Numerical modeling and experimental verification of

Microelectronic Engineering 170 (2017) 54–58

Contents lists available at ScienceDirect

Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee

Research paper

Numerical modeling and experimental verification of copper electrodeposition for through silicon via (TSV) with additives Hongbin Xiao, Hu He, Xinyu Ren, Peng Zeng, Fuliang Wang ⁎ State Key Laboratory of High Performance Complex Manufacturing, Changsha 410083, China School of Mechanical and Electrical Engineering, Central South University, Changsha 410083, China

a r t i c l e

i n f o

Article history: Received 3 June 2016 Received in revised form 12 December 2016 Accepted 28 December 2016 Available online 30 December 2016 Keywords: TSV Copper filling Electrodeposition Numerical modeling Additives

a b s t r a c t Since voids and seams are easily formed during the process of filling TSVs with high aspect ratio, good methods that can achieve the superfilling of TSVs are eagerly needed. This paper presents the numerical modeling of TSV filling concerning the influence of three additives (accelerator, suppressor and leveler). By changing the additives' doses and current density, the following three different simulation results were obtained: the pinch-off effect, seam-inside filling model and “V” shaped filling model. The corresponding distributions of the current density along the cathode surface were analyzed to investigate the filling mechanism. Moreover, TSV filling experiments in the presence of additives were also conducted to validate the proposed numerical model. The simulation results matched well with the experimental results. TSVs with a diameter of 20 μm and depth of 200 μm were fully filled in the appropriate conditions. © 2016 Published by Elsevier B.V.

1. Introduction 3D integration with through silicon vias (TSVs) has been widely known as a promising technology for future electronic systems. It provides the shortest vertical interconnections with a series of significant advantages, such as a wide bandwidth, lower energy consumption, higher density, smaller form factor, and improved electrical performance [1–4] compared with the conventional 2D integration approach. The vertical interconnects are formed through the wafer to enable communication among the stacked chips. Its applications are also quite wide, which include MEMS sensors, CMOS image sensors (CIS), memorizers, hybrid memory cubes (HMC), radio-frequency circuits and so on. For most TSV development, copper electrodeposition is one of the most important technologies for implementing the 3D interconnection [5,6]. However, since most TSVs are of high aspect ratios [7–9], and the TSVs filling experiment is a complex physicochemical dynamic process, it is really a difficult and challenging task to fill these TSVs with no voids or seams. To accomplish void-free filling, several additives such as bis disulfide (SPS), poly (PSG) and chloride ions (Cl−) are usually added to the plating bath [10–15]. At the same time, the bottom-up superfilling has been achieved as the join of these additives. On the other hand, numerical modeling for TSV filling has also been presented. The bottom-up superfilling model was first proposed by Moffat et al. and explained by

the curvature enhanced accelerator coverage (CECA) [16]. Some other numerical models of TSV filling have also been developed to explain the superfilling process [17–22]. Among these models, a general diffusion-adsorption theory has been widely used in explaining the superfilling phenomena [23–28]. However, numerical modeling and experimental verification for high aspect ratio (H:D = 10:1) TSV filling with three additives(accelerator, suppressor and leveler) have not yet been closely combined to deeply investigate the filling mechanism. In addition, experiments of filling high aspect ratio TSVs with a diameter of 20 μm and depth of 200 μm in the presence of additives have not yet been systematically conducted from published papers. In this paper, a new numerical model that is based on the ButlerVolmer equation and considers the process of adsorption, desorption and diffusion of the three additives (accelerator, suppressor and leveler) on the cathode surface is proposed for copper electrodeposition of TSVs. Three different simulation results for TSV filling were obtained by changing the additives' doses and current density: the pinch-off effect, seam-inside filling and “V” shaped filling. The corresponding distribution of current density along the cathode surface was also analyzed to help us better understand the filling mechanism of copper electroplating. Moreover, TSV filling experiments with additives were also conducted to validate the proposed numerical modeling. The simulation results match well with the experimental results. 2. Numerical model

⁎ Corresponding author at: School of Mechanical and Electrical Engineering, Central South University, Changsha 410083, China. E-mail address: [email protected] (F. Wang).

http://dx.doi.org/10.1016/j.mee.2016.12.030 0167-9317/© 2016 Published by Elsevier B.V.

Fig. 1 shows the planar model and boundary conditions of the TSV filling. The diameter (D) of the TSV is 20 μm and the depth (H) is

H. Xiao et al. / Microelectronic Engineering 170 (2017) 54–58

55

The effect of accelerator (Acc for short), suppressor (Sup for short) and leveler (Lever for short) on TSV filling are all considered in the model. The three additives can affect the distribution of current density along the cathode, which can affect the deposition rate in different parts of the via and the final filling result. The distribution of the three additives in the electrolyte can be shown by the concentration diffusion field. The equation of the concentration field is: ∇2 ci ¼ 0

ði ¼ Acc; Sup; LeverÞ

ð2Þ

where cAcc is the concentration of accelerator, cSup is the concentration of suppressor and cLever is the concentration of leveler. Considering the adsorption, desorption and diffusion of the three additives(accelerator, suppressor and leveler) on the cathode surface, the time-dependent surface coverage of the three additives can be shown as: 8 ∂θAcc > > ¼ kAcc ads cAcc ð1−θAcc −θLever Þ−kAcc des θAcc −DAcc s ∇2∇ θAcc > > > < ∂t ∂θSup ¼ kSup ads cSup 1−θAcc −θSup −θLever −kSup des θSup −DSup s ∇2∇ θSup > ∂t > > > > 2 : ∂θLever ¼ k Lever ads cLever ð1−θAcc −θLever Þ−kLever des θLever −DLever s ∇∇ θLever ∂t ð3Þ

Fig. 1. Schematic illustration of the mathematical model used in the TSV copper electroplating.

200 μm. The symmetrical boundary of the electrolyte is 50 μm in height and 200 μm in distance. The cathode includes the wafer surfaces, the TSV sidewalls and the TSV bottom, which are all marked as shown in Fig. 1. It's assumed that the mass transport is dominated by diffusion and electromigration. Convection is neglected. The concentration of ions (c) on the anode surface, including copper and additives ions, is defined as equal to that in the bulk solution (cbulk). The concentration of ions between anode and cathode varies with time, which is defined as ∂c ∂t ¼ D∇2 c. The ion concentration remains steady in the x direction, which ∂c ¼ 0. The main parameters used for the numerical modelis defined as ∂x ing are shown in Table 1. The plating bath of the model consists of copper ion, sulfate ion and additive ions. The ion flux is given by the Nernst-Planck equation:

! Ni ¼ −Di ∇ci −zi ui Fci ∇ϕl þ ci u

ð1Þ

where Ni is the ion flux, Di is the diffusion constant, ci is the ion concentration, zi is the valence (charge number), ui is the mobility constant, ϕl ! is the electric potential inside the electrolyte solution, u is the flow velocity of the electrolyte and F is Faraday's constant.

Table 1 Parameters used for the numerical modeling. Symbols cCu2+∞ kAcc_ads kSup_ads kLever_ads kAcc_des kSup_des kLever_des DCu2+ DAcc_s DSup_s DLever_s

where θAcc (θSup,θLever) is the surface coverage of accelerator (suppressor, leveler) on the cathode, kAcc_ads(kSup_ads,kLever_ads) is the adsorption coefficient of accelerator (suppressor, leveler), kAcc_des(kSup_des,kLever_des) is the desorption coefficient of accelerator (suppressor, leveler), DAcc_s(DSup_s,DLever_s) is the diffusion coefficient of accelerator (suppressor, leveler), and ∇T is the mathematical operator of the vertical edge interface. Since every area in the via has four different states (covered by accelerator, covered by suppressor, covered by leveler and covered by basic solution), based on Butler-Volmer equation and considers the process of adsorption, desorption and diffusion of the three additives (accelerator, suppressor and leveler) on the cathode surface, the current density on the cathode surface can be then calculated as: 8 cCu2þ −α Acc Fη > > ¼ i θ exp i > Acc 0 Acc Acc > c∞ 2þ R T > > Cu g > > > c 2þ −α Sup Fη > Cu > i0 Sup θSup exp > < iSup ¼ c∞ R T Cu2þ g cCu2þ −α Lever Fη > > > ¼ i θ exp i Lever Lever 0 Lever > > c∞ 2þ Rg T > > Cu > > c 2þ > −α basic Fη >i Cu > i0 basic 1−θAcc −θSup −θLever exp : Basic ¼ ∞ c 2þ Rg T Cu

where i0_Acc(i0_Sup,i0_Lever) is the exchange current density of the electrolyte containing accelerator(suppressor, leveler), i0_basic is the exchange current density of the basic solution without any additives, cCu2+ is the concentration of copper on the cathode surface, cCu2+∞ is the concentration of copper in the bulk electrolyte, and αAcc(αSup,αLever, αbasic) is the electrochemistry transmission coefficient. The total current density on the cathode surface is the sum of iAcc, iSup,iLeverand ibasic, which can be shown as: i ¼ iAcc þ iSup þ iLever þ iBasic

Values

ð4Þ

ð5Þ

3

500 mol/m 7.9 × 10−3 m3/(s·mol) 0.158 m3/(s·mol) 9 × 10−3 m3/(s·mol) 1 × 10−3 1/s 0.01 1/s 0.3 × 10−3 1/s 2 × 10−8 m2/s 3.24 × 10−4 m2/s 3.92 × 10−5 m2/s 1.46 × 10−5 m2/s

Substituting Eq. (4) into Eq. (5), we can achieve: i¼

" cCu2þ i0 c∞ 2þ

−α Sup Fη −α Acc Fη þ i θ exp θ exp 0 Sup Sup Acc Acc Rg T Rg T Cu −α Lever Fη þ i0 Lever θLever exp Rg T # −α basic Fη þ i0 basic 1−θAcc −θSup −θLever exp Rg T

ð6Þ

56


Fig. 2. (a) Simulated TSV filling profile, (b) distribution of current density along the cathode surface (average current density 0.23ASD, cAcc = cLever =1 ppm, cSup =0 ppm).

Through the Faraday's law, the local current density i can translated into the local deposition rate: ν¼

M Cu i FρCu n

ð7Þ

where MCu is the molecular weight of copper, ρCu is the density of copper, and n is the number of electrons transferred. In a more general way, the model shown above is usable for all TSVs with any other dimensions. However, the chemical parameters used in the numerical modeling, including the diffusion coefficient, adsorption and desorption coefficients of additives and so on, have to be adjusted. 3. Results and discussion 3.1. Simulation result Fig. 2 shows the simulated TSV filling profile (Fig. 2(a)) and the distribution of current density (Fig. 2(b)) along the cathode surface when the average current density is 0.23ASD, cAcc = cLever = 1 ppm and cSup =0 ppm. It can be seen from Fig. 2(a) that the pinch-off effect occurs in this condition. The opening of the via merges in advance and

stops the filling process in the middle and bottom of the via. Obviously, this filling result is quite unsatisfactory. The pinch-off in the opening of the via can impede the mass transportation process of copper ions into the interior of the via, which will result in large voids in the via. The current density decreases from the wafer surface to the TSV sidewall and to the TSV bottom as illustrated in Fig. 2(b). It is seen that the current density in the opening of the via is as high as 0.31ASD, which is more than 2 times of that in the bottom. Since the filling rate increases with current density, the filling rate in the opening of the via is much higher than in the TSV sidewall and TSV bottom, which results in the pinch-off effect at last. Fig. 3 shows the simulated TSV filling profile (Fig. 3(a)) and the distribution of current density along the cathode surface (Fig. 3(b)) when the average current density is 0.05ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup =10 ppm. It can be seen from Fig. 3(a) that the pinch-off effect weakens in the opening of the via, but a large seam forms in the middle of the via. This filling result is also unsatisfactory. As shown in Fig. 3(b), the current density on the wafer surface remains at approximately 0.052ASD. The current density on the TSV sidewall decreases gradually, with a minimum of 0.032ASD at first, and increases sharply later. This increasing tendency continues until the current density in the center of the TSV bottom reaches its peak at 0.069ASD. Actually, since the

Fig. 3. (a) Simulated TSV filling profile, (b) distribution of current density along the cathode surface (average current density 0.05ASD, cAcc = 0 ppm, cLever = 1 ppm, cSup = 10 ppm).


57

Fig. 4. (a) Simulated TSV filling profile, (b) distribution of current density along the cathode surface (average current density 0.025ASD, cAcc = 0 ppm, cLever = 1 ppm, cSup = 20 ppm).

current density in the TSV sidewall is smaller than in the opening and bottom of the via, the filling rate in the TSV sidewall is the slowest, which undoubtedly contributes to a large seam in the middle part of the via. Fig. 4 shows the simulated TSV filling profile (Fig. 4(a)) and the distribution of current density along the cathode surface (Fig. 4(b)) when the average current density is 0.025ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 20 ppm. It can be seen that the “V” shaped filling model is obtained, as shown in Fig. 4(a). This filling model is satisfactory. Because it can achieve superfilling of the TSVs when the filling process finishes. The current density increases from the wafer surface to the TSV sidewall and TSV bottom along the cathode surface, as shown in Fig. 4(b). The corresponding filling rate increases then from the wafer surface to the TSV sidewall and TSV bottom, which results in a “V” shaped filling model in the end.

3.2. Experimental validation To validate the proposed numerical model, TSV filling experiments were conducted. Blind vias with a diameter of 20 μm and depth of 200 μm inside the chip were etched by a Bosch type deep reaction ion etching (DRIE). Then TEOS CVD was used to deposit a silicon oxide layer (approximately 0.3 μm) for sidewall insulation. Lastly, a Cu layer is deposited as a seed layer using the PVD process. The basic solution of the TSV filling experiments consisted of 195 g/L CuSO4 ⋅ 5H2O, 32 ml/L H2SO4 and 0.24 ml/L chloride ion. Additives added to the basic solution include accelerator SPS (bis 3-sulfopropyl disulfide), suppressor AESS (aliphatic amine sulphonate) and leveler PN (a polyethylene imine alkyl salt). The TSV sample was immersed in DI water in the vacuum chamber for 5 min and placed in an ultrasonator for 5 s, as pre-processing to ensure that no air bubbles or impurities were inside the vias. To make the copper and additive ions inside and outside the vias be at equilibrium, the TSV sample was immersed in the plating bath for 10 min before electroplating. A scanning electron microscope (SEM) (TESCAN MIRA3 LMU) was used to observe the cross section and microstructure of the TSV after the electrodeposition finished. To observe the cross section, TSV samples were embedded into an epoxy resin and then polished to a mirror-like surface using a series of emery papers and the slurries of fine alumina. The electrochemical analysis of the basic solution with and without additives was performed by an electro-chemical workstation (Chenhua CHI660E). Cyclic voltammetry was performed in a 3-electrode cell with an Ag working electrode, a Pt counter electrode and a mercurous sulfate

reference electrode (SCE). The testing potential ranged from 0.6 V to −0.7 V vs. SCE with a scan rate of 5 mV/s. Fig. 5 shows cyclic voltammetry of the basic solution with and without additives. It can be seen that the current increases more sharply in the basic solution than in the basic solution with additives when the potential is −0.2 V vs. SCE, which implies that the existence of additives suppresses the deposition of copper. Moreover, at a more negative potential, the slope of the curve continually increases, which means that the inhibition effect on copper deposition is stronger at a more negative potential. Fig. 6 displays the cross-sectional SEM images of the via filling results under different conditions. Fig. 6(a) shows the filling result when the average current density was 0.2ASD, cAcc = cLever = 1 ppm and cSup = 0 ppm with a plating time of 15 h. It can be seen that the pinch-off effect occurred in this condition and a large part of the via is not filled. Fig. 6(b) shows the filling result when the average current density was 0.05ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 10 ppm with a plating time of 16 h. It can be seen that a large seam formed inside the via. Fig. 6(c) shows the filling result when the average current density was 0.03ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 20 ppm with a plating time of 30 h. It can be seen that superfilling of the TSV was achieved in this condition. There are no large voids or seams inside. Comparing the three filling results in Fig. 6, it can be speculated that there are two important factors that can influence the TSV filling result. One is the dosage of additives which include accelerator, suppressor and leveler. The other is current density. Only if the two factors are adjusted appropriately can superfilling be achieved.

Fig. 5. Cyclic voltammetry of the solutions: (a) basic solution and (b) basic solution with additives (0 ppm accelerator (SPS), 20 ppm suppressor (AESS) and 1 ppm leveler (PN)).

58


three different filling models have been obtained. The pinch-off effect was observed when the average current density was 0.23ASD, cAcc = cLever = 1 ppm and cSup = 0 ppm. A large seam appears when the average current density was 0.05ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 10 ppm. A “V” shaped filling result was obtained when the average current density was 0.025ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 20 ppm. The distribution of the current density along the cathode surface was analyzed to investigate the filling mechanism. Moreover, TSV filling experiments in the presence of additives were also included. TSV samples with a diameter of 20 μm and depth of 200 μm used in the experiments were fully filled in appropriate conditions (average current density 0.03ASD, cAcc = 0 ppm, cLever = 1 ppm, cSup = 20 ppm). The simulation results matched well with the experimental results. Acknowledgments Fig. 6. Cross-sectional SEM images of the via filling results under different conditions. (a) Average current density 0.2ASD, cAcc = cLever = 1 ppm, cSup = 0 ppm, plating time 15 h, (b) average current density 0.05ASD, cAcc = 0 ppm, cLever = 1 ppm, cSup = 10 ppm, plating time 16 h, and (c) average current density 0.03ASD, cAcc = 0 ppm, cLever = 1 ppm, cSup = 20 ppm, plating time 30 h.

To find out the forming process of the filling results in Fig. 6, experiments were conducted with shorter plating times. Fig. 7 displays the cross-sectional SEM images of the via filling results under the same conditions corresponding to Fig. 6 except for shorter plating times. Fig. 7(a) shows the filling result when the average current density was 0.2ASD, cAcc = cLever = 1 ppm and cSup = 0 ppm with a plating time of 5 h. Obviously, the filling rate in the opening of the via is faster than in the sidewall and bottom, which results in the pinch-off effect in the opening of the via, as shown in Fig. 6(a). Fig. 7(b) shows the filling result when the average current density was 0.05ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 10 ppm with a plating time of 8 h. It can be seen that the filling rate in the sidewall of the via is a bit smaller than in the opening and bottom, which may account for the seams, as shown in Fig. 6(b). Fig. 7(c) shows the filling result when average current density was 0.03ASD, cAcc = 0 ppm, cLever = 1 ppm and cSup = 20 ppm with a plating time of 4 h. It can be seen that the filling rate increases from the opening to the sidewall and bottom of the via, which finally forms a “V” shaped model. 4. Conclusion In this paper, a numerical model of TSV filling is presented concerning the influence of three additives (accelerator, suppressor and leveler). By changing the additives' doses and current density,

Fig. 7. Cross-sectional SEM images of the via filling results in different current densities. (a) Average current density 0.2ASD, cAcc = cLever = 1 ppm, cSup = 0 ppm, plating time 5 h, (b) average current density 0.05ASD,cAcc = 0 ppm, cLever = 1 ppm, cSup = 10 ppm, plating time 8 h, and (c) average current density 0.03ASD, cAcc = 0 ppm, cLever = 1 ppm, cSup = 20 ppm, plating time 4 h.

The work was support by the China Department of Science & Technology Program 973 (Contract No.: 2015CB057202). References [1] P. Garrou, C. Bower, P. Ramm, Handbook of 3D Integration: Technology and Applications of 3D Integrated Circuits, Wiley-VCH, New York, 2008. [2] M. Kawano, N. Takahashi, Y. Kurita, K. Soejima, M. Komuro, S. Matsui, IEEE Trans. Electron Devices 55 (2008) 1614–1620. [3] J.U. Knickerbocker, P.S. Andry, B. Dang, R.R. Horton, M.J. Interrante, C.S. Patel, R.J. Polastre, K. Sakuma, R. Sirdeshmukh, E.J. Sprogis, IBM J. Res. Dev. 52 (2008) 553–569. [4] S. Pozder, R. Chatterjee, A. Jain, Z. Huang, R.E. Jones, E. Acosta, Progress of 3D integration technologies and 3D interconnects, IEEE International Interconnect Technology Conference 2007, pp. 213–215. [5] X. Gagnard, T. Mourier, Microelectron. Eng. 87 (2010) 470–476. [6] G. Druais, G. Dilliway, P. Fischer, E. Guidotti, O. Luhn, A. Radisic, S. Zahraoui, Microelectron. Eng. 85 (2008) 1957–1961. [7] D. Henry, J. Charbonnier, P. Chausse, F. Jacquet, B. Aventurier, C. Brunet-Manquat, V. Lapras, R. Anciant, X. Sillon, B. Dunne, A. Hotellier, J. Michailos, Through silicon vias technology for CMOS image sensors packaging: presentation of technology and electrical results, Electronics Packaging Technology Conference 2008, pp. 35–44. [8] R. Puschmann, M. Boettcher, M. Ziesmann, I. Bartusseck, F. Windrich, C. Fiedler, P. John, C. Manier, K. Zoschke, J. Grafe, H. Oppermann, M.J. Wolf, K.D. Lang, Via last technology for direct stacking of processor and flash, Electronic Components & Technology Conference 2012, pp. 1327–1332. [9] J. Van Olmen, C. Huyghebaert, J. Coenen, J. Van Aelst, E. Sleeckx, A. Van Ammel, S. Armini, G. Katti, J. Vaes, W. Dehaene, E. Beyne, Y. Travaly, Microelectron. Eng. 88 (2011) 745–748. [10] M. Tan, J.N. Harb, J. Electrochem. Soc. 150 (2003) C420–C425. [11] M. Hasegawa, Y. Negishi, T. Nakanishi, T. Osaka, J. Electrochem. Soc. 152 (2005) C221–C228. [12] Y. Chiu, W. Dow, J. Electrochem. Soc. 160 (2013) D3021–D3027. [13] T.P. Moffat, L.Y.O. Yang, J. Electrochem. Soc. 157 (2010) D228–D241. [14] W.P. Dow, H.S. Huang, Z. Lin, Electrochem. Solid-State Lett. 6 (2003) C134–C136. [15] T.P. Moffat, D. Wheeler, D. Josell, J. Electrochem. Soc. 151 (2004) C262–C271. [16] T.P. Moffat, D. Wheeler, M.D. Edelstein, D. Josell, IBM J. Res. Dev. 49 (2005) 19–36. [17] T. Matsuoka, K. Otsubo, Y. Onishi, K. Amaya, M. Hayase, Electrochim. Acta 82 (2012) 356–362. [18] R. Akolkar, U. Landau, J. Electrochem. Soc. 156 (2009) D351–D359. [19] R. Akolkar, V. Dubin, Electrochem. Solid-State Lett. St. 10 (2007) D55–D59. [20] L. Yang, A. Radisic, J. Deconinck, P.M. Vereecken, J. Electrochem. Soc. 160 (2013) D3051–D3056. [21] M.J. Willey, A.C. West, J. Electrochem. Soc. 154 (2007) D156–D162. [22] T. Hayashi, S. Matsuura, K. Kondo, K. Kataoka, K. Nishimura, M. Yokoi, T. Saito, N. Okamoto, J. Electrochem. Soc. 162 (2015) D199–D203. [23] X. Li, T.O. Drews, E. Rusli, F. Xue, Y. He, R. Braatz, R. Alkire, J. Electrochem. Soc. 154 (2007) D230–D240. [24] Z. Zheng, R.M. Stephens, R.D. Braatz, R.C. Alkire, L.R. Petzold, J. Comput. Phys. 227 (2008) 5184–5199. [25] P.C. Andricacos, C. Uzoh, J.O. Dukovic, J. Horkans, H. Deligianni, IBM J. Res. Dev. 42 (1998) 567–574. [26] K. Kobayashi, A. Sano, H. Akahoshi, T. Itabashi, Trench and via filling profile simulations for copper electroplating process, IEEE International Interconnect Technology Conference 2000, pp. 34–36. [27] Y. Cao, P. Taephaisitphongse, R. Chalupa, A.C. West, J. Electrochem. Soc. 148 (2001) C466–C472. [28] A. Pohjoranta, R. Tenno, J. Math. Chem. 52 (2014) 1414–1440.