![Table 1. Chemical Agent Structures, Hydrolysis Half-lives, and JSAWM Thresholds.
Agent Short-Hand Chemical Structure Hydrolysis JSAWM
Half-Life* Thresholds
Sarin (GB) F-[O=P-CH3]-O-CH(CH3)2 21.3 hours 3.2 µg/L
Soman (GD) F-[O=P-CH3]-O-CH(CH3)-(C-(CH3)3) 2.3 hours 3.2 µg/L
Tabun (GA) (CH3)2-N-[O=P-CN]-O-C2H5 4.1 hours 3.2 µg/L
VX C2H5O-[O=P-CH3]-S-(CH2)2-N-(CH(CH2)2)2 82.1 hours 3.2 µg/L
EA2192 HO-[O=P-CH3]-S-(CH2)2-N-(CH(CH2)2)2 >9 years 3.2 µg/L
Mustard (H) ClCH2CH2-S-CH2CH2Cl encapsulates 47 µg/L
Lewisite (L) ClCH=CH-As-Cl2 rapid 27 µg/L
HCN HCN rapid 2.0 mg/L
BZ** C7NH12-O-[C=O]-COH(C6H5)2 2.3 µg/L
T-2 Toxin 8.7 µg/L
* at pH 7 to 7.5 and 20 to 25 oC.
demonstrated by vibrational spectroscopy.7-10 Hoffland et al.7 reported infrared absorbance spectra and absolute Raman cross
sections for several chemical agents, while Christesen measured Raman cross sections for sarin, tabun, mustard gas, and VX
(ethyl S-2-diisopropylamino ethyl methylphosphonothioate).11 Again, however these techniques also have limitations.
Raman spectroscopy is simply not a very sensitive technique, and detection limits are typically 0.1% (1000 ppm). While
infrared spectroscopy would have limited value in analyzing poisoned water, since the very strong infrared absorption of
water would obscure most other chemicals present. Nevertheless, efforts to overcome these limitations have been
demonstrated. Braue and Pannella8 quantified the G-series nerve agents (tabun, sarin, and soman) in terms of infrared
attenuated total reflectance using a circle-cell. And Alak and Vo-Dinh demonstrated the possibility of surface-enhanced
Raman spectroscopy (SERS) to identify CWAs by measuring several organophosphonates that simulate the nerve agents.12
However, quantitative measurements have not been demonstrated for the SER-active material used (silver coated on alumina
particles) or other SER-active media.13
Recently, we developed silver-doped sol-gels to promote the SER effect.14-17 The porous silica network of the sol-gel matrix
offers a unique environment for stabilizing SER-active metal particles, and the sol-gel provides a high surface area that
effectively increases the number of molecules observed within the Raman scattering volume. The silver-doped sol-gels have
been coated on the internal walls of standard glass vials to allow rapid SER analysis of small sample quantities (< 0.1 mL)
without preparation. We have used p-aminobenzoic acid (PABA) as a test chemical to demonstrate surface enhancements
greater than 106, reversible measurements in a flowing system, reproducible measurements from vial-to-vial and batch-to-
batch, and measurements in multiple solvents, including water.14-17 Recently, we used these vials to measure Tabun (GB) and
Sarin, and several hydrolysis products, pinacolyl methyl phosphonate (PMP from Soman), and methyl phosphonic acid
(MPA from all G-agents, Figure 1). Although a number of unique vibrational bands are observed (e.g. C-N stretch doublet
and P-C stretch), the G-agents were only observed for 5% concentrations, and all spectra required baseline corrections.
A C 790
C-N
P-C 2135, 2190
770 545
1290 B D 760
Wavenumber (∆cm-1) Wavenumber (∆cm-1)
Figure 1. Surface enhanced Raman spectra of ~5% v/v A) Tabun and B) Sarin, C) 1% v/v PMP and D) 10 ppm MPA
using sol-gel sample vials, 785 nm excitation, 1-min scan, and CCD detection. Performed at Aberdeen Proving Ground.
Proc. SPIE Vol. 4577 167](https://image.slidesharecdn.com/finalreportdaad1302c0015part5appaf-13347354740707-phpapp01-120418025232-phpapp01/85/Final-Report-Daad13-02-C-0015-Part5-App-A-F-20-320.jpg)
![dominance of the P-C and the CH2 stretches, and the disappearance of the P-O-C mode in the upper spectrum suggest the
molecule is oriented with the methyl-hydroxyl-hydroxyl face towards the surface. However, considerably more research
must be performed to verify these points.
Both chemical forms were measured as a function of concentration to determine the onset of monolayer coverage and
estimate expected detection limits (Figure 6). Below monolayer coverage the signal to concentration dependence should be
linear, and the S/N of any spectral measurement in this range can be used to predict the detection limit. In the spectra
presented here, the peak height was used as the signal, while the noise, as root-mean-squared (RMS) was measured between
4400-4600 cm-1. Since noise is distributed evenly throughout the spectrum when transformed, this region was used since it
does not have any contributions from signals or baseline offsets. Figure 6 shows a series of spectra for MPA along with a
plot of the 1050 cm-1 band intensity (the noise was identical for this series) as a function of sample concentration. A clear
discontinuity appears in the vicinity of 0.1 mg/mL (19 ppm), indicating the onset of monolayer coverage. A detection limit,
defined as a S/N of 3, was calculated for the 0.1 and 0.05g/mL samples at 2.4x10-4 and 2.5x10-4 g/L, respectively. A more
modest detection limit of 10.1x10-4 g/L was obtained using the 760 cm-1 band in the second series of concentration
measurements. These detection limits correspond to 46 and 190 ppb for a 3-min scan and 75 mW of incident laser power.
140 600
A 120 B 500
100
400
I (1050)
I (760)
80
300
60
200
40
20 100
0 0
0 0.2 0.4 0.6 0.8 1 1.2
[MPA] (mg/mL)
Wavenumber (∆cm-1)
Figure 6. A) Concentration dependence of MPA SERS measured in silver-doped TMOS). B) Concentrations are 0.01, 0.05,
0.1, 0.5, 1 g/L (1.88, 9.4, 18.8, 94, 188 ppm). I760 series (•) and I1050 series (∆).
Enhancement factors for cyanide and methyl phosphonic acid can be determined by estimating the number of molecules
contributing to the surface-enhanced and normal Raman spectra. The enhancement factor, EF, can be defined by the
following equation:
EF = (ISERS/INR)•(MNR/MSERS) •(PNR/PSERS) •(TNR/TSERS)1/2
where I is the spectral band intensity, M is the sample mass, P is the incident laser power, and T is the measurement time (or
number of scans) for the two measurements. For the normal Raman spectra a cylindrical scattering volume is assumed, based
on the laser area (2.8x10-7m2, 6x10-4m diameter spot) and the penetration depth (1x10-3 m).24 The density of KCN and MPA
as powders were measured at 0.572 and 0.516 g/cm3, indicating that 1.6x10-4 and 1.44x10-4 g produced the normal Raman
signals in Figure 5, respectively.
The number of molecules contributing to the SER spectra are those on the silver particles embedded in the sol-gel. The total
silver surface area can be determined from the average particle size, concentration, and the scattering volume. Previous
scanning electron micrographs showed the average silver particle size to be 40 nm in diameter (3.35x10-23m3).17 The silver
concentration is 0.12M based on the reactant molar concentrations and dilution factors. And the scattering volume is 7.6x10-
11 3
m , again based on a cylindrical scattering volume, defined by a laser area of 2.8x10-7m2 and a sol-gel thickness of 2.7x10-
4
m. This volume contains 1.23x10-6g of silver, equivalent to 3.5x109 silver particles with a collective surface area of 1.8x10-
5 2
m . However, it may reasonably be assumed that at least half of the silver surface is in contact with the sol-gel matrix and
unavailable for analyte interaction. If we assume monolayer coverage and that each CN molecule occupies 1.5x10-20m2, then
approximately 6.2x1014 molecules or 2.7x10-8g of CN contribute to the SER spectrum (2.0x10-19m2, 4.6x1013 molecules,
7.4x10-9g for MPA). Accordingly, the EF for cyanide equals 4.8x104 ((180/599) •(1.6x10-4/2.7x10-8) •(900/75) •(500/100)1/2).
The EF for MPA is considerably higher at 8.7x106 ((603/26) •(1.44x10-4/7.4x10-9) •(900/75) •(500/200)1/2).
Proc. SPIE Vol. 4577 171](https://image.slidesharecdn.com/finalreportdaad1302c0015part5appaf-13347354740707-phpapp01-120418025232-phpapp01/85/Final-Report-Daad13-02-C-0015-Part5-App-A-F-24-320.jpg)
![appeared and grew as a function of time, and that the higher the laser power the faster the growth. Figure 1 shows the
growth of the 1050 cm-1 band over the course of 30 minutes when using 150 mW of 785 nm excitation, while Figure 2A
shows that the growth can be fit with a first order exponential equation, namely I1050 = 0.3+0.5e-0.13t. Furthermore, the
760 cm-1 band could be fit with a first order decay equation with an identical exponential rate constant, i.e. I760 = 0.8-
0.8e-0.13t. The rates represent classical first order kinetics and their correspondence allows one to conclude that MPA is
being transformed one-for-one into a photo-generated product. At this time the photoproduct has not been positively
identified, but phosphonic (phosphorous) acid and phosphonate are likely candidates, since the symmetric P(OH)3
stretch occurs at ~1050cm-1. Our studies show that a reduction in laser power to 100 mW at the sample for MPA
essentially eliminates this degradation process. This laser power or lower was used for further measurements.
MPA Photodegradation
A B
Raman Intensity (relative)
0
Arbitrary Units
min
30
600
570 800
770 1000
970 1200
1170 1400
1370 0 10 20 30 0 10 20 30
Raman Shift (cm-1) time (min) time (min)
Wavenumber (cm-1)
Figure 1. Growth of 1050 cm-1 band as a function of time Figure 2. A) Exponential growth of 1050 cm-1 band and B)
due to exposure to 150 mW of 785 nm. Spectra are 5 sec exponential decay of 760 cm-1 band for spectral series in Figure 1.
each, collected every 100-sec from 0 to 30-min.
Methyl phosphonic acid is a diprotic acid that stepwise dissociates into two anions, MPA- and MPA=, according to the
following reactions:26
MPA MPA- + H+ pKa1 = 2.12 Reaction 2
MPA- MPA= + H+ pKa2 = 7.29 Reaction 3
The relative concentrations of MPA, MPA-, and MPA=, can be determined at any pH by expressing [MPA] and [MPA=]
in terms of [MPA-] using Reactions 2 and 3, and summing all three to equal the total starting concentration, here 2
mg/mL (0.021M, MW = 96.02), viz:
[MPA] + [MPA-] + [MPA=] = 0.021M Equation 1
substituting from Reactions 1 and 2:
([H+][MPA-])/K1a + [MPA-] + (K2a[MPA-])/[H+] = 0.021M Equation 2
rearranging:
[MPA-] = 0.021M/(1+[H+]/K1a + K2a/[H+]) Equation 3
The relative concentrations of MPA, MPA- and MPA= as a function of pH are shown in Figure 3. It is worth noting
that near neutral pH both MPA- and MPA= will be present. To confirm that the SER signal followed this pH
dependence, a starting solution, consisting of 20 mg of MPA in 10 mL HPLC grade water was prepared and brought to
pH of 2.0 using dilute nitric acid. From this solution, 2 mL were added to a SER-active vial, and the SER spectrum
recorded. At this pH a peak at 760 cm-1 was barely discernable. The 2 mL solution was returned to the starting solution
and the pH was re-measured to correct for any changes that the silver-doped sol-gel vials might cause. In most cases
the change was less than 0.2 pH units, and the pH is reported as the before and after average. Next, the pH of the
SPIE -2003-5269 119](https://image.slidesharecdn.com/finalreportdaad1302c0015part5appaf-13347354740707-phpapp01-120418025232-phpapp01/85/Final-Report-Daad13-02-C-0015-Part5-App-A-F-33-320.jpg)
![starting solution was adjusted to 3.25 using dilute KOH. Again 2 mL were added to a vial and the SER spectrum
recorded. At this pH a reasonably strong 760 cm-1 band was observed. This process was repeated as spectra were
recorded at pHs of 7.0, 7.4, 7.5, 7.9, 8.5, and 10.0. A total of 1 mL of KOH was added, diluting the total concentration
by 10%. Next, the pH of the starting solution was made acidic by adding dilute nitric acid dropwise. This time spectra
were recorded at pHs of 7.2, 6.9, 6.4, and 3.7. Figure 4 shows the SER spectra for representative pHs (spectra were left
out to simplify the figure), while Figure 3 shows the 760 cm-1 peak intensities as a function of pH. (The band intensities
were adjusted to compensate for dilution effects caused by the addition of HNO3 and KOH, then normalized to 0.021 M
for the most intense band observed at pH 3.7.) It is clear from Figure 3, that the 760 cm-1 band follows the MPA-
concentration as a function of pH and must be assigned to this anion. No bands were observed that corresponded to
MPA or MPA=. The lack of an MPA SER spectrum may be due to the absence of an attraction between the neutral
analyte and the electropositive silver surface. The same reasoning suggests that a strong SER spectrum should be
observed for MPA=, but it is not, and a satisfactory explanation has not been found.
0.020 - =
MPA MPA
MPA
Concentration [M]
0.015
pK1 = 2.12 pK2 = 7.29
0.010
0.005
0.000
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
pH
Figure 3. Concentration dependence of MPA, MPA-, and MPA= Figure 4. SER spectra of 0.02M MPA as a function of pH.
as a function of pH for a 0.02M sample. Intensity of 760 cm-1 Conditions: 100 mW of 785 nm, 36 scans (1 min), 8 cm-1,
band from Figure 6 as a function of pH. (■ for increasing basic recorded 2 min after sample introduction. pH 1.9 and 2.0 not
adjustment, ● for increasing acidic adjustment, error was measured apparent on this scale, pH 6.9 and 7.4 near identical to 7.0 and
at ~10% for pH 6.4). 7.5, and not shown for clarity.
Since these measurements involved the removal and 1.8
Raman Intensity (760 cm-1)
replacement of the SER-active vial in the sample holder 1.6
to remove and add sample, variation in the intensity as a 1.4
function of vial position was minimized by illuminating 1.2
the exact same height along the vial wall. But this does
1.0
not account for variability of the SERS response of the
sol-gel coating around the vial. To analyze this effect, a 0.8
vial containing MPA at pH 6.4 was rotated at ~ 40o 0.6
intervals at the original height, and 1/8” above and below 0.4
this value. Figure 5 shows the intensity of the 760 cm-1 0.2
band for the 27 positions. It was found that the average
0.0
value was 1.37±0.14, an RSD of 10% overall and 5% for
0 5 10 15 20 25 30
each height. An error bar is included in Figure 4 for the Measurement Number
pH 6.4 measurement.
Figure 5. SER spectra of 0.02M MPA at pH 6.4 measured
around a vial at three heights (9 points per height).
Conditions as in Figure 4, but 10-sec scans.
SPIE -2003-5269 120](https://image.slidesharecdn.com/finalreportdaad1302c0015part5appaf-13347354740707-phpapp01-120418025232-phpapp01/85/Final-Report-Daad13-02-C-0015-Part5-App-A-F-34-320.jpg)
![0.007
DPA DPA- DPA=
pH 795 0.006
Concentration [M]
525 0.005
1.35 DPA
1.71 0.004 DPA-
2.19 0.003 pK1 = 2.16 pK2 = 6.92 DPA=
3.83 795
0.002 525
0.001
0
0 2 4 6 8 10 12 14
pH
Figure 8. SER spectra of 1 mg/mL DPA as a function of pH. Figure 9. Concentration dependence of DPA, DPA-, and
Conditions: 100 mW 785 nm, 100 scans (44-sec). DPA= as a function of pH for a 0.006M sample. Intensity of
525 (■) and 795 (♦) cm-1 bands from Fig. 8 as a function of
pH.
The ability to detect dipicolinic acid (DPA) by SERS is demonstrated in Figure 10. Even at 1 mg/L the primary bands
are visible. The intensity of the dominant ring breathing mode observed at 1008 cm-1 for DPA in water (pH = 2.5 to
5.5) is plotted as a function of concentration in Figure 11. These values were also used to estimate limits of detection
based on the S/N of the 1008 cm-1 band using 100 mW of 785 nm laser excitation and a 10-min acquisition time.
Again, the lower the measured concentration, the lower the predicted LOD (see Figure 11 inset), and detection of 160
µg/L is possible. Gastrointestinal anthrax requires significant more spores than inhalation anthrax,28 and a limit of
detection might be placed at 1 million spores in 1 liter of water or 10 µg/L. Since each spore contains ~10% CaDPA by
weight,29 a goal for DPA might be 1 µg/L, indicating that the present measurements must be improved by nearly two
orders of magnitude.
2.5
2
1008 Band Intensity
1.5
A
1 conc (mg/L) lod-10min-100mw
B 1 0.17
C 10 0.16
0.5 100 1.03
D 1000 3.55
0
0 200 400 600 800 1000 1200
DPA Concentration (mg/L)
Figure 10. SER spectra of DPA in water at A) 1000, B) 100, Figure 11. Plot of SER intensity of 1008 cm-1 band of DPA
C) 10 and D) 1 mg/L. Conditions: pH of 2.5-5.5, silver-doped as a function of concentration using 175 mW of 785 nm.
sol-gel coated vial, 175 mW of 785 nm, 1-min, 8 cm-1. D) has Inset table includes LOD in mg/L for each concentration, but
been multiplied by x10 to make bands visible. for 100 mW and 10 min.
Due to the increased hazards of handling HCN gas, KCN salt was used for these experiments. Nevertheless, all sample
preparations were performed in a chemical hood. KCN completely dissolves in water, but its conjugate acid, HCN, is
formed and has a Ka of 6.15x10-10,30 viz:
HCN CN- + H+ pKa = 9.21 Reaction 4
SPIE -2003-5269 122](https://image.slidesharecdn.com/finalreportdaad1302c0015part5appaf-13347354740707-phpapp01-120418025232-phpapp01/85/Final-Report-Daad13-02-C-0015-Part5-App-A-F-36-320.jpg)
![Consequently, the cyanide concentration must be determined for each initial KCN concentration. Specifically, the
samples prepared with concentrations of 0.1, 1, 10, 100, and 1000 mg/L of KCN produced CN- concentrations of
6.3x10-3, 0.33, 6.9, 89, and 964 mg/L, at pHs of 8.16, 9.0, 9.67, 10.2, and 10.7, respectively. The pH dependence for the
HCN and CN- concentrations are shown in Figure 12. Thus, as the amount of KCN added to the solution decreases so
does the pH of the solution (becomes less basic) and according to Reaction 4, the relative amount of CN- to HCN also
decreases. For example, in the preparation of a 0.1 mg/L solution of KCN, the pH is shifted from 7 for pure water to
only 8.16, and only 6.3% of the starting material becomes CN-, or 6.3x10-3 mg/L. In comparison, for a solution of 1000
mg/L, the pH is shifted from 7 to 10.7 and 96% of the starting material becomes CN-. This is significant since the
cyanide ion is better able to adsorb onto the silver particles and become SERS active. SER spectra of 10, 100, and 1000
mg/L of KCN samples are shown in Figure 13 and a corresponding plot of the 2100 cm-1 band intensity as a function of
concentration is shown in Figure 14.
1.1
1.0
Concentration [mg/mL]
0.9 -
HCN CN
0.8
0.7
0.6
0.5 pKa = 9.21
0.4 A
0.3
0.2 B
0.1 C
0.0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
pH
Figure 12. Concentration dependence of HCN and CN- as a Figure 13. SER spectra of KCN in water at A) 1000, B) 100,
function of pH for a 1 mg/mL sample. Calculated intensity and C) 10 mg/L. Conditions: pHs of 10.7, 10.2, and 9.7, silver-
of 2100 (■) cm-1 band for a 1 mg/ml sample at pHs of 8.16, doped sol-gel coated vial, 100 mW of 785 nm, 1-min, 8 cm-1.
9.0, 9.67, 10.2, and 10.7. C) has been multiplied by x10 to make band visible.
140
120
2100 Band Intensity
100
80
conc (mg/L) lod-10min-100mw Condition
60 0.1 0.01 Au-pH 12 A
1 0.07 Au-pH 12
6.9 0.03 Au-pH 9.7
40 6.9 0.16 Ag-pH 9.7 B
20
89 0.22 Ag-pH 10.2 C
946 1.13 Ag-pH 10.7
0
0 200 400 600 800 1000 1200
CN Concentration (mg/L)
Figure 14. Concentration dependence of KCN SERS measured Figure 15. SER spectra of KCN in water at A) 10, B) 1,
under conditions in Fig. 11. Concentrations are 1, 0.1, and 0.01 and C) 0.1 mg/L. Conditions: pHs of 9.7, 12, and 12,
mg/ml. Intensities are measured for the CN stretch at 2100 cm-1. gold-doped sol-gel coated vial, 100 mW of 785 nm, 1-
Inset table includes LOD in mg/L for each concentration in Figs min, 8 cm-1. C) has been multiplied by x10 to make band
13 and 15, but for 100 mW and 10 min. visible.
The SER spectra of cyanide are dominated by the single CN stretch at 2100 cm-1, which occurs in normal Raman
spectra of solutions at 2080 cm-1. However, a low frequency mode occurs at 135 cm-1 due to a Ag-CN stretch (not
shown), indicative of a strong surface interaction. It is also observed that as the concentration decreases the CN stretch
shifts to 2140 cm-1. This shift has been attribute to the formation of a tetrahedral Ag(CN)32- surface structure,31 as well
SPIE -2003-5269 123](https://image.slidesharecdn.com/finalreportdaad1302c0015part5appaf-13347354740707-phpapp01-120418025232-phpapp01/85/Final-Report-Daad13-02-C-0015-Part5-App-A-F-37-320.jpg)
Final Report Daad13 02 C 0015 Part5 App A F
- 1. “sf” RTA P2001#05 DOD SBIR Phase II Contract Number DAAD13-02-C-0015 Final Report December 2005 Appendices A Lee, Y.H. and S. Farquharson, "Rapid chemical agent identification by surface-enhanced Raman spectroscopy ", SPIE, 4378, 21-26 (2001). B Farquharson, S., W.W. Smith, Y.H. Lee, S. Elliott and J. F. Sperry, "Detection of bioagent signatures: A comparison of electrolytic and metal-doped sol-gel surface-enhanced Raman media", SPIE, 4575, 62-72 (2002). C Farquharson, S., P. Maksymiuk, K. Ong and S.D. Christesen, "Chemical agent identification by surface- enhanced Raman spectroscopy", SPIE, 4577, 166-173 (2002). D Farquharson, S., A. Gift, P. Maksymiuk, and F. Inscore, “Rapid dipicolinic acid extraction from Bacillus spores detected by surface-enhanced Raman spectroscopy”, Applied Spectroscopy, 58, 351- 354 (2004). E Farquharson, S, A Gift, P Maksymiuk, F Inscore, and W Smith, “pH dependence of methyl phosphonic acid, dipicolinic acid, and cyanide by surface-enhanced Raman spectroscopy”, SPIE, 5269, 117-125 (2004) F Farquharson, S, A Gift, P Maksymiuk, F Inscore, W Smith, K Morrisey and SD Christesen, “Chemical agent detection by surface-enhanced Raman spectroscopy”, SPIE, 5269,16-22 (2004). G Inscore, F, A Gift, P Maksymiuk, and S Farquharson, “Characterization of chemical warfare G-agent hydrolysis products by surface-enhanced Raman spectroscopy”, SPIE, 5585, 46-52 (2005). H Farquharson, S, A Gift, P Maksymiuk, and F Inscore, “Surface-enhanced Raman spectra of VX and its hydrolysis products”, Applied Spectroscopy, 59, 654-660 (2005). I Inscore, FE, AD Gift, Stuart Farquharson, “Detect-to-treat: development of analysis of Bacilli spores in nasal mucus by surface-enhanced Raman spectroscopy”, SPIE, 5585, 53-57 (2005). J Farquharson, S, W Smith, C Brouillette, and F Inscore, “Detecting Bacillus spores by Raman and surface- enhanced Raman (SERS) spectroscopy”, Spectroscopy, June supplement, 8-15 (2005). K Inscore, F, A Gift, P Maksymiuk, JF Sperry, and S Farquharson, “Identifying surfaces contaminated with Bacillus spores using surface-enhanced Raman spectroscopy to detect dipicolinic acid”, in Applications of Surface-Enhanced Raman Spectroscopy, Ed. S Farquharson, CRC Press, Boca Raton, FL, accepted L Christesen, S, K Spencer, S Farquharson, F Inscore, K Gonser, J Guicheteau “Surface-enhanced Raman detection of chemical agents in water”, in Applications of Surface-Enhanced Raman Spectroscopy, Ed. S Farquharson, CRC Press, Boca Raton, FL, accepted. M Farquharson, S, F Inscore, S Christesen “Detecting chemical agents and their hydrolysis products in water”, in Surface-Enhanced Raman Scattering – Physics and Applications Eds. K Kneipp, M Moskovitz, and H Kneipp, Springer, accepted. N Inscore, F, S Farquharson, “Detecting hydrolysis products of blister agents in water by surface-enhanced Raman spectroscopy”, SPIE, 5993, 19-23 (2005). O Inscore, F, P Maksymiuk, S Farquharson, “Surface-enhanced Raman spectroscopic characterization of the chemical warfare agent vesicant HD and related mono-sulfides”, JRS, in preparation. P ROC curve data from measurements of CN, HD, and VX at the US Army’s Edgewood ChemBio Center. 74
- 2. Appendix A Rapid chemical agent identification by surface-enhanced Raman spectroscopy Yuan-Hsiang Lee and Stuart Farquharson* Real-Time Analyzers, 87 Church Street, East Hartford, CT 06108 ABSTRACT Although the Chemical Weapons Convention prohibits the development, production, stockpiling, and use of chemical warfare agents (CWAs), the use of these agents persists due to their low cost, simplicity in manufacturing and ease of deployment. These attributes make these weapons especially attractive to low technology countries and terrorists. The military and the public at large require portable, fast, sensitive, and accurate analyzers to provide early warning of the use of chemical weapons. Traditional laboratory analyzers such as the combination of gas chromatography and mass spectrometry, although sensitive and accurate, are large and require up to an hour per analysis. New, chemical specific analyzers, such as immunoassays and molecular recognition sensors, are portable, fast, and sensitive, but are plagued by false-positives (response to interferents). To overcome these limitations, we have been investigating the potential of surface-enhanced Raman spectroscopy (SERS) to identify and quantify chemical warfare agents in either the gas or liquid phase. The approach is based on the extreme sensitivity of SERS demonstrated by single molecule detection, a new SERS material that we have developed to allow reproducible and reversible measurements, and the molecular specific information provided by Raman spectroscopy. Here we present SER spectra of chemical agent simulants in both the liquid and gas phase, as well as CWA hydrolysis products. Keywords: Chemical warfare agent, simulant, hydrolysis product, SERS, Raman spectroscopy, sol-gels, vapor 1. INTRODUCTION Chemical warfare has been banned since the 1925 Geneva Protocol, yet the use of chemical agents has persisted.1 This can be attributed to the simplicity in manufacturing, ease of deployment, and the relatively low cost of chemical warfare agents (CWAs). These attributes make these weapons especially attractive to low technology countries and terrorists. Well known examples include the large-scale use of tabun (GA) during the Iran-Iraq war (1984-1948),2 and the release of sarin (GB) in the Tokyo subway in 1995. The latter is the first documented terrorist use of a chemical weapon.3,4 This ever-present threat was again substantiated by the United Nations Special Commission's report that described Iraq’s facilities for nerve agents, anthrax and small pox production.5-7 These uses of chemical weapons have motivated the development of fast and accurate analytical techniques to warn soldiers and the public at large. The development of these analytical techniques is challenging, in that these techniques must not only measure extremely low concentrations quickly (microgram/liter in < 1minute), but must also be capable of measuring both gas phase and liquid phase to be effective. The latter is required since chemical agents can also be used to "poison" water supplies.8,9 The earliest technologies developed for CWA detection were based on electrochemical, ionization, or colorimetric analysis (e.g. phosgene tape). Although these analyzers were easy to use, they were not generally agent specific and suffered from false-positives.1 More traditional laboratory methods have also been investigated, and in particular, combined gas chromatography and mass spectrometry (GC/MS) has been very successful at eliminating false-positives.10,11 However, GC/MS requires extraction, repeated calibration, and long analysis times (typically 20 to 60 minutes),11 making it labor intensive and less than desirable for field use. More rapid analysis of agents in the solid, liquid and gas phase has been demonstrated by vibrational spectroscopy.12-15 Hoffland et al.12 reported infrared absorbance spectra and absolute Raman cross sections for several chemical agents, while Christesen measured Raman cross sections for sarin, tabun, mustard gas, and VX (ethyl S-2-diisopropylamino ethyl methylphosphonothioate).16 Again, however these techniques also have limitations. Raman spectroscopy is simply not a very sensitive technique, and detection limits are typically 0.1% (1000 ppm). While infrared spectroscopy would have limited value in analyzing poisoned water, since the very strong infrared * To whom correspondence should be addressed, email:farqu@real-time-analyzers.com SPIE-4378-2001 21
- 3. absorption of water would obscure most other chemicals present. Nevertheless, efforts to overcome these limitations have been demonstrated. Braue and Pannella13 quantified the G-series nerve agents (tabun, sarin, and soman) in terms of infrared attenuated total reflectance using a circle-cell. And Alak and Vo-Dinh demonstrated the possibility of surface-enhanced Raman spectroscopy (SERS) to identify CWAs by measuring several organophosphonates that simulate the nerve agents.17 However, quantitative measurements have not been demonstrated for the SER-active material used (silver coated on alumina particles) or other SER-active media.18 Recently, we developed silver-doped sol-gels to promote the SER effect.19-22 The porous silica network of the sol-gel matrix offers a unique environment for stabilizing SER-active metal particles, and the sol-gel provides a high surface area that effectively increases the number of molecules observed within the Raman scattering volume. The silver-doped sol-gels have been coated on the internal walls of standard glass vials to allow rapid SER analysis of small sample quantities (< 0.1 mL) without preparation. We have used p-aminobenzoic acid (PABA) as a test chemical to demonstrate surface enhancements greater than 107, reversible measurements in a flowing system, reproducible measurements from vial-to-vial and batch-to- batch, and measurements in multiple solvents, including water.19-22 Here we present preliminary measurements of chemical agent simulants, in both the liquid and gas phases, as well as chemical agent hydrolysis products using our SER-active vials. 2. EXPERIMENTAL The chemical agent simulants employed were obtained at their purest commercially available grade from Aldrich (Milwaukee, WI) and were dissolved in water or methanol for analysis. All chemicals used to prepare the silver-doped sol- gels were spectroscopic grade and also purchased from Aldrich. The sol-gel vials were coated in a manner similar to that previously reported by adding ammonium hydroxide to a solution of silver nitrate, tetramethyl orthosilicate, and methanol.22 After mixing, 0.2 mL of the sol-gel solution was transferred into a glass vial (2 mL), dried and heated. The incorporated silver ions were then reduced using dilute sodium borohydride. The vials were washed and dried prior to the addition of a sample solution. The patent pending SER-active vials are commercially available from Real-Time Analyzers (Simple SERS Sample Vials, RTA, East Hartford, CT). Dimethyl metylphosphonate (DMMP), pinacolyl methylphosphonate (PMP) and methylphosphonic acid (MPA) were prepared in aqueous solution, while 2-chloroethyl ethyl sulfide (CEES) was prepared in methanol at 1 mM for SERS measurements. Neat samples were employed for normal Raman measurements. All samples were prepared in a chemical hood and transferred into plain or SER-active vials for analysis. Special precaution was followed for CEES, since it is a severe blistering agent.23 Once prepared, the vial was placed into the sample compartment of a Raman spectrometer for analysis. A prototype Fourier transform Raman spectrometer (RTA) was used for these measurements.24 The system consisted of a Nd:YAG laser (Brimrose) for excitation at 1064 nm, an interferometer built by On-Line Technologies (OLT, East Hartford, CT) for frequency separation, an uncooled InGaAs detector for signal detection (RTA), and an Intel 400 MHz Pentium II based laptop computer (Dell, Round Rock, TX) for interferometric control, data acquisition (OLT), and analysis (LabVIEW by National Instruments, Austin, TX). Additional components included a Notch filter (Kaiser, Ann Arbor, MI) and interferometer entrance and exit optics (Edmund Scientific, Barrington, NJ). Fiber optics were used to deliver the excitation beam to the sample and the scattered radiation to the interferometer (1 meter lengths of 200 and 365 micron core diameter, respectively, Spectran, Avon, CT). A second Notch filter (Kaiser) was used as a beam splitter to direct the excitation beam along the same axis as the collected radiation. A microscope object (20x0.4, Newport, Irvine, CA) was used to focus the beam into the sample and to collect the scattered radiation back along the same axis. In this co-axial backscattering arrangement, the excitation beam was passed through the outside of a glass vial and focused onto the silver- doped sol-gel film (0.1 mm thickness) containing the sample. 3. RESULTS AND DISCUSSION As a prelude to chemical agent measurements in water, we evaluated the quantitative performance of the SER-active vials by measuring PABA over the concentration range from 10-7 M to 10-2 M. Figure 1 shows the spectra for 7, 35, and 70 micromolar concentrations, while Figure 2 shows a plot of the 1450 cm-1 band intensity as a function of concentration. The SER response is linear over nearly three orders of magnitude to just over 10-4M, at which point the band intensity suggests that the silver surface is becoming saturated. SPIE-4378-2001 22
- 4. 2 10 1 10 A 0 10 B 10 -1 C -2 10 500 1000 1500 2000 -7 -6 -5 -4 -3 -2 -1 10 10 10 10 10 10 10 Wavenumbers (∆cm-1) Concentration (M) Figure 1. SER spectra of A) 70, B) 35, and C) 7 micromolar Figure 2. SER spectral intensity for p-aminobenzoic acid p-amino benzoic acid in water. Conditions: 80 mW of 1064 as a function of concentration using RTA SER-active vials. nm laser excitation, 100 averaged scans (1.5 min) at 8 cm-1 resolution. In an effort to demonstrate the broad capabilities of the SER-active vials to measure chemical agents, spectra of a nerve agent simulant: dimethyl methylphosphonate, a mustard gas simulant: 2-chloroethyl ethyl sulfide, and hydrolysis products: pinacolyl methylphosphonate and methylphosphonic acid were collected. DMMP is widely used by the U.S. Army as a chemical warfare simulant because its chemical structure, volatility, and water solubility are similar to those of nerve agents.25 DMMP is completely miscible and stable in water at room temperature.26 Figure 3 compares the SER spectrum to the normal Raman spectrum of DMMP. A number of the normal Raman bands are SER-active, such as the P-C stretching mode which shifts from 715 to 735 cm-1, and the C-H stretching modes at 2855, 2930, 2960, and 3000 cm-1, which shift slightly. Surprisingly, the P=O stretching mode at 1250 cm-1 virtually disappears. However, the most dramatic change is the appearance of an intense triplet in the SER spectrum near 1000 cm-1. The bands at 1000 cm-1, 1030 cm-1, and 1075 cm-1 likely involved the P-O-C bond. This is supported by the nearly identical triplets observed for the SER spectra of fonofos and fonofoxon.17,19 It is also worth noting that a band appears at 425 cm-1 in the SER spectrum, that may be unique to DMMP and useful for identification. The enhancement factor is estimated at 120,000 based on the normal Raman and SER P-C band intensity, taking into account the difference in sample concentrations and spectral acquisition conditions. A detection limit based on a signal-to-noise ratio of 3 can be estimated at 1.6 ppm. O Cl-CH2-CH2-S-CH2-CH3 = A CH3O-P-OCH3 _ CH3 A B B Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 3. A) SER and B) normal Raman spectra of Figure 4. A) SER and B) normal Raman spectra of 2- dimethyl methylphosphonate. Conditions: SERS as in chloroethyl ethyl sulfide. Conditions as in Figure 3. Figure 1, normal Raman, 500 mW and 200 scans. SPIE-4378-2001 23
- 5. 2-Chloroethyl ethyl sulfide, a blister agent simulant, has a chemical structure similar to the mustard gas (Cl-CH2-CH2-S-CH2- CH2-Cl), with only one terminal chlorine. Due to its low solubility in water, CEES was dissolved in methanol for the SER measurement. Again the prominent Raman modes are SER-active and even maintain relative intensity (Figure 4). The primary difference is that the SER bands appear to broaden, such that the triplet near 700 cm-1 becomes a doublet and the shoulders at 2875 and 2970 cm-1 become less defined. Again, the latter bands are assigned to C-H stretching modes. A single band at 700 cm-1, which is attributed to the C-S-C asymmetric stretch, dominates the reported infrared spectrum of mustard gas.12 A corresponding symmetric stretch is reported at 705 cm-1 in the Raman spectrum of mustard gas.27 Here a corresponding symmetric stretch appears, but as a doublet at 700 and 755 cm-1, presumably due to the loss in symmetry for CEES. The band at 655cm-1 can also be confidently assigned to a C-Cl stretch. The SER spectral bands at 620 and 730 cm-1 are probably due to the same modes, i.e. C-Cl and C-S-C stretches, respectively. The enhancement factor for CEES was somewhat less than DMMP at approximately 62,000, as is the estimated detection limit of 2.2 ppm. The ability to rapidly detect trace quantities of chemical agents in the gas phase would be invaluable as an early warning system. Although the Raman scattering cross-sections for the nerve agents suggest that remote detection by Raman-based LIDAR is unlikely,16 a SER-based system for perimeter monitoring could prove successful. As a preliminary measurement, we prepared a 10% by volume solution of CEES in methanol, exposed a SER-active vial to the equilibrium vapor phase in a sealed jar, and monitored the SER spectrum as a function of time. Initially, the vial was removed through a transfer chamber every hour to record the SER spectrum. After ten hours, spectra were recorded only every ten hours. As illustrated by Figure 5, the sol-gel performed as a dosimeter, in that the spectra increased as a function of exposure time. The most intense SER bands at 620 and 2930 cm-1 are discernable in the first few hours. The spectrum after 40 hours is nearly identical to the solution phase spectrum, except for a diminished intensity of the 730 cm-1 band. This may be due to methanol solvation effects or surface-orientation effects. Based on the relative concentrations of methanol and CEES and their partial pressures, we estimate the equilibrium concentration of CEES to between 1 and 2 micromolar. Although not shown, this concentration could be detected in one hour. O = A HO-P-OH _ CH3 O = CH3 _ B HO-P-O-CH-C-CH3 _ _ CH3 CH3 CH3_ Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 5. SER spectra of 2-chloroethyl ethyl sulfide Figure 6. SER spectra of A) methyl phosphonic acid and vapor as a function of time (10 hour increments to top, B) pinacolyl methylphosphanate (note unique band at which is 40 hours). Bottom trace is a blank. Spectral 546 cm-1). Spectral conditions as in Figure 1. conditions as in Figure 1. As previously stated, the analysis of chemical agents in water is important in identifying poisoned water. It is also important to decommissioning activities, in which agents are destroyed by hydrolysis (acid or base). Furthermore, any analytical technique used must be capable of distinguishing between parent CWA and hydrolysis products to assess safety or effectiveness of decommissioning. For example, soman has a hydrolysis half-life of ~2.3 hours at ambient temperatures and neutral pH,28 and forms hydrofluoric acid (somewhat toxic) and pinacolyl methylphosphonate (relatively non-toxic).29,30 PMP further hydrolyzes to form methyl phosphonic acid and 3,3-dimethyl-2-butanol (both non-toxic). The structural similarities between soman, PMP and MPA are expected to produce similar Raman, as well as SER spectra. Figure 6 compares PMP and MPA, but not the highly toxic parent CWA soman. As with DMMP, the P-C stretch, the P-O-C mode, and C-H stretches are readily apparent. Yet it is worth noting that the band positions are reasonably different. The former two bands appear at 764 and 1042 cm-1 for MPA, while they are at 788 and 1032 cm-1 for PMP. More importantly, a unique band at 546 cm-1, as yet unassigned, appears in the PMP spectrum. SPIE-4378-2001 24
- 6. 4. CONCLUSIONS We have successfully measured the SER spectra of chemical agent simulants: dimethyl metylphosphonate and 2-chloroethyl ethyl sulfide, and chemical agent hydrolysis products: pinacolyl methylphosphonate and methylphosphonic acid, using silver- doped sol-gel coated sample vials. Measurements were obtained in both aqueous and gas phase. The P-C stretching mode was SER-active for all four chemicals, allowing identification by class. Within this group, each chemical contained at least one unique spectral band that could be used for identification (Table 1). Furthermore, these bands do not appear to coincide with SER spectra reported for organophosphorus pesticides, the most likely source of false-positives. Although surface enhancement factors appear to be an order of magnitude better than those previously presented in the literature for similar chemicals,17 measurement sensitivity needs to be improved by 1 to 2 orders of magnitude to provide adequate warning of chemical agent use. Current research efforts to increase surface-enhancement, optical collection efficiency, and instrument design are being pursued to achieve the required sensitivity. Table 1. Enhancement factors, detection limits and unique SER bands fro chemicals studied. Agent Simulant Enhancement Detection limit Unique bands (cm-1) Dimethyl methylphosphonate 120,000 90 µM (1.6 ppm) 425 2-Chloroethyl ethyl sulfide 62,000 60 µM (2.2 ppm) 620 Methylphosphonic acid 110,000 3 µM (60 ppb) 764, 1042 Pinacolyl methylphosphonate 150,000 70 µM (1.4 ppm) 546, 788, 1032 5. ACKNOWLEDGEMENTS The authors would like to thank Drs. Janet Jensen and Steven Christesen of Aberdeen Proving Ground for encouraging this work. They would also like to thank Advanced Fuel Research for making their laboratory facilities available. 6. REFERENCES 1 “The Chemical Weapons Convention – A Guided Tour, the Organization for the Prohibition of Chemical Weapons” at http://www.opcw.nl/guide.htm. 2 Robinson, J.P. and J. Goldblat, "Chemical Warfare In The Iraq-Iran War" Stockholm International Peace Research Institute Fact Sheet, at http://projects.sipri.se/cbw/research/factsheet-1984.html (1984) 3 “Chemistry of GB (Sarin)” at http://www.mitretek.org/mission/envene/chemical/agents/sarin.html. 4 Tu, Anthony, “Overview of Sarin Terrorist Incidents in Japan in 1994 and 1995”, 6th CBW Protection Symposium, Stockholm, Sweden, 10-15 May 1998. 5 Staff Reporter, “Going out with a bang”, Newsweek, June 28, 1999. 6 See UNSCOM reports in http://www.un.org/depts/unscom (1999). 7 Treven, T., Saddam’s Secrets, Harper Collins (1999) 8 “Decaying Sarin-filled Rockets Spark Fears”, Jane’s Defense Weekly, 25(20),3 (1996). 9 “The Chemical Weapons Convention Redefines Analytical Challenge”, Analytical Chemistry News & Features, June 1 397A (1998). 10 Johnston, R.L., Hoefler, C.M., Fargo, J.C., and Moberley, B., “The Defense Nuclear Agency’s Chemical/Biochemical Weapons Agreements Verification Technology Research, Development, Test and Evaluation Program and its Requirements for On-Site Analysis”, AT-ONSITE, 5-8 (1994) 11 Black, R.M., Clarke, R.J., Read, R.W., and Reid, M.T., “Application of gas chromatography-mass spectrometry and gas chromatography-tandem mass spectrometry to the analysis of chemical warfare samples, found to contain residues of the nerve agent sarin, sulphur mustard and their degradation products”, J. Chromatography, 662, 301-321 (1994) 12 Hoffland, L.D., Piffath, R.J., Bouck, J.B.,”Spectral signatures of chemical agents and simulants”, Optical Engineering, 24, 982-984, (1985) 13 Braue, E.H.J., Pannella, M.G.,”CIRCLE CELL FT-IR Analysis of Chemical Warfare Agents in Aqueous Solutions”, Applied Spectroscopy, 44, 1513-1520, (1990) SPIE-4378-2001 25
- 7. 14 Tseng, C.-H., Mann C.K., and Vickers T.J., “Determination of Organics on Metal Surfaces by Raman Spectroscopy”, Applied Spectroscopy, 47, 1767-1771 (1993) 15 Staff Reporter, “US Army Tests Lidar to Detect Biological Toxins”, Photonic Spectra, pg. 50, December 1998. 16 Christesen, S.D., "Raman cross sections of chemical agents and simulants", Applied Spectroscopy, 42, 318-321 (1988) 17 Alak, A.M. and Vo-Dinh, T., “SERS of Organophosphorous Chemical Agents”, Analytical Chemistry, 59, 2149-2153 (1987) 18 Norrod, K.L., Sudnik, L.M., Rousell, D., and Rowlen, K.L., “Quantitative Comparison of Five SERS Substrates: Sensitivity and Detection Limit”, Applied Spectroscopy, 51, 994-1001 (1997). 19 Lee, Y. and Farquharson, S., “SERS Sample Vials Based on Sol-Gel Process for Trace Pesticide Analysis”, SPIE, 4206, 140-146 (2000). 20 Farquharson, S. and Lee, Y., “Trace Drug Analysis by Surface-Enhanced Raman Spectroscopy”, SPIE, 4200-16 (2000). 21 Lee, Y., Farquharson, S., and Rainey, P. M., "Surface-Enhanced Raman Sensor for Trace Chemical Detection in Water", SPIE, 3857, 76-84 (1999). 22 Lee, Y, Farquharson, S., Kwong, H., and Shahriari, M., “Surface-Enhanced Raman Sensor for Surface-Enhanced Raman Spectroscopy”, SPIE, 3537, 252-260 (1998). 23 see Material Safety Data Sheets for details. 24 Farquharson, S., Smith, W., Carangelo, R. C., and Brouillette, C., “Industrial Raman: Providing Easy, Immediate, Cost Effective Chemical Analysis Anywhere”, SPIE, 3859, 14-23 (1999) 25 Bennett, S., Bane, J., Benford, P., and Pratt, R., “Environmental Hazards of Chemical Agent Simulants”, Aberdeen Proving Ground, Maryland: Chemical Research and Development Center, CRDC-TR-84055 (1984). 26 Mabey, W. and Mill, T., Critical Review of Hydrolysis of Organic Compounds in Water under Environmental Conditions. Journal of Physics and Chemistry Reference Data, 7(2): 383-414 (1978). 27 Christesen, S., MacIver, B., Procell, L, Sorrick, D., Carabba, M, and Bello, J., “ Noninstrusive Analysis of Chemical Agent Identification Sets Using a Portable Fiber-Optic Raman Spectrometer”, Applied Spectroscopy, 53, 850-855 (1999). 28 Meylan, W.M. and Howard, P.H., J. Pharm. Sci., 84, 83-92 (1995) 29 Jenkins, A., Uy, O. and Murray, G., “Polymer-Based Lanthanide Luminescent Sensor for Detection of Hydrolysis Product of the Nerve Agent Soman in Water”, Analytical Chemistry, 71, 373-378 (1999). 30 Nassar, A., Lucas, S., and Hoffland, L., “Determination of Chemical Warfare Agent Degradation Products at Low-Part- per-Billion Levels in Aqueous Samples and Sub-Part-per-Million Levels in Soils Using Capillary Electrophoresis”, Analytical Chemistry, 71, 1285-1292 (1999). SPIE-4378-2001 26
- 8. Appendix B Detection of bioagent signatures: A comparison of electrolytic and metal- doped sol-gel surface-enhanced Raman media Stuart Farquharson,* Wayne Smith, and Yuan Lee Real-Time Analyzers, 87 Church Street, East Hartford, CT 06108 Susan Elliott and Jay F. Sperry University of Rhode Island, 45 Lower College Rd, Kingston, RI 02881 ABSTRACT Since September 11, 2001, the threat of terrorist attacks and biological warfare within U.S. borders has become a sobering reality. In an effort to aid military personnel and the public at large, we have been investigating the utility of surface- enhanced Raman spectroscopy (SERS) to provide rapid identification of chemical agents directly, and biological agents through their chemical signatures. This approach is based on the ability of Raman spectroscopy to identify molecular structure through the abundant vibration information provided in spectra and the ability of SERS to detect extremely low concentrations (e.g. part-per-billion) through the enhancement of Raman scattering by six orders of magnitude or more. Towards the goal of developing a portable analyzer, we have been studying the ability of two SER media to obtain continuous (i.e., reversible) and quantitative (i.e., reproducible) measurements. Here we compare measurements of nucleic acid-bases, adenosine monophosphate, and ribonucleic acid extracted from Escherichia coli, Bacillus subtilis and Staphylococcus aureus obtained by electrolytic SERS and metal-doped sol-gel SERS. The capabilities of these SER media are summarized in terms of rapid detection of B. anthracis and dipicolinic acid. Keywords: bioagent detection, SERS, RNA analysis, bacterial analysis, Raman spectroscopy 1. INTRODUCTION The recent distribution of anthrax through the U.S. postal system and the subsequent infection and death of several postal and national media employees, amplifies the need for methods to rapidly detect and identify this and other chemical and biological warfare agents (BWA). The primary methods currently used, immunoassays for screening and nucleic acid (NA) sequencing for positive identification of BWAs (bacteria, protozoa and viruses), have serious limitations.1,2,3 Immunoassay methods employ competitive binding of the bioagent (as an antigen) and its labeled (e.g. fluorescence) conjugate for a limited number of antibodies. Although this analysis method is fast and semi-quantitative, other chemicals may compete for the antibodies, interfere with the enzymatic reaction or interfere with the measurement (e.g. it fluoresces) resulting in a high number of false positive responses.1 Furthermore, the antibodies denature due to moisture and heat, limiting shelf life, and require sterile, often refrigerated storage. Positive identification of a BWA can be accomplished by sequencing deoxyribonucleic acid or ribonucleic acid (DNA and RNA).2,3 This requires enumeration of a nucleic acid sequence through polymerase chain reactions (PCR) or multiplication of the microorganism through culture growth to provide sufficient quantities of DNA or RNA for analysis. Unfortunately, PCR and culture growth require from several hours to several days.2,3 Consequently, a wide variety of technologies have been investigated for rapid identification of BWAs. The Department of Defense is actively monitoring 200 such technologies.4 This includes traditional methods, such as gas chromatographic separation coupled with ion mobility spectrometry detection,5 to exotic methods based on nature, such as monitoring toxin induced color changes in fish scales.6 Although all of these techniques have achieved varying degrees of success, none are yet capable of detecting and identifying BWAs in 10 minutes or less. Towards this goal we have been investigating the ability of SERS to detect sub-nanogram quantities of DNA or RNA (eliminating enumeration), determine relative NA base concentrations, and identify BWA taxonomy. * To whom correspondences should be addresses, e-mail:farqu@real-time-analyzers.com, www.real-time-analyzers.com SPIE 2001-4575 62
- 9. Raman spectroscopy has a rich history of investigating biochemical and biological processes.7 Some of the earliest laser- Raman studies demonstrated that the five NA bases, adenine (A), cytosine (C), guanine (G), thymine (T, in DNA) and uracil (U, in RNA), yielded distinct spectra with several bands suitable for identification and quantification.8 Furthermore, these studies included exceptional spectra of both DNA and RNA, for which the NA bases, as well as several phosphate bands were easily identified.9 However, since the Raman effect is very inefficient (very low conversion of incident radiation to inelastically scattered Raman radiation), these samples had to be highly concentrated. Fortunately, two phenomena exist that can increase the generation of Raman photons by six orders of magnitude or more, known as the resonance Raman and surface-enhanced Raman effects.10,11 Resonance Raman scattering occurs when the laser excitation wavelength is in resonance with an electronic transition of a molecule (within the absorption envelope).10 Excitation at ultraviolet wavelengths has been used to obtain resonance Raman spectra of amino acids and nucleic acids in whole bacteria.12,13 For example, excitation at 242 nm has been used to maximize the nucleic acid spectral band intensities, and minimize the amino acids band intensities. A peak at 1530 cm-1 was found to be proportional to the amount of the NA base cytosine, while a peak at 1485 cm-1 was proportional to the combined amount of the NA bases adenine and guanine. This quantitative behavior has been used to define an A+T/G+C base-pair ratio and provide a level of bacterial identification as taxonomic markers.13 In recent years SERS has also been used to analyze bacterial cell components,14 including amino acids,15 lipids,16 nucleic acids,15,17,18 and the adenine derivatives.19,20,21 SERS has proven to be one of the most sensitive methods for trace chemical analysis through the detection of single molecules,22,23 including DNA (dye labeled 17-mer).24 Since its discovery in 1974,25 the mechanism responsible for the large increase in scattering efficiency has been the subject of considerable research.26,27 Briefly, incident laser photons couple to free conducting electrons within a metal, which confined by the particle surface, collectively cause the electron cloud to resonate.26,28 These surface plasmons are known as the physical component of the SER effect. These surface plasmons can transfer energy to the molecular vibrational modes of molecules through interactions with the molecular electron orbitals.26,29 This interaction is known as the chemical component of the SER effect. This perturbation of the molecular polarizability generates surface-enhanced Raman photons.26 A number of methods have been developed to produce surfaces or solutions containing one of these metals with optimum roughness or diameter to promote SERS.30 These methods include preparation of activated electrodes in electrolytic cells, 11,31 activated silver and gold colloid reagents,32 and metal coated substrates.33,34,35 Selecting a SER-active medium for chemical and biological agent detection requires consideration of the method of deployment, and hence the method of sampling. Chemical aerosols or airborne bacteria will require a collection device to concentrate and continuously present the sample to the SERS medium. Poisoned water supplies will also require a flow through device for continuous monitoring, or a grab-sample device for periodic analysis. And contaminated surfaces will require a grab-sample extractive device. A SERS-based device used for continuous monitoring (air or water) must be reversible and reproducible if quantitative measurements are desired, while a SERS-based device used for periodic sampling (water or surfaces) must be reproducible. Both reversible and reproducible measurements have been performed using electrolytic SERS (E-SERS).36 But this requires a three-electrode sample cell and an electrolyte of known concentration to perform the necessary oxidation-reduction cycles (ORCs) to re-activate the electrode surface with new, uncontaminated sites from one measurement to the next. Colloids are severely limited, in that continuous measurements would require a continuous supply of colloids. For periodic measurements, vials of colloids, one per measurement, could be used. However, aggregate size and consequently SER intensity change with sample conditions (especially pH), and quantitative, reproducible measurements are unlikely. Substrates appear to have the greatest potential, and designs range from silver evaporated on titania particles34 to periodic gold pyramids evaporated between polystyrene beads.35 Most substrates require concentrating the sample on the surface through drying to obtain the largest signal enhancements, in effect making the measurements irreproducible and irreversible. However, successful measurements using flow systems have been obtained with glass posts, but manufacturing costs appear prohibitive. In an effort to overcome these limitations, we have developed metal-doped sol-gels to provide SERS measurements that are reproducible, reversible, and quantitative, and yet not restricted to specific environments, such as electrolytes, solvents, or evaporated surfaces.37,38 The porous silica network of the sol-gel offers a unique environment for stabilizing SER active metal particles, and the high surface area increases the interaction between the analyte and metal particles. The sol-gel can be coated on the end of fiber optics or on the internal walls of a glass flow tube for continuous measurements or standard glass sample vials for periodic measurements. Previously we measured 100 mg/L methylphosphonic acid (the primary hydrolysis product of nerve agents) in water with an estimated detection limit of 0.5 mg/L (100 parts-per-billion). We have also SPIE 2001-4575 63
- 10. demonstrated reversible and reproducible measurements of p-aminobenzoic acid (PABA) in a flow through system. Here we investigate the ability of the sol-gel SERS (SG-SERS) to measure the NA bases, adenosine monophosphate, and RNA extracted from E. coli, B. subtilis and S. aureus. The measurements are compared to those obtained by E-SERS. 2. EXPERIMENTAL The inorganic chemicals and solvents used to prepare samples were spectroscopic grade and purchased from Aldrich (Milwaukee, WI), Fisher (Pittsburgh, PA) or Pfaltz & Bauer (Waterbury, CT). The nucleic acid bases and dipicolinic acid were purchased from Sigma (St. Louis, MO). Normal Raman samples were measured to establish enhancement factors. In each case 1cm3 of sample was placed into a 1x1 cm glass cuvette weighed and measured. Unpacked densities were typically 6-7 g/cm3. For all SER measurements, including RNA, samples were prepared as ~0.1mg/mL (see Figure captions for exact concentrations) in 0.1M KCl and buffered to a pH of 9.2 with Na2B4O7•H2O. Adenine pH dependence measurements used pH buffer solutions at 4 (potassium acid phthalate), 6.9 (potassium phosphate monobasic/sodium phosphate dibasic), 9.2, (Na2B4O7•H2O) and 10.4 (tris-hydroxymethyl amino methane). Escherichia coli, Bacillus subtilis and Staphylococcus aureus cultures (250ml per 1000mL Erlenmeyer flask) were grown overnight in a Trypticase soy broth (TSB) medium containing 1% glucose in a shaking water bath at 37 oC. The bacteria were harvested by centrifugation for 10 minutes at 8,000 rpm in a GSA rotor at 5°C, then washed once in 0.85% saline. The gram-positive bacteria were concentrated to 20 ml and passed through a French pressure cell twice at 15,000 psi to break open the cells. RNA was extracted according to Protocol 4.41,39 to ensure pristine samples for initial measurements. Since this method takes approximately 4 hours, a streamlined method was developed. For vegetative bacteria, the specimen was boiled for 30 sec in 1 ml of distilled water to lyse the cells and release the RNA. For bacterial spores, the specimen was first incubated in 1 ml of saline solution containing 0.2 mg lysozyme and phosphate-buffered to pH of 6.24 for 1 hr at 37 oC. This solution was then boiled for 2-3 minutes in 4X loading buffer to release the RNA. For both specimens, RNA STAT-60TM was added to the supernatant, which was centrifuged at 12,000 g for 5 minutes to precipitate the ~15% water-soluble proteins. This procedure allowed extracting RNA for SER analysis in ~ 10 minutes. Electrophoresis shows high purity, while the existence of chemicals that could interfere with the SER measurements is still under investigation. The electrolytic sample cell has been described previously.36 Briefly, a three electrode design is incorporated into a Plexiglas structure containing a well for the reference electrode (a saturated calomel electrode, Cole Parmer, Vernon Hills, IL) and a 5mL sample well containing the silver working electrode and platinum wire counter electrode (0.5 mm wire, Alfa, Ward Hill, MA). A channel connecting the two wells contained a 2 mm diameter semi-porous membrane (10-20 micron pore, Ace Glass). The silver electrode was made from a 3 mm length of 2 mm diameter silver wire (Alfa) soldered to a copper wire lead encased in a 4 mm diameter Pyrex tube. A cap containing the silver electrode, platinum wire, and nitrogen purge and vent lines fixed the silver electrode surface 1 mm from a 1 mm thick glass plate attached to the bottom of the sample well. The potentiostat used to control the three electrodes was built in-house and has been described in detail elsewhere.36 A multifuntional analog, digital, and timing input/output interface card (DAQCard-1200, National Instruments) is used to both drive the electrolytic cell as well as read the current generated in the cell. A LabVIEW software program is used set the oxidation potential, reduction potential, SER measurement potential, hold times, and sweep rates. The amount of charge passed was plotted as a cyclic voltammogram. For all spectra presented, five oxidation-reduction cycles (ORCs), stepping from -0.3 VSCE to 0.3 VSCE and back to -3 VSCE at 50 mV/sec were used. The SG-SER measurements were accomplished by simply placing the identical samples prepared above into Simple SERS Sample VialsTM (RTA). These 2-mL, glass vials are internally coated with ~ 0.1 micron thick silver-doped sol-gel. A prototype Fourier transform Raman spectrometer (RTA) was used for these measurements.40 The system consisted of a Nd:YAG laser (Brimrose or Spectra Physics) for excitation at 1064 nm, an interferometer built by On-Line Technologies (OLT, East Hartford, CT) for frequency separation, an uncooled InGaAs detector for signal detection (RTA), and an Intel 400 MHz Pentium II based laptop computer (Dell, Round Rock, TX) for interferometric control, data acquisition (OLT), and analysis (LabVIEW by National Instruments, Austin, TX). Additional components included a Notch filter (Kaiser, Ann Arbor, MI) and interferometer entrance and exit optics (Edmund Scientific, Barrington, NJ). Fiber optics were used to deliver the excitation beam to the sample and the scattered radiation to the interferometer (2 meter lengths of 200 and 365 micron core diameter, respectively, Spectran, Avon, CT). A second Notch filter (Kaiser) was used as a beam splitter to direct the excitation beam along the same axis as the collected radiation. A microscope object (20x magnification, 0.4 numeric aperture, Newport, Irvine, CA) was used to focus the beam into the sample and to collect the scattered radiation back along the same axis. In this co-axial backscattering arrangement, the excitation beam passed through the glass plate onto the silver SPIE 2001-4575 64
- 11. electrode surface for E-SERS, through the vial glass wall and into the silver-doped sol-gel film for SG-SERS, or through the glass wall of the cuvette and into the solid sample for normal Raman spectroscopy. All E-SERS and normal Raman spectra were obtained with 750 mW of laser power at the sample, while all SG-SERS spectra were obtained with 75 mW of laser power at the system. Incident powers above 200 mW in some cases degraded the sol-gel. 3. RESULTS AND DISCUSSION The generation of surface-enhanced Raman scattering at electrode surfaces has been extensively researched, and the optimum sample conditions are well developed.27,29 Several researches incorporated electrodes into flowing systems and demonstrated that continuous monitoring of chemicals is possible.18 These successes suggested investigated the capability of E-SERS to measure the NA bases and RNA. The E-SERS measurements also provided a benchmark to compare and evaluate SG-SERS measurements. The molecular structure of adenine (as well as the other base pairs), which includes an aromatic nitrogen- containing heterocycle, is ideally suited to interact with the surface plasmons and contribute substantially to the chemical component of the SER effect.11,19 Even with excitation at 1064 nm, a 3-minute scan of 1.8x10-5M adenine yields high signal- to-noise (S/N) E-SER spectra and all of the bands are revealed with clarity (Figure 1, Table 1). Spectra of similar quality were obtained by SG-SERS and the principal spectral bands are easily observed. The identical 1.8x10-5M adenine sample was measured in the same 3-minute time frame, but with 1/10th the laser power. The lower power appears to reduce the S/N. 725 A A 735 pH 10 B B SG-SERS C 735 C D pH 4 Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 1. A) Normal Raman spectrum of pure adenine Figure 2. A) and C) E-SERS and B) and D) SG-SERS of powder, B) E-SERS and C) SG-SERS of 1.8x10-5M adenine at A) and B) pH 10.4 and C) and D) pH 4.0. Note adenine at pH 9.2. All spectra 8 cm-1 resolution, 200 scans consistent appearance of bands at 1270 and 1375 cm-1 as (3 min), and 1064 nm excitation. A) and B): 750 mW, C) the pH is changed to 10 for both SER media. E-SERS 75 mW. B) measurement potential of 1.1VSCE. used 750 mW, SG-SERS used 75 mW of 1064 nm excitation. The amount of adenine responsible for the SER spectra, as well as enhancement factors for the two SER media can be determined. The molecules producing the E-SERS spectrum are those on the electrode surface within the illumination area of the laser. (The solution concentration only determines the number of molecules available to adsorb to the electrode surface.) For the current experiments the laser illuminates an area of 2.8x10-7m2, or 5.6 x10-7m2 if we assume the ORCs increase the surface area by a factor of two. Furthermore, if we assume monolayer coverage on the electrode and each 3x5 angstrom molecule (lying flat) occupies 1.5x10-19m2, then there are ca. 4x1012 molecules contributing to the Raman scattering. This is ca. twice the number of molecules measured at electrode surfaces using either differential capacitance- potential curve measurements or rapid linear sweep voltammetry (e.g. 3x1018 molecules/m2 for pyridine and pyrazine).29 Thus the adenine spectrum in Figure 1 is due to 8.7x10-10g (6x10-12 moles)! A detection limit defined as a S/N of 3 can also be calculated. The S/N for a 3-minute scan is 844 for the 735 cm-1 band, suggesting a mass detection limit of 3x10-12g (2x10-14 moles). This is consistent with previous estimates for adenine by others of 2.5 x10-14 moles.15,30 However, sub- monolayer concentrations must be measured to verify this. The root-mean-squared (RMS) noise is measured between 4400- 4600 cm-1. Since noise is distributed evenly throughout the spectrum when transformed, this region does not have any SPIE 2001-4575 65
- 12. contributions from signals or baseline offsets. The measurement error is given as S±RMS, and for adenine this equals 2.34%. The number of molecules contributing to the SG-SERS are those on the silver particles that are embedded in the sol-gel. The total silver surface area can be determined from the average particle size (40 nm diameter), concentration (0.73% by weight, based on molar conc. and measured sol-gel density), and the scattering volume (a cylinder defined by the laser area: 2.8x10-7m2 and sol-gel thickness:10-4m). The 6.1x109 silver particles in this volume have a collective area of 3.1x10-5m2. However, it may reasonably be assumed that at least half of the silver surface is in contact with the sol-gel matrix and unavailable for analyte interaction. Then approximately 1.0x1014 molecules or 2.2x10-8g of adenine contribute to the SG- SER spectrum. The slightly lower S/N of 207 suggests a mass detection limit of 3.2x10-10g. Determination of the enhancement factors for the two SER media requires estimating the number of adenine molecules contributing to the normal Raman spectrum. Here a cylindrical scattering volume is assumed, again based on the laser area (2.8x10-7m2) and the penetration depth (1x10-3 m).41 The density of the sample was measured at 0.64 g/cm3, indicating that 1.8x10-4g (1.3x10-6 moles) of adenine produced the normal Raman signal. The enhancement factor, EF, is defined by the following equation: EF = (ISERS/INR)•(MNR/MSERS) •(PNR/PSERS) •(TNR/TSERS)1/2 where I is the spectral band intensity (here 735 cm-1), M is the sample mass, P is the incident laser power, and T is the measurement time (or number of scans). For the E-SERS measurement the enhancement factor is 2.2x105 (0.178/0.184) • (1.8x10-4/8.7x10-10)), while the SG-SERS enhancement factor is 1.0x105 (0.16/0.184) •(1.8x10-4/2.2x10-8) •(750/75) •(3/1.5)1/2). The lower enhancement for the SG-SERS may be real, or the available surface of the silver embedded in the sol-gel may have been overestimated. In addition to enhancing the Raman scattering efficiency to an extent similar to E-SERS, the SG-SER medium also yields an identical shift of the adenine ring-breathing mode from 725 cm-1 in the normal Raman to 735 cm-1. Furthermore, in the course of optimizing the E-SERS sample conditions, it was found that pH influenced the adenine interaction with the silver surface (Figure 2). In particular, the relative band intensities of the pyrimidine ring skeletal vibrations at 1270 and 1375 cm-1, and the imidazol ring skeletal vibration at1335 cm-1 change. At pH 4 adenine is protonated, presumably the imidazol ring, since the band at 1335 cm-1 increases in intensity, while the pyrimidine bands are virtually absent. Conversely at pH 10, the imidazol band decreases in intensity, while the pyrimidine bands appear. It is worth noting that the ring-breathing mode at 735 cm-1 changes little between pH 4 and 10, suggesting that the skeletal changes are more a function of molecule-plasmon interactions than reorientation of the molecule on the surface. Measurements of the identical pH series of adenine samples by SG-SERS yielded virtually identical spectral changes. This suggests that the sol-gel does not influence the measurement. This is critical to reproducing measurements and performing quantitative analysis. Next, the remaining NA bases were measured by both E-SERS and SG-SERS and compared. Previously we examined the optimum pH and electrode potentials for E-SERS measurements to determine if a common pH could be used that yielded good sensitivity for all the bases, and if variations in potential could be used to provide an added degree of selectivity between the bases. Primarily it was found that high quality spectra were obtained between pH 7 and 9.5, and that cytosine and uracil were best enhanced at potentials positive of the potential-of-zero charge (pzc, ca. 0.65VSCE for Ag), guanine and thymine near the pzc, and adenine negative of the pzc. In all cases the ring-breathing modes were the most intense and in general could be used to identify the NA bases (Figure 3, Table 1). Specifically, adenine has an intense band at 735 cm-1, cytosine at 797 cm-1, guanine at 653 cm-1, thymine at 784 cm-1, and uracil at 800 cm-1. The adenine, cytosine, guanine and thymine bands are sufficiently separated that their contributions to DNA should be determinable. Although adenine and guanine contributions to RNA should also be determinable, cytosine and uracil are highly overlapped, and unfortunately share the same potential dependence. Alternate, unique bands at 1183 cm-1 for cytosine and 1275 cm-1 for uracil might be suitable for calculating contributions. The SG-SER spectra of the remaining NA bases faithfully reproduced the E-SER spectra. In particular the primary identifying bands occur at virtually the same wavenumbers (see Table 1). However, the spectra for both cytosine and thymine contain an intense band at ca. 1040 cm-1. Initially this was attributed to the pH buffer, but samples prepared without either the buffer or the 0.1M KCl electrolyte yielded identical spectra containing this band. In fact, the E-SER and SG-SER spectra of thymine are virtually identical except for this band. Also, the SG-SERS of guanine contains an intense band at 1551 cm-1 that is not observed in the E-SER spectrum. This band may be due to a moderately intense band at 1553cm-1 in the normal Raman spectrum that is SG-SER active. It was also found that the SG-SERS of cytosine was considerably better than the E-SERS, while uracil showed the opposite relationship. It is also worth noting that all of the SG-SERS were obtained with 1/10th the laser power. Most importantly, the primary ring-breathing modes in the SG-SER spectra are sufficiently intense and unique to be used in determining contributions to DNA and RNA as outlined above. SPIE 2001-4575 66
- 13. A A B B C C D D Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 3. E-SERS of A) 2.1x10-3M cytosine at -0.3VSCE, Figure 4. SG-SERS of A) 2.1x10-3M cytosine, 200 scans, 1000 scans, B) ~1.0x10-5M guanine at -0.6VSCE, 500 B) ~1.0x10-5M guanine, 200 scans, C) 2.3x10-3M scans, C) 2.3x10-3M thymine at -0.6VSCE, 500 scans and thymine, 200 scans and D) 1.2x10-3M uracil, 500 scans. D) 1.2x10-3M uracil at -0.93VSCE, 500 scans. All spectra: All spectra: at pH 9.2, 75 mW 1064 nm at 8 cm-1. at pH 9.2, 750 mW 1064 nm at 8 cm-1. Table 1. Comparison of E-SER and SG-SER Spectral Band Positions for the NA Bases and Adenosine Monophosphate. Adenine Cytosine Guanine Thymine Uracil AMP E-SERS SG-SERS E-SERS SG-SERS E-SERS SG-SERS E-SERS SG-SERS E-SERS SG-SERS E-SERS SG-SERS 1647 1638 1634 1656 1655 1630 1587 1585 1510 1580* 1551* 1539* 1456 1456 1465 1460 1462 1480 1453 1459 1394 1398 1425 1431 1435 1399 1404 1392 1374 1375 1373 1383 1370 1335 1332 1311 1307 1333 1331 1353 1348 1331 1329 1265 1273 1280 1292 1278 1276 1275 1279 1271 1183 1195 1222 1232 1221 1219 1204 1205 1180* 1144 1097 1033 1029 1038 1040* 1035* 1051 1037* 1041 1035 963 963 957 1001 1000 961 944 884 819 817 859 866 735 737 797 799 784 782 800 800 727,38 742 630 630 653 664 667 684* 603 602 590 611 466 561 * Bands unique to SG-SERS. The next chemical to be analyzed by both E-SERS and SG-SERS was adenosine monophosphate (AMP). The E-SER spectrum yields bands due to the adenine chemical functionality at 727, 961, 1233, 1279, 1331, 1381 and 1486 cm-1. In addition, phosphate bands are observed at 860, 1097, 1453, 1587, and 1705 cm-1 (Figure 5). Other researchers have noted that the ribose component does not appear to contribute to the spectrum.19 The AMP spectrum also changes as a function of potential. As the electrode is swept more positive (here from -0.9 to -0.3VSCE) the phosphate bands at 860, 1097, 1453, and 1587 cm-1 increase in intensity compared to the adenine bands, while a band at 1705 cm-1 appears. The adenine bands at 1233, 1381 and 1486 cm-1 virtually disappear. These potential dependent spectral changes are consistent with earlier studies that show that phosphate is attracted to silver at potentials positive of the pzc, but repelled at potentials negative of the pzc.19 SPIE 2001-4575 67
- 14. The SG-SER spectrum of AMP is considerably different. The adenine bands virtually disappear, except for the two primary bands, which shift to742 and 1329 cm-1. While the phosphate band at 1459 cm-1 has gained considerable intensity. In addition two new intense bands appear at 684 and 1539 cm-1, as well as a moderately intense band at 1180 cm-1. The SG- SER spectrum has greater similarity to the E-SER spectrum at -0.3VSCE, and suggests that the silver particles embedded in the sol-gel behave as if at a potential positive of the pzc. AMP RNA A E. coli B B. subtilis S. aureus C Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 5. E-SER spectra of 0.20 mg/mL adenosine Figure 6. E-SERS of 0.1 mg/mL RNA from E. coli, 0.2 monophosphate at A) -0.3 and B) -0.9VSCE, and C) SG-SER mg/mL RNA from B. subtilis and 0.2 mg/mL RNA from spectra. Conditions: sample in 0.1M KCl buffered to pH 9.2, A) S. aureus. Conditions: 0.1M KCl, pH 9.2 -0.3VSCE, 750 and B) 750 mW, C) 75 mW of 1064, 64 scans (1-min) at 8 cm-1. mW of 1064 nm, 640 scans (10 min) at 8 cm-1. RNA samples extracted from E. coli, B. subtilis and S. aureus were next examined by both E-SERS and SG-SERS. E-SER spectra of these samples yielded quality spectra in 10 minutes, in which all of the major features can be identified (Figures 6 and 7). This includes guanine at 650 cm-1, adenine at 791 cm-1, cytosine and uracil combining at 790 cm-1, and phosphate at 1100, 1335 (in combination with adenine and guanine), 1465 and 1570 cm-1. Surprisingly, adenine, which demonstrated the greatest surface-enhanced Raman effect, does not dominate the ring-breathing mode portion of the spectrum. The intensities of the other base-pairs bands are of the same order of magnitude. This suggests that when the base-pairs are linked together, as in RNA, they are enhanced in concert. In fact, the relative intensities are very similar to a normal Raman spectrum of E. coli RNA, which shows the combined cytosine and uracil band at ca. twice the intensity of the adenine band, and ca. four times the intensity of the guanine band. Unfortunately, this means that the independent enhancement factors for the NA bases can not be used to estimate relative concentrations. For example, the relative 791 and 734 cm-1 bands for B. subtilis would indicate that the cytosine and/or uracil concentration was at least 20 times the adenine concentration, whereas each of the four RNA bases are known to contribute 15-35%. Nevertheless, it is worth noting that the three RNA samples yield different relative band intensities that were reproduced in numerous measurements. Although the relative concentrations of the NA bases for these samples have not been determined, these differences can be quantified. If it is assumed that the 650 cm-1 band represents 25% guanine, the 791 cm-1 band represents 25% adenine, and the 790 cm-1 50% cytosine plus uracil in the E. coli RNA spectrum, then the relative concentrations can be estimated for the other RNA samples. To aid this calculation, the three spectra were normalized to the phosphate band at 1100 cm-1, which has been shown to correlate to the total phosphate concentration and can be used as an internal standard. In addition a simple baseline correction was applied (Figure 7). This yields 15% adenine, 30% guanine and 55% cytosine plus uracil for B. subtilis RNA and 18% adenine, 25% guanine and 57% cytosine plus uracil for S. aureus RNA. The average S/N of these measurements was 26 with an average error of 8% of the value (S±N). It is also worth noting that the three RNA spectra show a marked shift in a band near 825 cm-1. This band is assigned to the symmetric stretch of the O-P-O ester linkage.9 The band appears at 815 cm-1 for S. aureus, shifting to 820 cm-1 for B. subtilis, and 830 cm-1 for E. coli. Others have used the normal Raman intensity of the band at 815 cm-1 as a direct indication of the amount of A-class helix present, while the intensity of the band at 830 cm-1 has been used as a direct indication of the amount of B-class helix present. However, the latter is more associated with DNA, than RNA. SPIE 2001-4575 68
- 15. A B OPO E.coli B. subtilis S. aureas G A C+U P Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 7. SER spectra of RNA from A) B. subtilis with contributions indicated and B) E. coli, B. subtilis and S. aureus with baseline correction and peak positions used to calculate % contributions indicated. G = guanine, A = adenine, C+U = cytosine plus uracil, P = phosphate (backbone), OPO = phosphate ester linkage (A- vs. B-class helix). SG-SER spectra of reasonable quality were also obtained for E. coli and B. subtilis, especially the latter (Figure 8). However, the spectra differ substantially from the E-SERS of the same samples. Both SG-SER spectra are dominated by adenine at 735 cm-1 and a band at 1030 cm-1. Although unassigned, the latter does appear in the RNA E-SER spectra. Bands at 1105 and 1565 cm-1 are likely due to phosphate, while bands at 1320 and 1470 cm-1 are less confidently assigned to phosphate. They are significantly less intense and somewhat shifted from their SG-SERS counterparts (1335 and 1455 cm-1). A band at 670 cm-1 may be due to guanine, which was observed at 664 cm-1 for SG-SERS of the pure sample. However, the SG-SER spectrum of AMP also had an intense 667 cm-1 band. A number of other bands occur at 890, 1070, 1165, 1245, 1290, 1420, 1505 cm-1 and remain unassigned. The SG-SER spectra are somewhat disappointing, in that only adenine and guanine contributions can be positively identified. This limits the ability to determine relative NA base concentrations and distinguish bacterial RNA. However, several of the unassigned bands may be due to the bases (e.g. 1030 and 1420 cm-1 due to cytosine). Further experiments will be required to clarify this point. A B E-SERS E-SERS SG-SERS SG-SERS Wavenumbers (∆cm-1) Wavenumbers (∆cm-1) Figure 9. E-SER (-0.3VSCE) and SG-SER spectra of RNA from A) E. coli and B) B. subtilis. Sample conditions as in Figure 6. E-SER spectra at 750 mW, SG-SERS at 75 mW. SPIE 2001-4575 69
- 16. A final comparison was made between the two SER methods by measuring dipicolinic acid (DPA). This chemical may be invaluable as a test for spore forming bacteria, specifically B. anthracis. 50 to 90% of B. anthracis sporilates. During spore formation dipicolinic acid is synthesized, and once completed, 10-15% of the dry spore weight is composed of the Ca2+ complex located in the spore core.42 Heating in water can be used to initiate germination, at which point the exosporium breaks and releases the Ca dipicolinate, which becomes dipicolinic acid in water. The structure of this chemical strongly suggested that it would be SER active. However, the E- B (Ax20) SER spectrum was unstable and varied considerably as a A DPA function of potential. A consistent spectrum was obtained at +0.6VSCE (Figure 10). This potential is not recommended for measurement, because the surface is actively dissolving in solution. The SG-SER spectrum was considerably more stable, of higher quality, and easily reproduced. Bands at 660, 825, 1010, 1390, 1430, 1570, C 1590, and 3075 cm-1 were observed. Enhancement factors were determined for the two media using the symmetric ring stretching mode at 995 and 1010 cm-1, for the normal Raman and SER spectra respectively. E-SERS yielded an EF of 5x103, while SG-SERS yielded an EF of 2x105 for D DPA. The S/N of the latter suggests a detection limit of 2.0x10-10g (based on adenine coverage, 75 mW and 10- min). The differences in SER activity for these two media may be attributed to the combined electrolytic potential of Wavenumbers (∆cm-1) the solution, chemical and metal.15 Again the E-SERS Figure 10. A) Raman spectrum of solid dipicolinic acid, suggests that the SG-SERS is at a potential positive of the B) Ax20, C) electrolytic SERS of 6x10-3 M dipicolinic pzc. While the instability in the E-SERS may also be acid in 0.1 M KCl at a potential of +0.7VSCE and pH of 4, associated with surface interactions of two carboxylic acid and D) sol-gel SERS of 6x10-3 M dipicolinic acid. groups of dipicolinic acid during the ORCs. Conditions for A and C as in Figure 1, C) 100 mW of 1064 nm, 50 scans, 8 cm-1. 4. CONCLUSIONS Towards the goal of developing a method and technology to rapidly detect and identify bioagents, we have been investigating surface-enhanced Raman spectroscopy as a tool to measure relative concentrations of nucleic acid bases in RNA and determine bioagent taxonomy. Initially, we investigated E-SERS, since this method has been extensively researched, and the optimum sample conditions are well developed. However, this method requires a three-electrode sample cell and electrolyte solution. Incorporation of an electrolytic cell into sample systems used to analyze for BWA as aerosols, in water or on surfaces can be designed using flow injection analysis technologies, but cross-contamination and plugging of sample lines seems inevitable. For this reason, we also investigated metal-doped sol-gels as a SER-active medium. Previous studies have shown this material to be active in all solvents, particularly water, capable of continuous measurements in flowing systems, and reproducible (quantitative) between coated sample vials. Here we compared SG-SER spectra to traditional E-SER spectra of the nucleic acid base pairs, adenosine monophosphate and RNA. High quality spectra of adenine, cytosine, guanine, thymine and uracil were obtained by both E-SERS and SG-SERS. Both methods yielded very similar spectra for the NA bases, including a pH dependent study of adenine. Enhancement factors and detection limits for adenine were determined as 2x105 and 1.6x10-11g, and 1x105 and 1.2x10-10g for E-SERS and SG-SERS, respectively (normalized to 75 mW and 10-min acquisition time). Fifty percent of the silver particle surface area in the sol- gel matrix was assumed covered by adenine, which may have been overestimated yielding a lower EF and higher detection limit. It should also be realized that each E-SER spectrum required several attempts to optimize the measurement conditions (pH, electrode potential, etc.). While each SG-SER spectrum involved no sample preparation, and often represents the first and only attempt to make the measurement. Quality spectra of RNA extracted from Escherichia coli, Bacillus subtilis and Staphylococcus aureus were obtained by E- SERS that were easily interpreted. Bands due to adenine, guanine, cytosine plus uracil, and phosphate were identified. The SER band intensity of the NA bases in the RNA samples were of the same order of magnitude, suggesting that their interaction with the silver surface is concerted as is their Raman enhancement. Interestingly, the relative SER band SPIE 2001-4575 70
- 17. intensities for RNA extracted from E. coli are very similar to those measured by normal Raman spectroscopy. Although the relative percent that each of the NA bases contributed to each RNA sample was not determined, reproducible band intensities allowed noting the following trends. The percent adenine decreases, while the combined percent cytosine and guanine increase for both B. subtilis and S. aureus compared to E. coli. Quality spectra were also obtained for the RNA samples by SG-SERS, but only a few bands were readily identified. Calculations of NA base concentrations by SG-SERS will require further research. In light of recent events, we summarize the capabilities of these SERS media in terms of rapid detection of B. anthracis and dipicolinic acid. However, these capabilities must be qualified. First and foremost, the level to which SERS can distinguish bacteria or viruses has not yet been determined. Development of a database of both DNA and RNA base concentrations for BWAs and common bacteria to establish the level of taxonomic identification is ongoing. Second, rapid collection of aerosol, water, or surface samples is being addressed by others, who report trapping particles on filters from 100 liters of air per minute. Third, although not presented here, we have developed methods to extract RNA or DNA from cells and spores for SER analysis within 10 minutes. Finally, we assume a detection limit of 3600 spores per 100 liters of air is required, although a 50% lethal dosage of anthrax has not been established. With these qualifications, a mass detection limit for RNA using SERS is estimated as follows. A single measurement is performed in ca. 20 minutes (140 liters collected in 1.4 min, RNA extracted in 8 min, spectral acquisition and analysis in 10 min). The average human breaths 7 liters per minute, therefore the analyzer must, at the very minimum, detect 5000 spores in 140 liters of air. One spore is approximately 2x10-18m3 (1x1x2 µm), and if a density of 0.75 g/cm3 is assumed, this corresponds to a mass of 1.5x10-12g. Each spore contains 4-12% RNA or 1.2x10-13g RNA for 8%. If we assume 2/3 of the RNA can be isolated for analysis during lysis, then the proposed instrument must be able to detect 4x10-10g RNA from 5000 spores per 70 liters of air within 10 minutes. As noted above, the mass detection limits for adenine were estimated at 1.6x10-11g, and 1.2x10-10g for E-SERS and SG-SERS, respectively. Although, these detection limits suggests that RNA from 5000 spores is detectable with the current instrumentation, it is highly likely that only a portion of an RNA segment (e.g. 120-nucleotide 5S rRNA) is in contact with the metal surface and will contribute to the SER effect. The S/N for the RNA spectra were 1/10th of the average S/N for the four individual RNA bases suggesting a 10% contribution. Furthermore, effective taxonomy will likely require knowing the NA base concentrations to 1% of the value (e.g. 25±0.25%). Again the average measurement error for the bases is 12%. These values suggest that the E-SERS is within a factor of 4 of the required detection limit, whereas the SG-SERS detection limit must be improved by 25 times. The same arguments can be applied to the detection of dipicolinic acid. If we assume a spore releases 10% by weight DPA during germination, then the proposed instrument must be able to detect 7.5x10-10g DPA from 5000 spores per 70 liters of air within 10 minutes. The detection limit for SG-SERS was estimated at 2.0x10-10g and suggest that the vials are suitable to perform a rapid screen for anthrax. A series of concentration dependent measurements are currently being performed to verify this assertion. Finally, we note that the measurements performed here employed an FT-Raman spectrometer. This instrumentation was chosen over dispersive Raman instrumentation for the high wavelength accuracy afforded by the HeNe reference laser (Connes Advantage43), which would allow reliable spectral subtraction, matching of observed spectra to stored library spectra, and confident use of chemometric approaches. Such data analysis is likely to be required to enhance BWA identification. However, this instrumentation sacrifices sensitivity due to a less efficient detector (InGaAs vs. Si), less efficient Raman scattering, and less efficient generation of plasmon modes. Substantial improvements can be obtained using visible excitation and Si detection and these measurements are underway. 5. ACKNOWLEDGEMENTS The authors are grateful to Drs. D. Cookmeyer and S. Tove of the U.S. Army Research Office (Contract Number DAAH04- 96-C-0078) for their interest and support of this research. The authors would also like top acknowledge Dr. R. Yin and J. Jensen for supporting development of the metal-doped sol-gels (Contract Number DAAD13-01-C-0019). They also thank Dr. Wilfred H. Nelson for assistance in spectral interpretations. SPIE 2001-4575 71
- 18. 6. REFERENCES 1. Roberts, W.L and Rainey, P.M., Clin. Chem., 39, 1872-1877 (1993). 2. Pasechnik, V.A., C.C. Shone, and P. Hambleton, Bioseperations, 3, 267-283 (1993). 3. Jackson, P.J., M.E. Hugh-Jones, D.M. Adair, G. Green, K.K. Hill, C.R. Kuske, L.M. Grinberg, F.A. Abramova, and P. Keim, Proc. Natl. Acad. Sci., 95, 1224-1229 (1998). 4. Jensen, J.L., N.C. Wong, W. R. Loerop, SPIE, 4775-03 (2001) 5. Snyder, A. P.et al. SPIE, 3853-15 (1999). 6. Danosky, T. R. and McFadden, P. N., in press (1997) 7. Woodruff, W.H., Farquharson, S., Science, 201, 831-833 (1978) 8. Lord, R.C. and Thomas, G.J.,Jr., Spectrochemica Acta, 23A, 2551-2591 (1967). 9. Thomas, G.J.,Jr., Biochim. Biophys. Acta, 213, 417-423 (1970) 10. Placzek, G., "Handbuch der Radiologie," 2, E.Marx, e.d., Akademische Verlagagescellschatt, Liepzig, 1934, UCRL Trans. No. 526 (1959). 11. Jeanmaire, D.L., and R.P. Van Duyne, J. Electroanalytical Chem., 84, 1-20 (1977). 12. Chada, S., Manoharan, R., Moenne-Loccoz, P., Nelson, W.H., Peticolas, W.L. and Sperry, J.F., Applied Spectroscopy, 47, 38-43 (1993). 13. Manoharan, R., Ghiamati, E., Chada, S., Nelson, W.H., and Sperry, J.F., Applied Spectroscopy, 47, 2145-2150 (1993). 14. Todd, E.A., Morris, M.D., Applied Spectroscopy, 48, 545-548 (1994). 15. Wentrup-Byrne, E., Sarinas, S., and Fredericks, P.M., Applied Spectroscopy, 47, 1192-1197 (1993). 16. Weldon, M.K., V.R. Zhelyaskov, and M.D. Morris, Applied Spectroscopy, 52, 265-269 (1998). 17. Kneipp, K. and J. Fleming, J. Mol. Structure, 145, 173-179 (1986). 18. Pothier, N.J. and Force, R.K., Applied Spectroscopy, 46, 147-151 (1992). 19. Ervin, K.M., E. Koglin, J.M. Sequaris, P. Valenta, and H.W. Nurnberg, J. Electroanal. Chem. 114, 179-194 (1980). 20. Kim, S.K., T.H. Joo, S.W. Suh, and M.S. Kim, J. Raman Spectrosc., 17, 381-386 (1986). 21. Pothier, N.J., and Force, R.K., Analytical Chemistry, 62, 678-680 (1990). 22. Kneipp, K., Y. Wang, R.R. Dasari, and M.S. Feld, Applied Spectroscopy, 49, 780-784 (1995). 23. Nie, S, and Emory, S.R. Science, 275, 1102 (1997). 24. Graham, D., W.E. Smith, A.M.T. Linacre, C.H. Munro, N.D. Watson, and P.C. White, Analytical Chemistry, 69, 4703- 4707 (1997). 25. Fleischmann, M., P.J. Hendra, and A.J. McQuillan, Chem. Phys. Lett., 26, 163-166, (1974). 26. Pettinger, B., J. Chemical Phys., 85, 7442-7451 (1986). 27. Surface-Enhanced Raman Scattering, Section Four: Theory, SPIE, MS 10, M. Kerker and B. Thompson Eds. (1990). 28. Wang, D.-S., and Kerker, M., SPIE (M. Kerker and B. Thompson Eds.), MS 10, 417-429 (1990). 29. Weaver, M.J., Farquharson, S., Tadayyoni, M.A., J. Chem. Phys., 82, 4867-4874 (1985). 30. Norrod, K.L., Sudnik, L.M., Rousell, D., and Rowlen, K.L., Applied Spectroscopy, 51, 994-1001 (1997). 31. Farquharson, S., Weaver, W.J., Lay, P.A., Magnuson, R.H., and Taube, H., J. Am. Chem. Soc., 105, 3350-3351 (1983). 32. Lee, P.C. and Meisel. D. “Adsorption and Surface-Enhanced Raman of Dyes on Silver and gold Sols,” J. Phys. Chem., 86, 3391-3395 (1982). 33. Li, Y.-S., and Wang, Y., Applied Spectroscopy, 46, 142-146 (1992). 34. Bello, J.M., D.L. Stokes, and T. Vo-Dinh, Analytical Chemistry, 61, 1779-1783 (1989) 35. van Duyne, R.P., J. C. Hulteen, D. A. Treichel, M. T. Smith, M. L. Duval, and T. R. Jensen, J. Phys. Chem. B,103,3854- 3863 (1999). 36. Farquharson, S., and W. W. Smith, W. H. Nelson and J. F. Sperry, SPIE, 3533-27, 207-214 (1998). 37. Lee, Y. H., W. Smith, S. Farquharson, H. C. Kwon, M. R. Shahriari, and P. M. Rainey, SPIE, 3537, 252-260 (1998) 38. Lee, Y.-H., S. Farquharson, and P. M. Rainey, SPIE, 3857, 76-84 (1999). 39. Current Protocols in Molecular Biology, Wiley Interscience,10.03-10.06 (1987) 40. Farquharson, S., Smith, W., Carangelo, R. C., and Brouillette, C., SPIE, 3859, 14-23 (1999) 41. Chase, D. B. and J.F. Rabolt, Fourier Transform Raman Spectroscopy, Acad. Press, Ch.1, p. 131 (1994). 42 . Brock, T.D., M.T. Madigan, J.M. Martinko, and J. Parker, Biology of Microorganisms, 7th Ed., Prentice Hall, p. 76-80 (1994). 43. Connes, J. Rev. Opt. Theor. Instrum., 40, 45 (1961). SPIE 2001-4575 72
- 19. Appendix C Chemical agent identification by surface-enhanced Raman spectroscopy Stuart Farquharson* and Paul Maksymiuk Real-Time Analyzers, 87 Church Street, East Hartford, CT 06108 Kate Ong and Steven D. Christesen U.S. Army, SBCCOM, Aberdeen Proving Ground, MD 21010 ABSTRACT The recent distribution of anthrax through the U.S. postal system and the subsequent infection and death of several postal and national media employees, amplifies the need for methods to rapidly detect, identify, and quantify this and other chemical and biological warfare agents. The U.S. military has also identified water supplies as a likely method of warfare agent deployment and is funding the development of a Joint Service Agent Water Monitor (JSAWM). In an effort to aid military personnel and the public at large, we are developing a portable analyzer capable of identifying and quantifying chemical agents rapidly, either "on-demand" or continuously. The approach is based on the ability of Raman spectroscopy to identify molecular structure through the abundant vibration information provided in spectra and the ability of surface-enhanced Raman spectroscopy (SERS) to detect extremely low concentrations (e.g. part-per-billion) through the enhancement of Raman scattering by six orders of magnitude or more. A key element to the analyzer design is a new SER active medium that is capable of quantitative, reversible measurements. The medium consists of silver or gold nanoparticles incorporated into a sol-gel matrix. The porous silica network offers a unique environment for stabilizing SER active metals and the high surface area increases the interaction between the analyte and metal particles. Here we present the use of new sol-gels that also selectively enhance chemicals based on polarity and charge. Base-line measurements of chemical agents and their hydrolysis products are presented and compared to the JSAWM goal of 3.0 micrograms per liter detection. Keywords: Chemical warfare agent, hydrolysis product, SERS, Raman spectroscopy, sol-gel, nanoparticle 1. INTRODUCTION Since September 11, 2001, the threat of terrorist attacks and biological warfare within U.S. borders has become a sobering reality. The simplicity in manufacturing, ease of deployment, and the relatively low cost of chemical warfare agents (CWAs) raises public concern that they may also be used by terrorists. Indeed, terrorists released sarin (GB) in the Tokyo subway in 1995.1 Countering terrorism and terrorist attacks requires recognizing likely deployment scenarios and having the required technology to rapidly detect the deployment event. One method of deployment has been long identified by the U.S. military: distribution through water supplies. To counter this threat, the Department of Defense is funding or monitoring the capabilities of 200 technologies with the goal of developing a Joint Service Agent Water Monitor (JSAWM) that is field portable.2 The JSAWM must be capable of identifying the chemical agents with no false-positives and quantifying the agents at microgram per liter concentrations within 10 minutes (Table 1).3 This includes the analysis of drinking water supplies, distribution and storage systems, as well as potable water supplies. The earliest technologies developed for CWA detection were based on electrochemical, ionization, or colorimetric analysis (e.g. phosgene tape). Although these analyzers were easy to use, they were not generally agent specific and suffered from false-positives.4 More traditional laboratory methods have also been investigated, and in particular, combined gas chromatography and mass spectrometry (GC/MS) has been very successful at eliminating false-positives.5,6 However, GC/MS requires extraction, repeated calibration, and long analysis times (typically 20 to 60 minutes),6 making it labor intensive and less than desirable for field use. More rapid analysis of agents in the solid, liquid and gas phase has been * To whom correspondence should be addressed, email:farqu@real-time-analyzers.com Vibrational Spectroscopy-based Sensor Systems, Steven D. Christesen, Arthur J. Sedlacek III, Editors, 166 Proceedings of SPIE Vol. 4557 (2002) © 2002 SPIE ·0277-786X/02/$15.00
- 20. Table 1. Chemical Agent Structures, Hydrolysis Half-lives, and JSAWM Thresholds. Agent Short-Hand Chemical Structure Hydrolysis JSAWM Half-Life* Thresholds Sarin (GB) F-[O=P-CH3]-O-CH(CH3)2 21.3 hours 3.2 µg/L Soman (GD) F-[O=P-CH3]-O-CH(CH3)-(C-(CH3)3) 2.3 hours 3.2 µg/L Tabun (GA) (CH3)2-N-[O=P-CN]-O-C2H5 4.1 hours 3.2 µg/L VX C2H5O-[O=P-CH3]-S-(CH2)2-N-(CH(CH2)2)2 82.1 hours 3.2 µg/L EA2192 HO-[O=P-CH3]-S-(CH2)2-N-(CH(CH2)2)2 >9 years 3.2 µg/L Mustard (H) ClCH2CH2-S-CH2CH2Cl encapsulates 47 µg/L Lewisite (L) ClCH=CH-As-Cl2 rapid 27 µg/L HCN HCN rapid 2.0 mg/L BZ** C7NH12-O-[C=O]-COH(C6H5)2 2.3 µg/L T-2 Toxin 8.7 µg/L * at pH 7 to 7.5 and 20 to 25 oC. demonstrated by vibrational spectroscopy.7-10 Hoffland et al.7 reported infrared absorbance spectra and absolute Raman cross sections for several chemical agents, while Christesen measured Raman cross sections for sarin, tabun, mustard gas, and VX (ethyl S-2-diisopropylamino ethyl methylphosphonothioate).11 Again, however these techniques also have limitations. Raman spectroscopy is simply not a very sensitive technique, and detection limits are typically 0.1% (1000 ppm). While infrared spectroscopy would have limited value in analyzing poisoned water, since the very strong infrared absorption of water would obscure most other chemicals present. Nevertheless, efforts to overcome these limitations have been demonstrated. Braue and Pannella8 quantified the G-series nerve agents (tabun, sarin, and soman) in terms of infrared attenuated total reflectance using a circle-cell. And Alak and Vo-Dinh demonstrated the possibility of surface-enhanced Raman spectroscopy (SERS) to identify CWAs by measuring several organophosphonates that simulate the nerve agents.12 However, quantitative measurements have not been demonstrated for the SER-active material used (silver coated on alumina particles) or other SER-active media.13 Recently, we developed silver-doped sol-gels to promote the SER effect.14-17 The porous silica network of the sol-gel matrix offers a unique environment for stabilizing SER-active metal particles, and the sol-gel provides a high surface area that effectively increases the number of molecules observed within the Raman scattering volume. The silver-doped sol-gels have been coated on the internal walls of standard glass vials to allow rapid SER analysis of small sample quantities (< 0.1 mL) without preparation. We have used p-aminobenzoic acid (PABA) as a test chemical to demonstrate surface enhancements greater than 106, reversible measurements in a flowing system, reproducible measurements from vial-to-vial and batch-to- batch, and measurements in multiple solvents, including water.14-17 Recently, we used these vials to measure Tabun (GB) and Sarin, and several hydrolysis products, pinacolyl methyl phosphonate (PMP from Soman), and methyl phosphonic acid (MPA from all G-agents, Figure 1). Although a number of unique vibrational bands are observed (e.g. C-N stretch doublet and P-C stretch), the G-agents were only observed for 5% concentrations, and all spectra required baseline corrections. A C 790 C-N P-C 2135, 2190 770 545 1290 B D 760 Wavenumber (∆cm-1) Wavenumber (∆cm-1) Figure 1. Surface enhanced Raman spectra of ~5% v/v A) Tabun and B) Sarin, C) 1% v/v PMP and D) 10 ppm MPA using sol-gel sample vials, 785 nm excitation, 1-min scan, and CCD detection. Performed at Aberdeen Proving Ground. Proc. SPIE Vol. 4577 167
- 21. Nevertheless, MPA was readily observed for a 10 ppm sample, with an estimated detection limit of 0.4 ppm (based on a signal-to-noise ratio of 3 for the 760 cm-1 band intensity). This measurement provides encouragement in that SERS may satisfy the needs of a JSAWM. Furthermore, MPA is also a hydrolysis product of VX and V-gas, and EA2192 (Figure 2), and may prove a valuable indicator of agent usage. O CH3 O CH3 O CH3 P C + H 2O HF + P C P + C H3C O CH3 H3C O CH3 H3C OH HO CH3 F OH OH Sarin MPAMME MPA 2-propanol Figure 2. Hydrolysis of Sarin to form hydrofluoric acid (HF), methylphosphonic acid, 1-methylethyl ester (MPAMME), methyl phosphonic acid (MPA) and 2-propanol. With this initial, albeit modest, success, we began analyzing chemicals with various sol-gel compositions that we have been developing. Here we describe four sol-gel compositions that select for 1) polar-positive, 2) polar-negative, 3) weakly polar- positive and 4) weakly polar-negative chemical species. The ability of these sol-gels to select and enhance Raman scattering is described for several test chemicals and MPA. 2. EXPERIMENTAL The chemicals analyzed, as well as all chemicals used to prepare the metal-doped sol-gels were obtained at their purest commercially available grade from Aldrich (Milwaukee, WI). The sol-gel designed to select for polar-negative species was prepared from a silver amine complex, tetramethyl orthosilicate (TMOS) and methanol. After mixing, 0.2 mL of the sol-gel solution was transferred into a glass vial (2 mL), dried and heated. The incorporated silver ions were then reduced using dilute sodium borohydride. The vials were washed and dried prior to the addition of a sample solution. In a similar manner, the sol-gel designed to select for polar-positive species was prepared from a gold salt, TMOS and methanol. The sol-gel designed to select for weakly polar-negative species was prepared from a silver amine complex, tetraethyl orthosilicate (TEOS) and methanol. And the last sol-gel designed to select for weakly polar-positive species was prepared from a gold salt, TEOS and methanol. All samples were prepared in a chemical hood and transferred into plain or SER-active vials for analysis. Normal Raman spectral measurements employed 1-mL pure samples that were placed in a 1-cm3 cuvette and weighed. This yielded a powder density that allowed accurate calculation of molecules in the optical collection field. SERS measurements employed 1-mg sample per mL water concentrations, unless otherwise stated. Once prepared, a 0.1 mL sample was placed into one of the four selective sample vials, which in turn was placed into the sample compartment of a Raman spectrometer for analysis. A prototype Fourier transform Raman spectrometer (RTA) was used for these measurements.18 The system consisted of a Nd:YAG laser (Brimrose) for excitation at 1064 nm, an interferometer built by On-Line Technologies (OLT, East Hartford, CT) for frequency separation, an uncooled InGaAs detector for signal detection (RTA), and an Intel 400 MHz Pentium II based laptop computer (Dell, Round Rock, TX) for interferometric control, data acquisition (OLT), and analysis (LabVIEW by National Instruments, Austin, TX). Additional components included a Notch filter (Kaiser, Ann Arbor, MI) and interferometer entrance and exit optics (Edmund Scientific, Barrington, NJ). Fiber optics were used to deliver the excitation beam to the sample and the scattered radiation to the interferometer (1 meter lengths of 200 and 365 micron core diameter, respectively, Spectran, Avon, CT). A second Notch filter (Kaiser) was used as a beam splitter to direct the excitation beam along the same axis as the collected radiation. A microscope object (20x0.4, Newport, Irvine, CA) was used to focus the beam into the sample and to collect the scattered radiation back along the same axis. In this co-axial backscattering arrangement, the excitation beam was passed through the outside of a glass vial and focused onto the silver-doped sol-gel film (0.1-0.3 mm thickness) containing the sample. 3. RESULTS AND DISCUSSION p-aminobenzoic acid (PABA) and phenyl acetylene (PA) and were used to refine the selectivity and SER-activity of the four different metal-doped sol-gels. PABA is a popular chemical used to evaluate the performance of SER-active media. Here the polar end groups can be used to test selectivity of the polar-negative and polar-positive sol-gels. PA is essentially non- 168 Proc. SPIE Vol. 4577
- 22. polar, but a high electron density in the cylindrical π cloud around the carbon-carbon triple bond allows testing the selectivity of the weakly polar-negative and weakly polar-positive sol-gels. As Figure 3 illustrates, PABA passes through the polar sol- gel and is enhanced by either the silver or gold particles. At 1 mg/ml the concentration of neutral PABA is ca. 20 times that of the ionized form (pKa = 4.8). For electropositive silver, the PABA anion is expected to interact through the carboxylate group, and the associated vibrational modes are expected to dominate the spectrum. Conversely, for electronegative gold, either form of PABA is expected to interact through the amine group. The clear differences in our spectra support this expectation. Furthermore, bands at 840 and 1405 cm-1, assigned to a COO- bend and stretch, respectively, are significantly more intense for silver than gold. Additional bands at 1140 and 1195 cm-1 are assigned to CH bending modes, while bands at 1450, 1500 and 1605 cm-1 are assigned to ring vibrational modes. A very similar SER spectrum for PABA on a silver-coated alumina substrate has previously been reported with similar assignments.19 For the gold-doped sol-gel, new bands appear at 690, 1355, and 1585 cm-1. The first band is assigned to a ring-H bending mode, the second band to a ring-N- stretching mode, and the third band to a possible NH2 scissors mode or ring mode. The second band is not observed in the normal Raman spectrum, but infrared bands occur at this frequency for aromatic ring-secondary amine stretching modes. The scissors mode occurs at this frequency in Raman spectra for several chemicals, but is absent in the PABA Raman spectrum. Alternatively, this mode may be the1600 cm-1 ring mode that has been shifted by the gold interaction. Again, a very similar SER spectrum of PABA has been reported, but surprisingly using silver (colloids)20,21 not gold as the enhancement medium. These researchers also assumed the primary interaction of PABA with silver was through the carboxylate anion, and made assignments accordingly. For example, they assigned the 1359 cm-1 to a COO- stretch, not to the amine group as we have. They also favor the ring stretching mode assignment for the 1582 cm-1 band. Finally, it should be said that other researchers have argued that the most dominant band in the SER spectra at 1450 cm-1, a ring vibration mode, suggests that PABA lies flat on the surface, and the π-orbitals dominate the surface interaction.22 A C C CH H2N COOH B D Wavenumber (∆cm-1) Wavenumber (∆cm-1) Figure 3. SER spectra of A) PABA using polar-negative and B) polar-positive sol-gels, and C) PA using weakly polar- negative and D) weakly polar-positive sol-gels. PABA is 1 mg/mL, PA is 1% v/v. Spectral conditions: 75 mw 1064 nm, 100 scans (1.5 min), 8 cm-1 resolution. Non-polar PA passes through the non-polar sol-gels and is also enhanced by both metals. The spectra are easily understood. For electropositive silver, PA interacts through the cylindrical triple bond π electron cloud and a -C≡C- doublet occurs near 2000 cm-1. The interaction is reasonably strong, since this band appears at 2112 cm-1 in the normal Raman spectrum. For electronegative gold, this interaction is unlikely and only very weak bands occur near 2000 cm-1. The remaining bands are at 1000 cm-1, 1200 cm-1 doublet and 1595 cm-1 all appear in the normal Raman spectra at virtually the same frequencies and are assigned to the symmetric ring-breathing mode, CH bending modes, and the trigonal ring-breathing mode, respectively. The polar/non-polar selectivity of the polar-negative and weakly polar-negative sol-gels was tested by adding a 1:1 molar mixture of PABA and PA. The selective enhancement is quite good (Figure 4). The spectrum obtained using the polar sol- gel represents 78% PABA and 22% PA, while the spectrum obtained using the weakly polar sol-gel represents 9% PABA and 91% PA. The band peak intensities at 2000 cm-1 for PA and 1450 cm-1 for PABA were used for these calculations, and are expanded in Figure 4 for clarity. Proc. SPIE Vol. 4577 169
- 23. A C B D Wavenumber (∆cm-1) Wavenumber (∆cm-1) Figure 4. SERS of 1:1 M/M of PABA and PA in A) polar-negative and C) weakly polar-negative sol-gels. The lower traces, compare the pure chemicals; B) 1 mg/ml PABA in polar-negative sol-gel and D) 1% PA in weakly polar-negative sol-gel, while the insets magnify the minority species for clarity (x5 in A and x10 in B). Spectral conditions as in Figure 3. Following this development of selective sol-gels that maintained SER activity, we measured cyanide and MPA (Figure 5). Not surprisingly, the best sensitivity for both hydrolysis products was obtained using the polar-negative sol-gel. The interaction of the cyanide anion with the silver surface is sufficient to shift the C≡N stretch observed at 2080 cm-1 in the normal Raman spectrum to 2145 cm-1 in the surface-enhanced Raman spectrum. Furthermore, the band is substantially broadened. This anion has been extensively studied by electrolytic SERS and this shift and broadening have been attributed to the formation of a tetrahedral Ag(CN)32- surface structure.23 A B Wavenumber (∆cm-1) Wavenumber (∆cm-1) Figure 5. Surface-enhanced (upper traces) and normal Raman spectra (bottom traces) of A) CN- and B) MPA in silver-doped TMOS. SERS conditions as in Figure 3, and 1 mg/mL. Note MPA yields two distinct spectra for neutral (top) and acidic pH (middle). The normal Raman spectra employed pure powders, 500 scans and 900 mW of 1064 nm. SER measurements of MPA with the polar-negative sol-gel yielded two unique spectral signatures that depended on solution pH (Figure 5). For more neutral solutions, the P-C stretch of MPA at 762 cm-1 dominates, and the CH2 stretch at 2922 cm-1 appears. The S/N is sufficiently high that the anti-Stokes Raman shift at -762 cm-1 is observed. For deprotonated MPA, an oxygen-surface mode appears at 325 cm-1 (as well as its anti-Stokes complement) suggesting a strong interaction. This results in substantial enhancement of the P-O-C mode at 1051 cm-1 (upper trace). Others report that this mode dominates the infrared spectra of nerve agents measured in water.8 Comparison of the two spectra suggests the following molecule-to- surface orientations. The appearance of the oxygen-surface and P-O-C modes in the upper spectrum of Figure 5B indicates that the tetrahedral molecule interacts with the silver surface through the deprotonated oxygen and is oriented end-on. The 170 Proc. SPIE Vol. 4577
- 24. dominance of the P-C and the CH2 stretches, and the disappearance of the P-O-C mode in the upper spectrum suggest the molecule is oriented with the methyl-hydroxyl-hydroxyl face towards the surface. However, considerably more research must be performed to verify these points. Both chemical forms were measured as a function of concentration to determine the onset of monolayer coverage and estimate expected detection limits (Figure 6). Below monolayer coverage the signal to concentration dependence should be linear, and the S/N of any spectral measurement in this range can be used to predict the detection limit. In the spectra presented here, the peak height was used as the signal, while the noise, as root-mean-squared (RMS) was measured between 4400-4600 cm-1. Since noise is distributed evenly throughout the spectrum when transformed, this region was used since it does not have any contributions from signals or baseline offsets. Figure 6 shows a series of spectra for MPA along with a plot of the 1050 cm-1 band intensity (the noise was identical for this series) as a function of sample concentration. A clear discontinuity appears in the vicinity of 0.1 mg/mL (19 ppm), indicating the onset of monolayer coverage. A detection limit, defined as a S/N of 3, was calculated for the 0.1 and 0.05g/mL samples at 2.4x10-4 and 2.5x10-4 g/L, respectively. A more modest detection limit of 10.1x10-4 g/L was obtained using the 760 cm-1 band in the second series of concentration measurements. These detection limits correspond to 46 and 190 ppb for a 3-min scan and 75 mW of incident laser power. 140 600 A 120 B 500 100 400 I (1050) I (760) 80 300 60 200 40 20 100 0 0 0 0.2 0.4 0.6 0.8 1 1.2 [MPA] (mg/mL) Wavenumber (∆cm-1) Figure 6. A) Concentration dependence of MPA SERS measured in silver-doped TMOS). B) Concentrations are 0.01, 0.05, 0.1, 0.5, 1 g/L (1.88, 9.4, 18.8, 94, 188 ppm). I760 series (•) and I1050 series (∆). Enhancement factors for cyanide and methyl phosphonic acid can be determined by estimating the number of molecules contributing to the surface-enhanced and normal Raman spectra. The enhancement factor, EF, can be defined by the following equation: EF = (ISERS/INR)•(MNR/MSERS) •(PNR/PSERS) •(TNR/TSERS)1/2 where I is the spectral band intensity, M is the sample mass, P is the incident laser power, and T is the measurement time (or number of scans) for the two measurements. For the normal Raman spectra a cylindrical scattering volume is assumed, based on the laser area (2.8x10-7m2, 6x10-4m diameter spot) and the penetration depth (1x10-3 m).24 The density of KCN and MPA as powders were measured at 0.572 and 0.516 g/cm3, indicating that 1.6x10-4 and 1.44x10-4 g produced the normal Raman signals in Figure 5, respectively. The number of molecules contributing to the SER spectra are those on the silver particles embedded in the sol-gel. The total silver surface area can be determined from the average particle size, concentration, and the scattering volume. Previous scanning electron micrographs showed the average silver particle size to be 40 nm in diameter (3.35x10-23m3).17 The silver concentration is 0.12M based on the reactant molar concentrations and dilution factors. And the scattering volume is 7.6x10- 11 3 m , again based on a cylindrical scattering volume, defined by a laser area of 2.8x10-7m2 and a sol-gel thickness of 2.7x10- 4 m. This volume contains 1.23x10-6g of silver, equivalent to 3.5x109 silver particles with a collective surface area of 1.8x10- 5 2 m . However, it may reasonably be assumed that at least half of the silver surface is in contact with the sol-gel matrix and unavailable for analyte interaction. If we assume monolayer coverage and that each CN molecule occupies 1.5x10-20m2, then approximately 6.2x1014 molecules or 2.7x10-8g of CN contribute to the SER spectrum (2.0x10-19m2, 4.6x1013 molecules, 7.4x10-9g for MPA). Accordingly, the EF for cyanide equals 4.8x104 ((180/599) •(1.6x10-4/2.7x10-8) •(900/75) •(500/100)1/2). The EF for MPA is considerably higher at 8.7x106 ((603/26) •(1.44x10-4/7.4x10-9) •(900/75) •(500/200)1/2). Proc. SPIE Vol. 4577 171
- 25. 4. CONCLUSIONS Here we present for the first time, surface-enhanced Raman spectra of Tabun and Sarin obtained using silver-doped sol-gels. However, the inferior enhancement suggested employing SER-active sol-gels varying in composition to improve both detection limits and selectivity of the target analytes. To this end, we successfully demonstrated the capabilities of four sol- gels that select for 1) polar-positive, 2) polar-negative, 3) weakly polar-positive and 4) weakly polar-negative chemical species. p-aminobenzoic acid was used to show that silver could be used to attract polar-negative chemicals or functional groups (carboxylate anion) and that gold could be used to attract polar-positive chemicals (amine), while a mixture of p- aminobenzoic acid and phenylacetylene was used to show that tetramethyl orthosilicate preferentially solvates polar chemicals and that tetraethyl orthosilicate preferentially solvates weakly polar chemicals. This increased sample control was applied to cyanide and methyl phosphonic acid, two hydrolysis products of chemical warfare agents. Exceptional results were obtained for methyl phosphonic acid, allowing measurement of 1x10-2 g/L for a 3-min scan and 75 mW of incident laser power with an estimated detection limit of 2.45x10-4 g/L, and an enhancement factor of 8.7x106. However, this detection limit is 76 times less sensitive than required for the JSAWM (3.2x10-6g/L for the G-agents). Finally, we note that the measurements performed here employed an FT-Raman spectrometer. This instrumentation was chosen over dispersive Raman instrumentation for the high wavelength accuracy afforded by the HeNe reference laser (Connes Advantage25), which would allow reliable spectral subtraction, matching of observed spectra to stored library spectra, and confident use of chemometric approaches. Such data analysis is likely to be required to identify the chemical agents, as well as distinguish hydrolysis products. However, this instrumentation, which employs 1064 nm excitation and InGaAs detection, sacrifices sensitivity. We believe that the measurement sensitivity can be improved by at least two orders of magnitude by using shorter laser excitation wavelengths (e.g. 532 nm). This would provide more efficient Raman scattering (fourth power dependence on laser excitation wavelength), more efficient generation of plasmon modes, and allow using more efficient detector material (Si vs. InGaAs). These modifications are underway. 5. ACKNOWLEDGEMENTS The authors would like to thank Dr. R. Yin and J. Jensen of the U.S. Army for Proc. SPIE Vol.this work (Contract Number supporting 4577 DAAD13-01-C-0019). They would also like to thank Advanced Fuel Research for making their laboratory facilities available. 6. REFERENCES 1 Tu, Anthony, “Overview of Sarin Terrorist Incidents in Japan in 1994 and 1995”, 6th CBW Protection Symposium, Stockholm, Sweden, 10-15 May 1998. 2 Jensen, J.L., N.C. Wong, W. R. Loerop, SPIE, 4775-03 (2001) 3 JSAWM Requirements at www.sbccom.apgea.army.mil/RDA/ecbc/rt/PRODSER/JSAWM/jsawm.html 4 “The Chemical Weapons Convention Redefines Analytical Challenge”, Analytical Chemistry News & Features, June 1 397A (1998). 5 Johnston, R.L., Hoefler, C.M., Fargo, J.C., and Moberley, B., “The Defense Nuclear Agency’s Chemical/Biochemical Weapons Agreements Verification Technology Research, Development, Test and Evaluation Program and its Requirements for On-Site Analysis”, AT-ONSITE, 5-8 (1994) 6 Black, R.M., Clarke, R.J., Read, R.W., and Reid, M.T., “Application of gas chromatography-mass spectrometry and gas chromatography-tandem mass spectrometry to the analysis of chemical warfare samples, found to contain residues of the nerve agent sarin, sulphur mustard and their degradation products”, J. Chromatography, 662, 301-321 (1994) 7 Hoffland, L.D., Piffath, R.J., Bouck, J.B.,”Spectral signatures of chemical agents and simulants”, Optical Engineering, 24, 982-984, (1985) 8 Braue, E.H.J., Pannella, M.G.,”CIRCLE CELL FT-IR Analysis of Chemical Warfare Agents in Aqueous Solutions”, Applied Spectroscopy, 44, 1513-1520, (1990) 9 Tseng, C.-H., Mann C.K., and Vickers T.J., “Determination of Organics on Metal Surfaces by Raman Spectroscopy”, Applied Spectroscopy, 47, 1767-1771 (1993) 10 Staff Reporter, “US Army Tests Lidar to Detect Biological Toxins”, Photonic Spectra, pg. 50, December 1998. 172 Proc. SPIE Vol. 4577
- 26. 11 Christesen, S.D., "Raman cross sections of chemical agents and simulants", Applied Spectroscopy, 42, 318-321 (1988) 12 Alak, A.M. and Vo-Dinh, T., “SERS of Organophosphorous Chemical Agents”, Analytical Chemistry, 59, 2149-2153 (1987) 13 Norrod, K.L., Sudnik, L.M., Rousell, D., and Rowlen, K.L., “Quantitative Comparison of Five SERS Substrates: Sensitivity and Detection Limit”, Applied Spectroscopy, 51, 994-1001 (1997). 14 Lee, Y. and Farquharson, S., “SERS Sample Vials Based on Sol-Gel Process for Trace Pesticide Analysis”, SPIE, 4206, 140-146 (2000). 15 Farquharson, S. and Lee, Y., “Trace Drug Analysis by Surface-Enhanced Raman Spectroscopy”, SPIE, 4200-16 (2000). 16 Lee, Y., Farquharson, S., and Rainey, P. M., "Surface-Enhanced Raman Sensor for Trace Chemical Detection in Water", SPIE, 3857, 76-84 (1999). 17 Lee, Y, Farquharson, S., Kwong, H., and Shahriari, M., “Surface-Enhanced Raman Sensor for Surface-Enhanced Raman Spectroscopy”, SPIE, 3537, 252-260 (1998). 18 Farquharson, S., Smith, W., Carangelo, R. C., and Brouillette, C., “Industrial Raman: Providing Easy, Immediate, Cost Effective Chemical Analysis Anywhere”, SPIE, 3859, 14-23 (1999) 19 Narayanan, V.A., J.M. Bello, J.D. Stokes, and T. Vo-Dinh, Analusis, 19, 307-310 (1991) 20 Laserna, J.J., E. L. Torres, and J.D. Winefordner, Analytica Chemica Acta, 469-480 (1987) 21 Torres, E.L. and J.D. Winefordner, Analytical Chemistry, 59, 1626-1632 (1987) 22 Suh, J.S., D.P. DiLella, M. Moskovits, J. Phys. Chem., 87, 1540-1544 (1983). 23 Benner, R.E., R. Dornhaus, R. Chang, and B.L. Laube" Correlations in the Raman spectra of cyanide complexes adsorbed at silver electrodes with voltammograms", Surface Science, 101, 341 (1980) 24 Chase, D. B. and J.F. Rabolt, Fourier Transform Raman Spectroscopy, Acad. Press, Ch.1, p. 131 (1994). 25 Connes, J. Rev. Opt. Theor. Instrum., 40, 45 (1961). 173
- 27. Appendix D focusing the 488 nm laser beam ;2 mm deep into the of California, Lawrence Livermore National Laboratory under contract # W-7405-Eng-48. bulk of the crystals to avoid contributions from potential depletion layers. As shown in Fig. 1b, the dependence of the Raman shift on the degree of deuteration is almost 1. J. J. De Yoreo, A. K. Burnham, and P. K. Whitman, Int. Mat. Rev. perfectly linear and ts very well with D 5 22.684cm*R 47, 113 (2002). 1 2452.6, where D is the degree of deuteration (in %) 2. C. E. Barker, R. A. Sacks, B. M. Van Wonterghern, J. A. Caird, J. and R is the spectral m ean of the PO 4 vibration in cm 2 1. R. Murray, J. H. Campbell, K. Kyle, R. B. Ehrlich, and N. D. Nielsen, Proc. SPIE-Int. Soc. Opt. Eng. 2633, 501 (1995). A linear correlation coef cient of 0.998 indicates an ex- 3. T. Suratwala, paper to be published. cellent linear dependence of the Raman peak shift with 4. Cleveland Crystals, Inc., http://www.clevelandcr ystals.com / degree of deuteration. This result shows that the shift of KDP.shtml#table. the PO 4 peak is simply caused by the linear increase in 5. E. A. Popova, I. T. Savatinova, and I. A. Velichko, Sov. Phys. Solid State 12, 1543 (1971). atomic mass due to isotope substitution, which decreases 6. I. P. Kaminow, R. C. C. Leite, and S. P. S. Porto, J. Phys. Chem. the length of hydrogen-like bonds. Solids 26, 2085 (1965). This excellent linear dependence allows us to map the 7. J. A. Subramony, B. J. M arquardt, J. W. Macklin, and B. Kahr, pro le of the D/H exchange layer at the surface of DKDP Chem. M at. 11, 1312 (1999). crystals by acquiring Raman spectra and determining the 8. H. Tanaka, M . Tokunaga, and I. Tatsuzaki, Solid State Commun. 49, 153 (1984). position of the PO 4 peak for various depths. This method 9. R. J. Nelmes, G. M. Meyer, and J. E. Tibballs, J. Phys. C 15, 59 is preferable over other m ethods such as determining the (1982). strength of the OD vibration directly (e.g., at 715 cm 2 1), 10. M . A. Yakshin, D. W. Kim, Y. S. Kim, Y. Y. Broslavets, O. E. because the position of the m ost intense peak in the Ra- Sidoryuk, and S. Goldstein, Laser Physics 7, 941 (1997). 11. I. Takenaga, Y. Tominaga, S. Endo, and M. Kobayashi, Solid State man spectrum can be measured more precisely than the Commun. 84, 931 (1992). intensity of some of the weakest peaks in the spectrum. 12. C. Krenn, personal communication. This is demonstrated in Fig. 2, where depth-dependent 13. M . Sharon and A. K. Kalia, J. Solid State Chem. 21, 171 (1977). Raman spectra (Fig. 2a) and the resulting exchange layer pro les for two DKDP crystals are shown (Fig. 2b). The spectra in Fig. 2a were obtained from a depth scan of a DKDP cr ystal with 75% degree of deuteration in the bulk, grown at 45 8C. The spectra start out as DKDP with ;30% deuteration close to the surface and approach the Rapid Dipicolinic Acid Extraction bulk DKDP spectrum within a few m icrometers of depth. The fact that the relative degree of deuteration does not from Bacillus Spores Detected extend to 0% D is due to the limited depth resolution of by Surface-Enhanced the Raman m icroprobe, which averages over ;4 mm in depth. Figure 2b depicts the resulting D/H exchange layer Raman Spectroscopy pro les for this and a second crystal grown at 63 8C, respectively. Both crystals had the same exposure to am- STUART FARQUHARSON* ALAN bient conditions and their m ain difference is the temper- D. GIFT, PAUL M AKSYM IUK, and ature at which they were grown. The different exchange FRANK E. INSCORE layer pro les indicate that crystals grown at different Real-Time Analyzers, Inc., East Hartford, Connecticut 06108 tem peratu re s hav e differen t proton co ndu ctiv ities, 1 3 which leads to a difference in their rate of deuterium depletion. The param eters controlling this behavior are currently the objective of a detailed study, the results of Index Headings: Dipicolinic acid; Bacillus spores ; Anthrax; Surface- which will be reported elsewhere. enhanced Raman spectroscopy. CONCLUSION In conclusion, we have shown that the shift of the to- INT RODUCTIO N tally symm etric PO 4 stretch mode in the Raman spectrum The anxiety caused by the distribution of anthrax en- of DKDP crystals scales linearly with degree of deuter- dospores through the U.S. postal system in October 2001 ation. This allows us to correlate Raman peak positions was exacerbated by the long time required for positive to deuteration levels in these crystals. We have presented identi cation of the Bacillus anthracis spores and the un- a new technique to determine D/H diffusion pro les in known extent of their distribution. Since that time, many DKDP frequency conversion crystals based on micro-Ra- methods capable of rapid eld analysis have been inves- man spectroscopy. This technique is fast, inexpensive, tigated to augment or replace the laboratory method of and w orks u nder vario us env ironm en tal co nd itio ns, growing microorganisms in culture media, which takes which will allow us to better understand and control deu- days to perform. 1,2 Prominent among these approaches terium depletion in DKDP cr ystals. are polymerase chain reactions (PCR), 3 imm unoassays, and detection of calcium dipicolinate as a biochemical ACK NOW LEDGM ENTS signature. PCR employs primers to separate organism- We would like to thank M. Runkel for rst discovering DKDP crack - ing, R. Floyd for providing DKDP cr ystals, and L. Chase and A. Burn- ham for their support and helpful discussions. This work was performed Received 10 October 2003; accepted 14 November 2003. under the auspices of the U.S. Department of Energy by the University * Author to whom correspondence should be sent. APPLIED SPECTROSCOPY 351
- 28. speci c nucleic acid sequences (e.g., capsular protein en- were separated and weighed at 5 to 15 mg, representing coding gene for Bacillus anthracis), 4 and polymerases to 0.5 to 1.5 million spores. The sample masses were con- amplify the segment until it is detectable. Recently, am- sistent with a previous determination of spore density at pli cation times have been substantially reduced, and 0.081 g/mL that indicated a high degree of entrained air. complete analysis can now be performed in an hour or All chemicals used to prepare the silver-doped sol-gel less. Immunoassay methods are also being developed that coated capillaries were also obtained and used as received use competitive binding of the bioagent (as an antigen) from Sigma-Aldrich. According to previously published and its labeled conjugate for a limited number of anti- procedures,17 two precursor solutions were prepared, bodies. Although analyses can be performed in under 30 mixed, and then drawn into 1-mm-diameter glass capil- minutes, a well-de ned anthrax antigen has not yet been laries. The silver amine precursor consisted of a 5/1 v/v identi ed,5–7 and consequently, the false-positive rate is ratio of 1 N AgNO 3 to 28% NH 3OH, while the alkoxide unacceptably high.8 precursor consisted of a 2/1 v/v ratio of methanol to te- A number of other m ethods are being developed with tramethyl orthosilicate. The alkoxide precursors were a focus on the detection of calcium dipicolinate (CaDPA) mixed with silver amine precursor in an 8/1 v/v ratio. and its derivatives as a B. anthracis signature. This is so Approximately 0.15 m L was drawn into the capillary, because only spore-form ing bacteria contain CaDPA and coating a 15-mm length. After sol-gel formation, the in- the most common potentially interfering spores, such as corporated silver ions were reduced with dilute sodium pollen and mold spores, do not. Relatively fast methods borohydride, which was followed by a water wash to re- have been developed to chemically extract CaDPA and move residual reducing agent. then detect it directly by uorescence 9 or indirectly by A 100 mL drop of a 50 mM DDA solution in ethanol, luminescence. 10,11 In the latter case, hot dodecylamine pre-heated to 78 8C, was added to each of the B. cereus (DDA) has been used to extract dipicolinic acid (DPA), particles to digest the spore coat. After 1 minute the re- and terbium has been utilized to form a highly lumines- sultant solution was drawn into a SER-active capillary cent DPA complex.11 Although m easurem ents have been that was immediately xed horizontally to an XY posi- performed in as little as ve minutes, it was found that tioning stage (Conix Research, Spring eld, OR) just in- as many as three concentration-dependent complexes can side the focal point of an f /0.7 aspheric lens. The lens form, each with different lifetimes. This, coupled with focused the beam into the sample and collected the scat- the fact that the Tb 31 cation produces the same lumines- tered radiation back along the same axis. A dichroic lter cence spectrum, m akes determinations of low spore con- (Omega Optical, Brattleborough, VT) was used to re ect centrations problematic. the excitation laser to the lens and pass the Raman scat- It has been long known that Raman spectra of Bacilli tered radiation collected by the lens. An f /2 achromat spores are dominated by bands associated with CaDPA 12 was used to collimate the laser beam exiting a 200-mm- and that these spectra may provide a suitable anthrax sig- core-diameter source ber optic, while a second f /2 ach- nature at the genus level.13 Since that time considerable romat was used to focus the scattered radiation into a 365 improvements in Raman instrumentation have led to lab- mm ber optic (Spectran, Avon, CT). A short-pass lter oratory m easurements of single Bacilli spores 14 and to was placed in the excitation beam path to block the sil- eld m easurements of spores captured from a m ail-sort- icon Raman scattering generated in the source ber from ing system.15 However, the single spore measurements re ecting off sampling optics and reaching the detector. required complex instrum entation that is not rugged, A long-pass lter was placed in the collection beam path while the eld measurem ents required milligram s of sam- to block the sample Rayleigh scattering from reaching ple. Furtherm ore, the Raman spectra of both measure- the detector. A 785 nm diode laser (Process Instruments ments contained uorescence contributions that would in- Inc., m odel 785-600, Salt Lake City, UT) was used to crease uncertainty in quanti cation. deliver 100 to 150 m W of power to the sample. A Fourier In related research, we demonstrated that nanogram transform Raman spectrometer (Real-Time Analyzers, quantities of DPA could be detected by uorescence-free, model IRA-785, East Hartford, CT) and a silicon photo- surface-enhanced Raman spectroscopy (SERS). 16 We also avalanche detector (Perkin Elmer model C30902S, Stam- demonstrated that m icroliter volumes of chemicals can ford, CT) were used to acquire the SER spectra. be detected by SERS using m etal-doped sol-gel-packed glass capillaries.17 Towards the goal of developing a rap- RESULTS AND DISCUSSION id, eld, SERS-based, anthrax spore detector, we have combined our previous research, and we now report that As an initial experiment, the SER spectrum of 1 g/L DPA can be extracted from a 10 mg B. cereus spore sam- of DPA in water was measured using the newly devel- ple using DDA in 1 minute and can be detected by SERS oped silver-doped sol-gel-coated capillaries (Fig. 1A). At in an additional 1 minute. this concentration, a high signal-to-noise ratio (S/N) is obtained in 1 min. In fact, a reasonable spectrum is ob- EXPERIMENTAL tained in the same time frame for 1 mg/L (Fig. 1B). The SER spectra are reasonably similar to the normal Raman Dipicolinic acid (2,6-pyridinedicarboxylic acid, DPA) (NR) spectrum obtained for a saturated solution of DPA and dodecylamine (DDA) were used as received from in 1 N KOH (Fig. 1C), and the following band shifts are Sigma-Aldrich (M ilwaukee, W I). Lyophilized B. cereus observed (NR to SER): 647 to 657 cm 2 1 , 817 to 815 spores, prepared according to the literature,13 were sup- cm 2 1 , 998 to 1008 cm 2 1, 1384 to 1382 cm 2 1, 1434 to plied by the University of Rhode Island and used as re- 1428 cm 2 1, and 1569 to 1567 cm 2 1 . Many of these bands ceived. M ultiple particles, approximately 0.1 mm 3 each, have been previously assigned,12,13 such as 998 cm 2 1 to 352 Volume 58, Number 3, 2004
- 29. F IG . 1. SERS of DPA in water using silver-doped sol-gel-coated glass F IG . 2. SERS of DPA extracted from ;10 mg B. cereus particle using capillary for (A) 1 g/L and (B) 1 mg/L. (C ) NR of saturated DPA in 1 100 mL of 50 m M hot DDA acquired in (A) 1 m inute and (B) 2 seconds. N KOH in a glass capillary. Spectral conditions: (A) and (B), 150 mW (C ) Attempted SERS of 50 mM hot DDA in ethanol using silver-doped of 785 nm, 1-min acquisition time; (C) 450 m W of 785 nm, 5-min sol-gel-coated glass capillary acquired in 1 min. Spectral conditions: acquisition time; both 8 cm 2 1 resolution. 150 mW of 785 nm, 8 cm 2 1 resolution. the symm etric ring stretch, 1384 cm 2 1 to the O–C–O with the appearance of suspicious material or intentional symmetric stretch, 1428 cm 2 1 to the symmetric ring C– mailing of comm on substances as an anthrax hoax. This H bend, and 1569 cm 2 1 to the asymmetric O–C–O method could also prove useful in detecting the location stretch. of anthrax endospores in mail distribution facilities if an- The rst B. cereus samples consisted of 2 m g of spores other veri ed attack should occur. in 2 m L of 5 mM hot DDA. The samples were main- Research continues to fully characterize the surface- tained at 78 8C for 40 min, and while hot, approximately enhanced Raman spectroscopy signal intensities as a 10 mL was drawn into a SER-active capillary. Since spec- function of sample concentration and to explore other tra of DPA were obtained for these initial samples, small- extractants that do not require the use of elevated tem- er spore masses, higher DDA concentrations, and shorter perature. heating periods were examined. In due course it was found that 10 mg of spores could be digested by 100 mL of 50 m M hot DDA in one minute and detected (Fig. ACK NOW LEDGM ENTS 2A). In fact the signal was suf ciently intense that it can The authors are grateful for the support of the National Science Foun- be observed in as little as two seconds (Fig. 2B). The dation (DM I-0296116 and DM I-0215819) and the U.S. Arm y (DAA D13-02-C-0015, Joint Service Agent Water M onitor program). amount of DPA that was extracted was estimated to be The authors are indebted to Chetan Shende for preparing the sol-gel between 5 and 10 mg/L by comparing the signal intensity capillaries. The authors also thank James Gillespie, Nicholas Fell, and of the 1008 cm 2 1 band to that m easured for DPA in water. Augustus Fountain for providing important background information, This is consistent with previous research that found that Mark Farquharson for laboratory support, and Professor Jay Sperry of the m ajority of the DPA is extracted from spores using the University of Rhode Island for supplying B. cereus spores. DDA 11 and that B. cereus spores contain approximately 10% DPA by weight.18 The S/N of 127 for the 1008 cm 2 1 1. V. A. Pasechnik, C. C. Shone, and P. Hambleton, Bioseparations 3, band in the 1-minute SER spectrum suggests a limit of 267 (1993). detection of approximately 250 ng of B. cereus spores 2. P. J. Jackson, M. E. Hugh-Jones, D. M. Adair, G. Green, K. K. Hill, based on a S/N of 3. Finally, it should be noted that DDA C. R. Kuske, L. M. Grinberg, F. A. Abramova, and P. Keim, Proc. did not produce a detectable SER spectrum, as shown in Natl. Acad. Sci. U.S.A. 95, 1224 (1998). Fig. 2C. 3. B. R. Glick and J. J. Pasternak, M olecular Biology: Principles and Applications of Recombinant DNA (ASM Press, Washington, D.C., 1994). CONCLUSION 4. C. A. Bell, J. R. Uhl, T. L. Had eld, J. C. David, R. F. Meyer, T. F. Smith, and F. R. Cockerill, III, J. Clin. Microbiol. 40, 2897 We have demonstrated that by combining rapid extrac- (2002). tion of dipicolinic acid from Bacillus cereus spores with 5. D. L. Gatto-Menking, H. Yu, J. G. Bruno, M. T. Goode, M. Miller, chemical identi cation by surface-enhanced Raman spec- and A. W. Zulich, Biosens. Bioelectron. 10, 501 (1995). 6. J. J. Quinlan and P. M. Foegeding, J. Rapid Methods Automation troscopy, as little as 10 mg of spores can be detected. In M icrobiol. 6, 1 (1998). fact, the entire measurement, from the time of adding hot 7. A. A. Hindle and E. A. H. Hall, Analyst (Cambridge, U.K.) 124, dodecylamine to the spores to the time when the dipi- 1599 (1999). colinic acid SER spectrum is acquired and analyzed, 8. M . S. Ascher, US Department of Health & Human Services could be performed in less than two m inutes. The ability (http://www.hhs.gov/ophp/presentations/Ascher.doc). 9. R. Nudelman, B. V. Bronk, and S. Efrima, Appl. Spectrosc. 54, 445 of this method to distinguish between spore-form ing bac- (2000). teria, such as Bacillus anthracis, and non-DPA containing 10. D. L Rosen, C. Sharpless, and L. B. McBrown, Anal. Chem. 69, powders could help prevent costly shutdowns associated 1082 (1997). APPLIED SPECTROSCOPY 353
- 30. 11. P. M. Pellegrino, N. F. Fell, Jr., and J. B. Gillespie, Anal. Chim. 15. S. Farquharson, L. Grigely, V. Khitrov, W. W. Smith, J. F. Sperry, Acta 455, 167 (2002). and G. Fenerty, J. Raman Spectrosc., paper accep ted (2003). 12. W. H. Woodruff, T. G. Spiro, and C. Gilvarg, Biochem. Biophys. 16. S. Farquharson, W. W. Smith, S. Elliott, and J. F. Sperry, SPIE-Int. Res. Commun. 58, 197 (1974). Soc. Opt. Eng. 3855, 110 (1999). 13. E. Ghiamati, R. S. Manoharan, W. H. Nelson, and J. F. Sperry, 17. S. Farquharson and P. Maksymiuk, Appl. Spectrosc. 57, 479 Appl. Spectrosc. 46, 357 (1992). (2003). 14. A. P. Esposito, C. E. Talley, T. Huser, C. W. Hollars, C. M. Schal- 18. F. W. Janssen, A. J. Lund, and L. E. Anderson, Science (Washing- dach, and S. M . Lane, Appl. Spectrosc. 57, 868 (2003). ton, D.C.) 127, 26 (1958). 354 Volume 58, Number 3, 2004
- 31. Appendix E pH dependence of methyl phosphonic acid, dipicolinic acid, and cyanide by surface-enhanced Raman spectroscopy Stuart Farquharson, Alan Gift, Paul Maksymiuk, Frank Inscore and Wayne Smith Real-Time Analyzers, 87 Church Street, East Hartford, CT 06108 ABSTRACT U.S. and Coalition forces fighting terrorism in Afghanistan and Iraq must consider a wide range of attack scenarios in addition to car bombings. Among these is the intentional poisoning of water supplies to obstruct military operations. To counter such attacks, the military is developing portable analyzers that can identify and quantify potential chemical agents in water supplies at microgram per liter concentrations within 10 minutes. To aid this effort we have been investigating the value of a surface-enhanced Raman spectroscopy based portable analyzer. In particular we have been developing silver-doped sol-gels to generate SER spectra of chemical agents and their hydrolysis products. Here we present SER spectra of methyl phosphonic acid and cyanide as a function of pH, an important factor affecting quantitation measurements, which to our knowledge has not been examined. In addition, dipicolinic acid, a chemical signature associated with anthrax-causing spores, is also presented. Keywords: Chemical warfare agents, agent detection, agent hydrolysis, SERS, Raman spectroscopy, homeland security 1. INTRODUCTION In the past decade, the Unites States and its allies have been challenged by a different kind of warfare, exemplified by the terrorist attacks of September 11, 2001. Suicide bombings and the use of chemical agents are the norm, and military personnel must consider a wide range of attack scenarios. Among these is the intentional poisoning of water supplies to obstruct military operations in Afghanistan and Iraq. Currently, colorimetric paper is used to detect agents on-site, while gas chromatography combined with mass spectrometry (GC/MS) is used in mobile support laboratories. However both methods have severe drawbacks. The paper changes color in response to contact with many chemicals besides CWAs, causing a high incidence of false positives,1 while GC/MS, although very chemically specific, requires hours to perform and constant re-calibration.2,3,4 Military operations would be greatly aided by a portable analyzer that can identify and quantify potential chemical agents at concentrations that impact safety. This includes the analysis of drinking water supplies, distribution and storage systems. To meet this goal, the Department of Defense has been investigating numerous approaches under the auspices of the Joint Service Agent Water Monitor (JSAWM) program.5 The JSAWM must be capable of identifying the chemical agents with no false-positives and quantifying the agents at microgram per liter concentrations within 10 minutes.6,7,8 Recently, we and others have been investigating the ability of surface-enhanced Raman spectroscopy (SERS) to measure chemical agents,9-12 bioagents,13-17 and their hydrolysis products in water. SERS employs the interaction of surface plasmon modes of metal particles with target analytes to increase scattering efficiency by as much as 1 million times.18 In 1987, the potential of this approach to measure CWAs was demonstrated by measuring a series of pesticides.19 Several of these organophosphonates have chemical structures similar to CWAs, in particular P=O functional groups. In our studies, we have been employing silver-doped sol-gels to promote the SER effect. The porous silica network of the sol-gel matrix offers a unique environment for immobilizing and stabilizing SER-active metal particles.20-23 The sol-gels have been coated on the internal walls of standard glass vials to allow rapid SER analysis of small sample quantities without preparation. We have measured over 100 chemicals with enhancements of 104 to 106, demonstrated reversible measurements in a flowing system, reproducible measurements from vial-to-vial, and measurements in multiple solvents, including water.20-23 Previously, we used these vials to perform preliminary measurements of cyanide (CN), methylphosphonic acid (MPA), and dipicolinic acid (DPA). MPA is a hydrolysis product of the nerve agents (e.g. sarin, Reaction 1) and may be a valuable indicator of nerve agent usage, particularly since the alkyl methylphosphonic acids are relatively more stable than their corresponding parent complexes.24 DPA is SPIE -2003-5269 117
- 32. a chemical signature of spore forming bacteria, such as Bacillus anthracis. And in light of the inability to rapidly detect the anthrax spores distributed through the U.S. mail in October, 2001, a number of methods are being developed to extract and analyze this signature. O CH3 O CH3 O CH3 P C + H 2O HF + P C P + C H3C O CH3 H3C O CH3 H3C OH HO CH3 F OH OH Sarin IMPA MPA 2-propanol Reaction 1. Stepwise hydrolysis of Sarin to form hydrofluoric acid (HF), isopropyl methylphosphonic acid (IMPA), then methyl phosphonic acid (MPA) and 2-propanol. In our previous SERS investigations, MPA and DPA were measured at 50 and 100 mg/L, respectively. In both cases, limits of detection (LOD) were estimated at 100 µg/L providing encouragement in that SERS may satisfy the needs of the JSAWM. Since it has been shown that pH can substantially influence the intensity of SER bands,25 which would clearly influence quantitative analysis, we undertook the present study to determine the severity of these effects for cyanide, methyl phosphonic acid, and dipicolinic acid. Furthermore, we previously observed a band at 1050 cm-1 for MPA,6 possibly due to an anion formed at basic pH. Here we investigate the source of this spectral anomaly. 2. EXPERIMENTAL All chemicals, including potassium cyanide, methyl phosphonic acid, dipicolinic acid, and those used to prepare the silver-doped and gold-doped sol-gel coated vials were obtained and used as received from Sigma-Aldrich. All samples were prepared in HPLC grade water (Fischer Scientific, Fair Lawn, NJ) for SERS measurements. The pH of these samples was adjusted using dilute nitric acid or potassium hydroxide, and verified using a pH electrode (Corning Inc., Corning, NY) that had been calibrated with pH 4.00, 7.00, and 10.00 buffered standards from Fischer Scientific. Once prepared, the samples were transferred into the silver-doped sol-gel vials (Real-Time Analyzers, Inc., East Hartford, CT). The vials were coated in a manner similar to that previously reported by adding ammonium hydroxide to a solution of silver nitrate, tetramethyl orthosilicate (TMOS), and methanol.20 Gold-vials were coated by adding nitric acid to a solution of gold tetrachloride, TMOS and methanol. The two precursor solutions were prepared, mixed, and transferred to 2-mL glass vials, dried and heated. After sol-gel formation, the incorporated metal ions were reduced with dilute sodium borohydride (1mg/mL), which was followed by a water wash to remove residual reducing agent. After the resultant analyte solution was introduced, the SER-active vial was immediately fixed horizontally to an XY positioning stage (Conix Research, Springfield, OR) just inside the focal point of an f/0.7 aspheric lens. The lens focused the beam into the sample and collected the scattered radiation back along the same axis. A dichroic filter (Omega Optical, Brattleborough, VT) was used to reflect the excitation laser to the lens and pass the Raman scattered radiation collected by the lens. An f/2 achromat was used to collimate the laser beam exiting a 200 µm core diameter source fiber optic, while a second f/2 achromat was used to focus the scattered radiation into a 365 µm fiber optic (Spectran, Avon, CT). A short pass filter was placed in the excitation beam path to block the silicon Raman scattering generated in the source fiber from reflecting off sampling optics and reaching the detector. A long pass filter was placed in the collection beam path to block the sample Rayleigh scattering from reaching the detector. A 785 nm diode laser (Process Instruments Inc. model 785-600, Salt Lake City, UT) was used to deliver 100 to 150 mW of power to the sample. A Fourier transform Raman spectrometer (Real-Time Analyzers, model IRA-785, East Hartford, CT), and a silicon photo-avalanche detector (Perkin Elmer model C30902S, Stamford, CT) were used to acquire the SER spectra. 3. RESULTS AND DISCUSSION In a previous study of MPA,6 aimed at developing a concentration calibration curve and determining limits of detection (LOD), we observed an anomaly at 1050 cm-1. Since it was found that the band intensity changed as a function of concentration, the band must be associated with a sample parameter. Two possible parameters, photon flux and pH, are examined here. The first parameter was investigated by irradiating a 1mg/mL MPA sample in a SER-active vial with laser powers of 200 mW and above and monitoring spectral changes. It was immediately found that the 1050 cm-1 band SPIE -2003-5269 118
- 33. appeared and grew as a function of time, and that the higher the laser power the faster the growth. Figure 1 shows the growth of the 1050 cm-1 band over the course of 30 minutes when using 150 mW of 785 nm excitation, while Figure 2A shows that the growth can be fit with a first order exponential equation, namely I1050 = 0.3+0.5e-0.13t. Furthermore, the 760 cm-1 band could be fit with a first order decay equation with an identical exponential rate constant, i.e. I760 = 0.8- 0.8e-0.13t. The rates represent classical first order kinetics and their correspondence allows one to conclude that MPA is being transformed one-for-one into a photo-generated product. At this time the photoproduct has not been positively identified, but phosphonic (phosphorous) acid and phosphonate are likely candidates, since the symmetric P(OH)3 stretch occurs at ~1050cm-1. Our studies show that a reduction in laser power to 100 mW at the sample for MPA essentially eliminates this degradation process. This laser power or lower was used for further measurements. MPA Photodegradation A B Raman Intensity (relative) 0 Arbitrary Units min 30 600 570 800 770 1000 970 1200 1170 1400 1370 0 10 20 30 0 10 20 30 Raman Shift (cm-1) time (min) time (min) Wavenumber (cm-1) Figure 1. Growth of 1050 cm-1 band as a function of time Figure 2. A) Exponential growth of 1050 cm-1 band and B) due to exposure to 150 mW of 785 nm. Spectra are 5 sec exponential decay of 760 cm-1 band for spectral series in Figure 1. each, collected every 100-sec from 0 to 30-min. Methyl phosphonic acid is a diprotic acid that stepwise dissociates into two anions, MPA- and MPA=, according to the following reactions:26 MPA MPA- + H+ pKa1 = 2.12 Reaction 2 MPA- MPA= + H+ pKa2 = 7.29 Reaction 3 The relative concentrations of MPA, MPA-, and MPA=, can be determined at any pH by expressing [MPA] and [MPA=] in terms of [MPA-] using Reactions 2 and 3, and summing all three to equal the total starting concentration, here 2 mg/mL (0.021M, MW = 96.02), viz: [MPA] + [MPA-] + [MPA=] = 0.021M Equation 1 substituting from Reactions 1 and 2: ([H+][MPA-])/K1a + [MPA-] + (K2a[MPA-])/[H+] = 0.021M Equation 2 rearranging: [MPA-] = 0.021M/(1+[H+]/K1a + K2a/[H+]) Equation 3 The relative concentrations of MPA, MPA- and MPA= as a function of pH are shown in Figure 3. It is worth noting that near neutral pH both MPA- and MPA= will be present. To confirm that the SER signal followed this pH dependence, a starting solution, consisting of 20 mg of MPA in 10 mL HPLC grade water was prepared and brought to pH of 2.0 using dilute nitric acid. From this solution, 2 mL were added to a SER-active vial, and the SER spectrum recorded. At this pH a peak at 760 cm-1 was barely discernable. The 2 mL solution was returned to the starting solution and the pH was re-measured to correct for any changes that the silver-doped sol-gel vials might cause. In most cases the change was less than 0.2 pH units, and the pH is reported as the before and after average. Next, the pH of the SPIE -2003-5269 119
- 34. starting solution was adjusted to 3.25 using dilute KOH. Again 2 mL were added to a vial and the SER spectrum recorded. At this pH a reasonably strong 760 cm-1 band was observed. This process was repeated as spectra were recorded at pHs of 7.0, 7.4, 7.5, 7.9, 8.5, and 10.0. A total of 1 mL of KOH was added, diluting the total concentration by 10%. Next, the pH of the starting solution was made acidic by adding dilute nitric acid dropwise. This time spectra were recorded at pHs of 7.2, 6.9, 6.4, and 3.7. Figure 4 shows the SER spectra for representative pHs (spectra were left out to simplify the figure), while Figure 3 shows the 760 cm-1 peak intensities as a function of pH. (The band intensities were adjusted to compensate for dilution effects caused by the addition of HNO3 and KOH, then normalized to 0.021 M for the most intense band observed at pH 3.7.) It is clear from Figure 3, that the 760 cm-1 band follows the MPA- concentration as a function of pH and must be assigned to this anion. No bands were observed that corresponded to MPA or MPA=. The lack of an MPA SER spectrum may be due to the absence of an attraction between the neutral analyte and the electropositive silver surface. The same reasoning suggests that a strong SER spectrum should be observed for MPA=, but it is not, and a satisfactory explanation has not been found. 0.020 - = MPA MPA MPA Concentration [M] 0.015 pK1 = 2.12 pK2 = 7.29 0.010 0.005 0.000 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 pH Figure 3. Concentration dependence of MPA, MPA-, and MPA= Figure 4. SER spectra of 0.02M MPA as a function of pH. as a function of pH for a 0.02M sample. Intensity of 760 cm-1 Conditions: 100 mW of 785 nm, 36 scans (1 min), 8 cm-1, band from Figure 6 as a function of pH. (■ for increasing basic recorded 2 min after sample introduction. pH 1.9 and 2.0 not adjustment, ● for increasing acidic adjustment, error was measured apparent on this scale, pH 6.9 and 7.4 near identical to 7.0 and at ~10% for pH 6.4). 7.5, and not shown for clarity. Since these measurements involved the removal and 1.8 Raman Intensity (760 cm-1) replacement of the SER-active vial in the sample holder 1.6 to remove and add sample, variation in the intensity as a 1.4 function of vial position was minimized by illuminating 1.2 the exact same height along the vial wall. But this does 1.0 not account for variability of the SERS response of the sol-gel coating around the vial. To analyze this effect, a 0.8 vial containing MPA at pH 6.4 was rotated at ~ 40o 0.6 intervals at the original height, and 1/8” above and below 0.4 this value. Figure 5 shows the intensity of the 760 cm-1 0.2 band for the 27 positions. It was found that the average 0.0 value was 1.37±0.14, an RSD of 10% overall and 5% for 0 5 10 15 20 25 30 each height. An error bar is included in Figure 4 for the Measurement Number pH 6.4 measurement. Figure 5. SER spectra of 0.02M MPA at pH 6.4 measured around a vial at three heights (9 points per height). Conditions as in Figure 4, but 10-sec scans. SPIE -2003-5269 120
- 35. With the above analyses in mind, a preliminary investigation of the SER spectral response for MPA (as MPA-) as a function of concentration was performed. A single vial was used for these measurements, beginning with 1 mg/L, followed by measurements of 10, 100 and 1000 mg/L. In all cases the pH was ~7, and 3 positions around the vial were measured per concentration. Since the 760 cm-1 band was not observed for concentrations of 1 or 10 mg/L using 100 mW of 785 nm, the laser power at the sample was raised to 200 mW beginning with the 10 mg/L concentration. Photo- degradation was largely avoided (and not observed) by exposing the sample for only 33 seconds per spectral acquisition. Representative spectra for 10, 100, and 1000 mg/L are shown in Figure 6, while a plot of the 760 cm-1 band intensity as a function of concentration is shown in Figure 7. These values were also used to estimate limits of detection based on the signal-to-noise ratio (S/N) of the 760 cm-1 band using 100 mW of 785 nm laser excitation and a 10-min acquisition time. As summarized in the Figure 7 inset, the lower the measured concentration, the lower the predicted LOD. Even if, as estimated, 210 µg/L could be measured using the silver-doped sol-gel vials, an improvement of a factor of 70 is still required to meet JSAWM goals of measuring 3 µg/L in 10-minutes. 7 6 760 Band Intensity 5 4 3 conc (mg/L) LOD ave stddev %dev 2 A 10 0.21 0.02 10.21 B 100 0.72 0.10 14.30 1 1000 3.12 0.40 12.80 C 0 0 200 400 600 800 1000 1200 MPA Concentration (mg/L) Figure 6. SER spectra of MPA in water at A) 1000, B) 100, Figure 7. Plot of SER intensity of 760 cm-1 band of MPA as a and C) 10 mg/L. Conditions: pH of 7, silver-doped sol-gel function of concentration using 200 mW of 785 nm. Inset table coated vial, 200 mW of 785 nm, 33-sec, 8 cm-1 resolution. includes average intensity, LOD, standard deviation and percent deviation for each concentration, but for 100 mW and 10 min. Similar to MPA, DPA is a diprotic acid (pKa1 = 2.16 and pKa2 = 6.92) and variations in pH will effect the relative concentrations of DPA, DPA-, and DPA=, and possibly the SER spectra and band intensities. This could prove significant if an acid or base is used to denature anthrax spores with the goal of extracting and analyzing DPA. The MPA pH study described above was mimicked for DPA, except that the starting solution consisted of 20 mg of DPA in 20 mL HPLC grade water (6.0x10-3M, MW = 167.1). The initial solution had a pH of 2.45, which was made basic by dilute KOH to pHs of 3.55, 4.33, 4.87, 5.59, 10.69 and 11.66. SER spectra were recorded at each pH using 100 mW of 785 nm and a 44-sec acquisition time. Next, one drop of concentrated nitric acid was used to remake the solution acidic at a pH of 2.00. Again sequential pH measurements were performed at 3.83, 5.10, 7.35 and 8.22. The solution pH was made acidic a third time, but to pH 2.19, 1.71, then 1.35. Throughout this process, no more than 20 drops of acid or base were added, and therefore the concentration was diluted by no more than 10%. Most of the spectral bands showed a minor decrease in intensity as a function of increasing pH values. However, the bands at 525 and 795 cm-1 showed the most dramatic changes, which occurred at acid pH. Figure 8 shows the SER spectra of DPA for the spectral region and pH range of interest. The identity of the DPA species was determined by plotting the normalized peak intensities, with the lowest value set to 0 and the highest to 0.006 M, as a function of pH, and overlaying these values on a plot of the relative concentrations for DPA, DPA-, and DPA=, as previously done for MPA (Figure 9). As can be seen the 525 cm-1 band clearly corresponds to DPA. The correspondence of the 795 cm-1 band to this species is less clear, as the band retains intensity until pH of 5.5. This can be attributed to contribution to the overlapping band at 810 cm-1, which does not change as a function of pH. The fact that most bands are observed at all pHs suggest that the primary interaction with silver is through the ring nitrogen. This is supported by the fact that the most intense band occurs at 1008 cm-1, attributed to a symmetric ring breathing mode, and that this interaction has been characterized for pyridine in numerous papers.27 SPIE -2003-5269 121
- 36. 0.007 DPA DPA- DPA= pH 795 0.006 Concentration [M] 525 0.005 1.35 DPA 1.71 0.004 DPA- 2.19 0.003 pK1 = 2.16 pK2 = 6.92 DPA= 3.83 795 0.002 525 0.001 0 0 2 4 6 8 10 12 14 pH Figure 8. SER spectra of 1 mg/mL DPA as a function of pH. Figure 9. Concentration dependence of DPA, DPA-, and Conditions: 100 mW 785 nm, 100 scans (44-sec). DPA= as a function of pH for a 0.006M sample. Intensity of 525 (■) and 795 (♦) cm-1 bands from Fig. 8 as a function of pH. The ability to detect dipicolinic acid (DPA) by SERS is demonstrated in Figure 10. Even at 1 mg/L the primary bands are visible. The intensity of the dominant ring breathing mode observed at 1008 cm-1 for DPA in water (pH = 2.5 to 5.5) is plotted as a function of concentration in Figure 11. These values were also used to estimate limits of detection based on the S/N of the 1008 cm-1 band using 100 mW of 785 nm laser excitation and a 10-min acquisition time. Again, the lower the measured concentration, the lower the predicted LOD (see Figure 11 inset), and detection of 160 µg/L is possible. Gastrointestinal anthrax requires significant more spores than inhalation anthrax,28 and a limit of detection might be placed at 1 million spores in 1 liter of water or 10 µg/L. Since each spore contains ~10% CaDPA by weight,29 a goal for DPA might be 1 µg/L, indicating that the present measurements must be improved by nearly two orders of magnitude. 2.5 2 1008 Band Intensity 1.5 A 1 conc (mg/L) lod-10min-100mw B 1 0.17 C 10 0.16 0.5 100 1.03 D 1000 3.55 0 0 200 400 600 800 1000 1200 DPA Concentration (mg/L) Figure 10. SER spectra of DPA in water at A) 1000, B) 100, Figure 11. Plot of SER intensity of 1008 cm-1 band of DPA C) 10 and D) 1 mg/L. Conditions: pH of 2.5-5.5, silver-doped as a function of concentration using 175 mW of 785 nm. sol-gel coated vial, 175 mW of 785 nm, 1-min, 8 cm-1. D) has Inset table includes LOD in mg/L for each concentration, but been multiplied by x10 to make bands visible. for 100 mW and 10 min. Due to the increased hazards of handling HCN gas, KCN salt was used for these experiments. Nevertheless, all sample preparations were performed in a chemical hood. KCN completely dissolves in water, but its conjugate acid, HCN, is formed and has a Ka of 6.15x10-10,30 viz: HCN CN- + H+ pKa = 9.21 Reaction 4 SPIE -2003-5269 122
- 37. Consequently, the cyanide concentration must be determined for each initial KCN concentration. Specifically, the samples prepared with concentrations of 0.1, 1, 10, 100, and 1000 mg/L of KCN produced CN- concentrations of 6.3x10-3, 0.33, 6.9, 89, and 964 mg/L, at pHs of 8.16, 9.0, 9.67, 10.2, and 10.7, respectively. The pH dependence for the HCN and CN- concentrations are shown in Figure 12. Thus, as the amount of KCN added to the solution decreases so does the pH of the solution (becomes less basic) and according to Reaction 4, the relative amount of CN- to HCN also decreases. For example, in the preparation of a 0.1 mg/L solution of KCN, the pH is shifted from 7 for pure water to only 8.16, and only 6.3% of the starting material becomes CN-, or 6.3x10-3 mg/L. In comparison, for a solution of 1000 mg/L, the pH is shifted from 7 to 10.7 and 96% of the starting material becomes CN-. This is significant since the cyanide ion is better able to adsorb onto the silver particles and become SERS active. SER spectra of 10, 100, and 1000 mg/L of KCN samples are shown in Figure 13 and a corresponding plot of the 2100 cm-1 band intensity as a function of concentration is shown in Figure 14. 1.1 1.0 Concentration [mg/mL] 0.9 - HCN CN 0.8 0.7 0.6 0.5 pKa = 9.21 0.4 A 0.3 0.2 B 0.1 C 0.0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 pH Figure 12. Concentration dependence of HCN and CN- as a Figure 13. SER spectra of KCN in water at A) 1000, B) 100, function of pH for a 1 mg/mL sample. Calculated intensity and C) 10 mg/L. Conditions: pHs of 10.7, 10.2, and 9.7, silver- of 2100 (■) cm-1 band for a 1 mg/ml sample at pHs of 8.16, doped sol-gel coated vial, 100 mW of 785 nm, 1-min, 8 cm-1. 9.0, 9.67, 10.2, and 10.7. C) has been multiplied by x10 to make band visible. 140 120 2100 Band Intensity 100 80 conc (mg/L) lod-10min-100mw Condition 60 0.1 0.01 Au-pH 12 A 1 0.07 Au-pH 12 6.9 0.03 Au-pH 9.7 40 6.9 0.16 Ag-pH 9.7 B 20 89 0.22 Ag-pH 10.2 C 946 1.13 Ag-pH 10.7 0 0 200 400 600 800 1000 1200 CN Concentration (mg/L) Figure 14. Concentration dependence of KCN SERS measured Figure 15. SER spectra of KCN in water at A) 10, B) 1, under conditions in Fig. 11. Concentrations are 1, 0.1, and 0.01 and C) 0.1 mg/L. Conditions: pHs of 9.7, 12, and 12, mg/ml. Intensities are measured for the CN stretch at 2100 cm-1. gold-doped sol-gel coated vial, 100 mW of 785 nm, 1- Inset table includes LOD in mg/L for each concentration in Figs min, 8 cm-1. C) has been multiplied by x10 to make band 13 and 15, but for 100 mW and 10 min. visible. The SER spectra of cyanide are dominated by the single CN stretch at 2100 cm-1, which occurs in normal Raman spectra of solutions at 2080 cm-1. However, a low frequency mode occurs at 135 cm-1 due to a Ag-CN stretch (not shown), indicative of a strong surface interaction. It is also observed that as the concentration decreases the CN stretch shifts to 2140 cm-1. This shift has been attribute to the formation of a tetrahedral Ag(CN)32- surface structure,31 as well SPIE -2003-5269 123
- 38. as to CN adsorbed to two different surface sites.32 Alternatively, the 2140 cm-1 band could be attributed to HCN, since this species dominates at lower concentrations. However, it is unlikely that this species would be attracted to the electropositive silver surface. Further, both peaks should be present at pHs between 8.5 and 10.5, but this is not observed. It has also been suggested that at concentrations near and above monolayer coverage, the CN- species is forced to adsorb end-on due to crowding, and at lower concentrations the molecule can reorient to lie flat.33 This suggests that the 2100 and 2140 cm-1 bands correspond to the end-on and flat orientations, respectively. As Figures 13 and 14 show, the intensity of the CN stretch for the 89 mg/L sample is nearly as intense as the 964 mg/mL band. This suggests that the Raman signal for the flat orientation is more enhanced. However, more extensive measurements are required to verify this point. Since resent research has suggested that cyanide may be more effectively detected on gold, measurements of KCN solutions were also performed using gold-doped sol-gel vials. Preliminary measurements are shown in Figure 15 for samples prepared from 0.1, 1, and 10 mg/L KCN. Since the pHs are 8.16, 9.0, 9.67, the resultant CN- concentrations are 6.3x10-3, 0.33, and 6.9 mg/L. Initially, only the highest concentration was observed, and the signal intensity was significantly better than the equivalent concentration measured using silver. In an effort to shift Reaction 4 to the left, transforming HCN to CN- (Le Chatelier’s principle), KOH was added to the lower concentration samples producing solutions with pH 12. This effectively forces all of the cyanide in solution to be CN-, or 0.1 and 1.0 mg/L, respectively. More importantly, the CN stretch is now observed in the SER spectra. The band appears at 2125 cm-1, as has been previously reported for gold.12 As calculated for MPA and DPA, LODs can be estimated from this data. For the three concentrations of cyanide on silver, the LODs are 0.16 to 1.1 mg/L for 100 mW of 785 nm laser excitation and a 10-min acquisition time. For gold, pH adjusted, the LODs are10 to 70 µg/L, an improvement of more than 10 times silver. Nevertheless, either substrate is sufficient to meet the JSAWM goals of measuring 3 mg/L in 10-minutes, as the requirements form cyanide are much less stringent than the nerve agents. 4. CONCLUSIONS Here we examined the surface-enhanced Raman spectral response for methyl phosphonic acid, dipicolinic acid and cyanide as a function of pH. It was determined that the most prominent peak at 760 cm-1 reaches a maximum intensity between pH 3 and 7, and corresponds to the MPA- species. Neither the MPA nor MPA= species appear to generate a SER spectrum, and consequently no spectra were observed below pH 2 or above pH 8. In this study, we also found that higher laser powers could cause photodegradation of MPA, signified by the exponential growth of a band at 1050 cm-1, which is tentatively assigned to phosphorous acid. Unlike MPA, DPA was observed at all pHs. This is attributed to the dominant interaction of the pyridine functional group with silver. Minor spectral changes were observed at acid pHs and were assigned to neutral DPA. Like MPA, SER spectra of cyanide were pH dependent. No spectra were observed for the HCN species, while CN- was best observed at pHs more basic than 8. Preliminary concentration studies for the three analytes, allowed estimating limits of detection for MPA, DPA, and CN using 100 mW of 785 nm and a 10-min acquisition time of 210, 165 and 70 µg/L respectively. Although the latter value suggests that cyanide can be measured at sufficiently low concentrations to meet JSAWM goal, improvements by 100 to 200 times are required for MPA and DPA. It is clear from this study that pH of the sample is important and must be taken into account when developing concentration calibration curves as well as fieldable analyzers. Future work will include tailoring the sol-gel with specific functional groups to dictate sample pH, and thereby optimize sensitivity. ACKNOWLEDGMENTS The authors are grateful for the support of the U.S. Army (DAAD13-02-C-0015, Joint Service Agent Water Monitor program). The authors would also like to thank Dr. Steve Christensen of the U.S. Army, SBCCOM, for helpful discussions, and Mr. Chetan Shende of Real-Time Analyzers for assistance in development of the gold-doped sol-gels. REFERENCES 1 Erickson, B., Analytical Chemistry News & Features, June 1, 397A (1998). SPIE -2003-5269 124
- 39. 2 Johnston, R.L., Hoefler, C.M., Fargo, J.C., and Moberley, B., AT-ONSITE, 5-8 (1994) 3 Black, R.M., R.J Clarke, R.W. Read, and M.T. Reid, J. Chromatography, 662, 301-321 (1994) 4 USA Army Center for Health Promotion and Preventive Medicine (USACHPPM) Technical Guideline 230A. 5 Jensen, J.L., N.C. Wong, W. R. Loerop, SPIE, 4775-03 (2001) 6 Hoenig, S.L. Handbook of Chemical Warfare and Terrorism, Greenwood Press, Wesport, CT, (2002) 7 Munro, N.B., S.S., Talmage, G.D. Griffin, L.C. Waters, A.P. Watson, J.F. King, and V. Hauschild, Env. Health Persp., 107, 933-974 (1999). 8 Holstege C.P., Kirk M., Sidell F.R., Crit. Care Clin.,13, 923-42 (1997). 9 Farquharson, S., P. Maksymiuk, K. Ong and S. Christesen, SPIE, 4577, 166-173 (2001). 10 Lee, Y. and S. Farquharson, SPIE, 4378, 21-26 (2001). 11 Spencer, K.M., J. Sylvia, S. Clauson and J. Janni, SPIE, 4577, 158-165 (2001). 12 Tessier, P., S. Christesen, K. Ong, E. Clemente, A. Lenhoff, E. Kaler and O. Velev, Applied Spectroscopy, 56, 1524-1530 (2002). 13 Farquharson, S., W.W. Smith, S. Elliott and J.F. Sperry, SPIE, 3533, 207-214 (1998). 14 Farquharson, S., W.W. Smith, S. Elliott and J.F. Sperry, SPIE, 3855,110-116 (1999) 15 Farquharson, S., W.W. Smith, Y.H. Lee, S. Elliott and J.F. Sperry, SPIE, 4575, 62-72 (2002). 16 Guzelian, A.A., J. Sylvia, J. Janni, S. Clauson and K.M. Spencer, SPIE, 4577, 182-192 (2001). 17 Shende, C., F. Inscore, A. Gift, P. Maksymiuk and S. Farquharson, in press. 18 Weaver, M.J., S. Farquharson and M.A. Tadayyoni, J. Chem. Phys., 82, 4867-4874 (1985). 19 Alak, A.M. and T. Vo-Dinh, Analytical Chemistry, 59, 2149-2153 (1987). 20 Lee, Y. and S. Farquharson, SPIE, 4206, 140-146 (2000). 21 Farquharson, S. and Y. Lee, SPIE, 4200-16 (2000). 22 Lee, Y., S. Farquharson and P. M. Rainey, SPIE, 3857, 76-84 (1999). 23 Lee, Y, S. Farquharson, H. Kwong and M. Shahriari, SPIE, 3537, 252-260 (1998). 24 Wang, J., M. Pumera, G. Collins and A. Mulchandani, Analytical Chemistry, 74, 6121-6125 (2002). 25 Dou, X., Y.M. Jung, Z.-Q. Cao and Y. Ozaki, Applied Spectroscopy, 53, 1440-1447 (1999). 26 Data supplied by S. Christesen and K. Ewing. 27 Kerker, M. and B. Thompson, Eds., SPIE, MS 10 (1990). 28 Inglesby, T.V., D.A. Henderson, J.G. Bartlett, JAMA, 287, 2236 (2002) 29 F.W. Janssen, A.J. Lund, and L.E. Anderson, Science, 127, 26, (1958). 30 Lide, D.R., Ed. Handbook of Chemistry and Physics, CRC Press, 77th Ed. (1996-7) 31 Billmann, J., G. Kovacs and A. Otto, Surf. Sci., 92, 153 (1980). 32 Murray, C.A. and S. Bodoff, Phys. Rev. B, 32 671 (1985). 33 Kellogg, D. and J. Pemberton, J. Phys. Chem., 91, 1120 (1987). SPIE -2003-5269 125
- 40. Appendix F Chemical agent detection by surface-enhanced Raman spectroscopy Stuart Farquharson, Alan Gift, Paul Maksymiuk, Frank Inscore, and Wayne Smith Real-Time Analyzers, 87 Church Street, East Hartford, CT 06108 Kevin Morrisey and Steven D. Christesen U.S. Army, SBCCOM, Aberdeen Proving Ground, MD 21010 ABSTRACT In the past decade, the Unites States and its allies have been challenged by a different kind of warfare, exemplified by the terrorist attacks of September 11, 2001. Although suicide bombings are the most often used form of terror, military personnel must consider a wide range of attack scenarios. Among these is the intentional poisoning of water supplies to obstruct military operations in Afghanistan and Iraq. To counter such attacks, the military is developing portable analyzers that can identify and quantify potential chemical agents in water supplies at microgram per liter concentrations within 10 minutes. To aid this effort we have been investigating the value of a surface-enhanced Raman spectroscopy based portable analyzer. In particular we have been developing silver-doped sol-gels to generate SER spectra of chemical agents and their hydrolysis products. Here we present SER spectra of several chemical agents measured in a generic tap water. Repeat measurements were performed to establish statistical error associated with SERS obtained using the sol-gel coated vials. Keywords: Chemical agents, chemical agent detection, SERS, Raman spectroscopy 1. INTRODUCTION In the autumn of 2001 terrorism within U.S. borders became a sobering reality. While extensive efforts are being implemented to secure the homeland, U.S. and Coalition forces in Afghanistan and Iraq are constantly faced with terrorist attacks. In addition to car-bombings, the military has identified several non-traditional attack scenarios, including poisoning of water supplies by chemical warfare agents (CWAs). To counter this threat, the Department of Defense has been investigating numerous approaches to detect such attacks under the auspices of the Joint Service Agent Water Monitor (JSAWM) program.1 The JSAWM must be capable of identifying the chemical agents with no false-positives and quantifying the agents at microgram per liter concentrations within 10 minutes.2 This includes the analysis of drinking water supplies, distribution and storage systems. Currently, colorimetric paper is used to detect agents on-site, while gas chromatography combined with mass spectrometry (GC/MS) is used in mobile support laboratories. However both methods have severe drawbacks. The paper changes color in response to contact with many chemicals besides CWAs, causing a high incidence of false positives,3 while GC/MS, although very chemically specific, requires up to an hour to perform and regular re- calibration.4,5,6 Vibrational spectroscopy has also been investigated as a potential method of rapidly detecting CWAs,7- 11 as early as 1985 in the case of infrared spectroscopy8 and 1988 in the case of Raman spectroscopy.11 Again, however these techniques also have limitations when it comes to measuring trace poisons in water. Infrared spectra would be dominated by the very strong absorption of water, which would obscure absorptions by most other chemicals present. Whereas Raman spectroscopy is simply not a very sensitive technique, and detection limits are typically grams per liter. Surface-enhanced Raman spectroscopy (SERS) offers several advantages over conventional vibrational methods, and that may provide the necessary sensitivity required for detecting trace quantities of chemical agents in water. SERS employs the interaction of surface plasmon modes of metal particles with target analytes to increase scattering efficiency by as much as 1 million times.12 In 1987, the potential of this approach to measure CWAs was demonstrated by measuring a series of pesticides.13 Several of these organophosphonates have chemical structures similar to CWAs, in particular P=O functional groups. In the past few years we and others have further explored the ability of SERS to detect CWAs,14-17 and even bioagents. 18-21 We have been employing silver-doped sol-gels to promote the SER effect SPIE-2003-5269 16
- 41. in these studies. The porous silica network of the sol-gel matrix offers a unique environment for immobilizing and stabilizing SER-active metal particles.22-25 The silver-doped sol-gels have been coated on the internal walls of standard glass vials to allow rapid SER analysis of small sample quantities without preparation. We have measured over 100 chemicals with enhancements of 104 to 106, demonstrated reversible measurements in a flowing system, reproducible measurements from vial-to-vial, and measurements in multiple solvents, including water.21-25 Previously, we used these vials to perform preliminary measurements of cyanide (CN) and methylphosphonic acid (MPA). Most of the nerve agents form MPA during hydrolysis, while Tabun forms CN, a chemical agent in its own right. In another paper including in these proceedings, we examined the limits of detection (LOD) for MPA by measuring a series of concentrations down to 50 mg/L and estimated a limit of detection of 100 microg/L.26 These measurements provide encouragement in that SERS may satisfy the needs of the JSAWM. To further establish the viability of SERS, in particular silver-doped sol-gels, here we present analysis of cyanide, mustard, and VX in tap water. The measurements performed at the U.S. Army’s Edgewood Chemical Biological Center, Aberdeen, MD, also included numerous repeat measurements to establish reproducibility. 2. EXPERIMENTAL 2.a. General. All chemicals, including potassium cyanide, 2-chloroethylethyl sulfide, and those used to prepare the silver-doped and gold-doped sol-gel coated vials were obtained and used as received from Sigma-Aldrich. All samples were prepared in a chemical hood using HPLC grade water unless otherwise noted (Fischer Scientific, Fair Lawn, NJ) for SERS measurements. Once prepared, the samples were transferred into the silver-doped sol-gel vials (Real-Time Analyzers, Inc., East Hartford, CT). The vials were coated in a manner similar to that previously reported by adding ammonium hydroxide to a solution of silver nitrate, tetramethyl orthosilicate (TMOS), and methanol.22 The two precursor solutions were prepared, mixed, and transferred to 2-mL glass vials, dried and heated. After sol-gel formation, the incorporated metal ions were reduced with dilute sodium borohydride (1mg/mL), which is followed by a water wash to remove residual reducing agent. After the resultant analyte solution was introduced, the SER-active vial was fixed horizontally to an XY positioning stage (Conix Research, Springfield, OR) just inside the focal point of an f/0.7 aspheric lens. The lens focused the beam into the sample and collected the scattered radiation back along the same axis. A dichroic filter (Omega Optical, Brattleborough, VT) was used to reflect the excitation laser to the lens and pass the Raman scattered radiation collected by the lens. An f/2 achromat was used to collimate the laser beam exiting a 200 µm core diameter source fiber optic, while a second f/2 achromat was used to focus the scattered radiation into a 365 µm fiber optic (Spectran, Avon, CT). A short pass filter was placed in the excitation beam path to block the silicon Raman scattering generated in the source fiber from reflecting off sampling optics and reaching the detector. A long pass filter was placed in the collection beam path to block the sample Rayleigh scattering from reaching the detector. A 785 nm diode laser (Process Instruments Inc. model 785-600, Salt Lake City, UT) was used to deliver 100 to 150 mW of power to the sample. A Fourier transform Raman spectrometer (Real-Time Analyzers, model IRA-785, East Hartford, CT) and a silicon photo- avalanche detector (Perkin Elmer model C30902S, Stamford, CT) were used to acquire the SER spectra. 2.b. Edgewood Chemical Biological Center. The surface-enhanced Raman spectral Table 1. Generic Tap measurements at the US Army’s Chemical Biological center presented here were all Water Composition. . performed on September 12, 2003. To expedite measurements, a plate was machined to Compound mg/L hold up to 12 SER-active sample vials (Figure 1). The plate fit a standard XY plate reader NaHCO3 100 that could be programmed. Pure KCN, bis-(2-chloroethyl)sulfide (distilled mustard gas, CaSO4 27 HD), and ethyl S-2-diisopropylamino ethyl methylphosphonothioate (VX), were obtained MgSO4•7H2O 6.7 on-site and used to prepare 1 g/L tap water solutions in a chemical hood with appropriate NaNO3 1 safety equipment. Simulated tap water was prepared by adding 10 chemicals most often Fulvic Acid 1 K2HPO4 0.7 found in tap water at appropriate concentrations to distilled water (Table 1). SER KH2PO4 0.3 measurements were also performed in a chemical hood. For added safety, the FT-Raman (NH4)2HSO4 0.01 instrument was placed outside the laboratory and 30 foot fiber optic and electrical cables NaCl 0.01 were used to allow remote SERS measurements and plate manipulation. For each FeSO4 0.001 experiment 1g/L samples were prepared and added to 9 individual vials, which were then pH 7.6-7.8 loaded on the plate. In some cases a tenth vial was included as a blank. SPIE-2003-5269 17
- 42. A software program was written that allowed selecting the sequence that the vials were measured, the number of positions along the length of the vials to measure (1 to 5), and the number of scans to co-add. During sample analysis, the program displayed the vial being analyzed, the point being analyzed, and the spectrum as it was being acquired. Once all the data was collected a second software program was written to rapidly analyze the data. The spectra collected for all the vials on a plate could be loaded at one time, and then the spectra for each point could be displayed simultaneously or separately. The user could then select the Raman peak to analyze in terms of peak height or area. This was accomplished by selecting points on either side of the peak to define a baseline of zero. The peak height or area could then be computed for all of the spectra loaded and then exported to a spreadsheet for statistical analysis. A B C D Figure 1. A) Vial Holder: 6 slots to hold 2 vials each, end-to-end. B) Measurement Configuration Program: user selects vials to measure, sequence, number of points per vial (1 to 5), and number of scans per point. C) Spectral Acquisition Program: shows spectrum being collected, which vial and position. D) Spectra Analysis Program: user selects spectra to analyze by plate, vial, and point (s), as well as two wavenumbers defining the peak and the baseline to subtract. The image is of 5 repeat measurements of 10 mg/L KCN in generic tap water, 16 sec each, 100 mW of 785 nm. 3. RESULTS AND DISCUSSION Raman and surface-enhanced Raman spectra were obtained for potassium cyanide, bis-(2-chloroethyl)sulfide, and ethyl S-2-diisopropylamino ethyl methylphosphonothioate, representing three classes of chemical agents, cyanides, mustards, and nerve agents, respectively. Spectra were also obtained for 2-chloroethyl ethyl sulfide (CEES) a structural analogue to HD, which was included in the study to aid in assigning spectral bands. KCN salt was used for cyanide experiments to avoid the increased hazards of handling HCN gas. KCN completely dissolves in water forming its conjugate acid, HCN, according to its Ka of 6.15x10-10,27 and at a concentration of 1 mg/mL results in a pH 10.7 solution. This is significant in that only CN- appears to interact sufficiently with silver to produce a SER spectrum, and no spectral signal is observed below pH 7.26 Figure 2 shows the SER and normal Raman spectra for KCN. The SERS spectrum of 1mg/ml KCN in water shows a single intense somewhat broad, feature at SPIE-2003-5269 18
- 43. 2100 cm-1 assigned to the single C≡N stretch. The band is much sharper in the normal Raman spectra of the solid KCN salt at 2074 cm-1. This band does broaden and shift to 2080 cm-1 in solution (not shown). However, the observed SERS frequency is attributed to interaction with silver, and low frequency mode at 135 cm-1 attributed to a Ag-CN stretch (not shown) supports this conclusion. Cl-CH2-CH2-S-CH2-CH3 A A B B Figure 2. A) SER and B) NR spectra of KCN. Conditions: Figure 3. A) SER and B) NR spectra of CEES. Conditions: A) A) 1 mg/ml in tap water, 100 mW of 785 nm at sample, 1- 1% v/v (10 mg/ml) in MeOH, 100 mW of 785 nm, 1-min min acquisition time, B) solid, 300 mW of 785 nm, 5-min. acquisition time. B) neat, 300 mW of 785 nm, 5-min. All spectra are 8 cm-1 resolution. Prior to measurements of HD, CEES was examined Cl-CH2-CH2-S-CH2-CH2-Cl by Raman and SER spectroscopy (Figure 3). CEES, also known as half-mustard, is essentially identical A to HD, except one of the chlorine end atoms is replaced by a hydrogen atom. Again, although not as toxic as HD, CEES is a blister agent and dilute aqueous samples were prepared using appropriate safety equipment. Both the Raman and SER spectra of CEES are similar and dominated by bands B between 600 and 800 cm-1. These are associated with C-Cl and C-S stretching modes, which are tentatively assigned to 648 and 747 cm-1 in the Raman spectra, respectively. The shoulder at 630 cm-1, the overlapped band at 660 cm-1, and the strong band at 692 cm-1, could also be due to these Figure 4. A) SER and B) NR of HD. Conditions: A) 1mg/ml in modes or their asymmetric counterparts. It is worth tap water, B) pure, both 100 mW of 785 nm, 1-min. noting that theoretical calculations assign the 692 cm-1 band to a C-S stretch, but the authors concede that it is in fact more likely a C-Cl stretch.28 It appears that the most intense Raman bands at 648, 692, and 747 cm-1 shift to 620, 660, and 720 cm-1 in the SER spectra, and are tentatively assigned as above. The width of these bands suggests that they overlap underlying spectral features. Additional bands in the Raman spectra occur at 972, 1034, 1049, 1263, 1286, 1423, 1442, 2865, 2935 and 2960 cm-1. Corresponding bands occur in the SER spectra at 964, 1015, 1054, 1286, 1410, 1447, 2865 and 2935 cm-1. Most of these bands are associated with alkane modes, specifically the bands at approximately 1040 cm-1 to a C-C stretch, 1290 cm-1 to a CH2 in-phase twist, 1440 cm-1 to a CH2 wag, 2865 cm-1 to a symmetric CH2 stretch, and 2965 cm-1 to an asymmetric CH2 stretch. The Raman and SER spectra of sulfur mustard were measured at the Edgewood center (Figure 4). Both spectra are largely similar to CEES. The C-Cl and C-S bands in the Raman spectrum of HD now occur at 640, 655, 700, 739, and 760 cm-1 and are more resolved, possibly due to the increased molecular symmetry. Theoretical calculations indicate that the first three bands are due to C-Cl stretching modes, and the latter two to C-S stretching modes.28 Only the C-Cl bands maintain significant intensity in the SER spectra occurring at 624 and 643 cm-1, which is attributed to the SPIE-2003-5269 19
- 44. expected strong interaction between chlorine and silver, and adds support to the assignment of this band to a C-Cl stretch. Weaker, overlapping bands occur at 670, 692 and 724 cm-1, the latter, possibly due to C-S stretching modes. Again, the alkane modes are apparent in the normal Raman spectra of HD, but only a broad feature at 1300 to 1450 cm-1 suggests CH2 contributions in the SER spectrum. Although the observed bands in the VX spectrum have not been assigned (Figure 4), a computer generated Raman spectrum29 predicts many of the same features with surprising accuracy and are used here. Two intense bands at 460 and 530 cm-1 closely match predicted bands at 463 and 546 cm-1 assigned to a CH3-P=O bend and a PO2CS wag. Three highly overlapped bands occur at 694, 745 and 771 cm-1 matching predicted bands at 713, 730, and 760 cm-1. The first two have been assigned to a C-S stretch and CH2 bend, respectively, while the latter has been attributed to either a P-C stretch or an O-C-C stretch. Although the 745 cm-1 band may A alternatively be assigned to a C-S stretch based on the previous measurements of CEES and HD. The relatively intense bands at 890, 1106, 1218, 1445, and 1465 cm-1 also match predicted bands at 880, 1108, 1216, 1440 and 1464 cm-1, that are assigned B to a C-C stretch, CH3 rock, N-C3 stretch, various C- H3 bends, and C-H bends, respectively. Both the computer generated and the measured spectra contain numerous other less intense bands. One is worth mentioning. A unique band appears at 370 cm-1 that is predicted at 368 cm-1 and corresponds to an O-P=O bend. The surface-enhanced Raman Figure 5. A) SERS and B) NR spectra of VX. Conditions: A) spectrum of VX is also rich with spectral features. 1% v/v (10 mg/ml) in MeOH, B) pure sample, both 100 mW of It has the unique low frequency band at 370 cm-1, as 785 nm, 1-min acquisition time. well as a second band at 380 cm-1 that is assigned to the S-P-O bend predicted in the normal Raman Table 2. Measured SER peak heights for the CN stretch at 2100 cm-1 for 1 mg/mL of sample in 3 stock solutions repeated 3 times, spectrum at 388 cm-1. Based on the measured and and measured 4 times per vial. predicted normal Raman spectra, the following SERS assignments are given: 460 cm-1 to the CH3- stock solution spot Vial 1 Vial 2 Vial 4 P=O bend, 544 cm-1 to the PO2CS wag, 738 cm-1 to . 1 137.54 130.04 128.19 a C-S stretch (based on arguments above), 890 cm-1 2 135.19 126.92 129.09 to a C-C stretch, 1101 cm-1 to a CH3 rock, and 1456 1 3 135.41 127.21 126.39 cm-1 to a C-H bend. 4 134.62 126.48 126.51 ave 135.69 127.66 127.55 The ability of SERS to measure chemical agents in Vial 6 Vial 3 Vial 5 water containing real-world chemical interferents 1 115.86 140.68 107.57 was tested by using the generic tap water described 2 112.36 144.02 115.12 in Table 1. The ability to reproduce measurements 2 3 113.76 145.46 115.59 was accomplished by preparing three separate water 4 108.94 117.63 112.14 stock solutions, which were used to prepare three ave 112.73 136.95 112.61 sample solutions each of 1 mg/mL KCN in the generic tap water. The samples, defined as vials 1- Vial 7 Vial8 Vial 9 9, were then measured at 4 points per vial in a semi- 1 111.98 158.04 110.23 random fashion, such that errors associated with 2 112.43 157.16 88.75 stock solution preparation and errors associated with 3 3 115.62 152.16 112.18 instrument drift could be identified. No trends were 4 116.55 150.84 110.77 apparent that signified such systematic errors. Each ave 114.15 154.55 105.48 spectrum collected consisted of 20 averaged scans CN AVG STDEV %ERR taking 16 seconds, at 8 cm-1 resolution. The laser 1 mg/mL pk ht 125.26 15.60 12.45 power at the sample was measured periodically during the day and it ranged from 102 to 105 mW. spacccc SPIE-2003-5269 20
- 45. The CN measurements, consisting of 36 data points, produced an average height of 125 for the 2100 cm-1 peak with a standard deviation of 15.6 or 12.5% (Table 2). The HD and VX measurements were performed precisely the same way (Tables 3 and 4). For HD, the 624 cm-1 peak was used for analysis and it had an average height of 5.3 with a standard deviation of 0.68 or 12.9%, while for VX, the 544 cm-1 peak was used for analysis and it had an average height of 10.51 with a standard deviation of 3.08 or 29.3%. The greater error in the VX measurements can be somewhat attributed to Vial 4, which produced lower SER signal intensities. But removing this vial from the data set changes the standard deviation to 23.3%, only a modest improvement. Table 3. Measured SER peak heights for the HD band at 624 Table 4. Measured SER peak heights for the VX band at 544 cm-1 for 1 mg/mL of sample in 3 stock solutions repeated 3 cm-1 for 1 mg/mL of sample in 3 stock solutions repeated 3 times, and measured 4 times per vial. times, and measured 4 times per vial. stock solution spot Vial 1 Vial 2 Vial 4 stock solution spot Vial 1 Vial 2 Vial 4 1 5.25 6.09 4.68 1 14.64 10.34 4.67 2 4.5 6.75 4.84 2 14.85 9.9 6.1 1 3 5.27 6.44 5.09 1 3 14.91 9.89 5.68 4 5.59 8.07 6.69 4 10.41 7.77 5.53 ave 5.1525 6.8375 5.325 ave 13.7025 9.475 5.495 Vial 6 Vial 3 Vial 5 Vial 6 Vial 3 Vial 5 1 5.21 5.74 5.75 1 10.58 9.42 12.93 2 5.36 5.3 4.49 2 6.97 12.1 9.65 2 3 5.09 5.08 4.51 2 3 7.27 12.06 11.12 4 5.65 5.94 3.79 4 6.89 12.54 7.46 ave 5.3275 5.515 4.635 ave 7.9275 11.53 10.29 Vial 7 Vial 8 Vial 9 Vial 7 Vial 8 Vial 9 1 4.57 5.89 4.13 1 11.27 7.83 16.3 2 5.83 4.97 4.27 2 13.58 8.12 16.14 3 3 5.44 5.05 4 3 3 13.71 9.08 15.12 4 5.28 5 5.09 4 10.97 8.75 13.97 ave 5.28 5.2275 4.3725 ave 12.3825 8.445 15.3825 HD AVG STDEV %ERR VX AVG STDEV %ERR 1 mg/mL pk ht 5.30 0.68 12.91 1 mg/mL pk ht 10.51 3.08 29.25 4. CONCLUSIONS In this paper we examined the ability of surface-enhanced Raman spectroscopy to reproducibly measure CN, HD and VX in tap water without chemical interference. Both normal and surface-enhanced Raman spectra were examined to select unique bands suitable to identify and quantify these chemical agents. For SER measurements, the 2100 cm-1 C-N stretch was used for CN, the 624 cm-1 C-Cl stretch was used for HD, and the 544 cm-1 PO2CS wag was used for VX. It was determined that 1 mg/mL samples of each of these chemicals measured 36 times in glass vials coated with a silver- doped sol-gel reproduced measurements with standard deviations of 12.5, 12.9 and 29.3%. It was further found that the 10 chemicals added to simulate generic tap water did not interfere with or alter the SER spectra. It should be noted that the concentrations used in this study were considerably greater than those required by the JSAWM program. Current work involves improving SER sensitivity and designing sampling systems with better reproducibility. This includes the development of fractal silver and gold structures within the sol-gel matrix and the development of chemically selective sol-gels. ACKNOWLEDGMENTS The authors are grateful for the support of the U.S. Army (DAAD13-02-C-0015, Joint Service Agent Water Monitor program). The authors would also like to thank Janet Jensen, Ronald Crosier, and Kristina Gonser for helpful discussions. SPIE-2003-5269 21
- 46. REFERENCES 1 Jensen, J.L., N.C. Wong, W. R. Loerop, SPIE, 4775-03 (2001). 2 JSAWM Requirements at www.sbccom.apgea.army.mil/RDA/ecbc/rt/PRODSER/JSAWM/jsawm.html 3 Erickson, B. “The Chemical Weapons Convention Redefines Analytical Challenge”, Anal. Chem., 70, 397A (1998). 4 Johnston, R.L., Hoefler, C.M., Fargo, J.C., and Moberley, B., “The Defense Nuclear Agency’s Chemical/Biochemical Weapons Agreements Verification Technology Research, Development, Test and Evaluation Program and its Requirements for On-Site Analysis”, AT-ONSITE, 5-8 (1994). 5 Black, R.M., Clarke, R.J., Read, R.W., and Reid, M.T., “Application of GC-MS and GC-tandem MS to the analysis of chemical warfare samples, found to contain residues of the nerve agent sarin, sulphur mustard and their degradation products”, J. Chrom., 662, 301-321 (1994). 6 USA Army Center for Health Promotion and Preventive Medicine (USACHPPM) Technical Guideline 230A. 7 Hoffland, L.D., Piffath, R.J., Bouck, J.B., ”Spectral signatures of chemical agents and simulants”, Optical Engineering, 24, 982-984, (1985). 8 Braue, E.H.J., Pannella, M.G.,”CIRCLE CELL FT-IR Analysis of Chemical Warfare Agents in Aqueous Solutions”, App. Spectrosc., 44, 1513-1520, (1990). 9 Tseng, C.-H., Mann C.K., and Vickers T.J., “Determination of Organics on Metal Surfaces by Raman Spectroscopy”, App. Spectrosc., 47, 1767-1771 (1993). 10 Staff Reporter, “US Army Tests Lidar to Detect Biological Toxins”, Photonic Spectra, pg. 50, December 1998. 11 Christesen, S.D., "Raman cross sections of chemical agents and simulants", App. Spectrosc., 42, 318-321 (1988). 12 Weaver, M.J., Farquharson, S., Tadayyoni, M.A., “Surface-enhancement factors for Raman scattering at silver electrodes”, J. Chem. Phys., 82, 4867-4874 (1985). 13 Alak, A.M. and Vo-Dinh, T., “SERS of Organophosphorous Chemical Agents”, Anal. Chem., 59, 2149-2153 (1987). 14 Lee, Y., Farquharson, S., “Rapid chemical agent identification by SERS”, SPIE, 4378, 21-26 (2001). 15 Farquharson, S., Maksymiuk, P., Ong, K., Christesen, S., “Chemical agent identification by surface-enhanced Raman spectroscopy”, SPIE, 4577, 166-173 (2001). 16 Spencer, K.M., Sylvia, J., Clauson, S. and Janni, J., “Surface Enhanced Raman as a Water Monitor for Warfare Agents in Water”, SPIE, 4577, 158-165 (2001). 17 Tessier, P., Christesen, S., Ong, K., Clemente, E., Lenhoff, A., Kaler, E., Velev, O., “On-line spectroscopic characterization of sodium cyanide with nanostructured gold surface-enhanced Raman spectroscopy substrates”, App. Spectrosc., 56, 1524-1530 (2002). 18 Farquharson, S., W. W. Smith, S. Elliott and J. F. Sperry, "Rapid biological agent identification by surface- enhanced Raman spectroscopy", SPIE, 3855,110-116 (1999). 19 Farquharson, S., W.W. Smith, Y.H. Lee, S. Elliott and J. F. Sperry, "Detection of biological signatures: A comparison of electrolytic and metal-doped sol-gel surface-enhanced Raman media", SPIE, 4575, 62-72 (2002). 20 Guzelian, A.A., Sylvia, J. Janni, J., Clauson S., and Spencer, K.M. “SERS of whole cell bacteria and trace levels of biological molecules”, SPIE, 4577, 182-192 (2001). 21 Shende, C., Inscore, F., Gift, A., Maksymiuk, P., Farquharson, S., “Rapid dipicolinic acid extraction from Bacillus spores detected by surface-enhanced Raman spectroscopy”, App. Spectrosc., 58, accepted. 22 Lee, Y. and Farquharson, S., “SERS Sample Vials Based on Sol-Gel Process for Trace Pesticide Analysis”, SPIE, 4206, 140-146 (2000). 23 Farquharson, S. and Lee, Y., “Trace Drug Analysis by SERS”, SPIE, 4200-16 (2000). 24 Lee, Y., Farquharson, S., and Rainey, P. M., "Surface-Enhanced Raman Sensor for Trace Chemical Detection in Water", SPIE, 3857, 76-84 (1999). 25 Lee, Y, Farquharson, S., Kwong, H., and Shahriari, M., “Sol-Gel Chemical Sensor for Surface-Enhanced Raman Spectroscopy”, SPIE, 3537, 252-260 (1998). 26 Farquharson, S, A Gift, P Maksymiuk, F Inscore, and W Smith, “pH dependence of methyl phosphonic acid, dipicolinic acid, and cyanide by surface-enhanced Raman spectroscopy”, SPIE, 5269-19 (2003). 27 Lide, D.R., Ed. Handbook of Chemistry and Physics, CRC Press, 77th Ed. (1996-7). 28 Sosa, C., R.J. Bartlett, K. KuBulat, and W.B. Person, “A theoretical study of harmonic vibrational frequencies and infrared intensities of XCH2CH2SCH2CH2X and XCH2CH2SH (= H, Cl)”, J. Phys. Chem., 93, 577-588 (1993). 29 Hameka, H.F. and J.O. Jenson, “Computer-generated predictions of the structure and the IR and Raman spectra of VX”, ERDEC-TR-065, May 1993. SPIE-2003-5269 22