pdf, 1.1 MB - University of Massachusetts Lowell

Precisioh terahertz relative redlectometry using a blackbody source
and heterodyne receiver
P. P. Woskov,a) D. R. Cohqa) and S. C. Han
Ksirius Superconductivity Inc., 1110 North GIebe Road, Arlington, Virginia 22201
A. Gatesman, FL H. Giles, and J. Waldman
University of Massachusetts Lowell,b) 4.50 Aiken Street, Lowell, Massachusetts 08154
(Received 23 December 1992; accepted for publication 26 October 1993)
Instrumentation for making localized, precise relative reflection measurements of metals and
superconductors in the terahertz frequency range is demonstrated. The results can be used to
determine the surface resistivity of these materials which is of particular importance to
development of high-temperature superconductors. Emission from a commerically available
1000 “C blackbody source was reflected from the materials under test and a reference reflector.
The reflected signals were detected by a Schottky diode heterodyne receiver using a
CO,-laser-pumped 214.5 pm CH,F, laser as the local oscillator. Double-sideband bandwidth
and receiver noise temperature were 2 GHz and 30 000 K, respectively. The use of a broadband
incoherent diagnostic source minimizes instrumentation sensitivity to coherent interference
effects. The heterodyne receiver provides for sensitive detection with good spatial and frequency
resolution. Unoptimized spatial resolution of four times the diffraction limit was achieved. The
potential for relative reflectivity measurement accuracy of better than 0.1% was demonstrated
with 30 min measurement times though systematic errors limited actual measurement accuracy
to about 0.3%.
I. INTRODUCTION
Accurate measurements of the surface resistivity of
high conductivity materials in the terahertz frequency
range (0.3-30 THz) are of great practical importance. Advances
in communications, computing, and electronics to
higher frequencies, greater bandwidths, and smaller miniaturization
are limited, in part, by ohmic losses in components
such as transmission lines, filters, antennas, and resonators.
Development of instrumentation for measuring
the terahertz surface resistivity of conductors used in such
components can lead to a better understanding of conductor
properties at these high frequencies, provide for a
means of quality control, and facilitate the development of
new materials. Of particular motivation to the present
work is the development of new high critical temperature
(T,) superconductors for applications to high-frequency
electronics. ‘9’
In the terahertz frequency range reflectometry becomes
more advantageous for surface resistivity measurements
in contrast to resonator Q measurements used at
lower frequencies. Cavity and microstrip resonators become
more difficult, if not impractical, to fabricate with
increasing frequency. The surface resistivity R, of a good
conductor is related to the conductor reflectivity R by
R,= (1 -R)Z/4, where Z is.the impedance of the dielectric
medium in which the measurement is made (Z=377
s1 for free space) .3 This simple relation has no dependence
on diffractive, dielectric, or coupling losses that become
more important and uncertain with increasing frequency
ajAl~~ at Plasma Fusion Center, Massachusetts Institute of Technology,
Cambridge, MA 02139.
“Submillimeter Technology Laboratory.
for the resonator based measurement techniques. Other advantages
of reflectometry include the capability for spatial
resolution, applicability over a wide frequency range, and
remote measurements. In addition, an important advantage
for high temperature superconductor measurements is
the apparent scaling of R, as R,- f”. This can make the
determination of much smaller microwave surface resistivities
than possible with the best demonstrated microwave
measurement techniques, by scaling from the terahertz
measurements.4
The main challenge in implementing a terahertz reflectometer
for conductor surface resistivity measurements is
that a measurement accuracy of the order of one part in
lo3 is needed to distinguish differences among the high
conductivity metals such as copper, gold, and aluminum.
An even better accuracy would be desirable if improvements
in high T, superconductors continue. These measurements
by necessity are relative to a known reference
reflector because of the practical difficulty of absolutely
calibrating a reflectometer system to the required accuracy.
In an earlier work we used a far-infrared laser as the
source for the diagnostic terahertz beam and a cryogenic
bolometer for detecting the reflected signals.5 Measurement
accuracy was limited, in part, by the coherent nature
of the diagnostic beam. Slight differences in sample positioning
caused significant signal variations due to coherent
interference effects. This could be a significant limitation
for a possible future spatial scanning terahertz reflectometer.
In this paper we present results using a 1000 “C! blackbody
as an incoherent diagnostic source and a heterodyne
receiver with a far-infrared laser local oscillator for signal
detection. Although a heterodyne receiver operates on a
single mode, it is not a coherent detector because detection
436 Rev. Sci. Instrum. 65 (2), February 1994 0034-6746/94/65(2)/436/7/$6.00 0 1994 American Institute of Physics

To computer
control
BS Beam splitter
C Black body source chopper
FP Flip mirmr
Ll Lens for collimating FIR laser beam
L2 Focusing lens
Diode
Mixer
Bolometar
dvm
Ml Fast off-axis parabolic mirror
M2 Focusing mirror
MS Flat mirror for alignment laser
M4 Fast off-axis parabolic mirror
FIG. 1. Incoherent source/heterodyne receiver terahertz reflectometer
instrument setup. The Dewar is replaced by a sample wheel and the flip
mirror is removed for the room-temperature measurements.
is over a wide instantaneous bandwidth, 2 GHz in this
case, which effectively averages over unwanted interference
peaks and minima. With respect to a bolometer detector it
has a combined advantage in speed, sensitivity, and spatial
resolution without requiring cryogenics.
II. EXPERIMENTAL
SETUP
Two experimental configurations were used. One for
room-temperature measurements and another with a
Dewar for measurements as a function of temperature. The
instrument configuration with the Dewar is shown in Fig.
1 The room-temperature configuration differed only in that
a room-temperature sample holder wheel took the place of
the Dewar. Also the flip mirror (FP) was removed and the
reference reflector (SR) was incorporated as one of four
samples mounted in the wheel.
The far-infrared (FIR) laser used for all the measurements
was a CO,-laser-pumped 214.5 pm ( 1.40 THz)
CH2F2 laser. The diverging, vertically polarized laser beam
was approximately collimated by a 61 cm focal length TPX
lens, Ll. A 6.0-cm-diam aperture directly after this lens
was used to spatially filter small side lobes in the beam
profile. In this way a relatively clean Gaussian beam with
a l/e2 diameter of 34 mm was established. Figure 2 shows
this profile as scanned by a bolometer - 1 m from lens Ll.
The smooth curve is a theoretical Gaussian fit to the
slightly noisy measured profile.
A Mylar beam splitter was used to reflect a part of the
laser beam to the Schottky diode mixer as the local oscillator.
This beamsplitter also passed the signal from the test
samples, therefore its reflectivity had to be no higher than
necessary to saturate the mixer. An optimum beam splitter
thickness of 0.5 mil (25.4 pm) was found, having a measured
reflectivity of 15% at 214.5 pm. The part of the laser
beam not reflected toward the mixer was used to monitor
the laser power with a bolometer.
A Melles Griot 95 mm diameter, 84 mm focal length
off-axis parabolic mirror focused the laser beam onto the
mixer. The mixer was a University of Virginia corner cube
mounted Schottky diode optimized for a wavelength of 192
pm. Though the whisker antenna parameters (length
L=4A, distance from corner d= 1.2A, A= 192 pm) of this
diode were also close to optimum for 214.5 pm, the laser
spot size that was achieved with the given beam diameter
0.1900
214.5 pm Laser Protile
0.1425 t
3
$f 9SOOE-02 -
P
‘3
2
Position (mm)
FIG. 2. The 214.5 pm laser beam profile 1 m from the collimating lens. The slightly noisy curve is the bolometer scan. The smooth curve is the best-fit
Gaussian.
Rev. Sci. Instrum., Vol. 65, No. 2, February 1994 Relative reflectometry 439

and parabolic mirror was much larger ( ws/A z 3) than the
optimum of wd/z- 1.82.6~
The intermediate frequency (IF) amplifiers covered a
frequency range of 1.4-2.4 GHz. The receiver accepted the
signal in this frequency range. both above and below the
laser frequency for a total double-sideband (DSB) bandwidth
of 2 GHz. A Hewlett Packard microwave Schottky
diode detected the IF signal following 60 dB gain.
A 30 cm focal length TPX lens L2 focused the receiver
field of view onto the test sample. The sample was oriented
so that the receiver field of view was reflected to 30 cm
focal length mirror M2. This mirror recollimated the receiver
field of view and directed it toward mirror M4 (an
off-axis parabolic mirror identical to M 1) which focused it
into the hot blackbody source. The hot blackbody was an
Infrared Industries model IR-563 source operated at a
1000 “C setting.
The chopper C with eccosorb7 covered blades modulated
the blackbody signal as seen by the receiver via the
sample reflection. The receiver would alternately view the
room-temperature chopper blades and hot blackbody. The
difference signal was synchronously detected with the
lock-in amplifier. Since this signal is seen via the sample
reflection it is proportional to the sample reflectivity assuming
everything else in the system including alignment
remains constant.
In the room-temperature setup a large wheel with four
equally spaced samples in its surface, was rotated by a
computer-controlled stepper motor to bring the samples
into position one at a time for a reflection measurement.
The samples were sized to fit into 32 mm diameter recesses
in the sample wheel behind a cover allowing a 22 mm clear
aperture for each sample. The second sample in one revolution
was designated as the reference reflector, the signal
from which was used to normalize the other signals. The
reference reflector for all the room-temperature measurements
presented here was an Edmunds Scientific goldcoated
glass mirror. The sample wheel paused for five
lock-in time constants of typically one second for each
reflection measurement. A data set consisted of measurements
made over many sample wheel revolutions.
In the cryogenic setup, a flip mirror FP was used to
alternately view a reflection from a reference and that from
the sample in the Dewar. Since the.room-temperature reflectivity
was already measured with the sample wheel
setup, the absolute calibration of the reference reflector and
window losses were not important. It was only necessary to
monitor the change in sample reflectivity relative to room
temperature. A polyethylene window mounted at a 20” angle
on the 25-mm-diam Dewar aperture also filtered out
near infrared and visible emission from the hot blackbody.
III. REFLECTOMETER
SYSTEM PERFORMANCE
The performance parameters of interest are the accuracy
of the reflection measurements, the spatial resolution,
and if the use of an incoherent, source improves reproducible
performance as compared to earlier use of a laser
beam.5 In the laser measurements it was found that the
instrument setup was extremely sensitive to sample align-
ment as well as to translation of less than a wavelength
along the beam path. In the present work there was a
significant improvement in the robustness of the setup to
sample positioning.
A. Accuracy of measurements
A figure of merit for the potential accuracy of the measurements
is the signal-to-noise ratio for measuring a thermal
noise level by a heterodyne receiver?
S/N= TBByT,,, $cGTl,
where TBB is the blackbody noise temperature, Tsys is the
receiver system noise temperature at the blackbody source,
Af is the bandwidth of the measurement, r is the signal
integration time for one measurement! and n is the number
of measurements. For the measurements presented here,
TBIA Tsrs 3 the signal-to-noise ratio is proportional to the
blackbody temperature as well as to the square root of the
bandwidth, integration time, and number of measurements.
The system noise temperature was measured by placing
a liquid-nitrogen blackbody source (eecosorb) behind
the chopper. The noise temperature in degrees K was then
determined by using the following equation:
216vdvm
Tws=q 9 r7 -293,
L*L ” lock-in
where v&, is the voltage on the digital voltmeter connected
to the receiver output, F/lo&-in is the lock-in amplifier
signal, and the factor 216 is the difference in degrees K
between room and liquid-nitrogen temperatures. The fattor
of 2.2 in the denominator converts the lock-in amplifier
reading which is a rectified mean voltage to a peak-to-peak
voltage. This factor was experimentally determined for a
square wave input to our EG&G model 5209 lock-in amplifier.
In making system noise temperature measurements it
was discovered that there was a significant air-path loss in
the range of 3.0-5.0 dB per meter. Unfortunately, the 214.5
pm laser line is near a water vapor absorption line. The
system noise temperature therefore depended on how far
from the mixer it was measured. The best noise temperature
achieved was 30 000 K at the location of the test
sample 53 cm from the mixer, but this value deteriorated to
about 77 000 K at the location of the blackbody. Using Eq.
( 1) this corresponds to a signal-to-noise ratio of -550 for
a single one second measurement.
Having the receiver mixer saturated by the laser local
oscillator is an advantage for measurement accuracy because
it makes the system less sensitive to laser power drift
and fluctuations. Saturated operation was experimentally
verified as shown in Fig. 3 where system noise temperature
versus local oscillator (LO) power is plotted. In the
present setup the receiver was operated at an LO power
level of 0.3 on the scale in Fig. 3, which corresponded to
-3 mW at the mixer. Reduced sensitivity to laser power
drift is illustrated in Fig. 4. The bolometer detector mon-
440 Rev. Sci. Instrum., Vol. 65, No. 2, February 1994 Relative reflectometry

sdev a523e-003
var 7264~~005
.----L---. 1 I I ,,--I
2 4 6 -8 10
LO Power (relatrve)
lime (mnutes)
FIG. 3. Receiver system noise temperature as a function of laser local
oscillator power. Maximum relative power value 1.0 corresponds to - 10
mW at the mixer.
itoring the laser power shows it drifting upward during the
33 min measurement time. The receiver signal, however,
viewing the constant hot blackbody source during the same
time is drifting upward by a far smaller amount. Rationing
the signals with a laser power monitor, as previously done
with direct laser probing,5 was not required with this reflectometer
configuration.
To determine the potential accuracy of this system for
reflection measurements the sample wheel was held fixed
FIG. 5. Raw data for measurements from a stationary gold mirror.
These data are shown in Fig. 5. One hundred points
make up each dataset. The mean of all three hundred data
points is 0.9997, which is in error by 3~ 10m4 from the
correct value of one. This error is consistent with the standard
deviation of the total data set divided by the square
root of the number of points as would be expected for
random error.’ This error is also of the same order as the
inverse of the signal-to-noise ratio as given by Eq, ( 1) .
When the sample wheel is rotated and when the samples
are changed, additional sources of error are introduced
on a gold mirror and the data acquisition system took data due to realignment error. These sources of error could
as though the sample wheel was cycling through four sam- eventually be engineered out of the instrument configuraples.
The data system recorded the lock-in voltage in the tion, however, they were an important factor in the present
order sample A-reference standard-sample B-sample C setup as illustrated in Fig. 6. The sample wheel had idenwith
delay for sample rotation and five time constants be- tical gold-coated mirrors in positions A, C, and for the
tween readings. The sample A-B-C readings were divided reference standard. A 303 stainless-steel sample with a
by the reference reading. The resulting ratios should all be smooth machined surface was placed in position B. The
equal to one since they are ratios of readings made on the gold reflectivities deviated from one by up to 0.3%. The
same fixed mirror. The deviation from one is the error in reflectivity of the stainless steel was - 1.1% less than the
the system due to random fluctuations.
gold mirrors. These results were reproducible for several
..-
F
t.4,
c*,
5
e
.d
d
Time (mmutes)
FIG. 4. Raw data and best linear fit to the bolometer laser power monitor
and the heterodyne receiver signal from the 1000 “C blackbody.
FIG, 6. Raw data and mean for rotating sample wheel measurements.
Rev. Sci. Instrum., Vol. 65, No. 2, February 1994 Relative reflectometry 441

7
2 3
*
4
2.2
N
3 2
'6 -
22
E % 1
m
0
0 50 100 150 200
Distance From Laser (cm)
5 10
Aperture Diameter (mm)
FIG. 7. Calculated and measured I/h 214.5 pm laser beam radius propagating
from the laser cavity output coupling hole.
test rufls moving the various samples around the different
samfile wheel positions.
Using the dc conductivities for gold and 303 stainless
steel, the reflectivity of the stainless steel should be 2.0%
less than for gold at 214.5 pm. However, it has been shown
that, the dc conductivity is not always accurate for determining
the high-frequency performance of metal
conductors.” Also our gold mirror which is a visibly thin
coating of gold on glass may not be of high enough quality
to give a true gold reference. It is important to note here
that the reflectometer only provides a precision rdative
reflection measurement that is only as good as the &owledge
of the reference reflector.
B. Spatial resolution
It was initially thought that the spatial resolution
would correspond to the focusability of the laser beam that
illuminates the Schottky diode mixer. It turned out that
the actual focusability of the receiver field-of-view in the
present setup was approximately four times larger than
this. This may be improved upon in the future because in
the present experiment the coupling between the laser
beam and Schottky diode mixer was not optimized, as
pointed out above.
The laser beam did propagate as a diffraction-limited
Gaussian beam. This is shown in Fig. 7 where a number of
bolometer profile scanned beam diameters are compared
with calculations for propagation of a Gaussian beam.”
One of the beam profiles was shown in Fig. 2.
It was not possible to scan the receiver field-of-view
profile because a point source with sufficient signal strength
was not available. Instead the field-of-view profile was de
termined by making the noise temperature measurement
through a series of different round room-temperature apertures.
Figure 8 shows the data for such a measurement
made at the location of the test samples. As can be seen in
these data, a Gaussian profile was a good fit to the fall off
in signal as the apertures were decreased. Since the integral
of a Gaussian is also a Gaussian, the receiver field-of-view
442 Rev. Sci. Instrum., Vol. 65, No. 2, February 1994
FIG. 8. Receiver field-of-view beam cross section measurements at the
location of the test samples.
envelope must also be Gaussian. However, this profile measurement
technique will not readily reveal higher-order
structure in the profile due to the possible presence of
higher-order, spatial modes.
The measured l/e2 field-of-view diameter at the test:
samples was - 10 mm, The smallest receiver field-of-viem
spot of 5 mm was measured at the location of the hot
blackbody at approximately the focal point of mirror M4..
Figure 9 compares the propagation of the measured receiver
field-of-view l/e* radius with a diffraction limitedi
calculation assuming the laser beam is reflected from the
Schottky diode mixer. The measured minimum spot sizes
are about a factor of 4 larger than this calculation. Future
improvements in the coupling between the Schottky diode
and laser local oscillator may improve the focusability of
the receiver field of view.
IV. SAMPLE MEASUREMENTS
Measurements were made on a total of seven thin-film.
superconductors. The reflectivity of all of these samples
2.0
----.--- GaussIan Beam Calculation
_.----. ;------.--.-..----..-.,
:
,*“
15 - ':
*. ,,*'
,.a'
10 - '.
:
:
,/'
,*a'
,,*'
;
;
;-
. ;-
: :
:
:
'-
;-
,:'
\j _
5- : . ,'
T' 1
I. ,,'
*i _
II ,/
'\ ;y -
.,'
2 -
0 L_L u .___ A .-IS I I 0 I t 1 -L-L-Lo
30 60 YO 120
Distance From Beam Splitter (cm)
FIG. 9. Calculated and measured propagation of the receiver I/2 radius
field of view from the beam splitter to the blackbody source.
Relative reflectometry
i

TABLE I. Summary on room-temperature measurements reflectivity relative
to gold mirror.
Sample No.
YBC-34
YBC-35
Type
Y-Ba-Cu-0 on LaAlOs
Y-Ba-Cu-0 on LaAIOa
Reflectivity
0.9626 *to.0066
0.9661 &O.C055
23-4 Tl-Ba-Ca-Cu-0 on LaAIO,
26-4 Tl-Ba-Ca-Cu-0 on LaAlO,
0.8332*0.0126
0.8596~0.0103
LDO07
LDOO9
Y-Ba-Cu-0 on MgO
Y-Ba-Cu-0 on LaA103
0.8338*0.01565
0.8380+0.01725
Sample No. 2 Y-Ba-Cu-0 on MgO 0.4749*0.02156
I L I I
200 300
Temperature (K)
was measured at room temperature and four were cooled
to cryogenic temperatures. The cryogenic measurements
were limited to a minimum temperature of 91 K by the
available Speac liquid-nitrogen Dewar.
Table I summarizes the room-temperature measurements.
The error shown corresponds to the standard deviation
of each dataset of typically 25 or more measurements.
There were deli&e differences among the different
types of superconductor films and between the same film
type but different fabrication run. The YBCO films
YBG34 and YBC-35 on LaA103 substrates had a high
room-temperature reflectivity of about 96%. The YBCO
films LDOO7 and LDOO9, one on MgO and the other on a
LaAlOs substrate, had a much lower room-temperature
reflectivity of about 83%. The thallium films 23-4 and 26-4
also had a low room-temperature reflectivity of about 83%
and 86%. A very thin lilm on a MgO substrate (sample 2)
had the lowest reflectivity of 47%.
One of the samples (LDO09) arrived late in the experimental
period after the room-temperature sample wheel
setup was removed. Its room-temperature reflectivity was
determined by placing it in the Dewar, then replacing it
with a gold mirror, and finally replacing that with sample
LDO07 which was previously measured in the sample
wheel. The ability to do this reproducibly in the Dewar was
difficult with previous direct laser measurements.’
The cryogenic measurements were made on four of the
samples, two YBCO films, YBC-34 and LDO07, and the
two thallium films. The two YBCO films did not go
through a superconducting transition because of the limits
of the Dewar. The reflectivity of these samples was observed
to only gradually increase a small amount as the
temperature decreased from room to 91 K. The two thallium
films were observed to go through a superconducting
transition at about 110 K. Figure 10 shows the thallim data
with 11 point smoothing. The increase in reflectivity of
sample 23-4 is larger, going from 83% to 89.6%. However,
the higher retlectivity at 91 K is achieved by sample 26-4 of
91.1% which implies that this sample had a higher conductivity.
Since illumination of superconductors with highfrequency
radiation above the band gap could reduce con-
Rev. Sci. Instrum., Vol. 65, No. 2, February 1994
FIG. 10. Reflectivity measurements of two thin-film thallium high T,
superconductors as a function of temperature. Eleven point smooth data
are shown.
ductivity, a Teflon filter was inserted into the beam path
between the hot blackbody source and the Dewar to further
filter the near-infrared energy. No change in the reflectivity
of sample 26-4 at 91 K was observed when this
was done.
V. DISCUSSION
The present measurements have demonstrated that
precision terahertz reflectometry using a blackbody source
and a laser heterodyne receiver is possible for making surface
resistivity measurements of good conductors. A number
of improvements to the present instrumentation can
significantly increase the precision of these measurements.
These improvements would address the present limiting
systematic and random sources of error. The systematic
errors in the room-temperature setup can be reduced by a
better engineered sample wheel for more reproducible sample
alignment and to eliminate wobble on rotation, With
this improvement measurement precision would be limited
by random error which can be reduced by increasing the
blackbody temperature and the receiver bandwidth. A
mercury arc lamp could have an effective blackbody temperature
as high as 4700 K (Ref. 12) and the receiver
bandwidth could readily be increased to 10 GHz. With
these improvements the signal-to-noise ratio could be increased
by over an order of magnitude over the present
results.
ACKNOWLEDGMENT
This work was supported by SIDO Innovative Science
and Technology Office, Contract No. DASG60-92-C-0090.
‘R. W. Ralston, M. A. Kastner, W. J. Gallagher, and B. Batlogg, IEEE
Soectrum Auaust, 50 ( 1992).
‘Special 1ssue;n Microwave Applications of Superconductivity, IEEE
Trans. Microwave Theory and Technioues 39. Sentember (1991).
L ,I .
“S. Ramo, J. R. Whinnery, and T. Van Duzer, Fields and Waves in
Communication Electronics, 2nd ed. (Wiley, New York, 1984), p. 291.
‘D. R. Cohn, S. C. Han, P. P. Woskov, B. L. Zhou, A. Ferdinand, R. H.
Relative reflectometry 443

Giles, J. Waldman, D. W. Cooke, and R. E. Muenchausen, J. Superconduct.
5, 391 (1992)
‘P. Woskov, D. R. Cohn, S. C. Han, R. H. Giles, and J. Waldman,
Proceedings of the 15th International Conference on Infrared and Millimeter
Waves, Lake Buena Vita, Florida, Dec. 10-14, 1990.
‘E. N. Grossman, Infrared Phys. 29, 875 (1989).
‘Trade mark of Emerson and Cuming for microwave absorbing material.
‘H. Z. Cummings and H. L. Swinney, Progress Opt. 8, 135 (1970).
‘P. R. Bevington, Data Reduction and Error Analysis for the Physical
Sciences (McGraw-Hill, New York, 1969).
“M. A. Ordal et aL, Appl. Opt. 24, 4493 (1985).
“J. D. Gaskill, Linear Systems, Fourier Transforms, and Optics (Wiley,
New York, 1978), Section 10-7.
*‘E. R. Mueller, University of Massachusetts, Lowell (private communication).
444 Rev. Sci. Instrum., Vol. 65, No. 2, February 1994 Relative reflectometry