Low temperature Cu–Cu direct bonding in air ambient by ultrafast surface grain growth

Fine-grain copper (Cu) films (grain size: 100.36 nm) with a near-atomic-scale surface (0.39 nm) were electroplated. Without advanced post-surface treatment, Cu–Cu direct bonding can be achieved with present-day fine-grain Cu films at 130℃ in air ambient with a minimum pressure of 1 MPa. The instantaneous growth rate on the first day is 164.29 nm d−1. Also, the average growth rate (∆R/∆t) is evaluated by the present experimental results: (i) 218.185 nm d−1 for the first-day period and (ii) 105.58 nm d−1 during the first 14-day period. Ultrafast grain growth and near-atomic-scale surface facilitate grain boundary motion across the bonding interface, which is the key to achieve Cu–Cu direct bonding at 130℃ in air ambient.

bonding below 200℃, a relatively high bonding pressure has to be used such as 64-162 MPa [12][13][14].The high external bonding pressure is the key to eliminate the voids resulting from the rough Cu surface at the bonding interface.However, the high external bonding pressure causes potential damage to integrated circuit chips.Hence, a combination of a low-roughness Cu surface (typically treated with chemical-mechanical polishing (CMP)) and controlled bonding ambient (vacuum, inert gas, or active acid bonding ambient) has been developed for Cu-Cu direct bonding below 200℃ with a relatively small bonding pressure [9][10][11][12][13][14][15].The controlled bonding ambient is very challenging for Cu-Cu direct bonding in current semiconductor processes.Also, CMP is a very costly process, and the CMP process would bring in various impurities such as Si, SiC and Cl ions between the Cu-Cu bonding interface [16].Thus, an untreated and low-roughness Cu surface for the Cu-Cu direct bonding process in air ambient would have a great advantage over the current reported Cu-Cu direct bonding processes.In this paper, a fine-grain Cu film was developed, which has a very low Cu surface roughness.With the near-atomic-scale Cu surface (without CMP surface treatment), Cu-Cu direct bonding can be done at low temperatures (≤200℃) in air atmosphere with a minimum bonding pressure of 1 MPa.
A Cu (500 nm)/Ti (200 nm) bi-layer was sputtered on Si substrates (1.5 cm × 1.5 cm), where Ti serves as the adhesion layer and Cu is the seed layer for the electroplating.Before the Cu plating process, the surface of the Cu seed layer was cleaned by the following two pretreatment processes.First, Cu (500 nm)/Ti (200 nm)/Si substrates were immersed in a micro-etching solution containing 0.36 M H 2 SO 4 and 6.25 M sodium persulfate for 30 s.This step would micro-etch the Cu surface and remove possible oxides from the Cu surface.After the micro-etching process, the surface-cleaned Cu films (Cu (500 nm)/Ti (200 nm)/Si substrates) were immersed and stored in dilute acid solution with 0.09 M H 2 SO 4 .The purpose is to preserve the cleaned surface before Cu electroplating.After the surface cleaning processes, 5-μm-thick fine-grain Cu films were electroplated on the Ti (200 nm)/Cu (500 nm)/Si substrates.The Cu plating parameter and process flow are described as follows.A yttrium-oxide-coated Ti mesh is used as the anode.The current density and the plating time are 2 A dm −2 and 12 min, respectively.The Cu plating bath is composed of 1.5 M CuSO 4 , 0.18 M H 2 SO 4 and 60 ppm chloride ion.To produce the fine-grain Cu film, two special additives were added to the Cu-plating solution, which are 300-500 ppm polypropylene glycol (PPG) and 7-9 ppm 3-methyl-2cyclopentenone-4-ol.PPG additive functions as an inhibitor for reducing the Cu depositing rate and helping to form a fine-grain microstructure.3-Methyl-2-cyclopentenone-4-ol is used as a leveller, which typically contains nitrogen heterocycles, sulfur heterocycles and thione groups.Levellers can suppress Cu deposition in certain irregular areas and reduce the surface roughness.
The as-plated Cu film was cross-sectioned using a focused ion beam (FIB, FEI Versa 3D) as shown in figure 1.The grain size was determined using the intercept procedure method (ASTM E112-13).In the FIB image of the fine-grain Cu film, 100 horizontal lines and 100 vertical lines were drawn, and the total number of Cu grains was counted along these lines.By taking the reciprocal of the total number of Cu grains per unit length, the average Cu grain size was determined to be 100.36nm.The roughness (particularly below 5 nm) can be accurately obtained by integrating the shape function of the surface topology [17].The shape function is a function of the roughness coefficient and Fresnel's coefficient at the intended measured air-surface interface [17].
In the present work, X-ray reflectivity (XRR) measurement was carried out on the surface of the fine-grain Cu film, which provides the reflection intensity and reflection coefficient at the air-Cu interface against the incident angle (2θ) of X-rays.With the reflection intensity versus 2θ relationship, as shown in figure 2, Fresnel's coefficient at the air-Cu interface can be obtained through recursive algorithms [18][19][20][21][22]. Also, using Sinha's approach, the roughness coefficient can be calculated by the reflection intensity and reflection coefficient obtained by XRR measurement versus the incident angle [18,[20][21][22][23].Note that the R 2 value is 0.99.Knowing the roughness coefficient and Fresnel's coefficient at the intended measured air-surface interface, the shape function can be obtained.Thus, the roughness (R a ) of the as-electroplated fine-grain Cu films can be calculated by integrating the shape function, which is 0.942 ± 0.08 nm.Yet, the roughness obtained by XRR measurement for the Cu surface can be inaccurate due to noisy signals in modelling experimental data.Thus, atomic force microscopy (AFM) analysis has been done on the as-plated Cu surface in a 1 × 1 μm 2 area.The surface roughness obtained from the AFM measurements is 0.39 nm, which has been included in table 2.
After the Cu plating process, the as-plated fine-grain Cu/Ti/Si substrates (1.5 cm × 1.5 cm) were cleaned immediately by dipping in a 0.5 wt% H 2 SO 4 solution, rinsing with deionized water and blow-drying with N 2 gas.Two surface-cleaned fine-grain Cu/Ti/Si samples were placed in a graphite bonding device to perform the Cu-Cu direct bonding process.A pressure of 1 MPa was applied to the Cu-Cu bonding samples.The bonding temperatures are set at 450, 350, 300, 200, 150 and 130℃ for 30 min.The bonding atmosphere is air ambient.
After the Cu-Cu direct bonding process, the Cu-Cu bonding samples were cross-sectioned and examined by FIB, as shown in figure 3. Remarkably, as depicted in figure 3f, Cu-Cu direct bonding can occur at temperatures as low as 130℃ in air ambient.No observable micro-voids are present at the bonding interface, as indicated by the white arrows in figure 3f, demonstrating a continuous bonding interface.The slightly distorted bonding interface in the bottom image is due to chipping and missing of the left-corner Si during the polishing process.For the higher bonding temperatures at 450, 350 and 300℃, as seen in figure 3a-c, the Cu grain boundary migration across the bonding interface can clearly be observed.It indicates that much faster grain growth would occur at the initial bonding interface at higher bonding temperatures.Electron backscatter diffraction (EBSD) analysis has been done on the cross-sectioned Cu-Cu bonding interfaces.The EBSD scanning images are shown in figure 4a,b.EBSD images can reveal the bonding quality, which depends on the diffusion across the bonding interface.As seen in figure 4a, the bonding interface for the 130℃ sample is relatively straight.Compared with the bonding interface at 130℃, we can observe that the curvature shows the bonding interface at 200℃ (figure 4d), as indicated by a dashed black ellipse.From the EBSD images, we can also observe that the grain size increases with the bonding temperature and that an annealing twin boundary forms.We also performed shear tests with a shear test tool  .And, the resistivity (ρ) can be calculated by the expression 4.532(R s )(L), where L is the measurement length (about 1 mm) between two electrical probes.Table 1 shows the slope of the I-V curve and the calculated resistivity (ρ) of the Cu-Cu bonding samples under three low bonding temperatures (130, 150 and 200℃).The measured resistivity of Cu-Cu bonding samples with different bonding temperatures is of the same order as the resistivity of pure Cu (1.7 × 10 -8 Ω m).Cu-Cu direct bonding has been demonstrated by numerous studies.All Cu-Cu direct bonding done below 200℃ is summarized in table 2, which is categorized with four key parameters, i.e. bonding temperature, pressure, ambient and surface roughness.The majority of Cu-Cu direct bonding below 200℃ has to be done with CMP surface treatment [3,12,15,24].The typical surface roughness of the CMP Cu surface is between 1.2 and 6.5 nm [12,15,17,24].Most Cu-Cu direct bonding without CMP surface treatment has to be done at a bonding temperature of over 200℃.One exception is that the Cu-Cu direct bonding can be achieved below 200℃; however, it requires a vacuum bonding ambient  [10].The present work significantly demonstrates that Cu-Cu direct bonding can be achieved at 130℃ in air ambient.Figure 5a-d shows the FIB cross-sectional images on the fine-grain Cu films stored at room temperature for 1, 5, 7 and 14 days.The self-annealing in the fine-grain Cu films can be clearly observed.With the intercept procedure method, the average grain size of the fine-grain Cu films stored at room temperature for 1, 5, 7 and 14 days was calculated to be 537.73,787.42, 1265.89 and 1578.52 nm, respectively.
The average grain size of the fine-grain Cu films is plotted against the storage time in figure 6a.In figure 6, results of previous research on the self-annealing behaviour of Cu films also included [25][26][27].The curves in figure 6b were fitted to the grain size data in figure 6a with the Ostwald ripening model, which is typically written as equation (1.1): where C 1 and C 2 are constants, γ is the grain boundary energy, κ is the grain curvature and R is the average radius of Cu grains.The product of γ and κ is the net driving force for Cu grain growth.Assuming that the average radius of Cu grains (R) is an exponential function of time (t), which can be expressed as C 3 (t + a) n , where a and C 3 are constants, then equation (1.1) can be rewritten as equation (1.2), where b is the integrated constant: By integrating both sides of equation (1.2) with t, equations (1.3) and (1.4) can be derived under two conditions for the value of n: .
By fitting the data curves in figure 6a with equations (1.3) and (1.4), the constants of a and b in those equations can be obtained.Also, we found that the R 2 value (with regression analysis) is over 0.95, as n equals 1.Thus, the constants of a and b fitted by equation (1.3) were chosen for equation (1.2).The curves of grain growth can be plotted, as shown in figure 6.With the curves in figure 6c, the instantaneous growth rate can be defined at any particular self-annealing time.Thus, we can    Cu produced through severe plastic deformation has been reported [28].The ultrafast diffusion was proposed to be the main mechanism for the grain growth model of fine-grain Cu [29].
In conclusion, the current work demonstrates that without any advanced post-electroplating surface treatment, Cu-Cu direct bonding can be achieved at 130℃ in air ambient with a minimum bonding pressure of 1 MPa.The bonding interface contains no micro-voids.The achievement of the present Cu-Cu direct bonding is attributed to two key features of the Cu electroplated films, which are the near-atomic-scale Cu surface and ultrafast self-annealing grain growth.The low Cu surface roughness eliminates potential micro-voids forming at the bonding interface.And then, the ultrafast grain growth facilitates the motion of the grain boundary across the bonding interface, which should be the key to achieve Cu-Cu direct bonding at the low bonding temperature of 130℃ in air ambient.By the Ostwald ripening model, the instantaneous growth rate (dR/dt) of the present fine-grain Cu films is calculated to be 164.29 nm d −1 on the first day.

3
royalsocietypublishing.org/journal/rsos R. Soc.Open Sci.11: 240459 samples (130, 150 and 200℃) was measured with a probe station and Keithley 2400 measuring unit.Two electrical probes were placed on the centre point of two Cu-bonded layers and separated by about 1 mm.I-V curves of all three bonding samples (130, 150 and 200℃) demonstrate ohmic behaviour, i.e. linear curves in I-V plots.The slope of the regression lines in the measured I-V plot represents the reciprocal of the sheet resistance (R s )

Figure 6 .
Figure 6.Grain size and growth rate of fine-grain Cu films with different self-annealing times.(a) Grain size.(b) Fitted curves of grain size.(c) Fitted curves of grain growth rate.

Table 1 .
The slope of I-V curve, and the calculated resistivity (ρ) of the Cu-Cu bonding samples under three low bonding temperatures(130, 150 and 200℃).theinstantaneousgrowth rate on the first day as 164.29 nm d −1 .Also, the average growth rate (∆R/∆t) is evaluated by figure 5a and found to be as follows: (i) 218.185 nm d −1 for the first-day period and (ii) 105.58 nm d −1 during the first 14-day period.Comparing the present work with previous results, as shown in figure6c, an ultrafast self-annealing grain growth behaviour is found in the present fine-grain Cu films.Hence, we believe that the ultrafast grain growth facilitates the motion estimate of the grain boundary across the bonding interface, which should be the key to achieve Cu-Cu direct bonding at the low bonding temperature of 130℃ in air ambient.A similar growth model of fine-grain

Table 2 .
Summary of present and previous Cu-Cu direct bonding results and parameters.