Tb
AC
1
Twcdtpbew
�OaJtaSTbRMK8
2
erahertz imaging system based on aackward-wave oscillator
drian Dobroiu, Masatsugu Yamashita, Yuichi N. Ohshima, Yasuyuki Morita,hiko Otani, and Kodo Kawase
We present an imaging system designed for use in the terahertz range. As the radiation source abackward-wave oscillator was chosen for its special features such as high output power, good wave-frontquality, good stability, and wavelength tunability from 520 to 710 GHz. Detection is achieved with apyroelectric sensor operated at room temperature. The alignment procedure for the optical elements isdescribed, and several methods to reduce the etalon effect that are inherent in monochromatic sources arediscussed. The terahertz spot size in the sample plane is 550 �m �nearly the diffraction limit�, and thesignal-to-noise ratio is 10,000:1; other characteristics were also measured and are presented in detail. Anumber of preliminary applications are also shown that cover various areas: nondestructive real-timetesting for plastic tubes and packaging seals; biological terahertz imaging of fresh, frozen, or freeze-driedsamples; paraffin-embedded specimens of cancer tissue; and measurement of the absorption coefficient ofwater by use of a wedge-shaped cell. © 2004 Optical Society of America
OCIS codes: 260.3090, 230.7020, 170.3880, 170.0110, 120.4290, 150.3040.
gvwpmitrxwtorWiTtuaaqcshwee
. Introduction
erahertz �THz� waves, at the gap between micro-aves and the far infrared, have long been terra in-
ognita, mainly because of the lack of sources andetectors to facilitate efficient study of their proper-ies, let alone to apply these waves to solve real-worldroblems. Recently, as sources and detectors haveecome not only available but also affordable andasier to use, more groups have been joining the THzave research community.1
When this research was performed, A. [email protected]�, M. Yamashita, Y. N. Oshima, Y. Morita, C.tani, and K. Kawase were with RIKEN, The Institute of Physicalnd Chemical Research, 2-1 Hirosawa, Wako�, Saitama 351-0198,apan. Y. N. Oshima was also with the Incubation Center, Pen-ax Corporation, Wako�, Saitama 351-0101, Japan. K. Kawase islso with the Laboratory of Terahertz Bioengineering, Graduatechool of Agricultural Science, Tohoku University, 1-1sutsumidori-Amamiyamachi, Sendai 981-8555, Japan. A. Do-roiu is on leave from the National Institute for Laser, Plasma, andadiation Physics, P.O. Box MG-36, Bucharest, Romania. Y.orita is now with the Research Institute for Applied Mechanics,yushu University, 6-1 Kasuga-koen, Kasuga-shi, Fukuoka 816-580, Japan.Received 7 April 2004; revised manuscript received 12 August
004; accepted 13 August 2004.0003-6935�04�305637-10$15.00�0
© 2004 Optical Society of AmericaTHz imaging applications have received constantlyrowing interest. Among the most prominent ad-antages that it offers we mention the ability of THzaves to penetrate a wide range of materials—lastics, wood, paper, fabric, semiconductors, andany others that are opaque to visible and near-
nfrared light or produce only low-contrast images inhe x-ray region. As the THz photon energy isoughly 6 orders of magnitude smaller than that of an-ray photon, its interaction with matter, particularlyith biological tissues, is considered to cause no de-
ectable damage. A comparison with the other sidef the electromagnetic spectrum, the microwaveange, highlights the advantage of THz waves:ith their shorter wavelength they provide an imag-
ng resolution that is sufficient in many applications.he existence of substance-specific absorption spec-ra in the THz range, especially as a result of molec-lar transitions, facilitates fingerprinting and bringsbout a whole area of spectroscopic detection, testing,nd analysis techniques. THz spectroscopy canualitatively and quantitatively characterize thehemical composition of a sample2 with applicationsuch as the noninvasive detection of illegal drugsidden in envelopes.3 Absorption of THz waves byater, although it usually is a limiting factor, can bexploited in other applications. We also have highxpectations regarding THz imaging in medicine.4
The first report of THz imaging, by use of what now20 October 2004 � Vol. 43, No. 30 � APPLIED OPTICS 5637
waetcpwaibe
uaiocpvsTes
rmoth
hstsecfea
anptitrcAtw
siaEd
2
Fsc
mt5aoFt
bltf�mwmcmtl
oFac
Fpm
Fs
5
ould be considered a rudimentary but at that timen ingenious apparatus, seems to be that of Hartwickt al. in the 1970s,5 although many researchers citehe more modern approach by time-domain spectros-opy of Hu and Nuss from the 1990s.6 Other ap-roaches such as photomixing of two laseravelengths to produce a continuous THz wave havelso been successfully applied to imaging.7 The THzmaging techniques evolve and expand rapidly, ofteny adopting methods used in other regions of thelectromagnetic spectrum.8Our research is different from others’ in that we
sed a backward-wave oscillator �BWO� as the radi-tion source, to our knowledge for the first time in anmaging configuration. We gave up the advantagef synchronous detection, which provides direct ac-ess to the signal evolution in time and thus to thehase, in favor of other qualities that the BWO pro-ides, such as low noise and good stability, permittinghort acquisition times at a high signal-to-noise ratio.he good waveform that it produces is valuable forxcellent focusing on the sample and thus for goodpatial resolution.Other continuous-wave generators for the THz
ange are available, such as those based on photo-ixing of two closely spaced laser wavelengths and
n nonlinear multiplication of microwave oscilla-ions. The BWO generator has the advantages ofigh output power and ease of operation.In a few words, the BWO operates as follows: A
eated cathode emits electrons that are focused by atrong magnetic field and drawn toward the anodehrough a comblike decelerating structure. As a re-ult, an electromagnetic wave is produced that trav-ls in the opposite direction �hence the name� andouples into a curved waveguide that takes it out intoree space. The output frequency depends on thelectron speed, which is determined by the voltagepplied between the electrodes.For detection we chose to use a pyroelectric sensor,
s it has sufficient sensitivity, is compact, and doesot need cooling as many THz detectors do. Theyroelectric sensor works on the principle that a crys-al with specific properties becomes electrically polar-zed when it is heated. The minute change inemperature owing to absorption of electromagneticadiation is thus transformed into an electricalharge that is subsequently collected and amplified.pyroelectric sensor is sensitive only to variations of
he radiative power, so modulation of the incomingave is generally necessary.In this paper we discuss details of our imaging
etup, including solutions of some technical and phys-cal problems that arise when such a system is builtnd used, as well as its basic imaging characteristics.xamples of various applications are included toemonstrate potential uses of the imaging system.
. Imaging Setup
igure 1 shows the optical setup of the imagingystem. The BWO used as the source generates a
ontinuous, linearly polarized, essentially monochro- i638 APPLIED OPTICS � Vol. 43, No. 30 � 20 October 2004
atic wave at a frequency that can be controlled inhe case of our source in the approximate range of20–710 GHz �17.3–23.6 cm�1� by the high voltagepplied to the electrodes of its internal tube. Theutput power depends on the frequency as shown inig. 2 and has a maximum estimated by the producero be �15 mW.
First the divergent beam from the BWO is reflectedy a glass plate coated with an indium tin oxide �ITO�ayer, which we use as a dichroic mirror9 to combinehe THz beam with a visible beam that is necessaryor the optical alignment. The conductive ITO layer250 nm thick, with a bulk resistivity of 1.5 � 10�6 �� ensures good reflectivity, �95%, for the THzaves and a transmittance in the visible range ofore than 80%. As the visible light source we have
hosen a miniature electric bulb, whose small fila-ent works much as a point source; this option
urned out to be more efficient than using a He–Neaser with additional diverging optics.
Next the THz beam is collimated and then focusednto the sample by two off-axis parabolic mirrors.or image acquisition the sample is xy scanned withlinear motor stage that moves it through the fo-
used beam. Additionally, the sample can be moved
ig. 1. Schematic of the imaging system. The numbers on thearabolic mirrors represent their effective reflected focal lengths inillimeters; all four mirrors are 76.2 mm in diameter.
ig. 2. Spectral characteristics of the BWO output obtained bycanning the voltage applied to its tube. The periodic variations
n the spectrum correspond to an etalon effect inside the source.
isa
cpatlBfpitm
ctTdagbobalai
3
OwtptcdTdi
rewdtsdsasb
pltrot
c
cqhtotj
ersi4tswwcmswfo
FTrtb
FwwTavsa
n the lateral and axial directions with a manual xyztage, so the area to be scanned is placed in the beamnd the focusing can be adjusted.The beam transmitted through the sample is again
ollimated and focused onto the detector with anotherair of off-axis parabolic mirrors. Detection ischieved with a pyroelectric sensor operated at roomemperature. For this purpose the beam is modu-ated with an optical chopper placed in front of theWO output window, and the signal from detector is
ed to a lock-in amplifier synchronized with the chop-er. Afterward the output from the lock-in amplifiers read, stored, and processed in a personal computerhat also controls the BWO high voltage and the xyotor stage.Imaging resolution is one of our most important
oncerns. For this reason special care has beenaken for correct alignment of the parabolic mirrors.he first two mirrors are especially critical, as theyetermine the spot size on the sample. The opticallignment and the precise THz wave tracking arereatly facilitated by the superposition on the THzeam of a visible light beam, which is achieved by usef the ITO-coated plate introduced above. The visi-le light is also convenient afterward, when samplesre placed on their holder; the bright dot that theight creates is helpful both for easily locating therea to be scanned and for adjusting the sample planen the focus.
. Reduction of the Etalon Effect
ne of the problems that occur when one is imagingith such a highly coherent source is the presence of
he etalon effect. Its origin is the back-and-forthropagation of the THz wave between two surfaceshat come into the beam’s path and form a resonanceavity. The effect manifests itself as bright andark interference fringes that appear in the image.he reflecting surfaces can be the sample faces, theetector, or the source, such that they cannot be eas-ly eliminated.
One method for reducing the etalon effect is toeduce the reflection on the surfaces involved, forxample, by tilting one of them, to direct the reflectedaves out of the system, as shown in Fig. 1 for theetector. Tilting the sample leads to a result likehe one shown in Fig. 3 in which a tissue sample wascanned with and without tilting. The etalon effectid not disappear completely, possibly because of re-idual reflection by scattering or because the tiltingngle was insufficient. At the same time, tilting theample introduces a longer optical path and produceslurring of the image, especially for thicker samples.Another way to reduce the etalon effect consists in
lacing a partially absorbing medium inside the eta-on that is causing problems.10 Doing so will reducehe total usable power to a fraction f but will alsoeduce the etalon effect to roughly f2, which meansverall suppression of the interference term relativeo the mean intensity to a fraction f.
A third method that we tried consists in demono-
hromatizing the BWO source. We could reduce the aoherence length by modulating the THz wave fre-uency, which was done directly by modulation of theigh voltage applied to the electrodes of the BWOube. Unlike the parameters in the previous meth-ds �tilting angle and attenuation ratio�, in this casehe modulation amplitude has to be carefully ad-usted to cancel the interference fringes.
Figure 4 shows a demonstration of reduction of thetalon effect by wavelength modulation. A polysty-ene plate was placed in the sample holder and tiltedlightly to produce horizontal fringes when it wasmaged, as shown in Fig. 4�a�. For the image in Fig.�b�, each of the columns of pixels was scanned whilehe BWO frequency was modulated with triangularignals of different amplitudes: the leftmost columnith amplitude zero and the columns to the rightith linearly growing amplitudes. The arrows indi-
ate where the fringe visibility falls to a local mini-um, determined by calculation of the relative
tandard deviation of the pixel values separatelyithin each column. The contrast of the etalon
ringes can thus be decreased considerably by choicef the correct modulation amplitude.
ig. 3. Reduction of the etalon effect by tilting the sample. �a�he first scan is made in the normal horizontal position, and theesultant image exhibits fringes as indicated by arrows. �b� Forhe second scan the sample was tilted at a 27° angle. The visi-ility of the interference fringes was clearly reduced.
ig. 4. Reduction of the etalon effect by modulation of the BWOave frequency. �a� The image of a polystyrene plate scannedithout any modulation exhibits intense interference fringes. �b�he same sample was scanned while frequency modulation waspplied with increasing amplitude from left to right. The fringeisibility fell to local minima, indicated by arrows, corresponding topectral widths of 0.14 and 0.65 GHz. The contrast wasmplified—to the same extent in both images—for clarity; the
ctual images are shown as insets.20 October 2004 � Vol. 43, No. 30 � APPLIED OPTICS 5639
4
A
TrtswtAsiwtpf
wpaTio
oGiW
tpbiw
Te
wfi3dsdmaao
lmiti
soe0peiibsvawc
B
Bts
FmfGt
FoAip
5
. Characteristics of the System
. Imaging Resolution
he THz spot size at the sample plane, which di-ectly determined the resolution of the imaging sys-em, was evaluated by the knife-edge method. Aharp metallic blade was placed in the focal plane,ith its edge successively along the x and y direc-
ions, and moved perpendicularly to its orientation.n example of the signals recorded in these mea-urements is shown in Fig. 5. Assuming a Gauss-an bell distribution of the power in the focal spot,e fitted the graphs with appropriate error func-
ions. The intensity arriving at the detector de-ends on the knife’s position as expressed by theollowing equation:
I� x� � I0
12 �1 � erf�2�ln 2
x � xc
d �� , (1)
here I0 is the full beam intensity, x is the knife’sosition, xc is the position of the distribution center,nd d is the distribution diameter at half-height.he so-called error function, erf, appears from the
ntegration of the Gaussian distribution and can benly calculated numerically or approximated.For the frequency of 593.5 GHz �at the peak power
f our BWO source�, the x and y diameters of theaussian distribution, measured at half of the max-
mum, were found to be 561 and 534 �m, respectively.e conclude that the THz spot size is �550 �m.For determining the diffraction-limited spot size,
he same knife-edge measurement method was ap-lied to the collimated beam between the first para-olic mirror and the second. The data show that thentensity distribution is approximately Gaussian,
ig. 5. Measurement of the focal spot size by the knife-edgeethod. The measured signal �gray dots� was fitted with an error
unction. When the knife is tangent to the circle at half theaussian bell height, the transmitted signal is 11.9% or 88.1% of
he full signal; these levels are shown by the two horizontal lines.
ith a beam diameter of 35.6 mm at half-maximum. e
640 APPLIED OPTICS � Vol. 43, No. 30 � 20 October 2004
he following relationship gives the theoretical diam-ter of the focal spot at half-maximum:
d ��fd0
2 ln 2
, (2)
here � is the wavelength, in our case 505 �m; f is theocal length of the focusing element, 76.2 mm; and d0s the diameter of the collimated beam at half-height,5.6 mm. The formula gives a theoretical,iffraction-limited spot diameter of 477 �m. The re-ult shows that our imaging system works close to theiffraction limit. The remaining difference from theeasured 550 �m can be explained by imperfect
lignment, by the finite size of the parabolic mirrors,nd by the non-Gaussian distribution of the BWOutput wave.Choosing a collimation mirror with a longer focal
ength, thus producing a wider collimated beam,ight improve the resolution. However, if the lim-
ted diameter starts clipping too much of the wave,he focal spot will have some diffraction rings aroundt, with detrimental effects on the image quality.
The actual shape of the focal spot cannot be readilyensed by use of the knife-edge method. One canbtain a better visualization by scanning a small ap-rture, such as a thin piece of aluminum foil with a.5-mm-diameter hole, in the same manner as a sam-le. Although the hole diameter is approximatelyqual to the spot distribution size, the images, givenn Fig. 6, are still able to show that the spot changesn size and shape slightly with frequency, probablyecause of the presence of electromagnetic modes in-ide the BWO tube. Processing of the images re-eals that the long diameters of the spots, measuredt half-maximum, are in the range 700–1000 �m,hich is of course larger than the spot size because of
onvolution with the hole transmittance function.
. Depth of Focus
oth the tolerance of the sample’s axial position andhe ability to obtain well-resolved images in thickamples depend on the depth of focus. The following
ig. 6. Images of a 0.5-mm hole in aluminum foil. The numbern each picture represents the BWO wave frequency in gigahertz.length of 1 mm is shown as a white bar above each spot. As the
mages were all scanned at the same speed, low output levelsroduced noisier signals.
xpression gives, in the Gaussian approximation, the
ae
woaqam
C
TsiidsTosbttsFtm
efstIt
rtrama
icimsmiiaFrp�twdw
tslds
nttw
D
Wwuatmsrs
F0inb
xial extent within which the beam diameter does notxceed its minimum value by more than 21�2 times:
z �dmin
2
�
ln 2, (3)
here dmin is the spot diameter at half-height. Forur imaging system the depth of focus as definedbove is 2.0 mm. Exceeding this limit leads to auick blurring of the image such that, for example, at10-mm axial displacement the spot size becomes �5m, hardly acceptable as imaging resolution.
. Noise and Imaging Speed
he imaging speed is determined by the time neces-ary for taking the measurement of each pixel, whichn turn depends on the noise level. For our system,n which a BWO source is paired with a pyroelectricetector, the nonattenuated beam produces a 20-mVignal, whereas the detector’s dark signal is �2 �V.his means a signal-to-noise ratio of 10,000:1 whilene can safely neglect the source instability, which ismaller than the detector noise, as will become clearelow. These measurements were made at a lock-inime constant of 10 ms; allowing a pixel measurementime three times longer than the lock-in time con-tant, one obtains an imaging speed of 33 pixels�s.or a scan of a practical size such as 100 � 100 pixelshe total image time amounts to a little more than 5in.The signal-to-noise ratio increases 101�2 times for
very decade of decrease in lock-in time constant, soaster scans are possible with increased noise. Afterome hardware limitations are overcome, we expecthe scanning speed to reach at least 1000 pixels�s.n the case of more-absorptive samples, the integra-ion time has to be increased correspondingly.
The measurement dynamic range, defined as theatio of the largest to the smallest transmissionhrough a sample, is related to the signal-to-noiseatio. If the minimum transmitted power produces
signal twice as large as the noise level, and theaximum transmission is unity, then our system hasdynamic range of 5000:1, that is, 37 dB.Another parameter that influences imaging quality
s the power stability of the radiation source. Ac-ording to our measurements the BWO instabilityntroduces a low-frequency noise in the image. We
easured the signal at full output power 10 times aecond over a 1638.4-s period �214 samples� approxi-ately 1 h after the source was turned on; the result,
n which the signal variation was largely magnified,s plotted in Fig. 7�a�. The maximum relative vari-tion in this time interval is less than �2%. Theourier analysis of this signal, shown in Fig. 7�b�,eveals that the BWO instability introduces an ap-roximate Brownian noise, with an amplitude of0.3 �V rms on a time scale of 1 s, and increases 10
imes for each decade of time scale. Similar resultsere obtained at other sampling frequencies andata lengths. This instability should be compared
ith the dc level of the BWO output, 20 mV, and withhe dc level of the detector noise, 2 �V. The signaltops having a Brownian behavior on time scales ofess than 1 s when it reaches the ac noise floor of theetector and of more than 1 h when it becomes moretable.The chopping frequency also determines the scan-
ing time. The upper limit comes from the speed ofhe pyroelectric sensor. We set the optical choppero 316 Hz for our scans, and all noise levels given hereere measured at this frequency.
. Comparison with Other THz Imaging Systems
e compared the performance of our BWO systemith that of two other THz imaging setups, one thatses a terahertz parametric oscillator �TPO� source11
nd another based on femtosecond time-domain spec-roscopy �TDS�. A sample containing closely spacedetallic wires as well as areas of uniform transmis-
ion was scanned with all three systems, and theesultant images were compared visually, as pre-ented in Fig. 8.
ig. 7. �a� The BWO output power varies in time by �2% over a.5-h interval. More information is obtained by Fourier process-ng of this variation, which shows typical behavior for Brownianoise �b�. The frequency dimension was expressed as time scaley simple inversion for ease of understanding.
The TPO imaging system to which we had access
20 October 2004 � Vol. 43, No. 30 � APPLIED OPTICS 5641
rhToldsatvadt0wd
wtdCTqdtscqtwtt
5
Ttifptb
ioa
ttet�ft
antl
A
RbvpoobTdso
Fttsratmm.
Fc�imaged through an 18-mm-thick block of Teflon.
Fmu
5
elies on transmission optics �plastic lenses for tera-ertz use� and a silicon bolometer as the detector.he image produced by this setup is inferior to thatbtained with the BWO in both resolution and noiseevel. The TPO output power stability is �20%, and,espite an averaging of 12 pulses�pixel, the imagetill exhibits more noise than with the BWO. Also,t a frequency of 1.5 THz, meaning a wavelength 2.5imes shorter than with the BWO, the resolution isisibly lower. This happens because the TPO is notpoint source but rather a line source, which pro-
uces an irregular wave that is difficult to concen-rate into a small spot. The measuring time was.24 s�pixel, compared with our 0.03 s�pixel. Theider tunability of the TPO, however, allows more-iverse spectroscopic imaging applications.The TDS system that we used in this comparisonas designed by our research group at RIKEN par-
icularly for THz spectroscopic imaging and was pro-uced in collaboration with Tochigi Nikonorporation. The performances of the BWO and theDS setups cannot be directly compared, as they areuite different in nature. From the large amount ofata obtained with the TDS system we extracted theransmission image at a single frequency. The mea-urement parameters selected for the performanceomparison were spectral range, 0.5–3 THz; fre-uency step, 0.010 THz; and pixel time, 2.3 s. Underhese conditions the image corresponding to 0.59 GHzas extracted and compared with that obtained with
he BWO. Again, especially with respect to resolu-ion, the BWO system produced visibly better results.
. Applications
o check the imaging and measurement capabili-ies of the apparatus we attempted several prelim-nary applications; we present the results in whatollows. In each case there is much room for im-rovement and expansion. The main purpose ofhese tests was to evaluate the quality of the BWO-ased imaging system.A classical application of the THz waves consists in
nspecting the contents of packages, envelopes, andther containers that are made from materials that
re transparent in the THz range. As a demonstra- O642 APPLIED OPTICS � Vol. 43, No. 30 � 20 October 2004
ion of the capability of a THz imaging system to seehrough visually opaque materials, we present twoxamples in Fig. 9: A cardboard box containing me-allic objects was scanned, and the resultant imageFig. 9�a� reveals its contents; the RIKEN logo, maderom aluminum foil �Fig. 9�b� , was imaged through ahick block of Teflon.
Figure 10 is the THz image of a so-called touch-nd-go payment card used in the Japanese railwayetwork. It shows in good detail the internal struc-ure of the card, including a six-loop antenna whoseines have a transversal period as small as 0.8 mm.
. Technical Applications
eal-time nondestructive testing applications haveeen found. The ability of THz waves to penetrateisually opaque media was exploited in the testing oflastic tubes as they come out of the production line,r during production. Defects such as inclusions ofther materials, cracks, holes, and deformations cane detected in real time. Figure 11�a� shows theHz image of a polyurethane tube with a defect pro-uced by insertion in the tube of a small piece of theame material. Although in visible light the insidef the tube is not accessible, the THz image looks as
ig. 10. THz image of a railway payment card. The large loopade from six thin wires is the antenna that allows the card to besed just by being placed near the reading–writing machine.
ig. 8. Comparison of three THz imaging systems. The abilityo resolve the metallic lines in this sample and the uniformity ofhe brighter areas indicate the advantages offered by the BWOystem in terms of resolution and noise level. The TPO image wasecorded at 1.5 THz; the TDS and BWO images were both recordedt 0.59 THz. The scanning steps were small enough not to limithe resolving power. The size of the imaged area is 10 mm � 10
ig. 9. �a� Metallic objects—a 5-yen coin, a screw, and a paperlip—inserted in a cardboard box are revealed by THz imaging.b� The shape of the RIKEN logo was cut from aluminum foil and
ther elements of the circuitry can be seen.
iisccwi1
tomctopfcFluppsstdsmtimawt
aa
B
TpmaWttwfttwtwwBaltd
wftc
Nsas
Fost
FmTt
Ftcttsrett
f the tube were transparent. The implementationn the production line does not require imaging; aimple longitudinal scan obtained as the tube is beingonveyed is sufficient for detection of inclusions in theore of the tube. For detection of defects in the tubealls a manifoldly reflected beam can sense the tube
n all directions simultaneously, as suggested in Fig.1�b�.Still in the technical field, another application
hat we carried out is the nondestructive detectionf defects in the seals of plastic packages for foods,edicines, etc. The seals of these packages can
ontain flaws such as areas that remain unat-ached, through which the product inside can leakut or matter from outside can penetrate into theackage. We could detect the presence of such de-ects by passing the seal linearly through the fo-used beam and measuring the transmitted signal.or high-speed scanning, in this application the
ock-in amplifier and the optical chopper are notsed. Instead, the signal from the detector is am-lified, bandpass filtered, and read into the com-uter. When a defect, either air or an aqueousolution, passes through the beam the transmissionuddenly changes and causes a change in the de-ected signal. Figure 12 shows an example of aefect fabricated as a water-filled channel across aeal. The detection limit, expressed as the mini-um channel diameter that produces a signal dis-
inct from noise, is a few tens of micrometers, whichs judged by quality standards to be sufficient. The
aximum scanning speed that we tried for thispplication was 800 mm�s. The signal amplitudeas found to depend much more on the defect’s size
han on the scanning speed; thus we estimate that
ig. 11. �a� Polyurethane tube with a defect. �b� Schematic of anptical adapter for probing the plastic tube from several directionsimultaneously. The THz beam enters and exits as indicated byhe arrows and is reflected by mirrors.
ig. 12. Defect in polyethylene packaging seals. The peak at theiddle is produced by a water-filled channel, 30 �m in diameter.he peak shape comes from the fact that only the ac component of
the signal, in a selected frequency band, is measured.
t higher speeds the detection limit will still becceptable.
. Absorption Coefficient of Water
here is a rich literature on the subject of the opticalroperties of water in various ranges of the electro-agnetic spectrum. Various techniques are avail-
ble, such as using a cell of adjustable thickness.12
e measure the absorption coefficient of water—andhe same can be done with other liquids—by usinghe wedge-cell technique.13 We produce a layer ofater whose thickness varies in our experiments
rom 0 to 564 �m by joining two glass slides such thathey touch each other on one edge and are spaced athe other and by filling the volume between the sidesith water; the water is held in place by surface
ension. The beam is focused normally on theedge, and the transmitted signal is recorded as theedge is scanned from the thin end to the thick end.ecause the glass thickness is constant, the only vari-ble absorption effect comes from water. If the eta-on effect inside the water is neglected, theransmitted intensity will have a simple exponentialependence on the position:
I� x� � I0 exp����x�, (4)
here x is the position of the focused beam startingrom where the water’s thickness is zero, I0 is theransmitted intensity at x � 0, � is the absorptionoefficient, and � is the wedge angle in radians.
Figure 13 shows an example of a typical scan.ote that a water layer’s thickness of 0.3–0.4 mm is
till not too large for the signal to be detected withoutny lengthening of the measurement time; for thiscan the lock-in time constant was 10 ms. We ob-
ig. 13. Water absorption measured by the wedge-shaped cellechnique. Measured data are shown as dots, and the continuousurve is the exponential decay fit. The dip at left is the shadow ofhe upper slide edge, corresponding to zero thickness. At thehinner part of the wedge the etalon effect inside the water can beeen as a slight waviness of the measured data. In this semiloga-ithmic representation the data should lie in a straight line. How-ver, a small additive offset makes the data curve upward at thehicker end; this offset was included as an additional parameter inhe fitting function.
ained for the absorption coefficient of water �temper-
20 October 2004 � Vol. 43, No. 30 � APPLIED OPTICS 5643
a9pfm
C
Abcmtwaws3rsstesdFtp
ht�mmoti
ozeWbi
sflsmba
flnskFpfaFrcicuapso
D
A
Fcpl
FadbIttsbutrium�.
5
ture, 23 °C; frequency, 593.5 GHz� a value of 181 �cm�1; the literature14 gives 174.8 cm�1. We sus-
ect that including the etalon effect in the fittingunction would permit enhancement of the measure-
ent accuracy.
. Biological Applications
nother area of applications consists in analyzingiological samples. Because of the large absorptionoefficient of water, one application is the measure-ent of the water content in plants,15 with applica-
ions in agriculture, for example. The pattern ofater vessels and water distribution in a leaf can benalyzed, as shown in Fig. 14, and the quantity ofater can be monitored continuously and nonde-
tructively. This scan was performed at a speed of0 ms�pixel, which for the 138 � 138 pixel size cor-esponds to a total acquisition time of �10 min. Thecanning step was taken as 200 �m, considerablymaller than the THz spot diameter, to ensure thathe resolution is limited only by the physical param-ters of the imaging system; however, at a 500-�mtep the image definition suffers only a slight degra-ation, while the acquisition time becomes 2 min.or the specific purpose of monitoring the water con-
ent of leaves, fewer pixels are sufficient, makingossible a much higher frame rate if needed.As a measurement detail, the leaf in Fig. 14 was
eld between two stretched sheets of a very thin plas-ic film, commercially known as Saran Wrappoly�vinylindene chloride�, also called cling film,anufactured by Asahi Kasei, 11 �m thick, usedainly for food packaging . The THz transmittance
f a single layer of this film is above 98%, includinghe reflection loss on both sides, which makes it andeal material for sample support.
For biological purposes there are ways to avoid thebstruction caused by water. Samples can be fro-en; the better transmission of ice permits a deeperxamination. Figure 15�a� shows such an example:e froze a slice of pig tongue and kept it in this state
y placing it on a thin stretched plastic membrane
ig. 14. The vessel structure in this freshly cut leaf could belearly imaged. The details allowed by the 550-�m spot size arereserved by a 200-�m scanning step. The transmission of theeaf in the spaces between vessels is �15%.
nside a vacuum chamber. By this technique the o
644 APPLIED OPTICS � Vol. 43, No. 30 � 20 October 2004
ample was maintained at temperatures below 0 °Cor several minutes without water from the air al-owed to condense on its surface, which allowed ahort scan to be performed. However, there is al-ost no detail in the transmittance image, probably
ecause most of the absorption still takes place in icend the contrast that is due to tissue structure is low.The freeze-drying technique is much better than
reezing. It completely eliminates water through aow-temperature, low-pressure process. This tech-ique makes possible the examination of much thickeramples and does not pose a melting problem; to ournowledge it was not applied in THz imaging before.igures 15�b�–15�d� were obtained from sampleslaced on a stretched plastic film and prepared by thereeze-drying technique. Clearly resolved featuresnd excellent contrast can be observed. The scan inig. 15�b� is the heart of a chicken, with the left andight ventricles and the intraventricular septumlearly identifiable. In Fig. 15�c� we present the THzmage of a pig tongue, showing the texture of the mus-ular tissue. The different tissue layers of the pigterine cervix in Fig. 15�d� are also apparent. Theverage transmission through the freeze-dried sam-les, even at 1.5–2-mm thickness, is 10–25%. Freshamples of similar thickness would be essentiallypaque.
. Medical Applications
pplications in medicine have been a continuous
ig. 15. THz images of biological samples. The problem of highbsorption in water was solved by �a� freezing or �b�–�d� freezerying the samples. For the frozen sample hardly any detail cane seen, although the ice does transmit much better than water.n comparison, the freeze-dried samples show clear structural de-ails and tissue texture. The images correspond to �a� a pigongue; �b� a chicken heart �1, right ventricle; 2, interventriculareptum; 3, left ventricle�; �c� a pig tongue similar to the first sampleut flipped over and rotated; and �d� the cervical canal of a pigterus �1, the myometrium; 2, the uterine cavity; 3, the endome-
bjective for the THz wave research community.
Vtdttasstbc
eobvpi
6
Wraipbpat
rettr
Fvtsw
iitslcasoi
latpBti
wTMi
R
1
1
FpabFba
arious setups and techniques have been at-empted, including continuous-wave sources16 andark-field imaging.17 Tissue samples prepared byhe paraffin-embedding method can be scanned ashey are, as the water has been removed and par-ffin is transparent in the THz range. Figure 16hows samples of cancer tissues that we scanned ateveral wavelengths in an attempt to identify spec-rally the healthy and the affected regions, as it haseen suggested18 that such identification may be-ome possible in the THz range.The information contained in the terahertz prop-
rties of the samples can be processed in a multitudef ways.19,20 At present, however, there appears toe agreement among most workers that the data pro-ided by THz measurements, although they areromising, are still insufficient for objective discrim-nation between benign and malign tissues.
. Conclusions
e have presented an imaging system for the THzange in which a backward-wave oscillator is useds the radiation source. The good wave-front qual-ty permits focusing the beam in a small area toroduce high resolution, close to the Gaussian-eam diffraction limit. The intense and stable out-ut of the BWO yields a high signal-to-noise ratio,llowing short acquisition times, even when a room-emperature-operated pyroelectric detector is used.
Although they are more difficult to align, we usedeflection optical elements instead of lenses. Theynsure correct focusing as well as easy and preciseracking with visible light, while they neither reducehe available power through absorption or spurious
ig. 16. Paraffin-embedded tissue samples. A liver-cancer sam-le was scanned at several wavelengths; we show here the resultst �a� 567 GHz and �b� 676 GHz. The cancerous areas, indicatedy arrows, have higher THz transmission and appear brighter.or the breast cancer sample in �c� the diseased area is indicatedy the dotted, curved line at the left. Its brightness and texturere different from those of the rest of the sample.
eflections nor produce an etalon effect of their own.
or the surfaces that are perpendicular to the beam,arious methods of reducing the etalon effect, such asilt of the reflecting surfaces, partial absorption in-ide the resonating cavity, and modulation of theavelength, have been presented.The application examples presented here give an
dea of the large extent of the possibilities that thismaging technique has: real-time nondestructiveesting, biological and medical imaging, and mea-urement of physical properties of substances. Theow noise and good resolution allowed us to obtainlear, sharp images and well-defined signals, of whichn accurate analysis is possible. We expect thatuch qualities will lead to an increase of the numberf THz imaging applications for research, medicine,ndustry, and other fields.
The tunability of the BWO source, although it isimited, can be used in spectroscopic applications,lone or in conjunction with the imaging feature ofhe system. We have just started exploring this op-ortunity; we expect that the narrow band of theWO spectrum will be useful in situations in which
he spectral resolution of a THz Fourier-transformnfrared system, for instance, is insufficient.
We thank the following individuals for providing usith the tissue samples used in the present research:akashi Sawai and Yasuhiro Miura of the Iwateedical University and Minro Watanabe and Jun-
chi Nishizawa of the Iwate Prefectural University.
eferences1. See, for example, Eleventh International Conference on Tera-
hertz Electronics, J. Nishizawa, K. Hiromasa, and H. Ito, eds.�IEEE, Piscataway, N.J., 2003�.
2. Y. Watanabe, K. Kawase, T. Ikari, H. Ito, Y. Ishikawa, and H.Minamide, “Component spatial pattern analysis of chemicalsusing terahertz spectroscopic imaging,” Appl. Phys. Lett. 83,800–802 �2003�.
3. K. Kawase, Y. Ogawa, and Y. Watanabe, “Non-destructive tera-hertz imaging of illicit drugs using spectral fingerprints,” Opt.Express 11, 2549–2554 �2003�, http:��www.opticsexpress.org.
4. X.-C. Zhang, “Terahertz wave imaging: horizons and hur-dles,” Phys. Med. Biol. 47, 3667–3677 �2002�.
5. T. S. Hartwick, D. T. Hodges, D. H. Barker, and F. B. Foote,“Far infrared imagery,” Appl. Opt. 15, 1919–1922 �1976�.
6. B. B. Hu and M. C. Nuss, “Imaging with terahertz waves,” Opt.Lett. 20, 1716–1718 �1995�.
7. K. J. Siebert, H. Quast, R. Leonhardt, T. Loffler, M. Thomson,T. Bauer, H. G. Roskos, and S. Czasch, “Continuous-wave all-optoelectronic terahertz imaging,” Appl. Phys. Lett. 80, 3003–3005 �2002�.
8. D. M. Mittelman, M. Gupta, R. Neelamani, R. G. Baraniuk,J. V. Rudd, and M. Koch, “Recent advances in terahertz imag-ing,” Appl. Phys. B 68, 1085–1094 �1999�.
9. T. Bauer, J. S. Kolb, T. Loffler, E. Mohler, H. G. Roskos, andU. C. Pernisz, “Indium-tin-oxide-coated glass as a dichroic mir-ror for far-infrared electromagnetic radiation,” J. Appl. Phys.92, 2210–2212 �2002�.
0. P. H. Siegel, Submillimeter Wave Advanced Technology, JetPropulsion Laboratory, California Institute of Technology,Mail Code 168-314, 4800 Oak Grove Drive, Pasadena, Calif.91109 �personal communication, December 2003�.
1. K. Kawase, J. Shikata, and H. Ito, “Terahertz wave parametric
source,” J. Phys. D 34, R1–R14 �2001�.20 October 2004 � Vol. 43, No. 30 � APPLIED OPTICS 5645
1
1
1
1
1
1
1
1
2
5
2. O. A. Simpson, B. L. Bean, and S. Perkowitz, “Far infraredoptical constants of liquid water measured with an opticallypumped laser,” J. Opt. Soc. Am. 69, 1723–1726 �1979�.
3. D. M. Wieliczka, S. Weng, M. R. Querry, “Wedge shaped cell forhighly absorbent liquids: infrared optical constants of wa-ter,” Appl. Opt. 28, 1714–1719 �1989�.
4. M. R. Querry, D. M. Wieliczka, and D. J. Segelstein, “Water�H2O�,” in Handbook of Optical Constants of Solids II, E. D.Palik, ed. �Academic, San Diego, Calif., 1991�, pp. 1059–1077.
5. S. Hadjiloucas, L. S. Karatzas, and J. W. Bowen, “Measure-ment of leaf water content using terahertz radiation,” IEEETrans. Microwave Theory Tech. 47, 142–149 �1999�.
6. K. J. Siebert, T. Loffler, H. Quast, M. Thomson, T. Bauer, R.
Leonhardt, S. Czasch, and H. G. Roskos, “All-optoelectronic646 APPLIED OPTICS � Vol. 43, No. 30 � 20 October 2004
continuous wave THz imaging for biomedical applications,”Phys. Med. Biol. 47, 3743–3748 �2002�.
7. T. Loffler, T. Bauer, K. J. Siebert, H. G. Roskos, A. Fitzgerald,and S. Czasch, “Terahertz dark-field imaging of biomedical tissue,”Opt. Express 9, 616–621 �2001�, http:��www.opticsexpress.org.
8. J. Nishizawa, “Exploring the terahertz range,” J. Acoust. Soc.Jpn. 57, 163–169 �2001�.
9. P. Knobloch, C. Schildknecht, T. Kleine-Ostmann, M. Koch, S.Hoffmann, M. Hofmann, E. Rehberg, M. Sperling, K. Donhui-jsen, G. Hein, and K. Pierz, “Medical THz imaging: an inves-tigation of histo-pathological samples,” Phys. Med. Biol. 47,3875–3884 �2002�.
0. T. Loffler, K. Siebert, S. Czasch, T. Bauer, and H. G. Roskos,“Visualization and classification in biomedical terahertz
pulsed imaging,” Phys. Med. Biol. 47, 3847–3852 �2002�.
