A previously described microfluidic electrochemical cell [ M. J. Willey and A. C. West , Electrochem. Solid-State Lett. , 9 , E17 (2006) ] has been used to characterize the rates of adsorption and desorption of polyethylene glycol (PEG) onto copper (Cu) during electrodeposition. Galvanostatic and potentiostatic response to the introduction or removal of PEG from the bulk solution is measured. Adsorption time constants are ∼0.5s or less for bulk PEG concentrations greater than 50ppm . For lower concentrations, adsorption is mass-transfer controlled for the conditions prevailing within the microfluidic cell, where the average diffusion layer thickness is ∼2μm . PEG desorption rates are much slower, typically between 10 and 100s . For most conditions, the electrochemical signal during desorption consists of two regions, where the first stage shows a slow linear variation and the second stage a fast exponential-type transition. In the first stage, it is hypothesized that the PEG layer blankets the electrode surface. Once the polymer-layer thins sufficiently, the layer becomes patchy, and the second desorption stage commences.

The adsorption and activation of bis ͑3-sulfopropyl͒-disulfide ͑SPS͒ has been studied using a microfluidic electrochemical device. The device provides for accurate transitioning of solution over the working electrode allowing for addition or removal of additives from a base plating electrolyte. The transition from a Cl − plating bath to a polyethylene glycol ͑PEG͒/SPS/Cl − bath shows a quick ͑Ͻ2 s͒ suppression of copper deposition similar to that seen with a PEG/Cl − bath. Acceleration at longer times, presumably by the activation of adsorbed SPS species, is a function of SPS concentration and applied cathodic potential. When copper is plated in the presence of SPS and Cl − for times around 100 s before the introduction of PEG, the activated SPS appears to inhibit the adsorption of PEG to the surface. System response to chloride ion addition or removal is relatively fast. Desorption of an activated SPS/PEG/Cl − layer occurs slowly, even at high cathodic potentials of −0.275 V, where current densities are over 50 mA cm −2 . Numerical simulations show that the time needed for PEG to saturate the bottom of a sub-100 nm feature is small relative to SPS activation times.

Three copper-plating suppressors are examined in three-additive baths: a polyethylene glycol ͑PEG͒, a polypropylene glycol ͑PPG͒, and a triblock copolymer of the two. Bis͑3-sulfopropyl͒-disulfide ͑SPS͒ is found to transition each to a state of nonsuppression, i.e., accelerate each, at a rate dependent on the suppressor molecule, the SPS concentration, and, to a lesser extent, the suppressor concentration. Using a planar microscale working electrode ͑d = 100 m͒, the kinetic currents of the plating reactions are observed without the influence of ohmic resistance, revealing far higher current densities than previously reported. Potentiostatic and galvanostatic experiments of additive adsorption at short times, t Ͻ 20 s, are compared quantitatively using a surfaceblocking model to transform the data to effective surface coverage, EFF , vs time. A major difference is found in SPS acceleration between galvanostatic and potentiostatic experiments, with the rate of change in suppression being proportional to the current density. This results in a constant rate of change in EFF under constant current but a self-reinforcing rate of change in EFF at constant potential. A simple additive model is introduced to characterize the results.The chosen material for interconnect manufacture in the semiconductor industry is electrodeposited copper, switched from aluminum in the late 1990s to achieve architecture scales below 250 nm. 1,2 The drive for increasing computing power continues, following the relation known as Moore's Law, and the characteristic scale of interconnects has fallen below 100 nm, approaching 30 nm and lower in the coming years. Electrodeposition of copper in highaspect-ratio features on this scale requires plating bath additives, which suppress and enhance deposition rates at the feature mouth and floor, respectively, to achieve bottom-up, void-free copper filling. This is referred to as superconformal filling or superfill.The additives required for superfill are a halide-ion promoter such as Cl − , a polyether suppressor such as polyethylene glycol ͑PEG͒, and a sulfur-containing molecule such as bis͑3-sulfopropyl͒-disulfide ͑SPS͒. Superfilling behavior results from an interaction of these additives, and this study focuses on plating behavior as a function of bath composition in order to formulate for decreasing feature sizes. It has been shown that the polyether suppressor and chloride ions coadsorb on the copper electrode surface, raising the overpotential required for a constant plating current. 3-7 This suppression becomes minor if the halide is not present and is a function of both halide and polyether concentration. 5 Other polyether molecules besides PEG ͓polypropylene glycol ͑PPG͒ and pluronic triblock copolymers of ethylene oxide ͑EO͒ and propylene oxide ͑PO͔͒ display the same characteristic suppression, although its degree is a function of the molecular structure. 8,9 A proportional increase of PO repeat units in the chain causes an increase in suppression. However, the presence of long, hydrophobic PO...

Patterned 100 nm trenches with an aspect ratio of 3.5 are filled at a nominal current density of −6.6mA/cm2 . A low acid copper plating bath is used with the accelerator bis(3-sulfopropyl)-disulfide (SPS) and one of three suppressor molecules: Poly(ethylene glycol) (PEG) 3350, poly(propylene glycol) (PPG) 725, and an ethylene–propylene–ethylene (EPE) oxide triblock copolymer at a molecular weight of 2000 g/mol. All suppressors result in superconformal filling, although the filling rates vary widely. EPE 2000 results in the most rapid filling, metallizing the features without voids in 5 s under some conditions. The superior performance of EPE 2000 is attributed to its high suppression strength, which is greater than either PEG 3350 or PPG 725 at the relevant plating potentials. EPE 2000 filling becomes more rapid as suppressor concentration decreases, down to 100 ppm, which is attributed to a strong correlation between EPE 2000 concentration and adsorption time. EPE 2000 performance is also improved as SPS concentration decreases, a result in contrast to literature observations on larger (500 nm) features. A simple expression is developed to demonstrate that the time scales of suppression, acceleration, and filling can account for this result.

A microfluidic device is used to measure electrode response to rapid changes in electrolyte composition. These changes in composition are achieved by switching the electrolyte flowing through an electrolyte channel containing a working and counter electrode. In the present investigation, the time response of the device was characterized by reducing tri-iodide ions at a rate controlled by mass-transfer to the electrode surface. Results are compared with finite-element simulations that assume an ideal two-dimensional parabolic flow. Simulated, theoretical, and experimental steady-state limiting current densities are all in excellent agreement. Experimental time constants are also in accord with simulations.