Clay Minerals, (2016) 51, 325–350 Correlations among the mineralogical and physical properties of halloysite nanotubes (HNTs) S T E P H E N H I L L I E R 1 , 2 , * , R I K B RY D S O N 3 , E V E LY N E D E L B O S 1 , TO N Y F R A S E R 1 , N I A G R AY 1 , H E L E N P E N D LO W S K I 1 , I A N P H I L L I P S 1 , J E A N RO B E RT S O N 1 A N D I A N W I L S O N 4 2 1 The James Hutton Institute, Craigiebuckler, Aberdeen AB15 8QH, UK Department of Soil and Environment, Swedish University of Agricultural Sciences (SLU), Uppsala SE-75007, Sweden 3 Institute for Materials Research, SCaPE, University of Leeds, Leeds LS2 9JT, UK 4 Withielgoose Farmhouse, Withiel, Bodmin, Cornwall PL30 5NW, UK (Received 10 January 2016; revised 11 May 2016; Guest editor: Jock Churchman) A B S T R AC T : Twenty one samples of relatively pure tubular halloysites (HNTs) from localities in Australia, China, New Zealand, Scotland, Turkey and the USA have been investigated by X-ray diffraction (XRD), infrared spectroscopy (IR) and electron microscopy. The halloysites occur in cylindrical tubular forms with circular or elliptical cross sections and curved layers and also as prismatic tubular forms with polygonal cross sections and flat faces. Measurements of particle size indicate a range from 40 to 12,700 nm for tube lengths and from 20 to 600 nm for diameters. Size distributions are positively skewed with mean lengths ranging from 170 to 950 nm and mean diameters from 50 to 160 nm. Cylindrical tubes are systematically smaller than prismatic ones. Features related to order/ disorder in XRD patterns e.g. as measured by a ‘cylindrical/prismatic’ (CP) index and IR spectra as measured by an ‘OH-stretching band ratio’ are related to the proportions of cylindrical vs. prismatic tubes and correlated with other physical measurements such as specific surface area and cation exchange capacity. The relationships of size to geometric form, along with evidence for the existence of the prismatic form in the hydrated state and the same 2M1 stacking sequence irrespective of hydration state (i.e. 10 vs. 7 Å) or form, suggests that prismatic halloysites are the result of continued growth of cylindrical forms. KEYWORDS: halloysite, nanotubes, HNTs, cylindrical, prismatic, polygonal, length, diameter, specific surface area. Although it is probably less abundant in nature than kaolinite, the mineral halloysite is of widespread occurrence in many soils (Dixon, 1989) and also forms important economic accumulations in deposits mainly of hydrothermal origin (Wilson & Keeling, 2016, this issue). Halloysite has fascinated investigators ever since it was first discovered; in obvious contrast to the other kaolin subgroup minerals, namely * E-mail: stephen.hillier@hutton.ac.uk DOI: 10.1180/claymin.2016.051.3.11 kaolinite, dickite and nacrite which occur primarily in planar platy or blocky forms, many halloysites adopt an unusual tubular morphology. Spheroidal halloysite is also very well documented especially from soils, but here our focus will be entirely on the tubular form of halloysite, increasingly referred to as halloysite nanotubes or ‘HNTs’ due to a rapidly expanding range of applications in a wide variety of technologies. Aside from its unusual morphologies, halloysite is also distinct amongst the other kaolin polytypes in that it is hydrated with H2O molecules positioned in the interlayer space between the fundamental 1:1 layer © 2016 The Mineralogical Society 326 S. Hillier et al. combination of tetrahedral and octahedral sheets that form the basic kaolin structure. Indeed, according to Churchman & Carr (1975) the single most important characteristic that identifies, defines, and distinguishes halloysite as a distinct kaolin mineral is the presence, or evidence of the former presence, of molecules of H2O in the interlayer space. Interlayer H2O in halloysite is exceedingly labile and in response to changing environmental conditions, both naturally in the field or subsequently in the laboratory, it is readily and irreversibly lost. In its fully hydrated state halloysite contains two interlayer H2O molecules accounting for 12.25 wt.% of the molecular formula unit which can be written as Al2Si2O5(OH)4·2H2O. With this full complement of interlayer H2O, halloysite has a primary basal spacing, as observed in XRD patterns, of ∼10 Å. Because the loss of the interlayer H2O is generally not a reversible process, depending on the handling and storage history, halloysites are most commonly observed with smaller H2O contents by mass and smaller basal spacings. In an essentially dehydrated state the layer spacing approaches the 7 Å basal spacing characteristic of all other kaolin polytypes. Due to this variability of basal spacing as a function of the degree of hydration it has been recommended that the approximate basal spacing should be appended to the name, e.g. halloysite (10 Å) and halloysite (7 Å) would signify hydrated and dehydrated specimens, respectively (Bailey, 1980). In the older literature some workers referred to fully hydrated halloysite as “endellite” and the dehydrated form as “metahalloysite”, but both of these terms are now regarded as obsolete (see Brindley (1984), p. 153, for a brief summary of halloysite nomenclature past and present). Most investigators now accept that the primary form of halloysite, i.e. at the time of its formation, is the fully hydrated 10 Å form and that other forms with less H2O and consequent smaller basal spacings have undergone some degree of postformation dehydration. Although there are earlier reports, Bates et al. (1950) are usually attributed with discovering the tubular morphology of halloysite perhaps because they immediately offered an explanation for the cylindrical form as being the result of a mechanism to compensate for the lateral misfit between a larger tetrahedral and a smaller octahedral sheet of the basic kaolin 1:1 layer type. More recently, Singh (1996) has suggested that rolling is energetically favored in halloysite because it offers the least resistance from Si-Si Coulombic repulsion. Putting aside the various explanations for the curvature of the layers and the consequent rolled or cylindrical form of halloysite, it was nonetheless evident, even in the earliest literature on halloysite, that in addition to cylindrical tubes, polygonal tubes with a prismatic structure exhibiting flat faces present another distinctive morphology of tubular halloysite (Bates & Comer, 1957; Hoffman et al., 1962; Chukhrov & Zvyagin, 1966; Bailey, 1990; Kogure et al., 2013). In the present manuscript it is demonstrated that there are systematic relationships between the XRD patterns, IR spectra, cation exchange capacity (CEC), particle size, specific surface area and porosity of tubular halloysites all of which are fundamentally related to the proportions and structures of cylindrical tubular vs. prismatic polygonal tubular forms present in any given sample. Furthermore, it is emphasized that prismatic forms are not rare morphologies, but occur widely in many halloysite samples. The understanding presented of the interrelation of form (cylindrical vs. prismatic) and properties should be useful in the wide variety of emerging applications of HNTs and the relationships demonstrated may also help to constrain models of tubular halloysite formation and growth. M AT E R I A L S A N D M E T H O D S Twenty one different samples of tubular halloysite were studied. The samples were obtained from halloysite occurrences in Australia, China, New Zealand, Scotland, Turkey and the USA including various commercially exploited deposits, (Table 1). Sample 1Au is from the Eucla basin in S. Australia and is equivalent to sample CLA-1 described previously by Keeling et al. (2011) and Pasbakhsh et al. (2013). Six samples originate from halloysite deposits in China, including the Zunyi region (2Ch), Dafang region (3Ch, 4Ch), the Bifa deposit (5Ch), the Wan Jiar deposit (6Ch) (Wilson, 2004; Wilson & Keeling, 2016, this issue), and one sample (7Ch) is from an unknown location in China. The former were collected by one of the current authors (Ian Wilson), and the latter was supplied by IMERYS, courtesy of Mr Jeremy Hooper. Three samples are from New Zealand. The sample supplied by IMERYS (8NZ) is believed to originate from the well-known Matauri Bay deposit, North Island, as are the other two samples. That supplied by Jock Churchman (9NZ) was also described by Joussein et al. (2005) where it corresponds to the sample shown in fig. 5a (incorrectly captioned as 5c in Joussein et al., 2005; Churchman pers. comm. 2008), whilst the exact source of the third sample (10NZ) is unknown. The sample from Scotland (11Sc) is from Mineralogical and physical properties of HNTs 327 TABLE 1. Origin and provenance of halloysite samples investigated. Sample Country Location Source of samples 1Au 2Ch 3Ch 4Ch 5Ch 6Ch 7Ch 8NZ 9NZ 10NZ 11Sc 12Tu 13Tu 14US 15US 16US 17US 18US 19US 20US 21US Australia China China China China China China New Zealand New Zealand New Zealand Scotland Turkey Turkey USA USA USA USA USA USA USA USA Ecula Basin, S. Australia Zunyi region, Guizhou Province Dafang region, Guizhou Province Dafang region, Guizhou Province Bifa deposit, Yunnan Province Wan Jiar depoist, Yunnan Province Unknown location Matauri Bay, North Island Matauri Bay, North Island Matauri Bay, North Island Hospital Quarry, Elgin, Scotland Unknown location Balikesir region Dragon Mine, Utah Dragon Mine, Utah Dragon Mine, Utah Dragon Mine, Utah Dragon Mine, Utah Dragon Mine, Utah Dragon Mine, Utah Bovill deposit, Idaho John Keeling Ian Wilson Ian Wilson Ian Wilson Ian Wilson Ian Wilson IMERYS IMERYS Jock Churchman Unknown Heddle’s halloysite IMERYS Unknown Applied Minerals. Inc Applied Minerals. Inc Applied Minerals. Inc Applied Minerals. Inc Applied Minerals. Inc IMERYS Sigma Aldrich, Lot No MKBQ2512V Iminerals Inc the Macaulay Collection at the James Hutton Institute; it originates from Hospital Quarry, Elgin, as described previously by Hillier & Ryan (2002) and was collected by Heddle (1882); a counterpart ‘Spec. 493.1’ of this sample is held in the National Museum of Scotland in Edinburgh. A recent visit to the site showed that this halloysite occurs in a brecciated zone of Permian sandstone. Two samples were obtained from deposits in Turkey, one (12Tu) of unknown provenance was provided by IMERYS whilst the other (13Tu) originates from the Balikesir region, halloysite deposits of which have been described previously by Ece & Schroeder (2007). The remaining eight samples are all from the USA. Seven originate from the Dragon Mine in Utah, five of which were supplied by Applied Minerals, who now own and operate the mine. Additional samples from the Dragon mine include one (19US) supplied by IMERYS, and one which can be purchased from Sigma Aldrich (20US). The remaining sample (21US) from the USA is from the Bovill deposit, Idaho, supplied courtesy of IMinerals Inc. Chemical analyses (by X-ray Fluorescence – XRF), all using fused glass beads but made in a variety of different laboratories, were available for all samples. Due to differences in reporting, such as including or not including loss on drying or loss on ignition, for ease of comparison the analyses are compiled in Table 2 normalized to a theoretical kaolinite composition. X-ray diffraction For powder XRD analyses, 3 g samples were McCrone milled for 12 min in ethanol and the resulting slurry spray dried at 60°C directly from the mill to form random powder specimens (Hillier, 1999). Note that drying at 60°C from ethanol results in all halloysites dehydrating to the 7 Å form. Specimens were loaded into a Siemens D5000 X-ray diffractometer and scanned between 2 and 70°2θ, counting for 2 s per 0.02° step using Co-Kα radiation selected by a diffracted-beam monochromator. Diffraction patterns were also recorded from 3 to 70°2θ in 0.0167° steps on a Panalytical Xpert Pro diffractometer using Ni-filtered Cu radiation and counting for 100 s/step using an Xcelerator detector. Peak positions, intensities and peak width as Full Width at Half Maximum (FWHM) were measured using Bruker Diffrac Eva software, using data obtained from the D5000. Quantitative mineralogical analysis of all samples was made by a 45.91 47.37 46.60 45.77 45.33 45.93 46.39 47.24 48.20 50.13 45.81 44.97 46.92 46.39 46.05 45.78 46.23 45.58 46.26 45.73 48.81 46.55 1Au 2Ch 3Ch 4Ch 5Ch 6Ch 7Ch 8NZ 9NZ 10NZ 11Sc 12Tu 13Tu 14US 15US 16US 17US 18US 19US 20US 21US Kaolinite 38.91 37.97 39.37 39.63 39.87 39.89 39.15 38.20 37.43 35.52 36.16 39.25 38.32 39.43 39.75 37.55 39.08 39.93 39.07 39.60 35.46 39.50 Al2O3 0.09 0.05 0.00 0.01 0.07 0.00 0.00 0.10 0.09 0.07 0.19 0.20 0.04 0.00 0.00 0.10 0.00 0.00 0.02 0.01 0.14 0.00 TiO2 0.86 0.41 0.05 0.09 0.34 0.11 0.28 0.40 0.25 0.26 3.02 1.12 0.64 0.10 0.07 2.52 0.43 0.00 0.51 0.22 0.95 0.00 Fe2O3 0.00 0.01 0.00 0.02 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.02 0.00 Mn3O4 0.17 0.03 0.00 0.09 0.05 0.00 0.04 0.08 0.00 0.00 0.23 0.09 0.05 0.07 0.05 0.00 0.08 0.34 0.02 0.07 0.16 0.00 MgO 0.00 0.18 0.02 0.25 0.17 0.11 0.13 0.01 0.00 0.07 0.13 0.12 0.06 0.09 0.13 0.09 0.22 0.11 0.16 0.36 0.23 0.00 CaO 0.09 0.03 0.00 0.03 0.20 0.00 0.04 0.00 0.00 0.00 0.48 0.30 0.02 0.00 0.00 0.00 0.00 0.03 0.00 0.04 0.22 0.00 K2O 0.00 0.00 0.00 0.15 0.00 0.00 0.00 0.00 0.07 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.00 0.02 0.05 0.00 Na2O All data normalized to theoretical kaolinite basis (86 wt.%) which is also shown for comparison. Note that elements beyond Si and Al may occur variably as isomorphous substitutions (e.g. Fe), exchangeable cations (e.g. Mg, Ca, K, Na) or in associated impurities. SiO2 Sample TABLE 2. Chemical compositions (wt.%, XRF) of the samples. 328 S. Hillier et al. Mineralogical and physical properties of HNTs full-pattern fitting method using experimental patterns as standards for all minerals identified. The method is as described by participant 18 in the 3rd (2006) Reynolds Cup competition (quantitative clay mineralogy) (Omotoso et al., 2006), but without the addition of an internal standard. For halloysite, standard patterns were extracted from the sample set, based on the best examples and on verification of purity by multiple analytical methods. Patterns for all other minerals were obtained from examples held in the Macaulay mineral collection. In all cases extensive cross checks on purity were performed. The presence or absence of kaolinite in the samples was also verified by formamide intercalation (Churchman et al., 1984). Samples were ground gently under a small amount of water to produce a slurry that was pipetted onto a glass slide; once air dry the specimens were scanned from 5 to 15°2θ on the Panalytical Xpert Pro diffractometer detailed above with a total scan time of 9 min. The sample was then sprayed with a formamide/water (10%) solution, allowed to stand for 40 min and then rerun on the diffractometer using the same 9 min scan; as such, the total exposure time to formamide was ∼50 min. Transmission electron microscopy Samples for transmission electron microscopy (TEM) were prepared by dispersing powders in methanol and drop casting onto holey carbon supports on copper TEM grids. This preparation method was compared with a procedure involving simply dusting powders onto TEM grids; for both methods similar results were obtained. The TEM specimens were examined using bright-field, dark-field and phasecontrast imaging, selected area electron diffraction (SAED) and energy dispersive X-ray (EDX) spectroscopy at 200 kV using an FEI CM200 field emission TEM/STEM. Electron diffraction patterns were recorded on photographic film. Scanning electron microscopy For SEM secondary electron imaging either powders or fractured surfaces of specimens were mounted onto standard aluminium stubs using double-sided carbon tabs as the adhesive, and sputter coated with a thin layer of platinum to reduce charging under the electron beam. Analysis was conducted using a Carl Zeiss Sigma VP Field Emission Scanning Electron Microscope (FEG-SEM) in variable pressure mode at an accelerating voltage of 15 kV. 329 For image analysis of particle-size distributions, specimens of a few milligrams were prepared by ultrasonic dispersion for 5 min into 100 mL of deionized water to which two drops of ‘Calgon’ (a dispersing agent) were added, resulting in dispersions with turbidity readings varying between 7 and 96 nephelometric turbidity units (ntu). A single drop of the dispersion was then placed on a nickel TEM grid (90 µm diameter square holes) supported on a silicon wafer and the drop allowed to air dry in a desiccator. The grid serves to break the drying droplet into smaller micro droplets and seems to assist in obtaining more homogenous dispersion of particles on the surface of the silicon wafer to facilitate subsequent image analysis. The wafers were coated with an Au/Pd mixture and examined by SEM. Image analysis was performed using the freely available ImageJ software (Rasband, 1997–2015). All images were converted to 8 bit binary, calibrated using the scale bars, adjusted manually for threshold, filtered manually to reduce noise, and converted to binary images. Distributions of particle length and diameter were measured on multiple images for each sample using the particleanalysis function of ImageJ by fitting ellipsoids. In addition, maximum and minimum lengths and maximum and minimum diameters of halloysite particles were measured by a manual survey of each set of images. These values were then used as limits to filter the semi-automated particle measurements obtained using ImageJ such that apparently larger or smaller particles than the manually determined limits were rejected. Particles with aspect ratios of 5% quartz along with similar amounts of cristobalite, whilst one of the samples from Turkey (12Tu) also has 4% quartz. Perhaps not unexpectedly, kaolinite is the most common impurity and its presence was confirmed in the same samples by the formamide test. One of the samples from the US Dragon Mine (US18) was deliberately included in the set because it appears to contain ∼20% kaolinite based 1.82 1.49 5.10 1.87 1.38 2.55 1.50 5.30 6.32 5.40 12.77 1.79 1.06 2.60 1.20 3.04 4.45 3.52 2.47 2.33 7.96 1Au 2Ch 3Ch 4Ch 5Ch 6Ch 7Ch 8NZ 9NZ 10NZ 11Sc 12Tu 13Tu 14US 15US 16US 17US 18US 19US 20US 21US 0.11 0.10 0.25 0.18 0.14 0.15 0.09 0.30 0.45 0.38 0.60 0.14 0.20 0.09 0.18 0.37 0.49 0.31 0.26 0.28 0.41 Max D 0.07 0.06 0.13 0.07 0.09 0.04 0.08 0.12 0.12 0.11 0.06 0.10 0.04 0.08 0.05 0.06 0.06 0.36 0.12 0.06 0.11 Min L 0.04 0.03 0.04 0.03 0.02 0.02 0.03 0.04 0.05 0.04 0.03 0.03 0.02 0.03 0.03 0.03 0.04 0.06 0.06 0.03 0.03 Min D 0.75 1.07 1.76 2.17 1.52 2.42 2.68 1.81 2.58 2.59 2.79 1.38 0.89 0.76 1.82 1.28 1.24 0.96 1.10 1.49 1.44 Skew 0.46 0.25 0.46 0.22 0.25 0.17 0.22 0.42 0.45 0.46 0.62 0.26 0.18 0.27 0.23 0.46 0.44 0.96 0.43 0.32 0.84 Med L 0.06 0.05 0.08 0.07 0.05 0.05 0.05 0.14 0.13 0.13 0.10 0.08 0.07 0.08 0.07 0.11 0.09 0.17 0.13 0.11 0.15 Med D 7.8 4.7 5.4 3.6 5.0 3.1 3.8 3.0 3.4 3.4 6.0 3.3 2.5 3.1 3.2 4.5 4.1 6.1 3.3 2.9 5.7 Med AR 0.44 0.23 0.45 0.23 0.26 0.17 0.22 0.43 0.47 0.47 0.63 0.27 0.18 0.26 0.22 0.44 0.42 0.95 0.43 0.31 0.81 Mean L 1.71 1.60 1.84 1.82 1.71 1.76 1.78 2.16 1.97 1.93 2.64 1.86 1.99 1.58 1.70 2.18 2.19 1.65 1.86 2.09 1.83 σ 0.06 0.05 0.08 0.07 0.05 0.05 0.05 0.13 0.13 0.13 0.10 0.08 0.07 0.08 0.07 0.10 0.10 0.16 0.13 0.10 0.15 Mean D 1.25 1.31 1.45 1.46 1.49 1.52 1.38 1.72 1.59 1.75 2.17 1.70 1.85 1.52 1.45 1.89 1.88 1.37 1.52 1.70 1.53 σ 7.4 4.2 5.4 3.5 5.1 3.2 4.1 3.3 3.5 3.6 6.1 3.4 2.6 3.1 3.2 4.3 4.0 6.0 3.4 3.0 5.3 Mean AR 1.59 1.58 1.64 1.70 1.61 1.58 1.59 1.62 1.68 1.71 1.78 1.59 1.47 1.52 1.50 1.67 1.61 1.47 1.56 1.60 1.77 σ 15.3 8.9 13.2 9.7 12.5 7.8 10.1 8.7 9.3 10.1 17.2 8.6 5.7 6.2 6.5 10.8 9.4 11.5 8.0 7.3 14.3 AR10% 676 209 1,176 911 394 716 264 408 420 495 1,046 266 199 363 259 510 378 181 507 310 757 N particles 84.1 96.6 44.0 63.9 86.1 79.3 84.6 23.2 22.6 22.1 13.5 40.4 42.2 45.3 58.8 25.4 21.0 24.0 27.5 29.7 23.2 SA m2 g−1 calc Also listed are skewness, aspect ratios and calculated specific surface areas. Standard deviations should be understood as multiplying or dividing (X/) the mean values. L – length; D – diameter; Skew – skewness; AR – aspect ratio; SA – surface area; σ – standard deviation. Max L Sample TABLE 3. Size data for the samples obtained by manual measurement (values for maximum and minimum lengths and diameters) and by image analysis (median and mean). 334 S. Hillier et al. Mineralogical and physical properties of HNTs 335 FIG. 5. Example of particle morphologies in high-magnification images of dispersed samples. Note the serrated tube end edges of some prismatic tubes (e.g. “E” in 11Sc), “spindle” like morphologies (e.g. “S” in 9NZ and16US), hexagonal outlines on some particle surfaces (e.g. “H” in 18US) intersection mould (“M” in prismatic particle near center of image in 10NZ), and cylindrical nature of smaller intersecting tubes (e.g. “C” in 17US). Note variation in the size of scale bars. on full-pattern fitting, a value that appears to be corroborated by the formamide test, but seem somewhat at odds with the electron microscopy observations which suggest a smaller kaolinite content. XRD characteristics Because the samples are all relatively pure halloysite it is instructive to compare and contrast various features of their random powder XRD patterns which can be related directly to the characteristics of the halloysite present. In the ‘as received’ state the samples exhibited various states of hydration/dehydration, some showing large proportions of 10 Å halloysite, others showing only peaks for the 7 Å form. Halloysites are well known to dehydrate rapidly (see Fig. 1 of Wilson & Keeling, 2016, this issue), so, because of the dependence of basal peak position on the rather variable, often unknown, collection and storage history of different samples, it made little sense to compare diffraction data across the set in the ‘as received state’. However, examples of diffraction patterns of one predominantly cylindrical form (5Ch) and one predominantly prismatic form (17US), as determined by SEM examination, are shown both hydrated (as received) and dehydrated (oven dried 105°C) in Fig. 6. Across the entire set of samples measurements were made on the random powder specimens that were prepared by spray drying at 60°C because, after this process, all samples are consistently in the 7 Å form. Nonetheless the observed basal spacings of the spraydried samples are quite variable ranging between 7.193 and 7.667 Å (Table 4). Measurements conducted on two standard kaolinites (KGa-2 and KGa-1b) gave 0.3 1Au 2Ch 3Ch 4Ch 5Ch 6Ch 7Ch 8NZ 9NZ 10NZ 11Sc 12Tu 13Tu 14US 15US 16US 17US 18US 19US 20US 21US KGa-2 KGa-1b trace 0.1 1.9 Alunite 4.3 2.9 1.1 trace 0.1 Gibbsite 0.6 6.8 4.6 Cristobalite 7.7 0.5 19.4 0.6 4.8 6.1 3.2 1.4 1.4 Kaolinite 99.7 100.0 100.0 98.5 99.8 100.0 98.9 98.8 86.0 88.2 99.4 88.0 99.2 100.0 100.0 90.4 99.0 80.1 98.8 89.5 93.5 Halloysite 7.394 7.566 7.481 7.667 7.478 7.593 7.403 7.424 7.395 7.371 7.193 7.610 7.420 7.350 7.478 7.425 7.555 7.197 7.350 7.478 7.325 7.184 7.169 d001 1.10 1.30 1.06 1.50 1.24 1.32 1.21 1.00 0.97 0.99 0.50 1.48 0.88 0.90 1.22 1.17 1.41 0.65 1.12 1.34 0.70 0.33 0.25 FWHM-001 0.252 0.366 0.260 0.354 0.326 0.359 0.332 0.267 0.235 0.249 0.083 0.339 0.311 0.350 0.347 0.186 0.213 0.137 0.223 0.338 0.172 0.053 0.011 ’CP’ index 1.4857 1.4835 1.4868 1.4834 1.4838 1.4844 1.4838 1.4866 1.4875 1.4871 1.4880 1.4852 1.4855 1.4839 1.4853 1.4871 1.4873 1.4879 1.4863 1.4864 1.4877 1.4891 1.4882 d06,33 yes no yes no no no no no no no yes no no no no no yes yes no no no NA NA 021,112 XRD and IR data are also given for two reference kaolinites for comparative purposes. Samples 11Sc, 12Tu and 17US contain traces of phosphate minerals, possibly woodhouseite in 12Tu and 17US. Sample 21 US contains traces of micas. trace? 1.9 0.4 0.5 0.6 1.4 0.4 0.6 7.2 5.8 0.6 4.0 0.8 0.1 0.1 Quartz Sample 0.91 0.38 0.81 0.26 0.41 0.28 0.39 1.02 1.17 1.30 1.41 0.57 0.54 0.50 0.35 1.14 1.06 1.19 1.08 0.55 1.20 1.07 1.13 IR OH ratio TABLE 4. Mineralogy by full-pattern fitting XRD analysis along with XRD measurements including ‘cylindrical/prismatic’ (‘CP’) index and IR ‘OH-stretching band ratio’, see text for details. 336 S. Hillier et al. Mineralogical and physical properties of HNTs 337 FIG. 6. Examples of random powder patterns of cylindrical (5Ch) and prismatic (17US) tubular forms of halloysite in both 10 and 7 Å form. Inserts illustrate measurement of the cylindrical/prismatic “CP” index from the 02,13 diffraction band, and reaction of sample 17US to treatment with formamide, which confirms the absence of kaolinite in this sample. Note the peak near 4.35 Å in the ( predominantly) 10 Å form of sample 17US, which is taken as the 021 peak. Because 17US is not entirely 10 Å, the broad peak near 3.5 Å is taken as the 004 (7 Å) compared to the 10 Å form of 5Ch which only shows the 10 Å 006 near 3.35 Å. The smaller peak near 3 Å in 17US is possibly 025. Indices assume a two-layer structure. values of 7.184 and 7.169 Å, i.e. slightly smaller than any spacing recorded for halloysite. Peak widths (Full Width at Half Maximum (FWHM)) are also rather variable, ranging from 0.501 to 1.498 Å, compared to 0.326 and 0.254 Å for the reference kaolinites (Table 4). In addition, peak positions and peak widths are correlated. Comparison of the XRD patterns of hydrated and dehydrated forms (Fig. 6) confirms the summary of Brindley (1984) that the patterns of the variously hydrated forms differ mainly in the positions of the basal reflections and rather little in the broadly spreading non-basal diffraction bands, though some effects on the latter are apparent. However, a comparison of either 10 or 7 Å patterns for the two different tube morphologies indicates that there are modulations of non-basal diffraction bands, particularly evident in the 20,13 band, for the prismatic types compared to the cylindrical ones. Thus the 20,13 diffraction band between ∼2.6 and 2.3 Å varies between a broad rounded maxima (cylindrical form) with a leading edge, reminiscent of the turbostratic two-dimensional diffraction bands observed for smectites, to a more modulated band with at least two distinct peak maxima in addition to the peaked leading edge of the band ( prismatic form). Importantly, these differences are apparent irrespective of the state of hydration (Fig. 6), although we have not observed prismatic samples with modulated bands that were in a completely hydrated (10 Å) form. The extent to which the modulation of the band is developed may be quantified by a simple index termed ‘CP’ (cylindrical-prismatic) the measurement of which is illustrated in the inserts in Fig. 6 with values for all samples tabulated in Table 4. The index is defined as background-subtracted minimum/(minimum + maximum) intensities where, due to the variable characteristics of the 20,13 band, the minimum and maximum intensities are deliberately and necessarily measured from the lowest ‘valley’ and highest peaks of the entire 20,13 band, rather than at fixed positions. In the spray-dried random powder preparations a small number of samples also show discernible but generally weak 021 and 112 reflections at d spacings of ∼4.26 and 4.09 Å in the tail of the 02,11 band situated between ∼4.5 and 3.8 Å. These include sample 1Au from Australia, sample 3Ch from the Dafang region China, sample 11Sc from Scotland, and samples 17US and 18US from the Dragon mine, Utah, USA (Fig. 7). Sample 17US from the Dragon mine also shows a relatively strong peak at ∼4.35 Å in the ‘as received’ 338 S. Hillier et al. and prismatic examples, and some of the larger prismatic varieties show obvious hexagonal plate-like features arranged in apparent crystallographic alignment with the tube axis, particularly at tube endings where the edges of plate-like terminations appear oriented at right angles to the tube axis. Similar features were documented by Dixon & McKee (1974) who interpreted them as a continuation of “hexagonal growth edges”. FTIR characteristics FIG. 7. Examples of diffraction patterns of the 02,11 band of several prismatic samples all showing evidence of 021 and 112 peaks in the tail of the band. All samples are in the 7 Å dehydrated form except sample 17US which is shown in both dehydrated (7 Å) and hydrated (10 Å) form. In the 10 Å form the 021 peak is at 4.35 Å. sample which was largely hydrated, this peak being indexed as the equivalent 021 peak (Bailey, 1990) of the 10 Å hydrated form (Fig. 7). One additional feature of the XRD patterns also worthy of mention is the variability in position and breadth of the 06,33 band. For the two reference kaolinites the center of gravity of this band was measured at ∼1.489–1.488 Å. In comparison, the values for the halloysites extend to smaller d spacings of ∼1.483 Å with the smallest spacings always also associated with the broadest peaks. TEM observations Transmission electron microscopy was only conducted on a selected number of samples (3Ch, 11Sc, 14US, 18US). It confirmed the apparent two-layer structure (2M1) revealed in electron diffraction patterns (Fig. 8) as shown previously by other workers, and also revealed further details of the textures observed by SEM. In particular, the electron-transparent central lumen is clearly visible in tubes from both cylindrical The IR spectra of the various halloysites studied also vary in their characteristics (Fig. 9) and examination of the OH-stretching region, in particular (3,800– 3,400 cm−1) shows that certain features make some spectra clearly distinct from others. Thus, several halloysite spectra differ very clearly from IR spectra of kaolinite by showing just two prominent OH-stretching bands, arising from structural OH groups, near 3620 and 3,690 cm−1 with the lower frequency 3620 cm−1 band obviously more intense than the higher frequency 3690 cm−1 band. In contrast, other spectra show one or two additional weak intermediate OH-stretching bands near 3670 and 3653 cm−1, which also occur in kaolinite spectra due to out-of-phase stretching which is split due to imperfect three-fold symmetry (Farmer, 1974). When only one intermediate band is apparent it is the higher-frequency 3670 cm−1 band. In such spectra with intermediate bands the OH-stretching band near 3690 cm−1, arising from the surface hydroxyls (Farmer, 1974) tends to show increased intensity relative to the inner OH-stretching band near 3620 cm−1 which arises from the inner hydroxyl, and both bands also have a tendency to sharpen. There are also indications that the vibrations giving rise to the 3620 and 3690 cm−1 bands absorb at slightly lower frequencies in spectra of this type compared to those where intermediate OH-stretching bands are not present. The observed change in relative intensity of the two main OH-stretching bands, together with the appearance of the weak intermediate bands, are features that suggest that the corresponding samples possess significant ‘well-ordered’ kaolinite-like character. Indeed, the weak OH deformation band near 935 cm−1 in these spectra (assigned to a bending vibration of surface OH groups) that is also present in kaolinite spectra, is not detectable in the halloysite spectra with no intermediate OH-stretching bands presumably due to uncoupling of the surface OH bonds (Farmer, 1974). Therefore, this feature also presents evidence for the presence of kaolinite-like Mineralogical and physical properties of HNTs 339 FIG. 8. TEM images of selected samples showing electron transparent lumens in both cylindrical (14US) and prismatic types (3Ch, 11Sc, 18US). Note details of tube endings. Also shown is an electron diffraction pattern from a cylindrical tube in sample 14US, which is consistent with the two-layer structure reported by earlier workers. character in those halloysite IR spectra in which it appears. In addition to bands relating to structural OH groups, the OH-stretching region also shows OH-stretching bands arising from H2O molecules. The medium intensity absorption band near 3550 cm−1 is assigned to the OH-stretching of halloysitic interlayer “hole water” (Costanzo et al., 1984; Giese, 1988) that probably forms weak hydrogen bonds with the oxygen atoms of the siloxane tetrahedral sheet and possibly with the surface hydroxyls of the octahedral sheet in the next 1:1 layer. The corresponding OHbending band is taken to be the weak, sharp band near 1650 cm−1. As far as kaolins more generally are concerned, it is worth noting that these distinctive bands due to “hole water” are the only bands that are 340 S. Hillier et al. FIG. 9. Examples of IR spectra of cylindrical (5Ch) and prismatic forms (17US, 11SC) of halloysite. The insert illustrates the measurement of the ‘OH-stretching band ratio’. entirely unique to halloysite. The weak and broad absorption near 3430 cm−1 is assigned tentatively to the OH-stretching of “associated” interlayer H2O (Costanzo et al., 1984), with the low frequency indicating the presence of relatively strong hydrogen bonding; the corresponding OH-bending band is near 1630 cm−1. The structural “hole water” OH-stretching band also appears to be sharper in spectra with kaolinite-like character suggesting a more ordered structure overall. The presence/absence of the intermediate OHstretching bands as well as the 935 cm−1 OH bending band along with an ‘OH stretching band intensity ratio’, measured as shown in the insert to Fig. 9 (but in absorbance rather than transmission), provides an empirical means of differentiating the two halloysite types (Table 4). To some extent, the proposed ‘OH-stretching band intensity ratio’ may be affected by the degree of hydration, and also by possible orientation effects but the relative reproducibility of the IR sample-handling procedure would certainly reduce any effect of sample orientation. It should also be noted that minor/trace amounts of gibbsite or kaolinite would serve to either reduce or increase the ratio, respectively. Despite the possible influence of these effects the ‘OH-stretching band intensity ratio’ is, nonetheless, a simple way to portray the classification of the different samples according to IR spectroscopy. Specific surface area and porosity by N2 adsorption Specific surface area and porosity measurements for the sample set are given in Table 5. The BET surface area ranged from ∼15 to 90 m2 g−1 and BJH specific pore volume from ∼0.06 to 0.30 mL3 g−1. Also included in Table 5 are measured values for the most 62.6 66.8 33.7 80.4 68.5 90.6 69.5 35.4 22.5 27.8 15.5 60.7 55.6 68.4 75.7 34.0 30.0 35.9 36.1 74.0 28.0 1Au 2Ch 3Ch 4Ch 5Ch 6Ch 7Ch 8NZ 9NZ 10NZ 11Sc 12Tu 13Tu 14US 15US 16US 17US 18US 19US 20US 21US 0.29 0.20 0.14 0.24 0.22 0.30 0.27 0.16 0.08 0.12 0.06 0.13 0.15 0.25 0.22 0.13 0.14 0.24 0.18 0.13 0.10 Vp (cm3 g−1) 14.4 8.9 16.6 12.7 12.7 15.4 12.0 N/A N/A N/A N/A 11.3 13.5 14.4 14.4 14.4 15.4 N/A 16.6 10.2 N/A MFPDA (nm) 18.7 12.2 16.5 12.1 13.1 13.4 15.7 18.4 14.6 17.3 15.7 8.5 11.0 14.8 11.8 15.5 19.3 26.6 20.5 7.0 13.7 Hyd. (nm) 0.124 0.113 0.060 0.151 0.126 0.216 0.121 0.061 0.037 0.044 0.026 0.080 0.081 0.156 0.135 0.053 0.050 0.057 0.074 0.081 0.043 Vp (cm3 g−1) (5–30 nm) 32.3 29.4 15.5 39.4 32.7 56.2 31.5 15.8 9.7 11.5 6.7 20.7 21.2 40.6 35.2 13.7 13.1 14.8 19.2 21.1 11.2 Lumen space (%) 1.65 5.76 0.46 9.37 4.61 7.56 7.96 1.02 0.14 0.84 − 0.07 4.70 1.54 5.68 6.45 − 0.06 − 0.01 1.06 1.52 3.72 − 0.20 Hysteresis (cm3 g−1 × Ps/Po) 8.0 6.3 4.3 7.6 7.2 6.7 6.1 3.4 2.1 2.0 1.6 6.2 7.7 6.7 8.8 3.5 3.5 3.8 2.9 9.3 4.2 CEC (cmol(+)kg−1) 5.1 7.0 4.9 5.8 5.2 5.8 5.3 4.0 6.3 7.4 6.4 4.7 6.4 4.7 5.3 3.7 4.4 4.8 4.6 3.6 CEC @50 m2 g−1 Pore volume with the range of lumen diameter (5–30 nm) is also given, along with the extent of hysteresis between adsorption and desorption isotherms, and CEC normalized to a specific surface area of 50 m2 g−1. Note: CEC for sample 11Sc is calculated not measured. SABET (m2 g−1) Sample TABLE 5. Specific surface area (SA), specific pore volume (Vp) and cation exchange capacity (CEC) data. Pore volume within the range of lumen diameter (5–30 nm) is also given along with Most Frequent Pore Diameter (MFPDA), hydraulic (Hyd.) pore diameter, the extent of hysteresis between adsorption and desorption isotherms, and CEC normalized to a specific surface area of 50 m2 g−1. Mineralogical and physical properties of HNTs 341 342 S. Hillier et al. FIG. 10. Examples of differential pore-volume plots. The peaks in the distributions near 17 nm are assumed to correspond to the central lumen of the tubes. Isotherms for the same six samples are shown in Fig. 11. frequent pore diameter (MFPDA, adsorption isotherm). This value was determined readily for the majority of samples which all show a peaked distribution within the accepted size range of the HNT lumen diameter (5–30 nm in the differential pore volume plots); six samples showed more gradually sloping, flatter distributions, however, with no obviously discernible maximum and are labelled ‘not measurable’ in Table 5. Some examples of the differential pore-volume plots are shown in Fig. 10. The range of most frequent pore diameter was 8.9–16.5 nm and the average, 13.5 nm. Hydraulic pore diameters (assuming cylindrical pores (4Vp/ABET) range from 7.0 to 26.6 nm with an average of 15.1 nm. The specific pore volume for pores within the size range of 5–30 nm in diameter is also tabulated as well as the % lumen space calculated for each sample as per Pasbakhsh et al. (2013), i.e. assuming a density for halloysite of 2.6 g cm−3 which corresponds to a structure completely devoid of any interlayer water. A final feature of note is that some samples showed obvious hysteresis between the adsorption and desorption isotherms, whilst others showed branches that were more or less coincident between adsorption and desorption. Examples of the different types are shown in Fig. 11, and the relative extent of hysteresis, quantified by calculating the difference between the areas under the two curve branches at relative pressures greater than 0.4, is given in Table 5. CEC Measured CEC values for 20 of the halloysites are also listed in Table 5, they vary between 2.0 and 9.3 cmol (+) kg−1. Table 5 also includes CEC values normalized to a specific surface area of 50 m2 g−1 which approximates the average BET surface area of the sample set, and this narrows the range of values to between 3.6 and 7.4. By a similar calculation, the CEC of sample 11Sc, which was not available in sufficient mass for CEC measurement is predicted, based on its small surface area, to be ∼1.6 cmol (+) kg−1. DISCUSSION Halloysite has long been described in a wide variety of different habits but its early recognized tubular form (Bates et al., 1950) is undoubtedly its most common form (Joussein et al., 2005). In addition, its current nomenclature in many technological applications as halloysite nanotubes or ‘HNTs’ fundamentally acknowledges the importance placed on its nanotubular shape (Churchman, 2015). Shortly after the tubular form of halloysite was recognized it was also recognized that the character of the tubes varied between different samples. Some tubes were cylindrical, i.e. with circular or elliptic cross sections, others had angular polygonal or polyhedral cross sections, Mineralogical and physical properties of HNTs 343 FIG. 11. Examples of the N2 adsorption and desorption isotherms for cylindrical and prismatic forms of halloysite. Cylindrical forms show obvious hysteresis. Differential pore-volume plots for the same six samples are shown in Fig. 10. reflecting elongated prismatic forms (Bates & Comer, 1957; Hoffman et al., 1962; Chukhrov & Zvyagin, 1966; Bates, 1971; Dixon & McKee, 1974) the faces of which are all of index (001) (Chukhrov & Zvyagin, 1966). Furthermore, it was noted that the prismatic forms (often termed ‘laths’ by some authors of this era) tended to correspond to specimens of ‘metahalloysite’, a now defunct term used historically for the dehydrated 7 Å form of halloysite. It was also often remarked that specimens of ‘metahalloysite’ typically showed more order in their XRD patterns than examples of the fully hydrated ‘endellite’, now termed halloysite (10 Å) and from which, either in nature or in the laboratory, ‘metahalloysite’ was believed to be derived by dehydration. Collectively, these observations led many early workers to suggest that the prismatic faces were developed by flattening during dehydration and that this process was implicitly linked to the increased order observed for ‘metahalloysite’. Such increased order was mainly reflected in sharper basal reflections and greater resolution of non-basal reflections. In addition, peaks that are characteristic of the two-layer (2M1) structure of halloysite (Bailey, 1990, 1993), or at least some tendency towards this mode of stacking (Kogure et al., 2013) as commonly recorded in electron diffraction patterns of all types of halloysite (Chukhrov & Zvyagin, 1966; Kohyama et al., 1978), were occasionally observed by XRD for prismatic specimens in the tail of the 02,11 diffraction band, a feature first noted by Honjo et al. (1954). The idea that dehydration was responsible for the prismatic tubular form of halloysite was reiterated by Bailey (1990), although he added that “at least some of the faces appear too regular and flat for such an origin”. Indeed the possible link between dehydration and cylindrical vs. prismatic form is still being considered, with Kogure et al. (2013), in a detailed study of a prismatic halloysite from Russia, suggesting that the tendency towards two-layer periodicity in the layer stacking sequence may have its origins in the conversion of originally cylindrical forms to prismatic ones by a “long-term ripening process” with dehydration as a key step. The samples examined in the present investigation include both cylindrical and prismatic tubular types (Figs 1, 2, 5 and 8). With respect to the XRD patterns of kaolinites, they all show the usual, though subtle, indicator characteristics of halloysite, including slightly larger d spacings and considerably broader basal peaks (Table 4). Both of these features may be related mainly to residual interlayer H2O molecules and, in the case of the samples composed primarily of cylindrical tubular forms, curvature of the layers as an another type of disorder is also likely to contribute to these diffraction characteristics. In addition, the prismatic polygonal forms are characterized by 20,13 bands that are modulated into peaks, although the degree of modulation, or peak definition, as measured 344 S. Hillier et al. by the proposed ‘CP’ index is not as pronounced as it is in samples of pure kaolinite (Table 4), including poorly ordered kaolinite, such that the index also appears to provide a ready distinction of pure kaolinite from samples containing halloysite. Cylindrical halloysites, in contrast, show 20,13 bands that are composed of a single broad scattering region following the peaked leading edge of the 20,13 band. This latter type of pattern is certainly the type described by Brindley & Robinson (1948) for “metahalloysite” as one consisting only of basal (00l) reflections and (hk) bands. In terms of explaining the different types of diffraction patterns, the modulation of the 20,13 band into peaks in the prismatic examples is believed to simply reflect the presence of flat faces and the construction of prismatic tubes from radial sectors, as shown by other investigators via examination in the TEM (Chukhrov & Zvyagin, 1966; Kogure et al. (2013). Thus, the diffraction band for prismatic forms contains some peaks that are of the type (hkl) (Brindley & Robinson, 1948). In contrast, for the cylindrical tubes, it is assumed that the XRD pattern is also affected by disorder due to the curvature of the layers and as such there are no reflections with index (hkl). Thus the proposed ‘CP’ index clearly distinguishes between predominantly cylindrical vs. predominantly prismatic halloysites as well as distinguishing both types from kaolinite. Distortions of the halloysite diffraction patterns due to layer curvature may also explain the broader shape and slightly smaller d spacings recorded for the 06,33 bands of predominantly cylindrical compared to predominantly prismatic forms. With respect to the 06,33 peak position the present authors speculate that rolling with the b axis aligned along the tube axis, as is commonly observed, might also result in some contraction of the structure in this direction. The presence of planar faces in prismatic samples will also explain why some halloysites appear to show more preferred orientation in X-ray powder patterns relative to others in terms of basal/non-basal intensity ratios when prepared by methods that do not guarantee random powder specimens, a point remarked upon previously by Dixon & McKee (1974). From the discussion above it should be clear that Brindley & Robinson (1948) equated the XRD pattern of all ‘metahalloysites’ with the type of XRD pattern corresponding to dehydrated cylindrical forms and, despite noting ‘additional peaks’ in some ‘metahalloysite’ samples, they preferred to reserve judgement on their significance. These extra peaks probably indicate that these other ‘metahalloysite’ samples contained tubes of predominant prismatic morphology. Indeed, a reluctance to associate any order with the diffraction patterns of halloysite persisted for some time with de Souza Santos et al. (1965) preferring the description ‘tubular kaolinite’ for a sample from Brazil, that appears in hindsight to be an excellent example of a prismatic halloysite. In a related publication, Brindley et al. (1963) had earlier identified type “C” and type “D” XRD patterns equating them with “rolled” and “tubular” halloysite, but the study was more focused on the potential for “order” to be more apparent than real due to halloysite frequently being admixed together with kaolinite. Again with hindsight, Brindley’s type “C” patterns relate to prismatic and type “D” to cylindrical forms of halloysite, respectively. With modern methods of full-pattern fitting of XRD data the various differences present in the diffraction patterns of halloysites relative to those of kaolinite are sufficiently distinct to allow successful quantification of mixtures of kaolinite and halloysite. Measurements of length and diameter by image analysis show that there is a wide variation in tube sizes amongst the samples. However, the range of values obtained are very comparable to the values for 13 samples measured by TEM (400 measurements per sample) by Bates & Comer (1957). Calculation of specific surface areas from the particle-size distribution data gave values that were reasonably similar to those measured by N2 adsorption and this provides further validation of the particle-size data obtained from the image-analysis method. Bates & Comer (1957) also found a strong correlation between mean tube length and mean tube diameter and commented that this was indicative of a “structural relationship between the two parameters”. For comparison, the data of Bates & Comer (1957) for mean particle length and diameter are shown alongside data from the present study in Fig. 12. Neither Bates & Comer (1957), nor Bates (1971) who drew additional attention to this relationship, offered any further explanation. Presumably, the structural relationship envisaged is simply one reflecting processes of crystal growth, i.e. halloysite tubes that have grown longer have also tended to grow thicker. However, we are reminded that exceptions will and do undoubtedly exist, e.g. the exceptionally long (30 µm length) and thin (0.03 µm diameter) patch halloysite, described in detail by Norrish (1995). In recognition that some samples are mixtures of different morphological forms, the proposed “CP” index provides a simple way to rank the samples and compare the XRD characteristics to other measurements, such as size, specific surface area and CEC. In terms of the particle-size data these comparisons show Mineralogical and physical properties of HNTs 345 FIG. 12. Mean length vs. mean diameter for the studied samples plotted along with data from Bates & Comer (1957) which show comparable relationships. that the ‘CP’ index is correlated with both mean and maximum diameters and lengths of halloysite particles. The correlation demonstrates that cylindrical halloysites are typically smaller than prismatic ones confirming the impression gained from cursory examination of the different forms under the SEM (Fig. 1). It is pertinent to note that Dixon & McKee (1974), in a detailed TEM study of a tubular halloysite from Wagon Wheel Gap, Colorado, also commented that smaller tubes were more circular in cross-section than thicker ones. Perhaps not unexpectedly, given the correlations with particle size, specific surface area and CEC are also well correlated with the ‘CP’ index measured from XRD data. Thus cylindrical halloysites tend to have larger surface areas and larger CEC values compared to prismatic examples. A correlation matrix for the main diffraction, size, adsorption and IR spectroscopic data of the present study is given in Table 6. Documentation of these systematic relationships between tube form and size suggest that prismatic tubular forms may be interpreted as tubular forms that have simply grown larger than cylindrical forms, a point we shall return to later. Furthermore, it is evident that prismatic forms are just as common as cylindrical ones, e.g. all the classic New Zealand examples from Matauri Bay show intermediate CP values (0.267– 0.235) consistent with the presence of a significant proportion of both cylindrical and prismatic tubular forms, as are readily observed by SEM (Figs 2 and 5). It would seem, therefore, that a further simplification and refinement of the classification of halloysite types recently summarized by Churchman (2015) into the four categories of tubular, platy, spheroidal and prismatic is called for. Thus rather than reserving the term prismatic for a distinct type of halloysite the tubular category should be subdivided into cylindrical and prismatic types, because the prismatic form is simply a common variant of the tubular form. Parenthetically, we note that none of the physical characteristics appears to be obviously correlated (Table 6) with the Fe content of the samples (Table 2). Presumably this reflects the fact that Fe may be present in either minor impurity phases and/or incorporated in halloysites via isomorphous substitution, and elucidating the potential effects of Fe on halloysite will require more detailed investigation. As far as the IR spectra are concerned, according to the review of Joussein et al. (2005), halloysite exhibits only two Al2OH-stretching bands at 3695 and 3620 cm−1. Our examination indicates that this is only the case for cylindrical tubular halloysites and that intermediate bands are evident in prismatic specimens, which, based on additional tests, including formamide intercalation, show no obvious evidence for kaolinite as a separate phase. This is an important point because a physical mixture of halloysite and kaolinite could produce a similar IR pattern with apparent intermediate bands. Presumably the intermediate bands observed 0.256 0.084 0.341 0.514 − 0.326 − 0.436 − 0.485 − 0.414 − 0.372 − 0.332 − 0.332 0.378 − 0.582 0.473 0.813 0.601 0.599 − 0.596 − 0.726 − 0.361 − 0.684 − 0.611 − 0.716 − 0.703 0.787 − 0.868 0.810 0.263 Mean L 0.074 0.144 − 0.842 − 0.747 − 0.532 − 0.681 − 0.672 − 0.585 − 0.548 0.811 − 0.697 0.804 0.719 0.000 Mean D 0.907 0.100 0.313 0.068 − 0.363 − 0.190 − 0.445 − 0.414 0.300 − 0.574 0.411 − 0.130 0.004 0.749 Mean AR 0.058 0.459 − 0.202 − 0.417 − 0.374 − 0.417 − 0.437 0.385 − 0.653 0.515 − − 0.017 0.004 0.535 0.000 AR10% 0.773 0.752 0.725 0.651 0.477 0.475 − 0.851 0.679 − 0.771 0.149 0.004 0.000 0.666 0.802 SAcalc 0.745 0.897 0.894 0.601 0.615 − 0.857 0.862 − 0.939 0.048 0.000 0.000 0.167 0.036 0.000 SABET Correlation coefficients below diagonal; corresponding p-values above diagonal. L, length; D, diameter; AR, aspect ratio; SA, surface area; Vp, pore volume. Fe2O3 Mean L Mean D Mean AR AR10% SAcalc SABET Vp Hysteresis CEC d001 FWHM-001 d06,33 CPindex IR OH ratio Fe2O3 0.685 0.576 0.239 0.322 − 0.686 0.516 − 0.625 0.026 0.108 0.013 0.769 0.380 0.000 0.000 Vp 0.678 0.571 0.580 − 0.893 0.808 − 0.894 0.062 0.001 0.001 0.105 0.060 0.000 0.000 0.001 Hysteresis 0.493 0.496 − 0.682 0.756 − 0.845 0.097 0.003 0.001 0.409 0.095 0.001 0.000 0.006 0.001 CEC 0.913 0.592 0.737 − 0.685 − 0.141 0.000 0.005 0.043 0.060 0.029 0.004 0.297 0.007 0.023 d001 0.578 0.706 0.639 − − 0.142 0.000 0.010 0.062 0.047 0.030 0.003 0.155 0.006 0.022 0.000 FWHM-001 TABLE 6. Pearson correlation matrix for the main physical and mineralogical properties measured. 0.858 0.903 − 0.091 0.000 0.000 0.187 0.085 0.000 0.000 0.001 0.000 0.001 0.005 0.006 d06,33 0.918 − 0.006 0.000 0.000 0.007 0.001 0.001 0.000 0.017 0.000 0.000 0.000 0.000 0.000 CPindex 0.030 0.000 0.000 0.064 0.017 0.000 0.000 0.002 0.000 0.000 0.001 0.002 0.000 0.000 IR OH ratio 346 S. Hillier et al. Mineralogical and physical properties of HNTs for prismatic halloysite are an indication of a generally more ordered pattern of hydrogen bonding of the layers, presumably a reflection of the fact that many layers are planar in the prismatic examples whereas in the cylindrical halloysites there is added disorder due to curvature of the layers which probably results in uncoupling of the surface OH bonds (Farmer, 1974). Additionally, the similarity of the IR spectra of prismatic forms of halloysite to those of kaolinite may be expected, based on the observations of Kogure et al. (2013) that, locally, the stacking sequence in halloysite is identical to that of kaolinite. It is also noteworthy that heating cylindrical halloysite samples in the laboratory does not result in the appearance of intermediate OH-stretching bands. The measured ‘OH-stretching band intensity ratio’ is also well correlated with the ‘CP’ index (Table 6) indicating that the features reflected in the XRD patterns of the halloysites are also reflected in the order/disorder of the hydrogen bonding of the layers as revealed by IR spectroscopy. Remarkably this was noted almost 50 years ago by Chukhrov & Zvyagin (1966) and additionally by Bates (1971) who also went further and demonstrated that the ratio of the two main OH-stretching adsorption bands was related to both the length and the order of halloysite tubes (“and laths”). These relationships with tube length and order are confirmed in the present study (Table 6), although it should be remarked that correlations to particle length are potentially subject to any prior processing that some of the beneficiated samples may have undergone, so it may be better practice to rely on correlations with particle diameter in order to assess the characteristics of such processed samples. Returning to the structure of halloysite, peaks characteristic of the tendency towards the 2M1 stacking sequence (112, 021) in the terminology of Bailey (1990, 1993) were only observed in the XRD patterns of some prismatic samples all with “CP” indices of