Au trilayer metallization scheme

D IAMOND AND RELATED MATERIALS Diamond and Related Materials 5 (1996) 1450-1456

ELSEVIER

Ohmic contacts to semiconducting diamond using a Ti/Pt/Au trilayer metallization scheme H . A . H o f f a, G . L . W a y t e n a a, C . L . V o l d a, J.S. S u e h l e u, I.P. I s a a c s o n °, M . L . R e b b e r t c, D.I. M a °'*, K . H a r r i s d a Naval Research Laboratory, Washington, DC 20375-5343, USA b National Institutes of Standards and Technology (NIST), Gaithersburg, MD 20899, USA ° NanoeIectronics Processing Facility Naval Research Laboratory, Washington, DC 20375, USA a Harris Diamond Corporation, Mount Arlington, N J 07865, USA

Received 9 February 1996; accepted 12 June 1996

Abstract

Ohmic contacts have been fabricated on a naturally occurring type IIb diamond crystal using an annealed Ti/Pt/Au trilayer metallization where the Pt served successfully as a barrier to Ti diffusion into the Au capping layer. However, a specific contact resistance could not be reliably determined using transmission line model measurements. Auger microanalysis revealed the presence of Ti on the diamond surface near the ohmic contact pads. The most likely origin of the Ti on the diamond surface was determined to be lateral diffusion from beneath the contact pads. This would have produced a nonuniform concentration of Ti across the diamond surface which, in turn, would have affected the diamond sheet resistance in a complicated way. Keywords: Surface characterization; Diffusion; Electrical properties; Diamond single crystats

1. Introduction

Diamond is currently being investigated for use in solid state devices for high power and high frequency applications that can operate reliably in a radiation environment, and at high temperatures (400-600°C) without a cooling system. Diamond devices that operate at elevated temperatures, including diodes, radiation sensors, thermistors and transistors, have already been demonstrated [ 1-5]. However, several important device components must be optimized. One very important component is the ohmic contact. A simple process for forming low-resistance, strongly adherent ohmic contacts on diamond has been developed [ 6 - 8 ] . Briefly, the process begins with the deposition of a thin layer of carbide forming metal (usually Ti, Ta, Mo or V) on diamond, followed by the deposition of a An cap layer to protect the metal from corrosion. After deposition, the contacts are annealed at high temperatures in a purified H2 ambient or under vacuum, which causes the transition metal to react With the diamond. Analysis of the resulting contacts at the metal-diamond * Correspondingauthor. 0925-9635/96/$15.00© 1996ElsevierScienceS.A. All rights reserved PII S0925-9635 (96) 00566-3

interface have revealed the formation of carbide precipitates, which are believed to produce the strong adherence and low contact resistance [9]. Contacts with resistances on the order of 10 -s W cm 2 have been fabricated using annealed Ti/Au bilayers [ t 0 ] . However, it has been shown using X-ray photoelectron spectroscopy (XPS) depth profiling and other techniques that a considerable amount of intermixing between the individual Ti and Au layers occurs after an anneal at a high enough temperature to cause the contact to become ohmic [ 10,11 ]. More significantly, Ti has been detected on the contact surface [10,11], where, as a result of partial removal from the interface and subsequent deposition on the Au surface with the formation of a titanium oxide, the contact resistance has been observed to increase [11]. To eliminate Ti diffusion through the Au layer to the Au surface, we decided to use a material between the Ti and Au that would act as a diffusion barrier while being a slow diffuser in Au and/or non-reactive to oxygen. Platinum is a reasonable choice because it does not readily form an oxide, Ti diffuses slowly through Pt as compared to other metals (e.g., Pd) [12], and it has been used successfully in Si and GaAs technology [11,13,14].

G.L. Waytena et aL/Diamond and Related Materials 5 (1996) 1450-1456

1451

The Ti/Pt/Au trilayer system has been used on Si [ 11 ], where it was determined that the Pt prevented the interdiffusion between Ti and Au, but only up to temperatures of 500°C. Ti/Pt/Au has also been used for metallization on diamond [15], but these contacts were not annealed, so a metal carbide was not formed and the contact was not ohmic. Our work represents the first attempt at using this trilayer metallization scheme for ohmic contacts on diamond. In this work, Pt was found to act as a diffusion barrier at annealing temperatures of 900 °C.

2. Experimental details 2.1. Sample preparation

A 4 x 4 x 0.25 mm < 100 > p-type semiconducting natural type IIb diamond from Harris Diamond Corporation was used as substrate material. This diamond is semiconducting because it is naturally doped with boron. Before metallization, the substrate was cleaned using a four acid etch process which consisted of (1) boiling for 1 h in a 1 : 1 : 1 H20:CrO3:H2SO4 acid solution to remove non-diamond carbon [16-1, (2) boiling for 1 h in a 3:1 HCI:HNO 3 acid solution to remove metals [ 16], (3) boiling for 1 h in a 3:2 H2SO4:HNO3 acid solution to remove any remaining non-diamond carbon [16], and (4) a 1 h bath at room temperature in 1:1:1 HF:HNO3:acetic acid solution [17]. The substrate was rinsed in distilled water after each acid bath, then rinsed in ethanol and blown dry with N2 after the final bath. Standard photolithography techniques and a Shockley transmission line method ( T L M ) [18] mask were used to pattern a series of rectangular contact pads with dimensions 300 x 120 gm separated from each other by distances of 2, 5, 10, 20, 30, 40, 50, and 60 gm. A Hitachi $800 scanning electron microscope (SEM) was used to take micrographs of the T L M pattern, with a magnification standard. The actual contact spacing was measured on the micrographs. A part of the Shockley T L M pattern is shown in the SEM image of Fig. 1. Notice that a few of the contacts in the series at the left top side of the micrograph are missing or the edges of the contact are folded over. This may have occurred because the optic patterning did not completely open the windows seen in the left corner of the micrograph. As a result, the deposited metal did not contact the diamond. After liftoff, the contacts were either lost or their corners were folded over. Metallization was accomplished by electron beam evaporation of 5-10 nm of Ti, followed by 50 nm of Pt, then 500 nm of Au in an ion-pumped, U H V evaporating chamber. The sample was annealed in a tube furnace in

Fig. 1. Scanningelectronmicrographof some of the coIumns of contact pads from which the transmissionline model measurementsweremade. vacuum (5 x 10 -s Tort) for 120 rain at incremental temperatures of 700, 800 and 900°C. 2.2. Characterization

After each incremental time at temperature, the sample was removed from the furnace, and the current-voltage (I-V) characteristics were checked. Data for the I - V measurements were collected using a Keithley d.c. voltage source and microammeter controlled through I E E E interfaces by an IBM PC. Tungsten probes were used to contact the pads. Auger electron spectroscopy (AES) data were obtained with a Perkin-Elmer P H I model 660 scanning Auger microprobe (SAM). For depth profiles, the electron beam was set at 50 nA and 3 keV, while the samples were sputtered with a 3 keV Ar + ion from a differentially pumped ion gun. The argon ions were incident on the surface at about 45 ° at 10 m P a argon pressure. The ion beam was rastered over an area of 2 x 2 mm 2 and the AES profiles are shown as peak-to-peak heights of the derivative AES signals versus sputter time. Surface spectra were obtained using 5 keV electrons with a 300 nA beam current.

3. Results 3.1. Contact resistance

Examples of the I-Vcurves taken after the 900, 800, and 700°C anneal between the 60 gm spaced contacts

G.L. Waytena et aL/Diamond and Related Materials 5 (1996) 1450-1456

1452

2.0

I !

< E

1 i

1 i

I _ i a

0.1 A

i..,,.,.b
................................ i................ !.............

o

0

"~ Im

0 -2.(

I -4

l VoltaOeg (V)

1 4

-0.1

Fig. 2. The current-voltage curve for the Ti/Pt/Au contact surface annealed at (a) 900, (b) 800, and (c) 700°C, for a contact spacing of 60 gm.

are shown in Figs. 2(a), 2(b), and 2(c), respectively. The curves taken after the 800 and 900 °C anneal appear linear. Fig. 3 shows a plot of the total resistance versus contact spacing taken from the 800°C annealed sample. The resistance deviates systematically from linear behavior. The data are better fit with an exponential or power function than a straight line. This was also the case for the 900 °C annealed sample.

background. Between 8 and 12 min, the large Au signal is still present, while the Pt and Ti signals overlap and reach their maximum. Oxygen was present in the film near the diamond/Ti interface. This was expected because the CrO3 and H2SO 4 diamond cleaning has been shown to contribute to a submonolayer of oxygen [19]. Since only trace amounts of oxygen were found in the film, it is likely that very little oxygen diffused along grain boundaries from the diamond surface. The phase diagram for the Ti-O system indicates that titanium oxides (TiO~) are likely to form. Fig. 6 shows an Auger spectrum from the 900 °C annealed sample taken between two pads separated by 10 gm. Argon, C and Ti peaks have been identified at 217, 275, and 388 eV, respectively. The C and Ar signals come from the diamond and argon sputtering gas, respectively. However, the Ti was unexpected. Additional spectra were collected around the contacts. Although the Ti signal decreased as the distance from the contact increased, it was still present on the opposite surface of the substrate. Titanium was also detected between the contacts after an 800°C anneal. Again, the concentration varied with distance from the contact.

3.3. Surface analysis

3.2. Chemical analysis A representative AES spectrum taken from the surface of one of the 900 °C annealed contacts is depicted in Fig. 4(a). The detail in the low energy region of the spectrum can be seen in the expanded spectrum in Fig. 4(b). Gold peaks have been identified at 44, 71, 150, 163, 240, 256, 2016 (arrowed) and 2100eV. However, there is no Ti signal at 387 eV (arrowed). An AES compositional depth profile from the 900 °C annealed contact is shown in Fig. 5. In this profile Ot (510 eV), Til (387 eV), C1 (272 eV), Pta (1967 eV), and Au3 (2024 eV) were monitored. The trilayer film surface and interface with diamond is at zero and approximately 12 min sputter time, respectively. During the first 8 min of sputter time, a relatively large Au and Pt signal is observed, while a Ti and O signal are just above the A20 Ec-

Previous resistivity measurements of Ti/Pt contacts on diamond did not mention the presence of Ti on the diamond surface, especially between the contacts [ 10,20]. Apparently, it was either not looked for, or did not occur in their systems. In an attempt to explore this issue, a Ti/Au bilayer on diamond was investigated. The bilayer was fabricated under similar conditions as the trilayer, except the deposition of the Pt was excluded in the metallization step, and the anneal process occurred at 850°C (which is the same temperature used by others [10]). AES analysis performed on the diamond surface about 50 gm from a Ti/Au contact before and after a 120 min anneal at 850°C in vacuum (5 x 10 -s Torr) is shown in Figs. 7(a) and 7(b), respectively. The AES spectrum taken before the anneal shows only an O and C peaks at 515 and 275 eV, respectively. However, after the anneal, there is a Ti signal at 388 eV. Titanium was also detected on the Ti/Au contact surface after the anneal.

O

4. Discussion

oe-

4.1. Contact resistance

~0.

0

20 40 Pad Spacing (urn)

60

Fig. 3. Total resistance versus contact pad spacing for a transmission line model pattern with Ti/Pt/Au contact on IIb diamond after undergoing an 800°C anneal.

As shown in Fig. 2 for a specific contact spacing of 60 grn, the Ti/Pt/Au contacts have rectifying characteristics for anneal temperatures 'below 800°C, and ohmic characteristics for .anneal temperatures at and above 800°C.

1453

G.L. Waytena et al./Diamond and Related Materials 5 (1996) 1450-1456 0

I

1

-

~

Z

0 20

258

496

734

(a)

972

1210

1448

KtNETICENERGY, I

i0

~

I

I

T

r

l

r

l

l

l

1

1

1

'

1686

1924

2162

2400

eV l

!

'

'

!

I

r

5

Z

0

=

20

l

l

l

98

l

l

176

@)

254

l

l

332

l

l

410

J

l

488

KINETICENERGY,

l

l

]

566

l

t

644

722

800

eV

Fig. 4. (a) AES spectra taken from the Ti/Pt/Au contact surface on lib diamond after a 900°C anneal; (b) expansion of low energy section of AES scan shown in (a). The energy, 387 eV, at which the Ti peak would occur is arrowed.

10

~d < LTj

5 <

0.0

4.0

8.0

12.0

16.0

20.0

24.0

28.0

32.0

36.0

40.0

SPUTTER TIME, Min Fig. 5. AES depth profile taken from the Ti/Pt/Au contact on lib diamond surface after a 900°C anneal. The trilayer surface and interface with the diamond is located at 0 and approximately 12 min of sputter time, respectively. The O1 (510 eV), Til (387 eV), C1 (272 eV), Ptz (1967 eV), and Au (2024 eV) were monitored.

1454

G.L. Waytena et aL/Diamond 10

I

t

J

1

i

~

and Related Materials 1

i

I

1

5 (1996) 1450-1456

1

i

i

1

I

I

i

~

',

i

I

1

'

i

Ti

~5 z

0

' I

I

20

1

I

98

I

176

I

t

254

332

I

t

410

I

~ 488

KINETIC ENERGY,

I

566

1

644

I

722

800

eV

Fig. 6. AES spectra taken between two neighboring Ti/Pt/Au contacts separated by 10 gm after a 900 °C anneal. Ar, C and Ti signals are present at 217, 275 and 388 eV, respectively.

I

10

l

l

t

l

T

I

I

I

l

l

t

l

I

l

l

l

l

l

A

~'5 pa

0

v

Z

0 (a)

q

t

20

t

1

98

1

10

t

1

r

t

1

I

254

332

l

t

l

l

l

l

41o 488 KINETIC ENERGY, eV

176

t

1

I

t

r

P

I

I

l

I

566

;

644

T

I

1

722

r

,

800

i

~

q<

~-5 Z

0

• t

2O

(b)

1

98

1

1

176

I

1

254

1

l

332

t

I

410

I

I

I

488

t

566

1

I

644

I

t

722

I

I

800

KINETIC ENERGY, eV

Fig. 7. AES spectra collected on the diamond substrate about i0 gm from a Ti/Au contact (a) before and (b) after a 120 min anneal at 850°C in vacuum. The C, O, Ti, and Ar peaks can be seen at 272, 508, 387 and 217 eV, respectively.

G.L. Waytenaet aL/Diamond and .Related Materials 5 (1996) 1450-1456

In order to calculate the specific contact resistance, ro, the total resistance, RT, is first calculated from the slope of the I-Vcurve. The total resistance between any two contacts, which is the sum of the contact resistance (Re) and semiconductor resistance, is given by [21]: RT = (rsL)/W+ 2Ro

(1)

where r s is the semiconductor sheet resistance, L is the contact spacing, and Wis the contact width. The total resistance is measured for the various contact spacings, and then plotted. The slope of this plot (rs/W) gives the sheet resistance, where W is measured independently, and the intercept at L = 0 gives 2Re. The contact resistance, in combination with the intercept at RT =0, can be used to calculate the specific contact resistance [21]. Measurements from the 800 and 900 °C annealed sample taken between Ti/Pt/Au pads yielded ohmic contact characteristics for all separation distances. However, a plot of RT versus Lwas not linear (Fig. 3). As a result, the specific contact resistance, which is calculated using the intercepts of the resistance versus contact separation plot, could not be accurately determined. 4.2. Nonlinearity in contact resistance

The nonlinearity in the RT versus L plot of Fig. 3 may be due to the presence of a nonuniform concentration of Ti between the contacts. This may affect the total resistance between the contacts by differing amounts for various contact spacings. To see how this can occur, consider the slope of the line de'scribed in Eq. (1). The slope is proportional to the diamond sheet resistance, assuming it is uniform across the diamond surface. In the case of our samples, the sheet resistance may be increasing as the spacing between contacts increases. This would occur if a material with a lower sheet resistance covered less and less area proportionately between the contacts as the pad spacing increased. Using AES, we found that the Ti signal decreased as the distance from the contact increased. Since the sheet resistance of titanium carbide is lower than that of diamond, the proportionate decrease in area covered by titanium carbide is consistent with the continuous increase in total resistance as contact separation distance increased. As a result, the slope of the resistance versus contact spacing curve will not be constant, and the RT versus L plot will not be linear. 4.3. Origin o f surface contamination

The source of the Ti is the contact metal. Titanium was found on the diamond surface, in higher concentrations between both the Ti/Au and the Ti/Pt/Au contacts, and on the Ti/Au contact surface, but only after the anneal. Therefore, the diamond surface was not

1455

contaminated during metal deposition. In addition, the amount of Ti decreased with increasing distance from the contacts, suggesting surface diffusion across the diamond. Titanium was not found on the A u surface of the Ti/Pt/Au contact, which indicates that, for a Ti/Pt/Au trilayer contact annealed for 120 rain at 900 °C, the Pt layer successfully served as a barrier to Ti diffusion. This was confirmed by the depth profile of this same sample (Fig. 5). Fig. 5 shows that annealing caused the Pt and Au to interdiffnse, and the Pt and Ti to interact. While this interdiffusion and interaction led to the consumption of Pt, enough remained to act as an effective barrier as can be seen by the small amount of Ti in the Au layer. These observations are consistent with predictions based on the binary phase diagrams for these systems. Platinum and Au are known to interdiffuse without forming intermetallic compounds at 900 °C, and Ti and Pt are very reactive at 900 °C and form several intermetallics, the most likely of which is Ti 3Pt in a Ti rich environment. The above observations suggest that the Ti beneath the contact underwent lateral diffusion out the sides of the contact during annealing, independent of the presence of the Pt layer. 4.4. Vertical diffusion barrier

The Pt layer served successfully as a diffusion barrier to Ti as described above. Platinum is believed to act as a barrier as follows. Ti is expected to move along crystalline grain boundaries present in the Au and Pt layers much more rapidly than through the bulk. However, the likely formation and subsequent presence of titanium-platinum intermetallics and TiOx, which would consume diffusing species, and the resulting loss of a significant volume of the grain boundary region, may suppress the diffusion of any further Ti. Consequently, only trace amounts of Ti would be detected in the Au layer or on its surface. The absence of oxygen and presence of Ti on the diamond surface (Fig. 6) suggests that the Ti is present as a carbide. If Ti were not bound to C after the anneal, it would have reacted with oxygen when the sample was removed from the furnace. We expect a carbide because the anneal was performed in vacuum, and oxygen, which may have been present after the acid cleaning, could have been desorbed from the diamond surface. In fact, we have found that the oxygen signal is absent from the AES spectrum taken from a IIb diamond substrate precleaned in boiling water, CrO~, and H2SO 4 solution after it undergoes a 900 °C annealed in vacuum [22].

5. Conclusion

Ohmic contacts on natural IIb diamond have been achieved by metallization with a Ti/Pt/Au trilayer fol-

1456

G.L. Waytena et at./Diamond and Related Materials 5 (1996) 1450-1456

lowed by a n 800 a n d 900 °C a n n e a l in vacuum. AES depth profiling data provided i n f o r m a t i o n a b o u t the intermetallic reactions between the t r a n s i t i o n metal Ti a n d barrier metal Pt. The A u a n d P t interdiffused. However, the Pt barrier was n o t fully c o n s u m e d a n d effectively served as a barrier to Ti diffusion t h r o u g h the A u layer to the contact surface. W h e t h e r or n o t this decreased the contact resistance is still unclear since there is a n o n - l i n e a r relation between the resistance a n d p a d s e p a r a t i o n p r o b a b l y due to the presence of the Ti on the d i a m o n d surface which laterally diffused from b e n e a t h the contact. O n e possible way to reduce the lateral diffusion of Ti is addressed by the use of a d o u b l e m a s k system ~-22].

References [1] H. Shiomi,Y. Nishibayashi and N. Fujimori, FMd-effect transistors using boron-doped diamond epitaxial films, Jpn. J. Appl. Phys., 28 (1989) L2153. [2] M.W. Geiss, D.D. Rathman, D.J. Ehrlich, R.A. Murphy and W.T. Lindley, High-temperature point-contact transistors and Schottky diodes formed on synthetic boron-doped diamond, IEEE Electron Dev. Lett., 8 (1987) 341. [3] G.Sh. Gildenblat, S.A. Grot, C.R. Wronski,A.R. Badzian, T. Badzian and R. Messier, Electrical characteristics of Schottky diodes fabricated using plasma assisted chemical vapor deposited diamond fills, Appl. Phys. Lett., 53 (1988) 586. [4] V.K. Bazhenov, I.M. Yikulin and A.G. Gontar, Synthetic diamonds in electronics (review), Soy. Phys. Semicon., 19 (1985) 829. [5] C.R. Zeisse, C.A. Hewett, R. Nguyen, J.R. Zeidler and R.G. Wilson, An ion-implanted diamond metal-insulator-semiconductor field-effecttransistor, IEEE Electron Dev. Lett., 12 (1991) 602. [61 K.L. Moazed, J.R. Zeidler and MJ. Taylor, A thermally activated solid state reaction process for fabricating ohmic contacts to semiconducting diamond, J. Appt. Phys., 68 (1990) 2246. [7] K.L. Moazed, R. Nguyen and J.R. Zeidler, Ohmic contacts to semiconductingdiamond, IEEE Electron Dev. Lett., 9 (1988) 350.

[8] G.Sh. Gildenblat, S.A. Grot, C.W. Hatfield, A.R. Badzian and T. Badzian, High-temperature Schottky diodes with thin-film diamond base, IEEE Electron Dev. Lett., tl (1990) 371. [9] M. Reset, C.A. Hewett, K.L. Moazed and J.R. Zeidler, High temperature reliability of refractory metal ohmic contacts to diamond, J. Electrochem. Soc., 139 (1992) 2001. [10] V. Venkatesan, D.M. Malta, K. Das and A.M. Belu, Evaluation of ohmic contacts formed by B+ implantation and Ti-Au metallization on diamond, J. Appt. Phys., 74 (1993) 1179. [11] H.A. Naseem, I. Meyyappan, C.S. Prasad and W.D. Brown, Au-based metallizafion on diamond substrates for multichip module applications, Int. J. Microcircuits Electronic Packaging, 16 (i993) 257. [12] T.S. Tisone and J. Drobek, Diffusion in thin film Ti-Au, Ti-Pd and Ti-Pt couples, J. Vac. Sci. Teehnol., 9 (1972) 271. [13] A. Piotrowski and E. Kaminsker, Ohmic contacts to III-V compound semiconductors, Thin Solid Fihns, 193/194 (1990) 511. [14] M.P. Lepselter,Beam-lead technology,Bell System Teehn. J., XLV (1966) 233. [15] Shiow-Hwa Lin, L. Sverdrup, K. Garner, E. Korevaar, C. Cason and C. Phillips, Electron beam activated diamond switch experiments, in Prec. of SPIE, Optically Activated Switching Iti, I873 (1993) 97. [16] LW. Glesener,A.A.Morrish and K.A. Snail,Electron beam modification of the Schottky diode characterisitics of diamond, Appl. Phys. Lett., 61 (1992) 429. [ 17] P. Pehrrson, Private Communication. [18] W. Shockley, A. Goetzberger and R.M. Scariett, Research and investigation of inverse epitaxiai UHF power transistors, Rep. No. AFAL-TDR-64-207,Air Force AvionicsLab., Wright-Patterson Air Force Base, OH, Sept. 1964. [19] Y. Mori, H. Kawarada and A. Hiraki, Properties of metal/diamond interfaces and effects of oxygen adsorbed onto diamond surface, Appl. Phys. Lett., 58 (1991) 940. [20] P.E. Viljoen,E.S. Lambers and P.H. Holloway, Reaction between diamond and titanium for ohmic contact and metallization adhesion layers, 3". Vac. Sci. Technol., B12 (1994) 2997. [21] D.K. Schroder, Semiconductor Material and Device Characterization, John Wiley & Sons, New York, 1990, pp. t14-121. [227 G.L. Waytena, H.A. Heft, J.S. Suehle,I.P. Isaacson, M.L. Rebbert, D.I. Ma and Christie Martian The use of a double mask system to prevent lateral Ti diffusionfrom a Ti/Pt/Au ohmic contact on diamond, J. Electrochem. Soc., (1996) in press.