This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON NUCLEAR SCIENCE 1

Ionizing Radiation Effects on Nonvolatile MemoryProperties of Programmable Metallization Cells

J. L. Taggart, Student Member, IEEE, Y. Gonzalez-Velo, Member, IEEE, D. Mahalanabis, Student Member, IEEE,A. Mahmud, Student Member, IEEE, H. J. Barnaby, Senior Member, IEEE, M. N. Kozicki, Member, IEEE,K. E. Holbert, Senior Member, IEEE, M. Mitkova, Senior Member, IEEE, K. Wolf, Student Member, IEEE,

E. Deionno, and A. L. White

Abstract—The impact of ionizing radiation on the retention andendurance of programmable metallization cells (PMC) ReRAMcells is investigated and presented for the first time, with addi-tional work on resistance switching. This study shows thatgamma-ray exposure has a minimal effect on the retention of PMCdevices, up to a total ionizing dose (TID) of 2.8 Mrad ( ),the maximum TID level tested. The retention of both high resis-tance states (HRS) and low resistance states (LRS) during expo-sure was tested. Endurance appears to be slightly reduced withgamma-ray exposure. The endurance was tested to maximum TIDof 4.62 Mrad ( ). DC response characterizations werealso performed on PMC devices after cumulative dose exposureswith 50 MeV protons and 100 keV electrons. The data show thatPMCs are most sensitive to proton irradiation incident from thebackside of the device. For the electron exposures, it is shown thatthe LRS is mostly unaffected, but the HRS drifts to lower resis-tance values with an increase in radiation exposure.

Index Terms—CBRAM, chalcogenide, conductive bridgingRAM, endurance, ionizing radiation,memory, nano-ionicmemory,non-volatile, PMC, programmable metallization cell, ReRAM,retention, total ionizing dose.

I. INTRODUCTION

P ROGRAMMABLE metallization cells (PMCs) [1] are anovel type of nano-ionic resistive random access memory

(ReRAM) [2], [3]. They are recognized as one of the mostpromising alternatives to Flash technology for nonvolatilememory applications [4]. PMC devices also exhibit propertiesthat suggest their suitability for neuromorphic computation[5], [6]. It has been shown that the basic functionality ofmost ReRAM technologies (i.e., both cation based and anionbased) is largely unaffected by exposure to ionizing radiation,

Manuscript received July 11, 2014; revised September 03, 2014; acceptedOctober 03, 2014. This work was supported in part by the Defense Threat Re-duction Agency under Grant HDTRA1-11-1-0055, in part by the Air ForceResearch Laboratory Det 8/RVKVE under Grant FA9452-13-1-0288, in partby The Aerospace Corporation’s Independent Research and Development Pro-gram, and in part by Sandia National Laboratories.J. L. Taggart, Y. Gonzalez-Velo, D. Mahalanabis, A. Mahmud, H. J. Barnaby,

M. N. Kozicki, and K. E. Holbert are with the School of Electrical, Computerand Energy Engineering, Arizona State University, Tempe, AZ 85287-5706USA (e-mail: jennifer.taggart@asu.edu).M. Mitkova and K. Wolf are with the Department of Electrical and Computer

Engineering, Boise State University, Boise, ID 83725 USA.E. Deionno and A. L. White are with The Aerospace Corporation, Los An-

geles, CA 90009-2957 USA.Color versions of one or more of the figures in this paper are available online

at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/TNS.2014.2362126

even at high total dose levels [7]–[10]. In the case of PMCs,previous works have demonstrated that dc resistance switchingis maintained after gamma-ray exposure to total ionizingdose (TID) levels as high as 10 Mrad( ) [9], [10]. Arecent study also presented the effects of ionizing radiation ondata retention in a 128 kbit conductive bridge random accessmemory (CBRAM), a commercial variant of PMCs [11], [12],but only to a TID up to 5 Mrad(Si) [13], [14]. In this work wereport for the first time the impact of high levels of ionizingradiation ( gamma-rays) on PMC data retention and cy-cling endurance. Investigations of the dc resistance switchingresponse of PMCs are also expanded to a broader class ofionizing particles (i.e., electrons and protons).In the results and discussion section, we present data showing

the impact of gamma-rays exposure on memory retentionin the PMC’s low resistance state (LRS) and high resistancestate (HRS). The results of endurance testing on gamma-irra-diated PMCs are also reported. Finally, the resistance switchingresponse of PMCs exposed to 100 keV electrons and 50 MeVprotons are presented. The experimental results reveal that noneof the critical specifications of PMCs are drastically modifiedby ionizing radiation exposure, which further demonstrates thehigh level of radiation tolerance in this technology. Some of theradiation-induced effects observed in the endurance testing areexamined in more detail in order to understand the broader im-pact of irradiation on actual memory arrays.

II. DEVICE STRUCTURE AND OPERATION

A. Technology Overview

PMC devices are a simple stacked structure. The deviceconsists of a thin film of nickel as the bottom inert electrode(cathode), interfaced to a solid-electrolyte layer of chalcogenideglass (amorphous ) photo-doped with silver and a thinlayer of silver for the top active electrode (anode). The silvercontact is partially covered with aluminum to ensure goodcontact when probing or packaging the devices. A micrographand cutaway diagram of the layers is shown in Fig. 1. Thistechnology relies on oxidation and reduction reactions to form aconductive silver filament through the electrolyte layer [1]–[3].When a positive bias is applied to the top active electrode, thesilver is oxidized (loses electrons), creating silver cations thatmigrate through the , and reduce (gains electron) onthe inert nickel (grounded) contact. In a short amount of time( ns), a conductive filament composed of metallic bondedsilver will form to complete a low resistance connection to the

0018-9499 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON NUCLEAR SCIENCE

Fig. 1. (a) Micro photograph top view of PMC device with a circular m di-ameter via size. (b) A cross sectional diagram shown at the cut line highlighted.The cross section shows the deposited layers of Ni cathode, ChG (Ag doped

) via and Ag anode. (c) Microphotograph of crossbar structure.

silver top layer [Fig. 2(c)] [15]. The resistance of the filamentcan be further decreased by applying a bias until a specifiedcompliance current is reached. The larger the compliancecurrent, the thicker the filament [16], [17]. To dissolve thesilver filament, a negative bias is applied to the top activecontact with respect to the inert contact. Under this condition,the silver is oxidized, and migrates back to the active silvercontact, resulting in a high resistance across the device. Theprocess of filament creation and dissolution is illustrated inFigs. 2(a)–2(f). When considering memory applications, theprogrammed low resistance state (LRS) can represent one bitvalue while the high resistance state (HRS) represents another.PMC devices can be programmed by sweeping the active

contact voltage to a positive value and erased by sweeping toa negative value. The programmed LRS is controlled by lim-iting the current. The higher the compliance current ( ), thethicker the filament, resulting in a resistance lower than that ofa filament formed at a lower current compliance. A typical cur-rent-voltage (I-V) curve is shown in Fig. 3. The resistance stateof the device can be measured by applying a small voltage (lessthan the programming threshold) and measuring the current.

B. Device Fabrication

The results presented in this paper are obtained on three dif-ferent batches of PMCs, all manufactured in the NanoFab fa-cility at Arizona State University. The fabrication process startsby depositing 100 nm insulating layer of on a silicon wafer

Fig. 2. (a)–(d) The process of filament formation when a positive bias is appliedbetween the anode and cathode. (e)–(f) The process of dissolving the filament byapplying a reverse bias. (a) HRS–Positive bias applied across anode to cathode.

ions migrates through ChG. (b) HRS– ions migrate to Ni and reducesby gaining an electron from the cathode. Filament begins to form. (c) LRS–In-creasing bias past threshold, pushes enough Ag through ChG to complete theconductive filament. (d) LRS–Setting the compliance current controls the fila-ment thickness/low resistance value. (e) HRS–Appling a reverse bias past theerase threshold oxidizes the Ag in the filament and pulls the ions into theAg top layer. (f) HRS–Devices returns to HRS once the Ag fillament is dis-solved.

Fig. 3. Typical current-voltage (I-V) curve for a PMC device. The red (circlemarker) line shows the I-V characteristic for forming (setting) the conductivefilament. The black (squaremarker) line is the I-V characteristic when dissolving(erasing) the Ag filament. The section of the plot marked with star markers isthe low resistance state (LRS) whereas the horizontal section of the curve near0 A is the high resistance state (HRS). The resistance of the PMC device duringthe I-V sweep is marked by the dashed blue line.

using a TorrVac VC-320 electron-beam evaporator. A 100 nmthin-film of nickel is evaporated without braking vacuum in theTorrVac. For the devices used during retention testing, the waferis removed from the TorrVac and the nickel layer is etched toform dog bone style cathode contact bars. The batches usedduring endurance testing and resistance switching testing have

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

TAGGART et al.: IONIZING RADIATION EFFECTS ON NON-VOLATILE MEMORY PROPERTIES OF PROGRAMMABLE METALLIZATION CELLS 3

a continuous, unaltered, thin-film layer of nickel as the cathodecontact. After nickel deposition and etching, the wafer is placedback into the TorrVac and 100 nm of is deposited and thenetched to form vias to the cathode contact. The wafer is thenplaced in a Cressington 308 thermal evaporator where 60 nmof followed by 30 nm of silver film is deposited. Thewafer is removed from the Cressington and placed under a UV( nm, mJ cm ) lamp for 1 hour to photo-dope thechalcogenide layer with the 30 nm of silver. After doping, thewafer is placed back into the Cressington where an additional35 nm of silver is deposited to create the device anode. In thefinal steps, a total of 800 nm of aluminum is deposited using theTorrVac to create the contact pads for the anode and cathode. Forthe crossbar style devices used during retention testing, 400 nmof aluminum is first deposited and lifted off to create the top dogbone contact with an additional 400 nm of aluminum depositedon the contact pads at the ends of the nickel and aluminum dogbones.The results presented on the resistance switching capabilities

were obtained on the same batch of devices used in [9]. The cy-cling results were obtained on devices from a different, more re-cent batch with a postprocessing annealing performed at 80for 1 hour. The crossbar devices were annealed in air at 120for 20 minutes.

III. EXPERIMENT SETUP

A. Memory Retention After –Ray Exposures

During a retention test, the value of the programmed re-sistance state is measured periodically to investigate thenonvolatility of the memory cell. The devices were packagedopen top, with no lid. Retention testing after TID exposure wasconducted on three PMCs programmed in the LRS and fivePMCs erased into the HRS. The four devices were programmedinto a LRS by sweeping the voltage at the anode of the PMCfrom 0 V to 0.5 V and limiting the maximum current to a Acompliance, as shown in the “Set” curve in Fig. 3. The fivedevices in HRS were erased by sweeping the voltage across theanode to cathode from 0 V to 0.5 V as shown in the “Erase”curve in Fig. 3. Two control devices were programmed to aLRS limited with a A compliance and were not exposed togamma radiation. Another two devices were used as the controlfor the HRS and were also not exposed to gamma-ray radiation.All tested devices were 5 m in diameter.The devices were exposed with all pins left floating in a Gam-

macell 220 at a dose rate of 478.5 rad . Theretention measurements were conducted after each dose step,using an Agilent 4156C semiconductor parameter analyzer con-nected, via low noise triaxial cables, to an Agilent 16442B testfixture with a 28 pin DIP socket fixture. The resistance state wasmeasured by sampling the current when 10mVwas applied con-tinuously from anode to cathode.

B. Cycling Endurance After –Ray Exposures

In a cycling or endurance test, the ability of a memory cell tobe set to the LRS and reset to the HRS repeatedly is investigated.This cycling test is performed by applying a train of voltagepulses to set and reset the PMC cell. The setup used to conductthe test is described in Fig. 4. A positive voltage pulse applied

Fig. 4. Endurance test set-up of the PMC. The input and output voltage signalswere monitored with an oscilloscope.

Fig. 5. Typical waveforms applied and measured: input signal from the AWGon the bottom, output voltage on top. The inset presents a closer view to thereset observed on the output signal. The Set pulse used is ( V ms), thereset is ( V ms), and the read pulse is ( mV ms).

on the anode of the device sets it to the LRS, whereas a negativevoltage pulse resets it to the HRS (Fig. 5). In this method, thedevice resistance is manipulated by applying a bias for a fixedamount of time. The longer the bias is applied, the lower theresistance will become as the filament broadens. For the 10 mspulsewidth used in this experiment, the LRS is approximately

(shown in Fig. 7), which corresponds to a compliancecurrent of 1 mA when using a compliance current method forprogramming [17].After the write and erase pulses, a small amplitude pulse is

applied to sense the resistance value of the PMC. An arbitrarywaveform generator (Tektronix AWG 520) was used to generatethe pulse train of write/read/erase/read signals. An oscilloscope(Agilent 54832D MSO) was used to retrieve the signals to ob-serve if the cells were set or reset, and to compute the resistancevalues after a given number of cycles. This is the first time thatthis type of characterization has been performed on irradiatedPMCs.Several tiles of bare devices were initially placed in the Gam-

macell 220 with contacts left floating. At periodic TID, a tileof devices was removed, leaving the remaining devices to beexposed to higher TID. The removed devices were cycled asdescribed above, until the HRS began to decrease. The devicesexposed to gamma-ray radiation were compared to the behaviorof control devices not exposed to radiation.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON NUCLEAR SCIENCE

Fig. 6. Retention of PMC devices initially programmed to a LRS withA, and devices erased into a HRS then exposed to an accumulative TID of

2.8 Mrad ( ). The control (Ctrl) devices were not exposed to radiation.

C. DC Resistance Switching After Electron and ProtonExposures

Exposure to 100 keV electrons was conducted on PMCs de-vices at The Aerospace Corporation. Devices were step stressedto increasing fluence (i.e., increasing dose) then swept from0.5 V to 0.5 V and back to 0.5 V at the end of each exposure

step to retrieve the resistance switching characteristics and ob-tain the typical and values [9]. The samples’ pins wereleft floating during exposure. A total of 24 devices with a diam-eter of 5 m were characterized. Those devices were divided inthree sets of samples switched/programmed with different pro-gramming currents ( A, A, A). For each

, six PMCs have been exposed to electrons and character-ized and two PMCs have been characterized as control parts.50 MeV proton exposures were conducted at the UCB-LBNLcyclotron, with PMCs exposed from the front side and from thebackside. Only one fluence of cm was used duringthe proton exposures. Similar to the electron tests, different setsof PMCs have been characterized before and after exposure with

A, A and A.

IV. RESULTS AND DISCUSSION

A. Memory Retention After –Ray Exposures

Results obtained on exposed (as a function of time and TID)and control devices (as a function of time) are presented inFig. 6. Control measurements were performed on four devices,two in the LRS and two in the HRS. For the exposed devices,five devices were erased into the HRS and four devices wereprogrammed to a LRS using A. It is shown that de-vices in both the LRS and HRS maintained state for TID up to2.8Mrad( ) (the maximum TID tested on the presentedparts).The hashed area in Fig. 6 defines the HRS. The HRS of indi-

vidual device varied between M to M . The thresholdof M is the median HRS of the 13 devices tested before ex-posure to radiation. The trends shown in Fig. 6 are themean stateof the devices tested for each condition. The whiskers show the

Fig. 7. Endurance of PMC exposed to Co-60 gamma-rays. HRS and LRS re-sistance as a function of the Write/Erase cycle number for a control PMC and aPMC exposed to 4.62 Mrad. Failure is defined when .

minimum and maximum range of values. As shown in Fig. 6,no abrupt state change (LRS to HRS or HRS to LRS) occurreddue to radiation exposure. The LRS of the exposed devices in-creased within the first 24 hours of the test but this behavior wasalso seen in the control devices. Exposure to does not ap-pear to have a significant effect on the retention behavior of theLRS. The HRS of the exposed devices remained stable duringthe entire test. Over time, the HRS control devices drifted tohigher resistances.One control device programed to the HRS and one pro-

grammed to the LRS, suddenly went low, shown by the longwhiskers around 4000 minutes in Fig. 6. This behavior has beenseen several times during the testing of these devices. Each timea control measurement was taken, the packaged tile was placedin the test fixture for the duration of the measurement, thenpromptly removed and placed in a light-tight box. The act ofplacing the package in and removing the package from the testfixture, may have resulted in a static discharge that caused twoof the control devices to program to a lower resistance state.The reason for this test procedure was to mimic the testingprocedure of the irradiated devices. After each dose step, theirradiated packages were removed from the Gammacell andplaced in the test fixture to obtain measurements. The packagewas then removed from the test fixture and placed back in theGammacell for the next dose.

B. Cycling Endurance After –Ray Exposures

Representative endurance plots (where the HRS and LRS re-sistance values are plotted as a function of the cycle number)for devices exposed to incremental TID are presented in Fig. 7.The trend lines for each total dose and control in Fig. 7 is themean of three devices for a total of 12 devices tested.For the control devices tested in this work, the maximum

number of cycles achieved was generally betweento cycles. As shown in Fig. 7, the control begins toexhibit a decrease in the HRS resistance after cycles. After

cycles, the ratio between the HRS and LRS of the con-trol reduced below 100, constituting a failure for this scenario.Irradiated devices were cycled up until cycles, the valueat which the control devices began to deteriorate.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

TAGGART et al.: IONIZING RADIATION EFFECTS ON NON-VOLATILE MEMORY PROPERTIES OF PROGRAMMABLE METALLIZATION CELLS 5

Fig. 8. Cumulative distributions of the resistance values obtained onPMCs after increasing exposure to 100 keV electrons, for TID up to12 Mrad( ): (a) of m devices programmed at A; (b)of m devices; (c) comparison between (black symbols), and (redsymbols with line). Resistance values are extracted on the dc I-V resistanceswitching characteristics at a voltage of 50 mV.

C. DC Resistance Switching After Electron and ProtonExposures

Fig. 8 shows the cumulative distributions of andswitched with an A for 100 keV electron expo-sure. Similar results were observed for devices programmedwith A and A. It is shown that doesnot vary with TID [Fig. 8(a)], whereas is observed to de-crease with exposure for TID higher than 100 krad ( )[Fig. 8(b)]. This decrease in is also observed on devicescharacterized with all three programming current compliancelevels. The five LRS data points that overlap into the HRS are

Fig. 9. Cumulative probabilities of and after 50 MeV proton frontside (Ag anode end) exposure. and values are obtained at 50 mV onthe dc resistance switching characteristic.

due to a device that switched to a high resistance during its pro-gramming sweep. This behavior may be due to a weak point inthe formed filament. It is clear from Fig. 8(c) that the majoritydistribution of HRS resistance (red symbols) and LRS re-sistance (black symbols) do not overlap even after exposureto electrons up to 12Mrad( ), further demonstrating theradiation hardness in the switching characteristics of PMCs.In Fig. 9 and Fig. 10, cumulative distributions obtained on

devices exposed to 50 MeV protons incident to the front sideand back side of the devices are presented. The devices wereswitched using a dc sweep with an of A [Fig. 9,Fig. 10(a)] and of A [Fig. 10(b)]. For the protonexposures, a decrease in was observed for both back andfront side exposed devices. A slight decrease of the is alsoobserved for the samples exposed from the back and switchedwith the higher programming currents ( A). The shift in

is seen to be more significant due to back side irradiationthan from front side. When exposed from the back side, pro-tons travel through m of Si, 100 nm of , and 100 nmof Ni before entering the active region. Nuclear in-teractions in the silicon substrate layer may induce cascades ofrecoiled nuclei that result in displacement damage and ioniza-tion in the device layers [18]. During front side irradiation, theproton is directly incident on the active layers of the PMC de-vice which consists of 35 nm of Ag, and 60 nm of .The 50 MeV protons pass through the active layers with littleinteraction during front side irradiation.

V. CONCLUSION

The impact of total ionizing dose on the retention and en-durance of PMC memory cells is investigated for the first timein this work. PMCs are used as memory cells in novel commer-cial NVM circuits (e.g., CBRAM technology), and the effectsstudied on PMCs could enable a better understanding of the im-pact of radiation on such memory circuits and any other applica-tions based on PMC or CBRAM technology. It has previouslybeen shown that for metal-oxide based ReRAM cells exposed togamma-rays and x-rays, the most sensitive state is the HRS [8].In this work, it is shown that gamma ray exposure levels upto 2.8 Mrad ( ) have little impact on the LRS and HRSretention of these devices. Similar to what has been observed onmetal-oxide based ReRAM, the LRS is maintained after several

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON NUCLEAR SCIENCE

Fig. 10. Cumulative distribution of and after 50 MeV proton backside (Ni cathode end) exposure. and values are obtained at 50 mV onthe dc resistance switching characteristic. (a) Programming current of A.(b) Programming current of A.

Mrad of total dose. This is a result that expands upon recentstudies [13], [14] of retention conducted directly on a commer-cial based CBRAM memory circuit (128 kbit EEPROMtype circuit), where it was found that for TIDs up to 5 Mrad noerrors appeared on the data stored in the memory array. The en-durance of PMCs was shown here to be impacted by TID with adecrease in the total number of set/reset cycles that can be per-formed on the cells after exposure. Nevertheless, no decrease indynamic range was observed on the exposed cells (the HRS andLRSmaintain the same resistance levels and the sameratio) for cycles below . The resistance switching of PMCsduring 100 keV electron exposures and 50 MeV proton expo-sures for high levels of TID is also presented in this work and ex-pands upon previous studies conducted with gamma–rays[9]. It has been shown that 100 keV electron irradiation reducesthe HRS with increasing dose. When exposed to 50 MeV pro-tons, no effect is seen during front side irradiation but the HRSdecreases after back side exposure. This shift is most likely dueto recoiled particles from the substrate layers that induces ion-ization in the active device layers. The results obtained duringthese studies confirms the ability to maintain functionality, theability to program the PMC devices to a HRS or LRS state, afterexposure to various radiation environments.

ACKNOWLEDGMENT

The authors would like to thank Dr. James Reed of DTRA,Dr. Arthur Edwards of AFRL, and Matthew Marinella andMichael McLain of Sandia National Laboratories for theirsupport of this work.

REFERENCES[1] M. N. Kozicki, M. Park, and M. Mitkova, “Nanoscale memory ele-

ments based on solid-state electrolytes,” IEEE Trans. Nanotechnol.,vol. 4, pp. 331–338, May 2005.

[2] I. Valov andM. N. Kozicki, “Cation-based resistance change memory,”J. Phys. D, Appl. Phys., vol. 46, no. 7, p. 074005, Feb. 2013.

[3] I. Valov, R. Waser, J. R. Jameson, and M. N. Kozicki, “Electrochem-ical metallization memories—fundamentals, applications, prospects,”Nanotechnology, vol. 22, no. 25, p. 254003, Jun. 2011.

[4] ITRS Workshop, “Assessment of the potential and maturity ofselected emerging research memory technologies,” [Online]. Avail-able: http://www.itrs.net/Links/2010ITRS/2010Update/ToPost/ERD_ERM_2201FINALReportMemoryAssessment_ITRS.pdf

[5] B. Swaroop, W. C. West, G. Martinez, M. N. Kozicki, and L. A. Akers,“Programmable current mode Hebbian learning neural network usingprogrammable metallization cell,” in Proc. IEEE Int. Symp. CircuitsSyst., 1998, vol. 3, pp. 33–36.

[6] S. Yu and H.-S. P. Wong, “Modeling the switching dynamics of pro-grammable-metallization-cell (PMC) memory and its application assynapse device for a neuromorphic computation system,” IEDM Tech.Dig, pp. 22.1.1–22.1.4., 2010.

[7] Y. Wang, H. Lv, W. Wang, Q. Liu, S. Long, Q. Wang, Z. Huo, S.Zhang, Y. Li, Q. Zuo, W. Lian, J. Yang, and M. Liu, “Highly stable ra-diation-hardened resistive-switching memory,” IEEE Electron DeviceLett., vol. 31, no. 12, pp. 1470–1472, Dec. 2010.

[8] M. J. Marinella, S. M. Dalton, P. R. Mickel, P. E. Dodd, M. R.Shaneyfelt, E. Bielejec, G. Vizkelethy, and P. G. Kotula, “Initialassessment of the effects of radiation on the electrical characteristicsof TaOx memristive memories,” IEEE Trans. Nucl. Sci., vol. 59, no.6, pp. 2987–2994, Dec. 2012.

[9] Y. Gonzalez-Velo, H. J. Barnaby, M. N. Kozicki, P. Dandamudi, A.Chandran, K. E. Holbert, M. Mitkova, and M. Ailavajhala, “Total-ion-izing dose effects on the resistance switching characteristics of chalco-genide programmable metallization cells,” IEEE Trans. Nucl. Sci., vol.60, no. 6, pp. 4563–4569, Dec. 2013.

[10] P. Dandamudi, M. N. Kozicki, H. J. Barnaby, Y. Gonzalez-Velo, andK. E. Holbert, “Total ionizing dose tolerance of Ag-Ge40S60 basedprogrammable metallization cells,” IEEE Trans. Nucl. Sci., vol. 61, no.4, pp. 1726–1731, Aug. 2014.

[11] “CBRAM is a registered trademark of Adesto Technologies Corpora-tion,” [Online]. Available: http://www.adestotech.com/CBRAM

[12] C. Gopalan, Y. Ma, T. Gallo, J. Wang, E. Runnion, J. Saenz, F.Koushan, P. Blanchard, and S. Hollmer, “Demonstration of con-ductive bridging random access memory (CBRAM) in logic CMOSprocess,” Solid-State Electron., vol. 58, no. 1, pp. 54–61, Apr. 2011.

[13] Y. Gonzalez-Velo, H. J. Barnaby, M. N. Kozicki, C. Gopalan, andK. E. Holbert, “Total ionizing dose retention capability of conductivebridging random access memory,” IEEE Electron Device Lett., vol. 35,no. 2, Feb. 2014.

[14] Adesto Technologies, “Gamma radiation tolerant CBRAM tech-nology,” 2013 [Online]. Available: http://www.adestotech.com/sites/default/files/documents/CS_CBRAM_001.pdf

[15] M. N. Kozicki, M. Balakrishnan, C. Gopalan, C. Ratnakumar, andM. Mitkova, “Programmable metallization cell memory based onAg-Ge-S and Cu-Ge-S solid electrolytes,” in Proc. NVMTS, 2005, pp.83–89.

[16] D. Kamalanathan, U. Russo, D. Ielmini, and M. N. Kozicki, “Voltage-driven on–off transition and tradeoff with program and erase currentin programmable metallization cell (PMC) memory,” IEEE ElectronDevice Lett., vol. 30, no. 5, pp. 553–555, Apr. 2009.

[17] M. N. Kozicki, C. Gopalan, M. Balakrishnan, M. Park, and M.Mitkova, “Non-volatile memory based on solid electrolytes,” in Proc.NVMTS, 2004, pp. 10–17.

[18] D. M. Hiemstra and E. W. Blackmore, “LET spectra of proton energylevels from 50 to 500 MeV and their effectiveness for single eventeffects characterization of microelectronics,” IEEE Trans. Nucl. Sci.,vol. 50, no. 6, pp. 2245–2250, Dec. 2003.