Magnetic properties of iron minerals produced by natural iron- and manganese-reducing groundwater bacteria

Abstract

Understanding the contribution of biogenic magnetic particles into sedimentary assemblages is a current challenge in palaeomagnetism. It has been demonstrated recently that magnetic particles produced through biologically controlled mineralization processes, such as magnetosomes from magnetotactic bacteria, contribute to the recording of natural remanent magnetization in marine and lacustrian sediments. Contributions from other, biologically induced, mineralization types, which are known from multiple laboratory experiments to include magnetic minerals, remain largely unknown. Here, we report magnetic properties of iron minerals formed by a community of iron- and manganese-reducing bacteria isolated from a natural groundwater deposit during a 2 yr long incubation experiment. The main iron phases of the biomineralized mass are lepidocrocite, goethite and magnetite, each of which has environmental significance. Unlike the majority of the previous studies that reported superparamagnetic grain size, and thus no remanence carrying capacity of biologically induced magnetite, hysteresis and first-order reversal curves measurements in our study have not detected significant superparamagnetic contribution. The biomineralized mass, instead, contains a mixture of single-domain to pseudo-single-domain and multidomain magnetite particles that are capable of carrying a stable chemical remanent magnetization. Isothermal remanent magnetization acquisition parameters and first-order reversal curves signatures of the biomineralized samples deviate from previously proposed criteria for the distinction of extracellular (biologically induced) magnetic particles in mixtures. Given its potential significance as a carrier of natural remanent magnetization, environmental requirements, distribution in nature and the efficiency in the geomagnetic field recording by biologically induced mineralization need comprehensive investigation.

Biogenic magnetic minerals, Magnetic mineralogy and petrology, Rock and mineral magnetism

1 INTRODUCTION

Iron is a transition metal with alternating redox states that can easily transform under conditions commonly encountered in nature. Iron-containing magnetic minerals are found in a variety of environments and may carry valuable information about the direction and intensity of the geomagnetic field and/or palaeoenvironmental conditions at the time of rock formation. For a reliable interpretation of the palaeomagnetic signal, understanding of magnetization recording processes, that is, the time and process in which the magnetization is acquired in a rock or sediment, is essential. Until recently, it has been assumed in palaeomagnetism that natural remanent magnetization (NRM) of sediments is carried dominantly by detrital magnetic particles. Palaeomagnetic signal recording has been thought to be achieved by the action of a geomagnetic torque that causes the magnetic moments of permanently magnetized particles to align parallel with the Earth's magnetic field during or soon after deposition (Johnson et al.1948; Irving 1957). Magnetizations acquired at about the time of deposition are referred to as a depositional remanent magnetization (DRM). In addition to the detrital particles carrying a DRM, sediments can contain authigenic particles that were grown in situ through the critical blocking volume, thus acquiring a chemical remanent magnetization (CRM). Authigenic growth of minerals often involves biotic activity (Lowenstam & Weiner 1989). Bacteria biomineralize iron oxides through two different pathways: biologically controlled (BCM) and biologically induced (BIM) mineralization (Lowensham 1981; Bazylinski et al.2007). In BCM, particles of magnetic iron oxide magnetite, or iron sulphide greigite are formed within the cells as membrane‐bounded structures called magnetosomes. These magnetosome grains are structurally well-ordered, have unique crystal morphologies and a narrow [magnetic single domain (SD)] size range. Their uniformity and typical physical arrangement in chains give rise to characteristic magnetic properties by which fossils magnetosomes can be identified in sediments (Roberts et al.2012). With recent development of magnetic techniques that enable their confident identification in sedimentary mixtures, BCM has been found to be much more common than previously thought, and is widely preserved in a range of sediment types (Roberts et al.2012). In pelagic carbonates, the NRM has been found to be often dominated by biogenic magnetite, rather than by detrital particles (Abrajevitch & Kodama 2009, 2011; Roberts et al.2013).

While our understanding of magnetosomes’ contribution to NRM of sedimentary rocks has been improving steadily over the last decade (e.g. Heslop et al.2013; Roberts et al.2013), the contribution of particles formed through BIM pathway remains poorly understood. Minerals that form by biologically induced mineralization processes typically precipitate outside the bacterium cell wall as a result of metabolic activity of the organism and subsequent chemical reactions involving metabolic byproducts (Lovley et al.1987). Laboratory experiments have demonstrated that BIM processes depend on the local chemical environment and may be dictated by pH, redox potential and availability of dissolved iron (Fredrickson et al.1998; Zachara et al.2002). The minerals that form are often characterized by broad particle-size distributions, lack of specific crystal morphologies, variable mineralogy and the inclusion of impurities in the mineral lattice. Among other biogenic phases, important magnetic minerals such as goethite, hematite and magnetite are common products of BIM processes (e.g. Konhauser 1998). As a rule, under laboratory conditions BIM magnetite crystals grow to less than 20 nm in diameter on average (Lovley et al.1987, 1990; Moskowitz et al.1989, 1993; Sparks et al.1990; Hanzlik et al.1996; Zachara et al.2002; Roh et al.2003, 2006; Kukkadapu et al.2005; Stapleton et al.2005; Lee et al.2008; Wu et al.2011, 2013), that falls into the superparamagnetic (SP) size range with non-permanent magnetization (Jimenez-Lopez et al.2010). However, extracellular SD or even larger magnetite has been described in cultures of iron reducing bacteria under specific chemical conditions (Vali et al.2004), temperature regimes (Zhang et al.1998) and after prolonged incubation (Li 2012). Extracellular SD magnetite in association with periplasmic lepidocroicite was reported in a biomineralization experiment involving an iron-oxidizing bacterium (Miot et al.2014). Given that the quantity of extracellular magnetite per unit of biomass could be several thousand times larger than that of magnetite formed by magnetotactic bacteria (Frankel 1987; Lovley 1991) and much longer reaction time in natural conditions leading to prolonged growth of the particles, biologically induced mineralization may be an important, but yet underappreciated contributor to the NRM of sediments. Our presently poor knowledge of their magnetic properties hampers identification of magnetic particles formed via BIM processes in sedimentary mixtures. While some researchers regard BIM particles as generally physically indistinguishable from iron phases formed inorganically (e.g. Frankel & Basilinski 2003), others suggest that the BIM particles can be told apart from the detrital (inorganic) based on their characteristic coercivity distribution (e.g. Egli 2004), or magnetization state features (e.g. Ludwig et al.2013). Our study aims to contribute to the growing database on magnetic properties of BIM by describing biomineralization induced by a community of iron and manganese reducing bacteria isolated from a natural groundwater deposit in a two year long incubation experiment.

2 SAMPLE

In a biochemical study described by Kondratyeva & Golubeva (2014), nine test funnels filled with porous substrate, nutrient solution and natural bacterial benthos from various surface and groundwater sources were incubated for 2 yr. In this paper, we investigate magnetic properties of biomineralized mass formed during the experiment in funnel #7, which contained natural iron and manganese bacteria isolated from the Tunguska groundwater deposit. The deposit, currently being explored as a major alternative source of water supply for the city of Khabarovsk (Russian Far East), is situated at the confluence of the rivers Amur and Tunguska, ∼10 km east from the city. The main artesian aquifer is a ∼65 m thick sequence of alluvial sand and gravel deposits of Pliocene to Quaternary age, which is overlain by 6–18 m thick clay-rich deposits (Kulakov & Shtengelov 2015). Groundwater of the aquifer is of the hydrocarbonate type, with high concentrations of Fe (10–50 mg l⁻¹) and Mn (1–8 mg l⁻¹), circum-neutral (pH = 6.0 to 6.35), and moderately reducing (Eh = −100 to 100 mV) (Kulakov 2011). Groundwater temperature is 6–9 °C. The sample for the biochemical study was collected from an exploratory well, at a depth of 35 m below the surface.

The following paragraph provides only a brief description of the experimental conditions and main analytical findings that are relevant to our biomineralized sample. For a full description of the analytical protocols and detailed data analysis, the reader is referred to Kondratyeva & Golubeva (2014).

Funnel #7 (of 250 ml capacity) was first filled to 1/3 of its capacity by quartz sand; followed by ∼1 cm³ of bacterial enrichment culture, and covered finally by zeolite. The enrichment culture was obtained by cultivating the inoculum from the Tunguska ground water deposit in Bromfield liquid medium in the presence of steel wire providing a source of iron. Before being used, the sand and zeolite (clinoptilolite from the Chuguevskoe deposit, Russia) were washed with distilled water and then baked at 110 °C. Pore space was then filled with a solution of Fe(NH₄)₂(SO₄)₂⋅6H₂O + MnSO₄⋅5H₂O. Concentrations of Fe and Mn in the solution were 3 and 1 mg l⁻¹, respectively. After 10 d, the funnel was drained and rinsed with 500 ml of distilled water. No bacterial cells were detected in the drained liquid, indicating that the bacteria adhered to zeolite. The funnel was then closed by a stopper, and was kept at room temperature (20–23 °C) in a natural light regime for 2 yr. At the end of the experiment, the material was washed in 1 l of sterile distilled water and air dried.

During the experiment, an approximately 1 cm thick layer of dense ochre-coloured substance formed at the interface between the sand and zeolite layers (Kondratyeva & Golubeva 2014). Scanning electron microscopy (SEM) images, energy dispersive spectroscopy (EDS) elemental spot analysis, and X-ray powder diffraction (XRD) spectra revealed that the substance mostly consist of insoluble iron oxyhydroxides. XRD analysis identified goethite as the dominant biomineral. Distinct globular and tubular structures morphologically resembling iron-encrusted cells of Leptothrix and Sphaerotilus species of chemoheterotrophic bacteria were observed in the images. Variable concentration of iron and sporadic appearance of sulphur in bacterial encrustations determined with elemental spot-analyses suggest spatial variability in the composition of the biomineralization.

3 METHODS

Magnetic measurements were made at the Center for Advanced Marine Core Research, Kochi University (Japan). Hysteresis and remanence cycles, isothermal remanent magnetization (IRM) acquisition and first-order reversal curves (FORC) were measured for three sample aliquots (of ∼0.01 gram) at room temperature with a Princeton Alternating Gradient Magnetometer (AGM) MicroMag2900 model. Hysteresis loops were recorded between peak magnetic fields of ± 1 T with a 2 mT field increment and a 100 ms averaging time. Saturation magnetization (M_s), saturation remanence (M_r), and coercivity (B_c) were obtained after subtracting the paramagnetic (diamagnetic) contribution. Remanence coercivity (B_cr) was obtained by demagnetizing the saturation IRM (SIRM) in a stepwise increasing backfield. IRM acquisition curves were acquired with 50 logarithmically scaled steps, up to a maximum peak field of 1 T. The curves were decomposed into coercivity components using the fitting program of Kruiver et al. (2001), which is limited to symmetric distributions in logspace. Magnetic components are characterized by the SIRM, which is proportional to the magnetic mineral concentration in a sample, the peak field at which half of the SIRM is reached (B_1/2), and the dispersion parameter (DP) for the corresponding cumulative lognormal distribution (Kruiver et al.2001). For each sample, 120 FORCs were measured using a protocol of Zhao et al. (2015) with irregularly spaced field steps with a saturating field of 0.5 T, an averaging time of 200 ms and a wait time of 1 s between successive measurements. FORC were processed using the algorithm of Zhao et al. (2015) with a smoothing factor (SF) of 3.

Low temperature magnetic properties of a 0.02553 gm aliquot were investigated with a Quantum Design SQUID magnetometer (MPMS, sensitivity of ∼10⁻⁷ mA m²). A remanence acquired at room temperature (RT IRM) in 5 T field was monitored on cooling to 10 K in zero field. Temperature-dependent low-field magnetic ac susceptibility was determined in an applied field of 0.5 mT at a frequency of 100 Hz.

High-temperature thermomagnetic curves were acquired with the horizontal Curie balance (Natsuhara Giken); using an applied field of 0.3 T. A 0.0097 g sample aliquot was progressively heated in air to 700 °C at a rate of 10 °C min⁻¹, and then cooled to room temperature at the same rate.

4 RESULTS

4.1 Hysteresis curves

The magnetization of sand and zeolite matrix both show a typical of paramagnetic materials linear dependence on applied field (Figs 1a and b). The lack of hysteresis attests to the absence of detrital ferrimagnetic phases in the matrix. Hysteresis loops of mineralized samples are open until to 250–300 mT (Fig. 1c, in the ± 0.5 T field range). The three studied subsamples have similar shapes albeit with minor variations in their coercivity and remanence ratio values that are indicative of short-range variations in the median magnetic grain-size and/or mineralogy. The loops have normal shapes; there is no evidence of wasp-waisted loops, or an abrupt slope change when crossing from positive to negative fields and vice-versa, that might have been expected for samples with a substantial SP fraction (Roberts et al.1995; Tauxe et al.1996). On the Day plot (Day et al.1977) that is widely used in palaeomagnetism to discriminate between domain states and, by implication, grain size of magnetite particles, all three studied subsamples fall far to the right of SD + multidomain (MD) mixing curves (Fig. 1d). Such points lying off standard Day plot curves cannot be interpreted unambiguously. Dunlop (2002b) suggested that such hysteresis ratio values may indicate broad size distributions, encompassing pseudo-single-domain (PSD), stable SD and the smallest (∼10 nm) SP particles; with a puzzling lack of larger SP grains just below stable SD size (15–25 nm). Aside from the SP grains, an alternative possibility for the high H_cr/H_c values is a contribution of a high coercivity phase. Samples containing goethite, which was identified in our samples by XRD and low-temperature magnetometry, are usually displaced to the right of the magnetite trends on the Day plot (Peters & Dekkers 2003).

Figure 1.

Field dependence of magnetization for (a) sand, (b) zeolite matrix and (c) biomineralized mass subsamples. Zeolite and sand samples have typical of paramagnetic minerals linear field dependence of magnetization, while the biomineralized mass samples show hysteresis loops. No correction for paramagnetic contribution was applied. (d) The Day plot of hysteresis parameters for the three subsamples of the biomineralized mass. Theoretical mixing curves are those expected for admixtures of multidomain (MD), single domain (SD) and superparamagnetic (SP) magnetite for different median sizes of MD and SP grains (Dunlop 2002a,b). The percentages along these curves refer to the volume fraction of SP or MD grains.

4.2 Low-temperature properties

During cooling, the remanent magnetization acquired at room temperature shows a steep decay at ∼110 K (Fig. 2a). This temperature range is characteristic of the Verwey transition of magnetite (Verwey 1939). In-phase magnetic susceptibility increases quickly from 3.2 × 10⁻⁶ m³ kg⁻¹ at 10 K to 4.9 × 10⁻⁶ m³ kg⁻¹ at ∼50 K, and then the rate of increase slows down (Fig. 2b). Such an increase in susceptibility is characteristic of lepidocrocite, an iron oxyhydroxide that is antiferromagnetic at low temperatures; in synthetic lepidocrocites a peak in-phase susceptibility occurs around 52 K (Hirt et al.2002). On further warming, susceptibility curve have distinct kinks at 110 K and 290 K, corresponding, respectively, to the Verwey transition of magnetite and the Neel temperature of poorly crystalline goethite (Dunlop & Özdemir 1997).

Figure 2.

Low-temperature magnetic properties of the biomineralized mass. (a) Cooling curve of a 5 T remanence in a zero field shows a distinct stepwise change in magnetization at ∼110 K, characteristic of the Verwey transition of magnetite. (b) In-phase magnetic susceptibility versus temperature shows maximums at ∼50 K (the Neel temperature of lepidocrocite), 110 K (the Verwey transition of magnetite) and ∼290 K (Neel temperature of poorly crystalline goethite).

4.3 High-temperature properties

The thermomagnetic run is irreversible (Fig. 3) indicating that magnetic phases were modified by heating. A strong increase in magnetization is observed on the heating limb at 250–300 °C. Such an increase is typical of thermal decomposition of iron oxyhydroxides lepidorocite (Gendler et al.2005) and goethite (Özdemir & Dunlop 2000). The magnetization gradually decreases at temperatures above 300 °C, with detectable inflections at ∼500, 620 and 670 °C. The cooling limb has only two inflections, at ∼670 and 570 °C, which are the diagnostic Curie temperatures of hematite and magnetite, respectively (e.g. Dunlop & Özdemir 1997). The magnetization decrease at 620 °C during warming most likely corresponds to the Curie temperature of maghemite. Maghemite is a thermally unstable phase that converts to hematite on heating. Reported values for the maghemite's Curie temperature vary from 470 to 695 °C depending on strain, structural imperfections and impurities of maghemite crystals (Dunlop & Özdemir 1997). In natural samples, maghemite usually have the Curie temperatures that are intermediate between those of magnetite and hematite. For example, the Curie temperature of 610 °C was reported by de Boer & Dekkers (1996) for natural maghemite; maghemite was also identified in burnt cave sediments by the Curie temperatures of ∼620 °C (e.g. Carrancho et al.2009; Kapper et al.2014). The disappearance of this phase on heating (no inflection at this temperature is detected on the cooling curve) is supportive of the maghemite identification. An identification of a phase responsible for the slight inflection at 500 °C on warming curve is more speculative. Poorly crystalline magnetite or an additional (grain-size) population of maghemite, observed occasionally in natural systems (e.g. de Boer & Dekkers 1996), may account for this feature.

Figure 3.

Thermomagnetic run. A peak increase in magnetization between 250 and 350 °C is due to thermal decomposition of iron oxyhydroxides. Stepwise changes in saturation magnetization at ∼560, 620 and 670 °C are characteristic of magnetite, maghemite and hematite, respectively.

While a magnetization increase in the 250–300 °C interval during heating is typical of thermal alteration of both lepidocrocite and goethite (Cornell & Schwertmann 2003), these two minerals have different thermal alteration pathways. Goethite transforms to hematite, either directly (Goss 1987), or with formation of a small amount of magnetite as an intermediate product (Özdemir & Dunlop 2000). Lepidocrocite transforms to metastable maghemite, that then converts to hematite on further heating (Gendler et al.2005). The distinct 620 °C Curie temperature on heating curve and its disappearance on cooling (Fig. 3) identify maghemite in our sample. The presence of maghemite, as an intermediate product during heating, is supportive of the identification of lepidocrocite in the biomineralized mass. Although the presence of a small amount of hematite in the biomineralized mass cannot be excluded, hematite, detectable by its characteristic Curie temperature in the thermomagnetic run, most likely forms as a final alteration product during lepidocrocite and goethite decomposition.

4.4 IRM acquisition

IRM acquisition plots and plots of their gradients (Fig. 4) for three subsamples are dominated by low coercivity phases. The shape of the curves differs slightly between the samples. Assuming that the IRM acquisition curve follows a log-normal distribution (Robertson & France 1994; Kruiver et al.2001), statistical analysis (Kruiver et al.2001) distinguishes three components in the low coercivity part of the spectra. Two components, Component 2 with log B_1/2 ∼ 1.2–1.3 and DP ∼ 0.3–0.35 and Component 3 with B_1/2 ∼ 1.5–1.6 and narrower distribution of ∼0.2–0.26, are the dominant components that in sum account for >90 per cent of the total SIRM of the samples. These two components likely represent two sub-populations of magnetic grains differing in the median grain-size or degree of non-stoichiometry. The lowest coercivity component, Component 1 with log B_1/2 ∼ 0.6 and DP ∼ 0.45 contributes <10 per cent to the total SIRM of the subsamples. This component most likely arises from thermal activation effects common in PSD and MD magnetite grains (Egli & Lowrie 2002); these effects usually result in left-skewed distributions (Egli 2003) that cannot be fitted with the Kruiver et al. (2001) approach. Component 1 is likely an associated component that accounts for the skewness, and thus does not represent an additional magnetic phase. Standardized acquisition plots (shown in Fig. 4) have a significant concave curvature at higher fields, which is indicative of a presence of a non-saturated high coercivity component (Kruiver et al.2001). Because of the lack of saturation, the IRM acquisition parameters of this component are impossible to determine with an acceptable precision. The contribution of the high coercivity component into the total 1 T SIRM does not exceed 1.5 per cent.

Figure 4.

IRM acquisition curves for the three subsamples of the biomineralized mass. The IRM is dominated by low-coercivity components; concave curvature at higher fields in the standardized acquisition plots indicates high coercivity (likely goethite) contribution.

4.5 FORC distributions

An FORC diagram describes the distribution of coercivities and local interaction fields for an assemblage of magnetic particles (Pike et al.1999; Roberts et al.2000). The vertical spread in an FORC distribution is a manifestation of magnetostatic interactions, while the B_c coordinate reflects the coercivity distribution, where the peak is a measure of the modal coercivity of the particle assemblage. A combination of grain-sizes and concentration of the particles in the sample produces distinct FORC signatures. SD-like moments are evident as peaks with closed contours; non-interacting SD grains are manifested as a narrow central ridge along the B_c axis with no vertical spread (Pike et al.1999; Egli et al.2010), whereas magnetostatic interactions cause vertical spreading and displacement of the FORC distribution peak below the B_j = 0 axis (e.g. Roberts et al.2000). MD moments are characterized by contours that diverge towards B_j axis. FORC diagrams of magnetite with PSD grain size have triangular-shaped contour lines intersecting the Bu axis (Pike et al.2001a; Muxworthy & Dunlop 2002). SP grains are characterized by a dominant distribution at the origin of the FORC diagram accompanied with a nearly vertical distribution in the lower quadrant of the diagram (Pike et al.2001b).

FORC diagrams of the three subsamples from the biomineralized mass differ (Fig. 5). Subsample ZG1 is dominated by an SD-like peak defined by closed contours with a maximum at 16 mT and peak width (defined as the full width at half maximum of the peak) of 18 mT. Outer contours intersecting Bu axis with a characteristic triangular shape indicate a small contribution from the PSD grains. In subsample ZG2, the SD-like peak with maximum at 14 mT and the width of 21 mT is shifted to lower coercivity. More prominent open outer contours indicate a larger PSD contribution compared to subsample ZG1. In subsample ZG3, there are two distinct peaks (Fig. 5). In addition to the SD-like peak, maximum of which is observed at 7 mT, there is a prominent peak at the origin of the FORC diagram that is characterized by increasingly divergent contours towards B_j axis, a behaviour typical of MD particles. The shift of the SD-like peak to lower coercivities accompanied with the increase in the MD-like magnetization observed in our samples are indicative of increasing grain-size of the PSD particles (Roberts et al.2000). Considered as a group, the FORC measurements of the three subsamples indicate variable grain-size of biomineralization, from SD to PSD and MD. No typical FORC signature of an SP population—a nearly vertical distribution in the lower quadrant of the diagram (Pike et al.2001a; Kumari et al.2015)—has been detected in our samples.

Figure 5.

FORC diagrams (a) with profiles of the coercivity (b) and interaction field (c) distributions for biomineralized mass subsamples (SF = 3 in all cases). FORC diagrams are indicative of broad particle size distributions, from interacting single domain, to pseudo-single domain and multidomain.

5 DISCUSSION

5.1 Magnetic mineralogy

A combination of low- and high-temperature measurements indicates the presence of several magnetic phases. A prominent rise in susceptibility with a peak at ∼50 K (Fig. 2b) and characteristic thermal decomposition pattern with formation of maghemite as an intermediate phase (Fig. 3) suggest that oxyhydroxide lepidocrocite is a significant contributor to the biomineralized mass. An inflection at the characteristic Verwey temperature (Figs 2a and b) identifies magnetite, while susceptibility change at ∼290 K (Fig. 2b) is indicative of non-stoichiometric goethite. These minerals are themselves important phases in palaeomagnetism and environmental magnetism studies, but can also be precursor phases for other magnetic minerals. Metastable oxyhydroxides, goethite and lepidocrocite, are known to undergo alteration to more stable phases over millions of years at ambient temperature, or faster at elevated temperatures, producing remanence-carrying phases such as magnetite, maghemite and hematite (e.g. Gehring & Hofmeister 1994; Gendler et al.2005; Till et al.2015).

5.2 Environmental context of oxyhydroxides

Goethite and lepidocrocite formed through BIM processes in our experiment are common minerals in natural environments and are often used as palaeoclimatic indicators. Lepidocrocite is a common phase in modern non-calcareous soils of cooler regions, where it predominantly occurs in environments characterized by a seasonal alternation of reducing and oxidizing conditions (Cornell & Schwertmann 2003). In contrast, in our experiment, seasonal variations played no role in biomineralization of lepidocrocite.

Goethite is by far the most common Fe oxide in surface weathering environments and soils (Cornell & Schwertmann 2003). It predominantly forms through oxidation of ferrous iron, either by direct precipitation from aqueous solution, or through formation of an intermediate ferrihydrate phase that with time transforms to goethite or hematite depending on climatic conditions. Fe²⁺ released from Fe-containing minerals during weathering or microbial reduction may be oxidized immediately in aerobic environment, in this case distribution of goethite will mirror primary iron distribution of the rocks. Alternatively, Fe²⁺ can be transported in solution and goethite will form when an aerobic environment is encountered. In this case, goethite distribution will be controlled by redox structure of the sedimentary column or soil profile. Such close association of goethite, which can be remotely recognized by its distinct yellow colour, with Fe content of rocks and oxygen availability is relied upon in mineral exploration (e.g. Sabins 1999). Contrary to general expectations, goethite resulting from BIM processes does not always require availability of oxygen, and may form in suboxic conditions (e.g. Frankel & Basilinski 2003).

In warmer climates soils, goethite is often found in close association with hematite; relative abundance of these two phases is thought to be largely controlled by moisture availability during soil formation (e.g. Yapp 2001; Cornell & Schwertmann 2003; Evans & Heller 2003). Higher average annual temperatures favour formation of hematite, whereas higher excess moisture and higher organic carbon promote goethite formation. As organic carbon content and moisture availability are generally related, the relative abundance of goethite can be used as an indicator of higher precipitation, whereas low values imply drier/warmer conditions. Relative abundances of goethite and hematite have been increasingly applied as a palaeoenvronmental proxy for precipitation in studies of marine sediments (Harris & Mix 1999; Clift 2006; Zhang et al.2007; Abrajevitch et al.2009; Abrajevitch & Kodama 2011) and palaeosol sequences (Sangode & Bloemendal 2004; Hyland et al.2015). Our experiment suggests that biologically induced mineralization is capable of producing significant amounts of goethite. Goethite formed by BIM processes may not be directly related to the atmospheric precipitation, but rather be associated with ground water circulation. If left unrecognized, biologically induced formation of goethite, likely confined to discrete (permeable) stratigraphic horizons, may introduce an error into climatic interpretation of goethite and hematite abundances in sedimentary sequences. While laboratory experiments (Konhauser 1998) have demonstrated that biomineralization is capable of producing iron minerals that are traditionally imparted with particular environmental significance, biologic and environmental requirements of BIM remain poorly understood (e.g. Larese-Casanova et al.2010), and the contribution of magnetic particles formed through BIM processes into natural sedimentary assemblages remains largely unknown.

5.3 Remanence carrying capacity

Contrary to multiple previous studies that have reported predominantly SP size, and thus, no remanence carrying capacity of biomineralization (Lovley et al.1987, 1990; Moskowitz et al.1989, 1993; Sparks et al.1990; Hanzlik et al.1996; Zachara et al.2002; Roh et al.2003, 2006; Kukkadapu et al.2005; Stapleton et al.2005; Lee et al.2008; Wu et al.2011, 2013), hysteresis, IRM acquisition and FORC characteristics of our biomineralized mass are indicative of remanence-carrying assemblage. Magnetic minerals identified in the biomineralized mass have different remanence-carrying capacity. Lepidocrocite, antiferromagnetic with a low Néel (∼52 K) temperature, does not carry remanent magnetization at ambient temperatures at the Earth's surface. However, lepidocrocite is metastable phase and transforms with time to goethite (Cornell & Schwertmann 2003), or to strongly magnetic maghemite and hematite during (burial or metamorphic) heating (Gehring & Hofmeister 1994; Gendler et al.2005). Thus, lepidocrocite formed through BIM processes can be a precursor phase for geologically important remanence-carrying minerals.

Goethite is a weakly antiferromagnetic mineral capable of carrying NRM (Dekkers & Rochette 1992). As goethite formation is often related to weathering processes, it usually carries secondary magnetizations acquired at the time of alteration. Often, these are recent remanences which are of minor importance for palaeomagnetic studies (e.g. Johnson et al.1984; Torsvik et al.2005; Abrajevitch et al.2014). However, goethite is known to retain stable remanence over long times thus allowing palaeomagnetic dating of diagenetic (e.g. Gehring & Heller 1989) and continental weathering episodes (e.g. Théveniaut et al.2007). Potentially, goethite biomineralization can contain information on timing of biological activity or fluid migration episodes.

Although the importance of oxyhydroxides as the carriers of geologically meaningful magnetization is rather limited, magnetite, another component in our biomineralized mass, is a strongly magnetic mineral capable of carrying stable remanent magnetization. Stability of magnetite remanence depends on its grain-size (e.g. Dunlop 1981). A distinct closed peak distributions in the FORC diagrams (Fig. 5a) indicates that the biomineralized mass contains a population of SD particles that are very important remanence carriers. Although detectable variations in magnetic properties of the three studied subsamples suggest that biomineralization products vary on a short spatial scale, some amount of SD magnetite was detected in all subsamples. Overall, the detection of the SD magnetite in our experiment suggests that in natural environments biologically induced mineralization is capable of recording and retaining the geomagnetic information.

As pointed out by reviewers of this article, there is a widely held expectation that products of biologically induced magnetizations should be predominantly of the SP size range, whereas our results suggest larger particle sizes. The difference can be accounted for by the duration of the biomineralization experiment. While in the studies that reported the SP size range of biomineralization products culture growth experiments lasted few days (e.g. 72 hr in Wu et al. (2013); 240 hr in Stapleton et al. (2005)) to few weeks (e.g. 16 d in Moskowitz et al. (1989); 30 d in Roh et al. (2003); 33 d in Zachara et al. (2002)), our biomineralization experiment lasted 2 yr. The longer experiment time allowed for time-dependent growth (ripening) of particles after nucleation. Particle growth is inherent in chemical systems and occurs through several mechanisms, such as (1) consuming molecular precursor from surrounding solution; (2) Ostwald (dissolution–re-deposition) ripening; and (3) fusion of several particles (e.g. Banfield et al. 2000; Hyeon 2003). When nanoparticles grow by the adsorption of ions from solution, the size of smaller particles increases faster than that of the larger ones (e.g. Hyeon 2003), which leads to an increase in the median size of the particle population. The fusion of several particles and the Ostwald ripening, when larger particles grow at the expense of dissolving smaller ones, also lead to an increase in the median particle size accompanied with a decrease in the fraction of smaller (nucleus-size) particles (e.g. Ratke & Voorhees 2013).

A direct observation of time-dependent particle growth during a biochemical experiment has been reported previously by Li (2012). They estimated that after 1 week of incubation the average size of the neo-formed magnetite crystals was 91 ± 44 nm, while magnetite crystals after more than 2 yr of incubation showed significantly increased crystal size of 292 ± 51 nm. Assuming that particle size distribution is approximately normal, a rough calculation shows that a fraction of particles with sizes below 30 nm (the SP/SD threshold for magnetite) after 1 week of incubation was ∼6.3 per cent, while after 2 yr the SP fraction dropped to a negligible value of 0.00002 per cent. Thus, the absence of a significant SP population in our biomineralized mass is in agreement with chemical evolution trends and with the direct evidence for particle ripening during the extended biochemical incubation experiment by Li (2012). Considering long reaction time in natural systems that allows for chemical particle growth, the widely held expectations that natural biomineralization products should be predominantly of the SP size range should be re-evaluated.

Remanent magnetization is acquired by grains formed through BIM processes at the time that they grow through the critical blocking volume. Such CRM (or grain growth remanent magnetization) overall parallels the external magnetic field and is proportional to it in intensity (e.g. Stokking & Tauxe 1990). Despite its faithful field direction recording, bacterial synthesis of magnetite can occur long after sediment deposition; such delay in remanence acquisition, if unrecognized, may introduce an error in high resolution palaeomagnetic studies. The efficiency of the CRM in recording of the geomagnetic field intensity is also likely to be different from this of other remanence acquisition mechanisms (Pucher 1969; Dunlop 1981). This distinction may be especially potent in studies of sedimentary rocks that contain multiple sources of magnetite particles, including detrital, biogenic (BCM and BIM) and authigenic sources. While recent studies have demonstrated that detrital and BCM (magnetosomes) particles record geomagnetic information in the different way and with the different efficiency (e.g. Patterson et al.2013; Ouyang et al.2014), the contribution of the extracellular (BIM) magnetite to the sedimentary palaeomagnetic record has been difficult to assess, primarily because such BIM particles have been proven hard to identify in natural mixtures.

5.4 Identification of biologically induced magnetite in mixtures

With recent development in numerical magnetic unmixing techniques (e.g. Kruiver et al.2001; Egli 2003) and FORC measurements (Pike et al.1999; Roberts et al.2000, 2014), several attempts have been made to distinguish between the detrital and authigenic magnetic particles in natural sedimentary mixtures. Egli (2004) suggested that these sources can be differentiated based on their coercivity values; extracellular (biologically induced) magnetite particles in lake sediments seem to have lower mean coercivity compared to the detrital. The IRM acquisition curves of our samples have skewed shape (Fig. 4). Although statistical analysis does not provide a unique solution, such a shape allows for two-component fitting; these components could be compared with the extracellular (EX) and detrital (D) components of Egli (2004).

A discrimination scheme based on characteristic FORC signatures of detrital and authigenic magnetite has been proposed by Ludwig et al. (2013). According to the authors, detrital magnetite contributions should be clearly distinguishable by its PSD signature, expressed as triangular FORC contour lines. Authigenic magnetite (collapsed magnetofossil chains and/ or authigenic precipitates), on the other hand, are characterized by interacting SD assemblages that are identifiable by oval contours over the upper quadrant of the FORC diagrams. The FORC distributions of our samples do not quite fit the proposed scheme (Fig. 5a). While subsamples ZG1 and ZG2 are dominated by closed suboval contours typical of the interacting SD systems, their outer contours are diverging and forming roughly triangular outline that, according to Ludwing et al. (2013) is typical of detrital particle signature. Subsample ZG3 deviates even more from the prediction, with its diverging but somewhat asymmetric contour lines suggesting the PSD-MD grain-size range, rather than SD.

Overall, the biomineralized mass is characterized by significant grain-size variations on short spatial scales. The IRM acquisition parameters and FORC distributions of the minerals formed through BIM span over characteristics thought to be typical for the detrital (higher coercivity, PSD particles) and authigenic (lower coercivity, interacting SD particles) magnetic grains. Accordingly, these parameters cannot be used for definitive discrimination between these particle sources.

6 CONCLUSIONS

We report magnetic properties of biologically induced mineralization produced in a two year long laboratory experiment by natural iron- and manganese-reducing bacteria. A combination of temperature-dependent properties identified lepidocrocite, goethite and magnetite in the biomineralized mass. These minerals are often used as environmentally sensitive indicators in palaeoecological studies. Magnetic properties of three subsamples of biomineralized mass differ, indicating short-spatial-scale variability of biologically induced mineralization. Unlike the majority of previous laboratory experiments that reported predominantly superparamagnetic size of biologically induced magnetite particles, the hysteresis and FORC measurements have not detected a sizable SP population in our samples. Magnetite particles produced in our experiment range from SD to PSD and MD. Our results indicate that biologically induced mineralization is capable of recording NRM (CRM). Given the widespread of iron and manganese reducing bacteria in nature and large volume of biomineralization they are able to produce, biologically induced magnetic minerals may be an important contributor to the NRM of sedimentary rocks. Environmental requirements of BIM, contribution of particles produced through BIM processes into sedimentary magnetic assemblages, and the efficiency of biomineralization in recording the geomagnetic field information require comprehensive investigation.

We sincerely thank Ann Hirt, Ken Kodama, an anonymous reviewer and the journal Editor Eduard Petrovsky for insightful suggestions that helped to improve the paper.