The input noise spectrum of a body-to-source-connected PD device is shown in Fig. 3-6, where the channel length is 0.2 mm, and VDS=0.1 V. We observed that the noise spectrum is dominated by 1/f noise. This suggests that the source-to-body connection can suppress the noise overshoot efficiently. This is because the body contact provides a low-resistance discharge path for body charges, where the source-body impedance reduces drastically by the parallel combination of RB and Zbody. The RB is the body path resistance and Zbody
(=rSB/(1+j(2πf⋅rSBCBB))) is the source-body impedance [19]. Once the source-body impedance reduces by body contact, the shot noise amplified gain would be reduced. Therefore, only 1/f noise would be observed in source-to-body-connected devices.
In MOSFETs, the 1/f noise magnitude is usually related to the interface trap density between the silicon film and the gate oxide. A model, proposed by Gross [20], demonstrated that the noise in MOSFET is due to the fluctuation of the carrier number in the channel and that these fluctuations induce local correlated fluctuations in the channel mobility. When the device operates in linear region and a uniform distribution of oxide/silicon interface traps in activation energy is assumed, the input-referred voltage noise can be written as [20]:
2 2 mobility, and scattering parameter, respectively. NT (Ef) is the trap number near Fermi energy Ef, and the traps lie in the oxide distributed over some region x0 near the interface. Nis the channel carrier number and is proportional to the gate drive voltage. From the equation (3-4), it is clear that the 1/f noise can stem from processing-induced variations in either the near-interface trap density NT, or in the scattering parameter S, for fixed operation conditions
and device area. If the induced mobility fluctuation term µ0SN <<1, the induced mobility fluctuation can be neglected, and the noise power spectrum is independence of the gate drive voltage. On the contrary, if the term µ0SN >>1, the noise magnitude is proportional to the square of gate drive voltage as shown in the Fig. 3-7. Hence, the 1/f noise component is dominated by induced mobility fluctuation. Fig. 3-8 shows the input noise spectrum versus channel length for a device operating at VDS=0.1 V. As shown in Fig. 3-8, the input noise spectrum is proportional to 1/L, which is consistent with the equation (3-4).
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
f
01/f 1/f
2excess noise S VG (V2 /Hz)
Frequency (Hz) VDS=0.1V
VDS=0.4V VDS=0.7V VDS=1.0V VDS=1.2V
VG-VT=0.4V
Fig. 3-1 Input referred noise spectra of a PD floating body SOI MOSFET with different drain biases.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
VDS=1.0V VG-VT=0.5V
S VG (V2 /Hz)
Frequency (Hz) L=0.12
L=0.2 L=0.3 L=0.45
1/f 2
Fig. 3-2 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 1.0V with different channel lengths.
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 1E-13
1E-12 1E-11 1E-10
101 102 103 104 105
VDS=1.0V VG-VT=0.5V
Coner Frequency (Hz)
S VG (V2 /Hz)
Channel Length L (µm) plateau
corner frequency
Fig. 3-3 Plateau and corner frequency of the input-referred noise for a PD floating-body SOI MOSFET at VG-VT = 0.5V and VDS = 1.0V with different channel lengths.
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0
5 10 15 20 25
VG-V
T=0.5V
Leakage Current I L (pA)
Channel Length L (um)
VDS=1.0V VDS=0.8V
Fig. 3-4 Drain-body junction leakage current as functions of channel length for a PD floating-body SOI MOSFET with different drain voltages.
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0
2 4 6 8 10 12
VDS=1.0V VG-VT=0.5V
Body Capacitance C bb (fF)
Channel Length L (um)
Fig. 3-5 Body capacitance as a function of channel length for a PD floating-body SOI MOSFET at VG-VT = 0.5 V and VDS = 1.0V.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
source-to-body-connected L=0.2µm VDS=0.1V
S VG (V2 /Hz)
Frequency (Hz)
Vg-Vt=0.1V Vg-Vt=0.3V Vg-Vt=0.5V Vg-Vt=0.7V Vg-Vt=1.1V
1/f
Fig. 3-6 Input-referred noise spectra of a PD source-to-body-connected SOI MOSFET at VDS = 0.1V and L=0.2mm with different VG-VT.
0.1 1 1E-13
1E-12 1E-11 1E-10 1E-9
S VG (V2 /Hz)
VG-VT (V) source-to-body-connected L=0.2µm f=10Hz
VDS=0.1V
Fig. 3-7 The magnitude of input-referred noise as a function of gate voltage for a PD source-to-body-connected SOI MOSFET at VDS = 0.1V, L=0.2mm and f=10Hz.
0.1 1 10 1E-13
1E-12 1E-11 1E-10 1E-9
source-to-body-connected VDS=0.1V f=10Hz
VG-V
T=0.7
S VG (V2 /Hz)
Channel Length L (µm)
Fig. 3-8 The magnitude of input-referred noise as a function of channel length for a PD source-to-body-connected SOI MOSFET at VDS = 0.1V and f=10Hz.
Chapter 4
Low-Frequency Noise in PD SOI MOSFETs at Various Temperatures
Because the SOI MOSFETs have higher thermal resistance, the channel temperature is higher than bulk devices with the same power dissipation. In addition, there is increasing demand for high-temperature electronics with CMOS technology. The temperature characteristics of SOI devices on analog applications should be studied [21], [22]. In this chapter, we use Temptronic TP03000 Thermal Controller to vary the temperature and investigate the influence on the low-frequency noise of PD SOI MOSFETs operating from linear region to saturation region. The noise characteristics of the devices with floating-body and source-to-body-connected structures have been discussed. The gate length and gate width of test devices are 0.2µm and 10µm, respectively.
The devices operate in linear operation are corresponding to a drain voltage VDS = 0.1V.
As the drain bias increases to about 0.4V, the devices enter the saturation region, both the data obtained before the kink occurring (VDS = 0.8V) and after the kink occurring (VDS = 1.2V) are shown.
4.1 Temperature Effect on Floating-Body PD Device
4.1.1 Operating in Linear Region
The low-frequency noise characteristics of a floating-body PD SOI MOSFET operating in the linear region (drain voltage VDS = 0.1V) is given in Fig. 4-1. At T = 0oC, owing to such a low temperature and low drain bias, both the junction generation current and impact
ionization current have very low values, resulting in a higher plateau and lower f0. Therefore, the noise overshoot only exists in low frequency range (f < 100Hz), and the 1/f noise dominates. As the temperature increases, the intrinsic carrier density (ni ∝ T1.5⋅exp(-1/T )) significantly increases, resulting in a larger junction thermal generation current. This large junction generation current contributes to a lower plateau and higher f0, thus the Lorentzian-like noise overshoot begins moving to higher frequency and exceeds the 1/f noise gradually, as shown in Fig. 4-1. With further increasing temperature to T = 150oC, the noise overshoot shifts to lower plateau and higher corner frequency due to the increase of junction generation current and a Lorentzian-like noise overshoot was observed explicitly.
4.1.2 Operating in Saturation Region and Pre-kink Region
A similar phenomenon is observed in Fig. 4-2 for a floating-body device operating in saturation region before kink effect occurring (The kink voltage is about 0.9V), but an obvious noise overshoot has already appeared at T = 0oC due to the higher junction generation current than that in linear region. Since the junction leakage increases more rapidly at higher temperature, the magnitude of noise overshoot would be decreased drastically with the rise of temperature, as shown in Fig. 4-3.
To further see the effect of the IL, α, and Cbb on the temperature behavior of the noise overshoot, we extract their values from the plateau and corner frequency of noise overshoot in Fig. 4-3 by using equation (3-1). We find that the body factor is nearly independent on temperature and drain bias, and its value is about 0.22. However, the IL and Cbb are strong functions of temperature, as shown in Fig. 4-4. As the temperature increases, intrinsic carrier density increases, resulting in the decreasing of the depletion width, and therefore the Cbb
increases monotonously. The exponential function of IL with temperature indicates the thermal generation current dominates the junction current. From the plot of log(IL) versus 1/T
shown in Fig.4-5, the activation energy Ea for junction leakage can be extracted: Ea = 0.21eV at low temperature (T ≤ 750C), and Ea = 0.42eV at high temperature. The equation of the junction leakage including ideal and nonideal cases can be expressed as:
2
The first term is the diffusion current, while the second term is the generation current due to the generation-recombination processes through the traps in the depletion region. If the junction leakage is dominated by the first term, owing to JR1 ∝ ni2 ∝ exp(-Eg/kT) (where bandgap energy Eg=1.1 eV), the value of Ea is approximately 1.1 eV. If the second term dominates, owing to the equation JR2 ∝ ni ∝ exp(-Eg/2kT), the value of Ea is approximately 0.55 eV. Since the extracted Ea in our devices is smaller than 0.55 eV, it suggests that the thermal generation current is trap dominated in the temperature range of 0-1500C [23].
4.1.3 Operating in Saturation Region and Post-kink Region
As the device operates in the post-kink region (VDS=1.2 V), the large junction leakage current makes the noise overshoot only exists in higher frequency (f > 1 kHz) and is approximately constant up to 75oC, as shown in Fig. 4-6. This is because the impact-ionization current generated near the drain overwhelms the thermal generation current and dominates the leakage current, and the impact-ionization current is less sensitive to temperature as compared to junction generation current [24]. Above 150oC, the drain to body junction generation current drastically increases and dominates leakage current, resulting in a lower plateau than the other temperature conditions. Except for the higher frequency, a pure 1/f noise has been observed over the range of measurement temperature. In addition, we find that the 1/f noise component is temperature-independent.
4.2 Temperature Effect on Body-to-Source-Connected PD Device
We may compare the noise measurements of floating-body SOI transistors with those of source-to-body-connected devices. As shown in Figs. 4-7 and 4-8, the input noise spectrum is dominated with 1/f noise component, and may contain small Lorentzian-like noise for some devices. This suggests that the noise overshoot can be suppressed by body contact. It is noted that the 1/f noise is independent of temperature as the case of floating-body device at high drain bias. From section 3.3 we know that the 1/f noise in SOI devices is dominated by induced mobility fluctuation, and the induced mobility fluctuation term µ0SN has a large effect on the variation of 1/f noise with temperature, yielding a relatively temperature-independent characteristic shown in Fig.4-7 and 4-8. We compare Figs. 4-6 and 4-8, the magnitudes of 1/f noise in floating-body devices and source-to-body-connected devices have approximate values. It suggests that the two types of devices have similar trap density. Therefore, the body contact structure wouldn’t affect the gate oxide quality [25].
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=10/0.2 µm S VG (V2 /Hz)
Frequency (Hz) T=0OC
T=25OC T=75OC T=150OC
VDS=0.1V VG-V
T=0.4V
1/f 1/f2
Fig. 4-1 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 0.1V with different temperatures.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=10/0.2 µm S VG (V2 /Hz)
Frequency (Hz) T=0OC
T=25OC T=75OC T=150OC
VDS=0.8V VG-VT=0.4V
1/f2
Fig. 4-2 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 0.8V with different temperatures.
-20 0 20 40 60 80 100 120 140 160
Fig. 4-3 Plateau and corner frequency of the input-referred noise for a PD floating-body SOI MOSFET at VG-VT = 0.4V and VDS = 0.8V with different temperatures. Figure also shows the magnitude of the input-referred noise at f = 50Hz for a PD floating-body SOI MOSFET at VG-VT = 0.4V and VDS = 1.2V.
-20 0 20 40 60 80 100 120 140 160
Body Capacitance C bb (fF) Leakage Current I L (pA)
Temperature (oC)
VDS=0.8V VG-VT=0.4V
Fig. 4-4 Drain-body junction leakage and body capacitance as functions of temperature for a PD floating-body SOI MOSFET at VG-VT = 0.4V and VDS = 0.8V.
0.0022 0.0024 0.0026 0.0028 0.0030 0.0032 0.0034 0.0036 0.0038 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10
VDS=0.8V VG-VT=0.4V
Ea=0.42eV
Leakage Current (A)
1/T (K-1)
Ea=0.21eV
Fig. 4-5 Leakage current versus 1/T for a PD floating-body SOI MOSFET at VG-VT = 0.4V and VDS = 0.8V.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=10/0.2 µm
S VG (V2 /Hz)
Frequency (Hz) T=0OC
T=25OC T=75OC T=150OC
1/f
V
DS=1.2V V
G-V
T
=0.4V
Fig. 4-6 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 1.2V with different temperatures.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=10/0.2 µm
S VG (V2 /Hz)
Frequency (Hz) T=25OC
T=75OC T=150OC
source-to-body-connected VDS=0.1V
VG-V
T=0.4V
1/f
Fig. 4-7 Input-referred noise spectra of a PD source-to-body-connected SOI MOSFET at VDS= 0.1V with different temperatures.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=10/0.2 µm source-to-body-connected
VDS=1.2V VG-VT=0.4V
S VG (V2 /Hz)
Frequency (Hz) T=25oC
T=75oC T=150oC
Fig. 4-8 Input-referred noise spectra of a PD source-to-body-connected SOI MOSFET at VDS= 1.2V with different temperatures.
Chapter 5
Degradation of Low-Frequency Noise in PD SOI MOSFETs after Hot-Carrier Stress
With the reduction of device size to deep submicrometer dimensions, the problem of the quality of the gate oxide and its interface with the silicon substrate due to hot-carrier effect has been of increasing concern in the fabrication of SOI CMOS devices. Because the 1/f noise is sensitive to oxide charge and interface traps, it is good to monitor the device reliability by measuring the 1/f noise. In this chapter, we investigate the influence of hot-carrier stress on the low-frequency noise of PD SOI MOSFETs including 1/f noise and shot noise induced kink-related excess noise. The noise characteristics of the devices with floating-body and body-contact structures have been discussed.
5.1 Experiments
The n-channel MOSFETs were fabricated on SIMOX substrates with 205nm thick Si active layers, and 400nm thick buried oxides. Briefly, 450nm LOCOS field oxide was used for device isolation. After VT-adjust implant and anti-punch through implant, a 4nm gate oxide and in-situ n+-doped polysilicon gate were formed. The floating-body and body-contact devices with nitride spacer and As+-implanted source/drain junctions are partially depleted.
After NiSi salicidation, the devices were metalized using a typical backend flow.
The current-voltage characteristics of devices were measured using HP4156A semiconductor parameter analyzer. The low-frequency noise measurements in normal and
reverse (i.e., source and drain interchanged) modes were performed using a BTA9812B noise analyzer in conjunction with an HP35670A dynamic signal analyzer. The hot-carrier stress was applied at a drain voltage of VDS = 4V and a gate voltage of VG = 2V with a stressing time ranging from 0 to 1000 sec.
5.2 Hot-Carrier Effect on Floating-Body PD Device
Fig. 5-1 shows the low-frequency noise characteristics of a floating-body PD SOI MOSFET operating in the linear region (VDS = 0.1V) before and after hot-carrier stress for 1000 seconds. The gate length and gate width of test devices are 0.4µm and 20µm, respectively. Owing to the low drain bias, the drain-body junction leakage is low, and the kink related excess noise was overwhelmed by the 1/f noise. After stress, there are two types of damages would affect the 1/f noise: interface states and oxide-trapped charge. In a stressed device, the energy transferred from the hot carriers to the Si/SiO2 interface would break the bond at the interface and thus leads to the creation of the interface state. In n-MOS devices, only acceptor states which become negatively charged would cause an influence on it. When a charge is injected into the oxide after hot carrier stress, it can create a trap; and they are charged positively or negatively depending on this trap [26]. After stressing the device with 1000 seconds, the interface trap and oxide charge stem from hot carrier effect would be increased. From the model in chapter 2, the 1/f noise is proportional to the trap density Nt near the Fermi-level.
(5-1)
VG t
S ∝N
Therefore, the 1/f noise would be increased after hot-carrier stress, and the similar phenomenon can also be observed in reverse mode. It is noted that, the magnitude of 1/f noise in the normal mode and the reverse mode have approximate values. However, in saturation condition (VDS = 2V), the magnitude of 1/f noise in the reverse mode is higher than that in the
normal mode, as shown in Fig. 5-2. Fig. 5-3 shows the hot-carrier degradation modes of a MOSFET operated in the linear and saturation regimes. In linear region, the effects of the trap are the same in both modes, nevertheless, in saturation region, the effect of the trap decreases in normal mode as the pinch-off region increases. For reverse mode operation, hot-carrier-induced interface traps are located on the source side, so the effect of the trap remains unchanged in saturation region [27], [28]. Therefore, the magnitude of the 1/f noise increased by hot-carrier stress in reverse mode is higher than that in normal mode.
Fig. 5-4 shows the low-frequency noise characteristics of a device operating in the saturation region and before kink effect occurring (VDS =1.2V). A Lorentzian-like noise overshoot was observed explicitly before and after stress in normal mode. The noise overshoot comes from the interactionbetween the shot noise of the junction leakage and the source-body impedance, as illustrated in chapter 3. After stressing the device, the increased interface-trap density near the drain junction causes the increase of gate-induced-drain-leakage (GIDL) current in the low field region as shown in Fig. 5-5, and provides an additional current path due to the interface-trap assisted band-to-band tunneling mechanism; therefore, the trap-assisted drain-to-substrate leakage increases [29]. As discussed in chapter 3, when the drain-to-substrate leakage increases, the plateau of the kink-related excess noise will decrease, and the corner frequency will increase as shown in Fig. 5-4. In reverse mode, because the interface traps are located on the source side, the GIDL current would not be changed as shown in Fig. 5-6. Hence, the drain-to-substrate leakage would not be affected by stress and remain constant. However, the excess holes in body region due to the floating-body effect can effectively recombine through the interface traps near source with excess electrons due to the parasitic bipolar action. As a result, the floating-body effect is suppressed, thus the kink-related excess noise is suppressed. From Fig. 5-4, we only observe the 1/f noise after stress with reverse mode.
5.3 Hot-Carrier Effect on Body-Contact PD Device
To reduce floating body effect, the body contact structure is widely used in SOI CMOS technology. Hence we are also interested to know the hot-carrier effect on the performance of body-contact devices. Figs. 5-7 and 5-8 show the threshold voltage and drain current degradation respectively as a function of stress time in floating-body and body-contact devices. The gate length and gate width of test devices are 0.4µm and 20µm, respectively. It is noted that the degradation in body-contact device was larger than that in floating-body device.
As the body voltage changes from high (floating-body device) to low (body-contact device), the vertical channel field increases, which means the channel electrons are pushed further toward the Si surface [30]. These hot channel electrons which closer to the Si surface would lead to greater Si surface damage. Therefore, the degradation which stems from hot-carrier stress was more serious in body-contact device. Figs. 5-9 and 5-10 show the low-frequency noise characteristics of a body-contact PD device operating in the linear and saturation region respectively. Because the body contact provides a low resistance leakage path, the shot noise amplified gain decreases, and the input noise spectrum is dominated with 1/f noise component.
As the same with the floating–body devices, the 1/f noise would increase after hot-carrier stress due to the generation of the interface states and oxide charge.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
VD=0.1V VG=1.56V
S VG (V2 /Hz)
Frequency (Hz) t=0sec
N t=1000sec R t=1000sec W/L=20µm/0.4µm
1/f
Fig. 5-1 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 0.1V in normal and reverse modes before and after stress.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=20µm/0.4µm V
D=2V V
G=1.56V
S VG (V2 /Hz)
Frequency (Hz) t=0sec
N t=1000sec R t=1000sec
1/f
Fig. 5-2 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 2V in normal and reverse modes before and after stress.
(a)
VDS
(b) S D
G
S D
G
S D G
VDS
S D
G
VDS
VDS
(c) (d)
Fig. 5-3 Schematic diagrams of a MOSFET operated in (a) the normal mode and linear region; (b) the reverse mode and linear region; (c) the normal mode and saturation region; and (d) reverse mode and saturation region. The position of the interface trap is indicated symbolically by the cross (µ)
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=20µm/0.4µm
VD=1.2V VG=1.56V
S VG (V2 /Hz)
Frequency (Hz) t=0sec
N t=1000sec R t=1000sec
1/f2
Fig. 5-4 Input-referred noise spectra of a PD floating-body SOI MOSFET at VDS = 1.2V in normal and reverse mode before and after stress.
0.0 0.5 1.0 1.5 2.0 1E-12
1E-11 1E-10 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 0.01
Normal
I D (A)
VG (V)
Vd=0.1V t=0min Vd=2V t=0min Vd=0.1V t=50min Vd=2V t=50min W/L=20µm/0.5µm
Fig. 5-5 ID-VG characteristic of a PD floating-body SOI MOSFET in normal mode before and after stress with W=20mm and L=0.5mm.
0.0 0.5 1.0 1.5 2.0 1E-12
1E-11 1E-10 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 0.01
W/L=20µm/0.5µm Reverse
I D (A)
VG (V)
Vd=0.1V t=0min Vd=2V t=0min Vd=0.1V t=50min Vd=2V t=50min
Fig. 5-6 ID-VG characteristic of a PD floating-body SOI MOSFET in reverse mode before and after stress with W=20mm and L=0.5mm.
1 10 100 1000 0.0
0.1 0.2 0.3 0.4 0.5 0.6
VD=0.1V
∆V TH (V)
Time (sec)
Body-Contact Floating-Body W/L=20µm/0.4µm
Fig. 5-7 Threshold voltage degradations as a function of stress time in floating-body and body-contact devices.
1 10 100 1000 0
10 20 30 40 50 60
W/L=20µm/0.4µm
|∆I / I| (%)
Time (sec)
Body-Contact Floating-Body VD=0.1V V
G=2V
Fig. 5-8 Drain current degradations as a function of stress time in floating-body and body-contact devices.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
VD=0.1V VG=1.555V
S VG (V2 /Hz)
Frequency (Hz) t=0sec
t=1000sec W/L=20µm/0.4µm
1/f
Fig. 5-9 Input-referred noise spectra of a PD body-contact SOI MOSFET at VDS = 0.1V before and after stress.
10 100 1000 10000 1E-15
1E-14 1E-13 1E-12 1E-11 1E-10 1E-9
W/L=20µm/0.4µm VD=2V VG=1.555V
S VG (V2 /Hz)
Frequency (Hz) t=0sec
t=1000sec
1/f
Fig. 5-10 Input-referred noise spectra of a PD body-contact SOI MOSFET at VDS = 2V before and after stress.
Chapter 6
Conclusion and Future Work
6.1 Conclusion
The development trends of MOS devices are scaling down the device size and enhance the speed of operation. However, it results in many obstacles for bulk MOSFETs and is
The development trends of MOS devices are scaling down the device size and enhance the speed of operation. However, it results in many obstacles for bulk MOSFETs and is