Applied Physics Express 4 (2011) 115701 DOI: 10.1143/APEX.4.115701
Schottky Barrier Inhomogeneities in Nickel Silicide Transrotational Contacts Alessandra Alberti, Fabrizio Roccaforte, Sebania Libertino, Corrado Bongiorno, and Antonino La Magna CNR-IMM, Zona Industriale Strada VIII N 5, 95121 Catania, Italy Received August 23, 2011; accepted September 28, 2011; published online October 27, 2011 Ni-silicide/silicon Schottky contacts have been realised by promoting low-temperature Ni–Si interdiffusion during deposition (50 C) and reaction (450 C) on an oxygen-free [001] silicon surface. A 14 nm transrotational NiSi layer was produced made of extremely flat pseudo-epitaxial domains (200 nm in diameter). The current–voltage (I–V ) characteristics (340–80 K) have indicated the presence of structural inhomogeneities which lower the Schottky barrier by 0:1 eV. They have been associated with the core regions of the trans-domains (wherein the silicide lattice is epitaxially aligned to that of Si) since their density (2:5 109 cm 2 ) and dimension (10 nm) fit the I–V curves vs temperature following the Tung’s approach. # 2011 The Japan Society of Applied Physics
n the pioneering studies focusing on the early stages of Ni–Si reaction it has been observed that, on a perfectly cleaned Si surface, a Ni–Si precursor layer is formed at the surface after room temperature deposition of a very thin nickel layer (2 nm).1) The precursor layer is limited to a few monolayers of nickel and originates from Ni atoms diffusing into the Si lattice.2) It has a disordered structure with Ni atoms having the short range environment similar to that observed in the NiSi2 lattice.1,3) This similarity promotes the transformation of the precursor layer to NiSi2 .1) The precursor layer is formed once nickel atoms come into contact with silicon, also if the deposited nickel layer exceeds the critical values: its presence is adduced as the reason why all the Ni-silicide phases have comparable Schottky barrier heights.4) When a supply of Ni is present on top of the precursor layer, the role of the diffusion layer, is usually found to be shadowed by the fast diffusion process of the excess Ni atoms, and a conventional phase sequence is observed with NiSi2 formed after the Ni-rich phases.1,5) However, we have recently shown6) that the thickness of the precursor layer can be further increased to 2 nm and the direct transition to NiSi2 guaranteed even in presence of a Ni supply (7 nm), provided the reaction process is sufficiently slowed down by using low-temperature annealing (220–350 C). We have additionally shown7) that the composition of the precursor layer can be also modified by slightly increasing the deposition temperature. A 6 nm precursor layer (amorphous), with a Ni:Si concentration ratio of 3, can be formed on an oxygen-free Si substrate by depositing nickel atoms at 50 C.8) From the amorphous mixed layer first transrotational Ni2 Si9) and, subsequently, transrotational NiSi domains originate10,11) during lowtemperature annealing (e.g., 260 C). The electrical behavior of the metal/Si interfaces having small structural inhomogeneities was carefully modelled by Tung.12) In the model, the conduction path of structural inhomogeneities of size comparable to the depletion region, having a low Schottky barrier (SB), can be pinched-off by the potential field of the surrounding ones having higher SB. Based on the findings of the Tung’s model, much effort has been dedicated to study several silicide/Si interfaces for standard phases such as poly-Ni2 Si, NiSi,13) and epy-NiSi2 .14) It was well assessed that the structure of the interface, at the atomic level,4,13–15) can play a crucial role in the Schottky barrier height (SBH) determination.
I
E-mail address:
[email protected]
In this work, we analyse the electrical behavior of Ni–Si transrotational layers and we correlate it to the structure of the pseudo-epitaxial domains. A 7 m epitaxial n-doped (1 1016 P/cm3 ) Si layer was deposited on a [001] nþ Si substrate by chemical vapour deposition. On top, an 800-nm-thick oxide layer was deposited and used to open, by a photolithographic procedure, circular diodes of 1 mm diameter. A back ohmic contact was realised on the nþ Si substrate. After a HF cleaning procedure, the diodes were loaded in a sputtering chamber (5 109 Torr), wherein the Si surface was further cleaned by a soft sputter etch procedure in Ar ambient. On the [001] Si surface, a 7 nm nickel layer was deposited by sputtering at T 50 C (100 W). The phase transition from Ni to Nisilicides has been promoted by spike annealing at 450 C in a pure N environment. A reference poly-NiSi layer was produced leaving a thin oxide layer at the Si surface and by annealing the diodes at 260 C for 2 h + 450 C for 120 s. The characterization was carried out by polar figure X-ray diffraction (XRD) analyses using a synchrotron radiation source [ID01 European Synchrotron Radiation Facilities (ESRF)] and by transmission electron microscopy (TEM) analyses. The electrical characterization was performed using a Bio-Rad DL4600 equipment. The structure of the as deposited layer was investigated by TEM analyses, and the result is shown in Fig. 1(a). The layer is amorphous and has a uniform thickness of 10 nm, with a flat interface with Si. Energy-filtered TEM analyses (not shown) have shown that the interface between the silicide and Si is oxygen free, as effect of the double cleaning procedure. The absence of an oxide layer at the interface and the lowtemperature input given during deposition (50 C) have allowed a mixing process to occur during deposition, resulting in a layer wherein Si and Ni atoms are simultaneously present in the form of an amorphous layer [Fig. 1(a)]. The composition of that layer7) resulted in a ratio of Ni and Si atoms close to 3. After spike annealing at 450 C [Fig. 1(b)], the layer thickness increased to 14 nm, as expected for the formation of NiSi, still maintaining an extremely flat interface with Si. Pole figure analysis related to the (202)/(211) NiSi planes, collected using a synchrotron X-ray source (not shown), confirmed the stoichiometry of the as-reacted layer as that of NiSi.10) It was found that the layer has a transrotational structure.6) The structure of a transrotational domain, as shown in the TEM plan-view image of Fig. 2, substantially differs from that of a polycrystalline layer since it is characterised by dark elongated narrow crossing features wherein the silicide lattice is in Bragg condition with respect to the e-beam. The
115701-1
# 2011 The Japan Society of Applied Physics
Appl. Phys. Express 4 (2011) 115701
A. Alberti et al.
(a)
(a)
(b) Fig. 1. Cross-sectional TEM analysis of (a) the as-deposited nickel layer (Tdeposition 50 C) and (b) the as-reacted NiSi layer (T ¼ 450 C). The asdeposited layer is amorphous and contains a mixture of Ni and Si atoms. The as-reacted layer is flat and uniform.
(b)
Fig. 3. (inset) Current density vs applied voltage of a Schottky diode with a transrotational NiSi layer contact compared with a conventional diode (poly-NiSi); (a) I–V curves related to the transrotational contact as a function of temperature; (b) fitting curves using Tung’s model (for simplicity, only the experimental data of the lower and the higher temperatures are superimposed). At low-temperatures, the I–V curves show a double feature due to the presence of structural inhomogeneities having low SBHs.
Fig. 2. Plan-view TEM analyses of the as-reacted silicide; (inset) large area e-diffraction analysis. The transrotational NiSi domains are characterised by bending contours crossing one to another at an average distance of 200 nm. The bending of planes along a section perpendicular to a bending contour is also sketched.
image was acquired after an alignment procedure on the [001] zone axis of the Si substrate and therefore evidences that there is a kind of relationship between the silicide and the Si lattices. This relationship has been clearly identified by the e-diffraction analysis shown in the inset of Fig. 2, which was collected over a large diode area, larger than that shown in Fig. 2. In the diffraction, the (020) planes of the NiSi lattice result aligned to the (220) planes of Si: the direction of the two G vectors of the reciprocal lattice is the same. On the basis of what is known on those structures,6,9–11) the dark narrow and elongated regions of the domains are bending contours.16–19) Outside the bending contours the silicide planes bend6) in a way similar to that sketched in the inset of Fig. 2. The silicide lattice is represented (by a rectangle) progressively bending once leaving the core region [from position (1) to (2) to (3)], to stress that, at the atomic level, the interface progressively changes along the cross section. It must be emphasized that the internal bending of planes does not imply any roughening of the layer, since both surface and interface of the silicide
remain extremely flat [Fig. 1(b)]. The domain core is the central portion of the domain wherein the bending contours cross each another establishing a double epitaxial relationship with the substrate: within this portion of the interface a pseudo-epitaxial match between the two lattices is arranged. By large area plan-view TEM analyses and by pole figure X-ray investigations it was found that the domain core has a width of 7–15 nm, while the distance between them is of 100– 300 nm. It has been found, by large area analyses, that the entire diode area is covered by transrotational domains, while the reference silicide is a conventional polycrystalline layer (not shown). The current density passing through a 1 mm diode at room temperature is shown in the inset of Fig. 3(a). The reference NiSi silicide shows an almost standard behavior, with a saturation current density (Js ) [thermoeletronic emission (TE)], corresponding to an SBH of 0.66 eV, very close to that reported in the literature for standard Ni-silicides.4,13,20) The ideality factor is close to 1.1, indicating a certain degree of uniformity of the barrier. On the other hand, the diode made of the transrotational silicide is characterised by a large reverse current and also by a direct current higher than that in the reference. If a classical TE model was applied, an SBH of 0:5 eV would be obtained by forcing a meaningless ideality factor close to 3. I–V curves were collected by reducing the temperature during analyses from RT down to 80 K (Fig. 3). The fact that, at low T and low V (
