Surface Evolution during the Chemical Mechanical Planarization of Copper

Surface Evolution during the Chemical Mechanical Planarization of Copper

Surface Evolution during the Chemical Mechanical Planarization of Copper 1 2,3 3,2 W. Che , A. Bastawros , A. Chandra 1 Saint-Gobain Inc., Boston MA, ...

212KB Sizes 0 Downloads 60 Views

Surface Evolution during the Chemical Mechanical Planarization of Copper 1 2,3 3,2 W. Che , A. Bastawros , A. Chandra 1 Saint-Gobain Inc., Boston MA, U.S.A. 2 Dept. of Aerospace Engineering, Iowa State University, Ames, IA, U.S.A. 3 Dept. of Mechanical Engineering, Iowa State University, Ames, IA, U.S.A. Submitted by P.M. Lonardo (1), Genoa, Italy

Abstract Stressed surfaces are configurationally unstable under chemical etching wherein they may evolve to reduce their total energy. This paper investigates how such an effect may influence the planarization rate in a Chemical Mechanical Planarization (CMP) process. Nano-wear experiments on electro-plated copper surfaces have been conducted with systematic exposures to chemically active slurry. The nano-wear experiments have been first performed to generate local variation of the residual stress levels, followed by chemical etching to investigate the variation of the wear depth and the evolution of surface topography. It is found that the residual stress caused by the mechanical wear enhances the chemical etching rate. Keywords Surface, Polishing, Nano indentation

1 INTRODUCTION Chemical Mechanical Planarization (CMP) has grown rapidly during the past decade as part of mainstream processing methods in submicron integrated circuit manufacturing because of its global or near-global planarization capability. Currently, CMP is widely used for interlevel dielectrics and metal layer planarization [1]. The main objectives of the CMP process are to planarize dielectric surface topography, to enable multilevel metallization and to remove excess of deposited materials, which produces inlaid metal damascene structures and shallow trench isolations. A common way to perform the CMP process is by sliding the wafer surface on a relatively soft polymeric porous pad flooded with a chemically active slurry containing abrasive particles of sub-micron diameter. Two simple phenomenological models have been widely accepted for the role of chemistry in the CMP process. The first is proposed by Kaufmann [2] for CMP of tungsten as an abrasion-passivation mechanism. The mechanical abrasion removes the passivated oxide layer on the top of an asperity followed by accelerated chemical dissolution of detached oxide material, while simultaneously another oxide layer forms on the exposed metal surface. The effect of the intact oxide in the recessed area is to protect the recessed surface from dissolution, thus achieving the planarization capability. This mechanism could also apply to noble metals, such as copper in alkaline solutions because of formation of passivating oxides [3]. Accordingly, corrosion inhibitors such as benzotriazole (BTA) are used in acidic slurries to form passivating layers. These passivating films would play the role of oxide layer in the abrasion-passivation mechanism [4, 5]. The second is proposed by Steigerwald [6] as an abrasion-dissolution mechanism. The role of chemicals is only to dissolve the abraded metallic debris left by abrasion. However, in industrial settings, the material removal rate is achieved by balancing the abrasion rate with the dissolution rate. It can be noted that both of these models separate the chemical and mechanical effects into two sequential Annals of the CIRP Vol. 55/1/2006

effects. However, various researchers have observed synergistic interactions between these two effects. It has been shown that a flat surface with residual stress near the free surface is unstable under chemical etching [7]. The surface long wavelength greater than a critical level will grow, while short wavelength will decay [8], [9]. In CMP, similar effects may exist and thereby further enable the planarization process. In this paper, we will focus on understanding the synergy between the chemical dissolution and mechanical abrasion in copper CMP by quantifying the relative material removal due to each mechanism and the wavelength and stress dependent selectivity of the process. The gained insights would enable the control of the relative rates of chemical dissolution and mechanical abrasion. 2

DRIVING FORCES FOR SURFACE INSTABILITIES A solid may lose or gain mass from its environment by a surface exchange process. Examples include vapor deposition, chemical etching, and phase transition. Many long-range and short range transport mechanisms influence the velocity of the interface motion. The longrange mechanisms are typically associated with heat conduction, mass diffusion and viscous flow. At the short range level, atoms must leave one phase, react on the interface, and join the other phase. As the surface interface of a nominally flat solid moves, it either remains flat or becomes wavy under different driving forces such as surface energy, elastic strain energy, electrochemical energy, among many other driving forces. Mullins [10] showed the effect of surface energy on the evaporation of a solid. If the solid surface is perturbed into a wavy shape, the surface energy causes the solid to evaporate faster at crests than at the troughs, so that the wave amplitude decays over time, and the surface flattens. Srolovitz [7] showed continued evolution of a wavy surface under a stress applied parallel to the surface. While the applied remote stress field is uniform, the local stress field is nonuniform: the magnitude of the

stress is larger at troughs than at crests. Consequently, the solid evaporates faster at the troughs than at the crests, so that the wave amplitude grows over time, and the surface roughens. When both surface energy and stress act together, the ratio of the surface energy to the elastic strain energy, defines a length scale, λcr, wherein, shorter waves decay, but longer waves grow. When such a stressed wavy surface is exposed to an active chemical etch solution, the driving forces F at any special point on the surface (i.e., the free energy reduction associated with the solid gaining per unit volume) will be [11],

F = g − w − Kγ

(1)

Here g is the electrochemical potential, w the elastic strain energy per unit volume (quadratic in stress so it does not depend on the stress sign), K is the sum of the principal curvatures and γ is the surface energy per unit area. The solid gains mass when F>0, and loses mass when F<0. While controlling g by itself does not affect surface stability, its relative amplitude to both w and Kγ will have a significant impact on controlling the evolution of surface morphology. The reaction rate R (namely, the volume of solid gained (R>0) or lost (R<0) per unit surface area per unit time) depends on the driving forces, F, either linearly near equilibrium, or exponentially far from equilibrium. Moreover, the proportionality constant for such dependence is known as the surface mobility, and it can also depend on the sign of the applied stress. Accordingly, the reaction rate at any point along the surface depends on the applied stress in two different ways; (i) through the driving force F, which depends on the stress through the elastic energy density. Since the elastic energy density is quadratic in the stress, the resulting reaction rate is the same for a solid under tension and under compression. (ii) Through the dependence of the surface mobility on the applied stress. In a typical etching process, as depicted in Figure 1, when the solid is under a compressive stress (Figure 1a), the etching rate is slower at troughs than at crests, so that the crests will catch up with the troughs, and the wave amplitude decays. Conversely, when the solid is under a tensile stress, the wave amplitude grows (Figure 1b). For biaxial stress field on a surface, Kim et al [8] found that the critical wavelength is

λcr = Eγ / σ

2

(2)

Here, E is the Young’s modulus of the surface material and σ is the principal stress in the direction of the waviness. fast

fast

(b)

slow

σ

Stable solid flattens out

σ σ

slow

fast

slow

Unstable solid roughens out

Figure 1: Scheme of chemical etching at a surface under spatially uniform or varied stress. (a) The reaction rate decreases under uniform compression and stabilizes flat surface. (b) The reaction rate increases under uniform tension and destabilizes flat surface.

From these considerations, it is quite clear that the evolution of a surface during reaction depends on how the reaction rate varies with the stress. Here, we will address the relevant operative mechanisms during a lab simulated CMP process. 3 EXPERIMENTAL PROTOCOL Copper film is electrodeposited onto oxygen free, 99.99% high purity copper (101-alloy series) discs in order to get a similar microstructure of copper used in semiconductor process. The plating bath is CuSO4·5H2O and H2SO4 [12]. Copper is deposited for an hour under constant voltage 0.5 V. A thick layer (more than 50 μm) is achieved based on the experimental conditions [12]. The specimen is kept at room temperature for more than a day to achieve the desired grain size of 1-2 μm range [13]. Finally the specimen is gently polished with 1 μm and 0.05 μm alumina particles to achieve initial roughness of about 4-5 nm over 20 μm window. The polished specimens are kept under vacuum. Before the nano-wear experiment, each specimen is then cleaned by a 2 wt% HNO3 for 1 minute to remove the native oxide layer. Nano-wear experiments are conducted using a triboindenter in wear mode. A diamond Berkovich tip with a tip radius of about 500 nm is utilized. A set of dry wear tests with different normal force is conducted on the copper surface on 10 μm window (about 5-10 times the grain size). Then, the surface is exposed to the chemically active slurry for different length of exposures. The same wear regions are scanned by SPM before and after chemical exposure with a 20 μm window to examine the effects of mechanical abrasion on chemical etching. Finally, a compound multi-pass wear test is carried out to examine the mutual effects of both mechanisms on the material removal. The test is composed of multi-pass dry wear, followed by chemical slurry exposure, then another set of multi-pass dry wear. 4 RESULTS 4.1 Stress Enhanced Chemical Dissolution If mechanical abrasion enhances the chemical dissolution, then residual stresses that evolve within the wear region should be the main effect. In general, the chemical reaction rate is controlled by the change in free energy, surface energy and elastic strain energy. Residual stresses will change the elastic strain energy and thereby the chemical reaction rate. To examine this effect, a set of dry wear tests are carried out on copper under normal force of 10-40 μN. The wear surface is exposed for a varying length of time (1-10 min) to two chemically active slurries, commonly used slurries in CMP; I (1 wt% HNO3 +0.5 wt% BTA) and II (0.6 wt% NH4OH). Typical results for both slurries are shown in Figures 2, 3. The surface evolution is depicted in Figure 2(a) after 10 min exposure to slurry-I, wherein the step increase represents the stress-enhanced etching. The experimental measurements depicted on Figure 2(b) show a wear depth increased by as much as ~ 30% under certain normal force. The nonlinear response shown for the wear depth as a function of the applied load is a manifest of the nonlinear stress field associated with the indentation field. This synergistic effect is further studied for slurry-II. Initially, 4 dry wear passes at 40 μN or 8 dry wear passes at 20 μN are employed to provide approximately the same wear depth but with different residual stress levels. The surface topographic evolutions are measured as a

(b) 20

Dry wear

h+Δh Stress enhanced wear

Wear depth (nm)

h

Slurry-I

16 12 8 4 0

Stress enhanced

• Dry wear 0

10 20 30 40 Normal force (μN)

50

Figure 2: Wear depth variation via abrasion-dissolution mechanism. (a) Schematic representation. (b) Experimental measurements of wear depth evolution. function of the exposure to the slurry. The time evolution of the wear depth is shown in Figure 3(a). Starting with almost the same level of cumulative dry wear level, it is evident that the increase in the wear depth depends on the level of the accumulated residual stress levels. It is evident also that the change of the wear depth with time follows an exponential decay rate until the whole residually stressed layer is consumed. A net enhancement of the dry wear trench by ~15% to ~30% is observed on different etching time. Figure 3(b) shows the time history of roughness evolution under chemical attack of slurry-II. The original surface showed very modest variation of its roughness, while the abraded region showed steady roughening evolution. It should be noted that the reported static etch rate slurry-II is close to 0 nm/min [6]. This fact ascertains that the increased etch rate arises from the stress assisted chemical dissolution. (a)

Wear depth (nm)

50

fast

60

medium

low saturation

10

▪ 0

500

Fn=20 μN Fn=40 μN 1000

Time (s) (b)

Slurry-II

12

wear + chemical exposure

10 Ra (nm)

25

8 6 chemical exposure only

4 Fn=40 μN

2 0

500 Time (s)

Multi-wear: wear depth approaches zero MRR is increased due to wear after etch

20 15

Multi-wear: wear depth approaches zero

10

0

20

Total step height

5

30

0

30

Relative change

low

40

0

4.2 Synergistic Role A compound multi-pass wear test is carried out to examine the mutual effects of both mechanisms on the material removal. The test is composed of 4-passes dry wear at 40 μN normal force, followed by chemical slurry exposure to slurry-II for 90 seconds, then another set of 4pass dry wear at the same force level. The experimental measurements of the resulting material removal variation are summarized on Figure 4. This figure shows the steady abrasion of dry wear, the instantaneous jump due to stress enhanced chemical dissolution, the chemical enhanced abrasion and the steady wear to follow after that. The first data point represents the initial unsteady wear due to interaction with the initial surface asperities and roughness. The Δ h is increased nearly by 15% due to residual stress enhanced etching (5th point). The fifth th wear path (6 point, immediately after exposure) shows the largest Δ h increase compared to other increments. This presents the synergistic interactions between chemical dissolution and mechanical abrasion in the material removal process. The wear rate caused by the interaction between chemical etching and mechanical abrasion is much larger than the pure mechanical abrasion rate or chemical etching rate alone.

Wear depth (nm)

(a)

1000

Figure 3: Time evolution of stress enhanced chemical dissolution. (a) Wear depth. (b) Surface roughness.

0

2 4 6 8 Number of wear passes

10

Figure 4: A compound multi-pass wear test with single th exposure to slurry-II after the 4 path for 90 s. 4.3 Roughness Evolution To examine the wavelength selectivity during the CMP process, a nano-wear test is conducted with 20 μN normal load, and then followed by chemical exposure. Topographic measurements are made on an area of 20 μm by 20 μm before and after chemical exposure with a 256 by 256 digital sampling. The surface frequency spectrum is analyzed by FFT. The experimental results for surface etched by slurry-I (1 wt% HNO3 + 0.5 wt% BTA) for 10 minutes are show in Figure 5. The Y-axis is related to the amplitude of the surface roughness, and the X-axis is the surface frequency. The amplitude of low frequency (frequencies at left part of intersection point) after etching is higher than that before etching, indicative of the growth of long wavelength surface roughness. The amplitude of high frequency (frequencies at right part of intersection point) after etching is lower than that before etching, indicative of decays of short wavelengths. Figure 5 indicates critical wavelength selectivity; λcr ~ 0.75 μm. Tests with exposure to slurry II (0.6 wt% NH4OH) slurry have showed similar results with λcr ~ 1.4 μm.

Amplitude (au)

80

Long wavelengths dominate

X X

60

After chemical exposure

X X X X

40

Before chemical exposure

XX

Short wavelengths decay

XX X XX X XX XX XX X XXX X XXX XXXXX X XXXX XXX X XXXXXXX XXXXXXX XX

20 0 0

1

XXXXX

2 3 4 5 6 Wave number (1/λ, μm-1)

7

Figure 5: Experimental data showing long wave grows, and short wave decays for mechanical wear followed by chemical etching (1wt% HNO3 +0.5wt% BTA). 5 DISCUSSION Systematic nano-wear tests have been designed and conducted on electroplated copper with and without chemical exposure. It has been observed that the synergy between chemical dissolution and mechanical abrasion plays an important role in material removal during the CMP process. The stress enhanced chemical etching increased the wear depth by as much as 30%. The percentage increase is determined by the applied pressure, properties and thickness of chemical product. The wavelength selectivity phenomenon due to residual stresses is also observed to enhance local planarization. The λcr represents an additional design parameter that would dictate the selection of the chemical strength of the slurry. By estimating the level of residual stresses from a simulated multi-track nano-wear test by finite element (Che [14]), we found that the average residual stresses normal to the wear direction is about 150 MPa for the process parameters utilized here (Copper surface with E = 123 GPa, ν = 0.33, σy = 70 MPa, σult = 220 MPa, indentation depth of 10 nm, tip radius of 400 nm and track overlap of 40 nm). For such values of residual stresses due to the mechanical wear abrasion and utilizing Equation (2), one can arrive to estimate the interface energy between the 1 wt% HNO3 + 0.5 wt% BTA and the 2 copper at about γ =0.14 J/m and between 0.6 wt% NH4OH slurry and the copper at about γ = 0.26 J/m2. These values are within the typical range for interface energies [8]. 6 CONCLUSION Residual stresses to mechanical wear seems to significantly influence the chemical etching process. Further studies are needed to render these observations into practical design space parameters for CMP processes. 7 ACKNOWLEDGMENT This work is supported by U.S. National Science Foundation under Grants No. DMI-0084736 and DMI0323069.

8 REFERENCES [1] Martinez, M. A., 1994, Chemical-Mechanical Polishing: Route to Global Planarization, Solid State Technology, 37/5: 26-31. [2] Kaufman, F.B., Thompson, D.B., Broadie, R.E., Guthrie, W.L., Pearson, D.J. and Small, M.B., 1991, Chemical-mechanical Polishing for Fabricating Patterned W Metal Features as Chip Interconnects, Journal of the Electrochemical Society, 138/11: 3460-3465. [3] Pourbaix, M., 1966, Atlas of Electrochemical Equilibria in Aqueous Solutions, Pergamon Press, Oxford. [4] Carpio, R., Farkas, J. and Jairath, R., 1995, Initial Study on Copper CMP Slurry Chemistries, Thin Solid Films, 266: 238-244. [5] Luo, W., Campbell, D. R. and Babu, S. V., 1997, Chemical-mechanical polishing of copper in alkaline media, Thin Solid Films, 311: 177-182. [6] Steigerwald, J. M., Murarka, S. P., Gutmann, R. J. and Duquette, D. J., 1995, Chemical Processes in the Chemical Mechanical Polishing of Copper, Materials Chemistry and Physics, 41: 217-228. [7] Srolovitz, D. J., 1989, On the stability of surfaces of stressed solids, Acta Metallurgica, 37: 621-625. [8] Kim, K. S., Hurtado, J. A. and Tan, H., 1999, Evolution of a surface-roughness spectrum caused by stress in nanometer-scale chemical etching, Physical Review Letters, 83: 3872-3875. [9] Yu, H. H. and Suo, Z., 2000, Stress-Dependent Surface Reactions and Implications for a Stress Measurement Technique, Journal of Applied Physics, 87: 1211-1218. [10] Mullins, W. W., 1957, Theory of Thermal Grooving, Journal of Applied Physics, 28: 333-339. [11] Herring, C., 1951, in The Physics of Powder Metallurgy, edited by Kingston, W.F., McGraw-Hill, New York, pp. 143-179. [12] Weisenberger, L. M. and DurKin, B. J., 1978, Copper plating, in ASM Handbook: Surface Engineering. Materials Park, OH: ASM International, 1994, 5: 167–176. [13] Gignac, L. M., Rodbell, K. P., Cabral, C., Andricacos, P. C., Rice, P. M., Beyers, R. B., Locke, P. S. and Klepeis, S. J., 1999, Materials Research Society Symposium Proceedings, 564: 373-378. [14] Che, W., 2005, Ph.D. thesis, Iowa state University. [15] Evans, C. J., Paul, E., Dornfeld, D., Lucca, D. A., Byrne, G., Tricard, M., Klocke, F., Dambon, O., Mullany, B.A, 2003, Material removal mechanisms in lapping and polishing, Annals of the CIRP, 52/2: 611-634. [16] Komanduri R., 1996, On material removal mechanisms in finishing of advanced ceramics and glasses, Annals of the CIRP, 45/1: 509-514. [17] Wang, C., Sherman, P., Chandra, A., Dornfeld, D., 2005, Pad Surface Roughness and Slurry Particle Size Distribution Effects on Material Removal Rate in Chemical Mechanical Planarization, Annals of the CIRP, 54/1: 309-312. [18] Brinksmeier, E., Lucca, D. A., Walter, A., 2004, Chemical Aspects of Machining Processes, Annals of the CIRP, 53/2: 685-699