COTS MOS Dosimetry on the MeMOSat Board, Results After 2.5 Years in Orbit
José Lipovetzky, Mariano Garcia-Inza, Macarena Rodríguez Cañete, Gabriel Redin, Sebastián Carbonetto, Martín Echarri, Federico Golmar, Fernando Gomez Marlasca, Mariano Barella, Gabriel A. Sanca, Pablo Levy, Adrián Faigón
CCOTS MOS Dosimetry on the MeMOSat Board, Results After 2.5 Years in Orbit
José Lipovetzky (1*, 2,3,4), Mariano Garcia-Inza (1,3), Macarena Rodríguez Cañete (1), Gabriel Redin (1), Sebastián Carbonetto(1,3), Martín Echarri (5*), Federico Golmar (3,5,6), Fernando Gomez Marlasca (2), Mariano Barella (2,3,5), Gabriel Sanca (6), PabloLevy (2,3,6), Adrián Faigón (1,3)(1) Facultad de Ingeniería, Universidad de Buenos Aires || Paseo Colón 850, Ciudad de Buenos Aires, Argentina || Phone +541152850819(2) Comisión Nacional de Energía Atómica || Bustillo 9500, Centro Atómico Bariloche, Bariloche, ArgentinaPhone +54294 444 5100 x 5349(3) Consejo Nacional de Investigaciones Científicas y Técnicas || (4) Instituto Balseiro, ArgentinaBustillo 9500, Centro Atómico Bariloche, Bariloche, Argentina(5) Centro de Micro y Nanoelectrónica del Bicentenario - Instituto Nacional de Tecnología Industrial (INTI), Argentina(6) Escuela de Ciencia y Tecnología, Universidad de San Martín (UNSAM), ArgentinaMails: [email protected] , [email protected] , [email protected], [email protected]*formerlyThis work was supproted by CONICET with PIP 2014-2016-GI-Faigon, UBACyT with 20020150200085BA, and ANPCyT with PICT2014-1812
Abstract : We present the results after 2.5 years in or-bit of Total Ionizing Dose (TID) measurements doneusing Metal Oxide Semiconductor (MOS) dosime-ters on the MeMOSat board. The MeMosat boardwas launched on July 19th 2014 at the BugSat-1“Tita” microsatellite developed by Satellogic to stayat LEO. We used as dosimeters p-channel Commer-cial Off The Shelf (COTS) MOS transistors with gateoxides of 250 nm. Before launch, a subset of transis-tors with similar drain current to voltage (I-V)curves where selected from a group of 100 devices.The temperature dependence of the (I-V) curves wasstudied to find the minimum temperature coefficientbiasing point. Then, a calibration subgroup of sen-sors was irradiated using a Co gamma source tostudy their response to TID, showing responsivitiesof ~75 mV/krad when the sensors are irradiatedwithout gate bias. Also, the post irradiation responseof the sensors was monitored, in order to include acorrection for low dose rate irradiations, yielding30 mV/krad. A biasing and reading circuit was devel-oped in order to allow the reading of up to 4 sensors.The threshold voltage was monitored during differ-ent periods of the mission. After 2.5 years in orbit,the threshold voltage of the sensor mounted on theMeMOSat Board had a V T shift of approximately 35mV corresponds to a dose of 1.2 krads. I. Introduction
Metal Oxide Semiconductor (MOS) dosimeters areMOS transistors which allow the quantification ofthe Total Ionization Dose (TID) through the shift ofthe Threshold Voltage ( V T ) caused by buildup ofpositive charge in the oxide and the generation of in-terface traps [1,2]. MOS dosimeters have been usedin space applications [3-6] and medical applications[7,8]. To allow high sensitivities, MOS dosimeters areusually manufactured in ad-hoc processes with gateoxides of hundreds of nanometers or even few mi-crometers, thicker than regulars MOS transistors [1-4,7,9]. Since the responsivity to TID is approximatelyproportional to the oxide thickness t ox [10,11]. However, several works have dealt with the use of Commercial Off The Shelf MOS transistors as MOS dosimeters in medical applications [12] or industrial applications [13,14]. This work proposes the use of a COTS transistor as a low cost sensor for on-board dosimetry in a Cubesat. II. Devices and initial calibration
The devices used in this work as sensors are COTS p-channel MOS enhancement transistors with a gateoxide thickness of ~250nm. The oxide thickness---not informed by the manufacturer---was estimatedusing gate tunnel current vs gate oxide voltage char-acteristics of the sensors during Fowler Nordheim urrent injection [13]. The MOS devices are pack-aged in a TO-72 metal case.It is known that the responsivity to TID of differentMOS dosimeters can have a high dispersion—up to30%---if are manufactured in different Si wafer [13-14]. This dispersion would introduce errors in thequantification of dose if the devices are not individu-ally calibrated. Thus, initially, a subset of 15 devicesfrom a batch of 100 MOS transistors where selectedusing as criterion to have similar drain current togate voltage (I-V) characteristics, i.e. similar V T andtransconductance values. These devices where cali-brated. A. Temperature Dependence
The temperature dependence of the I-V curve of thedevices selected for calibration was studied. The I-Vcurve of MOS dosimeters is affected by two maintemperature dependencies, the decrease of V T andthe decrease of the transconductance with tempera-ture. Usually, both effects compensate for a givendrain current value, known as Zero Temperature Co-efficient (ZTC) current (I ZTC ). Usually in MOSdosimetry, this point of the I-V curve is used to mea-sure the V T shift [4]. Fig.1 presents the I-V curvesmeasured at four different temperatures from 0 o C to60 o C, a temperature wider than the temperature fluc-tuations expected during the mission. It can be ob-served that the ZTC point of the curve is atI
ZTC =270 µA. However, due to circuital restrictions,the dosimeters where finally biased with a 200 µAdrain current to quantify the V T shift on the board.This introduced a small temperature dependence ofthe reading as will be seen later. Fig. 1. I-V characteristics of the selected devices.
B. TID Response
The response of the sensors to TID was studied toobtain the responsivity of the sensors. It is knownthat the responsivity of a MOS dosimeter depends onthe gate voltage sustained during the irradiation.Usually, a positive gate voltage will increase the re-sponsivity. However, a constraint imposed by thesatellite platform to the board with the dosimeterswas that the board would not be powered duringmost of the time of the mission. This implies that thesensors would be biased during most of the missiontime with zero volts between all device terminals. The response to irradiation of the sensors was doneusing a Co gamma rays source, at a dose rate of~0.5Gy/min (50rads/min). Some sensors where bi-ased with a gate voltage of zero volts (as would be onthe satellite), and others with a gate voltage of 9V.Fig. 2 presents the V T shift as a function of dose dur-ing irradiation up to 1.6 and 1.8 Gy with gate volt-ages of 9 V and 0 V respectively. The responsivitiesof the devices where 58mV/Gy and 6.6mV/Gy for9 V and 0 V bias. Figure 3 shows the shift of the I-Vcurve after irradiation in a device, showing how ashift in the gate voltage required to apply a givendrain current can be used as a dosimetric magnitude. Fig. 2. Response to TID during Co irradiation of onesample of MOS dosimeters with 0V and 9V gate bias.
The experiment of Fig. 2 was repeated with an incre-mental dose of 1.5Gy with zero volts bias to the 15devices of the set, yielding responsivities from6.6mV/Gy to 7.5mV/Gy, proving that each deviceneeds an individual calibration despite of havingsimilar initial V T values. This pre-use irradiation al-lows the initial calibration of each device. . Long Term Annealing. A difficulty with the use of MOS dosimeters is fad-ing, i.e. the long term recovery of V T with time, aproblem associated to the neutralization of oxidetrapped charge via tunneling or thermal excitation ofelectrons [1-4]. Annealing can introduce errors inthe quantification of TID if the sensors are calibratedat a high dose rate and then the dosimetry is carriedout at such a low dose rate that annealing causes asignificant recovery in V T , underestimating the ab-sorbed dose. This feature is also known as apparentdose rate dependency of the response of the sensor[4]. It has been reported that the final V T shift after ashort irradiation at a high dose rate followed by longterm annealing is approximately similar to the V T shift obtained after the exposure to the same dose inthe same annealing time at a much lower dose rate.To investigate into this, one of the calibrated deviceswas irradiated with the Co source up to 20.3 Gy us-ing a gate bias voltage of 0V before the launch of thesatellite. The device was kept on Earth during themission time with all terminals grounded, to evaluatethe recovery of V T after the mission time. Figure 3shows the I-V curves of the fresh MOS sensor, afterirradiation and after 2.5 years of annealing. Initially,V T shifts -152 mV. But then, after anneal, V T recovers92 mV, a 60%. Thus, the final after anneal V T shifthappensto be 60 mV, meaning that the real low doserate sensitivity of the dosimeter is only 3.0 mV/Gy.This is an important result, since the irradiation inspace will be performed in a scale time of years in-stead of minutes. Fig. 3, change in I-V curves after a 20.3Gy (2.03 krads) ir-radiation with 0 V gate bias and the recovery after 2.5years.
III. Memosat Board and Reading Circuit
Two COTS MOS transistors were used as sensor wasmounted on the MeMOSat-01 board, which was partof the payload of the BugSat-1, a microsatellite devel-oped by Argentine company Satellogic [15],launched fromthe Dombarovsky air base (Russia) inJune 19th, 2014. The BugSat-1 is on a LEO orbit with620 km of altitude and an inclination of 97.9 degreesand weights 22 kg.MeMOSat-xx is a reconfigurable platform for test-ing ReRAM nonvolatile memories in the satellite en-vironment. The first launched board, MeMOSat-01was designed to perform target specific tests on twoReRAM HfO , generating reports including experi-mental and system parameters, reported periodicallyto Earth via interaction with the satellite. MeMOSat-xx was the first step of LabOSat , a comprehensiveproject to perform experiments in space. The MeM-OSat-01 board is mounted close to the center of theBugsat-1 satellite, surrounded by a large amount ofmass in most directions. Fig. 4. The MemoSat board before mounting the MOSdosimeter.The sensors are mounted on the bottom corner of theboard.
The circuit designed to read V T shifts of the sensorwas designed taking in account that the platform ofthe satellite would not provide power during most ofthe time. Moreover, since the main objective of themission is to test the satellite itself, the MeMOSatboard would be turned on only during short inter-vals, during specific periods of time. Thus, the circuithad to ensure that the sensor would have all the timethe zero volts gate bias, except for very short readingtimes in which the reading drain current (I READ )would be applied to measure the V T shift. Also, to be ble to measure in future missions the dose in differ-ent points of the satellite, the circuit should allow thereading of four devices using the same analog to digi-tal converter (ADC).The schematic design of the circuit used in theboard is presented in Fig. 5. When the analogswitches S10 and S20 are open, the sensor (QRAD-FET) is in the “Exposure” mode, with all terminalsare grounded through 1 M Ω resistors. When theswitches are closed, the operational amplifier U2 ap-plies on the gate of the sensor the gate voltage re-quired to have a fixed drain bias current of I READ (9V-Vref)/Rref.
The gate voltage was reduced using a re-sistive divider and read using an a 12-bit ADC, yield-ing a resolution of 2.4mV per ADC count. The ana-log switches are in fact the switches of an analogmultiplexer, allowing the sequential reading of fourdevices with the same electronics.
Fig 5. Bias and reading circuit.
Device restrictions forced the use of I
READ = 200µA,which was, as explained before, not exactly the I
ZTC value. Thus, the temperature dependence of the volt-age reading was estimated from I-V curves at differ-ent temperatures, yielding ~296µV/ o C. IV. Results and Discussion
The sensors were mounted on the MeMOSat board,included on the satellite and launched. Beforelaunch, the initial V T value was read. The satellite be-gan to transmit dosimetric information after all theinitial checks andtests were performed on the rest ofthe platform, 164 days after launch. Only one of thetwo sensors resulted to be functional after this pe-riod. Figure 6 presents the V T values obtained during themission. There are two periods of time in which theinformation was not recorded, at the beginning ofthe mission and during the second year. It can be observed that V T has a clear tendency toreduce, with a high dispersion. This dispersion---of~+2ADC counts or ~+5mV---is attributed to ther-mal variations in the satellite, causing an uncertaintyin the reading of V T due to the fact that the read cur-rent was not set exactly to the ZTC point of the I-Vcurve of the device. This dispersion is consistent witha temperature amplitude of 32 o C, which agrees withthermal information provided by the satellite manu-facturer which reported less than 40 o C of tempera-ture amplitude on the board. Unfortunately, we didnot have available measurements of temperaturedone at the same times of V T readings. New editionsof the board include on board temperature measure-ments to allow thermal correction of dosimetricmagnitudes. The threshold voltage shift recorded during themission is approximately -35mV. Taking in accountthat the calibrated sensitivity of the sensor at lowdose rates was 3.0mV/Gy, the dose measured by thedetector is ~12 Gy---i.e.1.2 krads. This dose is surprisingly lower than what is usuallyexpected in a LEO orbit, higher than tens of Gy/year[16] in most satellites. However, the explanation forthe low dose measured on the experiment is thatsince the board is placed closed to the center of thesatellite, most of the mass of the rest of the systemserves as a shielding for most particles. According to[16] most high energy electrons will be shielded by alayer thicker than 300 mils (7.6 mm) of Al, beingmost of the dose caused, behind shielding, by highenergy protons. On the other hand the dose ratecausedbyhighenergyprotonsfrom theVanAllenBelts reduces approximately 10 times per inch of Alequivalent shielding. The shielding done by the atellite mass of 22 kg distributed around the sensorwould explain the low dose rate measured on theboard. Unfortunately, proprietary information of thesatellite manufacturer could not be disclosed to us tomake a more complete analysis based on MonteCarlo simulations. Fig. 6. Threshold Voltage Shift as a function of time forthe dosimeter on the MeMOSat board.
V. Conclussions and future work
The work presents thedesign and realization of MOSdosimetry on the MeMOSat board, mounted on theBugSat-1 satellite, on a 620 km low earth orbit. Thesensors used in this work where COTS MOS transis-tors, proving the feasibility of using suchdevices inlow cost experiments. The dose measured on theboard after 2.5 years is ~1.2 krad, a value compatiblewith the fact that the board was protected by themass of the rest of the satellite from most of the highenergy particles in the space environmentThe work is being continued, andthe MeMOSatboard has been replaced by a more complex experi-ment named LabOSat [20]. During the past years, theFOXFET, anew thick gate oxide MOS dosimeter wasdeveloped and tested in different application [9]. TheFOXFET proved to have lower fading and a muchhigher sensitivity, and will replace the COTS sensorsused in this work. However, low cost academicCubeSat projects might use COTS MOS transistorsas an effective way to estimate dose and perform ex-periments in orbit. The use of different sensors indifferent positions of the satellite to evaluate the rela-tive dose rates observed after different shieldingthicknesses. Also the long term fading needs to betaken in account for a reliable determination of thereal dose in the experiment.
References: [1] Andrew Holmes-Siedle, The space-charge dosimeter:General principles of a new method of radiationdetec-tion, Nuclear Instruments and Methods, Volume 121,Issue 1, 1 October 1974, Pages 169-179, ISSN 0029-554X, DOI: 10.1016/0029-554X(74)90153-0.[2] Holmes-Siedle, A. and Adams, L. RADFET: a review ofthe use of metal-oxide-silicon devices as integratingdosimeters, International Journal of Radiation Applica-tions \& Instrumentation. Part C, Radiation Physics \&Chemistry}, Vol. 28No. 2, pp. 235—244, 1986.[3] Adams, Leonard; Holmes-Siedle, Andrew, "The Devel-opment of an MOS Dosimetry Unit for Use in Space,"Nuclear Science, IEEE Transactions on , vol.25, no.6,pp.1607-1612, Dec. 1978[4] Holmes-Siedle, A.; Ravotti, F.; Glaser, M., "The Dosi-metric Performance of RADFETs in Radiation TestBeams," Radiation Effects Data Workshop, 2007 IEEE ,vol.0, no., pp.42-57, 23-27 July 2007[5] Andrew Holmes-Siedle, The space-charge dosimeter:General principles of a new method of radiationdetec-tion, Nuclear Instruments and Methods, Volume 121,Issue 1, 1 October 1974, Pages 169-179, ISSN 0029-554X, DOI: 10.1016/0029-554X(74)90153-0.[6] MacKay, G.F.; Thomson, I.; Ng, A.; Sultan, N., "Appli-cations of MOSFET dosimeters on MIR and BION satel-lites," Nuclear Science, IEEE Transactions on , vol.44,no.6, pp.2048-2051, Dec 1997[7] Rosenfeld, A.B.; Lerch, M.L.F.; Kron, T.; Brauer-Krisch, E.; Bravin, A.; Holmes-Siedle, A.; Allen, B.J.,"Feasibility study of online high-spatial-resolution MOS-FET dosimetry in static and pulsed x-ray radiationfields," Nuclear Science, IEEE Transactions on , vol.48,no.6, pp.2061-2068, Dec 2001[8] Rosenfeld, A.B."Electronic dosimetry in radiation ther-apy", Rad. Meas. 41 pp 134-153, 2006.[9] J. Lipovetzky, M.A. Garcia Inza, S. Carbonetto, M.J.Carra, E. Redin, L. Sambuco Salomone, A. Faigon ,“Field Oxide n-channel MOS Dosimeters Fabricated inCMOS Processes. ”, IEEE Transactions on Nuclear Sci-ence, may 2013. [10] Oldham, T.R. "Ionizing radiation effects in MOS ox-ides" ,World Scientific Pub Co Inc, 1999 .[11] Holmes-Siedle, A.G. and Adams, L., "Handbook ofradiation effects, 2002, Oxford University Press, USA.[12] L.J. Asensio, M.A. Carvajal, J.A. Lopez-Villanueva, M.Vilches, A.M. Lallena, A.J. Palma, Evaluation of a low-cost commercial mosfet as radiation dosimeter, Sensorsand Actuators A: Physical, Volume 125, Issue 2, 10 Janu-ary 2006, Pages 288-295, 2006[13] Lipovetzky, J.; Redin, E.G.; Faigon, A.; "ElectricallyErasable Metal–Oxide–Semiconductor Dosimeters" ć , G. and Goluboviv, S. and Pejovi ć , M. “Sensitiv-ity and fading of pMOS dosimeters with thick gate ox-ide”, Sensors \& Actuators: A. Physical, Vol. , NO. 2-3,pp.153—158, 1996.[19] Holmes-Siedle, A. G., Goldsten, J. O., Maurer, R. H.,& Peplowski, P. N. (2014). RADFET dosimeters in thebelt: The Van Allen Probes on day 365. IEEE Transac-tions on Nuclear Science, 61(2), 948-954.[20] “LabOSat: Low cost measurement platform designedfor hazardous environments”, M. Barella, , G. Sanca, F.Gomez Marlasca, G. Rodriguez, D. Martelliti, L. Abanto,P.Levy, F. Golmar, Proceedings of the 2016 Seventh Ar-gentine Symposium and Conference on Embedded Sys-tems (CASE): ISBN: 978-987-45523-9-6 Printed IEEECATALOG: CFP1646V-PRT IEEE XPLORE: CF-P1646V-ART(2016).273