Bulk and Surface Electrical Properties of BiSb on Flat and Ion‐Beam Nano‐Patterned InP Substrates

One route to the improvement of thermoelectric material performance is nanostructuring. Herein, BiSb thin films are grown by molecular‐beam epitaxy on two types of InP(001) substrate: flat epiready wafer, and nano‐patterned by ion bombardment and annealing (IBA). The effects of IBA on InP substrate are presented, and the structural and electrical properties of the BiSb films are measured. BiSb films on flat substrates are highly textured with (003) orientation, while films on IBA‐patterned substrates are less well ordered with mainly (012) orientation. A simple calculation method for nonplanar Hall measurements is proposed to explain the electrical conduction in rough films. The effects of BiSb film texture, roughness, and thickness on carrier density and mobility are provided. Sheet carrier density is independent of thickness at 77 K, which suggests conduction via surface states. Topological protection of such states may support nano‐patterning as a route to reducing thermal conductivity while maintaining electrical conductivity.


Introduction
Thermoelectric (TE) materials can be used to directly extract useful electrical power from temperature gradients.The TE figure of merit (ZT ) is defined as [1] ZT ¼ where σ is the electrical conductivity, k is the thermal conductivity, S is the Seebeck coefficient, and T is the absolute temperature.The higher the value of ZT, the more electrical energy can be gained from waste thermal energy.However, maximizing ZT is not a straightforward task, as values of σS 2 and k are typically well correlated [2] such that their ratio does not vary much as material parameters are changed.To increase ZT, it is necessary to increase the values of the electrical conductivity and Seebeck coefficient while decreasing or maintaining the thermal conductivity.Topological insulators (TIs, e.g., BiSb) have a bandgap, usually narrow, in their bulk.However, they may have a metalliclike surface supporting topologically protected gapless states. [3,4]Fundamental symmetries of these states mean that they are unaffected by nonmagnetic impurities. [5]This leads to suppressed backscattering and hence high surface conductance. [6]The ideal case of a TI with a completely insulating interior is hard to achieve because of the usual narrow gaps and the presence of defect-related residual bulk carriers which contribute to the conduction. [7]To minimize this contribution by means of increasing the surface-to-volume ratio, nanostructuring may be used. [7]This is important for investigating the topological conductivity of BiSb films, as the contribution of surface states to the total conductivity of a material cannot be accurately determined by using bulk crystals. [8]This could be done by preparing ultrathin membranes or nano-ribbons, but these have limited practicality for device manufacturing.Another approach is to increase the surface-to-bulk ratio by introducing many holes in bulk BiSb samples which can improve ZT because of two factors: 1) the contribution of the surface states to the conductance, and 2) the suppression of phonon thermal conductivity. [9]Whether these holes are periodic or not, the phonon contribution to thermal conductivity k will be minimized by scattering. [9]Similar nanostructuring could be done by depositing thin films of an appropriate thickness on nanostructured substrates, which is the motivation for this article.The large area and 3D nature of the substrate-TI interface causes phonon scattering, minimizing k, while the topologically protected interface and surface states maintain high σ.Many groups have investigated nanostructuring of solid surfaces by ion-beam-induced sputtering, [10][11][12][13] and we adopt the same approach here.
[16][17][18] Thin-film nanostructuring can also extend the presence of surface states when 0.22 < x < 0.35 beyond the bulk TI regions due to the quantum size effect that creates an external bandgap and clearly defines the DOI: 10.1002/pssb.202300337One route to the improvement of thermoelectric material performance is nanostructuring.Herein, BiSb thin films are grown by molecular-beam epitaxy on two types of InP(001) substrate: flat epiready wafer, and nano-patterned by ion bombardment and annealing (IBA).The effects of IBA on InP substrate are presented, and the structural and electrical properties of the BiSb films are measured.BiSb films on flat substrates are highly textured with (003) orientation, while films on IBA-patterned substrates are less well ordered with mainly (012) orientation.A simple calculation method for nonplanar Hall measurements is proposed to explain the electrical conduction in rough films.The effects of BiSb film texture, roughness, and thickness on carrier density and mobility are provided.Sheet carrier density is independent of thickness at 77 K, which suggests conduction via surface states.Topological protection of such states may support nano-patterning as a route to reducing thermal conductivity while maintaining electrical conductivity.boundary with the semimetallic region. [17]Gunes et al. [19] concluded that compared to the bulk Bi 1Àx Sb x sample, where 0 ≤ x ≤ 0.22, ball-milled nanostructured samples have broader semiconducting composition areas (0 ≤ x ≤ 0.5).
BiSb films are readily grown by molecular-beam epitaxy (MBE).Previous reports have shown that increasing the Sb ratio to 32% can degrade the quality of Bi 1Àx Sb x films grown on Si(111) that was observed by the disappearance of the reflection high-energy electron diffraction (RHEED) pattern. [8,20]This can be mitigated by increasing the substrate temperature, as Ueda et al. reported that temperatures of 200-250 °C are required for the epitaxial growth of these films. [17]n this article, we compare the properties of Bi 77 Sb 23 thin films grown on normal flat surface InP substrates (denoted by F) and those grown on Ar-sputtered nano-patterned InP substrates (denoted by P).We investigate the growth, structure, morphology, and electrical Hall measurements of the films for samples grown by MBE.The article is structured as follows.We first discuss Hall measurements on nonplanar films, since the BiSb "P" films are substantially nonplanar from the substrate nano-pattering.We then discuss experimental methodology and the crystallographic and morphological properties of the films.The P films are very different from the F films in both regards.Finally, we discuss electrical measurements on the two types of film.

Hall Measurements on Nonplanar Films
The effects of geometry on four-point contact Hall and sheet resistance measurements have been widely studied, [21] in particular the magnitude of any error in derived Hall coefficient or sheet resistivity as a function of local defect positioning.A good choice of geometry, for example, the cloverleaf pattern, minimizes the effects of local defects across as much as possible of a planar sample.However, the effects of nonplanarity across the whole sample have not been studied, to our knowledge.Here, we describe a simple modification to Hall and sheet resistance measurements for a uniformly nonplanar square sample with point contacts at the four corners.
We assume that two films are grown together in the same experiment on square substrates of side L. One is deposited on a flat substrate (denoted by subscript f ), while the other is grown on a patterned surface (denoted by p).The same volume of material V is deposited on each substrate for a given growth cycle, where and the A f,p and the d f,p are the films' surface areas and thicknesses, respectively.This is illustrated in Figure 1.The flat film's surface area is identical to the sample area, i.e., A f = L 2 neglecting the area of the corner contacts.Because the patterned surface has a 3D structure, its surface area is bigger than the equivalent flat one (A f ).We define the ratio which could readily be determined by, for example, atomic force microscopy (AFM).We assume that the geometry of the patterning does not involve much area of surface at steep angles relative to the average surface plane, and is spatially uniform, but otherwise specifies no details about the topography.If the thickness of a conformal film is comparable to the vertical scale of the patterning or surface roughness, χ of the film's surface will be very close to the substrate's χ value.As the film thickness increases, the film surface may become smoother than the underlying pattern (depending on the growth mode) and in such a case χ could be considered as the average value between the film (top, χ measured after growth) and the substrate (bottom, χ measured before growth).In all cases, the average film thickness on the patterned substrate will be smaller than that on the flat substrate since d p = d f /χ.We assume that the material resistivity ρ and bulk carrier concentration n are not affected by the nonplanar growth.With sheet resistances R f,p and sheet carrier concentrations N f,p , this means that A van der Pauw measurement of sheet resistance [22,23] involves measuring a conductor using four Ohmic contacts on its surface.For a square sample, the contacts are best placed at the corners.By measuring with all combinations of contacts and both field orientations, errors due to differing contact resistances, TE offset voltages, and shape discrepancies can be minimized.However, for a nonplanar film with morphology as described earlier, several parameters will change compared to the flat film case.According to the assumptions given earlier, we expect that R p = χR f and that R p = N f /χ.
For Hall measurement, a perpendicular magnetic field is also applied.On a nonplanar film, the carrier path may be more complicated than assumed in a standard Hall effect calculation since the field lines intersect the film at a range of non-perpendicular angles.A simplified approach is to note that the same magnetic flux ϕ = B f A f = B p A p passes through the flat and patterned samples, so that the average flux density is reduced on the nonplanar film owing to its increased effective area, and B p = B p /χ.The Hall voltages V f,p are given by where q is the electronic charge.Under the aforementioned assumptions, V f = V p = V H since the factors of χ cancel in the effective field and effective thickness for the nonplanar film.
The mobilities μ are given by In this case, 3. Methodology

InP Substrate Nano-Patterning
Semi-insulating InP(001) substrates were bombarded by Ar þ ions inside a high vacuum system.The substrates were degassed for 15 min at 200 °C, followed by Ar þ sputtering for 30 min at an Ar partial pressure of (2.5 AE 0.5) Â 10 À4 mbar and (2.6 AE 0.2) keV energy with an ion current of (6.6 AE 1.2) μA cm À2 .Note that this ion dose and energy is much higher than those used for surface cleaning of InP.
The surface resistance was measured before and after the patterning process to ensure that the samples remained insulating. [24]One patterned substrate was measured using AFM before and after annealing it for 1 h at 300 °C and then at 400 and 500 °C, to study the effect of thermal annealing on the patterned surface and to determine the highest temperature the samples could endure without destroying the pattern (see Supporting Information).

BiSb Film Deposition
The BiSb films were prepared by MBE, with chamber base pressure less than 10 À9 mbar.Prior to growth, substrates were prepared using two 20 min cycles of gentle Ar sputtering (5 Â 10 À6 mbar partial pressure and 0.5 keV energy, substrate temperature at 300 °C) to remove any oxide layers.This was followed by 30 min of annealing at 400 °C to recrystallize the substrate, which was observed to have no significant impact on patterned surface morphology.
Two separate effusion cells were loaded with 6 N grade elemental Bi and Sb.The Sb-and Bi-beam equivalent pressures were adjusted to be 3 Â 10 À8 and 1.4 Â 10 À7 mbar respectively, as measured by a shielded retractable ion gauge.The sample temperature was maintained at 320 °C to achieve high-quality films and a composition of around Bi 0.8 Sb 0.2 .

Characterization
Film structure was investigated using a Panalytical Empyrean X-ray diffractometer (XRD) at room temperature using Bragg-Brentano geometry to determine preferred orientation.Texture coefficient, TC (γ), was calculated using the relation [25,26] TCðγÞ ¼ where I hkl and I 0hkl are the intensity of the diffracted peaks in our experiment and the corresponding PDF card, respectively, and N is the number of studied peaks.Films were scanned using a Bruker Dimension Icon AFM in tapping mode.Analysis of topographic data was performed using the Gwyddion software. [27]he thickness of these films was determined by scanning over the boundary between masked and unmasked regions of a flat film (grown in the same deposition cycle as the corresponding patterned samples) and averaging the height profile from multiple images.The van der Pauw technique was employed when carrying out the electrical measurements, using the ECOPIA Hall measurement system (HMS-3000) at 300 and 77 K with a 0.55 T magnet.The sample holder uses springtensioned gold pins to connect the four corners of the square samples.The elemental composition of films was determined using a Zeiss SUPRA 55-VP field-emission-gun scanning electron microscope (FEG-SEM) with an Oxford Instruments energy-dispersive X-ray (EDX) spectrometer at an accelerating voltage of 7 kV.

XRD
Figure 2 shows the presence of distinct XRD diffraction peaks in both patterned and flat samples, suggesting that the films are crystalline.The peak positions showed good agreement with the powder diffraction card of rhombohedral structure Bi 0.8 Sb 0.2 (PDF 04-023-9518, space group: R-3m). [28]However, the difference in relative intensities between our experimental data and the PDF card reflects the presence of a strong preferred ) card [28] plotted on logarithmic intensity scale with arbitrary units.Peaks corresponding to preferred orientation are labeled in bold.
orientation toward certain planes, such as (003) and (012).To determine the ratio between both, the texture coefficient TC (γ) was calculated.The presence of other peaks in some samples shows that the films, especially those grown on patterned surfaces, contain a polycrystalline portion.This portion was calculated from the (104) peak only as it is the closest high-intensity peak to the (003) and (012) peaks.Note that the diffracted beam comes from the BiSb film and the InP substrate underneath.When 2θ increases, the film's contribution decreases compared to the substrate.
The texture analysis shows that films grown on flat InP substrates have higher textural purity than those grown on patterned surfaces.Table 1 shows the texture coefficient of each sample and relative abundance of their corresponding phases.In general, BiSb films grown on flat InP substrates have 98%-99% of the (003) texture, while those grown on nano-patterned surfaces have 88%-92% of the (012) texture with 6%-10% of (003) texture plus up to 2% polycrystalline phases.Similar textural dependency on the surface of the substrate was found by Chi et al., [29] where stoichiometric BiSb crystals tended to orient in (003) or (012) planes or stay random when deposited on various seed layers.The transition between the two textures also could occur by controlling the preparation conditions.For example, epitaxial Bi 1Àx Sb x films on Si(111) change from (003) to (012) when Sb concentrations exceed 8%-9%. [20]The Sb percentage in our samples exceeds this ratio (23%) where (012)-oriented films were expected.Siegal et al. annealed as-prepared Bi 0.8 Sb 0.2 films up to 295 °C to increase the ratio of (003) to (012) grains and improve the crystallinity of the specimen to %99%. [16]Our initial substrate temperature was slightly higher (320 °C), which likely changes the (012) texture to the (003) texture in flat samples.However, this high temperature was not enough to reorder the rough structure of the BiSb films that were grown on the patterned substrates.In other words, the substrate surface can stop a predicted texture transition from happening under certain circumstances.

RHEED
BiSb films on InP substrates were studied using the in situ RHEED technique, where Figure 3 shows RHEED patterns of one flat and one patterned InP substrates before and after the growth of the BiSb films.Panels (a,b) represent the fourfold symmetry patterns along the [001] and [011] directions of the flat InP (100) substrate.Kikuchi lines can be seen in the patterns with no intermediate surface diffraction streaks, which shows that the surface is highly ordered and has the same bulk symmetry with no surface reconstruction.Panel (c) shows patterned InP substrate RHEED, which has powder diffraction rings and some spots indicating the 3D nature of the surface.This suggests that Table 1.The texture coefficient TC (hkl) in the investigated samples.The percentage (%) of the (003), (012), and "others" phases is calculated by normalizing the texture coefficients.
Sample TC (003) TC (012) TC (others) % (003) % ( 012  the patterning process has likely amorphized the surface of the substrate, which was partially recrystallized after annealing.Panels (d,e) show the BiSb film grown on a flat InP substrate with a dominant sixfold rotational symmetry and no Kikuchi lines, which differs from the fourfold symmetry seen in the substrate and indicates epitaxial mismatch.The diffuse nature of streaks suggests the possibility of a small abundance of other textures.This has been confirmed through XRD analysis, which shows that F films are (003) oriented.This plane contains two principal nonsymmetrically equivalent directions (i.e., 100 and 110) with a 30°angle between them.Moreover, the InP substrate 011 patterns (e.g., panel (b)) coincide with either of the BiSb symmetry directions every 90°.In addition, at one angle (Panel (f )), the streaks are split (with an angle 2θ = 14°), which indicates the presence of some facets tilted 7°from the surface normal.RHEED analysis of BiSb films grown on patterned InP substrate was more complex: the film contains more than one crystallographic orientation that causes coexisting patterns to appear at various angles.In the bottom row of Figure 3, the BiSb film grown on the patterned InP substrate is presented.In general, a sixfold rotational symmetry feature (panels (g,h)) was observed overlapping with a fourfold symmetry feature every 90°.These are believed to belong to the (003) and ( 012) textures since the (003) plane has sixfold rotational symmetry and the (012) plane has fourfold-like rotational symmetry.Two of the opposite overlapped patterns showed a faceted feature (panel (i)) at 19°to the surface normal, which is larger than the tilt in the flat sample (7°).This is logical, as the films on the patterned substrate are rougher than those grown on the flat one.In panels (g,h), the pattern is formed of spots, not streaks, which is associated with transmission diffraction since the patterned samples have 3D surfaces.Panels (h,i) have clear powder diffraction rings that indicate the presence of a polycrystalline portion of the film on the surface.In summary, RHEED analysis confirms that the P films have a higher ratio of different phases and polycrystallinity, with a rougher surface, than the F films.This is consistent with the conclusions deduced from the texture analysis of the XRD patterns.

SEM and EDX
The prepared BiSb films had a shiny silver appearance, with a noticeable difference in "mirror finish" between F and P samples.SEM images of prepared films in Figure 4 show a clear difference in the topography between samples, where F films have more uniform topography with larger particle sizes than P films.Thicker films in panels (a,b) (IBS11F and IBS11P) have lower surface roughness with larger particles and more uniform topography.EDX analysis reveals the homogeneity of all the studied samples.The composition of each sample was determined at three positions across its diagonal, showing less than 1 at% variation in the composition.Table 2 shows the composition of the films determined by EDX for all four samples.Samples grown in the same growth have nearly the same composition, with less than 1 at% difference, reflecting the compositional uniformity of the grown films.Moreover, the composition in both growths was similar, with an average of 22.6 at% Sb and a standard deviation of 1.4 at%.

AFM
Patterning of the InP substrates was confirmed by the AFM topographs shown in Figure 5, and the root-mean-square (RMS) roughness of patterned substrates IBS11P and IBS12P was determined to be 36 and 30 nm, respectively.Both substrates contain a dense array of irregular nano-ripples with a depth of around 100 nm.Ar ion bombardment of InP (100) above 1 keV has been observed to produce irregular surfaces using this system, though for this experiment it is not a concern since the main purpose of the patterning is to increase the surface area, not to produce periodic nanostructures.The autocorrelation lengths may represent the in-plane length scales of the roughness.For the IBS11P substrate, these are determined to be 41 and 65 nm in the horizontal and vertical image directions, respectively, while for the IBS12P substrate, these are 33 and 85 nm.This increase in area was evaluated by dividing the substrate surface by its projection (surface ratio χ explained in Section 2) extracted from AFM topographs.The surface area was initially doubled by the patterning, with χ = 2.1 AE 0.1 for both substrates.Experiments showed that after 1 h of annealing at 500 °C, the morphology altered and χ fell from 1.7 to 1.2, as shown in Figure S1, Supporting Information.To preserve the patterning, sample temperature was maintained at ≤400 °C before and during film deposition.The thickness of the films was determined and reported alongside roughness and χ values in Table 2. Figure 6 shows 3D representations of BiSb film surfaces scanned by the AFM.Films that were grown on flat substrates (at the left) have threefold symmetry features (i.e., pyramids and 120°angles).In contrast, patterned films have rougher surfaces, especially the thinner film (IBS12P) which has a surface morphology that closely matches the pre-growth substrate.As the film thickness increases, the surface becomes smoother and has a smaller χ value (1.28 for IBS11P compared to 1.59 for IBS12P), and the rotational symmetry features of the BiSb film become more dominant instead.

Hall Measurements
The flat InP substrate is electrically insulating, with a resistance greater than the operable range of the HMS-3000 of >10 10 Ω.
After the patterning process, the spreading resistance was very high but within the measurement range (%3 Â 10 7 Ω), possibly due to the formation of an ultrathin conductive In-rich layer on the surface.This does not affect the electrical measurements of the film layer because the resistance of the BiSb film is four orders of magnitude lower.
To explain the electrical parameters, the BiSb film can be treated as bulk semiconductor with a metallic surface.In this case, the total conductivity of BiSb can be expressed as [30,31] where σ s is the surface conductivity, d s is the effective surface thickness, E g is the electrical bandgap, and k B is the Boltzmann constant.The first term in the equation represents the surface contribution to the conductivity, while the second term represents the bulk conductivity.Accordingly, surface conduction becomes dominant in two cases: 1) a decrease in film thickness increases the surface contribution, and at very low d, the effective E g may increase due to quantum confinement and reduce the bulk contribution; 2) at low T, the bulk contribution is reduced and surface conduction dominates.To separate both conductivity components and determine the equation parameters, the electrical conductivity of films of different thicknesses over a wide range of temperatures could be used. [17,20,31]Electrical parameters were derived from Hall measurements at 300 and 77 K, where values for σ, μ, N s , and N B are calculated from the total conduction of both surface and bulk BiSb.This means that N s and N B , here, are the averaged sheet and bulk carrier concentrations, respectively, which relate to the total conductivity σ.
In general, the type of InP substrate (flat or patterned) and the thickness of the film significantly impact the electrical parameters.Table 3 shows the electrical parameters of the prepared films, corrected for nonplanarity according to Section 2. The negative sign of the bulk and surface carrier concentration values (N B and N S ) denotes majority n-type conduction.
At 300 K, N B is not significantly influenced by thickness, with values of around 7.8 Â 10 19 and 2.8 Â 10 19 cm À3 for flat and patterned films, respectively.This indicates two points.First, both films behave as semiconductors, where the bulk conduction is dominant over the surface conduction because at high temperatures the bulk semiconducting interior gets enough carriers activated to dominate.Second, F-films, i.e., (003)-oriented, have more carriers than P-films, mostly (012) oriented.BiSb films  with less Sb percentage (0.12-0.16) have been reported to have smaller carrier concentrations but higher mobility than results from this study, where the carrier concentration of (003)-oriented Bi 84 Sb 16 was 2 Â 10 19 cm À3 with a mobility of 500 cm 2 V À1 s À1 at 300 K. [32] Other research shows that the carrier concentration of ball-milled Bi 88 Sb 12 increases from 1 Â 10 18 cm À3 at 50 K to 1 Â 10 19 cm À3 at 300 K, while mobility decreases from 1.5 Â 10 4 to 2.3 Â 10 3 cm 2 V À1 s À1 . [33]ilm thickness was observed to significantly affect bulk carrier concentration at 77 K.In this temperature range, the N B values are higher for thinner films, however, N S values are close for different thicknesses at around 2 Â 10 14 cm À2 for flat films and 4 Â 10 13 cm À2 for patterned films.This suggests that the metallic surface conduction is dominant and the bulk contribution is minimal at this low temperature.
The conductivity for the F samples and the thickest P sample (≥70 nm, at 300 K) exceeds 10 3 Ω À1 cm À1 near the values of MBE grown (003)-oriented Bi 1Àx Sb x films on GaAs(111). [17]or (012)-oriented ball-milled Bi 77 Sb 23 samples, the conductivity was measured as 2.65 Â 10 3 Ω À1 cm À1 . [19]The conductivity values increase with the corrected thickness for all samples because thinner films have larger confinement (high E g ) for carrier movement, which reduces mobility.As with low dimensionality, spatial confinement in the thinner P films displaces the bandgap edges and converts semimetals like Bi to semiconductors. [34,35]amples IBS11P and IBS12F are good examples to understand the effect of the patterning only, as both have an equivalent (corrected) thickness of 70 nm, but the former has 1.7 times more surface area.On one hand, both samples have similar conductivity values (around 1525 Ω À1 cm À1 at 300 K and 550 Ω À1 cm À1 at 77 K).Therefore, the lower conductivity observed in the P films might be compensated by depositing a larger corrected thickness d p , equivalent to that of the F films d f .But on the other hand, the patterned sample (IBS11P, (012) BiSb) has a smaller number of carriers with higher mobility.This could have resulted from two opposite conduction mechanisms.The first one is that the patterned samples have higher density of protected surface states, whose mobility is higher than that of bulk states.Consequently, more conduction occurs from the surface states, with fewer carriers of high mobility.In contrast, the F film has more ordered (003) semiconducting bulk, which has higher conductivity than its surface states, so a high number of bulk carriers contributes to the conduction, but with lower mobility than those of the surface states.An important difference between BiSb (012)-and (003)-oriented surfaces is the number of Dirac cones present, associated with electron pockets.It was predicted that three Dirac cones exist in the surface Brillouin zone of the (012) orientation, compared to just one in the (003) orientation. [36]This was confirmed experimentally using epitaxial growth and angle-resolved photoemission spectroscopy. [37]This will affect the surface conductivity of otherwise identical BiSb films, and is consistent with a higher surface conductivity in the IBS11P film compared to IBS12F, as discussed.Walker et al. found that when BiSb thin films are orientated  in the (012) direction instead of the (003) direction, the temperature-dependent conductivity of the films decreases, most likely, as a result of changes in the surface states and the quantum confinement. [20]Finally, MBE-grown BiSb in (012) orientation has been used in room-temperature spin-orbit torque devices showing enormous spin Hall effect (SHE). [38]This crystallographic orientation is essential for achieving the very large SHE, probably due to the multiple Dirac cones at this surface.
Our observation that ion-beam patterning changes the dominant orientation of an MBE-grown BiSb film from (003) to (012) may be helpful in optimizing thin-film preparation for applications in magnetic random access memory devices [38][39][40] and skyrmionbased devices. [41]

Conclusion
In this study, Bi 77 Sb 23 thin films were deposited by MBE on both flat and nano-patterned InP substrates.Nano-patterns were produced by argon ion sputtering, and increased the surface area of the substrates by up to a factor 2. BiSb films grown on flat InP substrates showed a highly textured (003)-oriented structure, while those grown on patterned films were mostly (012) oriented with a minority of (003) and polycrystalline phases.All flat and patterned films exhibited behavior characteristics of n-type semiconductors.Patterned samples had a smaller nonuniform thickness, which lowered their conductance and average conductivity compared to the corresponding flat samples.However, the conductivity of the two types of samples could be considered equal if they had the same corrected (average) thickness.Sheet carrier density is independent of thickness at 77 K, suggesting conduction via surface states.It may be possible to exploit these states to retain high electrical conductivity while lowering thermal conductivity via nano-patterning, hence increasing TE performance.The unique properties of BiSb(012) surfaces may also be exploitable in spin torque and skyrmion devices.

Figure 1 .
Figure 1.Illustration of films (green) grown simultaneously on flat and patterned substrates (blue).Vertical arrows represent incoming growth flux and, in Hall measurement, magnetic field B; these are identical for both samples.

Figure 2 .
Figure2.X-ray diffractometer diffraction pattern of the four studied BiSb samples the with Bi 0.8 Sb 0.2 (PDF 04-023-9518) card[28] plotted on logarithmic intensity scale with arbitrary units.Peaks corresponding to preferred orientation are labeled in bold.

Figure 3 .
Figure 3. Reflection high-energy electron diffraction patterns of the InP substrates and the BiSb films grown on them.a,b) The flat InP (100) substrate.c) The patterned InP substrate.d-f ) The BiSb film grown on a flat InP substrate.g-i) The BiSb film grown on a patterned InP substrate.

Figure 4 .
Figure 4. Scanning electron microscope images of BiSb samples grown on a,c) flat and b,d) nano-patterned InP substrates at 7 kV accelerating voltage and 10 4 magnification.

Table 2 .
The composition of the BiSb films as determined by EDX, with film morphology parameters calculated via AFM analysis.Thicknesses were measured for each sample set using an additional flat sample grown in the same deposition cycle.