Basic solution of Cu Electrodeposition - Introduction of Cu Electroplating Bath

1.2 Introduction of Cu Electroplating Bath

1.2.1 Basic solution of Cu Electrodeposition

Cu ECD baths are normally formulated using a highly stable acid electrolyte solution containing copper sulfate and sulfuric acid. The basic kinetics and properties of these solutions have been investigated for more than 80 years and well understood ⁷.

In integrated circuit damascene applications, the unique important criterion for copper sulfate concentration is to avoid the depletion of cupric ion during gap-filling in features processes. Typical cupric ion concentrations in use are in the range of 17.5-60 g/L and sulfuric acid is usually added into the ECD electrolyte (45-325 g/L) to maintain solution conductivity and improve wetting or oxide dissolution on seed surfaces.

Generally, more conductivity solutions result in a system where plating thickness distribution is less dependent on plating cell geometry, while low acid electrolytes result in a system with less dependence on seed layer resistivity.

1.2.2 Organic Additives of Plating Baths

Copper Electroplating can provide bottom-up filling or superfilling behavior by adding chemical additives such as poly (ethylene glycol) (PEG) suppressors together with chloride ions, inhibitors, and accelerators, into H₂SO₄-based electrolytes thus resulting in the void-free or super filling of narrow Cu trenches and vias in damascene process ^8-9. This is commonly assigned to the action of organic additives and chlorides ions added in small amounts to the electrodeposition (ECD) bath¹⁰. The mechanisms by which these additives lead to super-fill have been proposed in many investigations. The basic solution and followed by the additives will be introduced in next section.

At present, organic additives added to the electrolyte of Cu ECD lead to three basic categories.

Accelerator which can be called brightener or anti-suppressor, catalyzes and accelerates the conformal overfilling of vias and trenches by locally accelerating current at a given voltage where they are adsorbed. It contains sulfur-containing molecules, typically sulfonic acid groups or disulfides such as SPS (Bis- (sodium

sulfopropyl)-disulfide) with the chemical formula of NaSO3(CH2)3S-S(CH2)3SO3Na.

Accelerators are usually adopted in the plating bath in the concentration range of 1-25ppm ^{8-9, 11-12}.

Suppressors is a kind of surfactant or wetting agent which can be called the carrier, that suppresses Cu growth at the top edges of vias and trenches, thus slowing the deposition rate by competing with copper for electron transfer sites or crystal lattice sites on the metal surface. Surfactants containing long chain polymers such as polyethylene glycol (PEG) or co-polymers of polyoxyethylene and polyoxypropylene have average molecular weights more than 1000. ^13-17

Levelers (also called a grain refiner or over-plate inhibitor) are usually high-molecular weight polymers with amine (-NH3) or amide (-NH2) functional groups.

They are second class of current-suppressing molecules, which are usually added to the plating bath at a low concentration. Hence, unlike suppressors, the concentration of levelers at the interface is mass-transfer-dependent. Therefore, isolated locations such as the inside of a via (where mass transfer is limited) are less suppressed, while protruding surfaces or corners (where mass transfer by diffusion or migration is more efficient) are more suppressed 2-3, 8-9, 12, 17.

1.2.3 Relationship between PEG and SPS

Many researchers suggested one reaction equations for the (PEG, SPS)-containing baths to achieve void-free filling. 9,15, 18-21.

In the statement of Moffat et al, the I-V curves revealed that the competition between the effects of PEG and MPS during electrodeposition²¹. The linear sweep voltammetry (LSV) indicates the competition between inhibition provided by Cl-PEG-Cu²⁺/Cu⁺/Cu interaction and the catalytic effects of Cl-MPS-Cu²⁺/Cu⁺/Cu interaction. A possible reason of the LSV effect entails the following sequence of events.

Potential-driven desorption or disruption of the blocking Cl-PEG-based layer allows thiolate, or a derivative thereof, then to adsorb on the surface. This further disrupts the inhibiting function of the Cl-PEG-based monolayer. N. Kovarsky et. al agreed with this

Furthermore, M. Tan proposed that a transition time of the system to reach steady state was observed under both galvanostatic and potentiostatic conditions and found to be a strong function of SPS concentration ⁸. These experimental results provide evidence for slow adsorption and desorption of the accelerator and SPS incorporation into the deposit. Linear sweep voltammetry (LSV) indicates that behavior of the accelerator is potential dependent.

In the respect for the interaction of SPS in plating bath, the reaction was established in the system containing SPS only ¹⁹

4Cu(I) + SPS Æ 2Cu(I)MPS + Cu(II) [1]

In conclusively of the mentioned above, we would like to introduce the kinetics in SPS system to our two-component-additive system, in agreement with the assumption from Moffat.

The both effects of PEG and SPS on the deposition of Cu from dilute acid sulphate solutions have been studied for several years. When were PEG added alone in the plating bath, a polymer film adsorbed at a metal surface is expected to slow down deposition currents by imposing a dense barrier at the surface and thus lowering the concentration of cupric ion at inner Helmholtz plane. Cupric ions compete with adsorbed PEG and adsorbed cuprous ions for free adsorption site, the adsorbed PEG molecules do not further affect electrode kinetics. The proposed mechanism is :¹³

Cu²⁺ + e^- ÅÆ Cu⁺ads [2]

Cu⁺_ads + e^- ÅÆ Cu [3]

Where Cu⁺_ads denotes an adsorbed cuprous ion, which competes with PEG_adson Cu substrate.

When PEG-containing bath combine with SPS, the v-t curves will be different, as shown in Fig.1-3. During copper ECD, PEGs immediately adsorb on the interface in the potential range and then SPS gradually substitutes the sites of PEGs on the Cu surface.

18-19, 23 Ultimately, equilibrium between the adsorption of these two additives is accomplished. Possible reactions are:

SPS Æ (driven by ECD current) Æ 2MPS [4]

4MPS + 2Cu²⁺ Æ 2Cu(I)MPS + SPS [5]

Thus the complex Cu(I)MPSs gradually replace the adsorbed PEGs, and then reduce the inhibition as well as enhances depolarization. This model is called slow adsorption/desorption mechanism. A schematic diagram is shown below in the Fig. 1-4.

The relative information to SPS and MPS ¹⁷are listed in Table Ⅰ.

Fig.1-3. Evolution of v-t curves of PEG bath and SPS bath during ECD, respectively.

3-6 Illustration of the slow adsorption/desorption mechanism

Fig.1-4. Illustration of the slow adsorption/desorption mechanism¹⁷

0 100 200 300 400 500 600

0.14 0.16 0.18 0.20 0.22 0.24 0.26 0.28

After SPS 6ppm added

Cell voltage (V)

Deposition time (s) PEG-containing bath

Table Ⅰ. Relative information to SPS and MPS

1.2.4 Filling Model of Cu Line/Via

Accelerated bottom-up deposition has been explained by a local accumulation of the accelerator species at the deep of a feature, as the surface area within the feature decreases during deposition process ^{11, 24}. Recently, the bottom-up gap filling of Cu ECD with plating additives has been clarified by growth-accelerator species such as SPS and its byproduct such as MPS at the bottom of small features and slow diffusion of free MPS out of these features. The reduction of SPS to MPS provides a possible catalytic path for copper deposition through the formation of cuprous thiolate ¹⁰. The model of overfilling is described step-by-step in Fig. 1-5 ¹⁰. As current flow was applied in electrolyte, it is assumed that all additive species have reached an equilibrium level on all surfaces of the wafer. In the initial stage, applied currents should be expected are approximately equivalent on all surfaces. This effect is corresponded with the observing relatively small amount of bottom-up growth seen in the initial 5-10 sec of a filling process.

After a period of plating time, two effects might begin to contribute filling. First, the accumulation of accelerating mercapto species (or their more accelerating derivatives) within the features takes place on the surfaces. This accumulation results in surface area within feature and sidewall decreases and adsorbed mercapto species (which are neither

Name abbreviation Chemical formula

bis-(3-sodiumsulfopropyl

3-mercapto-1-propanesulfonate MPS ^-SO₃(CH₂)₃SH 3-mercapto-1-propane sulfonic

acid, or mercaptopropyl sulfonic acid

MPSA HSO₃(CH₂)₃SH

incorporated in the deposit nor desorbed into solution) are thereby increased. Current increases in the areas of geometric concentration (bottom’s corners) as chloride and suppressing polymer is displaced.

Too much accelerator in plating bath disrupts filling because the accumulation of accelerating species also take place outside the feature, and differentiation of the deposition rate from the via base is lost. The accumulation of catalytic species on the growing surfaces within features is strongly supported by the continued Cu growth above features in the absence of leveler. Discontinuing current flow to allow polymer re-equilibration does not interrupt this behavior. However, this disrupted by reversal of interfacial potential to a value causing oxidation or desorption of the adsorbed catalytic material, or by addition of a leveling additive that suppresses current at protruding geometries.^{10, 17, 24}

Fig.1-5. Scheme of the mechanism of the overfilling.¹⁷ suppressor.

accelerator

1.2.5 Aging Influence of Electrodeposition Bath for Gaps Filling (Degradation of Electrodeposition Bath)

K.H. Dietz proposed that PEG is quite stable in acidic solution ¹². However, it is interesting to learn that these organic additive or organic-copper complexes degrading in acid Cu plating baths. Once the kind of polyglycol chain is cleaved, thus can yielding shorter chain polyglycol fractions. A solvent extract of a fresh plating bath gives a high performance liquid chromatography (HPLC) scan with a narrow peak indicating a narrow molecular weight distribution of the poly (ethyleneglycol), as shown in Fig.1-6. After a couple of time, the original peak gets smaller, and wide distributions of lower molecular weight PEG species appears in the HPLC of the extract.²⁵

However, Koh et al. found that a large MW PEG polymer cleaved randomly into smaller molecules is activated by catalyzed oxidation or hydrolysis²⁵.

On the other hand, the following is a summary of the proposed mechanism about the effect of SPS in ECD bath. Accelerator SPS will crack into two MPS when driven by the overpotential of plating, and MPS is more active depolarization than SPS.^{19, 26} Moreover, the more and more existed MPS will make the electrolyte more ineffective for filling capability during aging process. As illustrated in Fig.1-7, after aging of plating cycle, it can be assumed that the entire reaction between the MPS/SPS and Cu²⁺/Cu⁺ are circulating and autocatalytic reaction systems, where the reaction products involve the initial reaction again and the reaction rate is slow initially.^{19, 26-27} In other words, two molecular MPS tend to oxidize to one SPS after more aging periods of plating, thus the relationship are listed below ^{18-19, 26}:

SPS Æ (oxidation) Æ 2MPS [6]

2MPS Æ (reduction) Æ SPS [7]

Furthermore, detail reactions with Cu ions involved is shown as followed 4Cu⁺ + SPS Æ 2Cu²⁺ + 2Cu (I)MPS [8]

4MPS + 2Cu²⁺ Æ SPS + 2Cu (I)MPS [9]

Therefore, the key depolarization effect of SPS is the oxidation of MPS, and then accelerate the cupric ions reduce to cuprous ions. The interfacial-adsorbed cuprous-thiolate complex allowed to move by surface diffusion, and also to break apart

into an adsorbed cuprous ion and an adsorbed disulfide product, and thus acted to accelerate the rate-limiting step⁹. However, T.O. Drews et al. provided another destructive oxidation reaction of SPS.

SPS Æ (further oxidation) Æ S-product [10]

The electrolyte will lose all the accelerating ability after SPS become S product which is some forms of derivatives of aged SPS. Koh confirmed the dissolved-oxygen damage to SPS and observed a similar phenomenon to the above results ²⁵, as shown in Fig. 1-8. The effect of accelerator will be depleted at atmosphere. We suggest that the most amounts of SPS in electrolyte will be decomposed to S-product and loss depolarization ability after many plating cycles faster at atmosphere.

Furthermore, K.H. Dietz also described the possible chemical formulas for that two reactions, as shown in Fig. 1-9. Since the formed compounds contains no divalent sulfur, the oxidation will result in a loss of depolarization.¹²

Fig.1-6. The HPLC scan of electrolytes before and after aged respectively. The carrier component reduced after aging process.²⁵

Fig.1-7. Schematic of assumed mechanisms for different filling aspects between MPS and SPS/aged MPS. (stoichiometry is ignored in this illustration). ²⁷

Fig.1-8. Comparison of the effect of ambient on the accelerator decomposition rate.²⁵

Fig.1-9. Possible Chemical formula of S-products ¹⁷

O

₃

S (CH

₂

)

₃

S S O O

O O

(CH

₂

)

₃

SO

₃

S-product

O

₃

S (CH

₂

)

₃

SO

₃

S-product

Chapter 2: Aging Influence of Organic additives (PEG and PEG-SPS containing) of Cu Electrolytes on Gaps Filling

2.1 Introduction

Several investigators have used an electrochemical method to study the inhibition effect of PEGs on electroplating^1-4. Kelly and West indicated that PEGs react with metal ions on cathodic surfaces, forming complex agents of PEG-Cl-Cu composites^1-2. Material characteristics determined by atomic force microscopy (AFM), surface-enhanced Raman spectroscopy (SERS) and secondary ion mass spectroscopy (SIMS)^5-7 also verify the presence of a monolayer of PEG-Cu-Cl film. This composite adsorbed on reacting surfaces constructs a diffusion barrier against the accumulation of cupric ions at the inner Helmholtz plane, thus reducing the number of Cu ions.

Moreover, Stoychev and Tsvetanov proposed that the inhibition effect of PEG originates not only from the formation of PEG-Cl-Cu (requiring the presence of chloride ions) on Cu surfaces, but also from the complexation of Cu ions and PEG in electrolytes⁸ with Cl ions not participating. The latter mechanism impedes the overall ionic transport in bulk electrolyte. In general, both reactions reduce Cu deposition rate.

In addition to the Cu ECD rate and morphology of plated surfaces, PEG-related composites also influence the effectiveness of both inhibitors and accelerators in superfilling. In particular, for PEG-containing electrolytes free of inhibitors and accelerators, gaps filling ability is only governed by the combination of PEG additives of different molecular weights (MWs) ⁹. In general, all or some PEG additives introduced to electrolytes should have high MWs to provide sufficient inhibition of electroplating¹⁰. The smaller-molecular-weight PEG with higher diffusion ability can enhance cupric ions migrating into deep features and be used to achieve bottom-up filling. The larger-molecular-weight PEG can provide enough inhibition effect of surface current to obtain denser and small-grained Cu deposition thus enhance the filling capability.

The effects of PEG and SPS on the deposition of copper from dilute acid sulphate solutions have been studied for several years^11-13. When electrolyte involves with accelerator (SPS), v-t curves change dramatically, i.e., a depolarization effect also occurs.

Therefore, by applying ECD into integrated circuits process, SPS plays can enhance diffusivity of ions in gap-filling ability.

However, few studies have addressed the impact of PEG and PEG-SPS aging on gap-filling. In this study, the gap-filling ability of PEG-containing electrolytes was investigated in relation to various aging stages with a fixed Cl^- concentration for various feature sizes. Real-time cell voltage transient and electrochemical analyses were performed to explore the degradation mechanism.

2.2 Experimental

(A) Sample preparation: Wafers were prepared by sputtering a 50-nm-thick Ta diffusion barrier and a 50-nm-thick Cu seed layer on SiO₂/Si substrates or patterned substrates with regular trenches (0.5 μm in depth and 0.35-1 μm in width).

(B) Apparatus of Cu ECD: The experiments on Cu ECD were carried out in a tank of non-conducting material. The counter electrode was a Cu plate with a size of 6 × 6 cm² and the working electrode was a wafer with a size of 1 × 3 cm². The distance from counter electrode to working electrode was about 10 cm. Contact to the electrode was implemented outside of the electrolyte with an alligator clip. Agitation air was introduced into the solution from a compressor. In Cu ECP processes, the standard electrolyte was composed of CuSO4⋅5H2O (purity > 99%,): 30 g/l, H2SO4 (97%): 50 g/l, chloride ions: 66 ppm, and deionized water (~18 MΩ). It was additive-free. All organic additives (PEG and SPS) used in this work were purchased from Fluka. The films were deposited under galvanostatic control at room temperature. The direct current power supply utilized in this study was Keithley model 2400, while cell v-t curves are in-situ recorded by PC via GBIP protocol.

(C) Electrochemical analyses: All D.C. and A.C. electrochemical polarization studies were made in three-electrode cells using a computer-controlled EG&G model 273A potentiostat. Potentiodynamic (PD) polarization curves were employed to analyze electrochemical behavior of Cu ECP. The counter electrode was platinum (Pt) and the working electrode was Cu with a constant surface area of 0.5 cm². Before each

which was used as the reference electrode.

2.3 Aging Influence of PEG Suppressors of Cu Electrolytes on Gaps Filling

Cross-sectional SEM images in Fig. 2-1 show 0.35 μm patterns filled with ECD Cu using single-PEG (PEG4000) electrolytes undergoing numbers of electroplating cycles (1 and 5 cycles), and using bias currents of 6 mA (2x10^-3 A/cm²), 15 mA (5x10^-3 A/cm²) and 30 mA (1x10^-2 A/cm²). With a low plating current of 2x10^-3 mA/cm², the filling quality for the damascene features and roughness of plated Cu surfaces are almost independent of the number of electroplating cycles. With a higher plating current of 5x10^-3 A/cm², worse gaps filling and rougher surfaces were obtained with aged electrolytes than those obtained with fresh electrolytes. This deterioration phenomenon with electroplating cycles becomes more significant with increasing plating current, as confirmed by SEM images for the plating current of 1×10^-2A/cm². Examining STD electrolytes using the same tests applied to PEG-containing electrolytes revealed that plating cycles have a negligible influence on the gaps filling capability and field, and roughness of plated surfaces (results not shown here). Hence, PEGs in electrolytes are expected to be the predominant constituents degraded during plating cycles. Notably, fresh STD electrolytes have the worst performance in gaps filling and surface smoothening among the three electrolytes discussed.

Fig.2-1. Side-view and cross-sectional SEM images for 0.35 μm trenches electroplated with PEG (PEG4000) electrolytes undergoing various aging stages with various bias currents.

To investigate gap-filling yield with different plating cycles under various conditions, hundreds of trenches, with sizes ranging from 0.35 to 1 μm, plated with PEG-containing electrolytes were analyzed. Figure 2-2 (a) shows plots of gap-filling yield, Y_G, as functions of plating cycles for the three current densities. Gaps filling yield decreases as the number of plating cycles increase, and was higher at higher current densities. In addition, the gaps filling yield increases with trench width for the three current densities, and a higher current density gives a higher yield. To examine the aging–induced gaps filling degradation, the decay rate of gaps filling abilityΔY_G Δn, which is defined as the change in gaps filling yield during specific cycles and obtained by analyzing the data in Fig. 2-2, is plotted in Fig. 2-3.

Fig.2-2. Gaps filling yields in trenches electroplated with PEG400-containing electrolytes operated at various bias currents as functions of (a) number of plating cycles, and (b) trench width.

Therefore, a higher ΔY_G Δn means a higher level of more serious gaps filling decay. Figure 2-3 reveals one interesting trend: gaps filling decay for narrower trenches is at a higher level than that for wider trenches. Moreover, excellent gap-filling in wide trenches is possibly obtained with electrolytes without additives, and hence gap-filling is not affected by the alteration (or decay) of additives in electrolytes. Nevertheless, Figure 2-3 reveals clearly that the decay rate of gap-filling increases with the bias current employed regardless of the electrolyte used or trench width considered.

Fig.2-3. Decay rates of gaps filling yield for PEG-containing electrolytes operated at various bias currents, as function of trench width.

Next, we will correlate aging-induced electrochemical reactions with gaps filling degradations. To date, two possible degradation mechanisms for PEGs during ECD have been proposed by Stoychev and Tsvetanov, and Dietz^{8, 14}: (1) the cracking of long-chain PEGs, and (2) increasing numbers of small clusters of Cu ions complexing with –(CH₂)- bonds of PEGs formed outside the double-layer region. Because electroplating systems can be modeled as an equivalent circuit, Nyquist plots resolve impedances that originate from various reactions in ECD ^2,11,15. For verifying our assumption, the electrochemical behavior of low-MW/short-chain PEG on a Cu surface should be analyzed. Nyquist plots in Fig. 2-4(a) depict a higher diffusion layer resistance of 4 Ω (calculated from the second semicircle of the curves) but a similar charge-transfer resistance of 2 Ω (first semicircle) for the fresh PEG200-containing

在文檔中銅電鍍與電解拋光於銅鑲嵌金屬連導線應用之研究 (頁 21-0)