Microelectronic Engineering 64 (2002) 351–360 www.elsevier.com / locate / mee Dependence of the minimal PVD TA(N) sealing thickness on the porosity of Zirkonீ LK dielectric films F Iacopi a,b , *, C Zistl c , C Jehoul d ,1 , Zs Tokei a , Q.T Le a , A Das a , C Sullivan d , G Prokopowicz d , D Gronbeck d , M Gallagher d , J Calvert d , K Maex a,b a IMEC, Kapeldreef 75, B-3001 Leuven, Belgium E.E Department, Katholieke Universiteit, Leuven, Belgium c AMD Saxony, Postfach 110110, D01330 Dresden, Germany d Shipley Company, Marlborough, MA, USA b Abstract The understanding of the parameters and mechanisms playing a role in sealing the surfaces of porous dielectrics is fundamental for an effective integration of these new materials in the interconnect processing This study focuses on the correlation between the porosity of Zirkonீ low-k dielectric films and the physical vapour deposition (PVD) Ta(N) thickness needed to achieve an efficient sealing of the dielectric surface A clear dependence of the minimal Ta(N) sealing thickness on the porosity of the Zirkon dielectric films has been observed Since the average pore size of the films remains approximately constant for the different porosity formulations, this effect is most probably attributed to a partial conversion of the initial ‘closed pore’ structure into an ‘open pore’ one for the highest porosity films 2002 Elsevier Science B.V All rights reserved Keywords: Porous low-k films; Zirkonீ; Cu drift barrier; Sealing; Ellipsometric porosimetry Introduction Dielectrics with decreasing permittivities (k values) will have to be employed in interconnects to comply with the scaling trend in CMOS technology, as reported in the International Technology Roadmap for Semiconductors [1] This is necessary in order to reduce signal transmission delays in integrated circuits arising from R–C couplings One widely-used method to obtain ‘ultra-low-k ’ * Corresponding author Fax: 132-16-281-214 E-mail address: francesca.lacopi@imec.be (F Iacopi) Affiliate at IMEC from the Shipley Company 0167-9317 / 02 / $ – see front matter PII: S0167-9317( 02 )00808-0 2002 Elsevier Science B.V All rights reserved F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 352 materials is to incorporate a certain amount of porosity in the already existing dielectric matrices, so that lower dielectric constants can be achieved following the trend of a two-phase material [2]: ln kc v1 ln k1 v2 ln k2 (1) For a porous film, k1 and k2 are the permittivities of the ‘dense’ material and that of air, respectively, while kc is the resulting dielectric constant; v1 and v2 are their respective volume fraction in the film This approach, though being effective for k value reduction, presents a number of drawbacks both from interconnect processing and device reliability point of view The presence of a high porous volume fraction in the films has detrimental influences on the mechanical [2,3] and thermal [4] properties of the dielectrics, and it makes them more sensitive to the processing conditions throughout their integration steps in interconnects [5] Moreover, the size of the pores of the currently proposed ultra-low-k dielectrics is invariably of the order of magnitude of a few nanometers, so that their sealing becomes an important issue In particular, the realization of an efficient sidewall sealing of porous dielectrics by depositing very thin (in the order of few nm as well [1]) Cu diffusion barriers has become a critical matter [6] Various parameters related to both porous dielectric films and barrier deposition process can have impact on the sealing efficiency Here the influence of film porosity on the minimal PVD barrier sealing thickness is investigated Experimental Experiments have been carried out by capping blanket Zirkonீ films (Zirkon LK is a trademark of the Shipley Company, L.L.C.) with physical vapour deposition (PVD) Ta(N) layers with thicknesses ranging from 10 to 60 nm The N flow during PVD deposition is such to yield a N-poor Ta(N) film (Ta N) Zirkon is an MSQ-based low-k dielectric, and is synthesized through a polymer templating technology that allows the preparation of films with tunable porosity [7] Films prepared with wt.% (‘dense’), wt.%, 22.5 wt.% and 33.75 wt.% porogen load have been compared in this study The uncapped dielectric films have been characterized by ellipsometric porosimetry (EP, [8]) to extract information about porosity and pore size distribution, and by atomic force microscopy (AFM) to compare surface roughness Sheet resistance measurements by four-point probe have been performed on the thin Ta(N) layers deposited on the different dielectric films, and wafer-level EP has been used to verify the sealing performance of these cap layers [5,6] This has been done by monitoring the ellipsometric angles D and C of the Ta(N) / dielectric stack during the adsoprtion process The samples have been brought to a moderate vacuum condition (10 22 Torr) and subsequently toluene was let into the chamber at a fixed flow-rate to reach the solvent saturation pressure at room temperature (P0 ) This measurement follows the principle that if a capping layer on top of a porous low-k dielectric constitutes an efficient sealing layer, no absorption of solvent into the dielectric stack is observed when the solvent pressure increases The measured curves are also compared to the computed ones showing the expected evolution of the ellipsometric angles in the case where the solvent was adsorbed into the dielectric through the Ta(N) layer The computed curves assume a complete filling of the porous volume and are calculated taking into account the transmission and reflection coefficients of the full stack (multiple interfaces) F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 353 Table Properties of Zirkon dielectric films prepared with increasing porogen load a Film load (wt.%) Total porosity (V%) k Value from Eq (1) Pore krl (nm) Roughness (RMS, nm) Integrity 10 nm Ta(N) Integrity 20 nm Ta(N) 22.5 33.75 17 22 35 45 2.7 2.57 2.26 2.04 Micro 1.2 1.5 1.8 0.43 0.42 0.37 0.52 Closed Closed Pinholes Porous Closed Closed Closed Pinholes a K values are derived through Eq (1), where the value of 2.7 for the dense MSQ film has been taken as a reference and the remaining values are calculated on the basis of the added film porosity measured by EP The last two columns show the sealing efficiency of thin Ta(N) cap layers Results and discussion 3.1 Characterization of uncapped dielectric films The first two columns in Table show the total porosity and average pore size measured on the Zirkon dielectric films prepared with increasing porogen load The k values derived through Eq.(1) for these films are reported in Table 1, where the value of 2.7 for the dense MSQ film has been taken as reference and the remaining values are calculated on the basis of the added film porosity measured by EP In Fig the pore radius distributions are also reported for each formulation The dense film presents about 17% porosity, essentially given by pores with radius smaller than nm (micropores) The increase in porogen load leads to an almost linear increase in total porous volume up to 45% Fig Pore size distributions of films prepared with different porogen load: wt.% (A), wt.% (B), 22.5 wt.% (C) and 33.75 wt.% (D) 354 F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 Fig Two-dimensional power spectral densities (2D PSD) from AFM measurements on films with different porogen load Films with the highest load differ from the rest by showing a larger contribution to roughness for finer features (smaller spatial wavelengths) porosity for the highest load The average radius of the mesopores does not show significant differences versus the various loadings (Table and Fig 1), and that the peak corresponding to the initial microporosity remains present in the pore size distributions Concerning surface roughness, no substantial difference has been noticed among the dense and the low and intermediate porosity films Films with the highest porogen load show a slightly higher RMS value (Table 1), caused by an increase in roughness for finer features (smaller spatial wavelengths, l) as can be seen from the power spectral densities (Fig 2) 3.2 Characterization of thin Ta( N) cap layers The graph in Fig reports sheet resistance values for 10 nm Ta(N) deposited on the different Zirkon films No significant differences are observed between Ta(N) layers on the 0, and 22.5 wt.% load dielectric films However, an increase in sheet resistance can be noticed on top of the 33.75 wt.% Fig Values of sheet resistance on 10 nm thick Ta(N) layers deposited on top of blanket Zirkon films: an increase is observed on top of the highest porosity films F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 355 Fig Behaviour of sheet resistance for Ta(N) layers with different thickness, deposited on top of dense Zirkon films and on top of the highest porosity films The latter shows a slight deviation from linearity, caused by a steeper increase for the 10-nm thick Ta(N) film load films In Fig the behaviour of sheet resistance for different thickness of Ta(N) layers on top of dense and highest porosity Zirkon films is also shown The graph shows that the behaviour differentiates only for the thinnest Ta(N) depositions (10 nm), where a deviation from linearity is observed for the highest porosity films In the last two columns of Table the sealing efficiency of thin Ta(N) films on top of the different porosity Zirkon films is reported Three distinct situations are observed concerning the integrity of the cap layers On top of the dense and the wt.% porogen load films 10 nm Ta(N) appear as a fully closed layer, as can be seen in Fig 5A): the ellipsometric angles measured on the Ta(N) / dielectric stack remain constant throughout the solvent flow process, indicating that no solvent has been adsorbed into the pores of the low-k film An equivalent capping layer on top of films prepared with 22.5 wt.% (Fig 5B) and 33.75 wt.% (Fig 5C) porogen load show an evolution in the D /C diagram corresponding to a complete filling of the porous volume in the dielectrics, as shown by the agreement between the experimental and computed curves There is though an essential difference between the adsorption kinetics of these two situations As shown by the graphs in Fig 6, while the adsorption isotherms for an high porosity uncapped film and an equivalent film capped by 10 nm Ta(N) layer are nearly identical (Fig 6A), the solvent adsorption process in the Ta(N) / 22.5 wt.% load film shows a considerable delay when compared to its uncapped dielectric counterpart (Fig 6B) This is seen in Fig as a delay versus pressure, but is an indication of a kinetic-limited regime and expresses in reality a time-delay effect When at least a few percent of the initial amount of pores in the dielectric still remain open after capping, no delay is seen, as in Fig 6A) When only a very small amount of defects / pinholes are present in the cap layer (roughly in the order of a few per mm and less), the adsorption of solvent in the stack can only happen very slowly: it is limited by the diffusion velocity of the solvent in the porous network (see Fig 7) The delay has indeed been found to be related to the position of the measurement point with respect to the nearest defect / pinhole Concerning the desorption phase, due to capillary forces the delay is observed even more pronounced, and a complete removal of the solvent from the stack at atmospheric pressure in a pinhole regime is achieved only after several hours of storage This is witnessed by the picture in Fig The surface of a Ta(N) / dielectric stack is 356 F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 Fig Measurements of the sealing performance of 10-nm Ta(N) cap layers through ellipsometric porosimetry: the cap layer is fully closed on top of films with wt.% load (A) while full solvent adsorption in possible through voids on top of films with 22.5 wt.% (B) and 33.75 wt.% (C) load, as also shown by the agreement between the experimental and computed evolutions in the D versus C diagram photographed after solvent adsorption and desorption processes: the solvent still trapped in the dielectric makes a few millimetre wide region around the pinholes in the cap layer (voids and defects in general) clearly visible to the naked eye In particular, in Fig a scratch is visible and regularly spaced defects created during sheet resistance measurements can be recognized Final discussion The shown results can be interpreted as follows The microporosity of the dense Zirkon films can be fully sealed by 10 nm PVD Ta(N) In the same way, films prepared with wt.% porogen load, F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 357 Fig Adsorption isotherms for films prepared with 22.5 wt.% (A) and 33.75 wt.% (B) load measured by ellipsometric porosimetry both for uncapped films and films capped by 10 nm PVD Ta(N) The adsorption process is substantially delayed in the case of the capped 22.5 wt.% film, indicating that the cap layer possesses only a very small amount of voids (pinhole regime) Fig Schematical representation of a pinhole regime: an isolated defect / void in the cap layer lets the solvent reach the dielectric and diffuse through the porous network at a speed that reaches few mm per hour 358 F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 Fig Example of pattern given by solvent trapped in the cap / dielectric stack around local defects in the cap layer (darker spots) Note that the regular pattern has been generated through sheet resistance measurements, while a scratch on the barrier layer can also be identified showing mesopores with krl51.2 nm (Table 1), can also be efficiently closed On the other hand, when the porosity of the film starts increasing, more and more difficulties are encountered in sealing the surface by the same capping thickness even though the pore size distribution does not change significantly First some pinholes appear, as in the case of films with 22.5 wt.% load, and on top of films with the highest porogen load 10 nm Ta(N) are plainly porous In the latter case, since the thin Ta(N) layer is systematically porous, an increase in sheet resistance value is also observed This can be explained by the following model The addition of nanoparticles (porogens) in the silsesquioxane resin precursor leads most probably to mesopores interconnected by microchannels given by the original microporosity of the MSQ resin, as also shown by the bimodal pore size distributions in Fig This is a condition in which the pore structure can be considered as closed, since the necking size is much smaller than the size of the mesopores, and pores appearing a the surface can be just closed by conformal deposition from the bottom, as shown in Fig 9A) When forcing an increasing porous volume fraction into the bulk of the dielectric films by higher porogen loads, there is an enhancement of the probability that the mesopores will end up so close to each other that their necking connection will be affected If the size of this interconnection becomes comparable to the size of the mesopores, the pore structure becomes open: as shown in Fig 9B), a conformal coverage is no longer possible These pores will need to be bridged by the PVD deposition, so that a substantially thicker Ta(N) layer is needed to achieve an efficient sealing In the case of films with the 22.5% porogen load, the frequency of these open pores is extremely low, giving rise to a pinhole condition when capped by a 10-nm Ta(N) layer Films with the highest porosity (33.75 wt.% load) show a substantially higher frequency of this sort of pores, which probably start to account for a few percent of the total porosity The slight increase of surface roughness for small spatial wavelengths observed only on films prepared with the highest load conditions further support this statement F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 359 Fig Pore structure in a Zirkon mesoporous dielectric film: sacrificial nanoparticles (porogens) generate mesopores connected by the original microporosity of the matrix MSQ material The mesopores appearing at the surface can be conformally covered by thin PVD deposition (A) When the porogen load percentage of the films is increased (B), part of this closed pore structure can be converted into an open one (with necking comparable to the pore size): these pores can only be sealed by bridging them with a thicker cap layer The pinholes remaining in the 10 nm Ta(N) cap on the 22.5 wt.% load film are then successfully closed by doubling the thickness of the cap layer, while in the highest porosity films such significantly large pores are generated by this mechanism that cannot be closed by 20 nm Ta(N) The frequency of these pores is small enough to appear as a pinhole condition (Table 1) Conclusions A clear dependence of the minimal PVD Ta(N) sealing thickness on the porosity of the Zirkon dielectric films has been observed Since the average pore size of the films remains approximately constant for the different porosity formulations, this effect is most probably attributed to a partial conversion of the initial closed pore into an open pore structure The increase in porogen load of the films enhances the probability of generating a certain amount of open pores, i.e pores whose interconnection starts to be comparable to the pore size This sort of pores is substantially more difficult to seal Data from sheet resistance and EP measurements agree very well, though EP appears definitively as a more sensitive technique for studies on sealing efficiency 360 F Iacopi et al / Microelectronic Engineering 64 (2002) 351–360 Acknowledgements Mikhail R Baklanov, Denis Shamiryan, Konstantin P Mogilnikov and Ivan Callant are gratefully acknowledged for scientific and technical help concerning the ellipsometric porosimetry systems References [1] International Technology Roadmap for Semiconductors (ITRS), update 2001 [2] Y Xu, D.W Zheng, Y Tsai, K.N Tu, B Zhao, Q.Z Liu, M Brongo, C.W Ong, C.L Choy, G.T Sheng, C.H Tung, J Electronic Mater 30 (4) (2001) 309 [3] C Jin, S Lin, J.T Wetzel, J Electron Mater 30 (4) (2001) 284 [4] C Hu, M Morgen, P.S Ho, A Jain, W.N Gill, J.L Plawsky, P.C Wayner, Appl Phys Lett 77 (1) (2000) 145 [5] F Iacopi, M.R Baklanov, E Sleeckx, T Conard, H Bender, H Meynen, K Maex, J Vac Sci Technol B20 (1) (2002) 109 [6] J.-N Sun, D.W Gidley, T.L Dull, W.E Frieze, A.F Yee, E.T Ryan, S Lin, J Wetzel, J Appl Phys 89 (2001) 5138; F Iacopi, Zs Tokei, D Shamiryan, Q.T Le, S Malhouitre, M Van Hove, K Maex, in: Proc Advanced Metallization Conference (AMC) 2001, Mat Res Soc., Warrendale, PA, USA, 2002, pp 61–66; D Shamiryan, M.R Baklanov, Zs Tokei, F Iacopi, K Maex, in: Proc Advanced Metallization Conference (AMC) 2001, Mat Res Soc., Warrendale, PA, USA, 2002, pp 279–285; Zs Tokei, J.J Waeterloos, F Iacopi, R Caluwaerts, H Struyf, J Van Aelst, K Maex, in: Advanced Metallization Conference (AMC) 2001, Mat Res Soc., Warrendale, PA, USA, 2002, pp 307–311 [7] M Gallagher, N Pugliano, J Calvert, Y You, R Gore, N Annan, M Talley, S Ibbitson, A Lamola, in: MRS Spring Meeting, San Francisco, 2001; M.R Baklanov, C Jehoul, C.M Flannery, K.P Mogilnikov, R Gore, D Gronbeck, G Prokopowicz, C Sullivan, Y You, N Pugliano, M Gallagher, in: Proc Advanced Metallization Conference (AMC), 2001, Mat Res Soc., Warrendale, PA, USA, 2002, pp 273–278 [8] F.N Dultsev, M.R Baklanov, Electrochem Sol St Lett (1999) 192 ... clear dependence of the minimal PVD Ta( N) sealing thickness on the porosity of the Zirkon dielectric films has been observed Since the average pore size of the films remains approximately constant... columns of Table the sealing efficiency of thin Ta( N) films on top of the different porosity Zirkon films is reported Three distinct situations are observed concerning the integrity of the cap... corresponding to the initial microporosity remains present in the pore size distributions Concerning surface roughness, no substantial difference has been noticed among the dense and the low and intermediate