Novel Cu/Cr/Ge/Pd Ohmic Contacts on Highly Doped n-GaAs
KARTIKA CHANDRA SAHOO,1CHUN-WEI CHANG,1YUEN-YEE WONG,1
TUNG-LING HSIEH,1EDWARD YI CHANG,1,3and CHING-TING LEE2
1.—Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu 300, Taiwan, ROC. 2.—Institute of Microelectronics, Department of Electrical Engineering, National Cheng-Kung University, Tainan 701, Taiwan, ROC.3.—e-mail: [email protected] The thermal stability of the Cu/Cr/Ge/Pd/n+-GaAs contact structure was evaluated. In this structure, a thin 40 nm layer of chromium was deposited as a diffusion barrier to block copper diffusion into GaAs. After thermal annealing at 350C, the specific contact resistance of the copper-based ohmic contact Cu/Cr/Ge/Pd was measured to be (5.1 ± 0.6)· 10-7X cm2
. Diffusion behaviors of these films at different annealing temperatures were character-ized by metal sheet resistance, X-ray diffraction data, Auger electron spec-troscopy, and transmission electron microscopy. The Cu/Cr/Ge/Pd contact structure was very stable after 350C annealing. However, after 400C annealing, the reaction of copper with the underlying layers started to occur and formed Cu3Ga, Cu3As, Cu9Ga4, and Ge3Cu phases due to interfacial
instability and copper diffusion.
Key words: Thermal stability, annealing, ohmic contact, diffusion
INTRODUCTION
Low-resistance ohmic contacts that are thermally stable are essential for GaAs-based microwave and millimeter-wave devices.1 Since IBM first engaged in the copper interconnection technology,2–4 copper metallization has attracted great attention in the silicon integrated circuit (IC) industry. Conven-tionally, gold is used as the contact and interconnect metal for GaAs microwave devices and circuits. Using copper in place of gold as the metallization metal for the GaAs devices has the advantages of lower resistivity, higher thermal conductivity, and lower cost. Copper diffuses very quickly into silicon without any diffusion barrier.5 It is generally con-firmed that the rapid diffusion results from singly ionized interstitial copper that migrates as a posi-tively charged ion in silicon.6 Similarly, copper is known to diffuse rapidly into GaAs via a kick-out mechanism in the absence of a diffusion barrier7to create deep traps that degrade device characteris-tics. Even though copper metallization has played
an important role in the silicon IC industry, there are few papers related to copper metallization for GaAs devices.8–10
The AuGeNi alloyed ohmic contact was commonly used as the ohmic contact to n-GaAs in the past. In this study, the feasibility of using novel Pd/Ge-based copper ohmic contacts on highly doped n-GaAs is investigated. In comparison with the AuGeNi ohmic system, the PdGe-based ohmic contact has the fol-lowing advantages: (1) a better surface morphology, and (2) a better contact edge definition due to the solid phase regrowth.11 In this Cu/Cr/Ge/Pd struc-ture, Ge/Pd was used to reduce the contact resis-tance. The refractory metal chromium was used as the diffusion barrier between Cu and the underlying materials due to its high melting point and low solubility in copper even at high temperatures.12 The thick Cu metal on the top was used to reduce the sheet resistance of the metal layers.
EXPERIMENTAL
The ohmic contact structure of Cr/Cu/Cr/Ge/Pd/ n+-GaAs was prepared for electrical and materials characterizations. The substrates were semi-insu-lating (100) GaAs wafers with Si-doped layers (Received January 23, 2007; accepted January 22, 2008;
published online February 27, 2008)
Journal of ELECTRONIC MATERIALS, Vol. 37, No. 6, 2008 Regular Issue Paper
DOI: 10.1007/s11664-008-0398-3 2008 TMS
(5 · 1018cm-3, 0.2 lm) grown by metalorganic chemical vapor deposition (MOCVD). After an ohmic contact pattern was defined by photolitho-graphy, the ohmic metals were deposited by electron beam evaporation. A palladium film of 50 nm thickness was evaporated on top of the n+-GaAs substrate first, and 125 nm germanium was evapo-rated on the Pd film. To avoid copper contamination in the Au metallized process, the substrate was transferred to another electron beam chamber for evaporation of Cr(40 nm)/Cu(150 nm)/Cr(15 nm) thin-film stacks. Note that for the test samples, a top Cr (15 nm) layer was deposited on the surface of Cu to prevent the copper layer from oxidizing during further study. After metal deposition, the samples were annealed in a nitrogen ambient, and followed by a series of material analyses. X-ray diffraction (XRD), Auger electron spectroscopy (AES), transmission electron microscopy (TEM), and four-point probe measurements were used to identify the phases formed and the interfacial reactions of the contact metals.
DISCUSSION
The specific contact resistances of the Cr/Cu/Cr/ Ge/Pd/n+-GaAs ohmic contacts were obtained by the transmission-line method (TLM) using a Keithley 2400 source meter. The TLM pattern, as illustrated in Fig.1, was designed in the process control monitor (PCM) in order to measure the ohmic con-tact resistance and to identify the ohmic concon-tact characteristics. In our measurements, the distances between TLM electrodes were 3 lm, 5 lm, 10 lm, 20 lm, and 36 lm, respectively. Figure 2 shows specific contact resistance as a function of annealing temperature after a 10-min anneal of the Cr/Cu/Cr/ Ge/Pd/n+-GaAs structure. The lowest specific con-tact resistance was achieved after 350C annealing for 10 min. At this annealing temperature, the specific contact resistance of the Cr/Cu/Cr/Ge/Pd/ n+-GaAs contact was measured to be (5.1 ± 0.6) · 10-7X cm2. In order to evaluate the metal sheet resistances during thermal processing, n+-GaAs substrates with Cr/Cu/Cr/Ge/Pd metal stacks on top of the entire surface were prepared. Four-point probe measurements (collinear structure, Napson RT-7) were used to measure the metal sheet resis-tances of the ohmic metal structure at different
annealing temperatures, and a 10· 10 mm2square sample size was used. Figure3 shows the metal sheet resistances of the Cr/Cu/Cr/Ge/Pd/n+-GaAs structure after annealing at 300C, 350C, and 400C for 30 min. The lowest metal sheet resistance was also achieved after the sample was annealed at 350C, which was mainly due to grain growth and a decrease in the defect density in the ohmic metals. The metal sheet resistance of the ohmic contact increased drastically after annealing at 400C, implying that significant atomic diffusion and the interfacial reactions between the Cu layer and the underlying films had occurred. To confirm that the low metal sheet resistance obtained from the four-point probe method was due mostly to the metal layers and not by the highly doped n-GaAs layer, the bulk resistance and the sheet resistance of the semiconductor without metal layers were also measured by the four-point probe measurement; the
L2 L3 L4 L5
L1
Contact Pad Active Region
Fig. 1. Transmission-line method (TLM) patterns, where L1,L2, L3, L4, and L5are 36 lm, 20 lm, 10 lm, 5 lm, and 3 lm, respectively.
300 320 340 360 380 400 1E-7 1E-6 1E-5 1E-4 m c-m h O( e c n at si s e R t c at n o C cif i c e p S 2 ) Annealing Temperature(°C) Cr/Cu/Cr/Ge/Pd/n+GaAs
after a 10-min anneal in N 2 ambient
Fig. 2. Specific contact resistivity as a function of annealing tem-perature after a 10-min anneal for the Cr/Cu/Cr/Ge/Pd/n+-GaAs ohmic structure. 300 320 340 360 380 400 0 500 1000 1500 2000 2500 m( e c n at si s e R t e e h S Ω ) q s/ Annealing Temperature(°C)
Fig. 3. Metal sheet resistance of the Cr/Cu/Cr/Ge/Pd/n+-GaAs ohmic structure after annealing at 300C, 350C, and 400C for 30 min.
Sahoo, Chang, Wong, Hsieh, Chang, and Lee 902
resulting data were 4.14 kX and 287.7 X/square, respectively. The lowest metal sheet resistance after metallization and annealing at 350C was 230 mX/square. By comparing the sheet resis-tances, we can see that the semiconductor sheet resistance is much higher than the metal sheet resistance. Since the resistance of the metal layer and the resistance of the semiconductor layer are in parallel with each other, the sheet resistance we measured is mostly contributed by the metal layers. This indicates that there is only negligible parallel conduction in the semiconductor layer. XRD was used to identify the interfacial reactions between the ohmic metals and n+-GaAs. Figure4 shows the XRD patterns of the Cr/Cu/Cr/Ge/Pd/n+-GaAs, as deposited and after annealing at 350C and 400C. From a careful inspection of the peak at about 43 deg after annealing at 350C reveals Cr3Ge, PdGe,
and Cu (111) peaks. This further indicates that Pd reacts with Ge to form large PdGe grains and reveals the reaction of Cr with Ge to form Cr3Ge.
The metal-like behavior of Cr3Ge with higher Cr
concentration could reduce the contact resistance of the proposed structure. The XRD data show no extra copper compound formed after annealing up to 350C, suggesting that the contact was quite stable up to 350C. This is consistent with the results of TLM measurements. When the annealing temper-ature was increased to 400C, the interdiffusion of the ohmic metals and the substrate material occurred. Extra compounds such as Cu3Ga, Cu3As,
Cu9Ga4and e-Ge3Cu formed after 400C annealing
as indicated in the XRD data. This result indicates that the multilayer structure was destroyed as a result of the strong reaction of Cu with Ga to form Cu9Ga4 and Cu3Ga, the reaction of Cu with As to
form Cu3As, and the reaction of Ge with Cu to form
Ge3Cu after the failure of the Cr diffusion layer due
to massive diffusion of the elements. It is apparent that the Cr diffusion barrier failed to prevent the Cu atoms from penetrating into the Pd/Ge layers, and the Cu atoms diffused through Cr barrier and reacted with the Pd/Ge metals and GaAs at this temperature. This is consistent with the drastic increase of the metal sheet resistance after annealing at 400C as shown in Fig.3. Additional evidence showing the stability of the contact after 350C annealing can be seen from the AES depth profiles shown in Fig. 5. As can be seen from this figure, the Auger depth profiles clearly indicate that the Cr layer was very stable and there was no Cu diffusion into the Pd/Ge layers after 350C anneal-ing. To further investigate the reactions at the interfaces after thermal annealing at 350C, cross-sectional TEM analysis was performed on the annealed samples. Figure6 shows the cross-sectional TEM micrograph of this ohmic contact after 10 min, 350C annealing. In Fig.6, the TEM image shows that the PdGe grains formed after
25 0 500 1000 1500 Cr GaAs(100) Cu3As ε-Ge3Cu Cu9Ga4 Cu3Ga Ge Cr Cu(200) Pd Cu(111) Cr 400°C30m 350°C10m as-dep. )ti n u. br a( yti s n et nI 2θ(deg) Cu PdGe Cr3Ge 30 35 40 45 50 55
Fig. 4. XRD results of the Cr/Cu/Cr/Ge/Pd/n+-GaAs ohmic structure, as deposited and after annealing at various temperatures.
0 1000 2000 3000 4000 5000 6000 7000 0 20 40 60 80 100 O Pd Ge Ga Cu As Cr )ti n u. br A( yti s n et nI
Etch Time (sec)
Fig. 5. AES depth profiles of the Cr/Cu/Cr/Ge/Pd/n+-GaAs ohmic structure after annealing at 350C for 10 min.
Fig. 6. Cross-sectional TEM micrograph of the Cr/Cu/Cr/Ge/Pd/ n+
-GaAs ohmic structure after annealing at 350C for 10 min.
350C annealing. Radulescu et al. suggested that the PdGe phase dominated in the Pd:Ge reaction after 322C annealing. At the low contact resistance, ohmic contact formation was mainly due to the PdGe phase formed.13Figure6 also shows no evidence of intermixing between Cu and Cr, indicating that Cr is a reliable diffusion barrier for Cu. Figure7shows the high-resolution TEM micrograph of the Cu/Cr/Ge/Pd contact structure after 350C annealing. This result shows that the interfaces between the contact metals were quite sharp, indicating again that the Cr is an effective diffusion barrier between Cu and underly-ing metals. Judgunderly-ing from the data of XRD, AES, and TEM described above, it can be concluded that Cr is a reliable diffusion barrier for Cu ohmic contacts on n+-GaAs, and that Cu/Cr/Ge/Pd is an effective low contact resistance ohmic contact to n+-GaAs.
CONCLUSIONS
A Cu/Cr/Ge/Pd ohmic contact was proposed and characterized. The Cu/Cr/Ge/Pd ohmic contact
structure achieved a lowest contact resistance of (5.1 ± 0.6)· 10-7 X cm2 after annealing at 350C. At this temperature, the reaction of Cr with Ge to form the intermetallic Cr3Ge compound could
reduce the contact resistance. However, the contact structures deteriorated after annealing at 400C due to the interfacial reactions between Cu and the underlying films. The metal sheet resistance, XRD, AES, and TEM analysis data also proved that Cr is a reliable diffusion barrier for the Cu-based ohmic contacts to n+-GaAs for annealing up to 350C and failed to be an effective diffusion barrier after annealing at 400C. The experimental results in this study suggest that Cu/Cr/Ge/Pd is an effective copper-based ohmic contact structure and can be used for future copper metallization of GaAs-based electronic devices.
ACKNOWLEDGEMENT
The authors would like to acknowledge the assistance and support from the National Science Council, and the Ministry of Economic Affairs, Taiwan, R.O.C., under Contracts NSC96-2752-E-009-001-PAE, 95-EC-17-A-05-S1-020, and 96-EC-17-A-05-S1-020.
REFERENCES
1. P.M. Asbeck, High-Speed Semiconductor Devices (New York: John Wiley & Sons, 1990).
2. K. Holloway and P.M. Fryer, Appl. Phys. Lett. 57, 1736 (1990).
3. K. Holloway, P.M. Fryer, C. Cabral Jr., J.M.E. Harper, P.J. Bailey, and K.H. Kelleher, J. Appl. Phys. 71, 5433 (1992).
4. D. Gupta, Mater. Chem. Phys. 41, 199 (1995).
5. R.N. Hall and J.H. Racette, J. Appl. Phys. 35, 379 (1964). 6. C.S. Fuller and J.C. Severiens, Phys. Rev. 96, 21 (1954). 7. F.C. Frank and D. Turnball, Phys. Rev. 104, 617 (1956). 8. C.Y. Chen, E.Y. Chang, L. Chang, S.H. Chen, and D.F. Chang,
Appl. Phys. Lett. 77, 3367 (2000).
9. C.Y. Chen, E.Y. Chang, L. Chang, and S.H. Chen, Electron. Lett. 36, 1318 (2000).
10. C.Y. Chen, E.Y. Chang, L. Chang, and S.H. Chen, IEEE Trans. Electron. Dev. 48, 1033 (2001).
11. L.S. Yu, L.C. Wang, E.D. Marshall, S.S. Lau, and T.F. Kuech, J. Appl. Phys. 65, 1621 (1989).
12. R.N. Hall and J.H. Racette, J. Appl. Phys. 35, 379 (1964). 13. F. Radulescu and J.M. McCarthy, J. Appl. Phys. 86, 995
(1999). Fig. 7. High-resolution TEM micrograph of the Cu/Cr/Ge/Pd ohmic
contact structure after annealing at 350C for 10 min.
Sahoo, Chang, Wong, Hsieh, Chang, and Lee 904
