Raman and QCM Studies of PPG and PEG Adsorption on Cu Electrode Surfaces

We investigate the behavior of Cu plating bath suppressor additives poly(ethylene glycol) (PEG) and poly(propylene glycol) (PPG) using normal Raman spectroscopy, surface-enhanced Raman spectroscopy (SERS), and electrochemical quartz crystal microbalance (QCM) measurements. Raman and SERS show a clear spectroscopic trend of increased intensity in higher wavenumber modes in the CH stretching region as the environment is changed from pure material to solution to surface for both PEG and PPG. The spectral changes associated with PEG are larger than those associated with PPG, suggesting that the relatively more hydrophilic PEG undergoes more conformational changes upon surface association relative to the more hydrophobic PPG. Calculations show that the observed spectroscopic trend is associated with increased gauche character in the polymer backbone. QCM measurements show PEG adsorbs to the surface only in the presence of Cl − , while PPG adsorbs to the surface both with and without Cl − present. In the presence of Cl − , PPG forms a denser surface layer (0.598 μ g/cm 2 ) compared to PEG (0.336 μ g/cm 2 ) on a Cu underpotential deposition (UPD) layer on Au. These differences are consistent with the increased hydrophobicity of PPG relative to PEG. process Polarizability The discrete FWHM − line broadening.QCMmeasurements were performed using a RQCM instrument, corresponding and teﬂon crystal holder, as de- 24 Au-coated sensing crystals were purchased from INFICON (2.54 cm chromium adhesion layer under Measurements were conducted at constant temperature (30 ◦ C)usingawatercirculatorfeedingajacketedelectrochemicalcell. Argon was bubbled into the electrolyte to remove dissolved oxygen gas and then passed over the surface of the electrolyte during measurements to maintain an inert atmosphere. The Au-coated crystals were used as the working electrode, a Cu wire served as the counter, and a Ag/AgCl electrode was used as the reference. Cyclic voltam- metry was conducted from 0.39 V to 0.02 V at mV/s, beginning measurements si- multaneous with the cyclic voltammetry. change in of

Copper (Cu) electrodeposition is an efficient process for creating various metal connections within modern microelectronic devices. 1,2,3 These connections include micron-sized connections in printed circuit boards, 4,5 nanometer Damascene interconnects in microchips, 6,7 through silicon vias 8,9 and advanced packaging appilcations such as redistrubtion layers and pillars. 10 Continual miniaturization of these features while improving performance requires precise fabrication control achieved by the inclusion of additives in the standard sulfuric acid and copper sulfate electrolyte. 3 Among these types of additives are suppressors, anti-suppressors or accelerators, and levelers. In addition to these additives, chloride (Cl − ) is typically included because it plays a significant role in aiding the adsorption of the different additives, as well as stablizing the Cu(I) intermediate at the surface. 11 Common suppressors are large alkyl-ether polymers, such as poly(ethylene glycol) (PEG) or poly(propylene glycol) (PPG) that act on the wafer surface to prevent over-deposition, which is important for keeping trenches from closing too early in the deposition process, as well as limiting the amount of excess material that must be removed in the proceeding chemical-mechanical polishing step of the dual damascene process. 11 While much work has been conducted on all types of additive, the suppressor additive behavior is particularly studied. Techniques including electrochemical methods, [12][13][14] scan-probe microscopy 15,16 and spectroscopy 17,18 have been employed to understand the suppressor additive behavior at the Cu surface during electrodeposition, both by itself and in the presence of other additives. It is commonly thought that the PEG suppressor interacts with the copper surface through complexation of its oxygen atoms with Cu(I) stablized by adsorbed Cl − , creating a PEG-Cu(I)-Cl − bridge. 17 PEG in the absence of Cl − demonstrates very little suppression indicating that the surface bridge complex is essential. PPG is another suppressor commonly used in Cu plating baths. In contrast to PEG, PPG has been previously shown to 1) cause a more negative suppression potential; 2) reach surface saturation slower; and 3) be de-sorbed by SPS antisuppressor quicker. 12 The origins of the difference in electrochemcial behavior between PEG and PPG is unknown, as are putative changes in the molecular conformation of either molecule going from the neat material, to solution, to the surface.
The physical properties of both PPG and PEG have been intensively studied. 19 PEG in its pure form is a semi-crystalline solid material at ambient temperatures. The polymer backbone is oriented in a very specific dihedral sequence in this state. There are three main dihedrals in the repeat unit of PEG: C-O-C-C, O-C-C-O, and C-C-O-C. In the solid state, these dihedrals are in a trans-gauche-trans sequence to form a helical structure, with a complete helix cycle measuring 19.3 angstroms. 19,20 PPG on the other hand, is a liquid at ambient temperature and posesses a much less defined structure in this pure form. Vibrational spectra of both molecules has also been the focus of substantial work. The C-H stretching region (2800-3000 cm −1 ) and the C-H bending region (1400-1500 cm −1 ) exhibit characteristic patterns, the analysis of which can inform understanding of molculear conformation. 21 This work examines PEG and PPG using Raman spectroscopy and surface enhanced Raman spectroscopy (SERS) in order to evaluate conformational changes in the polymers at the surface during Cu electrodeposition. Along with complementary electrochemical quartz crystal microbalance (QCM) gravitometric measurements, we provide insight into the orgin of the different behaviors exhibited by these polymers during Cu deposition.
A polycrystalline Cu disk (approx. 9.7 mm in diameter) was used as the working electrode, which was manually polished using 9, 3, 1 and 0.25 μm diamond suspensions (MEtaDi Supreme, Beuhler) and thoroughly rinsed and sonicated after each polishing step. The electrode surface was roughened in 0.1 M KCl by multiple oxidationreduction cycles from −1.0 V to +0.4 V vs. a copper counter/reference electrode at 100 mV/s followed by a 10 sec hold at −1.0 V. The electrode was removed under potential control during the last cathodic hold and immediately rinsed with Milli-Q water. The counter electrode for SERS measurements was a Cu wire, and the reference electrode was a "no leak" Ag/AgCl electrode (3 M KCl, eDAQ). All potentials in this work are reported relative to Ag/AgCl. Spectra were obtained at a constant potential of −0.2 V for 30 sec following an anodic pulse at 70 mA/cm 2 for 4 milliseconds.
Raman spectroscopy was conducted with a 632.8 nm He-Ne laser (Meredith Instruments) operating at 58 mW, using a spectrometer and setup described previously. 23 Raman spectra were obtained at room temperature (292 K), except for the PEG melting experiment (292-335 K). Temperature control for the PEG melting experiment was achieved with a Brisk Heat heating tape (120 V, 144 W) and thermocouple monitor. Typical spectral resolution is 3-5 cm −1 .
The Raman spectra of PEG and PPG molecules in different conformations were calculated using Turbomole software version 7.0. All electronic structure calculations were performed at the DFT level employing the B3LYP functional in combination with a def2-TVP basis set. In order to obtain the Raman scattering activity, the geometries were optimized and subsequently a normal mode analysis was carried out. The Raman scattering activity, I Ram , is defined as where α is the polarizability, Q is the normal coordinate, β is the anisotropy of the polarizability, α is the mean polarizability derivative, and β is the anisotropy of the polarizability tensor derivative. Polarizability derivatives have been calculated using a wavelength of 633 nm. The discrete Raman intensities were convoluted by a Gaussian function with FWHM of 20 cm −1 to mimic the experimental line broadening. QCM measurements were performed using a Maxtek RQCM instrument, corresponding software and teflon crystal holder, as described previously. 24 Au-coated sensing crystals were purchased from INFICON (2.54 cm dia., 5 MHz, AT-cut, chromium adhesion layer under gold). Measurements were conducted at constant temperature (30 • C) using a water circulator feeding a jacketed electrochemical cell. Argon was bubbled into the electrolyte to remove dissolved oxygen gas and then passed over the surface of the electrolyte during measurements to maintain an inert atmosphere. The Au-coated crystals were used as the working electrode, a Cu wire served as the counter, and a Ag/AgCl electrode was used as the reference. Cyclic voltammetry was conducted from 0.39 V to 0.02 V at 5 mV/s, beginning with the cathodic scan. Gravimetric measurements were collected simultaneous with the cyclic voltammetry. The change in frequency of the quartz sensing crystal was used to calculate mass according to the Sauerbray equation: 25 where f is the measured resonant frequency change (Hz), f is the intrinsic crystal frequency, m is the mass change, ρ q is the density of quartz (2.65 g/cm 3 ), μ is the shear modulus (2.95 × 10 11 dyn/cm 2 ) and A is the electrode area. C f was determined to be 59.1 cm 2 /μg by measuring the frequency change during Ag deposition from a solution of 0.1 M KNO 3 and 1 mM AgNO 3 . 24 This value is close to the supplied instrument parameter of 56.6 cm 2 /μg. The largest change in crystal resistance during experiments was 4 , corresponding to a 1.4% change, a value of which is within the limits of the assumptions of the Sauerbray equation. 26 Figure S1 shows the crystal resistance as a function of potential for the cyclic voltammetry applied during QCM measurements. The base electrolyte was comprised of 0.1 M H 2 SO 4 and 0.01 M CuSO 4 , with 50 ppm Cl − and 100 ppm suppressor (PEG or PPG) added when indicated.

Results and Discussion
PEG raman and SERS: CH stretching and bending regions.- Figure 1 reports Raman spectra of PEG obtained in the C-H stretching region (1a) and C-H bending region (1b) in three different environments: solid, in solution and on the surface. Both the normal Raman spectrum of the neat PEG and the normal Raman of the 50% solution of PEG in water are consistent with those reported previously. 18,[27][28][29][30][31] Four peaks are observed in this spectral region. Generally, peaks in this region are assigned to symmetric and anti-symmetric CH 2 stretches. Indeed peak 2 (2885 cm −1 ) and 4 (2940 cm −1 ) are assigned to ν s (CH 2 ) and ν a (CH 2 ), respectively. 32-34 Peak 1 (2840 cm −1 ) and the intensity at lower wavenumbers is assigned to combination modes from the bending region. [33][34][35] Previously studied PEGs may not always contain methyl (CH 3 ) end groups, but they are part of the material used in the present work. However, PEG with and without methyl end groups gives the same Raman spectrum in solid and solution (data not shown), hence no specific peaks are assigned to methyl end groups.
Peak 3 only appears with significant intensity in the solution spectrum. The origin of this band is unclear in the literature. A series of solutions containing from 10%-40% PEG was investigated with Raman spectroscopy and did not show any change in the relative intensity of peak 3. Additionally, the spectrum obtained in 50 g/L H 2 SO 4 + 1 g/L CuSO 4 + 50 ppm Cl − was identical with that obtained in H 2 O alone. PEG solutions in other non-polar solvents were also interrogated ( Figure S2). PEG in chloroform and PEG in acetonitrile both gave Raman spectra more similar to that of the solid than that of the PEG in water solution, with a significant lack of intensity in the region that contains peak 3. 36 A similar trend is observed in ATR-IR spectra comparing spectra obtained from an aqueous solution of PEG to those obtained in either deuterated methanol or acetonitrile ( Figure  S3). This suggests that the PEG in water spectrum is significantly influenced by hydrogen bonding interactions between the water protons and the ether oxygen units in the PEG backbones. Hydrogen bonds are known to form between water and PEG. [37][38][39] Consequently, peak 3 is assigned to a PEG CH 2 stretch mode affected by solvent interactions. It may also be due to a Fermi resonance. The mode cannot be assigned to a CH 3 -based mode from an end group, as both methyl terminated and hydroxyl terminated PEGs give identical spectra. Overtone modes are additionally limited to the lower end of the CH stretch region (< 2850 cm −1 ). Calculations (vide infra) show that this region does not possess any vibrational intensity. It is unclear whether the intensity in peak 3 is the result of the ν s (CH 2 ) shifting up or the ν a (CH 2 ) shifting down upon interaction with water. Figure 1a also reports the SERS obtained from a 100 ppm solution of PEG + 1 g/L CuSO 4 + 50 ppm Cl − in 50 g/L H 2 SO 4 aqueous solution at a Cu electrode at a potential of −0.2 V. Spectra obtained at potentials between −0.10 and −0.35 V were essentially identical, with similar intensity throughout the potential sweep. The SER spectrum reported is qualitatively similar to that found from both neat PEG and the 50% PEG solution and is also qualitatively consistent with prior reports, both in our group 17 and others who have used SERS to investigate PEG adsorption in Cu deposition electrolytes and processes. 18,[27][28][29] In the C-H stretch region, the spectra generally exhibit a trend where intensity moves to higher energy going from neat PEG, to the PEG solution, to SERS obtained from PEG on a Cu electrode. This shift is essentially a manifestation of more intensity in peak 4, associated with the ν a (CH 2 ), and less intensity in peak 2, associated with the ν s (CH 2 ). Figure 1b reports the Raman and SERS obtained in the CH bending region. The spectrum shows the presence of four bands, the assignments of which have been previously reported. 19 The four vibrational modes (peaks 5-8) in the solid spectrum are assigned to different symmetry forms of the CH 2 bending modes present. In solution, two of these modes are no longer present which can be attributed to a lack of crystal splitting in the aqueous environment. Additionally, the lower wavenumber mode, peak 5, has slightly increased intensity in solution. Peak 5 increases even more in the spectrum of PEG at the Cu SERS electrode also shown in Figure 1b, and now possesses greater intensity than peak 6 at higher wavenumber.  PEG Raman: melting process.-In order to investigate changes in the PEG Raman spectrum without complications from the solvent, the Raman spectra of neat PEG both above and below the PEG melting temperature were obtained. Figure 2a shows the Raman spectra for the CH stretching region of PEG at different temperatures, from 292 K to 335 K. Although the melting temperature of PEG, Mw = 2,000, is 315-329 K, there are significant changes observed in the spectra between 305 K and 310 K. It is common for polymers and other large molecules such as phospholipids to undergo a pre-melting event before the melting temperature threshold. 40 The melting process of PEG has been studied by Raman spectroscopy before and agrees with our spectra. 32 The CH stretching region of PEG during the melting process is highlighted by a few distinct spectral changes. First, the ν s (CH 2 ) shifts down from 2889 cm −1 to 2875 cm −1 , while maintaining the greatest relative intensity. Additionally, ν a (CH 2 ) does not shift in wavenumber, but drastically increases in intensity. Interestingly, the region between the two fundamental modes does not change intensity or shift wavenumber. This is distinctly different from the solution spectrum, thus further supporting the suggestion that peak 3 is caused by solvent interactions.
The extra thermal energy in the melted PEG allows C-O bonds to rotate about their axes from the lower energy trans state to the higher energy gauche state, thus causing more gauche conformations (TGG or GGG). The C-C dihedral is not expected to change as much as the C-O dihedral during heating/melting. 32 The increase in intensity of the ν a (CH 2 ) in the melting experiment is reminiscent of the changes occurring to PEG in both solution and in SERS obtained from the surface. This similarity suggests the increase in the ν a (CH 2 ) is also due to increased gauche characteristic of the polymer chain. The CH bending region, shown in Figure 2b, in the melting experiment follows a similar pattern as in the environment-dependent experiment, wherein peak 5 has increased intensity at higher temperature. This similarity also supports the idea that there is more gauche characteristic in both experiments at higher temperature and in the surface spectrum. This is supported by calculations, which will be discussed below.

PPG Raman and SERS: CH stretching and bending regions.-
A similar environment-dependent study was conducted for PPG, as it is used similarly to PEG for Cu electrodeposition suppression, but is significantly different than PEG in terms of hydrophobicity. 41 Figure 3 shows the (3a) CH stretching and (3b) CH bending region of the Raman spectrum for PPG as a pure material, in an aqueous suspension and at a Cu SERS electrode surface. PPG as a pure material is a liquid at room temperature, opposed to PEG which is a solid. Qualitatively, there are three main peaks (9)(10)(11) in the CH stretching region, occurring at ca. 2865 cm −1 , 2930 cm −1 , and 2970 cm −1 . These modes are assigned to ν s (CH 2 ), ν a (CH 2 ), and ν(CH 3 ), respectively. The Raman spectrum of neat PPG agrees with a previously reported spectrum. 42 PPG is sparingly soluble in water, as it is hydrophobic. 41 A 25% suspension of PPG in water yields a cloudy mixture. The Raman obtained from this suspension (Figure 3a) is little changed relative to the Raman spectrum of the pure material. However, there is some increased intensity in the highest wavenumber mode at 2970 cm −1 , peak 11. The PPG SER spectrum exhibits even more increased relative intensity in peak 11 and decreased intensity in peak 9, the lowest wavenumber mode at 2870 cm −1 . The CH bending region for PPG (Figure 3b) exhibits little change between the three environments considered here, and only one dominant mode is present, peak 12.
Relative to PEG, PPG exhibits less spectral change between the three environments considered here. PPG (M w = 2,000) is sparingly soluble in water with a solubility of 250 mg/L, leading to less spectral change between the Raman spectrum obtained from the pure material and that obtained from the PPG suspension. 43 In contrast, the Raman spectrum obtained from PEG changes substantially upon dissolution, with new intensity in mode 3, likely associated with H-bonding interactions between PEG molecules and water in solution. SERS obtained from PPG shows no change in the CH bending relative to neat PPG, while substantial new intensity at lower wavenumbers is found in the PEG SERS, mimicking to some extent the spectrum obtained from melted PEG. Finally, in the CH stretch region, SERS collected from PEG on an electrode surface exhibits substantially increased intensity at higher wavenumbers, while this effect is somewhat less pronounced in PPG. Figure 4 shows calculated Raman spectra of PEG and PPG molecules of varying conformation. Calculations of polymers with chain length used in the experiment are computationally not feasible with the level of theory employed in this work. Therefore, in order to model the experimental data, we used molecules containing five monomer units with methyl end groups. This size allowed capture of the qualitative trends of the experimental spectra while still enabling analysis of the 3N-6 normal modes in detail.

Calculated raman spectra.-
In this model, the ratio of end cap methyl groups to ether units is 2:5 for the molecules used in calculations, while this ratio is 2:45 for the PEG and 2:34 for the PPG used in experiments. Therefore, the presence of end cap CH 3 modes is much more substantial in the calculated spectra compared to the experimental spectra. Calculations were performed examining the effect of incorporating more gauche character into the C-O-C-C dihedrals of the PEG chain. This correlates to adjusting the C-O-C-C dihedral sequence from all-trans (TTTT) to alternating gauche-trans (GTGT) to all-gauche (GGGG). The other dihedral (O-C-C-O) was changed from GGGG in the all-trans model to TTTT in the alternating gauche-trans and all gauche models. The full OCCO-COCC dihedral sequences ranging from least gauche character to most gauche character of the C-O-C-C sequence are therefore GGGG-TTTT, TTTT-GTGT, TTTT-GGGG. For simplicity, the conformations will be referred to by the C-O-C-C dihedral sequence. The same changes were calculated for PPG. Figure 4a shows the CH stretch region of the calculated Raman spectra for PEG for the three different dihedral conformations considered. Clear changes are seen in the CH stretching region of the calculated Raman spectrum of PEG with changing conformation. First, a coupled ν s (CH 2 ) and ν s (CH 3 ) mode at 2850 cm −1 decreases in intensity and blue shifts moving from TTTT to GGGG to 2865 cm −1 . Additionally, a coupled ν s (CH 2 ) and ν a (CH 2 ) mode at 2900 cm −1seen as a shoulder in the TTTT (helix) model -increases in intensity and eventually becomes an independently resolved peak in the GGGG model. Additionally, a ν a (CH 2 ) peak grows in and shifts from 2960 cm −1 to 2965 cm −1 as more gauche character is present in the GTGT and GGGG models.
The most significant difference between the calculated spectra and experimental measurements in the CH stretching region is between 2900 cm −1 and 2940 cm −1 . The experimental spectra, most specifically the 50% PEG aqueous solution, exhibit substantial intensity at 2917 cm −1 , whereas no intensity is calculated in this region. Experimentally, this intensity is associated with the presence of an H-donor solvent, suggesting that H-bonds to PEG are responsible for the experimentally observed intensity. Other effects, such as Fermi resonances or overtones also cannot be strictly ruled out, and are beyond the scope of the calculations.
The calculated Raman spectra for PEG show an overall trend in the CH stretching region of increased intensity in higher wavenumber modes with increasing gauche character, complemented by blue shifts of CH 2 stretch modes. This behavior is in agreement with the experimental data and suggests the presence of more gauche character in the PEG, particularly as it associates with the electrode surface. Figure 4b displays the calculated Raman spectra for the different conformations of PEG in the CH bending region. The lower energy mode shifts from 1420 cm −1 in the TTTT conformation to 1440 cm −1 in the GGGG conformation. The higher wavenumber mode increases in wavenumber from 1460 cm −1 to 1470 cm −1 , from the TTTT conformation to the GGGG conformation, respectively. Despite these shifts, it is clear that the lower wavenumber mode increases in intensity relative to the higher wavenumber mode with increasing gauche character. This is consistent with the experimental CH bending data collected for PEG displayed in Figure 1b. This suggests good agreement between the two methods and gives confidence to the calculations. A deeper investigation into the origin of the frequency shifts would involve a careful analysis of the change in effective masses and the topology of the potential energy surface. This, however, is beyond the scope of this work. Figure 4c displays the calculated Raman spectra for PPG, demonstrating changes with different C-O-C-C dihedral conformations. The ν s (CH 2 ) intensity decreases with increasing gauche character, while red shifting from 2860 cm −1 to 2845 cm −1 . A coupled ν s (CH 2 ) and ν a (CH 2 ) mode at 2870 cm −1 is present, but not resolved until the all-gauche conformation upon which it exhibits significant intensity.
A coupled ν a (CH 2 ) and ν s (CH 3 ) mode gains intensity with increased gauche character, as well as blue shifts from 2915 cm −1 to 2920 cm −1 . A ν a (CH 2 ) and ν a (CH 3 ) coupled mode at 2980 cm −1 slightly decreases in intensity with a small blueshift to 2985 cm −1 . In this same region, a ν a (CH 3 ) mode increases in intensity becoming a significant shoulder at 3000 cm −1 . Figure 4d shows the calculated Raman spectra for the PPG CH bending region, which slightly change with changing conformation.
There is a small decrease in intensity in the lower wavenumber mode from the TTTT conformation to the GGGG conformation, with a small redshift from 1440 cm −1 to 1436 cm −1 , respectively. Additionally, there is an emergence of a higher wavenumber mode at 1470 cm −1 in the GTGT and GGGG conformation calculated spectra. This implies the intensity lost in the lower wavenumber mode is gained in the higher wavenumber mode as more gauche character is included in the molecule. This differs from the experimental data which shows no significant intensity changes or wavenumber shifting across the different environments.
In the CH stretching region, the calculated Raman spectra more closely resemble the experimental data for PPG than for the PEG case. There is no lack of intensity in the calculated PPG spectra compared to the experimental data, unlike the PEG situation. There are three dominant peaks in the CH stretching region in both the calculations and the experimental measurements. As mentioned previously, the lack of intensity in the middle of the PEG spectra is most likely due to solvent interactions between water and PEG. PPG is much less hydrophilic and soluble in water; therefore, there are less expected solvent interactions. The lack of discrepancy between the calculated spectra and measured spectra could be explained by the lack of solvent interactions. Additionally, there is a clear trend in the calculated PPG spectra of intensity moving to higher wavenumber modes, as well as blue shifting behavior with more gauche character. This is also in agreement with the experimental results. It is important to note that this trend is present in both the PEG and PPG results, but more significant in the PEG case. Again, the unique hydrophilic properties of PEG compared to PPG most likely play a substantial role in this behavior.
In conclusion, Raman and SER spectroscopy results demonstrate PEG and PPG exhibit conformation changes as a function of environment. Calculated Raman spectra suggest that surface confined molecules exhibit increasing gauche character. PEG appears to change more drastically from pure material to solution to surface, most likely due to its ability to interact with solvent water molecules, whereas PPG less so. The difference in hydrophilicity/hydrophobicity between PEG and PPG could explain such conformational behavior.   Figure 5a shows bulk deposition of Cu occurs at 0.05 V for the base electrolyte with no additives. When only PEG is added to the base electrolyte, Cu deposits at the same onset potential with a similar current density. This indicates that PEG without Cl − does not suppress Cu deposition, as expected. 14,44 Upon the addition of Cl − , there is a small cathodic shift of bulk Cu deposition to 0.045 V and significantly less current density indicating suppression is achieved, as expected based on the commonly accepted model for PEG adsorption wherein PEG depends on the Cu-Cl bridge complex to interact with the surface. 3,17 The QCM data in Figure 5c shows the corresponding mass density changes at the surface of the Au-coated QCM crystal during the  applied electrochemistry. The electrolyte without additives and that with only PEG added give similar mass behavior. The maximum mass achieved for the electrolyte without additives is 3.50 μg/cm 2 , whereas the electrolyte with only PEG added gives a value of 3.67 μg/cm 2 . This difference is most likely due to slightly different maximum current densities achieved at the lowest cathodic potential. Most notably, the electrolyte with both Cl − and PEG added gives a maximum mass density of only 1.29 μg/cm 2 . The smaller maximum mass density value indicates less copper mass is deposited and suppression is achieved with both Cl − and PEG present. Figures 5b and 5d are highlighted regions of Figures 5a and 5c, respectively, focusing on the underpotential deposition (UPD) region of the potential scans. The electrochemistry in the UPD region for electrolyte without additives shows characteristic UPD peaks of Cu on Au (polycrystalline) are present at 0.30 V and 0.12 V in the cathodic scan. 45 The UPD peaks are retained in the electrolyte with only PEG added, but the first peak is shifted cathodically to 0.25 V and the lower second peak is slightly diminished, possibly due to PEG interference. The electrochemistry for the electrolyte with both Cl − and PEG added show the first UPD peak is anodically shifted to 0.33 V, as expected for solutions containing Cl − . 46,47 Figure 5d shows the enhanced corresponding QCM data for the different electrolytes. Increased surface mass density is seen at ca. 0.30 V for the additive-free electrolyte and 0.25 V for electrolyte with only PEG added, whereas the initial mass density increase for electrolyte with Cl − and PEG occurs at ca. 0.33 V. All instances correspond to the respective UPD features in the voltammetry. 0.1 V of the cathodic scan will be used as a reference point for discussing surface mass density values, because it is just negative of the second UPD peak at 0.12 V where the surface structure is presumed stable and before bulk deposition begins at 0.05 V. 46 At this potential, the electrolyte without additives and that with only PEG added give mass density values of 0.246 μg/cm 2 and 0.205 μg/cm 2 , respectively. Electrolyte with only PEG added consistently yields a lower surface mass density value at 0.1 V than that of electrolyte without additives, again possibly due to PEG interference, but with no additional observed surface mass.
Prior literature on the surface mass density obtained without additives reports a somewhat lower mass density (0.159 μg/cm 2 ) than that obtained here. 48,49 The data in these reports was collected using electrolyte with 5 mM Cu 2+ and 0.05 M SO 4 2− , concentrations lower than those used here. In order to evaluate the effect of the higher concentration of acid and Cu, we performed measurements using the lower Cu 2+ and SO 4 2− concentrations and same 10 mV/s scan rate, and obtained a surface mass density value of 0.148 μg/cm 2 , similar to that previously reported. We suggest that the larger mass density values reported from the higher concentration solution may possibly be a result of additional Cu deposition just prior to bulk deposition onset and a higher SO 4 2− contribution. 50 The surface mass density obtained at 0.1 V for electrolyte with Cl − and PEG is 0.336 μg/cm 2 . Electrolyte with only Cl − added and no PEG present yielded a surface mass density of 0.215 μg/cm 2 at 0.1 V (data not shown). Thus, the additional mass from the Cl − and PEG-containing electrolyte is likely due to PEG adsorption on the surface. Figure 6 shows the cyclic voltammetry and QCM data obtained for electrolyte without additives, that with 100 ppm PPG added, and that with 50 ppm Cl − and 100 ppm PPG added. The bulk Cu deposition potential in electrolyte with only PPG added occurs at 0.05 V (Figure 6a) and exhibits a current density similar to that of additivefree electrolyte, indicating that PPG alone does not inhibit bulk Cu deposition. Results obtained from electrolyte with both Cl − and PPG show a cathodically shifted bulk Cu deposition onset potential of 0.035 V, accompanied by a decrease in current density. The cathodic shift in Cu deposition onset potential due to PPG suppression is larger than that due to PEG suppression (15 mV vs. 5 mV), consistent with prior reports. 12 PPG, in the presence of Cl − , is thought to interact with a Cu surface during deposition in a similar manner as PEG, through a Cu-Cl bridge complex. 51 Figure 6c shows the full scale QCM data obtained from the three electrolyte compositions. The maximum surface mass densities achieved in electrolyte without additives and that with only PPG added are 3.98 μg/cm 2 and 3.91 μg/cm 2 , respectively, whereas the maximum surface mass density for electrolyte with Cl − and PPG added is only 0.89 μg/cm 2 . The smaller surface mass density value indicates successful suppression of bulk Cu deposition when both Cl − and PPG are present, similar to the Cl − and PEG case. Figures 6b and 6d highlight the UPD region of the electrochemistry and QCM data for the PPG related experiments, respectively. In Figure 6b, the cathodic potential scan of electrolyte without additives demonstrates the characteristic Cu on Au UPD peaks at 0.27 V and 0.12 V, similar to the PEG related experiments and literature. 45,46 The results obtained from electrolyte with only PPG added show the first UPD peak cathodically shifted to 0.25 V and the second peak diminished in intensity. Additionally, the UPD in the anodic scan of the electrolyte with only PPG added is cathodically shifted to 0.30 V compared to that obtained from the electrolyte without additives which occurs at 0.33 V. The shifts in the UPD peak potentials in both the cathodic and anodic scans indicate that PPG possibly interacts with the surface and affects the surface structure of monolayer Cu formed in the UPD region. The UPD peak in the cathodic potential scan of electrolyte with both Cl − and PPG occurs at 0.32 V, as expected for solutions containing Cl − . 46,47 Figure 6d shows QCM results obtained from the PPG related electrolytes. The initial increase in surface mass density for the electrolyte without additives occurs at ca. 0.27 V, concurrent with the first UPD peak in the potential scan, resulting in a surface mass density of 0.227 μg/cm 2 at 0.1 V, a similar value for the electrolyte without additive trial for the PEG-related experiment (0.246 μg/cm 2 ). These values are represented as an average in Table I. The surface mass density measured for electrolyte with only PPG added increases at 0.25 V in the potential scan, but to a much larger value, eventually reaching 0.413 μg/cm 2 at 0.1 V of the cathodic scan. The surface mass density at 0.1 V obtained from only PPG containing electrolyte is much larger than that obtained from only PEG containing electrolyte (0.205 μg/cm 2 ) indicating that PPG associates with a Cu surface without the presence of Cl − .
Despite the apparent increase in mass in the presence of PPG without Cl − , there is no effect on either the bulk Cu deposition onset potential or the current density. The anodic scan also features additional surface mass density attributed to PPG interaction. A large decrease in surface mass density associated with bulk Cu stripping is observed at 0.14 V in the anodic scan of Figure 6d. Shortly thereafter in the anodic scan, there is an anomalous increase in surface mass density beginning at 0.20 V before a final decrease at 0.30 V. The increased mass is associated with the more anodic UPD peak and is nearly identical with that associated with UPD on the cathodic scan. We suggest that stripping of the bulk Cu leaves a surface that is free of UPD Cu, possibly due to PPG decoration. As the potential is scanned more positive, anions and UPD Cu displace the PPG, leading to the UPD-related mass density. This surface mass density is removed as the potential is made even more positive. Scan rate-dependent QCM measurements ( Figure S4), which could report on a competition between PPG and Cu(I) related species, did not yield significant changes in the mass values when scan rates between 5 mV/s and 20 mV/s were tested.
The initial surface mass density increase in the potential scan for electrolyte with both Cl − and PPG begins at ca. 0.32 V and reaches 0.598 μg/cm 2 at 0.1 V. The electrolytes containing Cl − and suppressor (PEG or PPG) produce the largest surface mass densities at 0.1 V Figure 7. PEG requires Cl − to facilitate interaction with the copper surface (left), whereas the hydrophobic PPG has the ability to directly interact with the copper surface (right). for their respective experiments; however, the surface mass density achieved for electrolyte containing Cl − and PPG gives almost twice that of electrolyte with Cl − and PEG (0.336 μg/cm 2 ), suggesting much more PPG interacts at the surface than PEG. Table I summarizes the experimental surface mass densities at 0.1 V in the cathodic scan for the six different QCM plots shown in Figures 5 and 6. The table shows that PPG addition to CuSO 4 /H 2 SO 4 /Cl − electrolyte gives a surface mass density of 0.598 μg/cm 2 while PEG addition to CuSO 4 /H 2 SO 4 /Cl − gives a value of 0.336 μg/cm 2 . Prior work shows that PEG likely adsorbs in a spherical shape rather than a rod-like shape with a calculated mass density of ca. 0.195 μg/cm 2 . 15 Using this same protocol, we calculate that PPG alone would give a mass density of 0.192 μg/cm 2 . Experimentally, PPG leads to substantially increased mass on the surface relative to PEG, compared to these calculated surface layers of each suppressor. This increased mass cannot be explained by increased Cu or other components, and must reflect increased PPG associated with the electrode surface. The origin of this increased PPG association likely resides in the increased hydrophobicity of the PPG molecule, which makes it more likely to associate with the electrode and less likely to remain in solution. PPG is expected to be dense and closely packed in aqueous environment, whereas PEG has the ability to interact with solvent water molecules. 41,52 The insight from the QCM measurement is consistent with that from the spectroscopy. In particular the conformation of PPG as a suspension in solution and the conformation of PPG on a Cu surface are more similar to each other than the conformations of PEG between those two environments. The PPG molecules are likely compact in aqueous solution due to their hydrophobicity. Consequently, the small change in conformation as determined by the spectrosocopy upon interacting at a Cu surface suggests that PPG is still compact at the surface. This compact state is reflected in the QCM results. On the other hand the PEG conformation changes considerably more on going from solution to a Cu surface, resulting in a less dense surface layer relative to PPG measured by QCM. This observation suggests PPG more readily associates with the surface, maintaining its gauche/trans conformation, while PEG changes conformation while interacting at the surface.
In conclusion, PPG interacts with a Cu UPD layer on Au absent Cl − , whereas PEG does not. Figure 7 depicts the difference between the Cl − mediated PEG interaction and the direct surface/hydrophobic interaction of PPG. In the presence of Cl − , more surface mass density is achieved with electrolyte containing PPG than with PEG, suggesting PPG forms a denser surface layer, forms multiple layers, or both. The unique ability of PPG to associate with the electrode in the absence of Cl − , while not inhibiting Cu electrodeposition, suggests that this PPG association is weak. The increased facility of PPG to form layers on the electrode relative to PEG is likely again a consequence of the increased hydrophobicity of the PPG molecule. Understanding the role hydrophilicity/hydrophobicity plays in suppressor behavior is imperative to developing advanced, high-performance suppressors. This work indicates that combining hydrophilic and hydrophobic components together in suppressor molecules allows tuning of adsorption D695 behavior, leading to enhanced interfacial control during copper electrodeposition.