Rapid and accurate analyses of silicon and phosphorus in plants using a portable

X-ray fluorescence spectrometer

Stefan Reidinger1*, Michael H. Ramsey2, Susan E. Hartley1

1Department of Biology, University of York, York, YO10 5DD, UK

2School of Life Sciences, University of Sussex, Falmer, Brighton, BN1 9QG, UK

*Corresponding author:

Dr Stefan Reidinger

Telephone: +44 (0)1904 328590

Email: stefan.reidinger@york.ac.uk

Total word count for the main body of the text: 4270

Word count Introduction: 1351

Word count Materials and Methods: 985

Word count Results and Discussion: 1583

Number of Figures: 8

Number of Tables: 3

1

Summary

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 solubilisation and potential volatilization, whilst other methods

may require large, expensive specialised equipment.

Here, we present a rapid, safe and accurate procedure for the simultaneous,

non-consumptive 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 non-consumptive analysis of multiple elements,

including those known to be difficult to measure such as Si, in large numbers

of samples.

2

Keywords

Silicon, phosphorus, herbage, elemental analysis, portable X-ray fluorescence

spectrometry (P-XRF)

3

Introduction

The elemental analysis of plants is an important tool for biologists in

disciplines as diverse as ecology, physiology or 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 bottle-neck 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. Hence the development of new

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 here on the

measurement of phosphorus (P) and silicon (Si), both key elements for plant

biologists and 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).

Si typically constitutes between 0.1 and 5% of the dry weight of plants (Jones

& Handreck, 1967). Despite being considered a non-essential 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 (Neuman & 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

4

contribution to sustainable agriculture has recently been recognised (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, since 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, e.g. 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, 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 solubilisation

and, particularly in the case of Si, volatilization (Hoenig, 2001; Baffi et al., 2002).

5

The accuracy of Si analysis by flame-AAS can be further decreased by matrix effects

and oxide formation in the flame (Harris, 1998), whereas the performance of ICP can

suffer from the dilution of the analytes 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

corrosive 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), and considering the extensive weighing, heating, cooling and filtration

steps involved, digestion-based methods are not only hazardous but also very time

consuming. Furthermore, due 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, non-

consumptive 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 like cement, glass or metals (Guerra, 1995;

Lemberge et al., 2000), and their suitability for determining the elemental

composition of plants has been demonstrated in several studies (e.g. Evans, 1970;

6

Gladney et al., 1989; Guohui & Shouzhong, 1995; Richardson et al., 1995; Marguí et

al., 2003; Queralt et al., 2005). However, despite several advantages of XRF over

digestion-based techniques, such as its non-consumptive nature and its often higher

measurement accuracy, particular 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 1-10g),

limiting their use in studies where only small amounts of sample material is 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 very compact (the size

of a small benchtop centrifuge) so can easily be moved and require very little lab 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

7

been tested systematically, nor has a routine protocol for such measurements in plants

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 seconds. The plant material does not need to

be digested prior to analysis, making sample preparation fast, safe, convenient and

cheap. Multiple elements can be determined simultaneously for the same sample, and

the method is non-destructive 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, 2008) 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

8

composition of materials such as paint, soils or rocks (Potts & West, 2008). However,

since no such software is commercially available for the quantitative measurement of

elements in plant material, we established an empirical calibration function for Si and

for P.

To test for the linearity of the Si calibration function we used synthetic methyl

cellulose (Sigma-Aldrich, product number 274429) to simulate the plant matrix and

precipitated silica powder (Fisher Scientific, product number S/0680/53) 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

below). 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 1a). Therefore, we established the

empirical Si calibration by using synthetic calibrators only.

Synthetic P calibration material containing 0, 0.25, 0.5, 0.75 and 1% P was

prepared by spiking methyl cellulose with sodium phosphate (Sigma-Aldrich, product

number S5136). 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 1b), 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

9

CRMs to establish a robust empirical P calibration, and accounted for elemental

interference using standard procedures (see results).

Method validation

To determine the bias of the analytical method we used four different plant

CRMs for Si and three plant CRMs for P (Table 1a). 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

homogenised 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 D. caespitosa leaves under running tap water, then dried them

in a fan assisted oven at 60° C for three days. Prior to grinding, the leaves were re-

dried for 1 hour and roughly chopped using a conventional kitchen food processor.

Grinding the leaf material for 90 seconds in a Pulverisette 23 ball mill (Fritsch GmbH,

Germany) with a 5 ml stainless steel bowl and a 10mm stainless steel grinding ball at

a rate of 50 beats sec-1 resulted in a fine and non-fibrous 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, particularly in the case

10

of Si which is mainly deposited close to the tissue surface. Increased grinding effort

may reduce the size of particles and hence their surface area, thereby reducing the

emitted fluorescence (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

Since X-ray fluorescence emitted from light elements such as Si and P is of

low energy and has low penetrating power, 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 since the powders

showed good capacity to be compacted together. We pressed (if not otherwise stated)

0.7g of dried and ground material at 11 tonnes for 2 seconds using a manual hydraulic

press (Specac, Orpington, UK) and a standard 13mm diameter die, resulting in a

cylindrical pellet of around 5mm thickness. Pellets of any other size can be produced

instead as long as their diameter exceeds 12mm. 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, UK). Instrument specifications and

measurement conditions are shown in Table 2. Even though this analyser can be used

as a hand-held instrument in the field, we used it in the laboratory in conjunction with

11

a test stand (Thermo Scientific SmartStand), 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 centilitres min-1. However,

this is not essential and P-XRF analyses can also be conducted without helium,

though this may increase the value of the detection limit of the method, particular for

light elements such as silicon or phosphorus.

Si and P analysis using chemical digestion

To compare results obtained by P-XRF with those of a digestion-based

technique, 5 plant samples (one D. caespitosa sample and two samples of Lolium

perenne and Triticum aestivum) were analysed for silicon by fusing dried leaf samples

(0.5g) in sodium hydroxide followed by analysis using the colorimetric

silicomolybdate technique (Allen, 1989). Phosphorus analyses were carried out after

triple digestion of 0.25g material from 3 plant CRMs (‘Coast Grass’, ‘Alpine Grass

Mixture’ and ‘Rosa’; Table 1a), 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

12

pellets containing 0, 0.5, 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10% Si for 90 seconds 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 (Thompson, 2009). We measured the number of kcps for 10 spiked

and 10 un-spiked calibration pellets and applied a linear regression to the data set

[% Si=(5.17 × kcps )+0.91 ] .

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 4 elements, for 5 replicated pellets of each CRM (Table 1b) 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

[% P=8.25×( PAg )−0.53 ×( S

Ag )+0.14 ×( KAg )−0.14 ]. 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.

13

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 around 6 kcps per 1% Si, and around 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 minutes period for 15

different unspiked synthetic calibration pellets or 15 pellets of the CRM ‘Bush

Branches and Leaves’ (Table 1b), respectively.

Bias

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 of the validation CRM materials (Table

1a) 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 CRM values can be large, consequently violating

this assumption. Therefore, a functional relationship estimation by maximum

likelihood (FREML) analysis was applied to each data set, 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).

14

For Si, we found a good relationship with a non-significant rotational

(proportional) bias and a small significant translational (constant) bias of 0.082% m/m

(Fig. 2a), suggesting that readings for Si are slightly high, by 0.08% m/m 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 chemical and those from the P-XRF analyses (Fig. 2b). Since 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 chemical

digestion 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.

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, a considerable time saving over using separate digestion-

based methods for different elements. 15

Repeatability and precision

To evaluate the repeatability precision of the method, we prepared 10 pellets

of the HRM material (D. caespitosa leaves; mean Si = 1.03% m/m, mean P = 0.18%

m/m), 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

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 1a) 10 times over a period of 3 months. Fresh pellets were made every

month because they start to deform some months after being pressed and stored in

sealed plastic bags, resulting in an uneven pellet surface that may influence the

measured Si concentration, though we did not find any evidence for this (data not

shown). The calculated RSD of 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 (see above), demonstrating that the measurements can be reliably

reproduced over time.

Sample mass

16

X-ray analyses are usually performed on samples that are ‘infinitely thick’, i.e.

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, since 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 1g, resulting in a pellet thickness of

approximately 0.7mm per 0.1g plant powder. We did not press pellets weighing less

than 0.1g since 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 (silicon: R2=0.06, p=0.134; phosphorus; R2=0.00, p=0.864),

demonstrating that the smallest amount of ground plant material required to be

pressed into a stable pellet (~0.1g) is sufficient to obtain reliable Si and P

measurements using P-XRF. Though 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 0.15cm (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 prior to pressing it into pellets is unnecessary. In comparison,

17

analyses of P and Si with conventional, digestion-based techniques require

approximately 0.25g sample for P and 0.5g 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.

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 around 40

samples into pellets within 1 hour. The analyses of these pellets using P-XRF took

approximately 25 minutes, including sample labelling and changing samples between

measurements. This shows that it is feasible to analyse elemental concentrations in up

to 200 plant samples a day using this method, which is considerably faster than using

conventional methods based on the time consuming chemical digestion of the plant

material (as usually around 50 samples per day for one element alone).

Conclusions

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 prior to 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

18

analyses are non-consumptive and the sample can be de-aggregated, re-pressed and

re-analysed at any time, and subsequently the same samples can be re-used for the

analysis of other aspects of plant chemical composition. The fact that the sample can

be reused enables the possibility of collecting smaller sample volumes, which has

obvious advantages in minimising the time and expense of collecting, storing and

processing plant material.

Further, 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. Second, 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. Third, 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, both in-situ and in-vitro, as well as plants.

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, and may also provide a much 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.

19

Acknowledgments

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.

20

References

Allen SE. 1989. Chemical analysis of ecological materials, 2nd edn. Blackwell,

Oxford, UK.

Argyraki A, Ramsey MH, Potts PJ. 1997. Evaluation of portable X-ray fluorescence

instrumentation for in situ measurements of lead on contaminated land. Analyst

122: 743–749.

Baffi C, Bettinelli M, Beone GM, Spezia S. 2002. Comparison of different analytical

procedures in the determination of trace elements in lichens. Chemosphere 48:

299-306.

Blevins LG, Cauley TH. 2005. Fine particulate formation during switchgrass/coal

cofiring. Journal of Engineering for Gas Turbines and Power, 127: 457-463.

Cordell D, Drangert J, White S. 2009. The story of phosphorus: Global food

security and food for thought. Global Environmental Change 19: 292-305.

Datnoff LE, Snyder CH, Korndörfer GH. 2001. Silicon in agriculture. Amsterdam,

Elsevier.

Evans CC. 1970. X-ray fluorescence analysis for light elements in plant and faecal

materials. Analyst 95: 919-929.

Fauteux F, Rémus-Borel W, Menzies JG, Bélanger RR. 2005. Silicon and plant

disease resistance against pathogenic fungi. FEMS Microbiology Letters 249: 1-6.

21

Fox RL, Silvia JA, Plucknett DI, Teranishi DY. 1969. Soluble and total silicon in

sugar cane. Plant and Soil 30: 81-92.

Garbuzov M, Reidinger S, Hartley SE. 2011. Interactive effects of plant-available

soil silicon and herbivory on competition between two grass species. Annals of

Botany 108: 1355-1363.

Gladney ES, Neifert PE, Bower NW. 1989. Determination of silicon in National

Institute of Standards and Technology biological standard reference materials by

instrumental epithermal neutron activation and x-ray fluorescence spectrometry.

Analytical Chemistry 61:1834–1836.

Guerra MF. 1995. Elemental analysis of coins and glasses. Applied Radiation and

Isotopes 46: 583-588.

Guntzer F, Keller C, Meunier JD. 2010. Determination of the silicon concentration

in plant material using Tiron extraction. New Phytologist 188: 902-906.

Guohui L, Shouzhong F. 1995. Direct determination of 25 elements in dry powdered

plant materials by X-ray fluorescence spectrometry. Journal of Geochemical

Exploration 55: 75–80.

Harris DC. 1991. Quantitative Chemical Analysis, 3rd ed. Freeman, New York, NY.

p. 634.

Hattori T, Inanaga S, Araki H, An P, Orita S, Luxová M, Lux A. 2005.

Application of silicon enhanced drought tolerance in Sorghum bicolor. Physiologia

Plantarum 123: 459-466.

22

Hauptkorn S, Pavel J, Seltner H. 2001. Determination of silicon in biological

samples by ICP OES after non-oxidative decomposition under alkaline conditions.

Fresenius' Journal of Analytical Chemistry 370: 246–250.

Haysom MB, Ostatek-Boczynski ZA. 2006. Rapid, wet oxidation procedure for the

estimation of silicon in plant tissue. Communications in Soil Science and Plant

Analysis 37: 2299-2306.

Hoenig M. 2001. Preparation steps in environmental trace element analysis - facts

and traps. Talanta 54: 1021-1038.

JCGM. 2008. International Vocabulary of Metrology – Basic and General Concepts

and Associated Terms. VIM, 3rd edition, Joint Committee for Guides in Metrology

(JCGM) 200:2008 (terms 2.18, 2.21 and 2.23).

Jones LHP, Handreck KA. 1967. Silica in soils, plants and animals. Advances in

Agronomy 19: 107–149.

Lemberge P, Van Espen PJ, Vrebos BAR. 2000. Analysis of cement using low-

resolution energy-dispersive x-ray fluorescence and partial least-squares

regression. X-Ray Spectrometry 29: 297-304.

Lopez Molinero A, Martinez L, Villareal A, Castillo JR. 1998. Silicon

determination by inductively coupled plasma atomic emission spectrometry after

generation of volatile silicon tetrafluoride. Talanta 45: 1211-1217.

Marguí E, Hidalgo M, Queralt I. 2005. Multielemental fast analysis of vegetation

samples by wavelength dispersive X-ray fluorescence spectrometry: Possibilities

and drawbacks. Spectrochimica Acta Part B 60: 1363-1372.

23

Massey FP, Hartley SE. 2006. Experimental demonstration of the anti-herbivore

effects of silica in grasses: impacts on foliage digestibility and vole growth rates.

Proceedings of the Royal Society B 273: 2299-2304.

Miles TR, Miles TR Jr, Baxter LL, Bryers RW, Jenkins BM, Oden LL. 1996.

Boiler deposits from firing biomass fuels. Biomass and Bioenergy 10: 125-138.

Nayar PK, Misra AK, Patnaik S. 1975. Rapid microdetermination of silicon in rice

plant. Plant and Soil 42: 491-494.

Neumann D, zur Nieden U. 2001. Silicon and heavy metal tolerance of higher plants.

Phytochemistry 56: 685-692.

Piper CS. 1942. Soil and plant analysis: a laboratory manual of methods for the

examination of soils and the determination of the inorganic constituents of plants.

University of Adelaide, Adelaide, Australia.

Potts PJ, West M. 2008. Portable X-ray Fluorescence Spectrometry - Capabilities for

In Situ Analysis. RSC Publishing, Cambridge, U.K.

Queralt I, Ovejero M, Carvalho ML, Marques AF, Llabrés JM. 2005. Quantitative

determination of essential and trace element content of medicinal plants and their

infusions by XRF and ICP techniques. X-Ray Spectrometry 34: 213–217.

Ramsey MH, Potts PJ, Webb PC, Watkins P, Watson JS, Coles BJ. 1995. An

objective assessment of analytical method precision: comparison of ICP-AES and

XRF for the analysis of silicate rocks. Chemical Geology 124: 1-19.

24

Richardson DHS, Shore M, Hartree R, Richardson RM. 1995. The use of X-ray

fluorescence spectrometry for the analysis of plants, especially lichens, employed

in biological monitoring. Science of the Total Environment 176: 97-105.

Ripley BD, Thompson M. 1987. Regression Techniques for the Detection of

Analytical Bias. Analyst 112: 377-383.

Thompson M. 2009. Standard additions: myth and reality. Analytical Methods

Committee Technical Briefs: 37.

Wardle DA, Walker LR, Bardgett RD. 2004. Ecosystem properties and forest

decline in contrasting long-term chronosequences. Sciences 305: 509-513.

25

Figure 1: Graph of (a) silicon concentrations in silicon-spiked methyl cellulose

calibrators and (b) phosphorus concentrations in plant (CRM) and spiked cellulose

calibrators, against the measured silicon and phosphorus fluorescence intensity in kilo

counts per second (kcps), respectively. For phosphorus (b), the slopes of the

regression lines differed significantly (analysis of covariance, P < 0.0001) between

CRM (6.69) and synthetic calibrators (5.07), indicating spectral interference between

P and other elements in the plant CRM calibrators.

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 (FREML). Error bars show

95 % confidence interval of each mean.

Percentage siliconmeasured=0.082+0.932× percentage siliconcertified . (b)

Relationship between silicon concentrations of 5 grass samples measured by P-XRF

and a traditional colorimetric technique (1 measurement per sample/technique),

showing poor agreement between percentage silicon values obtained from P-XRF and

colorimetric measurements (Pearson correlation, r = 0.644, n = 5, p = 0.644).

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 (FREML). Error bars show 95 %

confidence interval of each mean.

26

Percentage phosphorusmeasured=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 P-XRF and a

traditional colorimetric technique (3 measurements per sample/technique). A good

agreement was achieved between percentage P measurements taken by P-XRF

colorimetric analyses (Pearson correlation, r = 0.979, n = 9, p < 0.001)

Figure 4: Plot of residuals for (a) silicon and (b) phosphorus versus sample mass.

27

Table 1a: Mean phosphorus or silicon concentration values (± 1 standard deviation) of

certified reference materials used for method validation, and their supplier.

Reference material

Percentage phosphorus ± 1 Std. Dev.

Percentage silicon ± 1 Std. Dev. Supplier

NCS ZC73013 ‘Spinach’

0.212 ± 0.024 % China National Analysis Center for Iron & Steel

NCS ZC73014 ‘Tea’

0.099 ± 0.008 % China National Analysis Center for Iron & Steel

NCS DC73349 ‘Bush Branches and Leaves’

0.60 ± 0.07 % China National Analysis Center for Iron & Steel

NJV 94-4 ‘Energy Grass’

2.1 ± 0.24 % Swedish 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

28

Table 1b: Mean phosphorus concentration values (± 1 standard deviation) of certified

reference materials used for establishing the empirical P calibration, and their

supplier.

Reference material Percentage phosphorus ± 1 Std. Dev. Supplier

NCS ZC73013 ‘Spinach’

0.32 ± 0.02% China National Analysis Center for Iron & Steel

NCS ZC73014 ‘Tea’

0.43 ± 0.03% China National Analysis Center for Iron & Steel

NCS DC73349 ‘Bush Branches and Leaves’

0.10 ± 0.004% China National Analysis Center for Iron & Steel

IPE 108 ‘Parsley’

0.383 ± 0.0180% Wageningen Evaluating Programs for Analytical Laboratories

IPE 638 ‘Maize’

0.200 ± 0.0117% Wageningen Evaluating Programs for Analytical Laboratories

IPE 682 ‘Wheat’

0.0855 ± 0.0088% Wageningen Evaluating Programs for Analytical Laboratories

IPE 776 ‘Lettuce’

0.611 ± 0.0451% Wageningen Evaluating Programs for Analytical Laboratories

IPE 977 ‘Angelica’

0.829 ± 0.0544% Wageningen Evaluating Programs for Analytical Laboratories

29

Table 2: P-XRF instrument specifications and measurement conditions used.

P-XRF Unit Niton XL3t900 GOLDD

Calibration Method Empirical

Target X-Ray Tube Ag

Detector Silicon Drift Detector

X-Ray Tube Voltage Si 6.2 kV

P 6.2 kV

Tube Current 100 µA

X-Ray Spot Diameter 8 mm

Primary Filter Si OFF

P OFF

Element Lines Si Ka 1.740 keV; Kb 1.838 keV

P Ka 2.015 keV; Kb 2.142 keV

30