Ellipsometry, transmission, and photoluminescence characterization of Mn-doped ITO thin films deposited by DC magnetron sputtering

Manganese-doped indium tin oxide (ITO) thin films (0–12.8 at% Mn) were deposited by DC magnetron sputtering. The structural, electrical, and optical properties of the films were studied. Optical characterization was emphasized and included ellipsometry, transmission, and photoluminescence (PL) measurements. Features of the energy band structure of ITO and Mn-ITO were extracted from PL spectra and Tauc analyses of absorption data. We concluded that the fundamental bandgap of ITO is ∼2.8 eV. A separate deep valence band, ∼0.8 eV below the valence band maximum, was confirmed to be involved in higher energy optical absorption and emission transitions. Both of these observations were consistent with recent published theoretical studies and spectroscopic measurements (XPS, XES, etc). Mn addition was found to result in a decrease of the transition energies. Additionally, Burstein–Moss shifts were observed.


Introduction
Decades of intensive research on transparent conducting oxides (TCOs) have benefited wide ranging technologies, including flat panel displays, photovoltaic cells, LEDs, transparent electronics, smart windows, and wearable electronics [1][2][3][4].TCOs are electrically conductive, yet optically transparent due to their large bandgaps.Indium oxide (In 2 O 3 ) and tin doped In 2 O 3 (Indium Tin Oxide or ITO) are commonly used TCOs with high optical transmission and electrical conductivity that can be tuned by adjusting the tin content and the deposition conditions.Additional dopants have been added to In 2 O 3 and ITO in efforts to modify the material properties.In particular, manganese doped In 2 O 3 (IMO) and ITO (Mn-ITO) have found applications in organic light emitting diodes [5] and spintronic devices [6].A typical Mn doping level for these applications is 3%-15%, and the preferred growth methods are e-beam evaporation [5], sputtering [6][7][8][9][10], pulsed laser deposition [11], reactive thermal evaporation [12], sol-gel [13,14], and solid-state synthesis [15,16].
Most studies focus on understanding the correlation between growth dynamics and electrical, optical, structural, and magnetic properties of IMO and Mn-ITO thin films [17][18][19].However, there is limited published discussion of the effect of Mn addition on the energy band structure of ITO.In this study, we investigated the dependence of the structural and optical properties of Mn-ITO thin films on the Mn content.Limited measurements of electrical properties have also been carried out.We believe that a better understanding of the band structure evolution of Mn-ITO with increasing Mn concentration will aid other researchers as they seek to understand the electrical and optical properties of this material.
Our investigation of the optical properties and energy band structure of Mn-ITO benefits from decades of study of In 2 O 3 and ITO.In early studies of optical absorption in In 2 O 3 , Weiher and Ley observed transitions that they identified as direct-allowed corresponding to an energy gap of 3.75 eV, probably at k = 0 (the Γ point) [20].They also observed transitions that they identified as indirect-forbidden corresponding to an energy gap of 2.62 eV and concluded that the maximum of the valence band is at k ≠ 0. Subsequent work on In 2 O 3 and ITO confirmed the direct allowed transitions and reported an increase in the transition energy to over 4 eV with increasing tin doping, attributed to an increase in free electron concentration in the conduction band (Burstein-Moss effect) [21][22][23][24].A smaller energy gap was also reported, attributed to indirect-allowed transitions, which also exhibited a Burstein-Moss (B-M) shift from ∼2.7 eV to ∼3.3 eV with increasing tin doping.Shifts were also observed when annealing increased the free electron concentration [24].Many additional studies on In 2 O 3 and ITO have reported similar results.
More recently, several groups have reported theoretical studies of the energy band structure of In 2 O 3 [25][26][27][28][29][30][31].Spectroscopic measurements have also been used to investigate the band structure of In 2 O 3 and ITO [27][28][29][30][31].The spectroscopic techniques included x-ray photoemission spectroscopy (XPS), x-ray emission spectroscopy (XES), x-ray absorption spectroscopy (XAS), and resonant x-ray emission spectroscopy (RXES).The energy gap of undoped In 2 O 3 was determined to be from 2.7 eV to 3.1 eV, while for ITO it can increase to ∼3.5 eV with increasing tin doping, due to the increasing carrier concentration and the resulting B-M shift [27,31].However, despite much painstaking work, the nature of this energy gap remains controversial.For example, Piper et al maintained that RXES spectra provided evidence of the lack of an indirect bandgap in agreement with their band structure calculations, which showed both the conduction band minimum and valence band maximum to be at the Γ point [30].They concluded that In 2 O 3 is a 'forbidden' gap material (i.e., direct-forbidden transitions at ∼2.9 eV).Other workers claimed that In 2 O 3 is indirect, with a bandgap of 2.7-2.9 eV and with the valence band maximum slightly away from Γ in the Γ-H direction [25,26,29,32].However, they stated that the energy of the valence band at Γ is only 10 to 50 meV below the valence band maximum.This view was supported by careful optical absorption studies [33].There is agreement that the strong optical absorption in In 2 O 3 at ∼3.7 eV is due to direct-allowed transitions from low-lying valence bands below the valence band maximum [26][27][28][29][30][31][32].Furthermore, consistent with earlier work, this transition energy increases with tin doping.
Analysis of our photoluminescence and optical absorption data for ITO gave results consistent with the recent studies just reviewed: a bandgap of ∼2.8 eV and strong optical absorption at ∼3.8 eV, caused by direct allowed transitions from a deep valence band with some B-M shift.Addition of Mn reduced the PL intensity and slightly decreased the energy gap, but it did not appear to introduce new peaks in the PL spectra.

Materials and methods
2.1.Sample preparation 2.1.1.Deposition system Manganese-doped ITO films were prepared by simultaneous dc magnetron sputtering from ITO and manganese targets.An Edwards Auto 500 sputtering system with a metal chamber and a turbomolecular pump was used.The chamber contained three sputter guns (AJA International) accommodating 7.6 cm diameter targets, and a quartz-tungsten-halogen lamp substrate heater.There was a shutter above each gun so that each target could be pre-sputtered for a time before opening the shutter to begin deposition.The rotating substrate holder was 30.5 cm in diameter and the substrates were located near its perimeter and 4.8 cm above the targets.As the substrate holder rotated (14 rpm), each substrate passed sequentially over the substrate heater and each of the sputter guns.Seven rectangular openings (2.6 cm × 7.7 cm) in the substrate holder were oriented so that during rotation they passed lengthwise over the center of each target.This provided good film thickness uniformity since each of the seven substrates experienced identical deposition conditions.Film thicknesses were measured with a KLA-Tencor P-7 surface profiler (2 μm stylus).The thickness variation over a single 2.5 cm × 7.6 cm substrate was less than ±5% and the variation of the average thickness from sample to sample was typically ±1%.
The ITO target composition was 90 wt% In 2 O 3 and 10 wt% SnO 2 and was 99.99% pure.It was 6.35 mm thick and was bonded to a copper backing plate.The manganese target was 99.95% pure, 3.18 mm thick, and was bonded to a copper backing plate.Both targets were purchased from Kurt J Lesker Co.All samples prepared for this investigation were deposited in an oxygen/argon gas mixture (10.3% oxygen) in order to establish conditions that were favorable for the oxidation of manganese.Additionally, the oxygen content reduced the manganese deposition rate, making it easier to control the manganese content of the films.

Deposition process
Soda lime glass (microscope slides) and silica substrates were included in each deposition run.They were cleaned by rinsing with acetone, isopropyl alcohol, and deionized water.They were then immersed in 22 wt% KOH solution followed by thorough rinsing with deionized water and isopropyl alcohol.The substrates were dried with a nitrogen gas stream before loading into the deposition system.The chamber was pumped down to a pressure of 4 × 10 -6 Torr while heating the substrates to approximately 135 °C as they rotated over the quartztungsten-halogen lamp.We chose 135 °C because this was the maximum achievable temperature in our system.That temperature was measured using irreversible temperature indicating labels, which were placed adjacent to the substrate during the deposition.This temperature was the result of the heat provided to the substrate by the heater combined with that generated by the sputtering guns.
The oxygen/argon gas flow was established, resulting in a chamber pressure of 4.4 × 10 -3 Torr.The Mn and ITO targets were pre-sputtered for three minutes at their desired operating power levels with their shutters closed.Then the shutters were opened for 40 min to coat the substrates.The ITO sputtering power was 90 W for all samples discussed here.A manganese oxide test deposition with 100 W power was completed and film thickness measurements were completed in order to obtain data allowing us to estimate the Mn content of our ITO-Mn films.For purposes of this estimation, it was assumed that the manganese oxide deposition rate depends linearly on the power delivered to the manganese target.Manganese sputtering power values of 13 W, 26 W, and 52 W were used to produce ITO-Mn films with nominal manganese doping levels of 3.17%, 6.15%, and 11.60%, respectively.(In this paper, compositions expressed as % refer to atomic %.) Considering the rotational rate of the substrate holder (14 rpm) and the deposition rate of ITO (8.3 nm min −1 ), approximately 0.6 nm of ITO was deposited during each complete rotation.For the Mn source, in the case of maximum manganese doping (12.8%),only 0.076 nm of manganese oxide was deposited during each rotation, which was substantially less than a monolayer.Consequently, for all the films deposited in this study, a well-defined multilayer structure was not expected.

Composition and structural characterization methods
Rutherford Backscattering Spectrometry (RBS) was used to measure the manganese content in our films.The measurements were performed at the Edwards Accelerator Lab, which is part of the Department of Physics and Astronomy at Ohio University, Athens OH.A tandem 4.5 MV accelerator beam was used to deliver alpha particles to our films, which were mounted inside an ultra-high vacuum small target chamber.A surface barrier detector measured the ions back scattered from our films at 165 degrees from the incident beam.The signal spectrum was divided into 512 channels ranging from 0 to 4.5 MeV and our data was recorded between channels 290-360, which corresponds to energies of 1.95-2.4MeV.To calculate the precise Mn content for each film, Rutherford Universal Manipulation Program (RUMP) simulation software was used to match the experimental data [34].
The surface morphology and roughness of the films were then investigated by atomic force microscopy (AFM), using a Bruker Dimension Icon instrument, at room temperature and pressure.For the measurements we used NanoScope 9.4 software, set on 'ScanAsyst in Air' to use peak force in tapping mode.The scanning rate was 1.00 Hz with 1024 samples per line.
X-ray diffraction (XRD) measurements were performed with a PANalytical X'Pert PRO MPD theta-theta diffractometer with x-ray Cu K-alpha radiation.The operating conditions for the measurements were 45 kV and 40 mA.We used a flat stationary sample stage (part number: PW3071/xx).The diffractometer program was set to scan from 20°to 70°(2θ) at a rate of 0.017 deg/s.Each measurement took approximately 51 min.

Electrical characterization
Four-wire measurements of the electrical properties of our ITO and Mn-ITO films were carried out on square van der Pauw samples [35].Indium contacts were applied to the corners of each sample using a soldering iron.Resistivity and Hall effect measurements were attempted for films of all compositions using a Keithley Model 2400 SourceMeter, or a Keithley Model 2182 Nanovoltmeter in conjunction with a Keithley Model 224 current source.Electric current values ranging from 1 μA to 1 mA were delivered to the samples depending on their resistivity, and the resulting voltages were recorded.A magnetic field of 0.80 T was used for the Hall effect measurements.These measurements, plus the known film thickness, enabled determination of a film's resistivity, carrier concentration, and carrier mobility, as described in detail by Schroder [36].

Optical characterization
Ellipsometry was done with a J A Woolam variable angle spectroscopic ellipsometer (VASE), equipped with an HS-190 light source and a VB-400 Control module.This ellipsometer could work with wavelengths from 193-2500 nm.Transmission measurements were done with the same VASE instrument.Film thicknesses were extracted from the ellipsometry data.
Photoluminescence spectra of films deposited on silica substrates were recorded using a Jasco spectrofluorometer FP-8500 with a Xe lamp source.The selected excitation wavelength was 250 nm (energy = 4.96 eV) with a FWHM of 20 nm.

Results and discussion
3.1.Composition and structural properties 3.1.1.Thickness measurements and RBS analysis As mentioned previously, film thicknesses were determined both with a surface profiler and by extraction from ellipsometry data.The thickness measurements presented in table 1, show a good agreement between the two techniques.

AFM analysis
AFM images of undoped and Mn-doped ITO thin films deposited on silica substrates are shown in figure 1.An uncoated silica substrate is shown for comparison.In all cases the scanning area was 1 × 1 μm 2 .The uncoated silica substrate [figure 1(a)] showed a smooth surface with a roughness of ∼0.4 nm and no distinguishable surface structures.
Figure 1(b) shows the surface of a 332 nm thick film of undoped ITO with overlapping island-like grain structures.These slightly elongated structures had an average grain area of ∼2600 nm 2 , revealing a rather homogenous coating on the silica substrate.The surface of the film had a root mean square (RMS) roughness of ∼4.1 nm.
As we increased the Mn concentration from 0% to 6.1%, figure 1 and table 2 showed that Mn-doped ITO films displayed an increase of the grain area, while the 12.8% Mn doped sample showed a reduction.The RMS roughness also increased with increasing Mn concentration and then decreased for the 12.8% Mn doped sample.It has been reported that a grain consists of several crystallites each with different boundaries and defects [37].If this is the case, this could explain the larger size of the grains reported in table 2 compared with XRD results that follow.However, our AFM images did not confirm or disprove the presence of crystallinity.

= + +
For these calculations, the (440) peak at 2 q ; 51°was used, with n = 1 and λ = 1.5406Å.The results are presented in figure 3.While a lattice constant of 11.2 Å was expected for In 2 O 3 based on JCPDS Card No. 06-0416, a decrease was expected when ions with smaller radii, such as Sn 4+ and Mn 2+ (0.81 and 0.83 Å, respectively) substitute for relatively larger In3+ ions in the Mn-ITO films.The lattice constant of our undoped ITO (no Mn) was 10.15 ± 0.01 Å and did not agree well with reported values of 10.10 ± 0.01 Å by [12], but they deposited on oxidized silicon.Our result also did not agree well with [7] who also deposited on oxidized silicon.However, our value agreed better with the lattice constant of 10.12 Å for In 2 O 3 reported by [38,39].
The Debye-Scherrer formula was used to calculate the crystallite size (G) [40]: where K is a constant that depends on the geometry of the particles (for spherical shape K ∼ 1), λ is the wavelength of the x-rays (1.5406 Å), the full width at half-maximum (FWHM) is represented by β, and q is the diffracted angle.We estimated the average crystallite size of undoped ITO to be around 48.9 ± 11.1 nm, which is consistent with [8].The remaining values were 57 ± 20 nm, 44 ± 17 nm, and 40 ± 15 nm, for Mn contents of 3.7%, 6.1%, and 12.8%, respectively.Although, the errors in the measurement of crystallite sizes were significant, our results suggested that there was an initial large increase in crystallite size from 48.9 nm (ITO) to 57 nm (3.7% Mn) after which the sizes decreased monotonically for the 6.1% and 12.8% Mn concentrations.It was possible that the sample with 3.7% Mn happened to be in a metastable equilibrium state.
Figure 3 shows that the lattice constant decreased with increasing Mn concentration.As was observed earlier from the AFM images (figure 1), the undoped ITO sample showed some island-like structures (grains).Doping ITO with 3.7% Mn did not make a significant change in the grain boundaries.Though for the 6.1% doped  sample, smaller grains were formed within the larger island-like structures, and then 12.8% doped ITO showed even larger islands.We propose that larger concentrations of Mn (in the presence of oxygen during the growth) increased the disorder in the film.We further tried annealing these films for electrical measurements (not shown here).We observed cracking of the films deposited on silica during the annealing and cooling process which was potentially attributed to the added strain resulting from higher Mn concentrations.

Electrical properties
As mentioned previously, all samples prepared for this investigation were deposited in an oxygen/argon gas mixture (10.3% oxygen) in order to establish conditions that were more favorable for the oxidation of manganese sputtered from the metallic target.This high oxygen content enhanced the film transparency and also reduced the manganese deposition rate, making it easier to control the manganese content of the films.However, the film resistivity was higher than it would be with a lower, optimized oxygen flow because oxygenrich deposition conditions hindered the formation of donor defects (e.g., oxygen vacancies and indium interstitials).Moreover, it has been reported that under these conditions oxygen interstitials were introduced, and they acted as acceptors, compensating nearby tin donors by forming neutral clusters with them [41].Van der Pauw measurements on ITO films (0% Mn) deposited on both glass and silica substrates indicated a resistivity of 10.6 Ω cm.Hall effect measurements on these samples were unsuccessful, since the Hall voltage was small compared to the measurement noise.Manganese-doped samples were unmeasurable because their resistivities were too high.Manganese has been reported to substitute for indium as Mn 2+ , acting as an acceptor [7,14,42].Thus, it could compensate for tin and donor defects, increasing the resistivity.A thorough discussion of annealed samples is beyond the scope of this paper, however, it should be noted that, for all samples deposited on soda-lime glass substrates, annealing in a reducing atmosphere (3% H 2 /97% N 2 ) at temperatures from 375 °C to 610 °C reduced their resistivities, to < 1 × 10 −2 Ω cm for zero and 3.7% Mn content.Hall effect measurements were also possible on the annealed samples with 0%, 3.7% and 6.1% Mn, confirming n-type material and yielding carrier concentrations of 3.6 × 10 19 -5.7 × 10 20 cm −3 and electron mobilities of 2.1-16 cm 2 /(Vs).Films deposited on silica substrates cracked during annealing.It should also be noted that ITO films (0% Mn) deposited from this same oxide target with the same pre-sputtering conditions, but with no oxygen flow and no substrate rotation during deposition, were transparent and had resistivity of 6.2 × 10 −4 Ω cm, carrier concentration of 3.0 × 10 20 cm −3 , and electron mobility of 33 cm 2 /(Vs).

Optical properties 3.3.1. Transmission and ellipsometry
The optical transmission spectra in figure 4 showed that all samples absorb high energy photons.The onset energy for this absorption decreased monotonically with increasing Mn concentration.At low energies the average transmission of all samples exceeded 80%, exhibiting maxima and minima due to optical interference.The wavelengths at which the maxima and minima occur were consistent with the film thickness (∼360 nm) and its refractive index (n = 1.96-2.15)over the transparent wavelength range.
Spectroscopic ellipsometry is an indirect optical characterization method in which two parameters are measured, namely the amplitude, Ψ exp , and the phase shift, Δ exp , of the light interacting with the film.These data were then compared to a theoretical model.It should be noted that this model incorporated transmission data in addition to the Δ/Ψ data which were ordinarily used.The theoretical model for all our films was created in consultation with J A Woollam Co(Lincoln, Nebraska USA).Better fitting between the experimental and the theoretical values leads to more accurate determination of the optical constants, n and k, where k is the extinction coefficient.The absorption coefficient for all measurements was calculated by using .
Ellipsometry data was measured between 300-1100 nm and for two angles of incidence, namely 50 degrees and 70 degrees.Figures 5(a) and (b) show examples of modeling Delta and transmission data together for the 12.8% Mn doped ITO film.The thickness determined with a stylus profiler (table 1) was used as a starting point for modeling each sample's data.The model produced a refined value of each sample's thickness (table 1).The quality of the fit was determined by using mean square error (MSE) method (a value of 10 or less is considered a good fit).Our MSE values were 6.
Figure 5(c) shows the absorption coefficient extracted from the models of figures 5(a) and (b).The absorption coefficient data were analyzed by generating Tauc plots.These plots were used to determine optical transition energies by assessing our samples' agreement with this relation introduced by Tauc and supported by other researchers [43][44][45]: where B is a constant, hν is the photon energy, E g is the energy gap, and the exponent n is , 1 2 , 3 2 2 and 3 for direct allowed, direct forbidden, indirect allowed, and indirect forbidden transitions, respectively.The value of E g was determined by plotting ( h a n) 1/n versus hν and extrapolating the linear region to its intercept with the energy axis.Figure 5(d) shows a sample Tauc plot generated from the absorption coefficient data of 12.8% Mn doped ITO [figure 5(c)], assuming indirect allowed transitions.Transition energy values from the complete analysis of all samples are presented in figure 6.For all transitions except direct allowed the associated transition energy decreases with Mn addition.As discussed in section 3.3.3below, all four transition types were not expected to be relevant to our samples.

Photoluminescence analysis
Figure 7(a) shows the room temperature PL spectra of ∼360 nm thick films of undoped ITO and Mn doped ITO of all compositions on silica substrates.The overall PL intensity decreased with Mn addition.We tentatively attributed this to the formation of additional defects within the crystalline grains and at the grain boundaries, which acted as nonradiative recombination centers.There was a disproportionate decrease in PL intensity at ∼3.28 eV relative to that at ∼3 eV, further suggesting the introduction of a nonradiative recombination route with Mn addition.There was no clear energy shift of the peaks with Mn addition.
As reported by other groups, the PL emission from ITO covers a broad energy range [17,18,[46][47][48][49][50].Some researchers used lower energy photons for PL excitation and, as a result, they do not see the high energy (∼3.8 eV) emission.We saw less low energy PL (< 2.1 eV) than was reported in some previous papers [48,50].Unlike some research with nanoparticles [17], our films were much thicker and had grains much larger than the Bohr radius of In 2 O 3 (2.14 nm) [51], so quantum confinement was not considered in this paper.
The spectra were modeled by fitting multiple gaussian peaks to the experimental results.An example is provided in the supplementary materials (figure S1).Numerous peaks and features are evident in figure 7 of   which two identified shoulders are consistent with absorption transitions, as discussed in the next section.At the PL excitation energy (4.96 eV) the transmission of our films was 0.4%.Consequently, there was little excitation of the substrate, and no significant contribution of the substrate to the PL spectra.Further discussion of the optical properties follows.

Discussion of optical properties
We now discuss the PL spectra in greater detail, comparing PL features with transition energies determined from Tauc analyses where appropriate.Since the dynamics of the absorption and PL transitions differ, we expect agreement between absorption and emission transition energies only in special circumstances, as discussed in the following paragraphs.For example, the absorption process requires accessible empty states at an appropriate energy above some filled states, whereas the PL emission process requires accessible empty states at an appropriate energy below some filled states.Absorption and PL transitions at the same energy may involve very different sets of states.We begin by examining the PL spectra in figure 7 in conjunction with transition models and summarizing Tauc plot and PL data in figures 8 and 9.The addition of Mn clearly reduced the PL intensity across the spectrum, but it did not introduce any new peaks in the spectra of figure 7.However, there were changes in the relative strengths of the peaks, discussed below.We do not believe that excitons are involved in any of the observed PL transitions, since our deposition conditions were not conducive to high material purity and good crystalline perfection, and the spectra were measured at room temperature.The resulting local fields and thermal energy tend to break up excitons into free carriers [52].We begin by considering the undoped ITO PL spectrum in figure 7 for energies below 2.9 eV, where a number of peaks are evident.Beginning at 2.9 eV and going to lower energies, the shoulder first encountered appears to originate from a peak at ∼2.82 eV.This PL may result from transitions from the conduction band minimum (CBM) to the valence band maximum (VBM), as illustrated in figure 8(a), or perhaps another near-bandgap transition.This assignment supports theoretical studies, comprehensive spectroscopic measurements (XPS, XES, etc), and absorption coefficient measurements reported by several groups [27-29, 31, 33] who have concluded that the true bandgap of In 2 O 3 and ITO (i.e., not including B-M shift) is 2.7-2.9 eV.The next peaks at lower energies may be caused by transitions to or from acceptor (A) and donor (D) levels, both shallow and deep, as reported for other semiconductors [52][53][54][55].For example, the peak at 2.75 eV may be caused by a transition from a shallow donor level (D) to the VBM (figure 8(a)).The energy difference between this peak and the proposed bandgap is 2.82-2.75= 0.07 eV, which may be a donor ionization energy.The next peak is at 2.64 eV, and PL near this energy has also been reported in ITO nanoparticles and thin films by two groups [18,46,47].The lower energy PL, down to 2 eV, is assumed to involve states farther from the band edges.
There are numerous possible sources of energy levels in the bandgap of Mn-ITO.As discussed earlier, Sn and Mn atoms can replace In atoms in In 2 O 3 and act as donors and acceptors, respectively.A number of defects can also act as dopants in In 2 O 3 and ITO, as reviewed by Bierwagen [56] and Chatratin et al [57].Examples of donor defects include oxygen vacancies, indium interstitials, and complexes of these defects.Examples of acceptor   defects include indium vacancies and oxygen interstitials.Transitions involving these donor and acceptor levels may contribute to the PL spectra in figure 7. It is noteworthy that these spectra include emission over a large range of energies below the energy gap of ∼2.82 eV (E g ), suggesting the existence of a large range of dopant energy levels in the energy gap.The number and range of dopant energy levels may be increased by the fact that the In 2 O 3 Bixbyite crystal structure has two nonequivalent In sites [56,57], as well as nonequivalent interstitial sites.So, for example, In vacancies, In interstitials, and O interstitials in nonequivalent sites would likely have slightly different energy levels.A defect energy level may also depend slightly on the proximity of the defect to Sn donors, grain boundaries, the film surface, and the film's interface with the substrate.Interstitial hydrogen has also been identified as an unintentional donor dopant in In 2 O 3 and ITO [56].Desorbed water in the sputtering chamber can be a source of hydrogen in the deposition plasma.It must also be noted that phonon emission may be involved in some of the transitions.The LO phonon energy in In 2 O 3 and ITO has been reported to be 0.07 eV [20,48,58].
We next consider photoluminescence at energies greater than 2.9 eV.This requires transitions that either begin at energies above the CBM or terminate at energies below the VBM, or both.As previously discussed, n-type doping by tin or native defects can fill lower conduction band states, enabling transitions to and from states higher in the CB (the B-M effect) [21-24, 27, 31].Returning to the undoped ITO PL spectrum in figure 7, beginning at 3.5 eV and going to lower energies, the shoulder first encountered is at 3.28 eV.Luminescence in the vicinity of this energy is believed to result from transitions of electrons higher in the CB to the top of the VB (figure 9(c)).Although this undoped ITO film had a high resistivity (∼10 Ω-cm) and Hall effect measurements were not possible, we believe that this is in part due to high resistance of the grain boundaries and low electron mobility.After mild annealing (325C in N 2 for 60 min.),Hall effect measurements could be completed, establishing that the carrier concentration and electron mobility were ∼2 × 10 20 /cm 3 and 4.5 cm 2 /(versus), respectively.So, it is not unreasonable to expect some degree of B-M shift, even in unannealed material.The shift in our ITO is ∼0.46 eV, while that reported for a carrier concentration of ∼2 × 10 20 /cm 3 ranges from 0.3 eV to 0.5 eV [22,23,26].The assignment of the 3.28 eV PL to the B-M shifted bandgap (B-M shifted Eg) is in reasonable agreement with the direct forbidden Tauc plot analysis, as shown in figure 9(a).The structure in the PL emission at energies between 3.28 and 3.0 eV may be caused by transitions of electrons in the CB to acceptor levels.
Irmscher et al [33] carried out careful optical absorption studies of undoped bulk single-crystal In 2 O 3 samples.Their samples were n-type but had a low carrier concentration (∼10 17 /cm 3 ).Their Tauc analyses for direct forbidden and indirect allowed transitions both showed excellent linear fits to the absorption data, yielding transition energies of 2.80 eV and 2.76 eV, respectively.The low carrier concentration of their samples resulted in negligible B-M shift, so their 2.80 eV direct forbidden transition energy agrees well with our assignment of 2.82 eV PL to the ITO energy gap.Our direct forbidden transition energy for ITO (3.18 eV) was B-M shifted since electrons from the valence band could not be excited to the filled states near the CBM (figures 9(a) and (c)).
The addition of 3.7% Mn reduced the PL intensity at all energies, but the reduction was greatest at ∼3.28 eV (figure 7).We would expect that a reduction of the free electron concentration due to compensation by Mn 2+ acceptors would cause the 3.28 eV peak to shift to lower energy (page 125 of [52]), but this did not appear to happen.Perhaps there was incomplete activation of Mn acceptors due to the low deposition temperature.High deposition temperature or post-deposition annealing have been found necessary to activate other transition metal dopants in In 2 O 3 [59][60][61][62].
An alternative explanation for the PL between 2.9 eV and 3.5 eV is considered next.Walsh et al. [27] have established that the strong optical absorption observed in In 2 O 3 and ITO at energies > 3.55 eV is caused by electron transitions to the CB from a deep valence band that is 0.81 eV below the VBM.This explains the absorption and PL data summarized in figure 9(b) and is discussed further in the next paragraph.The deep valence band could possibly also participate in the 2.9-3.5 eV PL.In figure 7 consider the PL from 2.15 eV to 2.5 eV.If this were due to transitions from a band of deep donor levels to the VBM, then transitions from these donor levels to the deep valence band could produce PL 0.81 eV higher in energy, i.e., from ∼3 to 3.3 eV (figure 8(b)).This explanation seems less likely since the very strong PL from ∼3 to 3.3 eV would require the optical excitation to produce a substantial concentration of empty states in the deep VB.We expect electrons from the higher valence bands to rapidly move to the lower VB, reducing the steady state hole concentration there.We believe that this is the explanation for the relatively weak PL at 3.8 eV, discussed in the next paragraph.
Finally, we consider the peak at ∼3.8 eV in figure 7.This PL results from transitions of electrons from the conduction band to the deep valence band 0.81 eV below the VBM [27] (figure 9(d)).Harinath Babu et al [46,47] observed a distinct PL peak at 3.76 eV in ITO nanoparticles.

Conclusions
In this study we performed detailed optical and structural characterization of Mn-doped ITO thin films.Photoluminescence measurements at room temperature revealed a fundamental bandgap of 2.82 eV for ITO.This is consistent with numerous recently reported theoretical and experimental investigations, which claim the bandgap to be 2.7 to 2.9 eV.Measurements also showed a transition at ∼3.28 eV (∼0.46 eV above the bandgap) which is attributed to the Burstein-Moss effect.This transition was observed in both emission (PL) and absorption measurements, and the Tauc analysis supported a direct-forbidden transition, which is also consistent with a number of recent reports.With increasing Mn concentration, a monotonic decrease of the PL intensity was observed, and PL and Tauc analyses showed a small decrease in the transition energies.PL peaks at ∼2.75 eV, ∼2.64 eV and lower energies are believed to involve donor and/or acceptor energy levels in the bandgap, which result from dopants (Sn & Mn) and numerous defects.PL and absorption transitions were found at ∼3.8 eV, which is in agreement with numerous groups, and is consistent with the reported involvement of a separate deep valence band.

Figure 2 .
Figure 2. XRD measurements of ITO thin films with various Mn percentages concentration dopants.

Figure 3 .
Figure 3. Lattice constant as a function of Mn concentration measurements based on the peak at (440).

Figure 4 .
Figure 4. Transmission measurements as a function of Mn %.

Figure 5 .
Figure 5. Examples of ellipsometry and transmission measurements used to model the energy bandgap of the 12.8% Mn doped ITO film: (a) delta measurement, and (b) transmission measurement, showing how well the VASE model (solid lines) fits with the experimental data (dashed lines).All ellipsometry measurements were done at 50-and 70-degree angles.(c) absorption coefficient [α] as a function of wavelength.(d) Tauc plot of h a u versus energy for the indirect allowed band gap determination.

Figure 6 . 1 2
Figure 6.Transition energy values extracted from Tauc analyses of all undoped and Mn-doped ITO thin films.The exponent in equation (1) was 1 2 for direct allowed, 2 for indirect allowed,

3 2
for direct forbidden, and 3 for indirect forbidden.

Figure 7 .
Figure 7. Photoluminescence measurements of the undoped and Mn doped ITO thin films at room temperature.

Figure 8 .
Figure 8. Examples of transitions: (a) from CBM to VBM and from D to VBM and (b) from deep states in gap to deep VB.There is little dispersion in the valence bands, so for simplicity, they were drawn as horizontal lines [27, 29].

Figure 9 .
Figure 9. Transition energy values extracted from Tauc analyses on undoped and Mn-doped ITO thin films.The exponent in equation (1) was (a) 3/2 for direct forbidden and (b) ½ for direct allowed transitions.PL transition energy values are shown in the same plots.The error bars for values extracted from Tauc analysis were estimated to be constant.The PL error bars were estimated from both experimental and fitting analysis.Energy band diagrams (c) and (d) show the absorption (Tauc) and emission (PL) transitions corresponding to (a) and (b), respectively.
Figure 9(b)  shows good agreement between the PL energy and the direct allowed Tauc analysis.It has been argued that, in the absence of any B-M shift, the transition energy is 3.55-3.7 eV[26,29,63].This is consistent with the ∼2.82 eV bandgap plus the 0.81 eV offset of the deep valence band.Our measured transition energy (∼3.8 eV) indicated some B-M shift (figures 9(b) and (d)).

Table 1 .
RBS composition and thickness measurements for Mn-doped ITO samples.
3.1.3.XRD analysisThe diffraction patterns observed for the four compositions studied in this work are shown in figure 2. All could be indexed based on the cubic bixbyite structure of In 2 O 3 (JCPDS Card No. 06-0416).No additional peaks corresponding to any secondary phases were detected.

Table 2 .
Average grain area and RMS roughness of the undoped and Mn-doped ITO samples, derived from AFM analysis.