Impact of Gamma Radiation on Dynamic RDSON Characteristics in AlGaN/GaN Power HEMTs

ma e ials A icle Impac o Gamma Radia ion on Dynamic RDSON Cha ac e is ics in AlGaN/GaN Powe HEMTs Ped o J. Ma ínez 1, En ique Mase 1,* , Ped o Ma ín-Holgado 2, Yolanda Mo illa 2, Da id Gilabe 1and Es eban Sanchis-Kilde s 1 1Depa men Elec onic Enginee ing, Uni e si y o Valencia, 46100 Bu jasso , Spain 2Cen o Nacional de Acele ado es (CNA), Uni e sidad de Se illa, CSIC, JA, 41092 Se illa, Spain *Co espondence: emase @u .es Recei ed: 30 July 2019; Accep ed: 27 Augus 2019; Published: 28 Augus 2019   Abs ac : GaN high-elec on-mobili y ansis o s (HEMTs) a e p omising nex -gene a ion de ices in he powe elec onics ield which can coexis wi h silicon semiconduc o s, mainly in some adia ion-in ensi e en i onmen s, such as powe space con e e s, whe e high equencies and ol ages a e also needed. I s wide band gap (WBG), la ge b eakdown elec ic ield, and he mal s abili y imp o e ac ual silicon pe o mances. Howe e , a he momen , GaN HEMT echnology su e s om some eliabili y issues, one o he mo e ele an o which is he dynamic on-s a e esis ance (R ON_dyn ) ega ding powe swi ching con e e applica ions. In his s udy, we ocused on he d ain- o-sou ce on- esis ance (R DSON ) cha ac e is ics unde 60 Co gamma adia ion o wo di e en comme cial powe GaN HEMT s uc u es. Di e en bias condi ions we e applied o bo h s uc u es du ing i adia ion and some s a ic measu emen s, such as h eshold ol age and leakage cu en s, we e pe o med. Addi ionally, dynamic esis ance was measu ed o ob ain p ac ical in o ma ion abou de ice apping unde adia ion du ing swi ching mode, and how apping in he de ice is a ec ed by gamma adia ion. The expe imen al esul s showed a high dependence on he HEMT s uc u e and he bias condi ion applied du ing i adia ion. Speci ically, a ee cu en collapse s uc u e showed g ea s abili y un il 3.7 M ad(Si), unlike he o he s uc u e es ed, which showed high deg ada ion o he pa ame e s measu ed. The changes we e demons a ed o be due o apping e ec s gene a ed o enhanced by gamma adia ion. These new esul s ob ained abou R ON_dyn will help elucida e ap beha io s in swi ching ansis o s. Keywo ds: high-elec on-mobili y ansis o (HEMT); gallium ni ide (GaN); adia ion ha dness; assu ance es ing; adia ion e ec s; o al ionizing dose (TID) 1. In oduc ion Gallium ni ide (GaN) is a p omising ma e ial o nex -gene a ion powe de ices due o i s wide band gap, which allows a la ge b eakdown elec ic ield and he possibili y o ope a ing unde ha sh en i onmen al condi ions [ 1 – 3 ]. Such cha ac e is ics make hese de ices p omising o space applica ions, whe e empe a u e and adia ion a e key ac o s. Pa icula ly, he de elopmen o new GaN high-elec on-mobili y ansis o s (HEMTs) wi h g ea cha ac e is ics, such as low on- esis ance and pa asi ic capaci ances, allow hem o swi ch a highe equencies wi h high e iciency, making hem a ac i e. The inhe en adia ion ha dness, he capabili y o wi hs and highe b eakdown ol ages, and he highe ope a ing empe a u es will enable his echnology’s use in u u e space applica ions, such as elecommunica ions, Ea h obse a ion, and science missions [4,5]. These p omising ad an ages ha e pushed he esea ch ocus on he eliabili y o GaN de ices unde adia ion condi ions. In space en i onmen s, ene ge ic pa icles which impac semiconduc o de ices lose hei ene gy o ionizing and nonionizing p ocesses while hey a el h ough he de ices. Ma e ials 2019,12, 2760; doi:10.3390/ma12172760 www.mdpi.com/jou nal/ma e ials Ma e ials 2019,12, 2760 2 o 15 The ene gy loss causes he p oduc ion o elec on–hole pai s (ioniza ion) and displaced a oms (displacemen damage). Gamma i adia ion is one o he es s used o e alua e he ha dness o de ices o be used in ae ospace applica ions. Ou main objec i e he e was o s udy he deg ada ion induced by he o al ionizing dose (TID) e ec s o 60 Co γ - ay adia ion on GaN HEMTs. The esponse o gamma i adia ion is complex. Comp on elec ons induced by γ - adia ion c ea e elec on–hole pai s, hus changing he occupancy o aps. Rega ding his opic, many esea ch pape s ha e been published in ecen yea s showing di e en beha io s depending on he dose applied and he s uc u e o he HEMT being i adia ed [ 6 ]. In gene al, HEMTs i adia ed wi h gamma ays exhibi a nega i e h eshold ol age and a ansconduc ance dec ease, which can be explained by he c ea ion o ap s a es h oughou he s uc u e and, in some cases, an inc ease in he wo-dimensional elec on gas (2DEG) shee concen a ion [ 7 , 8 ]. Some au ho s ha e epo ed s ain elaxa ion a low-dose gamma i adia ion, which enhanced he channel mobili y [ 9 , 10 ]. In con as , o he au ho s [ 11 ] ha e epo ed a educ ion o 60% o he d ain cu en a a ound 70 k ad(Si). Thus, he de ec s gene a ed by he γ -i adia ion a e e y sensi i e o he s uc u e, ha ing de ec c ea ion a es dependen on he quali y o he sample and he doping le el. On he o he hand, i is well known ha ano he p oblem ela ed o he GaN HEMT s uc u e is he elec on apping e ec s which dec ease de ice pe o mance [ 12 ], in pa icula , he dynamic on- esis ance (R ON_dyn ) [ 13 ]. This apping educes he cu en ha he de ices can d i e below he de ice’s a ed cu en and could be a ibu ed o a apping e ec in di e en egions inside he de ice. In [ 14 ], i was epo ed ha apping could be a ibu ed o he de ice su ace and bu e laye and ha i was possible o dis inguish be ween bo h. These apping e ec s ha e been shown o be induced by he elec ical ield applied be ween he d ain and he ga e. Addi ionally, i was con i med ha he HEMT s uc u e design is a key ac o ega ding he R ON_dyn . Di e en s a egies can be used o mi iga e he inc ease o R ON_dyn , mainly, he use o a p-GaN egion close o he d ain ha is elec ically connec ed o he d ain edge [15] and he op imiza ion o he de ice bu e laye design [16]. Taking in o accoun he p oblems ega ding he e ec s o γ - adia ion and R ON_dyn , i is necessa y o e alua e he e ec o γ - adia ion on R ON_dyn i GaN HEMTs a e o be se iously conside ed o u u e space applica ions. In ac , in e na ional space agencies, such as he Na ional Ae onau ics and Space Adminis a ion (NASA) and he Eu opean Space Agency (ESA), ha e shown inc easing in e es in e alua ing he de elopmen o GaN de ices. Some es ing ac i i ies ha e been included as pa o a u u e adia ion quali ica ion p ocess o wide band gap de ices, as p esen ed in [17]. Analyzing he ele an li e a u e shows ha some con adic o y esul s ha e been epo ed and he easons o hese di e ences a e no e iden . Fu he mo e, a p e ious s udy epo ed changes in R ON_dyn [ 18 ] wi h low-dose gamma adia ion. The e o e, in his s udy, we in es iga ed no only he s a ic cha ac e is ics bu also he R ON_dyn beha io o wo di e en HEMT s uc u es subjec ed o gamma adia ion un il 3.7 M ad(Si). New esul s on his subjec a e shown in his wo k ela ed o he HEMT s uc u e and he biasing condi ions applied du ing i adia ion. These esul s show ha , depending on he s uc u e and he bias applied, apping can be inc eased (case o nega i e ga e bias) o educed (case o sho ed de ices), which p o ides mo e in o ma ion han p e ious s udies, whe e only s a ic cha ac e iza ion had been done. 2. Ma e ials and Me hods No mally-o comme cial AlGaN/GaN HEMTs on Si wi h a ol age a ing o 600 V we e used o he adia ion expe imen . We chose wo di e en HEMT s uc u es: One was a 600 V p-doped GaN ga e injec ion ansis o (HD-GIT) PGA26E07BA, manu ac u ed by Panasonic, Inc. (Kadoma-shi, Osaka, Japan). The HD-GIT has an addi ional p-GaN egion be ween he ga e and he d ain, in which he holes a e injec ed du ing he o -s a e, compensa ing he elec on apping and a oiding R ON_dyn . The o he was he GS66516T, a 650 V p-GaN ga e wi h a me al insula o laye (MISHEMT), manu ac u ed by GaN Sys ems, Inc. (O awa, ON, Canada). Table 1summa izes he key pa ame e s o he in es iga ed de ices. Ma e ials 2019,12, 2760 3 o 15 Table 1. Pa ame e s o he in es iga ed de ices. Pa ame e Symbol GaN MISHEMT GS66516T p-GaN HEMT PGA26E07BA D ain- o-sou ce b eakdown ol age BVDSS 650 V 600 V Con inuous d ain cu en (Tc =25 ◦C) ID60 A 26 A D ain- o-sou ce on- esis ance (Tj =25 ◦C) RDSON 25 mΩ56 mΩ Inpu capaci ance (1 MHz, 400 V) Ciss 520 pF 405 pF To al ga e cha ge QG12.1 nC 5 nC Fo he TID adia ion expe imen , 12 de ices om each HEMT s uc u e we e used. As shown in Table 2, di e en bias condi ions we e applied du ing i adia ion o e alua e he bias dependence o he TID esponse o he HEMTs. In addi ion, one sample o each de ice ype was selec ed as a e e ence (con ol de ice) o con i m he p ope ope a ion o he measu emen sys em; ha is, hese uni adia ed de ices we e subjec ed only o elec ical measu emen s a e each s ep, wi hou any bias o adia ion condi ions applied. Table 2. Bias condi ions applied du ing he i adia ion es . Condi ion De ice Type Sample Se ial Numbe Uni s Radia ion Ga e Bias 1D ain Bias 1 Con ol MISHEMT Q0 1 No N/A N/A p-GaN HEMT Q25 1 No N/A N/A Sho ed MISHEMT Q11, Q12 2 Yes 0 V 0 V p-GaN HEMT Q23, Q24 2 Yes 0 V 0 V D ain bias MISHEMT Q6–Q10, Q21, Q22 7 Yes 0 V 400 V p-GaN HEMT Q18–Q20 3 Yes 0 V 400 V D ain–ga e bias MISHEMT Q1–Q5 5 Yes −5 V 400 V p-GaN HEMT Q13–Q17 5 Yes −5 V 400 V (1) Bias condi ion du ing he i adia ion exposu e (No Applicable—N/A— o uni adia ed con ol samples). The es campaign was ca ied ou in he CNA-RadLab acili y a he Na ional Cen e o Accele a o s in Se ille, Spain. The gamma i adia ion con ained a 60 Co gamma sou ce wi h associa ed pho on ene gies o 1.17 and 1.33 MeV (mean alue: 1.25 MeV). The selec ed dose a e was 23.742 k ad(Si)/h, which is wi hin he “s anda d a e” window (0.36–180 k ad(Si)/h) o he ESA, acco ding o he TID Tes Me hod [ 19 ]. The dose a e was ob ained by measu ing he cha ge wi h wo TM30013 ioniza ion chambe s (PTW-FREIBURG, Ge many) and one mul ichannel elec ome e , Mul iDOS (PTW-FREIBURG, Ge many), and also conside ing he en i onmen al co ec ion ac o . The dose a e uni o mi y in he il e box was 98.5%. The de ices unde es (DUTs) we e moun ed on a p in ed ci cui boa d which was placed in o a 12 × 17 cm il e box o be subjec ed o adia ion, in compliance wi h he Eu opean Space Componen s Coo dina ion Basic Speci ica ion No. 22900 [ 19 ]. This con aine had 2 mm o aluminum and 1.5 mm o lead in he ou e laye and a 5 mm on co e o polyme hyl me hac yla e (PMMA) o achie e he cha ged-pa icle equilib ium. Six i adia ion s eps we e ca ied ou du ing he campaign. Pos -i adia ion elec ical measu emen s we e pe o med a e each exposu e s ep o all he de ices, including he con ol de ices. A he end o o al i adia ion, wo annealing s eps we e implemen ed. The i s s ep consis ed o oom- empe a u e annealing unde bias o 24 h. A e wa ds, accele a ed aging was ca ied ou , Ma e ials 2019,12, 2760 4 o 15 whe e he de ices we e baked a 100 ± 5 ◦ C unde bias o 168 h. In bo h annealing s eps he bias ol age applied on each DUT was he same as ha du ing he i adia ion s eps. Conce ning he measu emen s, wo ypes we e pe o med: I–V measu emen s we e done wi h a Keysigh Powe De ice Analyze B1505A (San a Rosa, CA, USA), and o he R ON_dyn measu emen s, a cus om ci cui (Figu e 1) was implemen ed [20]. Ma e ials 2019, 12, x FOR PEER REVIEW 4 o 15 Conce ning he measu emen s, wo ypes we e pe o med: I–V measu emen s we e done wi h a Keysigh Powe De ice Analyze B1505A (San a Rosa, CA, USA), and o he RON_dyn measu emen s, a cus om ci cui (Figu e 1) was implemen ed [20]. The implemen ed swi ching ci cui had he bene i o ully con olling he ime ha he ol age s ess was applied o he GaN HEMT. Basically, i consis ed o wo ansis o s connec ed in se ies be ween he d ain and he sou ce, wi h a esis i e load be ween bo h o hem. T ansis o Q1 was used o con ol he s ess/ apping ime. A esis i e load Rload was used o se he cu en le el when DUT was in on-s a e. Due o he inhe en pa asi ic induc ance (Lp) o he powe Rload, wo SiC diodes (C4D05120)—D1 and D2—o e ed a eewheeling pa h o he cu en when ei he Q1 o DUT was swi ched om on o o . Figu e 1. Ci cui diag am used o measu e RON_dyn. Fo he ansis o Q1, a SiC MOSFET (C3M0065090) (Wol speed, Du ham, NC, USA) was used in o de o ha e a low ou pu capaci ance (CDS) and low cu en peaks due he cha ge and discha ge o his pa asi ic capaci ance. The alues o he d ain- o-sou ce on- esis ance (RDSON) o he DUT we e ob ained by measu ing he de ice on-s a e ol age (VDSON) ac oss i and di iding by he cu en (IDS) h ough he DUT. The d ain cu en was measu ed using a coaxial shun esis o o 98 mΩ (SDN-414- 10), and o he VDS, a 300 V and 500 MHz passi e ol age p obe was selec ed. Due o he high ol age applied o he DUT, he ol age ac oss i ep esen s a la ge dynamic ange inpu signal o he oscilloscope inpu ampli ie acquisi ions, which can be o e loaded, and as a esul , an accu a e de e mina ion o he on-s a e ol age would no be achie able. To a oid ha p oblem, a ol age clamp ci cui oge he wi h he passi e ol age p obe was used. In pa icula , he ol age clamp used was he comme cial clp1500V15A1 om Sp ingbu o GmbH Emmendingen, Baden-Wu embe g, Ge many). The low ange (2 V) was selec ed in he ol age clippe in o de o ha e a as e esponse—200 ns in his case—conside ing he passi e ol age p obe and he ol age clippe . P ecise equency esponse compensa ion was done in he passi e ol age p obe o make up o he whole chain o he clippe and he ol age p obe. To con ol he “on- ime” o he DUT, a gene ic MOSFET isola ed d i e SI8271BB (Silicon Labs, Aus in, TX, USA) was used. This d i e was selec ed due o he minimum supply ol age needed o 3 V. This low ga e ol age was equi ed o d i e GaN de ices wi h VG = 4 V, which allowed us o see any change in he apping cha ges when measu ing he RON_dyn. This is because, a a highe ga e ol age, he 2DEG densi y a he AlGaN/GaN in e ace is highe , so he de ice is able o d i e low d ain cu en s wi hou being a ec ed by he apped cha ges in he su ace o he bu e . O he wise, i we had used a lowe ga e ol age alue, he densi y o he 2DEG would dec ease and we could see any changes in he RON_dyn due o he apping, e en a low cu en s. Thus, in bo h cases, he apped cha ges a e p esen ; howe e , in he case o he highe ga e ol age, i would equi e highe cu en s o see he e ec by measu ing he RON_dyn. Thus, ins ead o inc easing he cu en , which could induce o he p oblems such as sel -hea ing, which would change he dynamic esponse, we chose o use a lowe ga e ol age o see any changes in he de ice apping. 3. Expe imen al Resul s Figu e 1. Ci cui diag am used o measu e RON_dyn. The implemen ed swi ching ci cui had he bene i o ully con olling he ime ha he ol age s ess was applied o he GaN HEMT. Basically, i consis ed o wo ansis o s connec ed in se ies be ween he d ain and he sou ce, wi h a esis i e load be ween bo h o hem. T ansis o Q1 was used o con ol he s ess/ apping ime. A esis i e load Rload was used o se he cu en le el when DUT was in on-s a e. Due o he inhe en pa asi ic induc ance (L p ) o he powe R load , wo SiC diodes (C4D05120)—D1 and D2—o e ed a eewheeling pa h o he cu en when ei he Q1 o DUT was swi ched om on o o . Fo he ansis o Q1, a SiC MOSFET (C3M0065090) (Wol speed, Du ham, NC, USA) was used in o de o ha e a low ou pu capaci ance (C DS ) and low cu en peaks due he cha ge and discha ge o his pa asi ic capaci ance. The alues o he d ain- o-sou ce on- esis ance (R DSON ) o he DUT we e ob ained by measu ing he de ice on-s a e ol age (V DSON ) ac oss i and di iding by he cu en (I DS ) h ough he DUT. The d ain cu en was measu ed using a coaxial shun esis o o 98 m Ω (SDN-414-10), and o he VDS, a 300 V and 500 MHz passi e ol age p obe was selec ed. Due o he high ol age applied o he DUT, he ol age ac oss i ep esen s a la ge dynamic ange inpu signal o he oscilloscope inpu ampli ie acquisi ions, which can be o e loaded, and as a esul , an accu a e de e mina ion o he on-s a e ol age would no be achie able. To a oid ha p oblem, a ol age clamp ci cui oge he wi h he passi e ol age p obe was used. In pa icula , he ol age clamp used was he comme cial clp1500V15A1 om Sp ingbu o GmbH Emmendingen, Baden-Wu embe g, Ge many). The low ange (2 V) was selec ed in he ol age clippe in o de o ha e a as e esponse—200 ns in his case—conside ing he passi e ol age p obe and he ol age clippe . P ecise equency esponse compensa ion was done in he passi e ol age p obe o make up o he whole chain o he clippe and he ol age p obe. To con ol he “on- ime” o he DUT, a gene ic MOSFET isola ed d i e SI8271BB (Silicon Labs, Aus in, TX, USA) was used. This d i e was selec ed due o he minimum supply ol age needed o 3 V. This low ga e ol age was equi ed o d i e GaN de ices wi h VG=4 V, which allowed us o see any change in he apping cha ges when measu ing he R ON_dyn . This is because, a a highe ga e ol age, he 2DEG densi y a he AlGaN/GaN in e ace is highe , so he de ice is able o d i e low d ain cu en s wi hou being a ec ed by he apped cha ges in he su ace o he bu e . O he wise, i we had used a lowe ga e ol age alue, he densi y o he 2DEG would dec ease and we could see any changes in he R ON_dyn due o he apping, e en a low cu en s. Thus, in bo h cases, he apped cha ges a e p esen ; howe e , in he case o he highe ga e ol age, i would equi e highe cu en s o see he e ec by measu ing he R ON_dyn . Thus, ins ead o inc easing he cu en , which could induce Ma e ials 2019,12, 2760 5 o 15 o he p oblems such as sel -hea ing, which would change he dynamic esponse, we chose o use a lowe ga e ol age o see any changes in he de ice apping. 3. Expe imen al Resul s The e ec o gamma adia ion on bo h es ed s uc u es was di e en . In his sec ion, we analyze he changes o he s a ic and dynamic cha ac e is ics conside ing he e ec s o applied bias condi ions due o he po en ial inc ease o he concen a ion o ac i a ed GaN de ec s [21]. 3.1. T ans e Cha ac e is ics The main elec ical measu emen s add ess he d ain and ga e cu en s, which a e exp essed as a unc ion o he ga e and d ain ol ages. F om he I DS –V GS cha ac e is ics, di e en pa ame e s can be es ablished o p o ide in o ma ion abou he changes su e ed by he GaN HEMTs depending on he adia ion applied. The i s pa ame e analyzed was he h eshold ol age (V TH ), which was ex ac ed by he cu en ex apola ion me hod, whe e he h eshold ol age alue is he V GS alue co esponding o I DS =100 mA wi h a V DS =0.1 V. The h eshold ol age is usually measu ed a lowe cu en s; howe e , ou se up and he limi s de ined o he es (un il 3 A) o ced us o use a highe cu en alue (100 mA) o ha e be e p ecision o h eshold ol age measu emen . Figu e 2shows he a ia ions o he h eshold ol age o all he de ices. I is impo an o highligh he way he esul s a e plo ed, since each de ice had a di e en V TH alue. To p o ide a compa ison, i was necessa y o no malize his alue, di iding he V TH o each de ice a each s ep be ween he co esponding ini ial VTH, measu ed a 0 k ad(Si). Ma e ials 2019, 12, x FOR PEER REVIEW 5 o 15 The e ec o gamma adia ion on bo h es ed s uc u es was di e en . In his sec ion, we analyze he changes o he s a ic and dynamic cha ac e is ics conside ing he e ec s o applied bias condi ions due o he po en ial inc ease o he concen a ion o ac i a ed GaN de ec s [21]. 3.1. T ans e Cha ac e is ics The main elec ical measu emen s add ess he d ain and ga e cu en s, which a e exp essed as a unc ion o he ga e and d ain ol ages. F om he I DS –V GS cha ac e is ics, di e en pa ame e s can be es ablished o p o ide in o ma ion abou he changes su e ed by he GaN HEMTs depending on he adia ion applied. The i s pa ame e analyzed was he h eshold ol age (V TH ), which was ex ac ed by he cu en ex apola ion me hod, whe e he h eshold ol age alue is he V GS alue co esponding o I DS = 100 mA wi h a V DS = 0.1 V. The h eshold ol age is usually measu ed a lowe cu en s; howe e , ou se up and he limi s de ined o he es (un il 3 A) o ced us o use a highe cu en alue (100 mA) o ha e be e p ecision o h eshold ol age measu emen . Figu e 2 shows he a ia ions o he h eshold ol age o all he de ices. I is impo an o highligh he way he esul s a e plo ed, since each de ice had a di e en V TH alue. To p o ide a compa ison, i was necessa y o no malize his alue, di iding he V TH o each de ice a each s ep be ween he co esponding ini ial V TH , measu ed a 0 k ad(Si). The HD-GIT HEMT de ices did no su e any change wi h he accumula ed dose, bu he MISHEMT de ices displayed di e en beha io s depending on he bias applied du ing he i adia ion p ocess. The de ices subjec ed o d ain and ga e ol age du ing i adia ion p esen ed a posi i e inc emen o V TH be ween 5% and 10%, while he de ices s essed only wi h a posi i e ol age a he d ain expe ienced a highe posi i e d i o V TH be ween 10% and 25%. In con as , he de ices subjec ed o adia ion wi h sho ed pins su e ed a nega i e a ia ion o V TH a ound 5–8%. In all cases, once he high- empe a u e annealing s ep o 168 h was inished, all he de ices almos ully eco e ed hei o iginal alues. Figu e 2. No malized h eshold ol age o GaN HEMTs as a unc ion o o al dose. Figu e 2. No malized h eshold ol age o GaN HEMTs as a unc ion o o al dose. The HD-GIT HEMT de ices did no su e any change wi h he accumula ed dose, bu he MISHEMT de ices displayed di e en beha io s depending on he bias applied du ing he i adia ion p ocess. The de ices subjec ed o d ain and ga e ol age du ing i adia ion p esen ed a posi i e inc emen o V TH be ween 5% and 10%, while he de ices s essed only wi h a posi i e ol age a he d ain expe ienced a highe posi i e d i o V TH be ween 10% and 25%. In con as , he de ices subjec ed o adia ion wi h sho ed pins su e ed a nega i e a ia ion o V TH a ound 5–8%. In all cases, Ma e ials 2019,12, 2760 6 o 15 once he high- empe a u e annealing s ep o 168 h was inished, all he de ices almos ully eco e ed hei o iginal alues. This nega i e shi o V TH in de ices sho ed du ing i adia ion and he posi i e shi o he de ices wi h bias applied ma ched wi h he indings epo ed in [ 8 , 22 ]. The nega i e shi was due o he apped holes in he ga e dielec ic o he in e ace wi h AlGaN. The eco e y o V TH du ing he high- empe a u e annealing is explained by he elease o apped holes in he AlGaN laye and/o hei neu aliza ion by elec ons om he channel. Highe empe a u es esul ed in highe hole mobili y, allowing he holes o mig a e o ecombine and, he e o e, educing he hole impac on de ice ope a ion. Meanwhile, du ing i adia ion wi h posi i e bias applied, bo h hole and elec on apping ook place. Howe e , due o he high ol age applied, he elec ons illed he hole aps, and in his way, he elec on aps domina ed in assis ing he posi i e shi o he de ices subjec ed o bias ol age du ing i adia ion. Mo e in o ma ion can be ex ac ed wi h he second pa ame e analyzed, which is he a ia ion o e he ga e cha ac e is ics I GS –V GS (inpu cha ac e is ic), which is a way o e alua e he quali y o he ga e con ac . Figu e 3shows he e olu ion o I GS o e he accumula ed dose o all he de ices measu ed a he maximum ga e ol age applied o 3 V. Ma e ials 2019, 12, x FOR PEER REVIEW 6 o 15 apping ook place. Howe e , due o he high ol age applied, he elec ons illed he hole aps, and in his way, he elec on aps domina ed in assis ing he posi i e shi o he de ices subjec ed o bias ol age du ing i adia ion. Mo e in o ma ion can be ex ac ed wi h he second pa ame e analyzed, which is he a ia ion o e he ga e cha ac e is ics IGS–VGS (inpu cha ac e is ic), which is a way o e alua e he quali y o he ga e con ac . Figu e 3 shows he e olu ion o IGS o e he accumula ed dose o all he de ices measu ed a he maximum ga e ol age applied o 3 V. Figu e 3. No malized o wa d ga e cu en (IGS) o GaN HEMTs as a unc ion o he o al dose measu ed a VGS = 3V. As shown in Figu e 3, he HD-GIT de ices did no su e any changes, while he GaN MISHEMT de ices su e ed a high inc ease in he IGS wi hou dependency on he bias applied du ing i adia ion. This inc ease was p opo ional o he accumula ed dose and independen o he bias condi ion applied. This o wa d ga e cu en did no eco e i s ini ial alue a e he annealing p ocess. Only a sligh eco e y a ound 10% was obse ed o he s essed de ices, while sho ed MISHEMT de ices s ill deg aded a e he high- empe a u e annealing. Inc eases in IGS can be explained by adia ion-induced de ec s on he ga e insula o . The ga e insula o in MISHEMT is used o make no mally o de ices, which deple es he channel while g ea ly educing he ga e cu en . Howe e , he de ec s c ea ed du ing i adia ion educe he e ec i eness o his insula o , and holes can become apped in hese de ec s, hus inc easing he ga e cu en . This also explains he nega i e shi o he h eshold ol age men ioned abo e in he sho ed de ices. Focusing on MISHEMT de ices and measu ing he inc ease o IGS has shown ha he e a e no di e ences due o bias applied, bu one di e ence was de ec ed a his wo k, which is he displacemen o IG–VGS cha ac e is ics wi h he accumula ed dose. In o de o measu e his displacemen , he ga e ol age was measu ed when he ga e cu en c ossed 2 mA, which we e e o as VTH_IG o di e en ia e i om he VTH. In Figu e 4a, blue ma ke s show how his ol age was measu ed, and he no malized alues a e shown in Figu e 4b o all he de ices. This igu e demons a es ha he e is a clea di e ence be ween MISHEMT de ices subjec ed o i adia ion wi h nega i e ga e ol age and he o he MISHEMTs. The e was a signi ican ly la ge nega i e shi in he VTH_IG in he de ices subjec ed o adia ion wi h a nega i e ga e bias. This di e ence can be explained by he addi ional elec on aps c ea ed du ing i adia ion in he egion unde he ga e due o he 0 500 1000 1500 2000 2500 3000 3500 5000 6000 0 2 4 6 8 10 12 14 16 18 Dose (k ad(SiO 2 )) I G /I G @0k ad MISHEMT con ol p-GaN con ol MISHEMT @ 400V DS MISHEMT @ 400V DS A e age MISHEMT @ 400V DS /-5V GS MISHEMT @ 400V DS /-5V GS A e age MISHEMT Sho ed MISHEMT Sho ed A e age p-GaN (All bias) p-GaN (All bias) A e age Ann24h25ºC Ann168h100ºC Figu e 3. No malized o wa d ga e cu en (I GS ) o GaN HEMTs as a unc ion o he o al dose measu ed a VGS =3V. As shown in Figu e 3, he HD-GIT de ices did no su e any changes, while he GaN MISHEMT de ices su e ed a high inc ease in he I GS wi hou dependency on he bias applied du ing i adia ion. This inc ease was p opo ional o he accumula ed dose and independen o he bias condi ion applied. This o wa d ga e cu en did no eco e i s ini ial alue a e he annealing p ocess. Only a sligh eco e y a ound 10% was obse ed o he s essed de ices, while sho ed MISHEMT de ices s ill deg aded a e he high- empe a u e annealing. Inc eases in I GS can be explained by adia ion-induced de ec s on he ga e insula o . The ga e insula o in MISHEMT is used o make no mally o de ices, which deple es he channel while g ea ly educing he ga e cu en . Howe e , he de ec s c ea ed du ing i adia ion educe he e ec i eness o his insula o , and holes can become apped in hese de ec s, hus inc easing he ga e cu en . This also explains he nega i e shi o he h eshold ol age men ioned abo e in he sho ed de ices. Ma e ials 2019,12, 2760 7 o 15 Focusing on MISHEMT de ices and measu ing he inc ease o I GS has shown ha he e a e no di e ences due o bias applied, bu one di e ence was de ec ed a his wo k, which is he displacemen o I G –V GS cha ac e is ics wi h he accumula ed dose. In o de o measu e his displacemen , he ga e ol age was measu ed when he ga e cu en c ossed 2 mA, which we e e o as V TH_IG o di e en ia e i om he V TH . In Figu e 4a, blue ma ke s show how his ol age was measu ed, and he no malized alues a e shown in Figu e 4b o all he de ices. This igu e demons a es ha he e is a clea di e ence be ween MISHEMT de ices subjec ed o i adia ion wi h nega i e ga e ol age and he o he MISHEMTs. The e was a signi ican ly la ge nega i e shi in he V TH_IG in he de ices subjec ed o adia ion wi h a nega i e ga e bias. This di e ence can be explained by he addi ional elec on aps c ea ed du ing i adia ion in he egion unde he ga e due o he educed ba ie heigh s o he aps when nega i e bias was applied a he ga e. This induced an inc ease in he ga e cu en o he low ga e ol ages due o he need o ga e cu en injec ion o emo e he aps. Ma e ials 2019, 12, x FOR PEER REVIEW 7 o 15 as V TH_IG o di e en ia e i om he V TH . In Figu e 4a, blue ma ke s show how his ol age was measu ed, and he no malized alues a e shown in Figu e 4b o all he de ices. This igu e demons a es ha he e is a clea di e ence be ween MISHEMT de ices subjec ed o i adia ion wi h nega i e ga e ol age and he o he MISHEMTs. The e was a signi ican ly la ge nega i e shi in he V TH_IG in he de ices subjec ed o adia ion wi h a nega i e ga e bias. This di e ence can be explained by he addi ional elec on aps c ea ed du ing i adia ion in he egion unde he ga e due o he educed ba ie heigh s o he aps when nega i e bias was applied a he ga e. This induced an inc ease in he ga e cu en o he low ga e ol ages due o he need o ga e cu en injec ion o emo e he aps. (a) (b) Figu e 4. (a) Ga e Scho ky diode cha ac e is ics I GS –V GS (inpu cha ac e is ic) o V DS = 0.1 V. (b) No malized h eshold ol age measu ed o a ga e cu en o 2 μA. 3.2. O -S a e D ain Cu en Measu emen 0123 0 5µ 10µ 15µ 20µ 25µ 30µ 35µ 40µ 45µ 50µ 55µ 60µ 65µ 2 μ A I G (A) V GS (V) Q3-0k ad Q3-300k ad Q3-1259k ad Q3-1710k ad Q3-2166k ad Q3-3697k ad Q3-ann24h25C Q3-ann168h100C V TH_IG measu emen Figu e 4. ( a ) Ga e Scho ky diode cha ac e is ics I GS –V GS (inpu cha ac e is ic) o V DS =0.1 V. ( b ) No malized h eshold ol age measu ed o a ga e cu en o 2 µA. Ma e ials 2019,12, 2760 8 o 15 3.2. O -S a e D ain Cu en Measu emen In each s ep du ing he adia ion es , he d ain- o-sou ce leakage cu en was also measu ed. The alue o he leakage cu en measu ed a 500 V o d ain while he ga e was biased wi h 0 V is shown in Figu e 5. The HD-GIT HEMT i adia ed samples su e ed a small inc ease in he d ain leakage cu en ; con e sely, he GaN MISHEMT de ices expe ienced a much la ge inc ease, which was se en imes highe han i s alue be o e i adia ion. The leakage cu en s be ween he d ain and he sou ce inc eased wi h he accumula ed dose du ing i adia ion. Fu he mo e, hey eco e ed app oxima ely hei ini ial alues a e he high- empe a u e annealing p ocess. This inc ease in leakage cu en ma ched he ga e insula o deg ada ion, which a o ed hole apping due o he educed ba ie heigh o he aps, and i educed he e ec i eness o he channel deple ion. Addi ionally, his inc ease in d ain leakage cu en was lowe in he de ices subjec ed o nega i e ga e ol age du ing i adia ion. This was due o he compensa ion o he hole aps by he exis ence o addi ional elec on aps unde he ga e induced by he nega i e ga e bias du ing adia ion. Ma e ials 2019, 12, x FOR PEER REVIEW 8 o 15 was se en imes highe han i s alue be o e i adia ion. The leakage cu en s be ween he d ain and he sou ce inc eased wi h he accumula ed dose du ing i adia ion. Fu he mo e, hey eco e ed app oxima ely hei ini ial alues a e he high- empe a u e annealing p ocess. This inc ease in leakage cu en ma ched he ga e insula o deg ada ion, which a o ed hole apping due o he educed ba ie heigh o he aps, and i educed he e ec i eness o he channel deple ion. Addi ionally, his inc ease in d ain leakage cu en was lowe in he de ices subjec ed o nega i e ga e ol age du ing i adia ion. This was due o he compensa ion o he hole aps by he exis ence o addi ional elec on aps unde he ga e induced by he nega i e ga e bias du ing adia ion. Figu e 5. No malized d ain- o-sou ce leakage cu en when applying a posi i e ga e ol age o 500 V a he d ain wi h 0 V a he ga e. 3.3. Dynamic Resis ance Measu emen In o de o e alua e he RON_dyn o he de ices subjec ed o adia ion, wo di e en es s we e pe o med: he double-pulse es (DPT) and he mul i-pulse es (MPT). Bo h es s we e conduc ed wi h he ci cui se up shown in Figu e 1, and he d i ing signal es sequence o each es is de ailed in Appendix A. A e he analysis o he esul s ob ained by hese double-pulse measu emen es s, we ound ha o he HD-GIT HEMTs, he e we e no signi ican changes o e he RON_dyn independen o he bias condi ions applied du ing gamma i adia ion. A p oo o one measu ed sample is shown in Figu e 6. Howe e , he GaN MISHEMTs su e ed a change in i s RON_dyn depending on he bias condi ions applied du ing i adia ion. An inc ease was e idenced when he MISHEMT was subjec ed o i adia ion wi h d ain and ga e bias applied simul aneously. As an example o his beha io , he RON_dyn alues o he de ice Q3 a di e en i adia ion and annealing s eps a e shown in Figu e 7. These de ices su e ed an inc ease in RON_dyn p opo ional o he accumula ed dose. The o al accumula ed dose had a deg ada ion e ec on he de ice, which sligh ly inc eased he RDSON, bu a e he annealing p ocess, an e en highe deg ada ion was obse ed in he channel esis ance. 500 1000 1500 2000 2500 3000 3500 5000 6000 0 1 2 3 4 5 6 7 8 9 10 MISHEMT con ol p-GaN con ol MISHEMT @ 400V DS MISHEMT @ 400V DS A e age MISHEMT @ 400V DS /-5V GS MISHEMT @ 400V DS /-5V GS A e age MISHEMT Sho ed MISHEMT Sho ed A e age p-GaN (All bias) p-GaN (All bias) A e age I DSS /I DSS @0k ad Dose (k ad(SiO 2 )) Ann24h25ºC Ann168h100ºC Figu e 5. No malized d ain- o-sou ce leakage cu en when applying a posi i e ga e ol age o 500 V a he d ain wi h 0 V a he ga e. 3.3. Dynamic Resis ance Measu emen In o de o e alua e he R ON_dyn o he de ices subjec ed o adia ion, wo di e en es s we e pe o med: he double-pulse es (DPT) and he mul i-pulse es (MPT). Bo h es s we e conduc ed wi h he ci cui se up shown in Figu e 1, and he d i ing signal es sequence o each es is de ailed in Appendix A. A e he analysis o he esul s ob ained by hese double-pulse measu emen es s, we ound ha o he HD-GIT HEMTs, he e we e no signi ican changes o e he R ON_dyn independen o he bias condi ions applied du ing gamma i adia ion. A p oo o one measu ed sample is shown in Figu e 6. Ma e ials 2019,12, 2760 9 o 15 Ma e ials 2019, 12, x FOR PEER REVIEW 9 o 15 Figu e 6. Dynamic esis ance double-pulse measu emen s a 500 V o p-GaN Q13, subjec ed o d ain and ga e ol age du ing i adia ion. Figu e 7. RON_dyn double-pulse measu emen s a 500 V o GaN MISHEMT Q3 subjec ed o d ain and ga e ol age du ing i adia ion. The inc ease o RON_dyn was highe in he second pulses han in he i s ones. This shows ha apping was induced by ho elec ons du ing he semi-on-s a e in he swi ching e en s. The e o e, applying mo e swi ching pulses would inc ease he RON_dyn. To demons a e his beha io , he mul i- pulse es was pe o med on his de ice. The esul is shown in Figu e 8, e ealing he inc ease o he RON_dyn and p o iding in o ma ion abou di e en apping phenomena. In he p e-i adia ion measu emen , he beha io was an exponen ial inc ease o he RON_dyn ha s a ed a a ol age s ess o a ound 450 V. In he pos -i adia ion measu emen , a new apping beha io occu ed, which had a linea dependence on he ol age applied om 100 o 550 V. 0 100µ 200µ 300µ 400µ 0 50m 100m 150m 200m 250m 300m R DS ( Ω ) Time (s) Q13-0k ad Q13-1877k ad Q13-2626k ad Q13-3697k ad Q13-Ann24h25C Q13-Ann168h100C 0 100µ 200µ 300µ 400µ 0 50m 100m 150m 200m 250m 300m R DS ( Ω ) Ti m e (s) Q3-0k ad Q3-1877k ad Q3-2626k ad Q3-3697k ad Q3-Ann24h25C Q3-Ann168h100C Figu e 6. Dynamic esis ance double-pulse measu emen s a 500 V o p-GaN Q13, subjec ed o d ain and ga e ol age du ing i adia ion. Howe e , he GaN MISHEMTs su e ed a change in i s R ON_dyn depending on he bias condi ions applied du ing i adia ion. An inc ease was e idenced when he MISHEMT was subjec ed o i adia ion wi h d ain and ga e bias applied simul aneously. As an example o his beha io , he RON_dyn alues o he de ice Q3 a di e en i adia ion and annealing s eps a e shown in Figu e 7. These de ices su e ed an inc ease in R ON_dyn p opo ional o he accumula ed dose. The o al accumula ed dose had a deg ada ion e ec on he de ice, which sligh ly inc eased he R DSON , bu a e he annealing p ocess, an e en highe deg ada ion was obse ed in he channel esis ance. Ma e ials 2019, 12, x FOR PEER REVIEW 9 o 15 Figu e 6. Dynamic esis ance double-pulse measu emen s a 500 V o p-GaN Q13, subjec ed o d ain and ga e ol age du ing i adia ion. Figu e 7. RON_dyn double-pulse measu emen s a 500 V o GaN MISHEMT Q3 subjec ed o d ain and ga e ol age du ing i adia ion. The inc ease o RON_dyn was highe in he second pulses han in he i s ones. This shows ha apping was induced by ho elec ons du ing he semi-on-s a e in he swi ching e en s. The e o e, applying mo e swi ching pulses would inc ease he RON_dyn. To demons a e his beha io , he mul i- pulse es was pe o med on his de ice. The esul is shown in Figu e 8, e ealing he inc ease o he RON_dyn and p o iding in o ma ion abou di e en apping phenomena. In he p e-i adia ion measu emen , he beha io was an exponen ial inc ease o he RON_dyn ha s a ed a a ol age s ess o a ound 450 V. In he pos -i adia ion measu emen , a new apping beha io occu ed, which had a linea dependence on he ol age applied om 100 o 550 V. 0 100µ 200µ 300µ 400µ 0 50m 100m 150m 200m 250m 300m R DS ( Ω ) Time (s) Q13-0k ad Q13-1877k ad Q13-2626k ad Q13-3697k ad Q13-Ann24h25C Q13-Ann168h100C 0 100µ 200µ 300µ 400µ 0 50m 100m 150m 200m 250m 300m R DS ( Ω ) Ti m e (s) Q3-0k ad Q3-1877k ad Q3-2626k ad Q3-3697k ad Q3-Ann24h25C Q3-Ann168h100C Figu e 7. R ON_dyn double-pulse measu emen s a 500 V o GaN MISHEMT Q3 subjec ed o d ain and ga e ol age du ing i adia ion. The inc ease o R ON_dyn was highe in he second pulses han in he i s ones. This shows ha apping was induced by ho elec ons du ing he semi-on-s a e in he swi ching e en s. The e o e, applying mo e swi ching pulses would inc ease he R ON_dyn . To demons a e his beha io , he mul i-pulse es was pe o med on his de ice. The esul is shown in Figu e 8, e ealing he inc ease o he R ON_dyn and p o iding in o ma ion abou di e en apping phenomena. In he p e-i adia ion measu emen , he beha io was an exponen ial inc ease o he R ON_dyn ha s a ed a a ol age s ess