CEA-LETI 12µm pixels for uncooled infrared detectors

1
Welcome
CEA is a French government-funded technological research organization. Drawing
on its excellence in fundamental research, its activities cover three main areas:
Energy, Information and Health Technologies, and Defense and Security. As a
prominent player in the European Research Area, with an internationally
acknowledged level of expertise in its core competencies, CEA is involved in setting
up collaborative projects with many partners around the world.
Within CEA Technological Research Division, three institutes lead researches in
order to increase the industrial competitiveness through technological innovation
and transfers: the CEA-LETI, focused on microelectronics, information & healthcare
technologies, the CEA-LIST dedicated to technologies for digital systems, and the
CEA-LITEN devoted to new energy technologies.
The CEA-LETI is focused on micro and nanotechnologies and their applications,
from wireless devices and systems, to biology and healthcare or photonics.
Nanoelectronics and Microsystems (MEMS) are at the core of its silicon activities. As
a major player in the MINATEC innovation campus, CEA-LETI operates 8,000-m²
state-of-the-art clean rooms, on 24/7 mode, on 200mm and 300mm wafer
platforms. With 1,700 employees, CEA-LETI trains more than 240 Ph.D. students
and hosts 200 assignees from partner companies. Strongly committed to the
creation of value for the industry, CEA-LETI puts a strong emphasis on intellectual
property and owns more than 2,000 patent families.
For more information, visit http://www.leti.fr.
Within CEA-Leti, Optics and Photonics research activities are focused principally
on big industrial markets of photonics: all-wavelength imaging (visible, infrared,
THz), information displays, solid state lighting, optical data communications, optical
environmental sensors. The R&D projects are performed with industrial and
academic partners. The industrial partners of the Optics and Photonics department
range from SME to large international companies. The projects are merging
fundamental aspects with advanced technological and industrial developments;
nanosciences are connected with material sciences, optics, electronics and micro &
nano-fabrication.
2
3
Contents
Page 5
Bruno MOUREY
> Interview
Head of Optics and PhoTonics
Division
Page 7
2012 key figures
Page 9
Scientific activity
Page 11
1- InfraRed Imaging:
cooled detectors
Page 41
4- Optical environmental
sensors
Page 47
5- Optics and
nanophotonics
Page 53
6- Silicon Photonics
Page 61
7- Solid state lighting
(LED)
Page 23
2- InfraRed Imaging:
microbolometers
Page 33
3- Display components
Page 67
8- PhD Degrees awarded
4
5
Interview with Bruno MOUREY,
Head of Optics and PhoTonics Division (DOPT)
Dear Reader,
After two years in its current configuration, the DOPT operates at cruising speed. It is
currently organized into two main applicative sub-divisions:

Imaging
o Cooled infrared
o Uncooled infrared
o Visible
 Novel applications
o Micro displays
o Optical sensors
o Silicon photonics
o Solid state lighting
They rely on a third sub-division that includes all Specific technologies. The Department also
draws extensively on the capacities of the Leti’s mainstream CMOS and back-end clean
rooms.
The Department is covering the whole development chain of modern photonics, from material
science and technological processing to photonic systems. But it ambitions particularly the
development of photonic components in the applicative domains mentioned here before.
A large part of the DOPT developments are done under partnerships with industrial
companies. The volume of the industry-lead activity is steadily growing, particularly the last
two years. So one of our current challenges is to keep a reasonable part of sound advanced
research, to be able to prepare our tomorrow’s applied R&D.
I wish to thank all the team of our Department for their capacity to innovate technically and
scientifically, both in the advanced and in the industrial R&D projects, sometimes under the
constraints of short-term goals.
This scientific report will let you appreciate some of our innovative results of the year 2012.
Have a nice reading!
6
7
2012 Key Figures
185 permanent researchers
35 PhD students and Post-docs
105 scientific publications
57 patents filed in 2012
380 patents portfolio
20% under licence
Clean rooms dedicated to IR
and visible imaging, photonics
fab, packaging
Optics and optoelectronics
characterization facilities
Modeling and simulation
8
9
Scientific Activity
Publications
105 publications in 2012, including conference communications and 40 papers in scientific review
Prize and Awards
II-VI materials 2012 conference Best student Paper Award granted to Alexandre Gaucher.
SPIE Photonics Europe 2012 conference, Best Student Paper awarded to Alexandros Emboras
In 2012, France’s Ministry of Education distinguished Gerard Destefanis with the Knight of the
Order of Academic Palms for his research on mercury cadmium telluride (MCT) IR detectors
Experts
32 CEA experts: 2 research directors, 3 international experts
8 Researchers with habilitation qualification to independently supervise doctoral candidates
Scientific Committees
Editorial Boards of ISRN Nanotechnology
Members of Technical Programs and Steering Committees in major conferences in 2012: SPIE
Security and Défence, Edinburgh,- International Workshop on ZnO materials”, Nice- OPTRO,
Paris.
Books
Participation of JM Fedeli in the “Handbook of silicon Photonics”, Editors L Vivien and L Pavesi and
published by CRC Press.
Others books of CEA / LETI are mentioned on the website:
www.leti.fr/en/Discover-Leti/Books
10
11
1
InfraRed Imaging:
cooled detectors
Bulk CdZnTe growth
HgCdTe epi growth, doping and processing
HgCdTe focal plane arrays: SWIR, LWIR
HgCdTe avalanche single photon detectors
InGaAs SWIR focal plane arrays
Flip-chip assembly and wafer level cameras
12
Growth of bulk CdZnTe :
size increase and quality improvement
Research topics : CdZnTe, growth, single crystal
D. Brellier, E. Gout, G. Gaude
ABSTRACT: The growth of single crystal, bulk cadmium zinc telluride, used as substrate for mercury cadmium
telluride epitaxy, has been optimized in two ways. First, the size of the ingots has been increased to extract very
large substrates, giving the opportunity to make megapixel size infrared focal plane arrays. Moreover, an
excellent reproducibility of crystal singularity has been reached. The Second point concerns crystal quality
evaluated by double crystal X-ray diffraction and etch pit density with very good value.
Bulk cadmium zinc telluride (CdZnTe) is considered as the
ideal substrate for mercury cadmium telluride (HgCdTe)
epitaxy, for infrared applications. The rapid increase in the
infrared focal plane array (IRFPA) size toward megapixel
resolution requires the use of larger dimension substrates.
Besides, good crystal quality of CdZnTe is also necessary for
the realization of epitaxy with a reduced dislocation density,
a drastic requirement for detector performance.
In order to reduce the size and the density of microscopic
defects usually present in the substrates, as precipitates or
inclusions, which can affect the electro-optic properties of
HgCdTe photodiodes, the growth of CdZnTe under specific
conditions has been implemented.
Infrared transmission microscopy shows that defect size has
been reduced from few tens of microns to less than few
microns.
In this context, it is necessary to control the growth of
CdZnTe single crystal ingots, in a reproducible way.
The growth technique used at LETI/DOPT is the Vertical
Gradient Freeze (VGF) method; single crystal and (111)
oriented ingots are usually obtained.
Several ingots of 80 mm and 115 mm in diameter have been
obtained with a single crystal grain from the bottom to the
top (Fig.1). The ingots are fully transparent in the infrared
window, from 1 µm up to 25 µm wavelength range. In the
largest ingots, 63mmx63mm substrates have been extracted
(Fig. 2).
Crystal quality has been investigated using double crystal Xray diffraction, and also, chemical revelation of etch pit
density (EPD). Fig. 3 is a mapping of the rocking curve full
width at half maximum (FWHM) of a 63mmx63mm substrate.
The FWHM mean value is (36±8) Arcsec. The evaluation of
EPD on the (111)Te face leads to values in the low 104/cm².
80 mm
Figure 1 : Single crystalline CdZnTe ingots
Dimension (mm)
115 mm
FWHM (Arcsec)
Figure 3: FWHM mapping obtained by double crystal X-ray diffraction
This quality is comparable to that of the best material
currently available on the world market.
In the framework of DEFIR join laboratory between CEA-LETI
and Sofradir [1], our industrial partner is able to compete with
world leaders on large IRFPAs market.
Figure 2: Evolution of CdZnTe substrate size
Reference:
[1] : M. Vuillermet, D. Billon-Lanfrey, Y. Reibel, A. Manissadjian, L. Mollard, N. Baier, O. Gravrand, D. Brellier, G.
Destéfanis, Proc. SPIE 8541, Electro-Optical and Infrared Systems: Technology and Applications IX, 854109 (October 24,
2012)
13
Strain Determination In Quasi
Lattice-matched LWIR
HgCdTe/CdZnTe
Research topics : HgCdTe, strain relaxation, HRXRD
P. Ballet, X. Baudry, B. Polge, D. Brellier, J. Merlin and P. Gergaud (LETI/DTSI)
ABSTRACT: In this paper, we take advantage of the zinc distribution of (211)B CdZnTe substrates to probe the
lattice mismatch induced stress in HgCdTe layers grown by MBE. HRXRD is used to accurately determine the
strain-free lattice parameters of both CdZnTe and HgCdTe, together with the in-plane components of the stress
tensor. By using several wafers, the stress evolution is derived over a broad range of lattice mismatch. In
particular, stress relaxation is evidenced for a mismatch greater than 0.02% and 0.04% for tensile and
compressively strained HgCdTe respectively. Strain relaxation is correlated with substrate curvature and
rocking curve peak broadening which provide indirect evidence for plastic relaxation.
compressive strain
tensile strain
stress (MPa)
20
0
-20
-0.15
S1 [11-1]
S2 [-110]
-0.10
-0.05
0.00
0.05
0.10
0.15
a/a (%)
Figure 1 : Measured in-plane stress components as a function of
lattice mismatch for the five samples.
The correlation with rocking-curve FWHM is given in fig.2.
The FWHM is lower than 40 arcsecs in the elastic regime,
which indicates no significant additional broadening compared
to the rocking curve of the substrate. On the other hand, in
the relaxed range, the FWHM rapidly reaches 60 arcsecs. This
increase in FWHM is again an indication that plastic relaxation
occurs probably through the generation of misfit dislocations.
Rocking Curve FWHM (arcsecs)
The performance of HgCdTe-based devices is closely related
to the crystalline perfection of the HgCdTe thin film. Lattice
matching is therefore essential, and can be achieved, in
principle, by growing HgCdTe on CdZnTe substrates with an
appropriate zinc fraction. Practically, the zinc content of
CdZnTe substrates varies from one another and within a
single wafer. This is particularly true for (211) substrates
since vertical gradient freeze ingots are usually (111)
oriented leading to a 20° offset in the cutting of the (211)
wafers.
In this work we investigate the strain distribution in five
LWIR HgCdTe layers grown by MBE on 4cmx4cm CdZnTe
(211)B substrates. The substrates have been chosen so that
the overall zinc dispersion allows for the investigation of the
broader range of lattice mismatch induced strain, spanning
from tensile to compressive. The MBE growth has been
carried out in a RIBER 32P system. The HgCdTe thickness is
5 microns for all layers which is the maximum thickness for
which the X-ray diffraction from the substrate can still be
detected for the symmetric (224) reflexion.
In order to extract the in-plane stress components for the
strained HgCdTe layers, we have used the metric-tensor
formalism, first applied to HgCdTe for the measurement of
thermal stress and the extraction of the coefficient of thermal
expansion for HgCdTe [1].
The in-plane stress components are plotted in fig.1 for the
five samples. There are two different behaviours:
(i)
An elastic regime for which the stress evolution is
linear with lattice mismatch. This regime describes a
coherently strained HgCdTe layer for which the strain is
directly proportional to the lattice mismatch between HgCdTe
and CdZnTe.
(ii)
A clear saturation of the stress for absolute lattice
mismatch greater than 0.02% for tensile strain and 0.04%
for compressive strain.
This latter regime is found to be slightly asymmetric towards
compressively strained HgCdTe, consistently with the
asymmetry of the zero stress lattice-mismatch which is a
signature of the thermal expansion coefficient difference
between layer and substrate [1]. It is also found to be
slightly asymmetric in magnitude with an absolute saturated
stress of 12MPa for tensile strain and 18MPa for compressive
strain. This latter value is estimated as the mean value
between the two in-plane components of the stress tensor
because of the significant stress anisotropy measured for
compressively strained HgCdTe layers.
The saturation of the stress clearly indicates that strain
relaxation is taking place with an amount of relaxation
directly proportional to the extrapolated difference between
the lines defined in the elastic and relaxation regimes as
graphically described in fig.1. The onsets for strain relaxation
are found to correspond to zinc fractions of 3.9% and 4.7%.
70
60
50
40
30
20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
a/a (%)
Figure 2 : Evolution of the (224) rocking-curve FWHM as a function of
lattice mismatch.
As a conclusion, our results indicate that for low mismatch,
the stress scales linearly with lattice mismatch whereas for
larger mismatch a strain relaxation regime is evidenced. We
show that the onsets for relaxation could be accurately
determined with strong implication in the zinc dispersion
specifications for CdZnTe wafers.
Références :
[1] P. Gergaud, A. Jonchère, B. Amstatt, X. Baudry, D. Brellier and P. Ballet, J. Electron. Mater. 41, 2694 (2012).
14
Quantitative damage depth profiles
in arsenic implanted HgCdTe
Research topics: HgCdTe doping, ion implantation damages
C. Lobre, D. Jalabert (INAC), I. Vickridge (INP), L. Mollard, P. H. Jouneau (INAC) and P. Ballet
ABSTRACT: Rutherford backscattering (RBS) experiments under channeling conditions have been carried out on
arsenic implanted Hg0.77Cd0.23Te (MCT) layers. Accurate damage profiles have been extracted through a simple
formalism from implanted and annealed layers. Damages evolution with increasing ion implantation dose was
investigated by a complementary approach taking advantage of each characterization mean. Evidences of
irradiation induced annealing process during implantation have been pointed out as well as complexes involving
arsenic in annealed layers.
The p-on-n photodiode architecture leads to desirable device
performance and in particular, low dark currents allowing
higher working temperatures. This architecture requires a
shallow, well-controlled, zone of electrically active p-type
extrinsic dopant. Arsenic is the most used p-type doping
species for this architecture and ionic implantation provides a
suitable means to produce the shallow localized zone.
Nevertheless, quite an important challenge needs to be
overcome in order to activate the implanted arsenic: a strong
electrical n-type doping region related to irradiation damages
defects need to be totally removed.
In our work, correlations between Rutherford backscattering
(RBS) channelling experiments, bright field scanning
transmission electron microscopy (BF-STEM) and secondary
ion mass spectrometry (SIMS) were performed to study
defects induced by arsenic implantation in MCT and their
evolution during post implant anneals.
Damage profiles extracted from the RBS channeling
experiments, arsenic concentration profiles from SIMS and
defect visualization from BF-STEM imaging are summarized
in Figure 1. Damage profile shows a good agreement with the
defect contrasted image. In particular, the depth where the
contrasted zone of the images stops well corresponds to the
fast decrease of the defect profiles. This agreement adds to
the validity of our calculation method of the defect profiles.
When increasing ion implantation doses BF-STEM study
shows a drastic evolution of the defects morphology. Small
dislocation loops seem to coalesce during the implantation
process to form large extended defects. In addition, as shown
in figure 2, quantitative information deduced from the RBS
channeling experiment, bring out that increasing the
implantation dose produce logarithmically deeper defects
maxima. These observations indicate that the MCT
experiences an irradiation induced annealing process during
implantation. The kinetics of damage accumulation is then
controlled by a competition between damage accumulation
and dynamic annealing [1].
Figure 2: Displaced atoms density depth profiles for implantation
doses of 5 × 1013 at.cm-2; 2 × 1014 at.cm-2 and 2 × 1015 at.cm-2
(from left to right). Insert shows the evolution with implanted dose of
total disorder from the integration of damage profiles and maximum
of the damage profile.
Figure 1: BF-STEM image of implanted MCT layer with 2 × 1014 at.cm-3
As ions. Arsenic concentration profiles in green line (left scale). The
dechannelling center density depth profile calculated from
dechanneling experiment is shown in red (right scale).
Damage evolution has been studied for thermally annealed
layers. Fast damage correction has been observed. Damage
profiles of annealed layer associated with arsenic
concentration profile from SIMS indicate the presence of
complexes involving arsenic [1].
Our complementary approach, taking advantage of each
characterization mean, leads us to a better understanding of
damage accumulation in MCT during ion implantation. In
addition evidences of complexes involving arsenic have been
pointed out.
Références:
[1] C. Lobre, D. Jalabert, I. Vickridge, E. Briand, D. Benzeggouta, L. Mollard, P. H. Jouneau and P. Ballet, “Quantitative damage depth profiles
in arsenic implanted HgCdTe”, in preparation for publication.
15
Arsenic complexes optical signatures
in As-doped HgCdTe epilayers
Research topics: p-type doping, photoluminescence, hall effect, ionization energies.
F. Gemain, I.C. Robin, S. Brochen, P. Ballet, O. Gravrand and G. Feuillet
ABSTRACT: The optical signatures of arsenic complexes in As-doped HgCdTe samples grown by molecular beam epitaxy
were clearly identified using comparison between photoluminescence spectra, Extended X-Ray Absorption Fine
Structure (EXAFS) and Hall measurements. The ionization energies of the different complexes were measured both by
photoluminescence and temperature-dependent Hall measurements.
In the past decade, ex-situ p-type doping during molecular beam
epitaxy (MBE) growth of HgCdTe has been widely studied,
because it is the key toward the realization of complex infrared
detector heterostructures such as dual band sensors, hot
detectors, and p-on-n-devices [1,2]. For ex-situ doping, arsenic is
the most used impurity, as it has demonstrated very low diffusion
properties into HgCdTe and also because the available purity and
associated effusion technology meet today’s requirements for
MBE. Nevertheless, there are still many unknowns regarding this
kind of doping. One of the major uncertainties is the knowledge of
the As incorporation site before and after thermal activation.
According to EXAFS measurements carried out by P. Ballet and X.
Biquard [3], a new insight on As site transfer upon thermal
activation was proposed. It was shown that before p-activation
annealing, As incorporated atoms are shared between two
structures: a donor chalcogenide glass As2Te3 and an acceptor
structure AsHg8. The proportion between the two phases is
approximately equal before thermal activation. After p-activation
annealing, the As2Te3 donor is dissociated. Some of the As atoms
from the As2Te3 glass enforce the active AsHg8 acceptor, the other
part occupy Hg sites, creating the AsHg donor defect. Thus, 67% of
the incorporated As atoms are incorporated as AsHg8 acceptors
and 33% are incorporated in the crystal lattice in Hg sites (AsHg)
after thermal activation. Since the concentration of acceptors is
higher than the one of donors upon p-activation annealing, a ptype conductivity is measured.
To identify the As complex optical signatures, we compared PL
spectra of an HgCdTe epilayer with EXAFS measurements and
temperature-dependent Hall measurements. Gaussian functions
were used to fit PL spectra as shown in Fig. 1. The studied HgCdTe
layer was grown by MBE and arsenic doped using an arsenic
cracker cell with a cell and a cracker temperature of 260°C and
500°C respectively. The active layer is 6.7 µm thick with a
cadmium composition of 30.7%, corresponding to a cut-off
wavelength of 4.97µm at room temperature. In order to activate
the arsenic after growth, a p-activation annealing at high
temperature is performed on the sample. In order to fill the Hg
vacancies responsible of uncontrolled p-type doping, a low
temperature annealing under saturating mercury pressure is
performed. This treatment called n-type annealing reduces the Hg
vacancies which are acceptor defects to a few 1013 per cm3, well
below the residual doping of the MBE-grown HgCdTe layers.
Modulated PL measurements with a continuous-scan Fourier
transform infrared spectrometer were carried out on the MCT
epilayer. The PL measurements were performed between 8 K and
300 K using a 1064 nm wavelength Nd-YAG laser for excitation.
The signal was detected with a cooled MCT detector.
A comparison between the PL spectra of the HgCdTe epilayer
before p-activation and after p-activation annealing ( Fig.1) show
a significant evolution of the different PL peaks. The temperaturedependent PL study (Fig.2) allowed us to identify more precisely
these peaks. The HE peak is clearly assigned to the band-to-band
recombination. The ME PL peak at 15.2 meV below the HE peak is
only measured in the spectrum of the sample before p-type
annealing and disappears above 60K. This peak also disappear s
after p-type annealing and a new peak ME2, closer to the HE peak
is then measured. The energy difference between the HE peak and
the ME, ME2 and the LE peak respectively gives the associated
ionization energies.
Figure 1: Comparison of the PL spectra at 8K of an As doped HgCdTe
sample (a) before and (b) after p-activation.
Moreover, temperature-dependent Hall measurements not
described here were carried out on of both samples before and
after p-activation. Data were modeled by the charge-balance
equation [4]. The activation energy found for the sample before
p-type annealing was the same as the one measured for the ME
peak with PL measurements (15.2 meV). Then, the activation
energy
obtained
with
the
temperature-dependent
Hall
measurements of the sample after p-activation corresponds to the
ionization energy of the LE peak measured with PL (26 meV).
Last, the ratio between the acceptor and donor concentration
found by the Hall measurements modeling before and after pactivation corresponds to the ratio measured by EXAFS
measurements for the As2Te3 complex donor, the AsHg8 acceptor
and the AsHg donor defect detailed before.
Figure 2: Temperature-dependent PL study of PL peak positions (a) before
and (b) after p-activation.
That is why, we could clearly associate the ME PL peak to the
As2Te3 complex optical signature, the LE peak to the AsHg8
complex and the ME2 peak to the AsHg donor defect optical
signature with ionization energies of 15meV, 26meV and 11 meV
respectively. To conclude, very consistent results were found
between the different characterization techniques, leading to a
clear identification of the different arsenic complexes, the
determination of their activation energy and their concentration
before and after p-activation.
[1] J. Baylet, P. Ballet, P. Castelein, F. Rothan, O. Gravrand, M. Fendler, E. Lafosse, J. P. Zanatta, J. P. Chamonal, A. Million, and G. Destefanis, J.
Electron. Mater. 35, 1153 (2006).
[2] J. Rothman, G. Perrays, P. Ballet, L. Mollard, S. Gout, and J. P. Chamonal, J. Electron. Mater. 37, 1303 (2008).
[3] X. Biquard, I. Alliot and P. Ballet, J. Appl. Phys. 106, 103501 (2009).
[4]F. Gemain, I.C. Robin, M. De Vita, S. Brochen, Appl. Phys. Lett., 98, 131901 (2011).
16
Characterization of plasma etching
process damage in HgCdTe
Research topics: etched sidewalls, Auger electron spectroscopy, minority carrier lifetime
A. Gaucher, J. Baylet, J.Rothman, E. Martinez and C. Cardinaud (IMN,Nantes)
ABSTRACT: Exposure of n-type long wave infrared (LWIR) Hg1-xCdxTe (MCT) to CH4-H2 based inductively coupled
plasma (ICP) have been investigated in terms of microstructural and electrical damage. The results of an
investigation of the damage of etched sidewalls is presented. Auger electron spectroscopy (AES) has been used
to monitor the evolution of X beneath etched surfaces. Conductivity measurements and minority carrier
lifetimes have been studied on patterned photoconductors from which it is possible to extract a surface
recombination velocity (SRV). These studies have evidenced surface conductivities variations and SRVs shift of
several orders of magnitude, depending on the process used to make the sample.
Hg1-xCdxTe is the most flexible material used to make high
quality infrared detectors arrays. In the future, these arrays
will need to be larger, with higher resolution, and added
functionality, like dual-band detection, avalanche gain, or
higher operating temperature. To accomplish such features,
it is necessary to develop new pixel architectures, which
often include trenches or vias that have to be etched into
HgCdTe. However, HgCdTe is a very sensitive material which
tends to be damaged during the fabrication process,
especially etching steps [1]. Plasma etching is considered as
a good option to structure HgCdTe as it potentially combines
the smooth and damage-free mechanism of wet etching to
the good anisotropy of dry etching. However, the plasma
etching step has to be optimized to fulfill these requirements.
In order to do so, it is important to be able to characterize
the microstructural and electrical damages induced by
plasma etching in both N-type and P-type HgCdTe with
variable composition. We investigate the damage of etched
sidewalls, which actually represent the main part of
patterned etched surfaces in novel architectures [2].
which means that the potential microstructural damage
induced by etching are lower than the demonstrated
performances.
Conductivity and minority carrier lifetimes measurements
have been performed on patterned photoconductors of
variable width. As shown on Fig. 2, the measured lifetime
decreases with the photoconductor width. This shows the
etched sidewalls influence on the electrical properties.
55µm
30µm
20µm
t = 0.6µs
t = 0.2µs
t = 0.1µs
Decreasing width
AES is used to monitor the evolution of X beneath etched
surfaces. An Auger quantification method has been calibrated
to our material (Fig. 1).
1
0.9
0.8
Figure 2: Photoconductive decay evolution with the photoconductor
width
0.7
XCd
0.6
It is possible to extract a surface recombination velocity
(SRV) from lifetime measurements on a set of
photoconductors of variable width. Similarly, it is possible to
extract
a
surface
conductivity
from
conductivity
measurements. These studies have evidenced variations of
SRV and of surface conductivity between ion beam etched,
plasma etched and hybrid plasma and wet etched samples.
0.5
0.4
0.3
0.2
Real stack
0.1
Auger linescan
0
0
5
10
15
Depth (µm)
Figure 1 : Calibration of the AES quantification on a HgCdTe stacking
of variable X layers
Successful quantification was achieved in the range 0.2 < XCd
< 0.3 with a XCd discrimination limit of ΔXCd = 0.02 and an
analysis step of 10 nm. Unfortunately, no stoichiometric
evolution was detected beyond etched surfaces with AES
These characterization methods allow to precisely tune the
etching processes in order to reduce as much as possible the
etched induced damage. These results were presented at the
U.S. Workshop on the Physics and Chemistry of II-VI
materials which took place at Seattle (USA) on November
2012. The Best Student Paper Award was presented to the
author for this work.
References :
[1] J. Baylet, O. Gravrand, E. Lafosse, C. Vergnaud, S. Ballerand, B. Aventurier, J. C. Deplanche, P. Ballet, P. Castelein, J. P. Chamonal, A.
Million, G. Destefanis, “Study of the Pixel-Pitch Reduction for HgCdTe Infrared Dual-Band Detectors”, (2004) Journal of Electronic Materials,
Vol. 33, No. 6, pp. 690-700
[2] A. Gaucher, J. Baylet, J. Rothman, E. Martinez, C. Cardinaud, “Characterization of plasma etching process damage in HgCdTe” submitted to
Journal of Electronic Materials
17
Ultra low dark current FPA for
photon detection in the SWIR range
Research topics : IR detection, ultra low flux, astrophysics
O. Gravrand and L Mollard
ABSTRACT:
Ultra low flux detection is mandatory for next generation astrophysics observatories. In a common effort
between CEA-SAp, CEA LETI, and Sofradir (LETI startup), we have fabricated first European low flux IR retinas.
Ultra low dark currents (0.04e/s) have been demonstrated on a HgCdTe photodiodes arrays hybridized on a Si
dedicated ROIC.
Space based observatories for astrophysics are very demanding
in ultra low flux detection in the IR spectrum. Such low flux
levels represent the detection of a few photons only during long
integration times (typically 1e-/s during several minutes) and
therefore require ultra low dark current photodiodes (0.1e/s)
coupled to a very high performance ROIC stage in terms of noise
and leakage.
We report here first results carried out at CEA and Sofradir to
build such an ultra low dark current focal plane arrays (FPA) in
the short wave infrared range (SWIR) to meet these
requirements,. Those FPAs are dedicated to very low flux
detection in the 2µm wavelength range. In this purpose, Sofradir
has designed a source follower per detector readout circuit
(ROIC), 384x288, 15µm pitch, dedicated to ultra low flux
detection. Leakage currents have been measured down to 0.01
e/s on test pixels, with ultra low readout noises (less than 10
electrons).
material absorbing efficiently IR radiation in this wavelength
range. Both p/n and n/p structures have been evaluated. The
metallurgical nature of the absorbing layer have also been
examined as both molecular beam epitaxy (MBE) and liquid
phase epitaxy (LPE) have been processed. Therefore a large
process window has been explored in order to optimize detection
abilities.
High temperature characterizations have been performed at
CEA-LETI and showed state of the art dark currents. Low flux
characterizations have been carried out at CEA-SAp at low
temperature (from 40 to 160K, see figure 1) in a dedicated light
thigh cryostat. Dark currents as low as 0.04e/s (one electron
every 25 seconds) has been measured at 45K for 2µm cutoff
diodes, with fairly uniform current mappings. As an illustration of
the level of performance obtained, note that the first limiting
mechanism
was
electroluminescence
from
the
silicon
temperature probes (1µm wavelength radiation that may be
sensed by the tested photodiodes). Turning off those probes led
to a factor ten improvement in dark current.
Figure 1: Ultra low dark currents measured at CEA-Sap
This ROIC has been hybridized on different HgCdTe diode
configurations processed at CEA-LETI. HgCdTe is a narrow gap
Figure 2: examples of measured ultra low dark current mappings
Références :
[1] O. Gravrand et al, "Ultralow-Dark-Current CdHgTe FPAs in the SWIR Range at CEA and Sofradir", JEM 41(10) p2686 (2012)
[2] B. Fieque et al., "IR ROIC for very low flux and very low noise applications", SPIE8176-55 (2011)
18
p-on-n VLWIR HgCdTe
Focal Plane Arrays for space applications
Research topics: IR sensors, HgCdTe, p-on-n technology
L.Mollard, N.Baier, O.Gravrand
ABSTRACT: IR detector market is driven by numerous applications in military, security, science and space
domains, which all require the highest-performance sensors currently available. The most commonly used
material, with high performances, is based on the II-VI semiconductor HgCdTe. Historically, Middle Wavelength
(MWIR) and Long Wavelength (LWIR) were focused on defense applications, whereas Very Long Wavelength
(VLWIR) and Small Wavelength (SWIR) are more used for remote sensing projects. In this way, TV/4 arrays, 30
µm pixel pitch, have been manufactured for the VLWIR range, using p-on-n technology. The measured dark
current fits state of the art results.
Different HgCdTe technologies, which enable high-quality
photodiodes, are compatible with backside-illuminated
technology on focal plane arrays. The two most common
technologies at CEA/LETI are n-on-p and p-on-n planar
architectures. Compared to the n-on-p technology, this p-onn structure is characterized by lower dark current, due to the
minority carrier (hole) lifetime. It is also characterized by
lower series resistance, due to the majority carrier (electron)
mobility in the base layer. Such properties are interesting to
improve operability at high temperature and to increase the
format of array ([1]-[3]). For such architecture, the p-type
arsenic area is formed by controlled diffusion of As into an In
n-type doped HgCdTe epitaxial layer. Moreover, two
annealing steps under Hg-overpressure are required for the
formation of the p-on-n photodiodes. The first one is a high
temperature annealing to diffuse and activate the arsenic as
p-type dopant. It also enables to eliminate the radiation
damage resulting from arsenic implantation. The second one
is a low temperature anneal to fill Hg vacancies created
during the previous process step and restore the n-type
indium doping level. A cross-sectional view of our p-on-n
architecture is presented in Figure 1.
array, processed at CEA/LETI, is a TV/4 format, with
respectively 30 µm and 25 µm pitch for λc equal to 12.3 µm
and 15.1 µm. Measurements were made at 78K (λc=12.3
µm) and 50 K (λc=15.1 µm). Excellent performances have
been obtained with our VLWIR IRFPAs. Concerning
responsivity, the operability is equal to 99.93% (λc=12.3 µm)
and 98.7% (λc=15.1 µm) using a ±50% criteria with
dispersions respectively as low as 3.9% and 5.4 %. Using the
same conditions of measurements, the comparison of shotnoise limit and measured Root-Mean-Square (RMS) noise,
clearly shows that diodes are shot-noise limited for both cutoff wavelength. These impressive results in term of array
operability are confirmed by the value of the R0A product. R0A
is defined as the product of the dynamic resistance at zero
bias (R0) by the sensitive area (A), and is inversely
proportional to the dark current density. Figure 2 clearly
shows that compared to n-on-p technology a gain of two
orders of magnitude is obtained. In addition, dark current
measurements are consistent with values previously
reported. None of our devices was found with dark current
densities significantly higher than the Rule 07 model (a p-onn heuristic model), indicating that our diodes are state-ofthe-art.
These results demonstrate the high quality of our p-on-n
photodiodes on both LWIR and VLWIR wavelength. Currently
further studies and efforts are conducted to develop a new pon-n technological approach which could (maybe) enable us
to obtain even better performances.
Figure 1: Cross sectional view of our p-on-n planar photodiode
For VLWIR spectral band, achieving excellent HgCdTe
detector performance is extremely challenging. Indeed,
VLWIR photodiodes are essential for space applications, but
require large Focal Plane Arrays (FPAs) with low defects and
high response uniformity. For such small energy gaps, the
material becomes sensitive to small energy perturbations,
which could be a source of non-uniformity in dark current.
Such developments have been funded by French National
Space Studies Center (CNES) for IASI project ([4]-[6]). The
Figure 2: RoA product at 77K for p-on-n IRFPAs
References :
[1] L.Mollard, Journal of Electronic Materials, Volume 38, Number 8 (2009), 1805-1813
[2] L.Mollard, N.Baier Journal of Electronic Materials, Volume 40, Number 8 (2011), 1830-1839
[3] N.Baier,L.Mollard Proc. SPIE 8353, (May 1, 2012)
[4] L.Mollard, N.Baier, Journal of Electronic Material, 2013  Status of p-on-n As ion-implanted HgCdTe at DEFIR, (to be published)
[5] L.Mollard, N.Baier, Conference on Space telescopes and instrumentation, Overview of p-on-n planar IRFPAs at DEFIR(to be published).
[6] N.Baier,L.Mollard , Conference on Space telescopes and instrumentation, Evaluation of VLWIR p-on-n MCT FPAs(to be published).
19
Linear mode single photon detection
with HgCdTe APDs
Research topics : Avalanche photodiodes, single photon detection, HgCdTe
J. Rothman, G. Vojetta, K. Foubert and F. Guellec (LETI/DACLE)
ABSTRACT: ABSTRACT: Infrared single photon detection have been demonstrated using HgCdTe APDs hybridized
to an ultra low noise read-out integrated circuit (ROIC) designed at Leti. The device has been used to make the
first direct measurement of the gain probability density function (PDF) of a HgCdTe APD and to characterize the
(Poison) PDF of attenuated laser pulses. The gain PDF of the APD is associated with a high detection efficiency
which should enable new applications in science, information technology and imaging.
Figure 1: Single photons events detected with a HgCdTe APD
hybridized with a specially dedicated low noise ROIC
HgCdTe avalanche photodiodes (APDs) with a sufficiently low
cadmium composition are characterized by exceptional
amplification properties that promise to open new windows
for optical observations when only a few photons are
available during the characteristic time of observation, such
as active imaging, lidar, wavefront sensing and photon
counting. These properties are stable linear gains to values
larger than 1000, close to zero excess noise and response
times that are independent of the gain, due to carrier
multiplication which is only induced by the
electrons.
bandwidth and keeping the power consumption low.
DOPT is one of the front runners in the development and
optimization of this technology. Since the first demonstration
in 2006, the focus of the activity at DOPT has been on arrays
of detectors for laser illuminated active imaging. More
recently, an important effort have been made on the
development of detectors assemblies which can detect a
single photon impinging on the detector. The use of HgCdTe
for this application opens the path to enhanced performance
in terms of photon detection efficiency (PDE) and detection
rate. In addition, the linear amplification allows detecting
several photons that arrives at the same instant on a single
detector. This is essential to determine the quantum nature
of a multi-photon state and in applications with a high
dynamic range of signals, such as the detection of backscattered light in atmospheric LIDAR measurements. The
challenge in developing such detectors resides in the
minimizing the noise of the amplifier while maximizing the
frequency bandwidth.
A first clear demonstration of single photon detection using
HgCdTe was reported by DRS in 2011, using a ROIC with a
50 electrons of noise per characteristic time at a bandwidth
of 120 MHz. This detector allowed high SNR single photon
detection at gains higher than 200. We developed an ultra
low noise ROIC with about 10 electrons of noise per
characteristic time and with a low 7MHz bandwidth1. The
detection of single photons with this detector is illustrated in
figure 1. Each peak corresponds to the detection of a single
photon and the fluctuation in amplitude is due to the random
variation of the gain for each avalanche multiplication
process. The distribution in amplitude of the events yields a
direct measurement of the probability density function of the
gain in the APDs, for which we have made the first direct
estimation of the excess noise factor of the APDs. The shape
of the gain PDF implies also that high PDE values, in the
order of 80-90 %, can be achieved at high threshold values
compared to the average gain. In addition, measurements at
zero flux have shown that the rate of dark events is
sufficiently low for most time resolved photon counting
experiments.
Figure 2 shows a first demonstration of using the single
element photon counting device to detect determines the
photon number state of attenuated laser pulses. The Poison
distribution, broadened by the excess noise of the detector
can be detected for photons with an average of at least 5
photons and illustrates the high dynamic range of the
detector which enables, for example, the detection of
complex multi-photon quantum states of light.
── m=5.5
── m=1.25
── m=0.8
Figure 2: Probability density distribution function of Poisson
distributed photon states detected with a DOPT HgCdTe APD single
photon detector, measured at a gain of 70
References :
1. G. Vojetta, F. Guellec, L. Matthieu, K. Foubert, P. Feautrier and J. Rothman, Proc. SPIE 8375, 83750Y (2012)
20
A 15 µm pitch 640 × 512 pixel
hybrid InGaAs image sensor for
night vision
Research topics: Infrared Imaging, SWIR, InGaAs, night vision, low-noise read out circuit
E De Borniol, P Castelein and F Guellec (LETI/DACLE)
ABSTRACT: Through collaboration between III-V Lab and CEA-Leti, a 640 x 512 InGaAs image sensor with 15
µm pixel pitch has been developed. The photodiode array detects the light from the visible to the near infrared
wavelength (0.4 to 1.7 µm). The readout IC (ROIC) was design in a standard CMOS 0.18 µm technology. The
pixel circuit is based on a capacitive transimpedance amplifier (CTIA) input stage with two selectable charge-tovoltage conversion gains. It has been optimized for low noise performance in the high gain mode. The readout
noise is around 30 electrons for a dynamic range of 71 dB in high-gain mode and 108 electrons and 79 dB in
low-gain mode.
InGaAs detectors enable compact light-weight packaging and
easy camera integration. Thanks to a small band gap the
dark current is low and cooling is not mandatory. These
detectors are sensitive in the Short Wave Infrared (SWIR)
band which presents some very interesting characteristics. In
this band the detectors can see through the haze and more
importantly for night vision, they benefit from the emission of
light by the atmosphere, called airglow or nightglow. This can
be a real advantage in dark night conditions (moonless or
cloudy sky). A 15 µm pixel pitch 640 x 512 image sensor
sensitive in this spectral band has been developed through
collaboration between III-V Lab (InGaAs photodiode array)
and CEA-Leti (ROIC). The detector array and read out circuit
were optimized for low light measurement.
dynamic range of 79dB. The noise floor improvement in highgain mode comes with a dynamic range reduction.
Nevertheless the careful CTIA design optimization allowed
reaching a low noise floor of 30e- with CDS at room
temperature for a dynamic range of 71dB. This set of figures,
achieved with a 15μm pixel pitch, brings an improvement
compared to available measurement results of other lownoise hybrid InGaAs image sensors.
Table 1. Measured ROIC performances in high-gain and low-gain
modes.
High-gain
mode
Low-gain
mode
Power consumption
150 mW
150 mW
Conversion gain
17.6 µV/e-
1.9 µV/e-
Full well capacity
Read noise
without CDS
with CDS
105 ke-
1 Me-
92 e- to 110 e30 to 40 e-
166 e108 e-
Maximum dynamic Range
71dB
79dB
An example of image captured with the developed image
sensor in day light conditions is depicted figure 2.
Figure 1. 640x512 pixel @ 15μm pitch InGaAs module
In order to optimize the ROIC noise, a Correlated Double
Sampling (CDS) function has been implemented [1]. The
ROIC can work in three different operating modes: high
frame rate (120 fps with 4 video outputs), CDS (60 fps with 4
video outputs) and low power (60 fps with 2 video outputs).
The measured read out noise and full well capacity are
detailed Table 1 [2]. In low-gain mode the read noise is
about 108e- with CDS for a full well capacity of 1Me- giving a
Figure 2. High resolution image achieved with VGA InGaAs module.
References :
[1] « A low-noise 15-μm pixel-pitch 640×512 hybrid InGaAs image sensor for night vision »
Fabrice Guellec, Sébastien Dubois, Eric de Borniol, Pierre Castelein, Sébastien Martin, Romain Guiguet, Michaël Tchagaspanian, Anne Rouvié,
Philippe Bois ; Proc. SPIE 8298, Sensors, Cameras, and Systems for Industrial and Scientific Applications XIII, 82980C (February 9, 2012);
doi:10.1117/12.912105.
[2] « High-performance 640 x 512 pixel hybrid InGaAs image sensor for night vision »
Eric De Borniol, Fabrice Guellec, Pierre Castelein, Anne Rouvié, Jean-Alexandre Robo, Jean-Luc Reverchon; Proc. SPIE 8353, Infrared
Technology and Applications XXXVIII, 835307 (May 1, 2012); doi:10.1117/12.921086.
21
3D thermo-mechanical simulation
of fine pitch - high count ball grid array
flip chip assemblies
Research topics: Flip chip, Ball grid array, 3D System in Package, Micromechanical modeling
M.Fendler, W.Kpobie, N.Bonfoh*, C.Dreistadt*, P.Lipinski* (*ENIM, Metz)
ABSTRACT: Flip chip technology is increasingly prevalent in electronics assembly (3D System in Package) and is
mainly used at fine pitch for the manufacture of megapixel large focal plane detectors arrays. For the estimation
of the reliability of these assemblies, numerical simulations based on Finite Element Methods (FEM) appear as
the cheapest method. Due to the impossibility of direct modeling of very large assemblies containing more than
one million solder bumps, various simplifications were tested. In the present study, we proposed a model based
on a micromechanical description of the equivalent thermo-elastic properties of the interconnection layer of the
flip chip assembly composed of solder bumps and epoxy filler.
3D System in Package devices, with more functions
in the module, require the miniaturization of their
components and the need of complex and innovative
assembly technologies.
For infrared imagers, the conventional flip chip technology is
performed with indium solder micro-balls to interconnect the
silicon readout circuit with an array of detectors (II-VI or IIIV semiconductor alloys). With the increase of the resolution
of imagers, leading to large megapixel detector arrays at fine
pitches down to 15 µm, optoelectronic components are
subjected to severe thermo-mechanical loads related to their
process of assembly (soldering), and to the cryogenic
operating conditions (77K).
This heterogeneous structure must thus be modeled and
simulated in order to obtain the most reliable assembly at
lower cost in terms of associated technological efforts (choice
of materials, configurations of assembly). 3D modeling seems
more realistic to study the reliability of these flip chip
assemblies.
The megapixel assembly considered in the present work
contains in its interconnection layer over 1.3 million solder
bumps. Modeling such a structure is currently impossible
given the present computer technology. To overcome this
difficulty, the composite interconnection layer made of indium
solder balls and epoxy underfill, has been replaced by a
homogeneous equivalent material (HEM), Fig.1.
Then thermo-elastic properties of the HEM were determined
using the Mori-Tanaka multi-site approximation and its
constitutive law has been implemented in the finite element
code Abaqus® through the UMAT subroutine.
Results deduced from numerical simulations based on the
HEM approach appear in good agreement with experimental
measurements of deflection, measured with an optical
chromatic confocal surface analyser. Moreover, the required
CPU time for this type of simulations remains reasonable
(less than 3 hours) when using the conventional workstations
(Dell Precision T7500, Intel Xeon E5620, 8 CPUs, 24 Go RAM,
2.4 GHz).
Locally, to estimate stress and strain fields in the indium
bumps, two approaches were tested for structural zoom: submodeling and coupled modeling. Comparisons of the obtained
results show that the sub-modeling approach is more
relevant and requires less CPU time than the coupled one.
Sub-modeling applied to a megapixel array assembly
revealed that local stresses are maximal in the vicinity of the
interface between solder bumps and chip (Fig.2). Cracks are
therefore expected to initiate in this area of the flip chip
because of the presence of intermetallic compounds which
are formed during soldering.
Figure 2 Equivalent Von Mises Stress in the solder bumps and epoxy
underfill.
The next step of this study is to validate simulations with
deflection measurements at low temperature, down to 77K.
Figure 1: Homogenisation and localisation thermo-elasticity
References :
Kpobie, W., Ben Khlifa, S., Bonfoh, N., Fendler, M., Lipinski, P. Multi Site micro mechanical modelling of thermoelastic properties of
heterogeneous materials (2012) Composite Structures, 94, pp.2068-2077
Ben Khlifa, S., Bonfoh, N., Lipinski, P., Fendler, M., Bernabe, S., Ribot, H. Thermomechanical characterization of electronic components
(2010) 11th International Conference on Thermal, Mechanical and Multiphysics Simulation and Experiments in Micro-Electronics and MicroSystems, EuroSimE 2010, art. no. 5464592.
22
Ultra compact infrared cryogenic
wafer level camera
Research topics: Infrared camera on chip, compound eyes, 3D system in package
M.Fendler, G.Lasfargues, F. De La Barriere*, G.Druart*, N.Guerineau* (*ONERA, Palaiseau)
ABSTRACT: A compact infrared cryogenic multichannel camera has been performed with a wide field of view
equal to 120°. By merging the optics with the detector, the concept is compatible with both cryogenic
constraints and wafer-level fabrication. The optical system is limited by the diffraction. By cooling the optics, we
achieved a very low NETD equal to 15 mK, compared with traditional infrared cameras. A post processing
algorithm which aims at reconstructing a well-sampled image from the set of undersampled raw subimages
produced by the camera is validated on experimental images.
Nowadays, both civilian and military applications
require miniaturized and cheap optical systems. The
constraints on the size and weight of such systems are so
demanding that downscaled versions of traditional systems
with a single optical axis are reaching their limits because of
a loss of resolved points in the final image. A solution to
compensate for this problem can be found by looking at
nature, where small invertebrates have developed compound
eyes. Therefore, multi-aperture imaging seems to be a good
approach to design thin optical systems. All these concepts
tend to merge the optical system with the detector, leading
to very thin wafer-level camera modules.
Different multichannel approaches, which differ in the way of
dividing the information contained in the scene, can be
distinguished. We choosed the TOMBO principle: each
channel captures a low-resolution image of the overall Field
Of View (FOV). Providing non redundant information between
the subimages, a high resolution image is retrieved thanks to
a super-resolution image processing.
We performed a 4x4 multichannel wafer-level infrared
camera, working in the 3 − 5µm spectral range and directly
integrated on a cooled IRFPA (Fig.1) [1]. Each channel has
the same wide field of view equal to 120°, and a very short
focal length (the total track length of the camera is equal to
4.08 mm). It is efficiently prevented from crosstalk between
the channels thanks to the use of the cold shield and of a
field diaphragm array.
at CEA LETI MINATEC. These techniques are directly inspired
by the hybridization of an IRFPA [2]. Basically, an IRFPA
based on HgCdTe technology is made of an HgCdTe layer
(which is sensitive to infrared radiations) and of a Silicon
readout circuit. The HgCdTe layer and the Silicon read-out
circuit are connected with indium bumps. The idea is to
support the optical module at a short distance from the
HgCdTe layer with indium bumps also. We used the selfalignment properties of micro-bumps soldering, to guarantee
a sub-micronic optical positioning between the different
optical stages (detector and two microlenses arrays, Fig.2).
The whole assembling process is compatible with wafer-level
fabrication, and is adapted to the cryogenic environment
(thermo-mechanical behaviour).
Figure 2:3D stack of optical and detecting devices
The resolution of a single subimage can be highly improved
by applying a simple shift-and-add post processing algorithm
to the set of undersampled raw subimages acquired by the
camera (Fig.3) [1].
Figure1: Cooled Infrared Cam-On-Chip
The optical wafer-level module is assembled with the IRFPA
using advanced, innovative and precise techniques developed
Figure 3: High resolution image after post processing.
References :
[1] De la Barriere, F., Druart, G., Guerineau, N., Lasfargues, G., Fendler, M., Lhermet, N., Taboury, J.
Compact infrared cryogenic wafer level camera: design and experimental validation(2012) Applied Optics, 51(8), pp.1049-1060
[2] Lasfargues, G., Fendler, M., De La Barriere, F., Guerineau, N., Druart, G., Ribot, H., Moullec, J.-B.
Surface shaping and 3D interconnection of optical functions on photodetector, using an optimized reflow process of solder bumps.
(2012) Proceedings of Electronics System Integration Technology Conference, 4 th ESTC, to be published
23
2
InfraRed Imaging:
Microbolometers
Bolometer-based IR, THz, and multispectral
Focal plane arrays: small pitch
Pixel level packaging arrays
Curved focal plane arrays
24
CEA-LETI 12µm pixels
for uncooled infrared detectors
Research topics : Uncooled infrared detector – Microbolometers – 12µm
S. Becker, P. Imperinetti, J.-J. Yon, J.-L. Ouvrier-Buffet,
V. Goudon, A. Hamelin, C. Vialle, A. Arnaud
ABSTRACT: Recent developments at CEA, LETI have demonstrated the successful integration of 12µm pixels on a
commercial VGA ROIC
Shrinking the pixel pitch usually leads to a dramatic drop in
sensitivity of thermal infrared (IR) detector. Therefore, in
order to keep a signal-to-noise ratio acceptable, one way
followed at CEA, LETI has been to improve the
photolithographic resolution during the microbolometer
process in order to reduce the width of the thermistor
supporting legs. For these developments, the legs have been
defined as thin as 300nm, leading to a thermal insulation
greater than 200 MK/W.
The TV-format ROIC (640x480 pixels) used for this study was
initially designed for 17µm pixels, so an adaptation of the
electrical
contacts
between
the
CMOS
and
the
microbolometers has been necessary. The fill factor is
naturally reduced in this situation, see figure 1.
The cavity design has been adjusted to maximise the
absorption coefficient in the range [8-12µm]. The mean
absorbance value between 8 and 12µm has been measured
at 63%. Since the fill factor is ~50%, this outstanding result
proves that the optical cross section is greater than the
geometrical cross section. Simulations performed in Comsol®
environment have confirmed this behavior.
Complementary electro-optical tests have been carried out
using a f/1 optical aperture and a frame rate of 30Hz.
Responsivity
was
measured
using
two
blackbodies
respectively set at 293K and 303K. The FPA temperature was
regulated by a thermo-electrical cooler (TEC) at 303K. Noise
Equivalent Temperature Difference (NETD) and responsivity
have been measured with a capacitive gain of 6pF and an
integration time of 65µs.
Table 1 : Electro-optical performances
Responsivity
11.3mV/K
Temporal NETD
53.6mK
Pixel time constant
6.6ms
FOM
353mK.ms
The figure of merit (FOM) of IR uncooled detectors,
expressed as the product between the temporal NETD and
the time constant, is estimated around 353mK.ms. This FOM
is similar to that of commercial 17µm a-Si FPA.
Figure 1 : SEM photography of 12µm pixels on a 17µm pitch ROIC
The electrical contacts between the readout circuit and the
microbolometers have been achieved using a copper
damascene process provided to fill the openings in the CMOS
passivation layer with copper plugs. A reflective metallic film
connecting the copper plugs and the microbolometer inputs
has been deposited and patterned in order to manage the
mismatch between the ROIC pitch and the microbolometer
size. Such a damascene arrangement leads to a planar
surface pretty perfect that improves significantly the
conformation of the quarter-wave optical cavity and as a
result, the absorption of the pixel. At the same time, the
planar surface provided by the damascene option improves
the photolithographic resolution used to pattern the
thermistor supporting legs.
A component of the first batch has been associated with a
35mm – F/1 objective, proving excellent imaging quality.
Figure 2 : IR image obtained with 12µm pixels on TV-format ROIC
Latest pixel size reduction of uncooled IR-FPA at CEA, LETI,S Becker, P Imperinetti; JJ Yon ; JL OuvrierBuffet ; V Goudon ; A Hamelin ; C Vialle ; A Arnaud Proc. SPIE 8541, Electro-Optical and Infrared Systems: Technology and
Applications IX, 85410C (October 24, 2012)
References :
25
Broadband THz Uncooled AntennaCoupled Microbolometer Array.
Electromagnetic Design,
Simulations and Measurements
Research topics : THZ Antenna-coupled bolometer, electromagnetic simulation, FTIR
Duy-Thong Nguyen, François Simoens, Jean-Louis Ouvrier-Buffet, Jérôme Meilhan, and Jean-Louis
Coutaz (IMEP-LAHC, Université de Savoie)
ABSTRACT: Bolometer sensors are good candidates for THz imaging thanks to their maturity and capability to
sense THz waves on the whole spectrum. Starting with infrared microbolometer technology, uncooled antennacoupled microbolometer focal plane array are being developed at CEA-LETI with the objective of offering lowcost, real-time 2D terahertz imaging sensors. One of the major challenges is studying the optical coupling
mechanism of the detector in THz frequency range. We present results of the electromagnetic design and
characterization process of these focal plane arrays, concentrating on the spectral absorption.
II.
DESIGN AND SIMULATION
A schematic of one pixel of the detector array is shown in Fig.
1. The 50 µm pitch detector associates quasi-double-bowtie
antennas to a thermometer microbridge structure derived
from the standard infrared bolometers. The function of these
antennas is to couple the incident wave to the load
resistances, located at the center of the bolometer.
Absorptance Simulation
1
0.8
Absorptance
I.
INTRODUCTION AND BACKGROUND
For THz detection, standard uncooled microbolometer
sensors are not optimized and exhibit limited sensitivity [1].
Thus, starting from the well-mastered infrared technology;
developments are being undertaken by several institutes to
shift response of IR detectors to THz range [2] – [3]. Such
adaptation entails significant works of electromagnetic
modeling of the pixel and development of experimental
techniques to validate the designs in this poorly exploited
spectral range where measurement methods are most of the
time inexistent. This paper presents the design, simulation
and experimental validation of a broadband THz antennacoupled microbolometers developed at CEA-LETI.
0.6
0.4
0.2
0
0.6
0.9
1.2
1.5
1.8
2.1
2.4
2.7
3
Frequency [THz]
Figure 2: Simulated absorptance of the detector for cross polarization
Absorption is evaluated through interferogram measurement
of bolometer with a TDS where a chopper and a lock-in
amplifier are employed to enhance SNR. The simulated
absorption matches very well with experiment for considered
polarization. The same absorption peak as the simulation at
1.35 THz is present and the shape of the absorption curve
exhibits the same features.
Figure 3: FTIR absorptance measurement for one polarization
Figure 1: Simulation scheme of one antenna-coupled bolometer pixel
To enhance antenna gain, an equivalent quarter-wavelength
resonant cavity is realized under the antennas with an 11-μm
thick SiO2 layer deposited over a metallic reflector.
The 2D focal plane array (FPA) is a periodic arrangement of
the previously described antenna-coupled micro-bolometers.
Thus this 2D antennas arrays can be treated as a frequency
selective surface (FSS) and simulations are performed for a
single pixel using Floquet port with periodic boundary
condition. Knowing the total dissipated heat on the
microbridge and the power delivered at the Floquet port, one
can infer the absorption efficiency of the detector (Fig. 2).
References :
[1]
[2]
[3]
[4]
III.
CONCLUSION
Results of electromagnetic designs of antenna coupled
microbolometers are successfully compared to detector’s
spectral absorptance measurement [4]. Both modeling of
such complex structure and characterization difficulty in the
THz domain have been overcome and showed very promising
results.
These tools open the way to future enhancement of THz
uncooled bolometer imaging array performances and
absorption range through new optimized designs.
Lee A., & al. "Real-time imaging using a 4.3-THz quantum cascade laser and a 240×320 element focal-plane array,"
Lasers and Electro-Optics, 2006 and 2006 Quantum Electronics and Laser Science Conference. CLEO/QELS 2006.
Conference on , vol., no., pp.1-2, 21-26 May 2006
M. A. Dem’yanenko, & al. "Microbolometer detector arrays for the infrared and terahertz ranges", Journal of Optical
Technology, 76, 739 (2009)
N. Oda, "Uncooled bolometer-type terahertz focal plane array and camera for real-time imaging", C. R. Phys. 11, pp.
496–509, 2010.
D.T. Nguyen et al., Broadband THz Uncooled Antenna-coupled Microbolometer Array – Electromagnetic Design,
Simulations and Measurements, IEEE Transactions on Terahertz Science and Technology (2012)
26
Terahertz frequency agility of
uncooled antenna-coupled microbolometer arrays
Research topics : Multispectral IR-VIS-THz imager, sub-THz uncooled antenna bolometer
Jérôme Meilhan, François Simoens, Jérémy Lalanne-Dera, Serge Gidon, Gilles Lasfargues,
Stéphane Pocas, Jean-Louis Ouvrier-Buffet
ABSTRACT: Antenna coupled bolometer developed at CEA-Leti can address the whole THz range by proper
tailoring of the antennas while keeping the technological stack unchanged. This paper presents the flexibility of
the design regarding absorption frequency and presents its capability to sense frequencies below 1 THz.
I.
INTRODUCTION AND BACKGROUND
The development of Terahertz (THz) applications is restrained
by the availability of affordable and highly-sensitive
detectors. CEA-LETI took up this challenge by extending the
sensibility of Infrared (IR) bolometer to THz range. Key
feature of these detectors is the addition of tunable dipole
antennas to the standard IR bolometer structure processed
above a dielectric resonant cavity. Unlike other technological
approaches under development (for ex. [1]), the optical
absorber and thermometer are separate.
We report on the pixel structure developed so far and detail
the design of absorbing antennas. Multispectral application of
antenna bolometers is also demonstrated with the realization
of a multi-spectral imager. Finally, we present current studies
to extend sensitivity below 1THz for passive imaging and
demonstrate feasibility through preliminary simulation
results.
The flexibility of the developed antenna structure leads the
path to multispectral THz FPA. A monolithic 160x160
bolometers FPA assembling IR sensors surrounding 32x32
THz pixels has been developed [3]. Tested performances are
at state of the art of both imaging domains as NETD of IR
channel is better than 50mK and NEP of THz channel is
comparable to previous achievements [4].
II.
Real-time video acquisition has been
laboratory test setup illustrated in Fig.
optical test pattern standing in the way
Cascade Laser (QCL) beam optical path
Visible and THz ranges.
RESULTS
Figure 2 : Setup for Visible / THz real time imaging tests
performed in the
2 where a paper
of a THz-Quantum
is imaged both in
Simulated Absorptance of Pixel 850 GHz
1
Absorption
0.8
0.6
0.4
0.2
0
0.5
0.6
0.7
0.8
0.9
1
Frequency [THz]
Figure 2 : Absorption of adapted antenna bolometer for f < 1 THz
Figure 1 : Detailed structure of 2-stage antenna-coupled bolometer
The two-stage antenna structure (Fig. 1) provides efficient
coupling of the 2 cross-polarization components of incoming
radiation [2]. One polarization excites the longer bowtie
antenna (“DC antenna”). A lower stack level antenna (“CC
antenna”) collects the other polarization. A sub-antenna
located on the microbridge capacitively couples this
polarization-related current. Induced currents of both
antennas heat up the suspended membrane through Joule
dissipation in matched load resistances. From now on this
technological stack is consolidated and absorption frequency
is adjusted by proper tailoring of antenna dimensions and
loads. Designs optimized for 1.7 THz and 2.4THz central
frequency have shown excellent agreement between FEM
modeling and absorption measurements.
Imaging at frequency f below 1 THz can be addressed while
keeping thickness of the resonant cavity and µ-bridge
dimensions by integration of metallic structures in the vertical
technological stack. In Fig. 3, first simulations of a 70-μm
pitch antenna-coupled bolometer show better than 80 %
optical absorption, between 0.8-0.9 THz.
III.
CONCLUSION
Antenna-coupled bolometer solution for uncooled THz
imaging is on its way to reach maturity. The pixel structure
has been successfully adapted to different frequencies and its
integration in a monolithic multispectral FPA has been
demonstrated.
Foreseen
studies
will
address
the
demonstration of the structure to be tuned seamlessly to low
THz frequency range.
References :
[1] Oda N, “Uncooled bolometer-type Terahertz focal plane array and camera for real-time imaging”, C. R. Physique 11, pp.496–509 (2010)
[2] D.T. Nguyen et al., Broadband THz Uncooled Antenna-coupled Microbolometer Array – Electromagnetic Design, Simulations and
Measurements, IEEE Transactions on Terahertz Science and Technology (2012)
[3] M. Perenzoni., “A monolithic visible, infrared and terahertz 2D detector”, IRMMW-THz 2010
[4] F. Simoens et al, “Real-time imaging with THz fully-customized uncooled amorphous-silicon microbolometer FPAs”, SPIE 2012
27
Innovative monolithic detector for
Tri-Spectral (THz, IR, Vis) imaging
Research topics : micro-bolometer, THz, Infra-Red, photodiode, multispectral
S.Pocas, M. Perrenzoni *, F. Simoens, J. Meilhan, W. Rabaud, B. Delplanque, A. Arnaud
(*Fondazione Bruno Kessler, Povo, Italy)
ABSTRACT: Fusion of multispectral images has been explored for many years for security and used in a number
of commercial products. CEA-Leti and FBK have developed an innovative sensor technology that gathers
monolithically on a unique focal plane arrays, pixels sensitive to radiation in three spectral ranges that are
terahertz (THz), infrared (IR) and visible. This technology benefits of many assets for volume market:
compactness, full CMOS compatibility on 200mm wafers, advanced functions of the CMOS read-out integrated
circuit (ROIC), and operation at room temperature. The ROIC houses visible APS diodes while IR and THz
detections are carried out by microbolometers collectively processed above the CMOS substrate.
CEA-Leti and FBK have developed a multispectral, real time
imaging sensor [1]. The main challenge was to integrate
monolithically three different sensing principles on the same
substrate: the detector combines pixels sensitive in the
visible and NIR range (400-900nm), in the thermal infrared
band (8-12 µm) and in the terahertz frequency range from 13 THz. Such integration has benefited from mature thermal
infrared bolometer technology [2] and developments of
innovative THz antenna-coupled silicon microbolometer [3] at
CEA-Leti.
The specifications of the sensor have been set to suit low
price applications. This constraint has motivated the design
of a focal plane on a CMOS substrate fully compatible with
standard silicon processing and operation at room
temperature to avoid any expensive cooling equipment
(thermo electric coolers, liquid helium, etc.).
The IR and THz pixels are simultaneously manufactured
above the ASIC wafer. First, a THz metallic reflector is
patterned on the CMOS passivation layer. Then an 11µm
planarized thick oxide is deposited. This stack acts as a
quarter wavelength cavity which improves significantly the
optical coupling efficiency for the THz bolometer. The micro
bridges are fabricated above this cavity, applying process
steps directly transposed from the CEA-LETI standard
infrared bolometer flow chart. Copper-plugged TOVs
(Through silicon-Oxide Vias) connect the bolometer sensor
level down to the CMOS read-out-circuit (ROIC) upper metal
pads through the thick dielectric cavity.
The CMOS ROIC designed by FBK, supports both visible 2D
array pixels and the mandatory Application-Specific
Integrated Circuit (ASIC) for the 3-channel signal readout.
The FPA has been arranged with 32x32 THz pixels with a
50µm pitch surrounded by 160X160 IR pixels with a 25µm
pitch (Figure 2). The visible diodes are located underneath
each IR pixel buried inside the CMOS substrate.
Figure 2: Transversal view of the tri-spectral focal plane
All
the
electrical
tests
performed
on
this
new
microbolometers combination, show a high level of
operability which means a perfect control of the technological
manufacturing. Indded we obtained many dies per wafer with
an operability of 100% for the THz microbolometers and
>99.5% for the IR bolometers in the same Focal Plane
Arrays.
Figure 1: Schematic view of the Focal Plane Array
The performances of each bolometer conform to the state of
the art in term of sensivity. Indeed we measured a NETD
<50mK for IR pixel and a NEP~33pW @1.7THz for THz
pixels. Due to the high yield and good sensivity, real time
videos with a good quality have been performed in the three
spectral modes.
References :
[1] S. Pocas, M. Perenzoni N. Massari, F. Simoens, J. Meilhan, W. Rabaud, S.Martin, B. Delplanque, P Imperinetti, V. Goudon, C. Vialle, A.
Arnaud, “Innovative monolithic detector for tri-spectral (THz, IR, Vis) imaging”, Proc. SPIE 8544, (2012)
[2] J.J. Yon, A. Astier, S. Bisotto, G. Chaming’s, A. Durand, J. L. Martin, E. Mottin, J.L. Ouvrier-Buffet, J. L. Tissot, “First demonstration of
25µm pitch uncooled amorphous silicon microbolometer IRFPA at LETI-LIR, Proc. SPIE 5783, 432-440 (2005).
[3] F. Simoens, J. Meilhan, B. Delplanque, S. Gidon, G. Lasfargues, J. Lalanne Dera, D.T. Nguyen, J.L. Ouvrier-Buffet, S. Pocas, T. Maillou, S.
Barbieri , “Real-time imaging with THz fully-customized uncooled amorphous-silicon microbolometer focal plane arrays”, Proc. SPIE 8363,
(2012)
28
World first infrared image using Pixel
Level Packaging technology
Research topics : Infrared detector, microbolometer, pixel level packaging
G. Dumont, L. Carle, V. Goudon, C. Vialle, S. Becker, A. Hamelin, A. Arnaud, JJ. Yon
ABSTRACT: The success and the commercial availability of uncooled infrared detectors has recently attracted
new interest in the outlook of IR sensors of lower resolution for high volume markets such as home automation.
The cost objectives for these markets however require a technological breakthrough, particularly in regard to
the vacuum packaging of these components which remains an adverse cost driver for the microbolometer
technology. In this context, CEA-LETI has proposed and has committed in a pixel level packaging technology.
This paper presents the recent development at CEA-LETI on this technology that aims at encapsulating each
pixel under vacuum in the direct continuity of the bolometer process.
Pixel Level Packaging (PLP hereafter) process consists in the
manufacturing of IR transparent microcaps that cover each
microbolometer (i.e. each pixel) of the array [1](see Fig.1).
To be efficient, the microcap has to be hermetically sealed
under vacuum and maintain the vacuum level in the 10-3
mbar range requested for nominal IR detector operation. The
main point (regarding cost reduction objectives) is that the
PLP process is thoroughly carried out directly on the readout
integrated circuit and bolometers wafer, in a full collective
way.
compatibility of PLP with respect to bolometer and CMOS
technologies.
Moreover, resulting devices were used to take the world first
infrared images using a pixel level packaged detector (see
Fig.2). The optical performances measured on these devices
prove the vacuum quality inside the packaging and are
already compatible with IR sensors applications. [3]
Anti-reflection and sealing layer
Microcap
Reflector / Getter
Bolometer
Figure 1: SEM view of a microbolometer after PLP process
CEA-LETI has already successfully demonstrated that the PLP
concept, when applied on a single microbolometer pixel, can
provide the required vacuum below 10-3 mbar [2]. Since this
feasibility milestone, CEA-LETI has pushed forward the
development of this technology: the bolometers and PLP
processes have been implemented onto operational readout
circuits CMOS base wafers with a 34µm pixel pitch.
Electrical resistance measurements have been performed
directly on wafers at different steps of the overall process in
order to evaluate the impact of PLP over bolometer and
CMOS technologies. First measurements were done after the
bolometer process and before PLP. Second measurements
were performed after PLP. Therefore it has been possible to
follow the overall number of defective pixels before and after
PLP resulting in a PLP yield of 99.77%. This demonstrates the
References :
Figure 2: First infrared image made with a PLP device
Therefore PLP technology has been successfully carried out
on a 320x240 readout circuit leading to the first presentation
of a PLP image. This key result demonstrates that, as a very
low cost collective packaging technology, PLP is well suited to
address the high volume market of low-resolution IR sensors.
The next step aims at developing a fully dedicated 80x80
CMOS readout circuit to address the foreseen market.
57 patents filed in 2011
[1] M. Vilain, “Procédés et dispositifs de fabrication de détecteurs de rayonnement”, FR 2 822 541 French Patent, 380
(2001).patents
portfolio
[2] G. Dumont, W. Rabaud, X. Baillin, JL. Pornin, L. Carle, V. Goudon, C. Vialle, M. Pellat, A. Arnaud, “Pixel level
packagin
g for uncooled
20%
under
licence
IRFPA”, Infrared technology and Applications XXXVII, Proc. SPIE Vol. 8012, (2011).
[3] G. Dumont, W. Rabaud, J-J. Yon, L. Carle, V. Goudon, C. Vialle, Sébastien Becker, Antoine Hamelin, A. Arnaud, “Current progress on pixel
level packaging for uncooled IRFPA”, Infrared technology and Applications XXXVIII, Proc. SPIE Vol. 8353, (2012).
29
Pixel level packaging
technology performances
Research topics : Infrared detector, microbolometer, pixel level packaging
G. Dumont, L. Carle, V. Goudon, C. Vialle, S. Becker, A. Hamelin, A. Arnaud, JJ. Yon
ABSTRACT: The success and the commercial availability of uncooled infrared detectors has recently attracted
new interest in the outlook of IR sensors of lower resolution for high volume markets such as home automation.
The cost objectives for these markets however require a technological breakthrough, particularly in regard to
the vacuum packaging of these components which remains an adverse cost driver for the microbolometer
technology. In this context, CEA-LETI has proposed and has committed in a pixel level packaging technology.
This paper presents the optical and ageing performances measured on the first devices using this technology
manufactured at CEA-LETI.
Pixel Level Packaging (PLP hereafter) process consists in the
manufacturing of IR transparent microcaps that cover each
microbolometer (i.e. each pixel) of the array [1](see Fig.1).
To be efficient, the microcap has to be hermetically sealed
under vacuum and maintain the vacuum level in the 10-3
mbar range requested for nominal IR detector operation. The
main point (regarding cost reduction objectives) is that the
PLP process is thoroughly carried out directly on the readout
integrated circuit and bolometers wafer, in a full collective
way.
Anti-reflection and sealing layer
Furthermore, in order to evaluate the PLP vacuum integrity
along time, CEA-LETI has initiated some accelerated ageing
tests on unitary PLP pixels which are fully representative of
those of the IRFPA. For that, the thermal resistance of a PLP
test device has been regularly measured while being baked in
an oven at 90°C during one month. Indeed, as the thermal
resistance of a microbolometer is strongly linked to its
surrounding pressure, this is a good indicator to assess the
residual pressure inside PLP cavities. These measurements
were performed in-situ, enabling to keep the device under
test at 90°C all the time.
Microcap
Reflector / Getter
Bolometer
Figure 1 : SEM view of a microbolometer after PLP process
Figure 2 : Thermal resistance evolution during 30 days at 90°C
CEA-LETI has already demonstrated that the PLP can provide
the required vacuum below 10-3 mbar [2]. PLP technology
has also been implemented onto operational readout circuits
CMOS base wafers resulting in the first infrared images taken
with pixel level packaged detectors [3].
After one month ageing at 90°C, the thermal resistance
shows a degradation of only 3% with respect to its starting
value (see Fig.2). This excellent result is in agreement with
previous work [3], where no noticeable evolution of the
thermal resistance was measured after a full year at room
temperature.
A Noise Equivalent Temperature Difference (NETD)
measurement has also been performed on these first imaging
devices resulting in a value of 117mK. If we subtract the
acquisition chain noise in a quadratic way, the intrinsic
bolometer NETD is evaluated at 100 mK. This first result is
very promising as it already fit the required performances for
addressing the low-end and low-cost IR sensors market.
Moreover, the pixel time constant is 6ms and is therefore
compatible with classical operation at 60Hz.
Although this preliminary result must be improved to reach
the required lifetime expected for the foreseen application, it
is already a key result as it proves that PLP still features a
residual pressure below 10-2 mbar after this accelerated
ageing test. This result enforces PLP technology as a
conceivable packaging solution for uncooled IRFPA for very
high volume applications.
References :
[1] M. Vilain, “Procédés et dispositifs de fabrication de détecteurs de rayonnement”, FR 2 822 541 French Patent, (2001).
[2] G. Dumont, W. Rabaud, X. Baillin, JL. Pornin, L. Carle, V. Goudon, C. Vialle, M. Pellat, A. Arnaud, “Pixel level packagin g for uncooled
IRFPA”, Infrared technology and Applications XXXVII, Proc. SPIE Vol. 8012, (2011).
[3] G. Dumont, W. Rabaud, J-J. Yon, L. Carle, V. Goudon, C. Vialle, Sébastien Becker, Antoine Hamelin, A. Arnaud, “Current progress on pixel
level packaging for uncooled IRFPA”, Infrared technology and Applications XXXVIII, Proc. SPIE Vol. 8353, (2012).
30
Hemispherical infrared focal plane
arrays: a new design parameter
for the instruments.
Research topics: Curved focal plane array, infrared, large field of view instruments
M.Fendler, D.Dumas, P.Laporte* (*OBSPM, Meudon)
ABSTRACT: In ground based astronomy, mainly all designs of sky survey telescopes are limited by the
requirement that the detecting surface is flat whereas the focal surface is curved. We propose an ideal solution
which is to curve the focal plane array in a spherical shape, thanks to our monolithic process developed at CEALETI based on thinned silicon substrates which allows a 100% optical fill factor
In astronomy, projects such as wide survey ground
based telescopes and the miniaturization of spatial launched
instruments need a novel camera design. Correction of
aberrations is one of the primary challenges to achieve that
camera optical design breakthrough; nature gives us an
interesting solution, which has never been copied before, by
curving the focal plane of animal and human eyes. Indeed,
the curved shape suppresses aberration of curvature and
then leads to a simplification and a miniaturization of the
camera design (Fig.1) [1].
Figure 2: Concave hemispherical 320 x 256, 25µm pitch infrared
microbolometers array (80mm bending radius)
Figure 1: Performance (PSF) and camera design simplification
provided by curved curved focal plane
CEA-LETI has developed a process to curve detectors in
hemispherical shapes. The process allows curving the whole
detector without modifying the flowchart and saving a 100 %
fill factor (Fig.2) [2,3]. Infrared pixels did not suffer from
mechanical and electrical damages after the spherical
bending. The camera based on curved detector is fully
compliant with standard cameras in terms of integration and
electrical noise (Fig.3), and moreover, it offers better optical
quality as well as a simplification and miniaturization of the
optical design.
Figure 3: Electro-optical performances of the monolithic
curved focal plane array (see Fig.2)
The bending radius is now a new parameter for future optical
designs that might revolutionize telescope architecture, as
well as any kind of retina-based imaging systems needing
simplification and miniaturization of the optical design.
Indeed, all technological areas would benefit from this new
generation of detectors and novels applications such as
biomedical cameras, robotic eyes, small spectrometers, and
Unnamed Aerial Vehicles for civil and military surveillance…
References :
[1] Fendler, M., Dumas, D., Chemla, F., Cohen, M., Laporte, P., Tekaya, K., Primot, J., Le Coarer, E., Ribot, H. Hemispherical infrared focal
plane arrays: a new design parameter for the instruments. (2012) Proceedings of SPIE, High energy, optical, and infrared detectors for
astronomy, 8453, art.no 8453-60
[2] Dumas, D., Fendler, M., Berger, F., Cloix, B., Pornin, C., Baier, N., Druart, G., Primot, J., Le Coarer, E. An infrared camera based on curved
retina (2012) Optics Letters, Vol.37, No.4, pp.653-655
[3] Dumas, D., Fendler, M., Baier, N., Primot, J., Le Coarer, E. Curved focal plane detector array for wide field cameras (2012) Applied Optics,
51(22), pp.5419 - 5424
31
Complete THz system for reflection
real-time imaging with uncooled
antenna-coupled bolometer arrays
Research topics : High resolution Terahertz imaging components and system
F. Simoens, J. Meilhan, J. Lalanne-Dera, S. Gidon, G. Lasfargues, S.Martin, S. Pocas, J.-L. OuvrierBuffet, D. Guillaume (CEA INAC-SBT), V. Jagtap, S. Barbieri (Laboratoire MPQ, Université Paris 7)
ABSTRACT: CEA-Leti has developed a real-time reflection imaging system: it combines illumination by QCL
sources and a telescope that focuses the reflected radiation onto a camera that houses an uncooled 320x240
antenna-coupled micro-bolometer array driven by a FPGA card. High spatial resolution images of large
illuminated surfaces are acquired in real-time. Such demonstration performance is a premiere in the high
frequency range THz imaging field.
INTRODUCTION AND BACKGROUND
Practical terahertz (THz) systems require reduced acquisition
time that can be achieved by the use of focal-plane array
(FPA) cameras instead of single-point detectors. Antennacoupled micro-bolometer arrays developed by CEA-Leti fulfills
this demand, with prototyped 320x240 pixel sensors
optimized for detection in the 1.5-3.5 THz range [1].
Moreover, for most of the envisaged applications, real-time
reflection imaging is to be applied to analyze very absorbing
materials. This expectation has motivated the development
of a complete reflection active THz imaging demonstrator by
Leti in collaboration with other CEA divisions and French
universities.
I.
IMAGING TESTS RESULTS
An optical test pattern has been placed in front of a metallic
plane mirror. The THz video camera images the letters in
real-time with a better than 2 mm spatial resolution.
55mm
IMAGING DEMONSTRATOR DESCRIPTION
Figure 2 : Imaged test pattern for optical resolution estimation
This target has been covered by a nylon clean room coat:
tests have shown that still the imaging system sensitivity is
sufficient to image the reflected beam after propagation in
more than 1 m in air and two-way crossing through the
textile.
Figure 1 : CEA THz reflection active imaging demonstrator
The demonstrator combines a THz source, an optical system
and an uncooled bolometer camera.
The THz source consists in associated THz Quantum Cascade
Lasers (QCL) of the French Paris 7 University [2] integrated
in a CEA-INAC designed cryogenic pulse tube cooler. A
specific illumination optical setup guides the THz beam
towards the scene on a 40x60mm2 surface. Then the
reflected and backscattered radiation from the scene is
collected and focused by a Newton telescope onto the
sensitive surface of the 320x240 bolometer array embedded
in a dedicated vacuum sealed packaging. This focal plane
array (FPA) is driven by commercial FPGA and customized
front-end electronic cards housed in a camera box (Fig. 4).
The demonstrator is fully integrated in an autonomous and
transportable hermitically closed system. This housing allows
air drying of 3 of the whole 4 m beam optical path from the
source to the camera.
Figure 2 : Real time reflected image of the target hidden by nylon coat
CONCLUSION
Real-time imaging of a 40x60 mm² illuminated surface has
been demonstrated with a typical 2 mm resolution thanks to
the large surface of the antenna-coupled bolometer array and
to a telescope-like optical system. Forthcoming work will aim
at improving beam shaping and homogenization and at
demonstrating the capabilities of spectro-imaging on the
bolometer FPA with multiple-frequency illumination. These
encouraging achievements open the way to industrial transfer
for several applications where reflection imaging is required
to analyze opaque materials.
References :
[1] F. Simoens, J. Meilhan, B. Delplanque, S. Gidon, G. Lasfargues, J. Lalanne Dera, D.T. Nguyen, J.L. Ouvrier-Buffet, S. Pocas, T. Maillou, O.
Cathabard, S. Barbieri, “Real-time imaging with THz fully-customized uncooled amorphous-silicon microbolometer FPAs”, Proc. SPIE 8363,
83630D (2012)
[2] S. Barbieri., J. Alton, J. Fowler, H. E. Beere, E. H. Linfield, D. A. Ritchie, “2.9THz quantum cascade lasers operating up to 70K in continuous
wave”, Appl. Phys. Lett. 85, 1674 (2004)
32
33
3
Display components
Small pitch active matrix LCDs and
associated technologies
OLED small displays: electrical
modeling, active matrix, encapsulation
Quantum dot OLEDs
34
Development of 5 µm-Pixel Pitch
Active- Matrix for Transmissive
LCD Picoprojector
Research topics: LCD, Picoprojector, Microdisplay, Transfer on glass
F Templier, U Rossini, D Sarrasin, L Clerc, T Flahaut,
V Larrey and J Segura-Puchades (LETI/ DACLE)
ABSTRACT: Development of transparent active-matrix for picoprojector with 5 µm color-pixel pitch has been
shown. Transfer on glass of pre-processed SOI wafer, followed by silicon grinding and post-process on glass has
been achieved. This process seems to be very suitable for the fabrication of high-resolution picoprojectors with
transmissive LCDs.
Recently a great interest has been shown for picoprojectors
which can be used as stand-alone devices or ultimately
integrated in portable devices such as cameras or cellphones. To fabricate an easy-to-make and compact optical
system for picoprojector, transmissive LCD can be an
excellent solution. To achieve this, it is necessary to have a
transmissive, very small pixel-pitch (5 µm or less) activematrix. Silicon Integrated Circuit (IC) technologies provides
very small pitch such as 5 µm, however silicon is not
transmissive. Here we present [1] the design and
development of very-small pitch (5 µm, color-pixel)
transmissive
active-matrix
suited
for
LCD-based
picoprojector, made from SOI circuit transferred on glass
plate.
The original process is the following (Figure 1). First part of
the process consists of processing 200 mm SOI wafers with
conventional IC technology including 6-level copper
interconnect. Then wafers are transferred on standard
borosilicate 200 mm glass wafer and backside silicon is
removed by grinding. Subsequently, post-process is
performed to define vias, light-shield, both made using
standard dual-damascene process. Finally, the pixel
electrode is made by deposition of ITO and patterning with
plasma etching. All this post-process is made on standard
silicon-line equipment.
in each field during the insulation step all over the wafer to
compensate. After correction, pattern positions are measured
and we could find that compensation is fully efficient, which is
a major result of the development.
Use of copper interconnect provides very narrow line and
columns thanks to the high conductivity, which contributes to
the good aperture ratio combined with the very small pixel
pitch. Vias are opened using plasma etching. Vias filling and
metallization patterning are made using conventional
damascene process. ITO electrode is then deposited by PVD,
which is followed by annealing. Plasma etching of ITO pixel
electrodes leads to tapered edges, which helps good covering
of subsequent liquid crystal alignment layer, combined with
very good insulation despite the small spacing between
pixels. Quality of insulation could be assessed by electrical
measurements, which has been done using test patterns
which consists in ITO lines separated by a given distance
ranging from 0.25 to 1 µm. We could evidence that for a
spacing as small as 0.5 µm, defect density is around 0.05 m1
. Considering the actual pixel layout, this corresponds to less
than 0.5 defects per display, which is encouraging for a first
evaluation in development step.
Figure 2: Processed glass-wafer (left) and processed active-matrix
made from SOI wafer transferred on glass and with post-process
(right).
Figure 1: Step-by-step original process for transmissive IC active
matrix from SOI
After transfer of SOI wafer on glass, distortion occurs due
to the mismatch of mechanical properties (e. g. Young’s
modulus) between Si and Glass. This distortion could be
measured using the photolithography tool (ASM300
stepper). After measurement (we found a radial distortion),
the stepper computes the required correction and applies it
Design and fabrication of transparent active-matrix for
picoprojector with 5 µm color-pixel pitch and 70 % of open
aperture has been shown. Transfer on glass of pre-processed
SOI wafer, followed by silicon grinding and post-process on
glass consisting of 3 photolithography levels has been
achieved. Distortion after transfer on glass has been
measured and we show that it can be compensated. This
process seems very suitable for the fabrication of highresolution picoprojectors using transmissive, SOI-based LCD
[1].
References:
[1] F.Templier, U. Rossini, D. Sarrasin, L. Clerc, T. Flahaut, J. Segura-Puchades, V. Larrey, H. Wehbe-Alause, M. Marty, “Development of 5 µmpixel pitch active-matrix for transmissive LCD picoprojector”, The 19th International Display Workshops in conjunction with Asia Display 2012
(IDW/AD’12), December 4-7 2012, Kyoto, Japan (2012)
35
Transferring silicon on glass
Research topics: transfer, glass wafer, SOI, distortion
U.ROSSINI and T.FLAHAUT (LETI/ DTSI)
ABSTRACT: We observed a negative expansion coefficient when transferring Silicon on glass wafer. Such an
expansion is not afforded by the stepper correction capability, so we tried to understand the mechanism of
expansion and a way to correct it.
In the project aimed at achieving a transmissive LCD micro
display (see “Development of 5 µm-Pixel Pitch Active- Matrix
for Transmissive LCD Picoprojector” in this report), one
process step is the transfer of silicon backplane from an SOI
wafer on a glass wafer. The post process then consists in
making the electrical interconnects for each pixel through the
bulk oxide until the first metallic layer (metal 1). After
transfer of SOI wafer on glass, distortion occurs due to the
mismatch of mechanical properties (e. g. Young’s modulus)
between Si and Glass. To get the maximum open aperture
ratio for the pixel, we tried to avoid the oversizing of via and
then we tried to minimize distortion effects during transfer.
To evaluate the distortion we used mainly data coming from
the photolithography tool (ASM300 stepper). We measured
mainly a negative extension phenomenon, which couldn’t be
explained yet by temperature consideration as the transfer
step was done at room temperature and as slippage is not
possible after transfer during curing step.
We suggested that Young modulus difference between glass
and silicon was responsible of the distortion phenomenon, as
explained in the next section.
When bonding two wafers, the distortion ε inside each wafer
increases with decreasing the distance to the wafers
interface:
e=0
dz
h
Wafers Interface
previous experiment, it can be calculated from the previous
formula that 30mJ/m² was necessary. This value is
consistent with the 70 to 100mJ/m² released during the
transfer. The remaining energy is used to remove the air
between the wafers, controlling then the propagation speed
of the bonding front as explained in [1].
Furthermore, if we calculate this expansion when the wafers
have different thicknesses hi, we may find the following link
between the overall distortion  and the sticking energy :
3(hglassEglass  hSi ESi ) 2
(hglassEglass  hSi ESi ) * hglassEglass * hSi ESi
This means that
hi Ei =0
*   2
will give no expansion. As, for
example, silicon has a Young modulus twice the one of glass,
the silicon wafer must be half glass wafer thickness.
Controlling the expansion is important because this amount is
found to be the limit capability of the stepper in expansion
correction when doing the post process photolithography on
the glass wafers.
The last formula for the overall distortion has been
experimentally
verified
by
Vincent
LARREY
from
LETI/DTSI/SSURF department and reported in the document
[2] and following graph (Fig.1). Two values for the Young
modulus have been considered for both crystalline axes of
Silicon (1.1.0 and 1.0.0) and a sticking energy of 75mJ/m²
has been measured.
e(z)
e=ei
T
We suppose a transversal component T of the Van der Waals
forces occurring at the wafers interfaces during the bonding
of both wafers. For a slice wafer of surface S and infinitesimal
thickness dz, the mechanical energy J needed to induce the
distortion ε(z)i can be written as:
1
.J i  * S * z * Ei *  ( z )i2
2
where Ei represents the young modulus for glass and Silicon,
S is the wafer area, dz is the infinitesimal thickness.
Then integrating the previous elastic energy on the whole
thickness of the wafer and assuming that the tension T is the
same for both wafers, the total elastic energy stored in both
wafers during the bonding process can be written as:
Theory curves with 75mJ/m²
Theory axe 1.1.0
0
Theory axe 1.0.0
-5
)
m-10
p
(p-15
n
o
is
-20
n
a
p
xE-25
-30
Experimal values
-35
300
400
500
600
700
Silicon wafer Thickness µm
800
Eglass * ESi
1
J total  Sh *
( Eglass  ESi ) *  2
2
3
( Eglass  ESi )
Figure 1: Experimental measurement in blue versus theory (red and
green curves)
where h is the thickness of each wafers and ε the measured
expansion after transfer on glass.
In order to generate the -20ppm expansion measured in the
We can see effectively that the negative expansion is nearly
zero when the silicon wafer thickness tends to 350µm which
is half thickness of glass wafer (700µm).
References
[1] F. Rieutord, B Bataillou, H Moriceau, Dynamics of a bonding front, PRL 94, 236101 (2005)
[2] Berthod Loic, Vincent Larrey .Collage des hétéro-structures et des déformations de nappes (2012)
36
Enhanced Cholesteric
Liquid crystal phase modulator
Research topics: liquid crystal, cholesteric, birefringence, phase modulation
U.ROSSINI
ABSTRACT: Using the Poincare sphere tool, we describe a method to avoid residual birefringence when using
Cholesteric liquid crystal in phase modulator devices. This simple method consists of determining the correct
alignment direction for the liquid crystal for each internal side of the device.
Phase modulation is often made by using nematic liquid
crystal with homogeneous alignment and polarized light in
the direction of the extraordinary index of the liquid crystal.
However many applications in optics need to work with nonpolarized light and then two devices with orthogonal
orientations must used to act on natural light. A well-known
arrangement with only one nematic liquid crystal device uses
a quarter wavelength plate and a mirror but works only if
transparency of the device is not required.
Cholesteric liquid crystal has been proposed for phase
modulation on non-polarized light. This implies that the liquid
crystal must act in the same manner whatever the light
polarization. We can imagine that a polarized light will cross
the liquid crystal in each direction (extraordinary and
ordinary) if the Cholesteric pitch is short enough regarding
the device thickness (d). The average optical index will be
equal to no+1/2Δn independently of the polarization
direction. If the liquid crystal is activated in homeotropic
direction (perpendicularly to the surface) using voltage, the
optical index will be no.
Using the Poincare sphere (well-known method to represent
light polarization, circular polarized light at the pole until
linear polarized light on equator) we can observe that a
polarized light through the device becomes slightly elliptic
depending of polarization, effectively, the optical effect due
to Cholesteric liquid crystal may be summarized as the action
of a elliptical bi-refringent layer (represented by K vector on
the following graph). Then, the effect on a linear polarized
light, is like a rotation on the Poincare sphere around this k
axe when going through the liquid crystal.
The highest the twist (k*2), more the birefringent axe (k)
will be in vertical position and less will be the residual
ellipticity. However, by giving a particular twist (θ) to the
liquid crystal, we can avoid what is called the residual
birefringence in devices using Cholesteric liquid crystal.
Here Amp is the rotation around the elliptical birefringent
axe. If Amp is a multiple of 2, (N*2), the light will turn
back to initial position on the Poincare sphere and no
ellipticity will be induced, whatever the polarization entrance
position. Then the twist has to be:
 
N2
d 2n2
2
 2 k  
Experimentally, the specified value for the twist of the
Cholesteric liquid crystal can be obtained by buffing the
polyimide in the good direction. As an example, using
MLC2062 from Merck and adequate Cholesteric proportion in
device with thickness 5.1µm, the correction ε is computed to
be 26°, 36°, 62° for respectively k= 7, 5, 3. The correction
needed, will tend to be zero when the number of gyrations
increases to infinity (the k axe coincides with Y axe on
Poincare sphere).
2
k
1
Ellipticité induite
Elliptical
biréfringent
No correction
Angle
non adapté
with correction
Angle
adapté e
0
-1
-2
-3
-4
10
Polarisation
Entrance
Liquid crystal
Figure 1: Action of elliptical birefringent layer
30
50
70
90
Azimut
entrée
ENTRANCE
POLARISATION
(°)
Figure 2: DIMOS computing comparing ellipticity obtained with and
without correction
Using DIMOS (dedicated Liquid crystal software) we can
verify that applying the computed correction, the induced
ellipticity is effectively zero regardless the entrance
polarization.
U ROSSINI and L CLERC, METHOD OF PRODUCING A PHASE DEVICE BASED ON A TWISTED LIQUID CRYSTAL HAVING OPTIMIZED STRUCTURE
OPERATING UNDER UNPOLARIZED LIGHT, Patent WO 2013/023869
37
Electrical Modeling of a Full Stack
Fluorescent Dual Emitter White PIN OLED
Research topics: OLED, Electrical Modeling, Fit simulation
K. Bouzid, H. Kanaan, H. Doyeux
ABSTRACT:
Electrical modeling of a full stack fluorescent dual emitter (blue + yellow) emitter white PIN OLED is performed
using a commercial software and a set of material parameters obtained by extraction from single carrier devices
JV curves. Simulation outputs show JV characteristics matching experimental JV curves and a consistent
behavior.
It is widely recognized that the development and optimization
of OLED stacks could be speeded up by using electrical and
optical modeling of their behavior. Such numerical tools could
give, when mastered, preliminary results concerning
innovative structures or new materials or even help speeding
up stack optimization procedures. As for any simulation tool,
it is of prime importance for data input values to be as close
as possible to reality in order to obtain a quality model.
In the present work, an electrical model of a full OLED stack
(Figure 1) has been created using experimentally extracted
parameters as input. The full modeled stack is a top emitting
OLED made of following layers: Al / p-doped layer (29nm) /
Yellow-doped NPB (1%) (5nm) / Blue-doped SMB013 (5%)
(27nm) / Alq3 (5nm) / N-doped layer (22 nm) / Ag.
Then, simulations based on this model have been run in
order to match experimental JV curve with the numerical
output. The goal was to validate the fit procedure used for
parameter extraction as well as the full stack model behavior
itself.
Moreover, a particular importance is brought to the behavior
of the simulated OLED. To investigate this, electric field,
charge carriers and excitons profiles have been studied and
all these points were found out to be closely reflective of the
results expected from state of the art knowledge.
For instance, carriers profile (Figure 3) shows an
accumulation of electrons and holes at the SMB013/NPB
interface which seems coherent with the big gap between the
two material LUMOs and HOMOs but also with the differences
in carrier mobilities. It also shows that each carrier type
concentration is decaying starting from its injection electrode
in accordance with material transport properties, revealing a
pseudo-symmetry at the NPB/SMB013 interface that seems
to be the node of the structure, as intended by design.
Figure 1: Fluorescent Dual Emitter White PIN OLED: 0.44 cm²
rectangular diodes, built on 200 mm silicon wafers
One can observe in Figure 2 the superimposed simulated JV
curve matching the experimental one.
Figure 3: Free electron (full line) and free hole (dashed line)
distribution profiles at 5V
With this study ([1] and [2]), we demonstrated that a
modeling of a full stack fluorescent dual emitter OLED based
on parameters extracted from experimental devices was
possible. To our knowledge, this work is the first successful
attempt. Creating a model such strongly based on
parameters
extracted
from
experimental
device
characteristics will enable to build up a predictive model in
the future, allowing further performance enhancement.
Figure 2: Experimental (dots) and simulated (full line) current density
versus voltage curves.
References :
[1] K. Bouzid, H. Kanaan and H. Doyeux, “Determination of electrical properties of OLED materials through combined
measurements and simulation of single carrier devices”, ICOE 2012 conference proceedings.
[2] K. Bouzid, H. Doyeux, H. Kanaan, “Electrical Modeling of a Full Stack Fluorescent Dual Emitter White PIN OLED”, IDW
2012 conference proceedings
38
Development of InGaZnO ThinFilm Transistors for activematrix OLED displays
Research topics: Thin-film transistors, metal-oxide, OLED displays, active-matrix, XPS
F Templier, Thuy Ngyuen, B Aventurier,
G Rodriguez*, J-P Barnes*, O Renaud* (*LETI/ DTSI)
ABSTRACT: We have developed bottom-gate IGZO thin-film transistors for active-matrix OLED displays. We
investigated the relation between electrical properties and material composition, as a function of process
conditions. The fabricated devices are state-of-the-art, except for threshold voltage, which has to be improved.
Indium Gallium Zinc Oxide Thin-Film Transistors (IGZO TFTs)
have shown interesting properties for application in activematrix displays, such as a high mobility and a good threshold
voltage (Vth) stability [1].
LETI has started development of such devices, with very
promising first results [2, 3]. Bottom-gate IGZO TFTs have
been fabricated, and process conditions have been varied to
include different annealing conditions (Oxygen or Nitrogen
ambient). Their performances have been investigated by
electrical characterizations. It has been found that
characteristics strongly depend on annealing conditions. TFTs
with Oxygen annealing exhibit standard TFT characteristics.
In the meantime, TFTs without a post-annealing or with
Nitrogen annealing exhibited poor characteristics, more
particularly it looked as if channel could not be depleted in
the reverse mode, IGZO layer behaving almost as a
conductive layer. To understand the origins of this
phenomenon, IGZO films from these devices have been
analyzed by X-Ray Photoelectron Spectroscopy (XPS). We
have focused on the O1s peak (Figure 1), from which, after
deconvolution, we could extract the respective contributions
of oxygen contamination (peak O3 at 533.6 eV), oxygen
vacancies (O2 at 532.7 eV) and oxygen-metal binding (O1 at
531eV). It can be seen that IGZO layers after annealing in N2
have higher concentration of oxygen vacancies. This is
consistent with our electrical results since it is assumed that
conduction in IGZO films is the result of oxygen vacancies.
We have also investigated the effect of aging TFTs under air,
with or without passivation. Electrical and XPS analysis where
performed. Our work shows a relation between electrical
properties and chemical composition as characterized by XPS.
A first batch of IGZO TFTs has been realized and
characterized. Figure 2 shows typical transconductance
characteristics of such TFT.
Figure 2: Typical transfer characteristics of bottom-gate IGZO thinfilm transistor, at Vd = 0.1 and Vd = 10 Volts.
Device parameters have been extracted from the transfer
characteristics and are summarized in table 1.
Parameter
Ion/Ioff
Vth (V)
µ (cm2/Vs)
S (V/dec)
Ioff (at Vg = -20V)
Vd = 0.1 V Vd = 10 V
1,4. 108
-10,6
5,8
0,201
1.10-12
4,4. 109
-10,6
2,61
0,21
5. E-12
Table 1: Summary of IGZO TFT parameters extracted from transfer
characteristics, in linear (left) and saturation (right) regimes.
Conduction properties are very interesting, both regarding
mobility, on-current and subthreshold slope. In the meantime, leakage current is very low, giving outstanding ON/OFF
ratio. On the other hand, threshold voltage is well below
zero, and this has to be improved.
Overall, these first results are very encouraging. Our work
contributes to finding a high-performance and reliable TFT
technology for application in the domain of active-matrix
displays.
Figure 1: XPS O1s spectra of IGZO films after annealing under N2
(left) and in O2 (right)
References :
[1] T. Kamiya, K. Nomura, et H. Hosono, Science and Technology of Advanced Materials, vol. 11, no 4, p. 044305, 2010.
[2] T. Nguyen, B. Aventurier, G. Rodriguez, J.-P. Barnes, F. Templier “Characterization of transparent Indium Gallium Zinc Oxide
semiconductor for application in thin-film transistors” Workshop nanoTransparent Conductive Materials (nanoTCM), 14-15 June 2012,
Grenoble, France (2012)
[3] T. Nguyen, B. Aventurier, G. Rodriguez, J.-P. Barnes, O. Renault, F. Templier “Analysis of IGZO films for the development of active-matrix
OLED displays” 5th International Symposium on Flexible Organic Electronics (ISFOE12) 2-5 July 2012, Thessaloniki, Greece (2012)
39
Stability of 8-hydroxyquinoline
aluminum films encapsulated by a
single Al2O3 barrier deposited by
low temperature atomic layer
deposition
Research topics: Thin film encapsulation, OLED, organic electronics
T. Maindron, J.-Y. Simon, E. Viasnoff, D. Lafond
ABSTRACT: 100 nm thick AlQ3 films deposited onto silicon wafers have been encapsulated by mean of low
temperature atomic layer deposition of Al2O3 (20 nm). Investigation of the film evolution under storage test as
harsh as 65 °C/85% RH has been investigated up to ~ 1000 h and no severe degradation could be noticed. The
results have been compared to raw AlQ3 films which deteriorate far faster in the same conditions. For that
purpose, fluorescence measurements have been used to monitor the film evolution.
The approach of this work has been motivated by the
research for an encapsulation for OLED which was not based
on usual polymeric decoupling layers but rather on molecular
decoupling layers. Few examples of such molecular
decoupling layers exist. Lee S.-N. et al. have described the
use
of
the
molecule
2-methyl-9,10-di(2-naphthyl)anthracene (MADN) as a decoupling layer between sputtered
SiN and SiON layers [1]. The authors explained that the
integration of a MADN layer in the encapsulation stack can
planarize the surface of the (n−1)th inorganic barrier layer
before the deposition of the subsequent nth layer thus leading
to lower pinhole density in the nth layer. The whole WVTR is
thus reduced. Liu et al. have integrated AlQ3 films between
evaporated LiF layers for the encapsulation of top-emitting
OLED [2]. One of the main advantages of such molecular
layers is that they are easily vacuum-deposited so that the
encapsulation process may potentially be fully integrated in
the OLED fabrication line without any additional tool required
for polymer deposition. The deposition of Al2O3 films from
ALD onto amorphous molecular films has been found
however not to be an easy task. In this work, the AlQ3
molecule has been found to be a material of choice for that
purpose because it shows a high stability to the low
temperature ALD of Al(CH3)3/H2O process (Fig. 1).
Figure 1: (left) the AlQ3 molecule; (right) description of the
test develop in this work; AlQ3 thickness is 100 nm; Al2O3 is
20 nm
Two
other
molecular
films
of
4,7-diphenyl-1,10phenanthroline (BPhen) and 2,2 ′ ,7,7 ′ -tetra(N, N-ditolyl)amino-spiro-bifluorene (Spiro-TTB) have been also
tested: they clearly react during the deposition process and
crystallization of the organic layers appears clearly upon
oxide deposition (table 1).
Table 1: Description of samples tested in this work
BPhen and Spiro-TTB molecules have two different Tg, the
first one being lower than the deposition temperature of the
oxide, the second one being higher. Tg of the molecular films
does not seem to be a criterion of choice to expect a nice ALD
deposition of Al2O3 onto organic molecular layers. Chemical
reactions with chemical reagents in the ALD process are
rather suspected. With AlQ3, it has been possible to compare
the stability of the single AlQ3 films with Al2O3 encapsulated
AlQ3 films versus time (up to ~ 1000 h) upon storage in a
climatic environment as harsh as 65 °C/ 85% RH (Fig. 2).
Figure 2: Evolution of Si/AlQ3/Al2O3 (top) and Si/AlQ3 (bottom)
fluorescence intensities and fluorescence peak position at maximum of
fluorescence versus storage time at 65 °C/85% RH
Storage of Si/AlQ3 and Si/AlQ3/Al2O3 films has been
performed at 65 °C/85% RH. The results have been depicted
in Fig. 2 for Si/AlQ3/Al2O3 where the FL peak position at
maximum FL intensity of AlQ3 and the maximum FL intensity
have been plotted versus storage time. The raw AlQ3 film
does not withstand the severe climatic conditions: the films
crystallized in less than 15 h leading to an obvious FL
quenching as well as to a FL blue shift from 532 nm to 496
nm. On the contrary, the Si/AlQ3/Al2O3 films showed an
incredible stability up to 889 h. As a result, the FL peak at
532 nm remains unchanged and the FL intensity decreases
by ~ 20 % with a linear behavior. No crystallization could be
observed at the end.
In summary, it has been shown that low temperature ALD of
Al2O3 films could be implemented onto organic molecular
films like AlQ3 showing very good barrier properties.
However, the deposition of Al2O3 films onto organic molecular
films is highly dependent on the intrinsic nature of the
molecule to be encapsulated rather than its Tg [3].
References
[1] S.-N. Lee, S.-W. Hwang, C.H. Chen, Jpn. J. Appl. Phys. 46 (2007) 7432
[2] S. Liu, D. Zhang, Y. Li, L. Duan, G. Dong, L. Wang, Y. Qiu, Chin. Sci. Bull. 53 (2008) 958
[3] T. Maindron et al., Thin Solid Films, Volume 520, Issue 23, Pages 6876 (30 September 2012)
40
Langmuir–Schaeffer monolayers of
colloidal nanocrystals for cost
efficient quantum dot LED
Research topics: Quantum dot, Langmuir, Microprinting, QDLED
S Le Calvez, H Bourvon, H Kanaan, S Meunier-Della-Gatta,
C Philippot* and P Reiss* (*CEA/ DSM/INAC)
ABSTRACT : Quantum dot (QD) LEDs of high color purity, low turn-on voltage and low leakage current are
demonstrated using a solvent free method. First, a monolayer of QDs is formed at the air/water interface, which
is then transferred with a PDMS stamp onto the device. The method is applicable to large substrates and
reduces materials consumption as compared to other deposition techniques.
Colloidal semiconductor nanocrystals or quantum dots (QDs)
have the appealing property to emit a tunable wavelength
determined by their chemical composition and size. These
characteristics make them attractive for applications in
general lighting and flat-panel displays. Solution-processed
quantum dot light emitting diodes (QDLEDs) open the way
for low-cost fabrication of color saturated displays. The
electroluminescence spectra of QDLEDs exhibit a narrow
bandwidth (full width at half maximum 30 nm) and QDbased displays offer high color purity and saturation. Despite
these
unique
characteristics
and
recent
research
breakthroughs, some drawbacks need to be overcome to
achieve mass production of QDLEDs and QD displays. At
present, cost reduction and improved homogeneity of the QD
layers are among the main challenges. Because of their high
molecular weight, QDs cannot be evaporated like small
molecules used in organic light emitting diodes (OLEDs).
Prepared in colloidal solution, they are commonly spin-coated
onto devices with the goal to achieve one or a few
monolayers [1], the low thickness being required for efficient
charge injection into the emissive QD layer. The spin-coating
process involves a large loss of material and the used solvent
must not dissolve underlying layers, which limits its choice.
Also, in many cases poor homogeneity of the deposited
layers is observed, which has a strong impact on device color
purity and efficiency.
Here we present a new cost-efficient technique overcoming
all these issues. We implemented a new method combining
both stamping and the Langmuir–Schaeffer technique [2].
The main steps of our approach are presented in Figure 1.
First, in order to form a QD monolayer, a Langmuir trough in
hydrophobic Teflon is used. A known quantity of QDs is
dispersed in chloroform and spread on the surface of
deionized water. After the evaporation of chloroform, the
barriers are closed to compress the nanocrystals and form a
compact film. A PDMS stamp is then approached horizontally
to the surface and gently pulled out of the water. After
drying, the QDs are ready to be stamped on devices by
regular stamping.
Figure 1: Step-by-step description of the developed QD deposition
method. QDs are compressed and deposited on a PDMS stamp using
the Langmuir–Schaeffer technique. Subsequently the QDs are
transferred on organic materials used in a OLED stack. (HIL: hole
injection layer, HTL: hole transport layer, ETL: electron transport
layer, EIL: electron injection layer).
Finally, we implemented the described QD deposition
technique to develop QDLEDs. On an ITO-coated glass
substrate,
a
p-doped
layer
of
2,2’7,7-tetra(n,n-ditolyl)amino-spiro-bifluorene
(spiro-TTB)
doped
with
tetrafluorotetracyano-p-quinodimethane (F4TCNQ) has been
deposited. Next, a HTL of N,N-diphenyl- N,N’-bis(1naphthyl)-(1,1-biphenyl)-4,4-diamine(NPB) was evaporated.
The QDs were transferred onto this layer by using the
method described above. Finally, a layer of 2,9-dimethyl-4,7diphenyl-1,10-phenanthroline (BCP), an n-doped layer, and
silver were thermally evaporated. Optical characterizations of
the obtained diodes are given in Figure 2. The energy levels
of the materials used in our device are represented in Figure
2c. Thanks to our process deposition the QDs are deposited
as a close-packed film. The low value of the QDLED turn-on
voltage of 2.4 V shows that charge balance in the device is
equilibrated.
Figure 2: a) Photograph of the obtained QDLED under operation (bias
voltage: 4 V). b) Photoluminescence spectrum (blue) in hexane using
an excitation wavelength of 405nm and electroluminescence spectrum
(red) showing high color purity. Maximum intensities are measured for
620 nm (FWHM 25 nm) and 628 nm (FWMH 35 nm), respectively. c)
Energy level scheme of the materials used in the QDLED.
Moreover, the observed leakage current densities of 105 mA/
cm2 are similar to those obtained with diodes entirely
composed of small molecules. The maximum luminance of
170 Cd/m2 is reached at 6 V and the maximum EQE accounts
for 0.11% at 2.4 V.
Using this method, it is possible to obtain compact and
homogeneous QD monolayers on large substrates, which is a
key point for developing light-emitting devices of high color
purity. We have implemented this technique to realize a
QDLED showing narrow emission, low turn-on voltage, and
low leakage current. In view of its advantages, it is very
likely that the described technique will contribute to the
development of more efficient devices in the future, notably
through the optimization of the device architecture.
Combining cost-efficiency and simplicity, it also paves the
way for QDs to be used in displays for mass-production.
References :
H. Bourvon, S. Le Calvez, D. Vaufrey, S. Meunier Della Gatta, Mater. Res. Soc. Symp. Proc. 2011, 1286
H. Bourvon, S. Le Calvez, H.Kanaan, S. Meunier-Della-Gatta,C. Philippot , P.r Reiss, Adv. Mater. 2012, 24, 4414–4418
41
4
Optical environmental
sensors
Micro-hotplate IR emitters
Non-dispersive IR sensors
Integrated photoacoustic sensors
42
Air Convection on
a Micro Hotplate for Gas Sensor
Research topics : Gas Sensor modelization, Filament, Electro-Thermal convection,
S. Gidon, M. Brun, S. Nicoletti and P. Barritault
ABSTRACT: monitoring of indoor CO2 concentration is of particular interest to detect room occupancy in order to
optimise power consumptions of building. Key feature for a wider use of the sensing technic involved the
management of the power consumption that is related to the temperature uniformity of the micro hotplate. We
improve our electro thermal heater model, tacking in account the convective thermal effect that can be
described using the numerical simulation in stationary regime. In such a way we can propose an explicit thermal
flux law.
One approach to monitor the indoor CO2 concentration is to
use optical detection using specific absorption lines of CO2
molecules in the infrared domain close to 4.2 µm. Such
optical sensors include a detector, typically a microbolometer, an IR source – such as a hotplate– and a filter to
select the interesting band in the black body spectrum of the
emitter. All these components are made in well known planar
Si technology using MEMS approach (Si3N4/SiO2 as
supporting layer and TiN/Pt/TiN for and heater layer).
In the past, we have optimised the filament geometry
(conductive
track
width)
to
minimize
the
sensor
consumption, using a Comsol™ electro-thermal model [1],
figure 1. Though, to describe conductive and convective
thermal exchanges in air out of the micro-hotplate, we use
the proposed law by [2].
proposed in the thermal ComsolTM module. We choose the
boundary conditions using axial symmetry, wall on the
hotplate and outlet flux elsewhere. The surface of the
hotplate is fixed at 650°C everywhere.
The model provides the temperature and flow velocity in air
related to its density gradient, figure 3. Reynolds number
appeared sufficiently low (lower than 100) to argue in favor
of the used laminar model.
Figure 3. Air temperature around a 150µm diameter hotplate (left)
and air flow around (right).
Figure 1. Filament temperature profile obtains in a first 3D electro
thermal model (z dimension magnified by a factor 50, for better
accuracy).
We deduce the thermal convection flux density on the
hotplate surface, figure 4. We notice that the thermal flow is
slightly greater on the upper surface, but can be one decade
larger at the border of the disc. This phenomenon explains
the temperature non-uniformity observed experimentally on
devices already designed with a constant thermal flux value.
However, we experimentally observed that the temperature
uniformity was lower than expected [3], as reported figure 2.
Figure 2. Microscopy images of the realized heated filament, SEM view
(left) and IR radiation (right), showing the temperature distribution on
the hotplate.
That is why we investigate thermal dissipation related to
laminar air convection in stationary regime. We use the air
parameters using the laminar flow coupled equations
Figure 4. Flux density profile on the top hotplate surface (red),
compared to the initially used constant value (purple).
Finally, we conclude that the sensor consumption can be still
lowered [4].
References :
1. S. Gidon, and S. Nicoletti “Optimisation of Filament Geometry for Gas Sensor Application”, Comsol Conf 2010.
2. Pierre Barritault, Mickael Brun, Serge Gidon, Sergio Nicoletti, “Mid-IR source based on a free-standing microhotplate for autonomous CO2
sensing in indoor applications”, Sensors and Actuators A 172, 379– 385, 2011.
3. S. Gidon, M. Brun, S. Nicoletti and P. Barritault “Air Convection on a Micro Hotplate for Gas Sensor”, Comsol Conf 2012.
43
Low Power consumption CO2 sensor
for domotics applications
Research topics: CO2 detection, Optical Sensing, NDIR
P.Barritault, M.Brun, S.Nicoletti
ABSTRACT: We report the fabrication and the characterization of a NDIR sensor based on a micro-bolometer
detector and a MEMS IR-source. The accuracy and repeatability of the sensor was studied for CO2 and
hydrocarbons detection. The results show that the performances obtained are totally compatible with indoor air
quality requirements. Moreover the sensor has a very low power consumption (in the order of 1mJ per
measurement) which makes it suitable for autonomous applications. Finally, we emphasize the fact that this
sensor, because it uses a bolometer array, can be converted in a multi-gas sensor.
Non-dispersive infrared (NDIR) gas sensors are used in
various applications, for instance, monitoring of air quality in
office buildings, which represent a big market for low cost
sensors. NDIR gas sensors typically consist of an IR-source,
an optical path containing the gas sample, and an IRdetector in combination with a gas-specific optical filter.
Commercial sensors generally use thermopile or pyroelectric
detectors with a high power thermal source. As a
consequence the power consumption is well above the
maximum energy tolerable for autonomous devices. We
report here a MEMS-based, low-power-consumption gas
sensor that can be extended to multi gas detection. It
combines, a free standing micro-hotplate which behaves as a
black-body providing a large IR emission spectrum, a
commercial optical filter suitable for CO2 detection, and an
array of micro-bolometer for the signal detection. The use of
an array of bolometers combined with an array of specific
filters opens the way to multi-gas sensing.
has been varied from 0 to 3000 ppm.
Figure 2: Average signal output of the NDIR CO 2 sensor. Each point
corresponds to an averaged measured value and the error bars
correspond to associated standard deviations.
The repeatability of measurement is better than 1 mV with
negligible hysteresis. From the curve slope we can determine
the sensitivity: a variation of 1 mV on the detector signal
corresponds to a variation of 30 ppm at 1000 ppm of CO2.
Based on these data we calculated the sensor sensitivity in
V/ppm and the noise the ppm equivalent of the standard
deviation as a function of the CO2 concentration.. The results
are reported in Fig. 3.
0,00
35
-0,02
-0,03
Sensitivity
Noise
-0,04
-0,05
Figure 1: Experimental setup used for NDIR sensing
Figure 1 shows a lab version of the sensor prototype with the
assembly used for the measurements [1]. During the test the
IR source is driven at a power of 45 mW which is a good
trade-off between luminance and life time [2]. Taking in to
account that the hotplate source can be switched on and off
in less than 1 ms, the time constant of the device is settled
by the bolometer’s arrays. In good agreement with the
simulated data indicates, the response time of the bolometer
is 10 ms. The choice of pulses durations of 30 ms is a good
compromise between power consumption and signal. In this
case, the energy consumed by the IR-source for each
measurement is 1.35 mJ, only.
Figure 2 shows the average signal output of the sensor as a
function. During the measurements the concentration of CO2
25
Noise [ppm]
Sensitivity (µV/ppm)
45
-0,01
15
500
1000
1500
2000
2500
3000
[CO2]
Figure 3: Sensor sensitivity and the noise equivalent concetrattion of
the device.
To conclude, we have experimentally demonstrated the
feasibility of a CO2 NDIR sensor with accuracy and
repeatability equivalent to those commercial devices. These
performances are reached with very low power consumption:
1.35 mJ per measurement for the IR-source a resolution of
30 ppm at 1000 ppm, as required by ASRAE specifications.
By simply changing the wavelength, this sensor can be used
to detect other gas such as for example hydrocarbons.
Acknowledgments: This work was carried out in collaboration
with Schneider Electric in the frame of “HOMES” project
founded by OSEO
References :
1.
S. Fanget, H Grange, F. Palancade, G. Ganuchaud, M Matheron, S. Charlot, T Bordy, T Hoang, P. Rey, D. Mercier, P. Brunet-Manquat , P.
Robert, CO2 MEASUREMENT USING AN ALN/SI SAW SENSOR, Proc. Transducers T3P.034 (2011) 1136-1139.
2.
P. Barritault, M. Brun, S. Gidon, S. Nicoletti, Mid-IR source based on a free-standing microhotplate for autonomous CO2 sensing in indoor
applications, Sens. Actuators A 172 (2011) 379– 385.
44
Low power CO2 prototype at CES’13
Research topics: CO2 sensors, NDIR, Air quality monitoring
L.ANDRE, P.BARRITAULT, O.LARTIGUE, F.LAULAGNET
ABSTRACT: Nowadays industrial CO2 sensors are individually assembled. In addition, they require either AA
type batteries or cable energy alimentation. Hence, these sensors are not only expensive, but also energy
consuming and cumbersome. In the frame of the GREENET project with STMicroelectronics and supported by
French Research Agency ANR, we developed a prototype of sensor that was presented at CES (Consumer
Electronic Show) 2013 in Las Vegas. Performances are compliant with ASHRAE norms. Power consumption is
lower than today’s solutions. Last we developed a waferlevel source, paving the way to industrial low cost,
ultra-low power and small footprint solutions.
Energy harvesting in modern buildings as well as air quality
monitoring are two emerging needs [1] requiring low cost
gas sensors. In particular, CO2 sensors market is expected to
ramp up from 10 to 200M$ in the decade. In order to grab
this market, solutions need to meet the following:

Compliance with ASHRAE norms. It should be able
to read concentrations within the range 1003000ppm, with a precision of +-50ppm at
1000ppm.In particular no measurement drift shall
occur.

Low consumption is a strong advantage as no
cables, batteries nor maintenance are required

Small Footprint. In particular one shall concentrate
on solution where no AA batteries are necessary

Not expensive
Most solutions for CO2 sensors use NDIR (Non Dispersive
Infra-Red) detection (Fig.1): A source produces Infra-red
light which is collimated and filtered so that the transmitted
light corresponds to the absorption band of the gas under
analysis. For instance, in case of CO2, the light will be filtered
and absorbed at 4.26µm. Then, light is focused on a detector
such as a thermopile, a pyrometer or a bolometer. The
detector transforms the remaining optical power into electric
signal.
and producing Infra-Red photons. As a drawback, the air
surrounding the membrane is heated up as well, which is a
waste of energy.
In this project, we started with a packaging of the source:
the source is aligned and wire-bonded on a TO39 box. Then
the box is brazed under vacuum (10-3mBar). The source is
aligned and assembled on the focal plane of a large numerical
aperture geltech lens. The collimated beam is filtered using a
commercial optical filter and refocused on a thermopile
detector. A picture of the demonstrator is shown on Fig.2.The
source is supplied with 1.2V and 1.6mA during 40ms and the
thermopile is supplied with 2.8V and 1mA during 65ms. We
have tested our setup using a CO2 chamber. The chamber is
thermally stabilized and humidity is controlled. The result is
shown on Fig.3
Figure 2: Picture of the demonstrator 10cmxF18mm
Output thermopile signal for 3 different CO2 concentration
0ppm / 1500ppm / 3000ppm
600
550
Figure 1: Non Dispersive Infrared gas sensors principle
From this signal, following Beer-Lambert law, one is able to
calculate the concentration of the gas.
On the one hand, today’s solutions are potentially compliant
with ASHRAE norms (+-50ppm + 3% of reading [2]). On the
other hand, they are based on ABC calibration where the
concentration is assumed to be as low as 300ppm at night,
leading to unreliable measurements. These products are
individually assembled, which is not a mass and cost effective
production. Last they require a supply voltage of 5V or
greater and current of 3.7mA or greater [3].
In a previous project “HOMES”, we have developed a
waferlevel “black body” source: a Nitride Titanium suspended
membrane is supplied with a current heating up the device
500
)
V
(m
e
il
p
o
rm450
e
h
t
t
u
p
t
u
O
400
350
300
0
500
1000
1500
# measurment (1 measure /5s)
2000
2500
3000
Figure 1: Characterization thermopile output as a function of [CO2].
If footprint is slightly large, multipath cavities could highly
reduce it. Then, preliminary characterization shows possible
precision of 60ppm in the range 1000ppm. In addition,
vaccum packaging has strongly reduced the voltage x current
need. Last waferlevel technology makes possible the
production of cost effective CO2 sensors.
The demonstrator is today a solution for “carbon dioxide level
–measurement- through the IPv6 wireless and batteryless
GreeNet network” as reported by partner STM at CES [4].
References : f
[1] : lux research inc.
[2] : www.gassensing.co.uk/products/cozir-ambient/
[3] : www.senseair.se/products/oem-module/k30/
[4] : IBM, STm and Shaspa Advance Smarter Home Initiative, Press Release, CES Las Vegas, January, 8th, 2013.
45
New trends in optical gas sensing
Research topics: Optical Sensing, Mid-IR photonics, QCL, photoacoustic spectroscopy
M.Brun, P.Labeye, S.Nicoletti
ABSTRACT: a novel approach to develop miniaturised optical sensors is presented. This approach results from
the compilation of concepts and preliminary experimental results dealing with widely tunable MIR sources, with
the MIR Si-based PIC and MEMS gas detectors all integrated on a single chip.
The monitoring of air pollution and control of industrial
emissions will become in a near future a key requirement for
a better quality of life from both the points of view of the
comfort
and
healthiness.
Topics
like
environment
preservation, emissions control, and reduction of the energy
consumption in public and residential buildings are often
discussed on the media and R&D actions in this field are
nowadays strongly supported by public authorities.
Today’s main technological developments address the
possibility of realizing measuring devices able to effectively
detect the principal pollutants at lower costs. In this context,
optical detection offers a number of advantages related to
the possibility of addressing specific molecules by the choice
wavelength of the source. Indeed, each molecule has a
distinct absorption spectrum which represents a unique
“signature” of the chemical compound. The choice of a given
adsorption peak or of a characteristic band of the spectrum
allows to detect the pollutant and to estimate its
concentration. This approach has been developed in a
number of lab and commercial tools. These tools are
generally bulky and costly, limiting de facto their diffusion in
many application fields. The miniaturization of the gas sensor
with the associated cost reduction is a key issue to improve
sustainability and to address application areas with very high
socio-economic implications, such as climate change or air
quality controls.
This approach results from the compilation of concepts and
preliminary experimental results of a series of papers dealing
with the widely tunable MIR source [Capasso 2009], with the
MIR Si-based PIC [Soref 2008] and MEMS photo-acoustic cell
[Holthoff 2010]. Such an approach has never been realized
so far and it represents a crucial breakthrough for sensing
applications. In this approach, the use of QCL sources allows
to cover a region of the electromagnetic spectrum where the
absorption is the most intense. Figure 1 reports different
pollutant as a function of the most intense wavelength band.
When used in combination with high resolution detections
methods, QCLs can fulfil the requirements of sensitivity and
selectivity needed for trace pollutant detection in air.
This target is achieved by fabricating on the same chip a
miniaturized
photo-acoustic
cell
based
on
MEMS
microphones. In this context the μ-sensor is fabricated
through the extensive use of Si-based Photonic Integrated
Circuits (PIC) and IC/MEMS technology, which allow to merge
on the same chip different functionalities spanning from
integrated optics, fluidics, acoustics and electromechanical
transduction.
CEA-Leti and III-V Lab jointly develop a novel detection
architecture where a multi-wavelength quantum cascade
laser (QCL) source is associated with a photoacoustic cell, all
integrated on a single chip, to realize a miniaturized optical
gas sensor based [1].
Figure 2: Schematics of the µ-sensor assembly.
Figure 1: some of most common pollutants and typical detection
wavelength range
Figure 2 reports a schematics of the µ-sensor assembly
where
the
mutual
arrangement
of
the
different
subcomponents in the device. The key challenge consist in
the combination of different enabling technologies allowing to
merge the output of different sources, via the use of mid-IR
photonics, to handle the beam to transduction unit and to
detect the gas concentration via a suitable measurement
mechanism.
This will constitute a sensitive and selective TDLS system on
a chip able to detect/control several gases.
Références :
3.
S. Nicoletti, M. Brun, P. Labeye, M. Carras “Développement d’un détecteur optique de gaz intégré sur puce” Photoniques Numéro 60,
Juillet-Août 2012
4.
B. G. Lee, Mikhail A. Belkin, Christian Pflügl, Laurent Diehl, Haifei A. Zhang, Ross M. Audet, Jim MacArthur, David P. Bour, Scott W.
Corzine, Gloria E. Höfler, and Federico Capasso. "DFB Quantum Cascade Lasers Array" IEEE JQE, 45(5): 554, 2009.
5.
R. Soref, “Toward Silicon-based Longwave Integrated Optoelectronics”, Proceedings of SPIE Photonics 6898-5, 2008
6.
E. Holthoff, J. Bender, P. Pellegrino and A. Fisher, “Quantum Cascade Laser-Based Photoacoustic Spectroscopy for Trace Vapor Detection
and Molecular Discrimination”, Sensors 2010, 10, 1986-2002, 2010
…
46
47
5
Optics and
nanophotonics
Large nanophotonic systems modeling
Super-resolution optical discs
Solid immersion lenses
Holographic watermarking
48
Multilevel fast multipole method for
the design of large-scale
nanophotonic systems
Research topics: numerical method, electromagnetism, nanophotonic
M. Fall, S. Boutami, A. Glière, J. Hazart, B. Stout
ABSTRACT: We have developed a multilevel fast multipole method, adapted to the simulation of large-scale
nanophotonic systems. The method consists in the formulation of equivalent charge and currents, related to
scalar and vector potentials, at the surface of the scatterers. The calculation of interactions between boundary
elements of the surface is accelerated by the fast multipole algorithm. This method has been applied to the
design of various systems, including plasmonic systems, which can hardly be treated by classical methods.
Nanophotonic structures are generally simulated by volume
methods, as Finite-difference time-domain (FDTD) method,
or Finite element method (FEM). However, for large
structures, or metallic plasmonic structures which require a
fine mesh, the memory and time computation required can
increase dramatically, and make proper simulation infeasible.
Surface methods, like the boundary element method (BEM)
have been developed to reduce the number of mesh
elements. These methods consist in expressing the
electromagnetic filed in whole space as a function of electric
and magnetic currents at the surface of scatterers. Combined
with the fast multipole method (FMM) that enables a huge
acceleration of the calculation of interaction between far
mesh elements (Fig. 1), very large systems can thus be
handled.
What we performed is the development of an FMM on a new
BEM formalism [1], based on scalar and vector potentials
instead of electric and magnetic currents, for the first time to
our knowledge. This method was shown to enable accurate
simulation of metallic plasmonic systems [2], while providing
a significant reduction of computation requirements,
compared to BEM-alone (Table 1). Several thousands of
unknowns could be handled on a standard computer.
Table 1 : Memory and time computation comparisons for BEM-alone
and FMM-BEM (Gold plasmonic nanoparticles of different radii,
=516nm)
Radius
(
R
0
)
0.125
0.145
0.5
1
1.5
2
Number
of
unknowns
Memory (GB)
BEM/MLFMM
Time
(s)
BEM/MLFMM
13 936
15 952
37 824
68 528
134 608
196 752
1.945 /0.162
2.913 /0.157
/1.192
/3.909
/3.399
/4.321
381 /16. 56
1 076 /23.79
/75.85
/371.8
/2 643
/4 801
More complex nanophotic systems have been simulated, such
as a plasmonic lens consisting of a collection of gold
nanorods, as described in [3]. The memory required in this
case is a few GBs, instead of several tens of GBs with FDTD.
Figure 2: (a) plasmonic nanolens with nanorods with width ranging
from 25nm to 75nm [3]; (b) electric field at focal plane (BEM-FMM) ;
(c) electric field in section plane (BEM-FMM).
Figure 1 : (a) surface mesh of a particle ; (b) FMM acceleration by
adaptative domain meshing,and merging of interactions calculations ;
(c) Surface charges at exterior surface of a gold nanoparticle
(=516nm, radius /2).
Our perspective would be to extend the method to periodic
systems with large number of unknowns [4], like image
sensor matrices for example.
[1] F. J. Garcıa de Abajo, A. Howie, "Retarded field calculation of electron energy loss in inhomogeneous dielectrics", Phys.
Rev. B 65, 115418 (2002).
[2] M. Fall, S. Boutami, A. Glière, B. Stout, and J. Hazart,"Multilevel fast multipole method based on a potential formulation
for 3D electromagnetic scattering problems", to appear in JOSA A, (2013).
[3] L. Verslegers, P. B. Catrysse, Z. Yu, W. Shin, Z. Ruan, and S. Fan, "Phase front design with metallic pillar arrays", Opt.
Lett. 35, pp. 844-846 (2010)
[4] S.Boutami and M. Fall, "Calculation of the Free-Space Periodic Green’s Function Using Equivalent Finite Array", IEEE
trans. on antenn. and prop. 60, pp 4725-4731 (2012).
49
Making the semiconductor-metal
transition in InSb compound for double
density Blu-ray super-RENS-ROM discs
Research topics: InSb, semiconductor-metal transition, optical memories
B. Hyot, B. André, L. Poupinet
ABSTRACT: A semiconductor-metal transition is characterized by a sudden change in electrical properties but
also in optical behaviours, as a consequence of a change in electron behaviour. The ability to induce a reversible
semiconductor-metal transition in a material by varying conditions such as applied temperature or electrical
field, results in attractive changes in properties that have fuelled the curiosity of scientists. At LETI, we have
investigated the interest of such materials, in particular InSb compound, exhibiting the reversible
semiconductor-metal transition in the development of the next generation of optical Blu–ray discs, the so-called
super-resolution near field structure (super-RENS) discs.
With the advent of high definition television and the ever
increasing storage demands resulting from the Internet,
higher capacity storage systems are an ongoing requirement.
One of the most promising approaches is the use of the
crystal-to-amorphous transition or phase change recording.
This technology has realized optical disc memories such as
DVD (Digital Versatile Disc) and Blu-ray Disc and proven that
optical storage is an excellent medium for such storage
needs.
However, increasing the capacity of optical recording in the
far-field optical diffraction limit is restricted by the minimum
resolvable spacing given by the resolution limit or Abbe’s
criterion /4NA where  is the wavelength of the light and NA
the numerical aperture of the objective. Optical disk drive
towards higher storage densities in the far-field limit is thus
reaching a significant technological and cost limit with the
use of lasers in the UV range. To circumvent such issues, the
use of optical near-field recording techniques has been an
active area of research. Most techniques (metallic coated
cone shaped fibers, metallic diffusing tips, Solid Immersion
Lens,…) require to control very precisely the distance
between the near field source and the recording area. This
requirement is a brake on the removability and/or the
handling of the disc.
Nevertheless, when an optical property of the object changes
under the influence of the laser spot, detection of details
smaller than the optical diffraction limit becomes possible in
the far field. In optical disc systems the object can easily be
provided with nonlinear optical materials enabling superresolution. Hence super-resolution appears to be a promising
technique which implies a “super-resolving” structure using
only thin film technology and is a candidate to be a likely
successor to the next generation of Blu-ray discs since it
combines removal and backward read-compatibility with
earlier optical storage media.
At the beginning of the development of this new concept of
super-resolution in 1998, chalcogenide phase change alloys
and the amorphous-to-crystal transition are at the heart of
this new generation of optical discs.
Developments driven at LETI on so-called super-resolution
near field structure (super-RENS) discs take advantage of the
reversible semiconductor-metal transition in the InSb
material; the optically (thermally)-induced metallization of a
semiconducting InSb layer through the solid-to-liquid
transition being characterized by a sudden and huge change
of its optical properties.
First success in the video playback on HDTV (High Definition
TeleVision) display from 50 GB (Blu-ray capacity X2) InSbbased super-RENS-ROM discs including a high definition
video content with 1920X1080 pixels was realized recently by
the super-RENS consortium joining three partners: AIST
(National Institute of Advanced Industrial Science and
Technology, Japan), Mitsubishi Electric Co. (Japan) and LETI.
Fig. 1 shows a snapshot of the video demonstration realized
with a data transfer rate of 36 Mbps (Blu-ray transfer rate at
1X).
HDTV
TV HD
1920X1080
Lecteur
Optical
optique
pick up
Disque
InSbà based
SR
Drive
super-RENS disc
Figure 1: Snapshot of high definition video content from InSb-based
super-RENS-ROM disc corresponding to 50 GB per layer (Blu-ray
capacity X2) displayed on HDTV.
InSb-based technology satisfying a data transfer rate of 72
Mbps (Blu-ray transfer rate at 2X), we succeeded in replaying
simultaneously four high definition multiplexed videos
recorded on a 50 GB super-RENS-ROM disc (Fig. 2).
Figure 2: Simultaneous replay of four high definition multiplexed
videos on four HDTV systems.
This achievement indicates that super-RENS technology has a
high potential to realize higher density optical disc system
beyond capacity of Blu-ray system.
References :
[1] K. Nakai, M. Ohmaki, N. Takeshita, B. Hyot, B. André, and L. Poupinet, Jpn. J. Appl. Phys. 49 (2010) 08KE01
[2] B. Hyot, PSSb 249, N°10 (2012) 1992-1998
50
Submicron hollow spot generation
by solid immersion lens: structured
illumination and tolerances
Research topics : µLens, Laser beam shaping; Optical trapping; Optical vortices.
M. Brun, S. Olivier, S. Nicoletti
ABSTRACT: The recent demand for miniaturized high quality optical systems for strong light confinement has
pushed forward the development of solid immersion lenses (SIL), down to the micrometer-scale. These microsolid immersion lenses (μ-SIL) have a potential to be applied in many different fields, such as microscopy,
lithography, optical data storage, Raman spectroscopy, and fluorescent imaging. In order to maximize the
focusing capabilities of the μ-SIL, they should be placed at the focal plane of a high numerical aperture (NA)
objective lens, resulting in an immersed spot whose dimensions are reduced by a factor equals to the µSIL
refractive index. Doughnut shaped intensity distribution generation and alignment tolerances are presented.
We have developed an integrated version of micrometer size
SILs on 200mm Silicon wafers by combining standard
microelectronic and MEMs process. The µSILs is constituted
of SiO2 or SiN material with diameters up to 2µm (see figure
1a) and are integrated on a silicon nitride self standing
membrane released from the silicon substrate by TMAH wet
etching technique. This allows giving optical access to both
sides of the µlens for characterisation and applications. Using
these well controlled technological routes quasi perfect
hemispherical shape are achieved on the micro SILS and
collective realisation allows for µSILS matrix realisation with
very good performances uniformity.
Solid immersion effects have been experimentally verified
using high-resolution interference microscope (HRIM) at EPFL
as illustrated on figure 1b and c. As shown when the focused
spot is located outside the µSILs the spot appears broad and
blurry (1b) limited in size by diffraction effect. Although,
when located in the centre of the µsIls the spot size decrease
and
effects.
a) become brighter due
b) to immersion c)
Figure 1: a) Hemispheric surface of the SiO2 µSIL. b) µSil matrix
bottom view with focalised spot outside the SIL (on the down-left
corner). c) with focalised spot onto the µSIL centre.
By structuring the polarization state of an incident light beam
impinging on a SIO2 µSIL, immersed submicron-size hollow
focused spots can be obtained [1]. Such structured focused
spots are characterized by a doughnut-shaped intensity
distribution, whose central dark region is of great interest for
optical trapping of nano-size particles and super-resolution
microscopy and lithography. Immersion effect influence over
the size and intensity of the final spot has been
experimentally and theoretically investigated for azimuthal
polarized plane wave. Here again immersion effect decrease
the size of the central hollow beam down to 220nm ie 1.3
times smaller than the non-immersed spot. The measured
peak intensity is also increased by a factor 1.57 as shown in
the figure 2.
Figure 2: Measured intensity distributions of the doughnut-shape
hollow spots: on the top left, the non-immersed spot is mapped in a 2
μm circular measuring window. The immersed spot is shown on the
bottom left. Normalised intensity cross section is given on the right.
At a practical level, one point of concern certainly regards the
alignment sensitivity of such micro-sized SILs in the optical
system. Answering these questions is of great relevance for
any application that may profit from the µ-SIL spot size
reduction. Experimental and theoretical study show that the
µsIL a very robust to misalignments when moving laterally
the µSil respectively to the incident beam [2]. To quantify the
tolerance error, the correlation (normalized covariance)
between the misaligned and the ideal focused spots has been
computed. For azimuthal polarisation correlations as high as
95% are found for a displacement of 200nm and reach 80%
at 400nm considered as the bottom limit to maintain a
doughnut like focus spot. This is a very large misalignment
tolerance regarding to the radius of the µSIL of 1µm. The
study have revealed that this alignment tolerance is
dependant of the incident polarisation and even greater for
linear polarisation state (80% at 500nm). It is also
dependant on the refractive index and the size of the µSIL
[2]. This observation sets a practical guide for designing and
evaluating optical immersion systems operating with
objectives and different numerical apertures and/or µSIL
size.
References :
1- Myun-Sik Kim, Alberto C. Assafrao, Toralf Scharf, Arthur J. H. Wachters, Silvania. F. Pereira, H. Paul Urbach, Mickael Brun,
Segolene Olivier, Sergio Nicoletti, and Hans Peter Herzig, « Submicron hollow spot generation by solid immersion lens and
structured illumination », New Journal of Physics 14, (2012) 103024
2- A C Assafrao, M-S Kim, A J H Wachters, T Scharf, H P Herzig, S Olivier, M Brun, S F Pereira and H P Urbach, «Experimental
and theoretical investigation on the misalignment tolerance of micron-sized solid immersion lens.», J. Opt. 15, (2013)
025706
51
Gray tone image holographic
watermarking for analogical archiving
Research topics: Computer Generated Holography, Halftoning, Diffraction grating
C. Martinez, F. Laulagnet, O. Lemonnier
ABSTRACT: We have developed an original approach for the watermarking of holograms in gray tone images for
use in microscopic analogical image archiving. Our concept is based on the coding of Computer Generated
Holograms by elliptical patterns with size and polarity independent of the holographic data but constrained by
the visual representation of a gray tone image. Digital data can then be superimposed to the halftone image
representation in order to increase the analogic media storage capacity with low visual impact.
The increasing use of numerical imaging technologies in
everyday life has increased the needs for image storage
solutions. Materialized storage media such as optical disks,
hard disk drives and flash memory or dematerialized
solutions such as cloud computing offer valuable answers for
storing and sharing huge quantities of data. Although these
solutions are satisfactory for general consumer needs, they
give no guarantee regarding information access durability.
Analogue archiving media such as microforms offer a good
solution to data access durability concerns in the case of
graphical file format preservation. The human readable
approach releases the constraint of file and media format and
is less sensitive to global data loss as it allows progressive,
detectable degradation of the media. Storage capacity is one
of the main drawbacks of this solution. To release this
capacity constraint due to poor information density, a
modern approach of human readable media based on up to
date lithographic technologies has been proposed [1].
techniques. The hologram then allows the coding of both
visual and holographic image.
To improve the visual rendering of the gray tone image, the
diffractive patterns polarity is alternated between open
aperture on the metallic layer and opaque aperture on the
transparent substrate. According to Babinet principle the
hologram phase function is  shifted following the aperture
polarity.
Figure 2 shows the result of a gray tone image engraving
with aperture period 4 µm on a resolution grid 200 nm. The
figure also shows the result of the hologram recovery with a
650 nm Laser. A 69x69 QR Code is easily decrypted.
In the frame of collaboration with the French company
Arnano, specialized in micro engraving for aesthetic and data
archiving applications, we have developed an original
approach for image watermarking with computer generated
holography (CGH). Our principle is to use the complementary
diffraction behaviour stated in the Babinet principle to add
visual information to the Fourier hologram [2].
Figure 1: principle of elliptical Computer Generated Holography
As shown in Figure 1, principle of CGH generation is to code
the complex digital Fourier transform of an image in both size
and location of diffractive patterns distributed in a periodic
grid. Location d and size wx, wy of the patterns allow coding
the hologram phase and amplitude.
As hologram amplitude has little influence on hologram
recovery we modify the diffractive pattern size to render a
visual aspect to the hologram according to halftoning
Figure 2: Top left) Gray tone image. Top right) Microscopic view of the
halftone engraving. Bottom left) Hologram recovery. Bottom right)
Details of the hologram structure and recovery.
Typical application should be the embedment of the image
metadata or the image digital compressed file inside the
engraved halftone raw image representation to improve
analogical data archiving capacity and access time.
References :
1. C. Martinez, O. Lemonnier, F. Laulagnet, A. Fargeix, M.F. Armand, “Micro and Nano Structuring for Long Term Data Preservation,” . French
Symposium on Emerging Technologies for micro-nanofabrication (2010), We-L5.
2. C. Martinez, O. Lemonnier, F. Laulagnet, A. Fargeix, F. Tissot, and M. F. Armand, "Complementary computer generated holography for
aesthetic watermarking," Opt. Express 20, 5547-5556 (2012)
52
53
6
Silicon Photonics
MOS plasmonics
Hybrid optical sources
Modulators
Receivers
Photonic systems
54
CMOS Plasmonics
Research topics : Plasmonics, Silicon Photonics, Modulators
A. Emboras, R.M. Briggs,* A. Najar, S. Nambiar, P. Grosse, E. Augendre, C. Leroux, Hyun Seok Lee,
Chawki Awada,** S. Boutami, Fabrice Charra,** Ludovic Douillard,** B. de Salvo, H.A. Atwater*
and R. Espiau de Lamaestre (*Caltech, **CEA, DSM, IRAMIS)
ABSTRACT: We proposed and validated a new technological platform to fabricate field effect surface plasmon
devices. It relies on state of the art complementary metal-oxide-semiconductor (CMOS) processes available in
the Leti 200mm CMOS foundry, which are further tuned to meet to goal of low optical losses and electrically
reliable operation. Such a platform opens the way towards the high performance and low cost fabrication of
plasmonic devices.
The use of surface plasmon polaritons (SPP) in metal
nanostructures to control light at scales lower than its natural
wavelength has emerged in the past few years as a
promising way of decreasing footprint of optical functions in
integrated silicon photonics. Both the control over the optical
losses of plasmonic modes and the compatibility of SPP
device
fabrication
with
complementary
metal-oxidesemiconductor (CMOS) processes are required to meet the
promises of applications.
In 2012, we first showed that copper layers fabricated by
standard CMOS processes used for electrical interconnections
are particularly well suited to optical purposes, owing to the
huge efforts made in microelectronics to achieve a very good
copper material quality. We fabricated Cu layers which
exhibit optical losses close to the ones of silver, a metal very
often considered for plasmonics but not permitted in CMOS
environment [1].
measurements conducted in a MNOS channel waveguide
configuration coupled to standard silicon photonics circuitry
confirms the very low optical losses (0.39 dB.μm−1), in good
agreement with predictions using ellipsometric optical
constants of Cu.
Finally, in an effort to minimize the insertion losses between
forseen plasmonic devices and standard silicon photonic
circuitry, we have demonstrated a compact, efficient and
straightforward approach for coupling light from a silicon
waveguide to a vertical metal-insulator-silicon-metal (MISM)
plasmonic waveguide [3], by inserting an intermediate metalinsulator-silicon-insulator (MISI) coupler of optimal length
(Fig.2). The device fabrication is fully compatible with CMOS
microelectronics circuitry and fabrication processes, involving
damascene process to fabricate Cu waveguides/electrodes
and direct bonding to achieve vertical integration of the MISM
structure. This coupling scheme provides experimentally
measured coupling loss as low as 2.5 dB per coupler for a
coupler length of only 0.5 µm, despite the high optical
confinement of the MISM mode and its mismatch with the
silicon waveguide mode. This integrated device could be used
as a very low footprint electro-optical plasmonic modulator.
Figure 1 – Left: SEM cross section of the MNOS plasmonic waveguide,
Right: comparison of the reliability of Cu gate MNOS vs MOS.
The use of Cu for CMOS interconnects has led to the
development of efficient diffusion barriers, so that electronics
devices can be connected without detrimental effects on their
electrical reliability. Due to the optical loss issue, integration
of Cu within MOS plasmonic devices requires a diffusion
barrier that is transparent to photons. We proposed a MNOS
stack [2] as a reference building block for integrated field
effect plasmonic devices, where the “N” stands for an
ultrathin stoechiometric silicon nitride layer acting as an
optically
transparent
diffusion
barrier
layer.
We
experimentally proved that the insertion of this nitride layer
in a MOS stack lead to an increase in the electrical reliability
of a copper gate capacitance from 50 to 95%, while
preserving the low optical losses brought by the use of
copper as the plasmon supporting metal. Optical transmission
Figure 2 - Top: scheme of the fabricated MISM devices with integrated
MISI input and output coupler. Bottom, from left to righ
SEM cross section of the active MISM stack, microscope top view of
the device with input and output silicon waveguides as well as Cu
electrodes, coupling loss variation as a function of the coupler length.
References :
1. Hyun Seok Lee, C. Awada, S. Boutami, F. Charra, L. Douillard and R. Espiau de Lamaestre, Optics Express, Vol. 20, Issue 8, pp. 8974-8981
(2012)
2. A. Emboras, A. Najar, S. Nambiar, P. Grosse, E. Augendre, C. Leroux, B. de Salvo, and R. Espiau de Lamaestre, Optics Express, Vol. 20,
Issue 13, pp. 13612-13621 (2012)
3. A. Emboras, R. M. Briggs, A. Najar, S. Nambiar, C. Delacour, Ph. Grosse, E. Augendre, J. M. Fedeli, B. de Salvo, H. A. Atwater, and R. Espiau
de Lamaestre, Appl. Phys. Lett. 101, 251117 (2012)
55
Hybrid III-V/Si Photonic Crystal
Vertical-Cavity Surface-Emitting Lasers
Research topics: Hybrid IIIV/Si semiconductor lasers, Photonic Crystals.
B. Ben Bakir, C. Sciancalepore, D. Bordel, JM Fedeli, S. Menezo,
X. Letartre* and P. Viktorovitch* (*INL Lyon)
ABSTRACT: Long-wavelength vertical-cavity surface-emitting lasers (VCSELs) for photonics-on-CMOS
integration based on Si/SiO2 photonic crystals mirrors (PCMs) have been fabricated. Capable to operate
continuous-wave up to 40°C at low thresholds, such compact VCSELs show single-mode polarization-stable
operation at 1.55-µm with uncooled output powers in excess of 0.4 mW. Noticeably, the light molding obtained
through the engineering of Si/SiO2 photonic crystals allows for a tailored modal selection and full polarization
control. Furthermore, the high-throughput cost-effective Si-based process technology developed is ideally wellsuited for an industrial development perspective.
Recently, our research efforts were oriented to the realization
of ultra-compact vertical-cavity surface-emitting lasers as
next-generation emitters for silicon photonics applications.
This silicon-based photonic building block has been conceived
within a large-scale CMOS-compatible processing technology
mindset. In detail, the silicon patterning of 1-D photonic
crystal mirrors (PCMs) on 200-mm-wide semiconductor-oninsulator (SOI) wafers aims at both an efficient light
harnessing and optimal optical confinement, while the III-V
epitaxial layers providing light amplification are wafer-bonded
to SOI by state-of-the-art molecular bonding to ensure laser
performances, CMOS-compatibility and large-scale low-cost
fabrication.
Figure 1: (a) Schematic view of a III/V-on-Si VCSEL cavity employing
a double set of Si/SiO2 1-D photonic crystal mirrors (PCM). (b) Crosssectional SEM view of the double PCM-VCSEL cavity.
The VCSEL architecture for optical pumping is sketched up in
Fig. 1(a), while a scanning electron microscope crosssectional view of fabricated devices is shown in Fig. 1(b).
Briefly, an InGaAsP-based three-quantum-well active region
is embedded between two silica spacers measuring 900 nm
each, while the VCSEL cavity is vertically terminated by two
290-nm-thick 1-D Si/silica photonic crystal mirrors
characterized by a 50% in fill-factor and 910 nm lattice
period, respectively. The photonic crystal membrane has
been designed in order to enhance its spectral response both
in terms of modal reflectivity and optical confinement, aiming
at low-threshold VCSEL devices.
Concerning the device optical and thermal features reported
in Fig. 2, these VCSELs operate continuous-wave up to 40°C
with output powers in the mW range, being the temperature
limit mainly ascribable to the use of InGaAsP-based active
regions. Lasing thresholds within the sub-mW range and
side-mode suppression ratio > 25dB confirm the optimal
modal confinement - both vertical and lateral - achieved in
these VCSEL structures also thanks to the introduction of
photonic
crystal
heterostructures
within
the
silicon
membranes.
Figure 2: CW lasing behavior of double PCM-VCSELs at different
pumping powers and stage temperatures. (Left) Single-mode emission
with 26 dB of transverse SMSR is obtained up to a stage temperature
of 40°C. The estimated thermal tuning coefficient is 0.057 nm/K.
(Right) LL curves corresponding to different stage temperatures.
References :
[1] C. Sciancalepore, B. Ben Bakir, X. Letartre, J. Harduin, N. Olivier, C. Seassal, J.-M Fedeli, and P. Viktorovitch, “CMOS-compatible 1.55-µm
ultra-compact emitting VCSELs using double photonic crystal mirrors,” IEEE Photon. Technol. Lett., vol. 24, no. 6, pp. 455-457, 2012.
[2] C. Sciancalepore, B. Ben Bakir, C. Seassal, X. Letartre, J. Harduin, N. Olivier, J.-M Fedeli, and P. Viktorovitch, “Thermal, modal and
polarization features of double photonic crystal vertical-cavity surface-emitting lasers,” IEEE Phot. J., vol. 4, no. 2, pp. 398-410, 2012
[3] C. Sciancalepore, B. Ben Bakir, X. Letartre, J.-M. Fedeli, N. Olivier, D. Bordel, C. Seassal, P. Rojo-Romeo, P. Regreny, and P. Viktorovitch,
“Quasi-3D light confinement in double photonic crystal reflectors VCSELs for CMOS-compatible integration”, J. Lightw. Technol., vol. 29, no. 13,
pp. 2015–2024, 2011.
[4] C. Sciancalepore, B. Ben Bakir, X. Letartre, N. Olivier, D. Bordel, C. Seassal, P. Rojo-Romeo, P. Regreny, J.-M. Fedeli, and P. Viktorovitch,
“CMOS-compatible integration of III–V VCSELs based on double photonic crystal reflectors”, in Proc. 8th IEEE Int. Conf. GFP, Sep. 2011, pp.
205–207
56
Silicon optical modulators
Research topics : Optical Interconnect, Si Modulator
J. M. Fédéli, E Grellier, P.Rivallin,( LETI/DCOS) , S. Menezo, M Fournier, P Grosse, L. Vivien*,
D.Morini*, D Thomson**, G.T.Reed** (* IEF, ** University of Surrey)
ABSTRACT: In the frame of the FP7 Helios project, we have developed a full set of silicon modulators with
40Gbps operation using carrier depletion . Trade-offs and challenges for those types of modulators are reviewed
I. INTRODUCTION
Year after year, the different network interconnects, ranging
from long haul to on chip communications, continue to
increase in both complexity and bandwidth. As the
technology becomes more cost effective and bandwidth
requirements increase, optical links are becoming competitive
with their electrical counterparts for transmission over
shorter and shorter distances. For applications such as high
performances computing or data centres, silicon photonics is
the leading candidate for optical link transmission due to its
low fabrication costs, its CMOS compatibility and the
possibility for co-integration with electronics.
consumption and footprint.
We have fabricated several carrier-depletion-based MZIs
capable of 40Gbit/s modulation. In [1], we fabricated a
Mach-Zehnder based on a pn junction with an efficiency of
VπLπ=2.7 V.cm and optical loss of 4.5 dB/mm.
For the generation of modulated light, a continuous laser
source and an optical modulator is the main required
functionalities for any optical interconnect solution.. There
are multiple figures of merit for the optical modulator. Ideally
the modulator would have a high modulation speed, high
extinction ratio, low optical losses, small footprint, low
driving voltage, low power consumption, a wide operating
temperature range and the desired optical bandwidth.
Figure 1: Microscope image of a fabricated MZI and ring modulator [1]
In [2], we fabricated a device using a pipin junction with
significantly lower optical losses of 1dB/mm and an efficiency
of 3.5 V.cm. In Figure 2 we represent a lateral pipin phase
shifter. Thanks to the intrinsic region inside the 420nm-large
waveguide, this structure exhibits less absorption due to the
free-carriers than a pn phase shifter.
II. RESEARCH AND RESULTS
During the HELIOS project and in collaboration with ( the
Institut d’Electronique Fondamentale” and the University of
Surrey, CEA-LETI fabricated modulators based on the
plasma-dispersion effect where the silicon refractive index is
changed by free-carrier concentration variation.
The plasma dispersion effect is obtained in a carrier depletion
structure using reverse biased diodes or in a carrier injection
structure using forward biased diodes. In general, carrier
injection devices present lower optical losses and lower
driving voltage than carrier depletion ones. However the
photon lifetime with carrier injection limits the modulation
rate at few Gigabit/s unless pre-emphasis is used.
The variation of refractive index resulting from the plasmadispersion effect is converted into an intensity modulation of
the signal thanks to either a Mach-Zehnder interferometer
(MZI) or a resonant structure such as a microring resonator.
Mach-Zehnder interferometers are commonly used due to
their broad spectral bandwidth (and thus related tolerance
against temperature variations). They however require a
large interaction length. The choice of the active region
length is a trade off on which depend multiple key
characteristics of the modulator. A long region is needed for
a high extinction ratio and low driving voltages whereas a
small diode junction is desired for low optical losses, power
Figure 2: Schematic views a lateral pipin phase shifter [2]
Due to their relatively large size, and thus large capacitance,
MZI modulators exhibit power consumption of several pJ/bit
[1], which is too high for network-on-chip applications.
More power-efficient modulators can be realized with ring
resonator devices. In [3], we fabricated a modulator of 6µmradius with a measured capacitance of 16fF enabling
modulation power consumption of only 32fJ/bit. When driven
with a 1.4Vpp voltage swing, an extinction ratio of 1.6dB
(1.1dB at 40Gbit/s) and an additional loss of 10dB were
obtained.
III. FUTURE WORK
In order to address short links applications, the power
consumption should decrease in the range of tens of fJ. So
optimization of ring devices, development of slow waves
devices, and new modulator concepts are path for energy
efficient devices.
References:
[1] D. J. Thomson, F. Y. Gardes, Y. Hu, G. Mashanovich, M. Fournier, P. Grosse, J.-M. Fédéli, and G. T. Reed, “High contrast 40Gbit/s optical
modulation in silicon,” Opt. Express 19(12), 11507–11516 (2011).
[2] Melissa Ziebell, Delphine Marris-Morini, Gilles Rasigade, Jean-Marc Fédéli, Paul Crozat, Eric Cassan, David Bouville, and Laurent Vivien, "40
Gbit/s low-loss silicon optical modulator based on a pipin diode", Optics Express, Vol. 20, Issue 10, pp. 10591-10596 (2012)
[3] D. J. Thomson, F. Y. Gardes, D. C. Cox, J.-M. Fedeli, G. Z. Mashanovich, and G. T. Reed, "Self-aligned silicon ring resonator optical
modulator with focused ion beam error correction," J. Opt. Soc. Am. B 30, 445-449 (2013)
57
Ge PIN Waveguide Integrated
Photodiodes for Optical
Communications Applications
Research topics: Photodetectors, Silicon photonics, Germanium
L. Virot , J.M. Fédéli , J.M. Hartmann , L. Vivien*, P Crozat* (* IEF)
ABSTRACT: Germanium has been proven to be a good candidate for waveguide integrated photodiodes. We
present here the latest results on lateral PIN Ge photodiodes. Bandwidth over 50GHz at zero bias has been
achieved as well as responsivity over 0.8A/W at 1.55µm wavelength. Very low dark current of the order of 25nA
at -1V has been obtained. Those performances should allow operation at 40Gb/s for several optical
communications applications.
Despite the lattice mismatch between Silicon and
Germanium, integration of crystalline Ge on Si has been
possible thanks to developments in hetero-epitaxy [1,2].
Using a two-step Reduced-Pressure Chemical Vapor
Deposition (RPCVD) process, followed by optimized
annealing, it has been possible to obtain very good quality Ge
layers on Si, with reduced Threading Dislocation Density
(TDD). Those TDD contributes directly to dark current
generation when the photodiode is reverse biased. The Ge is
grown in 10x10µm Si cavities at the end of the Si waveguide
leading to butt-coupled photodiode. To define the PIN
junction, a lateral structure has been adopted and p-type and
n-type regions were defined by ion implantation process. The
implantation and annealing steps are of importance since it
will both define the quality of the contact but also the electric
field inside the junction. With lateral junction, only one etch
step is necessary to take contact on top of doped Ge. This
simplifies the process for large scale integration of the
photodiodes with modulators and passive devices. The best
dark currents at -1V reverse bias were measured at 6nA and
25nA for 1µm and 0.5µm (Fig. 1) intrinsic region width
design respectively; those values are the lowest reported
ones for this type of Ge photodiodes.
design, very high bandwidth has been obtained as shown on
the Fig.2. 40Gb/s operation at zero bias has also been
demonstrated [3] as shown by open eye diagram in Fig. 3.
For the largest intrinsic region width of 1µm, the -3dB optical
bandwidth was measured to be 28GHz and over 50GHz at
zero bias and -1V respectively.
Figure 2: Typical frequency response of PIN Ge photodiodes.
By processing three 200mm SOI wafers, yields of 100% (in
term of functional diodes) were obtained for 1µm intrinsic
region width design and over 97% for 0.5µm design.
The responsivity at 1550nm was around 0.8A/W and 0.5A/W
for 1µm and 0.5µm intrinsic region width respectively.
40Gbit/s
Figure 3: Eye diagram at 40Gb/s at zero bias.
Figure 1: Typical dark and photo current of PIN Ge photodiode of
0.5µm intrinsic region width design.
Thanks to optimized doping and annealing steps, the intrinsic
region can be reduced to improve the zero volt efficiency of
the photodiode. Thus with the smallest intrinsic region
By tailoring the intrinsic region width by optimized design and
process (ion implantation and annealing), both the
responsivity and frequency response can be adjusted to meet
the requirements of the targeted applications. Zero bias
operation should allow very high sensitivity, limited only by
the Transimpedance Amplifier noise performances.
References :
[1] J.M. Hartmann, A.M. Papon, V. Destefanis, T. Billon, “Reduced Pressure Chemical Vapor Deposition of Ge Thick Layers on Si (001), Si(011)
and Si(111)”, J. Crystal Growth 310 (2008) 5287-5296.
[2] J.M. Hartmann, A. Abbadie, J.P. Barnes, J.M. Fédéli, T. Billon, L. Vivien, “Impact of the H2 anneal on the structural and optical properties of
thin and thick Ge layers on Si; Low temperature surface passivation of Ge by Si”, J. Crystal Growth 312 (2010) 532-541.
[3] L. Vivien, A. Polzer, D. Marris-Morini, J. Osmond, J.M. Hartmann, P. Crozat, E. Cassan, C. Kopp, H. Zimmermann, J.M. Fédéli, “Zero bias
40Gb/s Ge waveguide photodetector on Si”, Opt. Express 20(2) (2012) 1096-1101.
58
16 channel receiver with 20 GHz Ge
Photodiodes
Research topics : receiver, Ge photodiodes, silicon photonics
J.M. Fedeli, L.Virot, J.M. Hartmann (LETI/ DTSI) , P.Grosse ,
W.Bogaerts*, L. Vivien** (* IMEC, Univ. Gent, ** IEF )
ABSTRACT : A 200GHz 16 channel receiver with polarization management was obtained with a 2D grating
coupler, 2xAWGs and 16 Ge photodiodes. PDL was below 1dB, BW above 20GHz, receiver sensitivity in the order
of 0.08 A/W.
Submicron silicon photonics have generated an increasing
interest
in
recent
years,
mainly
for
optical
telecommunications
or
for
optical
interconnects
in
microelectronic circuits. The rationale of silicon photonics is
the reduction of the cost and energy of communications
systems through the integration of photonic components and
an electronic integrated circuit (IC) on a common chip
(telecommunications applications), or the enhancement of IC
performances with the introduction of optics inside a high
performance chip (core to core communications), or low cost
sensors. The FP7 HELIOS project aimed to combine a
photonic layer with a CMOS circuit by using microelectronics
fabrication processes. A first goal was to develop high
performance generic building blocks for a broad range of
applications: WDM sources by III-V/Si heterogeneous
integration [1], fast modulators [2,3] and detectors [4],
passive circuits and packaging. One of the demonstrators is a
16x10 Gb/s receiver with different building blocks: A 2D
surface grating couples the light coming from a single mode
fiber SMF fiber into the circuit and separates the two
polarizations while transforming the TM polarization into TE.
Identical 200GHz 16 channel AWGs receive the two input
signals and demultiplexes the guided TE modes. The two 16
output waveguides are then connected to 16 Ge photodiodes
(figure 1)
Figure 1: Top-view with Optical Microscopy of the 16 channel receiver
with 2D couplers (left) and Ge PD (right)
We have developed a self-aligned process for the fabrication
of the waveguides using two photolithography steps with a
193 nm stepper and two Si dry etching steps for the
fabrication of gratings and waveguides on optical SOI
substrates (220nm Si on top of 2µm Buried Oxide). We then
defined cavities for the selective epitaxial growth of
Germanium at the end of the waveguides. After doping with
ion implantations to form a PIN Ge diode, damascene type
metallization was used to form the electrodes.
For 480nm x 220nm cladded waveguides, the losses were
found at 2.3 dB/cm. The optimal efficiency for the 2D grating
coupler was experimentally found to be 15% (~8dB coupling
losses) at 1550nm with a 3dB bandwidth of 55nm. The
minimum PDL was measured at ~1dB at ~1550nm. The 16
channels AWG with 200GHz separation performed with a
crosstalk at -15 dB, and the minimum center-channel
insertion losses was around 2.8 dB.
The Ge photodiode is a butt coupled PIN lateral type of 10µm
length with a sensitivity ~ 0.8 A/W. Capacitance is below 5fF
range and dark current of the order of 20nA (-0.5V). With an
applied voltage of -1V, the bandwidth of all dies is above 15
GHz while the mean value is around 20GHz which is
comfortable for 10GB/s operation and should be also enough
for 25Gbit/s operation in new receivers.
Figure 2 : Photodiodes measurement after AWG
The spectral characteristic of the receiver is shown in figure
2. With the losses and sensitivity of the basic blocks, the
overall receiver sensitivity is in the order of 0.08 A/W with a
channel separation of 1.6nm, corresponding to 200GHz.
References :
[1]
M. Lamponi, S. Keyvaninia, C. Jany, F. Poingt, F. Lelarge, G. de Valicourt, G. Roelkens, D. Van Thourhout, S. Messaoudene, J.-M.
Fedeli, G.H. Duan, “Low-threshold heterogeneously integrated InP/SOI laser with a double adiabatic taper coupler”, IEEE Photonics Technology
Letters, Volume: 24, Page(s): 76 – 78, 2012.
[2]
D. J. Thomson, F. Y. Gardes, Y. Hu, G. Mashanovich, M. Fournier, P. Grosse, J-M. Fedeli, and G. T. Reed, "High contrast 40Gbit/s
optical modulation in silicon," Opt. Express 19, 11507-11516 (2011)
[3]
M.Ziebell &al, “40 Gbit/s low-loss silicon optical modulator based on a pipin diode”, to appear in Optics Express in 2012
[4]
L.Vivien & al, “Zero-bias 40Gbit/s germanium waveguide photodetector on silicon," Opt. Express 20, 1096-1101 (2012)
[5]
W. Bogaerts, D. Taillaert, P. Dumon, D. Van Thourhout, R. Baets, and E. Pluk, “A polarization-diversity wavelength duplexer circuit in
silicon-on-insulator photonic wires.,” Optics express, vol. 15, no. 4, pp. 1567-78, Feb. 2007
59
Optical Orthogonal Frequency-Division
Multiplexing (OFDM) transmission using
full CMOS-compatible transmitter
Research topics : Optical OFDM, Hybrid III/V-on-Silicon Laser, Quadrature-Amplitude
Modulation
G. Beninca de Farias, S. Menezo, A. Descos, B. Ben Bakir
ABSTRACT: The direct-OFDM modulation of a hybrid III/V-on-Silicon laser, developed and fabricated in the
CMOS Photonics lab, is demonstrated for the first time. While classical On-Off Keying (OOK) modulation limits
the data-rate at 6Gbps, we demonstrate a record data-rate of 12.6Gbps in a back to back configuration. A
9.4Gbps transmission data rate is further achieved over 250m of Multi-Mode Fiber (MMF). This work paves the
way for a full-CMOS compatible optical transmitter.
In the optical fiber communication domain, the classical OnOff Keying (OOK) modulation is the technique used in most
transmission systems shorter than 100km. While it is has
very low complexity, the product bit-rate x distance is rapidly
limited by chromatic dispersion (in Single-Mode Fiber SMF transmission)
or
intermodal
dispersion
(in
MMF
transmission).
A
particular
multi-carrier
technique,
Orthogonal-Frequency Division Multiplexing (OFDM), is
gaining interest for optical communication for its robustness
against time-dispersive channels, and for allowing much
higher data-rates by using spectral-efficient higher-order
modulation formats (Quadrature-Amplitude Modulation –
QAM) [1]. The OFDM signal can be generated and processed
with a Digital Signal Processor (DSP). In a general sense, we
analyze the implementation of optical OFDM using an entire
photonic transceiver compatible with CMOS technology. In
order to do optical OFDM transmission, special requirements
such as linear and low-noise devices are necessary.
We experimentally demonstrate a direct-OFDM modulation of
a hybrid III/V-on-Silicon laser, achieving a data-rate of
12.6Gbps in optical back-to-back (with no fiber) [2] and
9.4Gbps after 250m of MMF. The modulation level-format of
each subcarrier is optimized according to the measured
Signal-to-Noise Ratio (SNR) and the targeted Bit Error Rate
(BER) is below 2.2e-3.
Our hybrid III/V-on-Silicon laser fabricated using direct
bonding technique [3], exhibits a threshold current of 90mA,
and 1mW of optical power coupled to the MMF at 220mA bias
current. The -3dB modulation bandwidth is measured to be
4.2GHz as shown in Figure 1, at 220mA bias. As a
benchmark, we evaluated direct OOK modulation of the laser.
An opened eye diagram is measured up to 6Gbps (cf. Figure
1), but no transmission could be made after 250m of MMF
due to the dispersion.
through a computer. The OFDM signal is defined off line in a
classic manner for OFDM signal generation: 1) the serial
incoming data is divided into parallel channels (42 in this
work); 2) the data is then encoded using QAM modulation; 3)
an inverse FFT (IFFT) generates each of the 42 sub-carriers
with phase and amplitudes defined by the QAM modulators,
and superimposes them. As indicated in the inset of Figure 2,
the OFDM signal baseband bandwidth is 4GHz. An Arbitrary
Waveform Generator (AWG) serves as a Digital to Analog
Converter (DAC), with 8bits resolution. The rms (respectively
peak, pk) laser modulating current is 14mA rms (respectively
57mA pk current). The achievable data-rate is evaluated
firstly in optical B2B (with no fiber), and after 250m of MMF.
At the receiver, a photo-receiver is used with an overall gain
of 600V/W is used. A digital oscilloscope serves as the Analog
to Digital Converter (ADC) also with 8 bits resolution.
Figure 2: Experimental Set-up
The SNR computation per subcarrier is consistently verified
by running a first acquisition and allocating Quadrature
Phase-Shift Keying (QPSK) symbols to the 42 data
subcarriers. From the SNR distribution which is reported in
figure 3, we compute the modulation order (e. g. the number
of bits) maximum that we can transmit so that the (BER) is
kept below 2.2e-3. The QAM mapping varies from QPSK (2
bits) to 32-QAM (5 bits) for the subcarriers with the best SNR
as shown in Figure 3. The mean BER for optical B2B is 8.2e4, and for 250m of MMF is 1.7e-3, which can still be
corrected with appropriated error correcting codes.
Figure 3: Experimental Results
Figure 1 : Hybrid Laser Characterization
The experimental set-up for the direct OFDM transmission is
shown if Figure 2. The OFDM signal is generated off-line
Recently, with an improved laser design, we could achieve a
data-rate of 21.5Gbps in optical B2B, and 12.5Gbps after
50km of SMF, with no optical amplification. These results will
be presented this year in an international conference.
[1] J. Armstrong, “OFDM for Optical Communications”, IEEE Journal of Lightwave Technology, Vol. 27, No. 3, pp. 180-204,
January 2009.
[2] S. Menezo, G. Beninca de Farias, B. Ben Bakir, A. Descos, N. Genay, “12.6 Gb/s Direct OFDM modulation of a Hybrid IIIV-on-Silicon Laser”, Group IV Photonics Conference (GFP) 2012.
[3] B. Ben Bakir, A. Descos, N. Olivier, D. Bordel, P. Grosse, J.L. Gentner, F. Lelarge and J-M. Fedeli, “Hybrid Si/III-V Lasers
with adiabatic coupling”, Group IV Photonics Conference (GFP) 2011.
60
CMOS Photonic Circuits for Passive
Optical Networks (PON)
Research topics : Fiber To The Home (FTTH), Optical Network Unit (ONU), CMOS
Photonics, Reflective Mach-Zenher Modulator (R-MZM)
S. Menezo, G. Beninca de Farias
ABSTRACT: Frequency Division Multiplexing/Frequency Division Multi-Access (FDM/FDMA) passive optical
networks are shown to provide a possible solution in terms of performance, manufacturability and cost for the
second next generation passive optical access systems (NG-PON2). The CMOS Photonics Lab studies such
solutions within the frame work of FAON and FABULOUS projects (partially funded by the French National
Agency for Research and the European Commission). The link capacities are experimentally evaluated, and the
implementation of the required optical network unit in silicon photonics is being demonstrated, as this CMOS
compatible technology is well suited for mass market applications.
The second next generation of access networks (NG-PON2) is
forecast to increase the global capacity of access networks
systems well beyond 40 Gbps downlink and 10 Gbps uplink,
in order to provide a sustained data capacity of 1 Gbps per
user. Those systems will also be able to reach between 64
and 1000 users per feeder, with a passive reach of 20 to 40
km. It is generally accepted that the use of wavelength
division multiplexing (WDM) will be required. A large family
of solutions proposes a pure WDM shared access technique
with one individual wavelength per user. Others advocate the
combination of WDM with another per wavelength shared
access mechanism such as, for instance, time domain
multiplexing and multiple access (TDM/TDMA) or orthogonal
frequency division multiplexing and multiple access
(OFDM/OFDMA). In previous papers [1], a polarization
independent reflective Mach–Zehnder Modulator (R-MZM)
was proposed, capable of providing optical carrier
suppression modulation. The latest is suitable for single
wavelength OFDMA or FDMA applications in the uplink
direction of passive optical networks (PONs). FDM/FDMA can
be a very interesting alternative to TDMA as it removes the
need for burst mode operation and also can decrease
significantly the bandwidth of the electronic stages in the
optical network unit (ONU), thus reducing its cost and power
consumption. By finding a good compromise between the
maximum RF bandwidth accessible by any ONU and the cost
of the electronic stages, it is possible to achieve
simultaneously a suitable service level (max. data rate) and
acceptable deployment costs. Most of all, the ONU could be
realized completely in silicon photonics, integrating
complementary metal oxide semiconductor (CMOS) and
bipolar CMOS (BiCMOS) electronics, thus reducing further its
cost and making it suitable for a mass market such as the
optical access market.
In [2], we further assess the architecture of the previously
proposed system through a theoretical analysis for the
downlink and an experimental test-bed for the uplink. From
this, specifications are given for the ONU modem, including
required photonics and electronics. Following the system
level analysis with both downstream (DS) and upstream (US)
links, we study the possibility of making the required ONU
exclusively using silicon photonics and off-the-shelf
electronics.
A possible silicon photonics implementation of the ONU
reflective MZM is represented in Fig. 1. Following the
clockwise (CW) light path from the input fiber, the light is
coupled into the silicon photonic die through a) a 2D surface
grating coupler. It then travels through b1) a first WDM
tunable filter, c1) a first Semi-conductor Optical Amplifier
(SOA), d) a silicon MZ modulator and c2) another SOA, b2)
another WDM filter and back to the optical fiber through the
2D grating coupler.
Figure 1: Reflective Mach-Zehnder Modulator (Transmitter of the ONU
for the uplink)
A tap is taken at the output of the MZ modulator and
detected by e1) a monitoring waveguide-coupled photodetector in order to properly bias the MZ modulator. The
counterclockwise (CCW) path is reversed.
The ONU receiver is simply made of a photo-detector with an
associated trans-impedance amplifier. The integration of the
listed components can be made fully monolithically on silicon
wafers in CMOS-compatible platforms. A full monolithic
integration will allow a significant reduction of packaging, and
therefore cost.
The first generation of demonstrators will be focused on the
photonic circuit.
The second generation of demonstrators will comprise the
photonic circuit with the CMOS electronic drivers. Leti will use
its 3D copper pillar technology for achieving the electro/optic
integration.
References :
[1] B. Charbonnier, N. Brochier, and P. Chanclou, “Reflective polarisation independent Mach-Zenhder modulator for
FDMA/OFDMA PON,” Electron. Lett., vol. 46, no. 25, pp. 1682–1683, 2010.
[2] B. Charbonnier, S. Menezo, P. O’Brien, “Silicon Photonics for Next Generation FDM/ FDMA PON”, Journal of Optical
Communications and Networks, VOL. 4, NO. 9, SEPTEMBER 2012.
61
7
Solid state lighting
Nanowire light emitters
Nanoimprint for solid state lighting
Thermal engineering
(LED)
62
Nanowire growth mechanisms
Research topics: nanowires, wide band gap semiconductors, epitaxial growth, substrates,
polarity
G. Perillat-Merceroz, R. Thierry, P. Ferret, G. Feuillet and P.H. Jouneau (CEA/DSM/INAC)
Semiconducting nanowires are potentially interesting building blocks for novel and efficient optical devices. It is
the aim of this work to assess the allegedly structural perfection of these nanostructures when grown on
different types of substrates. We take ZnO as a case study. Whatever the substrate, the overall morphology of
the nanostructures are similar (a wire on top a pyramid), and the wires have opposite polarity compared to
their originating pyramids. However the location of the inversion domain boundary does depend upon the
underlying substrate, pointing to different nucleation mechanisms. Since inversion domain boundaries are
structural defects, a particular attention should be paid to controlling these defects in the nanostructures.
Controlling
the
growth
and
the
morphology
of
semiconducting nanowires appears crucial if one wants to
optimize the performance of nanowire-based devices such as
photovoltaic solar cells, nano-generators, or light-emitting
diodes. With this in mind and as a case study, we
investigated the nucleation and growth mechanisms of ZnO
nanowires grown by metal-organic vapor phase epitaxy.
Different substrates were considered to try and determine
their possible influence on the structure of the nanowires
(namely O polar ZnO for homepitaxial growth and c plane
sapphire fro heteroepitaxial growth).
Whatever the substrate and for the same growth conditions,
the overall morphology of the nanostructures is identical,
consisting in a pyramid with a nanowire sitting on top
(fig.1a). Whatever the substrate, ZnO nanowires are Znpolar, as demonstrated by convergent beam electron
diffraction while the pyramids are of O polarity (fig.1b, c, d).
Examination of a fair population of nanowires grown either
homo-epitaxially on ZnO or heteroepitaxially on sapphire,
allowed us to draw a clear picture of the nucleation
mechanisms of the inversion domain boundary at play in the
two cases.
For growth on ZnO (Fig.2), as growth proceeds, the inversion
domain boundary is found to move up in order to remain at
the top of the O-polar pyramids. It is proposed that the local
segregation of aluminium impurities at the top of the
pyramid, originating from the ZnO substrate, could account
for the nucleation of inverted domains as discussed in other
materials such as GaN:Mg. On the other hand, for growth on
sapphire substrates (figure 3), the nucleation of Zn-polar
wires occurs at the top of the O-polar pyramids, within the
pyramid itself or at the sapphire/ZnO interface. In this last
case, the nucleation of inverted domains could be partly
attributed to atomic steps, but also to the non-polar
character of the sapphire substrate.
To conclude, this work stresses the fact that crystal polarity
governs both the nucleation and the shape of ZnO
nanostructures. We recently found out that similar
behaviours exist in the case of GaN nanostructures. Devicewise, a particular attention should be paid to controlling
these structural defects since they could be detrimental to
optical/ electrical efficiencies.
Figure 1 TEM image of the wire on a pyramid (on a ZnO buffer ) and
related CBED diffraction pattterns , showing that nanowire and
pyramid have opposite crystal polarity
Figure 2:TEM images.
Nanowire growth on ZnO
substrate for two different
growth durations showing
that the inversion domain
boundary moves up during
growth to stay at the top of
the underlying pyramid
Figure 3 : TEM images of the nanowire on pyramid structure in the
case of growth on sapphire. The inversion domain boundary is found
to nucleate on top of the pyramid, within the pyramid itself or at the
substrate interface
References :
G. Perillat-Merceroz, R . Thierry, P.H. Jouneau, P. Ferret, G. Feuillet
Compared growth mechanisms of Zn-polar ZnO nanowires on O-polar ZnO and on sapphire
Nanotechnology 23 (2012) 125702
63
GaN nanowire Selective Area Growth
Research topics : nanostructures for Solid State Lighting, LED
D. Salomon, B. Amstatt, A. Dussaigne, P. Ferret, M. Lafossas, G. Perillat-Merceroz, S. Landis,
B. Martin, R. Templier
C Bougerol*, B Gayral*, Xiaojun Chen*, J. Eymery*, C. Durand *
(*CEA/INAC or CNRS Institut Néel.)
ABSTRACT: So as to control the growth of nanowire (NW) heterostuctures for LEDs devices, it is highly
necessary to make sure that all nanowires will have similar morphologies and dimensions. It is shown here that
this is only possible through the use of selective area growth (SAG); It is further demonstrated that the shape of
the obtained nanostructures depends on their crystal polarity which, in turns, is highly dependent upon the
nature and/or the crystalline polarity of the underlying substrate.
An important development in our Solid State Lighting
activities concerns nanowire (NW) Light Emitting Diodes
(LEDs). The expected advantages of NW over the standard
2D growth are based on the following concepts: higher
material quality, larger active surface, 200 mm Si compatible
technology …
However, fundamental issues have to be addressed in order
to control the growth of nanowire (NW) based LEDs
heterostuctures: NW nucleation mechanisms and shape,
indium incorporation, doping issues, strain state... But, since
the optical and electrical properties of the nanowire LED
strongly depend upon the uniformity of the individual
nanowires, only selective area growth (SAG) can be
envisaged, as pointed out below.
Similar processes can be undertaken on pre-patterned
substrates in order to obtain organized growth of nanowires.
This is done here using bulk GaN substrates were polarity is
perfectly defined. A selectivity mask (Si3N4) is deposited on
the N-polar (resp. Ga-polar) surface (Fig.2.a, resp. 2.b) of a
free-standing GaN substrate. Opening in the mask (Nanoimprint or optical lithography) yields a perfect control on the
position and size of the nanowires. In this case, homepitaxial
growth leads to either flat top NWs for N polarity but to
pyramids for Ga polarity.
These results stress the critical issue of crystal polarity to
control the growth of nitride nanostructures, which is highly
critical for the properties and efficiency of nanowire based
LEDs.
In this work [1], we show that crystal polarity strongly
impacts the MOVPE (Metal Organic Vapor Phase Epitaxy)
growth of GaN NWs. Let us first consider the case of selforganised hetero-epitaxial growth on sapphire substrates.
Firstly, the growth of N-polar GaN wires has been reported
using a nitridation step of the sapphire substrate to induce Npolar crystal growth. N-polar GaN wires present a hexagonal
shape with flat top surface (Fig.1.a). Differently, the growth
of Ga-polar prepared sapphire surfaces induces pencil-shape
GaN wires (Fig.1.b) with pyramidal shape at their top. Note
that the vertical m-plane facet formation requires a
combination of lower V/III ratio and highly reduced V-III
precursor flows, which tend to yield smaller diameters (100200 nm) NWs than for growth on N-polar surfaces.
(a)
(b)
Figure 2 : Effect of polarity on GaN nanostructure shape : a) ordered
wires on N-polar GaN substrates, b) ordered nanopyramids on Gapolar GaN substrates
Figure 1 : (a) N-polar GaN wires on c-sapphire with flat top surface,
(b) Ga-polar GaN wires on c-sapphire with pyramidal tip.
References :
[1] C. Durand, Xiaojun Chen, C. Bougerol, G. Perillat-Merceroz, D. Sam-Giao, B Gayral, B. Amstatt, B. Martin, R. Templier, P. Ferret, and J.
Eymery. « The control of crystal polarity for MOCVD nanowire growth », International Workshop on Nitride Semiconductors 2012, October 1419, 2012, Sapporo, Japan.
64
Strain relaxation mechanims in
nanowire heterostructures
Semiconductor nanowires, Core-shell heterostructures, Misfit relaxation, Dislocation glide
G. Perillat-Merceroz, R. Thierry, P.Ferret, G. Feuillet, and P.H. Jouneau (DSM/INAC)
Strain relaxation processes have been investigated in the case of radial nanowire heterostructures.
Transmission Electron Microscopy observations of multiple quantum core-shell ZnO / ZnMgO heterostrutures
clearly revealed that, like in the case of 2 dimensional strained epitaxy, strain relaxation occurs via glide of
dislocations. But in this case dislocation glide occurs in pyramidal and prismatic planes, in order to release the
interfacial strain on the lateral facets of the wires. It is also apparent that the corners of adjacent facets
introduce inhomogeneous strain fields with specific relaxation processes at these locations.
Wide Band Gap Semi-conducting nanowires are widely being
investigated for their potential applications in the field of
optoelectronics, especially for photovoltaics and light
emitters. For LED applications, the core-shell configuration
(Cf. figure 1), where the light emitting quantum wells are
deposited radially around the initial rod, are especially
interesting mainly because of the greater optically active
volume compared to classical 2 dimensional epitaxial layers.
The use of nanowire based optical devices is commonly
thought to limit strain accommodation problems because of
the limited size of these objects and of the presumed easier
surface relaxation. Nevertheless, as in the case of 2D layers,
we could show that strain relaxation may occur because of
the lattice mismatch stress between the core and the shell
materials. In turns, strain relaxation induces dislocations that
are detrimental to device efficiencies. For instance, we have
demonstrated that internal quantum efficiencies of ZnO coreshell quantum wells varies from 1% to 54% whether
dislocations are present or not [1].
In order to understand and better control strain relaxation
effects in these quasi-1 dimensionnal structures, we carried
out a TEM analysis of plastic relaxation mechanisms in
ZnO/ZnMgO core-shell nanowires grown by MOVPE [2]. It
was shown that strain relaxation along the c and a directions
of the m plane facets of the wurtzite core-shell nanowires
occurred through the glide of dislocation half-loops from the
free surface to the interface. Glide occurs in pyramidal
planes, with the Burgers vectors of the dislocations having a
component both along c and in the c plane, and in prismatic
planes, with b=a. Some of these half-loops may close up in
order to accommodate the misfit at two consecutive opposite
interfaces (ZnO/ZnMgO and ZnMgO/ZnO) of the nanowire
heterostructure. The stress state in these one-dimensional
structures is different from 2D layers, because of the finite
dimensions of the m and a facets. Actually, the presence of
corners between these facets induces inhomogeneous strain
fields in the shell. These observations will help design
optimized core-shell nanowire heterostructures by an
adequate choice of core and layer thicknesses together with
alloy composition in order to avoid the formation of
detrimental defects such as dislocations and so as to control
the strain relaxation mechanisms.
a
b
Figure 1: STEM image
of a) cross section
and b) plan view of a
c-oriented ZnO/
ZnMgO core-shell
nanowire
heterostructure
Figure 2: TEM contrast
study of dislocations in
core-shell ZnO/ZnMgO
nanowire
heterostructures; the
different types of
dislocations appear
visible or invisible
according to the
diffraction vector g,
allowing for the
determination of their
Burgers vector b
Figure 3
a) Stress
representation
in wurzite coreshell nanowires
b) and c)
Activated glide
directions and
glides planes
References :
[1] R. Thierry, G. Perillat Merceroz, P. Ferret, P.H. Jouneau, G. Feuillet ,Core-shell multi-quantum wells in ZnO/ZnMgO nanowires with high
optical efficiencies, Nanotechnology 23(2012) 085705
[2] G. Perillat-Merceroz, R. Thierry, P.H. Jouneau, P. Ferret, G. Feuillet, Strain relaxation by dislocation glide in ZnO/ ZnMgO core-shell
nanowires, Applied Physics Letters, 100, 17, 173102 (2012)
65
NanoImprint Lithography (NIL)
technology for solid state lighting
Research topics: NanoImprint Lithography, 3D pattern, Light extraction, LED
V. REBOUD, S. LANDIS, Y. LEE, E. ROGNIN
Typical light emitting diodes (LEDs) structures are composed of multiple thin semiconducting layers with an
active region capable of emitting light. LEDs efficiency derives from four individual terms: the internal quantum
efficiency; the light extraction efficiency (from the unpackaged chip); the conversion efficiency of the
phosphors; and finally the packaging efficiency. NanoImprint Lithography (NIL) technology has been
implemented in LEDs manufacturing scheme to improve light extraction efficiency and simplify process steps.
Two approaches are currently investigated: patterning the top surface of LED devices with Photonic Crystals
(PhCs) and Patterning the Sapphire Substrates (PSS) on top of which the LED structure layers are grown.
In practice, light extraction efficiency is far from 100%. The
reflection at the LED/air interface can be considered as a loss
mechanism. However, the main loss channel in the LED is
caused by the total internal reflection that occurs when the
light goes from an optically dense medium to a less optically
dense medium (GaN to air for example). For gallium nitrite
based LED, the light extraction efficiency from a flat LED
surface to the air is only 4.61%. Numerous methods have
been proposed to circumvent the poor light extraction
efficiency: modifying the geometry of the chip by shaping the
sides of the LED chip into trapezoidal shape; roughening the
sides and the surface of the chip, which can enhance
scattering effect, leading however to a degradation of the
electrical characteristics, and a rise of the forward voltage.
Based on the aforementioned remarks, the LED lightextraction efficiency can be improved by controlling the
structuration of the LED surface, advantageously achieved by
the NanoImprint lithography with an accurate control of the
critical dimension of surface structures. A uniform size and
the periodicity of the structures (comparable to the light
emitting wavelength), called photonic crystal, allows the
control of the angular diagram of the emitted light [1]. The
photonic crystal [2] enhances the external quantum
efficiency of the LED by controlling light propagating modes
in the LED chip, thereby forcing the generated photons to
exit the chip and/or forming an efficient waveguide to couple
photons to the free space.
curable based resists to pattern LEDs. The conformal
contact during the patterning process was then achieved
reducing dramatically the number of patterning defects.
The nanoimprinted PhCs were then successfully
transferred into a GaN layer on top of 100 mm sapphire
substrate (Fig. 1).
Another major concern of LED manufacturing is the
substrate/GaN interface. The most commonly used
substrate for growing GaN films is sapphire. However, a
16% lattice mismatch between the GaN and the sapphire
exists, causing a large number of threading dislocations at
the interface. An approach to solve this problem requires
altering the crystal growth orientation of the epitaxial film
by resorting to etched sapphire surface structures with
pyramidal
shapes.
The
interrupted
and/or
bent
dislocations might reduce the density of the threading
dislocations. In addition, these surface structures can
display effects similar to the roughened surface to
enhance the light extraction efficiency for the multiple
reflections at the GaN/sapphire interface. Such substrate
patterning is known as the patterned sapphire substrate
(PSS) technique (Fig. 2). Typical PSS patterns are on the
order of micrometers and shrinking the size of the
patterns to nanometer scales is believed to significantly
improve light extraction efficiency, but also to simplify the
buffer layer growth and its thickness. We succeeded in
developing a specific 3D patterning process [3, 4] to tune
the imprinted pyramid shape angle in order to optimize
the optical function response. This approach presents the
unique advantage to transfer 3D structures at wafer scale
in a single step on bowed and warped wafers.
Figure 1: 100 mm GaN layer on Sapphire Substrate fully patterned
with Photonic Crystal.
The typical properties of LED substrates presents high bow,
varying TTV and surface defects after epi-growth.
The
structuration of such substrates is challenging for
conventional lithography methods, especially if high
resolution is requested to realize 2D or 3D periodic and quasi
periodic PhCs. To circumvent these issues, we developed
imprint process based on transparent soft stamp and UV
Figure 2: 100 mm Sapphire Substrate fully covered with reflective
patterns printed by NIL.
[1] V. Reboud, C. M. Sotomayor Torres, Enhancement of extraction efficiency in nanoimprinted optical device structures, Proceedings of SPIE The International Society for Optical Engineering 8428 , art. no. 84280B 2012
[2] V. Reboud, A.Z. Khokhar, B. Sepúlveda, D. Dudek, T. Kehoe, J. Cuffe, N. Kehagias, M. Lira-Cantu, N. Gadegaard, V. Grasso, V. Lambertini,
C.M. Sotomayor Torres, Enhanced light extraction in ITO-free OLEDs using double-sided printed electrodes, Nanoscale, 4 (11), 2012.
[3] S. Landis, V. Reboud, T. Enot, C. Vizioz, Three dimensional on 300mm wafer scale Nano Imprint Lithography processes, Oral presentation,
Micro and Nano Engineering Conference, Toulouse, France, September 16-20, 2012.
[4] V. Reboud and S. Landis, 3D micro-optics patterning by wafer-scale NanoImprint lithography on 8" and 12" wafers, NNT2012 wafer scale
imprint for optical, Invited paper, The 11th International Conference on NanoImprint & Nanoprint Technology, Napa, CA, United States,
October 24-26, 2012.
66
Thermal engineering for LED based
modules
Research topics: LEDs, thermal design, lead-frame
Gasse, B. Pardo , A. Lagrange
ABSTRACT: Thermal design optimization is mandatory to enable mass deployment of LED technology in the
general lighting industry. The first optimization presented here, led to a novel leadframe package, that made it
possible to obtain a junction to slug thermal resistance as low as 5.5 K/W. This work was completed by the
manufacturing of multi-LED modules using the same technology. Thermal characterization was performed to
assess the gain both at component and board level in comparison to standard technologies.
LED based lighting systems are intrinsically much more
efficient and reliable than most other lighting technologies,
but still, only almost 30% of the total energy they use is
converted into visible light and 70% is lost into heat.
Although the internal efficiency of LEDs improves at each
new generation, packaging and system engineers still have to
design solutions that allow for a large dissipation of heat out
of the device junction. Indeed, keeping a low junction
temperature (Tj), has a major impact on the product
performances (LED efficiency, reliability, colour maintenance
...).
Any module or lamp can be described as a network of
thermal resistances in series or in parallel, as shown below.
Within Consumerizing Solid State Lighting program (CSSL),
sponsored by the European Commission (Cf. website:
http://www.consumerizingssl.eu), the consortium goal is to
design and fabricate a cost effective LED based lamp as a
retrofit for the 60W incandescent bulb [1]. LETI’s main focus
was to reduce the junction –to-slug thermal resistance of the
LED package. To do so, together with our partner Boschman,
we designed and fabricated a new leadframe package shown
on Fig. 1.
300-1000 µm
+
n
QW
p
3 µm
+
Au/Sn
100 µm
+
Si doped
5 µm
-
GaN
With a similar thermal characterization, we have obtained a
thermal resistance of 13.7 K/W (total thermal resistance from
junction to board) compared to 19 K/W for a board of
ceramic based components on a Via Filled FR4 board, the
improvements coming from both component and board (see
Fig.3).
Reflective
Polymer
10 µm
10 µm
Figure 2: Multi-LED module using LF packages
Silicone Glue
Reflective
Polymer
200 µm
In addition to encapsulation processes, this work has been
pursued on multi-component modules on Insulated Metal
Substrates (see Fig. 2).
+ Die-Attach (Ag Glue) or Solder
+
Component Lead Frame
Solder Joint SAC
Component Lead Frame
Air
Solder Joint SAC
Figure 1: Novel leadframe package on star PCB and schematic
showing cross section of complete assembly.
This package is based on a thin Cu layer (to improve thermal
dissipation) embedded into a white reflective polymer (to
enhance light extraction).
After die attach and soldering of the package on Star Metal
Core PCB using SnAgCu solder, thermal resistances have
been measured using the T3Ster© equipment. In parallel,
Numerical simulations using Ansys software have been
carried out. A relatively good agreement between simulation
and measurement has been found and is discussed in ref [12]. A thermal resistance as low as 5.5 K/W has been
obtained, which is almost twice as low as commercially
available ceramic submounts.
Figure 3 : Thermal characterizations of multi-LED modules
As to the optimization of the LED properties, other works are
currently in progress, to better control and enhance blue to
white conversion in LED modules and better understand their
failure modes.
References:
[1] B. Pardo et al, Thermal experimental & simulation investigations on new lead frame based LED packages, Imaps 7th European Advanced
Technology Workshop on Micropackaging and Thermal management, La Rochelle, 1-2 february 2012.
[2] B. Pardo, A. Gasse, J. Jakovenko, R.J. Werkhoven, X. Perpiñà, T. van Weelden, P. Bancken, 3th IEEE international conference on Thermal,
Mechanical and Multiphysics Simulation and Experiments in Micro/Nanoelectronics and Systems - EuroSimE 2012 – Portugal, April 16-17-18,
Oral presentation – Selected and accepted to be published in a Special issue of Microelectronics Reliability.
67
8
PhD Degrees Awarded
Audrey BASTARD
Anne Laure BAVENCOVE
Julien BOIZOT
Hélène BOURVON
Stéphane BROCHEN
Clément CHAUVEAU
Alexandros EMBORAS
Frédérique GEMAIN
Vipul GOHRI
Baptiste GOUBAULT
Johannes GOUPY
Siddharth NAMBIAR
Duy Thong NGUYEN
Etienne ROGNIN
68
Audrey BASTARD
University of Grenoble INP (Institut National Polytechnique)
Physical analysis of materials for electrical Phase Change memories (PC Ram)
Phase Change Memories are suitable for the next generation of non-volatiles memories due to high
programmation speed and endurance. However, two major improvements need to be made in order
to enter memories market, the short retention time at high temperature, and the important electric
consumption.
This thesis focuses on the development of new phase change materials to replace the reference
material, Ge2Sb2Te5, unsuitable for embedded memories applications working at high
temperatures.
The behavior of binary compounds GeTe and GeSb has been investigated and compared to the
reference material during both the crystallization of the « as deposited » amorphous and the « melt
quenched » amorphous materials. Indeed it is important to study the « melt quenched » amorphous
state of the material to be as close as possible to the cycled material in the devices. So, the
crystallization mechanism of GeTe checked by the crystallization study of the amorphous « melt
quenched » by laser annealing is in agreement with the in situ TEM observation (thermal annealing)
of the crystallization.
The addition of “doping” elements in the binary compounds has also been performed to improve the
thermal stability of amorphous undoped materials. These “doping” elements allow a current reset
decrease, or a later formation of « voids » during cycling.
Anne Laure BAVENCOVE
University of Grenoble UJF (Université Joseph Fourier)
Fabrication of GaN nanowire-based light emitting diodes
This thesis aims at studying the intrinsic properties of InGaN/GaN nanowires (NWs) in order to
fabricate efficient light emitting diodes (LEDs). Two active region designs, obtained through different
growth techniques, have been extensively investigated. Axial NW-based LEDs emitting from the blue
(450 nm) to the red (620 nm) spectral range have been grown by MBE. In this case, single emitters
present diameters typically smaller than 100 nm. MOCVD allowed the fabrication of LEDs emitting
shorter wavelengths (< 430 nm) from Core/Shell heterostructures with typical dimensions in the
micrometer range. In both cases, the spontaneous growth has been conducted on Silicon (111)
highly conductive substrates in order to inject the current vertically into macroscopically contacted
devices.
Fundamental technological blocks needed to fabricate LEDs have been investigated through a wide
range of characterization techniques suitable for high aspect ratio structures. Thus, the effect of ntype (Silicon) and p-type (Magnesium) dopant species has been assessed thanks to optical
spectroscopy techniques, and these results have been confirmed by electrical measurements carried
out on single wires. Furthermore, low temperature cathodoluminescence has been widely used in
order to study the optical properties of InGaN-based active regions. After technological integration,
electro-optical characterizations resolved at the wirescale have revealed that the performances of
NW-based LEDs are mainly limited by the fluctuation of electrical and optical properties between
single emitters.
69
Julien BOIZOT
University of Paris VI
Lifetime Optimization of Organic Light-Emission Diode microdisplays
The present study deals with active-matrix OLED microdisplays, based on a white top-emitting biemitters structure. The optimization of these devices lifetime is the main point of this manuscript.
The luminous efficiency loss and the voltage drift induced while ageing of the device under constant
current driving conditions are indeed key parameters.
A first part consists in understanding the main degradation mechanisms known to operate in OLED
devices. A focus on intrinsic mechanisms is here chosen to improve devices lifetime. Extrinsic
mechanisms like encapsulation issues or other process optimization are not developed in this work.
We propose here a systematic study on the influence of OLED structure parameters on initial but
especially on aging performances. The optimization of anode electrical contact through plasma
treatments and a thin oxide interlayer show very interesting results for reducing operating bias and
voltage drift induced while aging under constant current. The enhancement of doping percentage in
doped injection layers also show significant improvement on devices performances, with the great
advantage of being a useful tool for controlling devices efficiency. We also find that an optimization
of the emission layers thicknesses could lead to great lifetime improvement. Those results are also
combined and confirmed by a Design Of Experiments meant to determine the influence of the main
process parameters on devices performances. Finally, we initiate the characterization of our OLED
devices using impedance spectroscopy measurements. From the modeling of single-layer structures
to the understanding of simple bipolar devices through analysis of capacitive evolution of full-stack
devices with time, we here show that this technic appears very useful for the understanding of
charge carrier dynamic and could help reducing charge accumulation.
Hélène BOURVON
University of Grenoble INP (Institut National Polytechnique)
Development of Visible and Near-Infrared Light Emitting Diodes and photodetectors
based on Colloidal nanocrystals
This work is devoted to the development of Visible and Near-Infrared Light Emitting Diodes and
photodetectors based on colloidal nanocrystals.
First, we report a wet strategy for solution processed QDLEDs. Using spin coated and inkjet printed
polymers, we have developed our first devices. Unfortunately, their efficiencies were quite low. To
overcome these issues, we have developed an original QDs deposition method based on stamping
and Langmuir processes, called Langmuir-Schaeffer Stamping (LSS). LSS is a very cost-efficient,
dry, simple and homogeneous method to deposit nanocrystals on any substrate, even on small
molecules deposited layers. Using LSS, we have developed efficient devices whose specifications are
reported here. Direct injection and Förster mechanisms are studied.
Furthermore we have fabricated infrared devices with PbS nanocrystals. Performances of our 10 9
Jones NIR photo-detectors and NIR-QDLEDs are presented in this thesis.
70
Stéphane BROCHEN
University of Grenoble INP (Institut National Polytechnique)
Electrical properties of zinc oxide single crystals
Zinc oxide (ZnO) is a II-VI semiconductor which appears as a very promising material for UV optoelectronic applications, in particular for the production of light emitting devices (LED). For these
applications, ZnO presents strong advantages: a high exciton binding energy (60 meV ), a wide
direct band gap (3.4 eV ), the availability of large diameter bulk substrates for homoepitaxial growth
of high quality thin films or nanostructures. However, the development of these applications is
hampered by the difficulty to dope ZnO p-type. The impurity to obtain an electrical conductivity
associated with positive charge carriers (holes), and therefore the production of ZnO pn junctions
have not yet been really identified.
In this thesis we have studied the physical mechanisms that govern the electrical properties of ZnO
single crystal and epilayers. The control of contacts (ohmic or Schottky) on different ZnO surfaces
allowed us to carry out both transport measurements (resistivity and Hall effect) and capacitance
measurements (C(V ), Deep Level Transient Spectroscopy (DLTS) and admittance spectroscopy).
Experimental tools and theoretical models used in this work are described.
We have clarified intentional or unintentional n-type doping mechanisms in ZnO single crystal
samples. We have also identified impurities and growth parameters responsible for the residual ntype doping. This understanding is a crucial and preliminary step for understanding the doping
mechanisms at stake in this material and is also necessary to achieve stable p-type conductivity,
which is still the main challenge for the realization of optoelectronic devices based on ZnO.
Clément CHAUVEAU
University Technology of Troyes (UTT)
Resonators arrays for silicon photonics, applications in wavelength division
multiplexing
The development of the micro-electronics industry has given access to very high data transmission
rates. Currently, these data rates are limited by the electrical interconnection bandwidth and it will
soon be necessary to use optical links to obtain higher data rates. To attain this objective, new
building blocks must be developed such as lasers, modulators, photo-detectors, wave-guides and
routing devices which must all be fully compatible with the CMOS processing.
This doctoral thesis concerns the study and development of new components based on circular
resonators arrays, which other alternative solutions to existing devices in the field of wavelength
division multiplexing for silicon photonics.
The study of single ring resonators over the entire surface of a wafer shows that the use of thermal
regulation is required to compensate for fabrication variations. Results of simulations and
experiments show that arrays of circular resonators allow broadband filtering with very low loss.
Based on this principle, an 8 channel multiplexer is demonstrated conforming to telecoms
specifications. This kind of device is a potential candidate for use in the development of wavelength
division multiplexing in silicon photonics.
71
Alexandros EMBORAS
University Technology of Troyes (UTT)
CMOS Integration of plasmon field effect devices
Compact and low energy consumption integrated optical modulator is urgently required for encoding
information into optical signals. To that respect, the use of plasmon modes to modulate light is of
particular interest when compared to the state of the art silicon based optical modulators, as they
could allow lower operation voltage and energy consumption.
The PhD work of A Emboras tackled to the integration of a Si field-effect plasmonic modulator within
a silicon photonics circuitry using the standard CMOS technology. First, he clarified the material
aspects and the technological sequence required in order to realize an integrated plasmonic
modulator in a way compatible with the requirements of the CMOS technology. In particular, the
Metal-Nitride-Oxide-Semiconductor (MNOS) stack has emerged as a very promising for electrooptical plasmonic devices’design, allowing both an electrically reliable and a low optical loss
operation. The latter performance was amplified by the choice of copper processed by using
standard interconnect patterning technique as a plasmon supporting metal waveguide. The final
modulator took advantage of those developments by inserting this MNOS stack inside a vertical
MNOSM active structure, where the back metal was fabricated by waferscale direct bonding
technique. An efficient and compact (0.5 µm length) optical butt coupling structure was developed
between a standard strip silicon waveguide and the previous plasmonic active stack. A Emboras
demonstrated that such couplers operates with optical coupling loss of just 2.5 dB per facets, ie a
value twice smaller compared to the case of direct coupling. By applying an oscillating voltage to the
very resulting very compact modulator structure (0.5 to 3 μm2), an electro-absorptive plasmonic
modulator was experimentally demonstrated at telecom wavelength. The electro-absorption
modulation is exhibiting a capacitive behavior, and its experimentally measured magnitude is in
agreement with electro-optical simulations. This first fully CMOS compatible experimental
demonstration of a plasmonic modulator is paving the way to future innovative designs.
Frédérique GEMAIN
University of Grenoble INP (Institut National Polytechnique)
Doping Spectroscopic Studies in II-VI Materials
In this thesis work the optical and electrical properties of dopants in HgCdTe, CdZnTe and CdS
layers were studied. These 3 II-VI materials are used in detection devices, for infrared light
detection in the case of HgCdTe and CdZnTe or for visible light detection in the case of CdS. Optical
characterizations of these II-VI layers were carried out mainly by photoluminescence and were
correlated with electrical measurements realized by temperature dependent Hall effect.
HgCdTe layers are the active layers in infrared detectors. A study of intrinsic doping with mercury
vacancies and of extrinsic doping with arsenic in HgCdTe layers, was carried out. Correlations
between optical measurements with electrical measurements led to identification of the activation
energies of the 2 mercury vacancy acceptors. Comparison between PL spectra of As doped HgCdTe
samples with measurements carried out by X-ray absorption (EXAFS) allowed to determine the
arsenic complexes optical signatures in HgCdTe. Besides, a modeling work about alloy disorder in
HgCdTe layers was also carried out to precisely fit the PL spectra.
We also studied CdZnTe substrate used for HgCdTe epitaxy. Comparisons of PL spectra with growth
parameters allowed us to understand the origin of intrinsic defects in this material.
Last, we studied CdS layers, the II-VI material used as transparent window and forming the n side
of p-n junctions in CdS/CdTe in solar cells. In this part, we studied the influence of different
deposition methods on the formation of intrinsic defects measured by photoluminescence. These
measurements were correlated with solar cells efficiency.
72
Vipul GOHRI
University of Grenoble
Development of top-emission Organic Light-Emitting Diodes for high luminance
monochrome and full-colour microdisplay applications
The present work reports the development of high luminance organic light emitting diodes(OLEDs)
device stacks for microdisplay applications. The devices are based on silicon complementary metaloxide semiconductor (CMOS) backplane. In the present treatise efforts are particularly focused on
reducing the luminance decay and the voltage drift during device operation.
In the first part of this study, high brightness and low operating voltage green OLEDs are reported.
The top emitting device stack comprises of fluorescent green emitter accompanied by charge
blocking layers and doped charge transport layers. The effect of different device structures,
configurations and organic materials on the initial and lifetime performance of the device is
presented.
In the second part of the study, device development of hybrid OLED stacks for high luminance full
color microdisplays is reported. The hybrid devices comprise of a solution processed and
photocrosslinkable hole transport layer (X-HTL) and an evaporated white OLED stack. It is shown
that homogeneous thin films of X-HTL can be deposited by optimization of the deposition
parameters. Hybrid device stacks are presented and the results indicate the feasibility of achieving
higher color saturation and efficiency for primary colors than the standard technique which uses
white stack and color filters. Preliminary investigations on the three emitter white stack for
improvement in the efficiency of hybrid OLED are also presented. Finally, the fine pixel level
patterning (resolution  5 μm) of X-HTL is demonstrated.
Baptiste GOUBAULT
University of Lorraine
High density interconnection and 3D integration: study of the mechanical and
electrical contact achieved by micro-tube insertion
In order to address industrial requirements and 3D integration issues, a lot of research is focused on
the main bonding technologies such as reflow soldering, thermo-compression, Direct Bond
Interconnect (DBI), Solid Liquid Inter Diffusion (SLID) and insertion [1, 2]. Over the last few years,
shortcomings caused by planarity, parallelism and roughness defects have been investigated and
mitigated. Additionally, bonding parameters such as high process loads, temperature and low
hybridization speeds are also critical. However, by decreasing the interconnection pitch, the size of
components, and increasing the number of stacked dies, all of these issues have to be revisited.
That is why the room temperature micro-tube insertion technology is a good candidate to address
all of these industrial difficulties. To prove it, this thesis deals with the mechanical, electrical, and
metallurgical behavior study of the golden micro-tube insertion into Al-0.5Cu pads. An experimental
and numerical analysis is carried out in order to determine the main mechanisms involved during
the assembly time. Thus, the mechanical and electrical study of a single micro-tube insertion into an
Al-0.5Cu has demonstrated the loading rate and alignment influence. Moreover, component
assemblies have confirmed previous results and demonstrated the parallelism impact. Finally,
accelerated aging and mechanical tests have proved the good mechanical, electrical and
metallurgical performances of those assemblies.
73
Johannes GOUPY
University of Grenoble INP (Institut National Polytechnique)
Development of microcalorimeters matrix for high resolution spectro-imaging in Xray astronomy
Future of the next camera onboard space observatories implies a major enhancement in number of
pixels and a very low operative temperature (below 0.1 K). In this evolution, the large number of
output wires from the cool detector is often responsible of the most important thermal load onto the
cold bath (cryostat).
In this context, the thermal insulation between the different detection circuits is the bottleneck for
these cameras. An innovative technological component, protected by a patent, has been developed
to tackle this problem. This device has both an excellent electrical resistivity and a very high
thermal resistivity. The proposed solution is a stack of thin superconducting layers at electrical
interconnections.
The thermal resistance at each interface relies on the elastic properties of the materials used, the
quality of the interfaces and temperature. The AMM model used in conjunction with the measured
material characteristics allows a theorical estimation of the thermal resistance per interface. The
measurements undertaken with superconducting connections with very high thermal resistivity are
very well described by this AMM model. We have measured thermal resistances as high as 3.3 105
K/W @ 200 mK for a multilayer of 62 interfaces built with titaniun nitride and niobium alternatively
on a 16 mm2 array. In the conditions foreseen for a 4000 micro-calorimeters camera operating at
50 mK in X-rays, this multilayer technique should allow a thermal load onto the cold bath that is
much lower that 1 mW for more than 8000 contacts.
This device can be used in the future, any time an excellent thermal insulation, associated with an
excellent electrical conduction is necessary.
Siddharth NAMBIAR
University of Grenoble UJF (Université Joseph Fourier)
Plasmon Assisted Si based Electo-Optical Devices
Although the optical properties of nanostructured metals were known for many decades, it is only in
the past few years that this field has attracted wide interest. This is in part due to the progress in
nanofabrication techniques. The field of plasmonics is often touted as the next generation platform
that could interface nanoscale electronics and Si photonics. With current electronic systems nearing
saturation, the migration to photonic systems would become inevitable. Crucial to achieving this
integration is to predict the electromagnetic response of these nanophotonic components.
Electromagnetic numerical tools are one way to understand the optical properties of these plasmon
based nanophotonic components. By and large the thesis deals with numerically analysing the
propagation and near field characteristics of plasmon based components for Si photonics. The two
principal EM modelling tools used in this regard are the boundary element method as well as the
finite difference time domain. Two main kind of active plasmonic active devices were investigated:
integrated modulators, and free space radiation photodetectors. The critical issue of an efficient
coupling of light into a very confined guided plasmonic mode was first investigated so as to isolate
the main modal governing contributions. Next, a new structure of plasmon assisted modulator was
proposed and a complete optical design taking into the technological constraints of a CMOS foundry
is provided and discussed. Finally a design optimizing the radiative coupling to the absorption of a
Ge dot, using a plasmonic dipolar antenna, is studied. In particular the radiative engineering of the
supporting SOI substrate is shown to have a tremendous effect on the final performance of the
device.
74
Duy Thong NGUYEN
University of Grenoble INP (Institut National Polytechnique)
Design, modeling, and characterization of innovative terahertz detectors
This PhD thesis aims to establish an electromagnetic modeling of the bolometer at terahertz
(THz) range that can facilitate the design of the detector from the uncooled infrared bolometer
technology. The envisaged application for the detectors lies in active THz imaging at room
temperature.
We have studied the optical coupling of a THz antenna-coupled bolometer operating in the range
1 – 5 THz. Simulations in receiving and transmitting modes have been performed to study the
optical characteristics of the bolometer. The combination of these two simulation types leads to a
powerful toolset to design terahertz bolometers. For the experimental aspect, measurements
have been performed by using Fourier-transform technique to study experimentally the
electromagnetic behavior of the bolometer. They are measurement of reflectivity of the focal
plane array’s surface and spectral response measurement. The results of measurement were
found to be in good agreement with the simulation. The understanding from the study in this
PhD helps us make improvement to the actual detector. Also the design of bolometer for low
frequency (850 GHz) has been proposed. This leads to a perspective of using bolometer for
terahertz imaging at the frequency where many characteristic of the terahertz radiation are
favourable for imaging application.
Etienne ROGNIN
University of Grenoble INP (Institut National Polytechnique)
Characterization and applications of flowing nanoimprinted thin polymer films
This thesis presents a theoretical and experimental work on nanoscale flows of polymer melts.
Different levelling dynamics emerge from the analytical and numerical study of the reflow of a
polymer film that is first nanoimprinted and then annealed above its glass transition
temperature, depending on the initial topography of the film. These concepts were applied to the
manufacturing of optical devices from the reflow of complex nanostructures. A method to
measure the Newtonian viscosity and the terminal relaxation time of a thin polymer film was
also developed. Finally, an exploratory work on a residual-layer-free nanoimprint process based
on de-wetting is presented. Emphasis was put on the accurate computation of the disjoining
pressure in stratified media with the modern Lifshitz theory based on the optical properties of
the interacting materials.
75
76
Thanks to…
Editorial committee
Alexei TCHELNOKOV
Gérard DESTEFANIS
Jean Marc FEDELI
Guy FEUILLET
Jean MARTY
Graphic Art
Valérie LASSABLIERE
Hélène VATOUYAS