Impact of atmospheric clutter on Doppler-limited gas sensors in the submillimeter/terahertz

Impact of atmospheric clutter on Doppler-limited gassensors in the submillimeter/terahertz

Ivan R. Medvedev,1 Christopher F. Neese,2 Grant M. Plummer,3 and Frank C. De Lucia2,*1Department of Physics, Wright State University, 3640 Colonel Glenn Highway, Dayton, Ohio 45435, USA

2Department of Physics, The Ohio State University, 191 West Woodruff Avenue, Columbus, Ohio 43210, USA3Enthalpy Analytical, Inc., 2202 Ellis Road, Durham, North Carolina 27703, USA

*Corresponding author: [email protected]‐state.edu

Received 24 November 2010; revised 17 March 2011; accepted 13 April 2011;posted 13 April 2011 (Doc. ID 138722); published 17 June 2011

It is well known that clutter (spectral interference) from atmospheric constituents can be a severe limitfor spectroscopic point sensors, especially where high sensitivity and specificity are required. In this pa-per, we will show for submillimeter/terahertz (SMM/THz) sensors that use cw electronic techniques theclutter limit for the detection of common target gases with absolute specificity (probability of false alarm≪10−10) is in the ppt (1 part in 1012) range or lower. This is because the most abundant atmospheric gasesare either transparent to SMM/THz radiation (e.g., CO2) or have spectra that are very sparse relative tothe 105 Doppler-limited resolution elements available (e.g., H2O). Moreover, the low clutter limit demon-strated for cw electronic systems in the SMM/THz is independent of system size and complexity. © 2011Optical Society of AmericaOCIS codes: 280.1545, 300.1030, 300.6320, 300.6370, 300.6390, 300.6495.

1. Introduction

As new sensor and analytical approaches are consid-ered in the submillimeter/terahertz (SMM/THz)[1–9], an important figure of merit is their suscept-ibility to atmospheric clutter. It is well known thatthis atmospheric clutter can be the limiting factorfor spectroscopically based atmospheric gas sensors,especially in the optical/infrared (Op/IR) [10–15]. Wewill show in this paper that clutter has a signifi-cantly smaller impact on cw electronic sensors inthe less studied SMM/THz. The principal reasonsfor this are the 102–103 smaller Doppler widthsand the fortuitous transparency at low pressure ofall of the abundant atmospheric gases. We will alsoshow in the SMM/THz that systems with frequencyaccuracy and intensity calibration can further reducethe impact of clutter.

Atmospheric clutter is also low in another applica-tion, radio astronomy, but for an additional reason.

Near the surface of the earth, any contributions fromtrace clutter species are pressure broadened so thateven their small contributions are at most a baselineeffect. At high altitudes above about 100km, wherethe pressure-broadened linewidths might be compar-able to astrophysical linewidths, their densities areso low as to be unobservable.

The constituents of the atmosphere vary widelyandmany lists have been compiled. For the work con-sidered here, we will use the results of [16], which areshown in Table 1. Different lists would provide re-sults that differ in detail but do not impact the over-all conclusions of this paper. We will consider twoclutter atmospheres: the clean troposphere and a pol-luted urban environment. We will consider six spe-cies, shown in Table 2, that are representative ofspectral types and appear on lists of toxic industrialchemicals (TICs) [17]. The first three are linear orsymmetric top species, with rather simple spectrathat are easier to display graphically, whereas thelatter are asymmetric rotors with more complexspectra. Some TICs that are appropriate for SMM/THz sensors have stronger spectra than these, and

0003-6935/11/183028-15$15.00/0© 2011 Optical Society of America

3028 APPLIED OPTICS / Vol. 50, No. 18 / 20 June 2011

others have weaker, ranging over several orders ofmagnitude as determined by their dipole momentsand partition functions. The latter are strong func-tions of molecular size and complexity.

2. Characteristics of SMM/THz Rotational Spectra

Because molecular spectra are complex and vary en-ormously in strength and density, the sensitivity andspecificity of spectroscopic sensors are strong func-tions of target gases. Similarly, the contribution of at-mospheric gases to background clutter also variesenormously.

In the SMM/THz, the principal spectral signatureis due to the rotational fine structure. Because rota-tional energy level spacings are small in comparisonto kT, a spectrumwith many lines results. Because ofthe small Doppler width, these lines are generallywell resolved, providing an extraordinarily specificfingerprint. The signature is complexly redundant—the seemingly random spectra of many species

can be calculated from a smaller set of highly accu-rate spectral constants. This set of constants is alsoan extraordinarily specific fingerprint, althoughidentification is usually done by matching a subsetof the sensor spectra directly with library spectrarather than by spectral analysis [3,4]. A similar fin-gerprint, in the form of vibration-rotation spectra, isalso used for specificity in the Op/IR, but the resolu-tion is reduced by 102–103 by the increase in Dopplerwidth and in many cases by broader instrumentfunctions.

There are a number of approaches for the charac-terization of clutter that can be used for the evalua-tion of its impact on systems. These approachesinclude: (1) synthetic spectra based on catalogs suchas HITRAN [18], JPL [19], or CDMS [20]; (2) synth-eses of mixtures based on experimentally measuredspectra of individual species; and (3) field evaluationof instruments in more or less well-defined clutterbackgrounds. Calculation of rotational spectra fromcatalogs must be done with care because often cata-logs do not contain the rotational spectra of excitedvibrational states or minor isotopologues that alsocontribute to the clutter. Such synthetic spectra canbe massively incomplete, especially in the cases oflarge molecules [21].

For this work we will use both synthetic and la-boratory experimental spectra. Many atmosphericgases do not have low-lying vibrational states, andcatalog synthesis is appropriate. This is extremelyuseful because very little of the complete, intensity-calibrated experimental spectra required for mixturesynthesis is available in the SMM/THz. For specieswith substantively incomplete catalogs, we have de-veloped intensity calibration approaches for theSMM/THz [21].

The density of rotational spectra in the SMM/THzis a strong function of molecular size and symmetry.As an example, Fig. 1 shows the spectrum of OCS,and Fig. 2 shows the spectrum of C2H3CN. The lowerpanel of Fig. 2 shows C2H3CN on a 30 times ex-panded frequency scale, revealing that its spectrumis a hundred times denser than that of OCS. In orderto include lines due to transitions in excited vibra-tional states that might not be included in catalogs,their spectra are derived from experimental spectrataken at 1mTorr [9] and scaled to account for the1ppt (1 part in 1012) dilution. To provide figures thatare on the same linewidth footing as the simulationsin other figures in this paper, the experimental linepeaks were converted into lineshapes by convolvingthese observed strengths with Gaussian lineshapes.If the OCS spectrum of Fig. 1 were similarly ex-panded, no new lines would appear, and the figurewould be dominated by white space. The sparsenessof the OCS spectrum arises because OCS is a linearmolecule. A Doppler-limited SMM/THz sensor canresolve 105 spectral channels in the 60GHz spectralrange of Figs. 1 and 2.

Figure 3 summarizes these results at a concentra-tion of 1ppt (1 part in 1012) for OCS and C2H3CN

Table 1. Molecular Concentrations in the Clean Troposphere and thePolluted Urban Air (ppb)a

Species Clean Troposphere Polluted Air

N2 780,840,000 780,840,000O2 209,460,000 209,460,000H2O Variable VariableAr 9,340,000 9,340,000Ne 18,000 18,000He 5200 5200Kr 1100 1100Xe 90 90H2 580 580CH4 1650 1650CO2 332,000 332,000N2O 330 330SO2 1–10 20–200CO 120 1000–10,000NO 0.01–0.05 50–750NO2 0.1–0.5 50–250O3 20–80 100–500HNO3 0.02–0.3 3–50NH3 1 10–25H2CO 0.4 20–50HCOOH 1–10HNO2 0.001 1–8CH3CðOÞO2NO2ðPANÞ 5–35Nonmethane hydrocarbons 500–1200aData from Seinfeld [16].

Table 2. Target Species Considered in this Paper

Name Formula Type

Cyanogen chloride ClCN LinearCarbonyl sulfide OCS LinearAcetonitrile CH3CN Symmetric topAcrolein C2H3CHO Asymmetric rotorAcrylonitrile C2H3CN Asymmetric rotorEthylene oxide C2H4O Asymmetric rotor

20 June 2011 / Vol. 50, No. 18 / APPLIED OPTICS 3029

(and a number of other gases and clutter atmo-spheres that will be discussed below). In these plots,the peak absorption coefficients are shown for thelines in the 210–270GHz window, sorted accordingto strength. In the specific case of OCS, the five stron-gest lines have strengths near 10−13 cm−1 at theassumed 1ppt concentration. This can be seen by in-spection of Fig. 1, which is plotted for a 1ppt concen-tration. The next ten strongest lines share the sameJ rotational quantum numbers as the five strongest,but are in a doubly degenerate excited vibrationalstate whose Boltzmann population is reduced byabout an order of magnitude. Also, included in thisplot are transitions due to isotopologues in naturalabundance, the largest of which is OC34S with anabundance of 4%. The spectrum of C2H3CN in Fig. 2shows not only a greater density of lines, but also amore gradual change in intensity. This is reflected inthe much more gradual decline in the peak absorp-tion coefficient in Fig. 3. This is due to both the muchwider range of strengths of rotational lines in asym-metric rotors and to the larger number of excited vi-brational states.

Species commonlyused todemonstrate rotationallyresolved spectroscopic sensors [1,2,5,7,8,10,11,13,22–28], such as hydrogen cyanide (HCN), carbon monox-ide (CO),nitricoxide(NO),acetylene(C2H2),ammonia(NH3), hydrogen chloride (HCl), and the light asym-metric rotorswater (H2O) andhydrogen sulfide (H2S),have line spacings ranging from about 100GHz tomore than 500GHz. These molecules have spectra103–104 sparser than the C2H3CN spectrum shownin Fig. 2. Additionally, because absorption strengthsdepend on thedipolemomentsandpartition functionsof molecules, lines strengths can vary widely, and thecommonly used demonstration species are oftenseveral orders ofmagnitudestronger thanmanyotherspecies of analytical interest.

While we have chosen the 210–270GHz region be-cause of technical convenience for the sensors, withthe exception of a few relatively light species, anySMM/THz region of comparable bandwidth can ac-cess suitable fingerprints for the vast majority of mo-lecules. The clutter in this spectral region is alsorepresentative.

3. Experimental Approaches to Sensors inthe SMM/THz

Clutter in sensors is not only a function of the scenar-io and spectral region, but also a function of systemimplementation. Because the SMM/THz is at the in-terface between optical and electronic techniques,technical approaches of widely varying suitabilityhave been proposed for the use as SMM/THz gas sen-sors. Moreover, the technical communities are ratherisolated from one another and comparisons are theexception rather than the rule. Accordingly, we willbriefly discuss the system used in this work as wellas other implementations that have been proposedas the basis of gas sensors in the SMM/THz.

A. Electronic cw Systems

The technical approach used in this paper is de-signed to provide an optimal basis for clutter rejec-tion in the SMM/THz [3,4,29]. It does this by usingan electronic frequency reference that provides in ef-fect absolute frequency calibration and zero instru-ment function. As a result, this system operatesvery close to the Doppler limit, and the conclusionsof this paper are primarily determined by the physicsof the SMM/THz rather than by limits associatedwith technical implementation.

Figure 4 shows the system [9]. Its spectral purityarises from the mixing of a low frequency(0–400MHz) sweeping synthesizer and a steppingsynthesizer (8:75–11:25GHz) to provide the drive

Fig. 1. Spectrum of carbonyl sulfide (OCS) at a concentration of 1ppt between 210 and 270GHz.


for a 24×multiplier chain for the 210–270GHz probe.The stepping synthesizer drives the 24× local oscilla-tor chain of the receiver as well. The sweepingsynthesizer typically is centered near 100MHz,and the resulting intermediate frequency (IF) is thuscentered near 2:4GHz. The signal power in the IF isdemodulated by a diode whose response as a functionof power is calibrated. The output of this diode is de-tected with a lock-in amplifier at twice the frequencyof a small frequency modulation imposed by thesweeping synthesizer, providing a second derivativelineshape. The dc level of the detector is also re-corded to provide normalization for the widely vary-ing (500%) power characteristics of the 210–270GHzsystem. Absolute spectral intensity calibration,traceable to molecular (not system) parameters of∼1% results. The gases in this system are containedin a folded absorption cell of about 1m length and2 cm diameter. So as to be Doppler rather than pres-

sure broadened, total pressure is typically in the1–10mTorr range.

The specificity of this system is such that it is pos-sible to sort out a mixture of 32 gases whose spectralcomplexity and density range from that of the rela-tively simple OCS shown in Fig. 1 to the more com-plex spectra of C2H3CN of Fig. 2 with absolutespecificity [9]. By “absolute specificity” we mean thatthe large number of identifiable spectral lines whosefrequencies are known to a fractional uncertainty of∼10−7 and whose fractional intensities are known to∼10−2 makes the accidental reproduction of this spec-tral fingerprint very unlikely. We have quantifiedthis as resulting in a probability of false alarm≪10−10. Actual calculations of this number are modeldependent and vary widely, but all satisfy the≪10−10

criteria.Although less sensitive and not power calibrated,

simpler systems also provide Doppler-limited spectra

Fig. 2. Spectrum of acrylonitrile (C2H3CN) at a concentration of 1ppt between 210 and 270GHz (upper panel) and a 2GHz segment on a30 times expanded scale (lower panel).


and similar clutter characteristics. These include socalled fast scan submillimeter spectroscopy techniquesystems, based on free running backward wave oscil-lator tubes [29] and free running solid state systemsbased on voltage tunable oscillators [4,30]. Indeed,the systems that have been used for the developmentof virtually the entire SMM/THz spectral databaseshare this Doppler-limited characteristic [31–33].

B. Other SMM/THz Technologies

Perhaps the most widely discussed approach for THzgas sensors is that of terahertz time domain spectro-scopy (THz-TDS) [2,8,13,22–24,34,35]. However,

these systems have orders of magnitude lower reso-lution (typically a few thousand times larger than theDoppler limit), determined both by the mechanicallimits of translating a delay line (which provides theresolution in a fashion analogous to Fourier trans-form IR spectroscopy systems) and more fundamen-tally due to the low spectral brightness of the sourcesthat severely limits sensitivity with increased sys-tem resolution [4]. As a result, these systems are ty-pically demonstrated on species such as ammonia,hydrogen chloride, and water at relatively high pres-sure, and the conclusions of this paper do not apply toTHz-TDS systems.

Optical down conversion with photomixers hasalso been demonstrated for the production of SMM/THz radiation [26,36] and applied to the SMM/THzsensor problem [5–7]. However, because the fre-quency and spectral purity of the SMM/THz radia-tion are determined by the difference between twolaser frequencies and both are degraded by the downconversion factor, it is often practice to pressurebroaden lines well beyond the Doppler limit and tochoose target species with relatively sparse spectra.

For example, in a recent study, cigarette smoke,HCN, CO, H2CO (formaldhyde), and H2O were stu-died at pressures of 0.75 and 15Torr [5]. In anotherexample, OC34S (in its natural abundance of 4%) hasbeen observed with a pressure-broadened resolutionof about 50MHz [6]. In another lower resolution ex-ample, the analysis of a mixture of HCN, SO2, andHCl in the presences of water vapor has been re-ported at pressures near 760 and 76Torr [7,37].

When operating in the lower of these pres-sure ranges, photomixer systems have considerably

Fig. 3. (Color online) Distribution of peak absorption coefficients for six gases at 1ppt concentration, as well as for clean and pollutedatmospheres. The intensity of the strongest spectral curve is plotted above the “1”, the second strongest curve above the “2,” etc. Becausethe density of points on these log graphs is very high (except at the extreme left), the individual points are represented by continuouscurves. The horizontal dotted arrows start at the intensity of the strongest curves of ClCN and extend until they meet the graphs of theclean atmosphere (black) and the polluted atmosphere (red). The vertical dotted arrows show on the horizontal axis the number of curvesin each atmosphere that are comparable in strength to the strongest curves of ClCN at a dilution of 1ppt. The resulting number of clutterlines of comparable strength (∼100 and ∼1000, respectively) should be compared to the 105 available resolution elements in the210–270GHz band. For comparison with Fig. 1, the dotted vertical lines locate the five and 15 strongest lines of OCS.

Fig. 4. (Color online) SMM/THz gas sensor based on cw electronicand frequency multiplication technologies.


higher resolution and specificity than the THz-TDSsystems as evidenced by the observation of somesplitings in the H2CO spectrum [5]. As a result, whilethe conclusions reached in this paper do not fully ap-ply to cw photomixer systems, some are much closerto cw electronic systems in their performance than toTHz-TDS systems. However, operation in this lowerpressure regime requires the use of sophisticated fre-quency stabilization and reference schemes such asfemtosecond laser generated optical combs, lockingto cavities, and third lasers, etc. [6,38].

While quantum cascade lasers have been demon-strated as high spectral purity cw sources in theupper THz and locked to optically pumped far-IR la-sers [39], they have more commonly been consideredas local oscillators for the upper THz than as sourcesfor gas sensors. Presumably, this is because: (1) theirfrequency of operation does not extend low enough tobe able to observe the rotational spectra of the large

majority of molecules, (2) broadband frequency con-trol, measurement, and calibration is challenging,and (3) the cryogenic requirements for approachingthe required frequency range are substantial.

4. Atmospheric Clutter

A. Clean Troposphere in the SMM/THz

Table 1 shows the constituents of the clean tropo-sphere from [16], and Fig. 5 shows the resulting clut-ter. The spectral density of the clean troposphere isalso shown in Fig. 3 as a thick black curve, with theassumption that the water concentration is 1%. Inthese graphs, the absorption coefficients are forDoppler-limited lines and 1mTorr pressure, adjustedfor the dilution shown in Table 1. For reference, wewill see below that common analyte gases at dilu-tions of 1ppt have absorption coefficients of ∼10−13.

Fig. 5. (Color online) Atmospheric clutter from a clean troposphere sample between 210–270GHz (upper panel) and on a 30 timesexpanded scale (lower panel).


At this level, Fig. 3 shows that there are about 200clutter lines in the 105 resolvable channels.

The clutter spectrumof Fig. 5was synthesized fromline list catalogs. Spectra of HNO2 andHNO3 were si-mulated by overlaying Doppler-limited Gaussians onthe catalog line peaks obtained from the JPL spectro-scopic database [19]. Similarly, the spectra of H2O,N2O, SO2, CO, NO, NO2, O3, and H2CO were simu-lated using the HITRAN database [18]. As notedabove, such syntheses can be incomplete because ofmissing excited vibrational states and isotopes. How-ever, this is not the case for the atmospheric species.Not only are the atmospheric species favorable clutterspecies with few low-lying vibrational modes, but be-cause they are atmospheric species, they have beenextensively studied and their catalogs aremuchmorecomplete. For example, the strongest clutter lines inthis region are due to HDO in natural abundance(0.015%) and these are included in the catalogs.More-over, even if they had been omitted there are onlyabout 10 of them in the 105 resolution elements. Withthe exception of HNO3 and HNO2, all of the otheratmospheric species are sparse and do not have low-lying vibrational states.

HNO3 has been extensively studied and the rota-tional spectra all of its vibrational states below 5 kT(1000 cm−1) have been analyzed [40–43], and thesehave been included in the simulation of Fig. 5. Theexcited vibrational states of HNO2 have not beenanalyzed but its concentration is 300 times smallerthan that of HNO3, so their contributions to clutterwill be very low.

The most notable feature of these graphs is that atthe Doppler limit they are mostly white space, devoidof clutter. Clearly the clutter limit is far below 1ppt,but we have insignificant spectroscopic knowledge todetermine where. SMM/THz sensors in the clean tro-posphere will be sensitivity limited rather than clut-ter limited. Moreover, the relative intensities of therotational spectral lines are known to high accuracyand can be calibrated or subtracted down to the noisefloor of the sensor, thereby gaining several additionalorders of magnitude relative to clutter. For this sub-traction approach, the precise frequency accuracy ofthe cw electronic sources is important [9].

B. Polluted Urban Troposphere in the SMM/THz

Contributions from species with relatively sparsespectra. While in the SMM/THz most of the earth’ssurface will have clutter characteristics similar tothat of the clean troposphere, many scenarios areless favorable. A common and well-documentedone is the polluted urban atmosphere shown inTable 1 [16]. Figure 6 shows the clutter from thesespecies, assuming the upper limits in concentrationsgiven in Table 1. Again, most of the species here(H2O, SO2, CO, N2O, NO, NO2, O3, NH3, HCOOH,HNO2, HNO3, and H2CO) have relatively sparsespectra in terms of the number of Doppler-limited re-solution elements occupied. Thus, while the spectrallines of these species will be stronger in the urban

atmosphere than shown in Fig. 5 for the clean tropo-sphere, their contribution to the clutter limit will stillbe small because their fingerprints are concentratedin such a small fraction of the available resolutionelements. As noted above, HNO3 is denser but stillreasonably sparse. Figure 3 shows the distributionof these clutter intensities. It also shows at the10−10 cm−1 level characteristic of the six target gasesat concentrations of 1 part per billion (ppb) that thereare fewer than 100 clutter lines. Indeed, even at the10−13 cm−1 level characteristic of analytes at the 1pptlevel, there are only about 3000 clutter lines in the105 available resolution channels. As expected, themost notable clutter features are due to HNO3.The spectra region in the lower panel of Fig. 6 waschosen at an HNO3 band head so as to maximizethe clutter concentrations. However, even on thiscompressed scale, at the 10−13 absorption coefficientlevel, the spectrum is mostly white space.

This graph, however, does not include contribu-tions from excited vibrational states of HNO3 above5 kT, the unstable PAN, or nonmethane hydrocar-bons, and these must be considered.

Rotational spectra in the excited vibrational statesof HNO3. A good estimate can be made based on ourexperimental spectroscopic studies of this molecule.The issue here is that as one goes to ever greater sen-sitivities, the rotational spectra in excited vibra-tional states (which are weaker by the vibrationalBoltzmann factor e−Ev=kT) can contribute to clutter.Indeed, when one looks at the experimental spectra(one cannot use a theoretical model here becausemany of the high lying vibrational states are heavilyperturbed and have not yet been successfully anal-yzed), the line density increases rapidly. However,it has been possible to observe and assign many ro-tational lines in the ν2 vibrational state of HNO3 at1700 cm−1. These lines are weaker than ground statelines by a factor of 5000. Since HNO3 was assumed tohave a concentration of 50ppb in this cluttered atmo-sphere, for an analyte whose spectroscopic propertiesare similar to nitric acid, its clutter limit would besubstantially below 10ppt. Since the ν2 assignmentswere done visually and without any attempt to sub-tract the clutter spectrum, the true limit is severalorders of magnitude lower.

Nonmethane hydrocarbons. Because the composi-tion of the nonmethane hydrocarbon mix is un-known, it is only possible to consider a worst-caselimit. For such an analysis, we first note that theirtotal concentration is about 1ppm. The worst-caseanalysis would be that this unknown mixture wouldcause all of the 105 channels to randomly have spec-tral intensities from one or more of the species.However, this increase in the number of lines corre-spondingly reduces the strength of individual lines.Thus, a worst-case assumption is to assume a ran-dom distribution of line intensities among the reso-lution channels of the sensor, but at an intensityreduced not only by their 1ppm total concentration,but also by the need to distribute their intensity over


105 resolution elements. Thus, if an analyte has a fin-gerprint of 103 lines, the absorption coefficients ofthese lines would be 100 times stronger than thatof the clutter lines, corresponding to a clutter limitof 10ppb. Again, if one uses the numerical techniquesconsidered above to use the information in all 103lines of the analyte, the detection limit would bemuch lower. This is probably a very conservative lim-it, because lists of typical nonmethane hydrocarbonsshow that many of the most abundant are symmetricand thus have no rotational spectra in the SMM/THz[44,45]. Additionally, the atomic make up of thesespecies typically leads to small dipole moments forthe nonsymmetric species.

A corollary of this is that if the analyte of interest isa very large molecule that by itself would approxi-mately fill all available channels, each of its lines willalso be reduced by this dilution factor. In this limit,the simple visual identification limit would also be∼1ppm. However, in this limit, numerical techni-

ques, based on the known relative intensities ofthe many (∼105) lines of the analyte would make pos-sible detection at much lower levels against the noiseassociated with the clutter from the many species.The study of this very large molecule limit andhow far into this limit numerical techniques to dealwith it are feasible are important research questionsfor the SMM/THz sensors.

The unstable species, PAN. The SMM/THz spec-trum of PAN is unknown, at least in part becauseit is not a stable molecular species. We believe thattwo factors will cause PAN to not contribute signifi-cantly to clutter in the SMM/THz. First, because it isunstable, we expect that atmospheric sampling tech-niques can be implemented to cause its decomposi-tion into smaller molecules with relatively sparsespectra. Indeed, if one were to undertake the labora-tory spectroscopy of this species, one would expect tohave to take special precautions to insure its stabi-lity. Second, if it survived, it is subject to the same

Fig. 6. (Color online) Atmospheric clutter in a polluted urban atmosphere between 210–270GHz (upper panel) and on a 30 timesexpanded scale (lower panel).


worst-case scenario as the nonmethane hydrocar-bons, except that its maximum concentration issmaller by a factor of about 30.

C. Collection and Preconcentration

All of the above assumes that the SMM/THz sensorsimply ingests an air sample and pumps the pressuredown to about 1Pa (∼10−5 atm). However, becausethis optimum pressure is so low, it is common touse some preconcentration technique (sorbents, cryo-trapping, etc.) to eliminate atmospheric components,such as nitrogen and oxygen, and increase the sensorsensitivity. If we assume that presentation is neutralbetween the target and clutter species, none of theabove results changes.

It is common in less specific sensors to use collec-tors and preconcentrators that reduce the concentra-tions of the clutter species relative to the targetspecies. However, since the clutter limits set hereare so low, we do not include any preselection factorin our considerations. We should note, however, thatbecause of the very low optimum pressure of cw elec-tronic sensors, preconcentration factors of 105 arefeasible even with the sampling of small (on the orderof 1 l) samples of atmospheric air. As a result, the dis-cussion of clutter limits below 1ppt is not a purelyacademic exercise. For example, using a sorbent asa preconcentrator we have recently demonstrateda general purpose, compact cw electronic system en-tirely contained in a 1 ft3 volume whose acetonitriledetectivity limit is ∼2ppt [46]. A somewhat largersystem or one with a different optimization couldbe considerably more sensitive.

5. Results and Examples

Although it is possible to use the information in Fig. 3to gain a statistical idea of the impact of clutter, it isuseful to use the spectra to directly look at selectedfingerprints in relation to the clutter. We will show

the results graphically and without the use of numer-ical techniques for target gas recovery. However, withthe high frequency accuracy of intensity-calibratedcw electronic SMM/THz systems, significant addi-tional gains can be achieved with numerical techni-ques. For example, using the system described abovewe have recently shown additional overlap rejectionof a factor of 100 [9].

There are too many combinations of clean andpolluted atmospheres, target gases, and level of gra-phical expansion for the gases discussed above fordisplay in this paper. While we have examined allof these combinations, we will only show a limitedset here. These will be selected to illustrate specificpoints, with selections weighted towards the mostchallenging cases.

Although the topic of this paper is the limit set byatmospheric clutter, it is useful to also consider thesensitivity limits of the spectrometer. While this lim-it is a complicated combination of spectrometer de-sign, cell size, and integration time, we have recentlydemonstrated a compact cw electronic system en-tirely contained in a 1 ft3 volume whose acetonitriledetectivity limit is ∼2ppt. This system used a sor-bent as a preconcentrator. While it is possible to se-lect sorbent materials that would selectively enhancethe target gas relative to the clutter gas (and thusresult in lower clutter limits than shown in thispaper), we have not added this variable to our con-siderations. Since it is possible to build a largerand more sensitive sensor system, such a preconcen-trator strategy may be useful, especially in heavilypolluted atmospheres.

A. Target Molecules Diluted in the Clean Troposphere inthe SMM/THz

Species with relatively sparse spectra. The lines ofOCS can be seen in Fig. 1 to come in clumps, spacedby 12GHz starting near 219GHz. Figure 7 of the

Fig. 7. (Color online) 1ppb of OCS (black) in the clean troposphere (red) between 218 and 220GHz.


region around 231GHz shows the individual lines ofone of these clumps, the smaller lines of which aredue to the aforementioned relatively low-lying vibra-tional states and 13C isotopic species. This region waschosen as a worst case because it coincides with aband head of the major atmospheric clutter species,HNO3. As predicted statistically by Fig. 3, even whenthe OCS concentration is reduced to 1ppt, there islittle or no clutter interference. The spectrum ofClCN and the impact of clutter on it are very similar,except that the lines of ClCN are about 25 timesstronger because of its larger dipole moment.

The spectrum of CH3CN consists of clumps of lines,with spacings of ∼18GHz, starting near 220GHz.These clumps are denser than those of OCS becausein addition to the contributions of excited states andminor isotopic species, the symmetric top CH3CNalso has a K structure. Figure 8 shows the portionof the spectrum of CH3CN in clutter between 256and 258GHz. For this figure, the cluster centeredon 257GHz was chosen as a worst-case because it co-incides with another band head of HNO3.

Inspection of Fig. 8 shows that the 1ppb concentra-tion could be reduced to 1ppt and still be above theclutter floor. Again, this is consistent with the statis-tical prediction of Fig. 3. While it is almost certainlytrue that our spectroscopic knowledge of the clutterspecies is such that these plots do not include all ofthe clutter down to the 10−15 level (0:001ppt), it isalso true that Fig. 8 is still on a compressed graphicalscale and the amount of open spectral space is muchgreater than displayed here.

Species with more complex spectra. The minimalimpact of clutter in the cw electronic SMM/THz sys-tems and their higher resolving power for target spe-cies allow us to consider species that are not typicallyused for sensor demonstration. Here we consider theTICs: acrolein, acrylonitrile, and ethylene oxide.

Figure 9 shows the spectrum of C2H3CN on an ex-panded scale near one of the HNO3 band heads.Although these lines are marginally weaker thanthe species shown in Figs. 7 and 8, the atmosphericclutter is again so sparse even at this smaller signallevel as to have no impact on the observation of thesespecies. The clutter limit for these difficult species isagain at near or below 1ppt. Similar spectral plots foracrolein and ethylene oxide lead to the same conclu-sion. Interestingly, for cw electronic systems operat-ing in the Doppler limit, the probability of falsealarm for these species is actually smaller than forthe sparser species because of the greater redundan-cies of their fingerprints.

Figure 3 provides a statistical overview. At 1pptconcentrations, all six of these species have at least15 lines whose absorption coefficients are 10−14 cm−1

or greater. In the clean troposphere, there are fewerthan 100 lines in the 105 channels at this level. At the10−13–10−14 cm−1 characteristic of 1ppt analyte le-vels, there are only a few hundred clutter lines.

B. Target Molecules Diluted in the Polluted Tropospherein the SMM/THz

Simple species in a polluted urban atmosphere.Fig. 10 shows the spectra of OCS, ClCN, and CH3CNin a polluted atmosphere between 267.8 and268:3GHz, a strongly cluttered region. This smallerinterval is used because the linewidths of the nowmuch stronger clutter lines need to be consideredfor the largest dilutions of the target lines. However,in each of these cases, the figures show that the con-centrations of each of the target gases could be re-duced to 1ppt and they could still be identifiedvisually. Because there are many similar intervals(most of which are less cluttered), these speciescan be identified with very high confidence (probabil-ity of false alarm ≪10−10) even at these ppt levels.

Fig. 8. (Color online) 1ppb of CH3CN (black) in the clean troposphere (red) between 256 and 258GHz.


More complex species in a polluted urban atmo-sphere. Figures. 11 and 12 show acrylonitrile on ex-panded scales in a region of maximum clutter and aregion of less clutter. Inspection of these figuresshows that in the region of maximum clutter, the1ppt visual recovery is marginal, but in the regionof less clutter, the 1ppt recovery is straightforward.Similar results can be obtained for either acrolein orethylene oxide.

All of the figures above have been plotted in theDoppler limit. Since there is a wide range of spectralintensities, it is useful to consider the impact of thesignificantly broader wings of strong spectral linesthat contain a Lorentizian component due to pres-sure broadening. Accordingly, in Fig. 13 we replottedthe Doppler spectrum of Fig. 12, but assumed a10mTorr pressure and a pressure broadening pa-rameter of 10MHz=Torr. Comparison of the two

figures shows that the higher pressure assumed inFig. 13 increases the strength of all of the lines byabout an order of magnitude. As expected, it alsobroadens the far wings of the lines. However, the fig-ure shows that it has little impact on the observabil-ity of the trace species. Additionally, although wehave not discussed the details of THz spectroscopicimplementations in this paper, it is common practiceto observe either the first or second derivative of thelineshape. This will further discriminate against theclutter contributions of the Lorentzian wings.

Figure 3 again provides a statistical overview. Atthe 1ppb level, all of the gases have 15 or more lineswhose absorption coefficient are 10−11 or greater. Atthis level there are a few hundred lines due to thepolluted atmosphere. At the 1ppt level, there areabout 5000 lines clutter lines of correspondingstrength that occupy the 105 available channels.

Fig. 9. (Color online) 1ppb of acrylonitrile (black) in a clean troposphere (red) on an expanded scale between 230 and 232GHz.

Fig. 10. (Color online) 1ppb of ClCN (green), OCS (black), and CH3CN (blue) in a polluted troposphere (red) between 267.8 and268:3GHz.


6. Clutter in the Op/IR

There has been a substantial effort to develop Op/IRgas sensors, and it is well known that atmosphericclutter can have an adverse effect. Since we haveshown above that atmospheric clutter has almostno impact on Doppler-limited SMM/THz sensorsdown to 1ppt concentration levels for the targetgas, it is useful to briefly explore the reasons for thisdifference.

Clutter is much more of an issue in the Op/IR thanin the SMM/THz for two reasons. First, the Dopplerwidth and instrument function are at least 102 andsometimes 104 larger. Secondly, gases that are abun-dant in the atmosphere, notably H2O and CO2, havestrong IR spectra, but little or no SMM/THz spectraat Doppler resolution.

As we have noted above, because the rotational sig-natures of acrolein, acrylonitrile, and ethylene oxideand other species similar to those shown in Figs. 2are complex and relatively dense, Op/IR systemsare typically demonstrated on simpler moleculeswith much sparser spectra. Among recent sophisti-cated systems are (1) optical comb/cavity ring downsystems that have been demonstrated on O2, H2O,CO, C2H2, and NH3 with pressure-broadened line-widths of a few gigahertz and instrument resolutionof 25GHz [11,28]; (2) optical comb observations ofCO2, CH4, C2H2, and NH3 at 250Torr to provide alinewidth of 1:8GHz to match the grating resolutionof 800MHz [10]; and (3) pure frequency comb obser-vations of I2 at 633nm with 1:2GHz instrumentresolution [47].

Fig. 11. (Color online) 1ppt of acrylonitrile (black) in a region of maximum clutter in the polluted troposphere (red).

Fig. 12. (Color online) 1ppt of acrylonitrile (black) in a region of less clutter in the polluted troposphere (red).


Even for these sparse species the impact of cluttercan be significant. For example, Thorpe et al. foundthat of the more than 170 spectral features of CO intheir 1:55–1:61 μm window, only one did not overlapwith interfering absorptions of H2O and CO2 [10].These two species have no interfering spectral linesin the ∼100; 000 resolution elements between210–270GHz.

Alternatively, it is possible to build sensors basedon the detection of rotationally unresolved bands. Forexample, a cavity ringdown observation of bands ofCH3NH2 and ðCH3Þ2 has been reported. However,if these gases were combined with gases from humanbreath H2O and CO2, clutter reduced the detectionlimit by about a factor of 50 [12].

7. Summary and Conclusions

The much smaller Doppler widths in the SMM/THzcombined with the fortuitous spectral characteristicsof the major (H2O and CO2) and many of the minoratmospheric constituents significantly reduce theclutter limit for point sensors in this spectra regionrelative to that of the Op/IR. In absolute terms, theclutter limit ranges from ≪1ppt for favorable targetspecies in the clean troposphere to ∼1ppt for lessfavorable species in a standard polluted atmosphere.These limits are for simple visual identification.Numerical approaches can further reduce these lim-its. In order to take advantage of the small Dopplerwidth and highly specific rotational fingerprinting,cw electronic techniques are substantially superiorto either THz-TDS or THz photomixer systems.Additionally, the good spectral purity of cw electronicsystems in not a function of the sensor size, cost, orcomplexity.

Because this limit is so low, the real challenges forDoppler-limited SMM/THz sensors will not be atmo-spheric clutter, but rather sensitivity and the poten-tial for interference among large molecules in theanalytical mixture. Consequently, one of the majorresearch issues for this field is the exploration oflarger species with very dense spectra and the devel-opment of techniques to deal with them.

We would like to thank the Army Research Officeand the Defense Advance Research Projects Agency(under contract W911NF-09-1-0428), and the Semi-conductor Research Corporation for their supportof this work.

We would also like to thank the editor for his sug-gestion that we consider the impact of the Lorentziancomponents of pressure-broadened lines.

References1. R. H. Jacobsen, D. M. Mittleman, and M. C. Nuss, “Chemical

recognition of gases and gas mixtures using terahertz waves,”Opt. Lett. 21, 2011–2013 (1996).

2. D. M. Mittleman, R. H. Jacobsen, R. Neelamani, R. G.Baraniuk, and M. C. Nuss, “Gas sensing with terahertztime-domain spectroscopy,” Appl. Phys. B 67, 379–390 (1998).

3. S. Albert, D. T. Petkie, R. P. A. Bettens, S. P. Belov, and F. C.De Lucia, “FASSST: a new gas-phase analytical tool,” Anal.Chem. 70, 719A–727A (1998).

4. I. Medvedev, M. Behnke, and F. C. De Lucia, “Fast analysis ofgases in the submillimeter/terahertz with “absolute” specifi-city,” Appl. Phys. Lett. 86, 154105 (2005).

5. D. Bigourd, A. Cuisset, F. Hindle, S. Matton, E. Fertein, R.Bocquet, and G. Mouret, “Detection and quantification ofmultiple molecular species in mainstream cigarette smokeby continuous-wave terahertz spectroscopy,” Opt. Lett. 31,2356–2358 (2006).

6. F. Hindle, C. Yang, G. Mouret, A. Cuisset, R. Bocquet, J.-F.Lampin, K. Blary, E. Peytavit, T. Akalin, and G. Ducournau,

Fig. 13. (Color online) 1ppt of acrylonitrile (black) in a region of less clutter in the polluted troposphere (red). This figure includes apressure-broadening contribution that assumes a total pressure of 10mTorr and a pressure-broadening coefficient of 10MHz=Torr.


“Recent developments of an opto-electronic spectrometer forhigh-resolution spectroscopy,” Sensors 9, 9039–9057 (2009).

7. A. Majewski, “Terahertz spectroscopy: high-resolution tera-hertz spectrometer sniffs out chemicals,” Laser Focus World44(4) (2008).

8. R. Guo, K. Akiyama, H. Minamide, and H. Ito, “Frequency-agile terahertz-wave spectrometer for high-resolution gassensing,” Appl. Phys. Lett. 90, 121127 (2007).

9. I. R. Medvedev, C. F. Neese, G. M. Plummer, and F. C.De Lucia, “Submillimeter spectroscopy for chemical analysiswith absolute specificity,” Opt. Lett. 35, 1533–1535 (2010).

10. M. J. Thorpe, D. Balslev-Clausen, M. S. Kirchner, and J. Ye,“Cavity-enhanced optical frequency comb spectroscopy:application to human breath analysis,” Opt. Express 16,2387–2397 (2008).

11. M. J. Thorpe, K. D. Moll, R. J. Jones, B. Safdi, and J. Ye,“Broadband cavity ringdown spectroscopy for sensitive andrapid molecular detection,” Science 311, 1595–1599 (2006).

12. D. Marinov, J. M. Rey, M. G. Muller, and M. W. Sigrist,“Spectroscopic investigation of methylated amines by a cavity-ringdown-based spectrometer,” Appl. Opt. 46, 3981–3986(2007).

13. J. Shao, L. Lathdavong, P. Thavixay, and O. Axner, “Detectionof nitric oxide at low ppb m concentrations by differentialabsorption spectrometry using a fully diode-laser-based ultra-violet laser system,” J. Opt. Soc. Am. B 24, 2294–2306 (2007).

14. P. Werle, “A review of recent advances in semicondutor laserbased gas monitors,” Spectrochim. Acta A54, 197–236 (1998).

15. T. J. Geyer, G. M. Plummer, and T. A. Dunder, “FTIR Technol-ogy Development,” http://www.epa.gov/ttnemc01/ftir/reports/entrop01.html.

16. J. H. Seinfeld, Atmospheric Chemistry and Physics of AirPollution (Wiley, 1986).

17. B. P. Petrucelli, T. L. Barker, and J. A. Woods, USACHPPMTechnical Guide 273 (U.S. Army Center for Health Promotionand Preventive Medicine, Aberdeen, MD).

18. L. S. Rothman, I. E. Gordon, A. Barbe, D. C. Benne, P. F.Bernath, M. Birk, V. Boudon, L. R. Brown, A. Campargue,J.-P. Champion, K. Chance, L. Coudert, V. Dana, V. M. Devi,S. Fally, J.-M. Flaud, R. R. Gamache, A. Goldman, D.Jacquemart, I. Kleiner, N. Lacome, W. J. Lafferty, J.-Y.Mandin, S. T. Massie, S. N. Mikhailenko, C. E. Miller, N.Moazzen-Ahmadi, O. Naumenko, A. V. Nikitin, J. Orphal, V. I.Perevalov, A. Perrin, A. Predoi-Cross, C. P. Rinsland, M.Rotger, M. Simecková, M. A. H. Smith, K. Sung, S. A. Tashkun,J. Tennyson, A. C. Vandaele, and J. Vander Auwera, “TheHITRAN 2008 molecular spectroscopic database,” J. Quant.Spectrosc. Radiat. Transfer 110, 533–572 (2009).

19. H. M. Pickett, R. L. Poynter, E. A. Cohen, M. L. Delitsky, J. C.Pearson, and H. S. P. Muller, “Submillimeter, millimeter, andmicrowave spectral line catalog,” J. Quant. Spectrosc. Radiat.Transfer 60, 883–890 (1998).

20. H. S. P. Muller, F. Schloder, J. Stutzki, and G. Winnewisser,“The Cologne database for molecular spectroscopy,” J. Mol.Struct. 742, 215–227 (2005).

21. S. M. Fortman, I. R. Medvedev, C. F. Neese, and F. C. De Lucia,“A new approach to astrophysical spectra: the complete ex-perimental spectrum of ethyl cyanide (CH3CH2 CN) between570 and 645GHz,” Astrophys. J. 714, 476–486 (2010).

22. M. van Exter, C. Fattinger, and D. Grischkowsky, “Terahertztime-domain spectroscopy of water vapor,” Opt. Lett. 14,1128–1130 (1989).

23. R. A. Cheville and D. R. Grischkowsky, “Far-infrared terahertztime-domain spectroscopy of flames,”Opt. Lett. 20, 1646–1648(1995).

24. H. Harde, J. Zhao, M. Wolff, R. A. Cheville, and D. R.Grischkowsky, “THz time-domain spectroscopy on ammonia,”J. Phys. Chem. A 105, 6038–6047 (2001).

25. Y. A. Bakhirkin, A. A. Kosterev, C. Roller, R. F. Curl, and F. K.Tittel, “Mid-infared quantum cascade laser based off-axisintegrated cavity output spectroscopy for biogenic nitric oxidedetection,” Appl. Opt. 43, 2257–2266 (2004).

26. G. Mouret, S. Matton, R. Bocquet, F. Hindle, E. Peytavit, J. F.Lampin, and D. Lippens, “Far-infrared cw difference-frequency generation using vertically integrated and planarlow temperature grown GaAs photomixers: application toH2S rotational spectrum up to 3THz.,” Appl. Phys. B79,725–729 (2004).

27. J. Manne, O. Sukhorukov, W. Jager, and J. Tulip, “Pulsedquantum cascade laser-based cavity ring-down spectroscopyfor ammonia detection in breath,” Appl. Opt. 45, 9230–9237(2006).

28. M. J. Thorpe, D. Hudson, K. D. Moll, J. Lasri, and J. Ye,“Cavity-ringdownmolecular spectroscopy based on an optical fre-quency comb at 1:45–1:65mm.,” Opt. Lett. 32, 307–309 (2007).

29. D. T. Petkie, T. M. Goyette, R. P. A. Bettens, S. P. Belov, S.Albert, P. Helminger, and F. C. De Lucia, “A fast scan submil-limeter spectroscopic technique,” Rev. Sci. Instrum. 68,1675–1683 (1997).

30. I. Medvedev, M. Behnke, and F. C. De Lucia, “Chemicalanalysis in the submillimeter spectral region with a compactsolid state system,” Analyst 131, 1299–1307 (2006).

31. F. C. De Lucia, “Spectroscopy in the terahertz spectral region,”in Sensing with Terahertz Radiation, D. Mittleman, ed.(Springer, 2003), pp. 39–116.

32. F. C. De Lucia, “Science and technology in the submillimeterspectral region,” Opt. Photonics News 14, 44–50 (2003).

33. F. C. De Lucia, “The submillimeter: a spectroscopist’s view,” J.Mol. Spectrosc. 261, 1–17 (2010).

34. R. A. Cheville and D. Grischkowsky, “Observation of pure ro-tational absorption spectra in the n2 band of hot H2O inflames,” Opt. Lett. 23, 531–533 (1998).

35. B. Ferguson and X.-C. Zhang, “Materials for terahertz scienceand technology,” Nat. Mater. 1, 26–33 (2002).

36. E. R. Brown, K. A. McIntosh, K. B. Nichols, and C. L. Dennis,“Photomixing up to 3:8THz in low-temperature-grown GaAs,”Appl. Phys. Lett. 66, 285–287 (1995).

37. A. Majewski, A. Turner, and M. Wraback, “A high resolutionterahertz spectrometer for chemical detection,” presented atthe 2008 International Symposium on Spectral Sensing Re-search, Hoboken, New Jersey, 23–27 June 2008).

38. S. Matsuura, P. Chen, G. A. Blake, J. C. Pearson, and H. M.Pickett, “A tunable, cavity-locked diode laser system for tera-hertz photomixing,” IEEEMicro. Th. Tech. 48, 380–387 (2000).

39. A. L. Betz, R. T. Boreiko, B. S. Williams, S. Kumar, Q. Hu, andJ. L. Reno, “Frequency and phase-lock control of a 3THz quan-tum cascade laser,” Opt. Lett. 30, 1837–1839 (2005).

40. R. A. Booker, R. L. Crownover, F. C. De Lucia, and P. Helmin-ger, “Millimeter- and submillimeter-wave spectra of the NO′

stretching mode (n6) in nitric acid,” J. Mol. Spectrosc. 128,306–308 (1988).

41. R. A. Booker, R. L. Crownover, F. C. De Lucia, and P.Helminger, “Millimeter- and submillimeter-wave spectra ofthe ONO′ bending mode (n7) in nitric acid,” J. Mol. Spectrosc.128, 62–67 (1988).

42. R. L. Crownover, R. A. Booker, F. C. De Lucia, and P.Helminger, “The rotational spectrum of nitric acid: the firstfive vibrational states,” J. Quant. Spectrosc. Radiat. Transfer40, 39–46 (1988).

43. D. T. Petkie, P. Helminger, M. Behnke, I. Medvedev, and F. C.De Lucia, “The rotational spectra of the 7191, 6191, and 72 vi-brational states of nitric acid,” J. Mol. Spectrosc. 233,189–196 (2005).

44. D. Kang, V. P. Aneja, R. G. Zika, C. Farmer, and J. D. Ray, “Non-methane hydrocarbons in the rural southeast United Statesnational parks,” J. Geophys. Res. 106, 3133–3155 (2001).


http://www.epa.gov/ttnemc01/ftir/reports/entrop01.html





45. A. K. Baker, F. Slemr, and C. A. M. Brenninkmeijer, “Analysisof non-methane hydrocarbons in air samples collected aboardthe CARIBIC passenger aircraft,” Atmos. Meas. Tech. 3,311–321 (2010).

46. C. F. Neese, I. M. Medvedev, G. M. Plummer, A. Frank, C. Ball,and F. C. De Lucia are preparing a manuscript to be called

“Compact submillimeter/terahertz gas sensor with absolutespecificity and ppt sensitivity.”

47. S. A. Diddams, L. Hollberg, and V. Mbele, “Molecularfingerprinting with the resolved modes of a femto-second laser frequency comb,” Nature 445, 627–630(2007).


Documents

Impact of atmospheric clutter on Doppler-limited gas sensors in the submillimeter/terahertz