Ordered Ferrimagnetic Form of Ferrihydrite Reveals Links Among Structure, Composition, and Magnetism

Ordered ferrimagnetic form of ferrihydrite reveals links among structure, composition, and magnetism

F. Marc Michela,b,1, Vidal Barrónc, José Torrentc, María P. Moralesd, Carlos J. Sernad, Jean-François Boilye, Qingsong Liuf, Andrea Ambrosinig, A. Cristina Cismasua, and Gordon E. Brown Jr.a,b

aSurface and Aqueous Geochemistry Group, Department of Geological & Environmental Sciences, Stanford University, Stanford, CA 94305; bStanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park, CA 94025; cDepartamento de Ciencias y Recursos Agrícolas y Forestales, Universidad de Córdoba, Campus de Rabanales, 14071 Córdoba, Spain; dInstituto de Ciencia de Materiales de Madrid (CSIC), 28049 Cantoblanco, Madrid, Spain; eDepartment of Chemistry, Umeå University, SE 901 87 Umeå, Sweden; fPaleomagnetism and Geochronology Laboratory (SKL-LE), Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China; and gSandia National Laboratories, PO Box 5800, MS 1415, Albuquerque, NM 87185

Edited* by W. G. Ernst, Stanford University, Stanford, CA, and approved December 30, 2009 (received for review September 4, 2009)

The natural nanomineral ferrihydrite is an important component of unequivocally by conventional crystallographic techniques. The many environmental and soil systems and has been implicated as most recent model, and arguably the most complete, was derived the inorganic core of ferritin in biological systems. Knowledge of its from analysis of the pair distribution function of x-ray total scat- basic structure, composition, and extent of structural disorder is es- tering data that suggested that the ferrihydrite structure, although sential for understanding its reactivity, stability, and magnetic be- disordered, could be described using a single structural phase (6). havior, as well as changes in these properties during aging. Here However, this model has since received criticism for giving a we investigate compositional, structural, and magnetic changes composition that is anomalously H-poor (7) as well as not fully that occur upon aging of “2-line” ferrihydrite in the presence of satisfying the observed diffraction features (7), calculated bond adsorbed citrate at elevated temperature. Whereas aging under valence sums (5), or the measured density (7, 8) of ferrihydrite. these conditions ultimately results in the formation of hematite, In addition, ferrihydrite, whether natural or synthetic, is generally analysis of the atomic pair distribution function and complemen- found to be antiferromagnetic with superparamagnetic behavior tary physicochemical and magnetic data indicate formation of an at ambient temperature [see (15) and references therein]. Several intermediate ferrihydrite phase of larger particle size with few de- studies also suggested an additional ferromagnetic-like compo- fects, more structural relaxation and electron spin ordering, and nent attributed to the presence of uncompensated surficial spins pronounced ferrimagnetism relative to its disordered ferrihydrite (16, 17), but this has been difficult to confirm due to the uncer- precursor. Our results represent an important conceptual advance tainty regarding the relationships between magnetic behavior and in understanding the nature of structural disorder in ferrihydrite basic crystal structure. The metastability of ferrihydrite, par- and its relation to the magnetic structure and also serve to validate ticularly at elevated temperatures, has led to added ambiguity

a controversial, recently proposed structural model for this phase. regarding the ordering temperature (i.e., Néel or Curie) because GEOLOGY In addition, the pathway we identify for forming ferrimagnetic direct measurement is not feasible (9). Nonetheless, it is generally ferrihydrite potentially explains the magnetic enhancement that accepted that ferrihydrite’s metastability is related in part to typically precedes formation of hematite in aerobic soil and inherent structural disorder, although the nature and extent of weathering environments. Such magnetic enhancement has been this disorder remain poorly defined. attributed to the formation of poorly understood, nano-sized ferri- Understanding the relationship between the structure and magnets from a ferrihydrite precursor. Whereas elevated tempera- magnetic properties of ferrihydrite, as well as changes in both tures drive the transformation on timescales feasible for laboratory as a function of aging, is of particular importance because it pro- studies, our results also suggest that ferrimagnetic ferrihydrite vides independent constraints on ferrihydrite structure and could form naturally at ambient temperature given sufficient time. compositional variations with aging. Being antiferromagnetic at ambient temperature (15, 18) and with only a weak ferrimagnetic- crystal structure ∣ disorder ∣ nano-sized ferrimagnets ∣ soil formation ∣ like component, ferrihydrite is normally not considered in the in- strain terpretation of magnetic enhancement in soils on Earth (19) or Mars (20), nor is it considered useful in the tailoring of functional he structural and physical properties of ferrihydrite, an exclu- ferrimagnetic nanomaterials, in contrast with the ferrimagnets Tsively nano-sized ferric oxyhydroxide, are of importance in ex- magnetite (Fe3O4) and maghemite (γ-Fe2O3) (21, 22). However, plaining its chemical reactivity and wide variety of occurrences. In as will be shown below, ferrihydrite aged at different tempera- both pristine and contaminated soils and sediments, ferrihydrite tures in the presence of selected anions undergoes a significant acts as a natural filter of inorganic contaminants through sorption magnetic enhancement corresponding to the formation of an in- reactions, thus affecting their transport and fate in the environ- termediate phase preceding its transformation into hematite ment. Biomineralization of ferrihydrite as the inorganic iron core (α-Fe2O3) (11, 23–25), and this phase may play an important in ferritin—the protein mainly involved in iron storage and home- role in the magnetic enhancement of aerobic soils. Additionally, ostasis in the human body—also occurs in a vast number of or- such understanding may lead to new insights about the for- ganisms (1). Bloom-forming marine diatoms, for example, use mation of biogenic magnetic phases in organisms. For example, ferritin for enhanced iron storage (2), which suggests that ferrihydrite may also have underlying importance in primary produc- ’ Author contributions: F.M.M. and V.B. designed research; F.M.M., V.B., J.T., M.P.M., Q.L., A. tivity in the world s oceans. A., and A.C.C. performed research; F.M.M., J.T., M.P.M., C.J.S., J.-F.B., Q.L., and A well-known example of a nanomineral (3), ferrihydrite has A.A. analyzed data; F.M.M., V.B., J.T., M.P.M., C.J.S., J.-F.B., Q.L., and G.E.B.J. wrote the no known crystalline counterpart formed in the laboratory or paper; and V.B. and J.-F.B. contributed new reagents/analytic tools. found in nature. As such, the basic crystal structure (4–7) The authors declare no conflict of interest. and physical properties of ferrihydrite [e.g., density, composition This Direct Submission article had a prearranged editor. (7, 8), and magnetic properties (9–14)] have remained controver- 1To whom correspondence should be addressed. E-mail: [email protected]. sial. A variety of structural models have been proposed for ferri- This article contains supporting information online at www.pnas.org/cgi/content/full/ hydrite (see (4) for review) but all have proven difficult to confirm 0910170107/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0910170107 PNAS ∣ February 16, 2010 ∣ vol. 107 ∣ no. 7 ∣ 2787–2792 Downloaded by guest on September 27, 2021 ferritin-derived ferrihydrite in humans, with varying degrees of to the period of significant magnetic enhancement (Fig. 1B), an structural order, has been related to different neurodegenerative extended set of diffraction maxima develop from regions formerly diseases (26). In this regard, biogenic ferrimagnetic crystals, dominated by diffuse scattering (see SI Appendix) with no indica- identified as magnetite, that have been found in human brain tion of a structural transformation prior to that involving the for- tissues (27–29) may originate from biodegradation of a ferrihy- mation of hematite. These changes signal the formation of the drite precursor. ferrifh phase. Here we have synthesized a series of ferrihydrite samples with The transition from fh to ferrifh is accompanied by changes in differing magnetic properties and structural order that allowed us particle size and composition. Relative to the precursor fh, the to determine compositional and structural changes occurring average particle size of the nonhematite fraction, as indicated with aging and to explain the enhanced magnetic properties by transmission electron microscopy, increases by approximately mentioned above (see Methods Summary and SI Appendix). We 200% (Fig. 2A) and corresponds to an approximately 50% de- used the atomic pair distribution function (PDF) derived from syn- crease in specific surface area (Fig. 2B, see SI Appendix). In ad- chrotron high-energy x-ray total scattering to discriminate be- dition, density (Fig. 2B) and total Fe (Fig. 2C) of the solid phase tween competing structural models and detect subtle structural during the initial 8 h aging period increased by 15% and 18%, changes in the transformation of ferrihydrite to hematite. Both respectively, and are related inversely to the amount of water lost − real- and reciprocal-space fitting of the total scattering show that (OH ,H2O) as determined by thermogravimetric analyses the precursor ferrihydrite (disordered with antiferromagnetic-like (TGA) (Fig. 2C). Continuous weight losses, corresponding to behavior) undergoes a series of changes that lead to an ordered the removal of loosely bound or surface-adsorbed water ferrimagnetic phase (ferrifh) with significantly larger particle size (<125 °C) and dehydroxylation of more strongly bound structural (10–12 nm), far fewer cation vacancies, and less lattice strain than water (>125 °C), are observed by TGA. For example, fh under- its disordered precursor. The detailed characterization results goes weight losses of 7% and 20% when heated to 125 °C and represent a critical advance in our understanding of ferrihydrite 1000 °C, respectively. After 8 h of aging, the losses are 2% and because we are now able to validate the structural model proposed 8%, respectively, and the final hematite product has a total by Michel et al. (6) and show how prior inconsistencies in its dif- weight loss of only approximately 2% (see SI Appendix). fraction characteristics, calculated bond valence sums, density, The Fourier transform of the normalized total scattering (a por- and composition can be explained. tion of which is shown in Fig. 1A) results in the PDF, or GðrÞ function (see Methods Summary), and gives the probability of Results & Discussion finding an atom at a given distance r from another atom (31). Ferrihydrite Structural Variations with Aging. As shown in Fig. 1A, the initial ferrihydrite (fh, t ¼ 0 h) is characterized by a small number of poorly resolved Bragg maxima and a significant underlying diffuse scattering component typical of so-called 2-line ferrihydrite (30). The presence of citrate (at a molar ratio with respect to Fe of 3%) slows the transformation of ferrihydrite to hematite while promoting the formation of the intermediate ferrifh phase (see SI Appendix). After 14 h aging time at 175 ° C in the presence of citrate, the structural transformation to hematite is complete. During the first 8 h of aging, corresponding

Fig. 2. Changes in particle size, physicochemical properties, and phase abundances as a function of aging. (A) Average particle size from TEM analysis. Bars indicate standard deviations. (B) Density (Open Diamonds) from pycno- Fig. 1. X-ray scattering and corresponding magnetic enhancement. (A)X- metry and specific surface area (Light Shade) from N2-BET as a function of ray total scattering intensities collected as a function of sample aging time aging. (C) Total Fe (Solid Squares) of solid phase and hydration weight losses (h) at 175 °C with citrate. Transformation to hematite (Arrows) begins at t ¼ by TGA (Shaded). Medium and dark shaded regions indicate weight losses 8 h (indicated by *) and is complete by 14 h. (B) Formation of a strongly mag- after heating to 1,000 °C and 125 °C, respectively. Note that data for netic intermediate phase indicated by increases in room temperature mag- t ¼ 14 h are representative of single-phase hematite. (D) Percent phase abun- netic susceptibility (χ) and saturation magnetization (Ms) at low temperature dances of precursor ferrihydrite (fh), ferrimagnetic ferrihydrite (ferrifh), and (5 K), with both reaching maxima at approximately 11 h. hematite (hm) from chemometric analysis.

2788 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0910170107 Michel et al. Downloaded by guest on September 27, 2021 The real-space distribution of atom-atom correlations in the PDFs of the octahedral (Fe1) site. During aging, filling of these cation also shows variations with aging (Fig. 3A, see SI Appendix) but with vacancies is indicated by intensity increases of specific pair cor- the following important advantage over the analysis of reciprocal- relations (Fig. 4B and C, see SI Appendix). Refinement of the Fe space scattering: The PDF is sensitive to variations in short- site occupancy values during fitting of the experimental PDFs is (r < ∼5 Å) and intermediate-range (r < ∼15–20 Å) order (31). consistent with these intensity increases and indicates that cation Indeed, real-space analysis of the PDFs allows phase identification sites in the ferrihydrite structure are only fully occupied for sam- and estimation of phase abundances, provides quantitative struc- ples with aging times ≥8 h (Fig. 4, Inset). Relative to this low tural information, and, in certain cases, helps to discriminate vacancy form of ferrihydrite and when considered along with the between competing structural models (see SI Appendix). observed changes in total Fe, the number of vacancies in the Fe2 Chemometric analysis provides a quantitative means of deter- and Fe3 sites in fh are 45–50% each, which is higher than pre- mining the number of phases consistent with the PDF data (see viously estimated (6). One plausible mechanism for charge- Methods Summary). The factor indicator (IND) function provides balancing this number of cation vacancies in the ferrihydrite compelling evidence for the presence of only three distinct solid structure is the incorporation of three protons as hydroxyls into phases in our samples. A linear combination of the three domi- the vacant Fe3þ sites. Evidence for such proton substitutions for nant orthonormal basis vectors extracted from a singular value Si4þ has been found from neutron diffraction of D-substituted decomposition of the 13 PDFs, in fact, represents 99.8% of hydrogarnets (32). Our TGA results indicate a level of hydration the variance of the PDF data. The multivariate curve resolution in fh consistent with an average total hydration of approximately (MCR) method was used to rotate the three dominant vectors 7.4 protons per unit cell, with 5.4 protons being an amount suffi- into real chemical space. These gave rise to individual PDFs cient to satisfy the approximately 45–50% vacancies in the Fe2 for the fh, hematite, and an intermediate phase that can be as- and Fe3 sites and 2 as hydroxyls on the O1 site (see next section cribed to ferrifh (Fig. 3B). There is a dramatic decrease in the and SI Appendix). concentration of the precursor ferrihydrite phase down to 14% Size-dependent structural relaxation in ferrihydrite is indicated by 8 h of aging time, with a concomitant increase in ferrifh, fol- by systematic and anisotropic variations in lattice dimensions lowed by hematite formation (Fig. 2B). In contrast, in the absence occurring during the transition from fh to ferrifh, consistent with of citrate, aging of fh results in transformation to hematite within a reduction in strain (Fig. 5). These trends are observed in both 2 hr, with very little of the ferrifh intermediate observed (23). real- and reciprocal-space fitting (see SI Appendix) and, during The three PDFs representing the three phases were fit with the initial 8 h aging, correspond to an approximately 1% increase only two structural models: (i) The single-phase model of ferri- in unit cell volume (see SI Appendix). If we consider the lattice hydrite (6), or (ii) the known hematite structure (Fig. 3B, see SI parameter changes in ferrihydrite as a function of size, the data Appendix). Surprisingly, the PDFs for both fh and ferrifh could be indicate that the sample with the smallest particles, i.e., that with satisfactorily fit by the structural model of ferrihydrite proposed the highest surface-to-volume ratio, is the most strained (Fig. 5, by Michel et al. (6). Additionally, the extracted structural para- SI Appendix). meters satisfy Pauling bond valence sums (see SI Appendix) and provide a remarkable reproduction of diffraction maxima for the Composition and Density of Ferrihydrite. The high defect concentra- sample aged for 8 h, thus providing strong evidence that the basic tions in the form of cation vacancies in fh result in substantial GEOLOGY structure of the intermediate phase is indeed that of ferrihydrite deviations between measured and predicted compositions and (Fig. 3C, i). Alternative structural models, such as maghemite densities. By combining detailed structural information with mea- and magnetite, are clearly inconsistent with the diffraction data sured total Fe and hydration losses from TGA, we are now able to C ii iii ð Þ þ (Fig. 3 , and ). Moreover, the basic ferrihydrite structure per- propose a composition for disordered fh of Fe8.2O8.5 OH 7.4 sists until rapid transformation to hematite occurs at t ¼> 11 h. 3H2O (see SI Appendix), which differs significantly from pre- Variations in the intensities of particular correlations in the viously suggested compositions (15), as well as that of ordered PDFs reflect subtle changes in the short- and intermediate-range ferrifh (Fe10O14ðOHÞ2 þ ∼H2O). The compositional changes order of fh that are attributable, in part, to an ordering phenom- result in a large density increase of ferrifh (4.85–4.9 gcm−3) enon involving structural relaxation and the filling of partially relative to its disordered precursor (4.0–4.3 gcm−3) (8, 33) that vacant cation sites (Fig. 4A). Our previous study of disordered is consistent with the filling of cation vacancies in the basic ferrihydrite (6) suggested the presence of vacancies in both the structural model of ferrihydrite (6) (SI Appendix). As will be octahedral (Fe2) and tetrahedral (Fe3) sites, but complete filling shown below, the large magnetic enhancement in the ferrifh

Fig. 3. Real- and reciprocal-space analyses. (A) Experimental PDFs for samples with aging times t ¼ 0 h(Lower)to14h(Upper). (B) Individual PDFs for fh, ferrifh, and hm obtained by the MCR method for the three phases (Circles). Calculated fits for ferrihydrite or hematite overlain (Red) with difference plots below (Gray). (C) Comparison of the reciprocal-space scattering data at t ¼ 8 h(Circles) with calculated diffraction profiles (Lines) based on fits of the PDF for ferrifh with models for (i) ferrihydrite, (ii) maghemite, and (iii) magnetite (¼hematite).

Michel et al. PNAS ∣ February 16, 2010 ∣ vol. 107 ∣ no. 7 ∣ 2789 Downloaded by guest on September 27, 2021 are accounted for by the model proposed by Michel et al. (6). The evidence presented here further supports the idea that ferrimagnetic behavior in ferrihydrite can result from the arrangement of magnetic moments at tetrahedral and octahedral Fe sites rather than from a new phase. 2 −1 Ms values of 20.4 and 58.5 Am kg correspond to 3 and 9μB per formula unit for samples at t ¼ 0 and 11 h, respectively. We explain this increase by proposing the existence of a ferrimagnetic structure in ferrifh consisting of magnetic moments in opposite directions and a net magnetization resulting from uncompensated electron spins probably at the octahedral positions (Fig. 6, see SI Appendix). Based on the proposed structure of ferrihydrite (6), if magnetic moments in all Fe3 (tetrahedral or A) and Fe2 (octahedral or B) sites are aligned with 4 spins in one direction and 6 spins opposing in Fe1 (octahedral or B) sites, as has been observed for hexagonal ferrites (34), a moment of 10μB is obtained (Fig. 6A). Alternatively, if magnetic moments in A and B sites are aligned antiparallel, as in magnetite, a resulting moment of 30μB would be expected (see SI Appendix). The nanometer size of the particles could account for a further reduction in the magnetization due to spin canting down to the experimental values measured for the ordered ferrihydrite sample prepared in this study (9μB). Recent density functional theory calculations Fig. 4. Evidence for Fe vacancies in ferrihydrite. (A) Intensity increases for provide additional evidence that the model shown (Fig. 6) cor- specific correlations (Red Arrows) during initial 8 h aging time are attributed responds to the magnetic groundstate (35). to filling of vacant Fe sites. Shifts in other correlations (Blue *) are due to the decreasing a-dimension of the unit cell during the fh → ferrifh transition. (B) Magnetic properties show strong size dependence. For example, Ms decreases linearly for magnetite/maghemite nanoparticles Element-specific and (C) site-specific partials calculated from the refined 2 −1 structure of ordered ferrifh (Fig. 3B). (Inset) Refined occupancies (%) from with decreasing crystallite size (77 and 12 Am kg for particles PDF analysis normalized for total Fe for three Fe sites in ferrihydrite during of 13.5 and 4 nm in diameter, respectively) due to surface and aging. internal spin canting (cation vacancy order-disorder) (36). The high field irreversibility observed for samples at t ¼ 0 and 1 h, intermediate can be explained by an ordering of the electron spin which is closely related to the existence of a certain degree of moments of iron in this low-defect structure. magnetic disorder, supports this assumption. Whether the origin of this magnetic disorder is related to the lack of crystallinity Origin of Ferrimagnetism in Ordered Ferrihydrite. Ms and χ of in the samples and/or the existence of some degree of cation va- the initial ferrihydrite are typical of 2-line ferrihydrite and signify cancy disorder or is just a result of the reduced symmetry and the initial presence of a parasitic, weakly ferrimagnetic phase. uncompensated magnetic interactions of the surface spins is dif- χ M 58 5 2 −1 ficult to determine (36). Moreover, filling A sites in a sample aged Smooth increases in and s (at 5 K) up to . Am kg (for t ¼ 0 t ¼ 11 – t ¼ 11 h; see SI Appendix) during aging are concomitant with from to h reinforces A B interactions and the align- particle growth and specific changes in composition, keeping ment of B moments, leading to an increase in the saturation magnetization. It can be concluded that Ms is therefore a struc- the basic structure of ferrihydrite. 57 ture-sensitive property for nanometer-sized particles and that it A prior study using external-field Fe Mössbauer spectro- depends on the degree of cation site occupancy and surface and scopy indicated the presence of tetrahedral Fe(III) in a magnet- internal spin canting. ically enhanced intermediate of ferrihydrite (24). Note that the authors of that study assigned one tetrahedral Fe plus another octahedral Fe to “hydromaghemite” and one-third of the octahedral Fe to a residual “6-line ferrihydrite.” These three Fe types

Fig. 6. Possible electron spin orientations and magnetic moments in ordered ferrihydrite. Based on the Michel et al. 2007 structure of ferrihydrite (6), Fig. 5. Evidence for strain in ferrihydrite. Size-dependent anisotropic 6 Fe3þ with spins in one direction and 4 Fe3þ opposing results in a magnetic changes in the c-(Circles) and a-(Diamond) lattice parameters indicate strain moment of 10μB. Alternatively, aligning the spins of the Fe3 (tetrahedral) in ferrihydrite. Zero strain (Dashed Gray Line) corresponds to a theoretical and Fe2 (octahedral) sites antiparallel to one another results in a magnetic crystalline ferrihydrite. moment of 30μB.

2790 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0910170107 Michel et al. Downloaded by guest on September 27, 2021 Although both the ordered ferrifh and ferrimagnetic magnetite and phosphate are present. Although the magnetic data for our and maghemite have strong magnetism, they are essentially dif- samples do not exclude the presence of maghemite per se (see SI ferent in several key aspects. First, temperature-dependent mag- Appendix), we find no x-ray diffraction evidence in this study [or netic measurements (37) revealed a ferrihydrite phase with a in similar experiments at 25–175 °C (23)] for its formation. This Curie temperature (TC ¼ 13 °C) that is close to the value esti- finding is surprising given the qualitative similarities between the mated by Berquó et al. (9) for Si-ferrihydrite. Due to the low TC, structures of ferrihydrite (6) and maghemite (15) and the fact the ambient-temperature Ms of the ordered ferrihydrite is highly that the stabilities of these phases have the following order: reduced. Second, the coercivity of the ordered ferrihydrite at 5 K ferrihydrite < maghemite < hematite (46). Our findings may is much higher (52.5 mT) than that of magnetite and maghemite also be important for understanding other transformation path- (typically only several tens of mTassuming that shape anisotropy ways such as disordered ferrihydrite → goethite (47, 48). is dominant). Therefore, the ordered ferrihydrite is characterized by both high coercivity (antiferromagnetism) and high magne- Summary and Conclusions tism (ferrimagnetism). Moreover, with increasing ordering, the This study has shown that there is no major difference in the over- antiferromagnetic-like behavior is gradually masked by the all structural topology between the disordered and ordered forms ferrimagnetism. of ferrihydrite, and thus we are now able to confirm that the basic Previous studies have attributed the ferromagnetic-like mo- structure proposed for ferrihydrite (6) is verified in that it can ment of ferrihydrite to the presence of surface-uncompensated reproduce both the real-space PDF and reciprocal-space diffrac- spins. Thus the ordered ferrihydrite (ferrifh) should be domi- tion data. We also propose a new composition for disordered ð Þ þ 3 nantly antiferromagnetic because the effect of the surface- ferrihydrite (Fe8.2O8.5 OH 7.4 H2O), derived from a combina- uncompensated spins is significant only for finer grained ferrihy- tion of TGA and total Fe measurements and constrained by the drite particles (fh) (e.g., 2–3 nm). In contrast, our results show crystal structure, which is no longer H-poor. In addition, we can that ferrifh is ferrimagnetic, and the disordered surface spins explain the anomalously low measured density of disordered decrease rather than increase the bulk Ms for fh. ferrihydrite as being due to cation vacancies. Specifically, the nature of disorder in 2-line ferrihydrite can be understood as con- Does Ferrimagnetic Ferrihydrite Occur in Nature? A temperature- sisting, in part, of a random distribution of iron vacancies in two dependent magnetic study (37) suggested the presence of an in- specific cation sites that are likely charge-balanced by the incor- termediate maghemite-like phase with a TC of approximately poration of structural protons. Moreover, the filling of cation 400° C. The present study, however, confirmed that the inter- vacancies during the transition from disordered ferrihydrite to mediate phase is essentially ordered ferrihydrite (ferrifh). The ferrimagnetic ferrihydrite leads to an ordering of the electron ferrihydrite → ordered ferrimagnetic ferrihydrite → hematite spin moments of iron and corresponding magnetic enhancement. (fh → ferrifh → hm) transformation model is parsimonious and Although the findings presented here represent a significant consistent with observations on the mineralogical and magnetic advance in confirming the general topology of the disordered properties of soils from different geographic areas, as discussed in ferrihydrite structure, a full description of the positions of all detail by Torrent et al. (38). Thus, the coexistence of ferrimagnets atoms is not yet available. For example, the surface structure

and hematite in soils suggests that magnetic enhancement and of ferrihydrite is virtually unexplored and may exhibit varying GEOLOGY hematite formation are concomitant, whereas there is a general degrees of restructuring relative to the particle interior. Even absence of ferrimagnets in soils dominated by formation of though we have assumed a periodic structural model with cation goethite (α-FeOOH) (38, 39). The alternative hypothesis that vacancies that reproduces the diffraction data, the surface struc- magnetic enhancement is preceded by microbially mediated Fe ture of disordered ferrihydrite is unlikely to be truly periodic, reduction (40, 41) - and Fe2þ in the soil solution reacts with a based on observed changes in cell parameters and strain with Fe3þ hydroxide phase to form magnetite (later oxidized to structural ordering and increasing particle size. Additionally, maghemite)—now appears unlikely because reduction of Fe3þ the positions and suggested charge-balancing role of protons to Fe2þ requires the soil to be water-saturated for significant in the ferrihydrite structure are yet to be confirmed. In light periods. Under these conditions, however, hematite is not formed of these unknowns, a significant future challenge will be deter- (42), and the magnetic enhancement is therefore low also due to mining positions of atoms, including protons, at both the particle the absence of ferrimagnets. surfaces and in the interior structural framework. Perhaps, in situ Magnetic measurements indicate that the grain size of the fer- temperature-dependent studies, for example using proton NMR rimagnets in soils lies mostly near the superparamagnetic (SP)— spectroscopy or neutron scattering, will allow the separation of single domain (SD) threshold (approximately 20–25 nm), with these contributions and provide further insight into these remain- little differences between soils from different areas (43, 44). This ing questions. observation suggests a common origin of nano-sized soil ferri- Our results reveal the relationship between crystal structure magnets over a wide range of climatic conditions and initial input and magnetic structure in ordered ferrihydrite, and we also iden- of natural magnetic phases. Although magnetic extracts of soils tify a pathway that may explain the magnetic enhancement in are commonly dominated by coarse-grained natural ferrimagnets aerobic soils that is associated with the formation of nano-sized that make characterization of nano-sized ferrimagnets in soils dif- ferrimagnets from a ferrihydrite precursor. Although the exis- ficult by conventional methods (45), transmission electron micro- tence of ordered ferrimagnetic ferrihydrite in soils or in ferritin scope images revealed the presence of 10–50 nm magnetic in biological systems has not yet been verified, confirmation of particles in soils from the Russian Steppes (41). Fine-grained this phase in these systems is a challenging goal stimulated by the ferrimagnets in soils thus encompass the size range of ferrifh present study. found in the present study. Additional permissive evidence for the presence of ferrifh in Methods magnetically enhanced soils is provided by the magnetic enhance- Suspensions of 2-line ferrihydrite were prepared by precipitating 0.01M ferric 1M ment observed in the laboratory during the ambient-temperature nitrate with potassium hydroxide to a final pH of 7. The initial solution contained citrate in a molar ratio citrate/Fe of 3% and was aged at 175 °C in aging of ferrihydrite in the presence of selected anions over hun- individual Teflon-lined vessels and under aerobic, dark conditions for periods dreds of days (23), which could be due to the formation of an ranging from 0–14 h. ordered ferrihydrite intermediate, in light of results from the pre- High-energy x-ray total scattering data were collected at beamline 11-ID-B sent study. Therefore, a similar transformation in soils at ambient (approximately 90 keV, λ ¼ 0.13702ð4Þ Å) at the Advanced Photon Source temperature is also possible, particularly if ligands such as citrate (APS), Argonne National Laboratory (ANL). PDFs were calculated from

Michel et al. PNAS ∣ February 16, 2010 ∣ vol. 107 ∣ no. 7 ∣ 2791 Downloaded by guest on September 27, 2021 the Fourier transform of the reduced structure function truncated at MCR analysis is different from principal component analysis (PCA) because −1 28 0.5 Å . For additional details regarding PDF analysis, and physical, che- the eigenvectors of PCA are abstract whereas the vectors of MCR have a chemical, and magnetic characterization, see SI Methods. mical meaning. MCR builds from the results of PCA or SVD (we used the lat- Chemometric analyses included an evaluation of the dimensionality of ter) by rotating abstract PCA/SVD eigenvectors into real chemical space. This the PDFs using the factor indicator function (IND) (49) from the results of vector rotation procedure generates a concentration profile for each compo- a singular value decomposition (SVD) of the 13 PDFs. The data matrix nent which, in our case, corresponds to the reaction times of the ferrihydrite. An×13 (n rows of atom-atom distances and 13 columns for each equilibration The MCR method was therefore essential in extracting the intermediate fer- period) was represented as rihydrite phase.

A ¼ USVT ¼ A þ net E ACKNOWLEDGMENTS. We thank Dr. Peter J. Chupas and Evan Maxey of the APS for assistance with x-ray data collection. This work was supported in part where U is a matrix of orthogonal (basis) output vectors of unit length, S is through the Stanford Environmental Molecular Science Institute (National Science Foundation Grant CHE-0431425) and, in part, from the U.S. Depart- the vector and the transposed (T) vector V is a matrix of input basis vectors. ment of Energy (DOE), Office of Biological and Environmental Research, k The IND function identifies the number ( ) of chemically relevant vectors re- Environmental Remediation Sciences Program and National Science Founda- producing A. The remaining vectors are associated with extractable errors (E). tion Grant EF-0830093 (Center for Environmental Implications of NanoTech- T ’ The resulting Anet matrix can be calculated with Um×kSk×kV13×k. A multivari- nology) (F.M.M., A.C.C., and G.E.B., Jr.). This work was partly funded by Spain s ate curve resolution (MCR) analysis (50) was carried out on the resulting Anet Ministry of Science and Innovation and European Regional Development matrix. The basis vectors U were rotated into real chemical space, and the Funds [Project MAT2008-01489 (M.P.M. and C.J.S.) and Project AGL2006- C03-02 (V.B. and J.T.)]. Support was provided by the 100 Talent Program concentration profiles of the different components were simultaneously re- of the Chinese Academy of Sciences and by National Nature Science Founda- solved starting from initial estimates obtained from evolving factor analysis tion of China Grant 40821091 (Q.S.L.). We are grateful for access to the APS- (49). All calculations were carried out in the computational language of MA- ANL which is supported by the U.S. DOE, Office of Basic Energy Sciences TLAB (The Mathworks, Inc.). under Contract DE-AC02-06CH11357.

1. Chasteen ND, Harrison PM (1999) Mineralization in ferritin: An efficient means of iron 27. Brem F, et al. (2005) Characterization of iron compounds in tumour tissue from tem- storage. J Struct Biol 126:182–194. poral lobe epilepsy patients using low temperature magnetic methods. Biometals 2. Marchetti A, et al. (2009) Ferritin is used for iron storage in bloom-forming marine 18:191–197. pennate diatoms. Nature 457:467–470. 28. Gossuin Y, et al. (2005) Looking for biogenic magnetite in brain ferritin using NMR 3. Hochella Jr MF, et al. (2008) Nanominerals, mineral nanoparticles, and Earth systems. relaxometry. NMR Biomed 18:469–472. Science 319:1631–1635. 29. Kirschvink JL, Kobayashikirschvink A, Woodford BJ (1992) Magnetite biomineraliza- 4. Jambor JL, Dutrizac JE (1998) Occurrence and constitution of natural and synthetic tion in the human brain. Proc Natl Acad Sci USA 89:7683–7687. – ferrihydrite, a widespread iron oxyhydroxide. Chem Rev 98:2549 2585. 30. Michel FM, et al. (2007) Similarities in 2-and 6-line ferrihydrite based on pair distribu- 5. Manceau A (2009) Evaluation of the structural model for ferrihydrite derived from tion function analysis of x-ray total scattering. Chem Mater 19:1489–1496. – real-space modelling of high-energy x-ray diffraction data. Clay Miner 44:19 34. 31. Egami T, Billinge SJL (2003) Underneath the Bragg Peaks: Structural Analysis of 6. Michel FM, et al. (2007) The structure of ferrihydrite, a nanocrystalline material. Complex Materials (Elsevier, Oxford). Science 316:1726–1729. 32. Foreman DW (1968) Neutron and x-ray diffraction study of Ca3Al2ðO4D4Þ3 a garne- 7. Rancourt DG, Meunier JF (2008) Constraints on structural models of ferrihydrite as a toid. J Chem Phys 48:3037–3041. nanocrystalline material. Am Mineral 93:1412–1417. 33. Towe KM, Bradley WF (1967) Mineralogical constitution of colloidal hydrous ferric 8. Hiemstra T, Van Riemsdijk WH (2009) A surface structural model for ferrihydrite I: Sites oxides. J Colloid Interf Sci 24:384–392. related to primary charge, molar mass, and mass density. Geochim Cosmochim Acta 34. Cullity BD (1972) Introduction to Magnetic Materials (Addison–Wesley, Reading). 73:4423–4436. 35. Pinney N, et al. (2009) Density functional theory study of ferrihydrite and related 9. Berquó TS, et al. (2007) High crystallinity Si-ferrihydrite: An insight into its Néel temperature and size dependence of magnetic properties. J Geophys Res-Sol Ea Fe-oxyhydroxides. Chem Mater DOI: 10.1021/cm9023875. 112:B02102. 36. Morales MP, et al. (1999) Surface and internal spin canting in gamma-Fe2O3 nano- – 10. Berquó TS, et al. (2009) Effects of magnetic interactions in antiferromagnetic ferrihy- particles. Chem Mater 11:3058 3064. drite particles. J Phys-Condens Mat 21:176005. 37. Liu QS, et al. (2008) Magnetism of intermediate hydromaghemite in the transforma- 11. Cabello E, et al. (2009) Magnetic enhancement during the crystallization of ferrihy- tion of 2-line ferrihydrite into hematite and its paleoenvironmental implications. drite at 25 and 50 °C. Clays Clay Miner 57:46–53. J Geophys Res-Sol Ea 113:B01103. 12. Coey JMD, Readman PW (1973) New spin structure in an amorphous ferric gel. Nature 38. Torrent J, Barrón V, Liu QS (2006) Magnetic enhancement is linked to and precedes 246:476–478. hematite formation in aerobic soil. Geophys Res Lett 33:L02401. 13. Pankhurst QA, Pollard RJ (1992) Structural and magnetic-properties of ferrihydrite. 39. Torrent J, Liu QS, Bloemendal J, Barrón V (2007) Magnetic enhancement and iron Clays Clay Miner 40:268–272. oxides in the Upper Luochuan Loess-paleosol sequence, Chinese Loess plateau. 14. Pannalal SJ, et al. (2005) Room-temperature magnetic properties of ferrihydrite: Soil Sci Soc Am J 71:1570–1578. A potential magnetic remanence carrier?. Earth Planet Sci Lett 236:856–870. 40. Dearing JA, et al. (1996) Magnetic susceptibility of soil: An evaluation of conflicting 15. Cornell RM, Schwertmann U (2003) The Iron Oxides: Structure, Properties, Reactions, theories using a national dataset. Geophys J Int 127:728–734. Occurrences and Uses (Wiley–VCH, Weinheim). 41. Maher BA, Alekseev A, Alekseeva T (2003) Magnetic mineralogy of soils across 16. Guyodo Y, et al. (2006) Magnetic properties of synthetic six-line ferrihydrite nano- the Russian Steppe: Climatic dependence of pedogenic magnetite formation. particles. Phys Earth Planet In 154:222–233. Paleaogeogr Palaeocl 201:321–341. 17. Zergenyi RS, et al. (2000) Low-temperature magnetic behavior of ferrihydrite. 42. Schwertmann U (1985) The effect of pedogenic environments on iron oxide minerals. – J Geophys Res-Sol Ea 105:8297 8303. Adv Soil S 1:171–200. 18. Gider S, et al. (1995) Classical and quantum magnetic phenomena in natural and 43. Geiss CE, Zanner CW (2006) How abundant is pedogenic magnetite? Abundance and – artificial ferritin proteins. Science 268:77 80. grain size estimates for loessic soils based on rock magnetic analyses. J Geophys Res-Sol 19. Mullins CE (1977) Magnetic-susceptibility of soil and its significance in soil science— Ea 111:B12S21O. Review. J Soil Sci 28:223–246. 44. Liu QS, et al. (2005) Determination of primary magnetic minerals of a weathered me- 20. Morris RV, et al. (2006) Mossbauer mineralogy of rock, soil, and dust at Meridiani tapelite xenolith from Zhouba region, North China, by combining thermomagnetic Planum, Mars: Opportunity’s journey across sulfate-rich outcrop, basaltic sand and runs and low-temperature measurements. Chinese J Geophys-Ch 48:876–881. dust, and hematite lag deposits. J Geophys Res-Planet 111:E12S15. 45. Chen TH, et al. (2005) Characteristics and genesis of maghemite in Chinese loess and 21. Raj K, Moskowitz R (1990) Commercial applications of ferrofluids. J Magn Magn Mater paleosols: Mechanism for magnetic susceptibility enhancement in paleosols. Earth 85:233–245. Planet Sci Lett 240:790–802. 22. Tartaj P, et al. (2003) The preparation of magnetic nanoparticles for applications in biomedicine. J Phys D Appl Phys 36:R182–R197. 46. Navrotsky A, Mazeina L, Majzlan J (2008) Size-driven structural and thermodynamic – 23. Barrón V, Torrent J (2002) Evidence for a simple pathway to maghemite in Earth and complexity in iron oxides. Science 319:1635 1638. Mars soils. Geochim Cosmochim Acta 66:2801–2806. 47. Burleson DJ, Penn RL (2006) Two-step growth of goethite from ferrihydrite. Langmuir – 24. Barrón V, Torrent J, de Grave E (2003) Hydromaghemite, an intermediate in the 22:402 409. hydrothermal transformation of 2-line ferrihydrite into hematite. Am Mineral 48. Penn RL, Erbs JJ, Gulliver DM (2006) Controlled growth of alpha-FeOOH nanorods by 88:1679–1688. exploiting-oriented aggregation. J Cryst Growth 293:1–4. 25. Popov VV, Gorbunov AI (2006) Effect of organic acids on the hydrothermal crystalliza- 49. Malinowski ER (2002) Factor Analysis in Chemistry (Wiley, New York). tion of FeðOHÞ3. Inorg Mater 42:769–776. 50. Jaumot J, Gargallo R, de Juan A, Tauler R (2005) A graphical user-friendly interface for 26. Bartzokis G, et al. (2007) Brain ferritin iron may influence age- and gender-related risks MCR-ALS: A new tool for multivariate curve resolution in MATLAB. Chemom Intell Lab of neurodegeneration. Neurobiol Aging 28:414–423. Syst 76:101–110.

2792 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0910170107 Michel et al. Downloaded by guest on September 27, 2021