Synthesis of Quaternary Ammonium Gemini Levelers and Their Action Mechanisms in Microvias Void-Free Copper Filling

Abstract

Developing a highly efficient leveler in acid copper electroplating solution is one of the primary tasks necessary for achieving superconformal filling of microvias and interconnections in printed circuit boards (PCBs). Two triethylenediamine-based Gemini levelers, both with terminal quaternary ammonium groups, are synthesized and named as GL1 (C8) after reaction of triethylenediamine with 1,8-dichlorooctane and GL2 (C6 with two C–O linkages) after triethylenediamine with 1,2-bis(2-chloroethoxy) ethane. Electrochemical experiments indicate that at 100 rpm and 1000 rpm GL2 combines with a suppressor and accelerator to exhibit greater potential difference of 23 mV than GL1 in 9 mV for Cu²⁺ reduction, demonstrating that GL2 has a stronger synergistic convection-dependent adsorption (CDA) effect. Microvias copper electroplating experiments confirm that acid copper electroplating solution containing GL2 achieve more effective superconformal void-free filling as it results in FP = 96.1%, while the solution containing GL1 results in FP = 70%. Theoretical calculations indicate that adsorption energy of GL2 is −1037.54 kJ·mol⁻¹, which is lower than GL1 (−1019.06 kJ·mol⁻¹). GL2 displays lower electron density compared to GL1, which facilitates its displacement by accelerator at the bottom. The lower adsorption energy of GL2 suggests the weaker adsorption ability and the stronger CDA behavior.

1. Introduction

In recent years, there has been a rapid evolution of electronic products towards miniaturization [1], compactness [2], convenience [3], and integration of 5G intelligence [4]. High-density interconnected (HDI) multilayer printed circuit boards (PCBs) are indispensable components in these advanced electronic devices and products [5]. The copper electrodeposition processes for achieving superconformal microvia fillings play a key role in enabling the electrical interconnections within PCBs and have emerged as essential technologies in the manufacturing of HDI structures [6,7]. Acid copper electroplating, offering benefits such as low operating temperature, cost-effectiveness, high deposition rates, and solution stability [8], has gained popularity as a primary technique for the superconformal filling of microvias. However, as via diameter continues to decrease and aspect ratio increases, superconformal copper filling faces significant challenges for the serious gradients of copper ions concentration and electric field inside and outside the microvias, and becomes one of the critical manufacturing processes [9,10].

To achieve the copper superconformal filling in microvias, additives in acidic copper plating solution are typically employed by their synergistic effects to suppress copper deposition at via mouth while accelerating deposition at the bottom, resulting in a “bottom-up” filling behavior. Additives can be broadly classified into three categories [10]: accelerator [11], suppressor [12], and leveler [13]. Accelerators are generally the low-molecular-weight organic compounds containing sulfur, such as (bis-(sodium sulfopropyl)-disulfide (SPS) [11] and sodium mercaptopropanesulfonate (MPS) [14]. Apart from enhancing copper deposition, accelerators impact the surface morphology and crystal orientation of the copper layer, resulting in a smooth and shiny appearance. Suppressors are typically the long-chain compounds composed of polymeric alcohols or polyethers, such as polyethylene glycol (PEG) [15] and ethylene oxide–propylene oxide block copolymers (EO/PO) [16]. Suppressors inhibit copper ions electroreduction at both the microvias mouth and bottom by forming a barrier layer through the adsorption of chloride ions, forming a PEG−Cu⁺−Cl⁻ complex [17] on the cathode surface. Levelers are usually the positively charged organic compounds containing nitrogen or nitrogen-containing heterocycles [11], including azo dyes such as Janan Green B [18], as well as nitrogen-containing heterocyclic compounds like 2-mercaptopyridine [19] and 2-amino-4-methylbenzothiazole [6]. It is well-known that diazo compounds, such as JGB, can undergo breakage of the nitrogen–nitrogen double bond to form N, N-dimethylbenzene-1,4-diamine, which may impact the stability of copper electroplating in electronic applications. The positively charged functional group of the organic compound exhibits strong adsorption ability on the cathode surface and preferentially accumulates at the high-current-density regions to suppress copper ions electroreduction there. Leveler plays a crucial role in achieving superconformal filling of microvias.

Several additive action mechanisms have been proposed to elucidate the concept of microvias superconformal filling. Moffat [20,21] reported a mechanism of curvature enhanced accelerator coverage (CEAC), which said that the continuous accumulation of SPS at the bottom of microvias promoted copper deposition therein and achieved the superconformal filling. West introduced the diffusion–adsorption mechanism [22,23], which suggested that suppressors were more adsorbed at the microvias mouth, increasing copper deposition rate at the bottom and allowing for a “bottom-up” superconformal filling. Akolkar reported the transient diffusion and surface adsorption (TDSA) mechanism [24], in which PEG behaved at a low diffusion rate and took a longer time to reach the bottom of microvias. In contrast, SPS, which behaved at a high diffusion rate, could arrive more quickly to the microvias bottom. Due to the higher coverage of SPS at the bottom, PEG preferentially achieved a greater coverage at the mouth of microvias compared to SPS. The preferred coverages of accelerator at the bottom and inhibitor at the mouth facilitated the realization of microvias superconformal filling. Furthermore, Dow introduced a convection-dependent adsorption (CDA) mechanism [25], which posited that additives demonstrated transfer adsorption behavior; at the strong convection area of microvias mouth the composite suppressor compound mainly took action to suppress the copper filling, whereas at the weak convection area the accelerator presented mainly action to accelerate copper filling, leading to superconformal copper filling. The CDA mechanism has been one of the popular accepted theories.

Gemini surfactants [26] are a class of symmetrical molecules consisting of two hydrophilic or hydrophobic groups connected by a spacer such as an alkyl chain. Compared to single-headed surfactants, Gemini surfactants exhibit higher adsorption efficiency in reducing oil/water interfacial tension [27]. Additionally, Gemini surfactants typically demonstrate strong interfacial activity and good wettability, making them widely applicable in skincare products [28], pharmaceuticals [29], life sciences [30], oilfield chemistry [31], and porous materials [32]. Triethylenediamine, characterized by a simple structure [33], low cost [34], and stability [35], has garnered significant research interest in recent years for its application in CO₂ absorption [36]. Compared to JGB, which contains a nitrogen–nitrogen double bond, triethylenediamine with a carbon–nitrogen single bond exhibits greater stability. Therefore, it is clear that the products synthesized from triethylenediamine and haloalkanes will exhibit a Gemini structure at both ends of the alkyl chain. These products can be easily modified to form positively charged quaternary ammonium structures and used as the leveler (designated as the Gemini leveler) in an acid copper electroplating solution. A Gemini leveler with quaternary ammonium structures can strongly adsorb onto the cathode surface, particularly in areas of high current density (corresponding to the mouth of microvias), to effectively inhibit copper ions reduction. Hence, the Gemini leveler represents a crucial candidate for achieving superconformal copper filling in microvias.

Up to now, there have been no studies on the application and action mechanism of triethylenediamine derivative in copper electroplating for microvias superconformal filling. Therefore, it is essential to cleverly design a straightforward synthetic route to produce hydrophilic quaternary ammonium structures as a Gemini leveler from triethylenediamine. Developing a novel acid copper electroplating process based on a Gemini leveler and investigating its effect on microvias copper filling, as well as the action mechanism of the Gemini leveler, can not only enrich the varieties of levelers but also holds significant practical and theoretical value for PCB and the semiconductor packaging manufacturing.

In this study, triethylenediamine is used as a substrate, while 1,8-dichoctane and 1,2-bis(2-chloroethoxy) ethane, respectively, served as the reactants to synthesize the following two quaternary ammonium Gemini levelers: one with an 8-carbon chain (GL1, Figure 1) and the other with a 6-carbon chain containing two C–O structures (GL2, Figure 1). Through chronopotentiometric experiments and cyclic voltammetry measurements, we study the polarizing effects of the Gemini levelers on the reduction of copper ions, the synergistic effects, and the convection-dependent adsorption behavior of the additives, to reveal the theoretical feasibility of microvias superconformal filling. Electroplating experiments are carried out to verify the performances of the quaternary ammonium Gemini levelers in microvias filling. Quantum chemical calculation is conducted to study the adsorption configurations, energies, and electrostatic potential of the Gemini levelers. Additionally, atomic force microscopy (AFM) and X-ray diffraction (XRD) analyses are used to characterize the surface morphology, surface roughness, and crystal orientation of the copper layer.

2. Experimental

2.1. Materials

All experiments were conducted at 298 K. The analytical-grade reagents of triethylenediamine, 1,8-dichoctane and 1,2-bis(2-chloroethoxy) ethane were bought from Macklin Chemical Technology Co., Ltd. (Shanghai, China); Copper(II) sulfate pentahydrate, sulfuric acid, and sodium chloride were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China); Polyethylene glycol (PEG, Mn: 8000), bis-(sodium sulfopropyl)-disulfide (SPS) and dimethyl sulfoxide-d6 (DMSO-d₆) were bought from Aladdin Co., Ltd. (Shanghai, China). Unless otherwise specified, all reagents were used without further purification.

2.2. Characterization

The ¹H NMR and ¹³C NMR spectra of Me₄Si were measured at 0 ppm using DMSO-d₆ as the solvent with a Bruker Bio-Spin GmbH 500 MHz NMR spectrometer. Mass spectrometry was performed using an Agilent 1290-6545XT ESI mass spectrometer (ESI, positive mode, 60–1000 amu).

A heating magnetic stirrer with an integrated oil bath was Supplementary Materials employed as the heating setup to maintain the temperature at 353 K. Triethylenediamine, 1,8-dichlorooctane, or 1,2-bis(2-chloroethoxy) ethane were introduced into a 50 mL three-necked flask, with 10 mL acetonitrile serving as the reaction solvent combined to reflux. A thermometer (range: 100 °C; accuracy: ±1 °C) was immersed in the solution to monitor the temperature throughout the reaction for 24 h.

Synthesis of GL1: We accurately weighed 0.5507 g (4 mmol) of triethylenediamine and 0.4497 g (2 mmol) of 1,8-dichooctane. Afterward, we stopped stirring and heating, and allowed the mixture to cool to room temperature. We isolated the white solid raw product and performed column chromatography using petroleum ether/ethyl acetate (8:2, v/v) for separation. We collected the liquid purified product and evaporated the solvent using a rotary evaporator to obtain the refined product, which appeared as a white solid powder of a weight of 0.2785 g and a total yield of 51%. ¹H NMR (500 MHz, DMSO-d₆), δ (TMS 0 ppm):3.33 (t,6H), 3.25–3.22 (m,4H), 1.7 (s,2H), and 1.3 (s,2H). ¹³C NMR (125 MHz, DMSO-d⁶), δ (TMS 0 ppm): 63.0, 51.5, 44.7, 25.3, and 20.8. ESI-MS: m/z C₁₈H₃₆N₄²⁺ [M]²⁺: 154.1, found: 154.1. The details are shown in Supplementary Materials Figures S1–S3.

Synthesis of GL2: We accurately weighed 0.4484 g (4 mmol) of triethylenediamine and 0.3741 g (2 mmol) of 1,2-bis(2-chloroethoxy) ethane, Afterward, we stopped stirring and heating, and allowed the mixture to cool to room temperature. We isolated white raw product and performed column chromatography using petroleum ether/ethyl acetate (8:2, v/v) as the eluent for separation. We collected the purified product and evaporated the solvent using a rotary evaporator to obtain the refined product, which appeared as a viscous white liquid of a weight of 0.2024 g and a yield of 45%. ¹H NMR (500 MHz, DMSO-d6), δ (TMS 0 ppm): 3.88 (s,2H), 3.60 (d,2H), 3.49 (t,2H), 3.43–3.37 (m,6H), and 3.04 (m,6H). ¹³C NMR (125 MHz, DMSO-d₆), δ (TMS 0 ppm):70.1, 69.8, 63.3, 52.5, and 52.0. ESI-MS: m/z C₁₈H₃₆N₄O₂²⁺ [M]²⁺: 170.1, found: 170.1 The details are shown in Supplementary Materials Figures S4–S6.

2.3. Electrochemical Experiments

Electrochemical tests were conducted using a three-electrodes system, with a CHI 760E potentiostat. A rotating disk platinum electrode (5 mm in diameter) functioned as the working electrode, while a saturated calomel electrode served as the reference electrode and a platinum sheet electrode acted as the counter electrode. The electrolyte solution consisted of 200 g/L CuSO₄·5H₂O, 50 g/L H₂SO₄, and 60 mg/L Cl⁻. For the chronopotentiometric test, a constant current density of 20 mA/cm² was applied. In cyclic voltammetry tests, the working electrode was swept from the open circuit potential (approximately 0.1 V) to −0.65 V, followed by a positive sweep to 0.8 V, and a final negative sweep back to the open circuit potential, with a scan rate of 20 mV/s. All solutions were prepared using ultrapure water, and the ultrapure water was prepared using a Milli-Q IQ ultrapure water system (Merck KGaA, Darmstadt, Germany, resistivity:18.25 mΩ·cm@298 K).

2.4. Microvias Copper Electroplating

The electroless plated-copper printed circuit board (PCB) samples were supplied by Shanghai Meadville Electronics Co., Ltd., (Shanghai, China). The samples had a vias diameter of 100 μm, a depth of 90 μm, a spacing of 200 μm between the vias, and a total of 400 vias. Titanium mesh coated with iridium oxide was used as the anode and nylon cloth as the anode bag. A rectangular PCB sample board was used as the cathode, and was 12 cm in length and 5 cm in width. The composition of the plating solution and the electroplating conditions are detailed in

Table 1. The bath composition and plating parameters.

Bath Composition and Plating Parameters	Optimum Condition
CuSO4·5H2O (g/L)	200
H2SO4 (g/L)	50
Cl− (mg/L)	60
PEG−8000 (mg/L)	500
SPS (mg/L)	2
GL1 or GL2 (mg/L)	5
Cathode current density (mA/cm2)	20
Electroplating time (min)	90

.

2.5. Calculation of the Adsorption Energies of GL1 and GL2

The computational methods used in this experiment were based on density functional theory (DFT) implemented in the CP2K software package (version 7.1). Molecular optimization was performed using a double-zeta basis set with polarization functions (DZVP). Goedecker–Teter–Hutter (GTH) pseudopotentials [37] were employed for H 1s orbitals and C, N, and O 2s2p orbitals. The cutoff energy for electron density plane waves (PW) was set to 800 Rydberg. The electronic exchange and correlation energy were calculated using the Perdew–Burke–Ernzerhof (PBE) functional, along with Grimme D3 dispersion correction [38,39]. The convergence criteria for the self-consistent field (SCF) electronic gradient and total energy were set to 3.0 × 10⁻⁷ and 1.0 × 10⁻¹³, respectively. The convergence criteria for structural optimization were as follows: MAX_FORCE = 1.5 × 10⁻⁵, RMS_FORCE = 1.0 × 10⁻⁵, MAX_DR = 6.0 × 10⁻⁵, and RMS_DR = 4.0 × 10⁻⁵. For the structural optimization and single-point energy calculations of the additive molecules on the Cu metal layer, wave function optimization utilized the standard diagonalization (DIAG) method. For the single-point energy calculations of the additive molecules, wave function optimization utilized the DIIS optimizer’s orbital transformation (OT) method.

The surface orientation of the Cu metal layer model was (111), consisting of four atomic layers, with each layer arranged in a 12 × 12 planar model. The adsorption configurations and energies of the additive molecules on different crystal planes were analyzed. The additive molecules were generated using the RDKit package [40] and were positioned at a distance of 4–6 Å from the metal layer surface after structural optimization using the MMFF force field. The dimensions of the simulation box were 25.24 Å × 21.86 Å × 26.18 Å, which included a 20 Å vacuum layer perpendicular to the surface, and periodic boundary conditions were applied in all three dimensions [41]. The formula for calculating the adsorption energy is given as follows:

$$E_{\mathit{add}}=E_{\mathit{Mol}/\mathit{surf}}-E_{\mathit{Mol}}-E_{\mathit{Surf}}$$

(1)

where E_Mol/surf is the total energy of the additive molecules and the metal layer together, E_Mol is the total energy of the additive molecule alone, and E_surf is the total energy of the metal layer alone. Note that both E_Mol and E_surf are computed based on the optimized adsorption structures.

2.6. Surface Morphology and Structure

Surface roughness of the copper layer was measured using an atomic force microscope (AFM, Oxford Instrument Asylum Research Cypher ES, Santa Barbara, CA, USA), employing a silicon probe coated with antimony in the tapping mode. Structural characterization of the copper layer was performed using a Rigaku Ultima IV X-ray diffractometer (XRD, Rigaku, Tokyo, Japan) with a Cu-Kα target, operating at a voltage of 40 kV and a current of 40 mA (λ = 0.15 nm). The 2θ range was set from 30° to 90°, with a scan rate of 10° per minute. The texture coefficient (TC) is defined as the ratio of the relative diffraction intensity of the crystal plane to the total diffraction intensity of all crystal planes, as described in Equation (1), where I_hkl and I_{0 hkl} denote the diffraction intensities of the electroplated copper sample and standard copper [42] (PDF#89-2838), respectively. As the TC value of a metal crystal plane exceeded the average value of 1/n (where n is the number of crystal planes examined), it was considered to exhibit preferred orientation, with higher TC values indicating a greater degree of preferential orientation.

$$\mathrm{T}{\mathrm{C}}_{\mathrm{hkl}}={\displaystyle \frac{I_{\mathit{hkl}}/I_{0\mathit{hkl}}}{\mathsf{\Sigma }{I}_{\mathit{hkl}}/I_{0\mathit{hkl}}}}\times 100\%$$

(2)

3. Results and Discussion

3.1. Synthesis and Characterization of Gemini Levelers

Triethylenediamine is reacted separately with 1,8-dichlorooctane and 1,2-bis(2-chloroethoxy) ethane through the Hofmann alkylation reaction to synthesize quaternary ammonium GL1, which contains an 8-carbon chain, and quaternary ammonium GL2, which features a 6-carbon chain and two C–O structures. The specific synthesis routes are illustrated in Figure 1.

The synthesized GL1 and GL2 show no impurities in ¹H NMR and ¹³C NMR spectra, indicating that the products are relatively pure. Both the Gemini levelers are the derivatives of triethylenediamine. GL1 features an 8-carbon alkyl chain with two quaternary ammonium structures at the both ends of alkyl chain, and appears as a white solid with a corresponding molecular weight of 336.2. Differently, GL2 contains a 6-carbon alkyl chain and two C–O structures and also has two quaternary ammonium structures at both ends of the alkyl chain, and it presents in a viscous white liquid with a molecular weight of 340.2. Notably, both Gemini levelers exhibit good solubility in water.

3.2. Galvanostatic Measurements

The galvanostatic experiment is conducted to investigate the polarizing behaviors and convection-dependent adsorption effects of GL1 and GL2 on copper ion reduction at a platinum rotating disk electrode (Pt−RDE) with a constant current density of 20 mA/cm². The chronopotentiometric curves are recorded by the sequential addition of additives to the acid copper plating solution, as shown in Figure 2. The Pt−RDE is rotated at speeds of 1000 rpm and 100 rpm to simulate the copper ion transport environments at the microvias mouth and bottom, respectively. By analyzing the changes in potentials, we evaluate the polarization behavior of copper ions reduction by additives. The potential difference (Δη), calculate as Δη = η_{100 rpm} − η_{1000 rpm}, is used to assess the convection-dependent adsorption effect of the additives during copper ions reduction and to predict the filling performance of microvias based on the value of Δη. A positive Δη indicates lower polarization behavior and higher copper deposition rate at the bottom to mouth, and the theoretical feasibility of achieving superconformal filling of microvias; while a negative Δη suggests the opposite and that superconformal filling cannot be realized.

The results indicate that in the virgin make-up solution (VMS, CuSO₄ 200 g/L, H₂SO₄ 50 g/L, Cl⁻ 60 mg/L) the electrode potentials at the rotation speeds remain almost the same at around −0.48 V, suggesting that there is no convection-dependent adsorption effect in VMS. After adding 500 mg/L of polyethylene glycol (PEG) at 500 s, the electrode potential sharply decreases from −0.48 V to −0.71 V, indicating that PEG exhibits as a suppressor of strong inhibition characteristics in copper ions reduction. The potential remains unchanged at both high and low rotation speeds, indicating that PEG also does not exhibit a convection-dependent adsorption effect. At 1000 s, the further addition of 2 mg/L of bis-(sodium sulfopropyl)-disulfide (SPS) causes the electrode potential to gradually rise to approximately −0.62 V. This depolarization effect indicates that the adsorption sites occupied by PEG are increasingly replaced by SPS, and SPS exhibits as an accelerator for promoting copper ions reduction. Notably, the electrode potentials at the rotation speeds align closely, indicating that the mass transport of both PEG and SPS is not controlled by a convection-dependent adsorption effect.

At 1500 s, 5 mg/L of GL1 and GL2 are again added, respectively, as shown in Figure 2a,b. The electrode potentials gradually shift negatively for both of the Gemini levelers, indicating a suppression effect on copper ions reduction. It is noticed that there happens to be a more pronounced negative shift at higher rotation speed. At 100 rpm and 1000 rpm, the potential difference for the solution containing GL1 is 9 mV (Figure 2a), whereas for the solution containing GL2 it reaches 23 mV (Figure 2b). These non-overlapping electrode potentials at the low and high rotation speeds suggest that the mixture of the three additives exhibits a convection-dependent adsorption effect during copper ions electroreduction. Compared to GL1, GL2 causes a larger potential difference, indicating a stronger convection-dependent adsorption effect, thereby increasing the theoretical likelihood of achieving microvias superconformal filling.

3.3. Cyclic Voltammetry Measurements

Cyclic voltammetry experiments are conducted to assess the differences in anodic dissolving charges associated with copper deposition under 1000 rpm and 100 rpm rotation speeds and the influence of additives. This analysis can provide insights into the polarization behavior and convection-dependent adsorption effects of copper ions reduction which are impacted by GL1 and GL2, as well as predicting the superconformal microvias filling abilities of the created solutions. Figure 3 displays the cyclic voltammetry curves at the platinum rotating disk electrode (Pt−RDE) at 100 rpm and 1000 rpm in VMS alongside the addition of various combinations of additives. The anodic charge from a definite solution for copper deposition is listed in

Table 2. Anodic stripping charges (mC/cm²) measured from Figure 3.

Solution Composition	Q100 rpm (mC/cm2)	Q1000 rpm (mC/cm2)	ΔQ%
VMS	471.5	474.0	−0.5
VMS + GL1	257.0	220.0	15
VMS + SPS	515.0	523.0	−1.5
VMS + SPS + GL1	230.0	162.0	30
VMS + PEG	16.05	15.85	1.0
VMS + PEG + GL1	2.85	2.75	3.0
VMS + PEG + SPS	144.5	140.0	3.0
VMS + PEG + SPS + GL1	24.5	17.5	29

. To quantitatively describe the convective diffusion effects of the additives on copper ions reduction, Equation (3) is defined as follows:

$$\Delta Q={\displaystyle \frac{Q_{100\mathrm{r}/\min }-Q_{1000\mathrm{r}/\min }}{Q_{100\mathrm{r}/\min }}}\times 100\%$$

(3)

where Q_{100 rpm} and Q_{1000 rpm} represent the anodic peak areas (charges) at 100 rpm and 1000 rpm, respectively. In contrast to the results where ΔQ < 0, a positive ΔQ > 0 indicates that during the electroplating of copper there exists a convection-dependent adsorption effect from the additives, with the deposition rate of copper at the microvias bottom higher than that at the microvias mouth. The larger the ΔQ value is, the stronger the convection-dependent adsorption effect of the additives will be, inferring the higher theoretical feasibility of achieving “bottom-up” superconformal filling of microvias.

As shown in Figure 3a and Table 2, in the virgin make-up solution (VMS, CuSO₄ 200 g/L, H₂SO₄ 50 g/L, Cl⁻ 60 mg/L), the anodic peak areas (charges) at 1000 rpm and 100 rpm are nearly identical, measuring 474.0 mC/cm² and 471.5 mC/cm², respectively. This indicates that the reduction of copper ions is controlled by the charge transfer. After adding 5 mg/L of GL1 to the VMS, the anodic charge significantly decreased at both high and low rotation speeds, suggesting that GL1 inhibits copper ions reduction. The anodic charges at the high and low speeds are 220 mC/cm² and 257 mC/cm², resulting in a difference of 37 mC/cm², indicating the convection-dependent adsorption effect associated with GL1. When only 2 mg/L SPS is added to the VMS (Figure 3b and Table 2), the anodic charges slightly increased to 523 mC/cm² and 515 mC/cm², with a minimal difference of 8 mC/cm², suggesting that SPS has a very weak convection-dependent adsorption effect and an acceleration effect on copper ions reduction. Upon further addition of GL1, the anodic charges at the both speeds decreased to 162.0 mC/cm² and 230 mC/cm², which are lower than those in VMS with only SPS or only GL1, indicating a synergistic inhibitory effect of SPS and GL1 on copper ions reduction. The charge difference is 68 mC/cm², with a ΔQ of 30%, which is higher than the ΔQ of 15% obtained in VMS with only GL1, demonstrating a synergistic convection-dependent adsorption effect for the mixture of GL1 and SPS. When GL1 is added to the SPS-containing VMS solution, the charge difference at 1000 rpm is 361 mC/cm² (523–162.0 mC/cm²), whereas at 100 rpm the charge difference is 285 mC/cm² (515–230 mC/cm²). This indicates that the synergistic inhibitory effect of GL1 and SPS is stronger at the higher rotation speed (referring to microvias mouth) than that at the lower speed (microvias bottom). After adding only 500 mg/L PEG to the VMS (Figure 3c and Table 2), the anodic charges at both speeds sharply decrease to 16.05 mC/cm² and 15.85 mC/cm², with a negligible difference of 0.2 mC/cm², indicating that the suppressor PEG exhibits a strong inhibitory effect on copper deposition and demonstrating an equal inhibition at both the microvias mouth and bottom and without convection-dependent adsorption effect. Further addition of GL1 reduces the anodic charges at the high and low speeds to 2.75 mC/cm² and 2.85 mC/cm², respectively, with a minimal difference of 0.1 mC/cm², suggesting that GL1 in conjunction with PEG inhibits copper ions reduction and does not exhibit a synergistic convection-dependent adsorption effect.

When PEG and SPS coexist in the VMS (Figure 3d and Table 2), the anodic charges at 1000 rpm and 100 rpm are recorded at 140.0 mC/cm² and 144.5 mC/cm², respectively. These values are higher than those in VMS with only PEG but lower than those with only SPS, indicating a mutual weakening effect of the two additives. The small charge difference of 4.5 mC/cm² suggests that PEG and SPS exhibit equal polarization at both the microvias mouth and bottom for their minimal convection-dependent adsorption effect. Upon further addition of GL1 to the solution containing PEG and SPS, the anodic charges at the high and low speeds are measured at 17.5 mC/cm² and 24.5 mC/cm², respectively, resulting in a ΔQ of 29% which is similar to the ΔQ of 30% when SPS and GL1 coexist. An ΔQ of 29% also indicates the convection-dependent adsorption effect of the three additives.

Under the same experimental conditions, the cyclic voltammetry (CV) experiments at Pt−RDE in virgin make-up solution (VMS, CuSO₄ 200 g/L, H₂SO₄ 50 g/L, and Cl⁻ 60 mg/L) with GL2 are presented in Figure 4, and the corresponding anodic charges are shown in

Table 3. Anodic stripping charges (mC/cm²) measured from Figure 4.

Solution Composition	Q100 rpm (mC/cm2)	Q1000 rpm (mC/cm2)	ΔQ%
VMS	471.5	474.0	–0.5
VMS + GL2	146.5	124.5	16
VMS + SPS	515.0	523.0	–1.0
VMS + SPS + GL2	162.5	85.0	48
VMS + PEG	16.05	15.85	1.0
VMS + PEG + GL2	2.50	2.40	4.0
VMS + PEG + SPS	144.5	140.0	–4.0
VMS + PEG + SPS + GL2	30.0	18.0	40

. As indicated in Table 3, when only 5 mg/L GL2 is present in VMS the anodic charges decrease to 124.5 mC/cm² and 146.5 mC/cm² at 100 rpm and 1000 rpm, respectively, resulting in a difference of 22 mC/cm². This observation indicates that GL2 inhibits the reduction of copper ions and has a convection-dependent adsorption effect. Compared to GL1, GL2 exhibits a stronger inhibitory effect on copper ions reduction. When both SPS and GL2 are included in the VMS, the anodic charges also decrease to 85.0 mC/cm² and 162.5 mC/cm² at both the rotation speeds, with a difference of 77.5 mC/cm². This finding is consistent with the effect of GL1. GL2 and SPS also exhibit a synergistic inhibitory effect on copper ions reduction, which is accompanied by a convection-dependent adsorption effect. It is very likely that the continuous accumulation of SPS at the microvias bottom achieves “bottom-up” superconformal filling. Furthermore, when PEG is incorporated with GL2, the anodic charges further diminish to 2.40 mC/cm² and 2.50 mC/cm² at both the rotation speeds, with a negligible difference of 0.1 mC/cm². This result is also akin to GL1, but GL2 collaborates with PEG to further suppress copper ions reduction, with a negligible convection-dependent adsorption effect. Finally, following the addition of 5 mg/L GL2 to VMS containing both 500 mg/L PEG and 2 mg/L SPS, the anodic charges at 1000 rpm and 100 rpm were measured as 18.0 mC/cm² and 30.0 mC/cm², corresponding to a difference of 12 mC/cm² and a ΔQ value of 40%. The largest ΔQ value of 40% reveals the strongest convection-dependent adsorption effect in VMS with the mixture of the three additives, demonstrating the theoretical feasibility of “bottom-up” superconformal microvias filling.

3.4. Copper Electroplating for Microvias Filling

The electroplating experiments can verify the theoretical guidance of copper superconformal filling based on the Gemini levelers. The filling performance (FP) is a significant parameter for assessing the effectiveness of copper microvias filling. The filling performance [42] (FP) is defined by the equation FP = (H₁/H₂) × 100%, where H₁ represents the distance from the top center of microvias to its bottom, and H₂ denotes the distance from the mouth to the bottom of microvias. Typically in practice, an FP value greater than 80% is required. The copper electroplating formula and conditions based on GL1 and GL2 are detailed in Table 1. The microvias have a diameter of 100 μm, depth of 90 μm, copper foil thickness of 10 μm, and electroless deposited copper layer of 0.2 μm. The electroplating experiment is carried out at a current density of 20 mA/cm² and a plating time of 90 min. The cross-section of the electroplated copper in the microvias is illustrated in Figure 5.

The cross-section diagram of microvia before copper electroplating is presented in Figure 5a. Figure 5b demonstrates that with the addition of 500 mg/L PEG and 2 mg/L SPS to VMS (200 g/L CuSO₄, 50 g/L H₂SO₄, and 60 mg/L Cl⁻), the copper layer thickness is only about 5 μm at bottom of the microvias and 30 μm on the surface. The resulting FP is calculated to be only 5%. Figure 5c presents the result from the formula with GL1, showing a surface copper thickness of 28 μm and a copper thickness of 70 μm within the microvias, with an FP value of 70%. Although the copper filling inside the microvias is without any void, it is not completely superconformal filling. Conversely, Figure 5d illustrates the result from the formula with GL2, showing the “bottom-up” superconformal filling of microvias to be void-free. The surface copper thickness is 28 μm, and the average FP value is as high as 96.1%. This demonstrates that the electroplating formula based on GL2 plays a higher superconformal filling efficiency. Therefore, we can understand that the practice of copper electroplating for microvias filling aligns perfectly with the electrochemical predictions. Notably, GL2, which features two C–O branched structure, exhibits superior microvias superconformal filling performance compared to GL1, which contains only the C–C alkane branched structure.

3.5. Adsorption Configurations and Energies of the Gemini Levelers

The two Gemini levelers, which possess identical quaternary ammonium structures but the different branch configurations, exhibit varying microvias filling performances. By employing quantum mechanical methods, we calculate the adsorption structures and energies of these additives on a Cu electrode surface to elucidate the molecule action mechanism of the Gemini levelers and to provide a further understanding of the relationship between the Gemini levelers functionalities and their corresponding filling effects.

Figure 6a,b illustrate the optimized adsorption configurations of GL1 and GL2 on a Cu(111) surface (front view). Calculations indicate that both the molecules of GL1 and GL2 at a concentration of 5 mg/L adsorb in a parallel orientation on the copper surface, with the positively charged nitrogen (N) serving as the primary adsorption atom and site. The adsorption energies of GL1 and GL2, calculated based on their configurations, are −1019.06 kJ·mol⁻¹ and −1037.54 kJ·mol⁻¹, respectively. GL2, which features a branch structure containing two C–O group and six carbon atoms, exhibits a lower adsorption energy, indicating a lower capacity for adsorption on the copper surface. Additionally, the electrostatic potential (ESP) plot serves as a valuable tool for visualizing preferential reaction sites and predicting molecular properties. As illustrated in Figure 6, the ESP maps reveal electron density distributions, where blue and red regions correspond to high and low electron density, respectively. Both GL1 and GL2 exhibit concentrated high electron density (blue regions) around their quaternary ammonium cations, suggesting strong electrostatic attraction to electron-rich areas of the copper surface during electroplating. GL2 with a 6-carbon chain contains two C–O groups (−1037.54 kJ·mol⁻¹, red region) displays lower electron density compared to GL1 with an 8-carbon chain (−1019.06 kJ·mol⁻¹, white region). This reduced electron density correlates with GL2’s lower adsorption energy, facilitating its displacement by an accelerator to achieve bottom-up filling. This result suggests that during the copper electroplating GL2 combines with the suppressor and accelerator to produce a stronger convection-dependent adsorption effect, resulting in a stronger inhibition impact at the microvias mouth; however the weaker adsorption strength of GL2 results in insufficiently supplied release at the microvias bottom. Consequently, this promotes the continuous accumulation of SPS at the microvias bottom and is in favor of copper deposition, which helps to achieve “bottom-up” superconformal filling [11].

3.6. Surface Morphology and Structure of Copper Layer

The appearances of copper coatings electroplated from the formulae, respectively, with GL1 and GL2 are both neat and bright. Quantitative characterization is conducted by atomic force microscopy (AFM) experiment (Figure 7). The root mean square roughness (Rq) and mean roughness (Ra) of the copper layer obtained from the solution containing GL1 are measured as 19.1 nm and 15.2 nm, respectively (Figure 7a). In contrast, the Rq and Ra values for GL2 are lower at 11.4 nm and 8.8 nm, respectively (Figure 7b), indicating a smoother surface of the obtained copper layer.

The X-ray diffraction (XRD) results of the electroplated copper layer are presented in Figure 8. The copper layers obtained from the plating solutions containing GL1 and GL2 exhibit a similar structure, with the crystal planes of (111), (200), and (220) at 43.4°, 50.5°, and 74.2°, respectively, indicating a face-centered cubic mixcrystal structure. Calculations reveal that the texture coefficients for the (111), (200), and (220) peaks related to GL1 are 46.5%, 27%, and 26.5%, respectively, while those for GL2 are 51.2%, 20.9%, and 28%.

4. Conclusions

The synthesis and investigation of highly efficient Gemini levelers for microvias filling possess significant practical and theoretical value. In this study, a Hofmann alkylation reaction is conducted using triethylenediamine as the substrate and 1,8-dichloro-octane and 1,2-bis(2-chloroethoxy) ethane as the reactants. We precisely designed and synthesized two Gemini levelers: GL1 and GL2. GL1 features an eight-carbon branch chain with two quaternary ammonium structures at the two ends of the chain, and GL2 features a six-carbon branch chain and two C–O groups with two quaternary ammonium structures at the two ends of the chain. Both acid copper plating solutions containing GL1 and GL2 exhibit a microvias filling effect. However, the solution with GL2 demonstrates higher superconformal filling efficiency, with an FP value of 96.1%, and a better copper layer quality.

The galvanostatic and cyclic voltammetry experiments confirm that the Gemini levelers mixed with inhibitor PEG and accelerator SPS in the acid copper plating solution exhibit a synergistic convection-dependent adsorption effect, demonstrating the theoretical feasibility of microvias superconformal filling. GL2 displays a greater convection-dependent adsorption effect than GL1, resulting in a more efficient microvias superconformal filling ability. Theoretical calculations indicate that both GL2 and GL1 adsorb in a parallel orientation of recumbent mode on the copper surface, with the positively charged nitrogen (N) serving as the primary adsorption atom and site. GL2 exhibits a lower adsorption energy, suggesting a lower capacity for adsorption on the copper surface and a lower electron density facilitating its displacement by the accelerator. Therefore, due to its stronger convection-dependent adsorption effect and weaker adsorption strength, GL2 is less likely to supply after being consumed at the bottom of microvias during copper electroplating, resulting in a greater efficiency for superconformal filling. This study has significant theoretical and practical value for the development of new levelers and for elucidating the molecular mechanisms of Gemini levelers in an acid copper electroplating bath for microvias superconformal filling.