Amorphous TeO₂ as p-type oxide semiconductor for device applications Open Access

There is presently great interest in industrially relevant p-type amorphous oxide semiconductors. So far, various p-type oxides have been found, but each has its limitations. The first p-type oxide, Cu₂O, used a noble metal (Cu) with a s-shell lying near the O 2p level to raise the valence band maximum (VBM) energy above the O 2p level, making it a moderately shallow state.¹ However, the bandgap of Cu₂O was too small to be fully transparent. Cu₂O was superseded by CuAlO₂ where strong Al-O interactions widened the optical bandgap.² However, the layered structure of CuAlO₂ and similar compounds^2,3 did not favor disorder. SnO is a p-type oxide with a low effective hole mass and 3 eV direct bandgap.^4,5 However, its 0.7 eV indirect gap gives extremely high leakage current densities to be useful. The ZnRh₂O₄ spinel has a wide gap and is p-type and amorphous (a-).⁶ However, its cost and high processing temperature inhibit its use in back-end-of-line (BEOL) CMOS devices with, for example, n-type a-InGaZnO₄ (IGZO),⁷ which use low (<400 C) processing temperatures. There were various data-based searches identifying, say, Bi₂BaTa₂O₆ or SnTa₂O₆ as having low hole masses.^8,9 However, experiments showed that these oxides had a relatively low p-type conductivity, due to compensation by intrinsic defects.^10,11 Thus, there is still no fully satisfactory p-type oxide.

Bulk TeO₂ is a glassy oxide already known for its large nonlinear optical coefficient.^12–14 Guo et al.^15,16 identified TeO₂ as a possible p-type oxide from its calculated low effective hole mass and the high calculated mobility of its upper valence states. This is due to the isotropic nature of the now filled Te 5s states. Zavabeti et al.¹⁷ measured various experimental properties of crystalline β-TeO₂, such as a hole mobility of 146 cm²/V.s, an effective hole mass of 0.51, and the weakly p-type behavior of the undoped material. Shi et al.¹⁸ produced β-TeO₂ by pulsed laser and by sol-gel deposition, but again undoped. Extrinsically doped TeO₂ by, say, As has not yet been produced or a high conductivity measured.

We show here that a-TeO₂ would have a short range order similar to that of β-TeO₂ with minimal like-atom bonds. We then show that its VBM state would lie close to the approximate p-type doping limit^19–22 and that As-doped a-TeO₂ doped would be a good uncompensated p-type semiconductor with its Fermi energy (E_F) lying in the valence band. This occurs because TeO₂ is mostly covalently bonded and maintains its bonding when doped to a p-type doping limit energy, beyond that typical of ionic oxides, so the conduction is not compensated by intrinsic defects.

Here, the electronic structures of the rutile, α-, β-, γ-, and a-TeO₂ phases are calculated by density functional theory (DFT) using Heyd–Scuseria–Ernzerhof (HSE) hybrid functionals²³ to give good bandgaps. The HSE method is also used to correct the band edge energies [electron affinity (EA) and ionization potentials (IP)] against the vacuum level in Table I with supercells having a 15 Å vacuum spacing between TeO₂ slabs.

TeO₂ has a parent rutile phase with sixfold Te sites, Fig. 1(a), and three polytypes. The rutile structure has a relatively narrow gap.²⁴ The α and β polytypes of TeO₂ are distorted rutile orthorhombic phases with much wider gaps of ∼3.5 eV, with partly layered structures consisting of covalently bonded chains of fourfold coordinated Te sites and twofold coordinated O sites.^14,15 The γ phase is less stable by 0.06 eV per formula unit (fu). The calculated properties of these polytypes are summarized in Table I and Figs. 1 and 2. We also give the total energies per TeO₂ fu. for each polytype in Table I. These values are consistent with those of Deringer et al.²⁵ 

The HSE band structures of two crystalline phases of TeO₂ are shown in Fig. 2. They are similar to those of earlier work,²⁶ except for the HSE correction. Figure 2(a) shows that the parent rutile phase is a narrow gap semiconductor with a bulk indirect bandgap of 0.72 eV. Due to this narrow gap, the calculated electron affinity is quite high, 6.09 eV.

β-TeO₂ forms the stable layer-like structure shown in Fig. 1(c). Its band structure in Fig. 2(b) is calculated in a similar fashion. Its direct bandgap at Γ is 3.29 eV.^15,17 It has a Te s-like valence band maximum, which gives a high theoretical hole mobility. The ionization potential or VBM energy is 6.45 eV below the vacuum level. The β phase is stabilized from the rutile structure by 0.2 eV/fu.

The valence band maximum seen in the calculated partial density of states (PDOSs) has Te s-like character in Fig. 2(c), as found earlier.^{15,16,18,26–28} The highest valence band is Te s-like. Below this lies the predominantly O 2p-like states centered on −2.5 eV below the VBM. This band then extends down to −7.5 eV, followed by the O 2s band. The conduction band is largely Te p-like. It has a direct bandgap.

It is important to confirm that the TeO₂ structure can withstand disorder to be useful as an amorphous oxide. Experimentally, TeO₂ is known to be glass-forming like SiO₂ based on β-TeO₂ units.^12,13 On the other hand, many of the recently deposited metal oxides like, for example, ZnO have more ionic bonding to oppose homopolar bonding. A-TeO₂ is more covalent, so some homopolar bonds are a strong possibility. Thus, we subject a 96-atom supercell of crystalline β-TeO₂ to an ab initio molecular dynamics (AIMD) calculation, in which the supercell is heated to 2000 K for 4ps, then cooled to 300 K in 20 ps, and then relaxed at 300 K for 5 ns. The structure is shown in Fig. 3(a), and the calculated radial distribution function (RDF) and partial RDFs are shown in Fig. 3(b). The Te-O nearest neighbor bond distribution is centered on 1.96 Å, while the O-O second neighbor peak lies at around 2.8 Å. These results are quite similar to the earlier AIMD results of Pietrucci et al.²⁶ and also reasonably similar to experimental results.^29,30 We see there are very few direct O–O bonds at 1.8 Å or direct Te-Te bonds in the total RDF at 3.8 Å in this relatively small supercell. Thus, a-TeO₂ is very strongly chemically ordered, despite an absence of ionic bonding; the main disorder arises within the weaker Te-O secondary bonds.²⁴ TeO₂ is the only polytype of Te and O on the solid phase diagram and has a narrow composition range,³¹ consistent with this. The existence of glassy a-TeO₂ and that it can be made as a plasma-deposited (PLD) film¹⁸ suggest that it can be made by atomic layer deposition (ALD).³² The calculated PDOS of a-TeO₂ is seen in Fig. 3(c), the s-like upper valence bands being slightly mixed with O 2p states.

Zavabeti et al.¹⁷ have shown that undoped crystalline β-TeO₂ has a high p-type mobility but have not yet shown this in doped films, perhaps because of safety concerns with As doping.

It is still necessary to show that doped p-type films with high conductivity can be made without self-compensation of carriers by native defects. For this, the absolute band edge energies of the TeO₂ phases referred to the vacuum level are shown in Fig. 4(a) and should lie within the approximate doping limit guidelines. It shows that the band edges of β-TeO₂ are much higher than those of α-TeO₂ and are then more compatible with p-type doping. This arises because of the layer structure of β-TeO₂ where the dipoles of the lone pairs act to raise the electrostatic potential within the bonding units.

Self-compensation itself can be checked by calculating the formation energy of the main native defect, in this case the oxygen vacancy, as a function of E_F within the bandgap, and as a function of the O chemical potential μ_O, between the limits of O-rich TeO₂ and O-poor TeO₂, as shown in Fig. 4(b). If the formation energy drops to near 0 eV for E_F lying within the bandgap, then an O vacancy can form spontaneously, effective doping will be compensated,^19–21 and the overall mobility will be low. Figure 4(b) shows that in general β-TeO₂ will be a good host. However, self-compensation could occur in O-poor films for E_F at the VBM.

As the average calculation based on “doping limits” is based on defects in mainly polar oxides whereas TeO₂ is a much more covalent oxide, we will directly calculate the effect of substitutional p-type doping in amorphous TeO₂. We, therefore, test compensation by substituting two As_Te sites for Te sites into the a-TeO₂ lattice and disorder the network by additional AIMD calculation. Two atoms are used to avoid spin-polarization effects. Figure 5(a) shows the resulting network for this converged calculation. Figure 5(b) shows that E_F is found to lie inside the top of the valence band and that the As acceptor state is shallow in this HSE calculation. This acceptor level is not seen to be deep as are many p-type dopants in oxides like ZnO or Ga₂O₃.³³ We attribute this behavior to the direct covalent bonding of the As atoms lying at fourfold Te sites, and also to the fact that substituting for Te atoms rather than for O atoms does not lead to strong polaronic effects at the acceptor site as would occur on O sites in say ZnO. Interestingly, however, separately substituting two acceptor N atoms on O sites as N_O sites would remain quite shallow, but these sites disproportionate to having different coordination, to form N₂⁻ - N₃⁺ valence alternation pairs³⁴ as in Fig. 5(c). Here the subscript means coordination and the superscript means charge.

We have shown by calculation that amorphous TeO₂ is a dopable p-type oxide and that it is not self-compensated by intrinsic defects. Further, we have found that the valence band edge energy lies within the “doping limits” guidelines. We have also noted that substitutional As or Sb atoms lying at Te sites will produce degenerate conduction with the Fermi energy lying below the valence band maximum. This compound should be producible by low-cost atomic layer or sputtering deposition.

The authors have no conflicts to disclose.

John Robertson: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). Xuewei Zhang: Formal analysis (equal); Investigation (equal). Qingzhong Gui: Investigation (equal). Yuzheng Guo: Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.