STUDY OF ELECTROPHORETIC DEPOSITION OF Pd METAL

12. - 14. 10. 2010, Olomouc, Czech Republic, EU

STUDY OF ELECTROPHORETIC DEPOSITION OF Pd METAL NANOPARTICLES ON InP AND GaN CRYSTAL SEMICONDUCTORS FOR H2-GAS SENSORS Karel ZDANSKY, Roman YATSKIV, Jan GRYM, Ondrej CERNOHORSKY a, Jiri ZAVADIL, Frantisek KOSTKA INSTITUTE OF PHOTONICS AND ELECTRONICS, Academy of Sciences, Praha, Cesko, [email protected] a FACULTY OF NUCLEAR SCIENCES, Czech Technical University, Praha, Cesko Abstract Depositions on surfaces of semiconductor wafers of InP and GaN were performed from isooctane colloid solutions of Pd nanoparticles in AOT reverse micelles. Metal nanoparticles in the colloid were monitored by SEM and by optical transmission spectroscopy. Deposited Pd nanoparticles were characterized by SEM and AFM. Diodes were prepared by making Schottky contacts with colloidal graphite on Pd deposited semiconductor surfaces and ohmic contacts on blank surfaces. Forward and reverse current-voltage characteristics of the diodes showed high rectification ratio and high Schottky barrier height giving evidence of small Fermi level pinning. It was found that several ppb of hydrogen in nitrogen gas can be detected by monitoring the change of diode current at a constant bias voltage. Properties of diodes made on GaN were compared with those made on InP. Keywords: Electrophoresis, InP, GaN, Pd nanoparticles, Schottky barrier, Hydrogen sensor 1.

INTRODUCTION

Various gas sensors are being used for controlling industrial processes, for detections of toxic environmental pollutions or for prevention against leaking of hazardous gases. Protection of human health against such gases as nitrogen oxides and the need of monitoring explosive gases like hydrogen in industrial production, distribution and traffic stimulate research oriented to finding new gas sensors. Large dimensions and high costs are the main disadvantages of conventional gas sensors. Besides, most of them are passive elements needing additional equipments or convertors for signal analysis or amplification. Thus, conventional sensors cannot work as intelligent ones and research of new sensitive, efficient sensors of active type has become an important topic in order to satisfy the needs of modern industry. Gas sensors utilizing Schottky type interface between metal and semiconductor play a significant role in the research due to their small sizes and expected low expenses. Palladium or platinum are metals suitable for making hydrogen sensors based on Schottky barriers. The reason is in catalytic affectivity of these metals for atom dissociation of hydrogen molecules adsorbed on metal surfaces. Part of polarized hydrogen atoms can diffuse through the metal to the interface with the semiconductor and change the barrier height and thus change the electrical properties of a sensor. Hydrogen sensors made by deposition of metals in vacuum show lower sensitivity at higher deposition energy [1]. It is caused by fixing the barrier height due to Fermi level pinning explained by a model of disorder-induced gap states (DIGS) [2]. Larger deposition energy causes larger disorder at the metalsemiconductor interface which induces larger density of gap states and hence stronger Fermi level pinning. Interfaces of Pd or Pt with n-type InP are suitable for making sensitive sensors of hydrogen provided they are not subjected to a strong Fermi level pinning. Electroless plating is a better method for making sensitive Pd/n-InP hydrogen sensors than thermal evaporation [3]. However, the best sensors were made by electrophoretic deposition of Pd nanoparticles (NPs) from colloid solutions prepared by reverse micelle


technique [4]. The sensors show very good rectifying current-voltage characteristics and large Schottky barrier height 0.83 eV in comparison with 0.60 eV reached by electroless plating. Even larger barrier height of 1.07 eV was achieved by electrophoretic deposition and partial surfactant removing in our lab [5]. Herewith we report on our research of layers of Pd NPs deposited on single crystal wafers of InP and GaN by electrophoresis from reverse micelle colloid solutions. The obtained results should be useful for finding new efficient sensors and other electronic and optoelectronic devices 2.

EXPERIMENTAL

2.1

MATERIALS

Colloid solutions of palladium NPs in isooctane were prepared by reverse micelle technique with surfactant of sodium bis-(2-ethylhexyl) sulfoccinate (AOT) from water solutions of Pd chloride (PdCl2) and a reducing agent hydrazine [5]. The chemicals PdCl2, hydrazine, AOT and isooctane were purchased from SigmaAldridge Company, USA. Wafers of intentionally not doped InP of 0.5 mm thickness, polished on one side for epitaxial growing were purchased from Wafer Technology Ltd., USA. They were of n-type conductivity with the concentration of 16

shallow donors 2.5x10

-1

cm , as we determined from capacitance-voltage characteristics of prepared

Schottky diodes. Wafers of intentionally not doped GaN of 0.5 mm thickness, polished on both sides were purchased from Kyma Technologies, USA. They were of n-type conductivity with the concentration of 17

-1

shallow donors 2x10 cm . Colloidal graphite in water was produced by Agar Scientific Ltd., England. 2.2

APPARATUS AND PROCEDURES

Spherical shapes and dispersions of NPs radii were checked by transmission electron microscopy (TEM) in film samples prepared by evaporation of isooctane from a drop of the colloid solution on a thin copper Athene grid. Typically, palladium nanoparticles measured by TEM were of 10 nm in diameter with 10 % dispersion [5]. Optical absorption spectra due to surface plasmons in metal nanoparticles dispersed in colloid solutions were checked on a split-beam photo-spectrometer Specord 210, Analytic Jena, Germany. The absorption peak due to palladium surface plasmons appears at the wavelength of about 280 nm [6]. Layers of metal nanoparticles on polished semiconductor wafers were prepared by electrophoretic depositions in the cell made from teflon with a graphite electrode plan-parallel to the sample wafer [7] with the 1 mm gap. The applied voltage of 150 V (electric field 300 V/cm) was chopped with 1:1 stop-go ratio at 10 Hz frequency. The deposition was made onto bulk InP or GaN polished semiconductor wafers of 2

dimensions 1×1 cm , 0.5 mm, placed on the negative potential (cathode). 2

Diodes were prepared as follows. A semiconductor wafer of 1×1 cm was provided with a whole area ohmic contact on the back side and the front side, deposited with nanoparticles, was punctuated with small spots of colloidal graphite to making Schottky barriers. Current-voltage (I-V) characteristics of prepared diodes were measured using the Keithley Source-Measure Unit 237, by using special software made in LabView programming. The sensitivity of diodes to hydrogen was detected by measuring time dependence of current at an adjusted voltage after alternative switching


flow of hydrogen-nitrogen-calibration-gas and air. Both, the compressed calibration gas and compressed air were purchased from Linde Gas, Cesko. 3.

RESULTS

3.1

InP

Fig. 1. SEM image of InP after 2 hours of electrophoretic deposition of Pd NPs. The scale 10 nm is shown with the bright bar at the botom.

Fig. 2. AFM phase image of InP after 2 hours of electrophoretic deposition of Pd NPs. The scale 310 nm is shown with the bar.

The SEM image of InP surface after 2 hours of electrophoretic deposition of Pd NPs can be seen in Fig. 1. It is obvious that the surface is not fully covered with Pd NPs. Also, most NPs are not separated but they are grouped in small clusters of about ten spherical particles with the diameter about 10 nm. It is not clear whether the groups were present already in the colloid or they arise during the electrophoresis process. By increasing the time of deposition by a factor of ten to about 20 hours, fully covered surface was obtained. The layer of Pd NPs of the fully covered surface was electrically conductive in lateral directions while the layer of partly covered surface, shown in Fig. 1 was not conducting in lateral directions. The AFM phase image of Pd NPs on the same sample is shown in Fig. 2. It confirms the result of the SEM measurements, only the individual Pd NCs in clusters are not resolved. The dimensions of the clusters are slightly larger on the AFM image than on the SEM one. It can be explained as follows. AFM pictures Pd NPs including their soft covers made of AOT, remains of reverse micelles in colloid solution, while in SEM images these soft organic parts are not seen. Smaller amount of large clusters seen in AFM image in Fig. 2 are not seen in SEM image in Fig.1, because the last image is more detailed and no large cluster was in the snapshot. In other images with smaller magnification (not shown) cognate large clusters were observed as well. To make Schottky diodes, contacts were painted with colloidal graphite onto the surface of InP, partly covered with Pd NPs. In Fig. 3 the SEM image shows a magnified graphite spot near its edge. Graphite contact is seen on the left side of the image while the smaller right part shows InP surface partly covered


with Pd NPs. It can be seen that the graphite contact does not form coherent whole piece but it consists of irregular particles of sizes in the range of about one micrometer, with openings (pores) among them. Current-voltage characteristics of a diode made on InP wafer with the graphite Schottky barrier on the side with Pd NPs and the ohmic contact on the other side are shown in Fig. 4. The diode possesses a very good 7

rectification ratio, about 10 at 1.5 V bias voltage. The linear part of forward I–V characteristic gives a large Schottky barrier height 0.85 eV, close to the Schottky-Mott limit (vacuum level alignment) for Pd metal workfunction and InP electron affinity, which shows on a negligible Fermi-level pinning. It gives a good chance for sensitive detection of charged atoms diffusing to the metal-semiconductor interface. The response development of a InP diode with Pd NPs to the flow of gas containing 0.1 % hydrogen in nitrogen can be seen in Fig. 5. The diode was forward biased with the constant voltage 0.1 V. The current -10

changed from the value 4.2×10 -5

to 2.3×10

A in the air ambient

A in the flow of H2/N2. It represents a

change of 55000 times; It means that the diode detection limit of hydrogen should be in the range of 15 ppb. The responce and recovery times are of the order of tens seconds. However, there is seen a distinctly slower process in the final part of the recovery development. We suggest that the slower process is

Fig. 3. A part of Schottky contact made by painting colloidal graphite onto the surface of InP covered with Pd NPs. The scale 1 µm is shown with the bright bar at the botom.

due to slow release of hydrogen diffused into the crystal lattice of Pd NPs during the flow of H2/N2 [7]. It takes about ten hours for the full recovery to the original current value.

-2

10

InP-Pd

-5

-4

10

-6

10

-5

10

I (A)

I (A)

InP-Pd forw.bias 0.1 V 0.1% H2/N2

10

-3

10

-6

10

-7

H2/N2

10

-8

-7

10

10

-8

10 10

-10

10

-10

10

air

-9

10

-9

0.0

0.5

1.0

1.5

U (V)

Fig. 4. Current-voltage characteristics of a Schottky diode made on InP with Pd NPs: i) forward (circles), ii) reverse bias (squares).

0

500

1000

TIME (s)

Fig. 5. Current of a diode made on InP with a layer of Pd NPs as a function of time after alternative opening and closing the flow of the blend 0.1 % H2 in N2 and air. The diode was forward biased with the constant voltage 0.1 V.


3.2

GaN

We deposited Pd NPs and prepared the Schottky diodes on GaN wafers by using the same procedures as described above in the case of InP. The AFM phase image of Pd NPs on the GaN surface is shown in Fig. 6. By comparing the image in Fig. 6 with the one in Fig. 2, we can see some differences. Larger contrast between images of NPs and the semiconductor surface

I (A)

-4

10 -5 10 -6 10 -7 10 -8 10 -9 10 -10 10 -11 10 -12 10 -13 10

GaN-Pd

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

U (V)

Fig. 6. AFM phase image of GaN after 2 hours of electrophoretic deposition of Pd NPs. The scale 310 nm is shown with the bar.

Fig. 7. Current-voltage characteristics of a Schottky diode made on GaN with Pd NPs: i) forward (circles), ii) reverse bias (squares).

in the case of GaN can be explained by its larger reflectivity. However, there is a difference between shapes of clustered NPs on InP and on GaN. Clusters on GaN are nearly circular while those on InP are distinctly elliptical. Difference of cluster shapes on different kinds of surfaces could indicate that the clusters grow during the deposition. The origin of NP clusters has to be clarified by a more detailed study. Current-voltage characteristics of a diode made on GaN wafer with the graphite Schottky barrier on the

-4

10

-5

10

H2/N2

GaN-Pd forw.bias 0.1V 0.1 % H2/N2

-6

I (A)

10

side are shown in Fig. 7. The diode possesses a very 7

good rectification ratio, more than 10 at 1.5 V bias

-7

10

voltage. The linear part of forward I–V characteristic

-8

gives a large Schottky barrier height 1.1 eV

10

-9

10

The response development of a GaN diode with Pd

air

-10

NPs on the flow of gas containing 0.1 % hydrogen in

10

nitrogen can be seen in Fig. 8. The diode was forward

-11

10

side with Pd NPs and the ohmic contact on the other

0

200

400

TIME (s)

biased with the constant voltage 0.1 V. The current -11

changed from the value 7.4×10

Fig. 8. Current of a diode made on GaN with a layer of Pd NPs as a function of time after alternative opening and closing the flow of the blend 0.1 % H2 in N2 and air. The diode was forward biased with the constant voltage 0.1 V.

-5

to 1.4×10

A in the air ambient

A in the flow of H2/N2. It represents a

change of 190000 times; it means that the diode detection limit of hydrogen should be in the range of 5 ppb. The responce and recovery times are of the order of tens seconds.


4.

DISCUSSION

It is obvious that the Schottky barriers made by contacts of colloidal graphite and Pd NPs on InP or on GaN 7

polished surfaces possess very good electrical properties. The rectification ratio of about 10 at 1.5 V bias voltage represents excellent result, in particular in the case of InP. We assume that its merit is in the separation between the conductor (Pd metal and graphite) and semiconductor (InP or GaN) by membranes of organic substances. Membranes are formed by remains of AOT reverse micelles in the case of Pd NPs and by a substance used for producing the commercial colloidal graphite in water. The values of Schottky barrier heights are close to the Schottky-Mott limit which is the condition for a negligible Fermi-level pinning. It gives a good chance for sensitive detection of hydrogen diffusing to the metal-semiconductor interface and dissociated by the Pd metal NPs. Reported Schottky barriers act as excellent detectors of hydrogen with the detection limit about 5 or 15 ppb H2 in N2 for the case of GaN or InP, respectively. The merit of this result is, besides exquisite electrical properties of the Schottky barriers, mentioned above, in porosity of the contacts made by the colloidal graphite. Hydrogen can diffuse through the whole area of the contact due to its porous nature and fill the whole interface between conductor and semiconductor of the Schottky barrier, thus maximally affecting its height. 5.

SUMMARY

Colloid solutions of Pd NPs in reverse micelles in isooctane solvent were prepared and used for deposition of Pd NPs on surfaces of n-type InP and n-type GaN single crystalline wafers. Deposition was provided by electrophoresis during 2 hours with chopped 1:1 go-stop applied voltage 150 V (electric field 300 V/cm) with the negative pole on the sample. It was found by SEM and AFM that surfaces were not fully covered with Pd NPs and most NPs were coalesced in small clusters. Contacts with Schottky barriers were made on surfaces with deposited Pd NPs by application of colloidal graphite and ohmic contacts were made on blank surfaces to make rectifying diodes. Prepared diodes showed excellent rectification with high Schottky barrier heights and they acted as very sensitive hydrogen sensors with the detection limit in the order of ppb units. Detection limit of GaN diodes was about three times better than that of InP diodes. Acknowledgement: We acknowledge financial support by COST Action MP0805, project OC10021 of Ministry of Education CR, by project KAN401220801 of Academy of Sciences CR and by project 102/09/1037 of the Czech Science Foundation. LITERATURE [1]

L. M. Lechuga, A. Calle, D. Golmayo, P. Tejedor and F. Briones, J. Electrochem. Soc. 138 (1991) 159-162.

[2]

H. Hasegawa: Metal-Semiconductor Interfaces, ed. A. Hiraki (IOS Press, Tokyo, 1995), p. 280.

[3]

H. I. Chen, Y. I. Chou and C. Y. Chu, Sens. Actuat. B 85 (2002) 10-18.

[4]

Y. I. Chou, C. M. Chen, W. C. Liu and H. I. Chen, IEEE Electron Device Letters 26 (2005) 62-65.

[5]

K. Zdansky, P. Kacerovsky, J. Zavadil, J. Lorincik and A. Fojtik, Layers of metal nanoparticles on semicoductors deposited by electrophoresis from solutions with reverse micelles. Nanoscale Research Letters, 2007, vol. 2, is. 9, ps 450-454.

[6]

K. Zdansky, J. Zavadil, P. Kacerovsky, J. Lorincik and A. Fojtik, Deposition of Pd nanoparticles on InP by electrophoresis: Dependence on electrode polarity. IREE Transactions on Nanotechnology, 2010, vol. 9, is. 3, ps 355-360.

[7]

P. Kovacik; Nanoparticle gas sensor, Master Thesis, Czech Technical University in Prague, Czech Republic, 2008.