Polymeric semiconductor devices

Polymeric semiconductor devices

Synthetic Metals, 28 (1989) C735-C745 C735 POLYMERIC SEMICONDUCTOR DEVICES J. H. Burroughes, C. A. Jones and R. H . Friend, Cavendish Laboratory, M...

467KB Sizes 1 Downloads 117 Views

Synthetic Metals, 28 (1989) C735-C745

C735

POLYMERIC SEMICONDUCTOR DEVICES

J. H. Burroughes, C. A. Jones and R. H . Friend, Cavendish Laboratory, Madingley Road, Cambridge C B 3 0 H E , U.K.

§1 A b s t r a c t We have constructed various semiconductor device structures with polyacetylene prepared by the Durham precursor route as the active semiconductor. Electrical characterisations of Schottky barrier diodes and metal-insulator-semiconductor structures show that these structures behave in aa manner similar to that of conventional inorganic semiconductor devices, and we present evidence for the formation of accumulation and depletion layers. We show that charge is stored in solitonlike localised states in both the depletion and accumulation layers, and we obtain quantitative agreement for the modulation in the optical transmission through the active semiconductor region with current models for these localised states.

§2 Introduction Despite the considerable advances that have been made in the understanding of the electronic properties of conjugated polymers, and in the control and handling of these materials, there has been relatively little work reported on their use as the active component in semiconductor device structures.

There are several straightforward reasons for this; the first of which is that most

conjugated polymers cannot be conveniently processed to the forms required in these devices. There is also a fundamental problem in the use of p or n type doped polymer, that the dopants are expected to be able to diffuse through the sample at room temperature (the doping reaction is usually achieved by diffusion of dopant into the sample at room temperature). Thus, in a bipolar device, the dopants in the p and n regions are able to diffuse together, and in a unipolar device such as a MISFET (Metal-Insulator-Semiconductor Field Effect Transistor) or a Schottky-barrier diode, the dopants may be able to move under the influence of the electric field applied to modulate the width of the depletion region. Schottky and p-n diodes have been studied by several groups [1-4], and rectification ratios (Iforward/Ireverse) of up to a few hundred have been reported. There are also reports of field-induced conductivity measured in MISFET structures [5,6]. We report here some of the results of the programme of work in Cambridge on the use of polyacetylene prepared by the Durham precursor polymer route [7]. Polyacetylene prepared in this way can be conveniently prepared as thin coherent films by spin-coating the precursor in solution onto the required substrate, followed by heat treatment to convert to the polyacetylene [8-11 ]. The polyacetylene films prepared in this way are unoriented, and in contrast to stretch-oriented films 0379-6779/89/$3.50

© Elsevier Sequoia/Printed in Tile Netherlands

C736 [12] the polymer is disordered, with short sequences of straight chain separated by chain bends or twists. We have shown elsewhere [9,11,13] that, in spite of the apparently short conjugation lengths, polyacetylene prepared this way supports similar, soliton-like excitations as oriented Durham or Shirakawa material. As we discuss later, the films of polyacetylene prepared in this way are p-doped (possibly with catalyst residues at chain ends) to a level suitable for device applications ( = 1016 cm-3), and these dopants do not appear to be mobile under applied electric fields. We present data for Schottky barrier diodes and MIS (Metal-Insulator-Semiconductor) structures. We can model the electrical response of these devices using conventional models developed for inorganic semiconductors, and can demonstrate the movement of depletion layers and the formation of accumulation layers in these structures. However, charge storage in conjugated polymers is not expected to be in band edge states; it is well established that there is a structural relaxation around charges added to the chains, both through chemical doping and with charge separation following photoexcitation. For polyacetylene the localised states are bond alternation defects, or solitons [ 14,15], and the clearest evidence for their formation is through the appearance of additional optical absorption below the band gap, associated with the vibrational and electronic excitations of the solitons [16]. For unoriented Durham polyacetylene the 'mid-gap' absorption feature due to transitions between the 'mid-gap' levels on the solitons and the band edges is seen at around 1.0 eV for chemically doped samples [11], and at 0.55 eV for photoexcited charges [13]. We expect, therefore, that the formation of a depletion layer in polyacetylene should remove charged solitons and reduce the 'mid-gap' optical absorption, whereas the formation of an accumulation layer should introduce 'mid-gap' states and increase the mid-gap optical absorption. This is what we find, and we indeed measure changes in the optical properties in quantitative agreement with our predictions.

§3 Fabrication Solutions of the precursor polymer in 2-butanone were used in concentrations of 1 gm of solid precursor to 10-23 ml of 2-butanone for the Schottky diodes and 23-100 ml of 2-butanone for the MIS devices. These solution concentrations give polyacetylene film thicknesses of between 200 /~ and lp.m. The solution was spin-coated onto the appropriate substrate at 2000 rpm for 60 seconds. Transformation of the precursor polymer to a trans-polyacetylene film was carried out under a vacuum of <10-5 mbar in an evaporator built into the glove box at a temperature of 80°C for 12 hours. This heat treatment produces a fully dense coherent film of trans-polyacetylene. These processes were carried out in an argon-filled glove box with an oxygen and water concentration of less than 5 ppm.

The device was transferred to the optical-access cryostat used for the

measurements without exposure to air. Once in the cryostat the sample was kept at a pressure of less than 10-6 mbar. For the construction of Schottky diodes the polyacetylene precursor was spin-coated onto a wide-field spectrosil substrate onto which a film of gold, 200/~ thick, had first been evaporated. After transformation of the polyacetylene film, an aluminium contact, also 200/~ thick, was evaporated without exposure to the glove box atmosphere. Gold forms an ohmic contact with ptype polyacetylene, and the rectifying function is formed with the aluminium contact [1]. For the

C737

MIS structures, the substrate used was a slab of ultra-pure silicon, of dimensions 16 x 16 x 0.3 mm, polished on both sides. On one surface the silicon was heavily doped with phosphorus (> 1019 cm-3), to a depth of 4000/~. Onto this surface a silicon dioxide layer, 2000/k thick, was grown to serve as the insulator for the MIS device. The n-type doped layer acted as the distributed gate electrode. The polyacetylene precursor was spin-coated onto the SiO2 layer and transformed. Finally a gold contact was evaporated onto the polyacetylene film.

10-3

10-6 j (A crn -2) 10-9

10-12 i

-2

t



-1

i

I

0

i

1

i

i

2

bias voltage (V) Figure 1:

The Schottky diode current density-bias voltage characteristics. Sample area = 0.636cm 2, n = 1.3, JSat.= 4.4 x 10-13 A cm -2, hu = 1.28 V.

§4 Schottky Diodes The current density versus voltage characteristics for Schottky diodes constructed as above are shown in figure 1, in which log(I) is plotted versus bias voltage for both forward and reverse biases. The ratio of forward to reverse current reaches a value of typically 5 x 105 at bias voltages of around 1.5 V. The usual parameterisation of the variation of current density with bias voltage is q V ] for V>3kT/q, and where n is the ideality factor which for a perfect diode is J = JSat.ex p ( nkT equal to 1. As seen in figure 1, we find a value for n of about 1.3 at low forward biases. At higher forward biases the current is limited by the bulk resistance. In the thermionic emission model for conduction across the junction, the saturation current density, JSat. is given by JSat = A*T2exp{qW/KT}, with A*, the effective Richardson constant, which for an effective electron mass of one equals 120 A / K 2 cm 2, and ~t' is the barrier height [17]. By extrapolating from the linear region to the In(J) axis in figure 1, we estimate that Jsat = 4.4 x 10-13 A cm -1, and hence

C738

qS=l.28 eV. We do not see the reverse current saturate, as expected in the thermionic emission model; the most probable cause for not observing this is the formation of a small interfacial layer between the polyacetylene film and the blocking contact [18]. The doping density may be obtained from the variation of the Schottky diode capacitance with reverse bias voltage, using the simple depletion model for the voltage dependence of the depletion

1

1 (2(*+V)~

width. In terms of the measured capacitance [17], this may be expressed as ~

= A--~ ~ereoqNa)

where Na is the acceptor concentration and ¢pis the built-in potential. Figure 2 shows a plot of 1/C2 versus bias voltage for two different frequencies. From the slope of the linear region, we estimate that Na= 9 x 10 15 cm-3, and from the extrapolation of the curve to the forward voltage axis we estimate ~ = 0.94 V.

3

x 1016

I

(F-2)

0



-2

I

I

-1

I

I

0

I

1

I

i

2

bias voltage (V) Figure 2:

1/C 2 versus bias voltage; Sample area = 0.636cm 2, Na = 9 x 1015 cm -3, ~ = 0.94 V.

The analysis of the electrical characteristics of the Schottky barrier between aluminium and polyacetylene indicates that the device is operating in the conventional way, with movement of the depletion layer with applied bias voltage. However, as discussed in §2, charge is stored in polyacetylene in 'soliton' localised states, which have non-bonding Pz 'mid-gap' states associated with them. We expect to see, therefore, a modulation in the optical properties of the active semiconductor region with applied bias voltage. An increase in the width of the depletion layer should remove charged soliton states, thereby reducing absorption within the gap and increasing the absorption in the region of the ~ to re* transitions, above 1.4 eV. The voltage modulated transmission (VM'I) spectrum between 0.4 and 1.8 eV is shown in figure 3. The positive signal for

C739

2

x 10-3 1

T(-3 V)-T(O) T(O) o -1

-2

--3 -4

t

0-4 Figure 3:

i

(38

t ,,

i

,,I

1.2 E n e r g y (eV)

i

1-6

[T(V) - T(0)] / T(0) versus photon energy (Ep), where V= -3V with respect to the blocking contact. The data shows an EB featfare for Ep < 1.4eV and an EA feature for Ep > 1.4eV.

21

17dQ

13a-Q (nF) •

/' / /

0 - ~,7 -75 Figure 4:

i

-,~5

i '3.5

i -1.5

\\

\ ]9

0 0-5 2-5 bias vottoge (V)

The differential VMT signal and the capacitance versus the bias voltage. The ac voltage and frequency were + 0.25V and 1.2KHz respectively.

C740 energies less than 1.4eV shows that the transmission does increase for reverse bias voltages. The position of the electro-bleaching (EB) peak at 0.55 eV is in broad agreement with photo-induced absorption experiments [ 13] on spin-coated Durham polyacetylene. The periodic structure in the modulated transmission spectrum below the bandgap is due to interference fringes (modulation of the finesse of the Fabry-Perot etalon formed by the polyacetylene film). Electro-absorption (EA) is seen at energies greater than 1.4 eV, as expected. The variation of the amplitude of the VMT response with applied dc bias should scale with aQ/aV, the differential capacitance, shown in figure 2. In figure 4 we replot this data, as C rather than 1/C2, and show also the VMT signal measured with a 0.5V a.c. modulation (1200Hz). The correspondence between these two quantities is evident, and we therefore associate the variation of the optical transmission with bias voltage in these devices with variation in the width of the depletion layer.

11

~6 if

101-

II

,,' 1.2

C (nF)

..........

,"

G (pS) 0.8

\ 0-4

,

-15

-10

,

-5

0

I

i

5

10

bias voltage Figure 5:

0

15

(V)

The capacitance and the conductance measured at 500Hz versus the bias voltage. Sample area = 0.615 cm2, oxide width = 2000/~ and the polyacetylene width = 1200A.

§5 MIS Structures The MIS structure allows the possibility of band bending at the insulator-semiconductor interface through the Fermi level to produce a surface charge layer which may be the same carder sign as the majority carders (accumulation layer) or as the minority carders (inversion layer) [19]. The formation of accumulation, depletion and inversion layers may be demonstrated through the behaviour of the device capacitance with respect to the bias voltage. The measured capacitance, C, is that of the series combination of the insulator capacitance, Ci, and the capacitance of the active CiCd region of the semiconductor, Cd, and is given by C = Ci~2"ff" Since Cd is large for the

C741

accumulation and inversion layers, C is equal to the geometric capacitance of the insulating layer, but C falls to a lower value for movement of the depletion layer. The capacitance and the conductance versus bias voltage curves for a polyacetylene MIS structure are shown in figure 5. As expected the capacitance (500Hz) for negative gate voltages flattens out to the geometric capacitance of the insulator, indicating the formation of an accumulation layer. The decrease in the measured capacitance for positive voltages indicates the formation of the depletion layer. Saturation of the capacitance for large positive biases sets in when the depletion layer extends across the polyacetylene film (estimated to be about 1200,~). At present we have not been able to find evidence in the C-V data for the formation of an inversion layer for these structures; there are many possible explanations for this, including a low thermal generation rate of minority carriers and the presence of electron-accepting trap states. The peak in the conductance corresponds to the flat band voltage [19], which for this device is at about 0V.

7 xlO-3 6 5

T(O)-T(V) T(O) 4 3 2 1 0

0.4

Figure 6:

I

I

0.6

0-8

I

1-0 Energy (eV)

1.2

IT(0) - T(V)] / T(0) versus the photon energy, for V = --10, -20, -30, and -40V with respect to the gate electrode. The data shows the EA feature for Ep < Eg.

If excess charge in the accumulation layer is expected to reside on the polyacetylene chains as soliton-like localised states, we expect a decrease in the device transmission below the band-edge as the device is driven towards accumulation. The VMT specu'um between 0.4 and 1.2eV in figure 6 shows that there is a decrease in the transmission through the device, for the device biased from 0V to -40V. The peak value of AT/T, at 0.8 eV, is about 0.64%. As with the Schottky diode, the variation in the VMT signal with dc bias voltage should scale with OQ/OV, the differential capacitance if all the charge at the polymer-insulatorinterface is stored in soliton-like states. This is shown to be the case in figure 7, which shows the variation of the VMT response ( a.c. modulation 0.25 V) and differential capacitance with bias voltage.

C742

IxlO--4

11

10 dQ

d tnT (v-l)

d-V (nF)9

I

I

-40

I

I

-20

~'r" .....

0

20

bias voltage (V) Figure 7:

The differential VMT signal and the capacitance versus bias voltage for the MIS structure.

1.6 ¸

x l 0 -a 1-2.

T(O)-T(-5OV) T(O)

0.8.

0.~

I

1000

1700

I

2400

I

3100

I

3800

4500

Wavenumber (cm-1) Figure 8:

The VMT signal for a voltage of-50V wrt the gate electrode between 1000 and 5000 cm -1 showing the low energy edge of the electronic transition and two sharp vibrational features at 1379 and 1281 cm "1.

C743

Besides the electronic transitions associated with the soliton state, there is extra infrared activity due to new vibrational modes which couple to motion of the charged soliton along the polymer chain [20]. These are seen in both doping [ 11] and photoexcitation [13] experiments in unoriented Durham polyacetylene. Figure 8 shows the results from an FFIR experiment on an MIS structure, showing the difference in transmission for the MIS device biased between 0 and -50V with respect to the gate. There are three peaks observable; the absorption seen above 4000 cm -I is the low energy side of the electronic absorption, as in figure 6, and two sharp vibrational features at 1379 and 1281 cm -I. We identify these as the two higher frequency translation modes of the soliton. We were not able to measure the VMT spectrum below 1000 cm -1 due to a very large absorption peak in the SIO2, and we were thus unable to observe the lowest frequency, 'pinned' translational mode. The frequency of the highest vibrational mode depends on the structure of the polyacetylene chain where the charge is localised. The value of 1379 cm -1 lies between that measured for photoinduced charges, 1373 cm -1 [13] and for dopant-induced charges, 1410 cm -I [11]. Also of relevance is the ratio of the peak intensity of the 'mid-gap' electronic feature to that of this vibrational mode; the value here is about 5. This scales directly with the effective mass of the soliton, which itself scales as the third power of the local energy gap [20]. Our interpretation of these results is that the charges in the accumulation layer are forced to sit on the chains nearest the interface with the insulator, and these have larger band-gaps than the chains selected in the photoinduced absorption experiment in which the charge density is much lower. The upward shift of the translational mode from 1373 to 1379 cm -l, the decrease in peak intensity of this mode in relation to the 'mid-gap' feature from 1:2 to 1:5, and the shift in the 'mid-gap' energy from 0.55 to between 0.7 and 0.8 eV are all consistent with an increase in the band-gap of the chains where the charge is stored by some 25%. §6 D i s c u s s i o n We have found that polyacetylene behaves as a conventional semiconductor when used as the active component in a semiconductor device. However, we have demonstrated that charge storage in depletion or accumulation layers is not in the band edges as with conventional semiconductors, but in soliton-like midgap states. Whilst for the depletion layer the chain deformation associated with these states is achieved by the extrinsic doping, for the accumulation layer the presence of the surface field forces soliton-antisoliton pairs onto previously undistorted chains (We can exclude the possibility of charging any extrinsic spin 1/2 'soliton' defects, present in a concentration of about 1019 cm -3 in trans-isomer Durham polyacetylene [9], since we can obtain high modulation depths for the optical transmission, ~- 1%, for structures with polyacetylene layers as thin as 200 A). The modulation of the optical properties by control of the charged soliton density through the voltage applied across the device provides a new spectroscopy for the investigation of the electronic structure of the active region in these semiconductor devices. The change in the transmission may be modelled using the optical cross-section, ~, estimated to be about 8x10 -16 cm -2 at the peak of the 'midgap' absorption [16]. We have thus, AT/T = t~Na, where N a is the areal soliton concentration. For the MIS device driven into accumulation the areal charge concentration is given by N a =

C744

ereoV/qd, where V is the applied voltage and d is the insulator width. For the MIS device shown in figure 6 we calculate AT/T = 0.34%, which is in good agreement with the experimental value of 0.64%. We can thus obtain quantitative estimates for the accumulation layer charge density from measurements of the optical properties of the structure. We obtain similar agreement for analysis of optical modulation in the Schottky diode. Turning to the spectral resolution of the absorption features, we now have a powerful means of characterising the polymer chains active in the operation of the device. We have shown for example a difference between the peak positions for the Schottky diode (0.55 eV) and the MIS structure (0.7 - 0.8 eV), which we attribute to the different structures of polyacetylene surfaces probed by the VMT experiment. For the MIS structure control and characterisation of the polyacetylene structure at the interface with the insulator is important if these devices are to be optimised.

Initial results

on polyacetylene MIS

structures,

with

spin-coated

poly(methylmethacrylate), PMMA, serving as the insulator shows the absorption peak to be at 0.55eV, and we consider that the polyacetylene is more ordered at the PMMA interface than at the SiO2 interface.

Acknowledgement We thank British Petroleum plc for financial support for this work.

References 1.

P. M. Grant, T. Tani, W. D. Gill, M. Krounbi and T. C. Clarke, J. Appl. Phys. 52, 2, 869 (1980)

2.

J. Kanicki, pp 544-660,"Handbook on Conducting Polymers", ed. T. J. Skotheim, M. Dekker, New York (1986).

3.

F.Garnier and G. Horowitz, Svn. Met. 18~ 693 (1987)

4.

H. Tomozawa, D. Braun, S. Phillips, A. J. Heeger, and H. Kroemer, Svn. Met. 22, 63 (1987)

5.

E. Ebisawa, T. Kurokawa, and S. Nara, J. Appl. Phy~, 54, 3255 (1983)

6.

H. Koezuka, A. Tsumura, and T. Ando, Svn. Met, 18 699 (1987)

7.

J. H. Edwards and W. J. Feast, Polvmer Common, 21,595 (1980)

8.

R. H. Friend, D. C. Bott, D. D. C. Bradley, C. K. Chai, W. J. Feast, P. J. S. Foot, J. R. M. Giles, M. E. Horton, C. M. Periera and P. D. Townsend, Phil. Trans. Roy. Soc. Lond. A314, 37 (1985).

9.

R. H. Friend, D. D. C. Bradley, C. M. Pereira, P. D. Townsend, D. C. Bott and K. P. J. Williams, Svn. Met. 13, 101 (1986).

C745

10.

D.C. Bott, C. S. Brown, C. K. Chai, N. S. Walker, W. J. Feast, P. J. S. Foot, P. D. Calvert, N. C. Billingham and R. H. Friend, Svn. Met. 14,245 (1986).

11.

R.H. Friend, D. D. C. Bradley, P. D. Townsend and D. C. Bott, Svn. Met. 17, 267 (1987).

12.

R.H. Friend, D. D. C. Bradley and P. D. Townsend, J, Phys. D 20, 1367 (1987).

13.

R.H. Friend, H. E. Schaffer, A. J. Heeger and D. C. Bott, J. Phys. C 20, 6013 (1987).

14.

W.P. Su, J. R. Schrieffer and A. J. Heeger: Phys. Rev. Lett. 42, 1698 (1979); Phys. Rev. B22, 2099 (1980); Err. B28, 1138 (1983).

15.

M.J. Rice, Phys. Lett. 71A, 152 (1979).

16.

J. Orenstein: In Handbook of Conducting Pglym¢r,s, Vol. 2, p 1297, ed. by T. Skotheim (Marcel Dekker, New York 1986)

17.

S.M. Sze, Physics of Semiconductor devices, Wiley-Interscient Publication.

18.

E.M. Rhoderick, Metal-Semiconductor Contacts, Clarendon Press.

19.

T. Ando, A. B. Fowler and F. Stem, Rev. Mod. Phy~, 54, 437 (1982).

20.

B. Horovitz, Solid State Commun, 41,729 (1982)