Documente Academic
Documente Profesional
Documente Cultură
00, 00 2014
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
I. INTRODUCTION
34
35
36
37
38
39
40
41
42
43
44
IE
E
Pr E
oo
f
Manuscript received January 29, 2014; accepted July 31, 2014. Date of publication; date of current version. The review of this paper was arranged by
Associate Editor A. Martinez.
The authors are with the Microelectronics and VLSI Group, Department
of Electronics and Communication Engineering, Indian Institute of Technology Roorkee, Roorkee 247667, India (e-mail: shivamvermaid@gmail.com;
shalupec@iitr.ac.in; bkk23fec@iitr.ac.in).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TNANO.2014.2346790
1536-125X 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
Q1
127
PMA MTJs offer low critical currents and higher thermal stability and scalability for STT switching. Lee et al. [11] demonstrated a PMA in Fe-rich CoFeB free layers due to the reduction
of demagnetizing field. The demagnetizing field decreases with
increase in perpendicular anisotropy and eventually leads to a
layer with perpendicular easy axis. The effective demagnetizing
(4Me ) field is expressed as
128
Fig. 1. (a) Typical STT MRAM cell. (b) Front view of the proposed STT
MRAM cell.
93
94
95
96
97
98
99
100
101
creating Verilog-A models based on experimental results. Section IV focuses on the design considerations (read, write margins and switching probability) of an STT MRAM cell from the
perspective of select devices. Section V describes the TCAD
analysis of the proposed buried source GAA device followed
by the load line analysis. Section VI analyzes and compares
the HSPICE simulation results of the proposed STT MRAM
cell with conventional STT MRAM cell. Finally, in Section
VII, conclusions are drawn based on the analysis done in the
previous sections.
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
131
132
133
134
(1)
IE
E
Pr E
oo
f
92
129 Q2
130
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149 Q3
150
151
152
153
The select device in a typical memory system allows a memory cell to be accessible for read and write. This section describes
the design considerations of STT MRAMs from the perspective
of select devices. Fig. 2(a) and (b) represents the two states of
the STT MRAM cell before the write operation is performed.
Fig. 2(a) shows the equivalent circuit with polarity of BL and
SL trying to write a 0 into the cell (P write). Likewise, Fig. 2(b)
shows the equivalent circuit with the polarities of BL and SL
trying to write a 1 into the cell (AP write). Hence, an STT
MRAM cell can be viewed as an NMOS transistor with RP or
RAP connected to it as load.
The switching of MTJ depends on the direction of current and
also a minimum threshold current (IC0 ) required for switching
to occur. Hence, the primary consideration is that whether the
select device can provide sufficient drive current for MTJ switching in either direction. The driving current must be sufficiently
high to surmount the switching threshold with high switching
probability for proper operation of the STT MRAM cell.
A write error occurs when the strength of the write current is
not enough, and there is a probability that the desired data may
not be written into the cell [15]. Furthermore, a read disturb
is the unintentional write due to the read current flowing in a
direction; this lowers the switching energy barrier [15], [16].
The write error rate (WER) and read disturb rate (RDR) for
STT-MRAM cell are determined from the switching probability.
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
TABLE I
PRACTICALLY FABRICATED MTJ PARAMETERS FOR CURRENT PULSEWIDTH OF 1 ns
Barrier/recording layer structure (nm)
MgO(0.85)/CoFeB(1.7)/Ta
Diameter
(nm)
40
TMR
(%)
113
RA
(m2 )
16
= E /k B T
R P (k)
R A P (k)
I L H 0 (A)
I H L 0 (A)
V W (V)
Reference
39
12.73
27.1
72
28
0.608
[7]
V W = J C 0 (RA) is the mean write voltage, where J C 0 is the switching current density and RA is the resistance area product. R A P = 27.1 k and TM R = 113% are the
zero-bias values for the PMA MTJ from [7].
are expressed as
200
IE
E
Pr E
oo
f
RMP =
Fig. 2. (a) Equivalent circuit of STT MRAM cell before MTJ switching with
AP resistance as load. (b) Equivalent circuit of STT MRAM cell before MTJ
switching with P resistance as load. (c) Top view of the vertical GAA NMOS
device. (d) Front view of the vertical GAA NMOS device.
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
WER = 1 PSW
(2)
IHL0 IP0
,
IHL0
RMAP =
ILH0 IAP0
ILH0
201
202
203
204
205
206
207
208
209
210
211
(5)
where IP0 and IAP0 are the read currents during P and AP
reading, respectively.
Write access time (tP and tAP ) is defined as the time required
for switching the state of MTJ at the particular write current
and write current pulsewidth. The switching time also depends
on the switching threshold at that write current pulse duration
(tPW ) [14].
218
219
220
212
213
214
215
216
217
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
TABLE II
TCAD RESULTS OF GAA DEVICE FOR DIFFERENT GATE LENGTHS
Lch
40
80
120
Vt
l i n (V)
DIBL (mV/V)
Io n (A)
Io n /Io f f
SS (mV/decade)
115
22.1
13.6
214.6
194
175.5
3.8103
3.3106
1.5107
103.0
63.4
60.3
0.279
0.300
0.307
TABLE III
TCAD RESULTS OF PLANAR NMOS DEVICE
W (nm) L c h (nm) V t
40
40
80
243
244
245
246
247
248
249
250
251
252
253
254
0.250
0.275
0.275
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
85.8
72.8
145.6
18.29
142.8
142.7
768.3
294.2
294.2
255
354
280
310
SS (mV/decade)
IE
E
Pr E
oo
f
242
80
120
120
l i n (V)
Vt
VDS
Vt sat
sat VDS lin
lin
(6)
where Vt lin and Vt sat are the threshold voltage in the linear
(very low VDS ) and saturation regions, respectively. The values
of VDS sat and VDS lin are 1.6 and 0.05 V, respectively. Vt sat
is the value of VGS on ID VGS curve at VDS = VDS sat required to get the same value of current, which is obtained when
VGS = Vt lin and VDS = VDS lin . The subthreshold slope (SS)
is the change in VGS required to alter the subthreshold drain
current by one decade (ten times). The corresponding results
are shown in Table II. Although the device with 40 nm gate
length has the largest current drive, but it severely suffers from
short channel effects. The DIBL, off-current, and subthreshold
slope parameters are comparatively large for the device with
40 nm gate length. It is because of lower electrostatic gate control at smaller gate length. Evidently, the device with 120 nm
gate length demonstrates the best performance. In order to compare with the conventional STT MRAM cell, TCAD simulations are carried out for planar NMOS transistor also. The gate
work function and oxide thickness of the planar NMOS are
4.61 eV and 1.5 nm, respectively. However, the source, drain,
and channel doping is same as that for GAA NMOS. The performance parameters calculated for planar transistor for different
device dimensions are placed in Table III. The performance of
the planar transistor is poor, especially for small gate length,
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
302
303
304
305
306
307
308
length are also included in the load line analysis. The minimum width (W = F = 40 nm) planar NMOS has a low Ion as
compared to GAA (Tables II and III). Hence, the comparison
would be more appropriate with a planar NMOS of width 80 nm
(W = 2F = 80 nm), that has Ion comparable to the GAA device. The value of VDS , as deduced from Fig. 2(a) and (b) is
expressed as
VDS = VDD ID RM TJ
309
Q4 310
311
312
(7)
(8)
348
349
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
350
351
352
353
354
355
356
357
358
359
360
361
IE
E
Pr E
oo
f
347
313
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
IE
E
Pr E
oo
f
Fig. 5. Load-line analysis of vertical GAA (40 nm nanowire diameter and L ch = 120 nm) and planar NMOS (gate length of 120 nm) at different V G S (= V D D )
corresponding to P and AP resistance as load. (a) For V G S = 1.6 V, (b) for V G S = 1.4 V, and (c) for V G S = 1.2. Red and black lines represent P and AP load
lines, respectively, in each of (a), (b), and (c). IL H 0 and IH L 0 are 72 and 28 A, respectively.
Fig. 6. (a) Cell layout of conventional 10F 2 STT MRAM cell with planar select device. (b) Comparison of ID V G S characteristics of calibrated BSIM CG
with TCAD results at V D S = 1.6 V. (c) Comparison of ID V D S characteristics of calibrated BSIM-CG with TCAD results.
Fig. 7. (a) Variation of TMR and read current difference with read voltage (WWL = V D D = 1.6 V and SL = 0 V). (b) Variation of read margin (RM P )
and write margins WM P with WWL = V D D . The analysis is done keeping BL = 0.4 V and SL = 0 V with the initial state of MTJ considered as AP.
(c) Comparison of dynamic power dissipation (with WWL = 1.6 V) of the two cells in terms of V D D .
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
TABLE IV
TIMING ANALYSIS OF THE PROPOSED 4F2 GAA CELL AND 10F2 PLANAR CELL
Bias voltages
Planar
Write 0
Hold
Read 0 (P read)
Hold
Write 1
Hold
Read 1 (P read)
Hold
SL
BL
WWL
I c e l l (A)
R M T J (k)
TMR
MTJ State
I c e l l (A)
R M T J (k)
TMR
MTJ state
0
0
0
0
27.1
1.13
AP (1)
0
27.1
1.13
AP (1)
0
1.6 V
1.6 V
55.1
12.73
P (0)
61.3
12.73
P (0)
0
0
0
0
12.73
P (0)
0
12.73
P (0)
0
0.4 V
1.6 V
25.2
12.73
P(0)
26.5
12.73
P(0)
0
0
0
0
27.1
P(0)
0
12.73
P(0)
1.6V
0
1.6V
92
14.46
0.14
AP (1)
95.8
14.41
0.13
AP (1)
0
0
0
0
27.1
1.13
AP (1)
0
27.1
1.13
AP (1)
0
0.4 V
1.6 V
15.7
22.4
0.76
AP (1)
16.3
22.20
0.74
AP (1)
0
0
0
0
27.1
1.13
AP (1)
0
27.1
1.13
AP (1)
423
the proposed 4F 2 and conventional 10F 2 cell. The corresponding results are tabulated in Table V. Undoubtedly, the proposed
cell demonstrates a better performance in terms of power dissipation and write margins. The leakage power dissipation for
the proposed cell is four to five orders of magnitude lower than
conventional cell (when the cell is not selected for writing). The
dynamic power dissipation at WWL = 1.6 V (VDD ) is higher;
although, here the point of consideration is that the dynamic
power always has a tradeoff with the P and AP write margins
(write currents). There is a larger tradeoff window between the
write margin and dynamic power dissipation in the case of proposed cell with GAA device, as it can operate with VDD as low
as 1.3 V (at tPW = 1 ns) by sacrificing the write margins. It is
clearly observed in Fig. 7(b) that the planar cell cannot operate
correctly at VDD = 1.3 V for tPW = 1 ns. Moreover, a comparison of the dynamic power dissipation between the cells in
terms of VDD in Fig. 7(c) shows that the proposed cell can be
optimized for low-power operation also.
424
VII. CONCLUSION
425
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
426
427
428
IE
E
Pr E
oo
f
GAA
Parameter
APPENDIX I
BSIM CG MODEL CALIBRATION
The important parameters and model calibration methodology for configuring BSIM CG for cylindrical GAA operation
are classified and discussed as follows:
A. Geometry and Material Parameters: The geometry
and material parameters of BSIM CMG are set for a vertical
GAA device operation, according to the parameters used for
TCAD simulation (see Table VI).
B. Threshold Voltage Calibration: The threshold voltage
of BSIM CG is expressed as
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
where Vth0 is the threshold voltage of the model based on surface potential calculations. Vth,SCE is the threshold voltage
degradation due to short channel effects, Vth,DIBL is used
to model the effect of drain voltage on threshold voltage, and
DVTSHIFT is to handle any additional shift in Vth [21].
The model equation for Vth,SCE in BSIM CG [21] is
Vth,SCE =
0.5DVT0
(Vbi st )
cosh {DVT1 (Le /)} 1
(A2)
458
459
460
461
462
463
TABLE V
COMPARISON BETWEEN PROPOSED CELL AND 10F2 PLANAR NMOS
Select Device
GAA
Planar
W M P (A)
W M A P (A)
t p (ns)
t A P (ns)
Power (pW) W W L = 0 V
23.3
17.6
34.0
29.0
0.43
0.52
0.25
0.32
16.2
1140
112.4
107.0
TABLE VII
THRESHOLD VOLTAGE PARAMETERS
TABLE VI
GEOMETRY AND MATERIAL PARAMETERS
GEOMOD
BULKMOD
ASYMMMOD
COREMOD
L
D
EOT
NGATE
NSD
PHIG
NBODY
Value
Description[21]
3
0
1
0
120 nm
40 nm
2 nm
0
1 1020 cm3
4.61 eV
1 1016 cm3
Cylindrical GAA
SOI substrate
Asymmetric device
Surface potential model
Gate length
Gate diameter
Gate oxide thickness
Metal gate
S/D doping concentration
Gate work function
Channel doping
VALUE
Description [21]
DVT0
DVT1
ETA0
DSUB
DVTP0
DVTP1
DVTSHIFT
2.2
0.177
1
0.80
0
0
0
SCE coefficient
SCE exponent coefficient
DIBL coefficient
DIBL exponent coefficient
Coefficient for drain-induced V t h shift
Exponent coefficient for drain-induced V t h shift
Additional V t h shift handle
TABLE VIII
MOBILITY AND DRAIN CURRENT PARAMETERS
IE
E
Pr E
oo
f
464
465
466
467
468
where
Vbi = (kB T /q) ln NSD (nb o dy /n2i ) ,
and
=
Vth,DIBL =
0.5ETA0
VDS
cosh (DSUB (Le /)) 1
U0
IDS0MULT
U0MULT
Description [21]
TABLE IX
PARASITIC CAPACITANCE, SUBTHRESHOLD, AND LEAKAGE CURRENT
PARAMETERS
(A3)
Value
CGEOMOD
CGEO1SW
COVS
COVD
DVT1SS
GIDLMOD
AGISL
AGIDL
Value
Description [21]
1
1
0.1 fF
0.087 fF
1.0234
1
1.05 1015
1.05 1015
REFERENCES
[1] ERD. (2011). International technology road map for semiconductors [Online]. Available: http://www.itrs.net
[2] T. Schloesser et al., 6F2 buried wordline DRAM cell for 40 nm and
beyond, in Proc. IEEE IEDM08, San Francisco, CA, USA, pp. 14,
Dec. 2008.
[3] H. Chung et al., Novel 4F2 DRAM cell with vertical pillar transistor(VPT), in Proc. IEEE ESSDRC, Helsinki, Finland, pp. 211214, Sep.
2011.
[4] Z. Fang et al., Fully CMOS-compatible 1T1R integration of vertical
nanopillar GAA transistor and oxide-based RRAM cell for high-density
nonvolatile memory application, IEEE Trans. Electron Devices, vol. 60,
no. 3, pp. 11081113, Mar. 2013.
[5] D.-L. Kwong et al., Vertical silicon nanowire platform for low power
electronics and clean energy applications, J. Nanotechnol., vol. 2012,
pp. 121, 2012.
[6] T. Kawahara, K. Ito, R. Takemura, and H. Ohno, Spin-transfer torque
RAM technology: Review and prospect, Microelectron. Reliab., vol. 52,
no. 4, pp. 613627, Apr. 2012.
490
491
492
493
494
495 Q5
496
497
498
499
500
501
502
503
504
505
506
507
508
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
560
561
562
563
564
565
566
567
568
569
570
571
572
573
IE
E
Pr E
oo
f
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
Q6 541
542
543
544
545
546
547
548
549
550
551
552
553
Q7 554
555
556
557
558
Q8 559
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
QUERIES
597
598
599
600
601
602
603
604
Q1.
Q2.
Q3.
Q4.
Q5.
Q6.
Q7.
Q8.
Author: Please check the acronym of word line. Is this OK as is or should be WL.
Author: Please check the edited sentence PMA MTJs offer . . . switching. for intended meaning.
Author: Please provide the full form of the acronyms BSIM CG and TMR.
Author: The variable j has been set as R MTJ. Please check.
Author: Please provide all the names of authors in Refs. [2][5], [7], [8], [10], [12][14], [16], [21], and [24].
Author: Please verify Ref. [18] as set.
Author: Please provide the city of the publisher in Ref. [22].
Author: Please provide the month information in Ref. [24].
IE
E
Pr E
oo
f
596
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
I. INTRODUCTION
34
35
36
37
38
39
40
41
42
43
44
IE
E
Pr E
oo
f
Manuscript received January 29, 2014; accepted July 31, 2014. Date of publication; date of current version. The review of this paper was arranged by
Associate Editor A. Martinez.
The authors are with the Microelectronics and VLSI Group, Department
of Electronics and Communication Engineering, Indian Institute of Technology Roorkee, Roorkee 247667, India (e-mail: shivamvermaid@gmail.com;
shalupec@iitr.ac.in; bkk23fec@iitr.ac.in).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TNANO.2014.2346790
1536-125X 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
Q1
127
PMA MTJs offer low critical currents and higher thermal stability and scalability for STT switching. Lee et al. [11] demonstrated a PMA in Fe-rich CoFeB free layers due to the reduction
of demagnetizing field. The demagnetizing field decreases with
increase in perpendicular anisotropy and eventually leads to a
layer with perpendicular easy axis. The effective demagnetizing
(4Me ) field is expressed as
128
Fig. 1. (a) Typical STT MRAM cell. (b) Front view of the proposed STT
MRAM cell.
93
94
95
96
97
98
99
100
101
creating Verilog-A models based on experimental results. Section IV focuses on the design considerations (read, write margins and switching probability) of an STT MRAM cell from the
perspective of select devices. Section V describes the TCAD
analysis of the proposed buried source GAA device followed
by the load line analysis. Section VI analyzes and compares
the HSPICE simulation results of the proposed STT MRAM
cell with conventional STT MRAM cell. Finally, in Section
VII, conclusions are drawn based on the analysis done in the
previous sections.
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
130
131
132
133
134
(1)
IE
E
Pr E
oo
f
92
129 Q2
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149 Q3
150
151
152
153
The select device in a typical memory system allows a memory cell to be accessible for read and write. This section describes
the design considerations of STT MRAMs from the perspective
of select devices. Fig. 2(a) and (b) represents the two states of
the STT MRAM cell before the write operation is performed.
Fig. 2(a) shows the equivalent circuit with polarity of BL and
SL trying to write a 0 into the cell (P write). Likewise, Fig. 2(b)
shows the equivalent circuit with the polarities of BL and SL
trying to write a 1 into the cell (AP write). Hence, an STT
MRAM cell can be viewed as an NMOS transistor with RP or
RAP connected to it as load.
The switching of MTJ depends on the direction of current and
also a minimum threshold current (IC0 ) required for switching
to occur. Hence, the primary consideration is that whether the
select device can provide sufficient drive current for MTJ switching in either direction. The driving current must be sufficiently
high to surmount the switching threshold with high switching
probability for proper operation of the STT MRAM cell.
A write error occurs when the strength of the write current is
not enough, and there is a probability that the desired data may
not be written into the cell [15]. Furthermore, a read disturb
is the unintentional write due to the read current flowing in a
direction; this lowers the switching energy barrier [15], [16].
The write error rate (WER) and read disturb rate (RDR) for
STT-MRAM cell are determined from the switching probability.
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
TABLE I
PRACTICALLY FABRICATED MTJ PARAMETERS FOR CURRENT PULSEWIDTH OF 1 ns
Barrier/recording layer structure (nm)
MgO(0.85)/CoFeB(1.7)/Ta
Diameter
(nm)
40
TMR
(%)
113
RA
(m2 )
16
= E /k B T
R P (k)
R A P (k)
I L H 0 (A)
I H L 0 (A)
V W (V)
Reference
39
12.73
27.1
72
28
0.608
[7]
V W = J C 0 (RA) is the mean write voltage, where J C 0 is the switching current density and RA is the resistance area product. R A P = 27.1 k and TM R = 113% are the
zero-bias values for the PMA MTJ from [7].
are expressed as
200
IE
E
Pr E
oo
f
RMP =
Fig. 2. (a) Equivalent circuit of STT MRAM cell before MTJ switching with
AP resistance as load. (b) Equivalent circuit of STT MRAM cell before MTJ
switching with P resistance as load. (c) Top view of the vertical GAA NMOS
device. (d) Front view of the vertical GAA NMOS device.
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
WER = 1 PSW
(2)
IHL0 IP0
,
IHL0
RMAP =
ILH0 IAP0
ILH0
201
202
203
204
205
206
207
208
209
210
211
(5)
where IP0 and IAP0 are the read currents during P and AP
reading, respectively.
Write access time (tP and tAP ) is defined as the time required
for switching the state of MTJ at the particular write current
and write current pulsewidth. The switching time also depends
on the switching threshold at that write current pulse duration
(tPW ) [14].
218
219
220
212
213
214
215
216
217
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
TABLE II
TCAD RESULTS OF GAA DEVICE FOR DIFFERENT GATE LENGTHS
Lch
40
80
120
Vt
l i n (V)
DIBL (mV/V)
Io n (A)
Io n /Io f f
SS (mV/decade)
115
22.1
13.6
214.6
194
175.5
3.8103
3.3106
1.5107
103.0
63.4
60.3
0.279
0.300
0.307
TABLE III
TCAD RESULTS OF PLANAR NMOS DEVICE
W (nm) L c h (nm) V t
40
40
80
243
244
245
246
247
248
249
250
251
252
253
254
0.250
0.275
0.275
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
85.8
72.8
145.6
18.29
142.8
142.7
768.3
294.2
294.2
255
354
280
310
SS (mV/decade)
IE
E
Pr E
oo
f
242
80
120
120
l i n (V)
Vt
VDS
Vt sat
sat VDS lin
lin
(6)
where Vt lin and Vt sat are the threshold voltage in the linear
(very low VDS ) and saturation regions, respectively. The values
of VDS sat and VDS lin are 1.6 and 0.05 V, respectively. Vt sat
is the value of VGS on ID VGS curve at VDS = VDS sat required to get the same value of current, which is obtained when
VGS = Vt lin and VDS = VDS lin . The subthreshold slope (SS)
is the change in VGS required to alter the subthreshold drain
current by one decade (ten times). The corresponding results
are shown in Table II. Although the device with 40 nm gate
length has the largest current drive, but it severely suffers from
short channel effects. The DIBL, off-current, and subthreshold
slope parameters are comparatively large for the device with
40 nm gate length. It is because of lower electrostatic gate control at smaller gate length. Evidently, the device with 120 nm
gate length demonstrates the best performance. In order to compare with the conventional STT MRAM cell, TCAD simulations are carried out for planar NMOS transistor also. The gate
work function and oxide thickness of the planar NMOS are
4.61 eV and 1.5 nm, respectively. However, the source, drain,
and channel doping is same as that for GAA NMOS. The performance parameters calculated for planar transistor for different
device dimensions are placed in Table III. The performance of
the planar transistor is poor, especially for small gate length,
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
302
303
304
305
306
307
308
length are also included in the load line analysis. The minimum width (W = F = 40 nm) planar NMOS has a low Ion as
compared to GAA (Tables II and III). Hence, the comparison
would be more appropriate with a planar NMOS of width 80 nm
(W = 2F = 80 nm), that has Ion comparable to the GAA device. The value of VDS , as deduced from Fig. 2(a) and (b) is
expressed as
VDS = VDD ID RM TJ
309
Q4 310
311
312
(7)
(8)
348
349
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
350
351
352
353
354
355
356
357
358
359
360
361
IE
E
Pr E
oo
f
347
313
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
IE
E
Pr E
oo
f
Fig. 5. Load-line analysis of vertical GAA (40 nm nanowire diameter and L ch = 120 nm) and planar NMOS (gate length of 120 nm) at different V G S (= V D D )
corresponding to P and AP resistance as load. (a) For V G S = 1.6 V, (b) for V G S = 1.4 V, and (c) for V G S = 1.2. Red and black lines represent P and AP load
lines, respectively, in each of (a), (b), and (c). IL H 0 and IH L 0 are 72 and 28 A, respectively.
Fig. 6. (a) Cell layout of conventional 10F 2 STT MRAM cell with planar select device. (b) Comparison of ID V G S characteristics of calibrated BSIM CG
with TCAD results at V D S = 1.6 V. (c) Comparison of ID V D S characteristics of calibrated BSIM-CG with TCAD results.
Fig. 7. (a) Variation of TMR and read current difference with read voltage (WWL = V D D = 1.6 V and SL = 0 V). (b) Variation of read margin (RM P )
and write margins WM P with WWL = V D D . The analysis is done keeping BL = 0.4 V and SL = 0 V with the initial state of MTJ considered as AP.
(c) Comparison of dynamic power dissipation (with WWL = 1.6 V) of the two cells in terms of V D D .
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
TABLE IV
TIMING ANALYSIS OF THE PROPOSED 4F2 GAA CELL AND 10F2 PLANAR CELL
Bias voltages
Planar
Write 0
Hold
Read 0 (P read)
Hold
Write 1
Hold
Read 1 (P read)
Hold
SL
BL
WWL
I c e l l (A)
R M T J (k)
TMR
MTJ State
I c e l l (A)
R M T J (k)
TMR
MTJ state
0
0
0
0
27.1
1.13
AP (1)
0
27.1
1.13
AP (1)
0
1.6 V
1.6 V
55.1
12.73
P (0)
61.3
12.73
P (0)
0
0
0
0
12.73
P (0)
0
12.73
P (0)
0
0.4 V
1.6 V
25.2
12.73
P(0)
26.5
12.73
P(0)
0
0
0
0
27.1
P(0)
0
12.73
P(0)
1.6V
0
1.6V
92
14.46
0.14
AP (1)
95.8
14.41
0.13
AP (1)
0
0
0
0
27.1
1.13
AP (1)
0
27.1
1.13
AP (1)
0
0.4 V
1.6 V
15.7
22.4
0.76
AP (1)
16.3
22.20
0.74
AP (1)
0
0
0
0
27.1
1.13
AP (1)
0
27.1
1.13
AP (1)
423
the proposed 4F 2 and conventional 10F 2 cell. The corresponding results are tabulated in Table V. Undoubtedly, the proposed
cell demonstrates a better performance in terms of power dissipation and write margins. The leakage power dissipation for
the proposed cell is four to five orders of magnitude lower than
conventional cell (when the cell is not selected for writing). The
dynamic power dissipation at WWL = 1.6 V (VDD ) is higher;
although, here the point of consideration is that the dynamic
power always has a tradeoff with the P and AP write margins
(write currents). There is a larger tradeoff window between the
write margin and dynamic power dissipation in the case of proposed cell with GAA device, as it can operate with VDD as low
as 1.3 V (at tPW = 1 ns) by sacrificing the write margins. It is
clearly observed in Fig. 7(b) that the planar cell cannot operate
correctly at VDD = 1.3 V for tPW = 1 ns. Moreover, a comparison of the dynamic power dissipation between the cells in
terms of VDD in Fig. 7(c) shows that the proposed cell can be
optimized for low-power operation also.
424
VII. CONCLUSION
425
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
426
427
428
IE
E
Pr E
oo
f
GAA
Parameter
APPENDIX I
BSIM CG MODEL CALIBRATION
The important parameters and model calibration methodology for configuring BSIM CG for cylindrical GAA operation
are classified and discussed as follows:
A. Geometry and Material Parameters: The geometry
and material parameters of BSIM CMG are set for a vertical
GAA device operation, according to the parameters used for
TCAD simulation (see Table VI).
B. Threshold Voltage Calibration: The threshold voltage
of BSIM CG is expressed as
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
where Vth0 is the threshold voltage of the model based on surface potential calculations. Vth,SCE is the threshold voltage
degradation due to short channel effects, Vth,DIBL is used
to model the effect of drain voltage on threshold voltage, and
DVTSHIFT is to handle any additional shift in Vth [21].
The model equation for Vth,SCE in BSIM CG [21] is
Vth,SCE =
0.5DVT0
(Vbi st )
cosh {DVT1 (Le /)} 1
(A2)
458
459
460
461
462
463
TABLE V
COMPARISON BETWEEN PROPOSED CELL AND 10F2 PLANAR NMOS
Select Device
GAA
Planar
W M P (A)
W M A P (A)
t p (ns)
t A P (ns)
Power (pW) W W L = 0 V
23.3
17.6
34.0
29.0
0.43
0.52
0.25
0.32
16.2
1140
112.4
107.0
TABLE VII
THRESHOLD VOLTAGE PARAMETERS
TABLE VI
GEOMETRY AND MATERIAL PARAMETERS
GEOMOD
BULKMOD
ASYMMMOD
COREMOD
L
D
EOT
NGATE
NSD
PHIG
NBODY
Value
Description[21]
3
0
1
0
120 nm
40 nm
2 nm
0
1 1020 cm3
4.61 eV
1 1016 cm3
Cylindrical GAA
SOI substrate
Asymmetric device
Surface potential model
Gate length
Gate diameter
Gate oxide thickness
Metal gate
S/D doping concentration
Gate work function
Channel doping
VALUE
Description [21]
DVT0
DVT1
ETA0
DSUB
DVTP0
DVTP1
DVTSHIFT
2.2
0.177
1
0.80
0
0
0
SCE coefficient
SCE exponent coefficient
DIBL coefficient
DIBL exponent coefficient
Coefficient for drain-induced V t h shift
Exponent coefficient for drain-induced V t h shift
Additional V t h shift handle
TABLE VIII
MOBILITY AND DRAIN CURRENT PARAMETERS
IE
E
Pr E
oo
f
464
465
466
467
468
where
Vbi = (kB T /q) ln NSD (nb o dy /n2i ) ,
and
=
Vth,DIBL =
0.5ETA0
VDS
cosh (DSUB (Le /)) 1
U0
IDS0MULT
U0MULT
Description [21]
TABLE IX
PARASITIC CAPACITANCE, SUBTHRESHOLD, AND LEAKAGE CURRENT
PARAMETERS
(A3)
Value
CGEOMOD
CGEO1SW
COVS
COVD
DVT1SS
GIDLMOD
AGISL
AGIDL
Value
Description [21]
1
1
0.1 fF
0.087 fF
1.0234
1
1.05 1015
1.05 1015
REFERENCES
[1] ERD. (2011). International technology road map for semiconductors [Online]. Available: http://www.itrs.net
[2] T. Schloesser et al., 6F2 buried wordline DRAM cell for 40 nm and
beyond, in Proc. IEEE IEDM08, San Francisco, CA, USA, pp. 14,
Dec. 2008.
[3] H. Chung et al., Novel 4F2 DRAM cell with vertical pillar transistor(VPT), in Proc. IEEE ESSDRC, Helsinki, Finland, pp. 211214, Sep.
2011.
[4] Z. Fang et al., Fully CMOS-compatible 1T1R integration of vertical
nanopillar GAA transistor and oxide-based RRAM cell for high-density
nonvolatile memory application, IEEE Trans. Electron Devices, vol. 60,
no. 3, pp. 11081113, Mar. 2013.
[5] D.-L. Kwong et al., Vertical silicon nanowire platform for low power
electronics and clean energy applications, J. Nanotechnol., vol. 2012,
pp. 121, 2012.
[6] T. Kawahara, K. Ito, R. Takemura, and H. Ohno, Spin-transfer torque
RAM technology: Review and prospect, Microelectron. Reliab., vol. 52,
no. 4, pp. 613627, Apr. 2012.
490
491
492
493
494
495 Q5
496
497
498
499
500
501
502
503
504
505
506
507
508
VERMA et al.: NOVEL 4F2 BURIED-SOURCE-LINE STT MRAM CELL WITH VERTICAL GAA TRANSISTOR AS SELECT DEVICE
560
561
562
563
564
565
566
567
568
569
570
571
572
573
IE
E
Pr E
oo
f
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
Q6 541
542
543
544
545
546
547
548
549
550
551
552
553
Q7 554
555
556
557
558
Q8 559
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
QUERIES
597
598
599
600
601
602
603
604
Q1.
Q2.
Q3.
Q4.
Q5.
Q6.
Q7.
Q8.
Author: Please check the acronym of word line. Is this OK as is or should be WL.
Author: Please check the edited sentence PMA MTJs offer . . . switching. for intended meaning.
Author: Please provide the full form of the acronyms BSIM CG and TMR.
Author: The variable j has been set as R MTJ. Please check.
Author: Please provide all the names of authors in Refs. [2][5], [7], [8], [10], [12][14], [16], [21], and [24].
Author: Please verify Ref. [18] as set.
Author: Please provide the city of the publisher in Ref. [22].
Author: Please provide the month information in Ref. [24].
IE
E
Pr E
oo
f
596