JOURNAL OF MOLECULAR SPECTROSCOPY
ARTICLE NO.
175, 37–47 (1996)
0006
A Terahertz Photomixing Spectrometer:
Application to SO2 Self Broadening
A. S. Pine, R. D. Suenram,* E. R. Brown,† and K. A. McIntosh†
*Molecular Physics Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899; and
†Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, Massachusetts 02173-9108
Received July 7, 1995; in revised form October 3, 1995
A new tunable cw far-infrared difference-frequency spectrometer has been developed by mixing two single-mode
visible dye lasers separated by terahertz frequencies in an ultrahigh-speed photoconductor. The photomixer device is
fabricated with submicrometer interdigital electrodes driving a broadband self-complementary spiral antenna on epitaxial
low-temperature-grown (LTG) GaAs material. The LTG-GaAs material has subpicosecond recombination lifetimes,
facilitating the ultrafast response of the mixer. The coherent far-infrared radiation is coupled out of the rear surface of
the GaAs substrate using an aplanatic Si lens and is detected by a conventional Si-composite bolometer at T Å 4.2 K.
We have obtained usable signal-to-noise ratios for radiation from 0.1 to 1.0 THz using 584-nm dye laser pumps with
the prospect of reaching 2 or 3 THz with pump wavelengths closer to the GaAs bandgap. The instrumental linewidth
is currently about 2 MHz, limited by the jitter of the dye lasers. Baseline normalization is necessary to compensate for
amplitude fluctuations of the lasers and for the strong fringes arising from standing waves between the source and
detector and within the detector dewar. The instrument has been used to obtain SO2 self-broadening coefficients for
pure rotational transitions in several Q branches. q 1996 Academic Press, Inc.
or Schottky-barrier diodes (16–19). Two CO2 lasers can be
mixed in metal–insulator–metal (MIM) point-contact diodes (20) and further mixed with microwaves to provide
greater tunability (21). Harmonic generation from microwave klystrons, Gunn oscillators, and other solid state
devices also access portions of this range as well as fundamental radiation from carcinotrons (22) and other backwardwave oscillators (BWOs) (23, 24). Most of these microwavetype systems use fast Schottky diodes as harmonic generators
or mixers for phase locking. Low-temperature bolometers
are generally used to detect the radiation from these farinfrared sources, although heterodyne detection has been
demonstrated for the far-infrared laser-sideband system (17)
and the BWOs have enough power for photoacoustic (23)
or, perhaps, pyroelectric detection. The resolution, precision,
and accuracy of all these other THz sources currently exceed
those of the photomixer described here, whose linewidth is
limited by laser jitter (which could be reduced with faster
intracavity tuning elements if necessary). Our principal interest in the photomixer is the relative ease and speed of tuning
the broadband visible pump lasers and the prospects for
pumping with compact, efficient, and inexpensive semiconductor diode lasers.
I. INTRODUCTION
A number of broadly tunable coherent mid-infrared radiation sources have been developed for high-resolution molecular spectroscopy using difference-frequency mixing of cw
dye (1–3), Ti:sapphire (4), and diode (5) lasers in various
bulk nonlinear optical crystals. Extension to the far-infrared
by mixing cw CO2 lasers in GaAs has been demonstrated
(6), but with the reduction of conversion efficiency by the
square of the difference frequency in the small signal regime.
The conversion efficiency can be increased to the Manley–
Rowe photon conservation limit with pulsed laser sources
(7–9), usually at the expense of lower resolution. Photomixing, on the other hand, is frequency independent below the
inverse lifetime of the material and RC time constants of
the device (10, 11) and shows great promise for generation
of far-infrared radiation in ultrafast photodetectors. We describe here a cw photomixer spectrometer based on beating
two visible dye lasers in a photoconductor fabricated with
submicrometer interdigital electrodes on low-temperaturegrown (LTG) GaAs (11–13). This may be regarded as the
direct frequency domain analog of the pulsed femtosecond
laser generation of THz electrical signals (14, 15).
Several other tunable coherent far-infrared laser and microwave sources have also been developed for high-resolution molecular spectroscopy. Conventional molecular farinfrared lasers, usually pumped by a high-power CO2 laser,
operate on discrete lines, but can provide tunable sidebands
when mixed with microwave radiation in fast point-contact
II. PHOTOMIXING
With two incident lasers of powers P1 and P2 and frequencies v1 and v2, the power detected by a photovoltaic or photoconductive device is given by
37
0022-2852/96 $18.00
Copyright q 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec
38
PINE ET AL.
FIG. 1. Far-Infrared photomixer spectrometer: Terahertz low-temperature-grown GaAs photoconductor, LTG-GaAs; aplanatic Si lens, Si; dashed line
for horizontal polarization; solid line for vertical polarization; dotted line for circularly polarized far-infrared; beamsplitters, BS (R Å 50%), bs (R Å
4%), Bs (R Å 95%), PBS (polarizing); half-wave plate, HWP; confocal Fabry–Perot interferometer, CFP; photodiode, PD.
P(t) Å P0 / 2[hmP1P2]1/2cos(vt),
[1]
where P0 Å P1 / P2 is the total cw power, v Å 2p(v1 0 v2)
is the angular frequency difference, and hm is the mixing
efficiency arising from the interference of the two laser
fields. hm ranges between 0 and 1 and approaches unity
when the beams have the same polarization and overlap
spatially, and when the relative phase fronts are aligned
parallel within !l/2d, where d is the aperture of the detector. These are not difficult conditions to satisfy experimentally, but small fluctuations in the alignment can be
a significant noise contribution, as we will show. In Eq.
[1], we neglect the sum frequency to which the detector
cannot respond.
The resulting photocurrent can then be shown (10, 11) to
be
i(t) Å i0{1 / 2[hmP1P2/P20(1 / v2t2)]1/2sin(vt / f)},
or number of electron–hole pairs generated per absorbed
photon and is near unity for GaAs, ht is the surface transmission which is É0.67 for the 3.65 refractive index of uncoated
GaAs, ha is the unmasked active area relative to the illuminated area which is É0.9 for the interdigitated electrodes
here, t is the effective electron-hole lifetime, v Å (v1 / v2)/2
is the average optical photon frequency, and ttr is the carrier
transit time between electrodes. The transit time for an electrode spacing of wg, an electron–hole mobility, me,h, and an
applied bias voltage, Vb (part of which may be built into a
pn or Schottky junction or lost due to contact or lead resistance) is
ttr É w2g /(me / mh)Vb.
The ac photocurrent in Eq. [2] generates an rf power at
frequency v of
[2]
Prf(v) Å 2i20RLhc(v)[hmP1P2/P20]/
where the dc photocurrent is
i0 Å hqhthaeP0t/hvttr
[3]
[4]
In these expressions, hq is the internal quantum efficiency
Here RL is the effective load resistance of the circuit or
impedance of the antenna, C is the device capacitance, and
hc(v) represents a (possibly frequency-dependent) antenna
coupling factor. Equation [6] is clearly maximized for equal
power in the incident beams. At low beat frequencies, the
maximum available rf power into a real load, such as a 50V coax termination, is i20RL/2 and is very simple to measure
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
[6]
[(1 / v2t2)(1 / v2R2LC 2)].
and the phase shift is
f Å tan01(1/vt).
[5]
01-23-96 17:00:04
mspa
AP: Mol Spec
39
TERAHERTZ PHOTOMIXER
FIG. 2. Frequency dependence of photomixer output power; (n) present data, 20 1 20 mm, 584 nm; (s) 20 1 20 mm, 780 nm; (h) 8 1 8 mm, 780
nm. The 780-nm data were obtained with a similarly calibrated bolometer at Lincoln Laboratory.
FIG. 3. Empty cell baseline or background scans (two lower traces, bottom trace displaced by 0.3 units for visibility); cell length is 50 cm; upper
trace is ratio of lower traces 10.9.
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f6911
01-23-96 17:00:04
mspa
AP: Mol Spec
40
PINE ET AL.
FIG. 4. Upper trace: a direct transmission spectrum of the rQ5 subbranch of SO2 at P Å 19 Pa. Lower trace: numerical second derivative to suppress
baseline fringes.
either from the dc current or with a real load driving a
broadband oscilloscope or an rf spectrum analyzer. In a typical device here, for P0 Ç 50 mW, we observe i0 Ç 0.5 mA;
so the available power is Ç6 mW. Typical low-temperature
bolometer detectors in the farq infrared have stated noiseequivalent-powers of õ1 pW/ Hz so we expect reasonably
high signal-to-detector-noise ratios.
III. EXPERIMENTAL DETAILS
The high-speed photoconductor mixer used here consists
of a 20 1 20 mm interdigital electrode array, with 0.2-mm
fingers separated by 1.8-mm gaps, coupled to a broadband
3-turn self-complementary spiral antenna fabricated on an
epitaxial layer of LTG-GaAs as described in Ref. (13). The
photocarrier lifetime of the material was t Ç 0.35 psec as
measured by time-resolved photoreflectance (25), using a
mode-locked subpicosecond Ti:sapphire laser, and the photomixer frequency response was obtained by beating cw
fixed-frequency (standing-wave) and tunable (ring) Ti:sapphire lasers (13). For application to high-resolution molecular spectroscopy in the THz regime, we have used a similar
pair of jitter-stabilized, continuously tunable, linear-scancontrolled dye lasers.
A schematic of our optical system is shown in Fig. 1. The
single-mode fixed (standing-wave) and tunable (ring) dye
lasers are both pumped by an argon laser providing Ç3 W
at 515 nm divided roughly equally by a beamsplitter cube.
The dye lasers generate 50 to 100 mW of power near the
peak gain of rhodamine 6G dye at Ç584 nm. We observe
no strong visible wavelength dependence of the photomixing
in the range of this dye. Both dye lasers are jitter-stabilized to
external confocal Fabry–Perot étalons using the commercial
edge-lock servo control. The tunable ring laser output is
horizontally polarized (dashed line), whereas the fixed dye
laser is vertically polarized (solid line); so the bulk of the
tunable radiation from the R Ç 95% beamsplitter (Bs) is
directed through a half-wave plate (HWP) to rotate its polarization to vertical. The two laser beams are then combined
with an R Ç 50% beamsplitter (BS) and focused onto the
LTG-GaAs photomixer with a 25-mm focal length lens. The
far-infrared difference frequency radiation is extracted from
the rear surface of the GaAs substrate and partially collimated by an aplanatic Si lens in physical contact with the
chip. The THz radiation then traverses a 50-cm sample cell
using teflon lenses as windows, whence it is detected by a
Si-composite bolometer operated at T Å 4.2 K. The open
air path is kept õ5 cm to minimize atmospheric water vapor
absorption in certain spectral regions.
Some of the laser diagnostic and control components are
also indicated in Fig. 1. The other half of the visible light
rejected by the beamsplitter combiner is directed to a wavemeter for coarse wavelength measurement ({0.001 nm), a
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec
41
TERAHERTZ PHOTOMIXER
FIG. 5. (a) Survey scan of SO2 at P Å 41 Pa near the rQ8 subbranch ratioed to an empty cell trace. Scan with lasers blocked plotted along zero
transmission baseline. (b) High-resolution scan of rQ8 subbranch head, J Å 13 and 14 lines separated by Ç2.5 MHz.
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f6911
01-23-96 17:00:04
mspa
AP: Mol Spec
42
PINE ET AL.
FIG. 6. Scan of SO2 rQ10 subbranch at P Å 4 and 140 Pa and zero baseline with lasers blocked.
high-resolution confocal étalon (not shown) of free spectral
range Ç300 MHz and finesse Ç100, or a fast Si photodiode
(MRD510) with bandwidth of Ç1 GHz for displaying beats
with an rf spectrum analyzer. This beat enables us to measure
the convolved jitter of the dye lasers pertinent to the farinfrared instrumental linewidth, while the 300-MHz confocal
étalon can apportion the contributions. The ring laser jitter
(full-width-half-maximum) is Ç1 MHz and the standingwave laser jitter is Ç3 MHz, which can be reduced to Ç2
MHz by electronic adjustments that increase the susceptibility of the laser to mode hops. Most of the remaining jitter
is at too high a frequency for the piezoelectric mirror mounts
to follow, but in principle could be reduced using faster
electro-optic phase modulators.
The drift of the fixed dye laser is stabilized relative to a
polarization-stabilized 633-nm He–Ne laser using a ramped
confocal Fabry–Perot étalon of FSR Ç 1500 MHz and finesse Ç50. This same étalon is used to simultaneously monitor and linearly control the scanning of the tunable ring laser
(26) and to generate calibration markers for the far-infrared
difference-frequency spectrum. The horizontally and vertically polarized laser beams are kept on distinct channels
by the polarizing beam splitters and separate photodiode
detectors. This linear scan control increases the precision for
interpolating between markers, and greatly aids the reproducibility of scans for baseline normalization and pressure
dependence, as we will show.
One or both of the lasers are mechanically chopped by
a rotating wheel (Chop) at 200 Hz for synchronous detection of the bolometer signal. Frequency modulation of
the lasers is also a possibility, but we are interested in
lineshapes and broadening coefficients in this study,
which are more directly obtained by amplitude modulation. The bolometer consists of a small Si thermistor chip
bonded to the front of a 5-mm-diameter, 0.1-mm-thick
sapphire substrate with a thin absorbing Bi coating on
the rear surface, with a combined 3-dB electrothermal
frequency response of Ç300 Hz. The bolometer assembly
is mounted behind the 5-mm exit aperture of a parabolic
Winston cone radiation collector. A cold (4.2 K) longwavelength (ú100 mm) pass filter, consisting of a wedged
crystal quartz plate coated with a garnet powder scatterer
and a black polyethylene film, is used to reduce background radiation and scattered visible light. The external
window on the bolometer dewar is wedged Teflon (which
we found to be slightly porous) or TPX plastic.
The noise-equivalent-power (NEP) of the bolometer as
measured by the manufacturer from the dc current – voltage characteristics
and the rms dark current noise is
q
Ç0.13 pW/ Hz at 4.2 K. Ideally, the radiation-absorbing
Bi coating has the proper surface resistivity, r0/(ns 0 1),
to eliminate reflections from the rear surface (27). Here
r0 is the impedance of free space, 377 V per square, and
ns is the sapphire substrate refractive index. However,
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec
43
TERAHERTZ PHOTOMIXER
TABLE 1
SO2 Self-Broadening Coefficients in MHz/kPa for rQ5, rQ8,
and rQ10 Transitions at T Å 296 K
a
b
Statistical uncertainties in parentheses are one standard deviation in terms of the last digits.
Blended lines constrained and averaged with adjacent lines.
even such an ideal coating does not affect reflections from
the window, filters, or front surface of the substrate or
transmission to the reflective metal mounting bracket, all
of which introduce standing waves or baseline channeling
due to interference of the far-infrared radiation, as discussed below.
We crudely calibrated the bolometer responsivity to radiation by illuminating it with Ç85 GHz millimeter radiation obtained from a Gunn diode oscillator, suitably attenuated and monitored with a commercial waveguide power
meter which has a minimum detectable signal of Ç0.01
mW. Coupling from the microwave W-band waveguide to
the bolometer utilized a conical horn and pair of Teflon
lenses. The transfer efficiency of this quasi-optical system
is estimated to be Ç64 { 10%, based on recapture of the
radiation in a pyramidal horn to the waveguide power
meter or video detector. At this frequency, we obtained a
responsivity of the bolometer down from the manufacturer’s specification by a factor of Ç22, with a corresponding
increase in the effective NEP. Though we scanned the
Gunn diode Ç1 GHz without any improvement, this may
be at a local minimum in the standing-wave pattern.
If we assume that the above 85 GHz calibration of the
bolometer is valid for frequencies up to the filter cutoff,
we find that our photomixer generates Ç0.5 mW at lower
frequencies to a few nW at Ç1 THz, as shown in Fig. 2.
For beat frequencies near 85 GHz, we confirmed this
power level using the waveguide power meter with the
pyramidal horn collector. Also shown in Fig. 2 are power
vs frequency measurements with photomixing devices of
the same size and smaller pumped by the Ti:sapphire lasers at Ç780 nm. The improved frequency response for
the smaller (8 1 8 mm) device is expected from its reduced
capacitance. However, the wavelength dependence of the
frequency response is somewhat surprising and may result
from the reduced absorption depth for visible light, or
from longer photocarrier lifetimes or transit times due to
nonparabolicity of the conduction band in the direct gap
and possible indirect gap excitation at the visible pump
wavelengths.
IV. SPECTROSCOPIC CONSIDERATIONS
All of the spectra presented here are normalized against
laser amplitude fluctuations by dividing the bolometer sig-
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec
44
PINE ET AL.
FIG. 7. Observed (solid line) and calculated (dashed) absorbance spectra of the rQ5 subbranch of SO2 at P Å 143 Pa fit to Voigt lineshapes; observed
0 calculated residuals (dotted) displaced by 00.1 for visibility.
FIG. 8. Broadening coefficients vs J for rQKa subbranches with Ka Å 5, 8, and 10. Present data with error bars connected by dashed lines. Calculated
values (30) connected by solid lines.
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f6911
01-23-96 17:00:04
mspa
AP: Mol Spec
45
TERAHERTZ PHOTOMIXER
TABLE 2
SO2 Self-Broadening Coefficients in MHz/kPa for Selected
R-Branch Transitions at T Å 296 K
a
Statistical uncertainties in parentheses are one standard deviation in terms of the last digits.
nal from the far-infrared radiation by the product of the
incident laser powers. We use real-time analog integrated
circuits for the multiplier and divider functions to minimize the stored data channels. Since this analog ratio technique requires closely matched time constants, the cw
power from the tunable dye laser is multiplied with the
chopped fixed dye laser power (derived from the photomixer current), and the ac product is integrated in a
second lock-in whose output is divided into the bolometer
lock-in output. Here we used a 40-ms time constant for
both lock-ins with a 12 dB/octave rolloff, and we digitized
the ratioed signal at 20 Hz for computerized data acquisition and processing. Frequency scan markers were recorded on a 101 finer grid to enhance the calibration
interpolation and precision. A typical scan contains 6000
data points recorded in 5 min. In principle, the normalization could be done by splitting the far-infrared beam into
a sample and reference channel with separate (or timemultiplexed) detectors, as in our mid-infrared differencefrequency spectrometer (1). However, the expense of a
second bolometer and the strong standing-wave patterns
in the separate beams make this less desirable.
An indication of the severity of the standing-wave pattern is shown in Fig. 3. There, the middle trace is a transmission scan from 135 to 180 GHz through an evacuated
cell. The highly structured baseline is not noise and is
quite reproducible, as seen from a second trace displaced
below by 0.3 units for clarity. The fine fringes arise from
the source to detector spacing, which is variable, and the
coarser fringes arise mainly from the optical elements
within the detector dewar. We have been unable to suppress these fringes by tilting the dewar or lenses or by
removing the Winston cone or changing filters or padding
the path with attenuators. Fortunately, the fringes become
less severe at higher frequencies. The top trace in Fig. 3
is the ratio of the bottom two (arbitrarily multiplied by
0.9), removing most of the fringe pattern. The reproducibility of the linear scan rate and the slow thermal or
mechanical variation of the far-infrared fringes make this
possible. However, a noise level of 1 to 2% remains on
the ratioed trace, which we attribute primarily to relative
alignment fluctuations of the two dye lasers due to dye
jet flow instabilities, air currents, and mirror vibrations,
though some may be contributed by laser frequency jitter
on the highly structured baseline discriminant. The
‘‘alignment’’ noise could be greatly reduced with a farinfrared sample/reference ratio as mentioned above, were
it not for the strong fringing in the separate paths. In this
frequency range, the detector noise, obtained by blocking
the lasers, is not visible on this scale. At higher frequency,
when the gain is increased to compensate for the weaker
far-infrared signal, detector noise becomes a factor.
In Fig. 4, we show a portion of the rQ5 subbranch of
SO2 (P Å 19 Pa) near 562 GHz. The upper trace is the
direct transmission spectrum and the lower its second derivative, which is obtained numerically by twice shifting
the trace about a linewidth and subtracting. This method
is effective in suppressing baseline variations when the
spectral features are much sharper than the fringes and
when it is not convenient to pump and fill the cell for each
trace. For lineshape studies, however, when the linewidths
may be comparable to the fringe spacing, derivative techniques are less reliable than an empty cell normalization.
In Fig. 5a, we show a 45-GHz scan of the SO2 (P Å 41
Pa) rQ8 subbranch region near 865 GHz normalized to an
empty cell trace, each recorded in 5 min with 6000 sample
points. Since each point represents 7.5 MHz, it was necessary to pressure broaden the lines slightly to increase the
instrumental response because of the 40-msec lock-in time
constant. The Doppler width for SO2 at this frequency
is only Ç1.3 MHz FWHM. Along the zero transmission
baseline, we also show the detector noise level in this
trace when the lasers are blocked. The rQ8 subbranchhead
is shown expanded in Fig. 5b at lower pressure with a
slower scan rate. A turn around is observed at J Å 14,
whose spacing from the J Å 13 line of Ç2.5 MHz is
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec
46
PINE ET AL.
clearly resolved with the instrumental linewidth of Ç2
MHz here.
The frequencies of these SO2 transitions are extremely
well predicted by the constants determined from the microwave and far-infrared FTIR data (28). The rms deviation of the measured spectral line frequencies from the
predictions is Ç0.6 MHz, which is close to the precision
expected for this dye laser system, judging from the reproducibility of successive traces. This precision reflects the
jitter and residual scan drifts and nonlinearities of the
particular pump lasers and the signal-to-noise ratio of the
measurements and is not inherent in the photomixing process. Of course, the precision limits the absolute accuracy
of the spectrometer, which in turn depends upon the calibration of reference absorption lines. Using the calculated
SO2 spectrum (28) as a reference, we obtain a free spectral
range of 1507.43(1) MHz for our scan control interferometer (CFP in Fig. 1).
In Fig. 6, we show 11.25 GHz scans of the rQ10 subbranch of SO2 just above 1 THz for pressures of 4 and
140 Pa normalized to an empty cell trace. The self-broadening is quite obvious at the higher pressure and the
branchhead is completely blended, effectively saturating
the transmission. Just above 1 THz, the detector noise
level, also shown here, is comparable to the ‘‘alignment’’
noise and establishes a practical limit to the frequency
range of the present spectrometer. Hopefully, smaller devices, faster materials, and longer wavelength pump lasers
will permit us to extend this range to 2 or 3 THz.
V. SELF-BROADENING OF SO2
For the self-broadening study, we recorded a series of
11.25-GHz scans at various pressures of SO2 up to Ç143
Pa (Ç1.1 Torr) of the rQKa subbranches for Ka Å 5, 8, and
10. The lowest pressure traces, near 4 or 5 Pa, were primarily limited to the instrumental and Doppler widths and
the transitions were fit with Gaussian profiles to establish
the effective ‘‘zero’’ pressure resolution limit. The observed lineshapes in the low-pressure traces were slightly
asymmetric, skewed by the scan rate being somewhat too
fast for the time constant. At higher pressures, this skewing was not observable and the lines were well fit to Voigt
lineshapes, fixing the Gaussian component to the average
zero pressure value. Pressures were measured with a capacitance manometer accurate to Ç1% (or 0.3 Pa if
larger).
In Fig. 7, observed and calculated absorbance traces
(log baseline/transmission) for the rQ5 subbranch for P Å
143 Pa of SO2 are shown along with the obs 0 calc residuals dispaced for visibility. These residuals are generally
within the random noise limit, dominated by the alignment
fluctuations in this case. A few R-branch lines are interposed in all these traces and are fit simultaneously with
the regular pattern of Q-branch lines. The Q branchheads
are strongly blended at the higher pressures, so the broadenings of the blended transitions are obtained at lower
pressures, sometimes with constraints that average adjacent lines and fix their relative positions and intensities.
Consequently, the broadening coefficients for the blended
transitions are more poorly determined than for the wellresolved transitions.
The broadening coefficients for the J transitions in the
three Q branches are presented in Table 1 with unreliable
values for the strongly blended branchhead lines suppressed. These broadening coefficients are obtained as a
weighted average (29) of the Lorentzian component halfwidths in the Voigt fits for the various pressure traces.
The uncertainties given in parentheses in the tables are
one statistical standard deviation (29) and do not include
systematic errors in the pressure or frequency calibrations.
Also, undetected blends of isotopic or hot band transitions
may yield some anomalous broadenings without much effect on the error bars.
The results in Table 1 indicate little J dependence for
the broadenings in a given rQKa subbranch, but the Ka
dependence is significant. This is better illustrated in Fig.
8 where we compare our results to some semiclassical
Anderson – Tsao – Curnutte calculations by Tejwani (30).
The Tejwani results exhibit very similar J and Ka dependence though the absolute values are considerably lower
than measured here. We believe the discrepancy can be
attributed principally to the neglect of dipole – quadrupole
and higher multipole forces by Tejwani and, to a small
degree, the use of an SO 2 dipole moment of 5.30 1 10030
Crm (1.59 D), some 3% smaller than (1.634 D) now generally accepted (31). A small sample of R-branch transitions given in Table 2 also exhibit a strong rotational
dependence for the self-broadening coefficients, similar
to the microwave data summarized in Ref. (30) and some
more recent infrared measurements (32, 33), warranting
an updated theoretical study.
ACKNOWLEDGMENTS
We are grateful to W. J. Lafferty for some suggestions on the SO2
spectroscopy. The NIST portion of this work was supported in part by the
NASA Upper Atmosphere Research Program. We also thank K. N. Molvar,
K. B. Nichols, and D. J. Landers for technical assistance.
REFERENCES
1. A. S. Pine, J. Opt. Soc. Am. 64, 1683–1690 (1974); 66, 97–108 (1976).
2. M. G. Bawendi, B. D. Rehfuss, and T. Oka, J. Chem. Phys. 93, 6200–
6209 (1990).
3. A. McIlroy and D. J. Nesbitt, Chem. Phys. Lett. 187, 215–219 (1991).
4. P. Canarelli, Z. Benko, R. Curl, and F. K. Tittel, J. Opt. Soc. Am. B
9, 197–202 (1992).
5. U. Simon, S. Waltman, I. Loa, F. K. Tittel, and L. Hollberg, J. Opt.
Soc. Am. B 12, 323–327 (1995).
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec
47
TERAHERTZ PHOTOMIXER
6. R. L. Aggarwal, B. Lax, H. R. Fetterman, P. E. Tannenwald, and
B. J. Clifton, J. Appl. Phys. 45, 3972–3974 (1974).
7. G. D. Boyd, T. J. Bridges, C. K. N. Patel, and E. Buehler, Appl. Phys.
Lett. 22, 553–555 (1972).
8. B. Lax, R. L. Aggarwal, and G. Favrot, Appl. Phys. Lett. 23, 679–681
(1973).
9. K. H. Yang, J. R. Morris, P. L. Richards, and Y. R. Shen, Appl. Phys.
Lett. 23, 669–671 (1973).
10. R. H. Kingston, ‘‘Detection of Optical and Infrared Radiation,’’
Springer-Verlag, Berlin, 1978.
11. E. R. Brown, F. W. Smith, and K. A. McIntosh, J. Appl. Phys. 73,
1480–1484 (1993).
12. E. R. Brown, K. A. McIntosh, F. W. Smith, M. J. Manfra, and C. L.
Dennis, Appl. Phys Lett. 62, 1206–1208 (1993).
13. E. R. Brown, K. A. McIntosh, K. B. Nichols, and C. L. Dennis, Appl.
Phys. Lett. 66, 285–287 (1995).
14. M. van Exter, C. Fattinger, and D. Grischkowsky, Opt. Lett. 14, 1128–
1130 (1989).
15. H. Harde and D. Grischkowsky, J. Opt. Soc. Am. B 8, 1642–1651
(1991).
16. D. D. Bicanic, B. F. J. Zuidberg, and A. Dymanus, Appl. Phys. Lett.
32, 367–369 (1978).
17. W. A. M. Blumberg, D. D. Peck, and H. R. Fetterman, Appl. Phys.
Lett. 39, 857–859 (1981).
18. J. Farhoomand, G. A. Blake, M. A. Frerking, and H. M. Pickett, J.
Appl. Phys. 57, 1763–1766 (1985).
19. K. L. Busarow, G. A. Blake, K. B. Laughlin, R. C. Cohen, Y. T. Lee,
and R. J. Saykally, J. Chem. Phys. 89, 1268–1276 (1988).
20. K. M. Evenson, D. A. Jennings, and F. R. Petersen, Appl. Phys. Lett.
44, 576–578 (1984).
21. I. G. Nolt, J. V. Radostitz, G. DiLonardo, K. M. Evenson, D. A.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
Jennings, K. R. Leopold, M. D. Vanek, L. R. Zink, A. Hinz, and
K. V. Chance, J. Mol. Spectrosc. 125, 274–287 (1987).
C. Demuynck, J. Mol. Spectrosc. 168, 215–226 (1994).
A. F. Krupnov, in ‘‘Modern Aspects of Microwave Spectroscopy’’
(G. W. Chantry, Ed.), pp. 217–256, Academic Press, New York, 1979.
G. Winnewisser, A. F. Krupnov, M. Yu. Tretyakov, M. Liedtke, F.
Lewen, A. H. Saleck, R. Schieder, A. P. Shkaev, and S. V. Volokhov,
J. Mol. Spectrosc. 165, 294–300 (1994).
E. R. Brown, K. A. McIntosh, K. B. Nichols, M. J. Manfra, and F. W.
Smith, Proc. SPIE Nonlinear Opt. High-Speed Electron. 2145, 200–
206 (1994).
M. J. Coulombe and A. S. Pine, Appl. Opt. 18, 1505 – 1512
(1979).
S. W. McKnight, K. P. Stewart, H. D. Drew, and K. Moorjani, Infrared
Phys. 27, 327–333 (1987).
W. J. Lafferty, G. T. Fraser, A. S. Pine, J.-M. Flaud, C. Camy-Peyret,
V. Dana, J.-Y. Mandin, A. Barbe, J. J. Plateaux, and S. Bouazza, J.
Mol. Spectrosc. 154, 51–60 (1992).
A. S. Pine, J. Chem. Phys. 97, 773–785 (1992).
G. D. T. Tejwani, J. Chem. Phys. 57, 4676–4681 (1972) and private
communication.
K. H. Hellwege and A. M. Hellwege, Eds., ‘‘Landolt–Bornstein Numerical Data and Functional Relationships in Science and Technology,
New Series, Group 2: Atomic and Molecular Physics, Vol. 6: Molecular
constants from Microwave-, Molecular Beam-, and ESR-Spectroscopy,
Section 2.6, Springer-Verlag, Berlin, 1974, sect. 2.6.
F. Kuhnemann, Y. Heiner, B. Sumpf, and K. Herrman, J. Mol. Spectrosc. 152, 1–12 (1992).
O. Fleischmann, B. Sumpf, J. Washull, and H.-D. Kronfeldt, in ‘‘Proceedings, 13th International Conference on High-Resolution Molecular
Spectroscopy,’’ Poznan, 1994.
Copyright q 1996 by Academic Press, Inc.
AID
JMS 6911
/
m4652f$$73
01-23-96 17:00:04
mspa
AP: Mol Spec