Computational Analysis of ZnO as a potential Transparent Conducting Oxide alternative,

DFT analysis of the effects of Zn-site and O-site dopants.

Author: Rohan Sawhney

Department of Nanoengineering, Jacobs School of Engineering, University of San Diego, California

Abstract:

With the increasing demand for electronics globally, and with no end to the expansion visible, the supply chain for electronics and their use of exotic, rare materials is a significant point of contention. While recycling older electronics and recovering used materials for reuse is a possibly viable strategy for the future, another possibility is switching to cheaper, more abundant materials to replace the scarce ones.

One such substitution could be of transparent conducting oxides (TCOs) used in most display technologies today, as well as touch interfaces. The required properties of TCOs are due to their use cases and their ability to be applied to see-through surfaces without affecting the display's image quality.

Zinc(II) Oxide (ZnO, Zn(II)O) provides this much-needed alternative due to the rareness of Indium (In)2, the main functional component of ITO. Indium itself is considered a byproduct of Zinc and Copper extraction, and Oxygen is not in short supply. Thus, if ZnO could be modified or treated to attain the required properties to replace ITO, it would greatly benefit TCO manufacturing and potentially reduce the cost of the devices themselves.

Introduction:

Zinc can be naturally found in the oxidised form of Zn(II)O. However, the precise crystalline form varies depending on the method of formation. These crystal structures display multiple morphologies, each having slight differences in their stability at room temperature. The most common morphologies are of Wurtzite and Zincblende.

Therefore the first task is to determine thenviability of each as a TCO by comparingntheir resting lattice energies to determine themost stable morphology with which to proceed. This can be done computationally, using a large supercell of each crystal and carrying out a relaxation using suitable pseudopotentials for Zn and O, with the Quantum Espresso (QE)5 suite of Density Functional Theory (DFT) calculations. Once the favourable morphology is found, this study can then commence with further DFT analysis to determine the Density of States (DoS) of the pure ZnO crystal.

Due to these restrictions, the dopants selected for this study are - Aluminium (Al), Gallium (Ga), Fluorine (F), Chlorine (Cl). The Al and Ga are Zn-site dopants, whereas the F and Cl are O-site dopants. In either case, these dopants are hypothesised to cause ZnO to act as an n-type semiconductor, thus increasing the electron density in the conduction band - increasing conductivity. To standardise the effects of any dopant concentration variance, each crystal is doped evenly

  • Wurtzite

    Wurtzite, a hexagonal crystal lattice with a P6(3)mc spacegroup.

  • Zincblende

    Zincblende, a cubic lattice with an F4(bar)3m spacegroup.

Method:

To compare the viability of Wurtzite and Zincblende to find the more stable crystal arrangement, which will then be used for further analysis. Both crystals were run through a complete relaxation cycle, and the final energies were measured. A relaxation cycle was carried out using the method recommended by Prof. Yang, with ibrav = 0, with the cell parameters specified in the input file. The cutoff energy was set to 28eV, and the k-point grid used was (10 10 6 1 1 1).

After determining the appropriate lattice, a Density of States computation of the pure ZnO was performed using the following task path:

  • Creating a 3x3x2 (72 atoms) supercell of the crystal in VESTA

  • Use the crystal parameters and fractional atomic coordinates to prepare a Quantum Espresso input file

  • Perform a “vc-relax” relaxation on the doped crystal lattice

  • Perform an “scf” and “nscf” calculation on the crystal. (System energy and wavefunctions)

  • Perform a DoS computation of the crystal structure using the “dos.x” calculation from the Quantum Espresso suite

Results and Data:

When comparing the crystal lattices of Wurtzite and Zincblende (Figure 0.0), we can find the lattice energy per atom (to standardise the results as Wurtzite has 4 atoms in its primitive lattice, and Zincblende has 2 atoms).

Relaxed Energy per atom of Wurtzite: -1.0798 keV

Relaxed Energy per atom of Zincblende: -1.0796 keV

This gives us a delta of -0.1512 eV per atom (rounded to 4 sig.figs.) in favour of Wurtzite. While this may not appear significant, as this is a per atom difference, this corresponds to -9.105e16 MeV for a mole of ZnO (81.38g/mol.). As Wutrzite has significantly lower lattice energy, it is a more stable morphology. Thus Wurtzite will be used for the doping DoS computational experiments.

  • Pure Wurtzite

    Total Lattice Energy of ZnO: -77.7420 keV

    Ef = 7.251eV

    The DoS can be seen graphed as Fig. 2.0 and clearly shows the band-gap between the conduction (upper) and valence (lower) band, with the Ef lying just above the valence band. This bandgap corresponds to the ~3.4eV found in the literature for ZnO. Using previous literature that measures the absorbance/transmittance of ZnO, we can see that this ~3.4eV bandgap excludes most wavelengths in the visible spectrum (with little absorption of the violet end of the spectrum, giving ZnO crystals a slightly yellowed appearance)

  • Zn-site Dopants

    Total Lattice Energy of ZnO-Ga: -79.8009 keV

    Ef (Ga-doped) = 8.629 eV

    Total Lattice Energy of ZnO-Al: -76.1142 keV

    Ef (Al-doped) = 9.187 eV

    The resulting DoS graphs can be seen compared to pure Wurtzite in Fig. 2.1, showing how the change in Ef and the bending downwards of the conduction band leads these doped variants to exhibit n-type semiconductor properties.

  • O-site Dopants

    Total Lattice Energy of ZnO-F: -77.9674 keV

    Ef (F-doped) = 8.630 eV

    Total Lattice Energy of ZnO-Cl: -77.7768 keV

    Ef (Cl-doped) = 9.479 eV

    The resulting DoS graphs can be seen compared to pure Wurtzite in Fig. 2.2, while these dopants substitute on the O-sites, being halogens (Group VII), they cause band bending of the conduction band downward, resulting in an n-type semiconductor.

Discussion:

 

Now that we have the DoS of each doped variant of Wurtzite and the DoS of pure Wurtzite, we can see the band bending caused by the dopants on ZnO. While the dopants act on different sites, all four increase the Fermi Energy while bending the conduction band downwards, thus turning the neutral ZnO into an n-type semiconductor.

Doping Effects:

  • Firstly, looking at the Zn-site dopants, both Al and Ga increase the fermi energy of the lattice, and as can be seen from the DoS distributions, shift the Ef to be within the conduction band, therefore increasing the probability of electrons existing in the conduction band, increasing conductivity.

    However, Aluminium appears to cause a lower amount of band bending as the conduction band lies roughly equal to that of pure ZnO (starting around 9-9.5eV), but due to the more significant change in the Ef in Al-doped ZnO (7.251eV to 9.187eV), this still places the Ef within the conduction band. On the other hand, Gallium only shifts the Ef upward slightly (7.251eV to 8.629eV) while much more severely bending the conduction band downward (now starting around 8-8.5eV), thus placing the Ef within the new conduction band. This difference is likely due to the relative electron densities

  • Next, looking at the O-site dopants, again, both the F-doped and the Cl-doped crystals act similarly to the Zn-site dopants in that they both increase the Fermi Energy, placing it within the bounds of the conduction band, and thus turning the Wurtzite into an n-type semiconductor.

    Similarly to their Zn-site counterparts, F and Cl act in different ways - with Fluorine causing a minor Ef change (7.251eV to bending effect on the conduction band, bringing it down to ~8-8.5eV, whereas the Chlorine causes a more considerable Ef increase (7.251eV to 9.479eV) while appearing to bend the conduction band upward slightly over that of pure Wurtzite. This still results in the conduction band still lying around the Ef for Cl-doped ZnO, thus causing increased conduction band electron density, increasing conduction.

Supply Considerations:

 

Aluminium is widely used in manufacturing due to its low cost and low mass (not to mention theoretically infinite recyclability) and can currently be found at ~$3.14/kg11. (As of the writing of this paper, the prices of most materials appear to be increased likely due to pandemic induced supply reduction). Gallium, however, is a much more niche element, while still being used in electronics (LEDs, Optical discs, etc.), with potential as a future III-V semiconductor in the form of GaN. This places it as a much more expensive alternative as a dopant at ~$350.00/kg12, over 100 times as expensive as Aluminium.

This makes Aluminium the best choice for a Zn-site dopant for increasing conductivity.

Due to its source being sea salt or brine, Chlorine unsurprisingly comes in at ~$0.85/g15, inexpensive and abundant. Fluorine, mainly due to its reactivity, is not usually used directly in its gaseous form, being transported as a fluorite with group-I or group-II metals. CaF2 (or Fluorspar16) is a fluorite that can be used in synthesis. This reagent can be found for ~$5.29/g, making the cost of Fluorine by mass ratio to be ~$1.70/g.

Unlike the Zn-site dopants, the O-site dopants have a much smaller price difference. Thus external factors, such as the ease of use or specific manufacturing processes, would have to determine the appropriate O-site dopant for a given use case.

Chlorine would likely make a more sustainable dopant than Fluorine due to its abundance.

Acknowledgements:

 

This paper was made possible by Prof. Kesong Yang and Prof. Sheng Xu and their generous support and guidance in troubleshooting and setting up the experimental procedure. I would also like to thank the San Diego Supercomputing Center for allowing access to their Expanse supercomputing resources, without which none of the simulations would have been possible. I would also like to thank the University of California San Diego and the Jacobs School of Engineering - Nanoenginnering Dept. for their support in enabling me to complete this research.

A special thanks to the software developers for the open-source code of Quantum Espresso and VESTA to allow for these DFT simulations.

A special thanks to the Materials Project for their expansive, open-source materials library that has been well-curated for ease of access and accuracy to material-specific spectroscopic data and crystallographic information.