Anal. Chem. 1980, 6 0 , 7737-7743
1737
Organophosphorus Compound Detection with a Supported Cuprous Oxide Island Film. 2. Alternating Current Copper Studies and Sensor Performance
+
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 alternating current properties modified by the exposure of a dlelectric supported discontinuous fllm of copper cup rous oxide to parts-per-million concentration levels of dilsopropyl methylphosphate (DIMP) are reported. I n addition, the merlt of integrating these fllms wlth the distrlbuted RC notch network sensor is investigated. The frequency-dependent resistance [R,(f)] is interpreted in terms of a simple equivalent clrcuR composed of frequency-Independent parameters whose values were determlned by uslng a nonlinear least-squares analysis. The parameter changes caused by DIMP exposure can be differentiated between those assoclated with the bulk conductlvity of the Islands compared to thelr gaps. A detection scheme based upon the dlstrlbuted RC notch network is evaluated. I n operation, changes made In the lumped notch tuning resistance to reestabllsh a preexposed notch depth dlrectly correlate wlth the independently measured reslstance change of the gas-sensltlve flim.
+
+
The importance of the copper cuprous oxide island film’s physical structure and its impact on explaining the direct current conductivity changes evoked by diisopropyl methylphosphonate (DIMP) adsorption has been considered (1). To complete the investigation, their alternating current behavior was examined. The bulk of the existing literature focuses on explaining the direct current charge transport mechanism operating in the gap between metallic islands and implicitly regards the discrete metallic particles as perfect conductors (2). Since the resistance of the individual metallic clusters may not be negligible, it is important to investigate conduction in both regions. Alternating current measurements afford this opportunity. It has been known since the 1930s that very thin, evaporated metallic films exhibit a decrease in resistance as the frequency of the applied potential increases (3, 4). This behavior is physically attributed to the presence of interisland capacitance, whose impedance decreases with increasing frequency (5-14). In an effort to interpret this phenomenon, the interisland capacitance has been modeled as a continuous distribution (5,6), as a discrete parameter (7),and as a mixture of the two (8). However, unless a film is ultrathin (bo ' ' ""%2 NURHRLIZED FREQUENCY
'
'"'T~s?~
Figure 3. Theoretical gain response for the first-order solution of the lossless distributed RC notch network (a= 17.798)and Several d i r Crete values of a (a = 11.0 and a = 25.0).
When the gas to he detected is adsorbed on the discontinuous metallic film, R will change. As a consequence, the notch and dielectric loss parameten will also change, assuming constant, or nearly constant, R. and G.(w). With respect to these two quantities, the notch parameter was found to he the most gas-sensitive, and the theoretical effects of varying a on the transfer function's gain for the lossless device is shown in Figure 3. (It is noted that simultaneous perturbations in GJw) only cause minor shifts in the position of the notch minimum along the normalized frequency axis.) An indication of how much R changes due to gas adsorption can he determined by changing the lumped notch tuning resistance (R.) to reestablish the correct value of a and, simultaneously, realize the condition of optimum null.
RESULTS AND DISCUSSION Alternating C u r r e n t Conductivity Studies. A Hewlett-Packard, Model 4192A, impedance analyzer was used to make the alternating current conductivity measurements. This instrument generates a 10-V peak-to-peak swept frequency signal spanning 5 Hz to 13 MHz, and it has the capacity to simultaneously apply a direct current bias level as large as 35 V. Its high accuracy (4'12 digits) and programmable features facilitate dynamic compensation of lead and test fixture parasitic impedances. The analyzer's equivalent circuit
ANALYTICAL CHEMISTRY, VOL. 60, NO. 17, SEPTEMBER 1, 1988
was used to calculate estimates of R,, R,, and C, in eq 1 by using the Rp(w) data (17). In general, the fit obtained was quite good and the parameters were calculated to within 10% or less (95% confidence limits). Analogous to the comparative technique used to analyze the coefficients in the phenomenological direct current conductivity model (I), the effect of DIMP exposure was quantified via the relation
- 0
2 -1
.. .. .. .. .. ...........> .
u w
E:
C
-:
.
m P'
E
w m C
*'\
.
101
s,
..........-
= ( x parameter,,&)
e -.e
- ( x parameter,,,-J
( x parameter,,,,,d)
2: a :
a:. B*x:*
=, ZR1 I
,,.,,,,
I b2
'
"""'
Ib3
~
"""'
rL4
' """'
FREQUENCY
I bs
1739
' """'
I bs
' "T ' " I 107
105
lHrl
Figure 4. Equivalent parallel resistance [ R , ( f ) ] versus frequency of a 122 A thick film with an applied 35-V direct current bias: (0)-20 "C, unexposed; (+) 90 "C, unexposed; (m) -20 "C, atmospheric pressure DIMP exposure (650 ppm); (0)90 "C, atmospheric pressure DIMP exposure (650 ppm).
measurement regards the unknown impedance (Figure 1) as a parallel combination of a conductance [GPO= l/RpO]and a susceptance [BP= oCp@]. An expression for Rp(w)in terms of R,, R,, and C, was calculated to be
Sixty-four pairs of alternating current impedance parameters [GPOand BPIcorresponding to frequencies in the range 5 Hz If 5 13 MHz were recorded for each isotherm characteristic. The measurements were accomplished by submersing the sample cell into the silicon oil bath ( I ) . Analogous to the direct current measurements, nine discrete temperature (-20, -10, 0, 10,20,30,50,70,90 "C) profiles were examined. The initial temperature for each sample's profile was -20 "C, and heat was progressively added until the 90 "C characteristic was measured. To generate each isotherm, two applied direct current bias levels (1.5 and 35 V), a 650 ppm DIMP exposure concentration, and an unexposed (control) sample were utilized. The selection of the two bias levels was strongly influenced by the direct current conductivity measurements (1). Since ohmic behavior was observed for applied direct current potentials less the 10 V, the initial characteristic was measured by using the 1.5-V bias. Correspondingly, nonohmic behavior was observed for applied direct current potentials greater than 10 V (I), and to ensure compatibility with the impedance analyzer, the second characteristic was measured by using a 35-V bias. Finally, since the direct current measurements (1)revealed a gas sensitivity-to-thickness correlation, three discrete film thicknesses (67, 122, and 163 8,) were examined. The alternating current resistance [Rp(f) = l/Gp(fl] characteristics reveal the expected decrease in RpOwith increasing frequency. The plots of Rp(f) for a 122 8, thick film with an applied direct current bias of 35 V are depicted in Figure 4. Qualitatively, these plots reveal how a particular response curve is affected by DIMP exposure. For example, the RpCf) versus frequency plots show that for a temperature of -20 "C, the unexposed and exposed responses are essentially indistinguishable (Figure 4). However, for a temperature of 90 "C, the effect of exposure becomes significant in the low-frequency limit. In addition, for identical evaluation conditions, the larger the direct current bias, the more pronounced is the shift between the exposed and unexposed film measurements. The qualitative analysis of the experimental results motivated a quantitative effort to determine if the data could be fitted to the equivalent circuit model (Figure 1). A Levenburg-Marquardt (18-20) nonlinear least-squares algorithm
I lc,t
temperature thickness applied bias
(4)
where x E (R,, R,, C,, and (R, + R J ] . The primary purpose of this analysis was to study changes in the conduction mechanism in the copper + cuprous oxide films due to DIMP exposure. It was anticipated that by separately observing changes in R,, R,, and C,, the effects due to DIMP adsorption in the different regions of the film could be distinguished. A secondary purpose was to identify the film thickness and operational parameters (direct current bias and temperature) for the RC notch network sensor which maximize gas sensitivity. The following trends were gleaned from the S,-parameter calculations: 1. For x E {R,, R,, C,, and (R, + R,)] a. The lSx135-V-bia > I S x l 1 . ~ - V - b irelationship ~ is valid for a given temperature, film thickness, and DIMP exposure level. b. For a given temperature and bias level, IS,l increases with an increased DIMP exposure. c. The IS,l parameter increases with increasing temperature for both bias and exposure levels for temperatures greater than 0 "C. d. The ISxl67-A-film > ISxll22-A-fih > ISx~163-A-filmrelationship is valid for either bias or DIMP exposure level for temperatures greater than 10 "C. e. The IS,l parameter's gas sensitivity (DIMP) is maximized for thin films (67 8,))a large direct current bias (35 V), and an operational temperature of 90 "C. 2. For temperatures greater than 30 "C and an applied direct current bias of 35 V, the ~SRJ < ISCI < ISR,+R,I< ~ S IR relationship is valid for each of the three fih thicknesses ana exposure levels examined. These results indicated that the 67 8, thick film was the most sensitive, and this fact motivated a more comprehensive study of the film's response to several discrete levels of DIMP exposure. The objective was to discern how the alternating current impedance parameters behaved with respect to DIMP exposure. To be consistent with the prior evaluations, the 67 A thick film was examined for three discrete temperatures (0,20, and 90 "C) and two applied direct current bias levels (1.5 and 35 V). The response to DIMP concentrations spanning 0.1-650 ppm were evaluated by using a 2.3% relative humidity carrier. For the most gas-sensitive response, the results of the nonlinear least-squares and S,-parameter calculations for the R, and R, quantities are depicted in Figures 5 and 6, respectively. These plots reveal that temperature markedly influences the electronic response of the metallic films. Temperatures below 20 "C significantly reduce changes in the equivalent circuit parameters over the entire DIMP exposure concentration range, However, at higher temperatures (90 " C), the effect of DIMP exposure becomes significant. The new element of information revealed by this investigation (for all temperatures and bias levels) is that, associated with the initial DIMP exposure (0.1 ppm concentration), the resistance of the metallic clusters (R,) decreases to a greater
1740
ANALYTICAL CHEMISTRY, VOL. 60, NO. 17, SEPTEMBER 1, 1988
'
0
. 9
.
9.
.. f f
..
-_ _..C
.. A-1
' ' ' ' ' ' 1
I
1 bo
3 ' ' ' ' ' 1
' """'
lbl
>
I
-
I & >
-
?PO
-.,or
3
-0
-
.IO
~-
I , , .
or T
lb2
' ' 1 ' ' ' '
1'11
-
3 s
> b L
_ _ - -
. o
I
D
, O L
I d ,
3
Figure 7. Natural logarithm of the cluster resistance (R,) versus reciprocal absolute temperature for a 67 A thick film with an applied 35-V direct current bias: (0)unexposed; (+) exposed to 650 ppm DIMP and evacuated to 10 mTorr pressure: (M) exposed to 650 ppm DIMP at atmospheric pressure.
#
'1b-2'
-
.
' i""l 103
under conditions where a change in the resistance of the discontinuous thin film is maximized when exposed to DIMP. With this motivation, the Rp(f)versus frequency plots (Figure 4) reveal two regions of particular interest 1. 2.
..
In addition, the influential frequency effect of the gap capacitance can be minimized by avoiding operation of the sensor in the frequency range where Rpcf)makes a transition between these limits. The experimental data for all three film thicknesses investigated suggests that the frequency range 4.5 x l o 4 Hz If 5 6.0 X lo5 Hz should be avoided. The two frequency intervals with the most potential are 1.
2.
* * * +
. . . ..' . Ibl
+.*.
...m
.2
1b-1
lim Rp(f) = R, f-m
1bo
DIMP C o n c e n t r a t i o n Ippm)
bz
1
77103
Figure 6. Percent change in the gap resistance (R,) versus DIMP concentration for a 67 A thick film with an applied 35-V direct current bias: (0)0 O C ; (+) 20 OC; (m) 90 OC.
extent compared to the gap resistance (R&. However, when the DIMP exposure level exceeds 0.5 ppm, the gap resistance change exceeds that of the metallic clusters. This feature is in consonance with the Auger spectroscopy observations ( I ) that the metallic islands are the preferential adsorption sites. As discussed in the Theory of the Distributed RC Notch Network Sensor, it would be desirable to operate the sensor
o5f
5 4.5 x 104 H~
6.0 x 105 HZ
< f < 1.3 x
107 H~
Since the R, parameter displayed the greatest sensitivity toward DIMP, its influence was incorporated in the sensor's design. This objective was accomplished by operating the sensor at the low-frequency limit (0 < f < 4.5 X lo4 Hz). Synergistically, this situation includes the effect of DIMP exposure on both the R, and R, parameters, while the influence of the gap capacitance (C,) is minimized. Finally, as a check on the physical efficacy of the equivalent circuit model, the experimental data affords a suggestive verification. Least-squares fitted plots of the natural logarithm of the impedance parameters {Rc,R,, C,, and (R, + R,)) versus reciprocal absolute temperature were generated. The R,-parameter plots are linear with a negative slope; the R,- and (R, + R,)-parameter plots are linear with a positive slope; and the C,-parameter plots are nonlinear. It is noted that the negative slope associated with the R, parameter (metallic cluster resistance) is indicative of metallic-like behavior. The distinctive positive slope of the R, (gap resistance)- and (R, + R,) (low frequency, direct current impedance limit)-parameter plots is consistent with the behavior observed in the direct current measurements (activated process). Further, the activation energies calculated from the (R, + R,)-parameter plots are consistent with the magnitudes and trends observed in the earlier direct current measurements for comparable film thicknesses and operating conditions ( 1 ) . Figures 7 and 8 illustrate the behavior of the R, and R, parameters for the most sensitive metallic film.
ANALYTICAL CHEMISTRY, VOL. 60, NO. 17, SEPTEMBER 1, 1988
1741
Table I. Calculated and Measured Tuning Parameters for the Distributed RC Notch Network Sensors Operated at 90 "C with an Applied 35-VDirect Current Bias and the Changes Evoked by Exposure to DIMP" 203 pm wide electrode gap (71 8, thick film)
unexposed calculated parameters fn, Hz Rn,
Q
measured parameters
Hz
fm
Rn,
Q
notch depth, dB distributed resistance measurements R, Q a = R/Rn DIMP exposure effectsa (detuned results)
4.839 x 103 2.626 X 10' 5.42 x 103 2.65 x 103 87.1 1.002 x 108 3.778 x 104
AR
exposed 5.5510025 X lo3 2.429 213 2 X 10' 6.09 x 103 2.31 x 103 87.2
ARn
Anotch depth, Aff
unexposed 1.47650 X lo2 2.62687 X 10' 1.23 X lo2 2.24 X 10' 60.7
8.734 x 107 3.774 x 104 -12.834 -1.107 -12.753 -13.892 -12.753
Afn
508 pm wide electrode gap (126 A thick film)
8.386 X los 3.743 x 107
exposed 1.257 413495 X lo2 2.397888748 x 10' 1.345 X lo2 2.050 X 10' 60.3 7.655 X lo8 3.734 x 107 -8.717 -2.439 -8.482 -10.544 -8.718
"650 ppm DIMP concentration mixed with a 2.3% relative humidity laboratory air carrier; values expressed as a percent.
parameters in Table I motivated the dielectric loss compensation calculation. To implement the dielectric loss compensation technique, discrete parallel plate test capacitors with dimensions identical with those of the two sensors were fabricated adjacent to each distributed RC notch network ( I ) . The H P 4192A impedance bridge was used to measure the dissipation factor (loss tangent-tan 6 ) and capacitance (C,)as a function of frequency. The equivalent series resistance [R,(w)]was determined from the following relationship: tan 6 = wR,(w)C,
I
'1
i'eo
2'70
2'10
isP
y'Qv
,'*
y ~ ~ o , ~ o o ~ ~ ~ ~ ~ , s ~ . ~
Figure 8. Natural logarithm of the gap resistance (R,) versus reciprocal absolute temperature for a 67-A thin film with an applied 3 5 4 direct current bias: (0)unexposed: (+) exposed to 650 ppm DIMP and evacuated to 10 mTorr pressure; (B) exposed to 650 ppm DIMP at atmospheric pressure.
Therefore, in consonance with the direct current conductivity studies ( I ) , the alternating current impedance analysis supports the position that thin films (e100 A), with a large applied direct current bias (>lo V) operated at a temperature just less than 90 "C, and excited with frequencies less than los Hz, are the most gas sensitive. This information formed the basis for the design and operation of the distributed RC notch network sensor. Distributed RC Notch Network Sensor Performance. The distributed RC notch network sensors discussed earlier (I)were fabricated considering the operational constraints and gas sensitivity optimizing features. For practical reasons, the third-order solution was used for the 203 pm wide electrode gap sensor, and the fifth-order solution was used for the 508 pm wide electrode gap sensor. For both sensors, the greatest change in the notch depth and frequency was observed to be consistent with the operational parameters identified earlier. That is, maximum sensitivity for a 650 ppm DIMP exposure was elicited for operation at 90 "C with an applied 35-V direct current bias. Table I summarizes the performance of the two sensors operating at the optimum conditions. The magnitude of the error observed between the calculated and measured
and plotted as a function of frequency. The ideal theory of the distributed RC notch network, along with the measured value of R and the calculated value of C, was then used to estimate the critical notch parameters (notch frequency and Rn). In conjunction with the R,(w) versus frequency plot, the estimated notch frequency was used to interpolate an estimate of R,(w). Accordingly, the ratio of R,(w) to R specifies the dielectric loss parameter [ ~ ( w ) ] With . this value of ~ ( w ) eq , 2 was used to calculate the dielectric loss compensated estimates of the notch frequency and notch resistance (Rn). The merit of the dielectric loss compensation calculation is reinforced by comparing the tuning parameter errors tabulated in Table 11. Clearly, the errors associated with the Af,, AR,,, and Aa parameters are markedly reduced. Since the 203 pm wide electrode gap sensor possessed superior DIMP exposure performance, a more complete calibration profile was measured a t the optimum operating parameters (90 "C and an applied 35-V bias). The results are summarized in Table 111. Finally, the response time to attain complete equilibration for a 650 ppm DIMP challenge was investigated. After a 60-min purge with laboratory air (50 mL/min; 2.3% relative humidity), the sensor was challenged with a 650 ppm DIMP exposure and the response was recorded every 30 s. As depicted in Figure 9, a complete response is attained in 2 min.
CONCLUSION The alternating current conductivity experiments revealed that the equivalent frequency-dependent resistance [R,(f)] decreased with increasing frequency. Exposure to the DIMP + water complex produced prominent changes in the R p ( f ) versus frequency data. A nonlinear least-squares computer
1742
ANALYTICAL CHEMISTRY. VOL. 60. NO. 17. SEPTEMBER 1. 1988
Table 11. Dielectric Loss Compensated Tuning Parameters and Errors for the Distributed RC Notch Network Sensor Operated at 90 "C with an Applied 35-V Direct Current Bias 2u:i um wide electrode zap 171 A thick film!
unexpusrd
exposed
3.548611 X I@' 3.54152794 X 1 0 . ' 1.500 172 109 X 10' 4.862923265 X IO3 3.fimfimziz x in'
3.099435 X Io" 3.5487004 X 1.50018491 X 10' ~ . s m 9 0 4 zx 103 3.60038675 X IO'
dielectric loss compensated tuning parameters R J f A I?
ncn. 2 (w/wJ
f., H z R., I1
508rm wide electrode gap (126 A thick film)
unexposed
rrpwed
z.122243 x 104 2.5189875 X IO-$ 2.50921241 X IO* 1.16196454 X 10' 3.86399858 X 10'
1.951635 X IO' 2.54949529 X 1 0 . ' 4.509908658 X 10' 1.2731294616 X 10' 3.8732C6131 X IO'
compensated tuning parameter errorsb 11.250 -4.623 4.847
9.159 -4.527 4.741
5.855 3.212 3.112
5.631 3.724 3.590
Af"
11.m
AN
-8.529 9.325
9.701 -8.565 9.367
7.176 -14.727 17.271
6.966 -14.508 16.970
Af"
a" AN
uncompensated tuning parameter errorsb
a"
'Calculated from [ R , V ) / R ] . 'Expressed as a percent. Table 111. Parameter Changes Evoked by DIMP Exposure for the Distributed RC Notch Network Sensors (203pm Wide Electrode Gap and 71 A Thick Film) Operated at 90 'C with an Applied 35-V Direct Current Bias
DIMP concn. ppm unexposed 1
10 50 100 300 650
[notch frequency
V). ( x IO3 H z ) / 3 f m(%)I I-..-,
[5.40/0.d271 [5.39/+.2121 [5.317/+.582] [5.36/4.767] 15.35/4.9521
[notch resistance ( R J (x103O)/ARn
,
(%PI
-.
[2.5669)-3:15531 [2.3i99/-l0.20Sl [2.:i5318/-11.1931] [2.43360/-11.5401
[notch depth (dR)/ Anotch depth (%PI
[electrode resistance ( R ) (XlO' W / A R (%)*I
.-..-,
.--~--~,
[84.2/-3:472] [78.31/-lO.Z31l [77.5/-11.153]
p.fi95/-3.i861 [8.987/-10.256] 18.8880-1 1.2441 [8.85l0/-1 1.6141 18.7690/-12.4331
[77.1/-11.6162]
176.4/-12.4141
'&posed value. 'Adjusted value to maximize notch depth upon exposure. algorithm was developed, and the experimental data were fitted to an equivalent circuit model. The equivalent circuit's frequency-independent components include a resistance (RJ in series with the parallel combination of a resistance (RJ and a capacitance (CJ. Conduction through the metal clusters is associated with R, and that through the gaps with R, and
c,.
With respect to DIMP exposure, the R, parameter exhibited the greatest change for concentrations less than 1 ppm. However, for increasingly greater DIMP exposure levels, the R, parameter was the most sensitive. In consonance with the direct current measurements ( l ) ,higher temperatures (90"0, thin films (