agricultural (Cordell et al., 2009) systems, and investigations into plant P uptake mechanisms, for example by plant mutualistic mycorrhizal fungi, are of particular interest.

The most commonly applied methods to determine Si and P are based on alkaline fusion or acid digestion of the plant material (but see Masson et al., 2007), followed by spectrometric analyses of the obtained filtrate, using atomic absorption spectrometry (AAS; e.g. Hauptkorn et al., 2001), inductively coupled plasma spectrometry (ICP, e.g. Lopez Molinero et al., 1998), or colorimetric techniques (e.g. Fox et al., 1969; Allen, 1989). However, the accuracy of all these methods depends on the total destruction of the plant matrix, a process that can lead to element losses due to incomplete solubilization and, particularly in the case of Si, volatilization (Hoenig, 2001; Baffi et al., 2002). The accuracy of Si analysis by flame-AAS can be further decreased by matrix effects and oxide formation in the flame (Harris, 1991), whereas the performance of ICP can suffer from the dilution of the analyte with a large excess of the flux required for total dissolution of Si without volatilization (e.g. lithium metaborate) (Ramsey et al., 1995). Also, the digestion of the plant matrix usually requires the handling of dangerous chemicals, such as hydrofluoric, nitric, sulphuric and perchloric acid (e.g. Piper, 1942; Nayar et al., 1975; Haysom & Ostatek-Boczynski, 2006; but see Guntzer et al., 2010); digestion-based methods are not only hazardous but also very time-consuming considering the extensive weighing, heating, cooling and filtration steps involved. Furthermore, owing to the consumptive nature of all digestion-based techniques, the sample is inevitably lost during the analytical process, potentially a major problem in studies where only small amounts of test material are available and analyses of other aspects of plant chemical composition are required, or where researchers wish to re-analyse samples at a later date.

X-ray fluorescence spectrometry (XRF) provides a much faster, safer, nonconsumptive and potentially more accurate method to determine Si and P concentrations in plant material. XRF works on the principle of excitation of inner orbital electrons by an X-ray radiation source. As the excited electrons relax to the ground state, they fluoresce, thereby ejecting photons of energy and wavelength characteristic of the atoms present. Today, XRF instruments are widely used for the elemental analysis of building materials such as cement, glass or metals (Guerra, 1995; Lemberge et al., 2000), and their suitability for determining the elemental composition of plants, including Si, has been demonstrated in several studies (e.g. Evans, 1970; Gladney et al., 1989; Handson & Shelley, 1993; Guohui & Shouzhong, 1995; Richardson et al., 1995; Marguíet al., 2005; Queralt et al., 2005). However, despite several advantages of XRF over digestion-based techniques, such as its nonconsumptive nature and its often higher measurement accuracy, particularly in the case of Si (Ramsey et al., 1995), XRF has been largely confined to industrial applications and is not routinely used by biologists for the elemental analysis of plants. This might partly be due to the higher purchasing price of XRF instruments than that for equipment typically used in digestion-based elemental analysis techniques such as AAS or ICP. Furthermore, many XRF analysers require large quantities of plant material for analysis (typically between 1 and 10 g), limiting their use in studies where only small amounts of sample material are available.

Recently, the analytical power of portable X-ray fluorescence spectrometers (P-XRFs) has increased dramatically, and P-XRFs are now frequently applied in mining, soil exploration and in the analysis of consumer goods (Potts & West, 2008). The use of P-XRF instruments in plant analyses may provide important advantages over floor-standing or benchtop XRF instruments, including their much lower purchasing price, their very low running costs and their ability to analyse small amounts of plant material. Furthermore, these instruments are compact (the size of a small benchtop centrifuge) so can easily be moved and require very little laboratory or storage space. Also, P-XRFs constitute a valuable instrument for many laboratories by allowing in situ and in vitro measurements of, for instance, the distribution of nutrients or metals in soils (Argyraki et al., 1997). However, despite the ability of P-XRF to provide an economical and practical alternative to conventional XRF analysers, and to more time-consuming and potentially inaccurate digestion-based techniques, the suitability of P-XRF for the elemental analysis of plants has not yet been tested systematically, and a routine protocol for such measurements in plants has not been established.

Here, we describe a method for the rapid and accurate determination of two elements, Si and P, in plant material through the use of a P-XRF spectrometer. The method involves a quick, simple and inexpensive laboratory-based sample preparation procedure in which dried plant material is ground, pressed into pellets and analysed by exposing the pellets to X-rays for 30 s. The plant material does not need to be digested before analysis, making sample preparation fast, safe, convenient and cheap. Multiple elements can be determined simultaneously for the same sample, and the method is nondestructive so samples can be re-analysed at a later date.

We first established an empirical calibration for Si and P, then evaluated the analytical performance of the method through calculations of measurement bias, repeatability and intermediate precision (JCGM 200, 2012) using certified reference materials (CRMs) from different plant species, and one plant house reference material. We compared Si and P concentration data obtained by the analysis of plant material using P-XRF with those obtained by a digestion-based colorimetric technique. We tested empirically whether changes in sample mass are accompanied by changes in Si and P measurement intensity.

Materials and Methods

Empirical calibration

P-XRF instruments are usually equipped with a quantitative analysis software that uses the Fundamental Parameters Method for the analysis of the elemental composition of materials such as paint, soils or rocks (Potts & West, 2008). However, no such software is commercially available for the quantitative measurement of elements in plant material, so we established an empirical calibration function for silicon (Si) and for phosphorus (P).

In order to test for the linearity of the Si calibration function we used synthetic methyl cellulose (product number 274429; Sigma-Aldrich Ltd., Gillingham, UK) to simulate the plant matrix and precipitated silica powder (product number S/0680/53; Fisher Scientific Ltd, Loughborough, UK) to spike the matrix with Si. We homogenized the spiked methyl cellulose powder by vigorous shaking and stirring to produce powders containing 0% (no silica added), 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9% and 10% Si. In XRF analysis, samples composed of several elements, such as plants, yield multiple spectral lines that can interfere with each other (see the Results section). However, initial tests showed that the fluorescence intensity emitted per unit Si did not differ between the simple matrix of the synthetic calibrators and the more complex matrices of the plant CRMs ‘Spinach’, ‘Tea’, ‘Bush Branches and Leaves’ and ‘Energy Grass’ (data not shown; Table 1). Therefore, we established the empirical Si calibration by using synthetic calibrators only.

Table 1.   Mean phosphorus or silicon concentration values (± 1 SD) of certified reference materials used for method validation, and their supplier
Reference materialPercentage phosphorus ± 1 SDPercentage silicon ± 1 SDSupplier
NCS ZC73013 ‘Spinach’ 0.212 ± 0.024China National Analysis Center for Iron and Steel
NCS ZC73014 ‘Tea’ 0.099 ± 0.008China National Analysis Center for Iron and Steel
NCS DC73349 ‘Bush Branches and Leaves’ 0.60 ± 0.07China National Analysis Center for Iron and Steel
NJV 94-4 ‘Energy Grass’ 2.1 ± 0.24Swedish University of Agricultural Sciences
IPE 101 ‘Coastal Grass’0.303 ± 0.0176 Wageningen Evaluating Programs for Analytical Laboratories
IPE 106 ‘Alpine Grass Mixture’0.396 ± 0.0201 Wageningen Evaluating Programs for Analytical Laboratories
IPE 114 ‘Rosa’0.188 ± 0.0092 Wageningen Evaluating Programs for Analytical Laboratories

Synthetic P calibration material containing 0%, 0.25%, 0.5%, 0.75% and 1% P was prepared by spiking methyl cellulose with sodium phosphate (product number S5136; Sigma-Aldrich Ltd., Gillingham, UK). Whilst testing for the linearity of the P calibration function, it became apparent that these synthetic calibrators emitted a lower fluorescence intensity per unit P than the tested plant CRMs (Table 2), and an inspection of the spectral lines showed interference between P and other elements present in the plant matrices, a common phenomenon in XRF analysis. We therefore used both synthetic and plant CRMs to establish a robust empirical P calibration, and accounted for elemental interference using standard procedures (see the Results section).

Table 2.   Mean phosphorus concentration values (± 1 SD) of certified reference materials used for establishing the empirical P calibration, and their supplier
Reference materialPercentage phosphorus
± 1 SD
NCS ZC73013 ‘Spinach’0.32 ± 0.02China National Analysis Center for Iron and Steel
NCS ZC73014 ‘Tea’0.43 ± 0.03China National Analysis Center for Iron and Steel
NCS DC73349 ‘Bush Branches and Leaves’0.10 ± 0.004China National Analysis Center for Iron and Steel
IPE 108 ‘Parsley’0.383 ± 0.0180Wageningen Evaluating Programs for Analytical Laboratories
IPE 638 ‘Maize’0.200 ± 0.0117Wageningen Evaluating Programs for Analytical Laboratories
IPE 682 ‘Wheat’0.0855 ± 0.0088Wageningen Evaluating Programs for Analytical Laboratories
IPE 776 ‘Lettuce’0.611 ± 0.0451Wageningen Evaluating Programs for Analytical Laboratories
IPE 977 ‘Angelica’0.829 ± 0.0544Wageningen Evaluating Programs for Analytical Laboratories

Method validation

In order to determine the bias of the analytical method we used four different plant CRMs for Si and three plant CRMs for P (Table 1). These CRMs were not previously used for establishing the empirical calibration and thus are independent test materials. We also included a house reference material (HRM) composed of a large homogenized sample of leaves of the grass Deschampsia caespitosa (L.) Beauv. to quantify the repeatability of the method, its intermediate measurement precision, and the minimum amount of plant material required to obtain sufficiently accurate measurements.

Preparation of HRM material

We washed the Dcaespitosa leaves under running tap water, then dried them in a fan-assisted oven at 60°C for 3 d. Before grinding, the leaves were re-dried for 1 h and roughly chopped using a conventional kitchen food processor. Grinding the leaf material for 90 s in a Pulverisette 23 ball mill (Fritsch GmbH, Idar-Oberstein, Germany) with a 5-ml stainless steel bowl and a 10-mm stainless steel grinding ball at a rate of 50 beats s−1 resulted in a fine and nonfibrous powder. Although we did not find any evidence in the present study that Si and P measurement intensities changed with increasing grinding effort (data not shown), the emitted fluorescence intensity can be affected by the particle size of the powdered material. For most elements, the emitted fluorescence intensity does increase with increasing grinding effort because the smaller the particle size of the test material, the higher its surface area. However, in the case of Si, the opposite can be true because Si is mainly deposited in epidermal plant tissues and increased grinding can cause a reduction in the relative contribution of epidermal tissues to the total surface area (Evans, 1970). For reliable comparisons between contrasting plant samples, grinding time should be adjusted according to the toughness of the plant tissue to ensure particle sizes are similar.

Pellet preparation

X-ray fluorescence emitted from light elements such as Si and P is of low energy and has low penetrating power, and hence the sample surface must be tight, flat and of equal density to obtain a repeatable photon flux from the sample to the XRF detector. We prepared the pellets without the addition of a binder because the powders showed good capacity to be compacted together. We pressed (if not otherwise stated) 0.7 g of dried and ground material at 11 tons for 2 s using a manual hydraulic press (Specac, Orpington, UK) and a standard 13-mm diameter die, resulting in a cylindrical pellet of c. 5 mm thickness. Pellets of any other size can be produced instead as long as their diameter exceeds 12 mm. We used this procedure for both the synthetic calibration and plant materials.

P-XRF spectrometer system

We performed all analyses using a commercial P-XRF instrument (Niton XL3t900 GOLDD Analyzer; Thermo Scientific, Winchester, UK). Instrument specifications and measurement conditions are shown in Table 3. Even though this analyser can be used as a hand-held instrument in the field, we used it in the laboratory in conjunction with a test stand (SmartStand, Thermo Scientific, Winchester, UK), which increases the instrument’s performance when analysing light elements with low energy fluorescence such as Si and P. To avoid signal loss by air absorption, the instrument was connected to a (portable) gas cylinder containing low-grade helium, and all measurements were carried out in a helium atmosphere with a flow rate of 70 cl min−1. However, this is not essential and P-XRF analyses can also be conducted without helium, although this may increase the value of the detection limit of the method, particularly for light elements such as Si or P.

Table 3.   Portable X-ray fluorescence spectrometer (P-XRF) instrument specifications and measurement conditions used
P-XRF unitNiton XL3t900 GOLDD
Calibration methodEmpirical
Target X-ray tubeAg
DetectorSilicon drift detector
X-ray tube voltageSi6.2 kV
P6.2 kV
Tube current10 μA
X-ray spot diameter8 mm
Primary filterSiOFF
Element linesSiKa 1.740 keV; Kb 1.838 keV
PKa 2.015 keV; Kb 2.142 keV

Si and P analysis using chemical digestion

In order to compare results obtained by P-XRF with those of a digestion-based technique, five plant samples (one Dcaespitosa sample and two samples of Lolium perenne and Triticum aestivum) were analysed for Si by fusing 0.5 g finely ground leaf material in sodium hydroxide (NaOH) at c. 600°C for 10 min, followed by analysis using the colorimetric silicomolybdate technique (Allen, 1989). Phosphorus analyses were carried out after triple digestion of 0.25 g material from three plant CRMs (‘Coast Grass’, ‘Alpine Grass Mixture’ and ‘Rosa’; Table 1), using the molybdenum blue method (Allen, 1989).

Results and Discussion

Empirical calibration

The linearity of the Si calibration function was confirmed by measuring the signal intensity in kilo counts per second (kcps) for two replicated methyl cellulose pellets containing 0%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9% and 10% Si for 90 s each (Fig. 1a). We then established an empirical Si calibration using two types of calibration materials, one containing 0% Si and one spiked to contain 10% Si. This calibration strategy is optimal for analytical systems where it can be assumed that the calibration function is linear (Ellison & Thompson, 2008). We measured the number of kcps for 10 spiked and 10 unspiked calibration pellets and applied a linear regression to the dataset inline image.

Figure 1.

Graph of (a) silicon (Si) concentrations in silicon-spiked methyl cellulose calibrators and (b) phosphorus (P) concentrations in plant (certified reference materials, CRM) and spiked cellulose calibrators, against the measured Si and P fluorescence intensity in kilo counts per second (kcps), respectively, for two replicated pellets of each calibrator type. For phosphorus (b), the slopes of the regression lines differed significantly (analysis of covariance, < 0.0001) between CRM (6.69, crosses) and synthetic calibrators (5.07, circles), indicating spectral interference between P and other elements in the plant CRM calibrators.

Whilst testing for the linearity of the P calibration function, it became apparent that the plant CRMs emitted higher fluorescence intensities per unit P than the synthetic calibrators (Fig. 1b), a phenomenon caused by spectral interference between the elements P, sulphur (S) and potassium (K) present in calibration CRMs, and silver (Ag) back-scatter from the instrument. To account for this interference, we established an empirical calibration for P by simultaneously measuring the fluorescence intensity in kcps for these four elements, for five replicated pellets of each CRM (Table 2) and synthetic calibrator, and applied a linear regression model to the data using the LINEST function in MS excel to model the P fluorescence intensity inline image. In this equation, P, Ag, S and K in parenthesis stand for the kcps values of the according elements.

Next, we uploaded the Si and the P equation of the best-fit line onto the P-XRF instrument, enabling the simultaneous analysis of both elements. Empirical calibrations for elements other than Si and P can be established and uploaded, allowing the user to measure a wide array of elements in a single plant sample within seconds.

Sensitivity and detection limit

The sensitivity of the instrument (i.e. net fluorescence intensity obtained per unit of analyte concentration), as calculated by the slope coefficient of the calibration graph was c. 6 kcps per 1% Si, and c. 5 kcps per 1% P. The detection limit was estimated as 0.014% for Si and 0.013% P, using three times the standard deviation of the percentage Si and P measured over a 10 min period for 15 different unspiked synthetic calibration pellets or 15 pellets of the CRM ‘Bush Branches and Leaves’ (Table 2), respectively.


In order to estimate the translational bias (i.e. constant over the whole analyte concentration range) and rotational bias (i.e. proportional to analyte concentration) of the method, we measured 10 pellets of each validation CRM material (Table 1) and fitted a linear functional relationship between the measured Si or P values and the certified values. Linear regression requires the predictor variable to be measured without error, but uncertainties on some CRM materials can be large (Table 1), consequently violating this assumption. Therefore, a functional relationship estimation by maximum likelihood (FREML) analysis was applied to each dataset, which provides estimates of the intercept and slope of the line and plus their standard errors that do not suffer from the biases introduced by the inappropriate use of regression (Ripley & Thompson, 1987).

For Si, we found a good relationship with a nonsignificant rotational (proportional) bias and a small significant translational (constant) bias of 0.082% (Fig. 2a), suggesting that readings for Si are slightly high, by 0.08% above the certified value of the CRMs. This translational bias may be caused by very low concentrations of Si present as an impurity in the methyl cellulose. The value could be subtracted from the Si concentrations measured in the samples to eliminate the bias and improve the accuracy, if required. Poor agreement was obtained between Si values from the digestion-based and those from the P-XRF analyses (Fig. 2b). Because we have demonstrated above that the P-XRF technique suffers from very little bias, we are confident that this lack of agreement arises from the low accuracy of the digestion-based technique, possibly due to an incomplete destruction of the plant matrix or Si volatilization during the extraction process. This suggests that the use of P-XRF for the determination of Si in plant material is not only faster and safer than conventional digestion-based techniques, but also superior in terms of measurement accuracy.

Figure 2.

(a) Estimation of the bias of the proposed analytical method for the determination of silicon by comparison between measured (10 pellets per certified reference material) and certified concentration values of reference materials using a functional relationship estimation by maximum likelihood. Error bars show 95% confidence interval of each mean. Percentage silicon measured = 0.082 + 0.932 × percentage silicon certified. (b) Relationship between silicon concentrations of five grass samples measured by portable X-ray fluorescence spectrometry (P-XRF) and a traditional digestion-based colorimetric technique (one measurement per sample/technique), showing poor agreement between the two different techniques (Pearson correlation, = 0.644, = 5, = 0.240).

We did not detect any rotational or translational bias for P (Fig. 3a), indicating that P-XRF can provide highly accurate measurements for this element. Phosphorus concentration data from the P-XRF analysis were closely correlated with those of the digestion-based analysis (Fig. 3b), suggesting that both methods can provide high accuracy. Nevertheless, P analysis using P-XRF is superior to conventional digestion-based methods in terms of safety and time expenditure, and does not lead to a loss of the sample material during the measurement process, allowing samples to be re-analysed if required. Furthermore, P-XRF potentially allows the simultaneous analysis of all elements between magnesium (atomic number 12) and neodymium (atomic number 60) within seconds, offering a considerable time saving over using separate digestion-based methods for different elements.

Figure 3.

(a) Estimation of the bias of the proposed analytical method for the determination of phosphorus by comparison between measured (10 pellets per certified reference material) and certified concentration values using a functional relationship estimation by maximum likelihood. Error bars show 95% confidence interval of each mean. Percentage phosphorus measured = 0.020 + 1.136 × percentage phosphorus certified. (b) Relationship between phosphorus concentrations of certified reference material ‘Rosa’, ‘Coast Grass’ and ‘Alpine Grass Mixture’ measured by portable X-ray fluorescence spectrometry (P-XRF) and a traditional digestion-based colorimetric technique (three measurements per sample/technique). A good agreement was achieved between the two different analysis techniques (Pearson correlation, = 0.979, = 9, < 0.001).

Repeatability and precision

In order to evaluate the repeatability precision of the method, we prepared 10 pellets of the HRM material (Dcaespitosa leaves; mean Si = 1.03%, mean P = 0.18%), measured each pellet once under identical experimental conditions and calculated the relative standard deviation (RSD). The RSD was only 2.45% for Si and 3.69% for P, which is very low considering that these values include variation due to incomplete sample homogenization, as well as variation in measurement and instrument performance. The repeatability caused by counting and instrument statistics alone was 0.63% for Si and 2.6% for P, as estimated by measuring one HRM pellet 10 times and calculating the RSD.

The intermediate measurement precision of the method over time was evaluated for Si by re-analysing one of the CRM materials (‘Bush Branches and Leaves’; Table 1) 10 times over a period of 3 months. Fresh pellets were made every month because they can start to deform some months after pelleting and being stored in sealed plastic bags, resulting in an uneven pellet surface that may influence the measured Si concentration; however, we did not find any evidence for this (data not shown). The calculated RSD of the CRM (which also includes the uncertainty due to sample preparation) was only 2.04% for Si, which is in the range of the repeatability of the method, demonstrating that the measurements can be reliably reproduced over time.

Sample mass

X-ray analyses are usually performed on samples that are ‘infinitely thick’, that is, the fluorescence emitted near the top of the sample pellet (the pellet surface not facing the detector) is completely absorbed by the sample itself and does not influence the measurements. However, as the amount of sample available for analysis is often limited in biological investigations, we empirically evaluated the minimum amount required to obtain sufficiently accurate Si and P measurements. From the HRM material we prepared 10 different types of pellets, differing in mass and thickness: 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and 1 g, resulting in a pellet thickness of c. 0.7 mm per 0.1 g plant powder. We did not press pellets weighing < 0.1 g because these pellets broke easily during handling. We analysed Si and P concentrations in four pellets of each type and plotted the concentration residuals against pellet mass to confirm that the residuals show no sign of curvature (Fig. 4a,b). Regression analyses showed that the measured Si and P intensity did not change with pellet thickness (Si: R2 = 0.06, = 0.134; P: R2 = 0.00, = 0.864), demonstrating that the smallest amount of ground plant material required to be pressed into a stable pellet (c. 0.1 g) is sufficient to obtain reliable Si and P measurements using P-XRF. Although the penetration power of secondary X-rays typically increases with the atomic number of the element in question, other studies have shown that even for very heavy elements such as lead, P-XRF fluorescence is only emitted from a maximum sample depth of 0.15 cm (Argyraki et al., 1997). This lack of detectable effects of pellet thickness on the measurements demonstrates the suitability of P-XRF to study variations in the elemental composition of plants on a relatively small scale, such as analysing the effects of leaf age or position of leaves within a plant on Si and P concentrations. Also, time-consuming weighing of the ground plant material before pressing it into a pellet is unnecessary. In comparison, analyses of P and Si with conventional, digestion-based colorimetric techniques require up to 0.25 g sample for P and 0.5 g sample for Si (Allen, 1989), but due to the destructive nature of these techniques the sample is inevitably lost and no further analyses can be carried out, nor can samples be re-analysed at a later date.

Figure 4.

Plot of residuals for (a) silicon and (b) phosphorus vs sample mass.

Sample processing and analysis time

We recorded the time required to process and analyse samples starting with ground plant material. Including the time spent cleaning the hydraulic press die between samples and labelling the sample bags, we were able to press c. 40 samples into pellets within 1 h. The analyses of these pellets using P-XRF took c. 25 min, including sample labelling and changing samples between measurements. This shows that it is feasible to analyse elemental concentrations in up to 200 plant samples per day using this method. This is considerably faster than using conventional methods based on the time-consuming chemical digestion of the plant material, which allow analysis of c. 50 samples per day for just one element.


We conclude that the use of P-XRF to analyse the elemental composition of plants is superior to digestion-based techniques for several reasons. First, the plant material does not need to be digested before analysis, thereby avoiding the time-consuming handling of expensive and hazardous chemicals. Second, XRF analyses provide measurement accuracies (both in terms of method bias and precision) that are rarely achieved by other digestion-based techniques such as AAS or ICP. Third, XRF analyses are nonconsumptive and the sample can be de-aggregated, re-pressed and re-analysed at any time; the same samples can also be re-used subsequently for the analysis of other aspects of plant chemical composition. The fact that the sample can be re-used makes it possible to collect smaller sample volumes, which has obvious advantages in minimizing the time and expense of collecting, storing and processing plant material.

Furthermore, the use of P-XRF instruments in plant analyses provides several advantages over conventional XRF analysers. First, the purchasing price of P-XRF is much lower than that for conventional analysers, and apart from very small amounts of helium used during the measurement process the analyses are cost-free. Secondly, P-XRF instruments are able to analyse smaller amounts of plant material, a prerequisite for many studies where the amount of sample material is limited. Thirdly, P-XRF instruments are very compact and easy to store, and they are a particularly valuable and versatile instrument for many laboratories because they can be used for the elemental analysis of soils as well as plants, both in situ and in vitro.

Thus, P-XRF clearly has the potential to be more accurate and convenient than digestion-based analytical techniques, particularly for difficult-to-analyse elements such as Si. They may also provide a more economical and practical alternative to conventional XRF analysers, thereby providing a significant advance for biologists requiring safe, rapid and accurate elemental analysis in plant ecology, agronomy and other areas of plant biology.


This study was funded by a grant from the Natural Environment Research Council to S.E.H. (NE/F003137/1). We are grateful to Pietro Caria and John Hurley from Niton for their support.