on June 30, 2018 Terahertzreal-timeimaging ...royalsocietypublishing.org/content/roypta/372/2012/20130111.full.pdf · antenna-andcavity-coupled bolometers FrançoisSimoensandJérômeMeilhan

rsta.royalsocietypublishing.org

ResearchCite this article: Simoens F, Meilhan J. 2014Terahertz real-time imaging uncooled arraybased on antenna- and cavity-coupledbolometers. Phil. Trans. R. Soc. A 372: 20130111.http://dx.doi.org/10.1098/rsta.2013.0111

One contribution of 11 to a Discussion MeetingIssue ‘Beyond Moore’s law’.

Subject Areas:microsystems, nanotechnology,electromagnetism, solid state physics

Keywords:antenna-coupled bolometer terahertz array,terahertz real-time imaging, terahertzuncooled two-dimensional camera

Author for correspondence:François Simoense-mail: [email protected]

Terahertz real-time imaginguncooled array based onantenna- and cavity-coupledbolometersFrançois Simoens and Jérôme Meilhan

CEA-Leti MINATEC, 17 rue des Martyrs, Grenoble Cedex 9 38054,France

The development of terahertz (THz) applications isslowed down by the availability of affordable, easy-to-use and highly sensitive detectors. CEA-Leti tookup this challenge by tailoring the mature infrared(IR) bolometer technology for optimized THz sensing.The key feature of these detectors relies on theseparation between electromagnetic absorption andthe thermometer. For each pixel, specific structuresof antennas and a resonant quarter-wavelength cavitycouple efficiently the THz radiation on a broadbandrange, while a central silicon microbridge bolometerresistance is read out by a complementary metaloxide semiconductor circuit. 320 × 240 pixel arrayshave been designed and manufactured: a better than30 pW power direct detection threshold per pixelhas been demonstrated in the 2–4 THz range. Suchperformance is expected on the whole THz rangeby proper tailoring of the antennas while keepingthe technological stack largely unchanged. This papergives an overview of the developed bolometer-basedtechnology. First, it describes the technology andreports the latest performance characterizations. Thenimaging demonstrations are presented, such as real-time reflectance imaging of a large surface of hiddenobjects and THz time-domain spectroscopy beam two-dimensional profiling. Finally, perspectives of cameraintegration for scientific and industrial applicationsare discussed.

1. IntroductionTerahertz (THz) radiation, loosely defined to be between0.3 and 10 THz, combines many unique properties:non-ionizing penetration through non-metallic and non-polar materials, high sensitivity to water content, spatialresolution 10 times shorter than for millimetre waveradiation, and specific spectral fingerprints for many

2014 The Author(s) Published by the Royal Society. All rights reserved.

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

2

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

substances. These features open the way to promising applications in many interdisciplinaryfields [1–4], many of which require two-dimensional imaging.

The first demonstrations of THz two-dimensional imaging [5] were based on optoelectronicTHz time-domain spectroscopy (THz-TDS). Since then, various different imaging modalities havebeen developed and demonstrated. However, most of them use a single detector and thereforesamples need to be scanned in two dimensions relative to the fixed terahertz beam, or vice versa.Mechanical scans limit inexorably the data acquisition rate and point-to-point measurementsare preferably applied to stationary objects. Additionally, such systems tend to be bulky andcomplicated.

The spread of applications using THz imaging would be greatly increased by the availability ofaffordable, compact, easy-to-use and highly sensitive detectors. In particular, large-format (many-pixel) focal plane arrays (FPAs) integrated in compact and hand-held cameras would enable fasttwo-dimensional image acquisition; the camera would operate in video mode, recording two-dimensional data in real time without the need for raster scanning.

Such features are possessed by thermal infrared (IR) array sensors based on bolometers[6]. These monolithic FPAs are collectively processed above complementary metal oxidesemiconductor (CMOS) read-out integrated circuits (ROICs) with standard silicon technology.The optically sensitive surface may gather a large number of pixels [7]. These sensors operate atroom temperature. And finally the CMOS ROIC provides advanced functions of signal processingbefore delivering video output. Since the 1990s CEA-Leti [8,9] has been developing such detectorswith the choice of amorphous silicon (a-Si) as the thermometer material, whereas most of the otherplayers have chosen vanadium oxide.

In 2005, the Massachusetts Institute of Technology group [10] demonstrated real-time THzimaging using a commercial uncooled bolometer IR camera in association with quantumcascade lasers (QCLs). Several other institutes and companies [11,12] tested this imaging set-up configuration, including the authors’ group [13]. In a common way, these tests showedthat significant improvements in sensitivity could be advantageously achieved by designingbolometer FPAs specifically for THz frequencies.

CEA-Leti took up this challenge by tailoring its proprietary bolometer technology foroptimized THz sensing. Such a development was initiated a few years ago with two constraintsas guidelines. First, the pixel structure had to be as close as possible to the existing a-Si IRbolometer stack in order to benefit from the maturity of this technology that has demonstratedhigh performance and high fabrication yield [14]. The second constraint involved the design ofa technological flow chart fully compatible with standard CMOS microelectronic equipment.Indeed, similarly to other running developments of THz arrays based on CMOS field effecttransistors [15,16] or Schottky diodes [17], bolometers can reach very-large-scale integration asa result of silicon process technologies. The combination of these two requirements meets thecriteria of low cost and low SWAP (size, weight and power) that any commercial camera hasto target. This trend is followed by the IR thermal bolometer arrays: the prices have decreasedsteadily and now meet the demands of large-volume applications, such as automotive [18] orhome appliances.

This paper summarizes the results of the Leti bolometer-based technology development. First,it describes the pixel array structure and the chips prototyped with two different FPA formats.Then it reports the modelling and the characterizations of the main figures of merit. The followingsection shows imaging tests carried out with a prototype of a compact camera integrating theFPA. Then a specific section focuses on the demonstration of terahertz imaging capabilities ofa complete real-time reflection imaging system. Perspectives of camera integration for scientificand industrial applications conclude this paper.

2. Detector technologyWithout modifying the three-dimensional pixel architecture of a standard IR bolometer, THzabsorption can be optimized by the adjustment of the equivalent impedance of the standard

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

3

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

resonan

t

dielect

ric ca

vity

200 mm CMOS

ASIC w

afer

l/2

resistive loads microbridge bolometer

DC antenna

CC antenna

~l/4n

metallic reflector

TOV

(a) (b)

(c)

Figure 1. (a) Uncooled THz antenna- and cavity-coupledmicrobolometer pixel structure. (b) Prototyped 320 × 240 pixel arraychip. (c) Camera prototype integrating the THz bolometer array in its vacuum packaging. (Online version in colour.)

microbridge structure. It can be done in a quite straightforward way by tuning the sheetresistance of a metallic film deposited on the existing suspended membrane. This method hasbeen applied for the first time by NEC [19], and then followed by INO [20]. Such cameras are nowcommercialized but still the lack of a quarter-wavelength cavity necessarily hampers the opticalcoupling efficiency. In addition, when THz frequencies smaller than 3 THz are being targeted,the mismatch between membrane size and wavelength will quickly degrade the published noiseequivalent power (NEP) (100 pW in the 3–4 THz range).

The CEA-Leti THz pixel three-dimensional [21] stack differs substantially from standardIR bolometers. The key feature of this architecture relies on the separation between the twomain functions of such sensors, i.e. the electromagnetic absorption and the thermometer.Complementary to an optimized quarter-wavelength cavity, antennas are coupled to thebolometer to perform optical radiation absorption.

Crossed quasi-double-bowtie resistively loaded antennas are designed to collect the THz fluxof any polarization (figure 1a). The longer bowtie antenna located on the suspended membrane,named ‘DC’ for direct coupling, is directly excited by the incident THz wave whose polarizationis aligned along the axis of this antenna. In order to couple the other impinging polarization, alarge antenna, capacitive coupling (CC), is located below the microbridge. The surface currentsinduced in this antenna are coupled via a capacitive mechanism to metallic planes deposited onthe suspended membrane. This two-storied antenna architecture permits the tuning of the opticalcoupling of both polarizations independently and with no mutual interferences, as shown later.

The residual radiation that is not absorbed by the antennas is partially recovered owing toa specific resonant quarter-wavelength cavity; as a result, the optical absorption is typicallyenhanced from 60–70% to 80–90% at resonance.

Surface currents created on the metallic antenna planes are coupled onto the central siliconmicrobridge that integrates matched antenna resistive loads. The Joule power dissipated in theresistive loads induces a bolometer resistance change that is read out at the pixel level by afront-end CMOS circuit.

The process above the integrated circuit (IC) of the complete THz bolometer pixel stack is fullyCMOS foundry compatible. THz arrays are collectively manufactured above the IC wafer usingonly the classical steps of microelectronic photolithography, deposition process and etching. Themain technological key [22] is the etching and metallization of through-oxide-vias (TOVs) throughthe thick dielectric cavity. TOVs electrically connect the bolometer resistance with the CMOSupper metallic contacts. This process ensures the fabrication of a monolithic two-dimensional

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

4

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

sensor with a large number of pixels. Moreover, the resulting vertical architecture means thatpixels are separated from the CMOS ROIC by the metallic reflector plane; such a feature preventselectromagnetic perturbation of the optical coupling by ROIC metallic lines as encountered in[23].

The antenna sizes and shapes can be chosen independently from the bolometric device tomatch the illumination characteristics, in particular the frequency range and the polarization.This arrangement is very versatile with respect to frequency [24]. When designing the pixel,the antennas and the resonant cavity can be sized to match the wavelength, whereas thecentral bolometer structure can be kept small; as a result, the response time and the heat tocurrent conversion efficiency are preserved. Crossed polarized antenna structures have beenimplemented to make the bolometer sensitive to both transverse electric (TE) and transversemagnetic (TM) polarizations, but it is possible to integrate only one of these two polarizationsin the pixel structure.

3. Prototyped terahertz bolometer arraysTwo formats of THz FPAs have been designed, processed, tested and used for imagingdemonstrations. First, a large 320 × 240 pixel chip has been developed and integrated in acompact video camera housing [25]. A second chip integrates THz bolometers with visible andIR pixels to act as a tri-spectral monolithic detector [26].

The 320 × 240 array (figure 1b) is composed of 50 µm pitch pixels designed for an optimizedsensitivity in the 2–4 THz range where most of the explosive fingerprints are located and QCLsare efficient [27,28]. Extensive three-dimensional finite-element method (FEM) simulations [29]were carried out with the commercial High Frequency Structure Simulator (HFSS) and Comsolcodes to design two kinds of pixels. They are optimized at 1.7 and 2.4 THz for CC polarizationwhile maintaining maximum absorption at 1.35 THz for the DC direction.

A dedicated CMOS application-specific integrated circuit (ASIC) has been designed by CEA-Leti teams to ensure low-noise measurement of the resistances of the 320 × 240 bolometers and toformat the resulting signals into a single video data stream.

Thanks to state-of-the-art Si microelectronic facilities and robust bolometer technology know-how, a very high yield has been achieved: more than 99.5% of the 320 × 240 pixels are functionalfor 56 out of the 63 chips available per 200 mm wafer.

The monolithic tri-spectral array detector integrates three types of sensors, all of themoperating at room temperature. For the visible channel, photodiodes are implemented withinthe ASIC while the IR and THz detectors consist, respectively, of a-Si 25 µm pitch and 50 µmpitch microbolometers. The central region of the FPA is occupied by 32 × 32 THz pixels while thesurroundings combine IR and visible pixels.

One major issue [30] was to work out a technological stack compatible with the simultaneousprocess of IR and THz bolometric pixels above the CMOS wafer. Another important challengewas the design of an ASIC that efficiently reads and outputs the signal generated by thethree channels [31].

In spite of this very challenging tri-spectral integration gathering very diverse pixels, asatisfactory fabrication yield has been achieved: several chips have been integrated in vacuumpackaging with diamond windows and tested for real-time imaging, as presented in §4.

As common features, the THz bolometric pixels of these arrays present thermal time constantsbetween 20 and 40 ms and thermal resistances of 50 MK W−1.

4. Modelled and measured figures of meritExtensive work on modelling of the figures of merit has been carried out, of both the collectionand thermometer functions. Following this design phase, specific experimental developmentsand tests of the prototyped chips were undertaken in order to assess the numerical results.

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

5

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

1.0DC polarization CC polarization

DC polarization

measurementsimulation

measurementsimulation

measurementsimulation

measurementsimulation

CC polarization

0.8

0.6

0.4

0.2

0

1.0

0.8

0.6

0.4

0.2

00.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0

frequency (THz) frequency (THz)

refl

ectiv

ityab

sorp

tanc

e

(b)(a)

(c) (d )

Figure 2. Spectral reflectivity of (a) DC and (b) CC antennas and spectral absorptance of (c) DC and (d) CC antennas. For eachspectral feature, data extracted frombothmeasurements and three-dimensional FEMsimulations are compared. (Online versionin colour.)

(a) Spectral absorptionOne major figure of merit of any FPA is the spectral absorption. Few publications show spectralabsorption characterized continuously in a large THz range. However, this feature representsthe optical collection efficiency and is then directly related to the FPA sensitivity. Actualinstruments and procedures are lacking. Hence, special efforts have been dedicated to developingexperimental methods to extract spectral features of our two-dimensional array sensor with thehelp of Bruker Optics Vertex 80 V Fourier transform spectrometers (FTSs).

First, the reflectivity of the sensor surface was measured. A preliminary referenceinterferogram is acquired with a metallic mirror, and a second spectrum is recorded in whichthe FPA replaces the mirror. Assuming that the reference mirror is perfectly reflective, the Fouriertransform of the ratio of sample to reference interferograms represents the spectral reflectivity ofthe array.

This measurement was first applied to a 160 × 120 pixel prototype array processed abovea raw Si wafer instead of an ASIC. Without an ASIC to ensure the multiplexing, specificmetallic meander lines at the reflector backplane level address electrically 16 pixels. Hence, thisconfiguration makes possible the reading of the signal sensed by individual pixels selected withinthe array. The pixels of this array are designed for optimized coupling at 1.35 and 1.7 THz,respectively, for the ‘DC’ and ‘CC’ antennas.

Figure 2a,b shows the spectral reflectivity Ropt(f ) measured, respectively, for the DC and CC.Polarization is controlled with a wire grid polarizer inserted in the set-up. This experimentalresult is compared with the |S11|2 parameter (dotted line of figure 2) extracted from three-dimensional FEM simulations in which each pixel of the array is considered as a Floquet cell [29].A very good match between simulations and measurements is observed over the whole frequencyrange: a distinct absorption peak arises (at 1.35 and 1.7 THz) as a dip in reflectivity and the shapesfollow the same trends. However, this measurement is not sufficient to assess the useful pixelabsorption efficiency that is the ratio between the thermal heating of the antenna loads and theincident wave power. Moreover, this method overestimates the spectral absorption deduced fromαopt(f ) = 1 − Ropt(f ). Indeed, some power losses in the stack—such as Joule dissipation within thedielectric cavity—do not contribute to the bolometric signal.

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

6

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

1.0(a) (b)

350array sigalideal response fordecoupled cross antenna300

250

200

150

100

50

0.8

0.6

0.4

0.2

00.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 0 20 40 60 80 100

frequency (THz)

CC antenna

THz beam //CC antenna

THz beam //DC [email protected] THz

DC antenna

ª × 4

angle (°)

abso

rpta

nce

inte

grat

ed s

igna

l (V

)

Figure 3. (a) Spectral absorption of CC and DC antennas of the 2.4 THz design FPA (dotted lines, simulations; solid line,measured) showing the optical collection ratio close to 4 between the two antennas at 2.5 THz. (b) Measured output signalintegrated over the array versus the angle of the linearly polarized incident beam (solid line) and computed output with theassumption of uncoupled cross antennas (triangles). (Online version in colour.)

In order to ascertain the bolometric conversion efficiency, direct measurement of the spectralresponse of the sensor has to be performed. The collimated beam delivered by the interferometeris deflected outside of the FTS and is focused on the detector array enclosed inside a vacuumlaboratory test vessel.

Resulting interferograms include spectral absorption of the devices convoluted to the spectralfeatures of the source and of the optical chain. This measurement is referenced to an interferogramobtained in a similar operation scheme with the FTS internal pyroelectric detector to retrieve therelative spectral response of the bolometer. The relative spectral response is normalized to themaximum peak.

As shown by figure 2c,d, good agreement is noted between the empirical and theoreticalcurves. One can observe comparable performance and a large absorption peak in the vicinityof 1.35 and 1.7 THz, respectively, for DC and CC antennas. The observed weak oscillations andslight shift of the experimental resonance comes mainly from the atmospheric signatures that alterthe signal during its transmission in air between the FTS and the sensor.

The absorption spectra extracted from modelling and measurements (figure 3) are inagreement with the ones obtained in the reflectivity configuration. These characterizations giveus an upper bound for the useful absorption of our device and locations of resonance peaks.

The developed procedures make possible the evaluation of the ‘useful’ spectral absorption ofthe THz array sensor that is an essential figure of merit when one seeks to estimate and to comparedevices’ performance.

(b) Noise equivalent powerAnother key performance metric for the detector is the NEP. Mathematically, it is the ratiobetween the output noise voltage spectral density vn and the detector voltage responsivity Rv .This quantity can be reported at various circuit levels, from the direct signal generated at thepixel output, i.e. the current/voltage change in the resistance for a resistive bolometer, to theoutput of the overall FPA imaging chain operated in video mode. The latter NEP requires only ana priori knowledge of the total impinging source power. This method is applied to the prototyped320 × 240 pixel array with the 2.4 THz design, and is operated at a video rate of 25 Hz. A row-wise reading scheme of the ROIC allows the current of the bolometer to be integrated by the 5 pFcapacitance of its associated trans-impedance amplifier during 125 µs per pixel.

The sensitivity measurement set-up that combines a QCL [32] emitting at 2.5 THz and aThomas Keating Instruments power meter is detailed in a previous paper [33]. In a general way,no filter is used during these experiments.

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

7

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

First, the responsivity at 2.5 THz corresponding to the CC polarization is measured. As a totalpower of 238 µW illuminates the device, the sum of the pixel voltages over the array is 3000 V.The FPA responsivity is calculated as the ratio of the integrated signal to the total impingingpower and is found to be RCC

vtotal = 12.6 MV W−1. The same measurement performed for thecross-polarization state, i.e. the DC antenna, leads to a lower total responsivity RDC

vtotal =4.3 MV W−1.

Imaging RMS noise vn including pixel noise, ROIC and digitizing chain contributions ismeasured to be 400 µV RMS. The smallest variation of impinging power that can be detectedis calculated as the ratio of RMS noise to the device responsivity. This detection powerthreshold reaches 32 pW for the CC polarization at 2.5 THz and is better than 30 pW at theresonance frequency.

During the design of a detector sensitive to unpolarized radiation, one of the main challengesis to obtain cross-polarized antennas featuring both high absorption efficiency and no mutualinterferences. Cross antennas built at the same technological level suffer from mutual couplingthat averages and lowers the absorption efficiency independently from the polarization state.As described in §2, the two-storied crossed antenna architecture was chosen to remove thisunwanted mutual coupling. This property has been experimentally tested by rotating the FPAangle θ from 0◦ to 90◦ with respect to the linearly polarized incident QCL beam—controlledwith wire grid polarizers. For θ = 0◦, the TM polarization of the QCL corresponds to theDC antenna orientation. As illustrated in figure 3b, the resulting output integrated signal ofthe device follows a mathematical function VDC

out cos(θ )2 + VCCout sin(θ )2. Moreover, the ratio of

integrated output signals is in correspondence to the optical absorption efficiencies of DC andCC antennas (figure 3a). These results demonstrate the decoupling property of this crossedantenna three-dimensional structure. However, one has to point out that, if needed, it ispossible to design a square law detector in which the CC and DC antennas spectral absorptionsare similar.

(c) Imaging application demonstrationsReal-time active imaging capabilities of the prototyped two-dimensional FPAs have been testedin several set-ups, in both reflection and transmission configurations.

First, simple optical arrangements were tested in which samples intercept the collimated beamof a THz QCL. As shown in figure 4a,b, the 320 × 240 THz array is able to deliver real-time imagesof objects, respectively a scalpel blade and Leti characters written with metallic ink on paper,both concealed in a postal envelope [28]. Speckle patterns arising from the QCL source coherencedegrade the image quality.

THz antenna- and cavity-coupled bolometers also present ultra-wideband detectioncapabilities. Even if the prototyped 320 × 240 FPAs are optimized for centre resonance frequenciesin the 2 THz range, significant absorption remains below 1 THz. Therefore, the array can deliverreal-time frames of the THz beam produced by longer wavelength sources. Two essay campaignsdemonstrated sensitive two-dimensional real-time profiling of the beam delivered by classicalTDS set-ups [34]. During these experiments, the beam generated by the photoconductive emitterpropagates in the ambient atmosphere and is focused on the FPA by off-axis parabolic mirrors. Infigure 4c,d [35], the beam delivered by the TDS is imaged with a 29 dB signal to noise ratio (SNR)at peak; detected integrated THz power of the source emitting as low as 25 nW has been resolvedwith sufficient SNR to observe tiny details of the beam shape.

The characterization of the second type of prototyped array described in §3 (figure 5b)showed state-of-the-art performance for each channel [26]. The figure 5a presents the visibleand THz channel outputs when a paper optical test pattern is standing in the way ofa QCL beam optical path. The 32 × 32 THz pixels of the array centre profile the QCLbeam while visible photodiodes image the object at the periphery. In figure 5c, thermal IRimaging by the IR pixels of the two-dimensional sensor is demonstrated separately using f/1germanium optics.

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

8

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

0.20

(a) (b)

(c) (d)

0.15

0.10

0.05

0

02000

2000

4000

0

2000

4000

4000 distance (mm)

distance (mm)

distance (mm)distance (mm)

TH

z in

tens

ity (

arb.

uni

ts)

0.20

0.15

0.10

0.05

01000 2000 3000 4000

TH

z in

tens

ity (

arb.

uni

ts)

Figure 4. Examples of frames extracted from video sequences. (a) Images of a scalpel blade and of metallic ink Leti characterswritten on paper (b), both concealed in an envelope. Profiles of the beam emitted by a TDS emitter before (c) and after (d)optical alignment. (Online version in colour.)

20

20

40

40

60

60

80

80

100

100

120

120

140

140160

160 20 40 60 80 100 120 140

20

40

60

80

100

120

140

160

(a) (b) (c)

Figure 5. Frames extracted from real-time visible and THz (a) and thermal IR video output (c) of the MUTIVIS tri-spectralmonolithic array sensor (b). (Online version in colour.)

5. Imaging tests in a reflection systemImaging demonstrations shown in the previous section have motivated the development of acomplete reflection active THz imaging demonstrator combining QCLs with the 320 × 240 THzFPA (figure 6a).

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

9

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

horn

THzsource

Fresnel lens

1 m

scen

e

THz two-dimensional

bolometercamera

(a) (b) (c)

(d) (e)

55 mm

without with

Figure 6. Schematic of the complete reflection active THz imaging system (a) and examples of 4 × 6 cm2 images (b–e).(Online version in colour.)

The illumination beam is delivered by multiple QCL ridges in order to optimize the availableoptical power. The number of QCL ridges operating in continuous wave mode was chosen tofit the cooling power of the cryocooler. An innovative pulsed tube (PT) cryocooler developedby CEA-INAC has enabled such integration. This PT provides a 20 W cooling power on thecold finger, ending up with the choice of a 30 K temperature operation and four combinedQCL ridges.

The source is capable of delivering continuously tens of mW with dedicated fast drivingelectronics and switches to gate the laser. The optical beam is then guided by a specific optical set-up that points the beam towards the scene. A metallic horn ensures beam mixing to provide betterillumination uniformity. A specifically designed Fresnel lens is interposed in the beam opticalpath to magnify the output of the rectangular horn.

The imaging plane is located at 1 m from the demonstrator housing, whose inside atmosphereis dried to minimize water vapour absorption. The illuminated area size is 40 × 60 mm2.A large plane mirror is used to illuminate the scene and to collect the part of the beam that isreflected and backscattered by the imaged objects. Then a folded Newton telescope composedof an 80 cm paraboloid mirror focuses the beam onto the camera FPA with an equivalent f/0.8relative aperture. The optical path from the QCL-based source to the camera is 4 m long, of which2 m are in air.

The camera comprises a dedicated vacuum-sealed packaging topped by a Zeonex window.The operation of the array under the vacuum atmosphere—typically 10−3 mbar—is necessary forhigh sensitivity, as for any bolometer sensor. The camera also includes the front-end electronicsmade of two different cards. The first card delivers low-noise analogue voltages and digitizes thechip output voltage. The second card, a field-programmable gate array (FPGA) board, drives thesensor and ensures synchronization between the camera and external devices.

Evaluation of optical quality and resolution has been performed with the help of absorbingoptical test patterns placed in front of a metallic plane mirror. The image of the Siemens star isflawed by specular interferences induced by the coherence of the source (figure 6b). Then theslow time response of the bolometer is turned into an advantage to blur the speckle patterns: byoscillating axially the Fresnel lens, a large improvement of image quality is obtained as illustratedin figure 6c.

A second imaging test has been carried out with a Leti writing test pattern of various sizes(figure 6d). Letter details of dimensions close to 2 mm are being resolved and represent theoutmost resolution that can be achieved by this reflection imaging system demonstrator. In spite

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

10

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

of imperfectly homogeneous illumination, the SNR is sufficient to perform real-time imagingof hidden objects because this ‘Leti’ test pattern has been successfully imaged (figure 6d) oncehidden below a nylon clean room coat [33].

6. Conclusion and perspectivesAn innovative monolithic two-dimensional array sensor technology has been developed withthe aim of fulfilling the need for affordable, easy-to-use and highly sensitive detectors for real-time THz imaging. Derived from the mature uncooled thermal IR bolometers, an innovativearchitecture of monolithic array based on the use of antennas and of a resonant quarter-wavelength cavity has been designed.

Thanks to the maturity of bolometer technology and to full compatibility with silicon CMOSprocesses, technological barriers have been lifted. Two formats of THz arrays optimized forsensing in the 1–5 THz range have been processed above CMOS ASICs: a 320 × 240 THz sensorand a tri-spectral chip. A high manufacturing yield has been observed with, for example, 56 out of63 chips available in a 200 mm wafer. A NEP better than 30 pW per pixel has been characterizedat 2.5 THz for the 320 × 240 FPA integrated in a camera and operated in video mode, makingthis uncooled sensor the state of the art of such devices. Three-dimensional modelling studiesof this sensor have shown good agreement with the measured features, such as the broadbandspectral absorption and non-mutual coupling between the two crossed polarizations that thispixel structure can collect. Real-time imaging tests have highlighted the ease of use and the highsensitivity and resolution achieved by this broadband array detector.

Further technological developments will target performance improvement of the prototypedsensor fed by ongoing progress in IR bolometer technology. Moreover, the confirmed goodmastery of the electromagnetic design of our structure motivates the development of anarray optimized in the 0.6–1 THz range where atmospheric and material transmission aremore favourable.

The short-term aim of developing a commercial camera for scientific use has proved to bevery valuable in helping to align any THz set-up in real time. Meanwhile, industrial applicationsrequiring penetrating imaging with high resolution are being investigated through tests ofthis array in a transmission or reflection configuration. A compact active imaging systemwould fulfil the demands of various fields of applications, such as homeland security, biologyand medical sciences, non-destructive controls for industry (e.g. pharmaceutical, aerospaceand microelectronics), quality control of food and agricultural products, global environmentalmonitoring, and the study of historical and archaeological work. If appropriate, the imagingfunctionality can be considered in combination with spectral analysis for chemical identification,provided by the sensor itself associated with multicolour illumination or possibly with a separatespectrometer.

Acknowledgements. The support from CEA as well as from the EU FP7 project MUTIVIS is gratefullyacknowledged. The authors also acknowledge the Leti team and academic institutions—in France, Universityof Savoie, University of Paris 7, University of Rennes, Ecole Normale Supérieure de Paris, and in Italy, theresearch centre ‘Fondazione di Bruno Kessler’—for their fruitful collaboration.

References1. Jansen C et al. 2010 Terahertz imaging: applications and perspectives. Appl. Opt. 49, E48–E57.

(doi:10.1364/AO.49.000E48)2. Mittleman DM, Jacobsen RH, Nuss MC. 1996 T-ray imaging. IEEE J. Sel. Top. Quantum Electron.

2, 679–692. (doi:10.1109/2944.571768)3. Withayachumnankul W et al. 2007 T-ray sensing and imaging. Proc. IEEE 95, 1528–1558.

(doi:10.1109/JPROC.2007.900325)4. Tonouchi M. 2007 Cutting-edge THz technology. Nat. Photonics 1, 97–105. (doi:10.1038/

nphoton.2007.3)

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

11

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

5. Hu BB, Nuss MC. 1995 Imaging with terahertz waves. Opt. Lett. 20, 1716–1718. (doi:10.1364/OL.20.001716)

6. Simoens F. 2014 THz bolometer detectors. In Physics and Applications of Terahertz Radiation (edsM Perenzoni, DJ Paul), pp. 35–75. Springer Series in Optical Sciences, vol. 173. Dordrecht, TheNetherlands: Springer Science+Business Media.

7. Tissot JL, Tinnes S, Durand A, Minassian C, Robert P, Vilain M. 2010 High performanceuncooled amorphous silicon VGA and XGA IRFPA with 17µm pixel-pitch. Proc. SPIE 7834,78340K-1-8. (doi:10.1117/12.864867)

8. Yon JJ, Astier A, Bisotto S, Chamings G, Durand A, Martin JL, Mottin E, Ouvrier-Buffet JL, Tissot JL. 2005 First demonstration of 25 µm pitch uncooled amorphous siliconmicrobolometer IRFPA at LETI-LIR. Proc. SPIE 5783, 432–440. (doi:10.1117/12.606487)

9. Yon JJ, Mottin E, Tissot JL. 2008 Latest amorphous silicon microbolometer developments atLETI-LIR. Proc. SPIE 6940, 6940–6961. (doi:10.1117/12.780538)

10. Lee AWM, Hu Q. 2005 Real-time, continuous-wave terahertz imaging by use of amicrobolometer focal-plane array. Opt. Lett. 30, 2563–2565. (doi:10.1364/OL.30.002563)

11. Behnken BN, Karunasiri G, Chamberlin DR, Robrish PR, Faist J. 2008 Real-time imaging usinga 2.8 THz quantum cascade laser and uncooled infrared microbolometer camera. Opt. Lett. 33,440–442. (doi:10.1364/OL.33.000440)

12. Oda O. 2008 Detection of terahertz radiation from quantum cascade laser, using vanadiumoxide microbolometer focal plane arrays. Proc. SPIE 6940, 69402Y. (doi:10.1117/12.781630)

13. Simoens F, Durand T, Meilhan J, Gellie P, Maineult W, Sirtori C, Barbieri S, Beere H, Ritchie D.2009 Terahertz imaging with a quantum cascade laser and amorphous-silicon microbolometerarray. In Proc. SPIE Millimetre Wave and Terahertz Sensors and Technology, Berlin, Germany, 1–3September 2009 (eds KA Krapels, NA Salmon), p. 74850M. Proceedings of SPIE, vol. 7485.Bellingham, WA: SPIE. (doi:10.1117/12.830350)

14. Crastes A, Tissot JL, Vilain M, Legras O, Tinnes S, Minassian C, Robert P, Fieque B. 2008Uncooled amorphous silicon 1/4 VGA IRFPA with 25 µm pixel-pitch or high end applications.In Proc. Infrared Technology and Applications XXXIV, Orland, FL, 16 March 2008 (eds BFAndresen, GF Fulop, PR Norton), p. 69401X. Proceedings of SPIE, vol. 6940. Bellingham, WA:SPIE. (doi:10.1117/12.779488)

15. Al Hadi R, Sherry H, Grzyb J, Zhao Y, Förster W, Keller HM, Cathelin A, Kaiser A, PfeifferUR. 2012 A 1 k-pixel video camera for 0.7–1.1 terahertz imaging applications in 65-nm CMOS.IEEE J. Solid-State Circuits 47, 1–14. (doi:10.1109/JSSC.2012.2217851)

16. Boppel S, Lisauskas A, Max A, Krozer V, Roskos HG. 2012 CMOS detector arrays in avirtual 10-kilopixel camera for coherent terahertz real-time imaging. Opt. Lett. 37, 536–538.(doi:10.1364/OL.37.000536)

17. Han R, Zhang Y, Coquillat D, Videlier H, Knap W, Brown E, O KK. 2011 A 280-GHzSchottky diode detector in 130-nm digital CMOS. IEEE J. Solid-State Circuits 46, 2602–2612.(doi:10.1109/JSSC.2011.2165234)

18. Bercier E, Robert P, Pochic D, Tissot JL, Arnaud A, Yon JJ. 2012 Far infrared imaging sensor formass production of night vision and pedestrian detection systems. In Advanced Microsystemsfor Automotive Applications 2012: Smart Systems for Safe, Sustainable and Networked Vehicles (ed.G Meyer), pp. 301–312. Berlin, Germany: Springer.

19. Oda N et al. 2012 Development of terahertz focal plane arrays and handy camera. Proc. SPIE8012, 80121B. (doi:10.1117/12.888992)

20. Bolduc M et al. 2011 Noise-equivalent power characterization of an uncooled microbolometer-based THz imaging camera. Proc. SPIE 8023, 80230C. (doi:10.1117/12.883507)

21. Peytavit E, Agnèse P, Ouvrier Buffet J-L, Beguin A, Simoens F. 2005 Room temperatureterahertz microbolometers. Proc. IRMMW-THz 1, 257–258.

22. Pocas S et al. 2013 Technological customization of uncooled amorphous siliconmicrobolometer for THz real time imaging. Proc. SPIE 8624, 862419. (doi:10.1117/12.2001259)

23. Luukanen A, Miller AJ, Grossman EN. 2004 Active millimeter-wave video rate imaging witha staring 120-element microbolometer array. Proc. SPIE 5410, 195–201. (doi:10.1117/12.543488)

24. Meilhan J, Simoens F, Lalanne-Dera J, Gidon S, Lasfargues G, Pocas S, Nguyen DT, Ouvrier-Buffet J-L. 2012 Terahertz frequency agility of uncooled antenna-coupled micro-bolometerarrays. In Proc. 37th Int. Conf. IEEE Infrared, Millimeter, and Terahertz Waves (IRMMW-THz),Wollongong, NSW, 23–28 September 2012. Piscataway, NJ: IEEE.

25. Simoens F, Meilhan J, Pocas S, Ouvrier-Buffet JL, Maillou T, Gellie P, BarbieriS. 2010 THz uncooled microbolometer array development for active imaging and

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from

12

rsta.royalsocietypublishing.orgPhil.Trans.R.Soc.A372:20130111

.........................................................

spectroscopy applications. In Proc. 35th Int. Conf. IEEE Infrared Millimeter and TerahertzWaves (IRMMW-THz), Houston, TX, 5–10 September 2010, pp. 1–2. Piscataway, NJ: IEEE.(doi:10.1109/ICIMW.2010.5612553)

26. Perenzoni M, Massari N, Stoppa D, Pocas S, Delplanque B, Meilhan J, Simoens F, Rabaud W.2012 A 160 × 160-pixel image sensor for multispectral visible, infrared and terahertz detection.In Proc. 38th European Solid-State Circuits Conference (ESSCIRC 2012), Bordeaux, France, 17–21September 2012, pp. 93–96. Piscataway, NJ: IEEE. (doi:10.1109/ESSCIRC.2012.6341264)

27. Simoens F et al. 2010 Development of uncooled antenna-coupled microbolometer arrays forexplosive detection and identification. Proc. SPIE 7837, 78370B. (doi:10.1117/12.865189)

28. Meilhan J et al. 2011 Active THz imaging and explosive detection with uncooled antenna-coupled microbolometer arrays. Proc. SPIE 8023, 80230E. (doi:10.1117/12.883898)

29. Nguyen DT, Simoens F, Ouvrier-Buffet J-L, Meilhan J, Coutaz J-L. 2012 Simulations andmeasurements of the electromagnetic response of broadband THz uncooled antenna-coupled microbolometer array. IEEE Trans. Terahertz Sci. Technol. 2, 299–305. (doi:10.1109/TTHZ.2012.2188395)

30. Pocas S et al. 2012 Innovative monolithic detector for tri-spectral (THz, IR, Vis) imaging. InProc. Millimetre Wave and Terahertz Sensors and Technology V, Edinburgh, UK, 24 September 2012(eds NA Salmon, EL Jacobs), p. 85440F. Proceedings of SPIE, vol. 8544. Bellingham, WA: SPIE.(doi:10.1117/12.974970)

31. Perenzoni M, Massari N, Pocas S, Meilhan J, Simoens F. 2010 A monolithic visible, infraredand terahertz 2D detector. In Proc. 35th Int. Conf. IEEE Infrared Millimeter and Terahertz Waves(IRMMW-THz), Houston, TX, 5–10 September 2010. Piscataway, NJ: IEEE.

32. Barbieri S, Alton J, Beere HE, Fowler J, Linfield EH, Ritchie DA. 2004 2.9THz quantum cascadelasers operating up to 70 K in continuous wave. Appl. Phys. Lett. 85, 1674. (doi:10.1063/1.1784874)

33. Simoens F et al. 2012 Complete THz system for reflection real-time imaging with uncooledantenna-coupled bolometer arrays. In Proc. 37th Int. Conf. IEEE Infrared, Millimeter, andTerahertz Waves (IRMMW-THz), Wollongong, NSW, 23–28 September 2012. Piscataway, NJ: IEEE.

34. Oden J, Meilhan J, Lalanne-Dera J, Roux J-F, Garet F, Coutaz J-L, Simoens F. 2013 Imaging ofbroadband terahertz beams using an array of antenna-coupled microbolometers operating atroom temperature. Opt. Expr. 21, 4817–4825. (doi:10.1364/OE.21.004817)

35. Meilhan J, Oden J, Cavalié P, Dillhon S, Roux J-F, Coutaz J-L, Tignon J, Simoens F. 2013 Room-temperature video-rate imaging of terahertz pulses generated by femtosecond lasers usingan array of micro-bolometers. In Proc. Int. Workshop on Optical Terahertz Science and Technology2013, Kyoto Terrsa, Japan, 1–5 April 2013. Kyoto, Japan: Kyoto University.

on July 17, 2018http://rsta.royalsocietypublishing.org/Downloaded from