Charge transport across metal/molecular (alkyl) monolayer-Si junctions is dominated by the LUMO level
Omer Yaffe, Yabing Qi, Lior Segev, Luc Scheres, Sreenivasa Reddy Puniredd, Tal Ely, Hossam Haick, Han Zuilhof, Leeor Kronik, Antoine Kahn, Ayelet Vilan, David Cahen
CCharge transport across metal/molecular (alkyl) monolayer-Si junctions is dominatedby the LUMO level
Omer Yaffe, Lior Segev, Tal Ely, Leeor Kronik, Ayelet Vilan, and David Cahen
Dept. of Materials & Interfaces, Weizmann Inst. of Science, Rehovot 76100, Israel
Yabing Qi and Antoine Kahn
Dept. of Electrical Engineering, Princeton University, Princeton, New Jersey, 08544 USA
Luc Scheres and Han Zuilhof
Lab. of Organic Chemistry, Wageningen Univ., Dreijenplein 8, 6703 HB Wageningen, The Netherlands
Sreenivasa Reddy Puniredd and Hossam Haick
Department of Chemical Engineering, Technion - Israel Institute of Technology, Haifa 32000, Israel (Dated: April 20, 2018)We compare the charge transport characteristics of heavy doped p ++ - and n ++ -Si-alkyl chain/Hgjunctions. Photoelectron spectroscopy (UPS, IPES and XPS) results for the molecule-Si bandalignment at equilibrium show the Fermi level to LUMO energy difference to be much smaller thanthe corresponding Fermi level to HOMO one. This result supports the conclusion we reach, based onnegative differential resistance in an analogous semiconductor-inorganic insulator/metal junction,that for both p ++ - and n ++ -type junctions the energy difference between the Fermi level andLUMO, i.e., electron tunneling, controls charge transport. The Fermi level-LUMO energy difference,experimentally determined by IPES, agrees with the non-resonant tunneling barrier height deducedfrom the exponential length-attenuation of the current. I. INTRODUCTION
Molecular electronics describes charge transport pro-cesses whereby molecules serve as active elements (e.g.,rectifiers, switches, sensors) or passive ones (resistors orsurface passivating agents) in electronic devices . Sincethe emergence of this field, it has expanded greatly andnow includes several types of device configuration (e.g.,two-terminal junctions , three-terminal junctions andelectrochemical devices ), where the molecules serve asdirect current carriers or indirectly affect the electricalproperties of a junction .Despite the diversity of investigated systems, funda-mental questions regarding the mechanisms for electri-cal current passing through molecules between two elec-trodes and the possibility of gaining predictive power andcontrol over the electrical properties of molecular junc-tions remain mainly unsolved. Detailed discussions onthe limitations of existing transport models can be foundelsewhere. In short, even for one of the simplest sys-tems, i.e. that of alkyl chains between Au electrodes,there is a large discrepancy between the average tunnelbarrier height, extracted from current - voltage ( I − V )measurements ( ∼ . and the barrier thatis expected from the experimentally determined elec-trode work function and alkyl monolayer ionization po-tential, i.e. the energy difference between the Fermilevel and highest occupied molecular orbital (HOMO) asfound by Ultraviolet Photoelectron Spectroscopy, UPS( ∼ In spite of this apparent discrepancy, HOMO-dominated transport is the prevailing concept for suchjunctions.
In the present study we challenge this concept and find that transport is dominated by the en-ergy difference between the Fermi level and the lowestunoccupied molecular orbital (LUMO) of the molecules.It would seem that having a semiconductor (SC) in-stead of a metal as one of the electrodes in a molecu-lar junction increases the difficulty in understanding themechanisms that govern charge transport through thejunction. Indeed, with moderately doped SCs, the SCdepletion layer has a large effect on the overall observedcurrent and if the band bending is very large, it com-pletely overwhelms any molecular effect. As we shallsee, though, the intrinsic asymmetry of a junction with aSC, with respect to charge carrier type, provides uniqueinformation on the energy levels involved in transport, in ways that are impossible when using only metal elec-trodes .To find how charge is transported through a mono-layer of saturated alkyl chains we measure and analyzeelectrical transport across heavy-doped semiconductor -molecular insulator / metal junctions. Our results pro-vide unequivocal evidence that charge transport in thesejunctions is controlled by the energy difference betweenthe metal Fermi level and the LUMO of the molecule, | E F − LUMO | , rather than by the energy difference withthe HOMO, | E F − HOMO | . This finding agrees withcomplementary spectroscopic measurements of the sam-ples without Hg contact, but challenges the prevailingconcept of HOMO-dominated transport. Highlydoped n ++ - and p ++ - Si-C n H n +1 /Hg junctions ( n =2 , , ,
18) were prepared by alkylation of freshly etchedP- or B-doped ( N d / a ≈ cm − ) Si(1 1 1) surfaces.Detailed descriptions of the preparation and character-ization of both ’long’ (C14-C18) and ’short’ alkyl chain a r X i v : . [ c ond - m a t . m t r l - s c i ] O c t - 1 . 5 - 1 . 0 - 0 . 5 0 . 0 0 . 5 1 . 00 . 00 . 20 . 40 . 60 . 81 . 0 Normalized G (a.u.)
M e t a l b i a s ( V ) n + + p + + cba
FIG. 1: Normalized G-V curves for p ++ - (solid line) and n ++ -(dashed line) Si-C H /Hg junctions. The conductance ofthe n ++ -junction is parabolic around 0 V, while for the p ++ -junction at negative bias on the metal there is a voltage range,where the conductance increases slowly (shaded area), fol-lowed by a sharp increase. The transition bias between theabove-mentioned regions is ∼ . ∼ .
12 eV). (a), (b) & (c) refer to voltage ranges, shownin Fig. 2 and discussed in the text, for the p ++ junction only. (C2) samples can be found elsewhere, and are given inthe Suppl. Inf. (section 1 and section 2). II. RESULTS AND DISCUSSION
In Fig. 1, the normalized conductance of p ++ -/n ++ -Si-C H /Hg junctions is compared on a linear scale.While conductance for the n ++ junction (dash) isparabolic around 0 V (i.e. similar to the behaviourexpected for Metal -Insulator -Metal (MIM) tunneljunctions , for the p ++ junction (solid) there is a neg-ative bias range where conductance increases slowly(shaded), followed by a sharp increase at ∼ . .
12 eV) .This behaviour is highly reproducible and indepen-dent of the molecular length (C H -C H ; see ref.27 for J − V curves of n ++ -Si data and Fig. 5 belowfor the p ++ -Si ones). The results of the p ++ junctionat negative applied bias can be understood by consid-ering work on Metal-inorganic Insulator- (near) degen-erate Semiconductor (MIS) tunnel diodes by Esaki, Sze,Dahlke and co workers, which forms the basis forthe present study. Schematic band diagrams of the rel-evant model are shown in Fig. 2 (top), along with anexperimentally measured semi log J − V curve of a typ-ical p ++ Si-C H /Hg junction. We identify the undu-lating behaviour at low to moderate negative bias (re-gion b of Fig. 2, which is equivalent to the shaded regionin Fig. 1) with a region where negative differential resis-tance (NDR) was predicted originally for the MIS system - 1 . 5 - 1 . 0 - 0 . 5 0 . 0 0 . 5 1 . 01 E - 41 E - 30 . 0 10 . 11 cb Normalized J (a.u.)
M e t a l b i a s ( V ) a FIG. 2: Schematic band energy diagrams for MIS structureswith p ++ semiconductor substrates and LUMO-dominatedtransport (Top) and semi-log J-V curve of a typical p ++ Si-C H /Hg junction (averaged over 12 scans) (Bottom) withthe (a), (b) and (c) labels corresponding to the diagramsabove. Three major bias regimes are considered in the dif-ferent energy schemes (top) and separated in the J − V plot by vertical dotted lines (bottom) , corresponding to a)metal Fermi above (i.e., closer to vacuum level) the SCgap ( V < − ( E g + ζ )); b) metal Fermi within the SC gap( − ( E g + ζ ) < V < − ζ ); and c) metal Fermi below the SCgap ( V > − ζ ), where E g and ζ are the forbidden gap andthe energy difference between the SC Fermi level and the va-lence band edge. TOP: φ t is the energy difference between thetunneling electrons (marked by a horizontal arrow) and themolecular level (spatially averaged), which is taken as the es-timated effective tunneling barrier height. The insulator levelis assumed to vary across the insulator. Electron tunnelinginto the VB (as in b) is practically identical to holes tunnelingin the opposite direction. Surface states, band-bending andimage forces are neglected for simplicity. studied by Esaki et al. That effect is smeared out in the J − V curve, because of the effect of interface states, aneffect that, together with band bending and image forces,is, for simplicity’s sake, neglected in the schematic banddiagrams. In the figure, only E F -LUMO (i.e., electrontunneling) dominated transport is considered (LUMO isdepicted by thick, black line) and other cases are dis-cussed below.As long as the metal Fermi level does not align withthe SC forbidden energy gap (i.e., regions a, c) transportis completely equivalent to tunneling in the more com-mon Metal-Insulator-Metal (MIM) tunnel junction. In regions a and c, the bias acts to reduce the averagedtunneling barrier by ∼ | V | /
2. With a moderate nega-tive voltage to the metal (Fig. 2) the metal Fermi levelmoves to energies that are in the forbidden band gapof the SC. For such a case, neglecting interface states-assisted tunneling (ISAT), Esaki and Stiles predicted aneffective increase in the tunnel barrier height φ t (i.e., φ t (b) > φ t (a,c)), and NDR, as has indeed been observedexperimentally. If ISAT is not neglected, instead ofNDR a ln( J ) − V plot will show undulating behaviour(as we show in Fig. 2, voltage range (b), for molecularjunctions), which is strongly influenced by the interfacestate density and distribution in the band gap. In ourcase ISAT has a clear influence on the J − V data and,therefore, we observe undulating J − V behaviour, insteadof pure NDR. Undulating J − V occurs because in regionb the highest energy electrons have no matching states totunnel into, and transport proceeds via deeper electrons,aligned with the SC VB. For those electrons the barrieris now increasing as ∼ | V | / | V | , as in band-to-band tunneling (ranges a,c). In con-trast to p ++ junctions, n ++ junctions are not predictedto show undulating J − V plots for LUMO-dominatedtunneling because the SC Fermi and the tunneling levelsare on the same side of the forbidden SC gap, regardlessof applied bias (see Fig. S3.1b of Suppl. Inf.). Simplyput, undulating J − V is predicted to occur if tunnelingproceeds via carriers of opposite type to the SC majoritycarriers. This prediction is confirmed experimentally inFig. 1, which shows a highly symmetrical G − V curvefor the n ++ -Si junction.If the | E F − HOMO | energy difference did determine thetransport barrier (i.e., hole, rather than electron tunnel-ing), then the special undulating J − V behaviour shouldbe observed in the n ++ junction for positive applied biason the metal and the p ++ junction should not presentsuch behaviour at all. The reason for the latter is thatfor a p ++ -SC - based junction and hole tunneling, • with negative bias applied to the metal, holes tun-nel from the SC valence band to the metal; • with positive bias applied to the metal, holes tunnelfrom the metal to the SC valence band.Neither of these flows involves tunneling of holes intothe SC band gap. Thus, the undulating ln( J ) − V be-haviour (Fig. 2) and the approximately constant conduc-tance (Fig. 1) for the p ++ -Si-alkyl/Hg junction and lackof it in the n ++ one (Fig. 1) provide direct evidence for | E F − LUMO | (or electron)-, rather than | E F − HOMO | (hole)-dominated transport in both doping types. To putthis in other words, if transport were HOMO-dominated,then the tunneling level and the SC Fermi level wouldbe below and above the n ++ -SC gap (at positive bias),while both the HOMO and the SC Fermi level wouldbe always below the SC gap for p ++ junctions. Thus,tunneling via HOMO predicts undulating J − V for n ++ junctions, which is not observed experimentally (cf. Fig.in the Suppl. Inf. for energy schemes that illustrate these alternatives). We note that, for an MIM junction, thistype of asymmetry cannot be observed, because it origi-nates in variation in available density of states for tunnel-ing. Indeed, Scott et al have used this exact asymmetryand demonstrated that inelastic tunneling is much morepronounced in a metal/molecule-p ++ Si junction than ina metal-molecule-metal junction The conclusion that transport is dominated by the | E F − LUMO | energy difference implies that the LUMOis closer in energy to the SC E F for both doping types.This can be verified by characterizing the energeticsof the molecularly modified Si, prior to metal con-tact deposition. To this end we use Ultra-violet Pho-toemission Spectroscopy (UPS),
Inverse PhotoemissionSpectroscopy (IPES),
X-ray Photoelectron Spectroscopy (XPS) and
Contact potential difference (CPD) measure-ments. We then study the Si band bending at equilibriumafter top metal deposition by measuring charge trans-port across molecular monolayers (Si-C H /Hg) that areso thin that they do not present an effective tunnel-ing barrier. Finally, we perform a molecular mono-layer length-dependent study of SiC n C n +1 /Hg junctionswhere n=14, 16 and 18. These monolayers are thickenough to make tunneling the dominant current trans-port process. From this length dependence study we ex-tract the tunneling barrier height and compare it to thebarrier that is expected from the UPS/IPES results.We begin with determining the band bending (BB) inSi in the p ++ - and n ++ -Si-C H systems. The positionof the Fermi level is known in our system from frequentlyrepeated UPS and IPES measurements of the Fermi stepon atomically clean metal (e.g. Au) surfaces. Given thenearly 20 ˚A thickness of the SAM layer, the Si valenceband maximum (VBM) and conduction band minimum(CBM) are not visible in UPS and IPES, and we mustuse here the XPS measurement of the position of theSi 2p core level, which is clearly detectable through theSAM layer, to evaluate the VBM position and the BB. In-deed, the energy difference between the core level and theVBM, called here (Si2p-VBM) is fixed, and we can usethe position of either one to determine that of the other.Combined UPS/XPS measurements done in our labora-tory on carefully prepared hydrogen-terminated Si sur-faces give (Si2p-VBM)= 98 . The Si 2p peaksare found at 99 . . ++ - and n ++ -Si- C H surfaces, respectively.Thus, simple considerations give:BB p − Si = ( E F − VBM)= 99 . − . . n − Si = E gap − ( E F − VBM)= 1 . − (99 . − . . ++ and n ++ -Si surfaces are therefore seen to FIG. 3: a) Combined UPS (left of E F line, blue) and IPES (right to E F line, red) spectra of C H -Si(1 1 1) surfaces ofheavy-doped p ++ - (upper) and n ++ - (lower) Si. The vertical black line marks the experimentally determined Fermi levelof both samples. The crossings of the tangent lines determine the edges of the monolayer edge-to-edge gap. While the UPSmeasurements yield a single unambiguous edge, which corresponds with the position of the theoretically calculated HOMO, theIPES spectra show two transitions, marking two possible band edges, which are noted as ++ and n ++ .The black vertical arrow marks the position of the estimated tunneling barrier relative to the Fermi level, extracted from lengthdependent J − V measurements of the p ++ Si-alkyl/Hg junction. b) and c) Proposed band diagrams of p ++ - and n ++ -Si-C H . be slightly depleted with BB= 0 . TheCBM in each case is obtained by adding the 1 . .
22 eV) UPS spectra, we find that the work func-tions of p ++ - and n ++ -Si-C H are (4 . ± .
1) eV and(4 . ± .
1) eV, respectively. Similarly, the CPD measure-ments performed in the dark also yield a 0 . ++ - orn ++ -Si).The difference between work functions of p ++ -and n ++ -Si-C H (0 . . similar to HOMO*and LUMO* in ref. 14. Fig. 3 presents the combined results of He(II)(40 . ++ (top) andn ++ -Si-C H . We define the edges of the monolayerband gap as the crossing between the tangent lines ofthe spectra. While the edge of the UPS spectrum is de-fined clearly (both for p ++ and n ++ ), from the IPESspectra, two different edges can be defined. The reasonis that the IPES incident electrons have a larger penetra-tion depth than the escape depth of the UPS electrons.As a result, the interface states are more pronounced inthe IPES spectrum than in the UPS spectrum. Thesestates, which result from hybridization between the alkylmolecular states and the Si VB and CB states at the in-terface are known as Induced Density of Interface States(IDIS) and were calculated to be localized in the firstthree carbons of the alkyl chain (with respect to the Sisurface). IDIS is probably the main physical origin forthe onset of the intensity in the IPES spectrum (Fig.3). As a result, the IPES spectrum has a “two slope”behavior, noted as ++ and n ++ samples. We view slope by IDIS, and byradiation damage. The HOSO and LUSO positionscan be found by fitting the calculated density of statesresults, as reported by Segev et al. for the Si-C n H n +1 system to the experiment. In their calculations Segevet al., identified the actual LUSO to be ∼ . ∼ . ++ andn ++ samples, the Fermi level is closer to the onset of theIPES spectrum (LUSO) than to the onset of the UPSspectrum (HOSO) and the estimated uncertainty of theexact position of the HOSO and LUSO is marked bya shaded area. The apparent difference in the edge toedge gap in Fig.3 for the molecular monolayers on p ++ and n ++ , is within this uncertainty. For p ++ -Si-C H samples, the energy difference between the Si E F andthe monolayer LUSO ( ∼ . E F and the HOSO ( ∼ ++ -Si-C H samples, the energy difference betweenthe Si E F and the monolayer LUSO ( ∼ E F and the HOSO ( ∼ . Possibly this isrelated to the above-hypothesised doping-dependent Si-Cbond polarization. Regardless of this result, which is cur-rently under further study, the key experimental findingis that, for both p ++ and n ++ Si systems, the Fermi levelof the system and the carrier transport level in the Si aresignificantly closer to the LUSO than to the HOSO.We now ask if the spectroscopic results presented inFig. 3 can efficiently predict the charge transport char-acteristics of the alkyl chain monolayer? To answer this,we compare the effective tunneling barrier height thatwe can deduce from the J − V measurements with theexpected barrier that we find from the IPES spectrumi.e. | E F − LUSO | energy difference. To this end we needto take under consideration the Si band bending in thejunction at equilibrium (zero bias), rather than the bandbending at the Si surface as determined by XPS for thefree molecularly- modified Si surface, because that band - 0 . 4 - 0 . 2 0 . 0 0 . 2 0 . 41 0 - 6 - 5 - 4 - 3 - 2 - 1 n + + S i - C H p + + S i - C H J (A/cm2)
M e t a l B i a s ( V ) p + + / n + + S i - H
FIG. 4: J − V measurements of p ++ and n ++ Si-C H /Hgjunctions (solid and dashed curves, respectively), compared tothe J-V behavior of p ++ and n ++ Si-H/Hg junctions (dottedcurve; p ++ and n ++ curves are identical). The error barsrepresent the standard deviation of multiple measurementson 8 different junctions. bending can change after forming a metal (Hg) contact onthe molecules. Because the molecular monolayer is verythin (4 −
20 ˚A) and supports a large tunneling current,there can be charge transfer between the metal and theSi substrate across the molecules, until electronic equi-librium is established (equalization of the Fermi levels ofthe Si and the metal). This charge transfer affects theSi band bending, but should not change the alignmentbetween the Si bands (at the surface) and the monolayerHOSO and LUSO, assuming that the molecules do notbecome charged.Elsewhere we showed that, due do the large surfacedipole that is introduced by the molecules, depositing Hgon a moderately doped n-Si-C n H n +1 sample results ina large band bending in the Si, up to the point wherethe Si surface is in strong inversion and charge trans-port in the junction is controlled by minority carrier re-combination. For the heavy-doped Si used here, the ef-fect of the band bending on transport is usually muchsmaller, because the depletion layer is thinner ( ∼
10 nmcompared to ∼ µ m for moderately doped Si) and al-lows some charge tunneling through it. This type ofcharge transport mechanism, which is intermediate be-tween thermionic emission and pure tunneling, is knownas thermionic field emission (TFE). To isolate the effect of alkyl molecules on the Si bandbending from tunneling through the molecules, we per-formed J − V measurements on a junction with an ul-trathin molecular monolayer, C H (Fig. 4), and com-pared the results to those obtained with freshly etched,H-terminated Si. Because the C H monolayer is so thin( ∼ . J − V characteristics of the Si-H/Hg junction weremeasured as a control (dotted curve) and found to beOhmic and identical for p ++ and n ++ junctions. Theresults for Si-C H /Hg are very different with the p ++ junction exhibiting Ohmic behavior with current densityeven higher than that of the Si-H samples. The n ++ junc-tion yields an exponential increase at both bias polarities,a behavior that is typical of charge transport via TFE. The high current density of the p ++ junction is consistentwith our assumption that the tunnel barrier presented bythe C H monolayer is negligible, and that the dipole di-rection works to slightly decrease the Si band bending.The same dipole, on the n ++ -Si surface should, therefore,increase the band bending. Indeed, when we fit the n ++ results at forward bias (positive bias on the metal) toan analytical expression for TFE developed by Padovaniand Stratton, we find a barrier height of 0 . ++ -Si band bendingvalue at equilibrium. Thus, for p ++ -Si junctions thereis negligible internal barrier in the Si, while such a barrierexists for the n ++ -Si junction, where it has a large effecton the overall J − V curve. While this does not necessar-ily mean that the molecular contribution to transport isoverwhelmed by band bending in n ++ -Si, as is the casewith moderately doped Si junctions, this band bendingdecreases the current density significantly. Due to thelarge internal barrier in the n ++ -Si, we limit ourselvesto the molecular length dependent study of the p ++ -Sijunction, and then only to the data for positive appliedbias on the metal, where according to Fig. 1, current ismostly due to band-to-band tunneling with the densityof states available for tunneling approximately constantwith applied bias. The reason is that, as we noted earlier,for negative applied bias the dominant charge transportmechanism is ISAT, i.e., transport is controlled by therecombination rate at the Si surface and cannot be de-scribed with conventional tunneling models. Fig. 5 presents the J − V behavior of p ++ -Si-C n H n +1 /Hg junctions (n=14,16 and 18) (the length dependenceof the n ++ -Si-C n H n +1 /Hg junctions were publishedelsewhere ). It is clear that the qualitative behaviorof the J-V curve is independent of chain length. It isalso clearly seen that, as expected for such junctions, thecurrent density decreases exponentially with the num-ber of methylene units in the alkyl chain. This typeof behaviour is generally described by the Landauerrelation,
In this model, the conductance of a singlechannel (G) is given as: G = G c exp ( − βl ) (1)where l is the tunnel barrier width, which can be takenas the length of the molecule or the thickness of themonolayer, β is the length-decay parameter and G c isthe contact conductance. From a G vs. l plot at a given - 1 . 0 - 0 . 5 0 . 0 0 . 5 1 . 01 0 - 4 - 3 - 2 - 1 p + + S i - C n H / H g ( n = 1 4 , 1 6 , 1 8 ) J (A/cm2)
M e t a l b i a s ( V ) S i - H S i - C 1 4S i - C 1 6S i - C 1 8
FIG. 5: Length dependence of current density vs. voltagecurves for for p ++ Si-C n H n +1 /Hg junctions ( n = 14 , , and18). Data for the Si-H/Hg junctions (dashed) are given as ref-erence. The current density exponentially decreases with themolecular length. Error bars represent the standard deviationof at least 8 different junctions. V app , we extract both β (slope) and G c (intercept). Westart by discussing the meaning of G c , followed by an in-terpretation the extracted β value. The extracted G c value (at V app = 0 . ++ -Si-alkyl/Hg junc-tions is 2 · − G , where G is the quantum conduc-tance ( G ≡ q /h = 77 . µ S). The Gc value of n ++ -Si-alkyl/Hg junctions was 4 · − G , very low com-pared to typical Au-alkyl MIM junctions. It reflects thelarge band bending in the Si that was extracted from the J − V behaviour of n ++ Si-C H /Hg observed in Fig. 4.The G c found for the p ++ junction is similar to that re-ported for an MIM junction with one chemi-contact , aresult that can be rationalized by recalling the near flatband conditions in the p ++ -Si-C H /Hg junction (Fig.4). To further verify the correlation between the spec-troscopic evidence (Fig. 3) and the transport results, weestimate the tunnel barrier ( φ t ) from the J − V curves.This barrier should, in principle, reflect the energy dif-ference between the Fermi level of the electrodes and theenergy of the nearest allowed state in the interfacial in-sulator, as detected by UPS and IPES. In addition, wecompare the J − V behavior of our p ++ -Si-alkyl/Hg junc-tion to previously reported results on the more commonAu-S-alkyl-S-Au junction. To that effect, we use the gen-eral WKB approximation for tunneling that describes therelation between β and φ t as ? : β = 2 (cid:113) m e φ t / ¯ h (2)where β is the decay parameter at equilibrium ( V app → h is Planck’s constant, q is the elementary chargeand m e is the electron mass. This is by no means anaccurate model that reflects the complexity of molecularjunctions. Nevertheless, as it is the basis for the Sim-mons model that is widely used for the extraction of φ t in molecular junctions, it allows direct comparison withprevious results. The difference in dielectric propertiesbetween vacuum and insulating materials is commonlyaccounted for by using an effective mass for the tunnel-ing electron, which is smaller than the free electron mass.Using complex band structure DFT calculations, Tom-fohr and Sankey calculated the dispersion relation of theforbidden band gap of an infinite alkyl chain and showedthat by using the dispersion relation in vacuum (Eq. 2)with an effective mass of 0 . m e , a lower limit of thebarrier is extracted. Thus, we use this value of effec-tive mass to extract an estimated value of φ t , which wecan compare to previous studies and to the spectroscopicresults presented in Fig. 3.The average extracted φ t in the low bias range (0 − . J − V curves in Fig. 5 is (0 . ± .
05) ˚A − .Inserting this value into Eq. 2, along with a 0 . m e ef-fective mass, yields a 2 . ± . | E F − LUSO | dominates charge transport(Fig. 1), we can locate the extracted φ t value on the IPESspectrum (marked by vertical arrow) in Fig. 3. This valueis in between the IPES edges and, because it is only thelower limit of the barrier, we conclude that we have rea-sonable agreement with the LUSO of the alkyl monolayer(edge ∼ . | E F − HOSO | difference of 4 eV(Fig. 3), decreasing further the possibility of significantHOSO-contribution to transport.The natural question is how far our observation forLUSO-dominated tunneling in alkyl-Si junctions is ap-plicable to alkyl junctions on any substrate? The an-swer depends on the position of the molecular levels ofthe alkyl chains relative to the electrode Fermi level.UPS/IPES measurements of alkyl thiols on Au yielda HOSO-LUSO gap of 7 .
85 eV and significantly smaller | E F − LUSO | ( ∼ .
35 eV) than | E F − HOSO | ( ∼ . | E F − HOSO | for alkyl thiols on Au ( ∼ TheHOSO-LUSO gap is similar to the gap measured for alkylchain monolayers on both Si and GaAs with dif-ferent binding groups (Si-C, Si-O-C and GaAs-PO -C),i.e., it is not influenced significantly by binding group andsubstrate type . The range of the edge-to-edge gap in allexamined samples is ∼ (7 . − .
8) eV, if the IDIS on-set is ignored. These values agree with values reportedfor polyethylene.
Thus we conclude that the HOSO-LUSO gap of alkyl monolayers, and their position rel-ative to the vacuum level is rather independent of thecontacting electrodes. Therefore, we expect tunneling tobe dominated by the | E F − LUSO | energy difference forany alkyl junctions, except maybe for electrodes of ex-tremely high work function ( > ∼ | E F − HOSO | barrier,especially for the extensively studied Au-(alkythiol oralkyldithiol)-Au systems. In most cases theevidence for hole tunneling is based on transport acrossrelatively short alkyl chains, up to C12, and is indirect. An exception is the measurement of thermo-power (See-beck coefficient), where the positive coefficient was takenas direct evidence for hole / HOSO-mediated tunneling. Remarkably, though, the thermo-power decreases withincreasing molecular length and, when extrapolated, ispredicted to change sign (i.e., change from HOSO- toLUSO-dominated, hole to electron tunneling) for alkylchains longer than C12. This length-dependence was ra-tionalized by invoking Metal Induced Gap States (MIGS)as dominating transport, rather than pure molecularstates. Such an explanation is very reasonable, becausewith the short dithiol molecules used in those experi-ments (Au-SC n H n +1 S-Au with n = 2 , , , , ,
8) MIGSare expected to extend out to several carbon atoms fromthe contacts.In early CP-AFM work on short mono and dithiol alkylchains the extrapolated zero molecular length resistanceincreased as the work function of the contacting metaldecreased and this was taken to indicate an increasingtunnel barrier. Similar to the thermo-power, this effectcan be due to the MIGS, rather than to alkyl-derivedenergy levels.In earlier work with Si as substrate, the type ofdominant charge carrier in the semiconductor was sug-gested to dictate the type of tunneling through (longer)alkyl chains. In later work on a system, similar tothe one studied here, the lack of temperature activa-tion was taken as indicating HOSO-mediated /electrontunneling , an effect that may well be present, but isoverwhelmed by the effects that we have shown here.Except for the last two reports, we note that the maindifference between ours and other experiments is thatSi-monolayer/Hg junctions (with a several 100 µ m diam-eter contact, rather than a nm-sized one) make it pos-sible to measure conductance through longer molecules,which allows measuring molecule-, rather than contact-dominated transport by reducing the contribution ofMIGS and IDIS interface effects to net transport. III. CONCLUSIONS
To conclude, the analogy to asymmetry in NDR fordegenerately-doped MIS of different doping types indi-cates that for both p ++ - and n ++ -type Si-alkyl chain/Hgjunctions charge transport is controlled by | E F − LUSO | (rather than by | E F − HOSO | ). Furthermore, by compar-ing charge transport through p ++ - and n ++ -Si-C2H5/Hgjunctions, we showed that there is considerable bandbending in the n ++ -Si junction, which has a large ef-fect on the charge transport (even though the Si is heavydoped). The p ++ -Si junction, though, is near flat-bandin equilibrium, with comparable transport probability tothat of MIM junctions. Using UPS and IPES measure-ments we elucidated the electronic structure of Si-alkylsystem and showed that | E F − LUSO | < | E F − HOSO | ,in line with the electron tunneling qualitative argu-ment. The tunneling barrier height estimated from theWKB approximation also fits well with the IPES-derived | E F − LUSO | difference. Acknowledgements
We thank the US-Israel Binational Science Foundation(DC, AK), the Israel Science Foundation, ISF, throughits Centre of Excellence programs (DC, LK), the LiseMeitner Minerva Center for Computational Chemistry(LK), the National Science Foundation (DMR-1005892)) and the Princeton MRSEC of the National Science Foun-dation (DMR- 0819860) (AK), NanoNed, funded by theDutch Ministry of Economic Affairs (project WSC.6972)(HZ) and the Monroe and Marjorie Burk Fund for Al-ternative Energy Studies (DC) for partial support. Atthe Weizmann Inst. this work was made possible in partby the historic generosity of the Harold Perlman family.OY thanks the Azrieli Foundation for an Azrieli Fellow-ship. DC holds the Sylvia and Rowland Schaefer Chairin Energy Research. J. R. Heath, Ann. Rev.of Mater. Res. , 1 (2009). B. Mann and H. Kuhn, J.Appl. Phys. , 4398 (1971). A. Aviram and M. A. Ratner, Chem. Phys. Lett. , 277(1974). A. Vilan and D. Cahen, Adv. Funct. Mater. , 795 (2002). L. Venkataraman, J. E. Klare, I. W. Tam, C. Nuckolls,M. S. Hybertsen, and M. L. Steigerwald, Nano Lett. ,458 (2006). F. Chen and N. J. Tao, Account.Chem. Res. , 429(2009). O. Tal, M. Krieger, B. Leerink, and J. M. van Ruitenbeek,Phys. Rev. Lett. , 196804 (2008). H. Yu, Y. Luo, K. Beverly, J. F. Stoddart, H. Tseng, andJ. R. Heath, Angew. Chem. , 5884 (2003). S. Ciampi, T. B¨ocking, K. A. Kilian, M. James, J. B.Harper, and J. J. Gooding, Langmuir , 9320 (2007). S. Ho Choi, B. Kim, and C. D. Frisbie, Science , 1482(2008). H. Haick, M. Ambrico, T. Ligonzo, and D. Cahen, Adv.Mater. , 2145 (2004). A. Vilan, J. Ghabboun, and D. Cahen, J. Phys. Chem. B , 6360 (2003). X. D. Cui, A. Primak, X. Zarate, J. Tomfohr, O. F. Sankey,A. L. Moore, T. A. Moore, D. Gust, L. A. Nagahara, andS. M. Lindsay, J. Phys. Chem. B , 8609 (2002). V. B. Engelkes, J. M. Beebe, and C. D. Frisbie,J.Am.Chem.Soc. , 14287 (2004). E. H. Huisman, C. M. Gu´edon, B. J. van Wees, and S. J.van der Molen, Nano Lett. , 3909 (2009). M. Paulsson and S. Datta, Phys. Rev. B , 241403 (2003). A. Vilan, J. Phys. Chem. C , 4431 (2007). H. B. Akkerman and B. de Boer, J. Phys. Condens Mat , 013001 (2008). H. B. Akkerman, R. C. G. Naber, B. Jongbloed, P. A. vanHal, P. W. M. Blom, D. M. de Leeuw, and B. de Boer,PNAS , 11161 (2007). X. D. Cui, X. Zarate, J. Tomfohr, O. F. Sankey, A. Pri-mak, T. A. Moore, D. Gust, G. Harris, and S. M. Lindsay,Nanotechnology , 5 (2002). W. Wang, T. Lee, and M. A. Reed, Phys. Rev. B ,035416 (2003). D. M. Alloway, M. Hofmann, D. L. Smith, N. E. Gruhn,A. L. Graham, R. Colorado, V. H. Wysocki, T. R. Lee,P. A. Lee, and N. R. Armstrong, J. Phys. Chem. B ,11690 (2003). C. Li, I. Pobelov, T. Wandlowski, A. Bagrets, A. Arnold,and F. Evers, J.Am.Chem.Soc. , 318 (2008). J. M. Beebe, B. Kim, C. D. Frisbie, and J. G. Kushmerick,ACS Nano , 827 (2008). O. Yaffe, L. Scheres, S. R. Puniredd, N. Stein, A. Biller,R. H. Lavan, H. Shpaisman, H. Zuilhof, H. Haick, D. Ca-hen, et al., Nano Lett. , 2390 (2009). S. R. Puniredd, O. Assad, and H. Haick, J. Am. Chem.Soc. , 13727 (2008). O. Yaffe, L. Scheres, L. Segev, A. Biller, I. Ron, E. Sa-lomon, M. Giesbers, A. Kahn, L. Kronik, H. Zuilhof, et al.,J. Phys. Chem. C , 10270 (2010). W. F. Brinkman, R. C. Dynes, and J. M. Rowell, J. Appl.Phys. , 1915 (1970). S. M. Sze and K. K. Ng,
Physics of Semiconductor Devices (New York: John Wiley & Sons, Inc., 2007), 3rd ed. W. E. Dahlke and S. M. Sze, Solid-State Electron. , 865(1967). L. Esaki, Nobel lecture (1973). L. Esaki and P. J. Stiles, Phys. Rev. Lett. , 1108 (1966). L. Freeman and W. Dahlke, Solid-State Electron. , 1483(1970). J. G. Simmons, J. Appl. Phys. , 1793 (1963). R. Stratton, J. Phys. Chem. Solids , 1177 (1962). L. Freeman, Ph.D. thesis, Lehigh University (1969). D. Janes and A. Scott, Birck and NCN Publications (2010). L. Segev, A. Salomon, A. Natan, D. Cahen, L. Kronik,F. Amy, C. K. Chan, and A. Kahn, Phys. Rev. B ,165323 (2006). F. J. Himpsel, G. Hollinger, and R. A. Pollak, Phys.Rev.B , 7014 (1983). G. Nesher, H. Shpaisman, and D. Cahen, J. Phys. Chem.B , 734 (2007). I. G. Hill, A. Kahn, Z. G. Soos, and J. Pascal, Chem. Phys.Lett. , 181 (2000). N. Sato, K. Seki, and H. Inokuchi, J. Chem.Soc. , Farad.Trans. 2 , 1621 (1981). C. I. Wu, Y. Hirose, H. Sirringhaus, and A. Kahn, Chem.Phys. Lett. , 43 (1997). A. V. Oliver Seitz, Advan. Funct. Mat. , 2102 (2008). F. Amy, C. K. Chan, W. Zhao, J. Hyung, M. Ono,T. Sueyoshi, S. Kera, G. Nesher, A. Salomon, L. Segev,et al., J. Phys. Chem. B , 21826 (2006). S. Avasthi, Y. Qi, G. K. Vertelov, J. Schwartz, A. Kahn,and J. C. Sturm, Surf. Sci. , 1308 (2011). E. Rhoderick and R. Williams,
Monographs in Electri-cal and Electronic Engineering. Metal-Semiconductor Con-tacts (Clarendon press, Oxford, 1988), 2nd ed. F. Padovani and R. Stratton, Solid-State Electron. , 695 (1966). D. M. Adams, L. Brus, C. E. D. Chidsey, S. Creager,C. Creutz, C. R. Kagan, P. V. Kamat, M. Lieberman,S. Lindsay, R. A. Marcus, et al., J. Phys. Chem. B ,6668 (2003). G. Heimel, L. Romaner, E. Zojer, and J. Bredas, Acc.Chem. Res. , 721 (2008). A. Salomon, D. Cahen, S. Lindsay, J. Tomfohr, V. B. En-gelkes, and C. D. Frisbie, Adv.Mater. , 1881 (2003). F. Chen, X. Li, J. Hihath, Z. Huang, and N. Tao,J.Am.Chem.Soc. , 15874 (2006). M. Magoga and C. Joachim, Phys. Rev. B , 4722 (1997),copyright (C) 2009 The American Physical Society; Pleasereport any problems to [email protected]. J. K. Tomfohr and O. F. Sankey, Phys. Rev. B , 245105(2002). Y. Qi, O. Yaffe, E. Tirosh, A. Vilan, D. Cahen, andA. Kahn, Chem. Phys. Lett. , 344 (2011). S. Kera, H. Setoyama, K. Kimura, A. Iwasaki, K. Oku-daira, Y. Harada, and N. Ueno, Surf. Sci. , 1192(2001). C. M. Whelan, C. J. Barnes, C. G. H. Walker, and N. M. D.Brown, Surf. Sci. , 195 (1999). A. Salomon, T. B¨ocking, O. Seitz, T. Markus, F. Amy,C. Chan, W. Zhao, D. Cahen, and A. Kahn, Adv. Mater. , 445 (2007). F. Thieblemont, O. Seitz, A. Vilan, H. Cohen, E. Salomon,A. Kahn, and D. Cahen, Adv. Mater. , 3931 (2008). H. Shpaisman, E. Salomon, G. Nesher, A. Vilan, H. Co-hen, A. Kahn, and D. Cahen, J. Phys. Chem. C , 3313(2009). J. Delhalle, J.Chem. Phys. , 595 (1974). M. Fujihira and H. Inokuchi, Chem. Phys. Lett. , 554(1972). I. Kitani, K. Yoshino, and Y. Inuishi, Jap.J.App. Phys. ,765 (1980). K. J. Less and E. G. Wilson, J. Phys. Condens Mat ,3110 (1973). J. M. Beebe, B. Kim, J. W. Gadzuk, C. D. Frisbie, andJ. G. Kushmerick, Phys.Rev.Lett , 026801 (2006). J. A. Malen, P. Doak, K. Baheti, T. D. Tilley, R. A. Segal-man, and A. Majumdar, Nano Lett. , 1164 (2009). Y. Selzer, A. Salomon, and D. Cahen, J. Am.Chem.Soc , 2886 (2002). A. Salomon, H. Shpaisman, O. Seitz, T. B¨ocking, andD. Cahen, J. Phys. Chem. C , 3969 (2008). The computational results show that the IDIS/LUMOstates ratio is much larger than the IDIS/HOMO states While these results suggest that the MIM G c could serveas a standard value to determine the effect of the bandbending on the conduction properties of a molecular MISjunction, that is unlikely to be generally valid. The reasonis that G cc