1 OA-4
Rapid Thermal Chemical Vapor Deposition
of
I n - S i t u
Nitrogen-Doped Polysilicon for Dual Gate CMOS
S.C. Sun,
L.S.
Wang, F.L. Yeh,
and
C.H. Chen
National Nan0 Device Laboratory and Department of Electronics Engineering
National Chiao Tung University, Hsinchu, Taiwan, R.O.C.
Abstract
A novel gate structure with excellent electrical properties and reliability has been fabricated by in-situ rapid thermal multiprocessing. Gate oxide was grown first by low pressure rapid thermal oxidation in N,O, followed by sequential rapid thermal chemical vapor deposition (RTCVD) of an ultrathin layer (6nm) of nitrogen-doped polysilicon and then undoped polysilicon. Results show the suppression of boron penetration and high device reliability.
Introduction
The dual gate CMOS structure using the p+ polysilicon for PMOSFET has been widely studied to improve its short- channel behavior. Unfortunately, boron penetration from heavily doped p+ polysilicon has caused deterioration of the gate oxide and unstable threshold voltage. To avoid the boron penetration, several structures have been proposed [ 1- 31. In this paper, for the first time, we report the use of a thin layer of nitrogen-doped polysilicon formed by in-situ RTCVD method that has effectively suppressed boron penetration through the ultrathin gate oxide.
Experimental
Fig. 1 shows the schematic cross-section of newly developed dual gate CMOS using nitrogen incorporation into RTCVD polysilicon gate. Gate oxide was grown at 1050 "C in a low-pressure (40 torr) N,O ambient to improve thickness uniformities and reliability [4]. Next, a 60
8,
nitrogen-doped polysilicon film was deposited by introducing SiH, andNH,
gas mixture under 0.4 torr at 750 "C. Then, after evacuating NH, from the chamber, a 3000
8,
undoped polysilicon filmwas deposited. All these steps were performed without
atmospheric exposure in a load-locked rapid thermal reactor.
Results and Discussion
Fig. 2 shows the dependence of deposition rate and nitrogen concentration of the polysilicon film on the NH, to
SiH, flow ratio. SIMS measurement reveals that nitrogen concentration is almost in direct proportion to the flow ratio. Deposition rate decreases slightly with increasing nitrogen concentration until nitrogen doping reaches 5x1OZ1 ~ m - ~ , then it drops rapidly. FTIR data shown in Fig. 3 provides additional evidence of nitrogen incorporation in the polysilicon film with a Si-N peak.
A significant increase in p+ polysilicon sheet resistance is observed for the nitrogen concentration greater than 5x10'' ~311.~. as shown in Fig. 4. This increase may be due to
the addition of nitrogen atoms at grain boundaries. Furthermore, AFM measurement has confirmed that both grain size and surface roughness are decreased with increasing nitrogen doping concentrations.
Fig. 5 shows the high-frequency C-V curves of MOS
capacitors for different nitrogen concentrations. Boron
penetration is obvious on the capacitor without nitrogen- doped layer. The flatband voltage shifts in the negative direction with increasing amount of nitrogen dopings. Boron penetration was effectively suppressed at nitrogen concentration of 5x10m ~ m - ~ . For doping higher than 3x1021 ~ m - ~ , the flatband voltages are far from the desired degenerately-doped values. This indicates a depletion of carrier concentration in the polysilicon/oxide interface, as evident in the quasi-static C-V curves shown in Fig. 6 .
Fig. 7 presents the SIMS depth profiles of boron
measured on three samples after 4x10" cmS2 BF, implant and 900 'C/30 min heat cycle. Sample of 0,-grown oxide without nitrogen doped layer clearly exhibits boron penetration. N,O-grown nitrided oxide sample without nitrogen layer retards the boron diffusion through the Si/SiO, interface, but can not stop the boron penetration. Sample with nitrogen layer displays no boron penetration. At the same time, the segregation of boron into the gate oxide can be reduced.
Fig. 8 demonstrates the effectiveness of nitrogen doped layer even at high BF, doses. Without nitrogen layer, a small flatband voltage shift occurs at ~ X I O ' ~ cm-,. It rises to almost 8 volts at 8x10'' cm*2. Fig. 9 shows the dependence of flatband voltage on gate oxide thickness with BF, dose as a
parameter. Susceptibility to boron penetration increases
drastically as the oxide thickness is scaled down to 43
A.
Fig. 10 shows the subthreshold characteristics of PMOSFET. FETs with nitrogen doped layer = 0, 5x10', and 5x1OZ1 ~ m - ~ have threshold voltage = 0.0, -0.8, and -1.0 volts, respectively. The subthreshold slopes are all around73 mv/decade. Fig. 11 compares the charge trapping
characteristics. Boron penetration into SiO, has resulted in a significant increase of hole trapping on the control sample. Nitrogen doped sample shows reduced charge trapping due to
suppressed boron penetration, resulting in improved QBD.
This improvement in reliability is also reflected in the Weibull plot of charge-to-breakdown as shown in Fig. 12.
Conclusions
We have demonstrated that in-situ nitrogen-doped RTCVD polysilicon film is highly effective in suppressing boron penetration, leading to smaller flatband voltage shift, improved charge trapping and reliability characteristics.
References
[ l ] [2]
[3]
[4]
F.A. Baker, et al., Tech. Dig. of IEDM, p.443, 1989
S . Nakayama, ECS Spring Meeting Proc., p.9, 1991
T. Kuroi,
er
al., Tech. Dig. of IEDM, p. 325, 1993 S.C. Sun, et al., MRS Symp. Proc., 342, p.181, 1994A s + BFZ*
I
i I 1
,undopcdPoly.s,,I
i i
I
1.5 8 2.0 P-Well N-Well,
Subslrale XW"C/ Wrmn B F ~ b s e (cm.') . 4 - \ -3-2x10"Fig. 1 Schematic croos-section of
nitrogen-depod dual-gate CMOS structure
o,o
I
. . . , . . ... . . . . . . _ . ,l 1 0 ~ ~
10' IO' I O '
N H ~ I H , Flow Ralio
Fig. 2 Deposition rate and nitrogen concentration as a function of
NH,/SiH, flow ratio
RTCVD Poly Si=3000
IN] = 5x1OZ1 cm.3
Si-Si=[ 100
I
Fig. 3 FTIR data of in-situ nitrogen-
doped RTCVD polysilicon
film
/
/ L
_ _ _ _ _ - _ _ - -IO'
'
IO" 10'9 IP I @ '
Nitrogen Concentration (atom/cm')
Fig. 4 P+ polysilicon sheet resistance as a f u n c t i o n of n i t r o g e n concentration 1.1
,
1.0 0.9 0.8 I 0.7(si:
0.8-
0.5 0.4 0.2!
0 1 0 0 3 2 1 .o 1 2 3 4 5 V G (VUIIP)Fig. 5 High-frequency C-V curves of p+ polysilicon gate capacitors with d i f f e r e n t n i t r o g e n d o p i n g concentrations [Nl=O 1 . 2 , I 1 0.8 0.6 0.4 0.2 0 n
.
- 2 - 1 0 1 2 3 vg (volts)Fig. 6 Quasi-static C-V characteristics of p+ gate capacitors
Poly-Si -.sio,+ si -.I
-
w0
1
0.JFig. 7
SIMS
depth profiles of boronof samples with and without [NI-
doped layer 8
M ) N z 0 Oiille Annerd! YO0 T. 60 inin
6
.
+IN14-
-
+[N]=SxlOmem-'-
0 2 4 6 8 1 0
BF2 Implant Dose ( X I O ' ~ ion/cm2)
Fig. 8 Flatband voltage as a function of BF2 implant dose for samples with
and without nitrogen-doped layer
3.0 -I
O3
I
O.Ot.---' . . ' ' ' ' '
30 40 50 M) 70 80 90 IW 110
M,O Oxide Thickness
(A)
Fig. 9 Dependence of flatband voltage shift on the gate oxide thickness.
Samples have 6nm nitrogen-doped
layer h 1 0 7 m 109 IO" 0 4 0 2 0 02 0 4 06 0 8 I I 2 -vgs (volts)
Fig. 10 Subthreshold current vs. gate voltage. Shifts in VTp are caused by boron penetration
Y
0
L 0.1
0 110 20 30 40 SO
Stress Time (sec)
F i g . 11 C h a r g e t r a p p i n g characteristics for capacitors with and without nitrogen-doped layer
f X ,020 cm-3
I 0. 10.1 lo" Id Charge-lo-Breakdown (Clem')
Fig.
12
Cumulative failure of charge-to-breakdown (QBD) in the TDDB test
for samples with and without nitrogen layer