View
221
Download
0
Category
Preview:
Citation preview
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
1/16
UV MET L SEMICONDUCTOR MET L
DETECTORS
A robust choice or Al, Ga)N based detectors
J-L. REVERCHON \ M. MOSCA
1
N. GRANDJEAN
2
F. OMNES
2
F. SEMOND
2
, J-Y.
DUBOZ
2
,
and
L
HIRSCH
3
1
Thales Research Technology, 91404 Orsay Cedex, France
2
CRHEA-CNRS, rue Bernard Gregory, Sophia Antipolis, 06560 Valbonne, France
3
IXL-CNRS-ENSEIRB, University o Bordeaux I, 33405 Talence, France
Abstract:
UV
detection is interesting for combustion optimization, air contamination
control, fire and solar blind rocket launching detection. Most o these applica
tions require that UV detectors have a huge dynamic response between
UV
and the visible, and a very low dark current in the range o the
UV
flux meas
ured. (Al,Ga)N alloys present a large direct bandgap in this range and there
fore can be used as an active region in such detectors. To take advantage o the
large Schottky barrier, the good alloy quality, and to avoid any doping prob
lems, we have developed MSM photodetectors. High quality material has been
grown with MOCVD and MBE on sapphire substrates. Stress management is
employed for aluminum contents up to 65 to reduce crack density. This is
correlated with non-ideal features like dark current, sub-bandgap response and
non-linearity between photocurrent and optical flux. The spectral selectivity
between UV and visible reaches five orders o magnitude. A geometry o in
ter-digitized fingers is optimized in regards to the peak response. The Schottky
barrier and a dielectric passivation result in dark currents lower than 1
fA
up to
30 V for a 100 x 100 1m
2
pixel. Consequently, detectivity is mainly limited by
shot noise and corresponds to a noise o 500 photons per second and per pixel.
Key words: UV solar blind detectors, Metal-Semiconductor-Metal detectors, stress man
agement in (Al,Ga)N, IBICC.
77
M.S. Shur and A. Zukauskas eds.), UV Solid-State Light Emitters and Detectors, 77-92.
© 2004 Kluwer Academic Publishers.
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
2/16
8 J
L
everchon et l
1
INTEREST AND SPECIFICATIONS FOR Al,Ga)N
S N
ACTIVE LAYER FOR UV DETECTION
1 1 Al,Ga)N for
UV
Detection
Due to their use for blue LEDs and lasers, GaN based materials are gaining
more and more importance and a great effort has been undertaken in order to
improve their quality. As a result, nitrides can now be considered for many
other applications such
as
high power-high frequency electronics and ultra
violet UV) detection [1,2]. As a direct band gap III-V semiconductor,
Al,Ga)N is well suited for detecting light at energies higher than its band
gap energy and providing a large rejection at lower energies. The band gap
energy varying from 3.43 eV GaN) to 6.2 eV AlN), makes it possible to
adjust the wavelength of absorption from 360 nm to 195 nm. In particular,
we will focus on wavelengths of about 280 nm for which sunlight is ab
sorbed by the ozone layer and never reaches the surface
of
the earth. Conse
quently, detectors sensitive in this range see only UV sources coming from
the earth and are said to be solar blind.
Fundamentally Al,Ga)N based devices suffer from difficulties such as a
large activation energy required not only for magnesium
p
doping but also
for n doping in high aluminum content alloys. For the same reasons, ohmic
contacts are also difficult to achieve. On the contrary, this large barrier gives
the opportunity to achieve high barrier Schottky contacts. This
is
a great ad
vantage for obtaining
of
low dark current in Schottky based detectors. Fi
nally the main difficulties come from the quasi absence
of
GaN or AlN sub
strates. Nitrides are traditionally grown on sapphire or SiC with a lattice and
thermal expansion mismatch inducing strain, dislocations and cracks. In Sec
tion 2 we will discuss how to avoid cracks and to reduce the non-ideal fea
tures attributed to related electrical defects.
1 2
Specifications for UV Detection
UV
is
in the range of energy involved in chemical bonding. Thus, UV detec
tion presents a great interest for combustion optimization, air contamination
control, UV A/UVB medical control, and fire/flame detection and in particu
lar solar blind detection. Most
of
these applications require stringent specifi
cations because of the low fluxes to be measured. Indeed UV radiation is
diffused by the Rayleigh mechanism especially when UV sources are far
away in the atmosphere. As a consequence, dark current must be as small as
possible in comparison to photocurrent. Moreover, as far as noise
is
con-
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
3/16
UV
Metal Semiconductor Metal etectors
79
cerned, a low current would diminish the shot noise and the lifnoise, which
are respectively proportional to current and current squared.
In this paper, we present Al,Ga)N based detectors which are in competi
tion with photomultipliers PM) and silicon based charge coupled devices
CCD). PMs are not available in large array configuration, are fragile, and
use a high bias. Large array CCDs are available with a huge detectivity and
even with photon counting mode but only when cooled to reduce dark cur
rent. One
o
the advantages
o
Al,Ga)N based detectors versus PM and
ceo would be the intrinsic spectral selectivity between uv and visible.
t
prevents use
o
interference filters whose sensitivity to non-normal incidence
is a drawback. In the case
o
Al,Ga)N based detectors, such interference
filters may be added to the intrinsic selectivity to obtain even larger rejec
tion.
Obviously, we require from photodetectors a responsivity as large
as
pos
sible. t means that gain the ratio between electron pairs created per photon
absorbed) may be close to one in the case
o
photovoltaic detectors and as
large
as
possible in the case
o
photoconductor or phototransistor structures.
Moreover, the proportionality between photocurrent and incident power
linearity), must be preserved. Concerning the response time, a short one
may be expected due to the low capacitance and transit time
o
device [3].
Capacitance can be estimated to be lower than
0 1
pF for 100 x 100
f m
de
vices. Nevertheless, because
o
the need for large detectivity in imaging with
low fluxes, a long integration time is necessary. Thus the time response is
not so important and needs only be reasonably fast for imaging at several
hundreds
o
hertz.
1 3 The Choice o Metal Semiconductor Metal Detectors
A first kind
o
semiconductor detectors is the photoconductor that may show
a high internal gain defined
as
the ratio
o
lifetime
o
carriers to transit time
between electrodes. As we will see in Section 1.5, the lifetime depends on
density and occupancy
o
deep levels, so that non-ideal and uncontrolled
behaviors may appear like, e.g. a non-linear dependency o the photocurrent
on the incident power. Moreover, photoconductors present an intrinsic high
dark current leading to lifnoise. Therefore, photoconductors are not suitable
as flame detectors in terms
o
dark current, noise and detectivity.
Al,Ga)N
p i n
photodiodes do not exhibit the above-mentioned draw
backs. The gain is limited to one, the current
is
low and the responsivity is
linear. But p-type doping with high Al content is difficult to obtain. Some
attempts to use p-GaN show that long wavelength contribution could be lim
ited by convenient band diagram design [4]. Another difficulty is to make
good ohmic contacts on such wide bandgap semiconductors even i they are
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
4/16
80
J- L Reverchon et a
highly doped. Generally, a high-temperature annealing cycle (700-900 °C)
is necessary to make the contact ohmic. For this reason, together with the
need for a good
n-type
conducting layer, good Schottky photodiodes are not
so easy to achieve at high aluminum content.
Consequently, the situation is far more convenient for (Schottky)
Metal-
Semiconductor-(Schottky) Metal (MSM) detectors that
don t
need either
doping or ohmic contacts. The only but important difficulty is a mismatch
between the substrates and (Al,Ga)N layers. A MSM consists only
of a
photoabsorbing epitaxial layer with two interdigitated Schottky metal con
tacts deposited on the semiconductor surface. In this paper, we will focus on
this planar technology whose simplicity contributes to robustness.
1 4 Electrical and Optical Characterization Tools
In most detectors, dark current is measured with a picoammeter ( 485, Keith
ley), but when necessary, dark current is measured with a source/meter
(6430, Keithley) in the fA range taking care
of
the connections (Guarded
Tri-axial Cable). For the photoresponse measurement, we use a Xenon lamp
filtered by a monochromator and the light is focused on the back side of the
detector for samples grown on
Ab0
3
,
and on the front side for samples
grown on
Si(lll
. The incident power is measured by a calibrated pyrome
ter. The detectors are biased with a voltage source and connected in series
with a transimpedance amplifier. The photocurrent is measured both in AC
conditions with a chopper and a lock-in amplifier (7220, EG G Instru
ments) and in DC conditions with a picoammeter (485, Keithley).
There-
sponsivity is calculated as the ratio of the photocurrent to the power incident
on the detector. All measurements are made at room temperature.
1 5 Non Ideal Features
n
MSM
Due to dislocations or cracks, some layers may present defects that are elec
trically active and lead to traps or recombination centers. For MSM based on
such material, the high quantity of defects and deep levels gives poor rectify
ing contacts. These levels give both channels across the junction and a bow
ing
of
the conduction band that diminish the depletion thickness at the
Schottky barrier. Finally, this injection via trap-assisted tunneling corre
sponds to a photoconductive behavior. But, in photoconductors, responsivity
depends on the lifetime
of
carriers. This lifetime has been linked in many
ways to traps or deep levels [5,6,7,8].
t
results in a strong nonlinearity
of
photoresponse versus absorbed optical flux. These spectra also present a sub
band-gap absorption and a reduced dynamics depending strongly on fre
quency when spectra are acquired with a chopped flux.
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
5/16
VMetal Semiconductor Metal etectors
8
2 MATERIAL GROWTH
We present here the structure used for back illuminated samples and the
conditions
o
growth by MOVPE and MBE. The choice
o
nucleation layer
and the efforts to eliminate cracks will be particularly stressed.
2 1 Sample Structures for UV Detection
One o the main difficulties encountered in the growth
o
nitride materials
has been the absence
o a lattice matched substrate. In the case o large array
detectors, we have also to take into account that a Readout Integrated Circuit
ROIC) on the front side obliges us to use a substrate transparent to UV.
Thus, even
i
GaN, AlN substrates or pseudo substrates have been improved
during the last few years, GaN ELOG Epitaxial Lateral OverGrowth)
[9]
or
f. ELOG [10] and bulk GaN [11] cannot be used. On the other hand HYPE
[12] or bulk AIN [13] substrate may be adapted to UV detection i a good
transparency to UV is guaranteed. Up to now, sapphire is still the substrate
o choice. The choice o the nucleation layer must provide good optical and
electrical qualities for Al,Ga)N. As far as optical properties are concerned,
an AlN buffer layer is the only solution to provide transparency at 280 nm.
GaN buffer layer can be used only
i
its thickness is sufficiently low to guar-
anty transparency to UV. After the buffer growth, cracking may arise from
the lattice mismatch between Al,Ga)N and sapphire and also between
Al,Ga)N layers with very different aluminum contents. Thus, one o the
greatest challenges is to manage this mismatch whereas we have to use lay-
ers as thick as possible to minimize dislocations. In our case, we use a thick
window layer o 1
f. m
transparent to UV to improve materials quality. Then
the active layer is grown with a thickness o 0.4
f. m
, sufficient to easily col-
lect carriers.
Figure 1 Left: cross section o sample structures for UV detection. ROIC is on the front side
and light comes from backside. Right: overview
o
interdigitized fingers
o
a MSM.
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
6/16
82
J
L
R
everchon
t a
2.
2 Sa
mples Gr
own by L
ow Pres
sure Meta
lorganic
Vapour
Phase Ep
itaxy
So
me sample
s are grow
n by low-p
ressure me
talorganic v
apour-phas
e epi
taxy (LP-M
OVPE) on
c-sapphire
substrates
in an Aix t
ron growth
chamber
AIX200 R
F. Trimeth
ylgallium,
trimethylalu
minum, an
d ammonia
are used
as
precursors.
GaN or AI
N buffer la
yers are 25
-nm and 10
-nm thick a
nd are
gr
own at 525
oc
and 8
90 °C, resp
ectively, in
a pure nit
rogen carri
er gas.
(
Al,Ga)N al
loys are gro
wn at 118
0°C with a
V/III ratio
between 20
00 and
310
0 in a pure
hydrogen c
arrier gas. N
H
3
flux is
2 /min and
the total fl
ux
is
5 1
/min . The g
rowth press
ure is low (
20 mbar) in
order to a
void parasit
ic re
actions between N H
3
and TMAI. Finally, the growth rate is 1
f Lm/h
for th e
window
layer Alo.6sGao
35
N) a
nd
1
.8 flm
/h for the ac
tive layer (A
l
0
.
5
Ga
0
.
5
N)
.
More
details are
given in R
ef.
14
.
We now pay
attention to
layers gro
wn
with
a GaN bu
ffer layer. W
e notice a
strong sub
band gap a
bsorption c
orre
spondin
g to deep l
evels (Fig .
2, left) eve
n if no crac
k networks
are presen
t.
All dev
ices grown
on this la y
er present h
igh dark cu
rrent with n
on-ideal fe
a
tu res of
photoco
nductors a
lready men
tioned. Fo
r example,
we notice
in
Fig. 2 (rig
ht) that the
dynamics c
an be reduc
ed when hi
gh bias is a
pplied and
participa
te to trap-a
ssisted tunn
eling across contacts. The frequency depend
ence als
o shows th
e long tim e
needed to n
eutralize su
b-band-gap
absorption
10
4
.8
i OHz
.
20V
Q
0
.6
/)
.
>.
10
2
· 8 Hz
.
2
V
/
)
E
0.4
·
, · . : · - - 80Hz 2V
/)
\ •
/)
Buffer GaN
c
10°
'
_
c
0
'
··
tl
0.2
I
- Buffer
AIN ..
Buffer GaN
a.
1-
uffer
AIN
/)
. . . ,
Q
1o
2
..
.0
0
:::
200
500
300
400
Waveleng
th (nm)
Waveleng
th (nm)
Figure 2
Left: tra
nsm ission spe
ctra for layers
g rown on a G
aN or AlN bu
ffer. Right: sp
ectral
response de
pending on bi
as and frequen
cy for sample
with GaN bu
ffer layer.
O
n the contr
ary, the bes
t samples h
ave been o
btained wit
h an AlN bu
ffer
lay
er. The tra
nsmission
is good dow
n to 280
nm showin
g the absen
ce of
d
eep levels.
This is con
firmed by t
he dynamic
s independ
ent of bias
and AC
or DC mo
de used for
spectral a
cquisition (
Figure 3).
Then, no d
eep level
con
tribute to s
ub-band-ga
p absorptio
n or trap a
ssisted tun
neling acro
ss the
Sc
hottky barr
ier. Conseq
uently, the
time respon
se is due o
nly to the
transit
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
7/16
VMetal Semiconductor Metal Detectors
83
time needed for carriers to cross the spacing between electrodes and non
ideal features disappear. We must stress here that some good results have
been observed with the layers grown with a GaN buffer layer in the past
[14]. We don't have any clear explanation for these differences. We can only
mention that materials quality has been shown to depend closely on growth
parameters and that the average aluminum content is closer to AlN than to
GaN in such layers.
10'
2
.
,.
.
i
10
3
OHz SV
......
190
Hz
/5V
>
80Hz/5V
·:;;
10
4
- 80Hz/20V
/)
r::
l 4(1.1 N
UA
~ u n
Al
la
oH ~ · m
0
c.
10 '
5
/)
Q)
0::::
10'
6
lluO
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
8/16
84
J L
Reverchori et
a
expansion consequences.
In
this way,
we
obtain layers without any cracks
and with exceptional electrical and optical properties.
0.8
I
10-
2
0.6
- I
t\
1
_., .;) o
a 1 1m
..
>- 10·3
Ill
::J
\ l
,..,
(fil
lc
N
im
>
/)
iii
0.4
3
i\IN
IUilnm
c
10
4
iii
0
/)
t.l nm
Q
0.2
·
/)
10-
5
::J
\apphin:
Q)
cr:
0.0
250300350400450500
Wavelength nm)
Figure
4 Transmission and response spectra of the layer grown by MBE with a GaN buffer
layer.
3 OPTIMIZATION OF PROCESSING
3 1 Surface Preparation and Metallization
Most MSM detectors were processed for defining interdigitated fingers by
optical lithography. The spacing equals 2 or 5 11m whereas the width varies
from 1 to 10 f.lm. The surface was deoxidized in HCl for one minute and
rinsed in de-ionized water during four minutes just before being introduced
into the Joule evaporation chamber (Plassys chamber MEB550S). The con
tact consists
of
10 nm
of
platinum followed by 100 nm
of
gold. Even
if
we
take care to limit time between cleaning and deposition, we can expect
(Al,Ga)N to be oxidized. Some studies have shown that an oxide could pre
vent leakage via dislocations. For example, some enhancement
of
Schottky
barrier height on (Al,Ga)N/GaN heterostructures by oxidation annealing has
also been reported [ 18]. It may explain the exceptionally low dark current
low obtained with some samples
(1
fA up to 35 V). Even
if
oxide presence
has not been investigated here,
we
have observed that a smooth etching
just
before deposition could increase leakage via induced defects and oxide
elimination. After lift off, annealing at 400 °C during 10 minutes in nitrogen
atmosphere is used only for mechanical requirements. Higher temperatures
would induce leakage currents.
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
9/16
VMetal Semiconductor Metal Detectors
8
3.2 Contact Passivation and Connection
If
we want to perform a fast evaluation
of
epilayers, we can use a one-step
processing for which contact pads are evaporated at the same time
as
elec
trodes. But in the case
of
a backside-illuminated device, we observed that
both the interdigitated electrodes and the contact pad areas contribute to the
overall photocurrent even if contact pads are placed at several tens
of
mi
crometers from the interdigitated area. In order to avoid the parasitic current
due to the contact pads, we developed another process where the Pt/Au
Schottky contacts are deposited on the (Al,Ga)N surface whereas the contact
pads are sputtered on a dielectric. Several dielectric films were tested for
their electrical passivation capability.
Si
3
N
4
(300 nm) and
Si
2
(300 nm),
were deposited by plasma-enhanced chemical vapor deposition (PECVD) at
300 °C. Benzocyclobutene (BCB) (1500 nm) was deposited by spin coating
and annealed under vacuum at 250 o for 30 min. The first two PECVD ma
terials show good passivation up to
fA
range at several tens
of
volts. Passiva
tion has been particularly efficient in the case
of
layers having developed
microcracks related to excess stress. In that case, we showed that both the
dark current and the responsivity strongly depend on the crack density. By
using our two-level process, we have reduced the parasitic effects
of
cracks
on the dark current.
4 TRANSPORT PROPERTIES
As we can see in Fig. 4, responsivity is limited to 0.04 A/W. This value is
.low compared to the absolute photovoltaic limit
e h v
that would be
0.22A /W at 280 nm. Indeed, MSM detectors have been fabricated by many
groups on GaN [19,20] or on (Al,Ga)N [3,14,21,22] and exhibit good per
formance but with the same limited collection
of
carriers between fingers.
Collection efficiency in MSM detectors
is
studied here with submicronic
lithography, the ion beam induced charge collection method (IBICC), and
numerical 2-dimensional calculations of the electric field distribution.
4.1 Submicronic MSM byE-beam Lithography
A way to improve collection
of
carriers
is
to reduce spacing and trapping
between the electrodes. Here, we will study the effects
of
spacing on both
the spectral response and the absolute value
of
the photoresponse. We com
pare sub-micron devices obtained by electron beam lithography (the width
equals 1 lm, and the spacing
is
0.6 lm) to interdigitized fingers defined by
optical lithography (the finger width and spacing equal to 2 lm) in terms
of
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
10/16
86 J L
Reverchon
eta
responsivity and spectral selectivity. Exceptionally, the (Al,Ga)N layers are
grown on a Si lll) substrate. Pt/Au Schottky contacts are evaporated and
lifted off. The dark currents are in the pA range for biases up to
10
V and 50
V for 0.6 f.lm and
2 11m
spacing, respectively.
The spectral responses of two different detectors are shown in Fig. 5
(left). The cut-off wavelength is 280 nm, with a 3 decades rejection ratio
between 280 and 300 nm for both spacings. For the
2 f..lm
MSM, the re
sponse presents a plateau from 300 to 365 nm corresponding to the GaN
layer that is grown underneath the active (Al,Ga)N region [21]. This compo
nent, not present in de measurements is due to a capacitive coupling between
the 2-D electron gas at the AlN/GaN interface and the electrodes. At a posi
tive bias
of
40 V, the responsivity is 0.044 A/W corresponding to a 20
quantum efficiency. As far as the
0.6 f.lm
MSM is concerned, the response
decreases more regularly, without any plateau, and shows an overall better
rejection of near-UV light. Thus, the parasitic response in the underlying
GaN layer is largely reduced for the applied de field and for the ac photo
voltage. This is due to a reduced coupling between the GaN layer and the
electrodes when the finger spacing is reduced.
\ I
3 ~
.i: 0.20
I :2
• 0 10 • .
' .
0
..
0:: 0 20 40 60
Bias )
280 320 360 400
Lamnda (nm)
Figure
5 Left: response spectra with 2
J.tm
and 0.6 J.tm spacing MSM; responsivity versus
bias is given in the inset. Right: contours of equi-values
of
electric field found from the 2
MSM geometries.
The variation
of
responses with bias is shown in the inset
of
Fig. 5 (left).
We verify that the dark current at a given bias generally varies as the inverse
of
the finger spacing, although deviations from this law can be seen. The
responsivity increases first sub-linearly ( ·
7
)
and then linearly with bias.
The knee at about 40 V for the 2 11m MSM and
10
V for the
0.6 f.lm
MSM
corresponds to a transition from photovoltaic to photoconductive behavior
for which the contacts start to inject current. We note that the responsivity is
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
11/16
VMetal Semiconductor Metal etectors
87
much larger in the 0.6-flm MSM than in the 2- lm one at the same bias, or
reaches a given responsivity value at a much lower bias.
4.2 IBICC Measurements
We now present experiments based on IBICC measurements on MSM fabri
cated on the same layer on Si lll). IBICC measurements consist offocusing
a 2 MeV
4
He+ microbeam down to a 1 11m
2
spot size with a low flux of less
than 400 ions per second. Ions are absorbed in the crystal and create about
10
5
electron-hole pairs per ion. One electrode (called anode) is grounded
while a negative bias is applied on the other electrode (cathode). For each
incident ion, a signal was obtained, with the pulse height proportional to the
number
of
collected charges. More details of the experimental procedure can
be found in Ref. 23. Figure 6 shows maps
of
collected charges at 75 V. In
Fig. 6 (left), we have selected the events that give rise to a small charge per
ion. We observe that these events are located at the edges
of
the anode. In
Fig. 6 (middle), we have selected the events that give rise to a large charge
per ion. These events are now located close to the cathode in the Schottky
depletion region. Regions in between fingers give rise to a moderate collec
tion. The collection efficiency is given
as
a function
of
position for different
voltages from 0 to 75 V in Fig. 6 (right). On the anode edges, the collection
efficiency increases rapidly with bias up to 30 V, and then remains almost
independent of bias. On the cathode the collection efficiency is increasing
with bias, and is almost flat below the electrode. As far as the region be
tween the electrodes
is
concerned, the decrease of the current when moving
away from the cathode presents an attenuation length of 5 f tm It is a typical
length for minority carriers already found on EBIC measurements [24] .
20
r IS
:
u 10
a Ill IS 20 B
O
d p
Figure
6
IBICC response at anode (left) and cathode (middle). Response is plotted versus
bias on the right-hand side.
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
12/16
88
J
L
Reverchon et al
Some precautions have to be taken before extending these IBICC results to
UV MSM detectors. For instance, the detector
is
uniformly illuminated by
photons whereas the beam is focused in 1 Jlm
• Nevertheless, we can think
that below the cathode, carriers are created in the depletion region so that
holes are easily collected even at low bias. Electrons drift towards the anode
where they are collected after the screening
of
the build-up field
of
the
Schottky diode.
4 3 Electric Field Calculation
In order to explain these results, we performed a 2-dimensional calculation
of the applied electric field in the structure using a commercial 2-D solver
Atlas-Silvaco . Parameters used for this calculation are described elsewhere
[21]. Figure 5 right) shows the distribution ofth electric field in the direc
tion perpendicular to fingers in the 2-Jlm and 0.6-Jlm MSMs for a bias
of
15
V. The comparison clearly shows that the high field region extends through
the whole spacing between fingers in the 0.6-Jlm MSM whereas it remains
confined to the electrode edge in the 2-Jlm MSM. t also shows that the ver
tical extension
of
the high field region is reduced when the spacing between
fingers is reduced. The calculated field distribution thus explains the larger
response and the reduced coupling to the GaN layer when the spacing is lim
ited.
In order to calculate the response value from the field distribution, we
made the assumption that electron-hole pairs are collected in high-field re
gions only. The high-field criterion was the following: Al,Ga)N alloys show
some localization with a typical energy
of
50 me V on a spatial scale
of
50
nm [21, 25]. Then, a field higher than
10
kV/cm is needed to collect carriers.
For a front side illumination, photons above band edge are absorbed in the
first 0.2
Jlm,
and the volume
of
the high-field region is just proportional to
the lateral extension
of
the high-field region beside fingers that are not trans
parent to UV. Because
of
a slight dependence on the structure parameters
such
as
doping or finger spacing, or on the field value used to define the high
field region, the response depends on bias
as
v with yin the range
of
0.65 to
0.72. The inset of Fig. 5 shows the calculated response as a function of bias
for both MSM. t varies as ·
7
and the absolute value is close to the meas
ured one up to biases where internal gain starts to appear. This variation is
intermediate between the extension
of
the depletion region in a vertical
Schottky diode r= 0.5) and the linear response y= 1 of a photoconductor
with an uniform field assumed.
We describe IBICC results with the same kind of hypotheses and simula
tions
as
those previously used. Incident ions are absorbed in the Al,Ga)N
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
13/16
V
Metal Semiconductor Metal etectors
89
layer on a scale that is larger than the layer thickness, so that we can con
sider that the electron-hole pair generation is uniform in the vertical direc
tion. Electrons and holes are efficiently separated where the field is high
enough to overcome localization [21,23]. At the cathode, the high field sepa
rates carriers, and holes are all the more easily collected since the distance to
travel is small. Electrons are swept towards the anode, once the build-in field
of
the Schottky diode is screened. When the bias increases, the high-field
region extends below the cathode and separates more and more electron
hole pairs.
4 4 Conclusion for the Geometry
o
UV Detectors
From IBICC studies, we have shown that it is interesting to increase the
cathode area. As far
as
the region between the electrodes is concerned, sub
micronic studies have shown that spacing between the fingers must be as
short as possible. For example, we can see in Fig. 7 (left) that the responsiv
ity increases with cathode area for a constant area and spacing between the
fingers. We see also this tendency in Fig. 7 (right) with a larger responsivity
for a lower spacing and larger area.
~ 5
>
u
~ 10
·u
E
w 5
0 10 20 30
Bias voltage (V)
~ 20
[ 15
c
10
E
w 5
0 10 20 30
Bias voltage (V)
Figure 7 Responsivity versus bias for different electrode area and spacing in MOCVD sam
ple (left) and MBE sample (right).
5 PERFORMANCES AND CONCLUSION
MSM detectors benefit from the large band-gap and Schottky barrier
of
high
quality undoped materials. The most impressive performance is the dark cur
rents that are still in the femtoampere range at
35 V
We couldn t measure
noise in the best samples. Thus we estimated shot noise, Johnson noise and
1/fnoise corresponding to this dark current. A conservative assumption for
the constant
of fJP f
noise ( 5 x 1-
5
)
shows that noise is dominated by shot
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
14/16
90
L everchon et a l
noise in the fA range. Then we obtain a detectivity
of
4xl
14
w
1
with a
frame rate
of
100 Hz. It corresponds to an equivalent power
of
2.5
fW
or 500
photons/second per pixel of 100 x 100
j.lm
2
• In our case, capacity is not
measured but may be estimated to
10
fF for 100 x 100
11m
pixels. MSM are
also well suited to work at high frequency. Furthermore, we can stress the
advantage of Al,Ga)N based devices which is the intrinsic selectivity be
tween UV and visible close to five orders of magnitude. We notice that this
dynamics of UV/visible is obtained without any antireflection coating that
would improve both the peak responsivity and dynamics.
So, we have shown that MSM photodiodes present all of the desirable at
tributes of a flame detector: fabrication simplicity, robustness, large
UV/visible rejection, high sensitivity, high speed, low dark current, low
noise, high detectivity. Theses performances approach the ones of photo
multipliers PM) and the best cooled charge coupled devices CCD). Now a
new challenge is to design a Readout Integrated Circuits capable
of
reading
1
fA
with an optimal collection of carriers at 10
V.
Risks of breakdown
in
circuits designed on a small area are important. In a first time, it may be eas
ier to find circuits for large linear array.
If we compare MSM to Al,Ga)N-based Schottky or p i n photodiodes,
we observe that spectral selectivity of
4 orders
of
magnitude has been
achieved between UV and visible with an excellent detectivity [26,27,28,29].
The latter devices require a low voltage which is an advantage to adapt to
standard ROIC. On the contrary, it is more difficult to achieve dark currents
as low as those of MSM owing to the mesa processing and remaining mate
rial difficulties dark current in the
nA
or pA range are typical). Conse
quently, different detectors may be adapted to different kinds of applications:
MSM for extremely low fluxes for which very low dark current is required
fA), and Schottky or p i n photodiodes for larger ones pA). In all cases,
the key reason for choosing Al,Ga)N-based device would be the spectral
selectivity between UV and visible light.
CKNOWLEDGEMENTS
This work was partially supported by DGA contract N° 00-34-068). One
author MM) wishes to acknowledge financial support from a Curie Re
search Grant G5TR-CT-2001-00064). Thanks are due to
R.
Me Kinnon
NRC) for numerical simulations and ONERA for technical support.
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
15/16
VMetal Semiconductor Metal Detectors
91
REFEREN ES
1 B
Gil, Group III Nitride Semiconductor Compounds Physics and applications
(Clarendon Press, Oxford, 1998).
2
Y A
Goldberg, Semicond. Sci. Teclmol. 14, R41 (1999).
3 E
Momoy,
F
Calle,
E
Mufioz, and
F
Omnes, Appl. Phys. Lett. 74,3401 (1999).
4. C.Pemot,
A
Hirano,
M
Iwaya,
T
Detchprohm,
H
Amano, and
I
Akasaki, Jpn.
J
Appl. Phys. Part 2, 39, L387 (2000).
5
F
Binet, J.Y. Duboz,
E
Rosencher,
F
Scholz, and
V
Hlirle, Appl. Phys. Lett. 69,
1202 (1996).
6 E Mufioz, E Monroy, J.A. Garrido, I Izpura, F.J Sanchez, M A Sanchez-Garcia,
E
Callera, B Beaumont, and P. Gibart, Appl. Phys. Lett. 71, 870 (1997).
7
J
A
Garrido,
E
Monroy,
I
Izpura, and
E
Mufioz, Semicond. Sci. Techno . 13,
563
(1998).
8
J
L
Reverchon,
M P
Poisson, and J
Y
Duboz, Semicond. Sci. Teclmol. 16, 720
(2001).
9 B
Beaumont and
P
Gibart, Proc. SPIE 3725, 2 (1999).
10 B
Beaumont,
J P
Faurie,
E
Frayssinet,
E
Aujol, and
P
Gibard, this volume.
11
V
Kirchner,
H
Heinke, D Hommel, J
Z
Domagala, and
M
Leszczynski, Appl.
Phys. Lett. 77, 1434 (2000).
12 Nikolaev,
I
Nikitina,
A
Zubrilov,
M
Mynbaeva,
Y
Melnik, and
V
Dmitriev,
MRS Internet J Nitride Semicond. Res. 581, W6.5 (2000).
13
L
J. Schowalter, Y Shusterman, R Wang, I Bhat,
G
Arunmozhi, and
G A
Slack,
Appl. Phys. Lett. 76, 985 (2000).
14
F
Omnes,
N
Marenco, B. Beaumont, Ph. De Mierry,
E
Momoy,
F
Calle, and
E
Mufioz, J Appl. Phys. 86, 5286 (1999).
15 F
Semond,
P
Lorenzini, N. Grandjean, and
J
Massies, Appl. Phys. Lett. 78, 335
(2001).
16 N Grandjean, J Massies,
P
Vennegues,
M
Leroux,
F
Demongeot,
M
Renucci,
and J Frandon, J Appl. Phys. 83, 1379 (1997).
17
M
Mosca, J.- L Reverchon, N Grandjean,
F
Omnes, J.-Y. Duboz, I Ribet, and
M
Tauvy, Mat. Res. Soc. Symp. Proc. Vol 764, Material Research Society (2003).
18
C M
Jeon and J.-
L
Lee, Appl. Phys. Lett. 82, 4301 (2003).
19 J C Carrano, T Li, D. L Brown, P
A
Grudowski, C J. Eiting, R D Dupuis, and
J C Campbell, Appl. Phys. Lett. 73,2405 (1998).
20. S
W
Seo, K K Lee, Sangbeom Kang, S Huang, William
A
Doolittle, N.
M
Jok
erst, and
A
S Brown, Appl. Phys. Lett. 79, 1372 (2001).
21. J
Y
Duboz, J L Reverchon, D. Adam, B Damilano, N Grandjean, F. Semond,
and
J
Massies,
J
Appl. Phys. 92, 5602 (2002).
22. N.
M
Wong, U Chowdhury, C L Collins, B Yang, J C Denyszyn, K S Kim, J
C Campbell, and R D. Dupuis, Phys. Stat. Sol. (a), 188, 333 (2001).
23.
L
Hirsch,
P
Moretto,
J.-Y.
Duboz, J.-L. Reverchon, B Damilano, N. Grandjean,
F
Semond, and J Massies, J Appl. Phys. 91, 6095 (2002).
24.
A
Cremades,
M
Albrecht, J Krinke, R Dimitrov,
M
Stutzmann, and H. P Strunk,
J Appl. Phys. 87, 2357 (2000).
25.
F
Semond, N. Antoine-Vincent, N. Sclmell,
M
Leroux, J Massies,
P
Disseix,
J
Leymarie, and
A
Vasson, Phys. Stat. Sol. (a), 183, 163 (2001).
26.
E
Momoy,
F
Calle,
E
Mufioz,
F
Omnes, B. Beaumont, and
P
Gibart, J Elect.
Mat., 28, 240 (1998).
8/17/2019 [doi 10.1007%2F978-1-4020-2103-9_6] Shur, Michael S.; Žukauskas, Artūras -- UV Solid-State Light Emitters and …
16/16
92 L Reverchon et a
27.
G
Parish, S Keller,
P
Kozodoy, J
P
Ibbetson,
H
Marchand,
P T
Fini, S
B
Fleischer,
S P
DenBaars,
U K
Mishra, and
E J
Tarsa, Appl. Phys. Lett. 75, 247
1999).
28.
V
Adivarahan,
G
Simin,
G
Tamulaitis,
R
Srinivasan,
J
Yang,
M
AsifKhan,
M
S Shur, and R Gaska, Appl. Phys. Lett. 79, 1903 2001).
29. N. Biyikli,
0
Aytur, I Kimukin,
T
Tut, and
E
Ozbay, Appl. Phys. Lett.
81,
3272
2002).
Recommended