ZnSb Films on Flexible Substrates: Stability, Optical Bandgap, Electrical Properties, and Indium Doping

Undoped and In‐doped ZnSb thin films are deposited on rigid glass and flexible polyimide (Kapton) by physical vapor deposition. Detailed structural and chemical characterization is performed along with measurement of electrical and optical properties. These properties are very similar for films on glass and Kapton. Flexible ZnSb films show remarkable stability of electrical and optical properties, which are unchanged after 104 cycles of linear bending with surface strain 0.18%. Only severe flexing after this treatment (torsional bending with surface strain 1.7%) causes progressive degradation of conductivity over tens to hundreds of cycles. The ZnSb optical direct gap is determined to be 0.89 ± 0.05 eV. The optical direct gap of β Zn4Sb3 was 1.07 ± 0.05 eV. Other absorption features in the films, including smaller indirect gaps, are discussed.


Introduction
Originating with efforts to increase the power-to-weight ratio of silicon solar cells for satellites, [1] the field of flexible electronics has been vastly extended to cover applications such as lighting, displays, medical sensors, and energy harvesting. [2][4][5] Inorganic materials, such as indium oxide for transparent electrodes, may suffer from cracking in flexible applications [4] and so it is essential to study behavior of such films under repetitive bending.Even if flexibility in use is not required, flexible substrates can be easily shaped and processed, minimizing waste in the production process. [6]The most common and simplest type of these tests DOI: 10.1002/aelm.202300403 is the static one, where the flexible substrate is bent over a rod of defined diameter. [7,8]But this method is not adequate for testing under real working conditions, for which repeated bending is essential. [9]Push-to-flex bending is one of the most frequent dynamic bending tests, regardless of not offering a constant radius of curvature. [7,10,11]inc antimonides are among the most promising thermoelectric (TE) material classes, because of excellent material properties, [12,13] the abundance and low price of Zn and Sb, and non-toxicity.Different groups studied the deposition of the ZnSb films on flexible substrates using sputtering [14][15][16] and screen-printing techniques. [17]But there is no available data showing its resilience under dynamic flexing conditions.On the other hand, physical vapor deposition (PVD) technique is becoming more favorable for flexible thermoelectric applications due to its high-throughput capabilities. [18]he optical bandgap of ZnSb and related compounds has been discussed in the literature since the 1960s. [13]A commonly quoted value for the gap of ZnSb is 0.5 eV.However, it is not clear from the earliest works whether this refers to a direct or indirect transition, while the spectral range and total transmission may also be limited. [19,20]Modern band structure calculations of ZnSb showed direct (indirect) gaps of 0.74 (0.37) eV [21] and 0.88 (0.60) eV. [22]Some measurements of bandgap have been inferred from the temperature dependence of electrical conductivity and others by optical absorption.Furthermore, bandgap determinations are affected by the presence of other Zn-Sb phases and the degree of crystallinity of the samples. [13][25] Additionally, a lower value of the bandgap, typically 0.3-0.35eV, has often been reported for polycrystalline ZnSb, but the character of this gap (direct, indirect or semi-localised states) is not clear. [13]ush-to-flex cyclic testing has not been applied to ZnSb flexible films, and there is a need to clarify the optical transitions in defective and multi-phase ZnSb material.In this paper, both questions are addressed.In this paper, ZnSb films have been grown on glass and polyimide (Kapton) flexible substrates by low-pressure physical vapor deposition (LP-PVD) with varying Zn/Sb flux ratio and with or without In doping.Fourier transform infra-red (FTIR) absorption measurements exploited the transparency of the substrates and thin film structure, and elucidated features of the optical absorption spectra on multi-phase ZnSb films.Furthermore, electrical and optical properties were measured before and after 10 4 bending cycles of push-to-flex treatment for ZnSb films on Kapton.

Grazing Incidence X-Ray Diffraction (GIXRD)
After annealing, the films were scanned by the GIXRD technique to identify their phases.Figure 1 shows the diffraction pattern of the pure ZnSb films ordered from low to high Zn percentage.The compositions shown were obtained from energy dispersive X-ray analysis (EDX).The sample with overall 23 % Zn (lowest panel, S1) contains about 46 % of Sb and 54% of  −Zn 8 Sb 7 .The three middle patterns are from nearly stoichiometric ZnSb films.It is clear that changes of overall Zn:Sb ratio of just a few % have a strong influence of the phase composition near 1:1.Pure ZnSb phase occurs in the film with 48% Zn (S3), with excess Sb crystallites separately.With a small excess of Zn (52.5 %, S10), ZnSb is still dominant, with Zn 4 Sb 3 appearing.Phase-pure Zn 4 Sb 3 appears when the overall stoichiometry is close to 4:3 (56.4 % Zn, S4).At still higher Zn-to-Sb flux ratio, more Zn then precipitates on the surface to form solid bulbs and increases the roughness (see atomic force microscope (AFM) of S2, Figure 2), which likely causes the low-angle hump observed in the S2 diffractogram.Table 1 gathers the phase quantification data.The table includes the overall composition derived from GIXRD phase quantification, which is in generally good agreement with the EDX composition.

Surface Morphology and Thickness (AFM)
Figure 2 shows the topography of the ZnSb on glass samples.In the middle row, ZnSb samples near stoichiometry (S3 and S10) show well-defined particles, though it is not clear why S4 is dissimilar.The top row shows the most off-stoichiometric samples.Very Sb-rich S1 shows a mottled surface topography, smoother than the Zn-rich S2, which has large Zn agglomerates.In the latter case, the underlying ZnSb granular film features can be seen.This structure is most obvious for the near-stoichiometric samples S3 and S10 (middle row).The moderately Zn-rich S4 appears smoother, as do the In-doped samples (bottom row), though in the latter case, there appear to be In clusters scattered across the surface.The overall appearance of In-doped films was less homogeneous than the undoped films.Indium surface segregation is a well known phenomenon in epitaxial growth. [26]Agglomeration of surface indium has also been observed in granular oxide thin films. [27]In the present case, indium surface segregation may have occurred via grain boundaries during growth and/or during post-growth annealing.As the melting point of indium is only 156.6 °C, it can be more mobile in its liquid phase, thereby forming surface agglomerations.
The size distribution of the surface particles from AFM images and the grain size from Williamson-Hall (WH) plots were studied for selected samples (See Supporting Information).The WH grain size is observed to be smaller than both the AFM mean particle size and the film thickness.So the overall picture is of granular polycrystalline films, where individual grains are not themselves single crystals.

Electrical Measurements
All In-doped and undoped films have a positive sign for the carrier concentration, reflecting their p-type nature.For the undoped ZnSb thin films, Hall measurements show great dependence on the ZnSb composition.Figure 3 shows the carrier concentration and mobility dependence on the Zn percentage in the studied films.In general, the very Sb-rich sample (Zn:23%) has the highest bulk carrier concentration and mobility values (2.5-1.9 × 10 20 cm −3 and 127-209 cm 2 Vs −1 at 300 and 77 K).By increasing the Zn content in the rest of the samples, the bulk concentration increases from 6.2 × 10 18 cm −3 at (Zn:48%) to 4.9 × 10 19 cm −3 at (Zn:56 and 64%) and the mobility falls from 126 cm 2 Vs −1 at (Zn:48%) to 23 cm 2 Vs −1 at (Zn: 64%) at 300 K.This could be dependent on the ZnSb/Zn 4 Sb 3 phases' ratio.Zhang et al. studied ZnSb samples with Zn varying from 50 to 70% [28] and Jang et al. with Zn vary from 53 to 57%. [29]Their samples contain a mixture of ZnSb and Zn 4 Sb 3 phases and have very close values for carrier concentrations and mobility at 300 K as shown in Figure 4a,b.The mobility of the charge carriers is obviously falling as Zn% increases (more Zn 4 Sb 3 replacing ZnSb).In contrast, the carrier concentration increases sharply above a Zn fraction of 54%, with a possible peak at 57% corresponding to exact Zn 4 Sb 3 stoichiometry.It is noteworthy that the samples in Figure 4a,b from several studies were grown and annealed under varied conditions that might have affected their electrical proper-ties.For instance, the carrier concentration and mobility values were reported to be affected by the annealing temperature. [14]he second observation, from Figure 3, is that the electrical parameters of the two Zn-rich samples (Zn ≥ 55%) are more temperature-dependent than the stoichiometric ones.Bulk  concentration rises by a factor of 2.5-5 from 77 to 300 K for the Zn-rich samples, but hardly changes for nearly stoichiometric samples.Mobility decreases at 300 K for off-stoichiometric samples but changes little for nearly stoichiometric ones.In semiconductors, the carrier concentration (n) depends on the activation energy ΔE according to the Arrhenius equation n = n o e −△E∕K B T , where n o is constant and K B is Boltzmann constant.If ΔE is small, more carriers will be excited when the temperature increases, and n will be more dependent on the temperature.This is likely the case in our off-stoichiometric samples, they have smaller ΔE in the temperature range from 77 to 300 K.This might be attributed to the presence of more gap states in the off-stoichiometric samples, in which the carrier transition dominates.This dominance may have arisen because thermal energy in that temperature range (order of 10 meV) is much less than the optical bandgap that represents band-to-band transition (0.3-1.1 eV, Section 2.4).

Hall Measurements of Flexible Films
Hall parameters of ZnSb films on Kapton were measured before and after 10 4 bending cycles.S3 was tested after 10 5 cycles additionally.The values of carrier concentration and mobility of the flexible films lie in the same range as those grown on glass (e.g., μ ≈ 130 and N B ≈ 5.6 × 10 −18 for S3 sample).Figure 4c shows the values of the bulk carrier concentration of the undoped (left) and In-doped (right) ZnSb flexible films labeled by their EDX compositions.One can notice also that, after applying 10 4 bending cycles, these values stay very close to their original values.For the most highly doped sample (6.8% In) it was possible to obtain Hall mobility and conductance before bending but not after.This indicates that repeated flexing has altered the behavior of the In within the polycrystalline structure.We speculate that this is caused by the mobility of excess In at grain boundaries, but further study is needed to explore this effect and its origin.
Figure 4d shows the carrier concentration and mobility dependence on the In percentage.As the In concentration increases, the mobility falls monotonically to reach about one-tenth of its value by 4% In.In general, the carrier concentration increases by increasing the indium percentage (but the Zn-to-Sb ratio can affect this value as in sample S8, which has a higher value than S6).
To highlight the effect of bending alone, the electrical parameters of the bent samples were normalized to their original values before bending.Figure 5d shows that, after bending, the electrical parameters stayed very stable, within 7 % of the original values (the change could be from the repeated measurements, not from the bending).This reflects the suitability of the undoped ZnSb and up to 5 % In-doped films for the flexible device application.
In addition, the stability of the electrical parameters after flexing these films (radius of curvature, r is down to 1.36 cm) advantages the LP-PVD in fabricating flexible ZnSb films with 0.1-0.3μm thickness over the screen-printing technique.In the latter, the resistance of 26 μm thick module rose sharply when bent below a 7-cm radius of curvature. [17]t is important to highlight that the above-mentioned stability is observed by regular controlled bending only; twisting these films manually has a detrimental impact on their conductivity.In a separate experiment, the undoped ZnSb sample (S3) and the In-doped (S5) were tested under torsional stress, which is much harsher.Figure 5e-f shows the progressive degradation in samples' conductivity when being twisted manually.The conductivity in both films decreased by twisting, quicker in the In-doped sample than in the undoped one.This is likely from the disorder accompanied by In-doping.In conclusion, ZnSb films can endure gentle bending, even with 10 4 cycles, but degrade very quickly with tens of manual twisting cycles.

Optical Bandgap
For ZnSb samples grown on glass, absorption from the FTIR spectrum was acquired, and Tauc plots are presented in Figure 6.The optical transitions were investigated using both direct gap (n = 2) and indirect gap (n = 0.5) plots.In most cases, the transitions could be seen in both plots at close values of energy.The nearstoichiometric samples (middle row, S3, S4, and S10) all show a broad absorption peak in the 0.4 -0.8 eV range-and a sharper absorption edge around 1.0 eV.The edge is readily fitted as either an indirect or direct gap.However, on the basis of the shape of the absorption edge (longer linear region for the exponent two plots) and calculated band structures for ZnSb (see below), we assign this feature to a direct gap.The broad lower energy features can be fitted to direct gaps in the range 0.26 to 0.36 eV.However, indirect gap fitting does not lead to a clear range of positive onset energies.Combined with the weaker absorption and broad energy spread of the features, we suggest that this optical absorption is not associated with a well defined gap energy.The bottom row shows In-doped samples, and a change of the shape  [28] and Jang et al. [29] a,b) Green dashed lines are guides to the eye.Panels (c and d) are electrical parameters for ZnSb films on Kapton.c) Scatter of carrier concentration against composition and bending history.d) Carrier concentration and mobility as a function of In doping and bending history.
of absorption spectra, compared to undoped material, is immediately apparent.The absorption edge at around 1.0 eV is still present, but the lower energy absorption now comprises at least two broad features.These are discussed below.The top row represents the most Sb-rich (S1) and Zn-rich (S2) samples, and the optical absorption features are much less well defined.Nonetheless, energy gaps can be extracted from the Tauc plots.For comparison, in panel (X), we have replotted data from Ref. [20] based on thinned bulk single crystal ZnSb.A very clear indirect gap appears at 0.52 eV, with an equally clear direct gap 0.89 eV.
Figure 7 gathers direct transition energies and plots them as a function of Zn concentration and In doping.Symbols and dashed lines highlight the main groupings.The largest direct gap is shown by the green dashed line.As Zn concentration increases, this gap increases but then saturates at 1.07 ± 0.05 eV with no clear dependence on In doping.[25] The very Sb-rich sample (S1) has a direct gap around 0.69 eV, which could be attributed to the  −Zn 8 Sb 7 phase dominant in this sample.This value is close to the theoretically calculated direct bandgap value 0.61 eV, found using density functional theory with generalized gradient approximation (for structural optimization) and the modified Becke Johnson (MBJ) potential (for electronic structure calculation). [22]he gap observed at 0.85, 0.87, and 0.90 eV (circles) is attributed to ZnSb.It is in good agreement with the calculated MBJ direct gap of 0.88 eV [22] and the early measurement [20] replotted in Figure 6 panel X.Including our fitting error, we can quote 0.89 ± 0.05 eV as the direct optical bandgap of ZnSb.The value is somewhat larger than the calculation of Bjerg et al. [21] which found 0.74 eV using the Engel-Vosko (EV) functional.For undoped samples, a low energy direct gap is observed (purple dashed line) at 0.31 ± 0.05 eV.However, this is associated with a broad low-energy absorption feature (Figure 6, middle row) or a weak change of curvature in absorption (top row).We therefore do not assign this to a well-defined energy gap.The value is smaller than the indirect gaps calculated with EV and MBJ functionals, 0.37 and 0.60 eV, respectively. [21,22]A similar transition at around 0.35 eV has been noted in polycrystalline ZnSb compounds [13] and in both ZnSb and mixed Zn 4 Sb 3 / ZnSb samples, [28] inferred from temperature-dependent electrical data.In our thin films the transition is present independently of the phase composition, consistent with these previous observations.The transition might be associated with band tails or other defect states, giving rise to the broad absorption features and range of onset energies reported.
The low energy transitions at 0.31 ± 0.05 eV are not seen in the In-doped films, but new transitions (0.49 to 0.66 eV) appear instead.These gaps are shown as triangles in Figure 7.The most Zn-rich undoped sample (S2) may show both transitions, but the absorption curve is rather smooth (Figure 6b).

Bandgap of Flexible Films
Direct and indirect Tauc plots were compared for all the films before and after bending.The Kapton sheets caused strong interference fringes that overlap with the absorption spectrum in the NIR range.The data were smoothed using a fast Fourier trans-form filter that easily removes the interference fringes without producing spectral artefacts.Figure 8 shows the smoothed direct and indirect Tauc plots of the flexible films before and after bending.The plots show clearly the matching between the data obtained before and after flexing the films.The determined gap values of the bent films fall within 0.02 eV their values before bending.Moreover, these values are very close to the gap values of the same films grown on the glass as illustrated in Figure 6.That suggests that using different substrates (glass and Kapton) does not have much impact on the optical properties of the grown films.Overall, growing ZnSb films on a flexible substrate such as Kapton gives the same optical properties as it does when grown on glass, even after flexing them for 10 4 cycles.

Conclusion
ZnSb films were grown on glass and flexible Kapton substrates using the LP-PVD technique.Structural, electrical, and optical data were discussed for undoped and In-doped films.Experimental optical bandgaps were determined for ZnSb and In-doped ZnSb thin films.The ZnSb direct gap of 0.89 ± 0.05 eV agrees with the bulk single crystal [20] and with MBJ calculations.For more Zn-rich samples with and without In doping, a gap of 1.07 ± 0.05 eV could be identified as that of  Zn 4 Sb 3 .Lower Figure 6.FTIR spectra of ZnSb films on glass, panels a-h).Tauc plots for direct (in black) and indirect (in red) optical transitions were obtained using the absorption data.The top row are Sb-rich (S1) and Zn-rich (S2), the middle row are close to stoichiometry, and the bottom row are In-doped.Panel (X) in the top right corner is a Tauc plot calculated from reference. [20]ergy transitions depend on the presence of In doping and stoichiometry.
ZnSb films grown on Kapton have similar electrical and optical parameters as those deposited on glass with the same incident beam fluxes in the same experiment.The stability of the electrical and optical properties was excellent under dynamic pushto-flex bending.After 10 4 bending cycles down to 1.36 cm radius of curvature (0.18 % surface strain), bulk carrier concentration, mobility, and conductivity were all unchanged.Furthermore, optical absorbance spectra were also identical.This compares well against the screen-printed films of Lee et al. [17] which showed degradation of electrical resistance for static bending of around 0.1% surface strain.Only after severe distortion (manually applied torsion cycles with surface strain an order of magnitude higher) was the electrical conductivity degraded.LP-PVD is a suitable method for the production of ZnSb films for flexible electronics.

Experimental Section
Substrate Preparation: ZnSb thin films were grown on different insulating substrates, namely glass, and flexible Kapton, a commercial polyimide film.Standard glass slides were cleaned in an ultrasonic bath and then washed with de-ionized water.After that, they were treated chemically with a diluted solution (0.03%) of SnCl 2 for 15 min, followed by 20 min-200°C-annealing in the air to form a layer of tin oxide as described by Pejova et al. [30,31] .The purpose of the SnO layer was to enhance the adhesion of the grown films with the substrate and improve their crystallinity. [30,31]he prepared slides were then kept under dry nitrogen until used in the deposition.The conductivity of the treated substrate was tested to ensure that it remains insulating and that the SnO layer was not electrically conductive.Kapton film substrates of 0.05 mm thickness were cleaned with isopropyl alcohol and dried with nitrogen.ZnSb films were grown on glass (square ≈ 9 mm × 9 mm) and/or Kapton substrates (strips ≈ 27 mm × 8 mm).
ZnSb Film Growth: ZnSb thin films were prepared inside a dedicated high vacuum system comprising two chambers.A PVD technique employing separate effusion sources for Zn, Sb, and In was used.The substrates were degassed in the first chamber using an infra-red lamp to 200 °C for a minimum of 20 min.In the second chamber, Zn, Sb, and In vapors were obtained by simultaneously heating the pure elements inside alumina crucibles using spiral Ta filaments.Different compositions were prepared by controlling the filament current of each element's cell individually, calibrating by measuring film compositions using EDX analysis.For In-doped samples, the Zn and Sb cell conditions were kept fixed at conditions used for the stoichiometric S3 sample, while changing the In cell's temperature.This method provided reliable composition control over a growth campaign of several weeks.During PVD, substrates were not heated and so were slightly above room temperature, owing to the radiant heat from the effusion cells.All as-grown films appeared uniform and mirror-like, but some changes in color and reflectivity were observed between different compositions.For example, Zn-rich samples look duller and less reflective than Sb-rich samples, which could be related to the rougher surface morphology observed by AFM.After deposition, all samples were annealed on a hot plate under dry nitrogen at 150 to 200 °C to improve crystallinity, conductivity and stability as explained below.Five ZnSb and five ZnSb(In) compositions were prepared.Sample thicknesses and compositions were tabulated.
Structural and Compositional Analysis: The films' elemental composition was determined by Zeiss SUPRA 55-VP field emission gun scanning electron microscope (FEG-SEM) with Oxford Instruments EDX spectrometer, using an accelerating voltage of 7 kV.For each sample, the composition was calculated from an average of three regions.The structures of the films were analyzed using a Panalytical Empyrean GIXRD diffractometer operating with 1°incident angle.Phase identification and composition analysis were performed using the full profile fitting feature of the HighScore Plus software with the PDF4 database. [32]The chemical compositions derived from EDX were used to constrain the phase com-positions obtained from GIXRD in the following way.ZnSb and Zn 4 Sb 3 were always included in the list of possible phases.Plausible additional phases (Sb, Zn, other Zn x Sb y , In2O3, etc.) were added and the fits refined to give overall compositions close to EDX.While the GIXRD-derived compositions should not be considered fully quantitative, they agree within a few % with the EDX compositions, giving confidence that the phase compositions were reliable.Williamson-Hall (WH) analysis was performed for near-stoichiometric films using selected GIXRD peaks (high intensity and low overlap).The sample surfaces were imaged using Bruker Dimension Icon AFM for better depth resolution in roughness and thickness measurements.The AFM was equipped with a SCANASYST-AIR silicon tip and was run in Scanasyst mode, while acquired datasets were then processed using the Gwyddion 2.59 software package. [33]The thickness of the films was measured from the films grown on glass (see Supporting Information).
Bandgap Measurements: The optical analysis was done by acquiring the absorbance spectrum in the infrared region.The FTIR absorption spectrum was obtained using a Bruker Vertex 70V spectrometer at room temperature.For an ideal single-phase polycrystalline semiconductor, to obtain the optical energy bandgap a Tauc plot could be used.By plotting (h) n against h and extrapolating the linear segment to the energy axis, one can determine the value of the bandgap, according to the following equation: [34,35] (h here, k is a constant,  is the absorption coefficient, h is the photon energy, and E g is the band gap.The exponent n is 2 or 0.5 for direct allowed or indirect allowed optical transitions, respectively.For multiple phases, this plot becomes more complex as absorptions of different phases overlap and it becomes more challenging to identify the value of the optical bandgaps and whether they were direct or indirect.[38] Here, it used a manual graphical approach to identify the gaps from the intersection of the linear fit of the optical transition segment with the baseline (either =0 or tail of low energy absorption as defined by a linear extrapolation) as explained by Makula et al. [36] Electrical Measurements: Electrical measurements of the samples were carried out using ECOPIA Hall measurement system HMS-3000 with a 0.55 T magnet, which used the Van der Pauw technique to extract the conductivity, mobility and carrier concentration.Square samples were connected using four gold spring-tensioned pins placed directly on the corners of the sample surface and Ohmic contact was confirmed for all samples.Measurements were taken at room temperature and while submerged in liquid nitrogen (300 K and 77 K) for glass samples and room temperature only for Kapton films, as it was intended for confirming their stability over flexing and when operating at room temperature.The repeatability of the measurements was confirmed (see Figure S2 (Supporting Information): five separate measurements were taken when the sample was annealed at temperatures ⩽ 200 °C and six-month gap).
Flexibility Testing: The durability of ZnSb films was examined by testing their electrical and optical properties before and after bending them for 10 4 cycles.A linear actuator was set up to repeatedly bend batches of films on Kapton, as shown in Figure 5a-c.As it was free bending, and no constant radius of curvature was maintained, the minimum radius of curvature r was estimated graphically to be 13.6 mm [Figure 5c].Using this value with 0.05 mm as the Kapton substrate's thickness h, the maximum bending-induced surface strain  b (r) could be calculated: [39]  b (r) = h 2r + h ≈ 0.18% Additionally, 8 mm square sections of the same samples (films deposited in parallel) were attached to 27 mm Kapton strips using Kapton tape and subjected to the same 10 4 cycles of bending.They were then sub-Figure 8. Tauc plots of ZnSb films on Kapton before and after bending.The data was obtained by smoothing the original absorbance data with an FFT filter.The y-axes are scaled differently.The solid lines represent direct and indirect Tauc plots and are colored differently as explained in the legends, while dashed lines represent the linear fittings of the data segments of the same color.The small peak at 0.45 eV comes from the instrumental setup, not from the films.jected to cycles of torsional deformation by hand, up to several hundred cycles.This results in cyclical buckling of the sheets in orthogonal directions with a much smaller minimum radius of curvature, as low as 3 mm which corresponds to a surface strain of 1.7%.Conductivity was measured as a function of the number of torsional cycles.

Figure 1 .
Figure1.GIXRD diffraction patterns of undoped ZnSb films after annealing.The patterns are ordered from low Zn content at the bottom to higher Zn percentage at the top.The sample code of each sample is written on the top left of each pattern, while some distinguished peaks are marked with their phase.Zn percentage at the right is determined from EDX.

Figure 2 .
Figure 2. AFM topographs surface scaled to the same vertical and horizontal dimensions.The images are 2μm × 2μm, and the brightness level was adjusted to be from 0 to 50 nm.The top row contains the Zn and Sb rich samples, the middle row shows the three films near stoichiometry, and the bottom row contains In-doped samples.

Figure 3 .
Figure3.Electrical parameters (carrier concentration, mobility, and conductivity) of ZnSb films on the glass as a function of their EDX composition.Measurements were taken at 300 K (red circles) and 77 K (blue squares).Scale breaks are applied to show the Sb-rich sample's data in the conductivity and concentration panels.

Figure 4 .
Figure 4. Electric parameters of ZnSb films grown on glass compared to Zhang et al.[28] and Jang et al.[29] a,b) Green dashed lines are guides to the eye.Panels (c and d) are electrical parameters for ZnSb films on Kapton.c) Scatter of carrier concentration against composition and bending history.d) Carrier concentration and mobility as a function of In doping and bending history.

Figure 5 .
Figure 5. Kapton samples during the flexibility test, moving from a) relaxed position to b) bent position, and c) illustrates the maximum compression and resultant radius of curvature.Panel d) shows the electrical parameters of the flexed samples, normalized to their values before bending.The horizontal lines are ±7% from unity.The conductivity of two ZnSb samples grown on Kapton under multiple manual twisting cycles, e-g).Panel e) shows the undoped S3 sample in red triangles with the right scale and the In-doped S5 sample in blue circles with the left scale.Panels f-g) show one sample before and during the twisting.

Figure 7 .
Figure 7. Map of optical transitions from FTIR spectra of ZnSb films on glass, extracted from Figure 6.Undoped ZnSb films are ordered according to Zn % content, while In-doped films are ordered with their In content.The green dashed line is a guide to the highest energy transitions observed in each sample.The lowest energy transitions are highlighted by the purple dashed line.Circles represent the ZnSb direct gap while triangles represent a transition observed in In-doped samples.

Table 1 .
Phase quantification from GIXRD patterns of annealed ZnSb films on glass and overall compositions from XRD phase compositions, together with overall composition obtained from EDX.