JOURNAL OF APPLIED PHYSICS 99, 104315 2006

Electronic transport of n-type CdSe quantum dot lms: Effect of lm treatment

Dong Yu, Brian L. Wehrenberg, Praket Jha, Jiasen Ma, and Philippe Guyot-Sionnesta
James Frank Institute, University of Chicago, 5640 S. Ellis Avenue, Chicago, Illinois 60637

Received 28 September 2005; accepted 8 March 2006; published online 31 May 2006 Solid lms of CdSe colloidal quantum dots are conductive when reduced by electrochemistry. The dark conductivity and the photoconductivity of these n-type CdSe nanocrystalline solids have been investigated as a function of lm treatment. When the lms are treated with NaOH, a carrier mobility of 102 cm2 / V s and a charging/discharging time of 10 ms are obtained. Under an excitation intensity of 0.75 mW/ cm2, a signicant photoconductivity of 103 S cm1 has been observed. The high mobility allowed demonstrating the operation of a eld effect transistor. 2006 American Institute of Physics. DOI: 10.1063/1.2192288
I. INTRODUCTION

Semiconductor nanocrystals,1,2 with tunable electronic and optical properties, have shown promises for applications in biological imaging,3,4 light emitting devices,5,6 quantum dots lasers,79 photodetectors,10 and solar cells.11,12 Some of the envisioned optoelectronic device applications using the nanocrystals are based on solid state materials where the nanocrystals are assembled into close-packed nanocrystalline solids. Although the optical properties of nanocrystals have been extensively studied, the electronic properties of these nanocrystalline solids are not as much studied. One major reason has been their very poor conductivity. The individual nanocrystals in the solid are indeed separated by a layer of organic ligands which signicantly inhibit charge transport. For example, for CdSe dots capped by trioctylphosphine oxide/trioctylphosphine, the separation between the dots is reportedly 11 .13 The conductivity of the nanocrystal thin lms is then14 as low as 1014 S cm1. This extremely low conductivity would seem to preclude the use of such solids in any device requiring some measure of electronic transport. While intrinsic nanocrystal lms are extremely resistive, it is now known15 that reduction using potassium or electrochemistry leads to many orders of magnitude increase in conductivity. The quantized level structure of the quantum dots also shows up in the conductance with a modulated conductance and mobility as a function of lling of the quantum states. Furthermore, by treating the lms with various molecular cross linkers of different lengths and conjugation, it was possible to signicantly improve the conductivity up to 102 S cm1. Annealing of the nanocrystalline solids can also reduce the distance between the nanocrystals and improve the conductivity although the quantized level structure might be lost in the process.16 Recently it was reported that treating intrinsic lms of quantum dots with bases such as NaOH and 1-butylamine led to orders of magnitude improvement of the low voltage photoconductivity.17 We present here a study of how various treatments affect the dark conductivity and photoconductivity of the reduced
a

lms. Also, we report on the demonstration of a eld effect transistor based upon an optimized lm preparation giving a high mobility.
A. Film treatment and mobility

Electronic mail: pgs@uchicago.edu

In the following experiments, CdSe nanocrystals with diameter 5.3 nm were used. They are synthesized with dimethylcadmium as an organometallic precursor, following the literature.18 The nanocrystal thin lms are made on a glass substrate previously patterned with interdigitated Pt electrode. The electrodes consist of 50 pairs of 5 m wide, 5 mm long, and 110 nm thick Pt pads with 5 m gap. These substrates are made by Abtech Scientic. To bind the nanocrystal lm to the substrate, the clean substrate is rst dipped into 1% volume ratio 3-aminopropyltriethoxysilane APS acetone solution and then 1% volume ratio 1,6hexanedithiol HDT acetone solution. The silane end of APS should bind to the glass with the amino end providing a binding site for the nanocrystals. Similarly, the dithiol is used to provide a good afnity between the Pt electrodes and the nanocrystals. Without these surface treatments, the nanocrystal lms occasionally falloff the substrate when dipped into subsequent solutions. The lm itself is made by simply drop casting a hexane:octane 10:1 solution of the nanocrystals on the substrate. Immediately after drying, the lm is then dipped into a cross-linking solution for treatment. The effect of this treatment is to render the lms insoluble in their original solvent and to vastly improve the electrochemical19 and conducting properties. It also leads to shrinkage of the lms, consistent with the expected shorter distance between dots. For example, a heptanediamine linker should reduce the interdot distance to 8 . While molecules were initially chosen to have two binding sites so that they would a priori chemically bond the nanocrystals,19 it has, since then, been realized that this is not necessary17 as any treatment that can replace the original capping molecules with much shorter groups will increase the particle-particle interaction and lead to insoluble lms. The solutions used for cross linking are all in methanol which is a solvent for the original caps trioctyl phosphine/trioctylphosphine oxide TOP/TOPO. We stud 2006 American Institute of Physics

0021-8979/2006/9910/104315/7/$23.00

99, 104315-1

104315-2

Yu et al.

J. Appl. Phys. 99, 104315 2006

FIG. 1. Atomic force microscopy AFM of a typical CdSe nanocrystal thin lm treated with 1,7-heptanediamine. The size of the picture is 3 m. The lower plot is the cross section in the middle of the top picture.

FIG. 2. Cyclic voltammetry of TOP/TOPO capped CdSe nanocrystal lms of 5.3 nm diameter with different treatments at 60 C. Solid lines represent conduction currents, dashed lines electrochemical currents, dots 1S bleaches, and triangles 1 P bleach. The rst three lms estimated by their maximum 1S bleaches are about 12, 15, and 28 layers. The last one is similar in thickness judged by the optical density before doping. a Treated with pure methanol: scan rate of 10 mV/ S and bias of 40 mV. b Treated with 1,7-heptanediamine in methanol solution: scan rate of 20 mV/ S and bias of 5 mV. c Treated with 1-butylamine in methanol solution: scan rate of 10 mV/ S and bias 10 mV. d Treated with NaOH in methanol solution: scan rate of 50 mV/ S and bias 5 mV.

ied pure methanol, or methanol with 1% volume ratio 1,7heptanediamine, 1% volume ratio 1-butylamine, and 0.08M NaOH methanol solution. For all solutions the dipping time was kept at 10 min and the treated lms were then baked at 100 C under vacuum for 1 h. This step removes the water and accelerates the recapping. Subsequently, the lms are moved into a N2 glovebox and rinsed with methanol to remove the extra ligands. After a second drying step at 70 C under N2 for 10 min, the lms are assembled in the electrochemical cells. With atomic force microscopy AFM, the close packed lms look smooth and the roughness is about the diameter the nanocrystals Fig. 1. The thickness of the lms is usually in the range of tens of monolayers, monitored by AFM or simply by the optical density. The cyclic voltammetry is taken at 60 C and the occupation number is monitored by measuring the spectral bleaches of the S and P excitonic transition. Figure 2 shows the typical cyclic voltammetry of TOPO/TOP capped CdSe nanocrystal lms with different treatments. The sample treated with pure methanol shows a large hysteresis in the optical bleach indicating a slow charging/ discharging rate and the lm remains insulating Fig. 2a. Therefore, although cross linking with pure methanol is sufciently effective to lead to lms that present an electrochromic response, the response is slow. This is presumably because methanol removes capping molecules without replacing them by a ligand that passivates the exposed Cd surface sites. This would lead to a large number of trapping sites which also lower the conductivity below the detectable range.

As reported previously,15 the lm treated with 1,7heptanediamine, for which the amine is a strong ligand to the Cd sites, shows a much faster and reversible optical response while the conductance improves drastically with a shoulder appearing at about the half lling of the 1Se shell and the rst wave on the reduction or oxidation curve Fig. 2b. Although 1-butylamine is monodendate, it leads to further improvements in the system response,17 and the lms exhibit fast and reversible electrochemistry while they are about 100-fold more conductive than with heptanediamine. As observed so far with the more conductive lms, the 1Se conductance shoulder is present but less pronounced Fig. 2c. The OH base can also passivate the Cd surface sites, facilitating the removal of the original ligands and presumably forming cadmium hydroxide complexes. It is the shortest ligand such that the nanocrystals can get to their closest approach. In contrast to the amine treated lms, the lm treated with NaOH are as conductive but cannot be charged completely. The 1S optical bleach is smaller than 0.005 optical density OD, indicating less than 10% of the possible charge in the 1Se state, and the 1 P bleach is not seen. The complete electrochemical doping of the whole lm requires easy ow of the counterions throughout the lm network, in concert with electron transfer to the nanocrystal lms. Since the electrochemical current is small Fig. 2d for these lms, it suggests that the counterions do not diffuse easily in these lms, possibly because they are too compact. The visible absorption spectra of the nanocrystal thin lm after NaOH treatment still show the excitonic features, therefore the dots are still distinct Fig. 3a. No obvious change in optical density further indicates that the nanocrystals are not

104315-3

Yu et al.

J. Appl. Phys. 99, 104315 2006

FIG. 3. The effect of NaOH treatment on the CdSe nanocrystal lms of 5.3 nm diameter. Solid lines are before treatment and dashed lines are after treatment with 0.08M NaOH methanol solution. a UV-visible absorption spectrum shows that the rst exciton peak redshifts of 11 nm while the optical density remains the same. b Infrared absorption spectrum indicates a 60% reduction of the optical density at hydrocarbon stretching bands. c AFM of a thin lm on glass before and after treatment. The cut is made by a razor blade on the right of the graph. The surface of the lm remains smooth while the thickness shrinks by 30% 40%. d A cartoon of how the NaOH treatment may remove the capping molecules and shorten the interdot distance.

signicantly washed off by NaOH treatment. The rst excitonic peak redshifts by about 11 nm. This redshift is a possible indication of increased interaction between the nanocrystals. Infrared spectra of the hydrocarbon stretching bands show that their optical density in the infrared spectrum is reduced by 60% after treatment Fig. 3b. The AFM shows a 30%40% shrinking of the lm thickness while the surface of the lm remains smooth Fig. 3c. These evidences all concur to show that a signicant part of capping molecules is removed and support the notion of a very compact lm. Temperature dependent conductivity provides further evidence of the increased coupling between nanocrystals in these NaOH treated lms. The n-type CdSe nanocrystal thin lm treated with NaOH follows variable range hopping VRH theory20 as well as all the samples prepared here. However the T1/2 conductance gives a temperature scaling of T* 2000 K Fig. 4 which is about 2.5 times smaller than for heptane diamine treated lm. In the VRH model, T* is inversely related to the static dielectric constant and the localization length. A dielectric constant of 4 was used to describe the VRH behavior in a heptanediamine treated lm.20 The lms treated with NaOH are denser and should have a higher dielectric constant, closer to bulk value of 10.2. In addition, T* may also be reduced by a larger localization length of the electron wave function. Whichever of these two effects dominate, the strongly reduced T* is indicative of a stronger coupling between the nanocrystals.

FIG. 4. Temperature and electrical eld dependent conductances of a CdSe nanocrystal thin lm with diameter of 5.3 nm treated with NaOH. A Conductance measured at 0.1 V bias as a function of temperature. The linear tting gives T* = 2.0 103 K. B Conductance as a function of electrical eld at different temperatures. The straight line corresponds to E* = 6.4 107 V / m.

The charging and discharging of the NaOH treated lms are very fast. As the electrochemical potential is stepped from a potential where the lm is insulating to about 0.2 V above where the lm starts to conduct, both the conductance and electrochemical current change abruptly Fig. 5. This behavior is rather well captured by a simple RC circuit model of the electrochemical gating switching. The conductance is proportional to the mobility and the charge Q. It should follow that G = Q 1 expt / during charging and G = Q expt / during discharging, where is the RC switching time of the gate and lm. The charging current should also follow I = Q / expt / with an opposite sign for discharging. Focusing on the rst 100 ms after the potential jump, the conductance and the electrochemical current are tted by exponentials Fig. 6. The time constants are all around 10 ms, slightly slower for discharging than charging and slightly slower for the conductance change than the electrochemical current change. The mobility is estimated from the conductance and the integrated charge, given by = l2G / Q, where the channel length l is 5 m. For the NaOH treated lm, the mobility is then measured to be 102 cm2 / V s. This value is not inconsistent with the estimation based upon the cyclic voltammetry Fig. 2d given

104315-4

Yu et al.

J. Appl. Phys. 99, 104315 2006

FIG. 5. Measurement of the response of CdSe nanocrystal thin lm treated with NaOH during charging and discharging at 60 C. a The electrochemical potential is stepped from 0.4 to 1 V. b The conductance of the lm is measured with a 5 mV bias between two working electrodes. c The electrochemical current pulses.

FIG. 6. Focus of Fig. 5 around the potential jump. The dashed lines are data and the solid lines are tting. a The charging of the lm: the conductance is tted by G = G0 1 expt / 1 with tting parameters G0 = 4.1 mS and 1 = 16.8 ms; the electrochemical current is tted by I = I0 expt / 2 + Ib with I0 = 9.4 A, 2 = 9.4 ms, and Ib = 1.1 A. b The discharging of the lm: the conductance is tted by G = G0 expt / 1 + Gb with G0 = 5.4 mS, 1 = 12.6 ms, and Gb = 0.17 mS; the electrochemical current is tted by I = I0 exp t / 2 + Ib with I0 = 10 A, 2 = 8.3 ms, and Ib = 0.8 A.

B. Photoconductivity

the poor denition of the electrochemical current, and the absence of independent conrmation of the charge occupation by the optical bleach data. For a eld effect transistor FET,21 there is a relation between the mobility and the cutoff of the frequency response 3 dB point f , given by f = VG VT / 2l2, where VG is the gate voltage and VT is threshold voltage. In our electrochemical experiment, VG VT is estimated to be about 0.2 V. With 102 cm2 / V s the cutoff frequency is then f 1.3 kHz corresponding to a switching time constant 2 f 1 of 0.12 ms. We note that this is much faster than the measured 10 ms. This discrepancy is attributed to the charging time which is more likely limited by the slower counterion diffusion time through the lm. With the same pulsed response method, the mobility and the charging/discharging time constants of CdSe nanocrystal thin lms treated with 1,7-heptanediamine and 1-butylamine were also determined. The result is summarized in Fig. 7a. The mobility and the switching time vary by orders of magnitude with different treatments. Only the 1,7-heptane diamine treated sample shows a time constant close to the calculation based upon the FET relation. The charging time for lms with higher mobility is slower than the limit. As for the NaOH treated lms, we attribute this deviation to the slower diffusion rate of the counterions. This is an intrinsic problem for an electrochemical cell. To achieve faster response, all solid state devices such as FET are necessary.

Photoconductivity has been observed previously in intrinsic semiconductor nanocrystalline solids22,23 but it has not been studied in an n-type system. Since reduced nanocrystals exhibit vastly increased mobilities, one would expect similar improvements in the photoconductivity. To investigate the photoconductivity of the lms, we used the same lm preparation on the interdigitated elec-

FIG. 7. a The charging time vs mobility for the CdSe nanocrystal thin lms treated with 1,7-heptanedamine circles, 1-butylamine triangles, and NaOH squares. The solid line is = l2 / VG VT with the channel length l = 5 m and VG VT = 0.2 V. b The photoconductivity vs mobility. The solid line corresponds to = en with n 0.01 e/dot.

104315-5

Yu et al.

J. Appl. Phys. 99, 104315 2006

FIG. 8. Photoconductivity of NaOH treated CdSe nanocrystal thin lm of diameter of 5.3 nm after excitation at different electrochemical potentials. The light source is a 25 W lament with a lter of 10 cm away from the sample. The light intensity at the sample is 0.75 W / cm2 with wavelengths of 560 630 nm. Circles stand for dark conductivity and squares for conductivity increase after excitation.

trodes as presented above. The light source is an incandescent light bulb controlled by a dc power supply. The light bulb is placed 10 cm away from sample. The light is ltered by a bandpass lter covering 560 630 nm, and the excitation intensity is measured to be 0.75 mW/ cm2. We rst discuss lms treated with 1, 7-heptanediamine. With an applied electrical eld of 103 V / m, corresponding to a 5 mV bias across the 5 m electrode, the photoconductivity of the insulating intrinsic lms is not observable in our setup. When the lms are slightly reduced, the illumination introduces an increase in conductivity of the order of 108 S cm1. Assuming the electrons are the majority carriers and using a mobility of 106 cm2 / V s as deduced earlier, the electron density is increased by n = / e 1017 cm3 or 0.01 e/dot. From the light intensity and the nanocrystals cross section24 of 3 1015 cm2, the generation rate G should be 7 excitons/ dot s. Thus, the carrier must have a lifetime = n / G 103 s. Similarly long lived delocalized electron states after photoexcitation have been observed in the CdSe nanocrystals capped with p-thiocresol and were attributed to efcient hole traps.25 Although here the nanocrystals are not capped with thiols, the reducing environment should result in efcient hole traps. The higher mobility lms treated with 1-butylamine and NaOH show much more pronounced photoconductivity. Figure 8 demonstrates the dark conductivity and photoconductivity at different potentials for a lm treated by NaOH. Jarosz et al. reported that the intrinsic lm treated with NaOH shows17 a photoconductivity of 108 S cm1 with an excitation power of 14 mW/ cm2 and an electrical eld of 4 107 V / m. The photoconductivity of the n-type lm also treated with NaOH is ve orders of magnitude larger with 103 S cm1 under a much weaker excitation intensity only 0.75 mW/ cm2 and with a much lower electrical eld of 103 V / m. The difference is because the electron traps, which signicantly lower the mobility, are lled by the electrochemical doping. The improved photoconductivity of lms with different treatments is summarized in Fig. 7b.

FIG. 9. Photoconductivity of the n-doped CdSe nanocrystal thin lm at low temperatures. a IV curves of the thin lm with a fairly high electronic density. b IV curves of the same lm with a low electronic density. c and d The plots of a and b in half log scale, respectively. Solid lines are dark current and dashed lines are photocurrent excited with a 3 V 0.2 A light bulb. The curves with circles are taken at 4.3 K, squares at 10 K, and triangles at 30 K. e and f Time trace of the normalized conductance of the lm with a low electronic density. The photoinduced conductance response is at least as fast as the 100 ms response time of the light bulb.

Recombination rates are slowed at low temperature, such that the ratio of the photoconductivity to the dark conductivity should increase. This is indeed what an initial set of experiments has revealed. To measure the photoconductivity at low temperature, we made an electrochemical cell that remained transparent after freezing. The interdigitated Pt electrode is covered by a thin glass slide with a 10 m spacer. It is vertically inserted in a reservoir of the tetrabutylamonium perchlorate/propylene carbonate TBAP/PC electrolyte solution, which rises into the gap between the electrode and the glass slide because of the capillary force. At low temperature the frozen electrolyte solution in the small gap makes a transparent glass. On the other hand, the spacing is still large enough to allow the diffusion of the electrolyte such that one is still able to perform electrochemistry and charge the nanocrystals before freezing. To illuminate the sample cell inside the dewar, a small light bulb Magnite 3 V 0.2 A with measured intensity at the sample of 30 mW/ cm2 is wrapped around this electrochemical cell with Teon tape and the whole cell is inserted into a liquid helium tank. Figure 9 shows the photoconductivity of a lm treated

104315-6

Yu et al.

J. Appl. Phys. 99, 104315 2006

with NaOH at low temperatures. The ratio of the photoconductivity to the dark conductivity is indeed much improved at the lower temperatures. At 4.3 K, the conductance of the lm with a low electronic density increases by four orders of magnitude Fig. 9d upon illumination. In summary, n doping allows large increases in mobility and corresponding increase in photoconductivity while lower temperatures improve the ratio of the photoconductance to the dark conductance.
C. Semiconductor nanocrystalline solid eld effect transistor

Electrochemical gating allowed studies of charging with well-dened potentials that can be to a good accuracy reported to an absolute scale, but it certainly does not present the convenience of solid state gating. Solid state FET has the great advantage of not requiring an electrolyte, it is easily compatible with a sealed or vacuum environment, the Fermi level can be adjusted at any temperature, and it does not suffer from speed limitation associated with the motion of bulky counterions. However, doping nanocrystals by the solid state gate method is quite difcult. The electrical eld required to singly charge only one monolayer of nanocrystals is so large E = / 0 4 108 V / m if taking SiO2 as the dielectric that it can easily break down the dielectric material. Given such a small charge density, a high mobility is necessary. Previous work showed that a polycrystalline CdSe thin lm FET could be made after sintering the nanocrystals thin lm at high temperature.26 Here, with NaOH treatment, the CdSe nanocrystal thin lms have a large enough mobility that we have been able to detect conduction in a solid state FET while the lms retain the quantum dot integrity. Our FET device is made by spin coating a polymethylmethacrylate PMMA layer27 on top of the NaOH treated nanocrystal thin lm, then evaporating a gold layer as a gate electrode. The thickness of PMMA layer is 1 m and it is baked at 170 C for 30 min under N2 to prevent the oxidation of nanocrystals. All the measurements are performed at room temperature and in a dark N2 chamber. Figure 10 shows the current-voltage characteristics of the device. As the gate voltage is set at +10 V, source-drain conductance is up to 0.5 nS Fig. 10b. The interface charge density is estimated to be q 3 108 C / cm2 using q = 0V / L with a PMMA dielectric constant = 4 and a thickness L = 1 m. Considering a monolayer of nanocrystals with 5 nm diameter as an n-type channel, this surface charge density corresponds to at most 0.05 e/dot. Although this is a small degree of charging, the photoconductivity experiments described above had showed that even 0.01 e/dot can change the conductance by orders of magnitude. In this FET device, the mobility is given by = Gl / wq, where G is the conductance, l is the channel length, and w is the channel width. Using the estimated q from above, the mobility is measured to be very small 2 107 cm2 / V s, considerably smaller than 102 cm2 / V s obtained for the electrochemically doped lm. This discrepancy can arise from a number of reasons, such as defects introduced in the device preparation, but we believe that it is mostly because many of the gate charges ll trap states, such that the fraction of q

FIG. 10. Current-voltage characteristics of a CdSe nanocrystal FET device. a Drain-source current as a function of the drain-source voltage for several values of the gate-source voltage. The scan rate is 0.5 V / s and the measurement is started right after the gate voltage is applied. The amplier saturates at 2 nA. b Source-drain voltage is xed at 4 V and the gate-drain voltage is scanned 0.1 V / s. c With drain electrode open, gate-source capacitive current as a function of the gate-source voltage. The scan rate is 0.5 V / s.

that goes into the nanocrystal quantum states is actually much smaller than 1. We veried that a negative gate voltage only reduces the conductance indicating that p-type conducting channel is not achieved in this device. We also veried that the gate to source drain current is small and it can be in part accounted by the 0.1 nF capacitance caused by the PMMA layer Fig. 10c. This nanocrystal lm FET is different from the conventional devices by its memory effect. The source-drain conductance is steadily decaying after a gate voltage is applied. This effect can be observed from the hysteresis of the sourcedrain current Fig. 10b and is more pronounced when a much higher gate voltage up to 60 V is applied Fig. 11. With a 60 V gate voltage, the source-drain conductance is up to 20 nS but the hysteresis is also bigger Figs. 11a and 11b. The capacitive current remains much smaller than the source-drain current Fig. 11c. After a large gate voltage is applied, the threshold voltage above which the lm conducts is increased Fig. 11a and the system remembers for hours. Memory effects have been observed for nanocrystal ensemble28 and carbon nanotube FET.29 Presumably, the large fraction of the charge that does not get into the quantum states is instead taken up by traps. These traps are weakly coupled to source and drain, and this leads to a large hysteresis and the memoryeffect. When and if well controlled these effects could potentially be used for memory

104315-7

Yu et al.

J. Appl. Phys. 99, 104315 2006

show much improved photoconductivity compared to the intrinsic lms, consistent with the high mobilities that are achieved. Finally, based on this treatment, a CdSe nanocrystal thin lm eld effect transistor is demonstrated. Optoelectronic devices based on quantum dot thin lms with optimized postcasting lm treatment will be able to provide improved performance such as higher mobility and faster switching time, while still retaining the unique properties of nanocrystals.
ACKNOWLEDGMENTS

One of the authors D.Y. was supported by the University of ChicagoArgonne National Laboratory Consortium for Nanoscience Research. Another author B.L.W. was supported by the U.S. National Science Foundation NSF under Grant No. DMR-0407624, and the work was supported by the University of Chicago MRSEC NSF-DMR under Grant No. DMR-0213745.
Semiconductor and Metal Nanocrystals, edited by V. I. Klimov Marcel Dekker, New York, 2004. 2 C. B. Murray, C. R. Kagan, and M. G. Bawendi, Annu. Rev. Mater. Sci. 30, 545 2000. 3 D. R. Larson et al., Science 300, 1434 2003. 4 M. Dahan et al., Science 302, 442 2003. 5 S. Coe, W. K. Woo, M. G. Bawendi, and V. Bulovic, Nature London 420, 800 2002. 6 S. K. Poznyak, D. V. Talapin, E. V. Shevchenko, and H. Weller, Nano Lett. 4, 693 2004. 7 V. I. Klimov et al., Science 290, 314 2000. 8 H.-J. Eisler, V. C. Sundar, M. G. Bawendi, M. Walsh, H. I. Smith, and V. Klimov, Appl. Phys. Lett. 80, 4614 2002. 9 C. Wang, B. L. Wehrenberg, C. Y. Woo, and P. Guyot-Sionnest, J. Phys. Chem. B 108, 9027 2004. 10 P. Bhattacharya, A. D. Stiff-Roberts, S. Krishna, and S. Kennerly, Int. J. High Speed Electron. Syst. 12, 969 2002. 11 B. ORegan and M. Grtzel, Nature London 353, 737 1991. 12 W. U. Huynh, J. J. Dittmer, and A. P. Alivisatos, Science 295, 2425 2002. 13 C. B. Murray, C. R. Kagan, and M. G. Bawendi, Science 270, 1335 1995. 14 N. Y. Morgan et al., Phys. Rev. B 66, 075339 2002. 15 D. Yu, C. Wang, and P. Guyot-Sionnest, Science 300, 1277 2003. 16 M. Drndic, M. V. Jarosz, N. Y. Morgan, M. A. Kastner, and M. G. Bawendi, J. Appl. Phys. 92, 7498 2002. 17 M. V. Jarosz et al., Phys. Rev. B 70, 195327 2004. 18 C. B. Murray, D. J. Norris, and M. G. Bawendi, J. Am. Chem. Soc. 115, 8706 1993. 19 P. Guyot-Sionnest and C. Wang, J. Phys. Chem. B 107, 7355 2003. 20 D. Yu, C. Wang, B. L. Wehrenberg, and P. Guyot-Sionnest, Phys. Rev. Lett. 92, 2168021 2004. 21 D. E. Neamen, Semiconductor Physics and Devices MacGraw-Hill, New York, 2002. 22 C. A. Leatherdale et al., Phys. Rev. B 62, 2669 2000. 23 D. S. Ginger and N. C. Greenham, J. Appl. Phys. 87, 1361 2000. 24 C. A. Leatherdale, W. K. Woo, F. V. Mikulec, and M. G. Bawendi, J. Phys. Chem. B 106, 761 2002. 25 M. Shim, S. V. Shilov, M. S. Braiman, and P. Guyot-Sionnest, J. Phys. Chem. B 104, 1494 2000. 26 B. A. Ridley, B. Nivi, and J. M. Jacobson, Science 286, 746 1999. 27 J. Puigdollers et al., Org. Electron. 5, 67 2004. 28 M. D. Fischbein and M. Drndic, Appl. Phys. Lett. 86, 193106 2005. 29 G. P. Siddons, D. Merchin, J. H. Back, J. K. Jeong, and M. Shim, Nano Lett. 4, 927 2004; T. Ozel, A. Gaur, J. A. Rogers, and M. Shim, ibid. 5, 905 2005. 30 B. L. Wehrenberg and P. Guyot-Sionnest, J. Am. Chem. Soc. 125, 7806 2003. 31 B. L. Wehrenberg, D. Yu, J. Ma, and P. Guyot-Sionnest, J. Phys. Chem. B 109, 20192 2005.
1

FIG. 11. Current-voltage characteristics of a CdSe nanocrystal FET device at high gate voltage. a Drain-source current as a function of the drainsource voltage for several values of the gate-source voltage. The scan rate is 1.5 V / s and the measurement is started right after the gate voltage is applied. b Source-drain voltage is xed at 4 V and the gate-drain voltage is scanned 1.5 V / s. c With drain electrode open, gate-source capacitive current as a function of the gate-source voltage. The scan rate is also 1.5 V / s.

applications. We note that the FET could not operate in the p-type regime. This is consistent with the fact that CdSe cannot be charged positive with holes stably injected, in contrast to PbSe.30 This is presumably because, at the surface of the CdSe nanocrystals, hole trapping is much more effective than electron trapping, and this may be explained by the rather oxidizing potentials required for hole injection. Although the proof of concept of a nanocrystal thin lm FET is therefore given here, the effective mobility achieved are still quite low 2 107 cm2 / V s. Higher mobility and better device performance should be expected by improving the sample preparation, controlling better the nature of the traps, optimizing the gate structure, or by choosing other materials with higher mobility, for example PbSe.31
II. CONCLUSION

The post-treatment of drop-cast quantum dot lms plays an essential role in their electro-optical properties. With butylamine, rather complete electrochemistry, with reversible charging of both S and P states, and good mobilities are observed. With a NaOH treatment, n-type CdSe nanocrystal thin lms show the highest mobility 102 cm2 / V s, observed at rather low charging level. By stepping the electrochemical potential, the conduction channel can be turned on and off with a switching time of 10 ms. The n-type lms