Computational Analysis of Metal Contact on Bi2O2Se with Se Surface Vacancies

In this work, a theoretical study on the electronic properties of the metal/Bi2O2Se interface is presented through the density functional theory calculation. Particularly, the effects of Cr, Pd, Pt, Au, and Bi on monolayer, bilayer, and trilayer Bi2O2Se are explored. Naturally created Se vacancies on the Bi2O2Se surface are also considered by constructing two interface structures: the metal/Bi2O2Se with the Se vacancies remaining or filled with metals. For the metal/monolayer Bi2O2Se, it is observed that the Bi2O2Se layer is fully metallized since Bi2O2Se strongly interacts with the metal. Meanwhile, for the metal/bilayer and trilayer Bi2O2Se, the Bi2O2Se layers remain semiconducting except for the layer right next to metal. Among the considered metals, regardless of whether the Se vacancies are filled with metal or not, the semiconducting layers in bilayer and trilayer Bi2O2Se form Ohmic contact with Bi. It is also found that filling the Se vacancies with the metals heavier than Se increases the interface distance between metal and Bi2O2Se, and hence results in the weak Fermi level pinning effect.


Computational Analysis of Metal Contact on Bi 2 O 2 Se with Se Surface Vacancies
Sukhyeong Youn and Jiwon Chang* DOI: 10.1002/aelm.202201221 mobility 2D material, recently, Bi 2 O 2 Se has emerged as a promising high mobility 2D material without such problems. Bi 2 O 2 Se exhibits a sizable band gap of 0.8 eV. [8] High mobility of ≈28 900 cm 2 V −1 •s at 1.9 K and ≈450 cm 2 V −1 •s at room temperature, respectively, has been successfully demonstrated in Bi 2 O 2 Se FETs with a good subthreshold swing of 65 mV dec −1 . [8] Bi 2 O 2 Se remains stable after exposure to air more than a few months, which guarantees the practical applicability of Bi 2 O 2 Se. [8] Moreover, it has been confirmed that single-layer Bi 2 O 2 Se flakes can be synthesized to cover a large area. [8][9][10] Density functional theory (DFT) calculations have also predicted excellent performances of monolayer and bilayer Bi 2 O 2 Se FETs in the ultra-scaled regime. [11,12] The application of Bi 2 O 2 Se in FETs essentially involves the formation of source/drain metal contacts on Bi 2 O 2 Se. In general, most 2D materials exhibit the Schottky contact at the interface with metals due to the strong Fermi level pinning, [13,14] and the ultrathin nature of 2D materials makes the conventional doping techniques inapplicable to 2D materials. So, a high contact resistance is the critical issue in the realization of high performance 2D material FETs. Therefore, understanding the electronic properties of the metal/Bi 2 O 2 Se interface is necessary to fully utilize the potential of high carrier mobility in Bi 2 O 2 Se. Several theoretical calculations using DFT have been carried out to investigate the band alignment of monolayer [15] and bilayer [16] Bi 2 O 2 Se with the Fermi level of various metals. In refs. [15] and [16] , the surfaces of monolayer and bilayer Bi 2 O 2 Se are fully Se-terminated without vacancies, and all Se atoms are passivated by hydrogen. However, it has been experimentally observed that the Se-terminated surface of Bi 2 O 2 Se has the 50% Se vacancies, which is supported by DFT calculations showing that the structure with 50% Se vacancies is the most energetically stable. [17,18] Therefore, Se vacancies on the surface of Bi 2 O 2 Se need to be considered as an essential structural feature. So, investigating the effect of Se vacancies on the interfacial properties is more realistic and important.
In this work, we consider various metals (Cr, Pd, Pt, Au, Bi) on monolayer, bilayer, and trilayer Bi 2 O 2 Se to explore the electronic properties at the interface of the metal/Bi 2 O 2 Se using the DFT calculations. The Se-terminated surface of Bi 2 O 2 Se with 50% Se vacancies is considered in two different cases where the Se vacancies are filled with metal or not. Through the band structure, projected density of states (PDOS) and local density In this work, a theoretical study on the electronic properties of the metal/ Bi 2 O 2 Se interface is presented through the density functional theory calculation. Particularly, the effects of Cr, Pd, Pt, Au, and Bi on monolayer, bilayer, and trilayer Bi 2 O 2 Se are explored. Naturally created Se vacancies on the Bi 2 O 2 Se surface are also considered by constructing two interface structures: the metal/Bi 2 O 2 Se with the Se vacancies remaining or filled with metals. For the metal/monolayer Bi 2 O 2 Se, it is observed that the Bi 2 O 2 Se layer is fully metallized since Bi 2 O 2 Se strongly interacts with the metal. Meanwhile, for the metal/bilayer and trilayer Bi 2 O 2 Se, the Bi 2 O 2 Se layers remain semiconducting except for the layer right next to metal. Among the considered metals, regardless of whether the Se vacancies are filled with metal or not, the semiconducting layers in bilayer and trilayer Bi 2 O 2 Se form Ohmic contact with Bi. It is also found that filling the Se vacancies with the metals heavier than Se increases the interface distance between metal and Bi 2 O 2 Se, and hence results in the weak Fermi level pinning effect.

Introduction
In the past few years, 2D materials have been in the spotlight for the excellent electrical properties. Especially, in the application of the field-effect transistors (FETs), 2D materials have attracted a lot of attention due to the ultra-thin nature providing an excellent immunity to the short channel effect. For the high-performance FETs, however, only a few 2D materials, including graphene and black phosphorus, have been reported to exhibit high carrier mobility. [1][2][3][4] But, graphene basically has a zero band gap, [5] which leads to the high leakage current in FETs, and black phosphorus suffers from the poor stability in ambient conditions. [6,7] In the continuous search for the high www.advelectronicmat.de of states (LDOS), we examine the band alignment of monolayer, bilayer and trilayer Bi 2 O 2 Se with metals to check the contact types (Schottky or Ohmic). The effect of Se vacancies on the band alignment and the Fermi level pinning is also discussed.

Determination of Exchange-Correlation
Bi 2 O 2 Se has a body-centered tetragonal structure with the I4/mmm space group as illustrated in Figure 1a. Alternating Bi 2 O 2 and Se layers compose the layered structure of Bi 2 O 2 Se. The geometry optimization of bulk Bi 2 O 2 Se is first carried out, and then the band structure is calculated for the optimized structure in Figure 1a. Regarding the choice of exchange-correlation functional, the hybrid functionals yield more accurate band gap value than the standard functionals such as the local density approximation (LDA) [19] and the generalized gradient approximation. [20] However, the supercell structure which we eventually investigate in this work is too large to use the hybrid functionals. Therefore, we carefully choose the standard exchange-correlation functionals and the pseudopotential to reproduce the experimentally measured band gap and the previously reported band structures of bulk Bi 2 O 2 Se. [8] It is found that the LDA [19] with the SG15 [21,22] and PseudoDojo [23] pseudopotentials for oxygen and the other atoms respectively, results in the band gap value of 0.72 eV, close to the experimental value of 0.8 eV, [8] as shown in Figure 1c. We do not consider the effect of spin-orbit coupling (SOC) since ref. [24] reports that SOC primarily affects the band gap size, not the overall shape of band structure. The optimized lattice constants of bulk Bi 2 O 2 Se are a = 3.86 Å and c = 12.12 Å which are also very similar to the experimentally measured values of a = 3.88 Å and c = 12.16 Å, respectively. [17] In the thin film structure of Bi 2 O 2 Se, it has been observed that the surface is terminated with Se. [17] Since Se is shared by the adjacent Bi 2 O 2 layers in the layered Bi 2 O 2 Se structure in Figure 1a, it is likely that the Se-terminated surface has a large amount of Se vacancies. Ref. [18] has proposed a surface model of 50% Se vacancies and experimentally confirmed the existence of 50% Se vacancies on the surface of the as-grown Bi 2 O 2 Se thin film through the high-resolution transmission electron microscopy. Moreover, DFT calculation has reported that the surface with 50% Se vacancies is the most energetically stable with the lowest binding energy. [18] Therefore, we need to consider Se vacancies on the surface of the Bi 2 O 2 Se thin film as an essential feature in the structure rather than typical defects in the other 2D materials like S vacancies in MoS2. Therefore, we adopt 50% Se vacancies in the construction of the Bi 2 O 2 Se thin film as seen in Figure 1b. To introduce 50% Se vacancies, the in-plane (xy-plane) unit cell size for monolayer, bilayer and trilayer Bi 2 O 2 Se is four times larger than for bulk. Then, we consider Se vacancies arranged in a row rather than diagonally since it is more observed and stable. [18] It is confirmed that in-plane lattice constants of the geometrically optimized monolayer, bilayer, and trilayer Bi 2 O 2 Se are close to the experimentally measured value of few layers Bi 2 O 2 Se. [25] From the band structures in Figure 1c, the band gap values of monolayer, bilayer, and trilayer Bi 2 O 2 Se are 1.82, 0.86, and 0.83 eV, respectively, which are consistent with refs. [9] and [18] .

Building of Heterostructures
To investigate the metal contact properties on Bi 2 O 2 Se, we consider Cr, Pd, Pt, Au, and Bi. The work function values of  [26] are distributed in a wide range. Therefore, from the simple estimation of the Fermi level position according to the metal work function and the Bi 2 O 2 Se electron affinity, [27] it is expected that the Fermi level can be located at different positions in the Bi 2 O 2 Se band gap for each different metal. Additionally, all selected metals can be simulated with the same exchange-correlation potentials used for Bi 2 O 2 Se and the supercell structure of metal/Bi 2 O 2 Se heterostructure can be constructed with minimal lattice mismatch and low computational cost.
We examine the effect of metal contact on monolayer, bilayer, and trilayer Bi 2 O 2 Se, respectively. For each metal/ Bi 2 O 2 Se heterostructure, two interface structures are explored. As discussed previously, there are 50% Se vacancies on the surface of the Bi 2 O 2 Se thin film. Thus, we consider two cases where the Se vacancies are filled with or not filled with metal. The former one, named 50% metals, corresponds to the typical metal deposition using the evaporation method while the latter, named 50% vacancies, to the direct transfer of metal on the top of Bi 2 O 2 Se. [28]

Contact Evaluation
To investigate the effect of metal on Bi 2 O 2 Se, band structure, PDOS, and LDOS calculations are carried out for each metal/ Bi 2 O 2 Se heterostructure. Figure 2 shows the optimized crystal structure, band structure, PDOS, and LDOS of Au/monolayer Bi 2 O 2 Se with 50% metals. From Figure 2a, the crystal structure of monolayer Bi 2 O 2 Se is significantly affected by Au. In Figure 2b, we plot the band structures projected on Bi 2 O 2 Se in Au/monolayer Bi 2 O 2 Se to check if the band structure of pristine monolayer Bi 2 O 2 Se is preserved. As same as for the crystal structure, the band structure is severely distorted due to the strong hybridization of Au with monolayer Bi 2 O 2 Se. Therefore, the band gap disappears and monolayer Bi 2 O 2 Se is metallized as also confirmed from PDOS in Figure 2c. LDOS with the crystal structure in Figure 2d shows a non-negligible amount of density of states (DOS) in monolayer Bi 2 O 2 Se, indicating the full metallization of Bi 2 O 2 Se. For all other cases of metal/monolayer Bi 2 O 2 Se, regardless of whether the Se vacancies are filled with metal or not, we observe a significant modification of the crystal structure and the band structure, which eventually leads to the metallization of monolayer Bi 2 O 2 Se as in Figures S1-S10, Supporting Information.
For the metal/bilayer Bi 2 O 2 Se, we show the optimized crystal structure, band structure, PDOS, and LDOS of Pd/bilayer Bi 2 O 2 Se with 50% vacancies in Figure 3. From the optimized crystal structure in Figure 3a, Layer 1, the layer right next to Pd, exhibits the deformation especially around the Se vacancies while the other layer, Layer 2, almost maintains the original structure. Compared with the metal/monolayer Bi 2 O 2 Se, the modification of Layer 1 is relatively less since Layer 2 atomically bonded with Layer 1 suppresses the substantial distortion of Layer 1 in bilayer Bi 2 O 2 Se. However, as shown in the band structures projected on Layer 1 in Figure 3b, we still observe a significant hybridization of Pd with Layer 1, leading to the metal-induced gap states in PDOS for Layer 1 in Figure 3d. On the other hands, overall band structure of Layer 2 is preserved as shown in the band structures projected on Layer 2 in Figure 3c. PDOS for Layer 2 in Figure 3d also confirms our observation. The energy range where DOS is very close to zero clearly indicates the existence of the band gap in Layer 2. Additionally, from LDOS in Figure 3e, there are considerable amount of DOS in Layer 1 over all energy range, but a certain energy range devoid of DOS in Layer 2. Therefore, in Pd/bilayer Bi 2 O 2 Se with 50% vacancies, Layer 1 is metallized while Layer 2 still remains semiconducting. In all other cases of metal/bilayer Bi 2 O 2 Se, the semiconducting property of Bi 2 O 2 Se is maintained only in Layer 2 in Figures S11-S20, Supporting Information. Then, we can determine whether the interface between metal and bilayer Bi 2 O 2 Se forms Schottky or Ohmic contact through investigating the amount of available states around the Fermi level as in the previous study. [29] Figure 3d,e shows the band gap alignment of Layer 2 with respect to the Fermi level, indicating the conduction band (CB) and valence band (VB) locations. If the Schottky contact is formed, the Schottky barrier (SB) height Φ SB,n for electron can also be estimated by identifying the CB minimum (CBM) and VB maximum of Layer 2. Figure 4a,b shows the band gap alignment for the metal/bilayer Bi 2 O 2 Se with 50% vacancies and with 50% metals, respectively. With 50% vacancies in Figure 4a, Au and Bi result in Ohmic contact for electron while Cr, Pd, and Pt lead to Schottky contact. Even if Pd, Pt and Au have high work functions, the Fermi level is pinned near CBM or above CBM. Among the low work www.advelectronicmat.de function metals, only Bi forms Ohmic contact, whereas for Cr, the Fermi level is located around the mid gap. Thus, the Fermi level location has little correlation with the work function value. Meanwhile, in the case of 50% metals in Figure 4b, Ohmic contact is achieved only with the low work function metal Bi. All the high work function metals Pd, Pt, and Au and the low work function metal Cr led to Schottky contact. Similar results are observed in the experiments reporting Schottky contact of evaporated Pd, Ti, [30] and Au [31] on Bi 2 O 2 Se, which corresponds to the case of 50% metal. Compared to the case of 50% vacancies, for Pd, Pt, Au, and Bi with 50% metals, the Fermi level position appears to be affected by the work function, indicating the weak Fermi level pinning, while the Fermi level is still pinned around the mid gap for Cr. To understand the difference in the Fermi level location between 50% vacancies and 50% metals, we investigate the average distance d interface between metal and Bi 2 O 2 Se at the interface. As plotted in Figure 4c, the distance increases for Pd, Pt, Au, and Bi while decreases for Cr with 50% metals. Since the atomic numbers of Pd, Pt, Au, and Bi are larger than Se, filling up the Se vacancies with those metals pushes metals away from Bi 2 O 2 Se. On the other hand, the distance is slightly reduced if the Se vacancies are filled with Cr which has a smaller atomic number than Se. Since the Fermi level pinning effect becomes weaker as the distance at the inter-  www.advelectronicmat.de face increases, [32] the relative positions of the Fermi level for Pd, Pt, Au, and Bi becomes more consistent with the work function values as the Se vacancies are filled with metal as seen in Figure 4b.
For the metal/trilayer Bi 2 O 2 Se, the optimized crystal structure, band structure, PDOS, and LDOS of Pt/trilayer Bi 2 O 2 Se with 50% vacancies are presented in Figure 5. Similar to the metal/bilayer Bi 2 O 2 Se, the crystal structure of Layer 1 in Figure 5a is distorted due to the interaction with Pt, which induces states and DOS within the band gap as observed in Figure 5b,e, respectively. However, the structures of Layer 2 and Layer 3 remain almost the same, thereby maintaining the semiconducting property in Layer 2 and Layer 3 in Figure 5c-e. Figure 5f also confirms that the effect of Pt metalizes only Layer 1. Following the same procedure used in the metal/ bilayer Bi 2 O 2 Se, we examine the band gap alignments for 50% vacancies and 50% metals as plotted in Figure 6a,b, respectively. The average distance at the interface is also calculated in Figure 6c. What is observed for the metal/trilayer Bi 2 O 2 Se in Figure 6 is generally very similar to the metal/bilayer Bi 2 O 2 Se in Figure 4. With 50% vacancies, Ohmic contact is obtained with Au and Bi while Schottky contact with the other metals. With 50% metals, we can achieve Ohmic contact only with Bi. As same as in the metal/bilayer Bi 2 O 2 Se, the interface distance increases if the Se vacancies are filled with metal which has a larger atomic number than Se, resulting in the weaker Fermi  level pinning effect. Therefore, with 50% metals, the relative locations of the Fermi level for Pd, Pt, Au, and Bi are roughly in the order of the work function values. Since the interface properties for the metal/trilayer Bi 2 O 2 Se and the metal/bilayer Bi 2 O 2 Se are very similar, we can conclude that the interaction of Layer 1 with a metal is critical in determining the interfacial property.

LDOS in
Additionally, we check the charge transfer and the dipole formation at the interface between metal and bilayer Bi 2 O 2 Se. The electron density difference between the self-consistent valence electron density and the superposition of the atomic valence electron density.is calculated. We only consider bilayer Bi 2 O 2 Se since monolayer Bi 2 O 2 Se is fully metallized and trilayer Bi 2 O 2 Se shows similar interfacial properties with bilayer Bi 2 O 2 Se. Figures S36 and S37, Supporting Information, are the electron different densities for 50% vacancies and 50% metals, respectively. Regardless of whether the Se vacancies are filled with metal or not, electrons are transferred from Pd, Pt, Au, and Bi to Bi 2 O 2 Se while Cr gains electrons from Bi 2 O 2 Se. We may expect that the low (high) work function metals can transfer (extract) electron to (from) Bi 2 O 2 Se. However, this expectation does not hold as in the other study [33] on monolayer MoS 2 . The low work function metal Bi transfers electron to Bi 2 O 2 Se resulting in Ohmic contact. On the contrary, another low work function metal Cr extracts electron from Bi 2 O 2 Se, and hence forms the high SB height even with the low work function as shown in Figure 4a

Conclusion
In summary, we explore the electrical properties of various metal contacts on monolayer, bilayer, and trilayer Bi 2 O 2 Se, respectively, using DFT calculations. Two cases are studied in which the 50% Se vacancies on the Bi 2 O 2 Se surface are either empty (50% vacancies) or filled with metal (50% metals). For all cases, Layer 1 is metallized due to the strong hybridization between Bi 2 O 2 Se and the metal. Therefore, in the metal/monolayer Bi 2 O 2 Se, the band structure of monolayer Bi 2 O 2 Se is substantially changed and the semiconducting property disappears. Meanwhile, in the metal/bilayer and trilayer Bi 2 O 2 Se, the Bi 2 O 2 Se layers other than Layer 1 still remain semiconducting. These semiconducting Bi 2 O 2 Se layers form Ohmic contact with Au and Bi for 50% vacancies while Ohmic contact only with Bi for 50% metals. It is observed that if the Se vacancies are filled with the metals heavier than Se, the interface distance increases, which reduces the Fermi level pinning effect.

Experimental Section
DFT calculations were performed with QuantumATK [34][35][36] based on the linear combination of atomic orbitals. The LDA [19] was selected for the exchange-correlation functional. The SG15 pseudopotential [21,22] was used for oxygen and the PseudoDojo pseudopotential [23] for the other atoms. The van der Waals correction was not included in this calculation. In Ref. [18], DFT calculations revealed that the binding energy of Bi 2 O 2 Se was much higher than those of the typical 2D materials such as MoS 2 , WS 2 , and WSe 2 . Thus, the effect of van der Waals correction was negligible in Bi 2 O 2 Se. In all cases, the density mesh cut-off was set to 75 Hartree. The k-point grid to sample the Brillouin zone was 13 × 13 × 5 for bulk while 7 × 7 × 1 for monolayer, bilayer, and trilayer Bi 2 O 2 Se. All the structures were geometrically optimized until the force and stress on each atom became <0.01 eV Å −1 and 0.1 GPa, respectively. In monolayer, bilayer, trilayer Bi 2 O 2 Se, and the metal/Bi 2 O 2 Se heterostructures, since the periodic boundary condition was imposed, a vacuum layer of ≈3 nm was added in the z-direction to remove the interaction between the adjacent supercells. In the construction of metal/Bi 2 O 2 Se heterostructure, since the electronic properties of Bi 2 O 2 Se were mainly focused on, the in-plane lattice constants of Bi 2 O 2 Se were fixed at the original values while those of metal were adjusted. It was confirmed that the effect of strain on the band structure of metal was negligible. To minimize the strain in metal, the 1 × 1 Bi 2 O 2 Se was matched with the 2 × 2 Au (100), √8 × √8 Cr (100), 2 × 2 Pd (100), and 2 × 2 Pt (100), respectively. The 2 × 2 Bi 2 O 2 Se was matched with the √13 × √13 Bi (111). In-plane supercells of matched metal and Bi 2 O 2 Se in each metal/ Bi 2 O 2 Se heterostructure are shown in Figures S31-S35, Supporting Information. The minimization of the lattice mismatch was focused on rather than rigorously considering the thermodynamic stability which could be addressed using molecular dynamics simulations. A k-point grid of 5 × 5 × 1 was adopted for Bi while 7 × 7 × 1 for all other metals. During the geometry optimization, all atoms in Bi 2 O 2 Se and three atomic layers of metal close to Bi 2 O 2 Se were allowed to move as seen in Figures 2a, 3a, and 5a.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.