Presentation done at IEEE SSDM in October 2009

1. Charge Localization During Program and
Retention in NROM-like Non Volatile
Memory Devices
Etienne Nowak, Elisa Vianello*, Luca Perniola, Marc
2007
Bocquet, Gabriel Molas, Rabah Kies, Marc Gely, Gerard
Ghibaudo+, Barbara De Salvo, Gilles Reimbold, Fabien
Boulanger
CEA/LETI-Minatec, 38054 Grenoble, France
* DIEGM, University of Udine, Italy +IMEP/INPG
Grenoble, France
etienne.nowak@cea.fr
Etienne Nowak et al. – SSDM 2009 1

2. Outline
Motivation
Methodology
Program operation
2007 Retention
Conclusion
Etienne Nowak et al. – SSDM 2009 2

3. Outline
Motivation
Methodology
Program operation
2007 Retention
Conclusion
Etienne Nowak et al. – SSDM 2009 3

4. Motivation (1/2)
12V NROM -14V
0V 4.5V 0V 7V
e- h+
Write (CHE) Erase (HHI) 4bit/cell 8Gbit product
R.Sahar et al, ISSCC 2008
Benefit:
2007
Higher information density thanks to physically
separated bits
Purpose of the work:
Extract information on pocket of trapped charges
in alternative trapping materials for NROM devices
Etienne Nowak et al. – SSDM 2009 4

5. Motivation (2/2)
Retention of cycled and uncycled
Si3N4 devices has been well studied
M. Janai et al., IEEE IRPS Tech. Dig., 2008, pp417-423
Few works have been done on
different trapping layers
T.Sugizaki et al., VLSI Tech Dig., 2003, pp.27-28.
2007
Intrinsic trapping properties of Si3N4,
HfO2, Al2O3 still not well understood
Maximum amount of trapped charge
Localization of the trapped charge
∆Vt loss mechanisms on different material
Etienne Nowak et al. – SSDM 2009 5

6. Outline
Motivation
Methodology
Program operation
2007 Retention
Conclusion
Etienne Nowak et al. – SSDM 2009 6

7. Devices under analysis
Control
N+ Poly N+ Poly N+ Poly Gate
HTO HTO Blocking
10nm HTO 10nm 10nm oxide
6nm 6nm Trapping
6nm Si3N4 HfO2 Al2O3 layer
5nm SiO2 5nm SiO2 5nm SiO2 Tunnel
oxide
2007
W/L=10/0.27 µm
Three different trapping layers are compared
LPCVD Si3N4 / ALCVD HfO2 / ALCVD Al2O3
Etienne Nowak et al. – SSDM 2009 7

8. Method to extract trapped charges information
Virgin Written
1E-4 VS=1.5V Reverse Read
VG
Source Current IS [A/um]
VD=1.5V Forward Read
1E-6 SiO2
∆VtR Si3N4/HfO2/Al2O3 Qcharged
1E-8 SiO2
1E-10
∆VtF VS y x Lcharged VD
L
1E-12
1E-14
2007 0 2 4 6 8 10
Gate Voltage VG [V]
1 - Measure ∆VtR and ∆VtF from the
experimental results
L. Perniola et al., IEEE TNANO, 2005 Etienne Nowak et al. – SSDM 2009 8

9. Method to extract trapped charges information
∆VtF
Charge Density Q charged [10 12cm -2] VG
SiO2
Si3N4/HfO2/Al2O3 Qcharged
SiO2
VS y x Lcharged VD
L
Virgin Written
1E-4 VS=1.5V Reverse Read
Source Current IS [A/um]
VD=1.5V Forward Read
1E-6
∆VtR
1E-8
1E-10
∆VtF
2007
∆VtR 1E-12
1E-14
0 2 4 6 8 10
40 60 80 100 120 140 Gate Voltage VG [V]
Effective charged Length 2L[nm] [nm]
Charged Length L charged
2 - Extrapolate the values of Lcharged and Qcharged from
an analytical map calculated through the ψS approach
L. Perniola et al., IEEE TNANO, 2005 Etienne Nowak et al. – SSDM 2009 9

10. Outline
Motivation
Methodology
Program operation
2007 Retention
Conclusion
Etienne Nowak et al. – SSDM 2009 10

11. Program
HfO2 Al2O3 Si3N4
Programming Window ∆VtR [V]
Programming Window ∆VtR [V]
12 Stress Stress 12
V =V =0V VS=VB=0V
10 S B 10
VG=10V VG=12V
8 V =5V 8
D VD=5V
6 6
4 4
2 2
2007
0 0
-6 -4 -2 0 -6 -4 -2 0
10 10 10 10 10 10 10 10
Stress Time t [s] Stress Time t [s]
Programming windows over 10 V for the 3
materials
Etienne Nowak et al. – SSDM 2009 11

12. Charge localization
Charge Density Qcharged [10 cm-2]
18
Gate
16
12
14
12
10 t ~ 0.01s Source Drain
8 Gate
Vg=10V Vg=12V ; Vd=5V
6
HfO2
4 Al2O3
2 Source Drain
Si3N4
2007
0
0 50 100 150 200 250
Effective charged Length Lcharged[nm]
1. Charge “initially” localizes at ~40-60 nm next to drain
2. After t~10 ms, Qcharged saturates, then Lcharged broadens
3. Not significant difference between the trapping layers
Etienne Nowak et al. – SSDM 2009 12

13. Ey evolution during program
Normal Field EY [MV/cm]
1.4 Lcharged 1.4 Lcharged
Normal Field EY [MV/cm]
Vd=5V Vg=12V Vd=5V Vg=12V
1.2 1.2
Gate
1.0 Gate 1.0
0.8 12
Qcharged to 20x10 cm
=0
-2
0.8
0.6 12
every 2.5x10 cm
-2
0.6 EY
0.4 0.4
Source 12 -2
Drain
Qcharged=17.5x10 cm
Source
0.2 EY Drain 0.2
0.0 Effective LengthLeff Effective Length Leff
0.0
-0.2 -0.1 0.0 0.1 0.2 -0.2 -0.1 0.0 0.1 0.2
Source
2007
Position X [um] Drain Source Position X [um] Drain
Maximum Qcharged and subsequent Lcharged
broadening explained by:
Decrease of Ey at the Si/SiO2 interface
Ey peak shift towards the source side
Etienne Nowak et al. – SSDM 2009 13

14. Outline
Motivation
Methodology
Program operation
2007 Retention
Conclusion
Etienne Nowak et al. – SSDM 2009 14

15. Gate
Retention operation
Source Drain
Programming Window ∆VtR [V]
120
5
Total Charge Variation
Qchargedx Lcharged [%]
4
100
3
2 T=25° T=125°
C C 80 T=25° T=125°
C C
HfO2 HfO2
1 Al2O3
Al2O3
60 Si3N4
0 Si3N4
0 1 2 3 4 5 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10
2007
Time t [s] Time t [s]
Lateral charge migration is the main ∆VtR loss
mechanism for the three materials at 25° C.
Charge loss is relevant only for Al2O3 at 125°
C
Etienne Nowak et al. – SSDM 2009 15

16. Retention Model to extract the ∆Lcharged
Drift-Diffusion equations
 ∂n ( x , t ) ∂ ∂ 2 n( x , t )
 = [µeff n(x , t )E (x , t )] + D
 ∂t ∂x ∂x 2

 ∂E ( x , t ) = qn( x , t )
 ∂x εr ε0 Charge density Shape1

Qcharged Drain
Shape2 side
Source
Gate side
2007
Lcharged Position
Source Drain
∂n( x, t )
V =0 =0
∂x
1D model of nitride
Drift-Diffusion of majority carriers
Effective mobility coefficient µeff
Etienne Nowak et al. – SSDM 2009 16

17. Retention Model to extract the ∆Lcharged
Effective charged Length
150
Drift-Diffusion, Shape1
increase∆ Lcharged [nm]
Drift-Diffusion, Shape2
Charge density
Qcharged
Shape1
Drain 100 Diffusion, Shape1
Shape2 side
Source
side
Lcharged Position
∂n( x, t )
50
V =0
∂x
=0
∆ Lcharged=A*ln(t)
A=α µeffQtotal/εrε
0
2007
-3 -1 1 3 5 7
10 10 10 10 10 10
Timet [s]
Drift predominant over Diffusion
∆Lcharged independent of the shape of the trapped
charges
Drift follows an empirical law:
Lcharged=Lcharged0+A*ln(t)
Etienne Nowak et al. – SSDM 2009 17

18. Data vs model: lateral charge migration
25 HfO2
Effective Charged Length
increase ∆Lcharged [nm]
Al2O3
20 A=3.7
Si3N4
15
10 A=1.7
5
A=0.55
0
2007 3 4 5
10 10 10
Time t [s]
Lateral migration for the three material follows a
logarithmic law with different A coefficient
Lowest drift observed for Si3N4
Etienne Nowak et al. – SSDM 2009 18

19. Outline
Motivation
Methodology
Program operation
2007 Retention
Conclusion
Etienne Nowak et al. – SSDM 2009 19

20. Conclusion
Comparative study of trapping properties in
Si3N4, HfO2 and Al2O3 in program and
retention conditions
Large window (~10 V) possible for all trapping materials
Maximum Qcharged is limited by electrostatics not by the
trapping layer properties
Method allows separating vertical vs lateral
2007
charge migration
Lateral migration, due to charge drift, is the main Vt shift
mechanism in retention mode for the three materials at
25°C
Log(t) dependence of lateral migration, and Si3N4 shows
the lowest drift
Etienne Nowak et al. – SSDM 2009 20

21. Thanks for your attention!
2007
Etienne Nowak et al. – SSDM 2009 21

22. Extraction Method
Analytical model
Based on Liu Surface
Potential model
Calculate ∆VtR and
∆VtF for a given
Qcharged and Lcharged
Extract Qcharged and
2007
Lcharged from measured
∆VtR and ∆VtF
[L. Perniola et al., IEEE Trans. on
Nanotech., Vol. 4, No. 3, pp. 360-
368, May 2005] Etienne Nowak et al. – SSDM 2009 22

23. Retention Model
Drift-Diffusion equation
 ∂n ( x , t ) ∂ ∂ 2n(x , t )
 = [µeff n(x , t )E (x , t )] + D
 ∂t ∂x ∂x 2

 ∂E ( x , t ) = qn( x , t )
 ∂x
 εr ε0
Diffusion equation Drift equation
2007
D=0 µeff = 0
∂n ( x , t ) ∂ 2n( x , t )  x
n( x , t )dx 
=D ∂n ( x , t ) ∂  ∫
∂t ∂x 2 = A* n( x , t ) 0∞ 
∂t ∂x  n( x , t )dx 

 ∫0 

µeff q ∫ n( x , t )dx
∞
µeff Qtotal
A =
* 0
=
εr ε0 εr ε0
Etienne Nowak et al. – SSDM 2009 23