An EPR study of point defects in zinc oxide thin films

We studied ZnO thin films deposited on glass slides by thermal evaporation in vacuum followed by heat treatment in open air or Ar flow. The samples were characterized using x-ray diffraction, Raman scattering, photoluminescence (PL), and electron paramagnetic resonance (EPR) spectroscopy. We observed a broad PL band in the visible region at spectral positions of 504 and 560 nm and a low-intensity band in the UV region at 390 nm. The EPR spectra display a clear first derivative structure at g=1.96 at temperatures below 200 K. The PL spectrum shows red-shifted valence to conduction band emission due to electron hole recombination through shallow surface states. We found an activation energy of E a = 4.2 meV using the Arrhenius plot and estimated concentrations of paramagnetic defects and spin–spin relaxation time constants of 1016 m−3 and 10–14 s, respectively.


Introduction
Zinc oxide (ZnO), which has a high technological application potential, is a semiconductor material used in numerous studies.Non-toxic semiconductor-based heterogeneous photo-catalysts such as ZnO have been scrutinized for decades.ZnO, with a wide band gap energy (3.37 eV) and high excitonic binding energy (60 meV), has excellent optical properties for solar cells and, can replace expensive indium tin oxide (ITO) plates and light-emitting diodes (LED) [1][2][3][4].Shuai et al [5] studied pulsed electron beam-deposited ZnO thin films heat treated under an external magnetic field to elucidate defect generation.Recent reports on the growth of ZnO and defect structures show that ZnO is still an up-to-date research field [6][7][8][9][10][11] Production of electron-hole pairs under UV radiation provides decomposition of organic impurities through the formation of hydrogen peroxide (H 2 O 2 ) and hydroxyl radicals (•OH) [11][12][13].Previous reports show that it is a well-established procedure for growing zinc oxide thin films on glass slides with defect structures by resistive heating of zinc powder under vacuum followed by heat treatment in open air or a controlled atmosphere.The evolution of zinc to a zinc oxide structure triggers the emergence of point defects such as oxygen vacancies (V O ), zinc vacancies (V Zn ), interstitial zinc (Zn i ), and interstitial oxygen (O i ).However, the assignment of green-light (GL) PL emissions to defects has been an unresolved controversial issue.The GL emissions are observed in a wide range of visible spectra at 525 nm [5], 496 nm and 559 nm [8], and 523 nm and 550 nm [14].The origin of GL is unclear despite an enormous number of studies.Vanheusden et al found a relationship between the density of charge carriers produced under light and green region photoluminescence, and proposed that the emitted radiation was due to singly ionized oxygen vacancies [15,16].In contrast, Djurisic et al [17][18][19] found no correlation between PL emission and EPR signals based on their analysis using SEM, XRD, EPR, and temperature dependent PL and PLE spectroscopy.Savchenko et al [20] studied the point defects in hydrogenated ZnO microparticles using EPR spectroscopy and conductivity measurements.They proposed four intrinsic defects which are responsible for the high conductivity of ZnO: (i) hydrogen interstitials (H i ), (ii) zinc interstitials (Zn i ) (iii) singly ionized oxygen vacancies (V O ), and (iv) shallow donors.
In this study, zinc oxide thin films were grown on glass slides by thermal evaporation of zinc powder in vacuum, followed by heat treatment in open air and Ar flow to control the oxygen content.We employed x-ray diffraction, Raman scattering, photoluminescence, and electron paramagnetic resonance spectroscopy to illustrate the intrinsic defect centers and contribute to the ongoing research.We observed GL PL, which might be due to point defects, such as interstitials and vacancies.Our primary focus is on identifying and characterizing the different types of defects present through g -factor determination.
A recent report by Zhang et al [8] included data comparable with the EPR/PL results, as in this work.They reported a g -value of ∼1.96 observed in the EPR spectra of ZnO, which is expected to originate from to singly ionized oxygen vacancies.We note that the PVD process employed in this work to grow thin films followed by heat treatment in an oxygen-rich/poor atmosphere provides better control of the defect structure.Therefore, we observe strong GL emission in the PL spectra, which is a better distinguished and clear low-temperature EPR signal assigned to singly ionized oxygen vacancies.

Experimental details
ZnO thin films were deposited on glass slides with dimensions (3 mm × 26 mm) in a vacuum chamber using the Physical Vapor Deposition (PVD) technique, followed by heat treatment in open air or in a flow of Ar to reduce the oxygen percentage.PVD has been described previously [21].The samples studied in this work are listed in table 1. Thin film samples were deposited on glass slides at a rate of 1-2 Å s −1 to a thickness of 400 nm, and one thin film was left as deposited without any heat treatment (sample 1) and one of them was ITO coated glass slide (sample 4) after deposition, heat treated at 600 °C for 2 h in open atmosphere or in a flow of Ar (samples 2-5).
x-ray diffraction (XRD) spectra were recorded using a Bruker D8 Advance diffractometer.Raman measurements were carried out on a Renishaw 250 mm focal length in a Via Spectrometer system with a CW laser line at 532 nm as the excitation source.The photoluminescence (PL) spectra were excited by a Xe lamp at 300 or 380 nm.The EPR spectra were measured on a Bruker EMXnano spectrometer in the X-band (9.64 GHz) at temperatures between 90 and 300 K with power P o at between 3.16 and 10 mW, unloaded quality factor Q U = 4160, center field = 329 mT, cavity volume V C = 2.05 × 10 -6 m 3 , temperature of the detector T d and temperature of the sample T s set equal to room temperature (300 K).Detailed information on the EPR is provided in the supplemental information.

Results and discussion
Using the manufacturer's specified values with the help of commonly accepted literature, the noise factor (F) was set to 1, the half-width at half-height of the single absorption line (Γ) was 0.1 mT, the spin state (S) was 1/2, and the bandwidth of the complete detecting and amplifying system (b) was 1 s −1 .We estimate the number of defects by an order of magnitude using the equation [22]: We found N min ≈ 10 9 and the concentration of paramagnetic defect centers was calculated as ∼10 16 m −3 .A Gaussian absorption band was assumed for the resonant transition of the electromagnetic field.The structure observed in the EPR spectrum corresponds to the first derivative line shape: where A o is the EPR intensity, and B 0 is the magnetic field at which resonant absorption occurs.The value of g is given as where h is Planck constant, β is electron Bohr magneton and f is frequency.However, the lines at resonance absorption broaden homogeneously owing to weak dipolar interactions between paramagnetic centers, assuming that they are sufficiently close.The interaction of the electromagnetic field, represented by a Gaussian profile, with paramagnetic defects at microwave frequencies in an external magnetic field produces the first derivative line shape.The EPR intensity is given as follows where N is the number of paramagnetic centers, / g g 4, P = B 1 is the amplitude of the magnetic component of the radiation, and T is the temperature of the sample [23].The spin-spin relaxation time (τ) is dependent on the temperature through the relation:     where we take g 1.96 = assuming isotropic conditions.We note that the range of time constants we calculated using the EPR line width and equation ( 5) is consistent with the results of transient absorption spectroscopy [25].
The XRD pattern in figure 1 for sample 2, which was heat-treated in Ar flow, completely overlapped with that of sample 5, which was heat-treated in the open air.The Raman spectra of samples 1 and 3 are shown in figure 2. This result shows that the XRD pattern is sensitive to diffraction from crystal structures but not to defect structures, whereas Raman scattering is sensitive to composition through lattice vibrations.The crystallographic planes of wurtzite ZnO were assigned to the diffraction peaks of the ZnO thin film heat-treated at 600 °C for 2 h   in an open atmosphere (sample 5).The zinc thin film deposited without heat treatment displays a structure at 561 cm −1 which disappears with the emergence of a strong Raman line at 437.2 cm −1 upon heat treatment, assigned to the oxygen vibration in ZnO.The observation of the vibrational mode at ∼582 cm −1 in the heattreated sample may be due to oxygen vacancies in the zinc-rich thin film [26].We show the steady-state PL spectrum for sample 3 in figure 3 for the two excitation wavelengths.The PL spectrum excited at 300 nm (4.13 eV) had a high-intensity broad asymmetric band in the low-energy visible region and a low-intensity shoulder at high energy below the band gap.The high-energy band disappeared, and the intensity of the lowenergy band region decreased when the spectrum was excited at 380 nm (3.26 eV).The PL spectra were identified by taking the first minimum of the second derivative of each spectrum.The spectral positions of possible light emission are marked by arrows along with emission wavelengths of 390, 504, and 560 nm.
The PL spectra excited at 300 nm, which is well above the band gap energy of zinc oxide (3.37 eV) shows redshifted valence to conduction band emission due to electron hole recombination through shallow surface states.The magnitude of the red shift was calculated to be 190 meV.This suggests non-radiative thermalization through the defect levels within the forbidden band gap.Excitation at a higher wavelength of 380 nm below the band gap at 368 nm substantially decreases the intensity of the GL emission.The EPR spectra at 300 K (room temperature RT) and 100 K are shown in figure 4 for (a) reference glass slide, (b) zinc-deposited glass slide, and then heat-treated at 600 °C for 2 h in a flow of Ar at 5 bar (sample 3), and (c) ITO-coated glass slide on which zinc was deposited and then heat-treated at 600 °C for 2 h in flow of Ar flow at 3 bar (sample 4).The clear first derivative feature at 350 mT observed for sample 3 heat-treated at 5 bar, which is absent in the spectrum of sample 4 heat-treated at 3 bar, indicates the connection between the heat treatment under different oxygen contents and the concentration of paramagnetic centers.No clear EPR signal was observed at room temperature because of the high conductivity of the samples [20].However, we observed a clear first derivative feature at g 1.96 = as the temperature decreased from 200 to 100 K.It should be noticed that only singly ionized oxygen vacancies (V O + or zinc interstitials Zn i ) were observed in the EPR spectra because these are paramagnetic defects.
However, we observed a single paramagnetic defect structure in the EPR spectra, whereas two emission lines were observed in the PL spectra.This type of two-structure PL spectrum for ZnO powder thin films was also observed by Zhang et al [8] at a slightly different spectral position: 559 nm (2.22 eV) and 496 nm (2.5 eV).Temperature-dependent EPR spectra are shown in figure 5.The broad and barely noticeable structure emerges at around 350 mT in EPR spectroscopy conducted at room temperature and its transformation to a strong first derivative-like structure as the temperature decreases, as shown in figure 5.The dominant first derivative structure in figures 4 and 5 around g 2.003 = is due to nearly free electrons.Less intense temperaturedependent structure at around g 1.96  = is shown in a larger scale inset in figure 5 [27].The features observed at between 348 and 352 mT in the low-temperature EPR spectra were fitted by the first derivative line shape for a temperature range from 200 to 100 K.The best fit parameters for line width B d ( ) and peak height (EPR intensity) were determined.We insert the numerical values for the constants h, k B , β and B 0 = 0.350 T, f = 9.64 GHz, and g 1.96 = into equation (3) and calculate the temperature-dependent intensity I = 2.23 × 10 19 × tanh(0.231/T),according to which the EPR intensity increases with decreasing temperature.When analyzing the EPR spectra, it should be noticed that the intensity of the signals becomes higher at lower temperatures because the intensity of the signals is much higher than that of the noise signals, which is why it appears to have lower noise at high temperatures.To lower the noise-to-signal ratio, noise depends only on the number of scans and higher scan loads.
Employing equation (4), we relate the time constant to the line width and plot the natural logarithm of the relaxation time constant against 1000/T.The slope of the straight-line fit through the data points determines the activation energy of E a = 4.2 meV.An Arrhenius plot is shown in figure 6.
Zinc rich thin film sample 3 was heat-treated after deposition under a low oxygen atmosphere, which is why it is reasonable to assume the presence of a high density of oxygen vacancies in the ZnO thin films.As observed in figure 5, the EPR displays the first derivative line shape at g 1.96 = at low temperatures, which might be due to paramagnetic V O + defects related to the green emission in the PL spectra at 504 nm, which dominates the entire PL spectrum.

Conclusion
We studied the defect structure in ZnO thin films and observed a broad PL band in the visible region at spectral positions of 504 and 560 nm and a low intensity band in the UV region at 390 nm.The EPR spectra display a clear first derivative structure at g 1.96 = at temperatures below 200 K.Heat treatment after deposition under a high flow rate of Ar, that is, an oxygen-poor atmosphere, initiates the generation of oxygen vacancies.We calculated the following material parameters: (i) concentration of paramagnetic defects of 10 16 /m 3 , (ii) spin-spin relaxation time of 10 -14 s, and (iii) activation energy of 4.2 meV.The measurements conducted in this study are summarized in table 2.

Figure 1 .
Figure 1.XRD spectra for as deposited thin films heat treated at 600 °C for 2 h in open air (sample 5) and in Ar flow at 5 bar (sample 2).

Figure 2 .
Figure 2. Raman spectra for as deposited thin films (sample 1), heat treated at 600 °C for 2 h in open air (sample 5) and heat treated at 600 °C for 2 h in Ar flow at 5 bar (sample 3).

Figure 3 .
Figure 3. PL spectrum for as deposited thin films heat treated at 600 °C for 2 h in Ar flow at 5 bar (sample 3).bottom is PL with an excitation wavelength equal to 380 nm, in the middle the excitation wavelength is 300 nm, top is the first minimum in the second derivative of each PL spectrum.

Figure 4 .
Figure 4. EPR spectra for reference glass slide (a) sample 3 (b) and sample 4 (c) at room temperature and 100 K.

Figure 5 .
Figure 5. EPR spectra for sample 3 at between 300 and 100 K. intensity-temperature-dependent structure at around g 1.96 = .

Figure 6 .
Figure 6.Natural Logarithm of EPR signal relaxation time plotted against 1000/T.

Table 1 .
The samples investigated in this study.All the samples were deposited at a rate of 1-2 Å/s to a thickness of 400 nm on glass slides (Sample 4 was ITO coated glass slide).Samples 2-5 were heat treated at 600 °C for 2 h after deposition.

Table 2 .
Summary of the measurements from this study.