Dielectric anisotropy in the integration of Cu-SiLK (TM) system

(1)

Dielectric anisotropy in the integration of Cu–SiLK

e

_system

Hai-Sin Tseng

a

, Bi-Shiou Chiou

a,b,*

, Wen-Fa Wu

b

, Chia-Cheng Ho

a a_{Department of Electronics Engineering and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan}

b_{National Nano Device Laboratories, Hsinchu, Taiwan}

Received 1 December 2006; received in revised form 27 February 2007; accepted 18 April 2007 Available online 13 May 2007

Abstract

As device density and performance continue to improve, low dielectric constant (k) materials are needed for interlevel dielectric (ILD) applications. The dielectric anisotropy of polymers with low k is an important property to consider for developing ILD. This is on-going research on the integration aspects of Cu–SiLKTM_{system. In this study, the dielectric anisotropy of SiLK polymer was evaluated with two} test structures: the metal–insulator–metal (MIM) parallel capacitor structure for the out-of-phase dielectric constant (k?) and

comb-and-serpentine interdigitated structure for the in-plane dielectric constant (kk). A k?of 2.65, a kkof 2.75, and a dielectric anisotropy of 3.77%

were obtained for SiLK. However, SiLK exhibits larger leakage current as compared to amorphous SiO2ﬁlms. The reliability issue on

Keywords: Dielectric anisotropy; Cu; SiLK; Cu-low k integration

1. Introduction

As interconnect feature size decreases and clock fre-quency increases, interconnect RC time delay and current density increment become the major limitations on achiev-ing high circuit speeds and reliability. Small capacitance (C) between interconnects is required to reduce the cross-talk, insertion loss, and RC delay associated with the metal interconnect system. Therefore, the interconnect with low dielectric constant (k) material are required. Polymers, with low processing temperatures, ease of application, and good surface planarization, have attracted much attention in the application for low dielectric constant materials. However, polymer thin ﬁlms are anisotropic due to the preferred

chain orientation in the ﬁlm plane and a 21% diﬀerence

between the in-plane dielectric constant and the out-of-plane dielectric constant was reported for a ﬂuorinated

polyimide (DuPont EPI-136M)[1,2]. Hence, the dielectric

anisotropy of low dielectric constant polymers is an impor-tant parameter in selecting low dielectric consimpor-tant materi-als. Besides, the integration of low dielectric constant materials is a critical parameter in controlling electrical performance because it aﬀects the propagation delay, crosstalk, and power dissipation of the integrated circuits

[3–8].

This is on-going research on the integration aspects of Copper–SiLK systems. SiLK (trademark of the Dow Chemical Company) is a low-molecular-weight aromatic thermosetting polymer. SiLK ﬁlms are one of the most attractive interlayer dielectrics because of their good sur-face planarization characteristics, low dielectric constant

and high toughness[9,10]. In the previous work [11], the

thermal characteristics of SiLK are investigated to evaluate the feasibility of SiLK for low dielectric constant materials. Besides, the electromigration in Cu with SiLK passivation

was studied and the mechanism is explored[5,11]. In this

study, the dielectric anisotropy of SiLK is investigated. In the application of low dielectric constant polymer, it is advantageous in process integration to employ an inorganic

*

Corresponding author. Address: Department of Electronics Engineer-ing and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan.

E-mail address:bschiou@mail.edu.tw(B.-S. Chiou).

www.elsevier.com/locate/mee Microelectronic Engineering 85 (2008) 104–109

(2)

liner such as SiO2or Si3N4. The introduction of liner helps

in obtaining interconnect patterns with better resolution,

enhancing the dielectric breakdown strength, etc. [12].

However, the liner may behave as a current leakage path. In this work, the leakage currents between SiLK and

inor-ganic liner SiO2are studied, and the pros and cons of using

SiLK as low dielectric constant materials are discussed. 2. Experimental procedures

Four-inch diameter p-type (1 0 0) Si wafers with nominal resistivity of 1–10 X cm were used as substrates. Metal– insulator–metal (MIM) parallel-plate capacitors were pre-pared to measure the out-of-plane dielectric constant. A 30 nm Ta barrier layer and a 600 nm Cu ﬁlm were sput-tered sequentially onto the substrate to serve as the bottom electrode. SiLK ﬁlms were then spin-coated, baked, and

cured (90 s at 150C followed by 60 s at 325 C followed

by 30 min at 400C) to a thickness of 650 nm. Aluminum

ﬁlms were then deposited as the top electrode. The

out-of-plane dielectric constant (k?) was calculated using the

fol-lowing equation:

k_? ¼ Cd=e0A; ð1Þ

where d is the thickness of the dielectric ﬁlm, C is the

mea-sured capacitance, e0is the permittivity of free space, and A

is the area of the electrode. The amorphous SiO2 ﬁlms,

deposited by the decomposition of tetraethyl orthosilicate, with 500 nm in thickness were deposited onto Cu electrode with PECVD (Multi-chamber PECVD,

STS-MULTI-PLEX CLUSTER SYSTEM, England) at 250C and

100 mTorr. The dielectric constant of amorphous SiO2ﬁlm

was also measured with an MIM structure.

For the evaluation of in-plane dielectric constant (kk)

and interface leakage current, an interdigitated comb and

serpentine structure is employed, as shown in Fig. 1.

Fig. 2gives the ﬂow chart for the preparation of specimens.

A 500 nm SiO2ﬁlms were grown on the Si substrate.

Con-ventional photolithography was used to obtain SiO2

trenches (300 nm in depth) with the interdigitated pattern

shown inFig. 1. Ta (30 nm) and Cu (600 nm) ﬁlms were

then sputtered sequentially to ﬁll oxide trenches and were

annealed at 450C for 60 min in vacuum. Chemical

Serpentine comb2

Comb1 Serpentine

Fig. 1. Comb and serpentine interdigitated test structure for evaluating in-plane dielectric constant and interface leakage current.

Growth of SiO2(~500nm)

Photolithography to obtain oxide trenches

Sputtering of metal (Ta(~30nm) followed by Cu(~600nm))

Vacuum annealing (60 min at 450o_C)

CMP to obtain a smooth surface

Coating of SiLK or SiO2

Lithography to open bond pads Electrical Measurement Overlayer deposition Yes No SiO2 Si substrate

Cu/Ta Cu/Ta Cu/Ta Cu/Ta

Air or SiLK or SiO2

Fig. 2. (a) Flow chart for the preparation of samples and (b) schematic diagram of the sample cross-section.

(3)

mechanical polishing was then employed to obtain a smooth specimen with cross-section shown schematically in Fig. 2b. Some specimens were then coated with SiLK

or SiO2. The thickness of the coating is650 nm. An

addi-tional photolithography was used to open bond pads for electrical testing. A C–V analyzer (model 590, Keithley,

USA) and a semiconductor parameter analyzer

(HP4155B, Hewlett–Packard Co., USA) were employed to measure the capacitance and the leakage current, respectively.

3. Results and discussion

As described in Section 2, the out-of-plane dielectric

constant (k?) is measured with an MIM parallel-plate

capacitor structure. The k?of SiLK and SiO2is 2.65 and

4.21, respectively. Interdigitated electrode structure, shown inFig. 1, has been used to determine the in-plane dielectric

constant (kk) [1,2]. In order to characterize the dielectric

properties of SiLK in a structure of its actual use, a multi-layer test structure as fabricated as shown schematically in

Fig. 2a. The capacitances between the metal line passivated

with air (i.e., without passivation), SiO2, or SiLK are

mea-sured. The interdigitated metal line structure is used to amplify the capacitance between the metal lines, as shown inFig. 2b. The length of the serpentine metal line is about 400 lm. The capacitance measured, C, includes the

capac-itance contributed by SiO2(Cbottom+ Cside) and the

dielec-tric passivation (Ctop). As shown schematically inFig. 3a,

C equals to the sum of Ctop, Csideand Cbottom, where Cside

is the line-to-line capacitance, and Ctopand Cbottomare the

fringe capacitance. The capacitance of specimens

passiv-ated with air (i.e., unpassivpassiv-ated), SiO2, and SiLK are

22.6 pF, 36.9 pF, and 30.2 pF, respectively. The accuracy of measurements is ±0.1 pF. The dielectric constant of

air is1. Because SiO2is amorphous and without preferred

orientation, the dielectric behavior of SiO2 should be

iso-tropic and the dielectric constant of SiO2 is 4.21 which

obtained from the MIM structure. Because (Cbottom+

Cside) are approximately constant for the three specimens,

Ctop= C (Cbottom+ Cside) is proportional to the

dielec-tric constant of the passivation. Hence, the dielecdielec-tric con-stant of SiLK can be interpolated from the C–k curve

shown in Fig. 3b. The dielectric constant k thus obtained

is 2.70 from SiLK. The k?of SiLK is 2.65. The diﬀerence

between the k?and the k obtained from CtopofFig. 3a

sug-gests that the dielectric behavior of SiLK is anisotropic.

The capacitance Ctop consists of relatively large fringe

capacitance from the extension of electric ﬁelds around the metal lines. However, it is beyond the scope of this research to analyze the electric ﬁeld distribution inside

the dielectric layer of the capacitor Ctop and to calculate

the in-plane dielectric constant kkof SiLK on the basis of

the gross dielectric constant k (2.70), the out-of-plane

dielectric constant k? (2.65) and the electric ﬁeld. Since

the gross dielectric constant k is within the range of k?and

kk. It is estimated that the gross dielectric constant k is

around the average of k?and kk. The kkthus obtained is

2.75.

The dielectric anisotropy is attributed to the preferred chain orientation in the plane of the polymeric thin film, resulting in properties in the film thickness direction differ-ent from those in the film plane. Cho et al. reported that molecular structure affected the dielectric anisotropy of polymers. Rigid rod-like polymers, such as fluorinated polyimide, EPI-136M, have a strong propensity to align parallel to the substrate when solution cast due to a sub-strate confinement effect, while flexible chain polymers, such as fluorinated poly(aryl ethyl) (FLARE-1.51), have a smaller propensity to align parallel to the substrate and

are more likely isotropic[1].Table 1summarizes the

chem-ical structure and the dielectric constant of four low k

poly-mers. Fig. 4 exhibits the percent anisotropy ((kk k?)/

k?· 100%) as a function of weight and/or length of the

monomer. The anisotropy data shown in Table 1 and

Fig. 4 are derived from three separate studies with three

diﬀerent structures of multilayer test vehicles (Ref. [1,2]

and this work). Polymers with low monomer weight (<400 g/mol) exhibit small anisotropy, but no speciﬁc trend

passivation 1 Air SiLK SiO 2 C (nF) k 0.025 0.030 0.035 0.040 0.045 0.050 measurement Metal Metal SiO2 Si substrate Ctop Cside Cbottom

C=Ctop+Cside+Cbottom

(Cbottom+ Cside) Ctop

2 3 4

Fig. 3. (a) Schematic diagram of the capacitance between metal lines, and (b) capacitance vs. dielectric constant plot for specimens with various passivations. The measured capacitance is C, C = Ctop+ Cside+ Cbottom,

and (Cside+ Cbottom) is constant, Ctop= C (Cbottom+ Cside) is

propor-tional to the dielectric constant of the passivation, so the dielectric constant of SiLK can be obtained by the C–k plot.

(4)

is observed between the anisotropy and weight of mono-mer, probably due to the experimental errors and/or struc-tural diﬀerences. However, the anisotropy increases with further increase of monomer weight, the higher the weight of monomer, the larger the dielectric anisotropy, as can be

observed fromTable 1 andFig. 4. Factors that aﬀect the

evaluation of the in-plane dielectric constant and the anisotropy include measurement error (<0.5% in this study), the estimate that the gross dielectric constant is

about the average of kk and k?, the structure of the test

vehicles, such as: aspect ratio between the electrode spacing and the dielectric thickness, the hierarchy of the various dielectric layers with respect to metallization, the depen-dence of the lateral capacitance on the thickness of the dielectric between metal trenches, etc.

A dielectric material reacts to an electrical field differ-ently from a free space because it contains charge carriers that can be displaced (i.e., polarized), and charge displace-ments within the dielectric can neutralize a part of the applied field, and, consequently, increase the amount of charge stored. There are various possible mechanisms for polarization in a dielectric material, such as: electronic polarization, atomic polarization, molecular (orientation)

polarization, and space charge polarization. The dielectric anisotropy of the polymer is resulted from the molecular polarization which has a relaxation time corresponding to the particular material system and, in general, cannot fol-low the electric ﬁeld when the applied frequency exceeds

1010Hz. In this study, the applied frequency is 105Hz,

hence, molecular polarization contributes to the anisotropy of SiLK.

The leakage current of specimens with SiLK or SiO2

overlayer is evaluated with the comb and serpentine

inter-digitated structure shown in Fig. 1. The metal lines were

coated with SiLK or SiO2as described in the experimental

procedures. Fig. 5 exhibits the comb current Icomb as a

function of serpentine voltage Vserp for specimens with

the diﬀerent coatings. It is obvious that the leakage cur-rents of specimens with SiLK overlayer are larger than

those of specimens with SiO2overlayer. There are various

paths for current to ﬂow, such as: through the interface of the overlayer and the underlayer, through the bulk of overlayer, and/or through the underlayer, as shown

sche-matically inFig. 6a. If the majority current ﬂows through

the bulk, the leakage current should be approximately inversely proportional to the length of the path, since the

Table 1

Chemical structure and dielectric constant of some low k polymers

Polymer Structure kk k? % Anisotropy,

Dk= (kk k?)/k? (%) Weight of monomer (i.e., n = 1) (g/mol) Length of monomer (nm) Note

SiLK 2.75 2.65 3.77 216 0.464 This work

and Ref.[9]

Fluorinated polyimide (DuPont EPI-136M)

2.790 2.305 21.04 1046 0.57 Ref.[1]

Poly(aryl ethyl) (allied signal FLARE-1.51) 2.565 2.618 1.45 328 0.424 Ref.[1] Aromatic polyimide, BPDA-PDAa 3.3 3.1 6.45 356 0.26 Ref.[2] a Poly(p-phenylene bihenyltetracarboximide).

(5)

resistance is proportional to the length. As shown

schemat-ically in Fig. 6b, Iserp should be approximately twice of

Icomb2. However, if interface current ﬂow dominates, then

there is no apparent relation between leakage current and path length. To identify the major current leakage path,

voltage was applied onto pad of comb1, and currents were measured at pads of comb2 and serpentine. As observed in

Fig. 7, Iserp is much larger than Icomb2, and this suggests

that the SiLK/SiO2interface is the major path for current

leakage ﬂow. 200 400 600 800 1000 0 5 10 15 20 25 EPI-136M [1] SiLK FLARE-1.5 [1] BPDA-PDA [1] % Anisotropy

Weight of monomer (g/mole)

0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0 5 10 15 20 25 FLARE-1.5 [1] SiLK BPDA-PDA[1] EPI-136M [1] % Anisotropy Length of monomer (nm)

Fig. 4. Percent anisotropy (kk k?)/k?· 100% as a function of: (a)

length of monomer and (b) weight of monomer of four low k polymers.

0 10 0 200 400 600 800 Icomb (pA) Vserp(V) SiO2 SiLK 2 4 6 8

Fig. 5. Icombas a function of Vserpfor specimens coated with SiO2or

SiLK.

SiO2 or SiLK

comb1 serpentine comb2

(5) (3) SiO2 Si substrate (4) (2) (6) (1) comb2 SiO2 Si substrate SiO2 or SiLK C2 R2 C1 R1 C5 R5 C6 R6 I(2) I(1) I(6) comb1 serpentine I(5)

Fig. 6. (a) Partial cross-section of the specimen and possible current leakage paths indicated by arrows. Current can ﬂow through the overlayer (1, 2), the interface between the overlayer and the underlayer (3, 4), and the underlayer (5, 6). That is Iserp= I(1)+ I(3)+ I(5), and I com-b2= I(2)+ I(4)+ I(6). (b) Equivalent circuit model when current ﬂown

through bulk layer is the major path. That is Iserp Ið1Þþ Ið5Þ, and

Icomb2 Ið2Þþ Ið6Þ. Rirepresents the resistance to current ﬂow via path i

and is proportional to the length of the path. Because the distance between comb1 and comb2 is about twice that between comb1 and serpentine, so Ið1Þ 2Ið2Þ, Ið5Þ 2Ið6Þ, and Iserp 2Icomb2, if current ﬂown through bulk is

the major leakage path.

0 10 0 200 400 600 800

leakage current (pA)

Vcomb1(V)

Iserp

Icomb2

2 4 6 8

Fig. 7. Leakage current Iserp and Icomb2 as a function of Vcomb1 for

(6)

Table 2 gives a comparison between Cu–SiLKTM _and

Cu–SiO2systems. The low dielectric constant of SiLK

ren-ders it a good candidate as interlevel dielectric, however, the larger leakage current, higher thermal impedance, and the poor electromigration resistance of Cu passivated with SiLK cast the reliability concerns for Cu–SiLK system. 4. Conclusions

The dielectric anisotropy of SiLK ﬁlms is studied. The

out-of-plane dielectric constant (k?), measured with an

MIM structure, is 2.65. The in-plane dielectric constant

(kk) was evaluated with a comb and serpentine

interdigita-ted structure and an estimate to the gross dielectric

con-stant on the basis of kkand k?. The kk obtained is 2.75.

The dielectric anisotropy of SiLK, 3.77% is attributed

to the molecular polarization and should fade away at high

frequencies (>1010Hz). The low dielectric constant renders

SiLK a good candidate to replace SiO2 for low dielectric

constant material. However, Cu passivated with SiLK exhibits larger leakage current, higher thermal impedance, and shorter electromigration lifetime than that passivated

with SiO2. Hence, there is a reliability concern for the

inte-gration of Cu–SiLK system. Acknowledgement

This work is sponsored by National Science Council, Taiwan, under the Contract Nos. NSC91-2216-E-009-023 and NSC-92- 2216-E-009-007.

References

[1] T.H. Cho, J.K. Lee, P.S. Ho, E.T. Ryan, J.G. Pellerin, J. Vac. Sci. Technol. B 18 (2000) 208–215.

[2] A. Deutsch, M. Swaminathan, M.H. Ree, C. Surovic, G. Arjavalin-gam, K. Prasad, D.C. McHerron, M. McAllister, G.V. Kopcsay, A.P. Giri, E. Perfecto, G.E. White, in: Proceedings, the Topical Meeting on Electrical Performance of Electronic Packaging, Monterey, Califor-nia, 1993, pp. 151–154.

[3] J.J. Senkevich, T. Karabacak, D.L. Bae, T.S. Cale, J. Vac. Sci. Technol. B 24 (2006) 534–538.

[4] J. Liu, M. Scharnberg, J. Bao, J. Im, P.S. Ho, J. Vac. Sci. Technol. B 23 (2005) 1422–1427.

[5] L.M. Gignac, C.K. Hu, E.G. Liniger, Microelectron. Eng. 70 (2003) 398–405.

[6] C.C. Ho, B.S. Chiou, Microelectron. Eng. 83 (2006) 528–535. [7] C.C. Ho, B.S. Chiou, Microelectron. Eng. 84 (2007) 646–652. [8] W.L. Sung, B.S. Chiou, J. Electron. Mater. 31 (2002) 472–477. [9] S.J. Martin, J.P. Godschalx, M.E. Mills, E.O. Shaﬀer, P.H.

Town-send, Adv. Mater. 23 (2000) 1769–1778.

[10] O. Demolliens, P. Berruyer, Y. Morand, C. Tabone, A. Roman, M. Cochet, M. Assous, H. Feldis, R. Blanc, E. Tabouret, D. Louis, C. Arvet, E. Lajoinie, Y. Gobil, G. Passemard, F. Jourdan, M. Moussavi, M. Cordeau, T. Morel, T. Mourier, L. Ulmer, E. Sicurani, F. Tardif, A. Beverina, Y. Trouillet, D. Renaud, in: Proceedings, International Interconnect Technology Conference, 1999, pp. 198– 199.

[11] H.S. Tseng, B.S. Chiou, W.F. Wu, C.C. Ho, J. Electron. Mater. 33 (2004) 796–801.

[12] J.D. Plummer, M. Deal, P.B. Griﬃn, Silicon VLSI Technology: Fundamentals, Practice and Modeling, Prentice Hall, 2000, p. 58. Table 2

Dielectric properties of SiLK and SiO2as well as reliability issues[11]for

Cu–SiLK and Cu–SiO2system

SiLK SiO2 Dielectric constant kk 2.75 4.2 k? 2.65 4.2 % Dielectric anisotropy (kk k?)/ k?· 100% 3.77 0

Dielectric leakage current Icombat

Vserp= 6 V

(pA) (Fig. 5)

568 112

For reliability issues[11]

Cu-passivated with SiLK

Cu-passivated with SiO2

Thermal impedance of Cu,C/W 1832 1604

Electromigration (EM) lifetime of Cu interconnects at 2.8· 106 A/cm2 225C 4.2· 105 s 5.6· 105 s 250C 2.1· 105 s 3.1· 105 s 300C 2.9· 104 s 5.4· 104 s

Activation energy Q for EM of Cu (eV) 0.71 0.89

Predicted EM lifetimeaof Cu interconnect at 2.8· 106 A/cm2 100C 6.49· 107 s 8.51· 108 s 25C 1.69· 1010 s 9.04· 1011 s

a _{The predicted EM lifetime is calculated on the Arrhenius equation:}

t_{T 1} t_{T 2}¼ exp Q kT1 Q kT2 .