Low Energy Twist Defects in Pentatwinned Silver Nanowires

Transmission electron microscopy investigation of networks of pentatwinned silver nanowires has identified the presence of nanoscale defects within individual wires, showing narrow regions of banded contrast normal to the nanowire axis (bamboo faults). Structural analysis using machine learning decomposition of scanning precession electron diffraction data identifies these bamboo faults as a pair of twist boundaries normal to the wire axis creating low energy coincident site lattices between the faulted region and the parent nanowire. This leads to a conservation of the relative misorientation of the twinned structure within the faulted region, leading to an apparent local rotation of the nanowire. Their presence after spraying and nonthermal processing suggests that they may form during nanowire synthesis. However, examination of the networks before and after cyclic straining finds an increase in the density of the defects, indicating that they may also form as a result of mechanical deformation.


Introduction
Ag nanowires (NWs) can be easily synthesized using solution chemistry via the polyol process. [1]This ease of fabrication, coupled with the relative inertness of elemental Ag, allows their use in a number of applications.NWs with diameters < 100 nm have attracted a great deal of interest as transparent conductors for thin-film optical device technologies because of their superior mechanical properties coupled with being good electrical conductors and their relatively small optical extinction. [2,3]These are advantages over current indium tin oxide (ITO), which has poor flexibility and is prone to brittle failure.Transparent films made by Ag NWs show <2% increase of their mean sheet resistance after 100 cycles of bending with no change in light transmission. [2]Individual Ag NWs also show an apparent size effect, where the mean strength of the Ag NWs is several times stronger than bulk Ag. [3] Hence, polymer films containing networks of metal NWs compare favorably with other candidate systems across all three of strength, transparency, and conductivity. [4]Therefore, a potential application area for Ag NWs is their use in sparse networks that are optically transparent and electrically conductive.Just above the percolation threshold, these networks display sheet resistance values <30 Ω □ À1 and optical transmittance > 90%. [5]If these networks are deposited on flexible transparent polymer films, the resulting supported networks can be used as stretchable and flexible transparent conductors with a range of applications in portable, conformable, wearable, and body-mounted electronic systems.
There have been a number of investigations of the performance of flexible transparent conducting films made from Ag NW networks after repeated cyclic loading.These studies indicate that repeated flexing of films leads to damage characterized by a gradual increase in electrical resistance that becomes significant >10 5 strain cycles. [6,7]However, the mechanisms by which this damage occurs are still unclear, beyond the observation that a small number of the NWs within the network appear to fracture during cycling [6,8] and this is the subject of ongoing study, which will be reported in a future publication.The mechanical properties of individual metallic nanopillars and wires with diameters < 1 μm have been the object of considerable study in recent years.15] A further complication to understanding the behavior of Ag NWs formed through the polyol process is their characteristic pentagonal cross section.This forms because each wire grows along a 110 direction through Ag attachment on the low energy f111g facets of each segment of a decahedral seed particle. [1]The resulting wire comprises five subcrystals each of which is a triangular prism bounded by f111g-type twin planes (Σ3 tilt boundaries) between the subcrystals and a f001g facet at the outer surface of the wire.The two f111g boundary planes intersect at an angle of 70.53°. [16]However, a geometrically complete pentaprism must contain five triangles with apical angle 72°.Thus, the observation that five triangular prisms are joined together by Σ3 tilt boundaries to form the pentaprism implies an angular mismatch of approximately 7.4°that must be accommodated through elastic deformation.
[11][12] In these studies, deformation has typically occurred through the passage of dislocations formed randomly at the free surface of the subcrystals, which may imply that the subcrystals behave as five discrete NWs.However, there are reports proposing that partial dislocations nucleate across pairs of twin planes to create a tetrahedral fault across the full wire. [21]The geometry of the wire and the degree of coherency then force similar nucleation in the other subcrystals leading to the decahedral structure. [21]Despite the differences in local dislocation structures, a recent study comparing the mechanical properties of single-crystal and pentatwinned Ag NWs found little difference in their strengths at the same diameter. [20]he situation for different loading configurations is more complex.Studies of pentatwinned wires deformed in bending show a preference for particular angles of deformation that maintain higher levels of coherency between the different subcrystals of the NW. [22]Similar behavior has also been reported for the deformation of parallel arrays of NWs. [23]ere, we investigate another defect type found in both undeformed NWs and strain cycled NW networks.To our knowledge, these defects have been previously unreported in the literature and their presence and formation during strain cycling may have an important role in the degradation of Ag NW networks in service.

Experimental Section
Networks of Ag NWs were deposited using an identical procedure to that reported in an earlier study of Ag NW networks on polyester films. [7]A summary of the method is presented here.Pentatwinned Ag NWs, synthesized by the polyol process, were acquired from commercial sources (NovaWire-Ag-A60, Novarials, Woburn, MA, USA) with a mean diameter of 60 nm and length 10 μm (manufacturers data).Two different substrates were used in this study: 1) In order to measure the change in electrical conductivity, polyethylene terephthalate (PET) substrates of 300 μm thickness were used (Thermo Scientific, Loughborough, UK); and 2) To support the NW networks and to allow observation of the NWs in the transmission electron microscope (TEM) before and after cyclic straining, a piece of porous polycarbonate (PC) film (pore size 8 μm, Millipore UK Ltd., Watford, UK) was used as the substrate.This was first covered by an electron transparent collodion film of approximate thickness 200 nm; this supports the Ag NW networks and ensures that the film was not damaged by the electron beam. [20]g NWs were deposited on the PET and PC/collodion substrates using an Infinity airbrush (Harder and Steenbeck, Norderstedt, Germany), with a spray-head to substrate distance of 12 cm.This will result in a coverage of approximately 0.06. [7]o promote electrical conductivity after spraying, we used the procedure of Hwang et al., [6] applying a normal pressure of 29 MPa for15 s under ambient conditions using a laboratory hydraulic press.This pressure treatment disrupts polymer surface films present on the NWs from the polyol manufacturing process and forms conducting pressure welds between the NWs.After application of the pressure, the PC/collodion film containing the Ag NW network is approximately 10 μm in thickness.
Cyclic loading experiments were carried out using an in-house designed and built testing rig (Figure 1c) reported in an earlier study. [7]The sprayed NW networks were first fixed onto a flexible PET film of thickness 100 μm using double-sided tape and then mounted onto the belt of the testing rig, which was of thickness 340 μm.The flexible belt is driven forward and backward around a spindle of radius, R = 4 mm, to generate repeated strain with a frequency of approximately 1 Hz.The magnitude of the cycled strain can be controlled by changing the radius of the spindle.With the 4 mm spindle, the maximum strain applied on the NW networks is calculated to be about 5% (see Supplementary Information for the formal relationship between spindle radius, film thickness, and strain magnitude).Networks were tested with up to 300 000 strain cycles.
While most studies highlight the behavior of individual wires under reasonably idealized loading conditions (usually in pure tension), there is the concern that this does not translate well to the actual loading the wires would be subjected to in a real device.To explore this further, a workflow was developed to allow sprayed networks of NWs to be imaged before and after being subjected to cyclic loading.After fabrication, the sample on the PC/colloidon substrate was put into the TEM (Figure 1a) to observe and characterize the Ag NW network after manufacture.These films were then mounted for cyclic testing as described above (Figure 1b,c).After testing, the sample was inserted once again into the TEM to characterize them after testing and compare specific NWs with their untested state (Figure 1d).The characteristic pores in the PC film were used as fiducial references to allow the same NWs to be identified and compared before and after testing.
Sample characterization before and after testing was conducted using a Tecnai TF30 (FEI, Eindhoven, The Netherlands) and Talos F200A (Thermo Scientific, Eindhoven, The Netherlands), operating at 300 and 200 kV separately.For scanning nanobeam electron diffraction experiments, Tecnai TF30 was controlled using a Nanomegas Digistar apparatus.The size of condenser aperture was 10 μm corresponding to a convergence semiangle of %5 mrad, and the microscope spot size was 8, leading to a probe diameter of approximately 1 nm.
High-resolution TEM images provide the data to calculate the density of the bamboo defects.The following equation determines the linear density, ρ, of the bamboo defects in the NW network (ρ) is calculated from the number of faults counted in a set of TEM images (n) using Equation (1).
where n is the number of faults counted in each TEM image (2.6 μm Â 2.6 μm in this study), A is the area of the image, and C is the coverage of the NWs.The mean and standard deviation of the linear density is calculated from 12 images; each image contains approximately 15 NWs.

Results and Discussion
Silver NW samples were studied in the TEM to explore the type and distribution of defects.An example image from a NW network sprayed onto a substrate is shown in Figure 2a.Within each NW, there are sparsely distributed areas of contrast that are believed to be defects aligned perpendicular to the NW axis.These widely spaced defects are henceforth described as bamboo faults.The presence of the faults in the as-sprayed networks suggests that they may be present within the NWs as a growth defect.However, observations of identical regions of the NW network after mechanical consolidation and cyclic flexural straining indicate that they are influenced by mechanical deformation.Figure 2b shows a region of the NW network as deposited on a PET film prepared for in situ comparison of samples before deformation and free from any defects.Figure 2c shows the same region after 50 000 loading cycles, where the maximum network strain was %5%; as described in the methods section, narrow bands of enhanced contrast have appeared.These are equivalent to the bamboo faults shown in Figure 2a but in this case the defects have formed during a period of cyclic mechanical deformation.It is notable that there is no evidence of any change in the local NW diameter around the features.Figure 2d shows another region of the NW network that contains bamboo defects within undeformed NWs; the complementary Figure 2e, of the same region after 150 000 deformation cycles, shows the defect to have disappeared (note that this NW appears to have fractured away from the defective region during strain cycling).Finally, Figure 2f shows the density of bamboo defects per μm of NW length measured through TEM observation before and after cyclic deformation of the NW network.Thus, we conclude that these defects are strongly influenced by cyclic deformation with defects being generated and possibly either being mobile or annihilating as cyclic loading occurs.
A change in contrast in a TEM image implies a change in the local crystal orientation, with the lattice rotating to a condition where more of the incident electron beam is scattered, rather than simply a change in the thickness of the sample.The contrast line is vertical to the axis of a NW.Considering the growth mechanism of the NW, any increase in thickness would be expected to be accompanied by a change in NW diameter and this is not observed in Figure 2c,d.[29] The decomposition factorizes the SPED data into a series of basis diffraction signals and maps, indicating the physical localization of these signals.This, in turn, allows the position of particular crystal orientations to be identified by comparison with the expected diffraction pattern geometry.This approach is needed because there is a high degree of coherency between the different subcrystals; given that they all share a common <110> direction   along the wire axis and the boundaries between subcrystals are highly coherent twin planes.Hence, there are many common reflections from each subcrystal (such as the common {220} in all Figure 3b-d).Therefore, conventional dark-field or virtual dark-field (VDF) imaging is unreliable for analyzing the location of particular orientations. [21,22]Similarly, the overlap of the different subcrystals along the electron beam makes template matching approaches equally unreliable.
Figure 3a is a stereographic projection along the [110] growth direction, showing the major orientations of the NW subcrystals with respect to the beam direction.Here, the beam travels along a path inclined at around 6 °to one of the ð001Þ facets of the NW, resulting in the other subcrystals having orientations close to, but not exactly aligned with 110, 114, and 111. Figure 3c shows a combination of the factors attributed to orientations close to 111 from within and around the region between the pair of defects, with the edges of the NW indicated by dotted lines.Outside the region bounded by the defects, the 111 oriented subcrystals lie along the left-hand edge of the NW, although there are two subcrystals with orientations close to 111 their proximity and structural similarity makes it difficult for the decomposition to separate them, hence the appearance of a single large region of the NW oriented this way.Within the region, the 111 signal appears to have moved to the right-hand side of the NW.Similar combined images with regions oriented close to 110 and close to 114 are shown in Figure 3b,d, respectively.In all three cases there is a clear change in the position of the orientations within the NW width.However, there is little or no variation in the projected shape of the wire, with the region between the two defects showing an abrupt transition.
On comparing the relative orientation change between the subcrystals in the unfaulted NW and the region located between the two faults, there is a sequential transformation in the crystal lattice rotation as one progresses around the NW, such that using the labeling indicated in Figure 3a the original orientation sequence ABCDE in the unfaulted wire becomes DEABC within the faulted region.This gives an apparent rotation of -144°within the faulted region.
SPED analyses from two other faulted regions are presented in the Supporting Information (SI).In the first (Figure S1, Supporting Information), a different sequence of image translations are seen, which are consistent with the faulted region undergoing a different set of lattice rotations of approximately 72°within each subcrystal.This transforms the faulted region orientation sequence from ABCDE to BCDEA or an apparent rotation of the faulted region through þ72°.The second of these (Figure S2, Supporting Information) is associated with a parallel lattice rotation of approximately 51°, which generates orientations of each of the subcrystals that are distinct from the original crystal lattice orientations.
We propose that the abrupt change in crystal orientation observed between the boundaries visible in the TEM images is explained by the presence of a pair of identical low energy planar crystal defects.We believe that these defects are semicoherent twist boundaries based on coincident site lattice (CSL) interface models.For rotation about the [110] direction, a series of semicoherent boundaries can be created by rotation of the lattice within each individual subcrystal. [30,31]The initial relative orientation of the five subcrystals is defined by the pentatwinning crystallography and the tilt associated with the f111g Σ3 boundaries separating them.Table 1 shows the initial relative orientation of the five subcrystals in a single NW and the resulting transformation of the apparent orientations following lattice rotations consistent with the presence of the three lowest energy semicoherent twist boundaries about the common [110] direction of the subcrystal axes predicted by coincidence site lattice theory for rotations.It can be clearly seen that the Σ3 and Σ9 rotations produce the two apparent shifts in lattice orientation sequence determined by the SPED analysis of the faulted regions in the NWs, consistent with the observations of Figure S1 and S3, Supporting Information, respectively.The Σ11 rotation generates new lattice orientations that are different from the initial subcrystals within the original NW, consistent with Figure S2, Supporting Information.
We note that each bamboo defect characterized is formed by a constant lattice rotation of all five of the subcrystals, hence the original 70.53°tilt misorientation between adjacent crystal lattices is conserved.However, while the boundary between the subcrystals in the original NW is of symmetrical twin character, the rotated crystal segments are separated by an asymmetric boundary, inclined from the symmetric twin plane by AE the angle of the semicoherent twist boundary.A schematic of the crystallographic relationships between the subcrystals (and unrelaxed atom positions) within the three faulted regions is presented in Figure S3 in the Supporting Information.Calculations of such asymmetric tilt boundary energies in Cu, with a similar low stacking fault energy to Ag, show that the boundaries, although more energetic than the symmetric tilt, are still expected to be relatively low energy configurations. [32,33]However, this implies that beyond the twist fault, the NW will have higher energy internal boundaries and consequently there is a driving force for the formation of an opposite sense twist boundary to reduce the internal energy of the NW.This could explain the "bamboo" character of the defect, with a relatively small, faulted region present between a pair of twist defects.A brief analysis of the defects found that the length of the faulted region is approximately constant.We hypothesize that the three twist faults are of similar energy with the higher energy of the Σ9 and Σ11 twist boundaries compensated by the lower energy of the 38.94°and 50.48°asymmetric tilt boundaries between the subcrystals when compared to the Σ3 twist which results in a 70.53°deviation from the symmetric boundary.A more systematic further study of the width of the defects and their associated lattice rotation is needed to confirm this.
Table 1.The effect of low-Σ rotation CSLs about the ½110 direction on the orientation of the individual subcrystals in a pentatwinned NW.The Σ3 and Σ9 twists lead to an apparent reorientation of the complete NW, whereas the Σ11 twist generates five new orientations.The only previously studied and modeled stacking fault in pentatwinned Ag NWs is the decahedral model described by Filleter et al. [21] This uses the slip associated with a pair of partial dislocations to bound the faulted region.For a subcrystal with a growth direction of ½110, the partials act not on the f111g planes forming the Σ3 boundaries with the adjacent subcrystal, but instead act on planes parallel to the subcrystal end terminations with index ð111Þ and ð111Þ.The partials are of type a 6 ½112 and a 6 ½112, respectively.A model of this fault in one of the subcrystals is shown in Figure 4a viewed along the wire axis; the corresponding fault generated in all five subcrystals is shown in Figure 4b.From the structural model, all of the regions in the structure that have an orientation close to ½110 are shown in Figure 4c.The notable feature here is that the materials located within the fault undergo little or no rotation compared to the material outside the fault.Hence, this would not produce the sorts of changes in the orientations seen in Figure 4.For comparison, a similar sequence of images in Figure 4d-f shows how the presence of the Σ9 twist results in an apparent rotation of the NW-compare this with Figure 3.The analysis presented here shows that the high contrast features visible in the NWs do not exhibit the expected crystal orientations predicted by the decahedral fault model.While there is evidence from other studies of axial loading of pentatwinned NWs that deformation on the ð111Þ=ð111Þ habit planes routinely occurs it seems that there may be additional faults that might exist in the pentatwinned NW system.The decahedral fault model does, however, provide the template for how deformation in one subcrystal of the pentatwinned structure can result in a common behavior in all of the subcrystals.
As discussed in a previous TEM study on bending Ag NWs, [22] there is a large energy cost to any transformation occurring within a single subcrystal because a macroscopic change in the subcrystal shape will need to be accommodated given the coherency within the wire.Lattice rotation appears to occur uniformly in each of the subcrystals, resulting in the consistent wire shape and the flat interfaces seen in the TEM micrographs.While the rotation presented in this study does not maintain Σ3 twin planes between subcrystals, there are still semicoherent asymmetric tilt boundaries that can minimize the energy of the NW and the defect.Note that the faults are bounded by two twist boundaries, hence no macroscopic permanent NW rotation is implied.
Therefore, a twist boundary fault on the {110} plane provides a simple explanation for the appearance of the faults, with the resulting apparent rotation of material within the fault matching that observed in the experimental data.However, this gives no insight into the mechanism for the formation of such defects in the NWs.Although the presence of these defects in the as-received NWs (Figure 1a) is consistent with their formation as a growth defect, our TEM investigations before and after cyclic straining (Figure 1b,c) suggest these defects may also form as a result of mechanical deformation.Indeed, Figure 2f shows an increase in the density of bamboo defects after strain cycling up to 300 000 cycles.However, the NW network has been subject to a cyclic tensile stress and such a deformation regime would be more likely to generate decahedral faults as suggested by Filleter [21] and reported by others as forming during the tensile extension of pentatwinned NWs. [20]We have not carried out an exhaustive search for the presence or absence of other defect types within the cycled NWs because this study is focused on characterizing the bamboo defects.We propose that while the stress applied to the whole network during cyclic straining is macroscopically tensile, the individual ligaments within the network are unlikely to be aligned such that the tensile stress is transmitted directly to each of them.Instead, a range of tensile, shear, bending, and torsional forces would seem reasonable for individual microscopic elements within the network, depending on the local orientation and constraint of the ligament.
To the best of our knowledge, there has been no systematic study of the local forces that occur within a NW network when subject to a tensile strain, to support the above argument.However, there has been some work looking at the mechanical behavior of random fiber networks with reference to the behavior of paper and nonwoven textiles.Although these are typically much denser networks than those used in transparent conducting films, some trends can be observed, which can be used to interpret our results.Bergstrom et al. reported on a discrete element model of a network made from uniform fibers with properties similar to wood fibers. [34]This considered a random network of fibers of diameter 16 μm, almost three orders of magnitude greater than our Ag NWs, but with an aspect ratio of 120 that is similar to the mean aspect ratio of the NWs in this study, A R = 166.If we assume a scale independence of the elastic properties, the important features that govern network behavior are the aspect ratio and the density of fiber junctions, which will scale with the relative areal density, or coverage of the network.The coverage range reported by Bergstrom et al. was >0.10, which is slightly higher than the coverage of our networks; however, their results and trends seen with changes in coverage give an important insight into the expected behavior of our networks.Their results show that at the lowest coverage, the majority of the elastic strain energy was stored by bending of the fibers but a small but significant amount (%7%) is stored in torsional deformation.A more detailed analysis showed that the torsion component increases to >60% close to the junction between the fibers.The trend shows that the torsional component of the stored energy increases as coverage decreased, indicating that the lower coverages used here would lead to an increase in the torsional component of the stored energy when the network is strained in tension.A further point to consider is that Bergrtrom's study assumes load is transferred through Coulomb friction at the wire junctions, whereas in our NW networks the fiber junctions are formed through pressure welding and would transfer load more efficiently, possibly further increasing the torsional loads; however, this would need to be confirmed by further modeling.
The effect of local torsion would be to create and propagate dislocations with Burger's vectors normal to the wire axis (screw dislocations), as opposed to the Burger's vectors associated with the decahedral fault that have a large component parallel to the wire axis.As with any increase in dislocation density, this raises the possibility of recovery or recrystallization as a mechanism for reducing the dislocation density and hence the stored elastic strain energy.During this process, dislocations either annihilate or rearrange themselves into energetically favorable configurations that in extremis become new interfaces or boundaries in the crystal structure.For this to be thermodynamically favorable, the new boundaries should be of a low energy such that the new configuration has a lower energy than the previous dislocation line energies, which is consistent with the low Σ, coincident site boundaries observed in this study.Importantly, the shear from nonannihilated dislocations is retained and so defines the misorientation of the new boundary.In the case observed here, the possibility is raised that the individual dislocations are rearranged to form a small volume bounded by a pair of defects that can be described either as rotation boundaries or as arrays of screw dislocations.This is a direct analogy to annealing twins, where a high dislocation density is rearranged into a region bounded by a pair of symmetric low energy tilt boundaries, which can also be considered as arrays of edge dislocations.
The implication of this analysis is that the faults are local regions of the NW that have undergone a common lattice rotation to generate a pair of low Σ interfaces (Table 1) in each subcrystal, as part of a recovery process.Given their complexity, it is unlikely that the boundaries form spontaneously during a single loading/unloading cycle; instead, they are probably the result of dislocation accumulation after multiple deformation cycles.Although the NWs and subcrystals are small enough for dislocation starvation and dislocation loss through image forces to occur in the pristine wires, [13] during mechanical loading dislocations can nucleate at the outer surface of the NW and extend into and be retained within the subcrystals, [14,15,19] with such populations recovering to generate low energy defect structures.It is not immediately clear why the defects are relatively sparse within the network.The cyclic deformation of 0-5% suggests these must be an efficient way to remove a large dislocation density from the NW.So, the formation of a fault may exhaust a relatively large volume of retained dislocations.There may also be substantial difference in the rate of dislocation accumulation along the wires, meaning specific regions (perhaps close to network junctions) are more likely to become faulted.There may also be repulsive energy between faults because an array of screw dislocations may induce a back-torsion in the wire (similar to the Eshelby twist [35] that separates the faults.Work in progress has found that the magnitude of strain cycling has a quantitative effect on mechanical and electrical performance of Ag NW networks, with a higher electrical resistance measured, after a given number of cycles, with increasing strain magnitude.However, given the nature of the SPED analysis we have not yet carried out studies of how strain amplitude affects the presence or nature of the bamboo defects and this is the subject of further study. Bulk FCC materials with nanotwinning have been reported showing high strength while maintaining good ductility (14% for Cu). [36,37]This behavior is attributed to the highly coherent structure of TBs.The twin structure provides a barrier to dislocation motion and hence increases the strength of the materials.Under cyclic deformation of nanotwinned bulk metals, dissociated screw dislocations can pass through the twin boundaries aligned parallel to the loading direction.Twin boundaries aligned perpendicular to the loading direction; slip is hindered but the boundaries are still traversed by the dislocations, which are present in slip bands. [38,39]These lead to a high strength with increased ductility compared with nanocrystalline metals because the twin boundaries are less efficient as crack nucleation sites than high-angle grain boundaries.In the pentatwinned NWs, the twin boundaries are parallel to the loading direction when a network is deformed, and this may explain the observation that there is little difference in the tensile strength of pentatwinned and single-crystal NWs. [20]However, to the best of our knowledge, there have been no reports on the torsion behavior of bulk nanotwinned metals, hence it is difficult to compare their behavior with our observations of the defects in NWs that are proposed to be the result of torsional loading.

Conclusions
Careful TEM investigation of regions within pentatwinned Ag NWs has identified an extended defect normal to the [110] axial direction.SPED analysis of these defects reveals that they consist of a pair of low energy twist boundaries, separated by a distance slightly smaller than the NW diameter with rotations of 38.94°, 50.48°, and 70.53°, consistent with known coincident site lattices of Σ11, Σ9, and Σ3, respectively.The uniform lattice rotation in each subcrystal, coupled with the symmetry of the pentatwinned NW structure, can lead to an apparent transposition of the subcrystal relative orientations within the faulted region, leading to an illusion that the inner segment of the faulted region has undergone a macroscopic rotation of À144°and þ72°for the Σ9 and Σ3 twist boundaries.It is proposed that the small extent of the defect is associated with an increase in the energy of the interfaces separating the five subcrystals within the faulted region.
The defects appear in NW networks imaged after spray deposition and are thus potentially defects associated with the NW growth mechanism.However, correlated TEM images of the networks before and after cyclic flexural deformation show that there is an increase in density of the defects after deformation.It is tentatively proposed that these defects may form after torsional deformation through the interaction of screw dislocations during a recovery process, analogous to the formation of decahedral defects that have been proposed to result from the passage of edge dislocations in these structures.These defects are relatively sparsely distributed, which may reflect the relatively small torsional component of the deformation processes that occur when a NW network is strained in tension.This and other questions concerning the relative prevalence of this defect and other defects, e.g., the decahedral defect associated with tensile deformation, require further study.

Figure 1 .
Figure 1.Schematic workflow of the TEM imaging procedure to study the deformation of Ag NWs after cyclic straining experiments.a) The Ag NW network on the collodion film mounted on a nanoporous PC substrate is imaged in the TEM prior to deformation.b) The specimen is removed from the TEM and mounted on a PET film.c) The PET film is mounted on the loading belt and cycled repeatedly around a spindle to generate a cyclic tensile strain for a predetermined number of cycles.d) The sample is removed and reinserted into the TEM for imaging.The pores in the PC substrate act as fiducial markers to allow the same area as imaged in (a) to be studied postdeformation.

Figure 2 .
Figure 2. TEM images of Ag NW networks.a) As-sprayed network showing mean diameter of the NWs to be approximately 60 nm and the presence of defects normal to the NW axis.b) Section of a network imaged prior to cyclic flexure loading; there are no defects present.c) Image of the NWs seen in panel (b) after 50 000 flexural cycles showing the presence of defects.The defect outlined in green was used in further SPED analysis.d) Section of a network imaged prior to cyclic flexure loading; there are defects present in the region indicated by the white oval.e) Image of the NWs seen in panel (d) after 150 000 flexural cycles; the defect seen prior to mechanical cycling has disappeared.f ) The mean linear density of the bamboo defects as determined from TEM images of the NW networks before and after cyclic deformation.

Figure 3 .
Figure 3. a) Schematic illustration showing the cross section of the pentatwinned silver NW, for the labeled electron trajectory relative to the f001g face of subcrystal A, superimposed upon a stereographic projection along the ½110 growth direction of the NW.The approximate relative orientations of the subcrystals B-E are indicated as the nearest low index direction in the A subcrystal parallel to the rotated subcrystals B-E.b) NMF decomposition factors (left) and loadings (right) corresponding to subcrystal sections oriented close to 110 in and around one of the faults.The approximate NW edges are noted in the factors and the scale bar corresponds to 5 nm.In the loadings, the key reflection types used to identify the orientations are indicated.c,d) The NMF decomposition outputs corresponding to 111 and 114 orientations, respectively.

Figure 4 .
Figure 4. Upper row: a) the projected structure along the NW axis for a single subcrystal undergoing decahedral faulting.b) The same view for the decahedral fault occurring in all five subcrystals.c) The positions of material oriented close to <110> across the decahedral fault.Lower row: d) the projected structure along the NW axis across the Σ-9 rotation fault in a single subcrystal.e) The same view in all five subcrystals showing the effect of an identical twist in each.f ) The location of the subcrystal oriented closest to the <110> direction across the boundary (cf., Figure 2b.).