Rectifying Schottky Contact in ZrN/Polycrystalline p-Ge

Fermi-level pinning (FLP) at the metal/Ge interface makes it difficult to control the Schottky barrier height, which forces an ohmic behavior on p-Ge and a rectifying behavior on n-Ge. This study first demonstrates the rectifying behavior on polycrystalline (poly) p-Ge on a glass substrate, using sputter-deposited ZrN contacts under 350 °C process. The rectifying characteristics depend on the poly-Ge quality derived from the grain size and the defect-induced acceptor density. The highest quality Ge is formed by solid-phase crystallization (SPC) of amorphous Ge deposited at 125 °C and exhibits the lowest reverse leakage current as well as the highest hole (the lowest electron) barrier comparable to a single-crystal Ge substrate. In contrast, the Zr contact exhibits ohmic behavior, suggesting the importance of ZrN in alleviating FLP. The forward and reverse currents are determined by Ge resistance and grain boundaries in Ge, respectively. These technologies will further aid the development of electronic devices on glass or plastic substrates with low heat resistance, including n-channel thin-film transistors.


I. INTRODUCTION
Low-Temperature processed complementary metal-oxidesemiconductor (CMOS) devices are crucial for the application of high-performance 3-dimensional large-scale integrated circuits and high-functionality mobile terminals.Ge is a promising material for these applications, with lower crystallization temperature, higher carrier mobility for both electron and hole, lower grain boundary (GB) barrier than Si, and compatibility with conventional Si processing.
In this study, we first demonstrate rectifying ZrN/poly p-Ge SBDs using our SPC and contact formation technique.We then clarify the correlation between the SBD and poly-Ge properties.These results open the possibility of Schottky barrier control in poly-Ge and provide useful findings for all poly-Ge-based electronic devices, including n-channel TFTs.

II. SAMPLE PREPARATION
We fabricated lateral SBDs on poly p-Ge formed by SPC with the process and structure shown in Fig. 1.The a-Ge layers (100 nm thick) on SiO 2 glass substrates were deposited using the Knudsen cell of a molecular beam deposition system.The deposition temperature T d-Ge ranged from 50 to 200 • C. The samples were annealed in N 2 at 450 • C for 5 h to induce SPC.The SPC-Ge layers were thinned to approximately 55 nm by a chemical mechanical polishing (CMP) system, consisting of a rotating polishing pad and a sample carrier with the addition of a slurry, assuming the fully-depleted devices [19].The photoresist was patterned into an open circular window (radius r 1 = 25-200 μm), and reactive ion etching (RIE) (16 nm depth) was performed for efficient current injection by embedding a Schottky electrode.After plasma sacrificial oxidation and oxide removal, ZrN (50 nm thick) and TiN cap (15 nm thick) were sequentially deposited using a rf magnetron sputtering system at room temperature.During the deposition, sputtering power density and Ar/N 2 gas flow rate were kept 11 W/cm 2 and 30/0.1 sccm, respectively.Both ZrN and TiN targets had a stoichiometry composition (a metal/nitrogen ratio of 1:1) and a purity of 99.9%.TiN/ZrN was patterned by the photoresist removal and postmetallization annealing was performed at 350 • C. Two concentric Al (100 nm thick) electrodes were simultaneously prepared as an inner contact on TiN/ZrN and an outer ohmic contact on Ge, using a thermal evaporation system and lift-off photolithography.The distance between inner (ZrN) and outer (ohmic Al) electrodes d = 10 μm.Finally, contact annealing was performed at 300 • C, which is lower than the postmetallization annealing temperature; therefore, the effect on the ZrN/poly-Ge contact is considered negligible.Here, we note that SPC-Ge surfaces after CMP and RIE were smooth (root mean square value ≤ 1.3 nm).For comparison, the SBDs were also fabricated on single-crystal p-Ge (100) substrates.The Ge surfaces were not passivated.

III. RESULTS AND DISCUSSION
Table 1 summarizes the thinned SPC-Ge properties.Electron backscatter diffraction and Hall effect measurements with the Van der Pauw method have been used to determine the grain size (GS) and electrical properties.It can be observed that the T d-Ge = 125 • C sample shows a larger GS, higher μ, and lower p than the other samples.Conversely, the resistivity ρ at T d-Ge = 200 • C is lower in comparison to the other samples, reflecting the balance between μ and p.More details of T d-Ge -dependent SPC-Ge are discussed elsewhere in the literature [15].
The current-voltage (I-V) characteristics of ZrN/poly-Ge SBDs exhibit clear rectifying behavior for all T d-Ge [Fig.2(a)].The forward current at V < 0 maximizes at T d-Ge = 200 • C. The series resistance of poly-Ge R Ge in the present device structure is expressed using the following equation: where t is the Ge thickness, r 1 is the inner electrode radius (ZrN), and r 2 is the outer electrode radius (ohmic Al).
The R Ge is proportional to ρ under the common device dimensions (t, r 1 , and r 2 ) and limits the forward current of ZrN/poly-Ge SBDs.In contrast, the reverse current at V > 0 V is minimized at T d-Ge = 125 • C. Reflecting the forward and reverse current, we obtained the best on/off ratio of 30 in ±2 V bias range at T d-Ge = 125 • C.This rectifying behavior suggests that the hole barrier height of ZrN/poly-Ge is naturally high.The effective hole barrier height BP,eff and the ideality factor n of ZrN/poly-Ge SBDs was estimated from the forward I-V characteristics using following equations: where I s is the saturation current, q is the elementary charge, k is the Boltzmann constant, T is the absolute temperature, A is the contact area, and A * is the is the Richardson constant of 40.8A cm −2 K -2 for p-Ge [41].We estimated BP,eff and n in the low voltage region (0.01-0.05 V) to avoid the influence of R Ge .BP,eff and n depend on T d-Ge and T d-Ge = 125 • C exhibits the highest performance ( BP,eff = eV and n 2.1) at T d-Ge 125 • [Fig.2(b)].This high BP,eff likely reflect poly-Ge quality, i.e., larger GS (lower GB density) and fewer defect-induced p.
The effective electron barrier BN,eff has been calculated as BN,eff + BP,eff = E g , where E g is the energy band gap of Ge (0.66 eV) [14] and is found to be as small as 0.13 eV at T d-Ge = 125 • C. The estimated BP,eff and BN,eff are comparable to those on the single-crystal Ge substrates [39].On the other hand, we should consider the dominant current conduction mechanism because the evaluated ns are high (>2).Our former study about Schottky contact on single crystalline Ge with the same carrier density as the present study (∼6×10 17 cm -3 ) showed a low n of 1.1 [42].It means the dominant conduction mechanism is thermionic emission.However, the present study (poly-Ge) shows a high n (2-2.2) even though they have a similar carrier concentration.We think the reason for the large n is grain boundary-related leakage current.For further electrical characterization of poly-Ge, we should estimate Schottky barrier heights from capacitance-voltage (C-V) measurement and compare the I-V results.However, it is difficult in the present structure because of the high series resistance of the  poly-Ge layer mentioned above.In the next study, we will increase Ge thickness and/or improve the device structure from lateral to vertical one [43] for successful C-V measurement and in detail discussion to clarify the grain boundary properties.
Identically, when we calculate electron or hole barrier heights using ( 2) and (3), we should carefully consider the effective current flow area, particularly lateral sample structure like this study.We confirmed that the geometrical electrode area can be used for barrier height calculation for our samples.Its details are described in the Appendix section.
To investigate the effect of the contact metal on the rectifying behavior in poly-Ge SBDs, we studied Zr as a non-nitride metal using the same process and structure as for ZrN [Fig.1].Figs.3(a) and 3(b) show that the Zr contact exhibits ohmic behavior, whereas the ZrN contact exhibits rectifying behavior.This suggests a strong FLP at the Zr/poly-Ge interface near the valence band edge, while it is alleviated toward the conduction band side at the ZrN/poly-Ge interface.These results are consistent with a previous study comparing Ti and TiN contacts on a single-crystal Ge substrate and further support the role of N-containing a-IL in alleviating FLP [38].The higher forward-current in the Zr contacts is possibly because the interfacial reaction in the Zr/Ge structure, i.e., Zr x Ge y germanide formation [44]  and the lateral diffusion of Zr through the GB in Ge [45], provides lower resistance, whereas ZrN is a stable material.
To understand the current behavior in ZrN/poly-Ge SBDs, we prepared the electrodes with different r 1 (25-150 μm) on the same substrate.Fig. 4(a) shows that both forward and reverse currents increase with increasing r 1 .The electrode perimeter 2π r 1 normalizes these currents [Fig.4(b)].This phenomenon is discussed separately for the forward and reverse currents.The R given by limits the forward current, as in Fig. 2(a).Since the distance between and outer electrodes is constant in this study (r 2 − 1 = d (10 and the ρ are constant on the same substrate, (1) is expressed of r 1 substituting r 2 = r 1 + d into (1).Using a Maclaurin expansion up to the primary term, (1) is approximated as Therefore, we can conclude that the forward current is proportional to r 1 is normalized by the electrode perimeter 2π r 1 .Similarly, in the ZrN/single-crystal p-Ge SBDs, the electrode perimeter normalizes the reverse current, indicating that the surface-state generation current dominates the reverse leakage current [37].We broke down the total reverse current into its perimeter and area components.The reverse current I R at a certain reverse bias V R is generally given by where A is the contact area, P is the contact perimeter, J A (A/cm 2 ) and J P (A/cm) are the areal and peripheral current densities.Thus, J A and J P can be estimated from the intercept and slope of the I R /A-P/A plot shown in Fig. 5(a), respectively.Fig. 5(b) shows that the J P s for the all T d-Ge s of poly-Ge is significantly higher than that for single-crystal Ge.This relationship is consistent with the reverse J P shown in Fig. 4(b).These results suggest that the GB-generation current unique to poly-Ge dominates the reverse current in addition to the surface state.Furthermore, both J P and J A of T d-Ge = 50 and 200 • C are larger than that of T d-Ge = 125 • C and bulk-Ge.These likely suggest larger J P and J A due to the small GS (high GB density) and the high p (high defect density), respectively.These results are consistent with the GS and p (Table 1).For a more accurate estimation of Schottky barrier height and a successful application of the Schottky contact with TFTs, the leakage current and n should be reduced.These can be reduced by surface passivation with an insulating layer such as SiO 2 /GeO 2 or Y 2 O 3 [14], [37] and GS enlargement (GB reduction across the electrode) using Sn incorporation [46], [47], [48] and GeO x underlayer insertion [16].

IV. CONCLUSION
We first demonstrated a rectifying behavior on poly p-Ge using the FLP-alleviating ZrN contact.This characteristic depends on the poly-Ge quality (GS and defect-induced p).In contrast, the Zr contact exhibits ohmic behavior, suggesting the importance of ZrN in alleviating FLP.In the ZrN/poly-Ge SBDs, R Ge limits the forward current, and the GB-generation current dominates the reverse leakage current.These results will help to understand the physics of FLP in metal/poly-Ge interfaces and expand the range of poly-Ge-based electronic device applications, including n-channel TFTs.

APPENDIX
We used (2) and (3) to estimate the BP,eff shown in Fig. 2(b) from the forward I-V characteristics.These equations are valid when the current flows through the entire area of the Schottky junction.To clarify the current flow area in the ZrN/poly-Ge junction, we estimated the transfer length L T , which indicates how long the current spreads from the electrode edge, based on a circular transmission line model (CTLM) [49].The total resistance R T of the CTLM structure, shown in Fig. 6

FIGURE 1 .
FIGURE 1. Fabrication process flow and schematic of the lateral poly-Ge SBDs.

FIGURE
FIGURE I-V characteristics of poly-Ge SBDs using Zr and ZrN contacts with (a) linear and (b) logarithm scales in I, where T d-Ge = 125 • C and r 1 = 75 μm.

FIGURE 4 .
FIGURE 4. Electrode radius dependence of ZrN/poly-Ge SBD properties where T d-Ge = 125 • C. (a) I-V characteristics and (b) Peripheral current density J P .

FIGURE 5 .
FIGURE 5. (a) Areal current density I R /A versus electrode perimeter/area ratio P/A obtained from ZrN/poly-Ge and ZrN/single-crystal Ge SBDs with different areas and (b) and peripheral components of current (J A and J P ), where reverse bias V R = 0.3 V.
(a) used in this study, is expressed as follows: R T = R sh 2π r 1 (d + 2L T )C, where R sh is the sheet resistance of Ge and C is the correction factor.R T /C-d plots are generally used to determine L T and R sh with different d and constant r 1 .Conversely, our structure shown in Fig. 6(a) has constant d and different r 1 .Therefore, we draw R T /C-1/r 1 plots.From the slope of these plots shown in Fig. 6(b), we can obtain the L T using R sh , where R sh is obtained by the 4-probe van der Pauw method.L T is estimated to be 387 μm for T d-Ge = 50 • C, 464 μm for T d-Ge = 125 • C, and 184 μm for T d-Ge = 200 • C.These values are larger than the electrode radius r 1 = 75 μm, indicating that the current flows through almost the whole poly-Ge/ZrN junction area.Therefore, we conclude that the BP,eff estimated from forward I-V characteristics, shown in Fig. 2(b), is reasonable.