A MODEL OF COPPER DEPOSITION FOR THE DAMASCENE PROCESS C. Gabrielli, J. Kittel, P. Moçoteguy, H. Perrot UPR15 du CNRS, Physique des Liquides et Electrochimie Université Pierre et Marie Curie 75252 PARIS, France ω
A. Zdunek†, P. Bouard , M. Haddix‡, L. Doyen₣ † AIR LIQUIDE, Chicago Research Center, Countryside, IL 60525, ω Centre de Recherche Claude Delorme, 78354 Jouy en Josas, France, ‡Dallas Research Laboratory, Dallas, TX 75243, ₣European Analytical Services, 91105 Corbeil-Essones, France M.C. Clech ALTIS Semiconductor; Corbeil Essones, France A model of copper deposition in the damascene process is proposed and analysed in terms of electrochemical impedance. Experimental impedance results are compared with theory for the copper sulfate-sulfuric acid bath containing chloride, PEG and MPSA and for an industrial copper interconnect bath. Good agreement was found with the proposed model. INTRODUCTION The damascene process for fabrication of copper on-chip metal interconnects requires electrodeposition into trenches or vias with width dimensions on the order of 130 nm or lower (1,2). To obtain void-free deposits, superconformal deposition, or superfilling, is necessary. These terms refer to the occurrence of more rapid electrodeposition in the bottom of a trench or via than toward its entrance. Superfilling is only obtained in the presence of a certain combination of additives in the electroplating bath. Among those studies available in the literature, Moffat et al. (3,4,5) have shown that baths with three additives in a classical copper sulfate-sulfuric acid deposition bath can be used to achieve superfilling in sub micrometer trenches and vias. The additives used were a polyether (polyethylene glycol, PEG), chloride ions, and a thiol (3-mercapto-1propanesulfonate, MPSA). From the kinetic point of view, Chassaing and Wiart (6) have shown that, for the copper sulfate-sulfuric acid deposition bath without organic additives, the deposition of polycrystalline copper takes place mainly by a direct charge transfer reaction at growth sites. The overall deposition rate is controlled by the crystallographic growth. When organic additives are present, adsorption mechanisms become the rate controlling steps to the copper deposition reaction. Chlorides ions are usually added to copper electrolytic baths to combine with generated cuprous ions and avoid Cu+ disproportionation. Cl- is known to adsorb on copper surfaces and complexes with the intermediate Cu+ ion. In this manner, Cl- reduces the importance of the Cu+ species with acceptable formation of a CuCl film. It is generally acknowledged that chloride ions are
necessary for the adsorption of PEG on the copper surface (7). The formation of a CuCl phase on the copper surface has been demonstrated by using radiotracer techniques (8). Organic additives often adsorb on the cathode surface. The presence of adsorbed molecules reduces the active surface of the electrode and the existence of adsorbed films has been reported with long chain PEG (9). According to Goldbach et al (9), addition of very low amounts of chloride to a polyalcohol-containing bath usually results in a strong inhibition of copper deposition by allowing the existence of bridge-like structures, for instance, Cu ions-Cl-polymer. Consequently, copper deposition occurs at higher rates than with only PEG. (9). Concerning the additive MPSA, it has been proposed that its slow adsorption competes with and displaces adsorbed PEG. The competitive adsorption of PEG and MPSA results in an inhibition of copper deposition by PEG adsorption and an activation by the adsorbing thiol, MPSA. MPSA seems to be incorporated into the deposited copper film (10,11). A model of copper deposition in the damascene process is proposed below and analysed in terms of electrochemical impedance. Few publications have reported investigation of the damascene process by impedance techniques (12, 13). Experimental results are compared with theory for the acid copper sulfate bath containing chloride, PEG and MPSA and for both fresh and aged industrial baths. THEORY The proposed model of copper deposition in the damascene process when mass transport is not a limiting step is: k1 Cu 2+ + (∗) + e− → Cu + (∗) +
−
[1]
k2
Cu (∗) + e → Cu
[2]
where (∗) is an active growth site and Cu + (∗) is an adsorbed cuprous ion on a growth site. The adsorption reactions for the various additives are: 3 PEG + Cu ++ + (∗) + e − → PEG-Cu + (∗)
k
MPSA+Cu
++
−
k4
+
+ (∗) + e → MPSA-Cu (∗)
[3] [4]
To simplify the model, the role of chloride is limited to enhancing adsorption of PEG and changing the copper deposition transfer coefficient. It has been shown that the CuCl film from chloride ion adsorption has a negligible hindrance to charge transfer (9). Thus, an adsorption reaction with chloride ion is not used in the model. The competitive adsorption of MPSA-Cu + (∗) and PEG-Cu + (∗) can be written: 5 MPSA-Cu + (∗)+PEG-Cu + (∗) → PEG+MPSA-Cu + (∗) + Cu + (∗)
k
[5]
where PEG-Cu + (∗) , and MPSA-Cu + (∗) are the adsorbed PEG and MPSA with surface coverage θ 2 and θ3 , respectively. In this model, PEG adsorption is potentialindependent. Finally, the incorporation of MPSA into the deposit is written: 6 MPSA-Cu + (∗) + e− → MPSA-Cu + (∗)
k
[6]
Thus, equations [2] and [6] represent two parallel reactions for copper deposition, one with incorporation of an additive. On the other hand, the adsorbed PEG, ( PEG-Cu + (∗) ), blocks the surface and does not participate in a charge transfer reaction or become incorporated into the deposit. Using the above model, an expression can be written for the faradaic current as a function of species concentrations, surface coverage of the adsorbed species, and the reaction rate constants:
I F = F ( k1 + k3c2 + k4 c3 )(1 − θ1 − θ 2 − θ 3 ) + k2θ1 + k6θ 3
[7]
The time-dependent equations for surface coverage follow:
dθ1 = k1 (1 − θ1 − θ 2 − θ 3 ) − k2θ1 + k5θ 2θ 3 dt dθ 2 = k3c2 (1 − θ1 − θ 2 − θ 3 ) − k5θ 3θ 2 dt dθ 3 = k4 c3 (1 − θ1 − θ 2 − θ 3 ) − k6θ 3 dt
[8] [9] [10]
where 1 – θ1 – θ2 – θ3 corresponds to the free copper electrode surface. The model was then solved for the steady-state condition (dθi/dt = 0) and the dynamic regime (electrochemical impedance), where the change in ∆θi with time is a function of the alternating current frequency (d (∆θi)/dt = jω∆θi). The measured electrochemical impedance, which takes into account the double-layer capacity, Cd, is equal to: Z (ω ) =
1 jω Cd
Z F−1 +
[11]
and the faradaic impedance, ZF-1, becomes, after linearization of the state equations,
Z F−1 =
∆I F = Rt−1 + F T1 ( − k1 + k2 − k3c2 − k4c3 ) + ( − k1 − k3c2 − k4c3 ) T2 + T3 ( − k1 + k5 − k3c2 − k4 c3 ) ∆E [12]
where,
Rt−1 = F ( b1k1 + b3k3c2 + b4 k4 c3 )(1 − θ1 − θ 2 − θ 3 ) + b2 k2θ1 + b6 k6θ 3
[13]
and
Ti =
∆θ i ∆E
i = 1, 2, 3
[14]
Also, the potential dependent rate constants, ki, are:
ki = kio exp(biE )
[15]
Figure 1 shows an example of the calculated impedance from this model using the following parameter values: c1 = c2 = c3 = 0.001 , k10 = 2,5 10−9 , b1 = −10V −1 , k10 = 6,8 10−9 , b2 = −30V -1 , k20 = 6, 0 10−10 , k30 = 1,15 10−6 , b3 = 1 , k40 = 1,5 10−12 , b4 = −15V −1 , k5 = 5 10−6 , k60 = 1, 75 10−12 , b6 = −10V −1 . The impedance predicted by the model shows four loops corresponding to four time constants, three that are capacitive and one that is inductive in the lower frequency range as seen in Figure 1. 50 40
Loop 1
30
Loop 2
1000 Hz
2
20 -Im( ∆ E/∆ I)/Ω cm
Loop 3
10 Hz
10
100 mHz
100 Hz
0
1 Hz 10 mHz
-10 -20 -30
1 mHz
-40 -50
Loop 4
0
20
40
60
Re( ∆ E/∆ I)/Ω cm
80
100
2
Figure 1. Calculated impedance from the model.
EXPERIMENTAL The impedance of the electrode was measured during superconformal copper deposition both in “standard” bath given by Moffat et al. (3) and in a fresh industrial bath containing proprietary additives. Electrochemical impedance spectroscopy (EIS) measurements were performed in a standard three-electrode cell with a copper plate (99.999%, Goodfellow) counter electrode and a saturated sulfate reference electrode (SSE). The working electrode was a copper disk (5 mm diameter) embedded in insulating epoxy resin, and was polished with 1200 grade SiC before each experiment. In order to eliminate mass transport control, a rotating disk was used.
The nominal electrochemical bath composition for the “standard” bath was 0.25 mol/L CuSO4, 1.8 mol/L H2SO4, 10-3 mol/L NaCl, 88 µmol/L PEG (3400 g/mol), and 10 µmol/L MPSA. In order to evaluate the influence of the additives, different combinations of chloride, PEG and MPSA concentrations were studied while the sulfuric acid and copper sulfate concentrations were held constant. Unless otherwise stated, the individual concentrations of the additives in this acid copper sulfate solution are those of the reference electrolyte. In order to minimize the influence of the ageing of the solution or of eventual by-products, fresh solutions were prepared and used within a 24-hour delay.
RESULTS EIS Results in a Standard Copper Bath The EIS spectra shown in Figure 2, in the standard copper bath containing Cl-, revealed four time constants. The high frequency (HF) loop, around 10 KHz, is representative of the charge transfer resistance (Rt) and of the double layer capacitance (Cd) at the cathode interface. The intermediate capacitive loop, around 10 Hz, also observed by other workers (6,12,14) may be associated with the presence of chloride in the solution. For instance, an EIS measurement performed under the same conditions as Figure 2 in an electrolyte without Cl- does not exhibit loop 2 (Figure 3a). Additions of Cl- and PEG (Figure 3b), or Cl-, PEG and MPSA (Figure 3c) give similar spectra to Figure 2.
10k
2
- Im aginary ( Ω .cm )
1
1k
10
1
0.1
0
1m
0
1
2
3
2
Real ( Ω .cm )
Figure 2. EIS spectra of copper deposition at 100 rpm and 25 mA/cm2 in acidic copper sulfate electrolyte with 10-3 M NaCl. To study more accurately the chloride ion effect on charge transfer kinetics, the working electrode rotation speed was increased from 100 to 2000 rpm so as to suppress diffusion limitation effects. The comparison between the deposition in the solution with or without chloride (Figure 4a and 4b) confirms that the last capacitive loop around 10Hz is related to this additive. The absence of chloride also results in an increased charge transfer resistance, as determined by the diameter of the HF loop: its value is around 2.5 Ω.cm2 and decreases to 1.5 Ω.cm2 when chloride is added to the electrolyte. Similar evolution of the double layer capacitance is observed: it varies from 70 µF.cm-2 to 20 µF.cm-2 when chloride is added to the solution, indicating an interfacial effect.
a) without Cl
2
- Im ( Ω .cm )
2 -
1k
1
100m
100 1
10m
0 0
1
2
3
2
Real ( Ω .cm )
2
- Im ( Ω .cm )
2
- Im ( Ω .cm )
-
b) with Cl and PEG
1
100m
10k
10 1
0 0
1
2
5
10k
1
10
100m
1
0 -
c) with Cl , PEG and M PSA
3
2
Real ( Ω .cm )
4
0
1
2
3
4
2
Real ( Ω .cm )
Figure 3. EIS spectra of copper deposition at 100 rpm and 25 mA/cm2 in acidic copper sulfate electrolyte, a) without additives, b) with 10-3 M NaCl and 88 µM PEG, and c) reference solution. The activation of the electrode area with increasing overpotential appears as a reasonable explanation for the low frequency inductive behavior in electrochemical impedance measurements performed on copper bath electrolytes with additives. However, it is not possible to make a clear distinction between a slow increase of the surface area due to the nucleation and development of growth centers and a slow removal of inhibiting species such as organic molecules (6). The faradaic impedance reduces to the transfer resistance with a high purity copper sulfate-sulfuric acid solution, which is consistent with the identical transfer coefficient in the two, one-step reactions of copper reduction. With the addition of chloride, the existence of a small capacitive loop suggests a change in the values of the transfer coefficient, where the loop of the second reaction is higher than the capacitive loop of the first one. 1
10k
100 1
0
a) without Cl 0
0.1
-
1
22
Real ( Ω .cm )
1k 10
0
0
1
-
0.1
1
2
b) with Cl
3
1
3
2
Real ( Ω .cm )
2
- Im ( Ω .cm )
1
2
- Im ( Ω .cm )
10k
2
- Im ( Ω .cm )
1k
2
- Im ( Ω .cm )
1
10k
1k
10 1
0 -
c) with Cl and PEG 0
1
2 2
Real ( Ω .cm )
3
10k
10 1k
1
0 -
d) with Cl , PEG and MPSA 0
1
2
2
Real ( Ω .cm )
3
Figure 4: EIS spectra of copper deposition at 2000 rpm and 25 mA/cm2 in acidic copper sulfate electrolyte, a) without additives b) with 10-3 M Cl-, c) with 10-3 M Cl-, and 88 µM PEG and d) in reference solution.
PEG addition also induces a visible modification of the EIS spectra: the LF inductive loop is significantly reduced. This EIS response clearly suggests that PEG is involved in an adsorption mechanism. It is well known that PEG inhibition efficiency is strongly correlated to the presence of chloride (9). This induces an important evolution of the impedance when increasing chloride concentration in an electrolyte already containing PEG (Figure 5), or when increasing PEG concentration in an electrolyte already containing chloride (Figure 6). The presence of the inductive loop, and the evolution of its radius size relative to that of the capacitive loops, might be used to evaluate the concentration of the two concerned additives.
1
100
10k 0
a) without Cl 0
-
1
0.1
1
22
100
10k 0
0
1
0.1
-
b) Cl / 50
3
Real ( Ω .cm )
1
2
3
2
Real ( Ω .cm )
1 2
2
- Im ( Ω .cm )
1
- Im ( Ω .cm )
1k
2
- Im ( Ω .cm )
1k
2
- Im ( Ω .cm )
1
10k 10
1k 0
1
-
c) Cl / 2 0
0.1 1
2
Real ( Ω .cm )
2
10k 1k 0
d) with Cl 0
10 1
-
1
2
Real ( Ω .cm )
2
Figure 5: EIS spectra of copper deposition at 2000 rpm and 25 mA.cm-2 in acidic copper sulfate electrolyte with 88 µM PEG and: (a) no chloride, (b) chloride diluted 50 times against reference solution, (c) chloride diluted twice against reference solution, (d) 10-3 M chloride (reference concentration). EIS Results in an Industrial Bath Figure 7 shows the results of the impedance measured in a fresh industrial solution before charge (no copper deposition). Figure 8 shows the impedance obtained just after charge end on the same solution aged in representative industrial operating conditions (charge at 16 mA/cm2 for 16 hours, corresponding to a total charge amount of 12.5A.h/L). In both cases, the impedance with four time-constants was found, 3 capacitive loops and a lowfrequency inductive loop. Thus, the model suggested in this work generally agrees with the impedance diagrams of “new” or standard bath solutions, but also on industrial bath solutions that have been aged and which contain by-products using conditions simulating superconformal copper deposition.
- Im ( Ω .cm )
1
10k 1k 0
10 0.1
a) without PEG 0
1
2
Real ( Ω .cm )
1
1k
10
100
0
1 0.1
b) PEG / 200
2
0
1
2
3
2
Real ( Ω .cm )
1 2
2
- Im ( Ω .cm )
1
- Im ( Ω .cm )
10k
2
2
- Im ( Ω .cm )
1
10k
10
1k 0
0.1 1
c) PEG / 20 0
1
2
Real ( Ω .cm )
10k 10
1k 0
2
1
d) with PEG 0
0.1
1
2
Real ( Ω .cm )
2
Figure 6: EIS spectra of copper deposition at 2000 rpm and 25 mA.cm-2 in acidic copper sulfate electrolyte with 10-3 M Cl- and: (a) no PEG, (b) PEG diluted 200 times against reference solution, (c) PEG diluted 20 times, (d) 88 µM PEG (reference concentration).
-Im( ∆ E/∆ I)/Ω .cm
2
Before charge 10 KHz
1
0 1 mHz 10 mHz -1 0
1
2
Re( ∆ E/ ∆ I)/ Ω .cm
3 2
Figure 7. EIS spectra of a fresh industrial solution (before charge).
4
100 mHz 10kHz
2
-Im(∆E/∆I) (Ω.cm )
2
0
10 mHz
-2 0
2
4
6
8
2
Re(∆E/∆I) (Ω.cm ) Figure 8. EIS spectra of an aged industrial solution (16 hours at 16 mA/cm2, with total charge amount of 12.5A.h/L), just after ageing charge end.
CONCLUSION Experimental electrochemical impedance results are in general agreement with the model proposed in this text. The tested reaction mechanism explains the occurrence of 3 capacitive loops and an inductive loop in the lower frequency range in the impedance diagram. The higher frequency loop is due to the charge transfer resistance in parallel with the double layer capacitance. The following capacitive loop is related to the presence of chloride in the bath solution. The low frequency capacitive and inductive loops are related to the accelerator and suppressor additives; the capacitive loop to the suppressor (e.g., PEG in the standard bath) and the inductive loop to the accelerator (e.g. MPSA in the standard bath). In addition, this model is applicable to industrial solutions containing proprietary additives and can be a basis for interpreting the role of the additives in copper interconnect bath solution ageing.
ACKNOWLEDGEMENTS The authors would like to acknowledge ALTIS Semiconductor for providing copper baths samples for the experimental work.
LIST OF SYMBOLS θi: fraction of active surface covered by species i, where: θ1 = surface coverage of Cu+ (*) θ2 = surface coverage of PEG - Cu+ (*) θ3 = coverage MPSA- Cu+ (*) + Cu (*): Cu+ adsorbed at a growth site PEG - Cu+ (*): complex of PEG and Cu+ adsorbed at a growth site MPSA- Cu+ (*): complex of MPSA and Cu+ adsorbed at a growth site (*) : active growth site at electrode surface. ci: concentration of species i in solution, where: c1 = (Cu++) c2 = (PEG) c3 = (MPSA) kj : kinetic constant of charge transfer reaction nr j (as defined in the brackets close to each reaction equation). IF: faradaic current F: Faraday constant (= 96485 C/mol) ZF: Faradaic impedance linked to charge transfer reaction between reactive species and electrode material Cd: double-layer capacitance Rt: charge transfer resistance corresponding to the diameter of loop 1 in EIS spectra REFERENCES 1. P. C. Andricacos, The Electrochem. Soc. Interface, 32-37, Spring (1999). 2. A. E. Braun, Semiconductor International, 58-66, April (1999). 3. T. P. Moffat, D. Wheeler, W. H. Huber, D. Josell, Electrochem. Solid-State Lett., 4, C26 (2001). 4. D. Josell, D. Wheeler, W. H. Huber, J. E. Bonevich, T. P. Moffat, J. Electrochem. Soc. 148, C767 (2001). 5. T. P. Moffat, J. E. Bonevich, W. H. Huber, A. Stanishevsky, D. R. Kelly, G. R. Stafford, D. Josell, J. Electrochem. Soc., 147, 4524 (2000). 6. E. Chassaing and R. Wiart, Electrochim. Acta, 29, 649 (1984) 7. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3472 (1998). 8. G. G. Lang, M. Ujvari, G. Horanyi, J. Electroanal. Chem., 522, 179 (2002). 9. S. Goldbach, W. Messing, T. Daenen, F. Lapicque, Electrochim. Acta, 44, 323 (1998). 10. K. R. Hebert, J. Electrochem. Soc, 148, C726 (2001). 11. S. Soukane, S. Sen, T. S. Cale, J. Electrochem. Soc., 149, C74 (2002). 12. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3477 (1998). 13. G. Fabricius and G. Sundholm, J. Applied Electrochem., 15, 797 (1984). 14. J.D. Reid, and A.P. David, J. Electrochem. Soc., 134 1389 (1987)
