`
`Low-Leakage and Highly-Reliable 1.5 nm SiON Gate-Dielectric
`Using Radical Oxynitridation for Sub-0.1 pm CMOS
`
`M. Togo, K. Watanabe, T. Yamamoto, N. Ikarashi, K. Shiba, T. Tatsumi, H. Ono, and T. Mogami
`Silicon Systems Research Labs, NEC Corporation
`1120 Shimokuzawa, Sagamihara, Kanagawa 229-1198, Japan
`
`ABSTRACT
`
`We have developed a low—leakage and highly—reliable
`1.5 nm SiON gate—dielectric. by using radical oxynitridation.
`In this development, we
`introduce a new method for
`determining ultra~thin SiON gate—dielectric thickness based
`on the threshold voltage dependence on the substrate bias in
`MOSFETs.
`It was found that radical oxidation followed by
`radical nitridation provides 1.5 nm thick SiON in which
`leakage current is two orders of magnitude less than that of
`1.5 nm thick SiO2 without degrading device performance.
`The 1.5 nm thick SiON was also found to be ten times more
`reliable than 1.5 nm thick SiOz.
`INTRODUCTION
`
`is
`A low-leakage and highly—reliable gate—dielectric
`essential for high—performance sub-0.1 pm CMOSFETS [1].
`Oxyniuidation is one of the key techniques to achieve low—
`leakage gate-dielectrics [2]. The radial process can be also
`useful to improve SiO2 quality [3]. We have investigated 1.5
`nm SiON gate-dielectrics by using radical oxynitridation.
`In
`this investigation, we will introduce a new method of gate—
`dielectrie thickness measurement, because it is important to
`measure electrical oxide equivalent thickness (TOXW).
`THICKNESS DETERMINATION
`
`Large gate leakage current of the ultra-thin SiO2 disturbs
`thickness measurement by using C-V characteristics (Fig. 1).
`A newly proposed thickness measurement method is based on
`the general relationship between the threshold voltage (VTH)
`and the substrate bias voltage (VB).
`In this method, the gate
`leakage current does not affect V1H of a transistor (Fig. 2).
`Fig. 3 shows V1H dependence on V3, which is expressed as
`the following general equation:
`(1)
`VTH = ((255509Nch)l/Z/Cox)(VB+Z¢F)1/2 1’ VFB+2¢F
`where Nch is the channel concentration and Cox is the gate-
`dielectric capacitance where the oxide electrical
`thickness
`(TOME)
`is a component part.
`In Fig. 3,
`the line slope
`consists of Nch and TQM” and the .y-intercept indicates VFB.
`To determine Toxgq of the ultra-thin Si02, we fabricated a 6
`nm thick SiO2 gate-dielectric NMOSFET using the same
`fabrication process as that used to make ultra-thin SiO2 gate—
`dielectric NMOSFETs. First, we obtained Cox by measuring
`the CV curve of the 6 nm thick SiOz, and obtained accurate
`Nch by using (1) and the Cox. Next, we calculated the ultra—
`thin SiO;
`thickness (Toxm) by using (1) and Nch.
`Tox,ale
`consists of an oxide thickness
`(Toxgq)
`and a parasitic
`thickness
`(TOX_P,,,), which is composed of poly-Si gate
`depletion and inversion layer quantization.
`T0x_pm was
`extracted by using the C-V curve for 6 nm-thick (=TOX1W,
`physical
`thickness, = Toxm) gate SiOz shown in Fig.
`1.
`Finally, we obtained the electrical oxide equivalent thickness
`(TOX_ ) of the ultra—thin SiO2 by using T0x_ele and T0x_ m (Tox,
`,q =
`on], - Toxpm).
`Fig. 4 shows that obtained
`ox.
`is
`almost the same as measured TOXW in SiOz. Thus this film
`thickness determination method 18
`a
`suitable means
`for
`
`obtaining TOMq of not only ultra-thin 5102, but also SiON
`with heavy nitrogen concentration and other types of gate-
`dielectric films having high 8 and a stacked structure.
`RADICAL OXYNITRIDATION
`
`We used radical oxygen and nitrogen from an electron—
`cyclotron resonance (ECR) plasma to form an ultra—thin SiON
`
`1 16
`
`0-7803-8305-4/00/$10.00 © 2000 IEEE
`
`film in an ultra—high vacuum (UHV) system where the base
`pressure was less than 1><10’9 Torr to produce a clean Si
`surface [4]. Four types of radical processes were tested as
`ways
`to form the films:
`radical oxidation (0*),
`radical
`nitridation after radical oxidation (O*—+N*), radical oxidation
`and nitridation simultaneously (O*+N*) and radical oxidation
`after radical nitridation (N*40*).
`Fig. 5 shows SIMS
`profiles of nitrogen in a gate oxide fabricated by the radical
`process [5]. Radical oxidation and nitridation enables us to
`control the nitrogen profile in an ultra-thin gate-oxynitride,
`NITROGEN PROFILE ENGINEERING
`
`IG—VG characteristics of 1.5 nm thick
`Fig. 6 shows ID,
`gate—dielectric NMOSFETs using 0* and O*—>N*. The use
`of 7% nitrogen decreased gate leakage current by two orders
`of magnitude without decreasing the drain current.
`Fig. 7
`shows the gate leakage current of SiON and SiO2 formed by
`the radical processes and the conventional thermal process at
`the same bias. The leakage current of SiON is much lower
`than that of SiOQ. The SiON film formed by the O*—>N*
`process with 7% nitrogen have lowest leakage current.
`Fig.
`8 shows cross-sectional TEM photographs of the SiON whose
`gate leakage current is shown in Fig. 7. Table 1 compares
`the dielectric properties obtained with different
`radical
`processes. Nitrogen concentration in the SiON film formed
`by O*—>N* is lower than that in the SiON film formed by
`N*—>O*. However, the dielectric constant of the SiON film
`formed by O*-—)N* is larger than that of one formed by
`N*—>O*.
`Fig. 9 and 10 show that
`the drivability of the
`NMOSFET fortned by the O*—>N* process with 7% nitrogen
`is comparable to that formed by 0* at the supply voltage
`regime.
`Fig.
`11
`shows T13D distribution under constant
`voltage stress.
`In Fig. 11(a), the 1.5 nm thick SiON formed
`by the O*aN* process was not broken within 1035, which is
`ten times more reliable than 1.5 nm thick SiOz formed by the
`0* process.
`Furthermore, as the figure (b) indicates, a 1.5
`nm-thick (=Toxgq) SiON formed by the O*——)N* process with
`7% nitrogen is more reliable than a 2.3 nm-thick (=T0X_Phy)
`SiO2 formed by thermal oxidation.
`Fig. 12 shows that the
`chemical shift of N Is orbital for SiON films depends on the
`radical process. These results indicate that the nitrogen state
`depends on the nitrogen profile in SiON and may affect the
`dielectric property and electrical reliability.
`CONCLUSIONS
`
`By using radical oxynitridation, we have successfully
`achieved a low-leakage and highly-reliable 1.5 nm SiON with
`no degradation of device performance.
`This 1.5 nm thick
`SiON has two orders of magnitude less leakage current than
`1.5 nm thick SiO2 and shows ten times more reliable than 1.5
`nm thick SiOQ. We have developed a new thickness
`determination method that is useful for measuring ultra-thin
`SiON thickness.
`
`ACKNOWLEDGMENTS
`
`The authors thank M. Fukuma, T. Kunio, T. Tashiro, and
`Y. Miura for their encouragement and helpful discussions.
`REFERENCES
`[1] H. S. Momose et al., lEDM Tech. Dig. p. 819, 1999.
`[2] H. Yang et al., lEDM Tech. Dig, p. 245, 1999.
`[3] M. Nagamine et a1., lEDM Tech. Dig, p. 593, 1998.
`[4] K. Watanabe et 31.. MRS 1999 Spring Meeting Abstract, p. 268, 1999.
`[5] K. Watanabe et al., MRS 1999 Fall Meeting Abstract, 1999.
`
`2000 Symposium on VLSI Technology Digest of Technical Papers
`TSMC 1126
`TSMC 1126
`
`
`
`
`
`0.02
`
`A
`2
`\
`s
`u 0 01
`2
`L5
`g
`E
`
`o
`
`NMOS L¢=0.8 u m
`vn=1v va=o~-av
`
`A
`E
`=
`3
`4’
`43
`
`.
`
`.1
`
`vG (v)
`
`(va+2¢,=)1’2 (V‘lz)
`
`Fig. l C-V curves Of thin and
`thick gate Oxide dielectrics.
`
`IG—VG characteristics
`Fig. 2 ID,
`as
`a parameter of substrate
`bias for NMOSFB’I‘ with thin
`oxide gate dielectric.
`
`Fig. 3 VTH as a function of
`substrate bias for NMOSFETs
`with thin and thick oxide gate
`dielectrics.
`
`A 6
`e
`r4
`,—
`g
`.E 2
`2
`8
`
`x
`
`(‘X
`\TEM a. Ellipsometry
`2
`4
`6
`Measured Tox_ph,(nm)
`
`,
`
`0
`
`Fig. 4 TOX,cq obtained by VTH
`dependence on substrate bias
`and
`To”phy measured
`by
`TEM and ellipsometry. To»cq
`= TOX~clc ' Toxrpam'
`
`0) U1
`
`N(n
`
`O
`
`
`
`
`
`NitrogenConcentration(atomicO/n)
`
` \Universal Mobilit
`
`\
`
`Oi_,N*
`7°/o (XPS)
`
`
`
`Fig. 5 SIMS profiles
`of nitrogen in SiON
`using radical processes
`(O*—>N*, O*+N*, and
`N*—>O*).
`
`Fig.6lD, IG-VG characteristics of
`NMOSFETs
`using
`radical
`processes
`(0* and O*—>N*).
`XPS spectra intensity determines
`nitrogen concentration.
`
`,
`
`Fig. 7 Gate leakage current vs.
`Tox_Oq of SiON and SiO2 using
`radical
`processes
`and
`using
`conventional
`thermal
`processes
`(Dry70, Wet»O and NOHDryio).
`
`Eel! (MVlcm)
`
`for
`Fig. 9 Comparison of pm-
`SiON and SiO2 using radical
`processes.
`
`Table 1 Comparison ofdielectricproperties.
`
`Radical Nitrogen Physical/Electrical Dielectric
`Process in SiO2
`hickness
`
`OX-phy/TOX-eq
`
`.
`
`i
`
`‘
`
`iV
`
`n
`
`.'
`
`v
`
`l‘ .
`’
`
`‘
`
`‘
`
`.
`,.
`
`i
`
`'
`‘
`
`i
`V
`" " I.
`
`i;
`
`x
`
`s
`
`..
`
`y N
`
`I'
`
`
`
`
`Cross-sectional
`8
`.1 Fig.
`’1_ photographs.
` (a) Radical nitridation after radical
`oxidation. N=7%, Tox‘cq=l.5 nm.
`(b) Radical oxidation after radical
`
`TEM
`
`'
`
`‘ nitridation. N=12%, T0X_aq=1.9 nm.
`M (c) Radical oxidation. TOX_W=1.5 nm.
`
`unlt)
`
`Tox.Phy'2.5 nm
`VG=-3.6+VFB ‘
`o
`
`0* dN‘ 7%
`TBD>1o°s
`
`102
`
`wt2
`
`10°
`Tan (sec)
`_
`Fig. 11 TBD distributions for SiON and SiO2 using radical processes
`and conventional thermal oxidation.
`(3) Same Tox‘cq.
`SiON using
`O*—>N* with 7% nitrogen was not broken within 1035 at —2.4V+VFB.
`(b) Same Toxphy.
`
`
`
`Intensity(arb.
`
`395
`400
`405
`Binding Energy (eV)
`Fig. 12 XPS Spectra of N
`is orbital for SiON using
`radical processes (N*—)O*
`and O*—>N*).
`
`2000 Symposium on VLSI Technology Digest of Technical Papers
`
`1 17
`
`0.6
`
`o
`
`0.4
`0.2
`LG (,1 m)
`Fig. 10 Comparison of ID for SiON
`and SiO2 using radical processes.
`‘
`
`