VOL. 28 NO. 4 (2016) ARTICLE

The application of Fourier transform infrared, near infrared and X-ray fluorescence spectroscopy to soil analysis

A.H. Jean Robertson, Charles Shand and Estefania Perez-Fernandez The , Craigiebuckler, , , AB15 8QH. E-mail: [email protected]

Introduction concentration has traditionally involved encompassing fundamental vibra- Soil is critical to life on earth and provides wet oxidation, although many labora- tions of both the organic and mineral not only the key functions of food and tories now use combustion with CHN components. The mineral components energy production but also many other analysers. There is now a drive towards present in soil are predominantly sili- important functions, such as water regu- using dry chemical methods such as cates, including clay minerals (alumi- lation and carbon sequestration.1 Soil is X-ray fluorescence (XRF) and infrared nosilicates), or carbonates, but can a highly complex material, consisting of (IR) spectroscopic analysis which mini- also include other types of minerals an extremely variable mixture of minerals, mise cost, improve speed and reduce such as sulfates. Clay minerals play an organic matter, liquids and gases. Soils environmental impact and waste. At the important part in the properties of a are heterogeneous on different scales: James Hutton Institute (and the legacy soil and have free OH groups present in the landscape they are heterogene- institutes of the Macaulay Institute and in their structures3 (i.e. without hydro- ous both horizontally and vertically and Scottish Crop Research Institute) we have gen bonding). The O–H stretching generally occur in distinguishable layers a long history of expertise in soil science bands for clay minerals (approximately termed horizons, as a result of modifica- and spectroscopic techniques, particularly 3700–3500 cm–1) are therefore sharp tion of inputs and outputs over time.2 On of IR spectroscopy in the mid-IR range. bands which are highly diagnostic. the microscopic scale, soil can be even In recent years we have been develop- Si–O stretching vibrations are strong more heterogeneous, moving, for exam- ing methods for Fourier transform (FT)-IR, bands found close to 1000 cm–1, and ple, over a very short distance from a near infrared (NIR) and XRF spectroscopic those for the carbonate anion are discrete mineral grain to a soil microbe. analysis of soil and creating databases close to 1400 cm–1. When consid - Soils can range from a mineral soil with of soil spectral data and calibrations to ered in combination with the bend - almost no organic matter content to an predict soil properties. Currently we are ing vibrations, below 900 cm–1, the entirely organic soil or peat. In the context working to move from laboratory-based individual minerals present can often of productive and sustainable agricul- analysis towards the development of be identified. For example, for carbon - ture there is a need to know soil prop- methodology for field-based soil analysis ates, it is relatively easy to differentiate erties such as pH, organic matter, and using FT-IR and XRF spectroscopy. between calcite (calcium carbonate), nutrient and pollutant concentrations. aragonite (a different form of calcium Traditionally, many soil analyses have Spectroscopic analysis of carbonate) and dolomite (a calcium involved wet chemical methods with soil magnesium carbonate). Silicate miner- liquid extraction and subsequent anal- Fourier transform infrared als can also be identified from their ysis of solutions with laboratory-based The highly complex and variable bending vibrations, e.g. quartz can be instruments. For example, the determina- nature of soil is reflected in the very readily identified by a sharp doublet at tion of the micronutrient concentration of different FT-IR spectra they produce in 797 c m–1 and 779 cm–1.4,5 soil involves extraction with acids (typi- the mid-IR region (4000–400 cm–1) The organic components present cally HF/HNO3 or HNO3/HCl) and subse- arising from the fundamental vibra- largely originate from vegetation and so quent analysis by inductively coupled tions of the components present. the IR spectral features of soil organic plasma (ICP)-based methods. Similarly, Usefully, the mid-IR spectra provide matter (SOM) have much in common the determination of soil organic matter an overall chemical profile of the soil, with IR spectral features of plants, www.spectroscopyeurope.com SPECTROSCOPYEUROPE 9 VOL. 28 NO. 4 (2016) ARTICLE

although plants degrade to form other spectra8 and can be more difficult to ments with soil chemical or physical organic compounds such as humic interpret. attributes under study, for predictive and fulvic acids. A broad O–H stretch purposes.9 As opposed to FT-IR spec- between approximately 3600 cm–1 and Near infrared troscopy in the mid-IR region, which 3200 cm–1 is characteristic of SOM and While the fundamental vibrations of has mainly been used qualitatively, NIR the C–O stretch for polysaccharide is molecular bonds and functional groups spectroscopy is mostly used as a quan- found at 1030 cm–1, which can coincide occur in the mid-IR region, the NIR titative technique that provides a very with the silicate stretch if minerals are region extends between 750 nm and rapid means of simultaneously meas- also present. The C–H stretching region 2500 nm (i.e. 12,500–4000 cm–1) uring multiple physical and chemical can reveal whether long chain mole - and consists of broad absorption properties of soil. cules are present or whether the organic bands, overtones and combina- material is more polysaccharide based. tions of the fundamental absorptions X-ray fluorescence In addition, functional groups present found in the mid-IR range. NIR spec- Unlike the IR spectroscopic methods in the SOM such as those arising from tra contain absorbance bands mainly which are molecular, XRF using labo - esters, carboxylic acid, amide (protein) due to chemical bonds of C–H (fats, ratory-based instruments has been a and lignin can be identified and hence oil, alkanes), O–H (water, alcohol) and prime and widely used method for the the SOM characterised.6 Interpretation N–H (protein). Other chemical bonds determination of the elemental compo- of FT-IR spectra therefore allows a rapid may exhibit overtone bands in the sition of soils. The main elements in assessment of the relative proportions NIR region, but are generally weaker. soil expressed on a weight/weight and nature of both SOM and clay miner- NIR spectra of soil therefore reflect basis are generally in the order O, Si, als present and also the determination the SOM and clay minerals well, but Al, Fe, C, Ca, K, Mg, Na, Ti, N and P. Soil of the underlying mineralogy. provide much less information on the also contains other minor elements, Generally, FT-IR spectroscopic analy- full mineralogy of the samples than including those which are essential for sis of soil samples in the laboratory is their counterpart mid-IR FT-IR spectra. plant growth such as Cu and Zn. The carried out on dried, milled samples, Although NIR spectra do not have XRF method is based on excitation with an optimal particle size <2 µm, in the resolution of the mid-IR spectra, of the sample with a beam of X-rays order to avoid problems of reflection of NIR has the advantage of being a more resulting in the emission of secondary the radiation, poor resolution of peaks energetic radiation that penetrates X-rays with energies characteristic of and sloping baselines. An effective deeper into samples thus allowing the the elements in the sample (see, for protocol for preparation of soil samples scanning of samples with little, or no, example, Figure 3). To obtain reproduc- for FT-IR spectroscopic analysis, to allow previous preparation. Nonetheless, ible and accurate data, soil samples are the production of reliable and reproduc- samples are usually dried and sieved to generally converted into a pressed disc ible spectra, has been developed at the 1–2 mm for homogenisation purposes. or fused using a flux before pouring the James Hutton Institute.7 Historically FT-IR The most common spectral collec- melt into a mould to produce a glassy spectroscopy of soils was carried out in tion mode used is diffuse reflection, disc when cooled. These discs have a transmission, using KBr discs prepared in which the light interacts with the flat surface that is relatively homoge - from the milled soil, but more recently, sample and re-radiates diffuse reflected neous and so the sample is in a repro - diamond attenuated total reflection energy, which is detected (and plotted ducible position in relation to the X-ray (DATR) sampling accessories have as log 1/Reflectance) at a 45° angle in source and detector. With technologi- become routinely used for soil analysis, order to minimise specular reflection. cal developments in devices for the having the advantage of allowing spec- Its ease of use has given NIR spec- production and detection of X-rays, tra to be recorded of unprocessed soil, troscopy some advantage over other portable XRF point and shoot instru- automatically producing spectra of opti- techniques for soil analysis; however, ments (PXRF) have become widely mum intensity and being non-destruc- the combinations and overtones that available and used on soil samples that tive. The spectra produced are similar form the spectra conceal the sum have received little or no preparation in appearance to transmission spectra. of the numerous bands that corre - apart from drying and milling. Because Alternatively diffuse reflection or DRIFTS spond to functional groups present of the ease of use of PXRF, it is easy (diffuse reflection infrared Fourier trans- in the sample, making its interpreta- to obtain numerical data. However, form spectroscopy) can be used for soil tion more complex (Figure 1 illustrates appropriate calibration and checks are analysis. In this case the IR radiation is the FT-IR and NIR spectra of the same required to ensure accuracy of element reflected from the external and inter - peat sample). For this reason, NIR has concentrations. However, because of nal surfaces of the sample, and this is traditionally relied on multivariate data the low energies and absorption by also non-destructive and requires little analyses techniques (chemometrics) air the determination of elements with sample preparation. DRIFTS spectra in for the extraction of information and atomic mass <12 by standard PXRF is the mid-IR region have a very differ - development of multivariate calibration challenging and requires the use of ent appearance to ATR or transmission models that correlate optical measure- purging with He or use in a vacuum.

10 SPECTROSCOPYEUROPE www.spectroscopyeurope.com VOL. 28 NO. 4 (2016) ARTICLE

Figure 1. FT-IR and NIR spectra of the same highly organic soil sample with the soil profile.

Application of the by wet chemical methods for a wide dataset had % C values ranging from spectroscopic techniques range of analytes and properties, some 1% to 50%. to the National Soils by more than one method. Scottish FT-IR spectra of the milled soils Inventory of Scotland soils have a great diversity and the were recorded on a Bruker Vertex dataset The James Hutton Institute holds the samples and data for the National Soil Inventory of Scotland (NSIS), a spatial dataset of soils sampled on a 20 km grid throughout Scotland.10 Both recently sampled (NSIS2, 2007–2009) and legacy samples (NSIS1, 1978–1988) from each of the discrete horizons of a pit profile at the 183 different sites were avail- able for the spectroscopic analysis. FT-IR spectroscopic and XRF analysis was carried out on the top horizons of each pit profile, while all the pit profile samples available were analysed by NIR spectroscopy. Prior to all the anal- ysis, the soil samples were air-dried at ~30°C, passed through a 2-mm sieve to remove stones, large roots or other debris, and representative subsam- ples taken with a spinning riffler or by coning and quartering. Further sample preparation, as described above, was carried out for each individual tech- Figure 2. Portable XRF instrument on stand with sample holder and film located under the nique. The samples were also analysed radiation shield. The inset (bottom right) shows portable XRF being used in the field. www.spectroscopyeurope.com SPECTROSCOPYEUROPE 11 VOL. 28 NO. 4 (2016) ARTICLE

Figure 3. PXRF spectrum of a reference soil (left). Comparison of Fe concentration measured in mineral and organic soils by aqua regia extraction/ ICP spectrometry, or directly by PXRF (see text for details). Figure 3. PXRF spectrum of a reference soil (left). Comparison of Fe concentration measured in mineral and organic soils by aqua regia extraction / ICP spectrometry, or directly by PXRF (see text for details).70 FT-IR spectrometer using a DATR and therefore an interesting overview of The future—moving accessory over the range from the effects that different analytical tech- towards field-based 4000 cm–1 to 400 cm–1 (ATR and niques can have on calibration perfor- spectroscopic analysis baseline corrected). High quality FT-IR mance was provided.11 Although most examples of soil IR spec- spectra were achieved for the Scottish We analysed the elemental compo - troscopy are laboratory-based, another Soil Spectral Database and they were sition of the dried and milled NSIS2 advantage is the potential adaptabil- shown to be highly reproducible topsoils by PXRF, using a Bruker ity of these techniques for field-based and representative by comparisons S1TurboSD and 25-mm diameter plas- assessment of soils, a particularly rele- between spectra of recently sampled tic cups with a 4-µm thick polypro - vant application nowadays given the and legacy soils from the same sites, pylene covering12 (Figure 2). The soil increasing need for accurate and inex- with differences relating primarily to samples were also analysed for inor - pensive spatial soil data for environmen- the SOM. This provided validation of all ganic element concentrations by tal monitoring and precision agriculture. the sampling and preparation proce - aqua regia extraction/ICP spectrom- NIR spectroscopic methodology for dures used and therefore showed that etry. In relation to PXRF, mineral soils field analysis of soils is probably the it was possible to look at change in behaved differently from organic soils most advanced of all three techniques13 the soil profile over time. In addition and a different calibration to the soils discussed here, but at the James Hutton to providing a comprehensive FT-IR programme supplied with the instru- Institute we are concentrating on the spectral database for Scottish soils, ment is required for organic soils. An potential for field-based FT-IR analysis. successful calibrations have also been example, for the determination of Fe is Hand-held FT-IR instruments have achieved for prediction of a number shown in Figure 3. Aqua regia extrac- relatively recently become readily avail- of soil parameters, including soil bulk tion data is often classified as “pseudo” able, with sampling accessories for both density. total but the efficiency of extraction for DRIFTS and DATR analysis, opening up NIR spectra were recorded on 2-mm many elements is far from complete. A the possibilities for field-based FT-IR sieved soils using a Foss NIRSystems comparison of data obtained by total spectroscopic analysis of soil. In situ anal- 5000 between 1100 nm and 2500 nm. dissolution of a typical mineral and ysis is on un-milled soil in the fresh state Successful NIR spectroscopy calibra- a typical peat soil using HF and aqua so in addition to particle size and inho- tions, using a NIR spectral database regia extraction showed that Fe is only mogeneity there are also moisture issues created from the NSIS soil samples, 84% and 87% extracted by aqua regia to deal with. Generally for ATR meas - have been developed for the prediction for the mineral and peat soil, respec- urements of un-milled spectra there is of a wide range of essential soil param- tively. This discrepancy partly explains a more sloping baseline, a lower inten- eters, some of which were measured the difference between PXRF and aqua sity of most mineral bands and a greater using different analytical techniques regia data shown in Figure 3. relative intensity of bands due to clay

12 SPECTROSCOPYEUROPE www.spectroscopyeurope.com ARTICLE

2. R.R. Weil and N.C. Brady. The Nature minerals and SOM, which can actually and Properties of Soils, Fifteenth Edition. be an advantage. Currently no methods Pearson, Boston, USA (2016). Why the or robust protocols for the field-based 3. R.E. Grim, Clay Mineralogy. McGraw Hill Book Company, New York (1953). S8 TIGER? FT-IR spectroscopic analysis of soils exist 4. V.C. Farmer, The Infrared Spectra of Minerals. and we are working to develop suitable Mineralogical Society, London (1974). methodology which will allow soil analy- 5. M.J. Wilson, Clay Mineralogy: Spectroscopic and Chemical Determinative Methods. sis in the field for a range of purposes Chapman and Hall, London (1994). doi: including soil monitoring, agricultural and http://dx.doi.org/10.1007/978-94-011- forensic work. 0727-3 6. R.R.E. Artz, S.J. Chapman, J. Robertson, The technological advances in XRF J.M. Potts, F. Laggoun-Defarge, S. Gogo, L. have permitted the widespread use of Comont, J.-R. Disnar and A.-J. Francez, “FTIR affordable PXRF analysers, opening up spectroscopy can be used as a screening tool for organic matter quality in regenerat- XRF analysis to a wider audience of soil ing cutover peatlands”, Soil Biol. Biochem. and environmental scientists and replac- 40, 515–527 (2008). doi: http://dx.doi. ing the traditional methods of soil anal- org/10.1016/j.soilbio.2007.09.019 7. A.H.J. Robertson, S.J. Hillier, C. Donald and ysis involving chemical dissolution and H.R. Hill, “The NSIS Team (2013). A Robust subsequent analysis by ICP methods.14 FTIR Database for Scotland”, in Proceedings PXRF can be used in the field directly on of the 3rd Global Workshop on Proximal Soil Sensing, 26–29 May 2013, Potsdam, the soil surface (Figure 2). At the James Germany, pp. 182–185. Hutton Institute we are investigating this 8. T. Nguyen, L.J. Janik and M. Raupach, approach but surface roughness, heter- “Diffuse reflectance infrared Fourier trans- 15 form (DRIFT) spectroscopy in soil studies”, ogeneity and wetness are known to Aust. J. Soil Res. 29(1), 49–67 (1991). doi: compromise performance. However, http://dx.doi.org/10.1071/SR9910049 often in soil analysis, semi-quantitative 9. B. Stenberg, R.A. Viscarra Rossel, A.M. Mouazen and J. Wetterlind, “Visible and near data or data that reflect composition infrared spectroscopy in soil science”, Adv. changes between sampling points are Agron. 107, 163–215 (2010). doi: http:// sufficient to meet the requirements. With dx.doi.org/10.1016/S0065-2113(10)07005-7 10. A. Lilly, J. Bell, G. Hudson, A.J. Nolan and W. an analysis time of <2 min it is possible to Towers, National Soil Inventory of Scotland analyse >20 soil samples per hour. The 1 (NSIS_1): Site Location, Sampling and detection limits for hand-held instruments Profile Description Protocols (1978–1988). Technical Bulletin, Macaulay Institute, are poorer by orders of magnitude than Aberdeen, UK (2010). traditional wet chemical method involv- 11. E. Pérez-Fernández and A.H.J. Robertson, ing analysis by ICP but they are often “Global and local calibrations to predict chemical and physical properties of a adequate for measurement of elements national spatial dataset of Scottish soils from in soil that are subject to legislation their near infrared spectra”, J. Near Infrared Spectrosc. 24, 305–316 (2016). doi: http:// governing maximum concentrations. EASY OPERATION dx.doi.org/10.1255/jnirs.1229 12. C.A. Shand and R. Wendler, “Portable Acknowledgements X-ray fluorescence analysis of mineral and REASON #5: The authors would like to thank the NSIS organic soils and the influence of organic matter”, J. Geochem. Explor. 143, 31–42 Easy operation team who carried out the soil sampling, (2014). doi: http://dx.doi.org/10.1016/j. processing and analyses; Angela Main gexplo.2014.03.005 The measurement of any and Renate Wendler for their support 13. R.A. Viscarra Rossel, S.R. Cattle, A. Ortega sample is easy using the and Y. Fouad, “In situ measurements of soil ™ with the FT-IR spectroscopy and XRF colour, mineral composition and clay content intuitive TouchControl analyses, and Tony Fraser for his exper- by vis–NIR spectroscopy”, Geoderma touchscreen user interface. tise and advice in relation to the FT-IR. 150, 253–266 (2009). doi: http://dx.doi. org/10.1016/j.geoderma.2009.01.025 Just place the sample in the Finally, this work was funded by the 14. D.C. Weindorf, N. Bakr and Y.D. Zhu, magazine and select any Rural & Environment Science & Analytical “Advances in portable X-ray fluorescence routine WDXRF application Services Division of the Scottish (PXRF) for environmental, pedological, and agronomic applications”, Adv. Agron. with a touch of the quick Government. 128, 1–45 (2014). doi: http://dx.doi. org/10.1016/B978-0-12-802139-2.00001-9 start button. References 15. A.R. Schneider, B. Cances, C. Breton, M. 1. FAO and ITPS, Status of the World’s Ponthieu, X. Morvan, A. Conreux and B. www.bruker.com/ Soil Resources. Food and Agriculture Marin, “Comparison of field portable XRF and S8TIGER-5 Organization of the United Nations and aqua regia/ICPAES soil analysis and evalu- ation of soil moisture influence on FPXRF Intergovernmental Technical Panel on Soils, WDXRF Rome, Italy (2015). http://www.fao.org/ results”, J. Soil Sediment. 16, 438–448 documents/card/en/c/c6814873-efc3- (2016). doi: http://dx.doi.org/10.1007/ 41db-b7d3-2081a10ede50/ s11368-015-1252-x www.spectroscopyeurope.com SPECTROSCOPYEUROPE 13