- Email: [email protected]
Combined ultrasonic vibration and chemical mechanical polishing of copper substrates
International Journal of Machine Tools & Manufacture 53 (2012) 69–76
Contents lists available at SciVerse ScienceDirect
International Journal of Machine Tools & Manufacture journal homepage: www.elsevier.com/locate/ijmactool
Combined ultrasonic vibration and chemical mechanical polishing of copper substrates Ming-Yi Tsai n, Wei-Zheng Yang Department of Mechanical Engineering, National Chin-Yi University of Technology, No. 35, Lane 215, Section 1, Chung-Shan Rd., Taiping City, Taichung County 411, Taiwan
a r t i c l e i n f o
abstract
Article history: Received 29 December 2010 Received in revised form 17 September 2011 Accepted 20 September 2011 Available online 4 October 2011
This paper introduces an ultrasonic, vibration-assisted, chemical mechanical polishing (UV-CMP) method and an ultrasonic, vibration-assisted, traditional diamond disk (UV-TDD) dressing method. A copper substrate is polished by traditional CMP and UV-CMP. UV-CMP combines the functions of traditional CMP and ultrasonic machining (USM) with small-amplitude, high-frequency tool vibration to improve the fabrication process and machining efficiency. The removal rate of the copper substrate, torque force, and polished surface morphology of CMP and UV-CMP are compared. The polishing pad is also dressed by traditional diamond disk (TDD) and UV-TDD. The pad cut rate, torque force, and pad surface profiles of TDD and UV-TDD are also investigated in experiments. Experimental results reveal that UV-TDD can produce twice the pad cut rate and reduce torque force compared to TDD. Consequently, a dressing time reduction by half is expected, and hence, the diamond life is extended. It is found that the removal rate of the copper substrate polished by UV-CMP is increased by approximately 50–90% relative to that of traditional CMP because in UV-CMP, a passive layer on the copper surface, formed by the chemical action of the slurry, will be removed not only by the mechanical action of CMP but also by ultrasonic action. In addition, the surface roughness improves and the torque force reduces dramatically. This result suggests that the combination processes of CMP/USM and TDD/ USM are feasible methods for improving polishing and dressing efficiency. & 2011 Elsevier Ltd. All rights reserved.
Keywords: Chemical mechanical polishing Diamond disk Pad Ultrasonic machining
1. Introduction Chemical mechanical polishing (CMP) has become a primary method for the planarization of semiconductor wafers. The CMP process involves simultaneous interactions among a polishing slurry, semiconductor wafer, polishing pad, and diamond disk. CMP is achieved by polishing wafers using a rotating polyurethane pad. The slurry that permeates the pad top contains submicron-sized abrasives. Reaction chemicals will gradually consume the wafer surface. The main requirements for the CMP process are the slurry, pad, diamond conditioner, clean equipment, and end-point detector. Among these components, the slurry plays a vital role in terms of removing material and ensuring high surface quality in the polished wafers. The polishing pad considerably affects the polishing results, which include aspects such as wafer material removal and uniformity. The diamond conditioner is employed periodically (and continually) to the dressing pad to regenerate the asperity structure of the pad [1,2]. CMP processing is a major factor that determines semiconductor manufacturing costs. One disadvantage of the CMP process
n
Corresponding author. Tel.: þ866 0423924505x7195; fax: þ886 0423920681. E-mail address: [email protected] (M.-Y. Tsai).
0890-6955/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijmachtools.2011.09.009
is the high cost of its consumables. The consumables (slurry, pad, and diamond disk) used in CMP processing account for over 60% of the total CMP cost. Up to 30% of supplied slurry is wasted owing to the centrifugal force of the platen; wasted slurry is also one of the reasons for the high cost of consumables [3,4]. Therefore, consumable cost reduction has become a prime focus of researchers. In addition, it is known that CMP can produce fine surfaces but with low machining efficiency. In order to overcome this problem, many researchers have been working toward developing new methods, including abrasive-free CMP [5], fixed-abrasive CMP [6], electrochemical mechanical polishing (ECMP) [7], and chemical mechanical grinding (CMG) [8], as well as new consumables such as pads, slurries, and dressers to achieve higher polishing efficiency [9–11]. Ultrasonic machining (USM) is of particular interest for the machining of non-conductive, brittle workpiece materials such as engineering ceramics. As the process is non-chemical and nonthermal, materials are not altered either chemically or metallurgically [12]. USM has been variously termed as ultrasonic drilling, ultrasonic cutting, ultrasonic abrasive machining, and rotary ultrasonic machining. The benefits of USM include considerably increased material removal rate, decreased cutting force, reduced tool wear, and improved surface finish [13]. Currently, ultrasonic vibrations are used successfully to enhance the machining
70
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
capability of micro-electrical discharged machining (EDM) to allow it to be used on titanium alloys [14]. It has been found that in micro-hole machining of a titanium plate, micro-ultrasonic vibration lapping enhances the precision of micro-holes drilled by micro-EDM [15]. The positive effects of ultrasonic vibration have been confirmed by many researchers [16–19]. In recent years, many studies have focused on ultrasonically assisted polishing. For example, Suzuki et al. [20,21] studied ultrasonic, two-axis, vibration-assisted polishing with five-axis, piezoelectric actuators developed in order to finish molds used to fabricate lenses with high numerical aperture. Kobayashi et al. [22] introduced an ultrasonically assisted polishing technique for silicon wafer edge treatment and developed a corresponding experimental apparatus with an ultrasonic, elliptic vibration pad holder. The surface roughness of wafer edges polished by the presented method improved by over 31.7% relative to a wafer without ultrasonically assisted polishing. In this study, a copper substrate is polished by traditional CMP and UV-CMP. UV-CMP integrates the functions of traditional CMP and ultrasonic machining. In the UV-CMP system we developed, the rotating polishing head can vibrate in the vertical direction. The polishing pad is also dressed by traditional DD and UV-TDD. The removal rate of the copper substrate, torque force, and polished surface morphology of CMP and UV-CMP methods are compared. In addition, the polishing pad cut rate, torque force, and pad surface profiles of TDD and UV-TDD methods are described by experiments. The combined USM and CMP process is expected to improve the machining efficiency.
2. UV-CMP system setup The UV-CMP system and process used in the polishing experiments are shown in Fig. 1 (schematically) and Fig. 2. The system includes a USM set, moving and rotating polishing head, rotating table, slurry supply, torque dynamometer, and PC-based controller. Its operating principle is that the exciting current created by the ultrasonic generator (power supply) in the control box is transported to a piezoelectric transducer (PZT) located in the polishing head to create a high frequency vibration. The power supply for the USM is more accurately described as a high power generator that offers the user control over both the frequency and the power of the generated signal. This electrical signal is supplied to the transducer for conversion into mechanical motion. The PZT changes the electrical signal into a mechanical vibration
due to the piezoelectric effect. The mechanical vibration is amplified and transported to the polishing head by an ultrasonic technique; therefore, the rotating polishing head with the copper substrate can vibrate in the vertical direction. The polishing head is vibrated ultrasonically and rotated at the same time while being pressed against the rotating polishing pad surface at a constant pressure. Simultaneously, the slurry is provided to the interface of the copper substrate and polishing pad. Thereafter, the copper surface material is removed by the combined action of the ultrasonic vibration and traditional CMP.
3. Experimental 3.1. Dressing and polishing conditions In this experiment, the power supply converts the low frequency (approximately 50 Hz) electrical supply into a high frequency (approximately 20 kHz) AC output. This electrical signal is supplied to the transducer for conversion into mechanical motion. The amplitude of the ultrasonic vibration can be adjusted in the range of 5–15 mm. During polishing, a power head applied a specific load to a copper substrate having dimensions of F100 0.5 mm and a dispenser was used to supply the slurry. The wafer was attached to a holder that was connected to a wafer carrier head that rotated at 40 rpm. A polyurethane pad was mounted on a table disk with a diameter of 300 mm that rotated at 42 rpm. The applied pressure was approximately 27.5 kPa. The polishing slurry used in the experiments was C7092, produced by the Cabot Company. The slurry flow rate was 150 mL/min. During conditioning, the diamond pad conditioner was attached to a rotating head (spinning at 40 rpm) via a connecting holder. A polyurethane pad was mounted on a table disk that rotated at a speed of 50 rpm. The oscillation speed of the conditioner was fixed at approximately 5 mm/s. The applied load was fixed at approximately 3 kg. 3.2. UV-CMP consumable preparation The polishing pad used was IC 1000 concentric grooves, manufactured by Rohm–Haas from polyurethane-based foam. The pads are skived, or cut, from a cake. Foam density, open or closed cells, cell shape, and intrinsic polymer properties determine the final properties of the pad. Polyurethane pads contain a distribution of asperities, pores, and grooves. The pore volume
Servo motor Piezoelectric transducer
Servo motor Controller Transducer
Holder
Polishing pad
Working zone
Zoomed-in view of working
Power supply
Servo motor
Wafer
Mechanical vibration Holder Wafer Slurry
Polishing pad Polishing plate
Zoomed-in view of wafer-slurry- pad contact Fig. 1. Detailed schematic diagram of the ultrasonic-assisted chemical mechanical polishing test setup.
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
71
Fig. 2. View of the VU-CMP: (a) view of the entire machine developed, (b) vibration head, (c) polishing system, (d) torque dynamometer, and (e) controller.
Table 1 Experimental conditions. Parameters Polishing conditions Dressing conditions Ultrasonic vibration conditions Workpiece Polishing pad Polishing slurry Dressing tool
Polishing pressure (27.5 kPa), platen/carrier speed (40/42 rpm), slurry flow rate (150 ml/min) Dressing load (3 kg), diamond/pad speed (40/50 rpm), dressing liquid (D.I. water) Frequency (20 kHz), amplitude (15 mm) Copper substrate (F100 0.5 mm) IC 1000 concentric grooves C7092 (Aluminum oxide ¼ 3 wt%, pH¼ 4 by adding NH4OH) Conventional diamond pad conditioner (DiaGrid dresser)
Fig. 3. Schematics of pad dressing mechanism, D: penetrating depth and E: amount of elastic deformation.
represents approximately 35% of the total volume, but this can be varied by altering the manufacturing conditions. The isolated pores are spherical and have diameters in the range of 20–70 mm. The asperities of the pad surface are important as they determine the area of contact with the surface of the wafer. A conventional diamond pad conditioner contains single crystalline diamond grits (of length 200 mm, for example) that are attached to a metal substrate such as stainless steel either by electroplating of nickel or brazing with an alloy. The conventional diamond pad conditioner used in this experiment is the DiaGrid dresser, manufactured by the Kinik Company of Taiwan; it is bonded with metal using a brazed nickel alloy. The diameter of the brazed diamond disk was 100 mm; the spacing between diamonds was 800 mm,
and the diamond protrusion was 100 mm, as presented in Fig. 2. The wafer used was a copper substrate with a thickness of 0.5 mm, and dimensions of 100 mm. The polishing slurry used in the experiments was C7092 (Cabot Co.). The settings of the main machining parameters for the present study are summarized in Table 1. 3.3. Measurement apparatus After the pad was dressed, its surface topology was measured using a stylus-type instrument for three-dimensional surface roughness measurement. The radius of the diamond tip probe, nose angle, and measuring force were 5 mm, 901, and 4 mN,
72
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
respectively. Each image was captured with over 50 scan curves and 20,000 points over a 0.7 1.5 mm area of the pad. The roughness values used in the next section are the averages of three measured values. The measured roughness data have approximately the same reliability as the statistically averaged value of 50 measurements. A scanning electron microscope (SEM) was employed to examine the morphological features of the pad, such as pores and grooves. The pad cut was measured using a linear variable differential transformer (LVDT) instrument.
The material removal rate was determined on the basis of weight loss before and after polishing. The oxide wafers were weighed on a balance with 0.1 mg precision and their surfaces were characterized by atomic force microscopy with 0.01 nm resolution. In addition, a Laser Doppler vibration meter system was used to measure the tiny amplitude of the ultrasonic vibration.
4. Results and discussion 4.1. Pad cut rates of TDD and UV-DD during dressing
Fig. 4. Schematics image of the pad created by the penetration of the diamond grit into the pad.
140
Pad cut rate (µm/hr)
120 100 80 60 40 20 0
TDD
UV-TDD Dressing system
Fig. 5. Variation in pad cut rate between the disk and the pad when the pad is dressed by TDD and UV-TDD.
In the CMP process, the dressing action occurs between the diamond disk and the polishing pad. During polishing, the reaction product gradually accumulates in the pores and asperities of the pad surface, leading to the so-called pad glazing [23]. Glazing is undesirable as it often leads to a decreased wafer removal rate over time when there is no pad conditioning, and increases the risk of inducing micro-scratches. Therefore, the functions of the diamond disk during dressing are to clean the debris from the pores and grooves of the pad surface, to remove slurry residues and polish by-products of the pad surface, and to regenerate the asperity structure of the pad. The wear mechanism of a pad dressed using a diamond during conditioning mainly involves the abrasive wear process and can be caused as shown in Figs. 3 and 4 [23]; where D is the penetrating depth, E is the amount of elastic deformation, and V is the relative velocity. When the dressing pressure of the diamond grit exceeds the pressure required for the elastic deformation of the pad, the diamond grit can penetrate into the pad surface. The pad material begins to wear and generates a groove with ridges on both sides because of the relative motion between the diamond grit and the pad. Fig. 5 shows the variation in the pad cut rate when the pad is dressed by TDD and UV-TDD. During the dressing process, the pad cut rate was one of the most important factors that influenced the surface condition of the pad. The amount of the pad surface that is removed per unit time is the pad cut rate. The pad cut rate is measured in micrometer per hour in this study, and it is given by the pad thickness removed divided by the dressing time. In the figure, the average pad cut rate of the UV-TDD is approximately 113 mm/h, which is markedly higher than that of the TDD, which has an average pad cut rate of approximately 54 mm/h. UV-TDD leads to a better overall dressing ability in the CMP process. This may be because UV-TDD contributes to the impact force on the
Fig. 6. Torque signals between diamond disk and pad as a function of dressing time when the pad is dressed by TDD and UV-TDD.
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
soft polishing pad surface to increase the penetrating depth of the diamond grit into the pad. In addition, a representative image of the torque force signal between the diamond disk and the pad generated by TDD and UV-TDD is shown in Fig. 6. A careful analysis of the torque profile indicates that the average torque force of UV-TDD, which is approximately 156 kg mm, was lower than that of TDD, which is approximately 180 kg mm, as shown in Fig. 7. This is probably because of the high frequency impact between the diamond grit and the polishing pad surface, which causes the cutting process in UV-TDD to become discontinuous and creates an ultrasonic impact action. This process causes the material to be penetrated by the diamond grit more easily and increases the effectiveness of the interaction between grits and the workpiece surface. Therefore, the torque force effects are decreased because less plastic deformation occurs in the cutting zone [24]. Chang and Bone [25] studied ultrasonically assisted drilling of an aluminum workpiece; they recorded an average of 20% reduction in thrust force when using ultrasonically assisted drilling. This result implies that when the UV-TDD is used, it can double the pad cut rate and reduce torque force; consequently, a reduction in the dressing time of half is expected, and hence, the diamond life is extended.
4.2. Pad surface topography of DD and UV-TDD during dressing
approximately 5.05 mm for UV-TDD is less than Ra of approximately 5.42 mm for TDD as shown in Fig. 9. It is found that UV-TDD can dress the pad more uniformly, substantially reducing the amount of slurry that is required for the CMP process. Usage of the TDD forms an uneven texture with variable asperities. The cross-sectional SEM images of the pores on the polishing pad after 20 min of dressing formed by TDD and UV-TDD are shown in Fig. 10. It is observed that some residual chips of pad material are trapped in the pores dressed by TDD. On the contrary, in the UV-TDD case, the pore is shown to have no residual chips. This may be because ultrasonic vibration increases the ability of the slurry to remove chips from the pores. Ichida et al. [28] described that the cavitation phenomena occurring during the USM process can separate debris or chips from the workpiece. During cavitation, innumerable small bubbles are produced in a liquid when it is subjected to a high frequency (low wavelength) ultrasonic vibration [27,28]. Thereafter, the bubbles explode rapidly to increase the flow on the slurry interface between the pad and diamond grit. The residual chips trapped in the pores decreased the area for slurry transport, and this should be taken into account during dressing in the CMP process. 4.3. Material removal rate of CMP and UV-CMP during polishing In the polishing process of the copper CMP, material on the copper surface are removed by combined chemical reaction of slurry and mechanical action by means of the abrasive particles in the slurry that adhere to the top of the pad surface [29,30]. The copper polishing mechanism can typically be caused as follows (as shown in Fig. 11): (1) the oxidant reacts with the copper
220
6.5
200
6
180
5.5 Ra (µm)
Torque (kg-mm)
Fig. 8 displays representative images of the pad surface measured using the stylus-type instrument after 20 min of dressing by TDD or UV-TDD. In the figure, comparisons of asperities in the pad, Ra of
160
5
140
4.5
120
4
100
TDD
UV-TDD
3.5
TDD
UV-TDD Dressing system
Dressing system Fig. 7. Variation in torque force between diamond disk and pad when the pad is dressed by TDD and UV-TDD.
73
Fig. 9. Variation in surface roughness when the pad is dressed by TDD and UV-TDD.
Fig. 8. Profiles of the pad surface formed by (a) TDD and (b) UV-TDD.
74
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
Fig. 10. SEM images of an IC 1000 pad formed by TDD and UV-TDD.
Cu atoms
Cu oxide Cu passive layer Slurry particle
Slurry particle
Cu passive layer
Fig. 11. Schematic of copper polishing mechanism.
MRR ¼ CPV p=c where MRR represents the material removal rate; P, the applied pressure; Vp/c, the relative velocity; and C, the Preston coefficient that is a proportionality constant based on the efficiency of the polishing system. It can be seen that the MRR of the copper substrate is improved greatly using UV-CMP. The MRR is increased by approximately 50–90% over that of traditional CMP. Lin et al. [26] have shown that the material removal rate of the ultrasonically assisted electrical discharge machining process is higher than that of conventional electrical discharge machining. In this experiment, velocity and pressure are the same under all conditions; thus, the relationship of the Preston coefficient between the UV-CMP and CMP can be given simply by CUV CMP E1.5 1.9CCMP. This may be because, for UV-CMP, a passive layer on the copper surface formed by chemical action of the slurry will be removed not only by the mechanical action of traditional CMP but also by ultrasonic action; along with this, the abrasive particles can more easily penetrate into the polishing zone. In general, the material removal by means of ultrasonic machining is believed to be a combination of four mechanisms [13]. (1) Hammering of abrasive particles in the work surface by
35 CMP
30 MRR (mg/hr)
surface in the slurry and forms Cu2O species or CuO species. (2) Anions of a complexing agent react with the Cu2O or CuO species, forming a passivation film on the copper surface. (3) Fine particles in the slurry contact the passive layer and bond with the molecules of the passive layer. (4) When fine particles are separated from the wafer surface, they remove passive layer molecules from the copper surface. Fig. 12 shows the variation in material removal rate (MRR) for CMP and UV-CMP at two different speeds (40 and 60 rpm). For traditional CMP, the MRR varied from 12 to 16 mg/h. For UV-CMP, the MRR increased to about 18–30 mg/h. This result is consistent with the Preston equation that can be given as [31]
UV-CMP
25 20 15 10
40
60
Plarform ratational speed (rpm) Fig. 12. Variation in material removal rates for CMP and UV-CMP.
the tool, (2) the impact of free abrasive particles on the work surface, (3) cavitation erosion, and (4) chemical action associated with the fluid employed. Therefore, the ultrasonic action can both enhance mechanical action because of the impaction of abrasive particles and chemical action due to cavitation erosion to help the flow on the slurry interface between the pad and wafer and contribute to removal of the reaction product and thermal energy created by the polishing process. This result indicates that the polishing efficiency and MRR of a copper substrate obtained by UV-CMP are greatly improved relative to traditional CMP. 4.4. Cutting force and wafer surface roughness of CMP and UV-CMP during polishing The changes of torque force between the polishing pad and the copper substrate were also studied, as shown in Figs. 13 and 14.
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
The torque profile indicates that the average torque of UV-CMP, which is approximately 41 kg mm, was lower than that of CMP, which is approximately 58 kg mm. Tawakoli et al. [24] studied
70
Torque (kg-mm)
60 50 40 30 20
UV-CMP
CMP Polsihing system
Fig. 13. Variation in torque force for CMP and UV-CMP.
75
ultrasonically assisted grinding compared to conventional grinding. A decrease of up to 60% of normal grinding forces (i.e., in conventional machining) has been achieved owing to plastic deformation of the machined surface in the case of using ultrasonic oscillation. Liang et al. [32] pointed out that ultrasonically assisted grinding processes not only decrease the grinding force by 50%, but also simultaneously reduce the surface roughness by 10%. The polished surface of the copper in the cases of traditional CMP and UV-CMP are shown in Fig. 15. The average surface roughness values, Ra, of the polished copper substrates are 2.378 and 1.448 nm obtained by traditional CMP and UV-CMP, respectively, as shown in Fig. 16. One can see that the surface roughness, Ra, of the copper substrate polished by UV-CMP is lower than that of traditional CMP. The reason for this is probably that the ultrasonic vibration of the polishing pad can prevent abrasive particles within the slurry from agglomerating. Accordingly, UVCMP improves the polishing effect on the wafer surface. Choi et al. [33] studied chemical-assisted ultrasonic machining and noted that the material removal rate increased to 200% of previous values and the surface roughness improved by up to 40%. Based on the observations regarding material removal, torque force, and surface roughness made here, UV-CMP can yield a higher wafer removal rate, give better surface quality, and reduce the required
Fig. 14. Torque signals between pad and copper substrate as a function of polishing time for CMP and UV-CMP.
Fig. 15. Polished surfaces of (a) CMP and (b) UV-CMP observed by the AFM. (a) Digital Instrument NanoScope. Scan rate: 1.001 Hz. Images data: height. (b) Scan size: 5 mm 5 mm. Number of samples: 256. Resolution: 0.01 nm.
76
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
3.5 3
Ra (nm)
2.5 2 1.5 1 0.5
UV-CMP
CMP Polishing system
Fig. 16. Variation in surface roughness when the wafer is polished by CMP and UV-CMP.
torque. The results obtained can help researchers to develop more efficient polishing methods in the future.
5. Conclusions This paper presents an experimental investigation of the characteristics of a copper substrate polished by CMP and UVCMP. The dressing characteristics of the polishing pad were also studied when dressed by TDD or UV-TDD. A series of dressing and polishing experiments for a polyurethane pad and a copper substrate were performed. Some interesting research results were obtained as follows. During conditioning, comparative experiments of the torque force demonstrated up to 15% reduction for the polishing pad dressed by UV-TDD. The pad cut rate of UV-TDD is twice that of TDD. The surface roughness of the dressed polishing pad obtained using UV-TDD is 5.05 mm, which is better than that obtained using TDD (5.42 mm). In the polishing process, we found not only an increase of up to 50–90% in the material removal rate for UV-CMP relative to traditional CMP but also a decrease in torque force to 40% of its previous value. In addition, the surface roughness for UV-CMP was 1.448 nm, which was better than that recorded for traditional CMP (2.378 nm). This result indicates that UV-CMP and UV-TDD have the potential to be effectively used as wafer polishing and pad conditioning methods in the future.
Acknowledgment The authors would like to thank Kinik Company and the National Science Council of the Republic of China, Taiwan, for financially supporting this research under Contract no. NSC-982622-E-167-005-CC5. The authors would also like to express their special gratitude to the General Manager Dr. J.C. Sung and President Y.L. Pai for their valuable suggestion and comment. References [1] M.Y. Tsai, Polycrystalline diamond shaving conditioner for CMP pad conditioning, Journal of Materials Processing Technology 210 (2010). [2] M.Y. Tsai, Y.S. Liao, J.C. Sung, Novel diamond conditioner dressing characteristics of CMP polishing pad, International Journal of Machine Tools and Manufacture 49 (2009) 722–729. [3] Y.J. Seo, W.S. Lee, P. Yeh, Improvements of oxide-chemical mechanical polishing performances and aging effect of alumina and silica mixed abrasive slurries, Microelectronic Engineering 75 (2004) 361–366.
[4] T.F. Bibby, J.A. Adams, K. Holland, G.A. Krulik, P. Parikh, CMP CoO reduction: slurry reprocessing, Thin Solid Films 308–309 (1997) 538. [5] V.H. Nguyen, A.J. Hof, H.V. Kraneburg, P.H. Woerlee, F. Weimar, Copper chemical mechanical polishing using a slurry-free technique, Microelectronic Engineering 55 (2001) 305–312. [6] P. van der Velden, Chemical mechanical polishing with fixed abrasive using different subpads to optimize wafer uniformity, Microelectronic Engineering 20 (2000) 41–46. [7] S. Jeong, S. Lee, H. Jeong, Effect of polishing pad with holes in electrochemical mechanical planarization, Microelectronic Engineering 85 (2008) 2236–2243. [8] S. Kamiya, H. Iwase, T. Nagaike, L. Zhou, H. Eda, S. Kimura, Microstructural analysis for Si wafer after CMG process, Key Engineering Materials 329 (2007) 361–366. [9] H. Kim, H. Kim, H. Jeong, H. Seo, S. Lee, Self-conditioning of encapsulated abrasive pad in chemical mechanical polishing, Journal of Materials Processing Technology 142 (2003) 614–618. [10] V.H. Nguyen, A.J. Hof, H.V. Kraneburg, P.H. Woerlee, F. Weimar, Copper chemical mechanical polishing using a slurry-free technique, Microelectronic Engineering 55 (2001) 305–312. [11] M.Y. Tsai, L.W. Yan, Characteristics of chemical mechanical polishing using graphite impregnated pad, International Journal of Machine Tools and Manufacture 50 (2010) 1031–1037. [12] R. Singh, J.S. Khamba, Ultrasonic machining of titanium and its alloys: a review, Journal of Materials Processing Technology 173 (2006) 125–135. [13] E.J. Weller, Nontraditional Machining Processes, Society of Manufacturing Engineering, Dearborn, Michigan, USA, 1983. [14] A.C. Wang, B.H. Yan, X.T. Li, F.Y. Huang, Use of micro ultrasonic vibration lapping to enhance the precision of microholes drilled by micro electrodischarge machining, International Journal of Machine Tools and Manufacture 42 (8) (2002) 915–923. [15] Z.C. Li, Y. Jiao, T.W. Deines, Z.J. Pei, C. Treadwell, Rotary ultrasonic machining of ceramic matrix composities: feasibility study and designed experiments, International Journal of Machine Tools and Manufacture 45 (1–13) (2005) 1402–1411. [16] H. Hocheng, K.L. Kuo, Fundamental study of ultrasonic polishing of mold steel, International Journal of Machine Tools and Manufacture 42 (2002) 7–13. [17] T.B. Thoe, D.K. Aspinwall, N. Killey, Combined ultrasonic and electrical discharge machining of ceramic coated nickel alloy, Journal of Materials Processing Technology 92–93 (1999) 323–328. [18] C. Zhang, R. Rentsch, E. Brinksmeier, Advances in micro ultraonic assisted lapping of microstructures in hard-brittle materials: a brief review and out look, International Journal of Machine Tools and Manufacture 45 (2005) 881–890. [19] K. Suzuki, T. Uematsu, M. Lwai, S. Ninomiya, S. Sano, T. Nakagawa, A New Complex grinding method for ceramic materials combined with ultrasonic vibration and electrodischarge machining, Key Engineering Materials 329 (2007) 439–444. [20] H. Suzuki, S. Hamada, T. Oondo, Y. Yamagata, T. Higuchi, Ultraprecision finishing of micro-aspheric surface by ultrasonic two-axis vibration assisted polishing, CIRP Annals—Manufacturing Technology 59 (2010) 347–350. [21] H. Suzuki, T. Moriwaki, T. Okino, Y. Ando, Development of ultrasonic vibration assisted polishing machine for micro aspheric die and mold, CIRP Annals—Manufacturing Technology 55 (2006) 385–388. [22] N. Kobayashi, Y. Wu, M. Nomura, T. Sato, Precision treatment of silicon wafer edge utilizing ultrasonically assisted polishing technique, Journal of Materials Processing Technology 201 (2008) 531–535. [23] M.Y. Tsai, Y.S. Liao, Dressing characteristics of oriented single diamond on CMP polyurethane pad, Machining Science and Technology 13 (2009) 92–105. [24] T. Tawakoli, B. Azarhoushang, M. Rabiey, Ultrasonic assisted dry grinding of 42CrMo4, International Journal of Advanced Manufacturing Technology 42 (2009) 883–891. [25] S.F. Chang, G.M. Bone, Burr size reduction in drilling by ultrasonic assistance, Robotics and Computer-Integrated Manufacturing 21 (2005) 442–450. [26] Y.C. Lin, B.H. Yan, Y.S. Chang, Machining characteristics of titanium alloy (Ti–6Al–4V) using a combination process of EDM with USM, Journal of Materials Processing Technology 104 (2000) 171–177. [27] Z. Liang, G. Zhou, S. Lin, Y. Zhang, H. Yang, Study of low-frequency ultrasonic cavitation fields based on spectral analysis technique, Ultrasonic 44 (2006) 115–120. [28] Y. Ichida, R. Sato, Y. Morimoto, K. Kobayashi, Material removal mechanisms in non-contact ultrasonic abrasive machining, Wear 258 (2005) 107–114. [29] Y. Wang, Y. Zhao, Modeling the effects of oxidizer, complexing agent and inhibitor on material removal for copper chemical mechanical polishing, Applied Surface Science 254 (2007) 1517–1523. [30] H. Lee, S. Joo, H. Jeong, Mechanical effect of colloidal silica in copper chemical mechanical planarization, Journal of Materials Processing Technology 209 (2009) 6134–6139. [31] J.M. Steigerwald, S.P.M. Urarka, R.J. Gutmann, Chemical Mechanical Planarization of Microelectronic Materials, John Wiley & Sons Inc. Press, New York, 1997. [32] Z. Liang, Y. Wu, X. Wang, W. Zhao, A new two-dimensional ultrasonic assisted grinding method and its fundamental performance in monocrystal silicon machining, International Journal of Machine Tools and Manufacture 50 (2010) 728–736. [33] J.P. Choi, B.H. Jeon, B.H. Kim, Chemical-assisted ultrasonic machining of glass, Journal of Materials Processing Technology 191 (2007) 153–156.
Contents lists available at SciVerse ScienceDirect
International Journal of Machine Tools & Manufacture journal homepage: www.elsevier.com/locate/ijmactool
Combined ultrasonic vibration and chemical mechanical polishing of copper substrates Ming-Yi Tsai n, Wei-Zheng Yang Department of Mechanical Engineering, National Chin-Yi University of Technology, No. 35, Lane 215, Section 1, Chung-Shan Rd., Taiping City, Taichung County 411, Taiwan
a r t i c l e i n f o
abstract
Article history: Received 29 December 2010 Received in revised form 17 September 2011 Accepted 20 September 2011 Available online 4 October 2011
This paper introduces an ultrasonic, vibration-assisted, chemical mechanical polishing (UV-CMP) method and an ultrasonic, vibration-assisted, traditional diamond disk (UV-TDD) dressing method. A copper substrate is polished by traditional CMP and UV-CMP. UV-CMP combines the functions of traditional CMP and ultrasonic machining (USM) with small-amplitude, high-frequency tool vibration to improve the fabrication process and machining efficiency. The removal rate of the copper substrate, torque force, and polished surface morphology of CMP and UV-CMP are compared. The polishing pad is also dressed by traditional diamond disk (TDD) and UV-TDD. The pad cut rate, torque force, and pad surface profiles of TDD and UV-TDD are also investigated in experiments. Experimental results reveal that UV-TDD can produce twice the pad cut rate and reduce torque force compared to TDD. Consequently, a dressing time reduction by half is expected, and hence, the diamond life is extended. It is found that the removal rate of the copper substrate polished by UV-CMP is increased by approximately 50–90% relative to that of traditional CMP because in UV-CMP, a passive layer on the copper surface, formed by the chemical action of the slurry, will be removed not only by the mechanical action of CMP but also by ultrasonic action. In addition, the surface roughness improves and the torque force reduces dramatically. This result suggests that the combination processes of CMP/USM and TDD/ USM are feasible methods for improving polishing and dressing efficiency. & 2011 Elsevier Ltd. All rights reserved.
Keywords: Chemical mechanical polishing Diamond disk Pad Ultrasonic machining
1. Introduction Chemical mechanical polishing (CMP) has become a primary method for the planarization of semiconductor wafers. The CMP process involves simultaneous interactions among a polishing slurry, semiconductor wafer, polishing pad, and diamond disk. CMP is achieved by polishing wafers using a rotating polyurethane pad. The slurry that permeates the pad top contains submicron-sized abrasives. Reaction chemicals will gradually consume the wafer surface. The main requirements for the CMP process are the slurry, pad, diamond conditioner, clean equipment, and end-point detector. Among these components, the slurry plays a vital role in terms of removing material and ensuring high surface quality in the polished wafers. The polishing pad considerably affects the polishing results, which include aspects such as wafer material removal and uniformity. The diamond conditioner is employed periodically (and continually) to the dressing pad to regenerate the asperity structure of the pad [1,2]. CMP processing is a major factor that determines semiconductor manufacturing costs. One disadvantage of the CMP process
n
Corresponding author. Tel.: þ866 0423924505x7195; fax: þ886 0423920681. E-mail address: [email protected] (M.-Y. Tsai).
0890-6955/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijmachtools.2011.09.009
is the high cost of its consumables. The consumables (slurry, pad, and diamond disk) used in CMP processing account for over 60% of the total CMP cost. Up to 30% of supplied slurry is wasted owing to the centrifugal force of the platen; wasted slurry is also one of the reasons for the high cost of consumables [3,4]. Therefore, consumable cost reduction has become a prime focus of researchers. In addition, it is known that CMP can produce fine surfaces but with low machining efficiency. In order to overcome this problem, many researchers have been working toward developing new methods, including abrasive-free CMP [5], fixed-abrasive CMP [6], electrochemical mechanical polishing (ECMP) [7], and chemical mechanical grinding (CMG) [8], as well as new consumables such as pads, slurries, and dressers to achieve higher polishing efficiency [9–11]. Ultrasonic machining (USM) is of particular interest for the machining of non-conductive, brittle workpiece materials such as engineering ceramics. As the process is non-chemical and nonthermal, materials are not altered either chemically or metallurgically [12]. USM has been variously termed as ultrasonic drilling, ultrasonic cutting, ultrasonic abrasive machining, and rotary ultrasonic machining. The benefits of USM include considerably increased material removal rate, decreased cutting force, reduced tool wear, and improved surface finish [13]. Currently, ultrasonic vibrations are used successfully to enhance the machining
70
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
capability of micro-electrical discharged machining (EDM) to allow it to be used on titanium alloys [14]. It has been found that in micro-hole machining of a titanium plate, micro-ultrasonic vibration lapping enhances the precision of micro-holes drilled by micro-EDM [15]. The positive effects of ultrasonic vibration have been confirmed by many researchers [16–19]. In recent years, many studies have focused on ultrasonically assisted polishing. For example, Suzuki et al. [20,21] studied ultrasonic, two-axis, vibration-assisted polishing with five-axis, piezoelectric actuators developed in order to finish molds used to fabricate lenses with high numerical aperture. Kobayashi et al. [22] introduced an ultrasonically assisted polishing technique for silicon wafer edge treatment and developed a corresponding experimental apparatus with an ultrasonic, elliptic vibration pad holder. The surface roughness of wafer edges polished by the presented method improved by over 31.7% relative to a wafer without ultrasonically assisted polishing. In this study, a copper substrate is polished by traditional CMP and UV-CMP. UV-CMP integrates the functions of traditional CMP and ultrasonic machining. In the UV-CMP system we developed, the rotating polishing head can vibrate in the vertical direction. The polishing pad is also dressed by traditional DD and UV-TDD. The removal rate of the copper substrate, torque force, and polished surface morphology of CMP and UV-CMP methods are compared. In addition, the polishing pad cut rate, torque force, and pad surface profiles of TDD and UV-TDD methods are described by experiments. The combined USM and CMP process is expected to improve the machining efficiency.
2. UV-CMP system setup The UV-CMP system and process used in the polishing experiments are shown in Fig. 1 (schematically) and Fig. 2. The system includes a USM set, moving and rotating polishing head, rotating table, slurry supply, torque dynamometer, and PC-based controller. Its operating principle is that the exciting current created by the ultrasonic generator (power supply) in the control box is transported to a piezoelectric transducer (PZT) located in the polishing head to create a high frequency vibration. The power supply for the USM is more accurately described as a high power generator that offers the user control over both the frequency and the power of the generated signal. This electrical signal is supplied to the transducer for conversion into mechanical motion. The PZT changes the electrical signal into a mechanical vibration
due to the piezoelectric effect. The mechanical vibration is amplified and transported to the polishing head by an ultrasonic technique; therefore, the rotating polishing head with the copper substrate can vibrate in the vertical direction. The polishing head is vibrated ultrasonically and rotated at the same time while being pressed against the rotating polishing pad surface at a constant pressure. Simultaneously, the slurry is provided to the interface of the copper substrate and polishing pad. Thereafter, the copper surface material is removed by the combined action of the ultrasonic vibration and traditional CMP.
3. Experimental 3.1. Dressing and polishing conditions In this experiment, the power supply converts the low frequency (approximately 50 Hz) electrical supply into a high frequency (approximately 20 kHz) AC output. This electrical signal is supplied to the transducer for conversion into mechanical motion. The amplitude of the ultrasonic vibration can be adjusted in the range of 5–15 mm. During polishing, a power head applied a specific load to a copper substrate having dimensions of F100 0.5 mm and a dispenser was used to supply the slurry. The wafer was attached to a holder that was connected to a wafer carrier head that rotated at 40 rpm. A polyurethane pad was mounted on a table disk with a diameter of 300 mm that rotated at 42 rpm. The applied pressure was approximately 27.5 kPa. The polishing slurry used in the experiments was C7092, produced by the Cabot Company. The slurry flow rate was 150 mL/min. During conditioning, the diamond pad conditioner was attached to a rotating head (spinning at 40 rpm) via a connecting holder. A polyurethane pad was mounted on a table disk that rotated at a speed of 50 rpm. The oscillation speed of the conditioner was fixed at approximately 5 mm/s. The applied load was fixed at approximately 3 kg. 3.2. UV-CMP consumable preparation The polishing pad used was IC 1000 concentric grooves, manufactured by Rohm–Haas from polyurethane-based foam. The pads are skived, or cut, from a cake. Foam density, open or closed cells, cell shape, and intrinsic polymer properties determine the final properties of the pad. Polyurethane pads contain a distribution of asperities, pores, and grooves. The pore volume
Servo motor Piezoelectric transducer
Servo motor Controller Transducer
Holder
Polishing pad
Working zone
Zoomed-in view of working
Power supply
Servo motor
Wafer
Mechanical vibration Holder Wafer Slurry
Polishing pad Polishing plate
Zoomed-in view of wafer-slurry- pad contact Fig. 1. Detailed schematic diagram of the ultrasonic-assisted chemical mechanical polishing test setup.
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
71
Fig. 2. View of the VU-CMP: (a) view of the entire machine developed, (b) vibration head, (c) polishing system, (d) torque dynamometer, and (e) controller.
Table 1 Experimental conditions. Parameters Polishing conditions Dressing conditions Ultrasonic vibration conditions Workpiece Polishing pad Polishing slurry Dressing tool
Polishing pressure (27.5 kPa), platen/carrier speed (40/42 rpm), slurry flow rate (150 ml/min) Dressing load (3 kg), diamond/pad speed (40/50 rpm), dressing liquid (D.I. water) Frequency (20 kHz), amplitude (15 mm) Copper substrate (F100 0.5 mm) IC 1000 concentric grooves C7092 (Aluminum oxide ¼ 3 wt%, pH¼ 4 by adding NH4OH) Conventional diamond pad conditioner (DiaGrid dresser)
Fig. 3. Schematics of pad dressing mechanism, D: penetrating depth and E: amount of elastic deformation.
represents approximately 35% of the total volume, but this can be varied by altering the manufacturing conditions. The isolated pores are spherical and have diameters in the range of 20–70 mm. The asperities of the pad surface are important as they determine the area of contact with the surface of the wafer. A conventional diamond pad conditioner contains single crystalline diamond grits (of length 200 mm, for example) that are attached to a metal substrate such as stainless steel either by electroplating of nickel or brazing with an alloy. The conventional diamond pad conditioner used in this experiment is the DiaGrid dresser, manufactured by the Kinik Company of Taiwan; it is bonded with metal using a brazed nickel alloy. The diameter of the brazed diamond disk was 100 mm; the spacing between diamonds was 800 mm,
and the diamond protrusion was 100 mm, as presented in Fig. 2. The wafer used was a copper substrate with a thickness of 0.5 mm, and dimensions of 100 mm. The polishing slurry used in the experiments was C7092 (Cabot Co.). The settings of the main machining parameters for the present study are summarized in Table 1. 3.3. Measurement apparatus After the pad was dressed, its surface topology was measured using a stylus-type instrument for three-dimensional surface roughness measurement. The radius of the diamond tip probe, nose angle, and measuring force were 5 mm, 901, and 4 mN,
72
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
respectively. Each image was captured with over 50 scan curves and 20,000 points over a 0.7 1.5 mm area of the pad. The roughness values used in the next section are the averages of three measured values. The measured roughness data have approximately the same reliability as the statistically averaged value of 50 measurements. A scanning electron microscope (SEM) was employed to examine the morphological features of the pad, such as pores and grooves. The pad cut was measured using a linear variable differential transformer (LVDT) instrument.
The material removal rate was determined on the basis of weight loss before and after polishing. The oxide wafers were weighed on a balance with 0.1 mg precision and their surfaces were characterized by atomic force microscopy with 0.01 nm resolution. In addition, a Laser Doppler vibration meter system was used to measure the tiny amplitude of the ultrasonic vibration.
4. Results and discussion 4.1. Pad cut rates of TDD and UV-DD during dressing
Fig. 4. Schematics image of the pad created by the penetration of the diamond grit into the pad.
140
Pad cut rate (µm/hr)
120 100 80 60 40 20 0
TDD
UV-TDD Dressing system
Fig. 5. Variation in pad cut rate between the disk and the pad when the pad is dressed by TDD and UV-TDD.
In the CMP process, the dressing action occurs between the diamond disk and the polishing pad. During polishing, the reaction product gradually accumulates in the pores and asperities of the pad surface, leading to the so-called pad glazing [23]. Glazing is undesirable as it often leads to a decreased wafer removal rate over time when there is no pad conditioning, and increases the risk of inducing micro-scratches. Therefore, the functions of the diamond disk during dressing are to clean the debris from the pores and grooves of the pad surface, to remove slurry residues and polish by-products of the pad surface, and to regenerate the asperity structure of the pad. The wear mechanism of a pad dressed using a diamond during conditioning mainly involves the abrasive wear process and can be caused as shown in Figs. 3 and 4 [23]; where D is the penetrating depth, E is the amount of elastic deformation, and V is the relative velocity. When the dressing pressure of the diamond grit exceeds the pressure required for the elastic deformation of the pad, the diamond grit can penetrate into the pad surface. The pad material begins to wear and generates a groove with ridges on both sides because of the relative motion between the diamond grit and the pad. Fig. 5 shows the variation in the pad cut rate when the pad is dressed by TDD and UV-TDD. During the dressing process, the pad cut rate was one of the most important factors that influenced the surface condition of the pad. The amount of the pad surface that is removed per unit time is the pad cut rate. The pad cut rate is measured in micrometer per hour in this study, and it is given by the pad thickness removed divided by the dressing time. In the figure, the average pad cut rate of the UV-TDD is approximately 113 mm/h, which is markedly higher than that of the TDD, which has an average pad cut rate of approximately 54 mm/h. UV-TDD leads to a better overall dressing ability in the CMP process. This may be because UV-TDD contributes to the impact force on the
Fig. 6. Torque signals between diamond disk and pad as a function of dressing time when the pad is dressed by TDD and UV-TDD.
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
soft polishing pad surface to increase the penetrating depth of the diamond grit into the pad. In addition, a representative image of the torque force signal between the diamond disk and the pad generated by TDD and UV-TDD is shown in Fig. 6. A careful analysis of the torque profile indicates that the average torque force of UV-TDD, which is approximately 156 kg mm, was lower than that of TDD, which is approximately 180 kg mm, as shown in Fig. 7. This is probably because of the high frequency impact between the diamond grit and the polishing pad surface, which causes the cutting process in UV-TDD to become discontinuous and creates an ultrasonic impact action. This process causes the material to be penetrated by the diamond grit more easily and increases the effectiveness of the interaction between grits and the workpiece surface. Therefore, the torque force effects are decreased because less plastic deformation occurs in the cutting zone [24]. Chang and Bone [25] studied ultrasonically assisted drilling of an aluminum workpiece; they recorded an average of 20% reduction in thrust force when using ultrasonically assisted drilling. This result implies that when the UV-TDD is used, it can double the pad cut rate and reduce torque force; consequently, a reduction in the dressing time of half is expected, and hence, the diamond life is extended.
4.2. Pad surface topography of DD and UV-TDD during dressing
approximately 5.05 mm for UV-TDD is less than Ra of approximately 5.42 mm for TDD as shown in Fig. 9. It is found that UV-TDD can dress the pad more uniformly, substantially reducing the amount of slurry that is required for the CMP process. Usage of the TDD forms an uneven texture with variable asperities. The cross-sectional SEM images of the pores on the polishing pad after 20 min of dressing formed by TDD and UV-TDD are shown in Fig. 10. It is observed that some residual chips of pad material are trapped in the pores dressed by TDD. On the contrary, in the UV-TDD case, the pore is shown to have no residual chips. This may be because ultrasonic vibration increases the ability of the slurry to remove chips from the pores. Ichida et al. [28] described that the cavitation phenomena occurring during the USM process can separate debris or chips from the workpiece. During cavitation, innumerable small bubbles are produced in a liquid when it is subjected to a high frequency (low wavelength) ultrasonic vibration [27,28]. Thereafter, the bubbles explode rapidly to increase the flow on the slurry interface between the pad and diamond grit. The residual chips trapped in the pores decreased the area for slurry transport, and this should be taken into account during dressing in the CMP process. 4.3. Material removal rate of CMP and UV-CMP during polishing In the polishing process of the copper CMP, material on the copper surface are removed by combined chemical reaction of slurry and mechanical action by means of the abrasive particles in the slurry that adhere to the top of the pad surface [29,30]. The copper polishing mechanism can typically be caused as follows (as shown in Fig. 11): (1) the oxidant reacts with the copper
220
6.5
200
6
180
5.5 Ra (µm)
Torque (kg-mm)
Fig. 8 displays representative images of the pad surface measured using the stylus-type instrument after 20 min of dressing by TDD or UV-TDD. In the figure, comparisons of asperities in the pad, Ra of
160
5
140
4.5
120
4
100
TDD
UV-TDD
3.5
TDD
UV-TDD Dressing system
Dressing system Fig. 7. Variation in torque force between diamond disk and pad when the pad is dressed by TDD and UV-TDD.
73
Fig. 9. Variation in surface roughness when the pad is dressed by TDD and UV-TDD.
Fig. 8. Profiles of the pad surface formed by (a) TDD and (b) UV-TDD.
74
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
Fig. 10. SEM images of an IC 1000 pad formed by TDD and UV-TDD.
Cu atoms
Cu oxide Cu passive layer Slurry particle
Slurry particle
Cu passive layer
Fig. 11. Schematic of copper polishing mechanism.
MRR ¼ CPV p=c where MRR represents the material removal rate; P, the applied pressure; Vp/c, the relative velocity; and C, the Preston coefficient that is a proportionality constant based on the efficiency of the polishing system. It can be seen that the MRR of the copper substrate is improved greatly using UV-CMP. The MRR is increased by approximately 50–90% over that of traditional CMP. Lin et al. [26] have shown that the material removal rate of the ultrasonically assisted electrical discharge machining process is higher than that of conventional electrical discharge machining. In this experiment, velocity and pressure are the same under all conditions; thus, the relationship of the Preston coefficient between the UV-CMP and CMP can be given simply by CUV CMP E1.5 1.9CCMP. This may be because, for UV-CMP, a passive layer on the copper surface formed by chemical action of the slurry will be removed not only by the mechanical action of traditional CMP but also by ultrasonic action; along with this, the abrasive particles can more easily penetrate into the polishing zone. In general, the material removal by means of ultrasonic machining is believed to be a combination of four mechanisms [13]. (1) Hammering of abrasive particles in the work surface by
35 CMP
30 MRR (mg/hr)
surface in the slurry and forms Cu2O species or CuO species. (2) Anions of a complexing agent react with the Cu2O or CuO species, forming a passivation film on the copper surface. (3) Fine particles in the slurry contact the passive layer and bond with the molecules of the passive layer. (4) When fine particles are separated from the wafer surface, they remove passive layer molecules from the copper surface. Fig. 12 shows the variation in material removal rate (MRR) for CMP and UV-CMP at two different speeds (40 and 60 rpm). For traditional CMP, the MRR varied from 12 to 16 mg/h. For UV-CMP, the MRR increased to about 18–30 mg/h. This result is consistent with the Preston equation that can be given as [31]
UV-CMP
25 20 15 10
40
60
Plarform ratational speed (rpm) Fig. 12. Variation in material removal rates for CMP and UV-CMP.
the tool, (2) the impact of free abrasive particles on the work surface, (3) cavitation erosion, and (4) chemical action associated with the fluid employed. Therefore, the ultrasonic action can both enhance mechanical action because of the impaction of abrasive particles and chemical action due to cavitation erosion to help the flow on the slurry interface between the pad and wafer and contribute to removal of the reaction product and thermal energy created by the polishing process. This result indicates that the polishing efficiency and MRR of a copper substrate obtained by UV-CMP are greatly improved relative to traditional CMP. 4.4. Cutting force and wafer surface roughness of CMP and UV-CMP during polishing The changes of torque force between the polishing pad and the copper substrate were also studied, as shown in Figs. 13 and 14.
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
The torque profile indicates that the average torque of UV-CMP, which is approximately 41 kg mm, was lower than that of CMP, which is approximately 58 kg mm. Tawakoli et al. [24] studied
70
Torque (kg-mm)
60 50 40 30 20
UV-CMP
CMP Polsihing system
Fig. 13. Variation in torque force for CMP and UV-CMP.
75
ultrasonically assisted grinding compared to conventional grinding. A decrease of up to 60% of normal grinding forces (i.e., in conventional machining) has been achieved owing to plastic deformation of the machined surface in the case of using ultrasonic oscillation. Liang et al. [32] pointed out that ultrasonically assisted grinding processes not only decrease the grinding force by 50%, but also simultaneously reduce the surface roughness by 10%. The polished surface of the copper in the cases of traditional CMP and UV-CMP are shown in Fig. 15. The average surface roughness values, Ra, of the polished copper substrates are 2.378 and 1.448 nm obtained by traditional CMP and UV-CMP, respectively, as shown in Fig. 16. One can see that the surface roughness, Ra, of the copper substrate polished by UV-CMP is lower than that of traditional CMP. The reason for this is probably that the ultrasonic vibration of the polishing pad can prevent abrasive particles within the slurry from agglomerating. Accordingly, UVCMP improves the polishing effect on the wafer surface. Choi et al. [33] studied chemical-assisted ultrasonic machining and noted that the material removal rate increased to 200% of previous values and the surface roughness improved by up to 40%. Based on the observations regarding material removal, torque force, and surface roughness made here, UV-CMP can yield a higher wafer removal rate, give better surface quality, and reduce the required
Fig. 14. Torque signals between pad and copper substrate as a function of polishing time for CMP and UV-CMP.
Fig. 15. Polished surfaces of (a) CMP and (b) UV-CMP observed by the AFM. (a) Digital Instrument NanoScope. Scan rate: 1.001 Hz. Images data: height. (b) Scan size: 5 mm 5 mm. Number of samples: 256. Resolution: 0.01 nm.
76
M.-Y. Tsai, W.-Z. Yang / International Journal of Machine Tools & Manufacture 53 (2012) 69–76
3.5 3
Ra (nm)
2.5 2 1.5 1 0.5
UV-CMP
CMP Polishing system
Fig. 16. Variation in surface roughness when the wafer is polished by CMP and UV-CMP.
torque. The results obtained can help researchers to develop more efficient polishing methods in the future.
5. Conclusions This paper presents an experimental investigation of the characteristics of a copper substrate polished by CMP and UVCMP. The dressing characteristics of the polishing pad were also studied when dressed by TDD or UV-TDD. A series of dressing and polishing experiments for a polyurethane pad and a copper substrate were performed. Some interesting research results were obtained as follows. During conditioning, comparative experiments of the torque force demonstrated up to 15% reduction for the polishing pad dressed by UV-TDD. The pad cut rate of UV-TDD is twice that of TDD. The surface roughness of the dressed polishing pad obtained using UV-TDD is 5.05 mm, which is better than that obtained using TDD (5.42 mm). In the polishing process, we found not only an increase of up to 50–90% in the material removal rate for UV-CMP relative to traditional CMP but also a decrease in torque force to 40% of its previous value. In addition, the surface roughness for UV-CMP was 1.448 nm, which was better than that recorded for traditional CMP (2.378 nm). This result indicates that UV-CMP and UV-TDD have the potential to be effectively used as wafer polishing and pad conditioning methods in the future.
Acknowledgment The authors would like to thank Kinik Company and the National Science Council of the Republic of China, Taiwan, for financially supporting this research under Contract no. NSC-982622-E-167-005-CC5. The authors would also like to express their special gratitude to the General Manager Dr. J.C. Sung and President Y.L. Pai for their valuable suggestion and comment. References [1] M.Y. Tsai, Polycrystalline diamond shaving conditioner for CMP pad conditioning, Journal of Materials Processing Technology 210 (2010). [2] M.Y. Tsai, Y.S. Liao, J.C. Sung, Novel diamond conditioner dressing characteristics of CMP polishing pad, International Journal of Machine Tools and Manufacture 49 (2009) 722–729. [3] Y.J. Seo, W.S. Lee, P. Yeh, Improvements of oxide-chemical mechanical polishing performances and aging effect of alumina and silica mixed abrasive slurries, Microelectronic Engineering 75 (2004) 361–366.
[4] T.F. Bibby, J.A. Adams, K. Holland, G.A. Krulik, P. Parikh, CMP CoO reduction: slurry reprocessing, Thin Solid Films 308–309 (1997) 538. [5] V.H. Nguyen, A.J. Hof, H.V. Kraneburg, P.H. Woerlee, F. Weimar, Copper chemical mechanical polishing using a slurry-free technique, Microelectronic Engineering 55 (2001) 305–312. [6] P. van der Velden, Chemical mechanical polishing with fixed abrasive using different subpads to optimize wafer uniformity, Microelectronic Engineering 20 (2000) 41–46. [7] S. Jeong, S. Lee, H. Jeong, Effect of polishing pad with holes in electrochemical mechanical planarization, Microelectronic Engineering 85 (2008) 2236–2243. [8] S. Kamiya, H. Iwase, T. Nagaike, L. Zhou, H. Eda, S. Kimura, Microstructural analysis for Si wafer after CMG process, Key Engineering Materials 329 (2007) 361–366. [9] H. Kim, H. Kim, H. Jeong, H. Seo, S. Lee, Self-conditioning of encapsulated abrasive pad in chemical mechanical polishing, Journal of Materials Processing Technology 142 (2003) 614–618. [10] V.H. Nguyen, A.J. Hof, H.V. Kraneburg, P.H. Woerlee, F. Weimar, Copper chemical mechanical polishing using a slurry-free technique, Microelectronic Engineering 55 (2001) 305–312. [11] M.Y. Tsai, L.W. Yan, Characteristics of chemical mechanical polishing using graphite impregnated pad, International Journal of Machine Tools and Manufacture 50 (2010) 1031–1037. [12] R. Singh, J.S. Khamba, Ultrasonic machining of titanium and its alloys: a review, Journal of Materials Processing Technology 173 (2006) 125–135. [13] E.J. Weller, Nontraditional Machining Processes, Society of Manufacturing Engineering, Dearborn, Michigan, USA, 1983. [14] A.C. Wang, B.H. Yan, X.T. Li, F.Y. Huang, Use of micro ultrasonic vibration lapping to enhance the precision of microholes drilled by micro electrodischarge machining, International Journal of Machine Tools and Manufacture 42 (8) (2002) 915–923. [15] Z.C. Li, Y. Jiao, T.W. Deines, Z.J. Pei, C. Treadwell, Rotary ultrasonic machining of ceramic matrix composities: feasibility study and designed experiments, International Journal of Machine Tools and Manufacture 45 (1–13) (2005) 1402–1411. [16] H. Hocheng, K.L. Kuo, Fundamental study of ultrasonic polishing of mold steel, International Journal of Machine Tools and Manufacture 42 (2002) 7–13. [17] T.B. Thoe, D.K. Aspinwall, N. Killey, Combined ultrasonic and electrical discharge machining of ceramic coated nickel alloy, Journal of Materials Processing Technology 92–93 (1999) 323–328. [18] C. Zhang, R. Rentsch, E. Brinksmeier, Advances in micro ultraonic assisted lapping of microstructures in hard-brittle materials: a brief review and out look, International Journal of Machine Tools and Manufacture 45 (2005) 881–890. [19] K. Suzuki, T. Uematsu, M. Lwai, S. Ninomiya, S. Sano, T. Nakagawa, A New Complex grinding method for ceramic materials combined with ultrasonic vibration and electrodischarge machining, Key Engineering Materials 329 (2007) 439–444. [20] H. Suzuki, S. Hamada, T. Oondo, Y. Yamagata, T. Higuchi, Ultraprecision finishing of micro-aspheric surface by ultrasonic two-axis vibration assisted polishing, CIRP Annals—Manufacturing Technology 59 (2010) 347–350. [21] H. Suzuki, T. Moriwaki, T. Okino, Y. Ando, Development of ultrasonic vibration assisted polishing machine for micro aspheric die and mold, CIRP Annals—Manufacturing Technology 55 (2006) 385–388. [22] N. Kobayashi, Y. Wu, M. Nomura, T. Sato, Precision treatment of silicon wafer edge utilizing ultrasonically assisted polishing technique, Journal of Materials Processing Technology 201 (2008) 531–535. [23] M.Y. Tsai, Y.S. Liao, Dressing characteristics of oriented single diamond on CMP polyurethane pad, Machining Science and Technology 13 (2009) 92–105. [24] T. Tawakoli, B. Azarhoushang, M. Rabiey, Ultrasonic assisted dry grinding of 42CrMo4, International Journal of Advanced Manufacturing Technology 42 (2009) 883–891. [25] S.F. Chang, G.M. Bone, Burr size reduction in drilling by ultrasonic assistance, Robotics and Computer-Integrated Manufacturing 21 (2005) 442–450. [26] Y.C. Lin, B.H. Yan, Y.S. Chang, Machining characteristics of titanium alloy (Ti–6Al–4V) using a combination process of EDM with USM, Journal of Materials Processing Technology 104 (2000) 171–177. [27] Z. Liang, G. Zhou, S. Lin, Y. Zhang, H. Yang, Study of low-frequency ultrasonic cavitation fields based on spectral analysis technique, Ultrasonic 44 (2006) 115–120. [28] Y. Ichida, R. Sato, Y. Morimoto, K. Kobayashi, Material removal mechanisms in non-contact ultrasonic abrasive machining, Wear 258 (2005) 107–114. [29] Y. Wang, Y. Zhao, Modeling the effects of oxidizer, complexing agent and inhibitor on material removal for copper chemical mechanical polishing, Applied Surface Science 254 (2007) 1517–1523. [30] H. Lee, S. Joo, H. Jeong, Mechanical effect of colloidal silica in copper chemical mechanical planarization, Journal of Materials Processing Technology 209 (2009) 6134–6139. [31] J.M. Steigerwald, S.P.M. Urarka, R.J. Gutmann, Chemical Mechanical Planarization of Microelectronic Materials, John Wiley & Sons Inc. Press, New York, 1997. [32] Z. Liang, Y. Wu, X. Wang, W. Zhao, A new two-dimensional ultrasonic assisted grinding method and its fundamental performance in monocrystal silicon machining, International Journal of Machine Tools and Manufacture 50 (2010) 728–736. [33] J.P. Choi, B.H. Jeon, B.H. Kim, Chemical-assisted ultrasonic machining of glass, Journal of Materials Processing Technology 191 (2007) 153–156.