Huge field-effect surface charge injection and conductance modulation in metallic thin films by electrochemical gating

0.34 MB·English

by  M. Tortello

Huge field-effect surface charge injection and conductance modulation in metallic
thin films by electrochemical gating
M.Tortelloa, A. Solaa, Kanudha Shardaa, F. Paoluccia,1, J.R. Naira, C. Gerbaldia, D. Dagheroa, R.S. Gonnellia,∗
aDipartimento di Scienza Applicata e Tecnologia, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
3
1Abstract
0
2 The field-effect technique, popular thanks to its application in common field-effect transistors, is here applied to

nmetallic thin films by using as a dielectric a novel polymer electrolyte solution. The maximum injected surface charge,
adetermined by a suitable modification of a classic method of electrochemistry called double-step chronocoulometry,
Jreached some units in 1015 charges/cm2. At room temperature, relative variations of resistance up to 8%, 1.9% and

61.6% were observed in the case of gold, silver and copper, respectively and, if the films are thick enough (> 25 nm),
1
results can be nicely explained within a free-electron model with parallel resistive channels. The huge charge injections

achievedmakethis particularfield-effect technique verypromisingfor a vast varietyof materialssuchas unconventional
]
lsuperconductors, graphene and 2D-like materials.
l
a
h
-
s Keywords: field-effectexperiments;electrochemicalgat- a surface charge nmax of the order of 1013 charges/cm2
2D
e
ing; surface electron states; conductivity of metals while withpolymericgatingtechniquesthis valuecaneas-
m
ilyreachsomeunits in1014 charges/cm2 asaconsequence
t.1. Introduction of electric fields as high as 100 MV/cm. By using a suit-
a
ablePES,Dhootandcoworkershaverecentlyachievedthe
m
One of the milestones of the technological revolution unprecedentedsurfacedensityofinjectedchargeof2 1015
-that has occurred in the last 50 years is the invention of charges/cm2 [5]. The electrochemical gating techni×que is
d
nthe field effect transistor (FET). As it is well known, the based on the formation of an electric double layer (EDL)
oworking principle of FET devices is based on the modula- betweenanelectrolytesolutionandthesurfaceofthesam-
ction of the transport properties of a material by means of ple under test: the polymer solvates positive andnegative
[
an applied electric field. This is important not only from ions in the electrolyte solutionandanexternalbias drives

1a technological point of view but also from a fundamen- them towards the oppositely polarized surface of the film
vtal one. Indeed, field-effect experiments allow tuning the or gate, thus forming the EDL. Therefore, the EDL acts
9
chargedensityofamaterialwithoutthesideeffectstypical asaparallel-platecapacitorwithextremelysmalldistance
6
of chemical substitutions or application of pressure, such between the plates (of the order of the polymer molecule
7
3as introduction of disorder or modification of the lattice size) [7] and thus a very large capacitance.
.structure. In this regard field-effect experiments allowed, Asalreadystatedabove,field-effectexperimentshavebeen
1
0for instance, enhancing the critical temperature of some performedinsomeexoticmaterialsinordertoinducedra-
3superconductors[1,2,3],inducingmetallicbehaviorinin- maticmodificationsoftheirpropertiesorevenphasetran-
1sulators [4], a metal-to-insulator transition in a colossal sitions [3]. However, although poorly studied, field-effect
:
vmagnetoresistivemanganite [5] or evena superconducting measurements have been carried out also in metals. The
Xiphase transition in materials like SrTiO3 [6], ZrNCl [7], little interest in this topic is mainly due to the fact that
and KTaO3 [8]. In order to achieve large effects, a high it seems not to be so much attractive from an applica-
r
aamount of charge has to be injected and therefore very tivepointofview,butalsobecausetheeffectiscommonly
intense electric fields are required. Instead of using the consideredtobedifficult,oralmostimpossible,toobserve:
standard solid dielectric which is nowadays a very com- indeed, the electronicscreeninglengthinthe semiclassical
monsolutionfor commercialdevices,the use ofapolymer model for a metal is less than one atomic radius. Nev-
electrolyte solution (PES) [4, 5, 6, 7, 8, 9, 10] is a very ertheless, field-effect induced modulations of the conduc-
promising technique to achieve these huge charge injec- tivity in metals have already been observed in the past
tions. Withasolid-dielectricdeviceitispossibletoinduce [1, 11, 12, 13, 14, 15, 16], also justifying a fundamental
interest in this subject.
∗Correspondingauthor Here we report on field-effect experiments performed on
Email address: [email protected] (R.S.Gonnelli) gold, silver and copper by means of the electrochemical
1NowatMaxPlanckInstituteforSolidStateResearch,Stuttgart gating technique. The novel PES we adopted allows sur-
(Germany).
Preprint submittedto Elsevier January 17, 2013
7 110
6 100
(a) (b) 5 90
m4 7800 nm
m 3 60
(cid:21)(cid:12)(cid:4)(cid:4)(cid:14)(cid:9)(cid:15) 2 50
(cid:22)(cid:10)(cid:12)(cid:4)(cid:13)(cid:14) (cid:11)(cid:10)(cid:12)(cid:4)(cid:13)(cid:14) 1 40
30
0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
0 1 2 3 mm4 5 6 7 mm
(cid:6) (cid:1)(cid:0)(cid:2)
(cid:5)
(cid:6)(cid:3) Figure 2: Left: an example of AFM image of the surface of a Ag
(cid:16)(cid:7)(cid:15)(cid:14)(cid:17)(cid:10)(cid:18)(cid:15)(cid:7)(cid:19)(cid:14) film. The flat right-hand side of the image is the glass substrate.
(cid:0)(cid:4)(cid:7)(cid:8)(cid:9)
Thesharpedgeshownherewasobtainedbyscratchingthefilm. The
white straight line indicates the cut along with the z-profile shown
intherightpanelwasmeasured. Thisprofileindicatesathicknessof
65±7nm. Theeffective thickness ofthisfilmthatresultsfromthe
correction for voids as described in ref.[17] and in the text is equal
(cid:11)(cid:10)(cid:12)(cid:4)(cid:13)(cid:14) (cid:16)(cid:7)(cid:15)(cid:14)(cid:20)(cid:7)(cid:23) to30nm.
(cid:6)
(AFM) which is also used to investigate the morphology
(c) (cid:0)(cid:4)(cid:7)(cid:8)(cid:9) of the film surface, together with field-emission scanning
electronmicroscopy(FESEM)[16]. Toaccountforthepos-
sible presence of voids between grainsthat canreduce the
effective cross sectional area of the film, we also applied
a correction proposed by Rowell [17] and based on the
fact that the temperature dependence of the resistivity
Figure 1: a) Scheme of devices for electrochemical gating: the por- depends only on the material and not on the form of the
tionof thefilmbetween thevoltage contacts iscovered bythePES
sample (bulk, polycrystal, film). The procedure requires
thatispolarizedbyavoltagesource(Vgate). Thefilmresistivityis
measuredinthestandardfour-probeconfiguration. (b)Cross-section comparing the experimental ρ(T) of a given film with the
sketchofthedevice: whenVgateisapplied,ionsareaccumulatedat resistivityofthepure(bulk)material,andprovidesascal-
theinterfacewiththemetallicfilmandachargewithoppositesignis ing factor F for the geometricalcrosssection. Inmost Au
inducedonitssurface,forminganEDL.(c)Photographofatypical
films, we found out that F 1, i.e. there is no need of
deviceusedinourfield-effectexperiments onmetallicthinfilms. ≃
correctionandthegeometricalcrosssectioncoincideswith
the effective one. In all the Cu and Ag films, instead, the
face charge injections as high as 4.5 1015 charges/ cm2 correction was necessary, which is consistent with AFM
·
[16](thus even larger than those obtained by Dhoot et al. and FESEM images that show a greater roughness. An
[5]) and far-from-negligible modulations of the conductiv- example of AFM image on a Ag film and the correspond-
ity wereobservedin allthe investigatedmetals. In partic- ing z-profile to determine its thickness is shown in Figure
′
ular, relative variations of the resistivity ∆R/R up to 8 2.
%, 1.9% and 1.6 % were achievedat room temperature in Voltage, current and gate pads were deposited on top
gold, silver and copper, respectively. If films are not too ofthefilmbyevaporatinggold,independentlyofthemetal
thin,suchastoavoidadominantroleofsurfacescattering, under study. As a substrate, we adopted either glass, or
′
the trend of ∆R/Rt as a function of n2D (where t is the amorphous SiO2 or Si3N4 on a Si wafer.
thickness of the film) can be nicely explained within the Themetallicthinfilmbehavesasthechannelofastan-
free-electronmodel. Moreover,thehugesurfacechargein- dard FET whose resistivity is measured with a standard
jections achieved make this technique promising for many four-probe configuration by inverting the current during
other materials such as graphene, unconventional super- each measurement or by using the ac technique. The
conductors and 2D-like crystals. PES acts as the dielectric and is therefore connected on
one side to the metallic film and on the other side to
the gate pad. A bias is applied to the gate by means
2. Experimental setup
ofa source-measureunit that, atthe sametime, measures
Figure 1(a) shows a sketch of the experimental setup. the electric current flowing through the PES. The PES
The device, fabricated in the planar configuration as in we used was obtained by a reactive mixture of bisphe-
ref.[10], consists of the metallic thin film under study, the nolAethoxylate(15EO/phenol)dimethacrylate(BEMA;
gate pad and the PES. Thin films are obtained by phys- average Mn: 1700, Aldrich), poly(ethylene glycol)methyl
ical vapor deposition (PVD) of the selected metal (gold, ether methacrylate (PEGMA; average Mn: 475, Aldrich),
silver or copper) at a pressure P 2 10−5 mbar. Their and lithium bis(trifluoromethanesulfonyl)imide (LiTFSI)
∼ ·
thicknessismeasuredbymeansofatomicforcemicroscopy inthepresenceof3%wtofa2-hydroxy-2-methyl-1-phenyl-
2
1.5
1-propanonfreeradicalphotoinitiator(Darocur1173,Ciba e (V) 1.0 (a) Gate voltage
g
a
SpecialtyChemicals). ThequantitiesofBEMAandPEGMA olt 0.5
v
are ina 3:7 ratio,andthe LiTFSI is the 10%wt ofthe to- e
at 0.0
tal compound. The PES was then polymerized by UV G
exposure using a medium vapor pressure Hg UV lamp, A) 0.004 (b) Gate current
w30itmhWa r/acdmia2t.ioAnllinthteenasibtoyvoenopthereastuiornfascweeorfetpheerfsoarmmpelde ionf ment ( 0.002
the controlled Ar atmosphere of a dry glove box with O2 e curr -00..000020
and H2O content < 0.1 ppm. Since the PES used here Gat
was originally developed for Li-ion battery applications, (c) Charge
1.5
the reproducibility of its chemical and physical properties C)
waswidelytestedbymeansofconductivitymeasurements, me ( 1.0
crosslinkdensitymeasurements,thermalhistoryandcyclic arg
h 0.5
voltammetry [19]. A careful preparation according to the C
abovereciperesultsinaveryhighreproducibilityandsta-
) (d) Resistance
bility of the PES over long times (several months). We ( 3.778
Figure1(b)showshowthefield-effectexperimentswork: c
Theapplicationofthegatevoltagecausesanaccumulation stan 3.776
si
of charge at the interface between the film and the PES, e 3.774
R
thus forming the EDL. A symmetric amount of charge is
0 1000 2000 3000 4000
induced at the gate pad, shown on the right hand-side of
Time (s)
Fig. 1(b). Figure1(c)showsaphotoofarealdeviceused
in these experiments. Figure 3: a) Gate voltage vs time. b) Current flowing through the
PESduringthe application and removal of gate voltage. c) Charge
movedduringthemeasurement, obtainedbyintegrationofthe cur-
3. Experiment and models rentshowninb). d)Effectofapplicationofthegatevoltageonthe
thinfilmresistance.
Inthissectionwedescribethemeasurementsperformed
on the devices previously shown, as well as the models
carefully determine the surface charge injected in the film
usedtodeterminetheinjectedsurfacechargeandtoquan-
by the polymer gating and then to calculate the effect of
tify the effect of surface doping on the conductivity of the
this charge on the transport properties of the whole film.
metallic thin films.
Both these topics are addressed in the following.
Thesequenceofoperationsandmeasurementsperformed
Infield-effectexperimentsperformedonverythinfilms
on a single device is summarized in Figure 3. At t = 0
of materials less conductive than metals a direct estima-
a voltage is applied between the gate pad and the film.
tionofchargeinjectioncanbedonebymeansofHall-effect
After a proper time (of the order of several hundreds of
measurementat the sample surface. In metallic thin films
seconds due to the rather slow dynamic of the PES) this
thisisnotpossibleasitwouldrequirehugemagneticfields
voltage is removedas shownin Figure 3(a). A sharppeak
because of the high intrinsic carrier density of the metal
in the gate current flowing through the PES appears im-
and the considerable thickness of the film (here of the or-
mediately after the application of gate voltage, indicating
derofsometensofnanometers). Intheabsenceofadirect
the formation of the EDL. This current decreases almost
method for estimating the total charge on the side of the
exponentially and after a few seconds it reaches a very
film we must rely on the determination of the additional
small constantvalue. When the gate voltage is removeda
chargeinduced onthe film surfacethroughthe quantifica-
similar peak appears with negative values of the current,
tionofthe EDL chargeonthe PESside. A roughintegra-
indicating the disruption of the EDL (Figure 3(b)). The
tionofthecurrentflowingthroughthe PESwouldgivean
time integralof this gate current as function of time (Fig-
overestimation of the charge that forms the EDL. In fact,
ure 3(c)) provides the total amount of charge that flows
determiningthechargeoftheEDLbyintegratingthegate
throughthe PESand,byapplyingaproperprocedure,al-
currentisnotcorrectifelectrochemicaleffectsarepresent,
lowsextractingthe(smaller)chargethatisinvolvedinthe
as pointed out in Ref. [18]. Indeed, in the currentvs time
formation of the EDL. Finally, this dynamical formation
graph it is possible to observe a slowly vanishing current
anddestructionoftheEDLwiththeconsequentchargein-
that always persists long after the application of the gate
jection atthe surface ofthe metallic thin film givesrise to
voltage (see Fig. 3(b)). This current is due to the flow of
relative changes in the film resistance of the order of10−3
charges necessary to maintain the gradient of ion concen-
as shown in Figure 3(d). As expected, the application of
trationwhendiffusionofelectroreactants[20]ortunneling
a positive bias to the gate induces a negative charge at
effects through the EDL [18] take place. However, being
the film surface (see Fig. 1(b)) giving rise to a reduction
severalordersofmagnitudesmallerthanthedirectcurrent
of the total resistance of the film. In order to explain the
flowing between source and drain, this gate current does
experimental results just described, it is first necessary to
3
not affect the resistance measurements.
1.5 1.6
Asuitablemethodofelectrochemistryusedtoevaluate 3 Qd 8.0 Qd 1.4
only the charge that forms the EDL (which, in principle, C) 1.0
cseoprraersaptoinngdsittforotmhethcehachrgaergienjreeclatetdedintotohtehemrepthaellnicomfilemna),, mCharge ( 12 Qc 77..05 0.5 Qc 011...802
is called double-step chronocoulometry [20]. The start- Ag 1V gate 6.5 Au 1V gate
0 0.0 0.6
ing point is the time dependence of the calculated charge 0 2 4 6 8 10 12 0 2 4 6 8 10 12
Qde(nt)ce=caRn0tbIGe(atn′)adlyt′zeshdobwonthinduFriign.g3th(ce).apTphliicsatQio(nt)adnedptehne- 20 t 0.5 Qd 2267 56 t 0.5 Qd 910
removal of gate voltage, giving two complementary pieces C) 15 25 4 8
ofinformationaboutthe chargeinthe EDL.The shape of mge ( 10 24 3 7
ar 23 2 Q
Q(t) showsindeed that two phenomena occur on verydif- Ch 5 Qc 22 1 c 6
ferenttimescales: aratherfastEDLcharging/discharging Ag 2V gate Au 2V gate
0 21 0 5
(thatisexpectedtoshowanexponentialtimedependence) 0 2 4 6 8 10 12 0 2 4 6 8 10 12
t0.5 t0.5
andothereffectsofelectrochemicalnaturethatshouldgive
a √t dependence [20]. By plotting Q as function of √t
Figure4: Fourdifferentexamplesofchronocoulometryplotsfortwo
∗
(see Fig. 4) it is very easy to determine the instant t differentmaterialsandtwodifferentgatevoltages. Ineachplot,thick
at which the √t behavior becomes dominant, i.e. when lines represent the electric charge obtained by integrating the gate
current in the charging (green bottom line) and discharging (blue
the exponential behavior has become negligible. We as-
top line) phases. Please note the different vertical scales (left for
∗
sume that the total charge injected in the film is Q(t ).
chargingandrightfordischarging).
This definition is different from the one used in standard
chronocoulometry, where the EDL charge is obtained by
the intercept of the linear fit of the Q vs √t dependence chargingonesbut,ofcourse,startathighQvalues: inthis
with the vertical axis. The physical reason is the follow- casetheEDLchargeisevaluatedasthedifferencebetween
ing: Standard chronocoulometry procedure implies that the initial Q at t= 0 and the one reached when the curve
the starting of the current that generates the EDL and becomes linear. Finally the EDL charges obtained during
that of the current due to electrochemical phenomena are the charging phase (Qc in the plots) and the discharging
concurrent and that the scale of time in which these cur- one (Qd inthe plots)areaveragedto obtaina singlevalue
rents develop is the same and very small. In the case of QEDL for every applied gate voltage. It turns out that
our viscous PES the diffusion of electroreactants is likely Qc and Qd are very similar to each other as can be seen
to be ratherslowand,consequently,itisreasonabletoex- in Fig. 4 (please note the different vertical scales) and, of
pect that the current due to electrochemical phenomena course, they increase at the increase of Vgate. Let’s now
startswithacertaindelayandreachesitsmaximumvalue suppose that QEDL, measured by this modified versionof
sometimes later compared to that which forms the EDL. double-step chronocoulometry, is the same charge (with
This is exactly the kind of approximation we adopted in opposite sign) Qi induced at the surface of the metallic
assumingthatQ =Q(t∗). Thisdouble-stepprocedure thin film. The induced surface charge density will thus be
EDL
has the advantage to return both the charge that builds n2D = Qi/eS, where S is the surface of the film covered
uptheEDLduringtheapplicationofgatevoltageandthe with the PES (gated area) and e is the electronic charge.
chargethatcomesfromthedissolutionoftheEDL,during Wemustnowfocusonwherethechargeinducedbythe
the removal of gate voltage, allowing a consistency check polymericgatingisdistributedandhowitaffectsthetotal
between the measured values at two different moments. resistivity of the film. Since the charge that we induce on
In Figure 4 four double-step chronocoulometry plots the metallic film is only (in a first, rough approximation)
are presented for two different metallic films (silver and two-dimensional,we can describe the thin film as a paral-
gold)andfortwodifferentgatevoltages. Eachgraphshows lel of a perturbed and an unperturbed 3D region, where
thecharging(thicksolidline,bottom)andthecorrespond- the three-dimensional charge density of the perturbed re-
ing discharging (thick solid line, top) curve plotted with gion follows a density profile n3D(z), where z is the axis
respectto the squarerootoftime. The mainfeaturespre- normaltotheinterfaceandthedensitydecaysonalength
viously described are clearly visible: a rapid increase (de- scaledefinedbyaparameterξ. Bothn3D(z)andξ depend
crease) of the charge is present in the first 15-40 seconds onthe materialbut for metallic films and within a simpli-
after the application (removal) of the gate voltage, corre- fied semiclassical model, one can imagine that the whole
sponding mainly to the phase of formation (destruction) injected charge is uniformly distributed in a surface layer
of the EDL. Subsequently the Q vs √t curves become lin- of thickness ξ so that n3D = n2D/ξ. Different choices
≃
ear with positive (negative) slope indicating that a phase arepossibleforξ: fromtheclassicalThomas-Fermiscreen-
dominated by the diffusion of electroreactants has been ing length that in good metals is of the order of 0.5 ˚A to
reached. The linear fits of the different curves are shown morecomplexexpressionsthattakeintoaccountquantum-
in Fig. 4 as red thin lines. The discharging curves (verti- mechanical screening phenomena. Often a simple rule of
calscaleonthe right)aresymmetricalwithrespectto the thumb is used that consists in assuming the thickness of
4
0.2
60
1E15 Cu film on glass, 41 nm thick
50 -2m) Ag Ag film on glass, 30 nm thick
40 n (c2D1E14 01235. 5VVVV V CAAuuuSGilOas1s.22.2 fresh m) 0.0
14-210 cm) 2300 1E130VG 5app1li0cat1io5n o2r0der25 AuGlass 2.2 R/R' x t (n -0.2
(D 10 D
n2
Au film on glass, 5 nm thick
0 Au film on SiN, 25 nm thick
-0.4 Au film on SiO3 4, 27 nm thick
2
-10 Au film on glass, 50 nm thick
-6 -4 -2 0 2 4 6
-5.00 -2.50 0.00 2.50 5.00 7.50 10.00
Gate voltage V (V) n (x 1014 electrons /cm2)
g 2D
Figure5: Inducedsurfacechargedensityn2D asfunctionofthegate Figure 6: Dependence of ∆R/R′×t on n2D as obtained for films
voltageVgfordifferentdevicesmadeofdifferentmetals(Au,Ag,Cu). of different metals (Au, Ag, Cu) with different thickness and on
Although most of the points aregathered around a common trend, differentsubstrates (indicated inthelegend). Thestraightlinesare
some data show much higher values of the surface charge density. onlyguidestotheeyes.
Linesareonlyguidestotheeye. Intheinsetthedependenceofn2D
on the sequence of application of Vg is shown. For details see the
text. sametime,therearesomeotherdata(relevanttodifferent
metals: Cu, Au) that for the same gate voltages indicate
one atomic layer for the region where the charge is per- muchhigher n2D values (circles,open triangles,left trian-
turbed by the field. However, for our purposes here, the gles). Some of these data (open triangles) were obtained
exact calculation of this quantity is irrelevant. In fact it in the device called AuGlass2.2 that however also gave
can be shown both by a simple classic approach and by charge densities that lie on the lower trend line (solid tri-
a much more complex perturbative self-consistent quan- angles). Evidently the large differences in the n2D values
tum model based on the Lindhard-Hartree theory of the measured at the same Vg do not depend on the particu-
electronic screening [21] that ξ does not enter the final lar metal or on the specific device. To better understand
expression for the relative variationof the total film resis- the origin of this behaviour we analysed the sequence of
tance. Vg applications in AuGlass2.2. The inset to Fig. 5 shows
Thesimplefree-electroncalculationofthisrelativevari- the values of n2D for different gate voltages as a function
ation∆R/R′,thatisbasedontheparalleloftheperturbed of the sequential number of the measurements. It is clear
surface region and of the unperturbed bulk one and as- that, for any gate voltage, the charge injection is maxi-
sumes a constant effective electron mass and relaxation mum in the first application and systematically decreases
time, gives in the following. Indeed, the highest n2D values measured
in AuGlass2.2 and shown in the main panel of Fig.5 were
∆R/R′ = R(Vg)−R0 = n2D. (1) obtainedinthefirstfewVg applications,i.e. inthe“fresh”
R(Vg) − nt device. Thesubsequentvaluesofn2D lieonthemaintrend
line common to other devices and materials (thick dash-
where Vg is the gate voltage, n is the unperturbed 3D dot line). Since the PES is, in itself, very stable over long
densityofchargecarriers,R0istheunperturbedresistance times,this behaviourmightbe ratherascribedto asortof
when Vg =0 and t is the total thickness of the film. “memory”effect–probablyrelatedtothe“loss”ofLiions
at the interface with the electrodes – that certainly limits
4. Results and discussion the performances of the PES. Record n2D values of the
order of 4.51015 cm−2 that at the beginning can be ob-
·
Figure 5 shows the dependence of the injected sur- tainedwithV 2 3Volt(seemainpanelofFig. 5)after
g
∼ −
face charge density n on V determined by the mod- some use of the device would be reached only by apply-
2D g
ified double-step chronocoulometry in different materials ing much higher V with the consequent risks of possible
g
and devices. For V . 2 V n depends almost linearly reactionsbetween the electrolyte and the sample. Finally,
g 2D
| |
on V and then it grows more than linearly for higher V . the fact that the highest charge injections were obtained
g g
Theinjectedchargedensitiesforpositiveandnegativegate inAufilmsonSiO butalsoonglassseemstoexcludeany
2
voltages are rather symmetric for V . 2 V while for effect of the substrate on the device performance, which
g
| |
V > 2 V the negative ones are smaller. It is interest- looks rather reasonable. However, for the time being we
g
| |
ing to notice that there are data of different metals (Au, do not have a sufficient statistics to properly separate the
Ag) and of different devices that are gathered around the “memory” effect from that of the substrate.
sameguidefortheeyes(thickdash-dottedline)but,atthe In Figure 6 the results concerning the film resistance
5
variations generated by the charge injection are shown. logical insulators and graphene [22] or other 2D-like ma-
′
According to eq. 1 the quantity ∆R/R t should be a terials could lead to a large and defect-free modulation of
×
linear function of n with a negative slope. As a matter their surface electronicproperties withextraordinarycon-
2D
of fact Fig. 6 shows that this is the case for the variety of sequences that now can only be imagined.
filmsofdifferentmetalsanddevicesstudiedhere. However
thedataforgold(fullsquares,trianglesanddiamonds)and The authors acknowledge Francesco Laviano for AFM
silver(fullcircles)lieonthesamestraightbluedashedline measurements on the films.
which has a higher slope than the green dash-dotted line
that approximates the data for copper (full pentagons).
References
The explanation appears quite simple: In eq. 1 the slope
′
of the ∆R/R t vs n2D curve is inversely proportional References
×
to the volume carrier density n of the metal. But in the
framework of the free-electron model the n values of Au [1] R. E. Glover, III and M. D. Sherrill, et al. Phys. Rev. Lett. 5,
and Ag are quite similar (5.90 and 5.86 in units of 1022 248(1960).
cm−3,respectively)whilenforCuis8.47 1022cm−3. The [[23]] JC..MHa.nAnhhnaretteatl.aSl.c,iZen.cPeh2ys8.4B,181532,3(01799(91)9.91).
·
higherbulkcarrierdensityofcoppercannicelyexplainthe [4] H.Shimotaniet al.,Appl. Phys. Lett.91,082106 (2007)
reducedslopeofthe∆R/R′ tbehaviourshowninFig. 6. [5] A.S.Dhoot,C.Israel,X.Moya,N.D.Mathur,andR.H.Friend,
× Phys. Rev. Lett.102,136402(2009).
Goldfilmswithveryloweffectivethickness(forexample5
[6] K.Uenoetal.,Nature Mater.7,855-858(2008).
nm, asinthe caseshownby redsquaresinFig. 6)deviate [7] J.T.Yeet al.Nature Mater.9,125-128(2009).
from the linear behavior at n2D greater than 2 3 1014 [8] K.Uenoetal.,Nature Nanotech. 6,408(2011).
cm−2. This could be due to the increased role t−hat·scat- [9] M. J. Panzer, C. R. Newman, and C. D. Frisbie, Appl. Phys.
Lett.86,103503 (2005)
tering phenomena at the film surfaces have in ultrathin
[10] A.S.Dhootetal.Proc.Natl.Acad.Sci.32,11834-11837(2006).
films. Intheseconditionsasimplefree-electronmodelthat [11] G.BonfiglioliandR.Malvano,Phys. Rev.115,330(1959).
neglects the surfacescatteringmaybe nolongerappropri- [12] G. Bonfiglioli,E. Coenand R.Malvano, Phys. Rev.101,1281
ate to represent the real physical situation. A reduction (1965).
′ [13] H.L.Stadler,Phys. Rev. Lett.14,979(1965)
in the absolute value of ∆R/R for a given n2D is indeed [14] G.Martinez-Arizalaet al.,Phys. Rev. Lett.78,1130(1997)
predictedbythealreadymentionedquantumperturbative [15] N.Markovi´cet al.Phys. Rev. B 65,012501(2001).
model [21], when the probability of electron reflection at [16] D.Daghero,et al.Phys. Rev. Lett.108,066807(2012).
thesurfaceisnotnegligible. Themaximumabsolutevalue [17] J.M.Rowell,Supercond. Sci. Technol.16,R17(2003).
′ [18] H.Yuanetal.,J.Am.Chem.Soc.132,18402(2010).
of∆R/R obtaineduptonowinourpolymer-gatingfield- [19] Jijeesh R. Nair et al., Reactive and Functional Polymers 71,
effect experiments at room temperature is of the order of 409-416(2011).
8%, 1.9% and 1.6% for Au, Ag and Cu, respectively. Low [20] G. Inzelt, in Electroanalytical Methods. Guide to Experiments
and Applications, editedbyF.Scholz,Springer-Verlag2010,p.
temperatureexperiments(downto4.2K)haveshownthat
′ 147.
∆R/R values up to 10% can be obtained in gold [16].
[21] M.Ominiet al.,unpublished.
In conclusion, we have shown that in polymer-gating [22] G.Savini,A.C.Ferrari,andF.Giustino,Phys.Rev.Lett.105,
field-effectexperimentsitispossibletoinducesurfacecharge 037002 (2010).
densities n up to 3 4 1015 charges/cm2 by using a
2D
− ·
suitable polymer-electrolytesolution. This value ofn is
2D
approximately one order of magnitude greater than that
obtained in almost all the similar experiments reported
in literature, and even largerthan the maximum achieved
sofar[5]. Apropermodificationofthe classicdouble-step
chronocoulometrymethodgivesusaveryversatiletoolfor
thedeterminationoftheamountofsurfacechargeinduced
in these experiments, even when the standard Hall-effect
techniquecannotbeadoptedasinthecaseofmetallicthin
films. The huge resistance shifts (up to 8%) observed at
roomtemperatureinthinfilmsofdifferentmetals(Au,Ag,
Cu) reveal the power of this technique and, maybe, could
suggest possible applications of these results. The linear
′
dependence of the quantity ∆R/R t on n observed
2D
×
in our thin films is explained very well by a simple free-
electron model with parallel resistive channels, that also
accounts for the observed differences in the slope of the
above dependence in metals with a different intrinsic car-
rier density. The future application of this technique to
conventional and unconventional superconductors, topo-
6