1731
Anal. Chem. 1988, 60, 1731-1736
Organophosphorus Compound Detection with a Supported Copper Cuprous Oxide Island Film. 1. Gas-Sensitive Film Physical Characteristics and Direct Current Studies
+
Edward S . Kolesar, Jr.*
Department of Electrical and Computer Engineering, Air Force Institute of Technology, Wright-Patterson AFB, Dayton, Ohio 45433-6583 Rodger M. Walser
Department of Electrical and Computer Engineering, T h e University of Texas at Austin, Austin, Texas 78712-1084
The electronlc propertles modified by the exposure of a dlelectric supported Island fllm of copper cuprous oxide to small concentrations (part-per-mllllon levels) of dlisopropyl methylphosphonate (DIMP) are reported. Electron mlcroscopy and electron dlffractlon measurements confirmed the fllm’s structure and composition. Auger spectroscopy verlfled the adsorption of DIMP. Isothermal dlrect current measurements suggest a space-charge conduction mechanism whereby carriers are thermally excited, fleld Induced, and transported via shallow trapping and impurlty centers.
+
One critical group of environmental contaminants are the organophosphorus pesticides and structurally affiliated compounds. These materials are unique among contaminants because their distribution is essential for their efficacy. Adding to their impact, these compounds are synthesized and valued for their deleterious effect and persistence. Thus, there is a need for personal monitoring instrumentation to detect subthreshold levels of these toxic compounds. A significant portion of the environmentally sensitive organophosphorus compounds contain either the phosphoryl or thiophosphoryl group. Since diisopropyl methylphosphonate (DIMP) is a phosphoryl-containing compound, has low toxicity, and has been used in previous studies concerned with detecting organophosphorus compounds, it was used as a model compound in this investigation (1-5). This paper reports the electronic properties that are modified when a dielectric-supported, island morphology, copper + cuprous oxide film is exposed to small concentrations (part per million levels) of DIMP. The motivation for analyzing this synergistic combination was inspired by two independently reported phenomena. Several researchers cite the use of copper complexes to catalytically hydrolyze phosphorus esters (6-9). The most significant study is from Wagner-Jauregg et al. (6). They found that Cu(I1) salts catalyze the hydrolysis of diisopropylphosphonate. Copper was found to be the best metal among those evaluated (nickel, cobalt, iron, and manganese). The overall reaction occurs in two steps. First, the copper complex binds the phosphorus ester reversibly. Second, the adduct-product is irreversibly broked down by hydrolysis. The property of the first reaction is most useful in this effort. That is, copper complexes may adsorb and desorb phosphorus esters in air. The second reaction is less likely to occur on the surface of a copper complex exposed to ambient air because of the low concentration of water. It is also well documented that many of the developmental problems associated with integrated circuit processing can be traced to impurity contamination (IO). A material’s surface
and electronic properties are strongly influenced by ambient conditions, but little progress has been made exploiting this phenomenon for the practical application of gas detection. Several investigators report that a discontinuous metallic film’s electrical conductivity is modified when it is exposed to select constituents of the atmosphere (11). The following possibilities, either acting alone or synergistically, may be attributable as the cause for the change in the electronic transport properties of these films: 1. The ambient atmosphere may induce a chemical reaction that results in an irreversible structural modification (for example, oxidation, reduction, or the alteration of surface stress or strain). 2. The ambient atmosphere may be adsorbed on the metallic islands and form a surface dipole layer that modifies the charge carrier’s tunneling barrier height or be adsorbed in the gap region and alter the dielectric’s trap density. The sorption-desorption reversibility characteristic associated with the changes observed in the electrical conductivity of the copper + cuprous oxide island films suggests that the second mechanism dominates the electronic transport process. The thermally evaporated, discontinuous copper + cuprous oxide films utilized in this detection scheme have large-valued resistances (typically 106-108 Q ) . The difficulties involved in designing a stable, two-terminal, direct current sensing device were resolved by integrating the gas-sensitive film with the distributed RC notch network. A distributed RC notch network can readily be configured to be compatible with large-valued, distributed resistances. Part 1 of this paper reports on the sensor’s fabrication, instrumentation arrangement, the physical characteristics of the gas-sensitivefilm, and the direct current behavior. Several critical parameters (film thickness, operating temperature, humidity level, and applied direct current bias) that act to optimize gas sensitivity are sighted. These parameters will be utilized in part 2, which discusses the RC notch network sensor theory, the film’s alternating current behavior, and the sensor’s performance.
SENSOR FABRICATION The distributed RC notch network sensor, illustrated in Figure 1,was fabricated by using standard integrated circuit processing techniques. Glass microscope slides were used as the substrate. A thermal evaporation system was utilized t o deposit the electrodes and gas-sensitive discontinuous metallic film. Depositions were accomplished at lo4 Torr by using a liquid-nitrogen trapped, oil diffusion pumped system (Veeco, Model VE-7700, Plainview, NY). A solenoid activated shutter was used to cover the filament during degassing and to terminate the deposition of a film at the desired thickness. High-purity (99.999%) copper wire (Johnson Matthey/Aesar, Seabrook, NH) was used as the source for the surface electrodes and gas-sensitive film. A piezoelectric quartz crystal thickness monitor was utilized to control the evaporation
0003-2700/88/0360-1731$01.50/00 1988 American Chemical Society
1732
ANALYTICAL CHEMISTRY, VOL. 60. NO. 17, SEPTEMBER 1. 1988
Table I. Critical Parameters for the Distributed RC Notch Network Sensor parameter
device structure 1
upper electrode gap width (pm) gas-sensitive film thickness (A) measured distributed gap resistance- (R,R) distributed capacitance film thickness (A) equivalent distributed gap capacitance (C, farads) characteristic device frequency (u,= 2 / R C ) (radians)
203 71 1.160 x 1 6 1285 49 x 10-12 3.5185 X 10"
device structure 2 508 126 8.506 X 1 6 4783 7.364 x 10-'0 3.1929
aExwrimental conditions: u n e x d . 35-VaDD1ied direct current bias. 20 "C.
1SlJllVl f l l l OltttClMC FILM
Ih
Flgure 1. Distributed RC notch netwolk physical structue. ChCun
symbol. and distributed parameter representation.
Instrumentation arrangement for measulng the dkect current conductivtty behavior and the transfer function (gain versus frequency) of the distributed RC notch network senscf. Figure 3.
Flpm 2. A 203 pm wlde elecwde gap dlsfibuted RC notch network sensor (two discrete sensors).
rate (typically 3-5 A d ) and monitor the average thicknesa of a deposition. Mounting was facilitated to ensure that the substrate's front surface temperature did not exceed 30 "C during deposition of the gas-sensitive film. Metal masks were used to define electrode arrangements and control the coverage of the copper + cuprous oxide films Surface electmde gaps of 203 and 508 pm were utilized to obtain values of R and C (Figure 1) that yield practical tuning o parameters. A pinhole and defect-free dielectric ( 1 ~ 5 o o A thickness range) support was realized by using an exploratory polymersilome linked system (Allied Chemical Corp., Aceuglasa M3R and M3RP spin-on glass materials, Buffalo, NY). Amorphous films of Si(rSi0, were obtained after a spin-on application and subsequent cure a t 300 "C. In order to account for dielectric losses, discrete parallel plate test capacitors with identical dimensions were fabricated on the same substrate. Table I summarizes the critical dimensions and parameters for the two sensor configurations. Figure 2 depicts a 203 rrm wide electrode gas sensor.
INSTRUMENTATION Point-by-point measurements of a distributed RC notch network's gain characteristic can be very tedious. especially when many points are needed to clearly define the vicinity of the null. This problem is compounded when tuning adjustments are simultaneously made to identify the notch resistance producing an optimum null. As illustrated in Figure 3, a solution was devised that integrates a swept-frequency function generator, gain-phase detector (both with digital displays), and an ascillaswpe operated
as an x-y recorder. To minimize loading effects. high impedance oseillosmpe probes were used to measure the semr's open-circuit voltage transfer function. The capability to apply a direct current bias to the gas-sensitive film was included along with an electrometer to permit independent, direct current measurements of the gas-sensitive film's resistance. A precision resistance decade box (0.01-Rresolution) was used to tune the optimum value of the notch resistance (RJ. The gain of the sensor's open-circuit voltage transfer function (20 log,,, V J Vi) was measured directly with the gain-phase detector, and a scaled, direct current analog output signal was used to drive the vertical channel iy axis) of the oscilloscope. Similarly, the function generator's direct current analog output signal (scaled to the swept-frequency interval) was used to drive the horizontal channel (x axis). Swept-frequency measurements were used to calibrate and display the Sensor's gain response. Manual frequency sweeps. combined with simultaneous adjustments of the notch resistance (R"), facilitated optimizing the sensor's response. The dynamic gas generation and delivery system is illustrated in Figure 4. A critical element in this system is the DIMP permeation tube (GC Industries, Model 23-7392,Chatsworth, CA). When operated a t SO "C, a broad concentration range spanning
ANALYTICAL CHEMISTRY. VOC.
Figure 5. TEM micrograph of R 100 h average thickness copper film ( 7 2 000
magnification).
0.1-fi50ppm of DIMP is available. In addition. the capacity to evaluate the sensnr’s performance relative to other gases (bottled AI. H,, N,, and O,, and lahoratory air) war incorporated. The laboratory air source was filtered through an activated charcoal hed to remove organic contaminants. Silica gel beds were used to remove water vapor. When required, humidity was added to the carrier via a gas dispersion hottle (deionized water) and monitored with an in-line hypxneter. Precision flowmeters with micrometer valve assemblies were used to estahlish the desired DIMP mncentration levels. To minimize adsorption effects, 316 stainless steel tubing, fittings. and valves were used throughout, except for the Pyrex glass components. Heating tape was used to provide a constant 26 O C thermal background for the gas mixture. A free-flowsample cell was fahricated from Pyrex, and its temperature (-20 to 200 “C) was thermostatically controlled hy suhmening it in a silicon oil bath. A pound g l a s joint provided a port for mounting the device under test.
RESULTS AND DISCUSSION Physical Characteristics. Two critical issues conceming the sensor’s operation are the physical structure and composition of the gas-sensitive film and verification that DIMP is adsorbed. The physical structure and composition is fundamental to the model used to explain the change of electrical conductivity induced by DIMP exposure. Similarly, independent confirmation that DIMP is adsorbed on the gassensitive film is critical for establishing a correlation hetween surface chemistry and electrical conductivity changes. Physical structure and composition were studed by direct replica transmission electron microscopy (TEM) and transmission electron diffraction (TED), respectively. Auger electron spectroscopy (AES) was utilized to verify DIMP adsorption. For the TEM and T E D investigations, a direct replica of the copper + cuprous oxide island structure was produced by using the standard lift-off technique. A micrograph of a typical discontinuous gas-sensitive film, whose average thickness is 100 A, is shown in Figure 5; the T E D pattern is shown in Figure 6. Qualitatively, the continuous rings indicate a random dispersion of polycrystallites, and their hroadening implies that the discrete crystallites are small, the lattice imperfections high, and there is a significant amount of strain or defects in the film. The film’s qualitative chemical composition was determined by indexing the TED. Consultation of the Joint Committee on Powder Diffraction Standards (JCPDS) file indicated that a mixed phase of copper and cuprous oxide is present (no evidence or cupric oxide). The motivation for comparing exposed and unexposed films using an AES investigation was accomplished to determine where DIMP is adsorbed. The AES spectra of a DIMP exposed, silicon dioxide surface loeated adjacent to the sensor’s electrodes and gas-sensitive film revealed silicon, oxygen, and
BO. NO. 17. SEPTEMBER 1. 1988
Figure 6. TED selected-area panern for the 100 A average film shown in Figure 5 (120-keV beam energy).
1733
lhicknes
copper
carbon transitions. A residual gas analysis indicated the presence of carbon monoxide, which likely accounts for the carbon transition. The AES spectra of the same sample with the Auger beam translated to the electrode gap region, revealed the superposition of the substrate, the adsorbed carbon monoxide, the copper + cuprous oxide film, and most importantly, a phosphorus transition that is only present for DIMP exposed samples. An examination of the sensor’s d i d copper electrodes revealed a similar result. Suhsequent short-duration (2 s). low-energy (500 eV) argon ion sputtering was used to probe the first few monolayers of the film’s surface. It was ohserved that the phosphorus transition was no longer resolvahle. In addition. phosphorus is not an observed spectral transition for an unexposed film sample. These observations strongly support the conclusion that DIMP is selectively adsorhed on the surface of the copper + cuprous oxide islands. Direct C u r r e n t Conductivity Studies. Direct current bias studies were used to screen several metallic candidates for DIMP sensitivity (650ppm of DIMP; 30% relative humidity; 22 OC; and a 100 A average film thickness). Of the discontinuous metallic films evaluated (AI, Ag, Au. Mg. Mn, and Ph). the copper + cuprous oxide films revealed conductivity increases on the order of 10-12%, while all the other candidates manifested either no response or a change of less than 1%. In addition, the copper + cuprous oxide films did not respond to either ultrahigh-purity AI or N,, or when either was used as a carrier for DIMP. When the copper films were initially exposed to just the filtered lahoratory air source, a slightly positive, irreversible resistance increase (approximately 3%) was ohserved. In addition, when lahoratory air with relative humidity 10 V (Figure 9) deDends on both the applied voltage and DIMP adsirption. Therefore, the following explanation is postulated for the transport mechanism. Qualitiatively, no carriers are excited at zero temperature and field since most trap and impurity sites below the Fermi level in the substrate are filled and few charge carriers can be transferred from one site to another by tunneling. At some critical temperature and near zero field, a finite population of carriers will be activated (thermally) above the Fermi level in the substrate. These carriers are free to hop from one trap site to another by a tunneling mechanism. This mechanism is activated because the charge carriers may not necessarily tunnel at a constant energy level. The density and distribution of the trap sites are attributable to the nature of the substrate (amorphous glass) and its electronic interaction with the carpet of the adsorbed DIMP + water complex. When carriers are thermally activated above the Fermi level of the substrate, this phenomenon distinctly implies that the tunneling current behaves according to the Arrhenius relationshiop (A(T,g)V= [ p exp(-Ea/kZ9]fl and represents the ohmic component of the total current. Since the experimental data reveal that the A coefficient and the activation energy (E,) are independent of temperature, the linear shifts between, the In A versus T'curves reflect the effect of the adsorbed DIMP + water complex and can be attributed to the preexponential temperature-independent coefficient (p). At low temperatures and when the field is increased, charges in the metal clusters may be injected into the trap sites in the insulator and hop from trap to trap and contribute to the current. Additionally, when the temperature is increased under the higher field conditions, the thermally activated mechanism can coexist with the nonactivated process. Consequently, the role of the adsorbed DIMP + water complex is to enhance the number and occupancy of the trap sites. Consistent with Hill's model (13),the experimental results suggest that the location of these sites is near the conduction band edge since the current is dramatically enhanced for conditions of high temperature and large fields. The observed decrease in conductivity associated with the desorption of the DIMP water complex is attributed to a reduced probability of occupancy or redistribution of the trap density near the conduction band edge of the substrate. For the nonohmic, high-field component of the phenomenological model, a space-charge-limited current mechanism is suggested by the v2 dependency (15,16). The space charge (Q) that can be forced into the insulator can be expressed as
+
Figure 10. Natural logarithm of the 8 coefficient versus reciprocal absolute temperature for three DIMP exposure conditions for 100 A average thickness films (average of six samples; the 10 mTorr profile represents an atmospheric exposure situation followed by an immediate evacuation).
nonlinear. These results suggests that the A ( T g ) V ( = [ p exp(-E,/kT)]V) term can be interpreted in terms of a thermal activation process. Similarly, the B(T,g)VLterm can be regarded as a nonactivated, field-injection (or hopping) conduction mechanism. Before discussing a transport model, it is acknowledged that, since the film's morphology is inhomogeneous (random distribution of island sizes and gap separations), it is reasonable to expect that a few preferential paths will carry most of the current. In addition, the distribution of potentials across the gaps in the dominate conduction paths implies that the corresponding field intensities will be pronounced a t locations of high resistance (narrowest gaps). Although it is not possible to precisely describe the spatial distribution of the internal microscopic fields, it is assumed that the microscopic fields of consequence are dependent upon the applied voltage. Therefore, only the variation of conductivity as a function of applied voltage is discussed, and no attempt is made to correlate the calculated value of the applied field (based on electrode geometry) with the absolute magnitudes of the internal microscopic fields. While the proposed electron transport model is largely qualitative, it does unify the pertinent phenomena that are critical for interpreting the experimental results. Hill (13) has analyzed direct current conduction mechanisms in dielectric-supported discontinuous metallic film structures and presents an argument that does not require direct tunneling between islands. He advocates a conduction process via a large density of closely spaced traps, impurities, or localized states near the conduction and valence band edges of the dielectric, while fewer, more-widely spaced states are located near the Fermi level. Hill proposes that electron transport between the islands is accomplished via successive electron tunneling jumps to neighboring trapping sites, and ultimately to an adjacent island. The fundamental advantage of this model is that it is not necessary to raise an electron to the conduction band of the substrate, and since the tunneling distance between traps is small, tunneling between islands whose spacings are greater than 30 A can be explained. Further, based upon the observed affinity of the metallic clusters for the DIMP + water complex, it is proposed that the number, distribution, and probability of occupancy of the trap and impurity states are favorably influenced by the adsorbed contaminant. It is also reasonable to anticipate that the corresponding fluctuating surface dipole (charge) produced by the adsorbed contaminants will act to modulate (lower)
+
Q = CV
(2)
where Q is the space charge injected into the substrate, C is the capacitance between the metallic clusters, and V is the applied voltage. For those charges that tunnel through the insulator via the trap sites, the space-charge-limited current (Isc)can be expressed as
1736
ANALYTICAL CHEMISTRY, VOL. 60, NO. 17, SEPTEMBER 1, 1988
I,, = Qt-' = CVt-l
(3)
where t is the transit time between trap sites. If D is the quantum mechanical tunneling transmission coefficient (tunneling probability), it is reasonable to expect that
t
D-'
(4)
In Neugebauer and Webb's (17) classical paper, they assumed that D was proportional to V when the potential barrier is not significantly perturbed by small fields. This approximation may also be applied in this situation since the fields used in this research were not large (=1.2 X lo4 V cm-I maximum). Mostovech and Vodar (18) report that for a potential barrier on the order of 1 eV and a distance ( d ) between trap sites of approximately 10 A, then for the barrier to be lowered (aE)by lo-' eV, a field [ F = (2AE)(de)-'] of at least 2 x 106 V cm-' is required. Thus, eq 4 can be expressed as
t = kV-'
(5) where k is a constant of proportionality. Substituting eq 5 into eq 3 yields an expression similar to the nonohmic term in eq 1. That is
I,, = B(Td?)V
(6)
where B ( T g ) is a coefficient of proportionality which is a function of temperature and DIMP exposure. The nonlinear plots of In B ( T g ) versus T-' suggest that B(Tg)is temperature dependent and significantly influenced by the DIMP water complex adsorption. This situation implies that the adsorbed contaminant favorably enhances the occupancy of the substrate's shallow trap density, and the experimental results reveal a maximum current flow under high temperature and large field conditions.
+
CONCLUSION The modification of the electrical conductivity of thermally evaporated, discontinuous films of copper cuprous oxide supported on a dielectric substrate caused by exposure to DIMP mixed with filtered laboratory air (2-3% relative humidity) was investigated. Direct replica transmission electron microscopy was used to verify the film's discontinuous island morphology. The associated electron diffraction measurements revealed that the f i s consist of a mixed phase of copper and cuprous oxide. Auger electron spectroscopy confirmed the preferential adsorption of DIMP on the gas-sensitive film. Isothermal, direct current versus voltage DIMP exposure profiles were least-squares fitted to an I = A V B v phenomenological model where B = 0.01A. With respect to DIMP
+
+
exposure, the [ M I A ] parameter, [ABIB] parameter, and their ratio revealed that temperatures just less than 100 "C, thin films (100A), and a large bias (greater than 10 V), evoked the strongest gas sensitivity. Plots of In A versus reciprocal absolute temperature are linear (temperature independent). Corresponding plots of In B are nonlinear (temperature dependent). The ohmic component suggests a thermal activation process. The nonohmic component is indicative of a spacecharge, field-injection mechanism. Both mechanisms operate together, which suggests that carriers are thermally excited, field induced, and transported via shallow trapping and impurity centers contributed by the substrate and carpet of the adsorbed DIMP water molecules. The direct current conductivity studies have a profound influence on the design and operation of the RC notch network sensor. In particular, maximum gas sensitivity and performance should be evoked when thin films (
