Sanghyuk
Lee
a,
Seungwon
Shim
a,
Hyunwoo
Jang
a,
Jae Kyeong
Jeong
b and
Youngho
Kang
*a
aDepartment of Materials Science and Engineering, Incheon National University, Incheon 22012, Korea. E-mail: youngho84@inu.ac.kr
bDepartment of Electronic Engineering, Hanyang University, Seoul 04763, Korea
First published on 2nd April 2025
Zn-doped In2O3 (IZO) has been extensively studied as a transparent conducting oxide (TCO) due to its favorable optical and electrical characteristics. In this work, to uncover the origin of degenerate n-type doping in IZO, we investigated point defects using density functional theory (DFT) calculations. Among the two possible configurations of Zn dopants, namely interstitial (Zni) and substitutional Zn(ZnIn), ZnIn is found to be energetically more favorable. While ZnIn acts as an acceptor, potentially compensating for n-type doping, it readily forms a defect complex, ZnIn–VO, by combining with oxygen vacancies (VOs), the dominant intrinsic defects in In2O3. This defect complex exhibits a substantial binding energy of approximately 1 eV and functions as a shallow donor. By evaluating carrier concentrations that can occur in IZO films, we demonstrate that the formation of ZnIn–VO is critical to maintaining or even enhancing significant n-type conductivities of IZO. By elucidating the doping behavior of IZO, this work provides critical insights to optimize its properties, thereby helping the advancement of optoelectronic and energy devices where IZO serves as a vital TCO.
Zn-doped IO (IZO) is another IO-based TCO that has been extensively studied for its advantages in carrier transport and optical transparency.12–15 Qiu et al. reported that IZO films deposited by radio frequency magnetron sputtering display fairly high carrier mobilities of ∼50 cm2 V−1 s−1, which is larger than the value for ITO prepared by the same technique.12 In addition, IZO films have higher work functions than ITO films, which can improve the performance of solar cells.12,16 The significant optical transparency is another merit of IZO films; their transmittance for the visible range is higher than 80%, which is comparable to ITO.14 In particular, IZO has relatively low Urbach tails near the conduction bottom, giving rise to low free-carrier absorption.12
Despite its promising properties, a complete understanding of the fundamental aspects of IZO remains elusive. In particular, the carrier-generation mechanism in IZO, which is essential for optimizing its electrical properties, is still not fully understood. Zn has a formal charge of 2+ in oxides, while In has a charge of 3+. When Zn substitutes for In sites in IO, this doping is likely to produce acceptors that counteract inherent n-type doping of IO films. However, IZO with Zn concentrations exceeding 1020 cm−3 consistently maintain significant free-electron concentrations, even surpassing 1020 cm−3.17,18 More intriguingly, Zn doping sometimes enhances the n-type conductivity.15,17,19 Additional generation of oxygen vacancies due to the charge difference between Zn2+ and In3+ has been suggested as a possible origin for the significant n-type conductivity of IZO in literature.15,20 In a previous work, point defects and n-type doping in IZO were theoretically investigated.21 However, the study reported low carrier concentrations of ∼1010 cm−3 under equilibrium conditions. To date, the detailed source of free carriers in IZO and the role of Zn dopants have not yet been identified. In addition, the impact of Zn doping on the atomic and electronic structures remains unexplored.
In this work, we investigate point defects in IZO to elucidate the impact of Zn doping on its electrical properties using density functional theory (DFT) calculations. By comparing the formation energies of Zn interstitials (Zni) and substitutional defects replacing In sites (ZnIn), we reveal that Zn preferentially forms ZnIn over Zni. As expected, ZnIn acts as an acceptor, which could potentially compensate for n-type doping of IO. However, we find that ZnIn readily forms defect complexes ZnIn–VO with oxygen vacancies (VOs), the dominant intrinsic defects in IO. These ZnIn–VO complexes exhibit a significant binding energy of approximately 1 eV and function as shallow donors. By explicitly calculating the carrier concentration considering experimental growth conditions of IZO films, we demonstrate that these defect complexes play a critical role in maintaining or even enhancing the n-type conductivity of IZO.
The formation energy of a defect with a charge (Dq) was obtained by using the following formula:31
![]() | (1) |
![]() | ||
Fig. 1 (a) Crystal structure of bixbyite In2O3 and local motifs of In and O. Local motifs for (b) In(8b), (c) In(24d), and (d) O(48e). |
Fig. 2(a)–(c) shows the atomic structures of VO, Zni, and ZnIn for their respective charge states, which are stable under n-type conditions with the EF positioned close to the conduction band minimum (CBM). An oxygen vacancy is an intrinsic donor in IO. Upon an V2+O is created, the surrounding In cations relax outward by 10% because of the lack of In–O bonds and the resulting repulsive interactions between the cations. We observe that the In neighbors of a VO defect does not relax inward significantly, even in the neutral and 1+ charge states, which contain electrons in the VO-induced defect state. Specifically, the relaxation amounts to −0.64% (inward) for V0O and 5.20% (outward) for V+O, relative to their equilibrium positions. This relaxation pattern implies the shallow nature of VO donors in IO. Namely, the electrons occupying the defect state are fairly delocalized in V0O or V+O, causing ineffective screening of the repulsion between the cations. This sharply contrasts with the relaxation of neighboring cations around deep VO defects observed in other oxides. For example, in ZnO, where VO is known to be a deep donor, the Zn ions surrounding VO relax inward by 12% in the neutral state.37 On the other hand, we find that the presence of V2+O leads to only minor modifications to the band structure, without introducing deep levels within the band gap (Fig. 2(d)). The slight change in the conduction band arises from the mixing between the host and V2+O-induced states. Nonetheless, the conduction band near the edge still maintains a large dispersion similar to that of pristine IO (green dashed line in Fig. 2(d)), supporting high carrier mobilities of n-type IO films.
When Zn is incorporated into IO, it can form either Zni or ZnIn. For Zni, there are two interstitial sites in the bixbyite structure: 8a and 16c. The Znis on these two sites have almost the same formation energies, with the difference being less than 0.1 meV, resulting in similar structural and electrical properties. In the following discussion, we focus on the Zni at a 16c site. Zni can act as a single or double donor, and Fig. 2(b) depicts the atomic structure of Zn2+i, where the Zn ion is surrounded by six O ions. Because of the large size of a Zn cation, Zni causes significant distortion in the surrounding lattice. In particular, it repels nearby In ions, causing a considerable contraction of In–O bond lengths from 2.19 to 2.14 Å. We notice that Zn2+i does not develop a defect level inside the band gap, as shown in Fig. 2(e). Instead, the defect state, which relates to the empty Zn 4s level, resides somewhere above the conduction band. Accordingly, the dispersion near the CBM remains largely unchanged compared to that of pristine IO.
On the other hand, Zn can substitute In at either the 8b or 24d site. Similar to Zni, the atomic sites for ZnIn do not lead to significant changes in the formation energies (with a difference of approximately 0.13 eV) and in the structural and electrical properties of ZnIn. We focus on the more stable ZnIn at the 24d site below. In contrast to Zni, ZnIn can serve as a single acceptor. Fig. 2(c) illustrates the atomic configuration of Zn−In, where the Zn ion is six-fold coordinated with O ions. Due to the smaller ionic radius of Zn compared to In, the substitutional Zn attracts nearby O ions, giving rise to Zn–O bond lengths (∼2.18 Å) shorter than the equilibrium In–O bond lengths (∼2.19 Å) along with a slight change in bond angles. For Zn−In, the defect state that accepts an extra electron lies in the vicinity of the VBM, as shown in Fig. 2(f). This defect state primarily arises from the hybridization between Zn and adjacent O states, exhibiting an antibonding character (partial density of states and charge density analysis are provided in Fig. S1 in the ESI†). As a result, when the transition from Zn0In to Zn−In occurs by accepting an excess electron, compensating for n-type doping, the Zn–O bond lengths somewhat increase from 2.14 to 2.18 Å on average. Similar to Zn2+i, Zn−In has little impact on the dispersion of the conduction band near the CBM.
We examine the formation energies of the isolated defects discussed above, as shown in Fig. 3. For readability, we present the formation energies for the most stable charge states of each defect at a given Fermi level, which correspond to the slopes of the formation-energy curves. For VO, it is observed that V2+O is the most stable over a wide range of the Fermi level. The stability transition from V2+O to V+O occurs at 0.24 eV below the CBM, while the stable region for V0O does not appear within the band gap. This is in good agreement with previous calculations.38 Consistent with the atomic structures, the higher stability of V2+O and V+O relative to V0O at the Fermi level close to the CBM suggests that VO is a shallow donor, which can donate free electrons to the conduction band at room temperature once it is created. Notably, VO displays formation energies less than 1 eV under oxygen-poor limit at EF ∼ CBM. Given that IO films are usually grown under oxygen-deficient environments, oxygen vacancies are therefore expected to be the dominant defects, leading to n-type conductivities.
![]() | ||
Fig. 3 Defect formation energies as a function of Fermi energy at the (a) O-rich and (b) O-poor limits. |
Zni is stabilized in the 2+ charge state at the EF within the band gap. Therefore, it can act as an effective double donor capable of producing two free electrons when formed. However, its formation energy is fairly high (>2.5 eV) at EF ∼ CBM, regardless of the growth condition, which is attributed to the high energy cost of the significant lattice distortion, as explained in the foregoing discussion. As a result, its concentration should be marginal in n-type IO films. On the other hand, ZnIn remains stable in the neutral state up to EF ∼ 1 eV, while its negative charge state becomes more energetically favorable at higher EF. As such, ZnIn can play a role as a compensator against n-type doping by accepting an excess electron from the host. The formation energy of ZnIn is quite close to that of VO under oxygen-poor limit at EF ∼ CBM and it can be smaller depending on growth conditions. This result seems to indicate that Zn doping deteriorates the n-type conductivity of IO films due to the formation of ZnIn acceptors, which contradicts experimental observations.
On the other hand, in extremely doped semiconductors with dopant concentrations exceeding 1019 cm−3, point defects may form defect complexes to decrease the enthalpy of the system despite the reduction of configurational entropy.39,40 In IZO, ZnIn can form a defect complex, ZnIn–VO, by combining with VO, a dominant intrinsic defect, as shown in Fig. 4(a). To assess the feasibility of defect-complex formation, we examine the binding energy (Eb) using the following equation:
Eb = Ef(ZnIn) + Ef(VO) − Ef(ZnIn–VO). | (2) |
![]() | ||
Fig. 4 (a) Atomic structure, (b) binding energy, and (c) band structure for (ZnIn–VO)+. In (c), the valence band maximum of pristine In2O3 is set to 0. |
The most stable charge state for each defect at a given Fermi level is considered when calculating the binding energy in eqn (2). As a result, Eb depends on the EF position. Notably, Fig. 4(b) shows that ZnIn–VO has large positive binding energies over 1 eV under n-type conditions, indicating that formation of the defect complex is highly favorable. The significant binding energy originates from the electrostatic attraction between V+O and Zn−In. (We examined other sites for VO in the supercell and confirmed that the configuration discussed above is the most stable.) Furthermore, the band structure of ZnIn–VO (Fig. 4(c)) shows that the defect complex is a shallow donor, which does not produce a deep level to trap electrons. As such, it is likely to be a possible n-type source that can explain the puzzling doping behavior of Zn doped In2O3.
To firmly confirm the role of ZnIn–VO, we explicitly evaluate the carrier (ne) and defect [Dq] concentrations considering experimental synthesis conditions of IO films. We calculate ne as
![]() | (3) |
![]() | (4) |
[V+O] + 2[V2+O] + [(ZnIn–VO)+] → [Zn−In] + ne. | (5) |
In eqn (5), we neglect the concentrations of hole and other defects such as Zn+i, which are expected to be negligible. Typically, transparent conductors based on undoped IO films exhibit carrier concentrations in the range of 1019–1020 cm−3 in experiments.8–10 To reflect this observation, we consider chemical potentials of In and O that yield a calculated carrier concentration of approximately 2.0 × 1019cm−3 for undoped IO. These chemical potentials fall outside the oxygen-poor limit previously discussed in Fig. 3, suggesting that the synthesized IO films exist in a nonequilibrium state, potentially leading to the occurrence of secondary phases. It is worth noting that degenerate semiconductor thin films, such as TCOs and P-doped Si used as source and drain materials in transistors,42,43 are usually produced using nonequilibrium growth techniques. Nevertheless, such films remain kinetically stable, enabling reliable device operation.
Fig. 5 shows the calculated ne and [Dq] as a function of the total Zn concentration (i.e., [Zn−In] + [(ZnIn–VO)+]). It should be noted that, between Zn−In and (ZnIn–VO)+, the latter is the dominant form; most of the Zn doped into IO forms the defect complex, while the concentration of isolated Zn−In is insignificant. This is attributed to the considerable binding energy of (ZnIn–VO)+, as mentioned above. At low Zn doping levels (1016–1018 cm−3), the predominant point defect is V+O, and the free-electron concentration ne closely matches [V+O]. However, as the Zn doping level increases beyond ∼1019 cm−3, the concentrations of isolated oxygen vacancies ([V+O] and [V2+O]) begin to decline, while the concentration of (ZnIn–VO)+ continues to increase, becoming the dominant defect. As a result, (ZnIn–VO)+ serves as a major donor that produces free carriers, enabling higher carrier concentrations than that of undoped IO. Our findings clearly demonstrate that Zn doping is not detrimental to n-type doping of IO films due to the favorable formation of the defect complex, which allows for the maintenance or even enhancement of carrier concentrations.
![]() | ||
Fig. 5 Calculated concentrations of V+O, V2+O, Zn−In, (ZnIn–VO)+, and free electrons as a function of Zn doping level. |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5cp00408j |
This journal is © the Owner Societies 2025 |