Robin
Carter
a,
Mikhail
Suyetin
b,
Samantha
Lister
c,
M. Adam
Dyson
d,
Harrison
Trewhitt
d,
Sanam
Goel
d,
Zheng
Liu
e,
Kazu
Suenaga
e,
Cristina
Giusca
f,
Reza J.
Kashtiban
d,
John L.
Hutchison
a,
John C.
Dore
c,
Gavin R.
Bell
d,
Elena
Bichoutskaia
*b and
Jeremy
Sloan
*d
aDepartment of Materials, University of Oxford, South Parks Road, Oxford, OX1 3PH, UK
bSchool of Chemistry, University of Nottingham, University Park, Nottingham, NG7 2RD, UK. E-mail: Elena.Bichoutskaia@nottingham.ac.uk
cSchool of Physical Sciences, Ingram Building and , University of Kent, Canterbury, Kent CT2 7NH, UK
dDepartment of Physics, University of Warwick, Coventry, Warwickshire CV4 7AL, UK. E-mail: j.sloan@warwick.ac.uk; Fax: +44 (0)24766 92016; Tel: +44 (0)24765 23392
eNanotube Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Higashi 1-1-1, Tsukuba, 305-8565, Japan
fNational Physical Laboratory, Hampton Road, Teddington, TW11 0LW, UK
First published on 7th March 2014
In common with rocksalt-type alkali halide phases and also semiconductors such as GeTe and SnTe, SnSe forms all-surface two atom-thick low dimensional crystals when encapsulated within single walled nanotubes (SWNTs) with diameters below ∼1.4 nm. Whereas previous density functional theory (DFT) studies indicate that optimised low-dimensional trigonal HgTe changes from a semi-metal to a semi-conductor, low-dimensional SnSe crystals typically undergo band-gap expansion. In slightly wider diameter SWNTs (∼1.4–1.6 nm), we observe that three atom thick low dimensional SnSe crystals undergo a previously unobserved form of a shear inversion phase change resulting in two discrete strain states in a section of curved nanotube. Under low-voltage (i.e. 80–100 kV) imaging conditions in a transmission electron microscope, encapsulated SnSe crystals undergo longitudinal and rotational oscillations, possibly as a result of the increase in the inelastic scattering cross-section of the sample at those voltages.
Density functional theory (DFT) type investigations and experimental investigations indicate that SWNT-embedded crystals,4,13–15 including p-type (electron donor) embedded iodine16,17 and, more recently, embedded graphene ribbons18,19 are all profoundly modified with regards to their corresponding bulk properties. Additionally, a combination of theoretical and experimental approaches have demonstrated that n-type (electron acceptor) properties exist for SWNT embedded, for example, with zinc,20 copper21 and cadmium22 halides. Charge-transfer behaviour between SWNTs and molecular scale species (i.e. such as these crystals) and single molecules or molecular scale ions23 is strongly correlated with local perturbations in the densities of states (DOS) of the encapsulating SWNTs although at least of comparable interest are the fundamental changes predicted to occur in the physical characteristics of embedded low-dimensional crystals. For example, we have reported band-gap ‘flipping’ for low-dimensional ‘tubular’ HgTe embedded within 1.35–1.49 nm diameter SWNTs4,15 and, most recently, we have demonstrated that SWNT embedded GeTe crystals exhibit unprecedented phase change behaviour on the smallest possible scale.24 In the present investigation, we describe structural and theoretical investigations into low-dimensional crystals of SnSe. Like many metal chalcogenides, confined thin films of SnSe have long exhibited significant potential in optoelectronic applications such as holographic recording systems25 and solar cells.26 Of particular interest in the current study is how this material, which typically exhibits the orthorhombic Pnma form in the bulk27 (Fig. 1, top), but which also forms a Fmm rocksalt form (Fig. 2, bottom) under high pressure or thin film epitaxy growth conditions28 responds to confinement in a narrower class of SWNTs (i.e. diameter <1.4 nm). Bulk SnSe has an indirect band gap of 0.90 eV and a direct band gap of 1.30 eV.29 In addition to exploring any changes in local crystal structure we also wished to explore how this material would interact with the encapsulating tubules in terms of the composite physical properties.
Fig. 1 (Top) [001] and [100] ball-and-stick representations of the layered orthorhombic Pnma form of SnSe (for unit cell data, see Table 1). (Bottom left) Perspective view of the Fmm rocksalt form of SnSe. (Bottom right) Space-filling representation of and undistorted 2 × 2 crystal fragment derived from the Fmm form within an (8,8) SWNT. |
In the present study we note that numerous individual images of the 2 × 2 SnSe nanocrystals are obtained with the crystal viewed parallel to <001> relative to an ideal 2 × 2 rocksalt structure (e.g.Fig. 2(a)–(c)). Systematic measurements of the lateral spacings of these encapsulated SnSe nanorods (Fig. 2(b)) relative to the centre point of the SWNT wall indicate that the obtained microstructure is undistorted and does not deviate significantly from the idealised 2 × 2 structure in spite of some mild curvature in the SWNT and d<200> lattice spacings (i.e. defined relative to the rocksalt SnSe, see also Table 1) measured along the SWNT corresponded to 0.299 nm (±0.014 nm), consistent with d<200> for rocksalt SnSe.28 However, as we discuss further below induced lateral and rotational oscillation of the encapsulated SnSe nanocrystal in response to 80 keV electrons reduced the relative precision of lateral d<020> measurements to 0.28 ± 0.01 nm.
Parameter | Pnma | Fmm | ||
---|---|---|---|---|
Calc. | Exp. | Calc. | Exp. | |
a (nm) | 1.166 | 1.150 | 0.603 | 0.599 |
b (nm) | 0.420 | 0.415 | 0.603 | 0.599 |
c (nm) | 0.447 | 0.445 | 0.603 | 0.599 |
α (°) | 90 | 90 | 90 | 90 |
β (°) | 90 | 90 | 90 | 90 |
γ (°) | 90 | 90 | 90 | 90 |
Sn⋯Se (nm) | 0.300 | 0.302 |
Calculated and experimentally determined lattice parameters for the orthorhombic Pnma and cubic Fmm forms of SnSe are reproduced in Table 1. The computed lattice parameters for both bulk structures are in excellent agreement with experiment. In terms of the relative lattice energies of the Pnma and cubic Fmm SnSe forms, both compute with a marginally different energy (i.e. −1027.8121 eV for the rocksalt form versus −1027.7648 eV for the Pnma form, respectively) indicating that the latter is marginally more stable although is less commonly observed. In terms of the DOS (Fig. 4(a) and (b)), we find that the Pnma bulk SnSe form computes with a significantly wider band gap (i.e. 0.9 eV) in comparison to rocksalt SnSe (i.e. 0.68 eV) presumably as a result of the slightly more open structure.
As an initial input for the optimisation of the 2 × 2 SnSe microstructure, we utilised atomic coordinates estimated from a ∼10 nm long sub-domain of an extended lattice image of a 16 nm filled SWNT (i.e.Fig. 1(b)). Following preliminary calculations, we found that a 2 × 2 × 6 atomic rod of SnSe (Fig. 3(a), top) was sufficient to equilibrate the structure with a reasonable level of precision with the central Sn4Se4 cubic cluster optimising (in vacuo) with lateral and longitudinal Sn–Se distances equivalent to the experimental case (i.e.Fig. 1(d) and (e)). A further optimisation with an “infinite rod” crystal of SnSe was found to produce similar lateral and longitudinal Sn–Se distances within experimental error (Fig. 3(a), bottom).
Fig. 3 (a) Top model shows an optimisation of 2 × 2 × 6 Sn12Se12 rod using periodic boundary conditions and SnSe atom column positions initially determined from the yellow box in Fig. 2(b) as input data. Following optimisation, the bonds distort symmetrically around the central Sn4Se4 cluster (left bonds shown only). The bottom model shows a similar simulation but for an effectively infinite SnSe nanorod. Following optimisation, identical longitudinal and lateral Sn–Se bonds are obtained. (b) Effect on optimisation of central Sn4Se4 cluster following confinement in two different diameter but structural equivalent SWNTs (as per (c) and (d)). In the top case, the Sn4Se4 cluster is confined within a ∼0.95 nm diameter (7,7) SWNT and the lateral Sn–Se bonds (in yellow) are ‘compressed’ to 0.264 nm and this composite had a slightly positive encapsulation energy. The bottom Sn4Se4 cluster was optimised within a more accommodating ∼1.09 nm diameter (8,8) SWNT. In this instance, the lateral Sn–Se bonds (in green) do not significantly distort away from the equivalent bonds in (a) and this structure was obtained with a more favourable −3.019 eV encapsulation energy. (c) and (d) Perspective and end-on views of (2 × 2 × 6)SnSe@(7,7)SWNT and (2 × 2 × 6)SnSe@(8,8)SWNT composites, respectively. Atomic positions determined for the second case were used as input date for the multislice image simulation embedded in Fig. 1(c). |
We next observed the effect of SWNT confinement on the local geometry of the 2 × 2 × 6 SnSe rod in terms of the DFT optimisation. Fig. 3(b) shows two cases where the central Sn4Se4 cubic cluster in a 2 × 2 × 6 nanorod is optimised in a narrower ∼0.95 nm diameter (7,7) SWNT and then in a slightly wider ∼1.09 diameter (8,8) SWNT. In the first case (Fig. 3(b), top), the longitudinal Sn–Se distances optimise to 0.300 nm, comparable to similar Sn–Se distances in the central Sn4Se4 cubic cluster in the optimised 2 × 2 × 6 SnSe nanorod (Fig. 3(a), top) and also to those of Sn4Se4 units in the infinite SnSe nanorod (Fig. 3(a), bottom). However the lateral bond distances in the 2 × 2 × 6 SnSe rod optimised within the narrower (7,7) SWNT are found to be somewhat compressed to ∼0.264 nm in comparison either to the experimental case (i.e. 0.28 nm, Fig. 1(e)) or those determined for either of the optimisations in Fig. 3(a) (i.e. ∼0.275–0.277 nm). When the 2 × 2 × 6 nanorod is optimised in the slightly wider (8,8) SWNT (i.e.Fig. 3(b), bottom), longitudinal and lateral Sn–Se separations are obtained that are more compatible with both theory and experiment.
Together with the structural optimisations of the 2 × 2 × 6 SnSe nanorods we have computed comparisons between the 1D densities of states (DOS) for the finite and infinite cases in order to compare them against the two bulk forms. Computing the DOS for the bulk Pnma and Fmm forms in the first instance (Fig. 4(a)) reveals that the layered orthorhombic form has a wider 0.93 eV band gap than the 0.68 eV gap determined for cubic rocksalt form, which is a less open structure. Given that the 2 × 2 × 6 nanorod structure is more closely related to this more densely packed form of SnSe, we argue that the modification of the electronic structure should be seen within the context of this archetype. Computing the DOS for both the finite and infinite rod variants of (2 × 2) SnSe produces only comparatively small differences in the electronic structure (Fig. 4(b)) and a small difference in the band gap for the finite 2 × 2 × 6 SnSe nanorod (i.e. 1.41 eV) versus the band gap determined for the infinite SnSe nanorod (i.e. 1.36 eV). Both computed gaps represent a significant expansion of the band gap corresponding to the ‘parent’ Fmm form.
Fig. 4 (a) A comparison of the densities of states (DOS) for the two bulk forms of SnSe. Lamellar Pnma SnSe exhibits a wider band gap (i.e. 0.93 eV) in comparison to Fmm SnSe (0.68 eV). (b) DOS computed for finite (i.e. 2 × 2 × 6) rod SnSe as depicted in the top model in Fig. 3(a) overlaid on the DOS computed for an infinite (2 × 2) SnSe nanorod. In this instance there is only a marginal difference between the estimated band gap for the 2 × 2 × 6 case (i.e. 1.41 eV) versus that determined for the infinite form (i.e. 1.36 eV) while both exhibit a significantly expanded band gap relative to either of the bulk forms. |
In terms of the properties of the surrounding SWNTs, both (7,7) and (8,8) SWNTs must be expected to be metallic and these would no doubt shield or mask both the DOS and also the expanded band gap of the encapsulated SnSe crystals. On the other hand, exciting electron transfer in the latter would permit the metallic tubes to act as a conduit for electron transfer from the encapsulates to elsewhere which may have implications for sensor devices based on these and similar nanocomposites.
Fig. 5 (a) First SnSe shear structure (I) in a ∼1.5 nm diameter curved SWNT obtained with the top section in focus. (b) Lower down in the same SWNT, a second shear structure (II) is observed at a different focus, indicating an overall tilt in the SWNT. The two insets at bottom show the relative shears of I and II (inset scale bar for II, 0.5 nm). These structures are significantly distorted relative to rocksalt SnSe (Fig. 1) and are apparently expanded but this may be due to staggering of the Sn and Se atoms. The two shear states I and II involve effectively 20° and −18° tilts away from an idealised (100) rocksalt lattice plane in a normal relationship to the SWNT axis. (c) Schematic model of shear inversion of I and II. Top right inset model indicates how staggering within the Sn and Se atom columns (observed within Pnma SnSe) can possibly lead to gross distortions in sheared rocksalt SnSe. Bottom right inset shows a multislice image simulation based on the main model. Fast Fourier transforms (FFTs) of shear structures I and II reveal mirror plane related diffraction behaviour which is useful in preferential dark field imaging techniques. |
A further consequence of these distortions is the observation of differential strain states insets that invert either side of a curve in a bent SWNT (i.e.Fig. 5(b) insets). This behaviour is similar though not identical to shearing behaviour that we reported for 3 × 3 AgI nanocrystals observed within ∼1.6 nm diameter SWNTs.36 In this case, the observed transformation was between rocksalt-like 3 × 3 AgI nanocrystals and a sheared version of this nanostructure. In the case of this new version of phase change behaviour, the process is shear inversion from one strain state (i.e. region I, Fig. 5(a)) to a second strain state (i.e. region II, Fig. 5(b)) that occurs either side of induced curvature in a SWNT in which the direction of shear is reversed (Fig. 5(b), inset and main model, Fig. 5(c)). The intermediate zone between the two sheared regions is filled with distorted and apparently amorphous SnSe. We note that this behaviour also differs from the recently reported electron beam induced crystalline-amorphous phase transitions in SWNT embedded GeTe nanocrystals.24
It is notable that both the shear states I and II inset in Fig. 5(b) involve small 20° and −18° distortions respectively relative to idealised 3 × 3 rocksalt fragments. This mode of shear inversion suggests a method for its automated identification if domains of similarly sheared nanocrystals can be induced in aligned bundles of SWNTs or similar. As shown in the inset in Fig. 5(c), a multislice image simulation (based on atomic coordinates extracted from strain states I and II in Fig. 5(b), inset) effectively reproduces the image contrast of the experimentally imaged shear inversion domains inset in Fig. 5(b). Fourier transforms of I and II produce two scattering patterns related by a mirror image. In real space dark field imaging, these patterns could be used to preferentially filter out either strain state allowing the other to be imaged selectively and dynamically, an effect that can also be reproduced offline digitally.
In Fig. 6(a), we show two modes of beam-induced oscillation in another section of 2 × 2 SnSe encapsulated within a ∼1.1 nm diameter SWNT. The section of crystal on the right hand side of the SWNT behaves like an oscillating coiled spring with this domain of the crystal apparently expanding and contracting over the ∼12 s period of image acquisition. The left hand section of crystal exhibits a more complex mode of oscillation in that this entire region of crystal rotates about the axis of the SWNT, and the crystal rotates from a projection equivalent to <110> for rocksalt SnSe to <100>. In the former orientation (i.e. top three images in Fig. 6(a)), the crystal appears to have three layers but in fact, as the corresponding models and simulations in Fig. 6(b) show, the 2 × 2 fragment is being observed at a 45° rotation relative to the corresponding <100> orientation. An animation, including both the images included in Fig. 6(a) and other additional images from the same image sequence, reveals the nature of oscillational modes in more detail than is possible with static 2D images is reproduced in the ESI (i.e. Fig. S1†).
One practical consequence of the observed vibrational behaviour of embedded SnSe and other crystallites within SWNTs has been that it has been difficult for us to obtain focal series of these encapsulated crystals that are of suitable quality for use to be able to obtained exit wave reconstructed phase images which was found to be more feasible under higher voltage (i.e. 300 kV), low electron dose imaging conditions.1,4,32,36 Reconstructed phase imaging38,39 has the advantage that further residual aberrations can be removed from the final phase image which can be obtained at even higher resolution than by using hardware aberration correction alone and also that more 3D information can be extracted from the resulting phase images which are sensitive to out-of plane phase shifts.40 Intriguingly, we do not see similar vibrational effects for empty SWNTs under the AC-TEM imaging conditions employed in this study. This suggests that there is more inelastic energy transfer between low kV electron beams (i.e. 80–100 kV) and the encapsulated crystals than at higher voltages which, at electron energies significantly greater than the threshold value of 86 keV, tend to cause significant knock-on damage with respect to the graphene carbon lattice of the encapsulating SWNTs.41 Most remarkably this indicates that the inelastic scattering cross-section for encapsulated nanowires is still significant even when they are as narrow as ∼0.8 nm, as is the case for 2 × 2 SnSe (Fig. 6(a)).
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4dt00185k |
This journal is © The Royal Society of Chemistry 2014 |