IN THE PAST four decades, microelectronic-device damage

Transcription

1 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 54, NO. 8, AUGUST SPICE Models of Fluorine-Ion-Irradiated CMOS Devices Henok T. Mebrahtu, Wei Gao, William E. Kieser, Xiaolei L. Zhao, Paul J. Thomas, and Richard I. Hornsey Abstract CMOS image sensors are attractive for space applications due to their low-power and system-on-chip features. The typical active pixel sensor (APS) is composed of a photodiode and several transistors. Using Fluorine +7 ionswithanenergy of 17 MeV, the effects of radiation are investigated on photodiodes and transistors manufactured using a standard 0.35-µm CMOS process. Simulation results show that the range of these ions overlaps with the active region of the device. Thus, the proximity effect of the ions on the performance of the device can be important. The tested photodiode showed a leakage current increase after it was irradiated with fluorine ions. The ideality factor of recombination current is observed to increase up to 4. Moreover, an increase in leakage current and absolute threshold voltage was observed in fluorine-ion-irradiated nmos and pmos transistors. In this paper, behavioral SPICE models are developed to analyze the contribution of these components to an overall increase in dark current of a CMOS APS. Index Terms CMOS images sensors, heavy ion, MOS transistor, photodiode, radiation effects, SPICE model. I. INTRODUCTION IN THE PAST four decades, microelectronic-device damage due to gamma-rays, energetic protons, neutrons, and highly energetic heavy ions have been well documented [1] [4]. It is observed that various characteristics of a photodiode, such as dark current, leakage current, and noise of the device, are affected by proton and neutron irradiation. These radiation sources have been reported to cause threshold-voltage shift, mobility, and transconductance changes in MOS transistors. In addition, energetic heavy ions have also been widely employed to assess the effects of single event and charge-collection mechanism on devices. The energy ranges, used in the studies, are typically greater than 100 MeV. In contrast, to the authors knowledge, little has been documented on effects of heavy ions in the lower energy spectrum (< 20 MeV). Effects of ions in this energy range can be significant because their range may overlap with the active regions of the device. Manuscript received August 4, 2006; revised March 29, This work was supported in part by Topaz Technology Inc. and in part by the Center for Research in Earth and Space Technology. The review of this paper was arranged by Editor T. Skotnicki. H. T. Mebrahtu is with the Department of Physics, Duke University, Durham, NC USA ( htm@phy.duke.edu). W. Gao is with the Computer Science and Engineering Department, York University, Toronto, ON M3J 1P3, Canada. W. E. Kieser and X. L. Zhao are with the IsoTrace Laboratory, University of Toronto, Toronto, ON M5S 3G4, Canada. P. J. Thomas is with Topaz Technology Inc., Toronto, ON M3J 1P3, Canada. R. I. Hornsey is with the School of Engineering, York University, Toronto, ON M3J 1P3, Canada. Digital Object Identifier /TED Photodiodes and transistors are the main components of CMOS active-pixel-sensor (APS) image sensors. CMOS-based imagers are attractive for space application that require low power, onchip integration of analog and digital components, and lower cost. Previous studies conducted on CMOS imagers have focused on the chip level [5] [7]. Thus, it is not trivial to determine and isolate the effects of contributions from individual chip components and thereby identify the critical source of failure in the chip and design-appropriate mitigation techniques. However, this paper addresses the effect of radiation at a component level. This paper is organized as follows. Section II provides some background material related to this paper. This is followed by results of ion range simulations on the devices under test. Experimental setup is described in Section V, followed by experimental results. The SPICE modeling works and conclusion are presented in Sections VI and VII, respectively. II. BACKGROUND A. Junction Photodiode The metallurgical p-n junction diode is a fundamental semiconductor junction that exists on a number of semiconductor devices, such as MOSFETs, bipolar transistors, photodiodes, and various other types of diodes. The dark current of photodiode detectors is known to increase with proton and neutron irradiation. This is associated with an increase in generation recombination centers in the bulk or at the surface of the devices. These centers also increase generation recombination probabilities, hence, decreasing the responsivity of the photodiode detector. The current flowing through the terminal of the p-n junction is a net effect of diffusion and generation recombination processes within the device. The diffusion component of the current at low- and high-carrier-injection properties of a long base p-n junction diode are well described by [8] ( ) J n,p = q D nn a,d 1 + 4n2 i (eqv/kt 1) 2L n Na,d 2 1 (1) where N a and N d are doping concentration of acceptors and donors, respectively. On the other hand, the generation recombination component of the current is estimated from a Shockley Read Hall model and is commonly expressed as J GR = J GR0 (e qv/2kt 1) (2) where J GR0 is slightly voltage-dependent /$ IEEE

2 1964 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 54, NO. 8, AUGUST 2007 TABLE I TYPICAL LAYER DIMENSIONS OF A 0.35-µm CMOS PROCESS (IN MICROMETERS) TABLE II ION RANGES IN A PROFILE OF 0.35-µm CMOSPROCESS In the standard diode model, these contributions are commonly shown as a single expression through the ideality factor n, where n is close to one for a diffusion-dominated diode and close to two for a generation recombination-dominated diode I = I o (e qv/nkt 1). (3) B. MOS Transistors MOS transistors are at the very heart of CMOS technology, so their behavior under irradiation has been the subject of study for the last four decades. MOS transistors were believed to be susceptible primarily to total ionizing dose damage. As a result, studies focused on this aspect of the device behavior, and there is a good understanding of responses of MOS transistors to the total ionizing dose damage [1], [9]. In these references, the electrical effects of charge traps in the oxide and at the SiO 2 /Si are discussed in detail. Moreover, the effects of ionization radiation in subthreshold current is widely reported [10], [11]. At moderate energy levels (< 100 MeV), however, the incident ions, in addition to the widely studied electron- and holerelated effects of ionization radiation, are believed to contribute additional effects in electrical properties of a device. In this paper, a physically based model of the observed moderate energy F-ion damage is developed. Moreover, to support data interpretation, a device-level understanding of ion range and ion-energy dissipation is simulated. III. STOPPING AND RANGE OF IONS IN MATTER (SRIM) SIMULATION The range of ions in the device is simulated using SRIM codes [12]. The layer profile for the simulation is extracted from a standard 0.35-µm CMOS process. In the simulation, silicon dioxide and aluminum are used as materials for dielectric and metal layers, respectively, while the passivation layer is assumed to be composed of silicon dioxide and silicon nitride. The layer properties of a typical 0.35-µm CMOS process are summarized in Table I. The ion range and dose of different ions species are simulated, namely, 12 C, 14 N, 16 O, 19 F, and 35 Cl. For these ions, our experimentally achievable charge state and energy and the corresponding ranges are tabulated in Table II (Si 3 N 4 passivation). Thus, for the purposes of studying the effects of moderateenergy ions, i.e., ions that carry adequate energy to the active region and remain embedded in the device, F ion is chosen as Fig. 1. Range of 17-MeV Fluorine +7 ions in 0.35-µm-process device. Simulation shown for ions at normal incidence. suitable candidate and offering a potential extreme-case-effect scenario due to its high chemical activity. The ion-range profile for MeV F ion is shown in Fig. 1. The figure shows ions that penetrate normal to the device and encounter all the metal layers. In contrast, ions which miss all the metal layers, their range is simulated to be around 8.9 µm. In either case, the simulation results show the ion range to be very close to the active region of the device. The energy-dissipation profile for the incident ion in the device is also shown in Fig. 1. For 17-MeV F +7 ions, most of the energy is deposited in the metal and dielectric layers above the active region with the Bragg peak appearing roughly in metal layer two. When the ion reaches the dielectric and fieldoxide regions at a depth of around 7.80 µm, slightly more than 90 ev/å per ion is dissipated as ionization-energy loss. Hence, one fluorine ion of the stated energy passing these areas delivers about rad(si) to the structure. IV. EXPERIMENT Heavy-ion-radiation data was collected from experiments conducted with F +7 ion radiation at the IsoTrace Laboratory of the University of Toronto [13]. The radiation dose delivered, by an incident ion, varies between the layers of a material. This is estimated as [14] ( ) de D = Φ (4) dx

3 MEBRAHTU et al.: SPICE MODELS OF FLUORINE-ION-IRRADIATED CMOS DEVICES 1965 where D is the dose in radians(si), de/dx is the linear energy transfer (LET) in electronvolts per angstrom, Φ is the ion flux per square centimeter, and the constant is a conversion factor to radians(si). The data for LET, at the SiO 2 /Si interface, is estimated from SRIM. For this setup, the dose measurement was determined with 16% uncertainty. A. Device Under Test Several n + -p photodiode arrays were fabricated using a standard 0.35-µm CMOS process. To enhance the measured current and average out any diode nonuniformities, several diodes are tied in parallel. The photodiodes in an array shared a common externally accessible anode and an internally grounded cathode terminal. The different sets of photodiode arrays were isolated via +1-V guard lines to suppress crosstalk effects between adjacent active areas. Each photodiode array contained 102 photodiode cells arranged in two rows, each with an active area of 5 30 µm 2. A 0.5-µm spacing was maintained between pixel cells within a row while the rows were designed 21.7 µm apart. The enhancement-type nmos transistor chip, fabricated on a different run of the same CMOS technology, contained an array of 39 transistors. All these transistors share a common gate, with their source and body terminals tied to ground, and had separate drain terminals. Each of these transistors was 4 by 3 µm in size. Moreover, the pmos-type transistors had similar transistor configuration to the nmos. This chip also contained 39 pmos-type transistors but with 6-µm channel length and 20-µm width. Fig. 2. Experiment setup for beam alignment. B. Experimental Setup A 10-µACs + beam with an energy of 27.5 kev was used to sputter a target of a Fluorine compound mixed with Nb powder. A F beam was generated from this target and was extracted from the ion source with 20-keV energy. The fluorine ions from the ion source were accelerated to a high-voltage terminal held at MV. The ions were then momentum analyzed in a spectrometer magnet with radius of 1.32 m at a field of T. The ion beam, 3 mm in diameter, was then transported to an end station described in the following. The end station, one element of which is shown in Fig. 2, consisted of a printed circuit board, test chips, electrodes, and Faraday cups configured in such a way that, for each irradiation location, the test chip and the cup, with an aperture in the base to collimate the beam, were aligned [15]. Furthermore, the full beam current was monitored using an additional Faraday cup (without an aperture at the base), as accurate Faraday-cup current measurements were important. In addition, various mitigation techniques were employed to protect the currentmeasurement cup and the circuitry from secondary electrons. Moreover, the device was isolated from all sources of illumination. V. RESULT A. Photodiode The dark-current characteristics of the irradiated photodiode are shown in Figs. 3 and 4, in which it is shown that Fig. 3. Dark-current characteristics of a photodiode array before and after F +7 -ion radiation. Fig. 4. Dark-current characteristics of a photodiode array before and after F +7 -ion radiation. the fluorine irradiation significantly altered the device characteristics. Reverse-biased dark-current characteristics of the diodes increased by an order of magnitude after the first set

4 1966 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 54, NO. 8, AUGUST 2007 Fig. 5. I DS V GS characteristics of MOS transistors irradiated with Fluorine-+7 ions (log scale). (a) nmos. (b) pmos. of measurements (Fig. 4). Similar behavior is also observed in small forward-bias voltages. This could indicate increased carrier generation and recombination rates in the diode and can be attributed to the enhancement of generation recombination centers in the device due to displacement damage. In reversebias mode, a decrease in carrier generation is observed with further irradiation, although similar effects can be slightly noticed in at small forward base (< 0.15 V). As discussed in the following, this is believed to be due to the passivation effect of the implanted fluorine ions; as the device receives more fluorine ions, some of the fluorine ions deactivate some of these generation recombination centers. In irradiated photodiodes, current in the diffusion-dominated region is also found to increase. This also agrees with the proposed increase in generation recombination centers where such an increase can result in a reduction of minority carrier diffusion length. At higher applied forward-bias voltages ( 0.6 V), the voltage drop in the device becomes significant since the voltage shown by the p-n junction is less than the voltage at the terminal by a factor of IR, where R is the resistance within the diode. This results in deviation of the diode characteristics from the expected ideal straight-line curve on a log scale. As shown in Fig. 3, this deviation increases with F-ion irradiation. The result suggests that resistance of the irradiated device has increased. Thus, at higher voltages, the current voltage (I V ) characteristics curve after irradiation lies below the preradiation values. B. CMOS Transistors The I V characteristics of the irradiated nmos and pmos transistors were changed significantly over all ranges of device operation. Fig. 5 shows I DS versus V GS characteristics of the transistors. Significant deviations from preirradiation characteristics can be noticed. The absolute threshold voltage is shown to increase in both types of transistors, although at much higher dosage compared to photodiodes. Moreover, the slope of the curves above threshold decreases in both transistors with increasing dose. The behavior of irradiated pmos transistors in the subthreshold region is also illustrated in Fig. 5. Current in the accumulation region of the transistor jumped from to A. In both tested nmos [Fig. 5(a)] and pmos [Fig. 5(b)] transistors, subthreshold curve shifted toward the direction of increased absolute gate voltage. Moreover, after irradiation, the slope of curves stretched out modestly. In contrast to the shift of subthreshold-current curves of an nmos transistors reported in literature [10], F-ion irradiated curves of the nmos transistor shifted in the direction of a higher gate voltage. Similarly, the subthreshold current of a pmos transistor shifts toward more negative gate voltage. Positive traps in the oxide tend to turn on an nmos transistor while they tend to turn off a pmos transistor, shifting the I V characteristics of the device toward more negative. Any negative charges trapped in the oxide would create an exactly opposite phenomenon. However, the observed shift in nmos and pmos transistors does not correspond with this observation. This implies that the dominant mechanism responsible for the observed I V characteristics shift for F-ion irradiated transistors is not charge related. Interface states created by total ionizing dose damage are characterized by voltage-dependent charge states. The extent of the subthreshold stretch out is used to estimate the amount of interface states and oxide traps in the subthreshold-current measurement technique [10]. In nmos transistors exposed to high level of ionizing dose radiation, rebound effects in threshold voltage are observed [9]. This is attributed to a large increase in interface states associated with the irradiation. However, the stretch out of the subthreshold current, which is shown in Fig. 5(a), is only modest. In SRIM simulations of fluorine, which is a highly electronegative element, irradiated devices fabricated under 0.35-µm process; a small fraction of the implanted ions remain in the gate-oxide insulation layer, while a large fraction remains close to the SiO 2 /Si interface. A fluorinated capacitor was

5 MEBRAHTU et al.: SPICE MODELS OF FLUORINE-ION-IRRADIATED CMOS DEVICES 1967 reported to exhibit a decrease in oxide capacitance [16]. In the cited paper, it was proposed that fluorine reacts with the weak and dangling bonds in the silicon-dioxide layer and surface by displacing oxygen ions from the bond. The displaced oxygen ions migrate to the SiO 2 /Si and polysilicon interfaces and grow an additional oxide layer. The proposed oxide-growth model agrees with the observations made in these experiments. The noncharge nature of the threshold-voltage shift and the shift in the subthreshold I V characteristics are possibly due to a decrease in oxide capacitance and carrier mobility. For an nmos transistor irradiated with F-ion, a positive shift due to oxide growth outweighs any negative shift associated with positive oxide traps in the gate-oxide layer. Moreover, such oxide growth can contribute to the large decrease in postthreshold current. The current increase observed in the accumulation region of irradiated transistors, the reverse-bias current of photodiodes, and the small decrease in leakage current can be related to the passivation effect of fluorine. The initial increase in leakage current is possibly due to enhanced carrier generation from displacement damage. As fluorine ions start to react with the dangling bonds, the carrier generation rate is abated. VI. MODELING F-ION IRRADIATION EFFECTS In this section, the fluorine-ion-induced characteristics changes, which is discussed in Section V, are modeled through a proposed SPICE-based device models. MOS transistors and photodiodes are the basic components found in CMOS APS. Thus, the models are useful to simulate circuit-level response of such CMOS image sensor exposed to similar irradiation conditions. In the following sections, appropriate mathematical models are discussed and corresponding SPICE model presented. A. Photodiode-Parameter Extraction We treat the diffusion and generation recombination current separately. The current of the photodiode, when not illuminated, can be approximated as linear combination of diffusion I df and generation recombination I gr current contributions as I I df + I gr. (5) A fit to the data is obtained by segmenting the data into forward- and reverse-bias regions. In the forward-bias region, the data is fitted to ( [ ] ) 1/2 I F p p 3 (e (V IR)/VT 1) 1 + p 2 (e p 4(V IR) 1) (6) where the first term corresponds to diffusion-current contribution, while the second term describes current due to generation recombination contribution. In this equation, the parameter 0.5p 1 p 3,2n i / p 3, p 2 numerically corresponds to Fig. 6. Recombination current parameter (p2) versus dose. diffusion-current coefficient of the diode, the doping concentration and generation recombination current contributions, respectively. The bulk and contact resistance drops are incorporated to the equation via the IR term. The reverse-bias region is particularly important for CMOS APS sensors. It is the operation region of the photo-sensing diode. The I V characteristic of the diode is dominated by carriers generated in the depletion region and an accompanying soft breakdown, both of which are voltage-dependent. After irradiation, the recombination-current term in (6) is found to have a high ideality factor. At early reverse bias (> 0.1 V), its contribution is higher than the observed current. Hence, the reverse-bias region is modeled separately, and the contribution of the forward-bias model to the reverse-bias region is isolated in the model. To parameterize voltage dependence of the reverse-bias diode, the data is fitted with the following equation: I r q 1 (e q 2V 1). (7) The parameters are extracted using nonlinear optimization routines in MATLAB. Both gradient-based, Gauss Newton, and direct Nelder Mead algorithms are used to compare local convergence results. The initial parameter values can be roughly estimated from processing technology and geometry of the devices. For 0.35-µm process and the geometry of the test devices, the preirradiation initial parameters are calculated to the following order of magnitude: p , p , p , p 4 20, and R 5. These initial parameters are varied up to an order of magnitude to search global convergence around initial estimates. These model equations describe the observed data up to 100 krad. Fig. 6 shows the extracted parameter change versus dose for one of the parameters (coefficient of the recombination-current term). At a higher dose, the various current components in the forward-bias region become difficult to isolate, and the convergence of the equations to the data degrades. With increased dose, the bias dependence of the voltage drop in the bulk of the device becomes important, and the approximation with constant resistive element R is a limitation.

6 1968 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 54, NO. 8, AUGUST 2007 Fig. 7. Behaviorial SPICE model of an irradiated (a) photodiode and (b) static-small-signal nmos transistor. B. SPICE Photodiode Model F-ion irradiation affects the various carrier-generation processes. Diffusion, generation recombination, doping, and mobility of carriers are changed at different rates. The diode model, as shown in Fig. 7(a), enables treatment of these factors separately. In comparison to the standard PSPICE diode, which lumps the diffusion- and recombination-current contributions to a single-diode component via an ideality factor, recombination generation and diffusion-current contributions are separated as three components: a voltage-controlled current source for the diffusion component I df and separate diodes for generation D g and recombination D r currents. The voltage dependence in the reverse-bias region is included via a reversely connected diode (D g ). The photo-generated current is represented by a current source (I ph ). The resistance inside the device and contacts is included as (R s ) in series. The effect of the modeled forward-bias recombination-current component in the reverse-bias models is isolated via an ideal diode (D ideal ) in front. The heavy ion behaviorial diode model, as shown in Fig. 7(b), and a standard SPICE diode model are compared in Fig. 8. The model parameters that are extracted by fitting the corresponding equations to the irradiated photodiode at 78 krad are as follows: for D gr diode, the Is and N parameters were A and 1, respectively; similar parameters for diode D dif were A and 1.995, respectively. Moreover, the parameters for R s, R sh, and C jo were 12.7 Ω, Ω, and F, respectively. As shown in Fig. 8, the standard SPICE diode fails to reproduce the data at low forward-bias voltages. However, at higher voltages, the diffusion component of the diode current dominates, and the difference between the two models diminishes. C. MOS Fitting Parameters We model static and small signal behavior of the transistor s operation under F-ion irradiation based on a standard static MOS SPICE model shown [Fig. 7(b)]. The model incorporates the model of F-ion-irradiated diode proposed in Fig. 7(a) as D rad. To simulate weak- and strong-inversion regimes of Fig. 8. Parameter fit for a photodiode irradiated to 78 krad. operation, an appropriate mathematical representation of I DS, as a function of gate, drain, and source biases, applicable to all transistor-operation regions is needed. For mathematical fitting purposes, a single expression for I DS is required. The EKV model representation of the I DS current provides such an empirical relationship and is valid in both the weak- and stronginversion regions of transistor operation [17]. For nmos, the drain current can be expressed as a function of gate, drain, and source bias by I DS = 2nµC ox ( W L )( kt q ) 2 ( ) (log e (V G V th )/n V S 2kT/q ( log e (V G V th )/n V D 2kT/q )) where V G, V D, and V S are the gate, drain, and source biases, respectively, V th is the threshold voltage, C ox is the oxide capacitance, and µ is the carrier mobility. The constant n is called the body factor and is equal to 1 + C D /C ox, where C D is the depletion capacitance. Equation (8) is fitted to I DS versus V GS to obtain preradiation values of mobility, body factor, and threshold voltage. For pmos, these extracted parameters are 262 ± 40 cm 2 /V s, 1.3, and ± V, respectively, while similar parameter extraction for nmos yield 389 ± 11 cm 2 /V s, 1.4, and ± V, respectively. To reduce the gate-voltage effect on channel mobility, the parameters are extracted for pmos V G 1.5 V and nmos V G 1.0 V. After F-ion irradiation, the I V characteristics of the device change. As a result, the different parameters in (8) change. As discussed previously, two dominant effects are considered: changes in threshold voltage and effect of gate voltage in the channel due to oxide-capacitance changes and degradation of carrier mobility in the channel of the device. Hence, a fit to postirradiation-transistor characteristics is achieved using two parameters: a 1 and a 2. (8)

7 MEBRAHTU et al.: SPICE MODELS OF FLUORINE-ION-IRRADIATED CMOS DEVICES 1969 The change in threshold voltage due to changes in oxide capacitance can be related by V th = Q D Cox 2 C ox (9) where Q D is the depletion charge. Similarly, the change in body-factor is related to change in oxide capacitance as n = C D Cox 2 C ox (10) where C D is the depletion capacitance. Combining (9) and (10), the following relationship between threshold voltage and body-factor is obtained: V th = Q D C D n. (11) Considering these two dominant degradation mechanisms and estimating C D and Q D from the manufacturer s data-sheet, the two parameters (a 1 and a 2 ) are extracted from irradiation I DS versus V DS data as ( ) V G + Q D n 0 +V I DS = a 1 log 2 C th0 a 2 Q D V D CD S 1 + e 2kT/q Fig. 9. (symbols) Irradiated nmos I V characteristics with (line) parameter fit. ( ) V G + Q D n 0 +V log 2 C th0 a 2 Q D V D CD D 1 + e 2kT/q (12) where V th0 and n 0 are preradiation threshold voltage and body factor values, respectively. The mobility and oxide capacitance of the transistor are numerically equal to C ox = C Da 2 1 a 2 (13) µ = a 1 (1 a 2 ) 2(W/L)C D ( kt q ) 2. (14) The curves obtained by fitting the two extracted parameters to the preirradiated and postirradiated I DS versus V GS data are shown in Fig. 9. The fitted curve captures the characteristics of the irradiated I V data well (R 2 = ). Mobilitydegradation and oxide-capacitance changes caused by F-ion implantation possibly account for the reduced slope of the curve, beyond the threshold voltage. In addition, oxidecapacitance-caused changes in body-factor and threshold voltage shift the I DS V GS curve to higher gate voltages. The extracted parameters (a 1 and a 2 ) versus radiation dose indicate decreases in mobility and oxide capacitance with F-ion dose up to a total dose of 10 7 rad (Fig. 10). This oxidecapacitance decrease due to F-ions could arise due to changes in dielectric constants of the oxide layer or changes in oxide thickness. However, the latter is considered the significant cause of the change for F-ion-irradiated dielectric oxides [16]. The observed capacitance change corresponds to about 30% increase in oxide thickness or 2 nm. Fig. 10. Extracted carrier mobility for nmos transistor. D. Dark Current in CMOS APS Dark current, which is the charge generated by sensors under zero illumination, is an important characteristic parameter of an image-sensor pixel. A typical CMOS APS circuit is shown in the inset of Fig. 11. It consists of a photodiode detector and three transistors for reset, buffer, and cell readout. In the simulation, a photodiode with an area of µm 2 is used. The transistors are all nmos of with W/L ratio of 4 µm/3 µm, respectively. The simulation predicts that the dark current increases by two orders of magnitude, at V photo node, just after 1 krad of F-ion irradiation. Moreover, as isolated in Fig. 11, the photodiode is found out to be the dominant source of dark current in F-ion-irradiated CMOS APS. VII. CONCLUSION Heavy ions in an energy range of 20 MeV inflicted significant damage on photodiodes and transistors manufactured using a 0.35-µm process. Indeed, the range of ions in this

8 1970 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 54, NO. 8, AUGUST 2007 Fig. 11. Simulation of dark-current contribution from CMOS APS components. A standard 3T CMOS APS pixel sensor is shown in the inset. energy range allows penetration close to the sensitive semiconductor regions and alters the electrical parameters of devices. For instance, parameter changes such as resistance increase due to implantation of ions were observed on the tested photodiode, and threshold-voltage increase and mobility degradation was observed in the tested nmos and pmos transistors. In fact, devices exposed to other heavy ion species of comparable energy range are expected to show similar resistance increase and mobility degradation. However, F-ion irradiation of transistors showed peculiar increase in the absolute threshold voltage of both nmos and pmos in contrast to ionization-radiationinitiated threshold-voltage shifts reported earlier. In addition, fluorine ions in the energy range studied caused significant displacement damage. The damage was reflected in increased current in the generation recombination-dominated regions of I V characteristics of the photodiode and increased leakage current in the nmos and pmos transistors. This is possibly due to increased carrier generation in the parasitic reverse-biased p-n junction at the drain and source nodes of MOS transistors. The generation recombination current parameters increased by almost an order of magnitude after a 1-krad dose indicating possible decrease of the carrier lifetime. Similar damages are also expected from other heavy ion species of comparable energy. Comparisons of parameter degradation induced by F-ion irradiation showed photodiode operation to be more susceptible than transistors. This observation has significant implication for CMOS APS sensors. Through the SPICE models developed for transistor and photodiode, it is found out that the photodiode element in the sensor is the most vulnerable component to a possible overall failure of a CMOS APS image sensor. For instance, this is important for large photodiodes typically used in space application. ACKNOWLEDGMENT The authors would like to thank J. S. Lee for the design and the Canadian Microelectronics Corporation for fabrication of the chips used in this paper. REFERENCES [1] T. Oldham and F. McLean, Total ionizing dose effects in MOS oxides and devices, IEEE Trans. Nucl. Sci., vol. 50, no. 3, pp , Jun [2] H. Hughes and J. Benedetto, Radiation effects and hardening of MOS technology: Devices and circuits, IEEE Trans. Nucl. Sci., vol. 50, no. 3, pp , Jun [3] F. Sexton, Destructive single-event effects in semiconductor devices and ICs, IEEE Trans. Nucl. Sci., vol. 50, no. 3, pp , Jun [4] J. Srour, C. Marshall, and P. Marshall, Review of displacement damage effects in silicon devices, IEEE Trans. Nucl. Sci., vol. 50, no. 3, pp , Jun [5] G. Hopkinson, Radiation effects in a CMOS active pixel sensor, IEEE Trans. Nucl. Sci., vol. 47, no. 6, pp , Dec [6] J.-P. David and M. Cohen, Radiation-induced dark current in CMOS active pixel sensors, IEEE Trans. Nucl. Sci., vol. 47, no. 6, pp , Dec [7] B.D.J.Bogaerts,Total Dose Effects on CMOS Active Pixel Sensors. [8] B. V. Zeghbroeck, Principles of Semiconductor Devices, [9] T. P. Ma and P. V. Dressendorfer, Ionizing Radiation Effects in MOS Devices and Circuits, T. P. Ma and P. V. Dressendorfer, Eds. New York: Wiley, [10] P. McWhorter and P. S. Winokur, Simple technique for separating the effects of interface traps and trapped-oxide charge in metal-oxidesemiconductor transistors, Appl. Phys. Lett., vol. 48, no. 2, pp , Jan [11] J. F. B. McLean, H. E. Boesch, and T. R. Oldham, Electron-hole generation, transport, and trapping in SiO 2, in Ionizing Radiation Effects in MOS Devices and Circuits, 1st ed. T. Ma and P. V. Dressendorfer, Eds. Hoboken, NJ: Wiley, [12] J. F. Ziegler, [Online]. Available: [13] IsoTrace-Laboratory, Univ. Toronto. [Online]. Available: physics.utoronto.ca/ isotrace/new/ [14] N. Kalkhoran, E. Burke, and F. Namavar, Charged particle radiation effects on bulk silicon and SIMOX SOI photodiodes, IEEE Trans. Nucl. Sci., vol. 42, no. 6, pp , Dec [15] H. Mebrahtu, Heavy ion radiation effects on CMOS image sensors, M.S. thesis, York Univ., North York, ON, Canada, [16] P. Wright and K. Saraswat, The effect of fluorine in silicon dioxide gate dielectrics, IEEE Trans. Electron Devices, vol. 36, no. 5, pp , May [17] E. V. C. Enz and F. Krummenacher, An analytical MOS transistor model valid in all regions of operation and dedicated to low-voltage and lowcurrent applications, Analog Integr. Circuits Signal Process., vol.8,no.1, pp , HenokT.Mebrahtureceived the B.Sc. degree in physics from the University of Asmara, Eritrea, and the M.Sc. degree in physics from York University, Toronto, ON, Canada, in He is currently working toward the degree in physics at Duke University, Durham, NC. He was a Junior Scientist in the VISion sensor Laboratory, York University, specializing in visionsensor research, where he is currently working in the development of simulation models for CMOSbased cameras and night-vision goggle devices. His research interests include device physics modeling and radiation response. Wei Gao received the B.S. degree in applied physics from Harbin Institute of Technology, Harbin, China, in 1990, and the M.S. degree in electrical and computer engineering from Queens University, Kingston, China, in She is currently working toward the Ph.D. degree in the Computer Science and Engineering Department, York University, Toronto, ON, Canada. Her research interests are in low-power and highdynamic-range CMOS image-sensor design.

9 MEBRAHTU et al.: SPICE MODELS OF FLUORINE-ION-IRRADIATED CMOS DEVICES 1971 William E. Kieser received the M.Sc. and Ph.D. degrees in nuclear structure physics from the University of Toronto, Toronto, ON, Canada, in 1971 and 1977, respectively. He worked in nuclear astrophysics at the University of Münster, FRG. In 1980, he began work in accelerator mass spectrometry (AMS) in the Iso- Trace Laboratory, University of Toronto, where he is currently the Assistant Director. His research interests include the development of new equipment for AMS, the application of radiocarbon analysis in biomedical and pharmaceutical studies, and the use of radio iodine analysis in oceanographic and atmospheric-circulation measurements. Xiaolei L. Zhao received the M.Sc. and Ph.D. degrees in accelerator mass spectrometry (AMS), an applied field of experimental nuclear physics, from the University of Toronto, Toronto, ON, Canada, in 1989 and 1993, respectively. He continued working as an AMS Isotope Analyst and carrying out research to explore new analytical techniques in AMS, IsoTrace Laboratory, University of Toronto. He is also interested in exploring other novel applications of tandem accelerators, such as studies of fragile negative ions of atoms and molecules. Richard I. Hornsey received the degree in engineering science and the Doctorate degree for research on the mechanisms and applications of liquid-metal ion sources from Oxford University, Oxford, U.K., in He spent a year as a Visiting Researcher for Hitachi, Tokyo, Japan, investigating the use of focused ion beams for semiconductor-device analysis. On returning to England, he spent three years as Wolfson Hitachi Research Fellow in the Cavendish Laboratory, Cambridge University, fabricating and analyzing quantum electronic devices. In 1994, he was with the Department of Electrical and Computer Engineering, University of Waterloo, where his research interests concerned solid-state image sensors, including thin-film technologies for large-area image sensors, and the technology and applications of CMOS-integrated electronic cameras. In 2001, he was with York University, Toronto, ON, Canada, to participate in the establishment of the new engineering programs. He is currently the Associate Dean of the School of Engineering, York University. He is also cross-appointed to the Departments of Computer Science and Engineering and Physics and Astronomy. He is also a member of the Centre for Vision Research, where his research focuses on integrated image sensors and vision systems. Dr. Hornsey was the recipient of Ontario Premier s Research Excellence Award in He is a licensed Professional Engineer in Ontario. Paul J. Thomas received the Ph.D. degree in physics in He is currently the President of Topaz Technology Inc., a consulting and R&D firm in Toronto, ON, Canada. His interests include the application of lasers and imagers to 3-D remote sensing.