A Reliable Way to Improve Electrochemical Migration (ECM) Resistance of Nanosilver Paste as a Bonding Material

Abstract

Electrochemical migration (ECM) of sintered nano-Ag could be a serious reliability concern for power devices with high-density packaging. An anti-ECM nano-Ag-SiO_x paste was proposed by doping 0.1wt% SiO_x nanoparticles rather than previously used expensive noble metals, e.g., palladium. The ECM lifetime of the sintered nano-Ag-SiO_x was 1.5 to 3 times longer than that of the sintered nano-Ag, due to the fact that the SiO_x could protect the Ag from oxidation. The thermo-mechanical reliability of the sintered nano-Ag-SiO_x was also improved by sintering under 5 MPa assisted pressure. The lesser porosity and smaller grain boundaries of the sintered nano-Ag-SiO_x could also be beneficial to retard the silver ECM. In the end, a double-sided semiconductor device was demonstrated to validate the better resistance to the ECM using the sintered nano-Ag-SiO_x.

1. Introduction

Die-attach materials are crucial to power devices. It bonds the chip to the substrate and performs the functions of conducting current and heat. Traditional lead-free and Sn-Pb solders harm the environment and have lower operating temperatures, thus nano-Ag paste as a die-attach material is a feasible option with excellent electrical and thermal conductivity [1,2,3].

However, silver is prone to electrochemical migration (ECM) and forms dendritic crystals, which result in short-circuit failure [4,5]. Generally, moisture is believed to play a crucial role in the migration of silver, and as a result, silver migration is described as a moisture migration phenomenon. However, recent studies have found that ECM failure of silver occurs even in high temperature and dry environments. The role of O₂ in dry environments is similar to that of H₂O in the classical moisture migration process [6]. Firstly, silver oxidized at the anode and form the intermediate species Ag₂O, which dissociated into silver cations and oxygen anions at a high temperature above 250 °C. Driven by the bias voltage, the silver cations move from anode to cathode through the insulating gap, while the oxygen anions move in the opposite direction to maintain charge neutrality. The continuous depletion of silver cations at the anode drives oxidation and dissociation until silver dendrites formed between the two electrodes. Subsequently, silver dendrites gradually appeared and result in short circuits [7,8].

It was concerned that the ECM of the sintered nano-Ag should be a failure risk for packaging wide band-gap (WBG) power semiconductors devices, which can theoretically be operated at extreme temperatures of 500 °C, not to mention the great increase of electrical bias [9]. Furthermore, the previous ECM of silver was only focused on thick film conductors in conventional printed circuit boards (PCBs) [4,10]. Double-sided cooling (DSC) silicon carbide (SiC) power devices have attracted more and more attention in industries, especially in electric vehicle applications, because of their superior heat dissipation capability and power density [11]. However, the DSC SiC devices have a very small electrode spacing, e.g., 250 μm [12]. When the great sintered nano-Ag was used as die attachment in the DSC SiC devices, it is crucial to avoid the ECM failure of sintered nano-Ag to guarantee the reliability of electric vehicles. Unfortunately, there is a current lack of research on the ECM of silver suppression for the DSC SiC power devices.

Inhibition of the ECM of silver, such as alloying silver with indium or copper to form anti-ECM Ag-In and Ag-Cu alloys, had been studied [13,14]. Recently, we proposed the method of alloying silver with palladium aimed at inhibiting the formation of silver ions through oxidation and decomposition in the anode to delay the ECM of silver [15]. However, these alloying methods need very high processing temperatures in order to improve the resistance to the ECM of silver by alloying more at the higher temperature, not to mention the extremely high cost of indium, palladium, and so on.

In this paper, a nano-Ag composite paste was proposed with the 0.1wt% SiO_x nanoparticles (NPs) as fillers. The conventional nano-Ag paste was compared as a reference. The lifetime for reaching the ECM failure was characterized in the first place. Then the thermo-mechanical reliability of the proposed nano-Ag-SiO_x paste was verified in the aspects of die-shearing strength and thermal impedance by thermal shocking tests because it is critical to ensure promising thermo-mechanical reliability besides the improvement of the resistance to the ECM. Microstructures and porosity were analyzed to clarify the effects of SiO_x doping on the ECM and the thermo-mechanical reliability. In the end, the proposed nano-Ag-SiO_x paste was used as a die attachment for a DSC power device. The improvement of the lifetime of the ECM was validated in the high power density demonstration.

2. Materials and Methods

2.1. Materials

The nano-Ag composite paste was made by ultrasonically mixing the 30 nm Ag NPs and the 30 nm SiO_x NPs with a specific weight ratio such as 0.1 wt%, accompanied by organic vehicles that included surfactants, binders, and thinners [16]. Figure 1a shows that the surfactant, binders, and thinners are dissolved by acetone in a beaker as a mixture. As seen from Figure 1b, in order to disperse the SiO_x NPs and the Ag NPs uniformly in the mixture, the mixture was stirred at 2000 rpm for 10 min using a planetary centrifugal mixer (THINKY ARE310).

2.2. Methods of Reliability Verification

Firstly, the capability of the sintered nano-Ag or nano-Ag-SiO_x against the ECM was verified in this work. The ECM samples were formed by stencil-printed the proposed anti-ECM nano-Ag composite paste on an alumina ceramic substrate (30 × 30 mm²) to form pairs of electrodes with identical gaps of 0.5 mm for ensuring uniform electrical bias distributions (Figure 1c). The as-printed ECM samples were then heated to 280 °C at a ramp rate of 5 °C/min and sintered at 280 °C for 30 min. These processes are shown in Figure 1d. The leakage currents between the electrodes are monitored by a Pico ammeter (RIGOL DM3068). The time when the leakage current first reaches 1 mA was defined as the lifetime of the ECM. The environment temperature during the ECM test was 25 ± 2 °C and the environment humidity during the ECM test was 50%.

Then thermal shocking tests were carried out according to the standard of JESD22-A104C to verify the reliability of the nano-Ag composite paste. The temperature swung from −40 °C to +125 °C and the soaking time at the extreme temperatures was 10 min. Each period was about 25 min. The degradation of shear strength and thermal resistance of the ECM samples was recorded every 100 cycles until 1000 cycles. The shear strength was measured by a die-shearing tester (XYZTEC, CONDOR 150) and the thermal resistance was recorded by a thermal resistance measurement system that used V_ge as temperature sensitive parameter [17,18].

For the thermal resistance measurement, a 3.92 × 3.88 × 0.07 mm³ Insulated Gate-Bipolar-Transistor (IGBT) (INFINEON, IGC15T65QE) was sinter-bonded on a 30 × 30 × 1 mm³ silver-plated direct-bond-copper (DBC) substrate under the same process as that of the ECM samples.

For the die-shearing tests, a 3 × 3 × 0.25 mm³ dummy silicon chip with bottom metallization layer of Al-Ti-Ni-Ag (1.2 μm) was sinter-bonded on a 30 × 30 × 1 mm³ silver-plated DBC substrate using the same process as the ECM samples, under an additional pressure of 5 MPa.

In the end, a DSC power device was demonstrated to verify the anti-ECM of the proposed nano-Ag composite paste. The fabrication process of the DSC demonstration is shown in Figure 2. Firstly, the top side of a silicon carbide (SiC) dummy chip of 4 × 4 × 0.17 mm³ was metalized with Ti-Ni-Ag using magnetron sputtering technology. Then the proposed nano-Ag composite paste was printed onto two silver-plated DBC substrates of 20 × 20 × 1 mm³ and the top Ti-Ni-Ag metallization of the SiC chip separately. Then the chip, the silver-plated copper block buffer of 2 × 2 × 2.2 mm³, and the silver-plated power terminals were assembled with the DBC substrates. The as-assembled DBC substrates were sintered at 280 °C and 5 MPa as the process of the previous samples for the die-shearing tests.

During the ECM test for the demonstrated DSC power devices, a direct current (DC) power supplier (Topower TN-XX702) was connected with the power terminals of the device. A controllable heating plate was used to provide the ambient temperature required for the ECM test. The leakage currents in the device were monitored by the Pico ammeter (RIGOL DM3068).

3. Results and Discussion

The ECM experimental lifespan findings of the suggested nano-Ag composite paste and the control samples of conventional nano-Ag paste are listed in

Table 1. ECM testing conditions and lifetime comparisons.

Temperature (K)	Electrode Spacing (mm)	Voltage (V)	Lifetime (min)
			Sintered Nano-Ag	Sintered Nano-Ag-SiOx
673	0.5	200	332	512
673	0.5	260	294	477
673	0.5	300	236	412
773	0.5	200	134	341
773	0.5	260	81	244
773	0.5	300	74	208

. According to Table 1, the nano-Ag-0.1%SiO_x paste has a substantially longer failure lifetime than the nano-Ag paste under identical conditions. The 0.1wt% SiO_x doping in the original nano-Ag paste serves a critical function in preventing the silver from migrating electrochemically. Furthermore, the lifetime reduces greatly when the temperature and the electric field are increased.

It was believed that the ECM behavior of the sintered nano-Ag at high temperatures initiates from the oxidation of the sintered anode and the decomposition of Ag₂O [6]. When the temperature is constant, the higher the electrode voltage applied, the faster the Ag⁺ formed after anodic oxidation. These Ag ions would reduce at the negative electrode as metallic Ag under the action of the electric field, resulting in the accelerated growth of conductive silver dendrites from the cathode and extended to the anode. As a result, the leakage current can be detected between the anode and the cathode. When the voltage remains constant, the higher the ambient temperature, the faster the ions travel. Then the produced Ag⁺ speeds up the creation of silver dendrites under the specific electric field. Therefore, the ECM lifetime of the sintered nano-Ag-SiO_x in the temperature range of 573–673 K is 1.5 to 3 times that of the sintered nano-Ag. It was likely that the SiO_x covered on the silver, as shown in Figure 3a,b, should be susceptible to oxidation. The oxidation electrode potential of SiO_x oxidation to SiO₂, i.e., ~0.06 V_SCE, was reported as more than one time lower than that of Ag oxidation to Ag₂O, i.e., 0.13 V_SCE [7,19]. As a result, silver oxidation was reduced and the ECM was delayed.

Figure 3c shows the change in shear strength following temperature cycling. The shear strength falls as the number of cycles increases. It should be noted that the shear strength of the Ag-SiO_x paste drops to less than 30 MPa after 600 cycles, but the shear strength of the nano-Ag control samples remains ~30 MPa after 1000 cycles. It was likely that the atomic diffusion rate of Ag into Ag, i.e., ~1.455 × 10⁻²³ m²/s [20], is much higher than that of Ag into SiO_x, i.e., ~8.76 × 10⁻²⁹ m²/s [21], resulting in an insufficient driving force for the sinter-bonding of Ag-SiO_x heterogeneous particles under the same pressure-free and sintering temperature conditions [22]. The Young’s modulus of the SiO_x nanoparticles, on the other hand, is much higher than that of the Ag nanoparticles, and the thermal expansion coefficient is much lower than that of the sintered nano-Ag, so the interfacial bonds of the sintered Ag-SiO_x were subjected to thermo-mechanical stress fatigue during temperature cycling, eventually leading to its shear strength dropping faster in the case of pressureless sintering [23].

It was known the driving force for the sintering can be increased greatly if assisted pressure was applied [24]. An extra 5 MPa pressure was applied to promote the sintering of the Ag-SiO_x paste for better long-term reliability. Fortunately, the shear strength of the pressure-assisted sintered Ag-SiO_x can be kept at ~35 MPa even after 1000 cycles. It was even higher than the shear strength of the nano-Ag control samples after 1000 cycles. It was believed that the pressure can ensure the better thermo-mechanical reliability of the nano-Ag-SiO_x paste.

Figure 3c also shows the change in thermal resistance after the temperature cycling. The thermal resistance grows steadily as the number of cycles increases. The thermal resistance of the nano-Ag paste improves by ~9% after 1000 cycles, while the thermal resistance of the nano-Ag-SiO_x paste sintered with 5 MPa pressure increases by ~12%. It was consistent with the variation of the die shear strength. Fortunately, the increase in thermal resistance of the nano-Ag-SiO_x pastes is still less than the common industrial failure criterion in power electronics, which is usually required as less than 20% of the thermal resistance after 1000 cycles of the temperature cycling.

Figure 4 shows the microstructures and porosity of sintered Ag and sintered nano-Ag-SiO_x before and after the thermal shocking aging. The porosity of the sintered nano-Ag-SiO_x before and after the thermal shocking aging is always lower than that of the sintered nano-Ag. It was believed that the nucleation of metallic Ag occurs preferentially adjacent to the SiO_x nanoparticles as seeds; this is called heterogeneous nucleation [25,26]. Therefore, for example, the porosity of the as-sintered nano-Ag-SiO_x, i.e., 21.8%, can be much lower than that of the as-sintered nano-Ag, i.e., 34.5%.

Furthermore, the grains of the sintered nano-Ag-SiO_x grew more than that of the sintered nano-Ag under a similar sintering profile. The larger grain size indicated a smaller grain boundary density as well. Considering that the grain boundaries are close to the free surface as fast diffusion paths [27], the more the grain boundary, the more the Ag hillocks could be formed by atomic diffusion or stress migration through the grain boundaries. The fewer grain boundaries of the sintered nano-Ag-SiO_x may be also beneficial to retard the silver ECM to some extent as a result.

For the ECM verification for device packaging applications at 400 °C under the applied voltage of 400 V, the ECM lifetime of the DSC device using the sintered nano-Ag-SiO_x and the sintered nano-Ag as die attachment is 240 h and only 96 h, respectively. The ECM of silver was suppressed successfully. As seen from Figure 5, although silver dendrites appeared after 240 h ECM testing of DSC devices using sintered nano-Ag-SiO_x, significant improvement in ECM lifetime was achieved by nano-Ag-SiO_x paste. Herein, the nano-Ag-SiO_x paste could be a promising option to be used as a bonding material against specific metal corrosion.

4. Conclusions

Silver ECM at high temperatures could be significantly retarded by the addition of SiO_x nanoparticles in the nano-Ag paste. The thermo-mechanical reliability of the proposed nano-Ag-SiO_x paste has been verified and proved to be improved by sintering under 5 MPa assisted pressures. It seems suitable as a new promising bonding material to prevent silver ECM failure because its die shear strength and thermal resistance degradation still did not reach the common failure criterion even after 1000 cycles of the thermal shock test. Finally, the ECM verification of the DSC power device further confirms the improvement of the sintered nano-Ag-SiO_x as die attachment against the sintered nano-Ag ECM failure at high temperatures. It is interesting to pay more attention to reducing the ECM failure of sintered nano-Ag in economical and promising ways in the future.