Electroless- and Electroplating of Cu(Re) Alloy Films for Self-Forming Ultrathin Re Diffusion Barrier

To inhibit the detrimental diffusion of copper into silicon devices, an effective barrier is strongly demanded. In this study, cop- per(rhenium) alloy ﬁlms were successfully electroless- and electroplated on silicon substrates for self-forming an ultrathin rhenium barrier layer. In pure copper ﬁlms, the interdiffusion of copper and silicon already occurred at 400 ◦ C, forming silicides and increasing the electrical resistivity of the ﬁlms. In comparison, not any silicide formation was observed in the electroless- and electroplated copper(rhenium) alloy ﬁlms with minor incorporations of rhenium for 0.10 and 0.39 at% until 600 and 500 ◦ C, respectively, attributed to the prior segregation of rhenium and the self-formation of rhenium barrier layer at the copper/silicon interface at 400 ◦ C. A zero solubility of rhenium in copper and a small solubility in palladium catalysts facilitate the separation of rhenium from copper. The kinetics of rhenium segregation follows uphill diffusion, and the diffusivity of rhenium in copper at 400 ◦ C is estimated to be 1.5 × 10 − 17 cm 2 /s.

To prevent the rapid diffusion of Cu in interconnect structures and the consequent failure of integrated circuits caused by silicide formations, 1 diffusion barriers such as practically used Ti, Ta and their nitrides with a low electrical resistivity, high stability and good interface adhesion are strongly demanded. [1][2][3] In recent years, more stable and diffusion-resistant barriers have further been intensively studied, including those of ternary components (e.g. Ru-Ti-N and Ru-Ta-N) [4][5][6] with an amorphous or distorted structure to diminish diffusion paths and those of layered structures (e.g. Ru/TaN and Ru/TaCN) [7][8][9] with large interface mismatches to elongate diffusion distances. Multi-component high-entropy materials and their stacking structures (e.g. (AlCrTaTiZr)N, AlCrRuTaTiZr, (AlCrTaTiZr)N x /AlCrTaTiZr, and multilayered (AlCrRuTaTiZr)N x ) [10][11][12][13] have also been proposed to achieve an extreme resistance to Cu/Si interdiffusion. However, for a practical application to next-generation interconnects below 20 nm, ultrathin (≤ 2.1 nm) robust barriers have rarely been reported.
Due to the difficulty in uniform depositions of thin barrier layers in nanoscale trenches, self-forming diffusion barriers (barrierless metallization) have been developed in the past few years. [14][15][16][17][18][19][20] The segregation of added minor elements (early Mg, 14 Ti, 15 Mn, 16,17 and recent Ag, 18 Ho, 19 Hf) 20 in sputtered Cu wires to Cu/dielectric interfaces under a thermal annealing will self-form an ultrathin barrier layer; the thickness of the self-forming barriers is determined by the content of the added elements and the temperature of annealing. In choosing appropriate alloying elements for forming barriers at low annealing temperatures, a low solubility and a high diffusivity in Cu (>self diffusivity of Cu) as well as a negative enough standard free energy of oxide formation or a large activity coefficient (>1), need to be considered. 21 However, the drawbacks of most developed barrierless Cu alloys caused by an improper selection of alloying elements yet include: (1) a deficient segregation (separation from Cu) of solutes due to a high solubility in or an easy reaction with Cu, (2) a low formation rate of barrier layers owing to a low diffusivity in Cu, (3) an early failure of interconnects or devices at high self-formation temperatures, and (4) a low resistance of self-forming barriers to Cu/Si interdiffusion.
Re with a high melting point, Young's modulus and elevatedtemperature strength has been recently considered for hightemperature applications. 22,23 Because of its strong bonding to resist an interdiffusion and low (almost zero) solubility in Cu for an easy phase separation, 24 Re might be a potential alloying element for barrierless metallization. A high temperature (750 • C) to endure a Cu/Si interdiffusion by an ReN x barrier layer which self-formed from a sputtered * Electrochemical Society Active Member. z E-mail: shouyi@dragon.nchu.edu.tw Cu(ReN x ) alloy film suggests the high thermal stability of Re. 25 For self-forming an ultrathin metallic Re diffusion barrier of high electrical conductivity by an electrochemical-compatible process as well as solving the problem of unconformable layer depositions in ultra-small trenches in a sputtering process, electroless-(EL) and electroplating (EP) of Cu(Re) alloy films were developed in this study. Thermal annealings were applied to the films for self-forming an Re barrier layer and for examining the interdiffusion resistance of the barrier. For comparison, EL and EP pure Cu films were examined as well.
Experimental P-type Si wafers (resistivity 7.5-12.5 -cm) as substrates for EL depositions and N-type high-conductivity Si wafers (1-5 × 10 −3 -cm) for EP were degreased and cleaned in acetone for 5 minutes and isopropyl alcohol for 1 minute under an ultrasonic vibration, and rinsed in deionized water. The cleaned substrates were then chemically etched to remove surface native oxides by using a 5% HF solution for 10 minutes. For EL depositions, the etched substrates were subsequently catalyzed in a sensitization solution composed of SnCl 2 ·2H 2 O (10 g/L) and HCl (40 mL/L) at 75 • C for 2 minutes, and in an activation solution of PdCl 2 (0.25 g/L) and HCl (2.5 mL/L) at 75 • C for 1 minute to deposit nanosized Pd catalysts. 26,27 A pure Cu or a Cu(Re) alloy film of about 150 nm thick was deposited on the catalyzed substrates in an EL Cu or EL Cu(Re) solution, as listed in Table I, at 50-60 • C or 70-80 • C, respectively. [26][27][28] Three mixed solutions with different EL Cu and EL Re composition ratios (EL Cu: EL Re = 10:1, 10:2 and 10:3) were attempted for the EL Cu(Re) depositions, and the thermal stability of the deposited alloy films under an annealing at 600 • C for 30 minutes in a vacuum of 10 −6 torr were preliminarily examined for selecting an appropriate EL Cu(Re) solution. For EP, a pure Cu or a Cu(Re) alloy film of about 200-250 nm thick was directly deposited on the etched substrates in an EP Cu or EP Cu(Re) solution, without a catalyzation, as also listed in Table I, at 25 • C under an applied current density of 0.75 A/cm 2 . For examining the thermal stabilities and interdiffusion-resistant abilities of the films, an annealing at 300-500 • C for 30 minutes was applied to the EL and EP pure Cu films, while an annealing at 400-700 • C for 30 minutes following a prior annealing at 400 • C for 30 minutes (for self-forming an Re barrier) was applied to the EL and EP Cu(Re) alloy films in a vacuum of 10 −6 torr.
Electron probe micro-analyses (EPMA, JEOL JXA-8800M; with a resolution of about 0.1 at%) were used to determine the elemental compositions of the as-deposited films, and scanning electron microscopy (SEM, JEOL JSM-6700F) was applied to observe the surface morphologies of the as-deposited and the thermally annealed films.  2 100 mg/L NaH 2 PO 2 10 g/L 2-ABT 50 mL/L K 4 Fe(CN) 6 57.3 mg/L EL Cu(Re) PEG #200 500 mg/L PEG #200 50 mg/L EL Cu + EL Re 10:1 PEG #2000 500 mg/L HCl 1 mL/L (EL Cu: EL Re) 10:2 EP Cu(Re) NaOH 5 g/L 10:3 EP Cu + HReO 4 12 g/L The crystal structures were analyzed by a glancing-incident-angle (1 • ) X-ray diffractometer (XRD, BRUKER D8 Discover), and the sheet resistance was measured by a four-point probe station (Keithley 2400) for electrical resistivity calculations. The depth profiles of elemental distributions and the bonding configurations of the 400 • C-annealed Cu(Re) alloy films were determined by using electron spectroscopy for chemical analysis (ESCA, ULVAC-PHI, PHI 1600; with a resolution of about 0.1 at% and 0.5 eV at the Ag 3d 5/2 peak).

Results and Discussion
Depositions and thermal stabilities of EL and EP pure Cu films.-For comparison, EL and EP pure Cu films were first deposited and examined. As shown in the surface morphologies in Figure 1a, typically, both the as-deposited EL and EP Cu films were continuous and smooth, composed of small grains with sizes of several tens nm.
After a 300 • C annealing, more densified films were formed, and, as plotted in Figure 1b, because of grain growth and defect elimination, the normalized sheet resistance of the films was slightly reduced (the calculated electrical resistivity of EL Cu decreased from 4.2 to 3.4 μ -cm; because an abnormally low resistance of EP Cu was measured due to the contribution of high-conductivity substrates, an normalized sheet resistance rather than an electrical resistivity was thus presented). As a result, in the XRD patterns shown in Figures 1c and 1d for EL and EP Cu, respectively, the broad diffraction peaks of the as-deposited films at 43.4 • , 50.6 • and 74.3 • (face-centered cubic, fcc; (111), (200) and (220) lattice planes, respectively) were more narrowed and sharpened after the 300 • C annealing. However, after a 400-500 • C annealing, the resistance began to rise (the resistivity of EL Cu increased to 3.8-7.5 μ -cm) due to the formation of some reaction products, as observed in Figure 1a, in consequence of Cu/Si interdiffusion. The XRD patterns confirmed the early failure of the  pure Cu films at 400 • C and verified the reaction products to be typical Cu silicides with the appearance of two small peaks at 44.7 • and 45.3 • (Cu 3 Si (320) and (312) lattice planes, respectively).
Depositions and thermal stabilities of EL and EP Cu(Re) alloy films.- Figure 2 and Table II present the preliminary evaluation results of EL Cu(Re) alloy films deposited in three mixed solutions with different EL Cu and EL Re composition ratios. From the surface morphologies and cross-sectional images of as-deposited alloy films in Figures 2a to 2d, a dense, smooth and continuous EL Cu(Re) 10:1 film with a grain size of 30-40 nm and a cluster size of 100-200 nm was observed. As measured by EPMA and listed in Table II, a very minor content of Re was successfully incorporated in the film, even for only 0.10 at% because of the higher reduction potential of Cu than Re (Cu +2 + 2e − → Cu 0 , E = +0.337 V; Re +3 + 3e − → Re 0 , E = +0.300 V). 29 Compared to an EL pure Cu film, the electrical resistivity of the EL Cu(Re) 10:1 alloy film obviously increased to 9.2 μ -cm, not only due to the Re addition but also some film discontinuity. With an increased EL Re ratio, although the Re content slightly increased to 0.12 at%, however, a more discontinuous and island-like EL Cu(Re) 10:2 film with a similar grain size but a much larger cluster size of 300-400 nm was deposited, owing to a more intense reaction and resulting in the further increase in resistivity to 12.3 μ -cm. With a higher EL Re ratio, a continuous EL Cu(Re) 10:3 film was obtained again, however with many small contaminant particles on the surface. Composition analyses indicated a reduced Re content to 0.07 at% but a markedly increased O content (from 2.1-2.3 at% in general) to 5.1 at%, revealing that, under a high EL Re ratio and an intense reaction, Re (with a higher oxidation potential than Cu) 29 more easily oxidized to form particle inclusions that would lead to the drastic increase in resistivity to 17.8 μ -cm as measured. After a direct 600 • C annealing, as observed in Figure 2e, some regions of the EL Cu(Re) 10:1 film and in particular the discontinuous EL Cu(Re) 10:2 film agglomerated into larger islands, accompanied with the formation of some small particles of Cu silicide (as verified by XRD analyses, not shown here), consequently elevating the resistivity to 19.7 and 25.7 μ -cm, respectively. By comparison, the annealed Cu(Re) 10:3 film exhibited a better continuity but a severer Cu/Si interdiffusion that caused the appearance of more and larger silicide particles and the higher resistivity of 29.3 μ -cm due to a lower Re content most of which was included in oxide particles with a poor diffusion resistance. The EL Cu(Re) 10:1 film with a relatively good morphology, low resistivity and high thermal stability, and an EP Cu(Re) alloy film shown in Figure 2f with a high Re content of 0.39 at% (±0.06%) were then selected for the further investigations below of self-forming an Re barrier layer at 400 • C and of thermal stability at 500-700 • C (denoted as 400 • C + 500-700 • C). The XRD patterns of as-deposited and thermally annealed EL and EP Cu(Re) alloy films presented in Figure 3 revealed a very small  shift of Cu (111) diffraction peak from 43.4 • to a lower angle for 0.1 • , possibly attributed to the expanded interplannar spacing caused by the incorporation of large-size Re atoms (1.97 Å, Cu 1.57 Å). Not a diffraction peak of Re phase was detected because of a very minor Re addition <0.5 at%. After an annealing at 400 • C, not any obvious differences but sharpened XRD peaks were found, indicating a grain growth rather than a change of crystal structure, as verified by the SEM surface morphologies of the thermally annealed Cu(Re) films shown in Figure 4. In Figure 4a, it was observed that the clustered small grains of the EL Cu(Re) film began to agglomerate into large and connected ones at 400 • C, although the grain growth rate seemed slower than that of the pure Cu films because of the pinned migration of grain boundaries by the added Re solute atoms. 30 By comparison, in Figure 4e, a special feature (texture) was formed on the 400 • C annealed EP Cu(Re) film with a higher Re content, very possibly owing to an Re (hexagonal close-packed) oriented grain growth or the surface precipitation of Re to form oxides. 25 After a further annealing at 500 • C (400 • C + 500 • C), as seen in Figure 4b, accompanied by some pores (due to grain growth induced film agglomeration) rather than any silicides, the irregular grains of the EL Cu(Re) film transformed into faceted ones with annealing twins. At 400 • C + 600 • C, in Figure 4c, the faceted grains and twins were more obvious; however, more and larger pores were observed, with the formation of some small Cu silicide particles that suggested a Cu/Si interdiffusion, as verified from the appearance of small Cu 3 Si diffraction peaks at 44.7 • and 45.3 • . At 400 • C + 700 • C, in Figure 4d, more and larger silicides were formed, indicating the complete failure of the film, as confirmed by the XRD analysis. For the annealed EP Cu(Re) film, the diffraction peaks of Cu 3 Si corresponding to the formed silicides in Figure 4f early appeared at 400 • C + 500 • C and were more obvious at 400 • C + 600 • C. Consistently, as plotted in Figure 5a, the normalized sheet resistance of the EL and EP Cu(Re) alloy films first decreased (the electrical resistivity of EL Cu(Re) decreased from 9.2 to 6.6 μ -cm) after an annealing at 400 • C because of a grain growth, then slightly increased (the resistivity of EL Cu(Re) increased to 7.6 μ -cm) at 400 • C + 500 • C due to a film agglomeration (for EL) or a silicide formation (for EP), and lastly, markedly increased (the resistivity of EL Cu(Re) increased to 14.0 μ -cm) at 400 • C + 600 • C when both the films failed.

Self-forming Re diffusion barrier layer.-
The normalized electrical resistivities (relative to as-deposited values) of as-deposited and thermally annealed EL and EP pure Cu and Cu(Re) alloy films in Figure 5b clearly indicate the superior endurance of Cu(Re) alloy films to a detrimental Cu/Si interdiffusion. The resistivities of EL and EP pure Cu films have begun to rise after an annealing at 400 • C and jumped to high values at 500 • C. In comparison, the resistivities of Cu(Re) alloy films decrease to the lowest values at 400 • C at higher descending rates than those of the pure Cu films, which is attributed not only to a grain growth but also the expelling of Re (segregation of Re to the Cu/Si interface to self-form an Re barrier layer), as suggested by ESCA analyses below. At 500 • C, the resistivities of both the Cu(Re) alloy films in particular the EL Cu(Re) very slightly increase, implying the inhibited Cu/Si interdiffusion by the self-forming Re barrier. More accurate Cu/Si interdiffusion behaviors can be further examined by using cross-sectional TEM observations (with compositional analyses); however, from the above SEM, XRD and electrical resistivity analyses, clearly, the Cu(Re) alloy films presented a higher-temperature endurance (roughly for 100-200 • C) than the pure Cu films did, still suggesting the effectiveness of the self-forming Re  barrier in resisting the interdiffusion. It was also noted that, at 400-600 • C, the ascending slope of resistivity of the EL Cu(Re) film is smaller than that of the EP Cu(Re), i.e., the Cu/Si interdiffusion in the EL Cu(Re) was less severe than that in the EP Cu(Re). Because a layer of densely packed Pd nanoparticles (with a size of several nm) that had been prepared on the surface of Si substrates for catalyzing the subsequent EL Cu(Re) deposition 26,27 might also inhibit the Cu/Si interdiffusion, the EL Cu(Re) thus showed a better thermal stability than the EP Cu(Re) without the Pd layer did. Figure 6a plots the ESCA elemental concentration distributions (depth profiles) in a 400 • C, 30-minute annealed Cu(Re) alloy film on an Si substrate. On the alloy film surface (slightly sputtered, ∼3 nm deep), as expected, Cu (binding energy 932 eV, Cu 2 O 952.3 eV) 31,32 and O (530 eV) 33 were detected, indicating a typical surface oxidation. A minor Re was also detected, suggesting the possible segregation of Re to the film surface. 25 In the film (below 3 nm) or the substrate (∼347 nm), only Cu or Si (99.1 eV) 34 of course was identified, respectively. At the Cu/Si interface (∼140 nm) of the 400 • C, 30-minute annealed Cu(Re) alloy film, in addition to the anticipated Cu, Si, and O (SiO 2 532.3 eV), 35 very low contents of Re (metallic state, Re 4f 7/2 at 39.9-40.5 eV and Re 4f 5/2 at 43.0 eV; ∼ 1 at%) 36 and Pd (335.3 and 340.5 eV; 37 ∼ 0.5 at%) were detected, as verified by their binding energy spectra (multiplex analyses of bonding configurations) in Figure  6b. Although the ESCA signals were low, the concentration profile of Re in Figure 6a still clearly indicated the near-zero content of Re in the film and the gradually increased content near the Cu/Si interface, revealing the segregation of Re and the self-formation of a thin Re layer. However, a minor content of ReO x might also possibly exist in the formed Re layer as very weak ESCA peaks of oxide-state Re (native oxide, Re 4f 7/2 at 41.7 eV and Re 4f 5/2 at 44.2 eV) 36 were mixed within the same binding energy range. To identify the exact Re state (metal or oxide), further detailed bonding configuration investigations are necessary. Figure 7 plots the experimentally measured distribution profiles of Re element concentrations in as-deposited (annealing duration of 0 s) and 400 • C annealed (1800 s) Cu(Re) alloy films. By comparing the two profiles, it is known that, during the thermal annealing, Re was separated from Cu and segregated at the Cu-Si interface, following uphill diffusion. Two important factors facilitate the uphill diffusion of Re from Cu to the Cu/Si interface or to the film surface: (1) the very low (zero) solubility of Re in Cu; in the Cu(Re) alloy system, a significantly raised free energy, G = H − T S, owing to a high mixing enthalpy, H mi x , cannot be compensated by the contribution of mixing entropy, −T S mi x , resulting in the instability of the alloy system but facilitating the phase separation of Re and Cu. (2) A small solubility of Re in Pd; 24 the Pd catalysts at the Cu/Si interface can act as preferred sites for the nucleation and growth of self-forming Re layer. A theoretical inverse thin-film diffusion solution is hence applied to curve-fit the experimentally measured distribution profile of Re concentration in the 400 • C annealed Cu(Re) alloy film, C (x,t) at time t and a distance from the Cu/Si interface toward the film surface x, under the assumption of uphill diffusion (piling up or accumulation, similar to a spinodal decomposition), using the following equation: 38,39 where C 0 is the original (average) concentration of Re before piling up, 0.10 at%, at time zero, S 0 = w (N 0 /V 0 ) V the atom density per unit area, nm −2 (for fcc Cu: N 0 = 4 atoms, V 0 = (0.3615 nm) 3 , w the thickness of one piled-up atom layer = 0.3615 nm, V = 1 nm 3 ), and λ the overall diffusion length (∼film thickness), 140 nm. From the best curve-fitting (prediction) result (at time 1800 s), as also plotted in Figure 7, the diffusivity of Re in Cu is estimated to be 1.5 × 10 −17 cm 2 /s at 400 • C, at the same order as that of immiscible Nb in Cu (∼4.2 × 10 −17 cm 2 /s at 400 • C; activation energy of Nb diffusion in Cu ∼ 52.6 kcal/mol). 40 Although the diffusivity of Re in Cu is slightly lower than the self diffusivity of Cu (∼1.0 × 10 −16 cm 2 /s at 400 • C; ) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 207.241.231.83 Downloaded on 2018-07-20 to IP activation energy ∼47.1 kcal/mol), 41 it is higher than the value of popularly studied Mn in Cu (∼2.2 × 10 −23 cm 2 /s at 400 • C; activation energy ∼ 91.4 kcal/mol). 40 Compared to miscible Mn, 16,17 immiscible Re with stronger bonding and a much smaller solubility in Cu, 24 which will yield a more spontaneous, rapid and complete phase separation from Cu, is another option for self-forming a thin, robust barrier layer at a low processing temperature. In addition, most Cu alloys for selfforming barriers are prepared by using a sputtering technique which suffers the non-uniform deposition in nanoscale trenches. [14][15][16][17][18][19][20][21] In this study, the electroless-and electroplating routes of Cu(Re) alloy films are expected to be more compatible to the current electrochemical metallization process.

Conclusions
In this study, Cu(Re) alloy films incorporated with small contents of Re for 0.10 and 0.39 at% were electroless-and electroplated on Si substrates. Facilitated by a zero solubility of Re in Cu and a small solubility in Pd catalysts, Re segregated to the Cu/Si interface, self-forming an ultrathin Re diffusion barrier layer at 400 • C. The segregation of Re follows uphill diffusion, and the diffusivity of Re in Cu at 400 • C is estimated to be 1.5 × 10 −17 cm 2 /s. Attributed to the self-forming Re layer and the coexisting Pd catalysts, the electrolessplated Cu(Re) alloy film did not fail until 600 • C which is higher than the failure temperatures of the electroplated Cu(Re) alloy film and the pure Cu films at 500 and 400 • C, respectively. The high-temperature endurance to Cu/Si interdiffusion and silicide formation suggests the high potential of electroless-plated Cu(Re) alloy film for application to barrierless metallization.