Electrical Characterization of …...semiconductor junctions, however a simple tunneling model (the case for metal insulator–metal junctions) cannot be ruled out, especially when

Ann. N.Y. Acad. Sci.

1006: 36–47 (2003). ©2003 New York Academy of Sciences.doi: 10.1196/annals.1292.002

Electrical Characterization of Metal–Molecule–Silicon Junctions

W. WANG,

a

T. LEE,

a

M. KAMDAR,

a

M.A. REED,

a

M.P. STEWART,

b

J.J. HWANG,

b

AND J.M. TOUR

b

a

Departments of Electrical Engineering, Applied Physics, and Physics, Yale University, New Haven, Connecticut, USA

b

Department of Chemistry and Center for Nanoscale Science and Technology, Rice University, Houston, Texas, USA

A

BSTRACT

: Direct assembly of molecules onto silicon surfaces is of particularinterest for potential employment in hybrid organic–semiconductor devices. Inthe study we report here, aryl diazonium salts were used to assemble covalentlybound molecular groups on a hydride-passivated, oxide-free

n

-type Si(111)surface. The reaction of 4-(trimethylsilylethynyl)benzenediazonium tetrafluo-roborate generates a molecular layer of 4-(trimethylsilylethynyl)phenylene(TMS-EP) on the

n

++-Si(111) surface. The monolayer modifies the electricalproperties of the interface and exhibits nonlinear current–voltage characteris-tics, as compared with the ohmic behavior observed from metal–

n

++-Si(111)junctions. The result of current–voltage measurements at variable tempera-tures (from 300 to 10K) on samples made with the TMS-EP molecule does notshow significant thermally-activated transport, indicating that tunneling is thedominant transport mechanism. The measured data is compared to a tunnel-ing model.

K

EYWORDS

: hybrid organic–semiconductor devices, silicon, diazonium, tunneling

INTRODUCTION

The ability to utilize single molecules as electronic devices has motivatedresearchers for years in the pursuit of miniaturizing electronic circuit elements.

1

Tostudy the electronic transport through molecular layers, a commonly used method isto form metal/molecule/metal junctions.

2–6

Another interesting research direction ofrecent years is to combine functional organic molecules with semiconductors.

7,8

Several groups have studied electronic transport properties for molecule/semicon-ductor junctions using Si,

9–11

GaAs,

12

and organic semiconducting materials

13

asthe substrates. The molecule/semiconductor system is of particular interest forpotential control of electronic properties at the interface. For example, it has beenreported that molecules can modify (in a controllable way) the Schottky barrier in

Addresses for correspondence: M.A. Reed, Departments of Electrical Engineering, AppliedPhysics, and Physics, Yale University, P.O. Box 208284, New Haven, CT 06520, USA. J.M.Tour, Department of Chemistry and Center for Nanoscale Science and Technology, Rice Uni-versity, MS 222, 6100 Main Street, Houston, TX 77005, USA.

[email protected] and [email protected]

37WANG

et al.

: METAL–MOLECULE–SILICON JUNCTIONS

molecule/GaAs systems

12

and molecule–organic semiconductor junctions

13

due tothe dipole moment formed at the interface.

Electrochemical grafting of aryl groups onto carbon

14

and silicon

15

electrode sur-faces from diazonium salt precursors is demonstrated elsewhere via heterogeneousphase reduction.

16

Aryl diazonium salts

17

can be used alone, in the absence of anexternally applied potential, to assemble covalently bound conjugated monolayerson the Si(111):H surface.

18

This type of organic monolayer on Si has demonstratedexceptionally high electrochemical passivation and chemical endurance.

18,19

In ourstudy, 4-(trimethylsilylethynyl)benzenediazonium tetrafluoroborate was used toform the organic layer of 4-(trimethylsilylethynyl)phenylene (TMS-EP) on the Si

FIGURE 1. A. Schematics of device configuration: n++ Si(111) is used as substrate.Backside metal contact is silver that forms an ohmic contact. B. Schematic showing amolecular junction formed in the device area (circled in A). The structure of molecule(TMS-EP) is shown in FIGURE 2A.

38 ANNALS NEW YORK ACADEMY OF SCIENCES

surface. Electrical characterizations such as temperature-dependent current voltage(

I

(

V

,

T

)) measurements were performed to investigate the charge transport mecha-nisms through such a TMS-EP layer.

EXPERIMENTAL

Electronic transport measurements on molecule/silicon junctions were performedusing microfabricated devices. A schematic diagram of such a device is illustrated inF

IGURE

1. Device fabrication starts with highly doped (arsenic as dopant; resistivity

ρ

=

0.001–0.005

Ω

cm)

n

-type Si(111) wafer. After 300nm SiO

2

is grown thermallyon the Si substrate, windows of various sizes (3

µ

m

×

3

µ

m to 200

µ

m

×

200

µ

m) areopened by standard photolithography and subsequent HF etching. The TMS-EPmolecular layer (the schematic for TMS-EP is shown in F

IGURE

2A) is assembledand covalently bound on hydride-passivated Si surface that was made by treating thesamples with NH

4

F for 15minutes, as illustrated in F

IGURE

2. The hydride-terminat-ed Si surface is exposed to a solution of the diazonium salt in anhydrous acetonitrilefor a few hours under an inert atmosphere (a nitrogen filled glove box with sub-1ppmH

2

O and O

2

) in the dark (F

IG

. 2A). The diazonium salts are believed to be activatedby a redox reaction at the open circuit potential of the substrate material in solution,which leads to the local generation of aryl radicals by loss of N

2

and ultimately theformation of irreversible surface–molecule bonds (F

IG

. 2B).To study the quality of the molecular layer, TMS-diazonium layers on large pieces

of Si wafers were analyzed with a single wavelength ellipsometer and a PHI 5700 X-ray photoelectron spectroscopy (XPS) instrument. T

ABLE

1 shows results of layerthickness measurements, indicating that partial multilayer character could be presentin the film. Ellipsometric measurements of molecular layer thicknesses were takenwith a single angle, single wavelength (632.8nm laser) Gaertner Stokes Ellip-someter. XPS spectra used a monochromic Al K

α

source at 350W and were collected

FIGURE 2. A. Synthesis of 4-(trimethylsilylethynyl)benzenediazonium tetrafluorob-orate from 4-(trimethylsilylethynyl)aniline and nitrosonium tetrafluoroborate. B. Reactionof 4-(trimethylsilylethynyl)benzenediazonium tetrafluoroborate salt with the Si(111):Hsurface to generate the TMS-EP monolayer.

39WANG

et al.

: METAL–MOLECULE–SILICON JUNCTIONS

at a 45

°

takeoff angle, giving a sampling depth of approximately 2nm (inelastic meanfree path). XPS spectral values were referenced to Si 2

p

value for a freshly H-termi-nated Si wafer and the C 1s value for adventitious hydrocarbon residue. F

IGURE

3shows the XPS data showing the Si 2

p

region of a TMS-EP-bonded sample madefrom a large piece of Si wafer. The main Si peak is at 99.5eV and the doublet struc-ture is due to 2

p

1/2

–2

p

3/2

spin-orbit coupling. As shown in this figure, the sample isfree of the intense SiO

2

signal that would appear at 103eV,

20

which indicates that themolecule/Si interface is free of silicon oxide.

The samples were then transferred under ambient conditions to a thermal evapo-rator to deposit the top electrode. Five nm Ti followed by 80nm Au were depositedunder a pressure of about 10

−

8

torr at room temperature. The most challenging stepin fabricating molecular junctions (vertical structure, similar to that shown in F

IG

. 1)is to make the top electrical contact. During the fabrication of metal–SAM–metaljunctions, metallic materials deposited on the top of molecules often either penetrate

T

ABLE

1. Surface thickness measurement results for the TMS-EP film

Method Thickness (Å)

Calculation (includes 1.85 Å Si–C bond) 12

±

1

Single wavelength ellipsometry 16

±

3

X-ray photoelectron spectroscopy 18

±

4

FIGURE 3. XPS data showing the Si 2p region of a TMS-EP monolayer reacted sam-ple. The main Si peak is at 99.5 eV and the doublet structure is due to 2p1/2–2p3/2 spin-orbitcoupling. The sample is free of the intense SiO2 signal that would show up at 103 eV.


through the thin molecular layer or contact directly with the substrate via defectsites (such as grain boundaries) in the monolayer, thus causing shorted circuitproblems.

21,22

Therefore a low-temperature evaporation technique is often used forthe top-side metallization in metal–SAM–metal junction fabrications to avoid thethermal damage to the molecular layer by flowing liquid nitrogen through a coolingstage during the evaporation.

2,23

In our study on diazonium/Si devices, we used bothlow temperature and room temperature evaporations and we observed device yieldsof about 6

%

(three working devices out of total 53 fabricated devices) and about 32

%

(30 working devices out of total 94 fabricated devices) for low and room temperatureevaporations, respectively. (We define a working device as a sample that is neitheropen, current less than 1pA at 1V, nor short, current greater than10mA at 0.5V, andshows nonlinear

I

(

V

) characteristics.)In this paper we focus on devices fabricated via room temperature metallization.

The fabricated devices were measured on a probe station and subsequently packagedfor I(V,T) measurements in a Janis cryostat. Two-terminal DC I(V) measurementswere performed using a HP4145B semiconductor parameter analyzer.

RESULTS

Current–Voltage (

I

(

V

)) Characteristics

F

IGURE

4A shows a typical

I

(

V

) characteristic for control samples without mole-cules that exhibits nearly ohmic behavior with a resistance of 10

Ω

(specific contactresistance of approximately 10

−

6

Ω

cm

2

) due to the highly doped silicon substrate.All of the metal/TMS-EP/Si junction devices showed nonlinear

I

(

V

) characteristicswith significantly suppressed current densities. Representative

I

(

V

) data (measuredat room temperature) for three molecular devices are shown in F

IGURE

4B. In thisfigure positive bias corresponds to electron injection from the bottom chemisorbedmolecule–silicon contact. Current densities for molecular junctions (

J

≈

5–10A/cm

2

at 1V; devices for F

IG

. 4B) are reduced roughly by five orders of magnitude as com-pared with that for the control sample (

J

≈

10

5

A/cm

2

at 0.4V; corresponding to2.5

×

10

5

A/cm

2

at 1V; device for F

IG

. 4A) due to the molecular layer acting as aninsulating barrier.

We observed device-to-device variations of current densities and

I

(V) shapes.FIGURE 5 shows a histogram plot for observed current densities (at 1V) of all theworking molecular devices. Note that J (abscissa) is on log scale in this plot.Although there is a wide range in current densities, most devices have J values of3−300A/cm2 at 1V. The device-to-device variation can be attributed to fluctuationsin the actual device size, formation of partial multilayer (TABLE 1), or difference incontact geometry introduced during fabrication. The detailed metal–molecule con-tact configuration has been reported to play an important role in the conductance ofmetal–molecular junction.24

Most I(V) characteristics showed asymmetric behavior with current at positivebias being larger than that at negative bias (as shown in FIG. 4B). For example, theaverage value of the asymmetric ratio, R = (I at 1V)/(I at –1V), was found to beabout 2.7 and about 24.9 for two different fabrication runs, which also indicates thatthere is a variation in fabrication process. Molecular devices have been observed to

41WANG et al.: METAL–MOLECULE–SILICON JUNCTIONS

FIGURE 4. A. Typical I(V) result for the control device showing ohmic behavior with aresistance of 10Ω. B. Representative I(V) data for three metal–TMS-EP–Si junction devices.


degrade.21 In our case, typically current densities decreased over a period of onemonth whereas I(V) shapes remained similar.

Schottky Barrier Model versus Tunneling Model

One tends to use a Schottky barrier model to explain I(V) of metal–molecule–semiconductor junctions, however a simple tunneling model (the case for metal–insulator–metal junctions) cannot be ruled out, especially when a highly doped semi-conductor is used as the substrate as in our case. Schottky barrier model and tunnel-ing model exhibit distinct temperature dependencies of their transportcharacteristics, as expressed in Eq. (1)25 and (2),23,26 respectively:

(1)

where A* is the Richardson constant, k is the Boltzmann constant, T is temperature,ΦS is the Schottky barrier height, V is the applied bias, n = 1/(1 – β) is the idealityfactor, and TN is the transmission probability for tunneling through the molecularlayer. Current density from the tunneling model is expressed by23,26

(2)

J A* T 2T N

q ΦS βV+( )–

kT--------------------------------⎝ ⎠

⎛ ⎞exp qVkT-------⎝ ⎠

⎛ ⎞exp 1–⎩ ⎭⎨ ⎬⎧ ⎫

,=

J e

4π2hd2------------------⎝ ⎠

⎛ ⎞ ΦreV2

-------–⎝ ⎠⎛ ⎞ 2 2m

h---------------– α Φr

eV2

-------– d

ΦreV2

-------+⎝ ⎠⎛ ⎞ 2 2m

h---------------– α Φr

eV2

-------+ dexp–

exp⎩

⎭

⎨

⎬

⎧

⎫,

=

FIGURE 5. A histogram plot for observed current densities of all the working molec-ular devices.


where m is the electron mass, h (= 2π ) is Planck’s constant, d is the tunneling bar-rier width, ΦT is the tunneling barrier height, and α is a unitless parameter thataccounts for either a non-rectangular barrier or an effective mass.23 The devicestructures that can be explained by the Schottky barrier model (Eq. (1)) require sig-nificant temperature dependence in their I(V) characteristics,25,27 whereas devicesthat follow the tunneling model exhibit no dependence on temperature.

To determine the conduction mechanism, we performed I(V,T) measurements onthree molecular devices that were chosen randomly and have different current den-sities. The I(V,T) results are summarized in TABLE 2 with their current densities andtemperature-dependencies. The measured I(V,T) data and the corresponding Arrhe-nius plot are shown in FIGURE 6 for device #3 (from TABLE 2). We observed no orlittle temperature dependence for any of the devices, indicating that the main con-duction mechanism is tunneling. Device #1 showed no temperature dependence inI(V) values measured from 300 to 10K, whereas devices #2 and #3 (measured from300 to 100K) showed little temperature dependence with a weak thermal activation.The thermal activation barrier height ΦTH (from I ∝ exp(–ΦTH/kT)) was determinedto be about 15meV and about 10meV for devices #2 and #3, respectively. Comparedwith such weak thermal barriers, thermionic Schottky barrier height for mole-cule/semiconductor systems is reported to be larger than 500meV.10

Schottky barrier type conduction is expected when a lightly doped semiconductormaterial is used as the substrate; here, tunneling is the dominant transport mecha-nism since a highly doped n-Si substrate is used in our study. The electron conduc-tion is expected to be tunneling when the Fermi levels of contacts lie within theHOMO-LUMO gap (HOMO: highest occupied molecular orbital, LUMO: lowestunoccupied molecular orbital) of a short-length organic molecule.28 The met-al/TMS–EP/Si structure studied in our case is similar to metal–insulator–metalstructure, hence tunneling is expected to be the dominant conduction mechanism.Since XPS data (FIG. 3) indicates that the molecule–Si interface is free of siliconoxide, the tunneling result obtained from I(V,T) characterizations is due to the TMS-EP molecular structure.

The measured I(V) data is fitted with the tunneling model, and a representativefitting result is shown in FIGURE 7 for a device having J ≈ 3A/cm2 at 1V. Obtainedfrom TABLE 1, 16Å was used as the tunneling gap width for this fit. For this device,from a nonlinear least squares fit using Eq. (2), the optimum fitting parameters werefound to be ΦT = 2.15eV and α = 0.90 (corresponding to m* = 0.81m) for the positivebias region and ΦT = 2.41eV and α = 0.85 (corresponding to m* = 0.72m) for thenegative bias region.

h

TABLE 2. Summary of I(V,T) measurement results

NOTE: There is no scaling correlation of current densities with device size.

Size (µm)

J (A/cm2 at 1 V)

Temperature Dependence

Thermal Activation Barrier (meV)

Device #1 5 × 5 10−3 no —

Device #2 10 × 10 0.1 weak 15

Device #3 25 × 25 10 weak 10


FIGURE 6. A. I(V,T) measurement result for device #3 (TABLE 2). I(V) data at tempera-tures from 300 to 110K in 10K steps are plotted on a log scale. B. Arrhenius plot generated fromthe I(V) data in (A) showing little temperature dependence with a weak thermal activation.


Since the dominant transport mechanism in our metal/TMS-EP/Si system is tun-neling, this indicates that there are few defects in our molecular layers. Otherwisesignificant temperature-dependent defect-mediated transport behavior would occur.

CONCLUSION

In conclusion, we have fabricated metal–molecule–Si junctions by directlyassembling covalently bound molecules (TMS-EP) onto highly doped n-typeSi(111) surface. Current–voltage measurements on these junctions showed nonlin-ear electrical characteristics with suppressed current densities, as compared withohmic behavior observed in the control samples (without molecules). The results oftemperature-dependent current–voltage measurements indicate that the dominantconduction mechanism is tunneling, rather than Schottky barrier type conduction.

ACKNOWLEDGMENTS

We thank James F. Klemic, Zhijiong Liu, and Ilona Kretzschmar for discussions.This work was supported by DARPA/ONR (N00014-01-1-0657), ARO (DAAD19-01-1-0592), AFOSR (17496200110358), and NSF (DMR-0095215).

FIGURE 7. Measured I(V) data () are compared with calculated values (solid curve)using tunneling model with the optimum fitting parameters of ΦT = 2.15 eV and α = 0.90(corresponding to m* = 0.81 m) for positive bias region (Fit (+)) and ΦT = 2.41 eV andα = 0.85 (corresponding to m* = 0.72 m) for negative bias region (Fit (–)). Current is plottedon log scale.


REFERENCES

1. AVIRAM, A., et al. Eds. 1998. Molecular Electronics: Science and Technology. Ann.N.Y. Acad. Sci. 852: The New York Academy of Sciences, New York.

2. METZGER, R.M., et al. 1997. Unimolecular electrical rectification in hexadecylquino-linium tricyanoquinodimethanide. J. Am. Chem. Soc. 119: 10455–10466.

3. REED, M.A., et al. 1997. Conductance of a molecular junction. Science 278: 252–254.4. CHEN, J., et al. 1999. Large on–off ratios and negative differential resistance in a

molecular electronic device. Science 286:1550–1552.5. COLLIER, C.P., et al. 2000. A [2]Catenane-based solid state electronically reconfig-

urable switch. Science 289: 1172–1175.6. PARK, J., et al. 2002. Coulomb blockade and the Kondo effect in single-atom transis-

tor. Nature 417: 722–725.7. ULMAN, A. 1996. Formation and structure of self-assembled monolayers. Chem. Rev.

96: 1533–1554.8. BURIAK, J.M. 2002. Organometallic chemistry on silicon and germanium surfaces.

Chem. Rev. 102: 1271–1308.9. BOULAS, C., et al. 1996. Suppression of charge carrier tunneling through organic self-

assembled monolayers. Phys. Rev. Lett. 76: 4797–4800.10. SELZER, Y., et al. 2002. The importance of chemical bonding to the contact for tunnel-

ing through alkyl chains. J. Chem. Phys. B 106: 10432–10439.11. LI, D., et al. 1998. Conduction properties of metal/organic monolayer/semiconductor

heterostructure. Appl. Phys. Lett. 73: 2645–2647.12. VILAN, A., et al. 2000. Molecular control over Au/GaAs diodes. Nature 404: 166–168.13. CAMPBELL, I.H., et al. 1996. Controlling Schottky energy barriers in organic electronic

devices using self-assembled monolayers. Phys. Rev. B 54: 14321–14324.14. DELAMAR, M., et al. 1992. Covalent modification of carbon surfaces by grafting of

functionalized aryl radicals produced from electrochemical reduction of diazoniumsalts. J. Am. Chem. Soc. 114: 5883–5884.

15. ALLONGUE, P., et al. 1998. Organic monolayers on Si(111) by electrochemical method.Electrochim. Acta 43: 2791–2798.

16. ALLONGUE, P., et al. 1997. Covalent modification of carbon surfaces by aryl radicalsgenerated from the electrochemical reduction of diazonium salts. J. Am. Chem. Soc.119: 201–207.

17. KOSYNNKIN, D.V., et al. 2000. Phenylene ethynylene diazonium salts as potential self-assembled molecular devices. Org. Lett. 3: 993–995.

18. STEWART, M.P., et al. 2003. One-step deposition of covalently bound aryl groups onPd, GaAs, and Si. Submitted.

19. HARTIG, P., et al. 2002. Engineering of Si surfaces by electrochemical grafting ofp-nitrobenzene molecules. Appl. Phys. Lett. 80: 67–69.

20. NIST X-ray photoelectron spectroscopy database version 3.2, <http://srdata.nist.gov/xps/>.

21. LEE, J.-O., et al. 2003. Absence of strong gate effects in electrical measurements onphenylene-based conjugated molecules. Nano Lett. 3: 113–117.

22. KAGAN, C.R., et al. 2003. Evaluation and considerations for self-assembled monolayerfield-effect transistors. Nano Lett. 3: 119–124.

23. WANG, W., et al. Mechanism of electron conduction in self-assembled alkanethiolmonolayer devices. Phys. Rev. B 68: 035416-1–035416-7.

24. DI VENTRA, M., et al. 2000. First-principles calculation of transport properties of amolecular device. Phys. Rev. Lett. 84: 979–982.

25. RHODERICK, E.A., et al. 1988. Metal–Semiconductor Contacts. Clarendon Press,Oxford.

26. SIMMONS, J.G. 1963. Generalized formula for the electric tunnel effect between similarelectrodes separated by a thin insulating film. J. Appl. Phys. 34: 1793–1803.

27. SZE, S.M. 1981. Physics of Semiconductor Devices. J. Wiley & Sons, New York.


28. RATNER, M.A., et al. 1998. Molecular wires: charge transport, mechanisms, and con-trol. In Molecular Electronics: Science and Technology. A. Aviram & M. Ratner,Eds.: 22-37. Annals of the New York Academy of Sciences, Vol. 852. The NewYork Academy of Sciences, New York.

Documents

Electrical Characterization of …...semiconductor junctions, however a simple tunneling model (the case for metal insulator–metal junctions) cannot be ruled out, especially when