Short Communication
Isotope effect of hydrogen release in
metal/oxide/n-silicon tunneling diodes
C.-H. Lin, F. Yuan, B.-C. Hsu, C.W. Liu
*Department of Electrical Engineering, Graduate Institute of Electronics Engineering, National Taiwan University, Taipei, Taiwan, ROC Received 2 July 2001; accepted 18 November 2002
Abstract
The metal/oxide/n-silicon tunneling diodes with hydrogen (deuterium) passivated Si/SiO2interface are stressed under
hole-injection conditions to investigate the mechanism of gate oxide degradation. Although the isotope effect on soft breakdown was previously observed in the deuterium-annealed metal/oxide/p-silicon devices, no isotope effect on the oxide soft breakdown was observed in the metal/oxide/n-silicon devices. However, the time evolution of electrolumi-nescence indeed shows the isotope effect on the interface states density at the Si/SiO2 interface of the
metal/oxide/n-silicon devices. This suggests that there is an isotope effect on the hydrogen (deuterium)-release at Si/SiO2 interface
under hole current stress from the gate electrodes, but the released hydrogen moves to the bulk Si (not oxide) due to the direction of the electric field. This can explain that the isotope effect is observed in the electroluminescence measure-ment, but not in the soft breakdown measurement. The hydrogen released to the bulk Si is not responsible for the soft breakdown, and the tunneling-hole-induced traps in the oxide may be responsible.
Ó 2003 Published by Elsevier Science Ltd.
Keywords: Ultrathin oxide; MOS; Deuterium; Degradation; Hole trapping
1. Introduction
The time-dependent degradation of metal–oxide–sil-icon (MOS) devices due to current stress has been ex-tensively studied since the early 1980s [1], and the reliability of ultrathin oxide becomes an important issue for future ultralarge scale integration technology due to the large gate leakage current. It is generally believed that gate oxide degrades after a critical density of elec-tron traps has been built-up in bulk oxide [2]. However, the mechanism of the trap generation is still a concerned issue these years. DiMaria and Cartier proposed that the trap creation was related to the Si/SiO2 interface
hy-drogen release (HR) by injected hot electrons [3], i.e., ‘‘HR model’’. The reliability improvement by replacing hydrogen with deuterium at the Si/SiO2 interface using
post-metallization anneal supports this model [4]. The improvement is due to the strong coupling between Si-D bending mode (460 cm1) and transverse optical
pho-nons in bulk Si (463 cm1) [5–7]. However, the anode/
hot hole injection model was imposed recently because no improvement in oxide reliability was observed in the devices with deuterium passivated Si/SiO2interface
un-der Fowler–Nordheim stress [8,9]. We have proposed that the oxide degradation is related to the HR model under low injected electron energy and high current density stress condition in n-channel MOS tunneling diodes [10]. In this letter, we investigate the degradation of the hydrogen- and deuterium-treated p-channel MOS (PMOS) tunneling diodes under positive gate voltage stress. The experimental results reveal that the deute-rium isotope effect is not observed on the oxide soft breakdown, while the isotope effect is indeed observed on the interface states density ðDitÞ monitored by
elec-troluminescence intensity. However, the breakage of the Si–H bonds does not contribute to the trap formation in the oxide due to that the released hydrogen moves into the bulk Si.
*
Corresponding author. Tel.: +886-2-2363-5251x515; fax: +886-2-2363-8247.
E-mail address:[email protected](C.W. Liu).
0038-1101/03/$ - see front matterÓ 2003 Published by Elsevier Science Ltd. doi:10.1016/S0038-1101(02)00461-6
Solid-State Electronics 47 (2003) 1123–1126
2. Experiments
The ultrathin gate oxide of the PMOS tunneling di-ode used in this study was grown by rapid thermal ox-idation (RTO) on n-type Si at 900 °C. The gas flows were 500 sccm nitrogen and 500 sccm oxygen at the pressure of 100–250 mbar. Before oxidation, the sample was cleaned by a HF dip. An in situ deuterium prebake at 900°C for 2 min was performed before the growth of the RTO. After the growth of the ultrathin oxide, the sample was in situ annealed in deuterium and nitrogen for 10 min each at 900 °C. This yields a deuterium concentration of 2 1020cm3in oxide measured by the
secondary ion mass spectroscopy using the same growth condition (Fig. 1 of Ref. [11]). The H2-treated samples
were processed with the same procedure except replacing deuterium by hydrogen. The wafer temperature was measured by a pyrometer with a close loop control. The thickness of oxide was measured by ellipsometry. The resistivity of the 100 mm n-type (1 0 0) wafers is 1–10 Xcm. PMOS diodes have Al gate electrodes with cir-cular areas defined by photolithography. In this experi-ment, the reliability measurement was carried out using an HP 4156A semiconductor parameter analyzer.
3. Results and discussion
Fig. 1 shows the time evolution of gate current for H2-treated PMOS tunneling diodes with oxide thickness
of 2.7 nm under constant voltage stress (CVS) at Vg¼ 2
V. The area size of the device is 3 104cm2. At positive
gate bias, electrons tunnel from the silicon substrate to gate electrodes, and holes tunnel from gate electrodes to silicon substrate. Note that the electron tunneling from
Si to Al cannot damage the Si/SiO2interface, since there
is no excess energy of electrons at the Si/SiO2interface.
The device reveals soft breakdown after2200 s stress, which indicates that some traps are built up in the bulk oxide due to the injected holes [13]. There is an apparent disparity between the current–voltage (I–V ) curves of the devices before and after stress (the inset of Fig. 1). Compared with the results of the H2-treated devices,
similar phenomena are observed in D2-treated devices
under the same stress condition. No deuterium isotope effect on soft breakdown is observed in the H2- and D2
-treated PMOS tunneling diodes under hole injection stress.
To confirm the observed results, more than 10 de-vices are measured to obtain the statistical data. Fig. 2 shows the Weibull plot of the charge to breakdown ðQBDÞ characteristics for both the H2- and the D2-treated
PMOS tunneling diodes with oxide thickness of 2.7 and 2.8 nm, respectively, under CVS at Vg¼ 2 V. The area
of the devices is 3 104 cm2. The Q
BD is calculated
from the integral of the gate tunneling current and the time-to-breakdownðTBDÞ as shown in Fig. 1 (labeled as
‘‘QBD’’). Both H2- and D2-treated devices have similar
QBDdistribution. No improvement of oxide reliability is
observed in the D2-treated devices. Based on the HR
model, the isotope effect is due to the strong coupling between Si–D bond bending mode and silicon optical phonon states, and as a result, the Si–D bonds become more difficult to be broken by injected electrons than Si– H bonds [5,10]. The isotope effect on soft breakdown is not observed in our experiments. Therefore, in the PMOS tunneling diodes, the degradation of the gate oxide may be dominated by other mechanisms.
Fig. 1. Time evolution of the gate current for H2-treated PMOS
tunneling diodes with oxide thickness of 2.7 nm under CVS at Vg¼ 2 V. The inset is the current–voltage curves before and
after stress.
Fig. 2. The Weibull plot of the charge to breakdown ðQBDÞ
characteristics for the H2- and the D2-treated PMOS tunneling
diodes under CVS at Vg¼ 2 V. No deuterium isotope effect is
observed.
One speculative mechanism is the HR from the Al gate. It is well known that the Al gates contain high concentration of hydrogen. The injected electrons from Si to Al gate may cause the HR from the gate, which will lead to the degradation of oxide. There will be no dif-ference between the reliability performance of the H2
-and D2-treated devices if the HR from the gate becomes
the dominant mechanism. The other possible mecha-nism is direct hole trapping in the oxide. Two kinds of current components exist in the PMOS tunneling diodes. The electron current tunnels from Si conduction band to Al and the hole current tunnels from Al to Si valence band. The hole tunneling from Al to Si can break the Si– O bond or Si–H bond [12] or be trapped in the bulk oxide by oxygen vacancy (O vacancy) [13,14] with the formation of the interface states and bulk tarp, respec-tively. While the electron tunneling from Si to Al cannot damage the Si/SiO2 interface, since there is no excess
electron energy at the Si/SiO2 interface. By theoretical
calculation, Yokozawa et al. have proposed that the O vacancy terminated with hydrogen initially can change its structure to be a new electron trap after capturing injected holes [13,14]. Therefore, if the injected holes have enough energy to break the Si–H bonds in the bulk oxide (hole trapped by O vacancy), the generation of traps will contribute to the leakage currents in oxide film [14], which will lead to the degradation of oxide. How-ever, these models are speculative and still under inves-tigation.
To further investigate the defect formation in the PMOS tunneling diodes, the light emission intensity at bandgap energy is measured [15]. The band-edge light emission can be observed when the oxide roughness [16] and phonons [17] can provide the necessary momentum. Fig. 3 shows the mechanism of light emission of the PMOS tunneling diodes. The device is biased at accu-mulation region, i.e., the positive voltage on the gate electrode. The holes tunnel from gate to the n-Si sub-strate, recombine with the electrons in the accumulation region, and then the light is emitted (radiative recom-bination). However, if the holes recombine with the electrons via the interface states, the emission intensity at bandgap energy decreases (non-radiative recombina-tion). Therefore, we can investigate the variation of the Dit by monitoring the time evolution of the emission
intensity. Note that it is difficult to get the information of the Dit in the MOS diode with ultrathin oxide from
capacitance–voltage measurement due to the significant leakage current. The time evolution of the emission in-tensity at the peak of the spectra was measured in the Fig. 4. The normalized emission intensity at the peak of the spectra for the D2-treated device decreases only9%
after 10,000 s constant current stress at 100 mA. While, the normalized emission intensity for hydrogen-treated device decreases33% after 10,000 s stress. For the H2
-treated device, the hydrogen bounded at interface (Si–H
bonds) was more easily released due to the injected holes than the deuterium. Therefore, the Dit increases and
the non-radiative recombination rate via interface states increases with the stress time. The deuterium isotope effect is observed on Dit, indicating that there must be
other origins of trap formation in the oxide of the PMOS device rather than the HR model under hole injection stress.
The speculative mechanism of the trap formation in the oxide of the PMOS device under hole injection stress is shown in Fig. 5. The hole tunneling from Al to Si can break the Si–H bond or be trapped in the bulk oxide by
Fig. 3. The mechanism of light emission of the PMOS tunnel-ing diodes.
Fig. 4. Time evolutions of the emission intensity at peak for both H2- and the D2-treated PMOS tunneling diodes under
constant current stress at 100 mA.
the oxygen vacancy (direct hole trapping) with the for-mation of the interface states and oxide traps, respec-tively. Since the released hydrogen moves forward into the bulk Si, not into the oxide, due to the direction of electric field, the released hydrogen cannot contribute to the trap formation in the oxide. The injected electrons from Si to Al gate may cause the HR from the gate. The origin of oxide traps may come from the direct hole trapping in the oxide or the HR from the Al gate, which induce the oxide soft breakdown. Therefore, no deute-rium isotope effect on soft breakdown is observed in the hole injection condition.
4. Conclusions
In conclusion, the degradation mechanism of PMOS tunneling diodes under hole injection is investigated. The soft breakdown has no isotope effect, but the elec-troluminescence has. The soft breakdown may be caused by the trap formation in oxide induced by the direct hole trapping in the oxide or the HR from the Al gate. The released hydrogen from the breakage of Si–H bonds at Si/SiO2 interface cannot contribute to the formation of
oxide traps due to the direction of the electric field.
Acknowledgements
This work is supported by National Science Council (89-2218-E-002-082 and 89-2218-E-002-054), and Insti-tute of Applied Science and Engineering Research, Ac-ademia Sinica, ROC.
References
[1] Maserijian J, Zamani N. Behavior of the Si/SiO2interface
observed by Fowler–Nordheim tunneling. J Appl Phys 1982;53(1):559–67.
[2] Degraeve R, Groeseneken G, Bellens R, Ogier JL, Depas M, Roussel PJ, et al. New insights in the relation between electron trap generation and the statistical properties of oxide breakdown. IEEE Trans Electron Dev 1998;45(4): 904–11. [3] DiMaria DJ, Cartier E. Mechanisms for stress-induced
leakage currents in thin silicon dioxide films. J Appl Phys 1995;78(6):3883–94.
[4] Lyding JW, Hess K, Kizilyalli IC. Reduction of hot electron degradation in metal oxide semiconductor tran-sistors by deuterium processing. Appl Phys Lett 1996;68(8): 2526–8.
[5] Van de Walle CG, Jackson WB. Comment on ‘‘Reduction of hot electron degradation in metal oxide semiconductor transistors by deuterium processing’’. Appl Phys Lett 1996; 69(16):2441–3.
[6] Hess K, Kizilyalli IC, Lyding JW. Giant isotope effect in hot electron degradation of metal oxide silicon devices. IEEE Trans Electron Dev 1998;45(2):406–16.
[7] Wei JH, Sun MS, Lee SC. A possible mechanism for improved light-induced degradation in deuterated amor-phous-silicon alloy. Appl Phys Lett 1997;71(11):1498–500. [8] Wu J, MacDonald RB, Li E, Tao J, Tracy B, Fang P. Anode hole injection versus hydrogen release: the mecha-nism for gate oxide breakdown. IRPS Tech Dig 2000:27–32. [9] Esseni D, Bude JD, Selmi L. Deuterium on interface states and SILC generation in CHE stress conditions: a compar-ative study. IEDM Tech Dig 2000:339–42.
[10] Lin CH, Lee MH, Liu CW. Correlation between Si–H/D bond desorption and injected electron energy in metal– oxide–silicon tunneling diodes. Appl Phys Lett 2001;78(5): 637–9.
[11] Liu CW, Lin CH, Lee MH, Chang ST, Liu YH, Chen MJ, et al. Enhanced reliability of electroluminescence from metal–oxide–silicon tunneling diodes by deuterium incor-poration. Appl Phys Lett 2001;78(10):1397–9.
[12] Chen Z, Garg P, Singh V, Chetlur S. Role of holes in the isotope effect and mechanisms for the metal–oxide–semi-conductor device degradation. Appl Phys Lett 2001;79(2): 212–4.
[13] Yokozawa A, Oshiyama A, Miyamoto Y, Kumashiro S. Oxygen vacancy with large lattice distortion as an origin of leakage currents in SiO2. IEDM Tech Dig 1997:703–6.
[14] Yokozawa A, Miyamoto Y. First-principles exploration of possible trap terminators in SiO2. Appl Phys Lett 1998;
73(8):1122–4.
[15] Liu CW, Lee MH, Chen MJ, Lin IC, Lin CF. Room-temperature electroluminescence from electron–hole plas-mas in the metal–oxide–silicon tunneling diodes. Appl Phys Lett 2000;76(12):1516–8.
[16] Liu CW, Lee MH, Chen MJ, Lin CF, Chern MY. Rough-ness-enhanced electroluminescence from metal silicon oxide tunneling diodes. Electron Dev Lett 2000;21(12):601–3. [17] Liu CW, Chen MJ, Lin IC, Lee MH, Lin CF. Temperature
dependence of the electron–hole-plasma electrolumines-cence from metal–oxide–silicon tunneling diodes. Appl Phys Lett 2000;77(18):1111–3.
Fig. 5. The speculative mechanism of the trap formation in the oxide of the PMOS device under hole injection stress.
