flexible non-volatile optical memory thin-film transistor ... · principle of a ﬂoating gate...

Flexible non-volatile optical memory thin-filmtransistor device with over 256 distinct levelsbased on an organic bicomponent blendTim Leydecker1, Martin Herder2, Egon Pavlica3, Gvido Bratina3, Stefan Hecht2*, Emanuele Orgiu1*and Paolo Samorì1*

Organic nanomaterials are attracting a great deal of interest for use in flexible electronic applications such as logiccircuits, displays and solar cells. These technologies have already demonstrated good performances, but flexible organicmemories are yet to deliver on all their promise in terms of volatility, operational voltage, write/erase speed, as well asthe number of distinct attainable levels. Here, we report a multilevel non-volatile flexible optical memory thin-filmtransistor based on a blend of a reference polymer semiconductor, namely poly(3-hexylthiophene), and a photochromicdiarylethene, switched with ultraviolet and green light irradiation. A three-terminal device featuring over 256 (8 bitstorage) distinct current levels was fabricated, the memory states of which could be switched with 3 ns laser pulses.We also report robustness over 70 write–erase cycles and non-volatility exceeding 500 days. The device was implementedon a flexible polyethylene terephthalate substrate, validating the concept for integration into wearable electronics andsmart nanodevices.

Although silicon-based technology is still effectively leadingtoday’s electronics industry, the development of high-performance organic semiconducting nanomaterials has

been the subject of important research efforts in the past twodecades1–3. Such materials are undoubtedly an interesting alter-native, because their properties can be tuned readily via ad hocchemical synthesis, and they can be fabricated using mild processingfrom solution using approaches that can be scaled up and applied to‘soft’ substrates, making them ideal candidates for the low-costfabrication of large-area electronic devices. They have shownexceptional versatility, enabling integration on flexible and/ortransparent substrates for a new generation of electronics4–7.Among the plethora of applications of semiconducting organicmaterials and more generally of π-conjugated (macro)molecules,memories have attracted particular interest because they combinehigh switching ratios8–10 and long retention times11, as well as lowoperating voltages12–14 and the possibility to be integrated ontoflexible substrates15–18.

To increase the data storage capabilities of electronic devices(hard drives, random access memories and so on), the mostpopular approaches so far have relied on a continuous scalingdown to enable the integration of an increasing number ofmemory cells per unit area. These strategies have shown theirlimits, as scaling down is hampered by photolithography, and fabri-cation complexity has increased dramatically over the years19,20. Analternative method is based on the development of memory cellswith increased storage capabilities, that is, multilevel memories. In1 bit storage devices, a finite yet limited degree of volatility is notdetrimental to the overall storage capacity as long as the twostates, typically represented by a current maximum and minimum,are still distinguishable. The main challenge is to obtain the highestpossible difference in current between an on state and an off state.

When multilevel memories are considered, the problem shiftsdramatically to ensuring the stability of the states, as each state willnecessarily be close to others. In a 4 bit memory cell (16 currentlevels), for example, a one order of magnitude differencebetween each level is not achievable, because this would mean aswitching ratio as high as 1015. Importantly, all methods where ahigh switching ratio is simply based on a threshold voltage shift(that is, not characterized by a change in charge carrier mobilityor current), with the ‘read voltage’ within a threshold voltagewindow, are prone to data loss and incorrect reading of thelevels in multilevel storage21,22.

Two- and three-terminal resistive switching memories based onpolymers, organic small molecules and ferroelectric materials havedemonstrated impressive switching ratios of up to 108 and low pro-gramming voltages23–25. Unfortunately, this approach relies on twostable states of polarization and is therefore not well suited to use inmultilevel memories26. On the other hand, memories designedaround electrets, that is, dielectric materials (usually polymers), orusing polymer/electrolyte pairs exhibiting quasi-permanent electriccharge or dipole polarization, have displayed the ability to performas multilevel memories with different applied programmingvoltages. Although the observed switching ratios of over 108 havepushed them to the forefront of research on organic memories27–30,their unacceptably high programming voltages (up to ±200 V)and, most importantly, their low retention times (<104 s) are stillmajor obstacles towards their integration into everyday electronics31,32.Storing charges in a metal or semiconductor layer by exploiting theprinciple of a floating gate located within the dielectric can beachieved with metallic nanoparticles, with each particle countingas a charge-storage site that is independent and isolated fromother sites. Although not yet extensively explored, this approach isrelevant for the fabrication of multilevel memories with acceptable

1ISIS & icFRC, University of Strasbourg & CNRS, 8 allée Gaspard Monge, 67000 Strasbourg, France. 2Department of Chemistry & IRIS Adlershof,Humboldt-Universität zu Berlin, Brook-Taylor-Straße 2, 12489 Berlin, Germany. 3Laboratory for Organic Matter Physics, University of Nova Gorica,Vipavska 13, SI-5000 Nova Gorica, Slovenia. *e-mail: [email protected]; [email protected]; [email protected]

ARTICLESPUBLISHED ONLINE: 20 JUNE 2016 | DOI: 10.1038/NNANO.2016.87

NATURE NANOTECHNOLOGY | VOL 11 | SEPTEMBER 2016 | www.nature.com/naturenanotechnology 769

© 2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

mailto:[email protected]



http://dx.doi.org/10.1038/nnano.2016.87

http://www.nature.com/naturenanotechnology

programming voltages (up to ±40 V), but so far the retention timesfor long-term multilevel storage (104 s) have been disappointing33.

Compared to electrical programming of organic memories,photoprogramming offers the fundamental advantage of the photo-writing being orthogonal to the electrical readout. Photochromicmolecules have shown promising results as molecular switches,but rather mediocre (semi)conducting behaviour and non-negligible fatigue34–36. Most studies on organic memories based onphotochromicmolecules have focused on three types of photochromicsystem: azobenzenes, spiropyrans and diarylethenes (DAEs)37–40.Optical switching between two independently addressable statesleads to changes both at the molecular and the macroscopic scale.At the molecular level, reversible modifications of the geometricaland electronic structure can be observed, including changes in thedipole moment, π conjugation and bandgap41, which can be trans-lated into corresponding adaptations of the macroscopic propertieson switching, including changes in shape, aggregation behaviourand conductance. Although changes in conductance are often asought-after property, concomitant morphological changes at themesoscale could lead to stability issues, in particular given thatmaintaining high order in the film is crucial for device performance.The optimal solution thus consists of a photochromic molecule in

which the two switching states simultaneously exhibit (1) the lowestpossible change in supramolecular organization, (2) the largestmodulation of the electronic properties, (3) sufficient stability ofboth interconverting isomers, (4) a fast switching induced by thelight stimulus, and (5) efficient switching in the solid state. All ofthese characteristics can be met by DAEs42. In earlier studies,organic memories based on layered device architectures composedof DAE and triarylamine derivatives proved to be a valuableapproach for the production of organic multilevel memories with30 reported current levels43,44. However, the fabrication of multi-layers clearly requires multistep processing, and the devices needseveral minutes of irradiation at high intensity to obtain a sufficientswitching yield, making them unsuitable for applications in the elec-tronic industry. Furthermore, electrical switching exhibits limitedreversibility, retention time and storage capacity, thereby hamperingany technological applications in memory devices. Another bigchallenge is the implementation of such devices on flexiblesubstrates with a view to creating foldable smart electronics.

Here, we use bifunctional organic thin-film transistors (OTFTs)with an active layer consisting of a blend of a DAE moiety, namelyDAE-Me (Fig. 1a), with a reference semiconducting polymer suchas poly(3-hexylthiophene) (P3HT) serving as the electroactive

SS

S S

Vis UV

DAE-Me_o

DAE-Me_c

2010−6 100

101

102

103

104

105

106

10−5

10−2

10−3

10−4

10−3

−20 −40 −60 0

Mob

ility

(cm

2 V−1

s−1

)

10 20 30DAE-Me in the blend (wt%)

40 500

Switc

hing

diff

eren

ce (A

)

Switching ratio

VGS (V)

20 −20

0

−100

−200

−300

−400

−500

−40 −600

DAE-Me_oDAE-Me_c

DAE-Me_oDAE-Me_c

I DS (

μA)

10−11

10−3

0 −60VGS (V)

|I DS|

(A)

a

b

c d

VGS (V)

P3HT + DAE-MeAu source, drain

SiO2n++Si

Figure 1 | Electrical characterization of the bottom-gate/bottom-contact P3HT/DAE-Me devices. a, Chemical formulae of the used DAE, in its openand closed form. Bottom right: Structure of the devices on a SiO2 substrate. b, IDS–VGS curve for the open and closed forms (after 10 min irradiation atλ = 313 nm), at VDS = −60 V. Inset: the same curve represented on a logarithmic scale. c, Switching ratio (IDS,DAE-Me_o/IDS,DAE-Me_c) and switching difference(IDS,DAE-Me_o − IDS,DAE-Me_c) as a function of gate voltage. d, Average mobility μFET of the devices as a function of the weight percentage of DAEs in the blend.

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.87

NATURE NANOTECHNOLOGY | VOL 11 | SEPTEMBER 2016 | www.nature.com/naturenanotechnology770




nanomaterial. The dual functionality in our devices relies on thelarge photoinduced modulation of the energy levels of theDAE-Me photoswitch in the polymer matrix, which allows opticalcontrol over charge transport. When the DAE-Me is in its openform (DAE-Me_o), the corresponding highest occupied molecularorbital (HOMO) level lies outside the P3HT’s bandgap, and it there-fore simply acts as a scattering centre for charge carriers. However,the closed isomer (DAE-Me_c) features a HOMO level that is situ-ated within the bandgap of the polymer matrix and can thereforeaccept holes. This principle of photoswitchable charge trappinghas already been proven in an earlier seminal study and was recentlyextended to both p- and n-type small molecules45–47. These worksserved as proof-of-principle experiments that provided evidence forthe basic mechanisms of the interaction between DAEs and P3HT.Essentially, these fundamental works revealed that when the energylevels of the components are properly engineered, the DAEs act ashole-accepting levels within the matrix of a polymer semiconductor.As a result, the insertion of DAE-Me within a P3HT matrix offersremote control of the output drain current in an OTFT by light.

In this Article we present a three-terminal memory device wherethe active layer is composed of an organic bicomponent DAE/P3HTthin film. We show that such a combination of nanomaterials can besuccessfully used in real technological applications to fabricate non-volatile organic-based memories with a high density of informationstorage. The choice of a three-terminal transistor device geometryas a memory element47 was made due to its superior technologicalrelevance over diode (two-terminal) devices. Transistors are able toimplement a signal addressing function in two-dimensionalmemory arrays that would not be possible if diode-only circuitswere employed. Significantly, these memory devices can besupported on flexible substrates: after 1,000 bending cycles, thecurrent measurements revealed no particular changes eitherfor the device with DAEs in the open form or for the devicewith DAEs in the closed form, proving the excellent dataretention characteristics.

By virtue of the photochromic DAE molecules dispersed in thepolymer matrix and acting as hole-accepting levels, the rate ofcharge carriers collected at the drain electrode over time can be pre-cisely tuned by the light intensity and duration at specific wave-lengths. This irradiation dose dictates the population of the openversus closed DAE isomers in the blend, therefore allowing toexactly control the device output current. With this set-up we canrealize a cognitive device, in which light irradiation triggers thelearning process, and the stability of the particular bit of infor-mation is ensured by the long retention times of the molecularDAE system.

Results and discussionThree-terminal memory devices on SiO2. We first exploredbottom-gate bottom-contact field-effect transistors using a SiO2

layer as the gate dielectric (Fig. 1a). The bicomponent active layercomprising the photochromic DAE moiety in a P3HT polymermatrix was spin-coated from a solution in chloroform and thenelectrically characterized in a nitrogen-saturated atmosphere in aglovebox. We have also tested several combinations of p-typematrices other than P3HT, such as polyfluorene- and isoindigo-based polymers, but found that the quantum yield of theisomerization reaction was poor when compared to that obtainedin P3HT (data not shown here). To confer a high startingcrystallinity to the polymer active layer and enhance its electricalperformances, octadecyltrichlorosilane (OTS) chemisorbed self-assembled monolayers were formed on a SiO2 dielectric surface48.OTS-treated devices featured smooth DAE-Me/P3HT films with aroot-mean-square roughness (Rrms) of 0.54 nm, reduced from0.74 nm for films supported on pristine silicon dioxide (calculatedfrom the atomic force microscopy data plots in Supplementary

Section 5). Films on OTS-treated silicon oxide resulted incurrents increased by one order of magnitude. Transfer curves(IDS–VGS) were used to quantify the devices’ field-effect mobility(µFET), threshold voltage (Vth) and Ion/Ioff ratio. Examples ofoutput curves are provided in Supplementary Section 2. Theelectrical characterization revealed devices featuring classicalp-type transfer characteristics and high field-effect mobilities forP3HT (0.01 cm2 V−1 s−1, Supplementary Section 2). The currentratio measured in the active layer comprising the photochromicunit in the open or the closed form (‘switching ratio’) wasextracted from the IDS–VGS curves. In particular, IDS was measuredin the dark once the DAE isomer in the bicomponent film hadbeen turned into its open (DAE-Me_o) or in a majority into itsclosed (DAE-Me_c) isomer upon irradiation with monochromaticlight at λ = 546 nm (‘erase’) or 313 nm (‘write’), respectively. Tobetter compare our results with the scientific literature, we used theratio IDS,DAE-Me_o/IDS,DAE-Me_c (measured at the same VDS and VGS)as a figure of merit. As is evident from Fig. 1b, a threshold voltageshift towards negative bias (ΔVth of approximately −15 V) wasobserved when the DAE molecules were converted from their opento their closed isomer. The resulting difference in threshold voltagebetween the open and closed forms enables a switching ratio of >105

for Vth(open form) <VGS <Vth(closed form) (Fig. 1c). Nevertheless,in contrast to single-level memories, their multilevel counterpartsrequire fine-tuning of each current level. Accordingly, to increasethe reliability of the device, the ‘dynamic range’ between the extremeoff and on levels (IDS,DAE-Me_o – IDS,DAE-Me_c at fixed VDS and VGS)should be maximized. Hence, further measurements wereperformed at VGS = −60 V.

Different amounts of DAE added to P3HT (wt%) were tested toexplore the effect of the mutual ratio on the electrical performancesand switching capabilities (Fig. 1d). We found that the mobilitiesdecreased when the amount of DAE in the blend was increased(both in the open and closed forms). The highest mobilities werefound at 2 wt% DAE-Me in the blend, with a small yet significant15% decrease in mobility following irradiation with ultravioletlight. The amplitude of the light-induced switching process (interms of µFET) of the devices was improved when a greateramount of DAE was introduced into the blend spin-coated ontothe device. As mentioned above, although a high switching ratiois preferable (as was obtained with DAE-Me and P3HT blendedin equivalent amounts), it is crucial to preserve a high electricalperformance to achieve multilevel storage with a high number ofdistinguishable current levels. A good compromise betweenmobility and switching ratio was found with 20 wt% DAE in theblend. The difference in device threshold voltage between openand closed forms remained constant (ΔVth = Vth[DAE-Me_o] −Vth[DAE-Me_c] ≈ +15 V) on increasing the quantities of DAE-Me inthe blend (see Supplementary Fig. 3a for values). This differencecan be ascribed to the change in the number of trapped chargesfrom open to closed forms due to the presence of the hole-acceptinglevels provided by the DAE-Me in the closed form. Furthermore, ageneral threshold voltage shift can be observed following theaddition of DAE-Me for both forms. Although DAE-Me in itsopen form can only act as a scattering centre for charges, itspresence will occupy an increasingly effective area in the channelwith increasing wt%. The associated reduction of the crystallineportions of the film will therefore give rise to this shift. ForDAE-Me in its closed form, the increasing threshold voltage stemsfrom the increased amount of charge-trapping sites within the film.

As long as the quantum isomerization yield of the photochromiccomponent is high enoughwithin a givenmatrix, our blending approachis versatile as it allows for tailoring to any desired DAE/polymer pairdepending on the application at hand and the materials in use34,45–47,contrary to the separate layer approach that is typically used in thecontext of organic two- and three-terminal memories43,44.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.87 ARTICLES





The endurance of the semiconductor–DAE blend was tested bydynamic write–erase cycles at constant bias. For this, and throughoutthis Article, continuous measurement of the IDS current as a functionof time will be referred to as dynamic characterization. To determinethe sole contribution of the DAE to the change in IDS, the decayobserved due to bias stress on P3HT was estimated from the curvebefore illumination and deduced from the IDS–time plot. Additionalexplanations and fitting parameters are provided in SupplementarySection 3. After over 70 write–erase cycles (10 s cycles after 20 minof measurement without illumination), the switching fatigue wasfound to be negligible. This is a good indication of the stability ofthe devices as they can undergo a large number of write–erasecycles with a minor impact on their switching ability (Fig. 2a,b).The retention capabilities of the devices were tested by measuringIDS at a fixed VDS of −10 V and VGS of −60 V in the open andclosed forms after different storage times in the dark. Only negligiblemodification of the current was found after 500 days in the dark,providing clear evidence for the excellent retention capability (Fig. 2c).

The readout accuracy of the memory was evaluated by means ofcurrent measurements in the dark with switching cycles performedwith a monochromator set-up (Fig. 2d). Each of the current levels1, 2, 3 and 4 was reached with a different irradiation time at 313 nm,and level 0 was recovered by irradiation at 546 nm. The standarddeviation was calculated for each level and compared to the totalcurrent difference between the memory with DAE in the openform and the memory with DAE in the closed form (for the values

see Supplementary Table 3). For each level, the standard deviationwas below 0.5% of the total gap (80 μA), thus demonstrating excellentreadout accuracy.

Switching with ultrafast light irradiation. A laser set-up withultrashort time pulses was used to investigate the photoinducedswitching with illumination times associated with modern optics andelectronics. The device was exposed to a 3 ns laser pulse (λ = 313 nm)at tunable intensity at intervals of 10 s, leading to a progressivecurrent decrease controlled by the number of pulses. Each pulsegenerated a ΔIDS step decrease due to the increasing number ofphotogenerated hole-accepting levels, that is, DAE-Me_c. Eachcurrent step had a certain height, and this allowed us to estimate thesignal-to-noise ratio (SNR) calculated from the r.m.s. amplitude ofthe noise over 3 × 80 consecutive points (Supplementary Section 6).Within the first current levels the SNR exceeds 50, and this decreasesto 2.8 in the final steps. The decrease results from the gradualreduction of the step height with a noise value that remains constant.We have successfully built a 400-current-intensity-level device (Fig. 3),achieving a data storage capacity of over 8 bit (256 levels). By reducingthe light intensity using filters, we have shown that the step in heightis dependent on the areal power density (Supplementary Section 4),because the number of molecules undergoing photoisomerizationis proportional to the density of the photons hitting the devicearea. In particular, our device can already operate at an incidentpower <1 μJ cm−² s−1. The number of steps and hence the storage

1,5301,200

0.01 0.1 1

Time (days)

10 100 1,000 0 5

546 nm 313 nm

Level 0

Level 1

Level 2

Level 3

Level 4

10Step

15

1,500 1,800 1,560

400

0

−400

ΔID

S (nA

)

700

−40

−120

−100

−80

−60

−35

−30

−25

−20

−15

−10

0

−5

0

−700

ΔID

S (nA

)I D

S (μA

)

I DS (

μA)

Time (s)Time (s)

DAE-Me_oDAE-Me_c

a b

c d

Figure 2 | Data retention of the memory device. a,b, Experimental dynamic IDS–time curves, corrected for bias stress, at VDS = −10 V and VGS = −60 V,with writing for 1 s at 313 nm and erasing for 5 s at 546 nm. A total of 70 cycles are shown in a, and only 5 in b. ΔIDS is defined as IDS − IDS,t(initial).c, Measurement of IDS after different times of storage of the device in the dark at fixed VDS = −10 V and VGS = −60 V to illustrate the retention capability ofthe devices in the open and closed forms. d, Static switching cycles at fixed illumination times. The current was measured for a single point in thedark at VGS = −40 V and VDS = −10 V. Level 0 → Level 1: 10 s at 313 nm; Level 1 → Level 2: 20 s at 313 nm; Level 2 → Level 3: 30 s at 313 nm;Level 3 → Level 4: 60 s at 313 nm; Level 4 → Level 0: 615 s at 546 nm.






density is therefore only limited by the noise in the signal (leadingto a maximum of 5,000 steps in the presented experiment) with atheoretical maximum given by the number of steps equal to thenumber of DAE molecules in the channel.

Three-terminal device memories on a flexible polyethyleneterephthalate (PET) substrate. In view of the ultimate goal ofdemonstrating the possible implementation of the memory

devices in novel electronic applications, the entire procedure(other than the OTS treatment) was transferred to a flexiblesubstrate, trading the rigid inorganic silicon-based layers forpolymeric sheets of PET, poly(methyl methacrylate) (PMMA) andpoly(4-vinylphenol) (PVP) (Fig. 4a). The bottom gate configurationwas selected again to benefit from unfiltered irradiation of the activelayer. Although the PMMA dielectric ensured low current leakage,the PVP layer deposited on top made spin-coating of the

−984

−972

1,000−400

−500

−600

−700

−800

I DS (

nA)

I DS (

nA)

−786

−780

I DS (

nA)

−470

−468I DS (

nA)−586

−584

I DS (

nA)

−900

−1,000

2,000 3,000

450 490

4,000

Time (s)

Time (s)1,200 1,230

Time (s)

3,620 3,650Time (s)

2,750 2,780Time (s)

Figure 3 | IDS–time curves (corrected for bias stress) with irradiation of the device with 3 ns laser pulses (313 nm) every 10 s, underlining the ability tobehave as a multilevel memory. Curves were plotted at VGS = −60 V and VDS = −10 V. Current spikes in the curve are attributed to photogenerated carriersduring irradiation. Given their short duration (3 ns), illumination steps are not represented to scale.

Bending cycles1,8001,200

Time (s)

Time (s)

600

−13.5

−3.25

−3.30

−3.35

−3.40

−13.0

−12.5

−11.5

−12.0

6004002000

0.6

0.8

1.0

I DS/I D

S,0

I DS (

μA)

I DS (

μA)

800 1,000

DAE-Me_oDAE-Me_c

1,000500

a b

c d

1,500

P3HT + DAE-MeAu source, drain

PVPPMMAAu gate

PET substrate

Figure 4 | Integration of the memory unit onto a flexible PET substrate. a, Structure of the devices on PET. b, Experimental dynamic IDS–time curves,corrected for bias stress, with write–erase cycles (313 nm for 20 s/546 nm for 140 s) at VGS = −60 V and VDS = −10 V. c, IDS–time curve corrected for biasstress showing small irradiation steps at 313 nm, underlining the ability to behave as a multilevel memory (5 to 20 s incremental irradiation steps). d, Draincurrent as a function of bending cycles involving devices based on either DAEs in the closed form (after 10 min irradiation at 313 nm) or in the open form.Each cycle consisted of 5 s of bending at a bending radius of 8.0 mm, followed by 5 s of resting time.






chloroform-based solution possible. The devices were electricallycharacterized and the field-effect mobilities were comparable tothose on SiO2 without OTS (over 10−3 cm² V−1 s−1). The switchingcapabilities were tested over multiple dynamic switching cycles.We found that the devices could be switched between closed andopen states, albeit at a slower pace than devices on SiO2 (Fig. 4b).The small increases in drain current during switching areattributed to photogenerated carriers. Dynamic IDS–time curveswith small illumination steps at 313 nm indicate the ability of thedevice to work as a 4 bit multilevel memory (16 distinct levels,Fig. 4c). Furthermore, electrical characterization was performedon the device after successive bending cycles (Fig. 4d). Over 1,000cycles, the current measurements revealed no particular changes,either for the device with DAEs in the open form or for thedevice with DAEs in the closed form. Longer bending steps atdifferent bending radii (29.0, 17.5, 10.0, 8.0 and 6.0 mm) wereperformed. Post-bending performances were still comparable tothose observed before bending at each bending radius for 60 s(Supplementary Fig. 5), with only minor losses in terms of chargecarrier mobility and a slight shift in the threshold voltage, provingthat with further optimization, the studied organic three-terminalmemory devices could be introduced into flexible electronics.

ConclusionsWe have shown that it is possible to build three-terminal memorydevices with unprecedented features. Above all, the ability todefine many distinct current levels paves the way for creatinghigh-density memories with 8 bit memory units (256 levels). Theillumination times using a laser set-up are on a nanosecond time-scale, giving rise to a fast response that is necessary for real-life elec-tronics. It has been shown that this approach can also be successfullyapplied on flexible substrates. Furthermore, the excellent data reten-tion capabilities of the devices (retention times of 107 s) and thereliability after many write–erase cycles make this method idealfor future applications in (flexible) electronics. Due to the versatilenature of our blending approach, the air-sensitive P3HT could bereplaced by an alternative air-stable p-type semiconductor withsimilar energy levels, leading to potentially higher currents (hencean increased possible number of steps) without the need for encap-sulation. Alternative polymer/DAE pairs should be selected inaccordance with their respective energy levels, ideally placing theHOMO level of the closed DAE as deep as possible within thebandgap of the polymer while the HOMO level of the open DAEshould be outside the bandgap of the polymer. Furthermore, theinterplay between small molecule and polymer should be con-sidered carefully to avoid aggregation of the DAE. Both the energeticlevels and the phase segregation can be tuned by functionalization ofthe photochromic units. Moreover, attaining multilevel memoryrequires the use of a semiconducting (macro)molecule exhibitingmobilities of >10−3 cm2 V−1 s−1 to ensure a detectable photomodu-lated change in the device current after hundreds of cycles.Furthermore, although in principle the optimal scenario consistsof DAE/semiconductor pairs with orthogonal absorption bands,in our case the use of films with a thickness of just a few tens ofnanometres makes this requirement less stringent, even when lowirradiation doses are used.

Our findings are of paramount importance for the realization oflight-controlled cognitive nanodevices where the irradiation historyrepresents the learning process and the non-volatile state retentionensures that the data are preserved over time in a given memorystate. The possibility of developing such devices on a flexiblesubstrate and to write using ultrashort low-intensity laser pulses isof great importance for the development of high-performingoptically gated electronic nanodevices for potential applications inmemories, logic circuits and more generally in optoelectronicsand optical sensing.

MethodsMethods and any associated references are available in the onlineversion of the paper.

Received 26 September 2015; accepted 26 April 2016;published online 20 June 2016

References1. Kelley, T. W. et al. Recent progress in organic electronics: materials, devices, and

processes. Chem. Mater. 16, 4413–4422 (2004).2. Braga, D. & Horowitz, G. High-performance organic field-effect transistors. Adv.

Mater. 21, 1473–1486 (2009).3. Han, S. T., Zhou, Y. & Roy, V. A. L. Towards the development of flexible non-

volatile memories. Adv. Mater. 25, 5425–5449 (2013).4. Benight, S. J., Wang, C., Tok, J. B. H. & Bao, Z. A. Stretchable and self-healing

polymers and devices for electronic skin. Prog. Polym. Sci. 38, 1961–1977 (2013).5. Pang, C., Lee, C. & Suh, K. Y. Recent advances in flexible sensors for wearable

and implantable devices. J. Appl. Polym. Sci. 130, 1429–1441 (2013).6. Baeg, K. J., Caironi, M. & Noh, Y. Y. Toward printed integrated circuits based

on unipolar or ambipolar polymer semiconductors. Adv. Mater. 25,4210–4244 (2013).

7. Ahmad, S. Organic semiconductors for device applications: current trends andfuture prospects. J. Polym. Eng. 34, 279–338 (2014).

8. Chiu, Y. C. et al. High-performance nonvolatile organic transistor memorydevices using the electrets of semiconducting blends. ACS Appl. Mater. Inter. 6,12780–12788 (2014).

9. Smithson, C. S., Wu, Y., Wigglesworth, T. & Zhu, S. A more than six orders ofmagnitude UV-responsive organic field-effect transistor utilizing abenzothiophene semiconductor and Disperse Red 1 for enhanced chargeseparation. Adv. Mater. 27, 228–233 (2015).

10. Yoon, S. M. et al. Fully transparent non-volatile memory thin-film transistorsusing an organic ferroelectric and oxide semiconductor below 200 °C. Adv.Funct. Mater. 20, 921–926 (2010).

11. Radha, B., Sagade, A. A. & Kulkarni, G. U. Metal–organic molecular device fornon-volatile memory storage. Appl. Phys. Lett. 105, 083103 (2014).

12. Liu, X. H. et al. The effect of oxygen content on the performance of low-voltageorganic phototransistor memory. Org. Electron. 15, 1664–1671 (2014).

13. Nougaret, L. et al. Nanoscale design of multifunctional organic layers for low-power high-density memory devices. ACS Nano 8, 3498–3505 (2014).

14. Lee, S. et al. Overcoming the ‘retention vs. voltage’ trade-off in nonvolatileorganic memory: Ag nanoparticles covered with dipolar self-assembledmonolayers as robust charge storage nodes. Org. Electron. 14, 3260–3266 (2013).

15. Kim, S. J. & Lee, J. S. Flexible organic transistor memory devices. Nano Lett. 10,2884–2890 (2010).

16. Cosseddu, P., Lai, S., Casula, G., Raffo, L. & Bonfiglio, A. High performance,foldable, organic memories based on ultra-low voltage, thin film transistors. Org.Electron. 15, 3595–3600 (2014).

17. Kim, R. H. et al. Non-volatile organic memory with sub-millimetre bendingradius. Nature Commun. 5, 3583 (2014).

18. Sekitani, T. et al. Organic nonvolatile memory transistors for flexible sensorarrays. Science 326, 1516–1519 (2009).

19. Liu, X. et al. Advancements in organic nonvolatile memory devices. Chin. Sci.Bull. 56, 3178–3190 (2011).

20. Bez, R., Camerlenghi, E., Modelli, A. & Visconti, A. Introduction to flashmemory. Proc. IEEE 91, 489–502 (2003).

21. Sala, F., Gabrys, R. & Dolecek, L. Dynamic threshold schemes for multi-levelnon-volatile memories. IEEE Trans. Commun. 61, 2624–2634 (2013).

22. Pirovano, A. et al. Reliability study of phase-change nonvolatile memories. IEEETrans. Dev. Mater. Res. 4, 422–427 (2004).

23. Nili, H. et al. Nanoscale resistive switching in amorphous perovskite oxide(a-SrTiO3) memristors. Adv. Funct. Mater. 24, 6741–6750 (2014).

24. Han, Y., Cho, K., Park, S. & Kim, S. Resistive switching characteristics of HfO2-based memory devices on flexible plastics. J. Nanosci. Nanotechnol. 14,8191–8195 (2014).

25. Zhang, W. B. et al. Thermally-stable resistive switching with a large ON/OFFratio achieved in poly(triphenylamine). Chem. Commun. 50,11856–11858 (2014).

26. Heremans, P. et al. Polymer and organic nonvolatile memory devices. Chem.Mater. 23, 341–358 (2011).

27. Chiu, Y. C. et al. High performance nonvolatile transistor memories ofpentacene using the electrets of star-branched p-type polymers and theirdonor–acceptor blends. J. Mater. Chem. C 2, 1436–1446 (2014).

28. Baeg, K. J. et al. High-performance top-gated organic field-effect transistormemory using electrets for monolithic printed flexible NAND flash memory.Adv. Funct. Mater. 22, 2915–2926 (2012).

29. Das, B. C., Pillai, R. G., Wu, Y. L. & McCreery, R. L. Redox-gated three-terminalorganic memory devices: effect of composition and environment onperformance. ACS Appl. Mater. Inter. 5, 11052–11058 (2013).








30. Kumar, R., Pillai, R. G., Pekas, N., Wu, Y. L. & McCreery, R. L. Spatially resolvedRaman spectroelectrochemistry of solid-state polythiophene/viologen memorydevices. J. Am. Chem. Soc. 134, 14869–14876 (2012).

31. Chiu, Y. C. et al. Multilevel nonvolatile transistor memories using a star-shapedpoly((4-diphenylamino)benzyl methacrylate) gate electret. NPG Asia Mater. 5,e35 (2013).

32. Chou, Y.-H., Chang, H.-C., Liu, C.-L. & Chen, W.-C. Polymeric charge storageelectrets for non-volatile organic field effect transistor memory devices. Polym.Chem. 6, 341–352 (2015).

33. Zhou, Y., Han, S. T., Sonar, P. & Roy, V. A. L. Nonvolatile multilevel data storagememory device from controlled ambipolar charge trapping mechanism. Sci. Rep.3, 2319 (2013).

34. Orgiu, E. & Samori, P. Organic electronics marries photochromism: generationof multifunctional interfaces, materials, and devices. Adv. Mater. 26,1827–1845 (2014).

35. Hayakawa, R., Higashiguchi, K., Matsuda, K., Chikyow, T. & Wakayama, Y.Optically and electrically driven organic thin film transistors with diarylethenephotochromic channel layers. ACS Appl. Mater. Inter. 5, 3625–3630 (2013).

36. Matsui, N. & Tsujioka, T. Carrier mobility of photochromic diaryletheneamorphous films. Org. Electron. 15, 2264–2269 (2014).

37. Tsujioka, T., Hamada, Y., Shibata, K., Taniguchi, A. & Fuyuki, T. Nondestructivereadout of photochromic optical memory using photocurrent detection. Appl.Phys. Lett. 78, 2282–2284 (2001).

38. Andersson, P., Robinson, N. D. & Berggren, M. Switchable charge traps inpolymer diodes. Adv. Mater. 17, 1798–1803 (2005).

39. Raimondo, C. et al. Optically switchable organic field-effect transistors based onphotoresponsive gold nanoparticles blended with poly(3-hexylthiophene).Proc. Natl Acad. Sci. USA 109, 12375–12380 (2012).

40. Taguchi, M., Nakagawa, T., Nakashima, T., Adachi, C. & Kawai, T. Photo-patternable electroluminescence based on one-way photoisomerization reactionof tetraoxidized triangle terarylenes. Chem. Commun. 49, 6373–6375 (2013).

41. Russew, M. M. & Hecht, S. Photoswitches: from molecules to materials.Adv. Mater. 22, 3348–3360 (2010).

42. Irie, M., Fukaminato, T., Matsuda, K. & Kobatake, S. Photochromism ofdiarylethene molecules and crystals: memories, switches, and actuators.Chem. Rev. 114, 12174–12277 (2014).

43. Shallcross, R. C., Korner, P. O., Maibach, E., Kohnen, A. & Meerholz, K.Photochromic diode with a continuum of intermediate states: towards highdensity multilevel storage. Adv. Mater. 25, 4807–4813 (2013).

44. Korner, P. O., Shallcross, R. C., Maibach, E., Kohnen, A. & Meerholz, K. Opticaland electrical multilevel storage in organic memory passive matrix arrays. Org.Electron. 15, 3688–3693 (2014).

45. El Gemayel, M. et al. Optically switchable transistors by simple incorporation ofphotochromic systems into small-molecule semiconducting matrices. NatureCommun. 6, 6330 (2015).

46. Borjesson, K. et al. Optically switchable transistors comprising a hybridphotochromic molecule/n-type organic active layer. J. Mater. Chem. C 3,4156–4161 (2015).

47. Orgiu, E. et al. Optically switchable transistor via energy-level phototuning in abicomponent organic semiconductor. Nature Chem. 4, 675–679 (2012).

48. Chua, L. L. et al. General observation of n-type field-effect behaviour in organicsemiconductors. Nature 434, 194–199 (2005).

AcknowledgementsThis work was supported financially by the EC through the MSCA-ITN project iSwitch(GA no. 642196) as well as ERC projects SUPRAFUNCTION (GA-257305) andLIGHT4FUNCTION (GA-308117), the Agence Nationale de la Recherche through theLabEx CSC (ANR-10-LABX-0026_CSC), the International Center for Frontier Research inChemistry (icFRC) and the German Research Foundation (via SFB 658).

Author contributionsP.S. and E.O. conceived the experiment and designed the study. M.H. and S.H. designed theDAEs and carried out their synthesis and electrochemical characterization. T.L., E.O., G.B.and E.P. designed and performed the time-response measurements. E.O. designed and T.L.performed the device experiments. T.L., E.O. and P.S. co-wrote the paper. All authorsdiscussed the results and contributed to the interpretation of data, as well as contributing toediting the manuscript.

Additional informationSupplementary information is available in the online version of the paper. Reprints andpermissions information is available online at www.nature.com/reprints. Correspondence andrequests for materials should be addressed to S.H., E.O. and P.S.

Competing financial interestsThe authors declare no competing financial interests.





http://www.nature.com/reprints



MethodsMemories on SiO2. Transistors with a bottom-contact bottom-gate configuration(IPMS Fraunhofer Institute) were used. These consisted of n++-Si substrates with230 nm of thermally grown SiO2 as the gate dielectric (15 nF capacitance) andprepatterned pairs of gold electrodes with interdigitated geometry as the source anddrain. All solutions, samples and devices were prepared and measured in a N2-filledglovebox to avoid oxidative doping of the materials and ensure reproducibility of theexperiments. OTS (Sigma-Aldrich) treatment consisted of a 12 h immersion of theozone-treated wafer in a 10 mM solution of OTS in toluene with 30 min of heating at60 °C at the beginning of the immersion. The samples were then rinsed with tolueneand heated for 60 min at 60 °C. Subsequently, 100 μl of the P3HT-DAE-Me_osolution was spin-coated onto the OTS-treated wafer at 1,500 r.p.m. (accelerationof 4,000 r.p.m. s–1). P3HT (50 kDa; dispersity 2.4) synthesized by BASF(commercialized under the name ‘Sepiolid P200’) was purchased and used asreceived without further purification. The DAE-Me/P3HT solution consisted of ablend of 0.4 mg DAE-Me_o and 1.6 mg P3HT in 1 ml chloroform.

Memories on PET. A square PET film (thickness = 175 μm) was attached to a glasssubstrate using thermal tape, and 60 nm Au was subsequently evaporated on top ofthe PET to form the gate electrode. One side of the sample was protected from thenext steps of the procedure with a small strip of thermal tape. A layer of the dielectricpoly(methyl methacrylate) (PMMA) was spin-coated onto the gate electrode, andannealed for 1 h at 120 °C. An additional dielectric, poly(4-vinylphenol) (PVP) wasspin-coated on the PMMA and annealed for 2 h at 150 °C. The interdigitated goldsource and drain electrodes were vacuum-deposited on the dielectric through shadowmasks (layer thickness = 60 nm). Finally, the semiconducting P3HT/DAE blend wasspin-coated as the final layer of the sample. The DAE-Me_o/P3HT solutionconsisted of a blend of 0.4 mg DAE-Me_o and 1.6 mg P3HT in 1 ml chloroform.

Device characterization. The device characteristics were measured by contactingthe source, drain and gate electrodes and applying different voltages to obtainIDS–VGS graphs. Measurements were all performed using Keithley devices and thesoftware Labtracer. Each IDS–VGS graph (transfer curve) was constructed bysweeping the gate voltages from +40 V to −60 V, with one measurement every 2 V.The drain voltage was either −10 V (linear regime) or −60 V (saturation regime). IDSwas measured each time. Using the software Origin pro 8.0, transfer curves wereplotted and fitted to extract the value of the mobility and the threshold voltage. Theoutput curves were used to extract the value of the Ion/Ioff ratio by dividing the valueof the observed current in the saturation regime when the transistor is ‘on’(VGS = −60 V) by the value of the current when the transistor is ‘off’(VGS = +40 V).

Experimental data were analysed using standard field-effect transistorequations (for p-type semiconductors) for the saturation regime:

IDS = −W2L

μCTOT(VGS − VTh)2

The mobility of samples with two dielectrics (PVP and PMMA) was calculated usingthe value of the total capacitance, given by CTOT = (CPMMA × CPVP)/(CPMMA + CPVP).The thickness of each layer was measured using a profilometer.

Dynamic switching was observed by tracing the IDS–time curves at constant VGS

and constant VDS. The samples were illuminated while biased, and IDS wasmeasured. Illumination was performed using either a laser set-up (85 ± 17 μJ cm−²for each 3 ns pulse, memories on SiO2) or a monochromator set-up (74 μJ cm−² persecond of irradiation at 313 nm, memories on PET and static switching cycles ofmemories on the SiO2 substrate).


NATURE NANOTECHNOLOGY | www.nature.com/naturenanotechnology




flexible non-volatile optical memory thin-film transistor ... · principle of a ﬂoating gate...

Documents