OHMIC CONTACT TO n-GaN WITH TiN DIFFUSION ...

6 downloads 0 Views 370KB Size Report
electrical measurements, x-ray diffraction and SIMS. It has been ..... T.W.Weaks, Jr., M.D.Bremser, K.S.Ailey, E.Carlson, W.G.Perry, E.L.Piner, N.A.EI-Masry,.

OHMIC CONTACT TO n-GaN WITH TiN DIFFUSION BARRIER E.KAMJNýSKA*, A.PIOTROWSKA* M.GUZIEWICZ*, S.KASJANIUK*, A. BARCZ*, E.DYNOWSKA**, M.D.BREMSER***, O.H.NAM***, AND R.F.DAVIS*** *Institute of Electron Technology, AI.Lotnik6w 46, Warszawa, Poland, [email protected] "**Institute of Physics, PAS, AI.Lotnik6w 46, Warszawa, Poland ***Department of Materials Science and Engineering, North Carolina State University, Raleigh, NC 27695-7907 ABSTRACT The formation of n-GaN/Ti ohmic contacts with TiN diffusion barriers has been investigated by electrical measurements, x-ray diffraction and SIMS. It has been shown that the onset of the ohmic behaviour is associated with the thermally induced phase transformation of Ti into TiN at the GaN/Ti interface. It is suggested that the process is accompanied by an increase in the doping level in the semiconductor subcontact region. The presence of a TiN barrier is found to inhibit excessive decomposition of GaN and to confine the reaction between n-GaN and Ti. INTRODUCTION The most successful approach to fabricate ohmic contacts to n-type GaN involves the use of Ti-based metallization, such as TiAI [1], TiAu [2], TiAg [3], and recently also Ti/Al with Ni/Au overlayer [4]. It is believed that during annealing of GaN/Ti interfacial nitrogen, extracted from GaN, reacts with Ti to form TiN. The outdiffusing nitrogen atoms create a substantial concentration of nitrogen vacancies, and thus a heavily doped semiconductor subcontact region [1]. Although Ti-based contacts may yield excellent contact resistivity, they do not satisfy the requirement of thermal stability [5]. In recent years there has been significant progress in the development of thermally stable metallization schemes for Si VLSI and III-V semiconductor devices. The use of diffusion barriers in contact structures enabled stable and reliable contacts and became widely accepted as a standard technology. One of the most attractive materials used for barrier layer is TiN due to its thermodynamic stability, its high melting point and low resistivity. In electrical contacts the effectiveness of TiN as a diffusion barrier between gold, aluminum, or copper overlayers for wiring purposes and underlying films has been shown by several authors [6-9]. In this paper we present an approach for improving the reliability of Ti-based metallization to n-GaN by using a reactively sputtered TiN diffusion barrier to prevent the intermixing of GaN and contact metallization with overlayers. We correlated the electrical properties and the microstructures of n-GaN/Ti contacts, encapsulated with TiN layers. In an attempt to clarify the particular role of Ti and TiN during the formation of the ohmic contact, we compared the interfacial microstructure of pure Ti, pure TiN, and Ti/TiN contacts both as-deposited and annealed. EXPERIMENTAL PROCEDURE N-type, Si doped GaN films for this study were grown by OMVPE on 6H-SiC at 1050 0C [10]. Since the current transport across the potential barrier at the metal/semiconductor interface depends essentially upon the doping level of the semiconductor, to follow this

1055 Mat. Res. Soc. Symp. Proc. Vol. 449 01997 Materials Research Society

dependence the experiments were performed on four sets of samples with various doping concentrations ND namely: 3.7* 1016cm 3 (lp=342cm 2/Vs), 3.7*1017cm 3 (ýt=411cm 2/Vs), 2.0* 1018cm 3 (pa=257cm 2/Vs), 1.1 * 10' 9cm 3 (pt= I70cm 2/Vs). Special attention was paid to the surface preparation prior to metal deposition. First, the surface of GaN was cleaned by boiling in hot organic solvents and by plasma ashering. Next, the samples were soaked in buffered IF for 5 min., rinsed in H 20 DI, dipped in NH 4OH:H 20 1:10 for 15 s. and immediately loaded into the deposition chamber. The cleaning of the substrates was completed in the deposition chamber by heat treatment at 550 0 C for 10 min. Metallic films were deposited by rf magnetron sputtering at a base pressure of 1*10-7 Torr. 0 Before starting the deposition of metallization the substrates were cooled down to 220 C. Ti films of the thickness 15-100 nm and 80 nm thick TiN layers were sequentially deposited, without breaking the vacuum, using Ti target, by Ar and Ar/N 2 reactive sputtering, respectively. Ti films were deposited at a pressure p=3.3*10-3 Torr, at a rate 1.4 nm/s. The parameters of the deposition of stoichiometric TiN films were as follows: a gas flow ratio of N2/Ar = 0.18, p=3.7* 103 Torr, the substrate bias voltage U1 = -60V. The deposition rate was 0.2 nm/s. The resistivities of as deposited Ti and TiN layers were 71.5 jtncm and 38 pf~cm, respectively. A complementary study of Ti contacts encapsulated with Si3N4 film has been performed. Ti/Si3 N4 (100 nm/200 nm) double layers were deposited by Ar rf magnetron sputtering from Ti and Si3N4 targets, respectively. Heat treatments were performed using rapid thermal annealing (RTA) in N2 flow at temperatures up to 9000 C for 30s. During RTA the samples were protected by a piece of oxidized Si as a proximity cap. The electrical characterization involved the measurements of I-V characteristics and contact resistivity r,. Samples for I-V measurements were patterned of dots with diameter varying from 120 to 350 ptm and one large area contact. I-V measurements were taken between one of the dots and the large area contact. The contact resistance test pattern had 1OOx 100 jm contact pads with pad separations 2, 4, 6, 10 and 12 ptm. Mesa structures, required for r, measurements by transmission line method (TLM), were patterned with Si0 2 mask and dry etched using CC14/H2 plasma. The RIE process pressure and the gas flow ratio were 1.1*10-1 Torr and 9.3 sccm/12 sccm, respectively. The etching rate was 16.7 nm/min. at a rf power (13.56 MHz) of 0.4W/cm 2. The thermally induced transformations in the contact region were investigated by the complementary use of x-ray diffraction (XRD) and secondary ion mass spectrometry (SIMS). RESULTS Electrical properties 2 3 2 3 As-deposited n-GaN/Ti/TiN contacts were ohmic with r, 5.3*10 Kcm , 1.9*10 F cm , and 3 3 2 2*10-gcm for dopant concentration 3.7*10"7 cm , 2.0*1018 cm , and 1.1*10' 9cm' 3, respectively. Contacts deposited on material doped to 3.7* 1016 cm 3 exhibited non-ohmic behavior. Figure 1 shows the I-V characteristics of contacts formed on lightly doped n-GaN, before heat treatment and after annealing at temperatures up to 9000 C. Slightly non-linear after deposition, the behaviour of Ti/TiN contacts annealed up to 4000 C remained unchanged. However, as a result of annealing at 5000 C the contacts became rectifying, with a ideality factor n=l.19 and the barrier height OB;0.71 eV, as inferred from the forward biased I-V characteristics. After annealing at higher temperatures, contacts gradually lost their rectifying properties. After the 9000 C anneal they were ohmic with rj=6* 102ncm 2. In contrast, pure TiN

1056

contacts deposited on the similarly doped substrates were ohmic in the as-deposited state and

remain such after heat treatment up to 400 0C, with r,=8.8* 10"2ncm 2. Anneals at higher temperatures resulted in progressive degradation of their linear characteristics and in increase of the contact resistivity. The I-V characteristics of n-GaN/TiN contacts are plotted in Fig. 2. 0.30

1 ?t

0.

"b2 0.15

0.05

0

Z

E

.700 OC

8000 C -0.15

-0.05

/ n-GaNITiN16 n=3.7*10 cm-3

4000!

ý=140prm

4 as-dep. .t 0.50

-0.10

-0.30 -0.5 0

-0.25

0

0.25

0.50

UM

U[V] Fig. 1. I-V characteristics of as-deposited

Fig.2. I-V characteristics of as-deposited

and annealed n-GaN/Ti/TiN contacts.

and annealed n-GaN/TiN contacts.

The resistivity of Ti/TiN contacts heat treated at 900 0C as a function of doping level of n-type GaN is shown in Fig.3. For dopant concentration up to 2*1018 cm3, an important decrease of r, as a result of annealing is observed, and the contact resistivity exhibits an inverse proportionality to ND. For doping level of 1.1.*1019cm- 3 the contact resistivity practically does not change upon annealing. Thermally induced interactions at the n-GaN/Ti interface Since the ohmic behaviour of n-GaN/Ti contacts may be attributed to the annealing induced growth of TiN phase at the interface, special attention was paid to properly distinguish between the deliberately deposited TiN barrier layer and the newly formed TiN. To facilitate conclusive phase identification, a complementary study of Ti contacts annealed under Si3N4 cap was performed. The results of XRD analysis of n-GaN/Ti(15nm)/TiN(80 nm) contacts are presented in Fig. 4. The as-deposited Ti/TiN contact consisted of two separate layers: Ti and TiN (fcc) phases. Hexagonal (cph) Ti was found to have a preferred orientation, and a lattice parameter of c=4.726 A in the direction perpendicular to the surface (c=4.684 A for bulk Ti). Cubic (fcc) TiN film exhibited a preferred orientation, with a lattice parameter

1057

a=4.299 A (a=4.24 A for bulk TIN). The sequence of thermally activated phase transformations in TiN- and Si3N 4-encapsulated contacts was similar. Up to 700 0C the lattice parameter of Ti progressively increased, while the lattice parameter of TiN decreased. During 800TC anneal, the TiN os-deposited 4000 phase transformation Ti-*TiN took place. .

0

.... theory20 . ...

.

0

...

0

n-GaN/TiN

100

0E

B=WFTN" XGaN=0"

"

44

=•

C)

a".

726A

2000

"1020

al

dTmN-'4.283A

10000 -1

-=4.752A

0-

ht @900°C

16

cT,

8000C

3000-

0/'/li

n-G a

-M 10-4

CD E

0

"46000-

\

110

:T=4

EN

-4000

e

W

dTN=4.250A

900C

.500(0 1017

1018

10 10 10

1019

1020

10

10.

41.20 42-00

Dopant Concentration [cm3] Fig.3. The dependence of r, of Ti/TiN contacts on the doping level of n-GaN; dots -experimental data points, dashed line - theoretical prediction for DB= 0 .44 eV.

44.00 2e (deg)

.

46.00

Fig.4. X-ray diffractograms of asdeposited and annealed GaN/Ti (15 nm)/TiN(80nm) contacts.

The observed phase transformations can be interpreted as follows. As indicated by the values of lattice constants, both the as-deposited Ti and TiN films are strained. Upon annealing, the nitrogen released from neighboring GaN at one side, and from TiN at the other, spreads throughout the Ti film. At lower temperatures nitrogen atoms enter the Ti lattice, filling the interstitial sites. As they are larger than interstitial sites, they cause further distortion of the Ti lattice. When the concentration of nitrogen exceeds the solubility limit, the structure transformation from hcp to fcc takes place [1 I]. The stress in TiN film was relaxed during annealing. The degree of decomposition of GaN during the formation of Ti-based ohmic contacts was studied with SIMS. Annealing in a N2 flow of the as-deposited GaN/Ti(50nm) /TiN(80nm) structure leads to the limited dissolution of the substrate which manifests in a migration of Ga atoms into the initial Ti layer with pronounced accumulation at the Ti/TiN interface (Fig.5.). The Ga content in the Ti layer can be estimated to be a few at.%, but is undetectable in the initial TiN cap. This suggests that the main supply of nitrogen required for the observed transformation of the Ti film into TiN originates from the ambience rather than from the substrate. This is confirmed by a similar measurement of a pure Ti(100nm) deposited on GaN (Fig.6.) Gradual decrease of the Ga concentration towards the surface seems associated with the inhibiting of the dissolution of GaN in the course of annealing and subsequent formation of TiN. Thermal stability of the GaN/TiN system is demonstrated in Fig.7. where the respective changes in the N, Ga and Ti profiles before and after heating to 900 0 C are insignificant. 1058

106

105"

IC 104" cc

ii10",

N

n-GaN/i/TiN -asdeposited

102.

N

n-GaN/Ti/TiN 900 0C, 30s, N2

103-j

ib 102

0

260

460 '600

800

0

200

Sputter time [s]

6•0

460

8600

Sputter time [s]

Fig.5. SIMS in-depth profiles for n-GaN/Ti/TiN contacts: a) as deposited, b) annealed at 900 0 C. 10'-

too

Ga

Ga

N

NN

n-GaN/Ti

lO,

103, r

n-GaN/Ti 900 0C, 30s, N2

as-deposited

b

a

1021

102

0

200 400 600 Sputter time [s]

800

0

200 400 600 Sputter time [s]

800

Fig.6. SIMS in-depth profiles for n-GaN/Ti contacts: a) as deposited, b) annealed at 900 0C. 10e .

.

I,

I,

,

106

1

.

.

.

=

-

-

a 104U)

a 103

102 0

200

400 600 Sputter time [s]

800

0

200

400 600 Sputter time [s]

800

Fig.7. SIMS in-depth profiles for n-GaN/TiN contacts: a) as deposited, b) annealed at 900'C.

1059

DISCUSSION AND CONCLUSIONS It is for a number of reasons that Ti plays a key role in producing low resistivity contacts to n-GaN. Ti is known to reduce the native Ga2 O3 on GaAs surface [12], thus we can expect a similar behaviour for the native oxide on GaN. The improvement of the diode characteristics of n-GaN/Ti/TiN contacts after annealing at 500 0 C can be interpreted as the reduction of the residual oxide and/or the annealing out of defects at the metal/semiconductor interface. In contacts annealed at higher temperatures, the barrier height is dominated by the interfacial reactions, leading to the formation of TiN at the metal/semiconductor interface. From the simple Schottky barrier theory, the barrier height at n-GaN/TiN interface should be equal to the difference between the work function of TiN (3.74 eV [13]) and the electron affinity of GaN (3.3 eV [14]). In Fig.3., together with the experimental data points, we have plotted a theoretical dependence for ru=f(ND), calculated from the simple thermionic and thermionic-field emission theory [15], for *1B=0. 44 eV. The discrepancy between the theoretical prediction and experimental results can be explained by the increase of the doping concentration in the semiconductor subcontact region. The assumption that the ND had increased up to -1.0*,109 cm"3 would explain the unchanged resistivity of contacts deposited on highly doped GaN, and the decrease of rc for ND=2.0* 108 cm 3 . For contacts formed on lightly doped substrates, the contact resistivity would be limited by the barrier between highly doped subcontact region and low-doped bulk, the latter being characterized by an inverse proportionality to ND [16]. TIN barrier inhibits excessive decomposition of GaN and thus confines the reaction between n-GaN and Ti. REFERENCES 1. M.E.Lin, Z.Ma, F.Y.Huang, Z.F.Fan, LH.Allen, H.Morkoc, Appl.Phys.Lett., 64, 1003 (1994). 2.M.A.Khan, J.N.Kuznia, A.R.Bhattarai, D.T.Olson, Appl.Phys.Lett., 62, 1786 (1993). 3. J.D.Guo, C.I.Lin, M.S.Feng, F.M.Pan, G.C.Chi, C.T.Lee, Appl.Phys.Lett., 68, 235 (1996). 4. Z.Fan, S.N.Mohammad, W.Kim, O.Aktas, A.E.Botchkarev, H.Morkoc, Appl.Phys.Lett., 68, 1672 (1996). 5. H.Morkoc, Proc.Int.Symp.Blue Laser and LEDs, Chiba Univ.,Japan, 23 (1996). 6. N.Kumar, J.T.McGinn, K.Pourezaei, B.Lee, J.Vac.Sci.Technol., A6, 1602 (1988). 7. J.Schulte, S.B.Brodsky, T.Lin, R.V.Joshi, Tungsten and other refractory metals for VLSI applications, Mat.Res. Soc. 1987 Workshop Proc. 367 (1988) 8. Jian Li, P.F.Chapman, F.Goodwin, Mat.Res. Soc.Conf Proc. ULSI-VIII, 75 (1993). 9. A.Piotrowska, E.Kamifiska, M.Guziewicz, S.Kwiatkowski, A.Turos, Mat.Res.Soc.Symp. Proc. 300, 219 (1993). 10. T.W.Weaks, Jr., M.D.Bremser, K.S.Ailey, E.Carlson, W.G.Perry, E.L.Piner, N.A.EI-Masry, R.F.Davis, J.Mat.Res., 10, 1011 (1996). 11. M.Wittmer, J.Vac.Sci.Technol., A3, 1797 (1985). 12. S.P.Kowalczyk, J.R.Waldrop, R.W.Grant, Appl.Phys.Lett., 38, 167 (1981). 13. V.S.Fomenko, Emission Properties of Materials, Naukova Dumka, Kiev, 1981. 14. R.J.Nemanich, M.C.Benjamin, S.P.Bozeman, M.D.Bremser, S.W.King, B.L.Ward, R.F.Davis, B.Chen, Z.Zhang, J.Bernholc, Mat.Res.Soc.Symp. Proc. 395, 777 (1996). 15. A.Y.C.Yu, Solid-State Electronics, 13, 239 (1970). 16. W.Dingfen, W.Dening, K.Heime, Solid-State Electronics, 29, 489 (1986).

1060

Suggest Documents