Oxidation of Al–Au intermetallics and its consequences ...The study of oxidation of intermetallics...

Oxidation of Al–Au intermetallics and its consequencesstudied by x-ray photoelectron spectroscopy

T. Sritharana)

School of Materials Science and Engineering, Nanyang Technological University,Singapore 639798, Singapore

Y.B. LiSchool of Materials Science and Engineering, Nanyang Technological University,Singapore 639798, Singapore; and Center for Composite Materials, School of Astronautics,Harbin Institute of Technology, Harbin 150001, People’s Republic of China

C. XuSchool of Materials Science and Engineering, Nanyang Technological University,Singapore 639798, Singapore

S. Zhangb)

School of Mechanical and Aerospace Engineering, Nanyang Technological University,Singapore 639798, Singapore

(Received 15 November 2007; accepted 30 January 2008)

Three common Al–Au intermetallics, Al2Au, AlAu2, and AlAu4, were oxidized in theair and characterized using x-ray photoelectron spectroscopy in terms of the elementalchemical state. It was found that there is an increasing trend of oxidation in theseintermetallics as the Au content increases. AlAu4 shows the greatest tendency tooxidize with two extra peaks appearing on the Au 4f spectra after long exposure timein air. The surface of AlAu2, although fully oxidized, reveals only one Au 4f peak shiftas depth increases. Al2Au was the least oxidizing compound, and the oxide is thin.The binding energies of Al 2p and Au 4f peaks were measured and reported. The Auatoms trapped in the oxide layers exhibit higher binding energy emissions comparedto those of elemental Au.

I. INTRODUCTION

Au wire ball bonding to Al pads is a common inter-connection method used in the microelectronics industry.A good Al–Au bond inevitably contains intermetallics atthe interface whose composition, thickness, and stabilityplay a key role in the reliability of the joint. Many re-searchers have studied the formation and growth of in-termetallics at Al–Au interfaces for decades; the worksby Campisano et al.1 and Majni et al.2 are frequentlycited. The Al–Au alloy system shows many possible in-termetallic phases but only some have been detected ineach study and, even then, the reports are not in agree-ment. One phase that forms initially was found to trans-form to other phases on continued diffusion-driven reac-tion. Such interface transformation has been a fascinating

topic for scientific investigation, and experimentalistshave predominantly used scanning electron microscopy(SEM) and x-ray diffraction (XRD) to examine them.Transmission electron microscopy (TEM) has beenadopted in some recent reports.3–6 The intermetallics ob-served at the interface vary with the relative amounts ofAu and Al available and also on the growth conditions.The phases Al2Au, AlAu, AlAu2, Al2Au5, and AlAu4 arereported to form in different combinations by variousstudies.

The bond reliability in these joints is thought to becontrolled by Kerkendall void formation,7–9 resultingfrom unbalanced interdiffusion during the annealing ordue to volume shrinkage that accompanies intermetallicformation and growth.7,10,11 Recently, Karpel et al.5,6

have reported that voids exist even in the as-formedbonds, which was attributed to solidification shrinkagethat occurs when the molten Au ball solidifies. Withincreasing integration and miniaturization, the wirediameter and the metallization thickness have becomeextremely small, and consequently, it has become evenmore critical to understand the connection betweeninterface quality and microstructure. Besides those

a)Address all correspondence to this author.e-mail: [email protected]

b)This author was an editor of this journal during the review anddecision stage. For the JMR policy on review and publication ofmanuscripts authored by editors, please refer to http://www.mrs.org/publications/jmr/policy.html

DOI: 10.1557/JMR.2008.0161

J. Mater. Res., Vol. 23, No. 5, May 2008 © 2008 Materials Research Society 1371

failure mechanisms reported in the literature, one mecha-nism that could purportedly cause failure is oxidation/corrosion of the intermetallic phases at the joint interface.The oxidation characteristics of Al–Au intermetallicshave not been examined. Recently, Piao et al.12–15 stud-ied the room-temperature oxidation behavior of Al–Authin films containing Al2Au and AlAu2 at the interface.They report that oxidation of these phases occurs byAl2O3 formation with consequent Au enrichment associ-ated with phase transformation from Al-rich to Au-richphases. Accelerated oxidation could occur when wirebonds are subjected to high temperatures in the range175–200 °C during the packaging process, particularly atthe bond periphery where the intermetallics are exposedto the ambient until encapsulation. Although, such phasetransformations may be harmless by themselves, they areaccompanied by volume changes and interdiffusion,which could produce harmful internal stresses or voids.Therefore, characterization of the intermetallic oxidationat elevated temperatures will be beneficial to assess jointreliability.

The study of oxidation of intermetallics at the actualAl–Au interfaces in wire bonds could be extremelydifficult. Because information on the oxidation behaviorof even the bulk Al–Au intermetallics is very limited, itwas decided to examine some common phases reportedin wire bonds, in single phase form, in a systematicmanner. This report details the results obtained in a studyof three intermetallic phases, synthesized and oxi-dized under controlled conditions, at temperatures typi-cally encountered in an electronics packaging process,and investigated by x-ray photoelectron spectroscopy(XPS).

II. EXPERIMENTAL DETAILS

Approximately 1.5 g of the intermetallics Al2Au,AlAu2, and AlAu4 were produced by arc melting stoi-choimetric amounts of 99.99% pure Au and Al granulesin an argon atmosphere. The samples were repeatedlyflipped over and remelted to ensure homogeneity. Partsof the final samples were ground to powder, and XRDwas used to confirm that the phases obtained are indeedthe intended intermetallic without any residuals or impu-rities. In addition, the melting points of the Al2Au,AlAu2, and AlAu4 samples measured by a differen-tial scanning calorimeter are 1055, 627, and 526 °C, re-spectively, which also match values quoted in the litera-ture.16 Hence, the samples are indeed pure intermetallics.The mass change of solid intermetallics was measuredwith a thermogravimetric analyzer (Mettler Toledo TGA/SDTA, Switzerland).

Nine specimens of size 4 mm × 2 mm were cut fromeach intermetallic, and one convenient face was polisheddown to 1-�m diamond paste surface finish and ultra-

sonically cleaned. The polished samples were heated at175 °C in an oven in air for durations of 0.25, 0.5, 1, 2,4, 8, 16, and 24 h with care to avoid contaminationand contact with the polished surface. XPS measure-ments were conducted in a Kratos AXIS Ultra spectrom-eter (UK) with a monochromatized Al K� x-ray source(1486.71 eV) operated at a reduced power of 150 W(15 kV and 10 mA). The base pressure in the analysischamber was 2.66 × 10−7 Pa. The core-level spectra wereobtained in Al 2p and Au 4f regions at a photoelectrontake-off angle of 90° measured with respect to the samplesurface and were recorded in 0.1 eV steps with a passenergy of 40 eV. The binding energy scale of the XPSspectrum was calibrated with the C 1s peak (neutralC–C peak at 284.5 eV). The depth profiling chemicalstates were analyzed on the same equipment for the 24 hoxidized samples only. An argon ion gun with acceler-ated voltage of 4 kV and filament current of 15 mA wasused to etch the samples at a gas pressure of 6.65 ×10−6 Pa. Ion bombardment was performed at an incidentangle of 45° to the surface normal. The Al 2p and Au 4fspectra were collected after each etching cycle of 3 min.Depth profiling was done until the oxygen peak reachedundetectable level.

III. RESULTS

A. Al2Au

The Al 2p and Au 4f spectra obtained on Al2Ausample surfaces oxidized for different times, up to 24 h,are shown in Figs. 1(a) and 1(b), respectively. The data inFig. 1(a) can be resolved into two component peaks atbinding energies (b.e.) of 75.8 and 73.5 eV. The reportedb.e. of oxidic Al peak matches well with 75.8 eV, and thusthis could be assigned to Al2O3. The other componentat 73.5 eV is near to that of Al species in Al2Au,12–15 andits intensity gradually reduces to a weak tail as oxidationis prolonged. Since the oxidic peak is the dominant oneeven in the as-polished sample, Al2Au is oxidized evenin the as-polished state.

In the Au 4f spectra shown in Fig. 1(b), two spin-orbitsplit components of the Au peak are clearly visible in allsamples at b.e. values of 89.1 and 85.4 eV, which cor-respond to the Au 4f5/2 and Au 4f7/2 values of 89.4 and85.7 eV reported in Al2Au by Fuggle et al.17 The b.e. andthe intensity of the component peaks do not seem tochange with increasing oxidation time, and no additionalcomponent peaks are required to achieve a good fit to theexperimental data. This implies that a strong signal isreceived from Al2Au phase in all samples although clearevidence of oxidation is also detected in Fig. 1(a).

The depth profiling curves (not shown) of Au in Al2Auat different oxidation durations confirm that there is nochange in binding energy indicating that the oxidation

T. Sritharan et al.: Oxidation of Al–Au intermetallics and its consequences studied by x-ray photoelectron spectroscopy

J. Mater. Res., Vol. 23, No. 5, May 20081372

treatment does not influence the valence of Au. Thedepth profiling information obtained in the 24 h oxidizedAl2Au sample is shown in Fig. 2 for Al 2p peaks. Ashoulder is clear at the surface (0 min), at the low b.e. end

in Fig. 2, indicating that two component peaks are over-lapping. The smaller peak with lower b.e. gradually be-comes dominant as the depth increases while the peakdominating on the surface diminishes. A good fit to

FIG. 1. Curve-fitted experimental data and their deconvolution of (a) Al 2p and (b) Au 4f XPS spectra on the surface of Al2Au oxidized todifferent durations.


J. Mater. Res., Vol. 23, No. 5, May 2008 1373

experimental data could be obtained in the deconvolutionwith two component peaks at 75.8 and 73.5 eV, as shownin Fig. 2. These peaks were observed in Fig. 1(a), too,and were assigned to Al 2p in Al2O3 and Al2Au, respec-tively. The evolution of the relative intensities of thecomponent peaks is in good agreement with the presenceof an Al2O3 layer on Al2Au whereby the oxidic peak isdominant at the surface, and the Al2Au peak increases indominance with increase in depth.

Under these conditions, it is possible to estimate theremaining oxide layer thickness from the ratio of theintensities of the peaks using the Strohmeier equation18:

d �nm� = 2.8 ln�1.4Io

Im+ 1� , (1)

where Io and Im are the intensities (peak areas) of theoxide and metal photoelectron peaks, respectively, whichcan be readily obtained from the quantification param-eters list when deconvoluting the peaks using theCasaXPS software (version 2.2.99, Casa Software Ltd.,UK). The oxide layer thickness calculated using Eq. (1)at each point in the depth profiling is also indicated inFig. 2. Accordingly, Al2Au forms an oxide layer of thick-ness 10.8 nm approximately when heated in air for 24 hat 175 °C.

In contrast, the two spin-orbital split Au 4f peaks donot change with depth, implying that the Al2O3 layer of10.8 nm does not significantly change the electronic en-vironment of the Au atoms or absorb the Au photoelec-trons, because the surface and subsurface signals areidentical.

B. AlAu2

The Al 2p and Au 4f spectra for AlAu2 samples oxi-dized for up to 24 h are shown in Figs. 3(a) and 3(b),respectively. For all oxidation times, the data could befitted well with a single peak at the b.e. of 75.8 eV, whichcould be attributed to the oxidic Al peak. This is in con-trast to the observation in Fig. 1(a) for Al2Au where theexperimental data could be resolved into two componentpeaks corresponding to the oxide and the intermetallic.The absence of the Al 2p component of AlAu2 on thesurface implies a thicker oxide layer in comparison toAl2Au. The Au 4f spectra in Fig. 3(b) show that twopeaks at the positions of 88.7 eV and 85.0 eV generate aperfect deconvolution to the data for all oxidation timeswithout any shift in b.e. This implies that the Au 4f signalof AlAu2 is received clearly in all samples despite thepresence of a thicker oxide layer here.

The spectra obtained in depth profiling are shown inFigs. 4(a) and 4(b) for Al 2p and Au 4f, respectively. Thescan on the surface (0 min) in Fig. 4(a) generates a singlepeak which can be assigned to the oxidic component of

FIG. 2. Curve-fitted experimental data and their deconvolution of Al2p depth profiling spectra of 24 h oxidized Al2Au. Note the shift of theb.e. from the oxidic value to the intermetallic phase.



Al at 75.8 eV after deconvolution. A small tail appears atthe low b.e. side in 3 min of etching, which evolves intoa significant peak at 73.9 eV after 15 min etching. Thisnew peak could be assigned to Al in AlAu2. The oxide

peak is weakened with increase of depth but is still de-tectable even after 90 min of etching, while the AlAu2 Al2p peak strengthens, signifying penetration through theoxide layer and reach of the underlying intermetallic. The

FIG. 3. Oxidized curve-fitted experimental data of (a) Al 2p and (b) Au 4f XPS spectra of AlAu2 oxidized to different durations. There are nooverlapping peaks.



remaining oxide layer thickness computed using Eq. (1)is also shown in the figure, except for the spectrum on theoriginal surface as the AlAu2 component is not detected.Hence, the full thickness of the oxide could not be com-

puted but is probably higher than in Al2Au since the alloypeak was obscured.

The evolution of Au 4f spectra shown in Fig. 4(b)manifests a chemical shift for the two peaks with increase

FIG. 4. Deconvolution of (a) Al 2p and (b) Au 4f depth profiling spectra of 24 h oxidized AlAu2. Note the shift in the peaks in (b) at the surface.



in depth. The b.e. decreases from 88.8 to 88.2 eV for Au4f5/2 and 85.1 to 84.5 eV for Au 4f7/2. Note that the shiftin b.e. is systematic and is higher than the resolution ofthe technique, which is typically ±0.2 eV. Piao12–15 re-ported the b.e. of the two Au 4f peaks in AlAu2 as88.6 and 84.9 eV, which are in good agreement with theb.e. values obtained in this study on the initial surface ofAlAu2. However, the surface is oxidized, and our resultsclearly show a shift in b.e. to higher values at the surface.Therefore, the lower b.e. values obtained after depth pro-filing should give a better estimate of the true b.e. inAlAu2. Consequently, we attribute 88.2 and 84.5 eV Au4f emissions in AlAu2. The increase in b.e. near thesurface signifies that the electronic environment aroundAu atoms is changed because of the thicker oxide layeron AlAu2.

C. AlAu4

Al 2p spectra for AlAu4 samples oxidized up to 24 h(not shown here), are similar to those of AlAu2 inFig. 3(a). Only the oxidic Al component at b.e. of 75.7eV could be detected at the surface of the samples indi-cating a thick oxide layer. The emission from the Alspecies in AlAu4 was obscured totally, indicating that theoxide on this compound is thick even at the as-polishedstate.

The Au 4f spectra obtained on the surface are shown inFig. 5. In the as-polished state, two distinct peaks at 87.8and 84.2 eV, corresponding to Au 4f5/2 and Au 4f7/2,respectively, fit the experimental data very well. Becausevalues of b.e. in AlAu4 are not reported in the literaturefor comparison and because the Al 2p spectrum clearlyindicates a thick oxide layer on these samples, we cannotattribute these b.e. values to those of Au 4f doublet inAlAu4. With increasing oxidation duration, it is clear thatadditional overlapping peaks become necessary to de-convolute the data. Each of the Au 4f doublets could beresolved into two overlapping components; one pair at88.8 and 87.8 eV and another pair at 85.2 and 84.2 eV.Thus, with increasing oxidation, the two additional peaksappear at the higher energy side of the Au 4f doubletobserved in the as-polished sample. There is no evidencefor a similar high b.e. component in the as-polished Al 2pspectra and hence these additional peaks are not artifactsof differential charging at surface regions. This addi-tional Au 4f doublet appears to represent an Au 4f emis-sion coming from a different chemical environment tothat in the as-polished sample.19 Since the presence of athick oxide layer is confirmed by the Al 2p spectra, thenew peaks at 88.8 and 85.2 eV must be emitted by Auatoms trapped in Al2O3, as was observed for the case ofAlAu2.

Depth profiling shows that the unetched surface scanshows only the oxidic Al 2p peak at 75.7 eV after de-convolution, as evident in Fig. 6(a). With increasing

FIG. 5. Curve-fitted experimental data and the deconvolution of Au 4fXPS spectra of AlAu4 oxidized for 24 h. Note the appearance andstrengthening of new extra peaks as the depth increases.



depth, a second peak appears as a shoulder at 15 minetching time and subsequently as a doublet at 33 min.After 15 min etching, the remaining oxide layer thick-ness, calculated using Eq. (1), is 5.1 nm. Comparison

with the Al2Au ion etching suggests that the initial oxidelayer is more than 10.8 nm thick. In 108 min, the oxidicpeak is still evident as a minor peak while the secondpeak is predominant. Deconvolution shows that the

FIG. 6. Curve-fitted experimental data and deconvolution of (a) Al 2p and (b) Au 4f depth profiling spectra of 24 h oxidized AlAu4 oxidized for24 h. Note the diminishing extra peaks in (b).



second peak should be at the b.e. of 73.7 eV for goodfitting to the data. Since this second peak was not de-tected at all at the surface, all surface Al atoms have beenoxidized to Al2O3. After 108 min etching, the photoelec-tron signal from the subsurface AlAu4 is received clearly,and thus the b.e. of Al 2p peak in AlAu4 can be taken asthat of the second peak of 73.7 eV.

The Au 4f spectra during depth profiling and theirdeconvolution are shown in Fig. 6(b). Each peak at thesurface could be resolved into two components; an over-lapping pair with b.e. of 88.8 and 87.8 eV and anotheroverlapping pair with values 85.2 and 84.2 eV. This issimilar to the spectra in Fig. 5(b) obtained at the surface.It is clear that the intensities of the two high energy peaksat 88.8 and 85.2 eV rapidly diminish in a short etchingtime of only 6 min. In the spectra for 33 and 108 minetched samples, deconvolution shows that single peaks at87.8 and 84.2 eV energies are adequate to fit the data.Having the additional high energy pair with a small in-tensity would not make any noticeable difference to thefit. Thus, the peaks at 87.8 and 84.2 eV could be attrib-uted to the Au 4f doublet in AlAu4 while the additionalpeaks at the binding energy of 88.8 and 85.2 eV could beattributed to a new emission from Au atoms trapped inthe thick Al2O3 layer.

D. Oxidation rate constant

Oxidation rate constant kp was calculated fromEq. (2)20:

��W�A�2 = kpt , (2)

where �W/A is the mass increase per unit surface areameasured by TGA, and t is time. The values obtained forthe three imtermetallics at three temperatures are shownin Table I along with that for pure Al for comparison.Note an increase in kp of all intermetallics compared toAl. The higher Au containing intermetallics show higherkp values, which implies faster oxidation.

IV. DISCUSSION

A. Binding energies of Al 2p and Au 4f peaks

Table II summarizes the b.e. values measured in thisstudy and compares it with those reported in the litera-ture. Our measurements are in good agreement with re-ported values in Al2Au and in Al2O3. In AlAu2, on theother hand, the Au 4f b.e. determined in this study islower by about 0.4 eV than the values reported by Piaoet al.12–15 This could be attributed to some interferencefrom the oxide layer as we indeed obtained the samevalues at the surface, but on depth profiling, the lowervalues were obtained, which must be the native emis-sions from the intermetallic phase. The effect of the

oxide layer in increasing the b.e. was more apparent inAlAu4 where distinct new peaks were required to fit theexperimental data. These new peaks were at the higherb.e. side of the peaks attributed to Au 4f in AlAu4, astabulated in Table II. No reports are available on meas-urement of b.e. in AlAu4, and thus we are not in a posi-tion to authenticate our results, but the depth profilingstudy is able to distinguish the native peaks of the inter-metallic from the additional emissions, which are be-lieved to be from the Au atoms trapped in the oxide layer.It is thought that a similar phenomenon also occurs inAlAu2, which causes a shift in b.e. values of the Au 4fpeaks when measured at the oxidized surface. Since itsoxide layer is thinner, it was not possible to distinguishthis effect as clearly as in AlAu4. In Al2Au, however, theoxide layer is very thin for any embedded Au atoms tocause a shift in the native Au 4f peaks of the phase, and,consequently, measurements made at the surface give anaccurate value for the b.e. representative of the phase.

B. Au atoms in the oxide layer of AlAu4

AlAu4 has not been investigated as much as Al2Au andAlAu2 but is likely to be the most predominant phase inwire bonds because of the presence of excessive Au inreal joints. XPS b.e. information for this phase is alsoscarce. The presence of two extra peaks at b.e. of 85.2and 88.8 eV in the Au 4f spectra of AlAu4 has not beenpreviously reported. The known (bulk) oxidized speciesof Au, Au2O3, has a Au 4f7/2 b.e. near 85.9 eV and isnormally formed only under exceptional conditions likeexposure of Au to highly reactive ozone,21–24 atomicoxygen,25 or oxygen plasma.26 Given that AlAu4 is the

TABLE I. List of oxidation rate constant kp.

Material

kp (kg2 m−4 s−1)

200 °C 300 °C 400 °C

Al 6.60 × 10−10 1.60 × 10−9 5.80 × 10−9

Al2Au 9.90 × 10−10 4.50 × 10−9 8.10 × 10−9

AlAu2 1.49 × 10−9 5.17 × 10−9 1.11 × 10−8

AlAu4 2.97 × 10−9 1.33 × 10−8 3.93 × 10−8

TABLE II. Binding energy values measured in this study and com-pared with those reported in the literature.

Intermetallic

Al 2p b.e. (eV) Au 4f b.e. (eV)

This study Literature This study Literature

Al2Au 73.5 73.712–15 89.1, 85.4 89.4, 85.717

AlAu2 73.9 88.2, 84.5 88.6, 84.912–15

AlAu4 73.7 87.8, 84.2Al2O3 surface

oxide layer 75.7 to 75.8 75.8Au in AlAu4

surface oxide 88.8, 85.2



gold-richest phase in Al–Au phase diagram, oxidationof AlAu4 to Al2O3 will result in the emergence of asignificant amount of elemental Au. The reported bind-ing energies of elemental Au 4f peaks are 87.5 and83.8 eV.27,28 The new peaks observed at 88.8 and85.2 eV in this investigation are significantly higher thanthe binding energy of elemental Au 4f peaks and thusmust be emitted from a different, strongly electron-withdrawing chemical environment. As a thick Al oxideis present on the surface, this binding energy shift mayresult from isolated Au atoms or nanoclusters of Au sur-rounded by oxygen defects in the oxidized layer. Suchimpurity atom-vacancy combinations are now offered asa possible reason for the emergence of strong magneticresponse in Co- or Fe-doped TiO2 dilute magnetic semi-conductors.29,30 Hence, this phenomenon may be a com-mon occurrence in oxides containing alien cations assubstitutional defects leading to shifts in the b.e. of thecation.

The b.e. shifts are characteristic of the physical andchemical environment of the analyzed species while theelectronegativity determines the type of interaction be-tween the various atoms in the system. The interaction ofAu with alumina has been investigated elsewhere.31–36

Serrano et al. observed the formation of Al2Au at theAu–Al2O3 interface at room temperature when Au issputtered onto an alumina substrate. Upon heating, thisAl2Au disintegrated and the emerging elemental Au wasincorporated into Al2O3, resulting in a solid solution ofAu in Al2O3. Serrano theorized that the Au atoms couldcluster into nanoparticles in the oxide matrix.37 Nguyenet al. coated thin gold films on Al2O3 ceramic substrateand reported the initial formation of AlAu2 as well assubsequent appearance of Al2Au upon heating at 500 °Cfor 3 h.38 Furthermore, Atsuko et al. proved that Audiffused through Al2O3 in an Au–Al2O3–Al sandwich toform Al2Au intermetallic at the Al interface.39 Thus, dis-solution of Au in Al2O3 has been suspected in somestudies, but its characteristic XPS emissions have notbeen reported. This study, while giving a strong indica-tion of dissolution of Au in Al2O3, also measures theAu 4f b.e. of such dissolved species. This value will beaffected by electronic interaction between Au and theneighboring atoms or vacancies. The b.e. of Au 4f inAl2O3 determined in this study is higher than that re-ported for elemental Au, and that attributed here forAlAu4. This implies that Au, under these circumstances,has a tendency to donate electrons to the neighboringatomic sites, implying the presence of O vacancies. SomeAu–O bonds may exist to stabilize the Au atoms, al-though no explicit evidence is available for any specificAu oxide species in our investigations. The Au 4f7/2

binding energy of 85.2 eV is less than the reported valueof 85.7 eV for Au2O3

40 suggesting that any such bondsare not as ionic as the bulk Au2O3 oxide.

C. Oxidation reaction in Al–Au intermetallics

Two processes could take place in oxidation of Al–Auintermetallics: the oxidation of Al and the phase trans-formation due to depletion of Al. As indicated in previ-ous XPS studies on Al–Au thin film alloys,12–15 Al2Autransforms to AlAu2 during exposure to atmosphere evenat room temperature. This process could be expressed by,

8Al2Au + 9O2 → 6Al2O3 + 4AlAu2 . (3)

Such transformation was however not observed in thepresent study of Al2Au. In Fig. 1(a), only Al2O3 is de-tected on the surface of Al2Au sample without any signof AlAu2. Then, the oxidation reaction of Eq. (3) has tobe re-written as

2Al2Au + 3O2 → 2Al2O3 + 2Au . (4)

The presence of elemental Au also could not be de-tected as no corresponding peak shift in Au 4f was ob-served for up to 24 h oxidation. Figure 2(a) shows thatthe oxide on Al2Au is very thin (about 10 nm), thus anyAu that is formed might be in minute amount below thesensitivity of the technique.

The AlAu2 oxidation reaction may lead to the forma-tion of more Au-rich phases such as AlAu4 or Al2Au5,but again such phases were not detected according tothe b.e. of the Au 4f peaks observed. The observed bind-ing energy shift in Au spectra upon depth profiling inFig. 4(b) suggests there is only a change of chemicalenvironment around the atoms due to oxide thickeninginstead of the formation of new phases. The oxide thick-ness on AlAu4 is the largest among the intermetallicsstudied here, causing considerable depletion of Al andthe appearance of extra Au peaks attributed to elementalAu trapped in the oxide. Hence, the reaction of AlAu4

oxidation could be given as

4AlAu4 + 3O2 → 2Al2O3 + 16Au . (5)

This reaction implies that 16 mol Au is produced for2 mol Al2O3. Hence, a significant amount of Au couldremain trapped in the oxide when AlAu4 oxidizes, com-pared to the other two intermetallics. In addition, thethicker oxide layer in this phase also obscures photo-electron emission from the pure AlAu4 beneath, makingphotoelectron emission from Au in the oxide layer easierto detect.

D. Oxidation tendency of Al–Au intermetallics

Note that the kp of AlAu4 is the largest among theintermetallics at every temperature, and the effect of in-creasing Au content is to increase kp. This implies thathigher Au containing intermetallics oxidize more readily.These results are in agreement with XPS investigations.



V. CONCLUSIONS

The following can be concluded from this detailedXPS study of oxidation of the three bulk intermetallicphases: Al2Au, AlAu2, and AlAu4.

The degree of oxidation increases in the order Al2Au <AlAu2 < AlAu4. The oxide layer thickness in Al2Au isestimated to be 10 nm. AlAu4 has the greatest tendencyto be oxidized among the intermetallics tested, suggest-ing that the oxidation of AlAu4 in wire bond could be aserious reliability problem

The binding energies were measured for the Al and Auspecies in the intermetallic phases and are shown inTable II. A binding energy shift was observed in the Au4f peak of AlAu2 on its oxidized surface, while at higherbinding energy, additional peaks were discovered on theoxidized surface of AlAu4. These were attributed toemissions from Au atoms dissolved in the oxide layer,which were at a higher binding energy than elementalAu. Such abnormality was not noted in Al2Au.

The binding energy of the two spin-orbit split com-ponents of Au 4f in AlAu2 are determined as 88.2 and84.5 eV, which are slightly lower than previously re-ported values. It is proposed that measuring the bindingenergy on the oxidized surface leads to erroneouslyhigher values because of emission from Au atoms en-trapped in the oxide layer.

The binding energies of the two spin-orbit split com-ponents of Au 4f in AlAu4 are determined as 88.8 and85.2 eV. Binding energies of these peaks are reported forthe first time in the literature.

ACKNOWLEDGMENTS

The authors acknowledge Prof. Roger Smart andDr. Madhavi Srinivasan for their invaluable suggestionson the manuscript.

REFERENCES

1. S.U. Campisano, G. Foti, E. Rimini, S.S. Lau, and J.W. Mayer:Kinetics of phase formulation in Au–Al thin films. Philos. Mag.31, 903 (1975).

2. G. Majni, C. Nobili, G. Ottaviani, and M. Costato: Gold-aluminum thin-film interactions and compound formation.J. Appl. Phys. 52, 4047 (1981).

3. C. Xu, T. Sritharan, and S.G. Mhaisalkar: Thin film alumi-num–gold interface interactions. Scripta Mater. 56, 549 (2007).

4. C. Xu, T. Sritharan, and S.G. Mhaisalkar: Interface transforma-tions in thin film aluminum-gold diffusion couples. Thin SolidFilms 515, 5454 (2007).

5. A. Karpel, G. Gur, Z. Atzmon, and W.D. Kaplan: TEM micro-structural analysis of as-bonded Al–Au wire-bonds. J. Mater. Sci.42, 2334 (2007).

6. A. Karpel, G. Gur, Z. Atzmon, and W.D. Kaplan: Microstructuralevolution of gold-aluminum wire-bonds. J. Mater. Sci. 42, 2347(2007).

7. X. Zhang and T.Y. Tee: Numerical and experimental correlation

of high temperature reliability of gold wire bonding to inter-metallics (Au/Al) uniformity. Thin Solid Films 504, 355 (2006).

8. H.S. Chang, K.C. Hsieh, T. Martens, and A. Yang: Wire-bondvoid formation during high temperature aging. IEEE Trans.Compon. Packag. Technol. 27, 155 (2004).

9. H.J. Ji, M.Y. Li, C.Q. Wang, H.S. Bang, and H.S. Bang:Comparison of interface evolution of ultrasonic aluminum andgold wire wedge bonds during thermal aging. Mater. Sci. Eng., A447, 111 (2007).

10. K. Hiroshi: Volume change due to intermetallic compound for-mation at the Al–Au bond in semiconductor devices. Jpn. J. Appl.Phys. 25, 934 (1986).

11. N.J. Noolu, N.M. Murdeshwar, K.J. Ely, J.C. Lippold, andW.A. Baeslack: Degradation and failure mechanisms in thermallyexposed Au–Al ball bonds. J. Mater. Res. 19, 1374 (2004).

12. H. Piao, M.S. Fuller, D. Miller, and N.S. McIntyre: A study of thinfilm Au–Al alloy oxidation in ambient air by x-ray photoelectronspectroscopy (XPS), x-ray absorption near edge structure(XANES), and secondary-ion mass spectrometry (SIMS). Appl.Surf. Sci. 187, 266 (2002).

13. H. Piao and N.S. McIntyre: High resolution XPS studies of thinfilm gold–aluminum alloy structures. Surf. Sci. 421, 171 (1999).

14. H. Piao and N.S. McIntyre: High-resolution valence band XPSstudies of thin film Au–Al alloys. J Electron Spectrosc. Relat.Phenom. 119, 29 (2001).

15. H. Piao, N.S. McIntyre, and G. Beamson: Electronic structures ofAu–Al thin-film alloys by high-energy XPS and XANES. J Elec-tron Spectrosc. Relat. Phenom. 125, 35 (2002).

16. Constitution of Binary Alloys, edited by M. Hansen andK. Anderko (McGraw-Hill, New York, 1958), p. 68.

17. J.C. Fuggle, E. Kallne, L.M. Watson, and D.J. Fabian: Electronicstructure of aluminum and aluminum-noble-metal alloys studiedby soft-x-ray and x-ray photoelectron spectroscopies. Phys. Rev. B16, 750 (1977).

18. B.R. Strohmeier: An ESCA method for determining the oxidethickness on aluminum alloys. Surf. Interface Anal. 15, 51 (1990).

19. B. Koslowski, H.G. Boyen, C. Wilderotter, and G. Kastle: Oxi-dation of preferentially (111)-oriented Au films in an oxygenplasma investigated by scanning tunneling microscopy and photo-electron spectroscopy. Surf. Sci. 475, 1 (2001).

20. P. Perez, M.F. Lopez, J.A. Jimenez, and P. Adeva: Oxidationbehaviour of Al-alloyed ZrSi2 at 700 °C. Intermetallics 8, 1393(2000).

21. H.G. Boyen, G. Kastle, F. Weigl, and B. Koslowski: Oxidation-resistant gold-55 clusters. Science 297, 1533 (2002).

22. D.H. Parker and B.E. Koel: Chemisorption of high coverages ofatomic oxygen on the Pt(111), Pd(111), and Au(111) surfaces.J. Vac. Sci. Technol., A 8, 2585 (1990).

23. D.E. King: Oxidation of gold by ultraviolet light and ozone at25 °C. J. Vac. Sci. Technol., A 13, 1247 (1995).

24. N. Saliba, D.H. Parker, and B.E. Koel: Adsorption of oxygen onAu(111) by exposure to ozone. Surf. Sci. 410, 270 (1998).

25. D.D. Eley and P.B. Moore: Adsorption of oxygen on gold. Surf.Sci. 76, 599 (1978).

26. J.J. Pireaux, M. Liehr, P.A. Thiry, and J.P. Delrue: Electron spec-troscopic characterization of oxygen adsorption on gold surfaces:II. Production of gold oxide in oxygen DC reactive sputtering.Surf. Sci. 141, 221 (1984).

27. K. Juodkazis, J. Juodkazyte, V. Jasulaitiene, and A. Lukinskas:XPS studies on the gold oxide surface layer formation. Electro-chem. Comm. 2, 503 (2000).

28. C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder, andG.E. Muilenberg: Handbook of X-ray Photoelectron Spectroscopy(Perkin-Elmer, Minneapolis, MN, 1978), p. 190.

29. J.E. Jaffe, T.C. Drouban, and S.A. Chambers: Oxygen vacancies



and ferromagnetism in CoxTi1–xO2–x–y. J. Appl. Phys. 97, 073908(2005).

30. J. Chen, P. Rulis, L. Ouyang, S. Satpathy, and W.Y. Ching:Vacancy-enhanced ferromagnetism in Fe-doped rutile TiO2. Phys.Rev. B 74, 235207 (2006).

31. J.W. Park, A.J. Pedraza, and W.R. Allen: The interface betweensputter-deposited gold thin films and ion-bombarded sapphiresubstrates. Appl. Surf. Sci. 103, 39 (1996).

32. P. Bruesch, R. Kotz, H. Neff, and L. Pietronero: Vibrational prop-erties of Al2O3 films on gold, aluminum and silicon. Phys. Rev. B29, 4691 (1984).

33. M. Gillet, A.A. Mohammad, K. Masek, and E. Gillet: Influence ofsurface structure on the growth of Au on �–Al2O3 (-1012). ThinSolid Films 374, 134 (2002).

34. J.G. Serrano and U. Pal: Synthesis and characterization of Aunanoparticles in Al2O3 matrix. Int. J. Hydrogen Energy 28, 637(2003).

35. T. Ishizaka, S. Muto, and Y. Kurokawa: Nonlinear optical and

XPS properties of Au and Ag nanometer-size particle-doped alu-mina films prepared by the sol-gel method. Opt. Commun. 190,385 (2001).

36. N. Mallya and V. Young: Angular distribution XPS studies onrecoil-implanted substrates: Au in Al2O3. J. Electron Spectrosc.Relat. Phenom. 24, 43 (1981).

37. J.G. Serrano, A.G. Galindo, and U. Pal: Au-Al2O3 nanocompos-ites: XPS and FTIR spectroscopic studies. Sol. Energy Mater. Sol.Cells 82, 291 (2004).

38. T.P. Nguyen, J. Ip, P.L. Rendu, and A. Lahmar: Improved adhe-sion of gold coatings on ceramic substrates by thermal treatment.Surf. Coat. Technol. 141, 108 (2001).

39. T. Atsuko, M. Kaoru, U. Tomohiro, and T. Kohei: Effects ofAl2O3 films on the reliability of Au/Al joint. Nippon Steel Tech-nical Report 72, 95 (1997).

40. D. Briggs and M.P. Seah: Practical Surface Analysis, Vol. 1, 2nded. (Wiley, New York, 1993).



Oxidation of Al–Au intermetallics and its consequences ...The study of oxidation of intermetallics...

Documents

Transcript of Oxidation of Al–Au intermetallics and its consequences ...The study of oxidation of intermetallics...