Micro-structural characterization of low resistive metallic Ni germanide growth on annealing of Ni-Ge multilayer

Magnetism and transport behavior of ultrathin films and multilayers have been driven by their manifold applications in the field of microelectronics, as interconnects in spintronics and as magnetic sensors in information storage devices.^1–4 Among these magnetic hetero-structures, ferromagnetic metal-semiconductor (FM/SC) systems are an important class, which have motivated many experimental and theoretical studies in the field of magnetism and nanotechnology.^5–10 Structure and magnetic properties of a FM/SC system, strongly depend on the structure and morphology of the interfaces.^11–14 

Germanium is an important material for metal oxide semiconductor field effect transistor (MOSFET) applications because of its high intrinsic mobility (factor of 2 for electrons and 4 for holes) as compared to Si.^7,15 In connection with the current Si-based technology, where metal silicides are used as contacts for the source, drain, and gate of the transistors, metal germanides appear as a natural candidate for making contacts.⁵ Design and processing of transistors require that the contact material should exhibit low contact resistance, low processing temperature and be resistant to thermal degradation. Hence, developing optimal contact materials at relatively low temperature is of supreme importance.

Nickel germanides are suitable candidates for interconnects in microelectronics as they form low resistive phases (monogermanide: NiGe) on annealing at relatively low temperature (∼270°C) as compared to other transition metal germanides^5,16 and such low processing temperature is effective to prevent thermal degradation of the gate material in transistors.⁷ Thermal annealing can cause solid state reactions between Ni and Ge with short bilayer periods (small number of repetition of bilayers) to produce nickel germanides.^7,17 Formation of NiGe on annealing mainly depends on atomic number ratio of Ni to Ge,^18,19 which can be achieved by growing individual Ni and Ge layers with proper thickness ratio. Considering the number density (no. of atoms per unit volume) of Ni and Ge as 9.1 × 10²² cm⁻³ and 4.41 × 10²² cm⁻³, respectively, we should have a thickness ratio d(Ge)/d(Ni) of 2.1:1 in a multilayer stack for getting NiGe alloy with an atomic ratio of 1:1.^18,19 Based on this estimate, we had deposited Ni-Ge multilayers with four bilayers of Ni and Ge with designed thicknesses of ∼ 100Å and ∼ 200Å, respectively.

The Ni-Ge multilayers were grown by DC/RF sputtering technique²⁰ with a base vacuum ∼5 × 10⁻⁷ Torr. The working vacuum with flow of Ar gas was ∼ 4 × 10⁻³ Torr during deposition. The multilayer was grown on p-type Si (111) substrate. Designed structure of the multilayer sample consisting of 4 bilayers of Ni and Ge can be represented as: Si (substrate) /[Ni_100Å / Ge_200Å] × 4. After deposition, the sample was annealed at 250°C under vacuum (∼10⁻³ Torr) for time intervals of 0.5 h, 1.5 h and 4 h.

The as-grown and annealed samples (after each successive annealing) were characterized by several techniques viz. grazing incidence x-ray diffraction (GIXRD), x-ray reflectometry (XRR), polarized neutron reflectometry (PNR) and Atomic Force microscopy (AFM). X-Ray data were collected at a CuKα (λ = 1.54 Å) laboratory source. The PNR data were collected at DHRUVA reactor, India.²¹ PNR measurements were performed on saturating the in-plane magnetization of the samples in a magnetic field of ∼2 kOe at 300 K. XRR and PNR are non-destructive tools which provide density profile as a function of depth with nanometer resolutions.^21–25 In case of XRR and PNR the reflected radiation is measured from a sample as a function of wave vector transfer [Q = 4 π sin(θ)/λ] perpendicular to the sample surface where λ and θ are x-ray or neutron wavelength and angle of incidence respectively. The reflectivity pattern is quantitatively related to the Fourier transform of the scattering length density (SLD) depth profile φ(z) averaged over the sample area. For XRR, φ(z) is proportional to electron SLD and in case of PNR, φ(z) consists of nuclear and magnetic SLDs²³ such that $φ z = φ n ( z ) ± φ m ( z )$ depth profile.^23,24 The sign (±) denotes the condition, whether the neutron beam polarization is parallel or anti-parallel to the applied field (sample magnetization) and are represented by the reflectivity patterns r⁺ and r⁻ respectively. All the Reflectivity data presented here have been plotted as a function of Q/Q_c, where Q_c = 4πsinθ_c/λ and θ_c is critical angle of incidence (below which reflectivity is unity).

PNR and XRR data obtained from as-deposited and annealed samples were independently analyzed by fitting model φ(z) profiles that best fit the data. The reflectivity was calculated using the dynamical formalism of Parratt²⁵ and parameters of the model were adjusted to minimize the value of weighted measure of goodness of fit, χ².²⁶ Using XRR and PNR together we identified the interface alloy stoichiometry as reported earlier for the binary systems.^19,22

Surface morphology of the as deposited and annealed samples at 250°C for 4 h has been quantified in terms of fractal parameters^27,28 of height difference correlation function. The AFM data were collected on samples of 2 × 2 μm² size using a ‘Solver P47H’ microscope. NSG10_DLC super sharp DLC tip grown on silicon with curvature 1-3 nanometer has been utilized in semi-contact mode. In AFM operation cantilever’s resonant frequency & force constants were 213 kHz and 10 N/m respectively.

The resistance of the samples were measured using four-probe technique in the temperature range of 10 K to 300 K. Bulk magnetization measurements were carried with a SQUID magnetometer and the morphology of the film surface before and after annealing were obtained from AFM.

Variation in resistance as a function of temperature, R(T), over the temperature range, 10 K - 300 K of the as-deposited and annealed samples are shown in Fig. 1. These measurements were carried out on samples of same dimension (∼5 × 5 mm²). For annealed samples the resistance values were observed to be lower than that of the as-deposited sample over the whole temperature range (10 K - 300 K). All the measurements showed that the samples were metallic in nature with a nearly flat resistance profile in the range of 10 K to 300 K. Hence one may compare the average resistances of the samples as a function of annealing temperature. We obtained R ∼ 17Ω for the as-deposited Ni-Ge multilayer sample. On annealing the sample at 250°C for 0.5 to 1.5 h the resistance reduced to ∼10 Ω. Further drop in resistance (∼8Ω) was observed on annealing the sample for 4 h. The overall drop is nearly a factor of two in the resistance of the sample after the final anneal, with respect to the as-deposited sample.

Fig. 2(a) presents the GIXRD data of the as-deposited sample indicating the Ni layers in the as-deposited sample were crystalline [marked by blue triangles]. Absence of any Ge peaks shows that Ge was amorphous. Fig. 2(b) shows the GIXRD data for the sample annealed for 4 h, showing several possible alloys phase peaks [green circles in Fig. 2(b)]. But the alloy could not be confirmed uniquely from such a broad GIXRD peak profile. The composition of the alloy phase was confirmed using dual PNR and XRR measurements on the same sample using formalism developed by us²² as discussed later.

Fig.2 (c) shows SQUID hysteresis curves for the as-deposited sample (red circles) and sample annealed (black triangles) for the 4 h at 250 °C. The SQUID data shown in Fig. 2 (c) are without normalization to sample volume (different size of the samples) and confirms ferromagnetic behavior of the multilayer in as-deposited and annealed state. We observed reduction in saturation magnetization of the multilayer on annealing, which is in-line with PNR measurements and consistent with the growth of non-magnetic Ni germanides at interfaces as seen by reflectivity techniques and discussed later. However samples were saturated well below a magnetic field of 2 kOe (applied during PNR measurements for saturating the sample) in both the cases.

Fig. 3 (a) shows PNR measurements from the as-deposited Ni-Ge multilayer and multilayer annealed at 250°C for 0.5 h, 1.5 h and 4 h along with fits (solid lines). Close (red) and open (black) circles are measured reflectivity for neutrons having spins parallel (+) and anti-parallel (-), respectively, with respect to sample magnetization direction. Neutron reflectivity profiles for different stages of annealing (Fig. 3(a)) are offset for better visualization. Fig. 3(b) and 3(c) show the detailed nuclear and magnetic scattering length density profiles [NSLD and MSLD] respectively for as deposited and annealed stages as obtained from PNR analysis.

The XRR data for the as-deposited sample and for the sample annealed for 4 h along with the fits (solid lines) are shown in Figs. 4 (a). The corresponding electron scattering length density [ESLD] is shown in Fig. 4(b) and 4(c). We ensured that the physical model obtained from PNR and XRR are consistent with each other and they fit the corresponding data within the error bars over the entire ‘Q’ range. To achieve best fit to PNR and XRR data we have assumed small variation in thickness of individual layers and roughness interfaces of the multilayer around their mean values [Fig. 3(b), 3(c) and Fig. 4(b) and 4(c)]. Fig. 5(a), 5(b) and 5(c) show the average NSLD, ESLD and MSLD of a bilayer (repetition of 4 such bilayer is multilayer) over the total number of bilayers. The estimated errors in the fitted parameters of the multilayers are in the range of 5 % - 7 %.

We obtained an average thickness of 145 Å and 216 Å for Ni and Ge layers respectively in the as-deposited state. From the NSLD profile (Fig. 3(b)) obtained from PNR one can clearly identify the growth of the interface alloy layer as a function of annealing time. The PNR data shows that on annealing the multilayer for 4 h at 250°C, the average thickness of Ni and Ge reduces to ∼ 26 Å and ∼111 Å at the expense of an alloy layer of thickness ∼113 Å which grow at both Ni/Ge (Ni grown on Ge) and Ge/Ni (Ge grown on Ni) interfaces [Fig. 5(a), blue diamond]. In addition, we found a reduction in interface roughness from ∼25 Å (as-deposited) to 8 Å on annealing the sample. The NSLD of interface alloy layer was ∼5.8 × 10⁻⁶ Å⁻² after annealing the multilayer for 4 h at 250 ⁰C. Similar structural parameters for as-deposited and annealed samples were also obtained from XRR measurements [Fig. 5(b), blue diamond]. From XRR data we found an ESLD of ∼ 4.8 × 10⁻⁵ Å⁻² for alloy layer at interface after annealing the multilayer for 4 h [Fig. 5(b)].

By simultaneously fitting the r⁺ and r⁻ data, the φ(z) + and φ(z) − SLD’s are generated, and the NSLD profiles from PNR can be extracted from φ_n = (φ⁺ + φ⁻)/2. Thus for estimation of the stoichiometry of alloy layer we have compared φ_n(z) obtained from PNR and φ_x(z) obtained from XRR data with theoretical value of SLD.^19,22 We obtained the stoichiometry of the alloy layer formed at the interface to be NiGe (mono germanide), which corroborates the GIXRD data.

We obtained very small magnetic moment density (∼30% of the bulk Ni) for Ni layers in the as-deposited sample from PNR measurements. It is evident from Fig. 2 (c) that the magnetic (Ni) layer thickness and magnetic moment density have reduced on annealing the multilayer up to 4 h. Reduction in magnetic moment density followed the trend of the bulk magnetization measurements, obtained from SQUID. In addition PNR data also shows that the NiGe alloy layers at interfaces are magnetically dead (there is no long range ordered magnetization). It is worth noting that the as-deposited Ni layer had a density ∼ 80 % of bulk density. After annealing the film for 30 minutes the Bragg peaks in PNR shifted to higher ‘Q’ indicating marginal reduction in the thickness of Ni-Ge bilayer. This we attribute to defects being annealed out and densification of the Ni layers in the Ni-Ge multilayer along with formation of metallic NiGe.^29,30

Fig. 5 (d) presents the variation of the thickness of the alloy layer as a function of annealing time of the sample at fixed temperature (250 °C). Variation of average resistance with annealing time at room temperature is shown in Fig. 5 (e). Increase in alloy layer thickness due to increase in annealing time is clear. Initial decrease in average resistance was observed on annealing the Ni/Ge multilayer was possibly due to removal of defects and growth of metallic NiGe alloy layer at the interfaces.³¹ Fig. 5(d) and 5(e) show a systematic correlation between the growth of alloy layer and reduction of resistance of the sample on annealing at 250°C for different time and suggest that the growth of alloy layer reducing the resistance of sample.

Accurate estimation of the layer thicknesses of the components in the Ni-Ge multilayer allows us to make an estimate of the resistivity of the NiGe alloy layer using a parallel network model as shown in Figs. 6. Fig. 6 (left panel) is a schematic of alloy formation at interfaces of a single bilayer of Ni-Ge on annealing. We have considered the Ni-Ge system as a parallel combination of resistors considering each Ni and Ge layer as an individual resistor (four Ni and four Ge resistors for the as-deposited state as shown in Fig. 6 (right panel). Ni has a much smaller value of resistivity (ρ_Ni ∼ 10 μΩcm) as compared to Ge (ρ_Ge ∼ 10⁷ μΩcm) at room temperature.³² For estimating net resistivity of the system, we considered the resistance values of each layer of the parallel network model and finally obtained the resistivity from the geometry of the samples. For this calculation, contribution of Ge layers can be neglected as 1/ρ_Ni > > 1/ρ_Ge. Hence Ge layers were assumed to be open paths for the flow of current between the contacts in the as-deposited case as shown in Fig. 3 (c). If ‘R’ is the net resistance of the system and r_Ni is the resistance of the individual Ni layers, then for the as deposited state we estimate: 1/R ≈ 4/R_Ni, giving R_Ni ∼ 64 Ω. After annealing, thickness of the Ni layers reduced and alloy layers of NiGe developed at the interfaces providing extra transport path. The enlarged view for the post annealing stage for a bilayer is shown in Fig. 3(c) (left panel) for better visualization. We have scaled the resistances of Ni layers after annealing (⁠$R Ni ′$⁠) by a factor t₀/t to account for the increase in resistance due to reduction in thickness of the layer, where t₀ and t are the thicknesses of Ni layer before and after annealing respectively. The net resistance (R) of the system after annealing can be written as: $1 R ≈ 4 R Ni ′ + 7 R NiGe$⁠. Since there are now seven conducting NiGe layers at the interfaces, Using the above relation, resistance of an alloy NiGe layer, R_NiGe is estimated to be ∼53 Ω. This gives an estimated resistivity of NiGe alloy layer as ∼59 μΩ.cm.⁵ This value clearly indicates metallic nature of the interface alloy layer.

For good adhesion of surface contacts it is important to ensure that during annealing of the multilayer the surface quality does not degrade. To study the effect of annealing on morphology of the surface we measured the height difference correlation function (g(r) =   < [h(r) − h(0)]² >), where the angular brackets, <  >, denote ensemble average.²⁸ We used height data from AFM and fitted the height difference function with the function for the self-affine fractals:^28,32 $g ( r ) =2 σ 2 1 − exp − r / ξ 2 H$ Where ‘ξ’ is the correlation length and is a measure of the lateral length scale of roughness, ‘σ’ is uncorrelated roughness and ‘H’ is the Hurst parameter which defines the fractal dimensionality (d) of the surface as: d = 3-H. Fig. 7(a) and 7(b) show three dimensional AFM images of as-deposited and sample annealed for 4 h. Fig. 7(c) and 7(d) show the corresponding height profile along a line at the centre of the AFM image. Fig. 7(e) shows the height difference correlation function (closed circles and closed triangle) with corresponding fit (solid lines) for as deposited and annealed samples at 250°C for 4 h. We obtained similar morphological parameters (σ, H and ξ) for both as deposited and annealed state which confirms no morphological degradation due to annealing. Both the surface showed H = 0.8, suggesting nearly two-dimensional surface morphology.³² 

In summary we confirm the formation of a low resistive, non magnetic germanide phase of composition NiGe at the interfaces of a Ni-Ge multilayer on annealing at 250°C (low processing temperature), using XRR and PNR measurements. Detailed evolution of structural and magnetic properties of the Ni-Ge multilayer system as a function of annealing time at constant temperature 250°C, have been studied at nanometer length scale. We estimated resistivity of the alloy layer to be ∼59 μΩ.cm, suggesting metallic nature of the alloy. From AFM study it is observed that the sample does not undergo morphological degradation due to annealing. The low formation temperature of metallic alloy germanide (NiGe) with low resistivity makes it a potential candidate for contacts in microelectronics.

The authors acknowledge the help of Dr. Shovit Bhattacharya (Technical Physics Division, BARC, INDIA) for annealing the samples. The authors also acknowledge the help of Mr. Swpan Jana (Solid State Physics Division, BARC, INDIA) for his help during deposition of the Ni-Ge multilayer.