Characterizing large-area electro-optic crystals toward two-dimensional real-time terahertz imaging
Fanzhen Meng,*Mark D.Thomson,Volker Blank,Wolff von Spiegel,
Torsten Löffler,and Hartmut G.Roskos
Physikalisches Institut,Johann Wolfgang Goethe-Universität,Max-von-Laue-Strasse1,D-60438Frankfurt,Germany
*Corresponding author:meng@physik.uni‑frankfurt.de
Received5June2009;accepted13August2009;
武汉光福7号
posted2September2009(Doc.ID112375);published15September2009
We have characterized the homogeneity of large-area(>10mm×10mm)CdTeð110Þand ZnTeð110Þcrys-
tals using a raster electro-optic scanning method to assess their usability in two-dimensional electro-
东芝m30optic terahertz(THz)imaging with parallel read out.The spatial variation in the detected THz signal
(at0.2and0:645THz,respectively)is due to nonuniform residual birefringence and scattering.For CdTe,
this depends critically on the growth method,and has an important contribution from slip planes in the
crystals,as is evident in the scanned images.For the highest-quality CdTeð110Þcrystals investigated,the
rms signal variations are less than15%,comparable to those for ZnTeð110Þ.For electro-optic scanning,
we introduce a hybrid measurement system based on a fs Nd:glass laser and a continuous-wave elec-
tronic THz source.©2009Optical Society of America
OCIS codes:300.6495,160.2100,110.6795.
1.Introduction
Radiation in the terahertz frequency range,typically defined as the range0:3–3THz,can be used for non-destructive measurements and has the ability to penetrate textiles,paper,and plastic materials.Ap-plications of imaging with THz radiation,in fields such as security and defense[1,2],inspection of defects in plastic pipes,identification of diseased skin tissue[3,4],inspection of textile-concealed goods [5],and surface-defect characterization[6,7],have been demonstrated.Recently,key advances have been achieved with advanced THz sources,including continuous-wave(cw)electronic multiplier THz sources[8,9],quantum-cascade lasers[10,11],nar-rowband THz-OPO systems[12,13],high-power laser-pumped semiconductor THz surface emitters [14–16],and electro-optic emitters[17,18].The emer-gence of such high-power THz sources expedites the development of practical THz imaging systems. Various detection methods have been used in these systems:electro-optical(EO)sampling[8,9,17,18], time-domain gating with photoconductive antennas [16],bolometric detection[12],direct detection with narrowband Schottky diodes[19],and GaAs and Si field-effect transistors[20,21].Many systems show a significant dynamic range—however,the majority of them still use a raster scanning method for2D imaging.Systems capable of real time(parallel)ac-quisition of2D THz images are highly desirable.This can be readily achieved with multipixel EO sam-pling,where the spatial field modulation of the unfocused THz beam is transferred to the polari-zation of an unfocused optical beam and detected with the help of a VIS/N
IR camera[22–26].The per-formance of such systems depends critically on the selection and quality of the large-area EO crys-tal used.
Assuming the EO crystal has sufficient trans-parency and nonlinear susceptibility for the optical and THz wavelengths used,the key selection criter-ion is to achieve phase-matching for the nonlinear EO process.In the past,most of the research with broadband THz pulses employed800nm femtose-cond(fs)Ti:sapphire lasers.For EO detection,
0003-6935/09/275197-08$15.00/0
©2009Optical Society of America
20September2009/Vol.48,No.27/APPLIED OPTICS5197
ZnTeð110Þis the most commonly used EO crystal [25,27–30]due to the natural phase-matching for wavelengths around800nm and THz frequencies ≲3THz.Recently,however,a variety of THz studies
based on lasers working at wavelengths around1μm were reported.For example,powerful THz genera-tion using a diode-pumped Yb:KGW[KGd(WO3)] solid-state fs laser was reported[16],while other groups have employed high-power Yb-doped fiber amplifiers[15,28,31–34].Also,narrowband T
Hz OPO systems based on nanosecond Q-switched Nd: YAG lasers were reported by other groups[12,13]. Compared with fs Ti:sapphire lasers,these lasers can be cheaper,more compact,have a more straight-forward operation,and/or have significantly higher output power.In these reports,the detection of the THz radiation was achieved using GaPð110Þ[27,34],ZnTeð110Þ[27],CdTeð110Þ[28,31–33],or photoconductive detectors[16].For EO sampling, although GaPð110Þshows near-perfect phase match-ing[27,34,35],CdTeð110Þwas found to be the most
efficient EO crystal[28,31–33]owing to its high EO
coefficient(about four times that of GaPð110Þ[25,34]).
In this paper,we study the applicability of large-
area CdTe crystals for2D EO detection with optical
wavelengths∼1μm.These results are relevant for
格伦 莱斯our target of building a real-time THz camera em-
ploying a THz OPO consisting of a Q-switched 1064nm Nd:YVO4laser and a PPLN parametric con-verter[36],which generates narrowband THz radia-
tion.We first examine the phase matching and
absorption properties of CdTe versus THz frequency
and optical wavelength to better define the range of
applicability.We then evaluate the potential perfor-
mance of large-area CdTeð110Þcrystals for2D cross-
polarizer-type EO detection[22,24]by measuring the
homogeneity of the single-pixel EO signal(with a
focused THz beam)with2D raster scans over the
crystal surfaces.This single-pixel THz setup is based
on a modification of our existing hybrid THz system
(with a cw electronic source and EO detection with an
asynchronous fs-pulse train[8,9])to use the1057nm
fs pulses from a Nd:glass laser(as opposed to a fs Ti:
sapphire laser).We compare the spatial uniformity
of the EO signals for CdTe crystals with different
growth/preparation conditions and compare this
with a large-area ZnTeð110Þcrystal.The uniformity
of the EO signals is affected by spatially varying
residual birefringence,associated with crystal im-perfections such as slip planes[37].The relative rms variations show minimum values(<15%)for CdTe obtained by the vertical-Bridgman technique.
2.CdTe Phase-Matching and Absorption Properties To evaluate the phase-matching coherence length L c [35]in CdTe,we used the refractive index data pub-lished for the optical[38]and THz[39,40]
ranges.In Fig.1we show a contour plot of L c versus both optical wavelength(λ¼900–1150nm)and THz frequency(ν¼0:5–3THz).As can been seen,at 0:5THz one achieves perfect phase matching at an optical wavelength ofλ∼1040nm,with L c remaining above1mm over the rangeλ∼950nm to>1150nm. With increasing THz frequency,the phase-matching
wavelength region shifts toward the visible,and the range narrows,as is characteristic for most EO crys-
tals(as dn THz=dν>0and dn opt;g=dλ<0).Hence,it is evident that CdTe possesses good phase-matching properties for the wavelength of∼1:06μm and
THz frequencies≲1:5THz,allowing one to take advantage of the stronger nonlinearity compared to GaP.
We note that the predicted coherence length de-
pends quite sensitively on the precise THz refractive index values used.Upon comparison of the CdTe data in[39–41],one observes a variationδn∼0:02, which can readily lead to a factor of2change in L c for situations where L c∼1mm(due to the large gradients in the L c surface in Fig.1).Moreover, t
he different literature data for the THz absorption coefficient vary significantly(which also affects the useful EO crystal length),whereas we found no study on the THz absorption versus CdTe growth technique and doping.In the materials community,it is well established that CdTe prepared using common mod-ern growth ,vertical-Bridgman,Sec-tion3)is inherently p-doped,due to an excess of Cd, and can be compensation-doped,using Group III atoms such as in[42,43].This compensation doping then allows one to achieve the highest electrical re-sistivity,and hence should result in the lowest THz absorption.
To investigate the THz dispersion/absorption of the CdTe crystals under study here,we performed THz transmission measurements on two CdTeð110Þcrys-tals,both from the same supplier(Keystone Crystal Corporation),only with and without indium compen-sation doping.The measurements were carried out using a THz time-domain spectroscopy(TDS)setup based on a1kHz amplifier laser(Clark-MXR CPA-2101)with ZnTe EO crystals for emission and
0.51 1.52 2.53
900
950
1000
1050
1100
11500.1mm
0.3mm
1mm
3mm
10mm
Fig.1.(Color online)Contour plot of the calculated coherence length of CdTeð110Þversus THz frequency(0:5–3THz)and optical wavelength(900–1150nm)based on refractive index spectral data from the literature[38,39].The inner and outer pair of black curves show the3and1mm border lines,respectively.
5198APPLIED OPTICS/Vol.48,No.27/20September2009
detection [17,18].The refractive index and absorption spectra of the two CdTe samples extracted from the TDS data are shown in Fig.2(each with thickness L ≈1mm —the precise thicknesses determined with fine mechanical calipers to within ≲5μm),along with measurements extracted from [39,40](crystal growth not specified).As can be seen,the refractive index data for the In-doped sample here agrees very well with that reported previously ,while the data for the undoped sample is shifted to smaller values by about δn ∼0:013.While this relative shift is compar-able with the precision in the thickness measure-ment,the negative direction is consistent with uncompensated dopant absorption expected from Drude theory [44].
A much more pronounced effect is apparent in the THz absorption coefficient α[Fig.2,bottom].For the In-compensation-doped sample,we measure similar results to those reported previously [40],with a low residual absorption α<0:05mm −1below 1THz ,which increases at higher frequencies due to one-/two-phonon absorption.For the undoped sample,there is a significant vertical shift to higher absorp-tion,with a residual level of α∼0:5mm −1below 1THz.We note that in another report on the THz properties of high-resistivity CdTe [45]the resi-dual low-frequency absorption was actually found to be closer to our undoped sample ,α∼0:5mm −1.However,in that report,a thinner CdTe sa
mple was used for this range (320μm)such that corrections for the Fresnel reflections would af-fect the derived bulk properties more strongly ,and no details of the growth or compensation doping method were specified.For near-IR wavelengths >1μm,the absorption of CdTe is negligible [41],and hence the effective absorption length for EO detection is domi-nated by the THz (field),L a ¼ðα=2Þ−1
[30].Hence,below 1THz we have L a ∼4mm and ∼8mm for the undoped and compensation-doped crystals,respectively .As shown in Fig.1,for wave-lengths near 1040nm the coherence length exceeds 10mm,such that in this regime the EO detection is absorption limited and the difference in THz ab-sorption between undoped and compensation-doped crystals can be important if thick crystals are used.For our target application,with wavelengths around 1060nm and THz frequencies of 1–1:5THz,both L c and L a come close to 1mm,such that a 1mm crystal thickness is a nearly optimal choice.
3.Details on Commercial CdTe Growth and Preparation
CdTe is a II –VI group compound with a zinc blende crystal structure,which ideally has no birefringence.The Cd –Te chemical bond has a high ionicity ,which leads to a low thermal conductivit
y and a very small formation energy for dislocations,twins and stacking faults [46].Many semiconductor growth techniques have been tested in the pursuit of high-quality CdTe single crystals [46,47].The vertical-Bridgman (VB)method in a high-pressure furnace is the most com-mon method used by commercial CdTe single-crystal suppliers.Tellurium precipitates [48]and inclusions [49],twins [50],grain boundaries [51],and slip planes [37]often appear in commercial CdTe crystals [52].These defects degrade the uniformity of the CdTe crystal and lead to residual birefringence.In general,high-quality large-area CdTe ð110Þcrystals are relatively rare on the market due to such growth/preparation difficulties.Moreover,the most common applications for CdTe crystals are for x-and γ-ray detectors,and as substrates for IR detec-tors (HgCdTe),hence many developments to improve the crystal properties may not be in the proper direc-tion for THz EO applications.
In this paper,we examine four CdTe ð110Þcrystals to find those with the most favorable properties for our application.The CdTe ð110Þsingle crystals obtained from different suppliers (which we label as C −n ,n ¼1−4):Keystone Crystal Corporation (USA,growth:VB,15×15×1mm,either undoped or In-compensation doped,C-1and C-2,respec-tively);Cradley Crystals (Russia,growth:VB,10×10×2mm,C-3);and Moltech GmbH (Germany ,growth:high-pressure vertical zone melting (HP-VZM),10×10×1mm,C-4).Note that by simple in-spection by eye,all CdTe crystals appear to have a very high surface quality and mirrorlike reflection.
Two-Dimensional EO Raster Imaging
In 2D EO THz imaging with cross-polarizer-type de-tection,the uniformity of the crystal plays an impor-tant role,as has been studied previously for (large-area)ZnTe crystals [22,24].A typical configuration places the EO crystal between two nearly crossed polarizers;the addition of a quarter-wave plate after the EO crystal provides a polarization bias such that the transmitted optical intensity
modulation
Fig.2.(Color online)THz refractive index and absorption coeffi-cient of two CdTe ð110Þcrystals (undoped and In-compensation-doped),including a comparison with literature data [39,40].
20September 2009/Vol.48,No.27/APPLIED OPTICS
5199
is essentially proportional to the THz field (not inten-sity).The resultant differential signal is given by ΔI ≈I 0 ð1=2ÞΓ0þδ Γ,where Γ¼n 3rk 0LE THz is the EO retardance in the crystal,Γ0is the residual bire-fringence,and δis the tilt angle of the second polarizer [22].
Although effective methods have been demon-strated to correct the distortion induced by residual birefringence (and scattering)in the EO crystal [22,24],these methods can only be applied with some restrictions.For instance,the polarization bias δmust be chosen large enough such that no image re-gions occur where the residual birefringence and polarization bias cancel (i.e.,ð1=2ÞΓ0þδ¼0,which physically eliminates the linear THz signal compo-nent).For EO crystals with large residual birefrin-gence variations,this then requires one to use a large δ,as opposed to that for optimal relative mod-u
lation depth [53].Moreover,the corresponding vari-ation in optical intensity on the camera prevents one from using its full capacity for most of the pixels (as the brightest pixels must still remain below satura-tion),which reduces the average signal-to-noise ratio (SNR).In addition,the correction methods are based on having direct access to both the reference (no THz)and the signal image frames.For certain detection modalities where only a differential signal is ac-quired (such as the single-pixel hybrid detection used here)this is not readily achieved.Hence it is impor-tant that the residual birefringence variations are kept to a minimum,even when correction methods are used.
In the experiments here,we simulate the situation of a 2D EO detection scheme by using single-pixel crossed-polarizer EO detection (with a focused THz beam)while raster-scanning the EO crystal.The scheme of the experimental setup is shown in Fig.3.In contrast to a typical fs optoelectronic THz system (with emitter and EO detection driven synchronously by the same fs pulses),here we employ a hybrid sys-tem [7–9]with a cw electronic THz source and asyn-chronous EO detection using a fs optical pulse train.
Two different narrowband terahertz multiplier sources were used here (both from Radiometer Phys-i
cs GmbH),with operation frequencies (f THz )of 0.20and 0:645THz (and cw output powers of ∼2mW and 0:5mW,respectively).The fs laser is a 1057nm Nd:glass laser (Time Bandwidth Products GLX-200)with a repetition rate (f rp )of 100MHz and 150fs pulse duration.The concept of the EO detection with an asynchronous source and pulsed laser can be de-scribed as heterodyne detection with the fs pulse train acting as a local oscillator (LO)[8,9].The mod-ulation of the optical polarization that occurs due to EO mixing with the THz field generates a sideband comb with frequencies at f THz −nf rp (extending down to baseband frequencies).The lowest modulation sideband for the 0:645THz source occurs at an inter-mediate frequency (IF)f IF ¼f THz −6450f rp close to 10MHz.This IF has a bandwidth of about 250kHz and drifts by less than 1MHz over an hour,under the highly stable condition of the temperature-controlled and vibration-isolated research environ-ment.To generate a phase-locked reference of the IF for lock-in detection,a fraction of the THz radia-tion is sent to a reference EO detection arm (using balanced detection with a 1mm CdTe ð110Þcrystal,Fig.3).The remainder of the THz radiation is sent to the single-pixel crossed-polarizer signal detection arm,where the different large-area EO crystals un-der study are mounted on a computer-controlled x –y scanner.The lock-in amplifier then demodulates the EO signal at the IF with the aid of the reference EO signal.As the cw THz phase is transferred to the IF electronic signal,the detected lock-in phase repre-sents the relative phase of the THz wave between the signal and reference detectors.
Therefore,coher-ent THz field detection is realized without actively synchronizing the laser and the THz source.A laser power of about 25mW was used in the signal arm,as limited by the saturation level of the photodiode.The scanned images of the different CdTe ð110Þcrystals are shown in Figs.4(a)–4(f),including mea-surements with 0:2THz and 0:645THz.The gray scales indicate the signal strength,which is (ideally)linearly proportional to the THz field magnitude (not intensity).In each image,the signals are normalized to their maximum values.While the use of 0:645THz more closely approaches our target applications in the range of 1–1:5THz,the use of 0:2THz allows us to perform a comparison with ZnTe (where the co-herence length is still above 3mm for 1057nm).A 0:2THz image of a ZnTe ð110Þcrystal (2mm thick-ness,diameter 25mm)is shown in Fig.4(g).As the focused optical beam samples the axial THz field,the lateral resolution of these scan images is dictated by the optical focal spot size (well below 100μm).The slanting striplike structures shown in Figs.4(a)–4(e)(VB growth)are due to residual birefringence in the crystal induced by slip planes formed during growth.The structure of this residual birefringence is also evident in transmission images obtained with a conventional optical cross-polarizer setup.
An
Fig.3.(Color online)Schematic of the hybrid THz system,includ-ing 2D raster scanning of the large-ar
ea EO crystals in the cross-polarizer signal detection arm.5200
APPLIED OPTICS /Vol.48,No.27/20September 2009
example is shown in Fig.5(for the CdTe crystal C-3),measured by illuminating the whole crystal surface with an expanded laser beam from the Nd:glass laser.The peripheral optical diffraction patterns in Fig.5also demonstrate that one should not use the entire EO crystal surface for 2D THz imaging.
Returning to Figs.4(d)–4(f),a significantly larger variation in signal strength is apparent for the CdTe ð110Þcrystal C-4,implying additional sources of inhomogeneity ,probably owing to inner structural vacancies,thickness variations,and/or grain bound-aries associated with the different growth technique for this crystal (HP-VZM,Section 3).To provide a more quantitative analysis of the signal homogene-ity ,in Fig.4(h)we plot a set of histograms for the sig-nal level in the 0:2THz measurements [normalized to the mean values,analysis regions as indicated in Figs.4(d)–4(g)].As can be seen,the signal distri-butions possess a fairly well-behaved statistical char-acter (except for some outlier points for the CdTe crystal C-4).The relative rms variations σin the THz signal obtained for the CdTe ð110Þcrystals are 13%(C-1),23%(C-3),and 49%(C-4),with 12%for the ZnTe ð110Þcrystchart控件
al.The rms values for the 0:645THz measurements (histograms not included here)are essentially the same.
In comparing the homogeneity of the various sig-nals,we should consider the differences in crystal thickness (1mm for C-1,C-2,C-4;2mm for C-3,ZnTe).We have performed simple modeling of the expected crossed-polarizer EO signal [22]with a ran-dom variation σΓ0in the net residual birefringence Γ0.It is straightforward to show that the resultant relative signal variation σis proportional to
the
Fig.4.(Color online)2D raster-scan images of the THz signal for different EO crystals,as detailed in text:undoped and In-compensation-doped 1mm-thick CdTe ð110Þ(C-1and C-2,respectively ,Keystone Crystal Corporation);2mm-thick CdTe ð110Þ(C-3,Cradley Crystals);1mm-thick CdTe ð110Þ(C-4,Moltech GmbH);and a 2mm-thick ZnTe ð110Þcrystal.(a)–(c)Measurements with 0:645THz;(d)–(g)with 0:2THz.(h)Histogram of relative EO signal strengths for the images in (d)–(g)(data region used for analysis indicated by dashed
rectangles).
罗马女
教皇Fig.5.Residual optical birefringence of the CdTe ð110Þcrystal from Cradley Crystals [(c)and (e)in Fig.4]measured by a con-ventional crossed-polarizer setup using the expanded Nd:glass laser beam. 20September 2009/Vol.48,No.27/APPLIED OPTICS
5201
豫公网安备41160202000603
站长QQ:729038198
关于我们
投诉建议