Dynamic tuning of chemical-mechanical planarization operation via sliding-mode theory

(1)

Dynamic tuning of chemical–mechanical

planarization operation via sliding-mode theory

Chia-Shui Lin

*

_{, Chien-Yu Chi}

Department of Mechanical Engineering, National Chiao Tung University, 1001 Ta Hsueh Road, Hsin-Chu 30050, Taiwan Received 27 November 2003; received in revised form 23 April 2004; accepted 26 May 2004

Available online 10 July 2004

Abstract

The development of chemical–mechanical planarization (CMP) to date relies on multi-stage polishing for better per-formance. The straight relationship between performance and operation provides a new method of control for CMP process. The strategy of dynamic tuning is proposed in this paper and one possible operation profile is established via sliding-mode theory. Because of the lack of operation mechanism of the equipment in existence, more elaborate ex-perimental verification of the strategy is yet to follow. We may need to work with some equipment suppliers. Based on the proposed strategy, lower dishing and more efficient copper step height reduction can be verified from some numer-ical simulations. Simulation results show better performance under the same throughput.

Keywords: Dynamic tuning; Chemical mechanical planarization; Dishing

1. Introduction

Chemical–mechanical planarization (CMP) was developed during the early 1980s. At that time, multilevel interconnect technology was being pushed to the limits of circuit density and perform-ance. Therefore, a better planarization technique was needed that would allow smaller geometries

for metal linewidth and spacing (wiring pitch), vias and contacts[1]. Among many planarization tech-niques, CMP is a widely accepted technique to provide a globally planarized surface for micro-electronic wafer fabrication. In recent years, the device levels and density have increased continu-ously; however, these densities have now brought resistance–capacitance (RC) time delays to a level that can appreciably slow circuit speeds and thus impede further advances in circuit performance. The emerging trend in LSI technology is therefore toward the replacement of aluminum or aluminum alloys by copper as the metal wiring material

*

Corresponding author. Tel.: 573-1796; fax: +886-3-572-0634.

E-mail address:chiaslin@mail.nctu.edu.tw(C.-S. Lin).

(2)

because of its low resistivity and high electromi-gration resistance[2–4].

Several research eﬀorts have been reported on modeling the CMP of interlevel dielectrics (ILD), shallow trench isolation (STI), and tungsten. How-ever, the CMP processes for copper damascene in-terconnects indicate unique characteristics and more complicated polishing mechanisms than con-ventional ILD, STI and tungsten CMP. The trans-lation of tungsten or ILD CMP to copper CMP is not simply a change in materials [5]. Thus optim-ized CMP becomes an important objective.

De-spite recent advances in CMP, some

manufacturing concerns associated with successful implementation of CMP remain to be overcome [6,7]. The physical interactions among the wafer, slurry and pad make the copper CMP very sensi-tive to operating conditions [6]. Most of the re-search work on CMP is focused on detailed removal mechanism and slurry chemistry [2,3, 8–11]. Much less work has been done on the opera-tional aspects of CMP, especially the process con-trol side of CMP operation. One of the most important variables: the setting of removal rate for the planarization process was studied by Chiu et al. [12]. In their research, an analogy between the ‘‘Soft Landing’’ of a spacecraft and the CMP process was established and the optimal time con-trol law was used. However, from the viewpoint of control engineering, optimal control is not so prac-tical in real world. Therefore, a robust operation region of parameters and an incorporated opera-tion strategy (three strategies) were proposed by Kao et al.[6]. To further improve the performance of CMP process, a more feasible method namely sliding-mode control will be employed to set the operation proﬁle of CMP process through ‘‘Dy-namic Tuning’’ method to enable the CMP process more close to soft landing.

Variable structure control (VSC) with sliding-mode control (SMC) was ﬁrst proposed and elab-orated in the early 1950s in the Soviet Union by Emelyanov and several coresearchers. In their pio-neer works, the plant considered was a linear sec-ond-order system modeled in phase variable form. Since then, VSC has developed into a general design method for a wide spectrum of sys-tem types [13]. Recently, the VSC and SMC are

almost synonymous nouns, and the key difference between VSC and SMC is the existence of sliding-mode in the system. From the viewpoint of math-ematics, sliding-mode is a system behavior in ‘‘hyperspace’’. It should be mentioned that the sys-tem performance will not be influenced by any matched disturbances, and the control strategy provides an effective and robust control for certain classes of nonlinear systems subject to modeling uncertainties in the sliding mode. One of the un-ique aspects of sliding mode is the discontinuous nature of the control action whose primary func-tion of each of the feedback channels is to switch between two distinctively different system struc-tures (or components) such that a new type of sys-tem motion, called sliding mode, exists in a manifold. This peculiar system characteristic is claimed to result in superb system performance which includes insensitivity to parameter varia-tions, and complete rejection of matched distur-bances. To this date, some theoretical problems of VSC are investigated deeply and comprehen-sively; some new synthesis methods of VSC are presented and are extended to many kinds of sys-tems, thus the distance between theory and prac-tice is reduced significantly. This control method is used in more and more systems because of its in-herent advantages[14].

In this paper, the problem statement will be pre-sented in Section 2. In Section 3, the design of SMC controller (setting operation proﬁle via SMC) and simulation results will be shown. The discussions will be presented in Section 4 followed by the conclusion.

2. Problem statements

2.1. Dynamic tuning (in situ/within-wafer control) Recently, the CMP systems have been widely studied and applied to the wafer manufacturing in semiconductor industry. Most of people who are devoted to do research on the CMP process fo-cus their eﬀorts on the process improvement, i.e. ﬁnding the perfect composition of process param-eters, namely, ‘‘Golden Recipe.’’ Nevertheless, it can not be realized in real world scenarios because

(3)

the existence of disturbances, noise, and uncertainty. The inherent characteristics/ﬂaws may cause some imperfections in products (that may be one of the reasons why control theories were developed for). Continuous control for CMP process seems neces-sary. Here, one question is raised: is it more com-plete if the ‘‘Golden Recipe’’ contains an appropriate operation pattern?

In the CMP process, process control covers consumables and equipment tool parameters to improve performance. Run-to-run control is the only one control method with specific control al-gorithm/law in the CMP process and is thought to be the only viable scheme in most semiconduc-tor manufacturing processes because of the lack of in situ measurements of the product quality of in-terest[15]. Besides, the method of run-to-run con-trol is based on statistic data of process performance. As the requirement of degree of planarization increases, within-wafer CMP control is necessary. Less effort has been done on this sub-ject except the works of Chiu et al. [12]. In the same way, from operational aspect of CMP proc-ess, within wafer control will be attempted in this paper. For control system design, a dynamic model which is not too complex and can describe the sys-tem behavior reasonably is necessary. Unfortu-nately, unless the conceptual description used in the works of Chiu et al.[12], there is not any dy-namic model with physical meanings of CMP process among the research efforts which have been reported on modeling the CMP of copper, ILD or STI. In the following section, some pro-posed mechanism/model by Chiu et al. [12] will be shown to support and clarify the operation as-pect of CMP process control.

2.2. Conceptual description of CMP process (dy-namics)

In general, the dynamic equation of the copper removal can be described as[12]

d~h

dt ¼ R; ð1Þ

dR

dt ¼ u; ð2Þ

where ~hrepresents the thickness of copper to be re-moved, R is the removal rate and u is the system input. For convenience, (1) and (2) are further changed into the following state equation:

_x¼ Ax þ bu þ dðx; tÞ; ð3Þ where x¼ ~h RT 2 R2_; A¼ 0 1 0 0 ; b¼ 0 1 and d¼ 0:1t 0:1 :

Note that the initial value of the system xð0Þ ¼ 8000½ 0T. The selection of disturbances is based on the investigations of run-to-run con-troller of CMP, just like the works of Da et al. [15]. In their research, the removal rate would drift down as the process proceeds because of pad wear-ing and slurry dilution. Although the problem of pad wearing can be adjusted by conditioning dur-ing the CMP process, the slurry concentration can be kept constant by pre-mix before delivering onto platen and they might not be too obvious during a single wafer polishing, the disturbance d is also added into the simulation to model the uncertain-ties during CMP process processing. Note that the coeﬃcients are assigned small values arbitrarily. In Section 3, the controller design and some simula-tion results will be made based on this model. 2.3. Mechanism/model with physical meanings

Among the research of CMP process, a detailed mechanism of chemical effects can not be derived yet, but some combined model are tried to include the chemical effect and mechanical effect[4,21]. In this work, the equation proposed by Luo et al. [4] will be employed, that is

R¼ ðKP þ BÞV þ RC; ð4Þ

where R denotes the polish rate, K is proportional constant, P and V are applied downward pressure and linear velocity of the wafer relative to the pol-ishing pad, respectively. Based on Preston equa-tion, two additional terms were included to represent polish rate data better. RC was inserted to represent the purely chemical reaction of the

(4)

slurry during the copper CMP process and B rep-resents the greater dependence of the polish rate on the relative velocity. Note that these constants were determined by the least squares procedure of experimental data and the unit of RCis the same to removal rate (polish rate) R.

Although many technical problems of the metal CMP process have been solved, a major remaining issue is the dishing of the metal line. Dishing may cause many drawbacks, like an increased resist-ance, a higher current density and lower planarity. Consequently, dishing reduces the speed of signal propagation through the circuit, and threatens the reliability of the metal line at high current den-sities. Recently, dishing has been categorized as a must-be-controlled parameter in CMP [20,22]. In the work of Nguyen et al.[20,22], a mathematical model of dishing phenomenon was established. That model accounts for the dependence of dish-ing on the feature size and polish time usdish-ing the morphology and the properties of the polishing pad and process parameters. In this work, the other copper dishing model which was proposed by Lin[5]will be employed to verify dynamic tun-ing method, that is

hD¼ ðS 1ÞPH n ln w w0

1 eKCuEV t_; _ð5:1Þ

where hDis the dishing height; P is applied down-ward pressure; H is the pad thickness; n is a con-stant defined as the conformity of the pad; hS0is the initial step height; KCuis the Preston constant for copper; E is YoungÕs modules of the pad; V is the relative velocity between the work piece and the pad; t represents the overpolish time; w and w0 are defined as the trench width and effective minimum width, respectively (the word ‘‘effective’’ means that pad can contact with the lower fea-ture). The factor S represents the removal rate se-lectivity of copper to barrier material, like tantalum, defined as S ¼RCu RTa ¼KCu KTa : ð5:2Þ

Note that(4) and (5)are both derived based on Preston equation; (5) was established mainly via the conformity (compression) of pad, i.e. this

model attributes planarization to properties of wa-fer-pad contact. Some proofs will be mentioned in the following sections via the above equations which have been established and veriﬁed by other researchers.

2.4. Control objective

From Eq.(5), it is clear that the process param-eters, like P and V, play important roles in the forming of dishing in addition to the feature size, polish time and properties of pad. One of the methods of reducing critical problems occurred in CMP process is to optimize these process pa-rameters. In this work, one new way is considered to improve the performance of CMP process, that is, dynamic tuning of the operation parameters during CMP process run in an appropriate form/ scheme.

The experimental data of Wijekoon et al. [18] indicate that at a constant speed (or pressure) the dishing varies linearly with pressure (or speed). To integrate these conclusions, it is clear that the defect of dishing can be further reduced if the for-mulation of proﬁle of operation parameters can be set appropriately in bulk-polish and over-polish stages. However, the higher loads of process pa-rameters (larger values of P and V) will increase the removal rate in (4). To maintain acceptable throughput and performance, the operation scheme has to be worked out. Note that it is not suﬃcient to improve the performance by adjusting mechanical loads (P and V). The chemical load (represents oxidizer concentration in this paper) is an important factor, too.

3. CMP operation via SMC 3.1. Design

From Chen and Chang[16], to design a control-ler via SMC, the procedure can be concluded into two steps, that is: ﬁrst, to specify a sliding function s(x) which will switch the system to move toward control goal while the system is in the sliding mode; Second, to determine the control input u

(5)

which will force the system to hit the sliding sur-face and generate the sliding mode in a ﬁnite time. There are three methods which are generally used in designing sliding function, Transformation Ma-trix method, eigenstructure Assignment method and Lyapunov-based method. Similarly, in estab-lishing control law, there are Hierarchical SMC and Integral SMC and so on. In this work, a con-venient and powerful method which is also pro-posed by Chen and Chang [17], namely, Virtual eigenvalue method is used in controller design. 3.2. Simulation

In SMC design, the ﬁrst thing that has to be ad-dressed is the convergence of the sliding function. InFig. 1, it is clear that the system reaches the slid-ing-mode (s = 0) in a ﬁnite time, about 0.6 min. This result represents that the design is successful, i.e. the control law works. Additionally, to visual-ize whole performance of the system, phase–plane analysis is a powerful tool.Fig. 2shows the phase– plane sketch of the system, and it reveals that the thickness (horizontal axis) decreases continuously and the variation of material removal rate (vertical

axis) is smooth. Besides, Fig. 3shows the plot of control input which also varies smoothly. The last one isFig. 4, and it shows time history of removed copper thickness and (copper-) material removal rate during CMP process separately. In Fig. 4, the more detailed proﬁle of these two important parameters are presented. The CMP process will ﬁnish in about 1.7 min (102 s), and the removal rate will decrease to about zero at the same time. 3.3. Control input concretization

To clarify the control input u, the mechanism which is presented in Section 2.3 is repeated here. From the system dynamic Eq.(3)and the material removal model (4), the control input is shown by diﬀerentiating(4), and(3)can be formulated as

_x¼ Ax þ BU þ dðx; tÞ; ð6Þ where x¼ ~h RT2 R2 ; A¼ 0 1 0 0 ; B¼ 0 0 KV KPþ B ; U¼ dP dt dV dt T ; d¼ 0:1t 0:1

(6)

and the initial value of the system is x(0) = [8000 0]T. The elements of U,dP

dt and dV

dt, can be transferred to the design/actuators of equipment. Note that con-straints of applied downward pressure and linear

relative velocity are not considered in(6). Because of the lack of model of chemical factors in CMP process, the detailed equation is diﬃcult to be ex-pressed here. However, from the research by Kao

Fig. 3. Control input (the maximum value is equal to 216,000 A˚ /min2). Fig. 2. Phase–plane plot.

(7)

et al.[6], some experiments were carried out over a wide range of oxidizer concentration. They indi-cate that the copper removal rate of copper CMP will follow a speciﬁc trend as the concentra-tion increases. Fortunately, the trend, shown in Fig. 5, is similar to the operation proﬁle which is shown above (Fig. 4). With the constraints of downward pressure and relative velocity, the chemical factor, RCcan be introduced into the

op-eration mechanism. The control input and input matrix B can be reformed as

U¼ dP dt dV dt dRc dt T and B¼ 0 0 0 KV KPþ B 1 :

Note that the relation between oxidizer concentra-tion and RC(with specific slurry) was proposed by Luo et al.[4], as shown inFig. 6. Note that reduc-tion of removal rate at higher H2O2concentration in both figures,Figs. 5 and 6, is due to the forma-tion of passivaforma-tion layer (copper oxide film).

Because of the important interdependence of chemical and mechanical eﬀects, it has to be men-tioned that a phenomenon in changing the process parameters. The phenomenon has been proposed in the works of Steigerwald et al. [11] and Kao et al.[6], that is, two polishing regimes, a dissolu-tion rate limited regime (chemical reacdissolu-tion control-led regime) and an abrasion rate limited regime (mechanical abrasion controlled regime), exist in CMP process. Besides, the eﬀects of slurry chemis-try on dishing have been proposed by Nguyen et al. [9]. They revealed that less dishing would

Fig. 4. Realistic trajectory (the maximum removal rate is equal to 13,866 A˚ /min).

Fig. 5. Removal rate (A˚ /min) as a function of oxidizer concentration for diﬀerent pressure and speed settings: P = 4 p.s.i. and V = 93 rpm; P = 3 p.s.i. and V = 93 rpm; P = 3 p.s.i. and V = 63 rpm[6].

(8)

be obtained from higher oxidizer concentration in the slurry.

To integrate the research achievements [6,8,9,11,19,20,22], the control inputs integration of control inputs described must be coordinated in harmony with each other. In other words, Sec-tion 2.4 can be implemented with these considera-tions of chemical eﬀects. The dynamic tuning operation of CMP process is then coordinated, as shown in Fig. 7. Fig. 5is used to illustrate the concept of dynamic tuning operation.

It must be mentioned again that most superior feature of dynamic tuning operation is the higher degree-of-freedom/flexibility of the process para-meters. If all of the parameters can be assigned arbitrarily during the process, the possibility of fig-uring out the optimal operation condition/profile is much larger.

The acceleration stage in Fig. 7 means the in-creasing of removal rate (before about 0.2 min of

second plot ofFig. 4). It is followed by decelera-tion stage after the peak value of removal rate is achieved. Because all of the operation parameters (means P, V and RC here) are all changeable, in-creasing and dein-creasing of removal rate can be re-alized by tuning them. Besides, the operation can be designed with more appropriate sequence to ob-tain better performance. The proposed sequence is to set the system at mechanical abrasion controlled regime in the beginning. At this regime, the passi-vation layer is formed and the rate of forming is higher than the mechanical abrasion rate. There-fore, it can be expected that lower region of wafer surface is protected by the passivation layer. Con-tinuously, the load is increased (larger values of parameters) to raise the removal rate until the peak value of removal rate. Note that relative ve-locity will be the major factor in increasing removal rate because of the lower inﬂuence on dishing. In the representation of Fig. 7, the values of opera-tion parameters, P, V and RC, can be assigned with 3.5 p.s.i., 93 rpm, 300 nm/min, respectively (this set of parameter represents the constraint which is as-signed in this work). As soon as the peak value is achieved, these parameters can be decreased for suppressing dishing, especially the applied down-ward pressure P. Moreover, higher oxidizer con-centration will cause less dishing from the results shown by Nguyen et al.[9]because a more eﬀective passivation layer is formed with higher oxidizer concentration and protects the recess areas.

3.4. Veriﬁcation

From the descriptions above, it is obvious that operation profile can be set in many ways if upper limit is the only one constraint for each input. Therefore, a new problem will be raised, what is the best tuning/operation profile? It will not be discussed in this paper. Here, one possible operation profile will be chosen to verify the ef-fect of dynamic tuning operation of CMP proc-ess. Three process parameters, P, V and the oxidizer (H2O2) concentration, will be tuned in specific profile dynamically during single wafer polishing. The bounds of P, V and the oxidizer concentration are set by 1.7–3.5 p.s.i., 40–95

Fig. 7. Dynamic tuning operation of CMP process. Fig. 6. Eﬀect of H2O2concentration on the value of RCwith

(9)

rpm and 2–5%, respectively. Note that relative velocity is transferred to platen/carrier rotational speed.

The dishing model(5)is used to verify the ability of dynamic tuning method and the result is shown inFig. 8. It is clear that dishing is suppressed from

Fig. 8. (a) Comparison of dishing between dynamic tuning and corresponding constant input (copper removal). (b) Variations of parameters (Kcu, B and RCare normalized forms).

(10)

1106 to 950 A˚ inFig. 8(a)while the dynamic tuning is applied to process operation/control. The im-provement is more than 14% with maintaining the copper removal (throughput) of constant input. Fig. 8(b) displays the sketch of variations of parameters during the over-polish stage. Note that the variations of Kcu, B and RC are caused by changing the oxidizer concentration.

To complete simulate the strategy of dynamic tuning, the step height reduction (bulk copper pol-ishing) must be included in. But the validity of the available model of step height reduction proposed by Lin[5]and Fu and Chandra[8]are not consist-ent with the experimconsist-ent data proposed by Wi-jekoon et al. [18], as shown in Fig. 9. The ﬁgure reveals that the dishing will exist before the end of process and linearly increases with overpolish/ time. This experiment data points out that initial value of dishing will not be zero (i.e. non-zero step height will exist before transition to dishing).

However, the step height in the models by Lin and Fu will decrease with time exponentially dur-ing bulk copper polishdur-ing stage. Both models [5,8]conflict with the experiment data[18]. There-fore, the verifications will be done by some exper-iments in future research. The illustration of the difference of surface height in the bulk-polish stage and over-polish stage is shown inFig. 10 (regard-less of oxide erosion).

4. Discussion

In CMP process, the chemical and mechanical effects have to be operated in harmony for main-taining acceptable performances and throughput. The procedure is a optimization problem. To op-timize the recipe (a set of specific process parame-ters) becomes a major work of improving the process. For most research on CMP process, the general characteristics of the process are described roughly. However, some limitations (due to con-stant input) will be led into the process in typical operation. Based on the existing knowledge and the typical operation, some new operation meth-ods are figured out, like two or three stage polish-ing. The multi-stage method is widely used in real production but it is not good enough in the grow-ing requirement of planarization. In dynamic tun-ing method, the degree-of-freedom/flexibility of process parameters will be increased (while break-ing away from the typical operation). That will in-crease the possibility of formulating the set of process parameters. The opportunity of further improving the performance of CMP process will be feasible.

The distinguishing feature of operation by dy-namic tuning is the process parameters varying during single wafer polishing. It involves within-wafer control of fabrication process. Although it is not an easy subject in process control of semi-conductor production, the tendency of progress of fabrication technique seems to be inevitable. From the simulations shown in Section 5, the op-eration of dynamic tuning method reveals out-standing improvement in dishing. The feature of deriving better performance and maintaining throughput simultaneously is the other important

Fig. 9. Eﬀect of over polish on the extent of dishing[18].

Fig. 10. Illustration of the diﬀerence of surface height in the whole process.

(11)

advantages of this strategy. Moreover, setting the operation proﬁle via sliding mode theory, not only the violent switching is avoided but the lower loads of parameters are employed. It will make this method much more applicable in real world.

Except the numerical simulations, significant evidences for proving the capability of the opera-tion profile have been revealed (Note that the pro-file is the removal rate curve inFig. 4). Because the recipe of low removal rate will be employed in the final stage of this operation, it will have better per-formance, low copper dishing[12,18,19]. The ‘‘de-liberate’’ (low load on the process) operation at final stage will reduce the damages caused by over-polish effectively, and the over-polish time will be easier to control [8,9,20]. Additionally, the operation profile of SMC design is closer to the shape on concept of soft-landing because the process almost stops near the endpoint of polish-ing. In case of the measurement uncertainties, a correct realization of soft-landing should be more effective using SMC approach.

To control the removal rate or the other para-meters precisely is a difficult objective because of the existence of uncertainties and disturbances of the prediction model or the working environment. One important point about dynamic tuning method needs to be emphasized, that is, even though the re-moval rate can not be performed perfectly with the designed trajectory, the performance can be im-proved by using the strategy of dynamic tuning (as described inFig. 7). (This point is shown in the dif-ference betweenFig. 8(a) and 4.) In other words, Dynamic Tuning operation approach operating is more important than what is the exact operating profile. It is the future work to determine what the best operation profile is, how to implement the com-plete control scheme via sliding-mode theory. The dynamic response of the operation of actual CMP tools needs to be looked to ensure proper following of soft landing model in real world.

5. Conclusion

In this work, the concept of dynamic tuning of operating CMP process is presented and one

pos-sible operation profile is proposed via sliding-mode theory. This strategy is verified by numerical simulations. The simulation results show 14% re-duction in dishing during over-polish stage. Future experiment needs to be designed to confirm this improvement.

However, the simulation veriﬁcation of step height reduction needs to be done in future work when there are better models which describes con-tinuous transition from step height from bulk-pol-ish stage to dbulk-pol-ishing during over-polbulk-pol-ish stage. Otherwise, experiments can be designed to com-pare directly step height reduction between con-ventional CMP operation and ‘‘dynamic tuning’’ CMP operation.

Dynamic operation provides a new aspect of CMP operation and it might be a candidate for within wafer control in CMP process. It is the hope of this research that dynamic tuning may be considered for improving CMP planarization in the area of step height reduction and dishing reduction.

References

[1] Texas Engineering Extension Service (TEEX), CMP, TEEX, 2001.

[2] T. Matsuda, H. Takahashi, M. Tsurugaya, K. Miyazaki, T.K. Doy, M. Kinoshita, J. Electrochem. Soc. 150 (9) (2003) G532–G536.

[3] Y. Ein-Eli, E. Abelev, E. Rabkin, D. Starosvetsky, J. Electrochem. Soc. 150 (9) (2003) C646–C652.

[4] Q. Luo, S. Ramarajan, S.V. Babu, Thin Solid Films 335 (1998) 160–167.

[5] Y. Lin, Solid State Technol. 43 (6) (2000) 111–121. [6] Y.-C. Kao, C.-C. Yu, S.-H. Shen, Microelectron. Eng. 65

(2003) 61–75.

[7] J.M. Steigerwald, S.P. Murarka, R.J. Gutmann, Chemical Mechanical Planarization of Microelectronic Materials, Wiley, New York, 1997.

[8] G. Fu, A. Chandra, IEEE Trans. Semiconduct. Manufact. 16 (3) (2003) 477–485.

[9] V. Nguyen, H. VanKranenburg, P. Woerlee, Microelec-tron. Eng. 50 (2000) 403–410.

[10] X. Wang, H. Wang, Y. Liu, in: Proceedings of the 6th International Conference on Solid-State and Integrated-Circuit Technology, 2001, pp. 369–371.

[11] J.M. Steigerwald, S.P. Murarka, R.J. Gutmann, D.J. Duquette, J . Vac. Sci. Technol. B 13 (6) (1995) 2215–2218. [12] J.-B. Chiu, C.-C. Yu, S.-H. Shen, Microelectron. Eng. 65

(12)

[13] John Y. Hung, Weibing Gao, James C. Hung, IEEE Trans. Ind. Electron. 40 (1) (1993) 2–22.

[14] S. Pan, H. Su, X. Hu, J. Chu, in: Proceedings of the 3rd World Congress on Intelligent Control and Automation, 2000, pp. 2977–2981.

[15] L. Da, V.G. Kumar, A. Tay, A.A. Mamun, W.K. Ho, A. See, L. Chan, in: Proceedings of the 2002 IEEE International Symposium on Intelligent Control, 2002, pp. 740–745. [16] Y.-P. Chen, J.-L. Chang, Design of Variable Structure

System (Chinese version), second ed., Chuan Hua, Taipei, 2002.

[17] Y.-P. Chen, J.-L. Chang, Int. J. Sys. Sci. 31 (4) (2000) 417– 420.

[18] K. Wijekoon, S. Tsai, M. Chandrachood, B. Brown, F. Redeker, S. Nanjangud, G. Amico, SEMI/Japan Technical Symposium, 1998.

[19] Y.-C. Kao, C.-C. Yu, S.-H. Shen, Microelectron. Eng. 65 (2003) 61–75.

[20] V.H. Nguyen, P. van der Velden, R. Daamen, H. van Kranenburg, P.H. Woerlee, IEDM Technical Digest. (2000) 499–502.

[21] J. Luo, D.A. Dornfeld, IEEE Trans. Semiconduct. Man-ufact. 14 (2) (2001) 112–132.

[22] V.H. Nguyen, R. Daamen, H. van Kranenburg, P. van der Velden, P.H. Woerlee, J. Electrochem. Soc. 150 (11) (2003) G689–G693.