Tracking the Verwey Transition in Single Magnetite Nanocrystals by

Apr 18, 2016 - Variable-temperature scanning tunneling spectroscopy revealed a sharp Verwey transition in individual ∼10 nm magnetite nanocrystals ...
0 downloads 0 Views 1MB Size
Download PDF
Letter pubs.acs.org/JPCL

Tracking the Verwey Transition in Single Magnetite Nanocrystals by Variable-Temperature Scanning Tunneling Microscopy Amir Hevroni,† Mukund Bapna,‡ Stephan Piotrowski,‡ Sara A. Majetich,‡ and Gil Markovich*,† †

School of Chemistry and Center for Nanoscience and Nanotechnology, Tel Aviv University, Tel Aviv 69978, Israel Department of Physics, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, United States

‡

S Supporting Information *

ABSTRACT: Variable-temperature scanning tunneling spectroscopy revealed a sharp Verwey transition in individual ∼10 nm magnetite nanocrystals prepared by the coprecipitation technique and embedded in the surface of a gold film. The transition was observed as a significant change in the electronic structure around the Fermi level, with an apparent band gap of ∼140−250 meV appearing below the transition temperature and a pseudogap of ∼75 ± 10 meV appearing above it. The transition temperature was invariably observed around 101 ± 2 K for different nanocrystals, as opposed to 123 K typically reported for stoichiometric bulk crystals. This suggests that the lowering of the transition temperature is an intrinsic finite size effect, probably due to the presence of the surface.

A

Electronic structure probes such as photoelectron spectroscopy or tunneling spectroscopy are surface sensitive, and the surface electronic structure of magnetite is expected to be modified relative to the bulk,12,19−21 especially in the case of a long-range charge-ordered phase. Density functional calculations of thin magnetite films reveal a Verwey transition even at subnanometer thickness of (001) oriented films covered with gold.21 However, these calculations also show that the surface layer of magnetite could be insulating, which further underlines the importance of sampling the interior states of the thin film or NC. Tunneling spectroscopy of single magnetite NCs could potentially probe the interior electronic structure of the NC. This can be accomplished through the double-barrier tunnel junction (DBTJ) configuration, which typically forms when a colloidal nanoparticle is placed on a conducting substrate and probed by a metallic tip of a scanning tunneling microscope (see Figure 1a). The electronic structure of semiconductor quantum dots was previously studied using scanning tunneling microscopy and spectroscopy (STM and STS, respectively) techniques.22,23 By tunneling electrons of different energies through a NC between a tip and conductive substrate in an asymmetric DBTJ configuration (Figure 1a, Rgap ≫ Rsub), one may extract information about the particle’s core density of states (DOS). The technique was shown to be effective in probing spin polarization and magnetization dynamics in individual superparamagnetic NCs.24,25 However, there have

dvances in fabrication and scanning probe techniques have enabled detailed studies of electronic phase transitions in nanoscale systems.1,2 The Verwey transition in Fe3O4 was the first to be associated with charge ordering,3 followed by studies of electronic transitions in other metal oxides,4 metal chalcogenides,5 as well as organic conductors. The Verwey transition was described as a first-order metal−insulator transition accompanied by a structural phase transition where the (T > 123 K phase) cubic symmetry of the Fe3O4 crystal is broken by a small lattice distortion on cooling through the transition temperature.6 Since Verwey’s seminal paper in 1939, many aspects of the Verwey transition in bulk magnetite crystals have been studied.7 It is understood that the driving forces for the transition are the strong electron−electron and electron−lattice interactions in the system. Long-range charge ordering is believed to dominate below the transition temperature (TV), while short-range order is sustained well above it.8 The long-range order manifests itself by opening a gap in the electronic density of states (DOS) around the Fermi energy level (EF). This gap has been detected through various methods,9−11 including tunneling spectroscopy.12,13 Some photoemission experiments suggest that a reduced gap in the DOS, attributed to short-range ordering, still exists well above TV.9,10 More recently, the Verwey transition was observed in assemblies of magnetite nanocrystals (NCs)14 and in thin films,15 where the effects of grain size,16,17 surface,12 and overall particle shape18 were studied. Still, the exact nature of the transition and its manifestation at finite nanometric scales remain under debate, particularly the lattice and electronic structures of the crystal in the low-temperature phase. © XXXX American Chemical Society

Received: March 22, 2016 Accepted: April 18, 2016

1661

DOI: 10.1021/acs.jpclett.6b00644 J. Phys. Chem. Lett. 2016, 7, 1661−1666

Letter

The Journal of Physical Chemistry Letters

This work is, to the best of our knowledge, the first experimental report on the DOS evolution (and phase transition) in individual nanocrystals as a function of temperature using STM. Performing STS experiments on the same NC while slowly varying the STM stage temperature required a special method of fixing the magnetite NCs to the conducting surface. This was achieved by embedding the NCs within a gold surface, as illustrated in Figure 1a (see Experimental Methods for details). The STM topography images of the small “craters” in the gold film exhibited small bumps (∼1−2 nm high) which were suspected as magnetite NC tops. The craters always had elongated shapes, probably due to surface deformation during the peeling process of the gold film from the Si substrate. A comparison of I(V) curves taken on top of the bumps and out of the bumps revealed a few orders of magnitude difference between their zero bias conductivities (dI/dV). The lower conductivity measured on top of those bumps confirmed that they were the magnetite NCs. Numerous experiments were performed in which each sample was cooled and a specific NC was selected and probed through small temperature steps across the transition temperature. A significant change in the band structure was observed around 101 K in 5 of the 15 particles studied by STS across TV. The band gap value was visually estimated from the dI/dV curves as the distance between the up-turn points of the curve from nearly zero slopes. The change in the local density of states was from a ∼130−250 meV gap at 95 K to 75 ± 10 meV at 115 K. As an example, Figure 1b displays a sequence of detailed temperaturedependent I(V) measurements performed on a specific NC. It reveals a sharp transition in the band gap magnitude, from ∼130−140 meV to ∼75−80 meV when a particle is heated from 101 to 102 K. This can be clearly seen in the dI/dV plots shown in Figure 2, where also a slow increase in band gap with decreasing T below TV can be seen. It should be noted that the changes in the I(V) curves occurred only when the temperature was varied and not during topography scans at constant temperature. Consequently, the observation of the sharp transition in the DOS can be clearly associated with the Verwey transition in those NCs. The Verwey transition is observable only for magnetite near perfect stoichiometry (up to ∼5% deviation in the Fe3+:Fe2+ = 2:1 ratio).7 Hence, the observation of this transition in some NCs proves that at least part of them have a near-perfect Fe3O4 stoichiometry. To make sure that I(V) characteristics would not be affected during topography scans by NCs or contaminants sticking to the tip, we decided to use the special sample preparation technique, which had the particles firmly attached to the embedding gold film. In addition, we used very low tunnel currents to keep relatively large tip−sample separations and avoid variations in tip properties (and to have the DBTJ as asymmetric as possible, to be discussed later). An additional data set for a different NC exhibiting the Verwey transition is displayed in Figure 3, where the lowtemperature band gap seems to be larger (∼240 meV) than the gap at the previous particle. A >200 meV gap was observed also in another NC exhibiting the Verwey transition (see Supporting Information, Figure SI7). In both NCs it seems that the small apparent gap above TV is not a full gap, and there is some finite low density of states within this pseudogap, in contrast to the apparently full gap below TV. In the literature, the gap below TV for bulk magnetite and/or magnetite thin films has been reported to be about

Figure 1. (a) Schematic of the DBTJ structure of the tip−NP− substrate structure, where Rgap is the tip−particle vacuum gap resistance and Rsub is the resistance of the NP−gold substrate interface, which is primarily due to the organic ligands around the NP. (b) Evolution of the I(V) curves measured over a specific magnetite NC (marked with the arrow) along with an STM topography image of this particle (inset).

been only few reports of the temperature evolution of the electronic properties in single, isolated NCs of any kind.16,26 In view of the apparent importance of long-range charge ordering in determining the electronic properties below TV, an intriguing question arises: How small could the system be for the Verwey transition to be observed? Poddar et al.14 were able to observe a sharp Verwey transition in arrays of Fe3O4 NCs of average size around 5.5 nm. However, the size distribution of the particles used in their work was broad and it is not clear whether the observed DOS switching was driven by particles of particular sizes or by all of them. Poddar et al.26 presented scanning tunneling spectroscopy results, providing evidence that the transition occurs in isolated Fe3O4 nanocystals of diameters measuring only several unit cells. More recently, Yu et al. used fabricated nanoscale junctions to perform electron tunneling measurements on individual ∼10 nm nanoparticles trapped between gold electrodes.16 They reported the gradual opening of a ∼300 meV gap in the DOS well below the supposed TV and no gap above it. Measurements on ensembles of magnetite NCs showed that the transition temperature is lowered from ∼120 K to ∼100 K as the particle size is reduced.17 While this could be explained by changes in stoichiometry, as observed for bulk crystals,27 more data on the Verwey transition in small magnetite NCs is required to distinguish between stoichiometry and finite size effects. Single-particle measurements are particularly important for removing inhomogeneous broadening effects in particle ensembles due to variations in size, shape, crystallographic orientation, or charge distribution. Here we describe temperature-dependent STS experiments performed on individual chemically synthesized Fe3O4 nanocrysals using a liquid nitrogen-cooled variable-temperature STM (VT-STM) system. 1662

DOI: 10.1021/acs.jpclett.6b00644 J. Phys. Chem. Lett. 2016, 7, 1661−1666

Letter

The Journal of Physical Chemistry Letters

Figure 2. (a) dI/dV plots of the same I(V) curves and (b) estimated band gap value versus T extracted from this set of curves. Each curve is the average of 5 successive measurements.

Figure 3. (a) dI/dV curves from a single particle (different from the particle shown in Figure 1) showing a wide gap at 95 K and a narrow gap at 115 K, along with an STM topography image of the particle (inset) and (b) the temperature evolution of the estimated gap width for the same NC.

∼100−150 meV by photoemission measurements9,10 and ∼200 meV by STM.12,13 Recent theoretical calculations based on Xray crystallography data predicated a gap as wide as ∼500 meV.28 In most of the NCs, the I(V) curves did not change on cooling or heating through 101 K (between 95 and 125 K). In one of these cases, which we studied in detail, the hightemperature (full) band gap value appeared larger than the typical pseudo-band gap observed for the particles that exhibited the transition (see Supporting Information, Figure SI8). In that case, the band gap value was ∼100−120 meV, which is intermediate between the pseudo- and full-band-gap values of the two phases observed for the particles exhibiting the transition. This may be the result of off-stoichiometry and/ or excessive defects in the NC. All the data mentioned above were taken at relatively large tip−sample separation produced by setting the tunnel current to 10 pA, which corresponds to Rgap of ∼100 GΩ. At lower gap resistances (10 nm)17,31 or small NCs which were sintered into larger ones.32 One possible reason for this is that NCs of ∼10 nm would still be superparamagnetic around TV, which makes the observation of small changes in the magnetic moment of the NCs difficult. It is interesting to note that several reported TV values for magnetite NCs of ∼10 nm size were consistently around 100 K, in spite of being synthesized using very different methods.14,17 It can thus be concluded that the change in TV from about 123 K for bulk, stoichiometric crystals to ∼100 K for ∼10 nm NCs is not due to stoichiometry deviations, but to an intrinsic size effect. The same conclusion was reached by Hyeon and co-workers in their recent study of the size dependence of TV in NC assemblies.17 1663

DOI: 10.1021/acs.jpclett.6b00644 J. Phys. Chem. Lett. 2016, 7, 1661−1666

Letter

The Journal of Physical Chemistry Letters In contrast to the constant value of TV for our NCs, the band gap magnitude in the insulating phase seemed to vary between various particles, as can be seen in the difference between the data presented in Figure 3 (∼240 meV band gap) and Figure 2 (∼140 meV band gap). This could be a matter of extrinsic effects, such as defects and slight stoichiometry variations. Because the low-temperature phase is associated with longrange charge ordering, it can be expected that this phase would be more sensitive to (surface) defects in the NCs, while the high-temperature phase has only short-range (nearest neighbor) charge ordering and would thus be less sensitive to defects in general. It should be noted that extended defects, such as antiphase boundaries, common in thin films, are not typically observed at high-resolution TEM images of the NCs (Supporting Information, Figure SI1), and to date, the magnetic properties characterization of such NCs indicated a single-domain behavior33 even in single NC measurements.24 Structural phase transitions in nanocrystalline systems are well-known to have NC size-dependent transition temperature, such as in the ferroelectric−paraelectric transition in perovskites.34,35 Various mechanisms were suggested to be responsible for finite size effects in perovskite NCs, primarily involving surface effects, such as “negative pressure” and electrostatic effects.36,37 Because both the Verwey transition and the ferroelectric lattice distortion are strongly influenced by collective effects of electrostatic interactions within the crystal, it is expected that also finite size effects would have similar origins in both systems. The current work provides a detailed look at the changes in the electronic structure of magnetite NCs around TV. Our results conform to most of the recent literature about the band gap magnitude at the insulating phase of bulk magnetite, ∼ 200 meV. The situation at the high-temperature phase is more controversial, ranging between zero gap to a finite gap above TV, which gradually closes as the temperature increases. We observe ∼75 meV gaplike electronic structure above TV, but with a relatively low DOS within this gap. The sharpness of the observed Verwey transition in ∼10 nm single NCs is surprising. Considering the commonly accepted long-range ordered nature of the charges in the low-temperature phase, and that the unit cell for magnetite is ∼0.8 nm, it is remarkable that the lowtemperature phase can form in a particle spanning only 10−15 unit cells. Most previous experiments on magnetite NCs dealt with ensembles of particles and could not address this issue. The broader transition observed for an ensemble of ∼10 nm NCs is probably the result of inhomogeneous broadening.17 The recent single-particle study by Yu et al. did not reveal such a sharp Verwey transition,16 possibly due to differences in the preparation method. In summary, the Verwey transition in individual ∼10 nm magnetite NCs was studied using VT-STM, which is shown to be an effective tool in detecting an electronic phase transition in single nanostructures. The method is capable of temperaturedependent measurements with a high-temperature resolution, down to 1 K resolution. This enabled us to detect the sharp change in the electronic structure of individual magnetite NCs synthesized by the coprecipitation method. A surprisingly sharp change in the band gap width was observed around 101 K, where below it a ∼ 200 meV gap appears around the Fermi level. A pseudo-band gap of ∼75 meV exists above the transition temperature. It is believed that the major cause for the decreased TV in the ∼10 nm magnetite NCs relative to bulk crystals is a finite size effect caused by the presence of a surface.

The insulating band gap value appears to be sensitive to individual particle size or composition.

■

EXPERIMENTAL METHODS

■

ASSOCIATED CONTENT

Sample Preparation. The magnetite NCs were prepared using a coprecipitation technique reported previously.38 The NCs produced had a size distribution of ∼10 ± 3 nm and were coated with oleic acid and suspended in toluene. A small diluted sample of the particle suspension was spin-coated on a silicon substrate which was subsequently coated with 200 nm of gold using electron beam evaporation. (See also Supporting Information Figure SI2.) Flat stainless steel chips where then glued to the gold layer on the silicon chip and were peeled off along with the gold layer when the glue was dry. The STM imaging and STS were performed on the exposed flat gold surface, in which magnetite NPs are embedded. This embedding procedure ensures immobilization of the particles and a relatively good substrate−particle electrical contact. The substrate was then placed in the UHV chamber and briefly treated with an argon ion beam to remove some of the oleic acid on top of the NC and a few angstroms of the NCs’ surface layer, which may have been converted to maghemite because of the earlier exposure to air. STM Measurements. The sample was then placed on the VTSTM stage (Omicron VT-SPM) which was connected to a temperature-controlled liquid nitrogen operated cryostate. Typical STM topography images were acquired with tunneling current set points of about 0.1 nA and sample bias of −0.7 V. A characteristic topography image of the sample is shown in the inset of Figure 1a, where typical “craters” were observed in the gold film, in which the magnetite NCs were entrapped. I(V) measurements were performed on NCs identified within the “craters” using a bias range between −0.4 and 0.4 V, with ∼1.6 mV steps and imaging current settings ranging from 0.01 to 1 nA. Such measurements were performed on individual particles as well as the gold background at various temperatures ranging from 94 to 130 K. The temperature was measured by a sensor located on a cooling block attached to the sample holder while the actual sample temperature was a bit higher because of heat flow from the rail holding the sample holder. A calibration curve was used to estimate the temperature difference between the cooling block and sample, which was +5 K around 100 K. The overall sample temperature estimate acuracy was ±3 K with