⇦Chinese Physics Letters, 2017, Vol. 34, No. 11, Article code 118501 Influence of Total Ionizing Dose Irradiation on Low-Frequency Noise Responses in Partially Depleted SOI nMOSFETs * Chao Peng(彭超)1,2**, Yun-Fei En(恩云飞)1, Zhi-Feng Lei(雷志锋)1, Yi-Qiang Chen(陈义强)1**, Yuan Liu(刘远)1, Bin Li(李斌)2 Affiliations 1Science and Technology on Reliability Physics and Application of Electronic Component Laboratory, China Electronic Product Reliability and Environmental Testing Research Institute, Guangzhou 510610 2School of Electronic and Information Engineering, South China University of Technology, Guangzhou 510641 Received 19 July 2017 *Supported by the National Postdoctoral Program for Innovative Talents under Grant No BX201600037, the Science and Technology Research Project of Guangdong Province under Grant Nos 2015B090901048 and 2015B090912002, and the Distinguished Young Scientist Program of Guangdong Province under Grant No 2015A030306002.
**Corresponding author. Email: 576167714@qq.com; yiqiang-chen@hotmail.com Citation Text: Peng C, En Y F, Lei Z F, Chen Y Q and Liu Y et al 2017 Chin. Phys. Lett. 34 118501 Abstract Total ionizing dose effect induced low frequency degradations in 130 nm partially depleted silicon-on-insulator (SOI) technology are studied by $^{60}$Co $\gamma$-ray irradiation. The experimental results show that the flicker noise at the front gate is not affected by the radiation since the radiation induced trapped charge in the thin gate oxide can be ignored. However, both the Lorenz spectrum noise, which is related to the linear kink effect (LKE) at the front gate, and the flicker noise at the back gate are sensitive to radiation. The radiation induced trapped charge in shallow trench isolation and the buried oxide can deplete the nearby body region and can activate the traps which reside in the depletion region. These traps act as a GR center and accelerate the consumption of the accumulated holes in the floating body. It results in the attenuation of the LKE and the increase of the Lorenz spectrum noise. Simultaneously, the radiation induced trapped charge in the buried oxide can directly lead to an enhanced flicker noise at the back gate. The trapped charge density in the buried oxide is extracted to increase from $2.21\times10^{18}$ eV$^{-1}$cm$^{-3}$ to $3.59\times10^{18}$ eV$^{-1}$cm$^{-3}$ after irradiation. DOI:10.1088/0256-307X/34/11/118501 PACS:85.30.-z, 61.80.-x, 07.87.+v © 2017 Chinese Physics Society Article Text The high insulating properties of silicon-on-insulator (SOI) substrates, in particular with the use of high resistivity material, lead to high performance mixed-signal circuits.[1,2] The low-frequency noise, which can directly impact analog or radio frequency (RF) integrated circuits, is an importance parameter for these kinds of applications.[3] The most commonly low-frequency noise is flicker noise which has a $1/f$-like spectral intensity. There is a long lasting debate about the mechanism of the flicker noise.[4-6] The $\Delta N$ model contributes the origin of flicker noise to the fluctuation in the carrier number, which is caused by the carrier tunneling between the conduction channel and the traps located in the oxide close to the interface. However, the $\Delta \mu$ model implies that flicker noise is generated by the fluctuation in carrier mobility, which has resulted from the scatter mechanism related to the lattice quality or process defect. The aggressive shrinking of gate oxide thickness leads to a sharp increase of the gate tunneling current, which has been reported to be responsible for the linear kink effects (LKEs) in advanced SOI devices.[7] Low-frequency excess noise associated to LKEs is reported, which is manifested as a Lorentzian spectrum ($1/f^{2}$-like, but not $1/f$-like).[8-10] Meanwhile, the SOI technology is widely used in the radiation environment because of its major advantages over bulk substrates (including their latch-up immunity and less sensitive volume of charge collection) thanks to the complete dielectric isolation of transistors.[11-13] Radiation will introduce trapped-charge and interface traps in the oxide, which can alter the low-frequency responses of the MOS device.[14-16] However, the radiation induced low-frequency degradation in SOI MOSFETs is not thoroughly studied. In this study, we focus on the influence of total ionizing dose (TID) irradiation on low-frequency responses of partially depleted (PD) SOI MOS transistors. Both the Lorenz spectrum noise, which is related to the linear kink effect, and the $1/f$-like low-frequency noise at front and back gates are investigated. All the devices used in our experiment are 130 nm PDSOI NMOS with $\sim$2 nm gate oxide, 100 nm top silicon film and 145 nm buried oxide. A T-shape gate is introduced as the body contact and the structure diagram of the MOSFET is shown in Fig. 1. The shallow trench isolation (STI) connected with the buried oxide at the bottom is introduced for isolation. Different transistor sizes were chosen, $W/L=5$ μm/0.13 μm, 10 μm/10 μm. The threshold voltage $V_{\rm th}$ of the device is approximately 0.3 V. All the samples are DIP ceramic packaged. The operating supply voltage is 1.2 V. The irradiation experiments were conducted in Xinjiang Technical Institute of Physics and Chemistry, Chinese Academy of Sciences, and $^{60}$Co $\gamma$-ray was used as the radiation source. The dose rate was 100 rad(Si)/s. During radiation exposure, the devices were under ON state with gate biased at 1.2 V and the other terminal grounded. The temperature was monitored and kept at room temperature. Low-frequency noise measurements were obtained prior to irradiation and after step irradiations up to 500 and 1000 krad(Si). The noise test system consists of an Agilent 35670 A dynamic signal analyzer, an HP4156 semiconductor parameter analyzer, and an impedance matching network. The entire tests were taken within half an hour following exposure.
cpl-34-11-118501-fig1.png
Fig. 1. Structure diagram and cross section of the PDSOI NMOS device with T-shape gate body contact.
Figure 2 shows the normalized drain current noise power spectral density $S_{\rm Id}/I_{\rm d}^{2}$ and input referred noise power spectral density $S_{\rm Vg}$ as a function of the front gate voltage before irradiation for a 5 μm/0.13 μm device. At $V_{\rm fg} < 1$ V, $S_{\rm Vg}$ is a constant and the line ($S_{\rm Id}/I_{\rm d}^{2}$ versus $V_{\rm fg}$) is parallel with the line ($g_{\rm m}^{2}/I_{\rm d}^{2}$ versus $V_{\rm fg}$). Based on the $\Delta N$ model, the input referred noise power spectral density of MOSFET in the linear region can be calculated by[4] $$\begin{align} S_{\rm Vg} =\frac{q^2kTN_{\rm t} }{\alpha _{\rm t} WLC_{\rm ox}^2 f}.~~ \tag {1} \end{align} $$ The normalized drain current noise power spectral density can be expressed as $$\begin{align} \frac{S_{\rm Id} }{I_{\rm d}^2 }=\Big(\frac{g_{\rm m}}{I_{\rm d} }\Big)^2S_{\rm Vg},~~ \tag {2} \end{align} $$ where $N_{\rm t}$ is the trap density near the interface in the gate oxide (eV$^{-1}$cm$^{-3}$), $\alpha _{\rm t}$ is the tunneling constant (approximately equal to 10$^{8}$ cm$^{-1}$), $g_{\rm m}$ is the transconductance, $C_{\rm ox}$ is the capacitance per unit area of the front gate, $I_{\rm d}$ is the drain current, $T$ is the temperature, while $W$ and $L$ are the width and length of the MOSFET, respectively. The $\Delta N$ model predicts that $S_{\rm Vg}$ is independent of gate voltage. The experimental data at $V_{\rm fg} < 1$ V are consistent with the $\Delta N$ model. According to Eq. (1), $N_{\rm t}$ is extracted to be $3.7\times10^{17}$ eV$^{-1}$cm$^{-3}$ before irradiation for the 5 μm/0.13 μm device, while at $V_{\rm fg}>1$ V, the test results are consistent with the $\Delta \mu$ model, which is manifested as $S_{\rm Vg}$ increases with the gate voltage. Based on the $\Delta \mu$ mode, the input referred noise power spectral density of the MOSFET in the linear region can be calculated by[4] $$\begin{align} S_{\rm Vg} =\frac{q\alpha _{\rm H}}{C_{\rm ox} WLf}(V_{\rm fg} -V_{\rm th}),~~ \tag {3} \end{align} $$ where $V_{\rm fg}$ is the gate voltage, and $\alpha_{\rm H}$ is the silicon lattice quality parameter near the front channel, which can be extracted to be $\sim$2.7 $\times$ 10$^{-4}$ before irradiation.
cpl-34-11-118501-fig2.png
Fig. 2. (a) Normalized drain current noise power spectral density and (b) input referred noise power spectral density as a function of front gate voltage before irradiation for a 5 μm/0.13 μm device.
Figure 3(a) shows $S_{\rm Vg}$ as a function of frequency before and after irradiation at the front gate for a 5 μm/0.13 μm NMOS. A $1/f$-like noise power spectrum is observed at $V_{\rm fg}=0.4$ V and 1.2 V. No significant change in the $1/f$ noise is observed after the total dose is up to 1 Mrad(Si). It implies that irradiation does not increase the gate oxide trap density or change the silicon lattice quality. Since the gate oxide is very thin, the radiation induced oxide traps can be ignored. Similarly, no increase of $S_{\rm Vg}$ at $V_{\rm gt}=V_{\rm fg}-V_{\rm th}=0.6$ V is observed for the 10 μm/10 μm device. However, when overdrive voltage $V_{\rm gt}$ is up to 1.1 V, radiation induced low-frequency noise increase is observed for the 10 μm/10 μm device (Fig. 3(b)). Figure 4 shows the drain current noise power spectral density as a function of frequency at different front gate biases before irradiation for a 10 μm/10 μm NMOS. The overdrive voltage $V_{\rm gt}$ increases from 0.9 V to 1.1 V during the test. At $V_{\rm gt}=0.9$ V, the low-frequency noise is manifested as a $1/f$-like spectrum. As the gate voltage continues to increase, the valence-band electron tunneling occurs through the ultrathin gate oxide and leaves excess holes in the floating body. It leads to the linear kink effects which feature as a second peak of the transconductance at $V_{\rm fg}=1.25$ V, as shown in Fig. 4(b).
cpl-34-11-118501-fig3.png
Fig. 3. Input referred noise power spectral density as a function of frequency before and after irradiation at front gate: (a) $W/L=5$ μm/0.13 μm, and (b) $W/L=10$ μm/10 μm.
cpl-34-11-118501-fig4.png
Fig. 4. (a) Drain current noise power spectral density as a function of frequency at different front gate biases before irradiation. (b) Front gate transconductance as a function of gate voltage before and after irradiation. Here $W/L=10$ μm/10 μm.
Simultaneously, the low-frequency noise in the linear kink region changes from $l/f$-like in a Lorentzian noise spectrum, i.e., a flat low-frequency plateau with constant amplitude, followed by an $l/f^{2}$ roll-off, as shown in Fig. 4(a). Simoen et al.[17] suggested that the physical origin of this type of noise relies on the capture/release of excess holes by the generation-recombination (GR) center in the depletion region. The LKE introduces a Lorenz spectrum noise to replace the $1/f$ noise. Thus the observed increase of low-frequency noise in Fig. 3(b) is the Lorenz spectrum noise, instead of $1/f$ noise.
cpl-34-11-118501-fig5.png
Fig. 5. Drain current noise power spectral density as a function of front gate overdrive voltage before and after irradiation for a 10 μm/10 μm device.
Figure 5 shows the drain current noise power spectral density as a function of front gate overdrive voltage before and after irradiation for a 10 μm/10 μm device. The occurrence of the Lorenz spectrum noise leads to a noise overshoot at $V_{\rm gt}=0.95$ V in Fig. 5, which is in accordance with the second peak of the transconductance. The noise overshoot is significantly enhanced after 500 krad(Si) irradiation. Based on the GR noise model, the LKEs related Lorenz spectrum drain current noise power spectral density can be expressed as[18] $$\begin{align} S_{\rm Id} =g_{\rm m}^2 \frac{q^2N_{\rm t} w_{\rm d} }{C_{\rm ox}^2 WL}\frac{4}{(\tau _{\rm c} +\tau _{\rm e} )\Big[\Big(\frac{1}{\tau _{\rm c} }+\frac{1}{\tau _{\rm e} }\Big)^2+(2\pi f)^2\Big]},~~ \tag {4} \end{align} $$ where $w_{\rm d}$ is the width of the depletion region in the body, $N_{\rm t}$ is the GR center density in the depletion region, while $\tau_{\rm c}$ and $\tau_{\rm e}$ are the capture and emission time constants, respectively. According to Eq. (4), $S_{\rm Id}$ is proportional to the effective area density of the GR center in the depletion region, i.e., $w_{\rm d}\times N_{\rm t}$. The radiation induced positive trapped charge in the STI and buried oxide will deplete the nearby body and will introduce an extra depletion region. It is equivalent to an increase of the effective area density of GR center. The radiation activates extra hole recombination centers and accelerates the consumption of the accumulated holes in the floating body. Thus an increase of the excess Lorentzian noise is observed (Fig. 5). Simultaneously, the consumption of extra holes results in the attenuation of the LKE, which is manifested as a weakened second peak of transconductance after irradiation (Fig. 4(b)).
cpl-34-11-118501-fig6.png
Fig. 6. Input referred noise power spectral density as a function of frequency at the back gate (a) before and (b) after irradiation for a 5 μm/0.13 μm device.
Figure 6 shows $S_{\rm Vg}$ as a function of frequency before and after irradiation at the back gate for a 5 μm/0.13 μm NMOS. Since the material quality near the back channel is worse than that in the front channel, the low-frequency noise level at the back gate is much higher than that at the front gate, as shown in Fig. 6(a). The same $S_{\rm Vg}$ at different back gate biases ($V_{\rm gt}=-5$ V, 0 V and 5 V) indicates that the $1/f$ noise at the back gate follows the $\Delta N$ model. According to Eq. (1), the trap density in the buried oxide near the back channel is calculated to be $2.21\times10^{18}$ eV$^{-1}$cm$^{-3}$ at $V_{\rm gt}=0$ V and $f=1$ Hz. Radiation induced trapped charge in the buried oxide leads to the increase of $S_{\rm Vg}$ at the back gate. The trap density in the buried oxide can be extracted to be $3.59\times10^{18}$ eV$^{-1}$cm$^{-3}$ after 500 krad(Si) irradiation. In summary, the influences of the TID effect on low-frequency noise in a 130 nm PDSOI MOSFET have been investigated in detail. According to the pre-irradiation test, the flicker noises at the front gate obey the $\Delta N$ model at $V_{\rm fg} < 1$ V and obey the $\Delta \mu$ model at $V_{\rm fg}>1$ V. The radiation does not increase the trap charge density or change the silicon lattice quality due to the ultra-thin gate oxide. Thus no flicker noise degradation is observed at the front gate for a 5 μm/0.13 μm device. However, the buried oxide is relatively thick, and the radiation induced trapped charge leads to a significant increase of flicker noise at the back gate. It is worth noting that the low-frequency noise increase at the front gate is observed for a 10 μm/10 μm device. The observed increase of low-frequency noise is proved to be the Lorentzian spectrum noise related to the LKEs, but not the $1/f$ noise. According to the GR noise model, the trapped charge in STI or buried oxide leads to the activation of an extra GR center and finally results in the enhancement of the Lorentzian noise. Furthermore, our research also proves that the low-frequency noise is a strong technology sensitive parameter, which in some cases can be used as a predictive or diagnostic tool for device radiation reliability. References Radio-Frequency Characteristics of Partial Dielectric Removal HR-SOI and TR-SOI SubstratesComprehensive study on low-frequency noise characteristics in surface channel SOI CMOSFETs and device design optimization for RF ICsOn the flicker noise in submicron silicon MOSFETsLow-frequency noise in advanced CMOS∕SOI devices1/f noise in MOS devices, mobility or number fluctuations?"Linear kink effect" induced by electron valence band tunneling in ultrathin gate oxide bulk and SOI MOSFETsElectron valence-band tunneling-induced Lorentzian noise in deep submicron silicon-on-insulator metal–oxide–semiconductor field-effect transistorsGate-induced floating body effect excess noise in partially depleted SOI MOSFETsLow-frequency noise overshoot in ultrathin gate oxide silicon-on-insulator metal–oxide–semiconductor field-effect transistorsEnhanced Radiation Sensitivity in Short-Channel Partially Depleted Silicon-on-Insulator n-Type Metal-Oxide-Semiconductor Field Effect TransistorsRadiation Effects in Advanced Multiple Gate and Silicon-on-Insulator TransistorsRadiation effects in SOI technologiesThe low-frequency noise behaviour of silicon-on-insulator technologiesUnified model of hole trapping, 1/f noise, and thermally stimulated current in MOS devicesLow frequency noise and radiation response in the partially depleted SOI MOSFETs with ion implanted buried oxideHigh-energy proton irradiation induced changes in the linear-kink noise overshoot of 0.10 μm partially depleted silicon-on-insulator metal-oxide-semiconductor field-effect transistorsThe kink-related excess low-frequency noise in silicon-on-insulator MOST's
[1]AubertonHerve A J 1996 SOI: Materials to Systems (New York: IEEE)
[2] Cheng S et al 2017 Chin. Phys. Lett. 34 068101
[3] Tseng Y C et al 2001 IEEE Trans. Electron Devices 48 1428
[4] Simoen E and Claeys C 1999 Solid-State Electron. 43 865
[5] Jomaah J and Balestra F 2004 IEEE Proc.-Circuits Devices Syst. 151 111
[6] Vandamme L K J, Li X S and Rigaud D 1994 IEEE Trans. Electron Devices 41 1936
[7] Mercha A, Rafi J M, Augendre E and Claeys C 2003 IEEE Trans. Electron Devices 50 1675
[8] Lukyanchikova N B et al 2003 J. Appl. Phys. 94 4461
[9] Dieudonne F, Jomaah J and Balestra F 2002 IEEE Electron Device Lett. 23 737
[10] Mercha A, Simoen E, Van Meer H and Claeys C 2003 Appl. Phys. Lett. 82 1790
[11] Peng C et al 2013 Chin. Phys. Lett. 30 098502
[12] Simoen E et al 2013 IEEE Trans. Nucl. Sci. 60 1970
[13] Schwank J R et al 2003 IEEE Trans. Nucl. Sci. 50 522
[14] Simoen E and Claeys C 1996 Solid-State Electron. 39 949
[15] Fleetwood D M et al 2002 IEEE Trans. Nucl. Sci. 49 2674
[16] Liu Y et al 2015 Chin. Phys. B 24 088503
[17] Simoen E et al 2004 J. Appl. Phys. 95 4084
[18] Simoen E, Magnusson U, Rotondaro A L P and Claeys C 1994 IEEE Trans. Electron Devices 41 330