1
In e acial Mechanism o E icien Resis i e Swi ching in Ruddlesden-
Poppe Pe o ski es o Non- ola ile Memo ies
Au ho s: Anku Solanki
1, 2
, An onio Gue e o
3,
*, Qiannan Zhang
1
, Juan Bisque
3
and Tze Chien
Sum
1,
*
5
A ilia ions:
1
Di ision o Physics and Applied Physics, School o Physical and Ma hema ical Sciences,
Nanyang Technological Uni e si y, 21 Nanyang Link, Singapo e 637371
10
2
Depa men o Science, School o Technology, Pandi Deendayal Pe oleum Uni e si y,
Gandhinaga 382007, India.
3
Ins i u e o Ad anced Ma e ials (INAM), Uni e si a Jaume I, 12006 Cas elló, Spain
15
*co esponding au ho s: Tzec[email p o ec ed]g, [email p o ec ed]
Abs ac : Ion mig a ion, one o he causes o hys e esis e ec s, is he bane o halide pe o ski e
op oelec onics. He ein, we le e age his unwelcome ai o unlock new oppo uni ies o
esis i e swi ching using laye ed Ruddlesdsen-Poppe pe o ski es (RPP) and explica e he
20
unde lying mechanisms. The ON/OFF a io o RPP-based de ices is s ongly dependen on he
laye s and peaks a  = 5 demons a ing he highes ON/OFF a io o ~10
4
and minimal ope a ion
ol age in 1.0 mm
2
de ices. A long da a e en ion e en in 60% ela i e humidi y and s able
w i e/e ase capabili ies exempli ies hei po en ial o memo y applica ions. Ac i a ion s udies
du ing impedance spec oscopy measu emen s e eal ha he high esis i e s a e is ela ed o
25
impeded cha ge injec ion om he con ac s. A chemical eac ion be ween mig a ing ions and he
ex e nal con ac s modi y he cha ge ans e ba ie a he in e ace. The in e play be ween
pe o ski e composi ions and ex e nal elec odes con i m ha ions popula ion, mig a ion and hei
eac i i y wi h ex e nal con ac s a e essen ial o esis i e swi ching. Ou indings unco e he
o igin o esis i e swi ching and a new amily o acile ma e ials and de ices o swi ching and
30
memo y applica ions.
35
2
In oduc ion
2019 ma ks he 10
h
yea anni e sa y o pe o ski e sola cells (PSCs) in en ion by Miyasaka.
1
F om humble beginnings o 3.8% e iciency, a whopping 24.2% o da e places hem in he same
5
league as es ablished polyc ys alline Si, and CIGS ino ganic hin ilm sola cells. The pe o ski e
e e has also spawned a pa allel ligh emission ield, yielding pe o ski e ligh emi ing diodes
(PLEDs) wi h ex e nal quan um e iciency > 21% in a sho de elopmen al span.
2
Ac oss bo h
hese ields, cu en hys e esis emains an open challenge in pe o ski e-based op oelec onics.
3, 4
In PSCs, cu en hys e esis lead o de ice ins abili y and imp ecise PCEs de e mina ion; while
10
colo pu i y (in addi ion o de ice ins abili y) is a majo d awback in PLEDs.
5, 6
Following ex ensi e esea ch, he possible sou ces o cu en - ol age hys e esis na owed down
o h ee
7
: (i) cha ge apping/de apping om bulk and su ace de ec s, (ii) mig a ion o he
excess ions due in e s i ial de ec s and (iii) e oelec ic pola iza ion in pe o ski es. Speci ically,
15
ion mig a ion ac oss he bulk pe o ski e and accumula ion a he in e ace unde any ex e nal
impulse like elec ic ield o ligh illumina ion has been e idenced by di e en echniques.
8, 9
In e acial chemical/physical eac ion gi es ise o an ionic beha io dis inc om he bulk. This
leads o ecombina ion losses a he in e aces, hus limi ing he de ice pe o mance.
10, 11
While cu en hys e esis is exec a ion o mos op oelec ical de ices, i is a boon o o he
20
applica ions such as logic ga es
12
, synap ic de ices
13
, neu omo phic elec onics
14
, da a s o age
15,
16
and swi ching
17, 18
, whe e s ong cu en hys e esis is a p ima y equisi e. Howe e , he
mechanism o con ol he hys e esis is s ill unde in ense deba e. Ini ially, he o ma ion and
up u e o he me al ilamen s a ei he elec odes a e belie ed o be he sou ce o cu en
swi ching in pe o ski e-based esis i e andom access memo y (ReRAM) de ices.
16, 19, 20
25
Howe e , u he s udies show ha he edis ibu ion o iodine acancies o m conduc ing
channels unde an ex e nal applied ield and is p oposed as he main sou ce o esis ance change
in pe o ski es.
18, 21, 22
On he o he hand, some epo s sugges ha he in e acial ene gy ba ie
a pe o ski e/elec ode in e ace gi e ise o cu en swi ching.
23
Recen ly, Ge e al.
demons a ed sil e iodide-induced swi ching in cesium-based memo y de ices.
24
Ne e heless,
30
he exac o igin o hys e esis in pe o ski es emains con en ious. A clea unde s anding o he
unde lying causes is essen ial ei he o mi iga ing hys e esis in op oelec onic de ices o
op imizing o esis i e swi ching (RS) in memo y applica ion.
He ein, we le e age he hys e esis in uddlesdsen-poppe (PEA)
2
(MA)
n-1
Pb
n
I
3n+1
(PEA =
phenyle hylammonium, MA = me hylammonium) o ganic-ino ganic hyb id pe o ski es o RS
35
h ough uning he s uc u al laye s and judicious selec ion o he ex e nal con ac s. We elucida e
a uni e sal wo king mechanism ha elies on he accumula ion o ions mig a ed o he con ac s
unde an applied elec ic ield, and he subsequen chemical/physical in e ac ion o hese
accumula ed ions wi h he ex e nal con ac s ha con ol he swi ching beha io . Impedance
spec oscopy (IS) expe imen s e eal he kine ics o he ac i a ion p ocess ha in ol es a
40
educ ion in he cha ge ans e esis ance connec ed o he pe o ski e/con ac in e ace h ough a
chemical eac ion. The op imal composi ion o MA
+
and PEA
+
molecules in he  = 5 sample
3
yields he bes RS pe o mance wi h ON/OFF a io o ~10
4
a 0.15 SET ol age compa ed o he
 = 1, 3, 7 and 3D MAPbI
3
samples in ITO/PEDOT:PSS/pe o ski e/PCBM/Ag de ice
con igu a ion. The  = 5 sample also exhibi good da a e en ion capabili y and en i onmen al
s abili y. Su ace po en ial and cu en swi ching in ba e hin ilm unde spec oscopic
measu emen s ule ou he me al ilamen o ma ion equi emen o RS. We also alida e ha
5
he con ol on he supply o ions and he chemical/physical in e ac ions wi h ex e nal con ac s
and hus swi ching beha io h ough he di e en ex e nal con ac s con aining Ag, Au and SnO
2
.
Resul s
He e, we u ilize di e en laye ed uddlesdsen-poppe (RP) hyb id pe o ski es (Fig. 1a) i.e.,
10
(PEA)
2
(MA)
n-1
Pb
n
I
3n+1
((PEA)
2
PbI
4,
( = 1) (PEA)
2
MA
2
Pb
3
I
10
( = 3), (PEA)
2
MA
4
Pb
5
I
16
( = 5)
and (PEA)
2
MA
6
Pb
7
I
22
( = 7)), and pu e h ee-dimensional pe o ski e MAPbI
3
( = ∞/3D) o
demons a e he excellen hys e esis p ope ies o esis i e andom access memo y (ReRAM)
applica ion. These laye ed RP pe o ski e ilms a e polyc ys alline in na u e and  ep esen s he
dominan n componen o each sample as e alua ed om XRD and UV-Vis abso p ion spec a
15
(Fig. S1 and S2). The mo phology o hese pe o ski e ilms as displayed in Fig. S3 con i m he
good quali y o he samples wi h no ob ious pinholes. Fig. 1b depic s he bes ReRAM de ice
con igu a ion used o his s udy ITO PEDOT:PSS pe o ski e PCBM Ag (see SI o he de ails
o de ice ab ica ion). ITO PEDOT:PSS is he bo om elec ode as a non-hys e e ic elec ode and
sil e is used as op elec ode wi h ~1.0 mm
2
de ice a ea. A hin laye o PCBM (10 ± 2 nm) was
20
inse ed be ween pe o ski e and Ag o minimize hei di ec con ac which can cause elec ode
and pe o ski e deg ada ion.
Fig. 1c shows a ep esen a i e semi-log scale I-V cha ac e is ics o pe o ski e ReRAM de ices
measu ed unde DC ol age sweep (0 V → (V
app
) V → 0 V → - (V
app
) V → 0 V) (see Fig. 3a
25
and S4 o o he de ices). A ol ages < 0.15V his de ice is in a high esis ance s a e (i.e., HRS
o OFF s a e >10
5
Ω) wi h measu ed low cu en le els. Beyond a h eshold ol age, he cu en
inc eases wi h ime, b inging he de ice o a low esis ance s a e (i.e., LRS o ON s a e). This
swi ching o he de ice o he LRS wi h a cu en le el o 1 mA a highe bias is e med he
"SET" p ocess. The onse ol age (~0.15 V) is mo e ob ious in he log-log plo (Fig. 1d). The
30
ansi ion om HRS o LRS is a he sha p. We a ibu e his o he induced chemical physical
in e ac ions be ween he mig a ing ions and he ex e nal con ac s. An ini ial high cha ge ans e
esis ance om he con ac o he pe o ski e laye is educed by su ace eac ion a highe bias,
which is co obo a ed by impedance spec oscopy esul s (mo e discussion la e ). Abo e he SET
ol age, he supp ession o he cha ge ans e ba ie om he elec ode ini ia e he ca ie
35
conduc ion in he bulk, which is e iden om he Ohmic egime (slope ~1.2, Fig. 1d) in he
LRS. Subsequen ly, unde nega i e applied bias, he de ice esis ance changes om a LRS o he
HRS by a "RESET" p ocess o he non-conduc ing s a e wi h a cu en le el o 10
-8
- 10
-6
A (Fig.
1c).
40
4
Fig. 1: De ice elec ical p ope ies: (a) C ys al s uc u e o uddlesdsen-poppe pe o ski e ( = 1, 3, 5, 7) and 3D
pe o ski e, (b) schema ic de ice s uc u e o Glass/ITO/PEDOT:PSS/pe o ski e/PCBM/Ag. A hin laye o PCBM
(~10 nm), was inse ed be ween he pe o ski e and op elec ode o con ol Ag mig a ion, (c) Semi-log I-V 5
cha ac e is ics o  = 5 ((PEA)
2
MA
4
Pb
5
I
16
) based ReRAM de ice wi h he scan a e o ~ 100 mV/s (See Fig. 3a and
Fig. S4 o o he s) – ed a ows indica e he scan di ec ion, (d) Log-log I-V plo p esen ing he SET ol age in HRS
and Ohmic conduc ion in LRS. (e) T ansien cu en a di e en applied bias, ( ) de ice endu ance es o 500
cycles (applied ol age pulse -1.5/+1.5 V and ead cu en a 0.15V) (g) da a e en ion unde ~5% ela i e humidi y
(RH) (inside N
2
glo ebox) and ambien condi ion wi h 60% RH. 10
The selec ion o he op imum de ice ope a ing condi ions is essen ial because a highe applied
bias can elec ically sho he de ice while a lowe bias can lead o unde -pe o mance. F om
Fig. 1e, he OFF s a e emains wi h low cu en and nea ly in a ian a <0.15 V, while he
cu en g adually inc eases a highe applied bias (>0.15 V) and inally swi ching o he ON
15
s a e. The excellen swi ching s abili y o hese de ices (s able cu en measu ed a 0.15V) is
con i med by con inuously igge ing hem om SET o RESET s a e by a ol age pulse sligh ly
highe han he sa u a ion ol age (i.e., -1.5/+1.5 V) (Fig. 1 ). Ope a ional s abili y o ou de ice
is clea ly demons a ed om he 500 I-V scans (Fig. S6). S ong da a e en ion abili y, a e y
impo an c i e ion o memo y applica ion, is measu ed o up o 250 hou s (Fig.1g). Howe e ,
20
he cu en in HRS inc eased g adually while exposed o he high humid ambien en i onmen
(~60 % RH). The op ical images o he de ice be o e and a e wo days exposu e o he ambien
5
(Fig. S7) con i m he deg ada ion o Ag, p obably con e sion in o AgO
x,
which could pene a e
in o he pe o ski e and eac wi h i , hus educing he de ice esis ance in OFF s a e.
25
Howe e , he e ec o ambien wa e canno be excluded since i has been epo ed ha
o ma ion o he wa e adduc wi h he pe o ski e sha ply inc eases he conduc i i y.
26
5
In pe o ski e ReRAM de ices, wo di e en mechanisms ha e been widely a ibu ed o he SET
and RESET p ocesses: (1) Fo ma ion/ up u e o he conduc ing ilamen om me al elec ode
and (2) De ec s edis ibu ion (p edominan ly iodide acancies, iodine in e s i ials due o lowe
ac i a ion ene gy be ween ~0.4 - 0.8 eV o he mig a ion) leading o he ma ked hys e esis.
18, 27
Howe e , i is di icul o disc imina e hese mechanisms in a de ice solely based on he I-V
10
cha ac e is ics. In o de o e eal o ule ou he necessi y o he me al ilamen o ma ion o he
swi ching mechanism, we ca ied ou Scanning Kel in p obe mic oscopy (SKPM) and
conduc i e a omic o ce mic oscopy (c-AFM) o ba e pe o ski e ilms. Ou esul s clea ly show
ha swi ching occu s e en wi hou he op elec ode p esen , he e o e discoun ing he need o
me al ilamen o ma ion in he swi ching mechanism. A comple e discussion can be ound in he
15
suppo ing in o ma ion (Fig. S8). Based on hese indings, we conclude ha ionic edis ibu ion
can o m he conduc ing channels o RS.
22
Howe e , i is impo an o no e ha such conduc ing
pa hways a e ansi o y and will no o m s eady s a es as seen om he consis en d op in
su ace po en ial (Fig. S8d). Hence, ions edis ibu ion alone o LRS/HRS will s ill no explain
he long e en ion obse ed in Fig. 3g. An in e play be ween he pe o ski e and ex e nal con ac s
20
is a wo k in hese ReRAM de ices.
Fig. 2: Impedance Spec oscopy: a, b) Complex impedance spec a o a de ice pola ized close o he ac i a ion
ol age a VDC = 0.2 V du ing IS measu emen s. IS equency scans a e epea ed 15 imes o e one hou (i.e., 1 scan 25
akes 4 minu es). The IS spec a epe i ion numbe is deno ed by = 4×n minu es (whe e n is an in ege ) in he g aph
(c) Equi alen ci cui use o i he impedance da a (d) Cu en inc eases du ing epe i ion o he IS equency scans

6
and he de ice ge s ac i a ed. (e) Fi ing esul s o he esis ance as a unc ion o he cu en , (cu en and esis ance
e o s a e below 5 %). ) Capaci ance equency plo o he spec a shown in a, b).
To in es iga e he o igin o he s ong and e e sible swi ching be ween he LRS and HRS, we
pe o med impedance spec oscopy, ha allows he sepa a ion o di e en con ibu ions o
5
conduc ion by he coupling o esis ances and capaci ances.
28
He e, he bes de ice con igu a ion
(i.e.,  = 5) is used. A DC bias sligh ly highe han he SET ol age swi ch o LRS is applied. We
moni o he de ice beha io a 0.2 V by epe i ion o IS spec a measu emen as he cu en
g adually inc eases. The complex impedance plo is shown as a unc ion o he ime (IS scan
numbe ) in Figs. 2a,b o di e en ange o impedance alues. Fig. 2a show he ini ial HRS ( =
10
0) ha co esponds o e y high impedance. The occu ence o wo a cs om ReRAM de ice is
he ypical esponse o pe o ski e de ices and is usually desc ibed by he equi alen ci cui
shown in Fig. 2c. The high equency (HF) a c con ains in o ma ion on he bulk conduc i i y o
he pe o ski e ilm and he low equency (LF) one ela es o he pe o ski e/con ac s in e ace
(ion mig a ion and accumula ion).
29
As shown in Fig. 2a, he HF a c is cha ac e ized by low
15
impedance alues o 0.2 MΩ, associa ed o he bulk pe o ski e, while he LF a c p o ides la ge
i ing alues o 4.1 MΩ, which is ela ed o he esis i e elemen s a he pe o ski e/con ac
in e ace. Rema kably, as he de ice is ac i a ed, he LF a c disappea s (i.e., only HF a cs
emains in Fig. 2b). The cu en ab up ly inc eases 2 o de s o magni ude (Fig. 2d, ime = 480 s,
8 minu es). Resul s show ha he bulk-conduc ion (HF) emains i ually unchanged, while he
20
d ama ic change in cha ge injec ion cha ac e is ics a he in e ace leads o he absence o he
la ge LF a c seen in Fig. 2a. Ins ead o he LF a c, some induc i e e ec s in he low- equency
ange a e obse ed, co ela ed wi h elec ochemical eac ions and co osion in se e al ma e ials
in ba e y con igu a ion.
30, 31
By epe i ion o he IS measu emen s abo e SET ol age, he e is a
ise o he cu en and he HF esis ance dec eases wi h ime (Fig. 2d), howe e , he induc i e
25
e ec is always p esen . Fig. 2e explici ly demons a es a linea co ela ion be ween he sum o
all esis ances (R
se ies
, R
HF
and R
LF
) obse ed in he IS measu emen s and he injec ed cu en .
The slope is con i med o be -0.9, indica ing an Ohmic esponse when all he esis ances a e
aken in o accoun , obeying Ohm’s law.
30
Addi ional in o ma ion abou cha ge dis ibu ion in he de ice is ob ained by plo ing he
capaci ance- equency plo (Fig. 2 ) whe e cha ge accumula ion a he pe o ski e/ex e nal
con ac s in e ace is ela ed o he o ma ion o an ionic double laye connec ed o he LF a c.
32, 33
In pe o ski es, double laye capaci ance a ises om he accumula ion o iodine ions/ acancies a
he elec ode wi h ypical alues in he ens o µF/cm
-2
. These alues a e only de ec ed in he
35
OFF s a e whe e he LF signal connec s he p esence o ions a he con ac s (C
DL
~ µF/cm
-2
) wi h
impeded cha ge ca ie injec ion om he con ac s (R
LF
~ M). The ea e , when he de ice
app oaches he SET p ocess, he capaci ance inc eases d ama ically up o alues o 10 mF/cm
2
a
equencies o 10 mHz, mo e han h ee o de s o magni ude, highe han ypical double laye
and Helmhol z laye capaci ances. Wi h common capaci ance alues o con e sion eac ion in
40
ba e ies he capaci ance esul s poin s o a chemical eac ion o iodide ions wi h he ex e nal Ag
in e aces ha lead o o ma ion o AgI.
34
As he SET p ocess ac i a es, he LF a c disappea s
wi h subsequen o ma ion o induc i e ea u es, which ha e been p e iously co ela ed wi h he
eac ion a elec ode/ac i e laye in e ace, esul ing in he lowe se ies esis ance o he de ice.
45
Nex , I-V cha ac e is ics in di e en ReRAM de ices a e sys ema ically in es iga ed o unco e
he in e play be ween he pe o ski e and ex e nal con ac s. We ab ica ed a whole se ies o
7
de ices con aining pe o ski es wi h di e en  alues ( = 1, 3 , 5, 7, ∞/3D) wi h con ac s
modi ica ion. Only, selec ed I-V cu es a e shown he e o cla i y and o he s a e p esen ed as
suppo ing in o ma ion (Fig. S4, 10, 11). Based on he I-V cha ac e is ics (Fig. 3a), he OFF-
s a e cu en clea ly depends on he pe o ski e composi ion (i.e.,  alues). In RPP hin ilms,
PbI
6
oc ahed al laye s a e he p edominan conduc ing channels, he e o e o e all bulk
5
conduc i i y in hin ilm depends on he numbe o laye s. Be ween wo elec odes, hese laye s
a e (i) un-in e up edly aligned in (110) plane pa allel o each o he om he bo om o op
elec ode; and/o (ii) pa ially c oss-linked; (iii) highe  alue laye s a e connec ed ia lowe 
alues laye s.
35, 36
In ei he case, cha ge anspo ac oss he pe o ski e laye is a ec ed by PEA
laye s, which ha e ela i ely poo cha ge ca ie mobili y o con ol he OFF s a e. The balance
10
in he composi ion o 3D oc ahed al and PEA laye s and hei well dis ibu ion in  < 5 leads o
e y high esis ance in he OFF s a e. On he o he hand, in highe o de laye ed pe o ski es ( >
5), he dominan p esence o he 3D laye s leads o well-connec ed channels ac oss he elec odes
and cha ge anspo dominan hough hese channels and o s a e esis ance is educed o show
a ela i ely highe cu en . All hese de ices a low ol age (< 0.4V) a e in high esis ance s a e
15
(HRS o OFF s a e i.e. >10
5
Ω) wi h measu ed low le el o cu en . The SET ol age is
app oxima ely 0.4, 0.3, 0.2, 0.22, 0.2 V o  = 1, 3, 5, 7 and 3D based ReRAM de ices,
espec i ely (Fig. 3b). The SET ol age o hese de ices a e es ima ed om he de ia ion o he
cu en om slope 1 in HRS (Fig. 1e).
18
This end is in ag eemen wi h he enhanced ionic
anspo obse ed o he 3D pe o ski es compa ed o 2D and mixed pe o ski es.
37
The e o e,
20
we es ablish a connec ion be ween ionic mig a ion and SET ol ages in ReRAM con igu a ion as
discussed below. The maximum ON/OFF a io ~1 x 10
4
and minimum alue o SET ol age in
 = 5 indica es ha RS is mos e icien as compa ed o o he laye ed pe o ski es (Fig. 3b). The
a ia ion o he ON/OFF a io o di e en  alue RP pe o ski es wi h simila hickness (~
270 ± 20 nm) is also a seconda y e idence o con i m ha RS in laye ed RPP is no caused by
25
me al ilamen o ma ion. O he han he  alues, he elec onic conduc i i y in RPP o OFF
s a e cu en is also de e mined by he e mina ing end o pe o ski e s uc u e and he in e ac ion
a pe o ski e/con ac in e ace, which con ol he cha ge ans e o he conduc ion band. Fo
example, he p esence o iodine o MA acancies (V
I
and V
MA
) a CH
3
NH
3
PbI
3
/TiO
2
in e ace
modi y he ene gy landscape and hinde he cha ge ex ac ion.
38
Simila ly, a high local
30
concen a ion o e minal iodine ions in he p esence o oxida i e condi ions (high hole
concen a ion a he Ag con ac ), o m me al-halides (i.e., AgI o AuI
3
) bonds locks he iodine a
he con ac and modi ies he injec ion p ope ies o he con ac educing he cha ge ans e
esis ance.
39
35
Se e al op elec odes (Ag, Au, ZnO/Ag, MoO
3
/Ag, Al…) and bo om elec odes (ITO, P ,
PEDOT:PSS) ha e been used in he li e a u e in combina ion wi h ei he iodide o b omide based
pe o ski es. Whils some elec ode/pe o ski e combina ions p o ide be e RS han o he s, a
comple e pic u e o he mos sui able combina ion is s ill lacking. To unco e he mechanism,
a ious de ice con igu a ions wi h di e en pe o ski e composi ion and ex e nal con ac s (Fig.
40
3c, 3d and Fig. S10, 11) we e ab ica ed and hei I-V cha ac e is ics a e in es iga ed. The lack
o cu en swi ching in de ices wi h op ba e Au, hin PCBM/Au and wi h hick PCBM/Ag
highligh s he signi icance o he popula ion o mig a ing ions ( ela i ely less mig a ing ions in 
= 1 compa ed o  = 5) and hei eac i i y wi h he elec odes (highe eac i i y o Ag compa ed
o Au). He e we p opose ha he un a o able chemical eac i i y o Au wi h I
-
equi es highe
45
8
concen a ion o ions, p o ided by he 3D pe o ski e, and di ec con ac wi h pe o ski e a e
needed o he RS. The e o e, we show ha he ion’s popula ion, hei mig a ion and eac i i y
wi h he con ac s is essen ial o achie e he memo y e ec .
5
Fig. 3: Elec ical p ope ies o di e en ReRAM de ices: (a) I-V cu es o de ices con aining pe o ski es wi h
di e en  indices ( = 1, 5, ∞/3D) in a ITO/PEDOT:PSS/Pe o ski e/PCBM/Ag con igu a ion (see Fig. S9 o
 = 3 and 7). (b) Summa y o SET ol ages and ON/OFF a ios o di e en de ices. (c) I-V cu e o  = 1, 5,
∞/3D based de ices con aining ba e Au op elec ode in he con igu a ion glass/ITO/PEDOT:PSS/PV/Au (d) I-V
cu e o de ices con aining di e en bo om elec odes and a hick laye o PCBM (80 ±10 nm) be ween 10
pe o ski e ( = 5) and Ag.
Las ly, we a e able o une he cu en swi ching quad an by con olling he chemical/physical
eac ion o mig a ing ions a pe o ski e/con ac in e aces, hus e i ying ou indings. The
in oduc ion o hick PCBM laye (80 ± 10 nm) in ou bes de ice c ea es a mo e obus physical
15
ba ie ha p e en s he in e ac ion be ween I
-
and Ag o o m AgI. Hence, he absence o RS in
hese de ices (Fig. 3c). On he o he hand, he non-in e ac i e na u e o PEDOT:PSS wi h
mig a ing ions does no con ibu e o he cha ge ans e and ON s a e e en ion and he e o e no
cu en swi ching a nega i e applied bias (no e he highly symme ic cu e). Al e na i ely, i he
PEDOT:PSS laye is eplaced by SnO
2
whils keeping he hick PCBM laye , he in e ac ions o
20
SnO
2
wi h mig a ing ions lead o a swi ching beha io a nega i e bias (Fig. 3d). This is
p obably due o I
-
ions coo dina ing wi h Sn
3+
de ec s p esen in SnO
2
ha yields he RS.
O e all, we conclude ha bo h elec ode ma e ials (Ag and SnO
2
) may lead o chemical
9
in e ac ions wi h mig a ing ions ha can be uned o p epa e ReRAM de ices. Al e na i ely,
PEDOT:PSS and PCBM o e a physical ba ie o mig a ing iodine ions and do no p o ide
s ong in e ac ions, making he laye s ine o ReRAM esponse. Selec ion o ma e ials can be
based on hese simple design ules o build up a comple e lib a y.
5
Based on he unde s anding om a combina ion o impedance spec oscopy and he esul s o
di e en con igu a ions, we p opose a gene al swi ching mechanism ha elies on he in e play
be ween wo kine ics; (i) ion mig a ion in he pe o ski e and (ii) e e sible chemical in e ac ions
wi h he ex e nal con ac s. This mechanism elies on he cu en unde s anding o he kine ic
componen s o he impedance esponse ha inc eases he capaci ance in a phenomenon ela ed o
10
he LF induc i e e ec and he chemical eac i i y o mig a ing ions wi h ex e nal con ac s.
40
Fig. 4 ep esen s he bes pe o ming con igu a ion desc ibed in his wo k. He e, he e e sible
eac ion o o ma ion and dissolu ion o a AgI monolaye a he in e ace is su ice o modi y he
injec ion p ope ies o con ol he cha ge ans e . The e o e, he selec ion o con ac s and hei
edox po en ials a e he p ima y ac o s o de e mine he chemical eac ion a he in e ace and
15
hus he equi ed onse po en ial o SET p ocess o obse e he memo y e ec . Simila ly, he
high deg ee o he e e sibili y o he eac ion is impe a i e o high ON/OFF a io in ReRAM
de ice.
20
Fig. 4: P oposed swi ching mechanism: (a) OFF S a e: Iodine acancies (VI) e mina ion a he in e ace limi he
cha ge ans e om Ag o he pe o ski e laye , (b) SET p ocess: unde an applied elec ical he pe o ski e
e mina ion is modi ied by VI mig a ion o o m a hin laye o AgI, bene icial o he e icien cha ge ans e o
o m LRS. (c) RESET p ocess: a he e e se pola i y o he applied bias, VI mig a es o he opposi e di ec ion and 25
end up pilling a he elec ode leading o he HRS. These p ocesses a e e e sible based on he eac ion a he con ac
in e aces.
Conclusions
30
He e, we demons a ed and in es iga ed he RS in RP pe o ski e using di e en ReRAM de ice
con igu a ions. Op imal  = 5 composi ion shows he bes pe o mance wi h minimal SET
ope a i e ol age and highes ON/OFF a io. The eliable endu ance o 500 cycles and long da a
e en ion o 250 hou s in an ine en i onmen wi h subsequen 50-hou da a e en ion unde
60% RH shows he excellen po en ial o RP pe o ski es o esis i e memo y applica ions.
35
Mic oscopic s udies e eal he su ace conduc i i y swi ching in he absence o me al ilamen
highligh he impo ance o he chemical/physical eac ion a he in e ace o con ol he RS. We
p opose a gene al me hod o une/enhance memo y e ec ReRAMs in he desi ed I-V quad an
by selec ion o app op ia e RP pe o ski e and elec ode ma e ials. Ou indings unco e new