The forensic comparison of trace amounts of soil on a pyjama top with hypersulphidic subaqueous soil from a river as evidence in a homicide cold case

Supplementary material: Scanning electron microscopy (SEM) images and energy-dispersive X-ray (EDX) spectroscopy spectra of samples mentioned in this paper are available at https://doi.org/10.6084/m9.figshare.c.4560989

On the night of 4 January 1983 or in the early hours of the morning on 5 January 1983, 10-year-old Louise Bell was abducted through her bedroom window at 5 Meadow Way, Hackham West in South Australia (Fig. 1). Her body was never found. However, 2 months after her disappearance the pyjama top worn by Louise when she disappeared was found in the front yard of a house near her home, as shown in Figure 1.

During the original South Australian Police (SAPOL) forensic investigations between 1983 and 1984 the pyjama top was ‘vacuumed’ to extract particles of soil, geological and biological materials to be characterized. Professor Cann of the University of South Australia identified significant numbers of foraminifera with an assemblage of different species in the vacuumed residue sample. Professor Cann also identified and characterized foraminifera in a range of control sediment samples taken in the Onkaparinga estuary (Fig. 1) between Ford F and Ford G during February and March 1984 (David 2016). He made comparisons with the foraminifera identified in the vacuumings taken from the pyjama top. Professor Cann re-confirmed his previous findings in Police Statement of Witness Reports dated 24 April 2015 and 26 September 2015 in which he listed the following assemblage of different species of foraminifera, which he and his group (Nash et al. 2010) have found to occur commonly in the Onkaparinga River estuary: Ammonia beccarii, Elphidium, Triloculina or Quinqueloculina and Trochammina inflata.

Diatoms were microscopically identified and photographed in the vacuumings from the pyjama top by Dr David William Cruickshanks-Boyd. Photographs of the diatoms were sent to Dr David Thomas, who had previously completed a PhD thesis on diatoms in the Onkaparinga estuary (Thomas 1978), and he was able to positively identify the various diatoms and the genus depicted in the photographs. Dr Thomas concluded that the occurrence of these diatoms were consistent with those previously identified in the Onkaparinga River but that they were not exclusive to the Onkaparinga River.

The exhibits and control soil samples from the Onkaparinga River were placed into long-term storage by SAPOL.

As part of a cold case investigation in 2012, a DNA profile was obtained from swatches of material removed from that pyjama top (Fig. 2) using advances in DNA science, principally by means of the Low Copy Number (LCN) DNA method (Mitchell et al. 2019). That profile was compared to a DNA profile obtained from the accused (Mr Dieter Pfennig) by two laboratories and neither could exclude the accused as a contributor to the DNA profile obtained from the material on the deceased's pyjama top. This information led to Pfennig being charged with the murder of Louise Bell on 19 November 2013. In January 1983, Pfennig lived approximately 10 min walking distance from the Bell residence (Fig. 1). At that time the accused was 34 years of age and his daughter knew Louise Bell as they were in the same year at Mitchell Park High School and participated in several common school and sporting activities. Pfennig was a teacher at this school at that time.

In 1983, Pfennig was also heavily involved in canoeing, especially on the Onkaparinga River, as shown in Figure 1. The Onkaparinga River estuary, which is 10.5 km long, is situated approximately 35 km south of Adelaide, at the base of a 560 km² catchment that extends from the Mt Lofty Ranges to Gulf St Vincent at Port Noarlunga (Fig. 1).

As part the ongoing cold case investigation, the Centre for Australian Forensic Soil Science (CAFSS) was approached by SAPOL in early 2015 to investigate if trace soil samples left on the pyjama top could be used to establish whether the samples had originated from control soil/geological samples taken by SAPOL from a range of control locations. Given that there was an insufficient amount of the previously collected vacuumings of soil and geological materials left to conduct further soil forensic investigations on the vacuumed residues, the questions that had to be answered by soil forensic investigators were as follows: (i) were there any remaining traces of soil particles on the pyjama top? (ii) If present, could the traces of soil particles left on the pyjama top be characterized to such an extent that they could be compared to archived soil samples collected by SAPOL investigators from known locations in the Onkaparinga River Estuary in 1983–84? (iii) If comparable, would it be possible to determine how the traces of soil particles from the Onkaparinga River estuary were transferred to the pyjama top?

The dominant soil type found in the Onkaparinga River estuary and other South Australian estuaries, are subaqueous acid sulphate soils (Fitzpatrick et al. 2008c; Fitzpatrick 2013c), which classify as ‘Hypersulphidic subaqueous soils’ in accordance with the Australian acid sulphate soil classification key (Fitzpatrick 2013c). Acid sulphate soils (ASS) are soils and sediments which either contain sulphuric acid or have the potential to form sulphuric acid because they contain iron sulphides (in particular pyrite, FeS₂) in an amount that can have significant impacts on other soil characteristics (Dent & Pons 1995). Subaqueous soils accommodate submerged soil materials in tidal settings with an arbitrary depth of up to 2.5 m below the water surface (Isbell & National Committee on Soil & Terrain 2016). Hypersulphidic material or soil has a low pH-buffering and acid-neutralizing capacity because when incubated for 8 weeks or longer, the pH will decrease to below pH 4 (Isbell & National Committee on Soil & Terrain 2016). When acid sulphate soils with hypersulphidic material dry, oxidation of pyrite may cause strong acidification (pH < 4) and form sulphuric material. Acid sulphate soils in estuaries are diverse but their geochemistry and mineralogy can be identified with high precision (e.g. Fitzpatrick et al. 2008c; Fanning et al. 2017), and as such can potentially provide important clues to provenance, transport pathways and depositional conditions. Despite the widespread occurrence of acid sulphate soils in Australia and globally (Fitzpatrick et al. 2011; Fanning et al. 2017), to our knowledge there is only one published forensic case (Fitzpatrick et al. 2008a, 2009) where acid sulphate soil has been used as forensic evidence.

The aim of soil forensic analysis is to associate earth materials taken from questioned items, such as clothing, shovels or vehicles, with a specific control location or the crime scene. Earth materials are powerful, perhaps ideal, pieces of contact trace evidence that help in criminal investigations, as outlined by Fitzpatrick (2013a, b). A wide variety of earth materials such as soils, rocks, minerals and human-made mineral particles (e.g. bricks) can be used to indicate or compare provenance, and therefore be used as intelligence and, subsequently, evidence to narrow areas of search during an investigation. Evaluative comparison of soil on one article of evidence compared to another, or compared to a known location, can and has been used as evidence in courts of law (e.g. Kugler 2003; Pye 2007; Ritz et al. 2008; Ruffell & McKinley 2008; Fitzpatrick et al. 2009; Murray 2011; Fitzpatrick & Raven 2012; Fitzpatrick 2013a, b).

Forensic earth scientists such as soil scientists and geologists are now using advanced techniques that have the ability to acquire information from very small samples. Consequently, earth forensics is being used increasingly in criminal investigations. This paper expands on preliminary published extended abstracts by Fitzpatrick et al. (2016, 2018), and provides more details on the methods used, developed and applied for the successful use of combined conventional laboratory X-ray diffraction (XRD), synchrotron micro-XRD (µ-XRD), scanning electron microscopy (SEM), pedology and transference experiments.

This paper demonstrates how pedological and laboratory approaches, especially those involving traditional XRD, synchrotron XRD and SEM methods, have been critical in developing predictive soil–regolith models, from microscopic to landscape scales, to help determine the source of trace amounts of acid sulphate soil particles on a pyjama top associated with the murder of 10-year-old Louise Bell in 1983. This investigation also highlights the critical importance of: (i) utilizing existing archived control samples collected by police in 1983 from the Onkaparinga River estuary to make comparisons with questioned samples on the pyjama top; (ii) consulting existing soil maps to provide background information on the likely nature and properties of the control soil samples collected in 1983; and (iii) conducting soil transference experiments on the pyjama-top fabric to study soil patterns transferred onto fabric when dipped in muddy water.

The Centre for Australian Forensic Soil Science (CAFSS) staff members are regularly subpoenaed to testify in court. To meet these responsibilities, CAFSS has developed guidelines for conducting criminal and environmental soil forensic investigations (Fitzpatrick & Raven 2016). The guidelines provide a systematic approach and detail the use of appropriate standard methods for sampling, characterizing and examining soils for forensic comparisons. These guidelines also assist CAFSS in its mission by ensuring efficiency and accountability in the proper handling, storage and tracking of soil evidence, which is essential to evidence collection and ultimately prosecution.

Forensic soil characterization requires a multidisciplinary approach, combining pedological (descriptive and spatial information) and analytical (mineralogical and chemical) information. When examining soil evidence, there are a range of stages involving screening soil tests (Stage 1) and then more detailed tests (stages 2 and 3), which will assist in providing more reliable answers. This final Stage 4 involves building coherent soil–landscape models of information from microscopic observations to the landscape scale, which may involve soil classification and the use of soil, geological and vegetation maps. The progression of a soil examination through each of the four stages will depend on factors such as: (i) the purpose of the investigation; (ii) amount of sample available; and (iii) the results from previous forensic soil or sediment investigations, such as in cold cases or the early stages of examination. Not all stages may be required for all investigations. New advanced analytical methods may have to be developed (e.g. in this case investigation the development of a synchrotron micro-X-ray diffraction method to analyse in situ trace amounts of soil particles on the pyjama-top fabric) and/or specific experiments may be required to be conducted such as soil transference experiments (i.e. on the pyjama-top fabric to study soil patterns transferred onto fabric when dipped in muddy water).

As stated previously, there was insufficient left-over vacuumed residue material from the pyjama top available to conduct further detailed soil forensic investigations. Through careful examination of the ‘vacuumed pyjama top’ at the Forensic Science South Australia laboratories on 3 June 2015, Fitzpatrick and Raven identified three areas with brownish stain/soil deposits, especially in the hem. as shown in Figure 2a–c. As a consequence, Ms Julie Halsall (Forensic Officer, Forensic Science South Australia) was guided by Fitzpatrick and Raven to cut the following four swatches from the pyjama top for a detailed soil forensic examination: (i) hem area with traces of a brownish stain, as shown in Figures 2a and 3; (ii) and (iii) mid right front side area with traces of sporadic brownish stains, as displayed in Figures 2b, c, 4 and 5; and (iv) rear side on the upper right with unstained areas of fabric taken as a ‘blank fabric sample’ (Figs 2d & 6).

The positive findings by Professor Cann, Dr Cruickshanks-Boyd and Dr Thomas in 1983–84 on the identification of assemblages of foraminifera and diatoms, respectively, on the pyjama top and comparisons with control sediment samples taken in the in the Onkaparinga estuary provided the impetus to conduct detailed soil forensic investigations on archived samples collected by SAPOL from the Onkaparinga estuary. Based on the previously published forensic geological and chemical reports, we were able to select a representative range of 14 known control soil and sediment samples that were collected by SAPOL in the Onkaparinga River estuary and surrounding tidal marsh floodplain area between 1983 and 1984 (Table 1). The 14 control soil samples previously collected in the Onkaparinga River estuary were used to determine the major similarities and differences between control samples, and whether they were suitable for comparison with the minute quantities of questioned soil observed on the pyjama top.

Initial visual examination and subsequent microscopical examination of the materials (Stage 1) were performed with a Wild Leitz M420 stereo microscope illuminated with Schott LED light source. A Lumenera Infinity 4 digital camera was used to photograph the samples under the stereo microscope. Larger items were photographed with a 10 megapixel Canon 40D digital camera. Low- and medium- to high-power light optical micrographs of the very small particles of questioned soil on the three swatches cut from the pyjama top are shown in Figure 3 for the hem area (front and back), and Figures 4 and 5 for mid right front side area. The close-up photographs shown in Figure 6 are of the unstained fabric area (i.e. blank fabric sample). Medium- to high-power light optical micrographs of four representative control samples from the Onkaparinga estuary are shown in Figures 7,, 8,, 9,, 10.

Morphological descriptions of all soil materials were conducted according to the USDA Field Book for Describing and Sampling Soils, Version 3.0(Schoeneberger et al. 2012) and Australian Soil and Land Survey Field Handbook (McDonald & Isbell 2009). Soil morphological descriptors such as colour, texture, quartz grain shape, ped structure, segregations (organic or brick fragments), effervescence class and water repellence class (WR) are some of the most useful properties for visual soil characterization (e.g. Fitzpatrick et al. 2003) and assessing soil conditions (e.g. Fitzpatrick et al. 1999).

Soil colour was determined on all dry questioned and control samples using Munsell soil colour notation (Munsell Soil Color Book 2009) (Table 1). Soil colour is usually the first property recorded in a morphological description of soils (and may be the only feature of significance to a layperson). Soil colour provides an indicator of drainage or redox status because soil colour relates to soil aeration and organic matter content (Fitzpatrick et al. 1999).

The following soil chemical properties were measured in this investigation: (i) electrical conductivity (1:5, soil:water) using the standard method described in Rayment & Lyons (2011); (ii) pH testing in 1:1 soil:water (pH_w) (Rayment & Lyons 2011); and (iii) pH testing after peroxide treatment (pH_POX) (Rayment & Lyons 2011). Hydrogen peroxide (H₂O₂) is a strong oxidizing agent, and is used to encourage the oxidation of sulphide minerals (principally pyrite: FeS₂) and the subsequent production of acidity. Since peroxide is a strong oxidizing agent, it can be argued that the resultant pH measured is a worst-case scenario. In nature, the presence of carbonate minerals such as calcite (CaCO₃) may neutralize any acid produced; however, in some cases the carbonate may not fully dissolve due to slow dissolution rates (reaction kinetics).

Apart from the original research of Locard (1930) and the recent work of Murray et al. (2016, 2017), there has been little research focusing on the transfer of soil particles onto textile fabrics. Most testing has involved human-made particles such as powder, glitter, glass fragments, acrylic and wool fibres (McDermott 2013; Roux & Robertson 2013). For example, Bull et al. (2006a, b) built on the experiments of Pounds & Smalldon (1975a, b, c), who had originally explored the transfer and persistence of textile fibres. Bull et al. (2006b) documented the transfer and persistence of pollen, powder and metal particulates (glitter) on different types of materials; namely acrylic, cotton, denim, nylon, polyester and wool textiles. These workers conducted tests involving flint particles being flicked on swatches of fabric by striking the flint of a ‘Clipper’ lighter. Bull et al. (2006b) concluded that the fabric weave of the material played a larger part in the transfer and persistence of particulates than did the type particulate. The transference method had an initial effect in the quantity of particles transferred, but the persistence of these particles over time tended to level out. According to Bull et al. (2006b): ‘No discernible difference was identified with any variant of material type, moisture level or grain size with regard to persistence, spreading capability, tenacity, transfer or detection during experimentation’.

A series of laboratory transference experiments involving 20 different anthropogenic and natural soil types and five fabric types were used by Murray et al. (2016, 2017) to ascertain whether some trace soil patterns could universally occur across all soils and fabrics tested. Soil mineralogy and moisture content, and irregularities on the fabric surface (such as raised seams) and appendages (such as buttons and metal buckles), had a greater influence on the resulting trace soil patterns than the five fabrics tested. This influence was also dependent on the method of soil transfer used. Several workers (e.g. Morgan & Bull 2007; Fitzpatrick et al. 2009) noted that silt and clay size fractions of soil (<50–100 µm) have a strong capacity to transfer and persist. Morgan et al. (2009) stated that:

In order for trace evidence to have a high evidential value, experimental studies, which mimic the forensic reality are of fundamental importance. Such primary level experimentation is crucial to establish a coherent body of theory concerning the generation, transfer and persistence of different forms of trace physical evidence.

This statement, together with the work of Murray et al. (2016, 2017), provided the impetus to conduct laboratory soil transference tests to investigate and quantify the nature of the soil transfer patterns on the pyjama top using small pieces of unstained pyjama-top fabric cut from the pyjama top, as shown in Figure 2d and documented in Table 2. These tests provide for a better interpretation of the soil evidence discovered on the pyjama top and especially determine if a clothed victim has been dragged across a muddy soil surface.

Two transference tests were conducted in a plastic vial by shaking (end over end by hand for 3 min every 15 min for 1 h) in distilled water a mixture of 10% of control soil sample (CAFSS_138.16b and 138.19) and small swatch samples (20 mm × 10 mm) cut from the pyjama top with little visual staining (i.e. CAFSS_138.04).

The small (20 × 10 mm) swatch samples were removed from the container and dried in air (Table 2). These two new control swatch samples were allocated the following new labels:

• CAFSS 138.5a (i.e. control soil CAFSS_138.16b transferred to swatch CAFSS_138.04).

• CAFSS 138.5b (i.e. control soil CAFSS_138.19 transferred to swatch CAFSS_138.04).

Representative sub-pieces (c. 10 × 10 mm) were cut from all of the original pieces of ‘swatch samples’ shown in Figures 2,, 3,, 4,, 5,, 6 (i.e. CAFSS_138.01–CAFSS_138.04). Smaller sub-pieces (c. 10 × 5 mm) were also cut from all the original pieces of swatch sample labelled as the ‘blank fabric sample’, as shown in Figures 2d and 6, and used to conduct the two transference tests as outlined in Table 2 with the following new allocated labels: CAFSS_138.05a and CAFSS_138.05b.

The sub-pieces of swatch samples were mounted on aluminium SEM sample stubs using araldite. Samples were coated with carbon to make them electrically conducting for SEM analysis. All coated samples were examined using a FEI Quanta 450 FEG Environmental SEM (ESEM) with an electron beam energy of 20 keV and fitted with an EDAX Apollo X SDD (silicon-drifted diode: a type of solid-state detector sensor) energy-dispersive X-ray (EDX).

Imaging was performed using the secondary electron (SE) signal when information about surface topography was required. Imaging was also performed using the backscattered electron (BSE) signal, where information about composition and phase were required. Where the elemental composition of the sample was required, characteristic X-ray signals were also collected at selected positions for qualitative analyses.

The dried control bulk soil samples (Table 1) were all dry sieved through a 50 µm sieve. The <50 µm fractions were then thoroughly ground in an agate mortar and pestle. The resulting fine powders were lightly front-pressed onto silicon low-background holders for XRD analysis. XRD patterns were collected with a PANalytical X'Pert Pro Multi-purpose Diffractometer using iron-filtered Co Kα radiation, automatic divergence slit and X'Celerator Si strip detector. XRD patterns were recorded from 3° to 80° 2θ in steps of 0.017° 2θ with a 0.5 s counting time per step for an overall counting time of approximately 35 min. Qualitative analysis was performed on the XRD data using in-house XPLOT and HighScore Plus (from PANalytical) search/match software. Mineralogical phase identification was made by comparing the measured XRD patterns with the International Centre for Diffraction Data (ICDD) database of standard diffraction patterns using computer-aided search/match algorithms. A quantitative analysis was performed on the XRD data using the commercial package SIROQUANT from Sietronics Pty Ltd. The amorphous content was determined using very poorly crystalline tridymite as an analogue for the amorphous mineral component in samples.

Investigations were conducted on the powder diffraction beamline at the Australian Synchrotron to:

• conduct a proof of concept investigation to establish if the superior sensitivity and resolution of synchrotron XRD using the multi-well-sample-cassette high-throughput stage (Fig. 11) could successfully determine the mineralogy of very small amounts of control soils (i.e. CAFSS_138.16b and CAFSS_138.19 mounted in the small (4 mm diameter) and thin (0.5 mm thick) round wells), which were previously analysed using a conventional laboratory XRD instrument at CSIRO where larger amounts of sampled had to be used;

• determine if the mineralogy of very small amounts of control soils, which have been transferred to small swatch samples (20 × 10 mm) cut from the pyjama top with no visual staining (i.e. CAFSS_138.04), could be identified in situ using the high-throughput stage shown in Figures 11 and 12; the transference test was done by shaking a mixture of 10% control soil sample with distilled water for 1 h in a plastic vial.

• determine if the mineralogy of very small amounts of questioned soil, which comprises tiny soil particles and fragments, on the pyjama top could be identified in situ using the high-throughput stage shown in Figures 11 and 12.

The two control swatch samples (20 × 10 mm) were then gently mounted over a well in the multi-well-sample-cassette high-throughput stage shown in Figures 11 and 12. An X-ray transparent film of polyimide (Kapton®: c. 25 µm thickness) was fixed around the edges of the sample to keep the swatch within each sample well.

XRD data were collected on the powder diffraction beamline using a X-ray wavelength of c. 0.8254 Å with a MYTHEN strip detector. The detector consists of 16 modules, each covering approximately 4.8° 2θ and with approximately 0.2° 2θ gap between modules. To overcome the gap between modules, the diffraction pattern is collected at two detector positions offset by approximately 0.5° 2θ. This set-up on the beamline results in each pixel covering approximately 0.00375° 2θ, with a total collection angle of approximately 80° 2θ. The two diffraction patterns are then merged into one using the JAVA program CONVAS2 (Rowles 2010).

The high-intensity X-ray beam was collimated in the vertical direction to approximately 1 mm and in the horizontal direction to approximately 3 mm for all samples.

XRD data from the control samples were collected for approximately 10 min each for the samples (5 min at each detector position).

The visual morphology comparison of the questioned and control samples was conducted using the naked eye and by low- and medium-/high-power stereo-binocular light microscopy, as shown in Figures 2,, 3,, 4,, 5,, 6,, 7,, 8,, 9,, 10 and documented in Table 1. Photographs taken under low-power light microscopy of swatch samples cut from the pyjama top clearly show the occurrence of distinct patches of questioned brownish coloured stains of soil, which extends over the hem on the both the front and back surfaces of the pyjama top (Figs 3a, c, 4a & 5a). Medium- to high-power light optical micrographs of the swatch samples show distinct particles/mineral aggregates of brown soil, mostly deeply impregnated in gaps between the fibres of the pyjama top (Figs 3b, d, 4b & 5b). In contrast, both the low- and medium-/high-power light optical micrographs shown in Figures 6a and b of the swatch sample cut from pyjama top that has no visual brownish staining on the pyjama top show no evidence of brown soil particles/mineral aggregates in the fibres of the pyjama-top fabric.

The soil morphology data shown in Table 1 indicate that the control samples can be categorized into the following three groups based on similar morphological features:

• Control samples labelled CAFSS_138.06 to CAFSS_138.15 range from sandy to loamy sandy, loose to soft consistence with very many quartz segregations and few fine shell segregations (Figs 7 & 8).

• Control samples labelled CAFSS_138.16 and CAFSS_138.17 range from sandy clay loam to loamy sand, moderately hard to soft consistence with few to common quartz segregations and no to very few fine shell segregations (Fig. 9).

• Control samples labelled CAFSS_138.18 and CAFSS_138.19 are a sandy clay loam, moderately hard to soft consistence with few fine/medium quartz segregations and no shell segregations (Fig. 10).

• Control samples labelled CAFSS_138.06–CAFSS_138.15 have approximately neutral pH_w and pH_POX values with low EC values (i.e. non-saline to slightly saline).

• Control sample labelled CAFSS_138.16 has very low pH_w and pH_POX values (pH < 3.5) with very high EC values (i.e. very saline soil). This sample is likely to have formed by the slow oxidation of pyrite to form sulphuric acid in the soil matrix which has a low pH buffer and acid neutralizing capacity. The slow oxidation process (incubation) is likely to have occurred when in 1984 the police sampled this soil under water and stored it in a wet/moist condition (i.e. without being completely dried). This slow incubation process simulated the standard incubation method used to access the occurrence of hypersulphidic material in ASS profiles (Fitzpatrick et al. 2010; IUSS Working Group WRB 2014).

• Control samples labelled CAFSS_138.17–CAFSS_138.19 have neutral to mildly acidic pH_w values and acidic pH_POX values with high EC values (i.e. very saline soil). Control soils CAFSS_138.17–CAFSS_138.138.19 have pH_w values ranging from pH 5.62 to 7.50 (Table 1; i.e. >3.5), and pH_POX values ranging from 3.41 to 4.40. It is likely that these samples have also formed by the slow oxidation of pyrite to form sulphuric acid in the soil matrix, which has a moderately high pH buffer and acid neutralizing capacity (i.e. the pH did not decrease to below pH 4). The slow oxidation process (incubation) is likely to have occurred when in 1984 the police sampled this soil under water and stored it in a wet/moist condition (i.e. without being first dried). This slow incubation process simulated the current incubation method used to access the occurrence of hyposulphidic material in ASS profiles (Fitzpatrick et al. 2010; IUSS Working Group WRB 2014).

The mineralogical composition of all the finely ground control samples from the Onkaparinga estuary (labelled CAFSS_138.06–CAFSS_138.19) are shown in Table 3. The XRD patterns for all samples are given in Fitzpatrick (2015). The XRD results showed that the mineralogy from the powdered control samples all contain similar amounts of quartz (dominant), layer silicate clays (minor and trace amounts of chlorite, smectite, kaolin and illite), feldspars (minor albite and orthoclase) and rutile (trace) but can be categorized into the following three groups based on similar mineralogy (Table 3):

• Control samples labelled CAFSS_138.06–CAFSS_138.15 have minor and trace amounts of calcite, Mg-substituted calcite, aragonite and dolomite with no pyrite and jarosite/Na-jarosite.

• Control samples labelled CAFSS_138.16 and CAFSS_138.17 have minor and trace amounts of jarosite/Na-jarosite and Mg-substituted calcite with no calcite, aragonite and dolomite.

• Control samples labelled CAFSS_138.18 and CAFSS_138.19 have trace amounts of pyrite, calcite and aragonite with no jarosite/Na-jarosite.

The presence of jarosite/Na-jarosite and pyrite is considered strongly indicative of coastal–estuarine acid sulphate soils and sediments, which have sufficient iron, sulphate and organic matter for sulphate reduction to occur (i.e. to produce pyrite under original waterlogged conditions and jarosite/Na-jarosite under oxidized conditions following sampling and storage in the police archives for 32 years) as outlined in Fanning et al. (2017). Interestingly, Taylor & Poole (1931) sampled subaqueous (submerged) soils in Lake Albert in 1930 when assessing the agricultural potential of Lake Albert, which was being considered for drainage and development for irrigated pastures and cropping. Their original 1930s soil samples were retrieved from the CSIRO Land and Water soil archive in 2007, and reanalysed for pH for comparison with the original measurements made 78 years previously (Fitzpatrick et al. 2008b). In this case, the original 1930s results can be taken as the original pH values (pH 8.5); the pH values for 2007 were much lower (pH 2–4; and contained jarosite/Na-jarosite) than when the samples were collected, confirming the acidifying effects of exposure to the atmosphere of the subaqueous acid sulphate soils (ASS) with hypersulphidic material (Fitzpatrick et al. 2008b).

A desktop assessment from available maps was conducted to determine the likelihood of the occurrence of subaqueous coastal acid sulphate soils (ASS) with hypersulphidic material in the Onkaparinga estuary. This assessment used information contained in both the regional coastal acid sulphate soil map of the Gulf St Vincent by Fitzpatrick et al. (2008b) and the online ASRIS (Soil Resource Information System; Johnston et al. 2003) website data from the Atlas of Australian Acid Sulphate Soils (Fitzpatrick et al. 2011). These acid sulphate soil maps and accompanying database information were also used to determine the risk of the formation of sulphuric material should the soil profiles be disturbed in the Onkaparinga estuary. This assessment identified the high likelihood of the occurrence of subaqueous hypersulphidic soils within the Onkaparinga estuary and the formation of sulphuric material should these soil profiles be disturbed.

It was not possible to determine the mineralogy of the small questioned samples on the pyjama top using laboratory X-ray sources.

The synchrotron XRD results showed the mineralogy from the following two control samples transferred to the pyjama top:

• CAFSS_138.16b contains quartz, chlorite, smectite, kaolin illite/mica, albite, orthoclase, rutile, pyrite, jarosite, natrojarosite, gypsum and halite (12 minerals).

• CAFSS_138.19 contains quartz, chlorite, smectite, kaolin, illite/mica, albite, orthoclase, rutile, pyrite, gypsum and halite (10 minerals).

The XRD data collected at the Australian Synchrotron enabled the mineralogy of very small amounts of the questioned soil, which comprises tiny soil particles and fragments, on the pyjama top to be identified in situ using the high-throughput stage shown in Figures 11 and 12. The XRD patterns for these samples are given in Fitzpatrick (2015) and in Supplementary Figure S2.3. It was not possible to determine the mineralogy of these small questioned samples using laboratory X-ray sources.

SEM analyses were performed on a wide range of questioned soil particles and fragments taken from various locations on the swatches. Techniques used included SE and BSE imaging, and EDX spectroscopy. All images and spectral analysis are given in the Supplementary material and are shown in Supplementary Figures S1.1–S1.55.

Observations made by SEM analysis shown in Figures 13 and 14 of questioned soil fragments (located in-between/attached to fibres in the swatch sample cut from the pyjama top with a brownish stain/soil deposit extending over the hem to the undersurface: CAFSS_138.01) clearly show the occurrence of pyrite (FeS₂), diatoms and clumps of clay (layer silicates) with high concentrations of pyrite. The SEM and combined EDX results confirm the presence of small particles of pyrite in several samples, as shown in Figures 15 and 16. Pyrite was also identified in questioned samples by synchrotron XRD analysis (Fitzpatrick 2015).

The SEM observations of the questioned samples clearly indicated that most of the questioned soil material was deeply impregnated in gaps between the fibres of the fabric (Fig. 17; Figs S1.1–S1.36). In contrast, the transference shaking experiments have indicated that most of the control soil is on the surface of the fibres of the pyjama-top fabric (Fig. 18; Figs S1.48–S1.55).

Classifying soils for a particular purpose involves the ordering of soils into groups with similar properties (Fitzpatrick 2013c). Sufficient descriptive, chemical and mineralogical (XRD) data were acquired on all the control soil samples to characterize properties and classify the soils. Based on soil morphology and soil chemical data (Table 1; Figs 7,, 8,, 9,, 10), and mineralogical data (Table 3), classification of the control soil samples was made according to the World Reference Base for Soil Resources (WRB) (IUSS Working Group 2014) and the acid sulphate soil profile classification (identification key) used in Australia (Fitzpatrick 2013c). In Table 4, the classification of each group of soil material is presented. Four of the 14 control samples that were sampled in the Onkaparinga River are acid sulphate soils, which classify as hypersulphidic subaqueous soils (CAFSS_138.16) and hyposulphidic subaqueous soils (CAFSS_138.17–CAFSS_138.19) in accordance with the Australian acid sulphate soil identification key.

The soil morphological (SEM: presence of pyrite and diatoms) and mineralogical (XRD and SEM data confirming the presence of pyrite) data of the questioned soil samples on the pyjama top indicate that they are likely to classify into groups with similar ASS properties, which are reflected in their likely similar soil classification, as indicated in Table 4. As a consequence, four of the control soils (i.e. CAFSS_138.16– CAFSS_138.19) from the Onkaparinga estuary and three of the questioned soils in the swatch samples cut from a pyjama top (i.e. CAFSS_138.01–CAFSS_138.03) classify as hypersulphidic subaqueous soils and/or hyposulphidic subaqueous soils.

Consequently, a professional judgement can be made to establish the ‘degree of comparability’, as defined in Fitzpatrick (2013a, b) and Fitzpatrick & Raven (2016), of the soil from the questioned samples, which contain pyrite, to those from the control sites by comparing soil samples that have prominent ASS features (i.e. contain or may have contained pyrite when the soils were submerged under water (i.e. subaqueous soils)). As such, the three questioned samples on/in the swatch samples cut from the pyjama top (i.e. CAFSS_138.01, CAFSS_138.02 and CAFSS_138.03) have a ‘moderately strong degree of comparability’ with the four control soils from the Onkaparinga estuary previously collected by SAPOL (i.e. CAFSS_138.16, CAFSS_138.17, CAFSS_138.18 and CAFSS_138.19).

The soil morphology, soil chemistry, soil classification and XRD (laboratory and synchrotron) examination of the three swatches cut from the pyjama top with three questioned soil samples on/in the fabric and four control soil samples with prominent acid sulphate soil material features (i.e. pyrite minerals) provides compelling evidence that they have virtually identical origins (i.e. a saline estuarine environment similar to the Onkaparinga estuary).

The primary transfer of highly saline hypersulphidic soil material to the pyjama top is likely to have originated from being immersed under water in mud (i.e. hypersulphidic subaqueous soil), which in turn then dried on the pyjama top. However, because salt (namely halite) has not been identified on the pyjama top, it is likely that the pyjama top was subsequently leached via rain events or washed in freshwater to remove the water-soluble salts.

The mineral particles in the pyjama top are mostly deeply impregnated in gaps between the fibres of the fabric, which are likely to have originated under water with some force being applied on the pyjama top. This is substantiated by transference shaking experiments where the mineral particles were observed to be dominantly located on the surface of the fabric.

Sufficient descriptive, soil chemical and mineralogical (XRD from laboratory and synchrotron X-ray sources; and SEM) data were acquired on the four questioned soil samples on/in four swatch samples cut from a pyjama top and the 14 control soil samples previously collected by SAPOL in the Onkaparinga River estuary to determine the major similarities and differences between samples. Accordingly, we were able to establish the ‘degree of comparability’ of the soil from the questioned samples, which contain pyrite, to those from the control sites by comparing soil samples that have prominent acid sulphate soil (ASS) features (i.e. contain or may have contained pyrite when the soils were submerged under water (i.e. subaqueous soils)). Therefore, four of the control soils (i.e. CAFSS_138.16, CAFSS_138.17, CAFSS_138.18 and CAFSS_138.19) from the Onkaparinga estuary and three of the questioned soils in the swatch samples cut from a pyjama top (i.e. CAFSS_138.01, CAFSS_138.02 and CAFSS_138.03) classify as being ASS (i.e. hypersulphidic subaqueous soils and hyposulphidic subaqueous soils).

The three questioned samples on/in the swatch samples cut from the pyjama top (i.e. CAFSS_138.01, CAFSS_138.02 and CAFSS_138.03) were found to have a ‘moderately strong degree of comparability’ with the four control soils from the Onkaparinga estuary previously collected by SAPOL (i.e. CAFSS_138.16, CAFSS_138.17, CAFSS_138.18 and CAFSS_138.19).

Given the significant mineralogical (pyrite), chemical (pH) and morphological (colour, presence of diatoms) similarities between the three questioned soil samples from the pyjama top and four control samples, it is highly likely that the soil on the pyjama top originated from a subaqueous ASS (i.e. hypersulphidic subaqueous soils and hyposulphidic subaqueous soils) in an estuarine/saline environment similar to the Onkaparinga estuary.

The direct transfer of subaqueous acid sulphate soils (i.e. hypersulphidic subaqueous soil) in a saline estuarine environment to the pyjama top is referred to as ‘primary transfer’ of soil material. The primary transfer of highly saline hypersulphidic soil material to the pyjama top is likely to have originated from being immersed under water in mud (i.e. hypersulphidic subaqueous soil), which in turn then dried on the pyjama top. However, because salt (namely halite) has not been identified on the pyjama top, it is likely that the pyjama top was subsequently leached via rain events or washed in freshwater to remove water-soluble salts.

The mineral particles in the pyjama top are mostly deeply impregnated in gaps between the fibres of the fabric, which are likely to have originated under water with some force being applied on the pyjama top. This is substantiated by transference shaking experiments where the mineral particles were observed to be dominantly located on the surface of the fabric.

To conclude, soil morphology, soil chemistry and mineralogical (laboratory XRD, synchrotron XRD, SEM and EDX) examination of the questioned soil samples on/in the fabric of three swatches cut from the pyjama top and four control soil samples with prominent hypersulphidic acid sulphate soil features (i.e. pyrite minerals) provides compelling evidence that they have ‘virtually identical origins’ (i.e. a saline estuarine environment similar to the Onkaparinga estuary).

The soil forensic work on this investigation was often complex and painstaking, with the CAFSS team successfully linking the very minute particles of soil containing pyrite on the victim's pyjama top and archived dried acid sulphate soil that was originally sampled under water (submerged) in the Onkaparinga River. The CAFSS report and presentations during cross examination in the Adelaide Supreme Court provided a ‘predictive, soil–regolith model, from microscopic to landscape scale’, which established that soil particles/fragments found on the victim's pyjama top are likely to have originated from the Onkaparinga estuary.

During the 2016 trial, substantial non-DNA scientific evidence was presented, which included soil and geological evidence indicating the deceased's pyjama top had been submerged in a specific area of the Onkaparinga River (David 2016). This scientific evidence was supported by a circumstantial case involving the accused's movements and whereabouts at the time of the disappearance of Louise Bell, and his words and actions to various neighbours, friends and colleagues in the years subsequent. According to Mitchell et al. (2019):

The Low Copy Number (LCN) DNA evidence was pivotal in the prosecution case but was supported by other DNA evidence and accompanied by other non-DNA scientific evidence from geology, soil and botanic experts. Samples collected from the top contained plant life (diatoms), micro-organisms (foraminifera) and minerals (glauconite and pyrite) and soil. The presence of each of these led experts to conclude that the pyjama top had been immersed in the Onkaparinga River.

Dieter Pfennig was found guilty by Supreme Court judge, Justice Michael David in the Criminal Trial by Judge Alone for the murder of Louis Bell, aged 10 years, in February 1983 (R v Pfennig – No. 2). This Supreme Court Judgement Report (David 2016) was presented in the South Australian Supreme Court on 11 November 2016, and provides background and reasons for Justice David arriving at a ‘Guilty verdict’ (see pp. 40–43 and 90 in David 2016, where references are made to the soil forensics evidence provided by CAFSS during the trial). Judge David concluded that the soil forensic evidence data from the particles on the victim's pyjama top together with other geological evidence indicated: ‘that the pyjama top was submerged in the Onkaparinga River and later washed in tap water before being deposited on a nearby property as shown in Figure 1 is proved beyond reasonable doubt’.

In 2018 Dieter Pfennig lost an appeal against his murder conviction and was given a non-parole period of 35 years for the crime.

The authors wish to acknowledge Ms Julie Halsall (Forensic Officer, Forensic Science South Australia) who cut the four swatch samples from the pyjama top under the direction of Rob Fitzpatrick, Mark Raven and Ms Natasha Mitchell (Senior Forensic Scientist, Forensic Science South Australia). In addition, we thank Natasha Mitchell for providing the photographs of the pyjama top in Figure 2. We acknowledge the help received from Detective Brevet Sergeant Karen Rankin from the Major Fraud Section, South Australia Police (SAPOL) with the retrieval of the 14 control soil samples from the SAPOL forensic archives. In addition, we acknowledge the extensive help during the investigation and subsequent judicial process from Detective Sergeant Anthony van der Stelt from the Major Crime Investigation Branch, SAPOL and Ms Sandi McDonald SC, Deputy Director of Public Prosecutions. We thank Stuart McClure (formally CSIRO Land and Water, and The University of South Australia) for the scanning electron microscopy. The authors wish to acknowledge the Australian Synchrotron proposal ID PD/10158 for beamtime and the assistance of the beamline scientist, Dr Justin Kimpton. The authors thank Greg Rinder for the art and graphics of Figure 1. We thank Dr Peter Self from CSIRO Land and Water for constructive comments on the draft manuscript.

The authors’ research is supported by CSIRO Land and Water and the University of Adelaide.

RWF: Conceptualization (Lead), Data Curation (Equal), Formal Analysis (Equal), Funding Acquisition (Lead), Investigation (Lead), Methodology (Equal), Project Administration (Lead), Resources (Lead), Supervision (Equal), Validation (Equal), Writing – Original Draft (Lead), Writing – Review & Editing (Lead); MDR: Data Curation (Supporting), Formal Analysis (Supporting), Methodology (Supporting), Resources (Supporting).