Abstract
Cu/SiO2 hybrid bonding has been considered one of the most promising advanced packaging technologies. However, due to the inferior thermal conductivity and potentially lower bonding strength of SiO2 compared with atomic layer deposited Al2O3, Cu/Al2O3 hybrid bonding has become an attractive technology to be developed. In this work, activation and bonding conditions for Cu–Cu and Al2O3–Al2O3 hydrophilic bonding are comprehensively investigated, with a combined formic acid and inductively coupled plasma (ICP) activation strategy being proposed. Plasma activation conditions for cobonding of Cu–Cu and Al2O3–Al2O3 interfaces, such as plasma type and activation time, are optimized. The influence of formic acid treatment on Cu surface resistivity and its compatibility with Al2O3 bonding is investigated. Additionally, the influence of annealing processes on bonding quality is analyzed. It is observed that Ar ICP-treated Cu and Al2O3 surfaces exhibit minimum root-mean-square roughness values of 1.53 and 0.125 nm, respectively, at an optimized ICP activation duration of 100 s. The combined formic acid and ICP activation strategy achieves the lowest sheet resistance of 0.31 Ω/sq on Cu surfaces. After annealing, the shear force of Cu–Cu and Al2O3–Al2O3 bonding is enhanced by 51% and 163%, respectively. Through synergistic formic acid–ICP activation followed by annealing at 300 °C, tightly bonded Cu–Cu and Al2O3–Al2O3 interfaces are realized with shear strengths of 4.05 and 1.65 MPa. This work provides experimental references and guidance for Cu/Al2O3 hybrid bonding technology.


1. Introduction
In the post-Moore era, 3D integration is a key technology to support scaling and multifunctionality of integrated circuits (IC) and semiconductor devices. However, traditional integration solutions, such as solder bump or micro-bump bonding, are difficult to achieve high-density interconnection with pitches less than 10 μm. Hybrid bonding eliminates bumps through planarized direct bonding of metal-to-metal and dielectric-to-dielectric interfaces, thus providing many advantages, such as high interconnection density and low interconnection delay. Therefore, hybrid bonding injects new vitality into the research of applications such as dynamic random-access memory (DRAM), , CMOS image sensor, , and microelectro-mechanical system (MEMS). In previous works, Cu/SiO2 is commonly used for hybrid bonding. − However, the poor thermal conductivity of SiO2 limits its application in high-temperature, high-frequency, and high-power electronic devices, such as dielectrics in high-power semiconductor devices, heat sinks, and high-power LEDs. Compared with SiO2, atomic layer deposited (ALD) Al2O3 exhibits superior insulation and thermal conductivity and is expected to achieve higher bonding strength, , which makes it a promising alternative dielectric material in hybrid bonding. In terms of application scenarios, Al2O3 can be used alone as a dielectric layer in hybrid bonding, or it can be deposited on the surface of the traditional dielectric material SiO2 as a bonding contact layer to improve the mechanical strength and thermal conductivity. Therefore, it is necessary to develop the Cu/Al2O3 hybrid bonding technology.
The primary methods for Al2O3 bonding include plasma activation − and surface-activated bonding (SAB). − For plasma activation, Fan et al. demonstrated wafer bonding with 10 nm Al2O3 deposited by ALD using O2 plasma at 300 °C; Li et al. developed a sequential surface activation process employing O2 + N2 plasma to achieve wafer bonding with 50 nm ALD Al2O3 at 330 °C. For SAB, Utsumi and Takigawa investigated the crystal structure and chemical properties of sapphire–sapphire, sapphire–amorphous Al2O3 (α-Al2O3), and α-Al2O3–α-Al2O3 interfaces prepared at room temperature. The dominant approaches for Cu–Cu bonding include plasma activation, SAB, and wet treatment. , For plasma activation, Park et al. achieved bonding of 2 × 2 cm2 chips with Cu using Ar plasma at 300 °C, with experimental results indicating that Ar plasma treatment generates Cu2O on the Cu surface. The native oxides and oxides produced by plasma treatment adversely influence the electrical performance and mechanical stability of Cu interconnects. Hu et al. realized multichip-to-wafer bonding with Cu using Ar + N2 plasma at 300 °C, where N2 plasma leads to a Cu3N passivation layer on the Cu surface to suppress Cu reoxidation. For SAB, Kim et al. employed Ar fast atom beam bombardment to remove surface organic contaminants and native oxides, enabling Cu–Cu bonding. Wet treatment serves as an effective method for eliminating surface oxides on Cu. Suga et al. proposed a Cu–Cu bonding method based on in situ formic acid vapor treatment, which removes oxides while preventing Cu surface reoxidation. Jangam et al. explored localized in situ formic acid vapor treatment, achieving rapid oxide elimination within 10 s. Among the above activation methods, plasma activation has emerged as the mainstream technology for Cu–Cu and Al2O3–Al2O3 bonding due to its advantages of low cost, high efficiency, and strong material compatibility. Although Cu–Cu bonding and Al2O3–Al2O3 bonding based on plasma activation have been extensively studied separately, there is a lack of studies on the hybrid bonding strategy and optimization of compatible process conditions for the two materials. Furthermore, although introducing formic acid in plasma-activated bonding reduces surface oxides on Cu, the compatibility of formic acid treatment with Al2O3–Al2O3 bonding remains unknown. Therefore, systematic investigations into the cobonding of Cu–Cu and Al2O3–Al2O3 are essential.
In this work, conditions in the surface activation and bonding processes for Cu–Cu and Al2O3–Al2O3 hydrophilic bonding are comprehensively investigated, with a combined formic acid and ICP activation strategy being proposed and realized. The plasma activation conditions for hydrophilic bonding are optimized, the influence of formic acid on surface oxides on Cu and its compatibility with Al2O3 hydrophilic bonding are studied, and the impact of annealing on the bonding quality is analyzed. First, capacitively coupled plasma (CCP) and inductively coupled plasma (ICP) are compared to activate the surfaces of Cu and Al2O3. The surface roughness and morphology are characterized by atomic force microscopy (AFM), and contact angles are also characterized, with their variation with plasma activation duration being analyzed to determine the optimal conditions. The ICP activation mechanism is revealed by X-ray photoelectron spectroscopy (XPS). Second, the sheet resistance of Cu samples under different combinations of plasma and formic acid treatments is measured by the four-point probe method, identifying the conditions that yield the minimum sheet resistance. Moreover, it is proven that the samples treated with formic acid are compatible with the requirements of a low surface roughness and good surface hydrophilicity for hydrophilic bonding. Finally, the influence of annealing on bonding quality is investigated using shear strength tests, scanning acoustic microscopy (SAM), focused ion beam (FIB), and transmission electron microscopy (TEM) by characterizing the bonding strength and interface connection state before and after annealing.
2. Experimental Section
Figure illustrates the process flow of this work, which can be categorized into five sequential stages: sample preparation, sample pretreatment, thermal compression bonding (TCB), annealing, characterizations, and discussions. A detailed description of each step is provided as follows.
1.
Schematic diagram of the hydrophilic bonding process flow.
2.1. Sample Preparation
50 nm of Al2O3 and 500 nm of Cu were grown on 4-in. wafers by ALD and magnetron sputtering, respectively. In order to ensure the feasibility of the experiment, after Cu and Al2O3 deposition, a layer of photoresist is spin-coated on the wafer to protect the Cu and Al2O3 surfaces, and then the 4-in. wafers are cut into 1 × 1 cm2 samples using a wafer-dicing machine.
2.2. Sample Pretreatment
Prior to the pretreatment phase, the samples are sonicated in acetone, anhydrous ethanol, and DI water for 10 min to remove photoresist and contaminants, followed by drying with N2. A 50% formic acid solution is utilized to treat Cu surfaces through 20 min of immersion.
Surface activation is performed using both the CCP and inductively coupled ICP equipment. The chamber pressure is maintained at 0.6 mbar with a gas flow rate of 150 sccm, while ICP and RF powers are set to 200 and 40 W, respectively. Plasma activation durations are varied from 100 to 400 s.
2.3. TCB
Following pretreatment, TCB is performed as soon as possible to prevent the activated samples from being recontaminated and the Cu surface from being reoxidized. The TCB process is conducted at a temperature of 300 °C and a pressure of 2.5 MPa, with a holding time of 30 min. The entire procedure is carried out under an N2 atmosphere to prevent ambient air exposure.
2.4. Annealing
Annealing is performed using an oxygen-free oven, where the chamber air is evacuated prior to N2 introduction. The annealing temperature is maintained at 300 °C with a holding time of 5 h.
2.5. Characterizations and Discussions
Surface roughness is measured using AFM, while hydrophilicity is characterized via surface contact angle testing. XPS is employed to analyze surface chemical states, and the sheet resistance of Cu surfaces is quantified with a four-point probe. Bonding quality is evaluated through shear testing, SAM, and FIB-TEM.
3. Results and Discussion
3.1. Optimization of Plasma Activation Conditions
Figures (a) and (b) show the curves of root-mean-square (R q) roughness variations and surface morphologies of Cu surfaces as a function of plasma treatment duration, respectively. As observed, the surface roughness under ICP activation initially decreases and then increases with prolonged treatment duration. Specifically, the minimum R q value of 1.53 nm is achieved at 100 s of ICP activation. The decrease in R q before 100 s is primarily attributed to the removal of organic contaminants from Cu surfaces by Ar plasma, which enhances surface smoothness. Conversely, the subsequent increase in R q after 100 s results from excessive ion bombardment, leading to surface damage and aggregation. In contrast, Cu surfaces treated with CCP exhibit fluctuating R q trends, potentially associated with the nonuniform plasma density and ion energy generated by CCP. Notably, the lowest R q value of 2.66 nm occurs at 200 s of the CCP treatment. Comparative analysis of ICP and CCP impacts on Cu surface roughness and morphology reveals (1) unlike the nonmonotonic R q variation observed with CCP, ICP treatment enables easier identification of optimal activation time due to its surface roughness with monotonic change; (2) ICP activation exerts a more pronounced influence on R q reduction, achieving a lower minimum value. This stems from the higher density and more uniform plasma generated by ICP, which provides superior surface cleaning; (3) prolonged ICP treatment induces larger particle aggregation , on Cu surfaces. While moderate aggregation forms micronano structures that enhance capillary action and hydrophilic liquid spreading, excessive aggregation causes surface nonuniformity and roughness increase. The sustained R q rise after 100 s of ICP treatment may correlate with overaggregation phenomena.
2.
(a) Variation of Cu surface roughness with plasma activation duration; (b) surface morphology of Cu under different plasma activation durations; (c) variation of Al2O3 surface roughness with plasma activation duration; (d) surface morphology of Al2O3 under different plasma activation durations.
Figure (c) illustrates the R q variation curves of the Al2O3 surfaces with ICP and CCP treatment times. Both methods demonstrate an initial decrease, followed by an increase in R q. The minimum surface roughness achieved through ICP and CCP activation is 0.125 and 0.160 nm, respectively. The initial decrease results from organic contaminant removal, while the subsequent increase arises from excessive plasma-bombardment-induced surface damage. Additionally, Figure (c) reveals that ICP-activated Cu surfaces achieve minimum R q at 100 s, whereas CCP-activated surfaces require 200 s. The overall lower roughness and shorter optimized time under ICP activation are attributed to its higher cleaning efficiency compared with CCP. Figure (d) shows the surface morphologies of Al2O3 after being treated by plasma for different durations. Unlike Cu, Al2O3 surfaces exhibit no significant cluster phenomenon, indicating that plasma treatment primarily functions to clean the Al2O3 surface and enhance its surface energy.
Figure (a) illustrates the evolution of Cu surface contact angles with ICP and CCP activation durations (the test results of the surface contact angle are shown in Figure S1). It is observed that the untreated Cu sample exhibits a contact angle of 61.49 °, which rapidly decreases to 32.66 ° and 22.25 ° after 50 s of ICP and CCP activation, respectively. The rapid reduction in contact angles induced by both activation methods is primarily attributed to the following mechanisms: − (1) contaminants and the native oxide layers on the Cu surface create a low surface energy, which hinders the adsorption of water molecules, while Ar plasma can effectively remove these contaminants, exposing the intrinsic regions with high surface energy; (2) Ar plasma generates numerous Cu dangling bonds on the surface. These Cu sites react with hydrophilic functional groups to form Cu–OH structures, enhancing surface hydrophilicity and reducing contact angles. Notably, before 200 s, the contact angles of the surfaces activated by CCP are generally higher than those activated by ICP. However, after 300 s, the contact angles of the surfaces activated by CCP become lower than those activated by ICP. This reversal occurs due to the fact that compared with CCP, long-term ICP activation leads to more significant surface agglomeration, forming larger Cu particles. These particles impede the uniform spreading of the liquid, thus suppressing the reduction of the surface contact angle.
3.
(a) Variation curve of the surface contact angle of Cu with plasma activation time; (b) variation curve of the surface contact angle of Al2O3 with plasma activation time.
Figure (b) displays the evolution of Al2O3 surface contact angles with ICP and CCP activation durations (the test results of the surface contact angle are shown in Figure S1). It is observed that the as-deposited Al2O3 exhibits a contact angle of 60.48°, which stabilizes at 13.45–15.41° after 100 s of ICP activation and 13.36–15.41° after 200 s of CCP activation. The mechanisms underlying rapid contact angle reduction on Al2O3 surfaces are similar to those observed for Cu: (1) Ar plasma sputtering removes surface organic contaminants; (2) Ar plasma generates Al–O dangling bonds on the surface of Al2O3. These Al–O sites are attached by active −OH groups, forming Al–OH structures, which enhance the surface hydrophilicity. In addition, compared with CCP, the samples activated by ICP take less time to reach stability, indicating that the activation efficiency of ICP is higher.
Figures (a) and (b) present the C 1s peak fitting results for Cu and Al2O3 surfaces before and after ICP activation, respectively. Peaks observed at binding energies of 284.8, 286.4, and 289.1 eV correspond to C–C, C–O, and CO bonds, respectively. Compared with the as-deposited samples, the C signal intensity on the surface of the activated samples is significantly reduced, which indicates that plasma treatment can effectively remove organic contaminants. Figure (c) displays the Al 2p peak fitting results for Al2O3 surfaces before and after activation. From the figure, the peaks of the Al–O–Al bond at 74.6 eV and the Al–O bond at 73.9 eV can be observed. Quantitative analysis reveals that untreated samples contain 35.7% Al–O bonds and 64.3% Al–O–Al bonds, whereas plasma activation increases the Al–O bond proportion to 45% while reducing Al–O–Al bonds to 55%. This transformation indicates plasma-induced dissociation of Al–O–Al bonds, generating Al–O dangling bonds. The increased Al–O bond enhances surface hydrophilicity by facilitating the adsorption of hydrophilic functional groups.
4.
Surface C 1s peak-fitting spectra of (a) Cu and (b) Al2O3 in the as-deposited state and after plasma activation; (c) surface Al 2p peak-fitting spectra of Al2O3 in the as-deposited state and after plasma activation.
3.2. Analysis of Surface Oxides on Cu
Figure illustrates the sheet resistance values of Cu surfaces treated under six different conditions. Samples 1–6 correspond to untreated Cu, CCP-activated Cu, ICP-activated Cu, formic acid-treated Cu, formic acid + CCP-activated Cu, and formic acid + ICP-activated Cu, respectively. The measured sheet resistances are 1.01 Ω/sq (S1), 1.03 Ω/sq (S2), 0.68 Ω/sq (S3), 0.39 Ω/sq (S4), 0.41 Ω/sq (S5), and 0.31 Ω/sq (S6), respectively. Comparative analysis reveals that (1) CCP activation (S2) slightly increases sheet resistance compared to untreated Cu (S1) due to the formation of oxides on the Cu surface by CCP activation; (2) ICP activation (S3) reduces resistance below that of untreated Cu, indicating its superior oxide removal efficiency outweighs plasma-induced oxidation effects; (3) formic acid treatment (S4–S6) significantly lowers resistance compared to nonformic acid samples (S1–3), confirming effective dissolution of Cu oxide; (4) formic acid + CCP (S5) exhibits higher resistance than formic acid alone (S4), suggesting residual oxide formation during CCP processing; (5) formic acid + ICP (S6) achieves the lowest resistance, demonstrating synergistic cleaning of Cu oxides by formic acid and ICP plasma. These results demonstrate that combined formic acid + ICP treatment minimizes surface oxides, which is beneficial for improving electrical conductivity and bonding quality. ,
5.

Square resistance of Cu in different states.
Figures (a)–(c) present the O 1s XPS spectra of as-deposited, ICP-activated, and formic acid + ICP-activated Cu surfaces, respectively. The O 1s peak at 530.5 eV corresponds to Cu oxides (CuO/Cu2O), while the peak at 532.0 eV indicates organic contaminants. It is observed that the oxide content is quantified as 63% on untreated Cu surfaces, 61% after ICP activation, and 54% following formic acid + ICP activation. Figure (d) displays Cu LMM Auger spectra for the same samples. The Cu2O peak appears at 916.8 eV, and metallic Cu appears at 918.6 eV. From the figure, the rightward shift of Cu LMM peaks indicates the conversion of Cu oxides to metallic Cu. Combined O 1s and Cu LMM analyses demonstrate that ICP activation partially removes surface oxides, and formic acid + ICP treatment achieves further oxide reduction.
6.
O 1s spectra in (a) the as-deposited state; (b) ICP activation; and (c) formic acid + ICP activation states; (d) Cu LMM Auger spectra.
Figure (a) shows the comparison of the surface roughness of the samples before and after formic acid treatment. As shown in the figure, after formic acid treatment, the surface roughness of Cu increases by 0.27 nm, which may be related to the removal of the oxide layer on the Cu surface, leading to a greater degree of surface undulation. The surface roughness of Al2O3 decreases by 0.015 nm, which might be due to the dissolution of organic contaminants. Figure (b) presents the comparison of the surface contact angles of the samples before and after formic acid treatment. As can be seen, after formic acid treatment, the surface contact angles of Cu and Al2O3 decrease by 7.36° and 2.43°, respectively. In addition, the surface contact angles of the samples treated with formic acid + plasma are also smaller compared with those treated only with plasma (Figures S2–S5). This is because formic acid has the ability to remove surface oxides, clean surface contaminants, and increase surface activation energy. The analysis of the influence of formic acid on the surface roughness and hydrophilicity of the samples indicates that formic acid can improve the hydrophilicity of the samples without having an excessive adverse impact on the surface roughness.
7.
Comparison of (a) surface roughness and (b) surface contact angle of Cu and Al2O3 before and after formic acid treatment.
3.3. Analysis of Activation Mechanism
Figure shows the schematic diagrams of the respective mechanisms of formic acid treatment, CCP activation, and ICP activation. On the as-deposited surfaces of Cu and Al2O3, there are excessive organic contaminants, and the surface activity is low, which fails to meet the requirements of low surface roughness and high surface activity for hydrophilic bonding. In addition, the presence of native oxides on the Cu surface is unfavorable for improving the mechanical strength and electrical properties of Cu–Cu bonding. After formic acid treatment, it reacts with CuO on the Cu surface according to reaction , generating soluble Cu(HCOO)2 and H2O. Therefore, the oxides on the Cu surface are partially removed, and the sample surface is attached with −COOH active groups, which improves the surface hydrophilicity.
| 1 |
8.

Schematic diagram of the mechanisms of formic acid treatment, CCP activation, and ICP activation.
After plasma activation under appropriate conditions, the surface contaminants are effectively removed, and surface active groups Cu–OH and Al–OH are generated. Therefore, the surface roughness is better and the hydrophilicity is enhanced.
The action mechanisms of CCP and ICP have the following differences: in ICP, the radio frequency electric field is along the toroidal direction. The electrons accelerated by the electric field also move in the toroidal direction, having more chances to collide with Ar particles. Thus, a uniform, high-density plasma can be obtained over a large area. Driven by the toroidal electric field, the plasma moves in a spiral manner and bombards the sample surface laterally. In CCP, by applying a radio frequency voltage between two electrodes, the electrons are accelerated to ionize Ar between the electrodes, forming a plasma state. The plasma density is relatively low, and it bombards the surface vertically under the influence of the electric field. Therefore, ICP has better cleaning and activation efficiency than CCP. It can more effectively remove organic contaminants and native oxides on the Cu surface, generate more surface active groups, and is less likely to cause surface damage.
3.4. Analysis of Annealing Reinforcement
Figure (a) shows a bar chart illustrating the variation of shear force in Cu–Cu bonded samples with respect to ICP activation time (the results of the shear force test are shown in Figure S6). Notably observed from the chart is that the maximum shear force, measuring 2.68 MPa, occurs at an ICP activation time of 100 s. Beyond this 100 s activation period, the shear force exhibits a decreasing trend as the treatment time increases. This variation in shear force aligns with the pattern of changes in the Cu surface roughness with ICP activation time. Specifically, at an ICP activation time of 100 s, the surface roughness reaches its minimum, with an R q value of 1.53 nm, and no excessive agglomeration is observed on the surface, thereby resulting in a higher Cu–Cu bonding strength compared to that observed for other activation times. Conversely, when the ICP activation time exceeds 100 s, the surface roughness gradually increases, accompanied by a more pronounced agglomeration phenomenon. These combined factors lead to a deterioration in the interfacial contact state, subsequently causing a gradual decrease in the bonding shear force.
9.
Shear forces of (a) Cu–Cu and (b) Al2O3–Al2O3 bonded samples under different ICP activation times.
Figure (b) presents a bar chart depicting the variation of shear force in Al2O3–Al2O3 bonded samples with ICP activation time (the results of the shear force test are shown in Figure S7). It can be seen from the chart that the maximum shear forces, measuring 0.63 and 0.62 MPa, respectively, occur at ICP activation times of 100 and 200 s, with a subsequent decrease in shear force after 200 s. This variation pattern corresponds to the changes in the Al2O3 surface roughness with ICP activation time. At ICP activation times of 100 and 200 s, the Al2O3 surface roughness is at its minimum, with R q values of 0.125 and 0.128 nm, respectively, resulting in a higher Al2O3–Al2O3 bonding strength compared to other activation times. The research on the variation of shear force in both Cu–Cu and Al2O3–Al2O3 bonds with ICP activation time indicates that the optimal ICP activation time for achieving the maximum bonding strength is 100 s for both cases.
Figure presents a comparative analysis of shear forces in Cu–Cu and Al2O3–Al2O3 bonded samples before and after annealing. (The shear force test results of the bonded samples after annealing are shown in Figure S8.) Observed from the figure is that the shear force of Cu–Cu bonding increases from 2.68 to 4.05 MPa, representing a 51% enhancement. For Al2O3–Al2O3 bonding, the shear force rises from 0.628 to 1.65 MPa, marking a 163% improvement. Attributable to this improvement is the fact that annealing, compared to unannealed samples, facilitates more complete hydroxyl condensation reactions at the bonding interface. Thereby, transformed are hydrogen bonds into strong covalent bonds. Additionally promoted by annealing is interatomic interdiffusion, resulting in an increased effective bonding area. It is worth noting that under the same processing conditions, the shear strength of Cu–Cu bonding is greater than that of Al2O3–Al2O3 bonding. This is due to the fact that Al2O3 has a high Young’s modulus ranging from 350 to 400 GPa, which confers strong stiffness and resistance to deformation, making it more sensitive to defects and contaminant particles. In contrast, Cu has a lower Young’s modulus of about 120 GPa, which allows for stronger atomic diffusion and results in lower sensitivity to defects and impurities.
10.

Shear force of Cu–Cu and Al2O3–Al2O3 bonding before and after annealing.
Figures (a) and (b) show the scanning acoustic microscopy (SAM) images of TCB Cu–Cu samples before and after annealing, respectively, while Figures (c) and (d) show the SAM images of TCB Al2O3–Al2O3 samples before and after annealing, respectively. By comparing the SAM images of the sample interfaces before and after annealing, it can be observed that there are large voids and delaminated regions in the central area of the unannealed sample. Attributable to this phenomenon is incomplete hydroxyl condensation reactions at the bonding interfaces following TCB processing, resulting in residual moisture retention. − After annealing, the delamination phenomenon in the central areas of the sample largely disappears, with only minor unbonded regions remaining at edges. This improvement can be attributed to more complete hydroxyl condensation reactions at annealed interfaces, where moisture is effectively eliminated and hydrogen bonds are predominantly converted into strong covalent bonds. − The hydroxyl condensation reactions at Cu–Cu and Al2O3–Al2O3 bonding interfaces are represented by reaction and reaction . In addition, annealing also promotes the diffusion of atoms at the bonding interface. This process can fill microgaps, reduce the formation of voids, and enhance the interfacial connectivity.
| 2 |
| 3 |
11.

SAM images of bonded samples: (a) unannealed Cu–Cu bonding; (b) Cu–Cu bonding after annealing; (c) unannealed Al2O3–Al2O3 bonding; (d) Al2O3–Al2O3 bonding after annealing.
Figures (a) and (b) show the cross-sectional TEM images of Cu–Cu and Al2O3–Al2O3 bonding after TCB processing without annealing. In Figure (a), a distinct interfacial boundary accompanied by sporadic voids can be observed, indicating insufficient atomic diffusion of Cu before annealing. From Figure (b), it can be observed that debonding occurs at the Al2O3–Al2O3 bonding interface, indicating insufficient formation of strong chemical bonds in the nonannealed Al2O3–Al2O3 bonding. Additionally, during TEM observation, heat generated by high-energy electron beam irradiation induces thermal deformation in the FIB-prepared sample, ultimately leading to interfacial separation.
12.

Bonding interfaces without annealing observed by TEM: (a) Cu–Cu bonding; (b) Al2O3–Al2O3 bonding.
Figure (a) presents the TEM image of the Cu–Cu bonding interface after annealing, while Figure (b) shows a high-magnification TEM view of the region marked with the red box in Figure (a). From the figure, a high-quality bonded interface can be observed, which features intimate atomic connectivity without voids or delamination. Figures (c)–(e), respectively, present the elemental distribution maps obtained by energy-dispersive X-ray spectroscopy (EDX) for Cu, Ti, and Si. These maps confirm that the bonding substrate is Si, the bonding layer is Cu, and the adhesion layer is Ti.
13.

Cu–Cu bonding interface observed by TEM: (a) low and (b) high magnification TEM images; EDX spectra of (c) Cu, (d) Ti, and (e) Si elements.
Figure (a) presents a TEM image of the Al2O3–Al2O3 bonding interface after annealing, while Figure (b) shows an enlarged TEM view of the region marked by the red box in Figure (a). These images reveal the formation of a recrystallized interlayer at the bonding interface, with intimate interfacial connectivity and no observable voids or debonding. Figures c–d display EDX mappings for Al, O, and Si, respectively, confirming a Si substrate and an Al2O3 bonding layer. Through the above-optimized process, the Cu–Cu and Al2O3–Al2O3 bonding with a well-connected interface has been demonstrated. For hybrid bonding, the chemical–mechanical polishing process needs to be employed to optimize the Cu dishing depth so that Cu–Cu and Al2O3–Al2O3 can be connected simultaneously. This paper only provides a hybrid bonding strategy with new combinations and the collaborative optimization of Cu–Cu and Al2O3–Al2O3 bonding processes. The Cu/Al2O3 hybrid bonding will be presented in future work.
14.

Al2O3–Al2O3 bonding interface observed by TEM: (a) low and (b) high magnification TEM images; EDX spectra of (c) Al, (d) O, and (e) Si elements.
4. Conclusions
This paper proposes a hybrid bonding strategy for the combination of Cu and Al2O3 materials and explores their hydrophilic bonding process. The main work is as follows: (1) optimizing plasma activation conditions for Cu–Cu and Al2O3–Al2O3 hydrophilic bonding; (2) investigating the influence of formic acid on Cu surface oxide and its compatibility with Al2O3–Al2O3 bonding; (3) analyzing the impact of thermal annealing on bonding quality; and (4) achieving tightly interfaced Cu–Cu and Al2O3–Al2O3 bonding through combined formic acid pretreatment, ICP activation, and postannealing processes.
It is concluded from the study on the surface activation process that the optimized plasma treatment condition is 100 s of Ar ICP treatment. Under this condition, the surface roughness of both Cu and Al2O3 reaches the minimum values simultaneously, with the R q values being 1.53 and 0.125 nm, respectively. Moreover, the surface contact angles also reach a relatively low and stable state. Another notable finding is that after formic acid immersion, the surface contact angles of Cu and Al2O3 decrease by 7.36° and 2.43°, respectively, compared to untreated surfaces. Sheet resistance analysis confirms that the combined formic acid + ICP activation achieves the lowest sheet resistance of 0.31 Ω/sq due to synergistic oxide removal effects. Postannealing experiments indicate that annealing significantly improves interfacial connectivity and reduces voids. The shear force of Cu–Cu bonding increases from 2.68 to 4.05 MPa after annealing, representing a 51% improvement. The shear force of Al2O3–Al2O3 bonding rises from 0.628 to 1.65 MPa after annealing, achieving a 163% improvement. High-quality Cu–Cu and Al2O3–Al2O3 bonding with void-free interfaces is achieved through the formic acid + ICP activation and postannealing process. This work provides guidance for the realization of Cu/Al2O3 hybrid bonding.
Supplementary Material
Acknowledgments
This work was supported by the National Natural Science Foundation of China under Grant 62204150, the National Key R&D Program of China under Grant 2022YFB3603502, the Natural Science Foundation of Shanghai under Grant 23ZR1422500, and the Special Funds for Promoting High-Quality Industrial Development in Shanghai (JJ-ZDHYLY-01-23-0004).
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c11163.
The test results of the surface contact angle of the samples before bonding and the shear force after bonding (PDF)
L.W. performed the experiments and wrote the manuscript. K.B. and H.C. performed the experiments. K.R.and X.D. provided the methodology and supervised the research. L.Y. and J.Z. were responsible for project administration.
The authors declare no competing financial interest.
References
- Park, J. ; Lee, B. ; Lee, H. ; Lim, D. ; Kang, J. ; Cho, C. ; Na, M. ; Jin, I. . Wafer to Wafer Hybrid Bonding for DRAM Applications. In 2022 IEEE 72nd Electronic Components and Technology Conference (ECTC); IEEE, 2022, 126–129. [Google Scholar]
- Gao, G. ; Mirkarimi, L. ; Fountain, G. ; Wang, L. ; Uzoh, C. ; Workman, T. ; Guevara, G. ; Mandalapu, C. ; Lee, B. ; Katkar, R. . Scaling Package Interconnects Below 20m Pitch with Hybrid Bonding. In 2018 IEEE 68th Electronic Components and Technology Conference (ECTC); IEEE, 2018. [Google Scholar]
- Kagawa1, Y. ; Fujii, N. ; Aoyagi, K. ; Kobayashi, Y. ; Nishi, S. ; Todaka, N. ; Takeshita, S. ; Taura, J. ; Takahashi, H. ; Nishimura, Y. ;et al. Novel stacked CMOS image sensor with advanced Cu2Cu hybrid bonding. In 2016 IEEE International Electron Devices Meeting (IEDM); IEEE, 2016, pp.8.4.1–8.4.4. [Google Scholar]
- Lhostis, S. ; Farcy, A. ; Deloffre, E. ; Lorut, F. ; Mermoz, S. ; Henrion, Y. ; Berthier, L. ; Bailly, F. ; Scevola, D. ; Guyader, F. , et al. Reliable 300 mm Wafer Level Hybrid Bonding for 3D Stacked CMOS Image Sensors. In 2016 IEEE 66th Electronic Components and Technology Conference (ECTC); IEEE, 2016, pp. 869–876. [Google Scholar]
- Reddy J., Pratap R.. Si-gold-glass hybrid wafer bond for 3D-MEMS and wafer level packaging. J. Micromech. Microeng. 2017;27:015005. doi: 10.1088/0960-1317/27/1/015005. [DOI] [Google Scholar]
- Fujino M., Takahashi K., Araga Y., Kikuchi K.. 300 mm wafer-level hybrid bonding for cu/interlayer dielectric bonding in vacuum. Jpn. J. Appl. Phys. 2020;59:SBBA02. doi: 10.7567/1347-4065/ab4b2b. [DOI] [Google Scholar]
- Hu Z. J., Qu X. P., Lin H., Huang R. D., Ge X. C., Li M.. Cu cmp process development and characterization of cu dishing with 1.8 μ m cu pad and 3.6 μ m pitch in cu/SiO2 hybrid bonding. Jpn. J. Appl. Phys. 2019;58:SHHC01. doi: 10.7567/1347-4065/ab17c4. [DOI] [Google Scholar]
- Mudrick, J. P. ; Sierra-Suarez, J. A. ; Jordan, M. B. ; Friedmann, T. A. ; Henry, M. D. . Sub-10m Pitch Hybrid Direct Bond Interconnect Development for Die-to-Die Hybridization, In IEEE 69th Electronic Components and Technology Conference (ECTC); IEEE, 2019, pp.648–654. [Google Scholar]
- Ji L., Che F. X., Ji H. M., Li H. Y., Kawano M.. Wafer-to-wafer hybrid bonding development by advanced finite element modeling for 3-d IC packages. IEEE Trans. Compon. Packag. Manuf. Technol. 2020;10:2106–2117. doi: 10.1109/TCPMT.2020.3035652. [DOI] [Google Scholar]
- Fan J., Tu L. C., Tan C. S.. High-κ Al2O3 material in low temperature wafer-level bonding for 3d integration application. AIP Adv. 2014;4:031311. doi: 10.1063/1.4867089. [DOI] [Google Scholar]
- Li Y., Wang S., Sun B., Chang H., Zhao W., Zhang X., Liu H.. Room temperature wafer bonding by surface activated ALD- Al2O3. ECS Trans. 2013;50:303–311. doi: 10.1149/05007.0303ecst. [DOI] [Google Scholar]
- Ma C., Nikiforov A., Hegemann D., Geyter N., Morent R., Ostrikov K.. Plasma-controlled surface wettability: Recent advances and future applications. Int. Mater. Rev. 2023;68:82–119. doi: 10.1080/09506608.2022.2047420. [DOI] [Google Scholar]
- Xiao X., Wu J., Cao Q., Huang X.. A novel capacitively coupled plasma driven by hollow cathode radio-frequency discharges. Phys. Scr. 2023;98:035605. doi: 10.1088/1402-4896/acba58. [DOI] [Google Scholar]
- Moon H., Jung J., Park J., Chin-Wook C.. Generation of high-density plasma via transparent electrode in capacitively coupled plasma. Plasma Sources Sci. Technol. 2024;33(8):085004. doi: 10.1088/1361-6595/ad678f. [DOI] [Google Scholar]
- Shigetou A., Itoh T., Matsuo M., Hayasaka N., Okumura K., Suga T.. Bumpless interconnect through ultrafine Cu electrodes by means of surface-activated bonding (SAB) method. IEEE Trans. Adv. Packag. 2006;29:218–226. doi: 10.1109/TADVP.2006.873138. [DOI] [Google Scholar]
- Takigawa R., Higurashi E., Suga T., Sawada R.. Room-temperature bonding of vertical-cavity surface-emitting laser chips on Si substrates using Au micro bumps in ambient air. Appl. Phys. 2008;1:112201. doi: 10.1143/APEX.1.112201. [DOI] [Google Scholar]
- Matsumae T., Kurashima Y., Takagi H.. Surface activated bonding of Ti/Au and Ti/Pt/Au films after vacuum annealing for MEMS packaging. Microelectron. Eng. 2018;197:76–82. doi: 10.1016/j.mee.2018.05.008. [DOI] [Google Scholar]
- Utsumi J., Takigawa R.. Crystallographic structure of aluminum oxide bonding interfaces prepared via room-temperature surface-activated bonding. Appl. Surf. Sci. 2025;694:162825. doi: 10.1016/j.apsusc.2025.162825. [DOI] [Google Scholar]
- Fan J., Lim D., Tan C.. Effects of surface treatment on the bonding quality of wafer-level Cu-to-Cu thermo-compression bonding for 3D integration. J. Micromech. Microeng. 2013;23:045025. doi: 10.1088/0960-1317/23/4/045025. [DOI] [Google Scholar]
- Koyama S., Hagiwara N., Shohji I.. Cu/Cu direct bonding by metal salt generation bonding technique with organic acid and persistence of reformed layer. Jap J. Appl.Phys. 2015;54:030216. doi: 10.7567/JJAP.54.030216. [DOI] [Google Scholar]
- Parka M., Baek S., Kimc S., Kim S. E.. Argon plasma treatment on Cu surface for Cu bonding in 3D integration and their characteristics. Appl. Surf. Sci. 2015;324:168–173. doi: 10.1016/j.apsusc.2014.10.098. [DOI] [Google Scholar]
- Hu, L. ; Goh, S. C. K. ; Lim, Y. D. ; Zhao, P. ; Lim, M. J. Z. ; Tan, C. S. . Multi-Die to Wafer Bonding Through Plasma-Activated Cu-Cu Direct Bonding in Ambient Conditions. In 2021 IEEE International 3D Systems Integration Conference (3DIC); IEEE, 2021, pp.1–5. [Google Scholar]
- Kim T., Howlader M., Itoh T., Suga T.. Room temperature Cu-Cu direct bonding using surface activated bonding method. J. Vac. Sci. Technol. 2003;21:449–453. doi: 10.1116/1.1537716. [DOI] [Google Scholar]
- Yang W., Akaike M., Suga T.. Effect of Formic Acid Vapor In Situ Treatment Process on Cu Low-Temperature Bonding. IEEE Trans. Compon. Packag. Manuf. Technol. 2014;4:951–956. doi: 10.1109/TCPMT.2014.2315761. [DOI] [Google Scholar]
- Jangam, S. ; Bajwa, A. ; Mogera, U. ; Ambhore, P. ; Colosimo, T. ; Chylak, B. . Fine-Pitch (≤10 μm) Direct Cu-Cu Interconnects Using In-Situ Formic Acid Vapor Treatment. In 2019 IEEE 69th Electronic Components and Technology Conference (ECTC); IEEE, 2019, pp.620–627. [Google Scholar]
- Kang Q., Wang C., Zhou S., Li G., Lu T., Tian Y., He P.. Low-Temperature Co-hydroxylated Cu/SiO2 Hybrid Bonding Strategy for a Memory-Centric Chip Architecture. ACS Appl. Mater. Interfaces. 2021;13:38866–38876. doi: 10.1021/acsami.1c09796. [DOI] [PubMed] [Google Scholar]
- Pasquariello D., Hedlund C., Hjort K.. Oxidation and Induced Damage in Oxygen Plasma In Situ Wafer Bonding. J. Electrochem. Soc. 2000;147:2699. doi: 10.1149/1.1393592. [DOI] [Google Scholar]
- Sun Q., Wang J., Wang X., Zhou X., Tang X., Akira K.. Cu aggregation behavior on interfacial reaction of Sn-3.0Ag-0.5Cu/ENIG solder joints. Mater. Lett. 2023;348:134659. doi: 10.1016/j.matlet.2023.134659. [DOI] [Google Scholar]
- Arman A., Talu S., Luna C., Ahmadpourian A., Naseri M., Molamohammadi M.. Micromorphology characterization of copper thin films by AFM and fractal analysis. J. Mater. Sci.: mater. Electron. 2015;26:9630–9639. doi: 10.1007/s10854-015-3628-5. [DOI] [Google Scholar]
- Litton D., Garofalini S.. Modeling of hydrophilic wafer bonding by molecular dynamics simulations. Appl. Phys. 2001;89:6013–6023. doi: 10.1063/1.1351538. [DOI] [Google Scholar]
- Navarro E., Bréchet Y., Barthelemy A., Radu I., Raskin J., Pardoen T.. Adhesion and separation models for direct hydrophilic bonding. J. Appl. Phys. 2015;117:085305. doi: 10.1063/1.4913481. [DOI] [Google Scholar]
- Cheng K.-Y., Wang W.-F., Wu M.-C., Hsieh K.-C.. Hydrophilic direct bonding and debonding of ultrathin glass substrates for high-temperature devices. Jpn. J. Appl. Phys. 2020;59(7):070901. doi: 10.35848/1347-4065/ab98d7. [DOI] [Google Scholar]
- Li D., Cui X., Du M., Zhou Y., Lan F.. Effect of Combined Hydrophilic Activation on Interface Characteristics of Si/Si Wafer Direct Bonding. Processes. 2021;9:1599. doi: 10.3390/pr9091599. [DOI] [Google Scholar]
- Takeuchi K., Ninomiya T., Kubota M., Kawano M., Takagi T., Masaaki N., Kuroda T.. et al. Hydrophilic Bonding of SiO2/SiO2 and Cu/Cu using Sequential Plasma Activation. ECS Trans. 2023;112(3):95–101. doi: 10.1149/11203.0095ecst. [DOI] [Google Scholar]
- Yang W., Lu Y., Zhou C., Zhang J., Suga T.. Study of Cu Film Surface Treatment Using Formic Acid Vapor/Solution for Low Temperature Bonding. J. Electrochem. Soc. 2018;165:H3080–H3084. doi: 10.1149/2.0121804jes. [DOI] [Google Scholar]
- Yang, W. ; Shintani, H. ; Akaike, M. ; Suga, T. . Low temperature Cu-Cu direct bonding using formic acid vapor pretreatment. In 2011 12th International Conference on Electronic Packaging Technology and High-Density Packaging (ICEPT-HDP); IEEE, 2011, pp.72–75. [Google Scholar]
- He R., Fujino M., Yamauchi A., Suga T.. Combined Surface Activated Bonding Technique for Hydrophilic SiO2-SiO2 and Cu-Cu Bonding. ECS Trans. 2016;75:117. doi: 10.1149/07509.0117ecst. [DOI] [Google Scholar]
- Kim Y., Nguyen T., Choa S.. Enhancement of the Bond Strength and Reduction of Wafer Edge Voids in Hybrid Bonding. Micromachines. 2022;13:537. doi: 10.3390/mi13040537. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sakuma, K. ; Yu, R. ; Polomoff, N. ; Darling, L. ; Chowdhury, P. ; Raghavan, S. . Voids-free Die-level Cu/ILD Hybrid bonding. In 2023 IEEE 73rd Electronic Components and Technology Conference (ECTC); IEEE, 2023, pp.800–805. [Google Scholar]
- He R., Yamauchi A., Suga T.. Sequential plasma activation methods for hydrophilic direct bonding at sub-200 °C. Jpn. J. Appl. Phys. 2018;57:02BD03. doi: 10.7567/JJAP.57.02BD03. [DOI] [Google Scholar]
- Hooge, J. ; Jois, C. ; Netzband, C. ; Son, I. . Modeling of Copper Hybrid Bonding Anneal. In 2024 IEEE 74th Electronic Components and Technology Conference (ECTC); IEEE, 2024, pp.1231–1235. [Google Scholar]
- Alford T., Tang T., Thompson D., Bhagat S., Mayer J.. Influence of microwave annealing on direct bonded silicon wafers. Thin Solid Films. 2008;516:2158–2161. doi: 10.1016/j.tsf.2007.06.118. [DOI] [Google Scholar]
- Shigekawa N., Kozono R., Yoon S., Hara T., Liang J., Yasui A.. Effects of post bonding annealing on GaAs//Si bonding interfaces and its application for sacrificial-layer-etching based multijunction solar cells. Sol. Energy Mater. Sol. Cells. 2020;210:110501. doi: 10.1016/j.solmat.2020.110501. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.







