micromachines

Article Comparison of and Treatments for Ambient Room-Temperature Wafer-Scale Au–Au Bonding Using Ultrathin Au Films

Michitaka Yamamoto 1,2,* , Takashi Matsumae 1 , Yuichi Kurashima 1, Hideki Takagi 1 , Tadatomo Suga 3, Toshihiro Itoh 2 and Eiji Higurashi 1,3,* 1 National Institute of Advanced Industrial Science and Technology (AIST), 1-2-1 Namiki, Tsukuba, Ibaraki 305-8564, Japan; [email protected] (T.M.); [email protected] (Y.K.); [email protected] (H.T.) 2 Graduate School of Frontier Sciences, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa-shi, Chiba 277-8563, Japan; [email protected] 3 School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan; [email protected] * Correspondence: [email protected] (M.Y.); [email protected] (E.H.)

 Received: 20 December 2018; Accepted: 7 February 2019; Published: 13 February 2019 

Abstract: Au–Au surface activated bonding is promising for room-temperature bonding. The use of Ar plasma vs. O2 plasma for pretreatment was investigated for room-temperature wafer-scale Au–Au bonding using ultrathin Au films (<50 nm) in ambient air. The main difference between Ar plasma and O2 plasma is their surface activation mechanism: physical etching and chemical reaction, respectively. Destructive razor blade testing revealed that the bonding strength of samples obtained using Ar plasma treatment was higher than the strength of bulk Si (surface energy of 2 bulk Si: 2.5 J/m ), while that of samples obtained using O2 plasma treatment was low (surface 2 energy: 0.1–0.2 J/m ). X-ray photoelectron spectroscopy analysis revealed that a gold oxide (Au2O3) layer readily formed with O2 plasma treatment, and this layer impeded Au–Au bonding. Thermal ◦ desorption spectroscopy analysis revealed that Au2O3 thermally desorbed around 110 C. Annealing ◦ of O2 plasma-treated samples up to 150 C before bonding increased the bonding strength from 0.1 to 2 2.5 J/m due to Au2O3 decomposition.

Keywords: heterogeneous integration; wafer bonding; low-temperature bonding; Au–Au bonding; ultrathin Au films; Ar plasma treatment; O2 plasma treatment

1. Introduction Low-temperature wafer bonding is increasingly required due to the increasing use of heterogeneous integrations [1–3]. An effective approach to such bonding is to use intermediate layers [4,5]. Gold (Au) is a good candidate material for these bonding layers because of its high electrical and thermal conductivity, good deformability, and high oxidation resistance. Several Au–Au bonding techniques have been investigated to achieve low-temperature bonding, including thermocompression bonding [6–12], atomic diffusion bonding [4,13,14], and surface activated bonding (SAB) [15–19]. With SAB, the surfaces to be bonded are activated by plasma pretreatment and then bonded at low temperature (<150 ◦C). Furthermore, the use of Au for the bonding layer enables bonding in ambient air because gold oxide (Au2O3) is the only metal oxide that has a positive formation enthalpy (+19.3 kJ/mol) [20]. The use of SAB with little contact pressure at room temperature in ambient air using ultrathin (<50 nm) Au films with smooth (root mean square (RMS) <0.50 nm) surfaces was recently shown to result in reliable Au–Au bonding [17].

Micromachines 2019, 10, 119; doi:10.3390/mi10020119 www.mdpi.com/journal/micromachines Micromachines 2019, 10, 119 2 of 12

While Ar plasma is commonly used to activate Au surfaces by physically bombarding them with Ar or , O2 plasma is commonly used to clean surfaces because it can eliminate organic contaminants by chemical reaction [21,22]. It is also used for pretreatment before Au–Au bonding; O2 plasma treatment has been reported to more effectively improve Au wire bond interfacial adhesion than Ar plasma treatment [23,24]. However, the Au–Au bonding may be weakened by the mixing of O2 gas into the plasma processing gas [25,26]. The effect of plasma treatment, and especially the effect of O2 plasma treatment on Au–Au bonding, thus needs further investigation. This paper reports the investigation of Ar plasma vs. O2 plasma as a pretreatment method for room-temperature wafer-scale Au–Au bonding using ultrathin Au films (<50 nm) in ambient air. The main difference between Ar plasma and O2 plasma is their surface activation mechanism: physical etching and chemical reaction, respectively. The investigation included an analysis of Au surfaces treated with Ar or O2 plasma.

2. Materials and Methods Silicon substrates with Au film and a Ti adhesion layer were used for analysis and bonding. A Ti adhesion layer followed by Au thin film was deposited on a Si substrate or a thermally oxidized Si ◦ substrate by direct current (DC) sputtering at a substrate temperature of 25 C. The Ar or O2 plasma used to activate the surfaces was generated by applying 200 W radio frequency (RF) power at 60 Pa. Both plasmas were generated using the same equipment; only the process gas was changed. The effect of each plasma treatment on the surface properties was determined by evaluating the surface roughness, hydrophilicity, electrical characteristics, and chemical state of the Au surfaces before and after each treatment. Surface roughness was measured using an atomic force microscope (AFM, Hitachi High-Tech Science Co., Tokyo, Japan, L-trace) and Si wafers with Au thin film (thickness: 15 nm) and a Ti adhesion layer (thickness: 5 nm). The AFM measurement was performed with a scan area of 500 nm × 500 nm and a resolution of 256 × 256 pixels. Hydrophilicity was evaluated by measuring the contact angle of the Au surfaces as measured with a contact angle meter (Kyowa Interface Science Co, Ltd., Saitama, Japan, PCA-11). Pure water (1.5 µL) was dropped on the Au surfaces, and the contact angle was calculated using the half-angle method. The electrical characteristics were evaluated by four-point measurement of the sheet resistance using a resistivity processor (NPS Inc., Tokyo, Japan, Model Σ) and Si wafers with Au thin film (thickness: 15 nm) and a Ti adhesion layer (thickness: 5 nm). The effect of the plasma treatment time was measured by irradiating plasma onto the same wafer several times. The surface chemical states were determined using X-ray photoelectron spectroscopy (XPS) analysis and thermal desorption spectrometry (TDS) analysis. The XPS analysis was performed using an X-ray photoelectron spectroscope (JEOL Ltd., Tokyo, Japan, JPS9200-T). A non-monochromatic Mg Kα source ( energy: 1253.6 eV) was used at an operating power of 10 kV × 10 mA. Smoothing, Shirley-type background subtraction, and charge correction using the Au 4f7/2 binding energy were performed. The TDS analysis was performed using a thermal desorption spectrometer (ESCO Ltd., Tokyo, Japan, EMD-WA1000S). Thicker Au films (thickness: 300 nm) with Ti adhesion layers (thickness: 5 nm) deposited on thermally oxidized Si substrates were used to avoid the effect of diffusion from the adhesion layer or Si substrate during annealing. The temperature was increased at a rate of 30 ◦C/min and was corrected by measurement using a dummy Si substrate. Four-inch Si wafers with Au thin film (thickness: 15 nm) and a Ti adhesion layer (thickness: 5 nm) were bonded after Ar or O2 plasma treatment in ambient air by overlapping two wafers and pinching the wafers together with tweezers. Bond front propagation was observed using an infrared (IR) transmission setup. The effects of plasma treatment time and air exposure time after plasma treatment on self-propagation of the bonding area and on bonding strength were investigated. The effects of annealing before bonding on the self-propagation of the bonding area and on bonding strength were also investigated by annealing wafers that had been plasma-treated up to 150 ◦C for 10 min in vacuum and then bonding them together in ambient air at room temperature. Micromachines 2019, 10, 119 3 of 12 Micromachines 2019, 10, x 3 of 11

Bonding strength strength was was measured measured using using destructive destructive razor razor blade blade testing, testing, also also known known as the as crack the crack openingopening or double cantilever cantilever beam beam method method [27,28]. [27,28 This]. This testing testing method method is an is effective an effective way wayto evaluate to evaluate thethe energyenergy of the bonded bonded surfaces surfaces (i.e., (i.e., surface surface ener energy).gy). It is It performed is performed by inserting by inserting a razor a razor blade blade betweenbetween thethe bondedbonded waferwafer pairs pairs and and observing observing the the crack crack length length along along the the bonded bonded interface. interface. The The surface energysurface γenergy(J/m2 γ) (i.e.,(J/m2 half) (i.e., the half bonding the bonding energy energyG) was G) calculatedwas calculated under under plane plane strain strain conditions conditions and the assumptionsand the assumptions of beam of theory beam usingtheory using 2 3 γ 3∙ ET T, w γ = · b ,(1) (1) 32 L4 where E (Pa) is the Young’s modulus of the wafer, Tb (m) is the blade thickness, Tw (m) is the wafer where E (Pa) is the Young’s modulus of the wafer, T (m) is the blade thickness, T (m) is the wafer thickness, and L (m) is the crack length [27]. In this experiment,b Si substrates with a thicknessw of thickness,525 µm and and a bladeL (m) with is the a crackthickness length of [10027]. µm In thiswere experiment, used. For calculation, Si substrates we with assumed a thickness that Si ofis isotropic,525 µ withm and Young’s a blade modulus with a thicknessof elasticity of E 100 = 169µm GPa were [29,30]. used. Although For calculation, the measurement we assumed should that Si is isotropic,be conducted with using Young’s beam-shaped modulus ofsamples elasticity in orderE = 169 to accurately GPa [29,30 determine]. Although the the surface measurement energy [31], should bewe conducted used whole using bonded beam-shaped wafer pairs samples because in the order fabr toication accurately of beam-shaped determine samples the surface is particularly energy [31], we useddifficult whole for bondedweakly bonded wafer pairs wafer because pairs. theThe fabricationcrack length of along beam-shaped the bonded samples interface is particularly was observed difficult forwith weakly an IR camera bonded in wafer a standard pairs. clean The crackroom lengthatmosphere. along Since the bonded values obtained interface from was observedEquation (1) with are an IR cameraaffected in by a standardthe crack clean detection room atmosphere.resolution, Sinceanisotropic values mechanical obtained from properties, Equation and (1) arehumidity affected by theenvironment crack detection [28,30], resolution, the comparison anisotropic of absolute mechanical energy properties, values with and published humidity environmentresults requires [28 ,30], thecareful comparison consideration. of absolute energy values with published results requires careful consideration.

3.3. Results

3.1.3.1. Surface Analysis Analysis

3.1.1.3.1.1. AFM Measurement Measurement The measured RMS RMS surface surface roughness roughness for for different different plasma plasma treatment treatment times times (30, (30,60, 120 60, s) 120 are s) are plottedplotted in in Figure Figure 1.1 .The The roughness roughness before before plasma plasma treatment treatment was 0.38 was nm. 0.38 Both nm. the BothAr and the O2 Arplasma and O 2 plasmatreatments treatments of 60 s or of less 60 shad or lesslittle had effect little on effectsurface on roughness, surface roughness, while both while treatments bothtreatments of 120 s of 120increased s increased it. This it. indicates This indicates that a short that aplasma short plasmatreatment treatment time is important time is important for achieving for achievinglow- low-temperaturetemperature bonding. bonding.

FigureFigure 1. 1Root. Root mean mean square square(RMS) (RMS) surfacesurface roughness vs. vs. plasma plasma treatment treatment time. time.

3.1.2.3.1.2. Contact Angle Angle Measurement Measurement The measured contact contact angles angles for for different different plasma plasma treatment treatment times times (30, (30,60, 120 60, 120s) are s) shown are shown in in FigureFigure2 .2. The The Au Au films films had had been been exposed exposed to air to forair over for 1over month, 1 month, and the and measurement the measurement was performed was immediatelyperformed immediately after plasma after treatment. plasma treatment. Typical imagesTypical ofimages the measured of the measured contact contact angle areangle shown are in Figureshown3 in. Figure Figure3 b,c3. Figures was taken 3b and 30 s3c after were plasma taken 30 treatment. s after plasma Before treatment. treatment, Before the contact treatment, angle the was aboutcontact 80–100 angle ◦was, and about the Au80–100°, surfaces and werethe Au hydrophobic. surfaces were Both hydrophobic. treatments Both reduced treatments the contact reduced angle, whichthe contact means angle, that which the Au means surfaces that becamethe Au surfaces hydrophilic became with hydrophilic plasma treatment. with plasma The treatment. contactangle The for

Micromachines 2019, 10, x 4 of 11 contact angle for Au surfaces with O2 plasma treatment was less than 5°, which is smaller than that with ArMicromachines plasma treatment.2019, 10, x The treatment times were the same. 4 of 11 The changes in the contact angle with the air exposure time after Au film deposition and after plasmacontact treatment angle forare Au plotted surfaces in withFigure O2 4.plasma Previo treatmentus studies was reported less than that5°, which Au film is smaller is hydrophilic than that immediatelywith Ar plasmaafter deposition treatment. and The that treatment the Au times surf wereace quickly the same. becomes hydrophobic after exposure The changes in the contact angle with the air exposure time after Au film deposition and after in air [32,33]. The measured change in the contact angle after deposition agrees well with the results plasma treatment are plotted in Figure 4. Previous studies reported that Au film is hydrophilic of previous studies. The contact angles of Au surfaces that received Ar plasma treatment increased immediately after deposition and that the Au surface quickly becomes hydrophobic after exposure rapidlyMicromachines Micromachinesafter exposure2019, 201910, 119, in10, xair, similar to the results for the Au film immediately after deposition.4 In4 of of 11 12 in air [32,33]. The measured change in the contact angle after deposition agrees well with the results contrast, the rate of increase was much lower for Au surfaces that received O2 plasma treatment. of previouscontact angle studies. for AuThe surfaces contact withangles O 2of plasma Au surfaces treatment that wasreceived less than Ar plasma 5°, which treatment is smaller increased than that ◦ Aurapidly surfaceswith Arafter withplasma exposure O 2treatment.plasma in air, treatment Thesimilar treatment to wasthe resulttimes less thans were for 5the the, whichAu same. film is immediately smaller than after that deposition. with Ar plasma In treatment.contrast,The the The changes rate treatment of increase in the times contact was were much angle the lower same.with for the Au air surfaces exposure that time received after Au O2 filmplasma deposition treatment. and after plasma treatment are plotted in Figure 4. Previous studies reported that Au film is hydrophilic immediately after deposition and that the Au surface quickly becomes hydrophobic after exposure in air [32,33]. The measured change in the contact angle after deposition agrees well with the results of previous studies. The contact angles of Au surfaces that received Ar plasma treatment increased rapidly after exposure in air, similar to the results for the Au film immediately after deposition. In contrast, the rate of increase was much lower for Au surfaces that received O2 plasma treatment.

Figure 2. Measured contact angle vs. plasma treatment time. Figure 2. Measured contact angle vs. plasma treatment time. Figure 2. Measured contact angle vs. plasma treatment time.

Figure 2. Measured contact angle vs. plasma treatment time.

(b) (a) (c) (a) (b) Figure 3. Typical images of measured contact angle: (a) without plasma treatment; (b) with(c) 30 s Ar Figure 3. Typical images of measured contact angle: (a) without plasma treatment; (b) with 30 s Ar plasma treatment; (c) with 30 s O2 plasma treatment. plasmaFigure treatment; 3. Typical (c) with images 30 sof O measured2 plasma treatment.contact angle: (a) without plasma treatment; (b) with 30 s Ar Theplasma changes treatment; in the (c) contact with 30 s angle O2 plasma with treatment. the air exposure time after Au film deposition and after plasma treatment are plotted in Figure4. Previous studies reported that Au film is hydrophilic immediately after deposition and that the Au surface quickly becomes hydrophobic after exposure in (b) air [32,33]. The measured(a) change in the contact angle after deposition agrees well with the(c) results of previous studies. The contact angles of Au surfaces that received Ar plasma treatment increased rapidly after exposureFigure in3. Typical air, similar images to of the measured results contact for the angle: Au film (a) without immediately plasma aftertreatment; deposition. (b) with 30 In s contrast, Ar plasma treatment; (c) with 30 s O2 plasma treatment. the rate of increase was much lower for Au surfaces that received O2 plasma treatment.

FigureFigure 4. Changes 4. Changes in contact in contact angle angle after after exposure exposure in airin airfollowing following Au Au film film deposition deposition and and plasma plasma treatment.treatment.

3.1.3. 3.1.3.Sheet Sheet Resistance Resistance Measurement Measurement

Figure 4. Changes in contact angle after exposure in air following Au film deposition and Figure 4. Changes in contact angle after exposure in air following Au film deposition and plasma plasma treatment. treatment. 3.1.3. Sheet Resistance Measurement 3.1.3. Sheet Resistance Measurement The relationship between the measured sheet resistance and the plasma treatment time is plotted in Figure 5. Before plasma treatment, the sheet resistance was about 4.7 Ω. While Ar plasma treatment Micromachines 2019, 10, x 5 of 11 Micromachines 2019, 10, x 5 of 11 MicromachinesThe relationship2019, 10, 119 between the measured sheet resistance and the plasma treatment time is plotted5 of 12 in FigureThe 5. relationship Before plasma between treatmen thet, measured the sheet resistancesheet resist wasance about and the 4.7 plasmaΩ. While treatment Ar plasma time treatment is plotted didin notFigure change 5. Before the resistance,plasma treatmen 30 s Ot,2 theplasma sheet treatment resistance increased was about it 4.7 to Ωabout. While 5.4 Ar Ω .plasma Additional treatment O2 plasmadiddid notnot treatment changechange thethetime resistance,resistance, did not increase 3030 ss OO2 2it. plasmaplasma treatmenttreatment increasedincreased itit toto aboutabout 5.45.4Ω Ω.. AdditionalAdditional OO22 plasmaplasma treatmenttreatment timetime diddid notnot increaseincrease it.it. 6.0 )

□ 6.0 /

5.5) Ω □ / 5.5

5.0Ω 5.0 4.5 4.5 4.0 4.0 ArAr plasma plasma Sheet resistance ( resistance Sheet 3.5 O2OAr2Ar plasmaplasma plasma plasma Sheet resistance ( resistance Sheet 3.5 3.0 O2O2 plasmaplasma 3.00 50 100 150 0Plasma 50treatment time 100 (s) 150 Plasma treatment time (s) Figure 5. Measured sheet resistance vs. plasma treatment time. Figure 5. Measured sheet resistance vs. plasma treatment time. Figure 5. Measured sheet resistance vs. plasma treatment time. 3.1.4.3.1.4. XPS XPS Analysis Analysis 3.1.4. XPS Analysis TheThe Au Au surface surface conditions conditions were were evaluated evaluated by by using using XPS XPS analysis analysis to to investigate investigate Au Au surfaces surfaces after after ArAr or or OThe2 O plasma2 Auplasma surface treatment. treatment. conditions The The treatment were treatment evaluated time time was by was set using to set 60 XPS to s. 60 The analysis s. relative The relativeto investigate peak peakintensities intensitiesAu surfaces in the in Au after the 4fAu Arregion or 4f regionO are2 plasma plotted are plottedtreatment. in Figure in Figure The 6a,b, treatment6 anda,b, andthe relativetime the relative was peak setpeak to intensities 60 intensitiess. The inrelative the in O the peak1s O region 1s intensities region are plotted are in plotted the in Au Figurein4f Figureregion 6c. The6 arec. Theresults plotted results plotted in Figure plotted in (a) 6a,b, in show (a) and show that the the thatrelative Au the surfaces Aupeak surfaces intensities that received that in received the Ar O plasma 1s Ar region plasma treatment are treatment plotted were in purewereFigure Au, pure 6c. while The Au, results whilethose thoseplotted plotted plotted in (a)(b) in showshow (b) showthat that the thatan Au Au an surfaces2 AuO3 2peakO3 thatpeak appeared received appeared for Ar for theplasma the Au Au treatmentsurfaces surfaces that were that receivedreceivedpure Au, O O2 plasmawhile2 plasma those treatment. treatment. plotted The Thein results (b) results show plotted plotted that inan in (c) Au (c) show2 showO3 peakthat that the appeared the O O2 plasma2 plasma for treatmentthe treatment Au surfaces increased increased that thethereceived oxygen oxygen O contamination2 contamination plasma treatment. intensity. intensity. The results plotted in (c) show that the O2 plasma treatment increased the oxygen contamination intensity. Au Au Au 4f Au Au 4f Au Au 4f Au 4f

Au2O3

Au2O3 Intensity (a.u.) Intensity (a.u.) Intensity (a.u.) Intensity (a.u.)

92 88 84 80 76 92 88 84 80 76 Binding energy(eV) 92Binding 88 energy(eV)84 80 76 92 88 84 80 76 Binding energy(eV) Binding energy(eV) (a) (b) (a) (b)

Ar plasma O2Ar plasma plasma

O2 plasma Intensity (a.u.) Intensity Intensity (a.u.) Intensity

540 536 532 528 524 Binding energy(eV) 540 536 532 528 524 Binding(c) energy(eV) (c) FigureFigure 6. 6. X-rayX-ray photoelectron photoelectron spectroscopyspectroscopy(XPS) (XPS) spectra spectra of of Au Au surfaces surfaces that that received received different different plasma plasmatreatments:Figure treatments: 6. X-ray (a) relative photoelectron(a) relative peak intensity peak spectroscopy intensity in the Auin (XPS)the 4f region Au spectra 4f ofregion surface of ofAu thatsurface surfaces received that that received Ar received plasma Ar treatment; plasmadifferent (plasmab) relative treatments: peak intensity (a) relative in the peak Au 4f intensity region of in surface the Au that 4f region received of Osurface2 plasma that treatment; received (Arc) relativeplasma peak intensities in the O 1s region of surface that received Ar and O2 plasma treatments.

Micromachines 2019, 10, x 6 of 11

Micromachinestreatment;2019 ,(b10), relative 119 peak intensity in the Au 4f region of surface that received O2 plasma treatment;6 of 12 (c) relative peak intensities in the O 1s region of surface that received Ar and O2 plasma treatments.

3.1.5.3.1.5. TDS TDS Analysis Analysis TheThe results results of of the the TDS TDS analysis analysis are are plotted plotted in in Figure Figure7 .7. There There was was little little desorption desorption from from the the Au Au surfacesurface with with the the Ar Ar plasma plasma treatment. treatment. In In contrast, contrast, with with the the O O22 plasmaplasma treatment,treatment,mass-to-charge mass-to-charge ratio ratio (m(m/z/z)) peakspeaks ofof 1616 andand 1818 werewere detected detected at at annealing annealing temperatures temperatures around around 70 70◦ C,°C, and andm m/z/z peaks peaks of of 16 16 andand 32 32 were were detected detected at at annealing annealing temperatures temperatures around around 110 110◦C. °C. Since Sincem/ zm/z16, 16, 18, 18, and and 32 correspond32 correspond to ◦ O,to HO,2 O,H2 andO, and O2 ,O it2, appearsit appears that that H 2HO2O was was desorbed desorbed around around 70 70 C°C from from the the Au Au surfaces surfaces that that received received ◦ OO22plasma plasma treatment,treatment,while whileO O22 waswas desorbeddesorbed aroundaround 110110 °C.C.

M/zm/z 1616 (O)

M/zm/z 1818 (H2O)

M/zm/z 3232 (O2) intensity (arb.unit) Ion intensity

25 125 225 325 Annealing temperature (oC)

(a)

M/zm/z 1616 (O)

M/zm/z 1818 (H2O)

M/zm/z 3232 (O2) Ion intensity (arb.unit) Ion intensity

25 125 225 325 Annealing temperature (oC)

(b)

FigureFigure 7. 7.Results Results of of TDS TDS analysis: analysis: ( a(a)) with with Ar Ar plasma plasma treatment; treatment; ( b(b)) with with O O2 2plasma plasma treatment. treatment.

3.2.3.2. BondingBonding Results results 3.2.1. Effect of Plasma Treatment Time 3.2.1. Effect of Plasma Treatment Time To determine the effect of the plasma treatment time, Au–Au bonding with or without plasma To determine the effect of the plasma treatment time, Au–Au bonding with or without plasma treatment was performed. The treatment times were 30 s or 60 s because a longer treatment time treatment was performed. The treatment times were 30 s or 60 s because a longer treatment time increased the surface roughness. Bonding was performed immediately after plasma treatment. Without increased the surface roughness. Bonding was performed immediately after plasma treatment. plasma treatment, bonding front propagation did not occur, and bonding did not succeed. With plasma Without plasma treatment, bonding front propagation did not occur, and bonding did not succeed. treatment, bonding front propagation occurred, and bonding succeeded. IR images of typical bond With plasma treatment, bonding front propagation occurred, and bonding succeeded. IR images of front propagation between two Si wafers for a plasma treatment time of 60 s are shown in in Figure8. typical bond front propagation between two Si wafers for a plasma treatment time of 60 s are shown The contrast and brightness of the images were adjusted, and the edges of the bonding area were in in Figure 8. The contrast and brightness of the images were adjusted, and the edges of the bonding marked to make it more visible. With Ar plasma treatment, the bonding area quickly expanded across area were marked to make it more visible. With Ar plasma treatment, the bonding area quickly the entire 4-inch wafer after pushing the center. With O2 plasma treatment, the bonding area expanded expanded across the entire 4-inch wafer after pushing the center. With O2 plasma treatment, the much more slowly. bonding area expanded much more slowly.

Micromachines 2019, 10, x 7 of 11

The results of the destructive razor blade testing are summarized in Table 1. With Ar plasma treatment (30 s and 60 s), the bonding was sufficient to break the substrate, meaning that strong bonding was achieved. With O2 plasma treatment, although much of the area was bonded, the bonding strength (surface energy) was only about 0.1–0.2 J/m2, lower than the surface energy of Au (1.6 J/m2) [34]. Debonding thus occurred between the Au–Au bonding interfaces. Transmission microscope (TEM) images of a bonded sample that received Ar plasma Micromachinestreatment are2019 shown, 10, 119 in Figure 9. Bonding was achieved at the atomic level, and part of the grain7 of 12 boundary grew beyond the bonding interface.

(a)

(b)

Figure 8. 8. ImagesImages ofof increasing increasing bonding bonding area: area: (a) with (a) with60 s Ar 60 plasma s Ar plasma treatment; treatment; (b) with ( 60b) s with O2 plasma 60 s O 2 plasmatreatment. treatment.

TheTable results 1. Results ofthe of room-tempera destructiveture razor wafer-scale blade testing Au–Au are bonding summarized with Ar or in O Table2 plasma1. With treatment. Ar plasma treatment (30 s and 60 s), the bonding wasPlasma sufficient treatment to breaktime (s) the substrate, meaning that strong bonding Types of was achieved. With O plasma0 treatment, although much of the area was bonded, the bonding strength plasmas 2 30 60 (surface energy) was(Without only aboutplasma) 0.1–0.2 J/m2, lower than the surface energy of Au (1.6 J/m2)[34]. Bonding Wafer broken Wafer broken DebondingAr plasma thus occurred between the Au–Au bonding interfaces. failed (High bonding strength) (High bonding strength) Bonding O plasma 0.2 J/m2 0.1 J/m2 Table2 1. Results of room-temperaturefailed wafer-scale Au–Au bonding with Ar or O2 plasma treatment.

Plasma Treatment Time (s) Types of Plasmas 0 30 60 (Without Plasma) Bonding Wafer broken Wafer broken Ar plasma failed (High bonding strength) (High bonding strength) Bonding O plasma 0.2 J/m2 0.1 J/m2 2 failed Micromachines 2019, 10, 119 8 of 12

Transmission electron microscope (TEM) images of a bonded sample that received Ar plasma treatment are shown in Figure9. Bonding was achieved at the atomic level, and part of the grain boundary grew beyond the bonding interface. Micromachines 2019, 10, x 8 of 11

(a) (b)

FigureFigure 9.9. TransmissionTransmission electronelectron microscopemicroscope (TEM)(TEM) imagesimages ofof Au–AuAu–Au bondingbonding interfaceinterface afterafter 3030 ss ArAr plasmaplasma treatment.treatment. ArrowsArrows indicateindicate originaloriginal interface:interface: ((aa)) lowlow magnification;magnification; ( b(b)) high high magnification. magnification. 3.2.2. Effect of Air Exposure 3.2.2. Effect of Air Exposure To investigate the effect of air exposure after plasma treatment, bonding was performed after To investigate the effect of air exposure after plasma treatment, bonding was performed after air air exposure. The bonding of Au films immediately after deposition was also compared. Because exposure. The bonding of Au films immediately after deposition was also compared. Because Au Au films immediately after deposition have clean and activated surfaces, bonding can be performed films immediately after deposition have clean and activated surfaces, bonding can be performed at at room temperature. This bonding technique is usually called atomic diffusion bonding [4,13,14]. room temperature. This bonding technique is usually called atomic diffusion bonding [4,13,14]. The The effect of air exposure time on the occurrence of self-propagation of the bonding area and on the effect of air exposure time on the occurrence of self-propagation of the bonding area and on the bonding strength of atomic diffusion bonding for Au films immediately after deposition is summarized bonding strength of atomic diffusion bonding for Au films immediately after deposition is in Table2. Self-propagation occurred even after 1 h of air exposure. Though the bonding strength summarized in Table 2. Self-propagation occurred even after 1 h of air exposure. Though the bonding decreased as the exposure time increased, sufficient bonding strength (greater than the surface energy strength decreased as the exposure time increased, sufficient bonding strength (greater than the of bulk Si (2.5 J/m2)[35]) was obtained even after 1 h of air exposure. surface energy of bulk Si (2.5 J/m2) [35]) was obtained even after 1 h of air exposure. TableThe effect 2. Effect of ofair air exposure exposure time onfor the surface occurrence activated of self-propagation bonding is summarized of the bonding in area Table and 3. on In the casesthe where bonding self-propagation strength of atomic did diffusion not occur, bonding bondin forg Auwas films achieved immediately by pushing after deposition. the whole wafer with tweezers. With Ar plasma treatment, self-propagation did not occur after 30 min of air exposure. Moreover, the bonding strength greatly decreased asAir the Exposure air exposure Time time increased. With O2 plasma Evaluation Points treatment, while self-propagationWithin 5 min occurred even after 10 min 1 hour of air exposure, 30 min the bonding strength 1 h was Occurrencelower under of all conditions. Yes Yes Yes Yes self-propagation BondingTable strength2. Effect of air exposureWafer broken time on the occurrenceWafer of broken self-propagation ofWafer the bonding broken area and on 2 >2.5 the(J/m bonding) strength(High of bondingatomic diffusion strength) bond(Highing bondingfor Au films strength) immediately(High bondingafter deposition. strength)

Evaluation Air exposure time Thepoints effect of air exposureWithin 5 timemin for surface 10 activatedmin bonding 30 is min summarized in Table1 h 3. In the cases whereOccurrence self-propagation of did not occur, bonding was achieved by pushing the whole wafer with Yes Yes Yes Yes tweezers.self-propagation With Ar plasma treatment, self-propagation did not occur after 30 min of air exposure. Bonding Wafer broken Wafer broken Wafer broken Moreover, the bonding strength greatly decreased as the air exposure time increased. With O plasma strength (High bonding (High bonding (High bonding > 2.5 2 treatment,(J/m while2) self-propagationstrength) occurred evenstrength) after 1 hour of airstrength) exposure, the bonding strength was lower under all conditions. Table 3. Effect of air exposure time on the occurrence of self-propagation of the bonding area and on the bonding strength of surface activated bonding.

Types of Evaluation Air exposure time plasmas points Within 5 min 10 min 30 min 1 h

Occurrence of Yes Yes No No self-propagation Ar plasma Wafer broken Bonding strength (High bonding > 2.5 2.0 0.1 (J/m2) strength)

Occurrence of O plasma Yes Yes Yes Yes 2 self-propagation

Micromachines 2019, 10, 119 9 of 12

Table 3. Effect of air exposure time on the occurrence of self-propagation of the bonding area and on the bonding strength of surface activated bonding.

Types of Evaluation Air Exposure Time Plasmas Points Within 5 min 10 min 30 min 1 h Occurrence of Yes Yes No No self-propagation Ar plasma Bonding strength Wafer broken >2.5 2.0 0.1 (J/m2) (High bonding strength) Occurrence of Yes Yes Yes Yes self-propagation O2 plasma Bonding strength 0.1 0.1 0.1 0.1 (J/m2)

3.2.3. Room-Temperature Wafer-Scale Au–Au Bonding After Annealing

Analysis of Au surfaces that received O2 plasma treatment revealed that Au2O3 formed and that ◦ ◦ O2 desorbed around 110 C. This indicates that bonding after annealing at over 110 C increases the ◦ bonding strength of samples that receive O2 plasma treatment. Here, annealing up to 150 C after plasma treatment was investigated: Ar or O2 plasma was irradiated onto Au surfaces for 60 s, and then the wafers were annealed at 150 ◦C for 10 min in vacuum. After annealing, the wafers were bonded in ambient air at room temperature. The entire annealing process took about one hour. Self-propagation did not occur in either the Ar or O2 plasma-treated samples. The wafers were thus bonded by pushing the entire wafers with tweezers. Although the bonding strength of samples that received Ar plasma treatment exceeded the surface energy of bulk Si (2.5 J/m2), the annealing reduced the bonding strength, as evidenced by the breaking of the wafers during blade testing for wafers that had not been annealed. The bonding strength of samples that received O2 plasma treatment increased substantially (from 0.1 J/m2 without annealing to 2.5 J/m2 with annealing). To investigate whether Au2O3 was desorbed, XPS analysis of annealed Au surfaces after O2 plasma treatment was also performed. The results did not reveal an Au2O3 peak, meaning that Au2O3 formed by O2 plasma treatment impedes Au–Au bonding. To investigate the change in surface roughness due to annealing, AFM analysis was also performed. The results did not reveal any changes in surface roughness.

4. Discussion While Ar plasma treatment improved room-temperature wafer-scale Au–Au bonding strength, ◦ O2 plasma treatment did not. The O2 plasma treatment reduced the contact angle to less than 5 and increased the sheet resistance. XPS analysis showed that the O2 plasma treatment formed Au2O3 on the Au surface. This indicates that the changes in contact angle and sheet resistance were both apparently caused by oxidation of the Au surfaces and that the Au2O3 formed by the O2 plasma treatment impeded Au–Au bonding. ◦ Because TDS analysis showed that Au2O3 desorbed at around 110 C, bonding after annealing up to 150 ◦C was performed. While the bonding strength of samples that received Ar plasma treatment was reduced by the annealing, that of samples that received O2 plasma treatment was improved from 2 0.1 to 2.5 J/m . This also demonstrates that the Au2O3 formed by O2 plasma treatment impeded Au–Au bonding. The decrease in bonding strength with Ar plasma treatment may have been due to the annealing processing time (~1 h). The effects of air exposure after Au film deposition or plasma treatment on the contact angle and bonding strength were also investigated. The changes in surface hydrophilicity due to contact angle measurement of water on Au surfaces that received Ar plasma treatment were similar to those on Au surfaces immediately after deposition. The bonding strength of samples that were exposed in air following Ar plasma treatment decreased more rapidly than that of samples bonded immediately Micromachines 2019, 10, 119 10 of 12 after deposition. Further research on the effects of air exposure and the relationship between surface conditions and bonding strength is necessary. Previous studies found that O2 plasma treatment improved Au–Au bonding strength [23,24], ◦ possibly because Au2O3 desorbed due to the high bonding temperature (150 C) or ultrasonic vibrations.

5. Conclusions

Pretreatment using Ar or O2 plasma was investigated for ambient room-temperature wafer-scale Au–Au bonding using ultrathin Au films. Surface activation by Ar plasma is mainly due to physical etching by Ar ions, while surface activation by O2 plasma is mainly due to chemical reaction. The O2 plasma treatment increased the sheet resistance of the Au surfaces but did not increase the bonding strength. XPS analysis revealed that Au2O3 formed on the Au surfaces that received O2 plasma treatment, and TDS analysis revealed that annealing at over 110 ◦C is necessary for desorption of ◦ Au2O3. Annealing up to 150 C before bonding increased the bonding strength of samples that received O2 plasma treatment, demonstrating that Au2O3 impedes strong Au–Au bonding. On the other hand, Ar plasma treatment increased the Au–Au bonding strength enough for the Si substrate to be broken in destructive razor blade testing. Although an Au surface that receives Ar plasma treatment is more activated than one that receives O2 plasma treatment, bonding should be performed immediately after plasma treatment.

Author Contributions: Investigation, M.Y., T.M., Y.K., H.T., T.S., T.I. and E.H.; Methodology, M.Y. and E.H.; Validation, M.Y.; Data curation, M.Y.; Writing—original draft, M.Y.; Writing—review & Editing, T.M., Y.K., H.T., T.S., T.I. and E.H.; Funding acquisition, T.H. and E.H.; Resource, H.T., T.S., T.I. and E.H.; Supervision, T.S., T.I. and E.H.; Project administration, E.H. Funding: This work is partly supported by New Energy and Industrial Technology Development Organization (NEDO). Conflicts of Interest: The authors declare no conflict of interest.

References

1. Morrow, P.R.; Park, C.-M.; Ramanathan, S.; Kobrinsky, M.J.; Harmes, M. Three-dimensional wafer stacking via Cu-Cu bonding integrated with 65-nm strained-Si/low-k CMOS technology. IEEE Electron Device Lett. 2006, 27, 335–337. [CrossRef] 2. Higurashi, E. Low-temperature bonding technologies for photonics applications. ECS Trans. 2013, 50, 351–362. [CrossRef] 3. Howlader, M.M.R.; Okada, H.; Kim, T.H.; Itoh, T.; Suga, T. Wafer level surface activated bonding tool for MEMS packaging. J. Electrochem. Soc. 2004, 151, G461. [CrossRef] 4. Shimatsu, T.; Uomoto, M. Atomic diffusion bonding of wafers with thin nanocrystalline metal films. J. Vac. Sci. Technol. B Nanotechnol. Microelectron. Mater. Process. Meas. Phenom. 2010, 28, 706–714. [CrossRef] 5. Niklaus, F.; Stemme, G.; Lu, J.-Q.; Gutmann, R.J. Adhesive wafer bonding. J. Appl. Phys. 2006, 99, 031101. [CrossRef] 6. Andrijasevic, D.; Austerer, M.; Andrews, A.M.; Klang, P.; Schrenk, W.; Strasser, G. Hybrid integration of GaAs quantum cascade lasers with Si substrates by thermocompression bonding. Appl. Phys. Lett. 2008, 92, 10–13. [CrossRef] 7. Taklo, M.M.V.; Storås, P.; Schjølberg-Henriksen, K.; Hasting, H.K.; Jakobsen, H. Strong, high-yield and low-temperature thermocompression silicon wafer-level bonding with gold. J. Micromechanics Microengineering 2004, 14, 884–890. [CrossRef] 8. Tanida, K.; Umemoto, M.; Morifuji, T.; Kajiwara, R.; Ando, T.; Tomita, Y.; Tanaka, N.; Takahashi, K. Au Bump Interconnection in 20 µm Pitch on 3D Chip Stacking Technology. Jpn. J. Appl. Phys. 2003, 42, 6390–6395. [CrossRef] 9. Unami, N.; Sakuma, K.; Mizuno, J.; Shoji, S. Effects of excimer irradiation treatment on thermocompression Au-Au bonding. Jpn. J. Appl. Phys. 2010, 49, 10–13. [CrossRef] 10. Zhang, G.G.; Ang, X.F.; Chen, Z.; Wong, C.C.; Wei, J. Critical temperatures in thermocompression gold stud bonding. J. Appl. Phys. 2007, 102, 063519. [CrossRef] Micromachines 2019, 10, 119 11 of 12

11. Jing, E.; Xiong, B.; Wang, Y. Low-temperature Au–Si wafer bonding. J. Micromechanics Microengineering 2010, 20, 095014. [CrossRef] 12. Al Farisi, M.S.; Hirano, H.; Frömel, J.; Tanaka, S. Wafer-level hermetic thermo-compression bonding using electroplated gold sealing frame planarized by fly-cutting. J. Micromechanics Microengineering 2017, 27, 015029. [CrossRef] 13. Shimatsu, T.; Uomoto, M.; Kon, H. Room Temperature Bonding Using Thin Metal Films (Bonding Energy and Technical Potential). ECS Trans. 2014, 64, 317–328. [CrossRef] 14. Uomoto, M.; Shimatsu, T. Necessary Thickness of Au Capping Layers for Room Temperature Bonding of Wafers in Air Using Thin Metal Films with Au Capping Layers. ECS Trans. 2016, 75, 67–76. [CrossRef] 15. Higurashi, E.; Imamura, T.; Suga, T.; Sawada, R. Low-temperature bonding of laser diode chips on silicon substrates using plasma activation of Au films. IEEE Photonics Technol. Lett. 2007, 19, 1994–1996. [CrossRef] 16. Kurashima, Y.; Maeda, A.; Takagi, H. Room-temperature wafer scale bonding using smoothed Au seal ring surfaces for hermetic sealing. Jpn. J. Appl. Phys. 2016, 55, 016701. [CrossRef] 17. Higurashi, E.; Okumura, K.; Kunimune, Y.; Suga, T.; Hagiwara, K. Room-Temperature Bonding of Wafers with Smooth Au Thin Films in Ambient Air Using a Surface-Activated Bonding Method. IEICE Trans. Electron. 2017, E100–C, 156–160. [CrossRef] 18. 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. [CrossRef] 19. Matsumae, T.; Kurashima, Y.; Umezawa, H.; Mokuno, Y.; Takagi, H. Room-temperature bonding of single-crystal diamond and Si using Au/Au atomic diffusion bonding in atmospheric air. Microelectron. Eng. 2018, 195, 68–73. [CrossRef] 20. Bond, G.C. Gold: A relatively new catalyst. Gold Bull. 2001, 34, 117–119. [CrossRef] 21. Raiber, K.; Terfort, A.; Benndorf, C.; Krings, N.; Strehblow, H.-H. Removal of self-assembled monolayers of alkanethiolates on gold by plasma cleaning. Surf. Sci. 2005, 595, 56–63. [CrossRef] 22. Hook, D.A.; Olhausen, J.A.; Krim, J.; Dugger, M.T. Evaluation of Oxygen Plasma and UV Ozone Methods for Cleaning of Occluded Areas in MEMS Devices. J. Microelectromechanical Syst. 2010, 19, 1292–1298. [CrossRef] 23. Peng, H.Y.; Devarajan, M.; Lee, T.T. Comparison of Argon and Oxygen Plasma Treatments on LED Chip Bond Pad for Wire Bond Application. Int. J. Sci. Eng. Res. 2014, 5, 908–912. 24. Peng, H.Y.; Devarajan, M.; Lee, T.T.; Lacey, D. Investigation of oxygen followed by argon plasma treatment on LED chip bond pad for wire bond application. Solder. Surf. Mt. Technol. 2015, 27, 129–136. [CrossRef] 25. Yamamoto, M.; Higurashi, E.; Suga, T.; Sawada, R.; Itoh, T. Properties of various plasma surface treatments for low-temperature Au–Au bonding. Jpn. J. Appl. Phys. 2018, 57, 04FC12. [CrossRef] 26. Koo, J.-M.; Lee, J.-B.; Moon, Y.; Moon, W.-C.; Jung, S.-B. Atmospheric pressure plasma cleaning of gold flip chip bump for ultrasonic flip chip bonding. J. Phys. Conf. Ser. 2008, 100, 012034. [CrossRef] 27. Maszara, W.P.; Goetz, G.; Caviglia, A.; McKitterick, J.B. Bonding of silicon wafers for silicon-on-insulator. J. Appl. Phys. 1988, 64, 4943–4950. [CrossRef] 28. Vallin, Ö.; Jonsson, K.; Lindberg, U. Adhesion quantification methods for wafer bonding. Mater. Sci. Eng. R Reports 2005, 50, 109–165. [CrossRef] 29. Brantley, W.A. Calculated elastic constants for stress problems associated with semiconductor devices. J. Appl. Phys. 1973, 44, 534–535. [CrossRef] 30. Fournel, F.; Continni, L.; Morales, C.; Da Fonseca, J.; Moriceau, H.; Rieutord, F.; Barthelemy, A.; Radu, I. Measurement of bonding energy in an anhydrous nitrogen atmosphere and its application to silicon direct bonding technology. J. Appl. Phys. 2012, 111, 104907. [CrossRef] 31. Bagdahn, J.; Petzold, M.; Reiche, M.; Gutjahr, K. Characterization of directly bonded silicon wafers by means of the double cantilever crack opening method. In Proceedings of the Fourth International Symposium on Semiconductor Wafer Bonding: Science, Technology, and Applications, Electrochemical Society Proceedings Vol. PV97, Paris, France, 30 August–5 September 1997; Volume 36, pp. 291–298. 32. Smith, T. The hydrophilic nature of a clean gold surface. J. Colloid Interface Sci. 1980, 75, 51–55. [CrossRef] 33. Zhu, J.; Zangari, G.; Reed, M.L. Tailoring the wetting properties of surface-modified nanostructured gold films. J. Phys. Chem. C 2011, 115, 17097–17101. [CrossRef] Micromachines 2019, 10, 119 12 of 12

34. Mezey, L.Z.; Giber, J. The surface free energies of solid chemical elements: Calculation from internal free enthalpies of atomization. Jpn. J. Appl. Phys. 1982, 21, 1569–1571. [CrossRef] 35. Tong, Q.-Y.; Kim, W.J.; Lee, T.-H.; Gösele, U. Low vacuum wafer bonding. Electrochem. Solid-State Lett. 1999, 1, 52. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).