Femtosecond Laser Simulation Facility for SEE IC Testing

Transcription

1

2 Femtosecond Laser Simulation Facility for SEE IC Testing Andrey N. Egorov, Alexander I. Chumakov, Oleg B. Mavritskiy, Alexander A. Pechenkin, Dmitry V. Savchenkov, Vitaliy A. Telets, Andrey V. Yanenko Abstract The new SEE laser simulation facility based on femtosecond laser source with tunable pulse duration is presented, and its most important features are discussed. The influence of laser pulse duration on simulation results is observed. Index Terms Laser simulation facility, single event effect, femtosecond pulse duration. U I. INTRODUCTION SING ultra-short laser pulses is a well-proven technique to simulate Single Event Effects (SEEs) in semiconductor devices for space applications. Recently we have presented the PICO series of SEE laser simulation facilities based on solid-state picosecond laser sources [1 3]. The most versatile PICO-4 laser simulation facility [3] incorporates the wavelength tunable picosecond laser source, capable of simulating ionization tracks produced by single ions with various penetration depths in a wide range of semiconductor devices (Si, GaAs, etc.). According to our experience in laser simulation tests, the most appropriate wavelength to be used for front-side SEE testing of Si and GaAs devices is about nm, which gives rather good efficiency of charge generation together with adequate length of simulated ionization tracks [4,5]. The pulse duration of PICO laser sources is in ps range, which is short Manuscript received February 6, Andrey N. Egorov is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(495) , ext. 2031, anegor@spels.ru). Alexander I. Chumakov is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(495) , ext AIChumakov@mephi.ru). Oleg B. Mavritskiy is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(495) , ext. 8294, OBMavritskiy@mephi.ru). Alexander A. Pechenkin is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(495) , ext. 6944, aapech@spels.ru) Dmitriy V. Savchenkov is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(495) , ext. 5031, dvsav@spels.ru) Vitaliy A. Telets is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(499) , VATelets@mephi.ru) Andrey V. Yanenko is with National Research Nuclear University MEPhI, Moscow, Russia (telephone: +7(495) , ext. 5011, avyan@spels.ru) enough for SEL or SET simulation tests of mainstream Si devices due to extended characteristic time of these effects. However, this may not be the case if SEU or SET effect is simulated in the modern ultrafast devices, having very short internal memory cell switch times (such as flip-flops, registers or SRAMs). Simulations [6,7] show, that there may be an increase in SEU thresholds when using laser pulses longer, than several picoseconds. Therefore, it may be very interesting to use shorter laser pulses in order to determine whether the pulse duration in particular case affects the SEU simulation results. Also, the possibility to vary the pulse duration in one facility could provide a flexible instrument for SEU simulation of modern ultrafast devices. Here we introduce the newly designed FEMTO-T experimental SEE simulation facility utilizing femtosecond laser source with tunable pulse duration. It can be used in combination with PICO series laser simulators, or as a standalone setup for the investigation of wide class of SEEs in modern Si and GaAs ultrafast space applications oriented devices. The main features, distinguishing this facility from PICO-4 laser simulator, are discussed. Some comparative results of simulation tests under femtosecond and picosecond laser irradiation, performed by local laser technique [8], are also presented. II. LASER SIMULATION FACILITY DESCRIPTION Fig. 1 presents the general configuration of FEMTO-T laser simulation facility. The main distinction of the new system from PICO-4 laser simulator [3] is the incorporation of AVESTA s MPAP 2500 laser source. This type of laser produces femtosecond laser pulses with output energy sufficient for local laser technique application, rather good energy stability and laser beam quality, that is very important for focusing into a micron-sized spot, and the variable pulse duration from hundreds of femtoseconds to several picoseconds. Femtosecond pulses of low energy are generated by TiF-50 Ti:Sapphire master oscillator, pumped with second harmonic of built-in CW diode-pumped Nd 3+ :YAG laser. From the output of master oscillator, the pulses pass to the stretcher, which increases their duration up to 200 ps for further amplification. The amplification subsystem consists of regenerative amplifier (RA) and additional multi-pass amplifier (MA), combined with pulse picker, producing the output pulses with up to several hundreds of µj with /14/$31.00@2014 IEEE 247

3 TABLE I MAIN PARAMETERS OF FEMTO-T LASER SIMULATION FACILITY Parameter Unit Value Wavelength nm 870 Laser pulse energy nj Pulse duration fs Minimum laser spot diameter a) µm 1.2 Maximum pulse frequency Hz 100 Output beam quality M 2 <1.6 Objective s magnification XYZ translation stage resolution µm 0.15 a) measured for 100 objective Fig. 1. Schematic diagram of FEMTO-T SEE laser simulation facility. controlled frequency or in single shot mode. To pump the RA and MA, additional frequency-doubled Q-switched diodepumped Nd 3+ :YAG laser is used. After amplification, the output pulses duration is reduced in tunable compressor to the value necessary for SEE simulation. Beam expander, located at the output of laser source, enlarges laser beam diameter to match the input pupil of the objective, thus achieving the smallest possible laser spot diameter in the focus point. To monitor the output pulse duration and shape the single-shot ASF-200 optical autocorrelator is used. The rest of the FEMTO-T optical scheme is similar to described in [3] and includes variable attenuator, focusing unit (microscope) with illuminator, CCD camera, XYZ translation stage, and control PC with all necessary interfaces. The most important parameters of FEMTO-T are presented in Table I. It should be noted, that laser beam can be focused to the spot as small as 1.2 µm in diameter, that is very close to the optical diffraction limit. Also, the new design of XYZ translation stage is implemented, which resulted in enhanced mechanical stability and greater DUT scanning speeds. The process of simulation tests is fully automated by using originally designed PC software that can be adjusted for specific parameters measurements. III. EXPERIMENT AND RESULTS SEU threshold energy J depends on laser wavelength and pulse duration τ p. In particular, if we fix the wavelength, then the following approximation [7] can be used: ν 1 τ ν p J = J 0 1 +, (1) RC where J 0 is threshold laser energy for τ p 0; RC is characteristic ionization response time, and ν is the approximating exponent coefficient with typical value about 1.5. One can see that for the pulse durations comparable or greater than RC the threshold laser energy increases. Thus, for the ICs with characteristic times shorter than several tens of picoseconds the utilization of femtosecond laser pulses for SEE testing becomes preferable. To reveal the possible influence of laser pulse duration on the results of SEU simulation tests, we have performed the measurements of the SEU laser energy thresholds for SY55852U CML ultrafast flip-flop IC with pulse duration τ p =200 fs. The flip-flop operated at 3.3 V power supply voltage in static storage mode, and the change of output pin logical state at the impact of laser pulse indicated the occurrence of SEU. The schematic diagram of the test circuit is presented in Fig. 2. In our experiment, in accordance with local irradiation technique [8], we have thoroughly scanned the whole IC chip from the front side with the femtosecond laser beam about 30 µm in diameter, and found a compact single area (see Fig. 3), susceptive to SEU. Then, the SEU threshold energies of laser pulse were measured using various laser spot diameters from 1.2 µm to 100 µm. Both 0 and 1 initial states of flip-flop were tested, and lowest threshold energy value was chosen. The obtained dependence (see Fig.4) is rather typical for this technique. Based on this dependence the size of SEU sensitive area in this IC can be calculated using RPP model, and the corresponding calculations show that it is rather small (not more than several µm). The minor fluctuations of the curve s smoothness can be attributed to the local inhomogeneity of optical losses due to the presence of metal layers. These results were practically independent on the laser pulse duration in all available for this simulation facility range. Fig. 2. Schematic diagram of the IC test circuit. 248

4 10 µm Fig. 3. The location of SEU sensitive area under test on SY55852U crystal. In addition, to estimate the value of optical losses we have registered the ionizing response in the power circuit of IC under laser irradiation for various laser pulse energies. In this experiment the test circuitry of the IC was switched to photodiode mode with zero bias voltage (power supply pin of the IC was grounded). The typical oscillogram of the voltage pulse on a current sensing resistor R c and the measured dependence are presented in Fig. 5 and 6 respectively. This information could be further used for laser energy to LET calibration by the accelerator tests. Finally, the whole test procedure was repeated using PICO-4 simulation facility with τ p =25 ps, tuned to the same wavelength (870 nm). The corresponding ionization response amplitude vs. laser pulse energy dependence is also presented in Fig. 6. The most important result of this experiment is that, although we were able to observe the ionizing response in the most sensitive area with similar amplitude and time characteristics, no SEU was registered up to the laser pulse energies, close to the damage threshold of the IC structure. Furthermore, the SEU effect was not observed using shorter wavelengths available for PICO-4 facility (down to 700 nm). The possible explanation for this experimental fact is that internal charge generation and collection dynamics in SY55852U IC for femtosecond laser pulse irradiation differs greatly from that for longer time range pulses, but for some reasons this can t be observed in the power supply circuit (for example, due to the influence of the capacitance of the chip itself or elements of measurement scheme). In favor to this hypothesis one can note that the pulse duration of the ionization response (see Fig. 5) is about 10 ns that is much greater than the flip-flop switch time characteristics and, a fortiori, the laser pulse duration. Obviously, such significant difference in the behavior Fig. 4. SEU threshold energy J vs. laser beam diameter D dependence in SEU sensitive point for IC SY55852U. Fig. 5. Oscillogram of the SY55852U ionization response U in power supply circuit under pico- and femtosecond laser irradiation for laser pulse energy of about 0.05 nj. 249

5 and Testing Facility Based on Wavelength Tunable Picosecond Laser," Radiation Effects Data Workshop (REDW), 2012, IEEE, pp [4] D.V. Savchenkov, A.I. Chumakov, A.G. Petrov, A.A. Pechenkin, A.N. Egorov, O.B. Mavritskiy, A.V. Yanenko, Study of SEL and SEU in SRAM Using Different Laser Techniques, presented at 14th European Conference on Radiation Effects on Components and Systems (RADECS), [5] A.V. Nechaev, O.B. Mavritskii, A.N. Egorov, P.K. Skorobogatov Study of ionizing response of semiconductor structures under laser irradiation in wavelength range nm, Bulletin of the Russian Academy of Sciences: Physics, vol. 70, no. 9, pp , [6] A. Douin, V. Pouget, F. Darracq, D. Lewis, P. Fouillat, and P. Perdu, Influence of Laser Pulse Duration in Single Event Upset Testing, IEEE Trans. Nucl. Sci., vol. 53, pp , [7] A.I. Chumakov and V.V. Gontar, Predicting the Failure Threshold of Dose Rate for ICs Exposed to Pulsed Ionizing Radiation of Arbitrary Pulse Shape, Russian Microelectronics, vol. 33, no. 2, pp , [8] A.I. Chumakov, A.A. Pechenkin, D.V. Savchenkov, A.S. Tararaksin, A.L. Vasil ev and A.V. Yanenko, Local laser irradiation technique for SEE testing of ICs, in Proc. 12th European Conference on Radiation and Its Effects on Components and Systems (RADECS), 2011, pp Fig. 6. Ionizing response U in IC SY55852U power supply circuit (for current sensing resistor R cs=50 Ohm) vs. laser pulse energy J. under femto- and picosecond laser irradiation can be observed only in devices with ultrafast internal switching times (for example, modern nanoscale process SRAMs or GaAs UHF devices). Further investigations are planned to be done to collect more experimental data for various other types of devices for space applications. IV. CONCLUSIONS In this work, we introduce the new FEMTO-T SEE laser simulation facility, based on femtosecond laser source. The main features of this facility are: 870 nm wavelength (optimal for Si and GaAs devices front-side SEE testing), very small focused laser beam diameter and the ability to change the output laser pulse duration in 70 to 1500 fs range. Using ultra-short laser pulses with tunable pulse duration may be helpful to determine whether the pulse duration affects the SEU simulation results. To illustrate this, we have experimentally studied the influence of laser pulse duration on the results of SEU simulation tests in SY55852U CML ultrafast flip-flop IC using local laser irradiation technique. It was shown, that SEU in this IC could be simulated only using femtosecond laser pulses. REFERENCES [1] A.I. Chumakov, A.N. Egorov, O.B. Mavritsky, A.V. Yanenko, Evaluation of moderately focused laser irradiation as a method for simulating single-event effects, Russian Microelectronics, vol. 33, no. 2, pp , [2] A.N. Egorov, A.Yu. Nikiforov, A.I. Chumakov, A.V. Yanenko, O.B. Mavritskiy, A.A. Pechenkin, Laser Facilities for Radiation Effects Simulation and Testing, presented at RADLAS 2011, Suresnes, France, Sep. 16, [3] A.N. Egorov, A.I. Chumakov, O.B. Mavritskiy, A.A. Pechenkin, D.O. Koltsov, A.V. Yanenko, "PICO-4 Single Event Effects Evaluation 250