Submillimeter spectroscopy for chemical analysis
with absolute specificity
Ivan R. Medvedev,1Christopher F. Neese,1Grant M. Plummer,2and Frank C. De Lucia1,*
1Department of Physics, Ohio State University, 191 W. Woodruff Avenue, Columbus, Ohio 43210, USA
2Enthalpy Analytical, Inc., 2202 Ellis Road, Durham, North Carolina 27703, USA
* Corresponding author: email@example.com
Received December 21, 2009; revised March 9, 2010; accepted March 19, 2010;
posted April 2, 2010 (Doc. ID 121787); published May 5, 2010
A sensor based on rotational signatures in the submillimeter (SMM) region is described. This sensor uses
frequency synthesis techniques in the region around 10 GHz, with nonlinear diode frequency multiplication
to 210–270 GHz. This provides not only a nearly ideal instrument function, but also frequency control and
agility that significantly enhance the performance of the spectrometer as a sensor. The SMM frequencies
provide significantly stronger absorptions and broader spectroscopic coverage than lower-frequency micro-
wave systems. Among the characteristics of the sensor are absolute specificity, low atmospheric clutter, good
sensitivity, and near-term paths to systems that are both compact and very inexpensive. © 2010 Optical
Society of America
OCIS codes: 280.1545, 300.1030, 300.6320, 300.6370, 300.6390, 300.6495.
A resurgence of interest in spectroscopic sensors has
been fueled by increases in performance made pos-
sible by advances in laser systems and applications
in medicine, environmental monitoring, and national
security [1–8]. Most of these new approaches make
use of the optical–infrared (Op–IR) spectral regions
[9–12]. The submillimeter (SMM) spectral region,
while much less well known, has also seen significant
technological advances . It is possible to use fre-
quency agile, synthesized electronic sources that can
be multiplied into the SMM, with power outputs on
the order of 1 mW, a power capable of saturating the
probed molecules and that corresponds to 1014K in a
1 MHz Doppler linewidth. In the SMM the thermal
excitation of many rotational energy levels provides a
fingerprint that is complexly redundant and resolv-
able, for even moderately large molecules, because of
the small Doppler limit. These fingerprints, con-
tained in more than 105resolvable spectral channels,
lead to absolute specificity, even in complex mixtures
In this Letter we demonstrate new SMM sensors
that are comparable with their optical counterparts
[1,2,4,7,8] in terms of fractional concentration (ppx,
where x is millions for 106, billions for 109, trillions
for 1012, etc.) sensitivity and are orders of magnitude
more favorable than time domain terahertz (TDS-
THz) sensors [16,17]. If the sensitivity figure of merit
is the amount of sample, the lower optimum pressure
??10 mTorr? of SMM sensors increases their sensi-
tivity by 2–3 orders of magnitude relative to Op–IR
sensors. These SMM sensors provide far greater
specificity than either, with good speed.
To illustrate the rotational fingerprint, Fig. 1
shows a Doppler-limited rotational spectrum re-
corded with a backward-wave-oscillator-based fast
scan submillimeter spectroscopy technique (FASSST)
system . The spectral segment in the bottom
panel uses only a small fraction of the available spec-
tral information, shows unambiguous identification
of eight of the species in the mixture, and required
only about 10 ms to acquire.
The more compact solid-state sensor system shown
in Fig. 2 probes a low-pressure ?1–10 mTorr?,
Doppler-limited sample in the 210–270 GHz region.
To provide the required frequency, a 24? frequency
multiplication scheme was adopted. Drive for this
system was provided by a fast sweeping 0–400 MHz
synthesizer, scanning around a nominal 100 MHz
center frequency, which is subsequently mixed with a
stepping synthesizer in the 8.75–11.25 GHz range to
provide a tunable source offset from the receiver’s
local oscillator by its intermediate frequency. A YIG
(yttrium iron garnet—a kind of microwave filter
based on a magnetically tunable resonance) filter was
used to select the desired sideband and suppress the
many intermodulation products that grow rapidly
with frequency multiplication. This frequency control
system provides more uniform line shapes and inten-
sities than previous free-running systems [14,18]. We
will see below that the frequency stability and agility
provided by this approach are very important for the
Fig. 1. (Color online) 50 GHz of a spectrum of a mixture of
20 gases on a highly compressed scale taken with a
0.4% of this spectrum is expanded to show individual lines
(bottom). Comparison with the library spectra of eight
gases shows their presence in the gas mixture (middle).
May 15, 2010 / Vol. 35, No. 10 / OPTICS LETTERS
0146-9592/10/101533-3/$15.00© 2010 Optical Society of America
speed, sensitivity, specificity, and clutter rejection of
the SMM sensor.
Because molecular saturation often limits probe
powers, we chose a heterodyne detector that makes
possible a system that approaches fundamental noise
limits for low-power probes. Since saturation is deter-
mined both by the square of the molecular dipole mo-
ment and the pressure, optimal power levels for each
molecule and operating condition vary widely. In gen-
eral we chose a low probe power, a few microwatts, to
ensure that under all circumstances we are in a non-
saturation limit. For species with smaller dipole mo-
ments, it is possible to increase the probe power to in-
crease sensitivity. The local oscillator for this receiver
is driven directly by the stepping synthesizer through
a similar 24? multiplier chain. The resulting inter-
mediate frequency signal sweeps around 2.4 GHz
(100±? MHz 24?) and is demodulated by a Schottky
diode. An FM modulation, somewhat smaller than
the Doppler linewidth, with lock-in detection was
used to suppress baseline due to standing waves and
other power variations. These power variations can
be 5–10 over the 210–270 GHz band. To obtain the
desired 1% absolute amplitude calibration, measure-
ments of the DC levels produced by the Schottky di-
ode were used. Library lines of known absorptions al-
lowed us to establish the modulation efficiency. Both
the source and the detector operate in the fundamen-
tal waveguide, and the power is propagated quasi op-
tically with horns and lenses through a 2.5 cm diam-
eter, 1.2 m long, stainless steel folded cell. Because
only a static sample of 1–10 mTorr is required,
vacuum is provided by a small turbo and diaphragm
The great strength of SMM rotational spectroscopy
is its specificity, especially when applied to complex
mixtures. Accordingly, we recorded a spectrum at
1 mTorr for each of 32 gases and made mixtures from
the gases in this library. This number is considerably
larger than is typically used in demonstrations of
spectroscopic sensors. While the chosen molecules
are generally favorable species for rotational spec-
troscopy, they range over about 2 orders of magnitude
in both strength and spectral density and contain 17
gases included in a common list of toxic industrial
chemicals . Many of these species do not have re-
solved rotational spectra in the Op–IR, and most are
considerably less favorable than species typically
chosen for demonstrations. Additionally, the near
transparency at low pressure in the SMM of H2O and
CO2(molecules that provide significant clutter in
Op–IR sensors ) makes this approach remarkably
clutter free. Whereas H2O and CO2have atmospheric
concentrations of ?10000 and 332 ppm, respectively,
the first molecule in order of abundance that has
even a 1% spectral density at Doppler resolution in
the SMM is nitric acid, with an atmospheric concen-
tration in the range of 0.0002–0.05 ppm .
210–270 GHz region far exceeds that necessary for
absolute specificity, we chose a subset of six lines for
each gas and observed a 2–6 MHz snippet for each
line—a total of 192 snippets and ?0.5 GHz of spec-
tral space. These results are shown in Fig. 3. The
snippets are recorded in order of the analyte in the
library, to facilitate visual identification of detected
gases. Automated analyses, based on least-squares
(LSQ) techniques and shown in the right-hand panel
of Fig. 3, provide quantitative results by comparison
of the spectra shown in Fig. 3 with previously re-
corded libraries. These LSQ analyses require less
than 1 s of computation time. Each of the 14 gases
placed in the mixture is identified in red in Fig. 3.
Those identified in blue are marginal detections that
result either from contamination or chemical reac-
tions among the analytes.
Insert a of Fig. 3 shows an enlargement of the
ClCN fingerprint region. No ClCN is present in this
mixture, and the signal in this region is from the in-
terloping spectra of other species, CH3CN, C2H3CN,
C2H5CN, and C2H3F. However, because of the inten-
sity calibration of the system, the fit for the concen-
tiplication, heterodyne receiver gas sensor. In this system a
frequency synthesizer near 10 GHz drives a 24? diode-
based frequency multiplier to provide the local oscillator of
a heterodyne receiver. A sideband generation scheme pro-
vides synchronized, but offset in frequency, drive to a simi-
lar multiplier chain for probe power to the folded 1.2 m ab-
(Color online) Overview of a SMM frequency mul-
fingerprint regions selected for each gas for the family of 32
gases considered, with enlargements of specific fingerprint
regions. Right, results of a quantitative LSQ analysis. A
digital lock-in recovers a near first derivative line shape
that results from a small FM modulation on the probe
(Color online) Observed spectra at each of the six
OPTICS LETTERS / Vol. 35, No. 10 / May 15, 2010
tration of these four species fully accounts for these Download full-text
spectral features. Accordingly, the fit properly con-
cludes that although there is spectral intensity in the
ClCN fingerprint, less than 0.01 ?Torr of ClCN is
present. The partial pressures of each gas recovered
by the LSQ fit ranged from 0.02 to 0.13 mTorr. The
0.89 mTorr. This pressure is within the accuracy of
the pressure measurement of the fill of the sample
cell, nominally 1 mTorr.
Because the pressure is low, all of the lines have
widths determined by Doppler broadening. Thus the
species are spectroscopically noninteracting, and it is
possible to determine their minimum detectable con-
centration individually. If we first consider the stron-
gest species in this figure, CH3CN, a sample pressure
of 6?10−5Torr produces a signal-to-noise ratio S/N
of ?500 with 0.01 s per point of integration time.
Each of the 6 lines has ?10 points. If each point re-
ceives 1 s of integration time and a fit is used to pro-
cess all of the information in the ?60 points on the
lines, the S/N for this recovery becomes ?40,000.
Thus, for a 1/1 S/N in the recovery a pressure of
?1.5?10−9Torr would be required. To obtain a ppx
measure, the fact that an air pressure of ?0.1 Torr
can be added without significantly broadening or de-
creasing the line provides a detectivity level of
15 ppb. Figure 3 also shows that the weakest of the
32 gases ?C2H4Cl2? is about 30 times weaker, result-
ing in a detection level of ?500 ppb. The most favor-
able gas, HCN, is about 5 times stronger, resulting in
a detection level of ?3 ppb. Similar results have been
obtained more directly on samples diluted in air.
A comparison of the results presented here with
those obtained in a wide variety of Op–IR experi-
ments show that they are similar in terms of ppx sen-
sitivity, with wide variation according to choice of
molecule and the trades of technical implementations
[1,2,4,7,8]. Because the optimum pressure (which is
proportional to the Doppler width and thus to fre-
quency as well) is 100–1000 times smaller in the
SMM, in terms of the minimum sample required,
according to this measure: ?10−14moles for HCN,
?5?10−14moles for CH3CN, and 10−12moles for
C2H4Cl2. For radicals, for which a strong electronic
transition is available, detection limits in the ppt
range have been reported .
Finally, there is a clear path to even simpler and
cheaper systems in the near to medium term. Inex-
pensive technology is commercially available up to
100 GHz courtesy of the wireless communications in-
dustry, and this upper frequency limit is increasing
rapidly. While we used instrumentation synthesizers,
etc., implementations based on chip level integra-
tions of these functions are available. Finally, the
never-ending increase in computational power will
further improve our ability to optimize the use of the
information content of high-resolution rotational
spectra. This combination of analysis power and
practicality will allow SMM rotational spectroscopy
to take its place alongside other methods as a general
analytical and sensor tool in the near term.
We thank the Army Research Office and the De-
fense Advanced Projects Research Agency (DARPA)
for their support of this work.
1. Y. He and B. J. Orr, Appl. Phys. B 85, 355 (2006).
2. M. J. Thorpe, K. D. Moll, R. J. Jones, B. Safdi, and J.
Ye, Science 311, 1595 (2006).
3. S. A. Diddams, L. Hollberg, and V. Mbele, Nature 445,
4. M. J. Thorpe, D. Balslev-Clausen, M. S. Kirchner, and
J. Ye, Opt. Express 16, 2387 (2008).
5. T. Gherman and D. Romanini, Opt. Express 10, 1033
6. A. A. Kosterev, A. L. Malinovsky, F. K. Tittle, C.
Gmachle, F. Capasso, D. L. Sivco, J. N. Baollargeon, A.
L. Hutchinson, and A. Y. Cho, Appl. Opt. 40, 5522
7. M. Kassi, M. Chenevier, L. Gianfrani, A. Salhi, Y.
Rouillard, A. Ouvrard, and D. Romanini, Opt. Express
14, 11142 (2006).
8. I. Vertrillard-Courtillot, T. Gonthiez, C. Clerici, and D.
Romanini, J. Biomed. Opt. 14, 064026 (2009).
9. F. Keilmann, C. Gohle, and R. Holzwarth, Opt. Lett. 29,
10. S. T. Sanders, Opt. Photonics News 16, 36 (2005).
11. J. Ye, L.-S. Ma, and J. L. Hall, Opt. Lett. 21, 1000
12. I. Debecker, A. K. Mohamed, and D. Romanini, Opt.
Express 13, 2906 (2005).
13. T. W. Crowe, W. L. Bishop, D. W. Porterfield, J. L.
Hesler, and R. M. Weikle, IEEE J. Solid-State Circuits
40, 2104 (2005).
14. S. Albert, D. T. Petkie, R. P. A. Bettens, S. P. Belov, and
F. C. De Lucia, Anal. Chem. 70, 719A (1998).
15. I. R. Medvedev, M. Behnke, and F. C. De Lucia, Appl.
Phys. Lett. 86, 1 (2005).
16. S. A. Harmon and R. A. Cheville, Appl. Phys. Lett. 21,
17. B. Ferguson and X.-C. Zhang, Nature Mater. 1, 26
18. I. Medvedev, M. Behnke, and F. C. De Lucia, Analyst
(Cambridge, U.K.) 131, 1299 (2006).
19. “Clean Air Act Ammendments of 1990,” 5.6130, 101st
Congress of the United States of America (April 3,
20. J. H. Seinfeld, Atmospheric Chemistry and Physics of
Air Pollution (Wiley, 1986).
21. G. Mejean, S. Kassi, and D. Romanini, Opt. Lett. 33,
May 15, 2010 / Vol. 35, No. 10 / OPTICS LETTERS