You are on page 1of 290

This page intentionally left blank

Optoelectronic Sensors
This page intentionally left blank
Optoelectronic
Sensors
Edited by
Didier Decoster
Joseph Harari
First published in France in 2002 by Hermes Science/Lavoisier entitled: Détecteurs optoélectroniques
© LAVOISIER, 2002
First published in Great Britain and the United States in 2009 by ISTE Ltd and John Wiley & Sons, Inc.
Apart from any fair dealing for the purposes of research or private study, or criticism or review, as
permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,
or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA.
Enquiries concerning reproduction outside these terms should be sent to the publishers at the
undermentioned address:
ISTE Ltd John Wiley & Sons, Inc.
27-37 St George’s Road 111 River Street
London SW19 4EU Hoboken, NJ 07030
UK USA
www.iste.co.uk www.wiley.com
© ISTE Ltd, 2009
The rights of Didier Decoster and Joseph Harari to be identified as the authors of this work have been
asserted by them in accordance with the Copyright, Designs and Patents Act 1988.
Library of Congress Cataloging-in-Publication Data
Détecteurs optoélectroniques. English
Optoelectronic sensors / edited by Didier Decoster, Joseph Harari.
p. cm.
Includes bibliographical references and index.
ISBN 978-1-84821-078-3
1. Optical detectors. 2. Image sensors. I. Decoster, Didier, 1948- II. Harari, Joseph, 1961- III. Title.
TK8360.O67D4813 2009
681'.25--dc22
2009011542
British Library Cataloguing-in-Publication Data
A CIP record for this book is available from the British Library
ISBN: 978-1-84821-078-3
Printed and bound in Great Britain by CPI Antony Rowe, Chippenham and Eastbourne.
Table of Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Chapter 1. Introduction to Semiconductor Photodetectors. . . . . . 1
Franck OMNES
1.1. Brief overview of semiconductor materials . . . . . . . . . . . . . 1
1.2. Photodetection with semiconductors: basic phenomena . . . . . 3
1.3. Semiconductor devices. . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4. p-n junctions and p-i-n structures . . . . . . . . . . . . . . . . . . . 5
1.5. Avalanche effect in p-i-n structures . . . . . . . . . . . . . . . . . . 7
1.6. Schottky junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7. Metal-semiconductor-metal (MSM) structures . . . . . . . . . . . 10
1.8. Operational parameters of photodetectors . . . . . . . . . . . . . . 11
1.8.1. Response coefficient, gain and quantum efficiency . . . . . 11
1.8.2. Temporal response and bandwidth . . . . . . . . . . . . . . . . 12
1.8.3. Noise equivalent power . . . . . . . . . . . . . . . . . . . . . . . 13
1.8.4. Detectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Chapter 2. PIN Photodiodes for the Visible and Near-Infrared. . . 15
Baudoin DE CREMOUX
2.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2. Physical processes occurring in photodiodes . . . . . . . . . . . . 17
2.2.1. Electrostatics in PIN diodes: depleted region . . . . . . . . . 17
2.2.2. Mechanisms of electron-hole pair generation . . . . . . . . . 19
2.2.3. Transport mechanisms. . . . . . . . . . . . . . . . . . . . . . . . 23
2.3. Static characteristics of PIN photodiodes . . . . . . . . . . . . . . 25
2.3.1. I/V characteristics and definition of static parameters . . . . 25
2.3.2. External quantum efficiency . . . . . . . . . . . . . . . . . . . . 27
vi Optoelectronic Sensors
2.3.3. Dark current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.4. Breakdown voltage. . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3.5. Saturation current. . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4. Dynamic characteristics of PIN photodiodes . . . . . . . . . . . . 34
2.4.1. Intrinsic limitations to the speed of response. . . . . . . . . . 34
2.4.2. Limitations due to the circuit . . . . . . . . . . . . . . . . . . . 37
2.4.3. Power-frequency compromise, Pf
2
“law” . . . . . . . . . . . . 41
2.5. Semiconductor materials used in PIN photodiodes
for the visible and near-infrared . . . . . . . . . . . . . . . . . . . . . . . 42
2.5.1. Absorption of semiconductors in the range 400-1,800 nm . 42
2.5.2. From 400 to 900 nm: silicon and the GaAlAs/GaAs family 43
2.5.3. From 900 to 1,800 nm: germanium, GaInAsP/InP… . . . . 46
2.6. New photodiode structures . . . . . . . . . . . . . . . . . . . . . . . 49
2.6.1. Beyond the limits of conventional PIN . . . . . . . . . . . . . 49
2.6.2. Photodiodes with collinear geometry . . . . . . . . . . . . . . 50
2.6.3. Waveguide photodiodes. . . . . . . . . . . . . . . . . . . . . . . 52
2.6.4. Traveling-wave photodiodes. . . . . . . . . . . . . . . . . . . . 53
2.6.5. Beyond PIN structures . . . . . . . . . . . . . . . . . . . . . . . 54
2.7. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Chapter 3. Avalanche Photodiodes . . . . . . . . . . . . . . . . . . . . . . 57
Gérard RIPOCHE and Joseph HARARI
3.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2. History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3. The avalanche effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.1. Ionization coefficients . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.2. Multiplication factors . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.3. Breakdown voltage. . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.4. Properties of avalanche photodiodes . . . . . . . . . . . . . . . . . 66
3.4.1. Current-voltage characteristics and photomultiplication . . 66
3.4.2. Noise in avalanche photodiodes. . . . . . . . . . . . . . . . . . 68
3.4.3. Signal-to-noise ratio in avalanche photodiodes . . . . . . . . 71
3.4.4. Speed, response time and frequency response
of avalanche photodiodes . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.5. Technological considerations. . . . . . . . . . . . . . . . . . . . . . 76
3.5.1. Guard ring junctions . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.5.2. “Mesa” structures. . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.5.3. Crystal defects and microplasmas . . . . . . . . . . . . . . . . 79
3.6. Silicon avalanche photodiodes . . . . . . . . . . . . . . . . . . . . . 80
3.6.1. Si N
+
P APDs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.6.2. Si N
+
PʌP
+
APDs . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.6.3. Si N
+
ʌPʌP
+
APDs . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Table of Contents vii
3.6.4. SiPt-Si N Schottky APDs. . . . . . . . . . . . . . . . . . . . . . 87
3.7. Avalanche photodiodes based on gallium arsenide . . . . . . . . 88
3.8. Germanium avalanche photodiodes . . . . . . . . . . . . . . . . . . 90
3.8.1. Ge APDs with N
+
P, N
+
NP and P
+
N structures for
1.3 μm communication . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.8.2. Ge APDs with P
+
NN
-
structures for 1.55 μm
communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.9. Avalanche photodiodes based on indium phosphate (InP). . . . 95
3.9.1. InGaAs/InP APDs for optical communications
at 2.5 Gbit/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.9.2. Fast InGaAs/InP APDs . . . . . . . . . . . . . . . . . . . . . . . 99
3.10. III-V low-noise avalanche photodiodes. . . . . . . . . . . . . . . 100
3.10.1. III-V super-lattice or MQW APDs . . . . . . . . . . . . . . . 101
3.10.2. Spin-orbit resonance APDs . . . . . . . . . . . . . . . . . . . . 102
3.11. Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.11.1. Si/InGaAs APDs . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.11.2. “Waveguide” MQW APDs . . . . . . . . . . . . . . . . . . . . 104
3.11.3. Low-noise APDs with a very thin multiplication region. . 105
3.12. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.13. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Chapter 4. Phototransistors . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Carmen GONZALEZ and Antoine MARTY
4.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.2. Phototransistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.2.1. Phototransistors according to their fabrication materials . . 112
4.2.2 Phototransistors classified by structure. . . . . . . . . . . . . . 114
4.3. The bipolar phototransistor: description and principles
of operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.3.1. The phototransistor effect. . . . . . . . . . . . . . . . . . . . . . 119
4.3.2. The response coefficient of a phototransistor . . . . . . . . . 124
4.3.3. Static electrical and optical gains of the phototransistor. . . 125
4.3.4. Dynamic characteristics of phototransistors . . . . . . . . . . 126
4.3.5. Noise in phototransistors . . . . . . . . . . . . . . . . . . . . . . 138
4.4. Photodetector circuits based on phototransistors. . . . . . . . . . 140
4.4.1. Amplification circuits . . . . . . . . . . . . . . . . . . . . . . . . 140
4.4.2. Nonlinear circuits. . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.5. Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.5.1. Galvanic isolation . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.5.2. Phototransistors for optical telecommunications . . . . . . . 145
4.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.7. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
viii Optoelectronic Sensors
Chapter 5. Metal-Semiconductor-Metal Photodiodes . . . . . . . . . 155
Joseph HARARI and Vincent MAGNIN
5.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
5.2. Operation and structure . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.2.1. Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.2.2. Materials used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.3. Static and dynamic characteristics. . . . . . . . . . . . . . . . . . . 165
5.3.1. Response coefficient . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.3.2. Dynamic behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.3.3. Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.4. Integration possibilities and conclusion . . . . . . . . . . . . . . . 177
5.5. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Chapter 6. Ultraviolet Photodetectors. . . . . . . . . . . . . . . . . . . . 181
Franck OMNES and Eva MONROY
6.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.2. The UV-visible contrast . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.3. Si and SiC photodetectors for UV photodetection . . . . . . . . . 190
6.3.1. UV photodiodes based on silicon. . . . . . . . . . . . . . . . . 191
6.3.2. SiC-based UV photodetectors . . . . . . . . . . . . . . . . . . . 194
6.4. UV detectors based on III-V nitrides . . . . . . . . . . . . . . . . . 195
6.4.1. Photoconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.4.2. Schottky barrier photodiodes based on AlGaN . . . . . . . . 202
6.4.3. MSM photodiodes . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.4.4. p-n and p-i-n photodiodes . . . . . . . . . . . . . . . . . . . . . 210
6.4.5. Phototransistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.6. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Chapter 7. Noise in Photodiodes and Photoreceiver Systems . . . . 223
Robert ALABEDRA and Dominique RIGAUD
7.1. Mathematical tools for noise . . . . . . . . . . . . . . . . . . . . . . 224
7.1.1. Known signals with finite energy or power . . . . . . . . . . 224
7.1.2. Random signals and background noise . . . . . . . . . . . . . 226
7.2. Fundamental noise sources . . . . . . . . . . . . . . . . . . . . . . . 227
7.2.1. Thermal noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.2.2. Shot noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
7.2.3. Multiplication noise . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.3. Excess noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
7.3.1. Generation-recombination noise . . . . . . . . . . . . . . . . . 232
7.3.2. 1/f noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Table of Contents ix
7.4. Analysis of noise electrical circuits . . . . . . . . . . . . . . . . . . 235
7.4.1. Representation of noise in bipoles . . . . . . . . . . . . . . . . 235
7.4.2. Representation of noise in quadripoles . . . . . . . . . . . . . 237
7.5. Noise in photodetectors . . . . . . . . . . . . . . . . . . . . . . . . . 239
7.5.1. Characteristic parameters . . . . . . . . . . . . . . . . . . . . . . 240
7.5.2. PIN photodiodes . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
7.5.3. Avalanche photodiodes . . . . . . . . . . . . . . . . . . . . . . . 244
7.6. Noise optimization of photodetectors. . . . . . . . . . . . . . . . . 245
7.6.1. Formulation of the problem . . . . . . . . . . . . . . . . . . . . 246
7.6.2. Concepts for photodetector-transistor matching. . . . . . . . 251
7.7. Calculation of the noise of a photoreceiver . . . . . . . . . . . . . 253
7.7.1. Basic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
7.7.2. Models of transistor noise . . . . . . . . . . . . . . . . . . . . . 255
7.7.3. Example calculation: a PIN-FET photoreceiver. . . . . . . . 259
7.8. Comments and conclusions . . . . . . . . . . . . . . . . . . . . . . . 266
7.9. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
List of Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
This page intentionally left blank
Preface
Photodetection is found in a large number of professional and
mass-market systems. There are numerous applications including:
fiber-based and free-space optical telecommunications, galvanic
isolation, solar cells, proximity detectors, etc. All of these applications
are based on the same process: the transformation of optical power
into an electrical signal, with this signal needing to be as large as
possible for an optical flux as weak as possible. When information
needs to be transmitted quickly, potentially at very high speeds, the
photodetector must react very fast. These basic considerations imply
that a certain number of performance requirements should be met in
order to satisfy the demands of the intended application. It is in this
context that various photodetecting structures have been conceived:
photoconductors, p-n and p-i-n photodiodes, avalanche photodiodes,
phototransistors, Schottky photodiodes, MSM (metal semiconductor-
metal) photodetectors. The range of wavelengths relevant to the
application also plays an important role. It is often this, through the
intermediary of the semiconductor bandgap, which dictates the nature
of the material(s) used.
This book gathers together the most detailed and significant
contemporary thinking on photodetection for wavelengths from the
near-infrared to the ultraviolet
1
. Its content not only gives the reader
1. The very particular nature of photodetection in the mid- and far-infrared requires
specific treatment and thus is not covered here.
xii Optoelectronic Sensors
the grounding to design simple photodetectors with specified
performance characteristics, but also discusses the state of the art in
photodetection. Chapter 1 begins with an introduction to
photodetection and the most well-known photodetector structures.
Among these, some are more commonly used than others, or offer
more exciting possibilities.
Because of this, each of Chapters 2 to 5 concentrates in great depth
on a specific photodetector type. In sequence, the reader will be able
to learn about:
– the PIN photodiode which is the photodetector on which the
majority of systems are based (Chapter 2);
– the avalanche photodiode, which is a refinement of the PIN
photodiode and which, under a sufficiently high bias voltage, allows
us to achieve a gain (Chapter 3);
– the phototransistor, which is another means of obtaining gain in
photodetection (Chapter 4);
– the MSM photodiode which is well suited to monolithic
integrated circuits such as MMICs (Microwave Monolithic Integrated
Circuits) (Chapter 5).
Chapter 6, however, is different. It is dedicated to ultraviolet
photodetectors. Given the significant role that this wavelength range is
likely to take in the near future, notably for environmental
applications, this subject deserves a separate chapter. In this case, all
types of photodetector are concerned, and it is the material which is
the source of improvement, notably with the appearance of AlGaN
materials with a very large bandgap.
Finally, Chapter 7 is entirely dedicated to noise, as this concept is
absolutely fundamental for the photodetection of low-intensity signals.
Preface xiii
The chapters of this book have all been written by specialists and
we take this opportunity to thank them sincerely and warmly for their
contributions.
Didier DECOSTER and Joseph HARARI
This page intentionally left blank
Chapter 1
Introduction to
Semiconductor Photodetectors
1.1. Brief overview of semiconductor materials
A semiconductor material is a continuous crystalline medium
characterized by an energy band structure corresponding, in the case
of an infinite crystal, to a continuum of states (which, in practice,
means that the characteristic dimensions of the crystal are
significantly larger than the lattice parameter of the crystal structure;
this applies as long as the crystal dimensions are typically larger than
a few dozen nanometers). In general terms, the energy structure of a
semiconductor consists of a valence band corresponding to molecular
bonding states and a conduction band representing the molecular anti-
bonding states. The energy range lying between the top of the valence
band and the bottom of the conduction band is known as the forbidden
band, or more commonly the bandgap. An electron situated in the
valence band is in a ground state and remains localized to a particular
atom in the crystal structure, whereas an electron situated in the
conduction band exists in an excited state, in a regime where it
interacts very weakly with the crystalline structure. What
Chapter written by Franck OMNES.
2 Optoelectronic Sensors
differentiates semiconductors from insulators is essentially the size of
the bandgap: we refer to semiconductors where the bandgap of the
material is typically less than or equal to 6 eV, and to insulators when
the bandgap is more than 6 eV: above this, the solar spectrum arriving
on the Earth’s surface is unable to produce inter-band transitions of
electrons situated in the valence band of the material. Semiconductor
materials are mostly divided into two large classes: elemental
semiconductors (group IV of the periodic table): silicon, germanium,
diamond, etc. and compound semiconductors: IV-IV (SiC), III-V
(GaAs, InP, InSb, GaN) and II-VI (CdTe, ZnSe, ZnS, etc.). Impurities
can be introduced into the volume of the semiconductor material and
can modify its electrical conduction properties, sometimes
considerably. An impurity is known as a donor when it easily releases
a free electron into the conduction band. The characteristic energy
level of the impurity is therefore in the bandgap, slightly below the
conduction band. For example, in the case of compound
semiconductors in group IV of the periodic table such as silicon, the
main donor impurities are those which, being from group V of the
periodic table (arsenic, phosphorous, etc.), are substituted in place of a
silicon atom in the crystal structure: since silicon is tetravalent, these
atoms naturally form four covalent bonds with the silicon atoms
around them, and also easily give up their surplus electron to the
crystal structure. These electrons become free to move, subject to a
weak activation energy provided by thermal agitation. In this case we
refer to n-type doping. In the case of silicon, a group III element
incorporated into the crystal structure of silicon naturally forms three
covalent bonds around it, and then completes its own outer-shell
electronic structure by capturing an electron from its fourth nearest-
neighbor silicon atom, again subject to a weak thermal activation
energy. Such an impurity is known as an acceptor, and doping with
acceptors is known as p-type doping. A hole carrying a positive
elementary charge and corresponding to a vacant energy state in the
valence band is therefore left in the crystal structure of the silicon. In
the case of III-V composites, the donors are mostly atoms from group
IV (silicon) substituted in place of group III elements, or group VI
elements (S, Se, Te) substituted in place of group V elements, and
acceptors are group II (zinc, magnesium) substituted in place of group
Introduction to Semiconductor Photodetectors 3
III elements. In the case of II-VI composites, the most commonly-
encountered donors belong to group VII (chlorine, etc.) substituted in
place of group VI elements, and acceptors belong to either group I
(lithium, etc.) or to group V (nitrogen, arsenic, phosphorous, etc). In
this last case, the group V element is substituted in place of a group VI
element in the semiconductor crystal structure, whereas group I
acceptors are substituted in place of group II elements. The chemical
potential, or Fermi energy, of an intrinsic semiconductor (i.e. one free
from n and p impurities) is found in the middle of the bandgap of the
material. When a moderate n-type doping is added, the Fermi level
rises from the middle of the bandgap towards the conduction band, by
an increasing amount as the level of doping rises. When the level of n-
type doping becomes large, the Fermi level can cross the bottom of
the conduction band and be found inside this band (Mott transition).
The semiconductor then behaves like a metal and for this reason is
called a semi-metal. In this case it is referred to as degenerate. In the
case of p-type doping, the semiconductor is said to be degenerate
when the Fermi level is below the top of the valence band.
1.2. Photodetection with semiconductors: basic phenomena
Photodetection in semiconductors works on the general principle of
the creation of electron-hole pairs under the action of light. When a
semiconductor material is illuminated by photons of an energy greater
than or equal to its bandgap, the absorbed photons promote electrons
from the valence band into excited states in the conduction band,
where they behave like free electrons able to travel long distances
across the crystal structure under the influence of an intrinsic or
externally-applied electric field. In addition, the positively-charged
holes left in the valence band contribute to electrical conduction by
moving from one atomic site to another under the effects of the
electric field. In this way the separation of electron-hole pairs
generated by the absorption of light gives rise to a photocurrent,
which refers by definition to the fraction of the photogenerated free
charge-carriers collected at the edges of the material by the electrodes
of the photodetecting structure, and whose intensity at a given
4 Optoelectronic Sensors
wavelength is an increasing function of the incident light intensity. On
this level we can distinguish between two large categories of
photodetectors based on the nature of the electric field, which causes
the charge separation of photogenerated electron-hold pairs:
photoconductors, which consist of a simple layer of semiconductor
simply with two ohmic contacts, where the electric field leading to the
collection of the charge-carriers is provided by applying a bias voltage
between the contacts at either end, and photovoltaic photodetectors,
which use the internal electric field of a p-n or Schottky (metal-
semiconductor) junction to achieve the charge separation. This last
term covers p-n junction photodetectors (photovoltaic structures
consisting of a simple p-n junction, and p-i-n photodetectors which
include a thin layer of semiconductor material between the p and n
region which is not deliberately doped), as well as all Schottky
junction photodetectors (Schottky barrier photodiodes and metal-
semiconductor-metal (MSM) photodiodes).
We will now briefly introduce the main physical concepts at the
root of the operation of the different semiconductor photodetector
families. Here the emphasis is placed on a phenomenological
description of the working mechanisms of the devices in question; the
corresponding formalism has been deliberately kept to an absolute
minimum in the interests of clarity and concision.
1.3. Semiconductor devices
Photoconductors represent the simplest conceivable type of
photodetector: they consist of a finite-length semiconductor layer with
an ohmic contact at each end (Figure 1.1). A fixed voltage of
magnitude V
B
is applied between the two end contacts, in such a way
that a bias current I
B
flows through the semiconductor layer, simply
following Ohm’s law. The active optical surface is formed from the
region between the two collection electrodes. When it is illuminated,
the photogenerated changes produced under the effect of the applied
electric field lead to a photocurrent I
PH
which is added to the bias
current, effectively increasing the conductivity of the device.
Introduction to Semiconductor Photodetectors 5
Figure 1.1. Diagram of a photoconducting device
The main point of interest in a photoconducting device is its
increased gain, the response of photoconductors being typically
several orders of magnitude greater than that of photovoltaic detectors
for a given material. On the other hand, its other operational
parameters (bandwidth, UV/visible contrast, infrared sensitivity) are
generally below that of other types of photodetectors, which often
greatly limits the scope of its potential applications (this is particularly
the case for photoconductors based on III-V nitrides, as we will see
later on).
1.4. p-n junctions and p-i-n structures
In p-n diodes, the metallurgical linkage of a region of a p-type
doped semiconductor and a region of n-type doping forms a p-n
junction, where the joining of the Fermi levels in equilibrium mostly
occurs through a flow of charge between the n and p regions. In
equilibrium we therefore find a region with no free charge carriers
immediately around the junction, similar to a charged capacitor, where
there are, on the n side, positively ionized donors and, on the p side,
negatively ionized acceptors (this zone is known as the space charge
region (SCR), where ionized donors and acceptors provide fixed
charges). The presence of charged donors and acceptors produces an
electric field in that region which curves the energy bands and, in
equilibrium, forms an energy barrier between the two regions: the
6 Optoelectronic Sensors
bottom of the conduction band and the top of the valence band on the
n side are below the corresponding levels on the p side (Figure 1.2).
Figure 1.2. Curvature of the energy bands and mechanisms of
photocurrent generation in a p-n junction
The width of the SCR is a decreasing function of the level of
doping in the material, while the height of the energy barrier is an
increasing function of it. An electron-hole pair produced in this SCR
(situation 2 in Figure 1.2) is therefore separated by the effect of the
internal electric field of the junction, and so does not recombine.
These are the charge carriers which contribute to the photocurrent, to
which we can add, to some extent, those generated at a distance from
the junction less than or equal to the diffusion length (situations 1 and
3 in Figure 1.2). The band structure of the junction implies that the
photocurrent will consist of minority charge carriers. For this reason,
the photocurrent flows in the opposite direction to the bias on the
diode, where the forward direction is defined as the direction of flow
of the majority charge carriers (from the n to the p region in the case
of electrons, and vice versa for holes). Moreover, the application of an
opposing external electric field (V
p
–V
n
< 0) allows us to increase the
height of the energy barrier in the vicinity of the junction, and also
increase the spatial extent of the SCR, which significantly improves
the efficiency of the separation of electron-hole pairs by increasing the
electric field within the junction.
Introduction to Semiconductor Photodetectors 7
We note that when the doping level is moderate, the width of the
SCR is important. This effect is beneficial in the case of p-n junction
photodetectors, where in order to increase the photoresponse it is
desirable to ensure that the mechanisms of electron-hole pair
generation through incident light take place predominately inside the
SCR. A simple means of increasing the spatial extent of the SCR is to
introduce between the n and p regions a thin layer of intrinsic
semiconductor material which is not intentionally doped: the structure
is therefore referred to as p-i-n. Such a structure is interesting because
it is possible to maintain high levels of doping in the n and p regions
without significantly reducing the extent of the SCR, whose width is
then largely determined by the thickness of the “i” layer. Additionally,
increasing the width of the SCR reduces the capacitance of the
structure, which makes p-i-n structures particularly well-suited for
high-speed operation.
1.5. Avalanche effect in p-i-n structures
When the reverse-bias voltage established at the terminals of a p-i-
n structure increases sufficiently that the electric field established in
the junction reaches values close to the breakdown field (in structures
of micron-scale thickness, this is generally the case when the bias
voltage at the terminals reaches a few dozen volts), the
photogenerated charge carriers in the SCR (which is effectively the
region that is not intentionally doped) are accelerated enough to
separate other secondary charge carriers from the atoms in the lattice
that they impact in the course of their motion: this is the avalanche
effect which results in a multiplication of the charge carriers in the
SCR. The gain is therefore greater than 1 for the generation of charge
carriers by light, and this gain can even typically reach 10 or 20 under
favorable conditions. This effect is exploited in what are called
avalanche photodiodes where the levels of n- and p-type doping are
generally adjusted to high values above 10
18
cm
-3
to maximize the
intrinsic electric field of the junction.
8 Optoelectronic Sensors
1.6. Schottky junction
A Schottky junction is formed by bringing a metal and a
semiconductor into contact. The basic phenomena which lead to the
formation of a Schottky junction with an n-type semiconductor are
summarized in Figure 1.3.
Figure 1.3. Formation of a Schottky junction (in an n-type semiconductor)
In thermal equilibrium, when the Fermi levels of the metal and the
semiconductor are equalized, a transfer of electronic charge occurs
from the semiconductor to the metal in the case where the work
function q.)
M
of the metal (q being the elementary charge) is greater
than the electron affinity X of the semiconductor, and a SCR appears
at the edge of the semiconductor of width x
d
next to the junction,
where the only charges present are the positively-ionized donors. A
curvature of the energy bands therefore occurs at the junction, which
leads to the appearance of an energy barrier between the metal and the
Introduction to Semiconductor Photodetectors 9
semiconductor, called a Schottky barrier, whose height is given to first
approximation by the expression:
q )
Bn
=q )
M
÷F
( )
[1.1]
In equilibrium, therefore, we find an intrinsic electric field
immediately next to the metal-semiconductor junction which is
comparable in form to that found in a p-n junction. Consequently, it is
the phenomenon of photogeneration of charge carriers inside and near
to the SCR which is responsible for the appearance of a photocurrent,
with the electron-hole pairs being separated by the effect of the
electric field in the Schottky junction. It is possible, as in the case of
the p-n junction, to modify the intensity of the internal electric field in
the junction by applying a bias voltage V between the semiconductor
and the metal of the Schottky contact (Figure 1.4).
Figure 1.4. Reverse-bias of a Schottky junction (n-type semiconductor material)
In the case of an n-type semiconductor, the application of a
negative voltage between the semiconductor and the metal electrode
of the Schottky contact has the effect of reverse-biasing the Schottky
junction, which leads to an increase in the height of the effective
barrier, along with an increase in the width of the SCR. This last effect
is of course favorable for photodetection. Indeed, it follows that the
majority charge carriers (electrons) cannot flow towards the Schottky
contact, and only the minority carriers (holes) generated by external
excitation (in particular photogeneration) can reach the Schottky
contact and hence produce an electric current: as in the case of the p-n
10 Optoelectronic Sensors
junction, we therefore find that the current flows in reverse through
the Schottky junction, that is, from the semiconductor towards the
Schottky contact. The illumination of Schottky photodiodes can occur
through the front or rear face (often this second option is chosen in the
case where the substrate material is transparent to the light to be
detected, as is the case for example with sapphire). In the case of
illumination through the front face, we resort to a semi-transparent
Schottky contact, characterized by a very small thickness of metal (of
the order of 100 Å) selected to ensure sufficient optical transmission:
while a thin layer of gold of 100 Å thickness transmits up to 95% of
the incident light in the infrared, the percentage transmitted in the
ultraviolet is around 30% in the range 300-370 nm. The gain of p-i-n
photodiodes (other than the specific case of avalanche photodiodes)
and Schottky photodiodes is at most 1, which would be the case if all
the photogenerated charge carriers were collected by the electrodes at
the ends of the device.
1.7. Metal-semiconductor-metal (MSM) structures
An MSM structure consists of two Schottky electrodes, often
interlinked in the form of a comb structure, leaving a free
semiconductor surface between the two contacts which forms the
active region in which light will be absorbed. A bias voltage can be
applied between the two electrodes, in order to break the initial
electrical symmetry of the contacts: one of the Schottky junctions is
reverse-biased, producing a SCR of increased width, and the other
junction is forward-biased.
The absorption of light near the reverse-biased junction creates
electron-hole pairs which are separated under the effects of the electric
field present in the SCR, thus creating the photocurrent. The other
electrode, consisting of a forward-biased (and hence transmissive)
Schottky junction, simply acts as a collection electrode. The band
diagram of the device under increased bias voltage (V
B
) is represented
schematically in Figure 1.5, in which L is the distance between two
adjacent contact fingers, )
0
is the height of the Schottky barrier and
Introduction to Semiconductor Photodetectors 11
I
ph
is the photocurrent. MSM photodetectors normally use
semiconductor materials which are not intentionally doped, are
chemically very pure and electrically very resistive. The SCRs
associated with Schottky junctions made of these materials are hence
of significant width which, for a given bias voltage, allows the electric
field of the junction to extend more easily into semiconductor regions
some way from the contact. It follows that photogenerated electron-
hole pairs are more easily separated and collected by the electrodes at
either end.
Figure 1.5. Energy band diagram for an MSM structure
under electrical bias; effect of illumination
1.8. Operational parameters of photodetectors
The main parameters which define the behavior of an ultraviolet
photodetector are respectively the response coefficient, the gain, the
quantum efficiency, the bandwidth, the noise equivalent power (NEP)
and the detectivity.
1.8.1. Response coefficient, gain and quantum efficiency
The response coefficient of a photodetector, R
i
, links the
photocurrent I
ph
to the power of the incident light P
opt
through the
relationship:
I
ph
= R
i
P
opt
[1.2]
12 Optoelectronic Sensors
It is important to note in passing that the response coefficient is a
quantity independent of the active optical surface of the photodetector
structure: indeed, the photocurrent as well as the incident optical
power are both, in the ideal case, proportional to the active optical
surface. At a given wavelength O, the flux ) of photons arriving on
the semiconductor surface, which is defined as the number of photons
reaching the active surface per unit time, is given by:
)= P
opt
O / h c
( )
[1.3]
where h is the Planck constant and c is the speed of light.
The quantum efficiency K is defined as the probability of creating
an electron-hole pair from an absorbed photon. Considering that all
the incident light is absorbed in the semiconductor material, the rate G
of electron-hole pair generation per unit time is thus given by:
G =K )=K P
opt
O / h c
( )
[1.4]
If we now introduce the gain parameter g which corresponds to the
number of charge carriers detected relative to the number of
photogenerated electron-hole pairs, then the photocurrent is given by
the equation:
I
ph
=q G g =qK P
opt
O / h c
( )
g = qK O / hc
( )
g
( )
P
opt
[1.5]
where q is the elementary charge (1.602 x 10
-19
C), from which we
obtain the expression for the response coefficient of the detector:
R
i
=q g K O / hc
( )
[1.6]
1.8.2. Temporal response and bandwidth
The speed of response of a photodetector may be limited by
capacitative effects, by the trapping of charge carriers or by the
saturation speed of charge carriers in the semiconductor. These
phenomena all lead to a reduction in the response of the photodetector
Introduction to Semiconductor Photodetectors 13
in the high-frequency domain. The cutoff frequency f
C
of the
photodetector is defined as the frequency of optical signal for which
the response coefficient is half that for a continuous optical signal.
The temporal response of a photodetector is characterized by the fall
time W
f
(or the rise time W
r
), which is defined as the time needed for the
photocurrent to fall from 90% to 10% of its maximum (or to rise from
10% to 90% of it). In the case of a transient exponential response with
a time constant W, the following relationship links the bandwidth BW
and the temporal response of the photodetector:
BP =1/ 2S W
( )
=2.2 / 2SW
m
( )
=2.2 / 2S W
d
( )
[1.7]
1.8.3. Noise equivalent power
The NEP is defined as the incident optical power for which the
signal-to-noise ratio is 1, and hence the photocurrent I
ph
is equal to the
noise current I
b
. In other words, it is the smallest optical power which
can be measured. It follows that the NEP parameter is given by the
equation:
NEP = I
b
/ R
i
in W
( )
[1.8]
In the case of white noise, the noise current I
b
increases as the
square root of the bandwidth of the photodetector device. It follows
that it is preferable and customary to use the following expression for
the NEP, normalized with respect to the bandwidth BW:
NEP* = NEP BW
( )
1 2
in W Hz
1 2
( )
[1.9]
In semiconductors, there are five sources of noise:
– shot noise, mainly due to the random nature of the collisions of
incident photons;
– thermal noise, due to random collisions of charge carriers with
the atoms of the crystal lattice, in permanent vibration due to thermal
motion;
– partition noise, caused by the separation of the electric current
into two parts flowing across separate electrical contacts;
14 Optoelectronic Sensors
– generation-recombination noise, caused by the random
generation and recombination of charge carriers, either band to band
or via trapping levels situated in the bandgap;
– 1/f noise, associated with the presence of potential barriers at the
level of the electrical contacts. This last type of noise dominates at
low frequencies.
1.8.4. Detectivity
This figure of merit is defined by the equation:
D = NEP
( )
1
= R
i
/ I
b
in W
1
( )
[1.10]
In general terms, the photocurrent signal increases in proportion to
the active optical area A
opt
, and in addition the noise current increases
with the square root of the product of the active optical area with the
bandwidth BW. It follows that the preferred method of comparing
between different photodetectors is to use an expression for the
detectivity normalized with respect to these parameters, written:
D* = D A
opt
BP
( )
1 2
= R
i
I
b
( )
A
opt
BP
( )
1 2
in W
1
cm.Hz
1 2
( )
[1.11]
The normalized detectivity is the most important parameter for
characterizing a photodetector because it allows direct comparison of
the performance of photodetectors using technologies and methods of
operation which are at first glance very different. It is clear from the
preceding definitions that the determination of the NEP and the
detectivity requires measurement of three parameters: the response
coefficient, the bandwidth and the noise current of the photodetector
device. The measurement of the noise current must be made in
darkness. The device is biased using a very stable voltage source, and
the entire measurement system must itself have an intrinsic noise level
considerably lower than the intrinsic noise of the photodetector device.
Chapter 2
PIN Photodiodes for the
Visible and Near-Infrared
2.1. Introduction
Photodiodes are optoelectronic devices with two electrodes and
asymmetric electrical characteristics. They are normally reverse-
biased (biased in the non-conducting direction) and used to convert
optical fluxes into electrical currents. The term PIN refers to the three
doped semiconductor layers in their active part, of types P, intrinsic
and N respectively, a structure which aims to optimize the
characteristics compared to those of simple PN junction photodiodes.
Figure 2.1. Schematic layout of PIN photodiodes: collinear optical flux and charge
transport on the left, orthogonal on the right
Chapter written by Baudoin DE CREMOUX.
16 Optoelectronic Sensors
Figure 2.1 shows the basic processes taking place in a PIN
photodiode:
– generation of an electron-hole pair through absorption of a
photon, most probably in the I region, either using radiative or non-
radiative processes;
– transport under the effects of the electric field present in the I
region: of the electron towards the N region and of the hole towards
the P region, then to the external circuit through the metallic contacts
deposited on the surfaces, which have a window where required.
Thus, the absorption of a photon in the I region leads to the
transport of an electron in the external circuit. Figure 2.1 shows the
possible arrangements for the optical flux and the electrical current.
Collinear geometry is the most commonly used. We use it as the
reference type for the analysis of photodiode characteristics. After this,
we will consider orthogonal geometry.
The spectral domain of interest, mostly from 400 to 1,700 nm,
determines the materials used to fabricate photodiodes. This design is
the most mature of the semiconductors, and as a result almost all
photodiodes used in this domain have a PIN structure, or their central
layer is at least weakly doped compared to the P and N layers.
This chapter is organized in order to first provide the reader with
the basic tools to understand the concepts of PIN photodiodes for a
given application, then it describes the devices which are currently
available or which should be available soon. With this aim in mind,
section 2.2 recalls some of the physics of semiconductors,
emphasizing the processes taking place in photodiodes, the pertinent
characteristics of the materials and the notations used later in the text.
Sections 2.3 and 2.4 establish the simple quantitative relations
between the structure of devices with typical geometries and their
static and then dynamic characteristics. We will stress the
compromises required among these characteristics if we are to keep a
particular geometry. Section 2.5 then describes the properties of
photodiodes with collinear geometry, currently used in a numerous
professional and mass-market systems. Finally, section 2.6 describes
PIN Photodiodes for the Visible and Near-Infrared 17
advanced devices with novel structures which push the boundaries of
the previously-discussed compromises. These are for the most part
laboratory devices which could find applications in real-world systems.
2.2. Physical processes occurring in photodiodes
2.2.1. Electrostatics in PIN diodes: depleted region
Depleted region
Figure 2.2. From top to bottom, profiles of fixed charge density,
electric field and potential across a reverse-biased PIN diode
In the neighborhood of a PN junction there is a region devoid of
free charge carriers (the depleted region) where fixed ionized
impurities can be found, negative on the P side and positive of the N
side, forming a space charge region (SCR). By integrating the Poisson
equation twice in the z direction:
d
2
V
dz
2
=
U z
( )
H
[2.1]
18 Optoelectronic Sensors
Where U(z) is the charge density and H is the dielectric constant, we
can obtain the profiles of the electric field E and the potential V across
the structure. In the case of an abrupt PN junction, the thickness of the
depleted region varies as —(V
BI
–V) where V
BI
is the built-in or
diffusion potential present in the absence of external bias applied to
the diode, and V is the voltage across the terminals of the diode. As a
result, the capacitance of the junction decreases when the reverse-bias
voltage increases in absolute value. The case of a PIN diode structure
is considered in Figure 2.2, showing in sequence the profiles across
the structure of the density of fixed charges, of the electric field and of
the potential.
In the realistic approximation that the densities of ionized
impurities are much greater than the density of residual impurities in
the I region (typically by a factor of 100), we draw the following
conclusions:
– the extension of the depleted region into the P and N regions is
negligible, and the depleted region is to all intents and purposes
coincident with the I region whose thickness, represented by d in the
rest of this chapter, is a very important parameter in the definition of a
PIN photodiode;
– the capacitance of the junction is independent of the reverse-bias
voltage. It is given, as for a conventional capacitor, by:
C
J
=H
S
d
[2.2]
where S is the area of the junction;
– the electric field is constant in the depleted region and, insofar as
V>>V
BI
, its value is
E =V d [2.3]
These approximations will be used throughout the remainder of the
chapter unless explicitly stated otherwise.
PIN Photodiodes for the Visible and Near-Infrared 19
2.2.2. Mechanisms of electron-hole pair generation
2.2.2.1. Radiative generation and optical absorption
Electronic transitions between permitted levels in the valence band
(VB) and levels in the conduction band (CB) can occur under the
action of photons transferring their energy to the electrons. Thus, a
photon can be absorbed, creating an electron-hole pair. This process of
radiative generation, shown schematically in Figure 2.3, is the inverse
of the process of recombination which gives rise to stimulated
emission of photons in laser diodes.
It follows from this that the only photons absorbed efficiently are
those whose energy hQ is greater than the width of the bandgap E
G
of
the material under consideration:
hQ _ E
G
= E
C
÷E
V
[2.4]
Figure 2.3. Radiative generation
The material is transparent to photons with lower energies. The
parameter E
G
is thus a criterion in the choice of materials when
designing a photodetector. Optical absorption is characterized by the
coefficient of absorption D (cm
–1
), according to the relationship:
dF
Fdz
=÷ [2.5]
20 Optoelectronic Sensors
where F (cm
–2
s
–1
) is the photon flux propagating in the z direction.
The level of radiative generation G
R
(cm
–3
s
–1
), the number of electron-
hole pairs created per unit volume and time, is therefore:
G
R
=DF [2.6]
It follows from these definitions that the level of radiative
generation in a homogenous material, for a photon flux F
0
falling on a
surface at z=0, can be written:
G
R
=DF
0
exp ÷Dz
( )
[2.7]
We see that 1/D (cm) gives an order of magnitude for the distance
required to absorb the radiation. Furthermore, the knowledge of
variation of D with photon energy (or with their wavelength) is
essential for the choice of materials for use in a photodetector. Figure
2.16 shows graphs for the main semiconductors used in the spectral
domain under consideration in this chapter.
The shape of these graphs and the magnitude of D depend on the
direct or indirect nature of the band structure of the material. This
nature is defined in Figure 2.4 which represents, for the two types of
material, the variations in energy of the electrons, with the wavevector
k describing their motion in the periodic crystal lattice.
Figure 2.4. Direct bandgap on the left and indirect bandgap on the right
PIN Photodiodes for the Visible and Near-Infrared 21
For materials with a direct bandgap (such as gallium arsenide) the
main minimum in the conduction band is found at the same vector k
as the maximum in the valence band. Furthermore, radiative
transitions must simultaneously conserve the energy and momentum
of the particles involved (the photon and electron). The momentum of
the electron is much greater than that of the photon, and as a result
radiative transitions are effectively vertical and connect with a high
probability those states at the top of the VB and those at the bottom of
the CB.
The opposite is true for materials with an indirect bandgap (such as
silicon), and radiative transitions require the involvement of the third
particle, a phonon, to provide the difference in momentum between
the initial and final states. These transitions are much less probable
and D grows much less quickly with the energy of the photons above
E
G
.
2.2.2.2. Non-radiative generation
As well as the radiative processes there are also non-radiative
processes of generation-recombination. Thus, in the SRH (Shockley
Read Hall) process shown in Figure 2.5 in generation mode, the
required energy is “borrowed” from the thermal motion of the crystal
lattice. This is one of the contributions to the dark current of
photodetectors, a detrimental parasitic phenomenon.
Figure 2.5. Non-radiative SRH generation
22 Optoelectronic Sensors
The transition takes place in two parts, via a “deep state” of energy
E
T
, which is close to the middle of the forbidden band and is the result
of a crystal defect (impurity, vacancy, etc.). The rate G
NR
(cm
–3
s
–1
) of
SRH generation-recombination can be written [SZE 81, p. 35]:
G
NR
=
V
n
V
p
Q
th
N
T
n
i
2
÷np
( )
V
n
n÷n
i
exp
E
T
÷E
I
kT
(
(
·
·
·
·
)
)

l
l
l
l
l
÷V
p
p ÷n
i
exp ÷
E
T
÷E
I
kT
(
(
·
·
·
·
)
)

l
l
l
l
l
[2.8]
with the following notation:
– V
n
and V
p
capture cross-sections for electrons and holes by the
deep state;
– v
th
, thermal velocity Q
th
= 3k T m*
( )
( )
;
– N
T
, density of deep states;
– n
i
, intrinsic electron density;
– n and p, electron and hole densities;
– E
I
, middle of the forbidden band;
– kT, Boltzmann factor.
In the depleted region of a photodiode we have np = n
i
2
. If we
further assume that V
n
=V
p
=V and E
T
= E
I
, and define an effective
lifetime W
eff
:
G
NR
= n
i
Q
th
N
T
2 = n
i
W
eff
[2.9]
We also know that:
n
i
2
= N
C
N
V
exp ÷E
G
kT
( )
[2.10]
where N
C
and N
V
are the equivalent densities of state for the CB and
VB. We can draw the following conclusions from this:
PIN Photodiodes for the Visible and Near-Infrared 23
– G
NR
increases very rapidly as the ratio E
G
/kT decreases;
– G
NR
is not an intrinsic characteristic of the material, but depends
for its purity on N
T
.
In an electrically neutral region of type p(n), where the electron
(hole) deficiency relative to thermodynamic equilibrium is Gn(Gp), if
we define Wn(Wp) as the lifetime of the minority electrons (holes), it
follows that:
G
NR
=GnVQ
th
N
T
=Gn W
n
[2.11]
SRH non-radiative generation is particularly active at the surface
of devices, where the density of defects is very high. It can be
reduced by a passivation treatment; for silicon the most effective is
the deposition of its natural oxide SiO
2
, silica. Another process of
non-radiative generation is shock ionization which occurs in a
semiconductor material under a strong electric field. It grows with
E
G
and the breakdown field is of the order of E
B
= 10
5
V/cm for the
materials considered. This effect is one cause of breakdown in
diodes, limiting the reverse-bias voltage that can be applied. It is
exploited as a gain mechanism in the avalanche photodiodes which
are described in Chapter 3, and whose structure is optimized with
this in mind.
2.2.3. Transport mechanisms
The charge carriers, electrons and holes generated by the
aforementioned processes, are transported in the device to the two
electrical contacts that transfer it to the external circuit – specifically,
to the load resistance across the terminals, across which the electrical
signal is measured.
24 Optoelectronic Sensors
2.2.3.1. Transport under electric field and resultant current
In the depleted region of a photodiode (normally reverse-biased) an
elevated electric field is present, and the electrons and holes are driven
at a speed which depends on the intensity of the field, as shown in
Figure 2.6.
Figure 2.6. Characteristics Q
n
(E) and Q
p
(E) in logarithmic coordinates
For electric fields such that E < 10
3
V/cm, the velocities are
proportional to E (linear regime) and we have:
Q
n, p
E
( )
= P
n, p
E [2.12]
where P
n
and P
p
(cm
2
/Vs) are the respective mobilities of electrons
and holes. For electric fields such that E > 10
4
V/cm, the velocities are
fairly constant and equal to a saturation velocity of the order of
v
s
= 10
7
cm/s. Between these two regions, and for certain materials,
the electron velocity passes through a maximum (high mobility
regime).
PIN Photodiodes for the Visible and Near-Infrared 25
2.2.3.2. Transport along the concentration gradient and diffusion
current
In the contiguous electrically neutral regions in the depleted region,
electrons and holes can be transported in the absence of an electric
field under the effect of their concentration gradients. Their fluxes F
n
and F
p
can be written:
F
n, p
= D
n, p
0n, p
0z
[2.13]
where D
n, p
= P
n, p
kT q cm
2
s
( )
are the diffusion coefficients.
The various recombination processes are also active in these
regions, such that the electrons and holes are only efficiently
transported over a distance known as the diffusion length and given
by:
I
n, p
= D
n, p
W
n, p
[2.14]
2.3. Static characteristics of PIN photodiodes
2.3.1. I/V characteristics and definition of static parameters
Figure 2.7 shows I/V curves, linking the current crossing a
photodiode to the voltage across its terminals for three illumination
conditions. From this we can extract the parameters commonly used to
characterize photodiode operation, whose relations to the structure of
the device are described in the following sections. The photocurrent I
P
produced by a photodiode is proportional to the illumination across a
large range of optical power P. The coefficient of proportionality is
the response coefficient R (A/W) (responsivity), not to be confused
with the sensitivity, which measures the minimum optical power
detectable by an optical detector, and which depends on the circuit
into which the photodiode is inserted and the specifications of the
signal to be received.
26 Optoelectronic Sensors
Figure 2.7. I/V characteristics of a photodiode, inserted in the circuit shown
and under several illumination conditions
The external quantum efficiency K
e
links the generated electron
flux to the incident photon flux. Thus, we have:
R =
I
p
P
=
q
hQ
K
e
[2.15]
where q is the charge of an electron and hQ is the energy of a photon.
The dark current I
0
, discussed in section 2.3.3, is caused by non-
radiative generation and/or the tunnel effect.
The saturation current I
M
is the photocurrent above which several
mechanisms limit the linearity of the response:
– the intrinsic mechanism, due to the screening of the electric field
in the depleted region by the space charge that is produced by the
mobile charge carriers created by the illumination (see section 2.3.4);
PIN Photodiodes for the Visible and Near-Infrared 27
– the circuit-dependent mechanism, caused by the voltage drop
across the load resistance (see the circuit in Figure 2.7).
A gain, characterized by the multiplication coefficient M defined
by:
0 p
I
M
I I
=
÷
[2.16]
can be obtained under a high reverse-bias voltage V
AV
in avalanche
photodiodes (see Chapter 3). The breakdown voltage V
B
is the highest
reverse-bias voltage the diode can support. For V=V
B
the dark current
grows very rapidly and the power dissipated in the diode can lead to
its destruction.
2.3.2. External quantum efficiency
The external quantum efficiency defined above can be thought of
as the product of an internal quantum efficiency K
i
and an optical
efficiency K
o
, the ratio of the photon flux entering the device to the
flux of photons incident on its surface. These two ratios, which are
both dependent on the wavelength, are considered in the following
text.
2.3.2.1. Optical efficiency
The optical refractive index n of the semiconductors under
consideration is in the region of 3, and as a result the incident light
rays are refracted and reflected at its surface. At normal incidence,
used as the reference situation, the Frensel transmission coefficient
determines the optical efficiency:
K
o
=
4n
n÷1
( )
2
[2.17]
which is of the order of 70% and depends little on the wavelength and
the angle of incidence. It is possible to achieve optical efficiencies
close to 100% at a given wavelength O, by depositing an antireflection
28 Optoelectronic Sensors
coating on the surface, for example of refractive index n and
thickness O 4 n
( )
. The optical efficiency therefore depends on O and
on the angle of incidence.
2.3.2.2. Internal quantum efficiency
The electron-hole pairs created by the absorption of photons
entering the device may, or may not, give rise to a current flow in the
external circuit, depending on where they are generated in the
structure. If generation is in the depleted region, the electrons and
holes are separated by the electric field, as shown in Figure 2.1,
leading to the passage of one electron for each absorbed photon: thus,
the internal quantum efficiency is 100%. In the case of generation in
the electrically neutral P and N regions, where the electric field is
weak, only the electrons (or holes) generated less than one diffusion
length away are transported into the depleted region, and the internal
quantum efficiency is in most cases weak.
It is therefore desirable to design a photodiode in such a way that
the radiation should be entirely absorbed in the depleted region.
Figure 2.8 shows a model which provides a good evaluation of the
internal quantum efficiency of a PIN photodiode, as a function of the
nature of the material on whose surface the radiation falls. The curves
1 and 2 represent the profiles of the radiative generation rate G
R
across
two structures. The structure for curve 1 is that of a homojunction PIN,
where all of the materials used have the same width of bandgap,
which is below the energy of the photons to be detected. The radiation
is assumed to be incident from the p side, and part of the radiation is
absorbed there, where the quantum efficiency is negligible. Another
part is absorbed in the i region, and has a quantum efficiency of 100%.
Finally, the contribution of the residual radiation entering the n region
is also negligible. The quantum efficiency can therefore be written:
K
i
=exp ÷Ddp
( )
1÷exp ÷Dd
( )

l
l
l
[2.18]
where the definitions of d
P
and d can be found in Figure 2.8.
PIN Photodiodes for the Visible and Near-Infrared 29
Depleted region
Figure 2.8. Optical generation rate across the structure of two PIN
photodiodes: curve 1 shows a homojunction PIN photodiodes;
curve 2 shows a heterojunction PIN photodiode
Curve 2 corresponds to a PIN photodiode, with the capital P
indicating that the material in the P region has a bandgap width greater
than the energy of the photons to be detected and which is therefore
transparent to them. The same reasoning provides us with the quantum
efficiency:
K
i
=1÷exp ÷Dd
( )
[2.19]
2.3.3. Dark current
The dark current is the sum of several contributions: non-radiative
generated currents requiring an input of energy (such as the SRH
process) and the tunneling current, which does not require such an
input. These two processes are represented in Figure 2.9 for a sharp
p
+
n homojunction.
30 Optoelectronic Sensors
2.3.3.1. Generation currents
According to equation [2.9], the dark current contribution of
generation in the depleted region is:
I
0
= SdG
NR
= Sdn
i
W
eff
[2.20]
Taking account of equation [2.11] – of the fact that the shortage of
electrons (holes) in the p(n) material near the depleted region is
2
, ,
i AD
n p n N G = and the fact that the thickness of the p(n) material is
L
n,p
(see section 2.2.3.2) – the generation contribution in the neutral p
and n regions can be written:
I
0
= Sqn
i
2
L
n
N
A
W
n
÷L
p
N
D
W
p
( )
[2.21]
This is the traditional expression for the reverse current of a diode
according to Shockley.
Figure 2.9. Profile of the VB and the CB of a reverse-biased p+n junction.
The generation process (shown here in the depleted region) requires an
input of energy; the tunnel process occurs at constant energy
PIN Photodiodes for the Visible and Near-Infrared 31
Thus, we see from equations [2.20] and [2.21] that the generation
current is primarily a function of n
i
and varies very rapidly with E
G
/kT.
2.3.3.2. Tunneling current
Using the tunneling effect electrons can also be transported at a
constant energy across the triangular potential barrier present at the
junction, as long as it is narrow enough. A calculation of the current
gives the following expression [STI 82]:
I
T
= S
2m*q
3
EV
4S
3
h
2
E
G
1 2
exp ÷
S 2m* E
G
3 2
4qhE
(
(
·
·
·
·
·
·
)
)

[2.22]
where E=V/d is the electric field in the depleted region, which is
constant for a PIN diode. From equation [2.22] we determine that, for
a given material, I
T
increases very rapidly with the voltage V across
the terminals of the diode, mostly as a result of the increase in E.
Additionally, I
T
increases when E
G
decreases, such that the
contribution of the tunneling current to the dark current is most
significant for photodiodes made of materials with a small bandgap.
2.3.4. Breakdown voltage
2.3.4.1. Zener breakdown via the tunneling effect
The very rapid variation of the tunneling dark current with the
reverse-bias voltage is the first process of breakdown. This mainly
applies to materials with a narrow depleted region and a small
bandgap; the two parameters which reduce the width and the height of
the potential barrier at the PI junction. The tunneling effect can be the
dominant one in the materials used for detection of the longest
wavelengths considered in this chapter (see section 2.5.3.2), if the
doping level of the i zone is too high [TAK 80].
32 Optoelectronic Sensors
2.3.4.2. Avalanche breakdown
We can define a critical electric field E
C
, of the order of
2 x 10
5
V/cm for the lightly-doped materials considered in this chapter,
where the multiplication coefficient becomes infinite and leads to
breakdown. In the case of PIN photodiodes, which are not designed to
take advantage of avalanche gain, breakdown can arise locally, most
often around the edge of the junction where the point effect increases
the electric field. The breakdown voltage is therefore approximated
below by:
V
C
= E
C
d [2.23]
2.3.5. Saturation current
2.3.5.1. Extrinsic limitation by the voltage drop across
the load resistance
In order to have a linear response, the photodiode must remain
reverse-biased (V<0) so that the carriers are separated by the electric
field in the depleted region. Thus, it can be seen in the simple bias
diagram of Figure 2.7 that for a bias voltage V
0
, the voltage drop in
the load resistance R
L
can cancel out the voltage V across the diode
terminals for a limiting current I
M
such that:
I
M
=V
0
R
L
[2.24]
2.3.5.2. Intrinsic limitation by the space charge of mobile
charge carriers
It is especially important to take into account the contribution of
mobile charge carriers to the electric field E in the depleted region for
high levels of incident optical power. Specifically, E must satisfy the
following conditions:
0 _ E _ E
B
[2.25]
in order to avoid breakdown of the junction.
PIN Photodiodes for the Visible and Near-Infrared 33
We will determine the maximum value I
M
of the photocurrent that
simultaneously satisfies these conditions for a heterojunction PIN
structure, with a coefficient of absorption in the i region such that Dd>>1,
and all the illumination is absorbed close to the Pi junction. We further
assume that the mobile electrons in the i region are transported with the
saturation velocity v
s
, independent of the electric field.
Figure 2.10 shows the density profiles of the fixed and mobile
charges and the electric field profile across the structure. Given the
hypotheses set out in this section, it follows that the charge density
due to electrons is ,
s
J U Q = where J is the current density, the
electric field variation in the depleted region is G E = Ud H and the
maximum current I
M
, obtained when G E = E
B
, can be written:
I
M
=C
J
Q
s
E
B
[2.26]
where C
J
is the capacitance of the junction.
Depleted region
Depleted region
Figure 2.10. Space charge density and electric field, for I=0 (dotted line)
and for I=I
M
(solid line). For I=I
M
, the electric field varies
from 0 to the breakdown field E
B
34 Optoelectronic Sensors
Note that, for this value of current, the voltage across the terminals
of the junction is
V = E
B
d 2 =V
B
2 [2.27]
We can easily show that, regardless of the components in the bias
circuit, the part of the curve representing the variation of the electric
field in the depletion region (marked :) rotates around a fixed point
as the photocurrent varies. Its position depends on the values of V
0
and
RL; Figure 2.10 provides an example where R
L
= 0 and the voltage
across the diode terminals is constant and equal to the bias voltage V
0
= V
C
/2.
2.4. Dynamic characteristics of PIN photodiodes
2.4.1. Intrinsic limitations to the speed of response
2.4.1.1. Transit time in the depleted region
The transit time of photocarriers in the depleted region is the first
limit to the response time of the photodiode. Its analysis is carried out
here with the same approximations as in section 2.2.1, which is a good
representation of the properties of real PIN diodes. Figure 2.11 shows
the electric charge profile in the depleted region at a time t after a brief
impulse signal, of energy W
0
, which was applied to the photodiode at
time 0. The packet of generated electrons is transported at speed v
s
and takes time W
t
=dv
s
to cross the depleted region.
Depleted region
Figure 2.11. Density of mobile charges produced in the depleted region of a
PIN photodiode in response to an optical impulse G (z, t)
PIN Photodiodes for the Visible and Near-Infrared 35
Figure 2.12. Impulse and frequency response of a
PIN photodiode limited by the transit time
The solution of Maxwell’s equations for the case R
L
=0 shows that
the current generated by the diode is the rectangular impulse shown in
Figure 2.12, of duration W
t
and of intensity:
I
0
= q hQ
( )
W
0
W
t
[2.28]
The frequency response can then be obtained by Fourier
transforming the impulse response. This is shown in Figure 2.12 and
given by:
I f
( )
= I
0
sin S f W
t
( )
S f W
t
[2.29]
The 3 dB cutoff frequency I f
( )
I
0
=1 2
( )
is:
f
t
=0.44 W
t
[2.30]
We see that this varies as 1/d. In the case where Dd <<1 and the
generation is uniform in the depleted region, the more accurate model
presented in section 2.4.2.2 shows that we obtain the following
numerical result:
f
t
=0.55 W
t
[2.31]
We see that regardless of the value of Dd, we can apply this
approximate numerical expression, which we will continue to use:
36 Optoelectronic Sensors
f
t
=1 2W
t
[2.32]
Also, equation [2.19] shows that if Dd <<1 we have:
K
i
=Dd [2.33]
Thus we see from equations [2.32] and [2.33] that the product K
i
f
t
only depends on the characteristics of the material in the i region:
K
i
f
t
=DQ
s
2 [2.34]
from which we conclude that, for the geometry considered and for a
particular wavelength and material, we must accept a compromise
between the internal quantum efficiency and the speed of response.
2.4.1.2. Diffusion time in the neutral regions
When the optical generation occurs in one of the electrically
neutral regions, n or p, the minority photocarriers must initially be
transported by diffusion to the edges of the depleted region, in which
they can be transported by the electric field. This process can be
characterized by a diffusion time W
d
which we can show has an order
of magnitude:
W
d
=d
2
D
n p
[2.35]
for electrons (holes) generated in the p(n) region. In practice the
diffusion time is always greater than the transit time in the depleted
region and has the effect of reducing the response speed of
photodiodes. Therefore, excluding some exceptions (see section 2.6.4),
it is desirable to design these devices such that the absorption of
photons takes place in the depleted region, both to optimize the speed
of response and the quantum efficiency.
2.4.1.3. Trapping at the Pi heterojunction
An additional parasitic mechanism can slow the speed of
heterojunction PIN photodiodes. The holes generated in the i
region, which has a small bandgap, can be trapped at the interface
PIN Photodiodes for the Visible and Near-Infrared 37
with the P region by discontinuity in the valence band. They can
then be emitted by thermo-ionic effects, although with delay, into
the P region. This effect is particularly marked for holes, whose
effective mass is greater than that of electrons. The proposed
remedy [FOR 82] is the use of gradual heterojunction or layers
with an intermediate-sized bandgap in order to reduce the potential
barrier presented to the holes and to make it easier for them to
cross through tunneling or thermal effects. We will assume below
that the structures of the devices are designed in such a way that
these effects are rendered negligible.
2.4.2. Limitations due to the circuit
2.4.2.1. Response of the basic circuit, optimum thickness
of the i region
In order to produce a signal, a photodiode is inserted into a
circuit whose components also contribute to limit the bandwidth of
an optical detector. The most simple circuit is shown in Figure 2.13,
consisting of the capacitance of the junction C
J
in parallel with a
load resistance R
L
, with a negligible impedance of the source at
high frequencies. This circuit is driven by the photodiode generated
current, whose response is independent of the voltage at its
terminals.
38 Optoelectronic Sensors
Figure 2.13. 3 dB cutoff frequency of a PIN photodiode, as a function of the thickness
of the depleted region. The material is assumed to be GaInAs/InP, the width of the
active region 10μm and the load resistance 50 :
The impulse and frequency response of such a resistor-capacitor
network are well known. The frequency response is given by the
expression:
I f
( )
=
I
0
1÷ 2S fR
L
C
J
( )
2
[2.36]
and the 3 dB cutoff frequency is:
f
RC
=1 2SR
L
C
J
( )
[2.37]
We see that f
RC
varies with d. The cutoff frequencies f
W
and f
RC
vary inversely with d, thus there is an optimal thickness for the
depleted region which maximizes the global cutoff frequency f
3dB
.
Figure 2.13 shows the cutoff frequencies variation with d for a typical
PIN photodiode.
PIN Photodiodes for the Visible and Near-Infrared 39
The optimal thickness of d is obtained when f
W
= f
RC
, giving:
W
t
=S R
L
C
J
[2.38]
The maximum cutoff frequency is then
f
3dB
=
Q
s
2 2d
[2.39]
2.4.2.2. Response of a real PIN photodiode
It is also interesting to consider the results of a more complete
analysis [BOW 87] of a photodiode made of GaInAs/InP, designed for
optical telecommunications at a wavelength of 1.3 μm, taking into
account:
– the finite value of the optical absorption coefficient: D = 1.3
10
4
cm
–1
;
– the different saturation velocities for electrons and holes Q
n
= 6
10
6
cm/s and Q
p
= 4.8 10
6
cm/s;
– the inevitable parasitic capacitances of the circuit: C
P
= 50 fF;
– a parasitic inductance L, caused by the connecting wires,
adjustable and kept as a free parameter.
The illumination is assumed to be incident at the N side of the
structure, and the diameter of the active region is assumed to equal
30 μm. Figure 2.14 shows the response curves of the photodiode
inserted in its circuit, with five examples of the parameters d and L
listed in Table 2.1 along with the performances obtained, the cutoff
frequency and the internal quantum efficiency. The load resistance
was kept constant: R
L
= 50 :.
40 Optoelectronic Sensors
Table 2.1. Values of the parameters for the curves in Figure 2.17
Figure 2.14. Frequency response of a PIN photodiode made
of GaInAs/InP at 1.3μm, inserted into the circuit shown.
The parameters of each curve are given in Table 2.1
Curves 1 and 2 show behaviors limited by the transit time and by
the circuit respectively. Curve 3 corresponds to a depleted region
whose width is close to the optimum, while curve 4 shows that its
response can be further improved by parasitic induction. For a higher
value of inductance (curve 5), the bandwidth is reduced, but a gain in
signal is obtained.
PIN Photodiodes for the Visible and Near-Infrared 41
2.4.3. Power-frequency compromise, Pf
2
“law”
For certain optical transmission systems, it is desirable to use a
distortion-free increased electrical power in the sinusoidal regime
across the resistance R
L
loading the photodiode. In particular, this
allows us to simplify the electronics of the receiver, and is made
possible by the current availability of powerful modulated optical
sources or optical preamplifiers. Given the maximum frequency to be
transmitted, the problem is now to design a photodiode, defined
mainly by the thickness d of the depleted region and by the area S of
the active region, as well as select the bias voltage V
0
in a circuit, as in
Figure 2.7. We assume that the incident optical power is modulated at
a level of 100%.
Figure 2.15. Excursion from the operating point of the photodiode in a plot of its I/V
characteristics: allowing us to determine the maximum RF electrical power. Points A,
A’ and A’’ are respectively the mean and extreme points
In order to maximize the RF power dissipated in the load, the
photocurrent must vary sinusoidally (see Figure 2.15) between 0 and
the maximum value I
M
given by equation [2.25], as well as the voltage
varying between the breakdown voltage V
B
given by equation [2.23]
and the voltage V
B
/2 (equation [2.27]). The maximum effective power
is therefore:
P
M
=
1
2R
L
V
C
4
(
(
·
·
·
·
)
)

2
=
E
B
2
d
2
32R
L
[2.40]
42 Optoelectronic Sensors
If we assume that the photodiode is optimized for its frequency
response, and we eliminate d from [2.39] and [2.40], it follows that:
P
M
f
3dB
2
R
C
=
E
C
2
Q
s
2
256
[2.41]
This expression shows that, for a fixed value of the load resistance,
the product P
M
f
3dB
2
only depends on the properties of the
semiconductor material used for the i region. We should note that the
numerical coefficient 1/256 is a maximal value which may in fact
depend on the detailed specifications of the nonlinearity of the
transmission system considered. Equation [2.41] is analogous to the
known result for electronic components such as transistors [JOH 64].
We can note that this limitation a photodiode can generate on the RF
power is of a purely electronic origin; in addition, thermal effects can
severely limit it, in particular for components designed for low
frequencies.
2.5. Semiconductor materials used in PIN photodiodes for the
visible and near-infrared
2.5.1. Absorption of semiconductors in the range 400-1,800 nm
The absorption functions of the main semiconductor materials or
semiconductor families used to build PIN photodiodes in the spectral
range considered in this chapter are shown in Figure 2.16. They can be
chosen as a function of the wavelengths to be detected. Other
materials have been studied in the past, but have not taken off in
practice for numerous reasons; these will be mentioned briefly later on.
PIN Photodiodes for the Visible and Near-Infrared 43
, , , , ,
Figure 2.16. Absorption spectra for semiconductor materials used in the
visible to near-infrared range
2.5.2. From 400 to 900 nm: silicon and the GaAlAs/GaAs family
These two materials have achieved a good technological maturity
under the demands of the integrated circuit market, and as such they
are the natural choice for this spectral domain. However, the shape of
their absorption functions demonstrates their difference in bandgap
nature, which is indirect for silicon and direct for gallium arsenide.
2.5.2.1. Silicon
Silicon has an indirect bandgap and its absorption coefficient
slowly increases below the corresponding wavelength (O
G
=
1,100 nm), reaching the value of 10
3
cm
–1
for O = 800 nm. The
interface with its natural oxide, SiO
2
, produces very few electronic
states in the bandgap, which is the main factor enabling the
development of planar and MOS technologies, and ultimately very
high density integrated circuits. On the other hand, this semiconductor
has not, as yet, contributed to the design of heterojunction
optoelectronics with competitive performance.
44 Optoelectronic Sensors
Figure 2.17 shows the spectral response of a silicon photodiode
designed for optical fiber telecommunications. The structure of the
device is chosen (with a depleted region thickness around 30 μm) in
order to increase quantum efficiency to between 600 and 900 nm
where weakly-attenuating fibers and high performance electro-
luminescent emitters are available. It is possible to extend the spectral
response slightly towards the infrared by increasing the thickness of
the intrinsic region, in particular for detection of the O = 1,060 nm
radiation from solid-state neodymium-doped YAG lasers. The drop in
efficiency at short wavelengths, due to optical absorption in the
surface layer, can also be minimized in the near-ultraviolet by several
methods.
The response speed is often limited to less than 1 GHz by the
transit time, in particular for components whose response has been
extended into the infrared. The dark current is mostly caused by
current generation in the depleted region, thanks to the effectiveness
of the surface silicon passivation, which is of the order 10
–7
A/cm
2
.
The maturity of silicon technology allows the construction of a
variety of photodiode types, potentially with multiple regions of
sensitivity, and offered at a low price. As a result, they are used for
short-range telecommunications in numerous professional systems
(automation, Lidar, etc.) or mass-market systems (optical disk readers,
photographic equipment, etc.). CCD and MOS imagers also form part
of the family of silicon-based photodetectors.
PIN Photodiodes for the Visible and Near-Infrared 45
,
Figure 2.17. Spectral response of photodiodes, made of silicon
for optical telecommunications and of experimental
GaAlAs/GaAs without antireflection treatment
2.5.2.2. Gallium arsenide and similar materials
Gallium arsenide has a direct bandgap; its absorption coefficient is
approximately 10
4
cm
–1
for O = 800 nm and increases very rapidly
below the corresponding wavelength O
G
(O
G
= 870 nm). Highly
developed for the construction of microwave integrated circuits and
diode lasers, it is available as a high-quality substrate for creating
defect-free heterojunctions with Ga
1–x
Al
x
As, Ga
1–x
In
x
As
1–y
P
y
and
Al
x
Ga
y
In
1–x–y
As alloys.
In fact, there has been interest in heterojunctions from the
GaAs/GaAlAs system for photodiode and solar cell fabrication since
the 1970s [ALF 70], and components showing good performance have
been demonstrated [LAW 79]. A spectral response function is given in
Figure 2.17, showing the typical shape for a heterojunction photodiode:
rapid cutoff at long wavelengths, absorption in the direct-bandgap i
material, internal quantum efficiency close to 100%, and cutoff at
short wavelengths due to absorption from the large-bandgap window
layer. The density of the dark current is of the order of 10
–8
Acm
–2
at
V
B
/2 and a product K
i
f
t
of the order of 20 GHz can be achieved for
conventionally-structured devices. Alongside heterojunctions, the use
of Schottky diodes based on GaAs has also been demonstrated. In
46 Optoelectronic Sensors
addition, the deposition of ITO (indium tin oxide), a conductive and
transparent material, has allowed the development of fast photodiodes
based on GaAs [PAR 87]. MSM (metal-semiconductor-metal)
photodiodes, discussed in Chapter 5, can also be thought of as
Schottky photodiodes.
Despite these encouraging and already well-known characteristics,
GaAs-based photodiodes have not seen significant development since
2001, except in solar cells. This is probably due to the cheap cost of
silicon components whose speed of operation is adequate for real-
world systems. This situation may change if the need arises for optical
transmissions above 1 Gbit/s within the spectral range covered in this
section.
2.5.3. From 900 to 1,800 nm: germanium, GaInAsP/InP…
This spectral range is mostly used for long distance transmissions
and broadband over silica fibers. These systems have driven the
development of photodetectors, initially in germanium – which
predated the emergence of optical telecommunications – and then in
various families of semiconductor compounds, with the GaInAs alloy
on InP substrate finally prevailing.
2.5.3.1. Germanium
Germanium has an indirect bandgap and its absorption coefficient
increases slowly below O
G
, then more rapidly for O < 1,500 nm where
its conduction band also has a direct minimum for k = 0 (see Figure
2.4). The need for photodetectors for telecommunications was the
drive for an important effort in the 1970s in the development of
germanium photodetectors [KAN 85]. However, several factors
finally brought an end to their application in this domain, along with
the increasing use of GaInAs/InP:
– non-existence of materials allowing the construction of
heterojunctions;
PIN Photodiodes for the Visible and Near-Infrared 47
– limitation of the spectral response around 1,550 nm for fast
photodiodes, due to the shape of the absorption spectrum;
– high dark current due to an insufficient control of the surface
preparation.
, , , , ,
Figure 2.18. Spectral response of state-of-the-art photodiodes in 2000: in germanium
for instrumentation and in GaInAs/InP for optical fiber telecommunications
Figure 2.18 shows the spectral response of a germanium
photodiode aimed at scientific instrumentation. The dark current
densities are of the order of 10
–4
Acm
–2
and the cutoff frequencies are
limited by the circuit to a few hundred MHz, depending on the active
surface.
2.5.3.2. Indium phosphate and related materials
Solid solution crystals from the Ga
x
In
1–x
As
y
P
1–y
family formed into
a crystal lattice on InP substrates (y # 2.2x) all have an indirect
bandgap, with Ga
0.43
In
0.57
As having the smallest bandgap of the
family (O
G
= 1,650 nm). Its absorption spectrum, as well as that of InP
(O
G
= 920 nm), is shown in Figure 2.16. The composition of the solid
solution can be selected to give a bandgap width anywhere between
those two values.
48 Optoelectronic Sensors
The technology of these materials was initially driven by the
development of electroluminescent emitters, electroluminescent
diodes and laser diodes, required for long distance transmissions. It
has been of benefit to photodetectors which have established
themselves as rivals to germanium photodetectors.
The main stages in the “planar” fabrication procedure for PIN
heterojunction photodiodes are shown in Figure 2.19. It is analogous
to that of the devices described above, but we note in particular that
the PIN junction emerges at the surface in the wider-bandgap InP
material, which contributes significantly to the reduction in the dark
current. The transparency of the substrate allows the possibility of the
illumination being in this direction (rear illumination).
Figure 2.19. Fabrication stages of PIN photodiodes in GaInAs/InP by “planar”
technology: 1) epitaxial deposition of the absorber and window layers,
2) deposition of the dielectric and etching of openings, 3) zinc diffusion
(p-type impurity), and 4) attachment of metallic contacts
Figure 2.18 shows the spectral response of a photodiode targeted
for optical telecommunications. This diode has had an antireflection
treatment applied which allows it to approach an external quantum
efficiency of 100% across a large spectral band. The dark current
density is of the order of 10
-5
Acm
-2
and, as we saw in section 2.4.2.2,
the cutoff frequency of a photodiode with an optimized structure can
surpass 20 GHz while retaining a high quantum efficiency. The
spectral response is limited to around 1,700 nm when the absorbing
layer is exactly matched to the InP substrate. This can be extended
beyond 2,000 nm with the use of unmatched materials, at the expense
PIN Photodiodes for the Visible and Near-Infrared 49
of the dark current, for certain applications (imaging, spectroscopy,
etc.) [LIN 92].
2.5.3.3. Other materials [PEA 85]
The integration of the GaAlAsSb family into GaSb substrates was
also considered by several groups, towards the end of the 1970s, for
the fabrication of photodiodes for optical transmissions. GaSb is a
direct semiconductor whose bandgap is close to that of GaInAs
(O
G
= 1,720 nm). By combining it with GaAlAsSb, it was possible to
demonstrate photodiodes with competitive quantum efficiencies,
although with a higher dark current (by a factor of ~10) than
GaInAs/InP photodiodes. On the other hand, a weaker additional
avalanche noise was achieved (see Chapter 3). Nevertheless, the
industrial significance of GaInAsP/InP, also used for
electroluminescent emitters, seems to have prevented investment in
research which might have taken this technology to new levels.
The HgCdTe (MCT, Mercury Cadmium Telluride) integrated into
CdTe substrates is the dominant material for detection and imaging in
the mid-infrared. Attempts were made to re-use the technological
achievements in the infrared for the construction of detectors for
telecommunications. Although the performances obtained were
comparable to those of GaInAs/InP, as in the case of GaInAsSb/GaSb,
the potential gains were insufficient to justify its industrial
development.
2.6. New photodiode structures
2.6.1. Beyond the limits of conventional PIN
The development in the last decades of new epitaxial techniques
for the integration of very thin layers (only a few atoms thick) of
semiconductor compounds has enabled the design of optoelectronic
devices, including photodetectors with far more complex structures
than the simple PIN photodiodes with three useful layers. In this way
it has become possible to exceed the limit of the product K
i
f
t
, which
50 Optoelectronic Sensors
we examined in section 2.4.1.1 (for the PIN photodiode geometries
discussed), whose order of magnitude is 30 GHz for indirect bandgap
materials. Several avenues were explored with this aim, all based on
making the radiation pass through the narrow i absorption region
multiple times (as shown in Figure 2.20) in order to increase the
internal quantum efficiency while maintaining a short electronic
transit time.
2.6.2. Photodiodes with collinear geometry
A first approach to increase the absorption of radiation by a layer
of thickness Dd << 1 is to cause the light to pass through the absorbing
layer again, after reflection from a mirror at the back (see Figure 2.20),
which results in a doubling of the absorbing thickness. Such a mirror
can be fabricated either by metalizing the rear face of the structure or,
as shown in the figure, by a Bragg semiconductor mirror. These
mirrors made from a stack of layers with an alternating sequence of
bandgaps (and hence refractive indices) of thickness O/2n allow
reflection coefficients approaching 100% to be achieved. They were
first used in the fabrication of vertical cavity surface emitting lasers
(VCSELs) and more generally in microcavity devices.
One technique for increasing the number of crossings of the
absorbing region is to insert the i layer between two mirrors, possibly
Bragg mirrors, to form a Fabry Perot cavity (or interferometer), as
shown in Figure 2.20.
Figure 2.20. Photodiodes of collinear geometry with an
improvement in the K
i
f
t
tradeoff
PIN Photodiodes for the Visible and Near-Infrared 51
The analysis of the behavior of such a structure [KIS 91] shows
that when the effective distance between the mirrors is an integer
number of half-wavelengths of light in the material, and the i layer is
placed at an antinode of the electromagnetic field, the absorption can
be considerably enhanced, as can the internal quantum efficiency.
This enhancement is obtained for a series of resonant wavelengths
satisfying the above condition, which gives rise to the name RCE
(Resonant Cavity Enhanced) photodiodes. However, this resonant
behavior has a marked effect on the characteristics of the device:
– the spectral response consists of a series of peaks which become
even narrower as the fineness of the cavity is increased, which is in
contrast to the normally broad response of conventional photodiodes
(see Figures 2.17 and 2.18). RCE photodiodes can be thought of as the
monolithic integration of a photodiode and an optical filter whose
selectivity can be exploited beneficially, in particular if it can be made
tunable;
– for a given wavelength, the resonance condition is satisfied only
for certain angles of incidence, such that the directivity is much higher
than that of conventional devices.
Analysis of the response speed [UNL 95] shows that a product K
i
f
t
approaching 100 GHz should be obtainable by RCE photodiodes. In
addition, other advantages of the use of very thin detector layers are
anticipated including:
– the reduction of the contribution of pair generation to the dark
current (equation [2.20]);
– the possibility of using partially incommensurate i materials
without loss of crystal quality if its thickness is below a critical value.
The experimental results obtained up to now for most of the
families of III-V materials have indeed confirmed the potential of
RCE photodiodes in terms of quantum efficiency and selectivity, but it
turns out to be difficult to show improvement in the K
i
f
t
product.
52 Optoelectronic Sensors
2.6.3. Waveguide photodiodes
Another approach consists of using an orthogonal geometry, giving
the device the shape of an optical waveguide, which is analogous to
that of a diode laser. This is the case in Figure 2.21, where we see an
absorbing i layer inserted between several other layers which have a
wider bandgap, are transparent and also have a lower refractive index.
Figure 2.21. Structure, width of the bandgap and doping of the layers, of a waveguide
photodiode with orthogonal geometry and with an improved K
i
f
t
tradeoff
In Figure 2.21, a geometric optics representation, the light rays
propagate in the structure through a series of total internal reflections,
and cross the i layer several times. In reality, since the dimensions of the
waveguide are of the order of the wavelength of light in the material, a
more rigorous analysis of the electromagnetic field shows that only
certain configurations (or modes) are supported. The radiation is
progressively absorbed with an effective coefficient of absorption
depending on the overlap between the guided mode and the absorbing
layer; this can be adjusted by the design of the waveguide, but is
difficult to reduce below 100 cm
–1
, leading to an internal quantum
efficiency of 95% for a waveguide of length 300 μm.
Numerous variations on this principle have been investigated, such
as the active part of the photodiode being localized on only part of the
waveguide. Figure 2.21 shows a structure with 5 active layers, allowing
independent adjustment of the transit time (by varying the thickness of
the i depleted region) and the waveguide thickness (which determines
the degree of coupling with the incident light and consequently the
external quantum efficiency [KAT 92]):
PIN Photodiodes for the Visible and Near-Infrared 53
– InP N layer (O
G
= 920 nm);
– GaInAsP N layer (O
G
= 1,300 nm);
– GaInAs i layer (O
G
= 1,650 nm);
– GaInAsP P layer (O
G
= 1,300 nm);
– InP P layer (O
G
= 920 nm).
In this way it has been possible to demonstrate a significant gain in
the merit factor K
i
f
t
# 30 for GaInAs. However, it should be noted that
the gain in internal quantum efficiency obtained, can in practice be
partially undone by the difficulty of coupling the radiation into the
mode(s) supported by the waveguide. Although it is in theory possible
to couple the mode from a single-mode fiber into a photodiode of
single mode with an efficiency of 100%, this can require coupling
optics of prohibitive cost.
2.6.4. Traveling-wave photodiodes
Waveguide photodiodes have the same limitation as conventional
structures, in terms of the power/frequency tradeoff discussed in
section 2.4.3. This limitation stems from the lumped-element circuit
analysis, which is a good approximation as long as the length of the
guide remains smaller than the wavelength of the electrical signal
(f = 100 GHz, O = 3 mm). Beyond this, it is necessary to use a
distributed element model and to think of a photodiode as a
waveguide coupled to a transmission line (see Figure 2.22),
terminated at both ends in its characteristic impedance in order to
maximize the bandwidth.
Such devices have been proposed and analyzed under the term
“traveling-wave photodiodes” [GIB 92]. It follows that, aside from the
transit time, the circuit bandwidth is limited by the mismatch between
the propagation speeds of the two guided waves, optical and
microwave – a limit which is not fundamental in nature, but which is
difficult to overcome in practice. The analysis of [GIB 92] predicts
values for the product K
i
f
t
of greater than 100 GHz; however,
54 Optoelectronic Sensors
fabrication difficulties appear, until now, to have seriously limited the
practical performances relative to the potential performance: the
challenge still remains to conceive and then construct structures
allowing simultaneous improvement in K
i
f
t
and Pf
2
(the figure of
merit).
Figure 2.22. Traveling-wave photodiode
2.6.5. Beyond PIN structures
Whatever their geometry, the photodiodes considered up to this
point have had, by definition, three active regions, with the i region
fulfilling both the role of radiation absorption and also the role of the
depleted region. It is possible to imagine photodiodes where these
functions are separated so that they can be optimized separately,
possibly with the additional use of waveguide photodiode or resonant
cavity concepts.
2.6.5.1. P
+
p
+
N

N
+
structures
The absorbing region, here p+, can be doped without reducing the
internal quantum efficiency and the speed, on the condition that it is
thin enough to ensure both efficient and rapid transport of
photocarriers by diffusion (see sections 2.3.2.2 and 2.4.1.2). An initial
example is that of electroluminescent diodes whose optimization leads
to a P
+
p
+
N

N
+
double heterojunction structure, where the emitting
layer is designated p+. Under reverse-bias the depleted region grows
into the N- region and the photodetection, quantum efficiency and
speed characteristics are sufficient for some applications
PIN Photodiodes for the Visible and Near-Infrared 55
K
e
=0.6, f
RC
=300 MHz
( )
[POU 80]. Very fast response speeds have
more recently been discovered to be at the expense of low quantum
efficiency K
e
=0.06, f
3dB
=300 GHz
( )
in a conventional geometry,
making more use of the high mobility regime of electron transport
[ITO 00].
2.6.5.2. P
+
IiIN
+
structures
The placement of the absorbing i region in a thicker I depleted
region, in order to equalize the transit time of electrons generated in
the i layer towards the N+ layer and of holes towards the P+ layer,
allows us in principle to gain a factor of 2 on the cutoff frequency f
t
[UNL 95].
2.7. Bibliography
[ALF 70] ALFEROV Z.I., ANDREEV V.M., ZIMOGOROVA N.S., TRET’YAKOV D.N.,
“Photoelectric properties of AlxGa1-xAs heterojunctions”, Sov. Phys.-Semicond.,
vol. 3, no. 11, p. 1373-1376, May 1970.
[BOW 87] BOWERS J.E., BURRUS C.A., “Ultrawide-band long-wavelength p-i-n
photodetectors”, J. Lightwave Technology, vol. LT-5, no. 10, p. 1339-1350,
October 1987.
[FOR 82] FORREST S.R., KIM O.K., SMITH R.G., “Optical response time of
In
0,53
Ga
0,47
7As/InP avalanche photodiodes”, Appl. Phys. Lett., vol. 41, no. 1,
p. 95-98, July 1982.
[GIB 92] GIBONEY K.S., RODWELL M.J.W., BOWERS J.E., “Travelling-wave
photodetectors”, IEEE Photonics Technology Lett., vol. 4, no. 12, p. 1363-1365,
December 1992.
[ITO 00] ITO H., FURUTA T., KODAMA S., ISHIBASHI T., “InP/InGaAs uni-travelling-
carrier photodiode with 310 GHz bandwidth”, Electron. Lett., vol. 36, no. 21,
p. 1809-1810, 12 October 2000.
[JOH 64] JOHNSON E.O., “Physical limitations on frequency and power parameters of
transistors”, IEEE Int. Conf. Rec. pt.5, p. 27-34, 1965.
[KAN 85] KANEDA T., “Semiconductors and semimetals”, Lightwave Communication
Technology, vol. 22, part D, chap. 3, p. 247-328, Academic Press, 1985.
[KAT 92] KATO K., HATA S., KAWANO K, YOSHIDA J., KOZEN A., “A high-efficiency
50 GHz InGaAs multimode waveguide photodetector”, IEEE J. Quantum
Electronics, vol. 28, no. 12, p. 2728-2735, December 1992.
56 Optoelectronic Sensors
[KIS 91] KISHINO K., UNLU M. S., CHYI J-I., REED J., ARSENAULT L., MORKOC H.,
“Resonant cavity enhanced (RCE) photodetectors”, IEEE J. Quantum
Electronics, vol. 27, no. 8, p. 2025-2034, August 1991.
[LAW 79] LAW H.D., NAKANO K., TOMASETTA L.R., “State-of-the-art performance of
GaAlAs/GaAs avalanche photodiodes”, Appl. Phys. Lett., vol. 35, no. 2, p.
180-182, July 1979.
[LIN 92] LINGA K.R., OLSEN G.H., BAN V.S., JOSHI A.M., KOSONOCKI W.F., “Dark
current analysis and characterization of InxGa1-x As/InAsyP1-y graded
photodiodes with x > 0.53 for response to longer wavelength (> 1.7 μm)”, J.
Lightwave Technology, vol. 10, no. 8, p. 1050-1055, August 1992.
[MAT 87] MATHIEU H., Physique des semi-conducteurs et des composants
électroniques, Masson, 1987.
[PAR 87] PARKER D. G., SAY P. G., HANSOM A.M., SIBBETT W., “110 GHz high-
efficiency photodiodes fabricated from indium tin oxide/GaAs”, Electron. Lett.,
vol. 23, p. 527-528, 1987.
[PEA 85] PEARSALL T.P., POLLACK M.A., “Semiconductors and semimetal”,
Lightwave Communication Technology, vol. 22, Part D, Chapter 2, p. 173-247,
Academic Press, 1985.
[POU 80] POULAIN P., DE CREMOUX B., “Diffusion limited transient response of
heterojunction photodiodes”, Japanese J. Appl. Phys., vol. 19, no. 4, p. L189-
L192, April 1980.
[STI 82] STILLMAN G. E., COOK L.W., BULMAN G.E., TABATABAIE N., CHIN R.,
DAPKUS P.D., “Long-wavelength (1.3 to 1.6 μm) detectors for fiber-optical
communications”, IEEE Trans. on Electron Devices, vol. ED-29, no. 9, p. 1355-
1371, September 1982.
[SZE 81] SZE S.M., Physics of Semiconductor Devices, 2
nd
edition, John Wiley &
Sons, 1981.
[TAK 80] TAKANASHI Y., KAWASHIMA M., HORIKOSHI Y., “Required donor
concentration of epitaxial layers for efficient InGaAsP avalanche photodiodes”,
Japanese J. Appl. Phys., vol. 19, no. 4, p. 693-701, April 1980.
[UNL 95] UNLÛ S., ONAT B.M., LEBLECI Y., “Transient simulation of heterojunction
photodiodes – Part II: analysis of resonant cavity enhanced photodetectors”, J.
Lightwave Technology, vol. 13, no. 3, p. 406-415, March 1995.
Chapter 3
Avalanche Photodiodes
3.1. Introduction
Photodetectors are one of two contrasting families of devices that
dominate the field of optoelectronics. Among these, avalanche
photodiodes (APDs) are very attractive devices because they benefit
from an internal gain due to the multiplication of the charge carriers
generated by absorption of incident light. This occurs when these
charge carriers cross a region of high electric field (>10
5
V/cm), thus
acquiring enough energy to ionize the atoms of the crystal lattice,
thereby creating new electron-hole pairs which are immediately
separated, and can themselves create other electron-hole pairs: this
leads step-by-step to an amplification of the primary photocurrent.
This mechanism, known by the term shock ionization or impact
ionization, is also at the root of the reversible breakdown of reverse-
biased p-n junctions.
In this chapter we will discuss the following subjects:
– theoretical consideration of the avalanche effect (ionization
coefficients, multiplication factor, etc.);
– technical conditions necessary for proper operation of APDs;
Chapter written by Gérard RIPOCHE and Joseph HARARI.
58 Optoelectronic Sensors
– APD structure;
– photoelectric properties of APDs;
– state of the art for APDs adapted to different ranges of
wavelength, based on Si, Ge, InGaAs, etc.;
– new low-noise APD structures based on superlattices or
heterogenous structures (Si/InGaAs).
3.2. History
The discovery of solid-state, gas and semiconductor lasers between
1958 and 1962, and the creation at the same time of the first
electroluminescent diodes, led to the birth of optoelectronics and to
the subsequent boom that it has seen, particularly in the field of
optical communications. however to take advantage of the
considerable potential of these new light sources, mostly
semiconductor lasers and light emitting diodes (with direct modulation
at high modulation frequencies, and short response times), there is a
need for fast and sensitive detectors operating at the emitted
wavelengths. Among the possible candidates, reverse-biased p-n
junction photodiodes offer numerous advantages: low bias voltages,
small footprint, robustness, ease of use and reliability. These make
them very attractive to users compared to photomultipliers which,
although benefiting from a high sensitivity thanks to a noise-free
internal gain process, suffer from two serious drawbacks for many
applications: their size and high operating voltage.
However, the electrical signal generated by the absorption of
incident light by a photodiode is, in most cases, too weak to be
directly usable and must be amplified. In boosting the signal to a
sufficient level, the amplifier stage limits the performance of the
receiver in two ways: by reducing the bandwidth of the system and by
lowering the global sensitivity due to the additional noise that it
introduces.
Avalanche Photodiodes 59
In order to eliminate these problems and to address the need for
amplification of the electric signal generated by the light incident on
the photodiode, in 1964 Johnson [JOH 64] highlighted the advantages
of primary photocurrent multiplication through the “avalanche effect”;
which was reported by McKay and McAfee [KAY 53] in 1953 during
their study of reversible breakdown in Si p-n junctions. This internal
gain mechanism gives enough improvement to the signal-to-noise
ratio to make the use of avalanche photodiodes (APDs) very
attractive for many applications, in particular for optical
telecommunications, despite the additional noise due to the random
nature of the photomultiplication mechanism.
Studies carried out in the 1960s focused on the development of
silicon avalanche photodiodes (Si APDs) for 1.06 μm telemetry and
for optical communications around 0.85 μm, and also the development
of germanium avalanche photodiodes (Ge APDs) whose photosensitivity
extends to 1.65 μm. The 1970s saw the optimization of Si APDs for
fiber-optic telecommunications at 0.85 μm through the use of “low-
noise” designs, as well as the exploratory study of gallium arsenide
(GaAs) APDs and those based on GaAs compounds. Since the future
of fiber-optic telecommunications is very much based in the near-
infrared due to the weak attenuation of optical fibers around 1.3 μm
(~2 dB/km) and at 1.55 μm (~0.2 dB/km), several laboratories worked
on improving Ge APDs. As the performance of these was limited,
even at 1.3 μm, the first investigations were carried out to evaluate the
possibilities offered by detectors based on III-V compounds, starting
with indium phosphate (InP). The development of working
InGaAs/InP APDs required important progress throughout the 1980s,
both in the epitaxial growth of heterostructures and in their design and
technology, which lead to the creation of devices tailored to the needs
of fiber-optic submarine links and in particular to those links that can
operate at bitrates of over 10 Gbit/s. From then on, investigations
focused on low-noise devices for high bitrates, such as superlattice
InGaAlAs/InAlAs APDs or heterogenous Si/InGaAs APDs.
60 Optoelectronic Sensors
3.3. The avalanche effect
The study of reversible p-n junction breakdown by McKay and
McAfee in silicon, and by Miller [MIL 55] in germanium, showed that
the growth of the dark current, observed when the bias voltage is
increased towards the breakdown voltage V
B
, which is a characteristic
of a given device, is caused by a mechanism commonly known as the
“avalanche effect”, similar to that observed in gas discharges.
The avalanche effect is shown schematically in Figure 3.1. A
primary charge carrier, electron or hole, which enters the space
charge region (SCR) where there is a fairly strong electric field E
(>10
5
V/cm), is strongly accelerated and can acquire a mean kinetic
energy greater than the characteristic ionization energy E
i
of the
semiconductor, before colliding with an atom of the crystal lattice,
creating an electron-hole pair through ionization and promoting an
electron from the valence band to the conduction band. The secondary
charge carriers (electron and hole) that are generated, as well as the
primary charge carrier, can then independently create new electron-
hole pairs, as long as the kinetic energy accumulated by each of the
charge carriers between successive collisions is greater than the
ionization energy E
i
. This process is cumulative and can under certain
conditions lead to the breakdown of the junction.
Figure 3.1. Diagram showing the principles of the shock ionization mechanism
Avalanche Photodiodes 61
3.3.1. Ionization coefficients
The ionization energy E
i
is, in the simplest case, that of a model
with two parabolic energy levels, greater than or equal to 3/2.E
g
,
where E
g
is the energy corresponding to the width of the bandgap. The
ionization energy E
i
depends on the type of charge carriers for a given
material. If it can be reduced by some means, this can lead to the
preferential ionization of one type of charge carrier, as we will see
later.
The ionization coefficients of electrons D
i
and of holes E
i
are
defined as the average number of ionizing collisions produced per unit
length (cm) by a charge carrier moving parallel to the constant electric
field E, and they quantify the ionizing power of the charge carriers.
Depending primarily on the electric field, they grow very rapidly with
the field. Their variations, determined experimentally, follow
empirical relationships of the form:
D
i
, E
i
= a
i
exp
÷b
i
E
(
(
·
·
·
·
)
)

[3.1]
where a
i
and b
i
are parameters that depend on the material and on the
type of charge carrier. Figures 3.2a and 3.2b show typical D
i
,
E
i
= f (1/E) curves, for silicon and germanium on the left and for InP
on the right, which are the base materials for APDs developed for
optical communications. Silicon is unusual in two ways: the electrons
are the most easily ionized carriers and the ratio of the ionization
coefficients k = E
i
/D
i
is very high for relatively weak electric fields
(E ~ 3 x 10
5
Vcm
–1
). The nature of the carriers triggering the
multiplication and the ratio k play an important role in the design of
devices, and directly influence the photoelectric performance of APDs.
It is worth noting that the variation seen in the experimental curves
published in the literature is a good illustration of the difficulties
inherent in determining the coefficients D
i
and E
i
from measurements
of photomultiplication or noise.
62 Optoelectronic Sensors
Figure 3.2. Ionization coefficients in silicon, germanium and
InP as a function of the inverse of the electric field 1/E
3.3.2. Multiplication factors
The impact ionization mechanism leads to a multiplication in the
number of charge carriers injected into the SCR, with this taking place
in the avalanche region where there is a high electric field E. For each
type of charge carrier, the multiplication factor M
n
or M
p
is defined as
the ratio of the number of charge carriers extracted to the number of
charge carriers injected. The following approximations are normally
used in the calculation:
– the ionization coefficients D
i
and E
i
only depend on the electric
field E, which itself is only a function of the x coordinate for a given
junction;
– recombination of carriers is negligible in the depleted region;
– mobile charge carriers do not disturb the electric field distribution;
Avalanche Photodiodes 63
– the levels of thermal and optical carrier generation in the depleted
region are not included.
Calculation of the multiplication factors M
n
and M
p
is carried out
by solving the continuity equations for the electron current J
n
and the
hole current J
p
under a strong electric field. The system of equations
describing steady-state behavior is as follows:
dJ
n
x
( )
dx

dJ
p
x
( )
dx
=D
i
x
( )
J
n
x
( )
÷E
i
x
( )
J
p
x
( )
[3.2]
J = J
n
x
( )
÷ J
p
x
( )
=C
ste
[3.3]
which can be rewritten as:
dJ
n
x
( )
dx
÷ D
i
x
( )
÷E
i
x
( )

l
l
l
J
n
x
( )
= E
i
x
( )
J [3.4]
Solving this differential equation for the case of multiplication
triggered by a pure injection of electrons from the P+ region, by
integrating J
n
(x) over the avalanche region (between x = 0 and x = w),
leads to the expression:

1
M
n
= D
i
0
w
¹
x
( )
exp ÷ D
i
x
( )
÷E
i
x
( ) ( )
0
x
¹
dx '

l
l
l
l
l
l
.dx [3.5]
with M
n
J
n0
, where J
n0
is the current of injected electrons at x = 0.
In the case of an injection of holes from the N
+
region, the
multiplication factor M
p
is given by a similar expression:

1
M
p
= E
i
0
w
¹
x
( )
exp ÷ D
i
x
( )
÷E
i
x
( ) ( )
x
w
¹
dx '

l
l
l
l
l
l
.dx [3.6]
64 Optoelectronic Sensors
with M
n
J
pw
, J
pw
being the current of injected holes at x = w.
In reality, carriers of both types are normally injected into the
avalanche region, leading to a multiplication factor somewhere
between the two which takes into account the “weighting” of each
contribution. It also follows that the experimental determination of the
ionization coefficients D
i
and E
i
from measurements of multiplication
factors is very sensitive, since in addition to a precise knowledge of
the electric field distribution E(x), it requires the separation of the
effects of both types of charge carrier. In the case where the ionization
coefficients are equal (D
i
= E
i
), the two expressions take the form:

1
M
= D
i
0
w
¹
x
( )
dx [3.7]
In the case of a PIN junction with a constant electric field in the
depleted region, for a given bias voltage V, the above equation
becomes:

1
M
=D
i
w [3.8]
where D
i
(E) is a function of the bias voltage V as w is constant, hence:
E =
V
w
[3.9]
3.3.3. Breakdown voltage
As the bias voltage increases, the electric field increases and
approaches the critical value for which the dark current grows
extremely fast: this is the volume breakdown of the junction, which
occurs for a bias voltage V
B
– the breakdown voltage. These
conditions occur when the multiplications factors M
n
, M
n
of which
occurs for:
Avalanche Photodiodes 65
1= D
i
x
( )
exp ÷ D
i
x
( )
÷E x
( ) ( )
dx '
0
x
¹

l
l
l
l
l
l
0
w
¹
dx [3.10]
and:
1= E
i
x
( )
exp D
i
x
( )
÷E x
( ) ( )
dx '
x
w
¹

l
l
l
l
l
l
0
w
¹
dx [3.11]
respectively.
In addition, Sze and Gibbons [SZE 66], studying the behavior of
P
+
N and N
+
P junctions in the avalanche regime, established a simple
semi-empirical relationship between the breakdown voltage V
B
and
the concentration N in the weakly-doped region of an abrupt junction,
or the concentration gradient a in a gradual junction. These relations
can be applied to different materials. The experimental results are
well-matched by the following expressions:
V
B
=60
E
G
1.1
(
(
·
·
·
·
)
)

3
2

N
10
16
(
(
·
·
·
·
)
)

3
4
[3.12]
for an abrupt junction and:
V
A
=60
E
G
1.1
(
(
·
·
·
·
)
)

6
5

a
3.10
20
(
(
·
·
·
·
)
)

2
5
[3.13]
for a gradual junction, where E
g
is the width of the bandgap of the
material in question, expressed in eV.
Finally, we point out the empirical relationship established by
Miller [MIL 55] between the multiplication factor M and the
breakdown voltage V
B
:

1
M
=
V
V
B
(
(
·
·
·
·
)
)

n
[3.14]
66 Optoelectronic Sensors
where n is an exponent which depends on the material, the type of
junction and the nature of the charge carriers triggering the
multiplication. This expression satisfactorily describes the variation of
the multiplication factor M with the bias voltage V, and it is often
applied to classical P
+
N and N
+
P junctions.
3.4. Properties of avalanche photodiodes
The behavior of avalanche photodiodes differs from that of
standard PIN photodiodes only in terms of the properties that are
involved in the avalanche multiplication mechanism: the gain, the
noise linked to the photocurrent, and the frequency response, these
being at operating voltages in the pre-avalanche regime of the junction,
in contrast to the low voltages at which PIN photodiodes are normally
used.
3.4.1. Current-voltage characteristics and photomultiplication
For a PIN junction, the photocurrent I
ph
generated by the light
incident on the active region barely varies with the operating voltage.
In contrast, for APDs, above a certain bias voltage the photocurrent I
ph
grows steadily, reaching its maximum value when the bias voltage is
equal to the breakdown voltage V
B
of the junction. The gain M is
given by the equation:
M =
I
ph
V
( )
I
ph
V
0
( )
=
I V
( )
÷I
d
V
( )
I V
0
( )
÷I
d
V
0
( )
[3.15]
where I
ph
, I and I
d
are respectively the photocurrent, the total current
flowing across the junction, and the dark current; V and V
0
are the
operating voltage and the reference voltage for which M= 1. In the
following, we will refer to I
ph0
and I
d0
as the primary photocurrent and
the primary dark current (i.e. the non-multiplied currents). These are
denoted as I
ph
(V
0
) and I
d
(V
0
) in equation [3.15]. This equation allows
us to plot the M(V) characteristics of an APD measured under well-
Avalanche Photodiodes 67
defined experimental conditions. These may have a decreasing or
increasingly strong influence on the measured values. In particular,
although an APD may behave fairly linearly for relatively low values
of gain (M< 10), for which M depends little on the incident light
power, this contrasts with significant saturation effects that are seen at
high gains. The maximum gain obtained (M
max
) is very sensitive to
the conditions of the measurement. A reduction in M
max
can be
partially explained by a reduction in the avalanche electric field
caused by an opposing electric field created by the mobile charge
carriers, which are present in significant numbers in the depleted
region. Also, close to breakdown, the applied bias voltage is reduced
because of the voltage drop across the equivalent resistance R of the
diode.
Taking these considerations into account, equation [3.14], which
applies for weak values of the current I, can for high currents be
written:

1
M
=
V ÷ I
d0
÷ I
ph0
( )
M R
V
B

l
l
l
l
l
l
n
[3.16]
for V= V
B
, M= M
max
and this can be simplified to:
M
max
~
V
B
n I
d0
÷ I
ph0
( )
R
[3.17]
This expression shows the value of devices with a weak dark
current (I
d0
<< I
ph0
) in order to promote photomultiplication. In these
conditions, the maximum gain is inversely proportional to the square
root of the primary photocurrent I
ph0
, and hence the power of the
incident light P
0
:
M
max
·
1
I
ph0
or M
max
·
1
P
0
[3.18]
68 Optoelectronic Sensors
Taking into account equation [3.12], which expresses the breakdown
voltage V
B
as a function of the concentration N, we find:
M
max
· N
3
8
[3.19]
Weak concentrations N in the depleted region, corresponding to
relatively high breakdown voltages V
B
, would be of benefit in
obtaining high gains M. However, for convenience of use, the diode
breakdown voltage is limited, where possible, to the range 50–150 V.
Finally, for devices with a strong dark current (I
d0
>> I
ph0
), the
limitation on the gain is imposed by the dark current and in these
conditions, which can occur with Ge APDs, we have:
M
max
~
V
B
n I
d0
R
[3.20]
3.4.2. Noise in avalanche photodiodes
Noise in detectors is discussed in Chapter 7 and we will limit
ourselves here to revisiting the main results for noise associated with
the operation of APDs in the avalanche regime. In the absence of
multiplication, the APD behaves like a traditional photodiode whose
noise is essentially shot noise caused by the dark current I
d0
and the
primary photocurrent I
ph0
across the junction, and whose spectral
density is:
i
2
=2 q I
d0
÷ I
ph0
( )
[3.21]
When the APD operates in the avalanche system, the currents
across the junction undergo a multiplication M, and hence the spectral
density of the noise is multiplied by M
2
. The random nature of the
impact ionization process results in a fluctuation in the gains
associated with the different carriers, contributing to the appearance of
Avalanche Photodiodes 69
extra noise linked to the multiplication. The spectral density of the
shot noise of the multiplied current can be written:
i
2
=2 q I
d0
÷ I
ph0
( )
M
2
F M
( )
[3.22]
where F(M), the excess noise factor, represents the contribution of
multiplication to the noise. The value of F(M) depends on the
characteristics of the p-n junction and on the nature of the primary
carriers. F(M) expressions for the two types of primary injection,
electrons and holes, have been derived by McIntyre [INT 66] under
the assumption of an avalanche region with a constant electric field.
These are expressed as a function of M and the ratio of the ionization
coefficients k = Ei/Di, which is considered constant. Instances of a
pure injection of electrons or of holes are of particular interest. The
excess noise factor for a primary injection of electrons can be
expressed by the following:
F
n
M
( )
= M
n
1÷ 1÷k
( )
M
n
÷1
M
n
(
(
·
·
·
·
)
)

2

l
l
l
l
l
l
[3.23]
and for a primary injection of holes:
F
p
M
( )
= M
p
1÷ 1÷
1
k
(
(
·
·
·
·
)
)

M
p
÷1
M
p
(
(
·
·
·
·
·
)
)

2

l
l
l
l
l
l
[3.24]
Where M
n,p
>> 1, these equations can be expressed in the
following simplified forms:
F
n
M
( )
= kM
n
÷2 1÷k
( )
[3.25]
F
p
M
( )
=
1
k
M
p
÷2 1÷
1
k
(
(
·
·
·
·
)
)

[3.26]
and for very high values of multiplication:
70 Optoelectronic Sensors
F
n
M
( )
÷kM
n
and F
p
M
( )
÷
1
k
M
p
[3.27]
In the specific case of ionization by only one type of carrier,
electrons or holes, the excess noise factor takes a minimum value of
F(M)=2, independent of M. Attempts can be made to approach this
value through modification of the ionization energy or by careful
design of the device. The typical variation of the noise factor as a
function of M for different values of k is shown in Figure 3.3.
Figure 3.3. Excess noise factor as a function of
different values of k for a pure injection of holes
In order to obtain low excess noise values, two conditions must be
met: the multiplication must be triggered by the more ionizing carrier,
and a material must be used whose ionization coefficients are very
different. Among the normal semiconductors (Si, Ge, GaAs, InP, etc.),
only silicon satisfies these two requirements. Furthermore,
consideration of the excess noise factor allows us to approach high
values of the ratio k = E
i
D
i
, with this parameter playing an important
role in the optimization of APDs. Finally, in the literature, the excess
noise factor F(M) is often written in the form:
F M
( )
= M
x
[3.28]
Avalanche Photodiodes 71
which comes directly from the spectral density of the noise, expressed
as:
i
2
=2 q I
ph0
M
2÷x
[3.29]
The exponent x, which depends on the value of the k ratio, allows
us to compare the different devices. This can often be found in device
specifications alongside the value of F(M) for a given M. For k = 1, the
excess noise factor is F(M) =M for the two types of injection, and the
exponent x therefore takes the value x = 1. For low values of k, the
multiplication is triggered by electrons, whereas for high values of k,
the injection of holes is favored.
3.4.3. Signal-to-noise ratio in avalanche photodiodes
In wideband applications, the limiting noise of traditional
photodiodes comes from the thermal noise associated with the load
resistance R
L
or, more commonly, from the noise of the first stage of
the amplification circuit required to make use of the electrical signal
produced by the photodiode. For some application types, the internal
gain of APDs allows us to improve the signal-to-noise ratio (SNR) of
the detector. This detector, consisting of the APD and amplifier,
includes two sources of noise outside the diode, independent of the
multiplication (the load resistance and the amplifying stage), given by
the equation:
i
ext
2
=
4k
B
T
e
R
L
'f [3.30]
where T
e
is the equivalent temperature of the noise, and also includes
sources of shot noise connected with the multiplication process:
i
M
2
= 2q I
d0
÷ I
ph0
( )
M
2
F M
( )
'f [3.31]
where 'f is the bandwidth of the system.
The mean squared value of the photocurrent is given by:
72 Optoelectronic Sensors
i
ph
2
=
1
2
m
2
I
ph0
2
M
2
[3.32]
where m is the modulation index of the incident light signal. The SNR
for a system with bandwidth 'f therefore follows the relationship:
S
B
=
1
2
m
2
I
ph0
2
M
2
( )
2q I
d0
÷ I
ph0
( )
M
2
F M
( )
÷
4k
B
T
e
M
2
R
L

l
l
l
l
l
'f
[3.33]
which can be rearranged into the following form:
S
B
=
1
2
m
2
I
ph0
2
( )
2q I
d0
÷ I
ph0
( )
F M
( )
÷
4k
B
T
e
M
2
R
L

l
l
l
l
l
'f
[3.34]
demonstrating the advantage of APDs over PIN photodiodes, since for
low values of F(M) the thermal noise contribution is divided by M
2
.
Figure 3.4. Variation of the SNR as a function of the gain M
for two semiconductors, silicon and germanium
As the gain increases, the SNR grows and reaches a maximum for
an optimal gain value M
opt
, which depends closely on the point at
Avalanche Photodiodes 73
which the multiplied shot noise and the thermal noise are equal, such
that, in the case where I
d
0 << I
ph
0:
M
opt
~
2k
B
T
e
qI
ph0
R
L
(
(
·
·
·
·
·
)
)

1
2÷x
[3.35]
For higher multiplication values, the SNR drops, with the
contribution of shot noise becoming the dominant factor. The
variation of the SNR with the gain is shown in Figure 3.4. The optimal
gain value M
opt
depends on the incident radiation power P
0
through
the equation:
I
ph0
=
qK O
h c
P
0
[3.36]
As a result, M
opt
will increase as P
0
decreases. Thus, APDs allow
us to improve the sensitivity of photodetectors as long as the shot
noise remains lower than the thermal noise entering the amplifier
stage.
3.4.4. Speed, response time and frequency response of avalanche
photodiodes
The ability of avalanche photodiodes to detect and amplify optical
impulses with short rise and fall times, or light signals modulated at
high frequency, is a topic of research for numerous applications. The
response time of an APD operating in the avalanche regime is limited
by:
– the transit time t
tr
of carriers in the SCR;
– the diffusion time t
Dif
, the time taken by the electron-hole pairs
created outside the depleted region to reach the p-n junction;
– the RC time constant of the APD/load system;
– the avalanche build-up time t
av
, the time needed to reach a stable
multiplication value for a given bias voltage.
74 Optoelectronic Sensors
The transit time t
tr
, the diffusion time t
Dif
and the RC time constant
all depend on the width w of the bandgap. The transit time t
tr
can be
expressed by the equation:
t
tr
=
w
Q
sat
[3.37]
where v
sat
is the saturation velocity for the slowest charge carriers.
In order to increase the speed of APDs for high-frequency
applications or broadband applications, we will seek to reduce the
thickness w. This reduction will in turn produce an increase in the
capacitance C
j
of the junction and a reduction in the response
coefficient, because the light will be absorbed in a thinner width of
material. This results in a compromise between the quantum
efficiency and the speed which must be treated on a case by case basis,
in a similar way to the tradeoff in PIN photodiodes (see section 2.4).
The diffusion time t
Dif
of electron-hole pairs towards the junction
is dominated by the lifetime of the minority charge carriers in the
lightly-doped region of the junction. Its value (between a few μs and a
few hundred μs depending on the material) makes some devices un-
usable for many applications. This limitation appears for certain Si
and Ge APDs. In practice, as far as possible, the thickness w of the
depleted region is adjusted so that all the light is absorbed there. This
result is approximately achieved for:
w=
2
D
O
[3.38]
where D
O
is the absorption coefficient for light of a given wavelength
O in a semiconductor.
The RC time constant is only significant at high frequencies. In this
case R~50 :, the value of the load resistance. The capacitance of the
junction Cj is related to the area A of the junction and the width w of
the depleted region by:
C
j
=
H A
w
[3.39]
Avalanche Photodiodes 75
where H is the permittivity of the semiconductor. As a result, we need
to minimize the area of the active region and that of the guard ring,
and select a value of w compatible with an acceptable transit time t
tr
.
This optimization is equivalent to that for PIN photodiodes (see
section 2.4).
As the multiplication process is not instantaneous, the response
time of APDs can be greater than that of PIN photodiodes of
comparable structure, and is very sensitive to the ionization conditions.
Hence, in the case of ionization by only one type of carrier, which is
similar to that of silicon, the impulse response time is controlled by
the transit time across the depleted region of the non-ionizing
secondary carriers created by the primary carriers. Under these
conditions, the bandwidth of the APD, independent of the
multiplication M, is not very different from that of a PIN photodiode.
When the two types of carriers contribute to the multiplication, the
steady-state value of M will be reached after a time which will be
longer if M is higher and if the ionization coefficients of the carriers D
i
and E
i
are close to each other. This is because the high number of
carrier-lattice interactions means the carriers “bounce” back and forth,
prolonging the time that secondary carriers are present in the
avalanche region. The avalanche build-up time t
av
is therefore
expressed by the equation:
t
aQ
= MW [3.40]
where W is the effective transit time of a carrier in the avalanche region
between two ionizing collisions. This depends very strongly on the
value of k =E
i
/D
i
and can become very small if D
i
and E
i
are very
different. Emmons [EMM 67] analyzed the frequency response of PIN
photodiodes operating in the avalanche system and established the
following expression:
W = N
b
E
i
D
i
(
(
·
·
·
·
)
)

W
1
[3.41]
where N
b
is a number which varies slowly between 1/3 and 2 as the
ratio k =E
i
/D
i
varies from 10
–3
to 1, and W
1
is the actual transit time in
76 Optoelectronic Sensors
the avalanche region. A low value of k minimizes the effective transit
time W and is also favorable for the response time and for the excess
noise factor. Emmons [EMM 67] showed that, for M>D
i
/E
i
, the gain-
bandwidth product is constant. The gain M varies with frequency
according to the following law:
M Z
( )
=
M
0
1÷ M
0
2
Z
2
W
2

l
l
l
1
2
[3.42]
where M(Z) and M
0
are the values of the gain as a function of the
frequency f = Z/(2S) and for a constant signal. The gain-bandwidth
product is given by:
M Z
( )
'f = N
b
k W
1
( )
1
[3.43]
and depends on the material through k and the characteristics of the
avalanche region through W
1
= l
a
/v
sat
, where l
a
is the length of the
avalanche region. It follows that the product will be much larger if the
ratio k of the coefficients of ionization is weak and if the avalanche
region is narrow. A high value of the gain-bandwidth product is
desirable as it allows a higher value of gain for a given frequency.
3.5. Technological considerations
APDs work best when they are biased with a voltage close to the
volume breakdown voltage in the active region. As a result there is a
very intense electric field (~4-5 x 10
5
V/cm) present in the avalanche
region, which must be as homogenous as possible throughout the
active region. Any inhomogenity present in the SCR that can cause a
local increase in the electric field will lead to a premature breakdown
which will degrade the APD performance. These inhomogenities can
be caused by the structure itself, as in a traditional planar junction
obtained by localized diffusion in a circular geometry, or linked to
imperfections in the crystal structure. Regardless of their origin, they
always need to be eliminated.
Avalanche Photodiodes 77
In the first case, the use of junctions with a guard ring or the use of
a mesa structure allows the suppression of peripheral breakdown, as
long as a geometry is used which is free from angular parts liable to
cause point effects. The suppression of crystal defects depends closely
on the progress of technology, particularly in the fields of
crystallogenesis and techniques of epitaxial growth and doping.
Figure 3.5. Cross-section of a planar photodiode
3.5.1. Guard ring junctions
For a junction constructed using planar technology (Figure 3.5),
the electrical field inside the depleted region is higher in the curved
region around the edge of the diffused area than in the active region,
as can be seen by solving Poisson’s equation [HAR 91a]. Breakdown
due to edge effects therefore occurs for a voltage sufficiently below
the breakdown voltage of the active region so that light falling on the
central region is not amplified or barely amplified. On the other hand,
if the illuminated region extends beyond the diffused area, the
increase in the injected global photocurrent due to the high electric
field present in a very narrow annular region around the edge can give
the illusion of a photomultiplication effect. Regardless, such a
photodiode cannot operate as an APD. In order to suppress peripheral
breakdown, two kinds of structures are commonly used, which have
annular guard rings modifying the electric field distribution around the
outer perimeter.
In the first structure (see Figure 3.6a), the guard ring is formed by
an annular diffused junction which is relatively weakly doped, and
quite deep in order to increase the radius of curvature and thus reduce
the peripheral field; the active region is obtained by a shallower
78 Optoelectronic Sensors
diffusion at a higher concentration, partially overlapping the guard
ring. The breakdown voltage of the gradual junction of the guard ring
is much larger than that of the active region, because for a given bias
voltage, the extent of the SCR is larger in the former than in the latter.
Figure 3.6. Cross-section of (a) an APD with a traditional
guard ring and (b) a PʌPN+ APD
In the second structure (see Figure 3.6b), the guard ring and active
region junctions are created by diffusion or ion implantation within a
weakly-doped substrate S. The n
+
–S guard ring junction which
surrounds the n
+
p junction of the active region is much deeper than
the latter, but the weak concentration in the ʌ substrate precludes
breakdown. This is the PSPN
+
structure. When the n
+
–S junction is
not very deep, or even superficial, the addition of a guard ring of the
first type is necessary to avoid peripheral breakdown. This is referred
to as a P
+
SPSN
+
structure.
3.5.2. “Mesa” structures
In mesa structures, the footprint of the active region of the APD is
defined by chemical etching or by dry etching followed by surface
reconstruction using a suitable chemical bath. This structure, shown in
Figure 3.7, offers the sometimes deceptive advantage of simplicity. It
allows us to reduce the capacitance of the device, while eliminating
the guard ring junction. However, the extension of the avalanche
region at high electric fields outside the edges of the etched region can
lead to migration of susceptible ion species, causing a growth in the
dark current which is often accompanied by instability of the diode. In
order to avoid these drawbacks, different passivation techniques have
Avalanche Photodiodes 79
been used with moderate success, which makes such a structure
difficult to commercialize on account of its lack of reliability.
Figure 3.7. Cross-section of a mesa APD
3.5.3. Crystal defects and microplasmas
Crystal defects such as dislocation, stacking faults, metallic
precipitation, etc., present in a PN junction introduce localized or
microplasma breakdowns: isolated regions of a few μm
2
where high
current densities cross, and which appear at well-defined bias voltages
which may or may not be close to the volume breakdown voltage.
Unless these microplasmas appear at a voltage very close to the
breakdown voltage, their presence decreases APD performance by
limiting the photomultiplication factor to that of the breakdown
voltage of the first microplasma. At higher voltages, the noise linked
to electrical discharges triggered by the random injection of charge
carriers in the microplasma is so significant that it precludes any
interest in the use of such diodes. In the case of Si APDs, inspection
under a microscope allows the microplasmas to be visualized,
appearing in the form of bright red spots. In the absence of
microplasmas, a diffuse, pale light uniformly illuminates the whole
active surface when the breakdown voltage is reached. Observation of
the current induced in a scanning electron microscope also allows the
detection of inhomogenities in the electric field. In addition, the
presence of microplasmas leads to vertical jumps of several μA in the
inverse characteristics I
d
(V), interspersed with resistive sections in the
relationship, thus enabling the electrical sorting of APDs through
systematic measurement of the I
d
(V) characteristics via probes.
Microplasmas can be eliminated using substrates with a low level of
80 Optoelectronic Sensors
dislocations and non-aggressive fabrication techniques tailored to the
specific materials used. The construction of high-performance,
reliable devices with a high fabrication yield depends closely on the
quality of the base material and on the standard of technology used in
each stage of the fabrication.
3.6. Silicon avalanche photodiodes
Silicon, an indirect-bandgap semiconductor, with a bandgap energy
E
g
= 1.1 eV at ambient temperature, is sensitive to light in the
wavelength range of 0.4–1.1 μm. The absorption length, considerably
less than 1 μm at O = 0.4 μm, varies between 10 and 10 μm in the
wavelength range 0.8–0.9 μm, the first transmission window for
optical fibers (attenuation ~ 2 dB/km), and is more than 200 μm at
O = 1.06 μm, which is the emission wavelength of the Nd lasers that
are very popular in military applications. Also, the ionization
coefficients of electrons and holes are very different, with electrons
having the higher ionization energy (D
i
>> E
i
). These characteristics
play an important role in the design of devices targeted at different
applications, with the aim of achieving a weak excess noise factor and
an acceptable compromise between sensitivity and response time.
3.6.1. Si N
+
P APDs
Si APDs with an abrupt-junction N
+
P structure are fabricated on a
P substrate with a concentration around 10
16
cm
–3
. As the width of the
depleted region is between 2 and 3 μm, these APDs are well-suited to
the detection of light signals with wavelengths below 0.7 μm (ruby
lasers emit at 0.63 μm). Commercialized in 1966, they were initially
developed for military applications such as laser telemetry at 1.06 μm,
taking advantage of an atmospheric transmission window allowing
targeting at 20–25 km with high precision (of the order of a meter).
Avalanche Photodiodes 81
For the first APDs we will study, the guard ring and active region
junctions are made by a double diffusion of phosphor impurities,
masked using a layer of silica obtained by thermal oxidation; the
fabrication conditions for the different stages of the construction
procedure have been developed to ensure high performance in terms
of the dark current and photomultiplication, along with a high
fabrication yield. The homogenity and reproducibility of phosphor ion
implantation used as a predeposition technology have greatly helped
the achievement of microplasma-free APDs. Additionally, the choice
of substrates with fairly high dislocation levels (~3,000 cm
–2
), but
which are stable under various thermal treatments at high
temperatures (T § 1,000–1,200°C), has allowed the attainment of
fabrication yields considerably higher than 50%, whereas they were at
best around 20% with the “zero dislocation” substrates advocated in
the literature.
These APDs, with a breakdown voltage V
B
~ 60 – 80 V, offer M(V)
characteristics which vary extremely rapidly as breakdown is
approached, with the exponent n in the Miller equation taking a value
close to 4. At short wavelengths absorbed in the N
+
layer, the
multiplication is initiated by holes. For O > 0.7 μm, the multiplication
is mostly initiated by electrons and takes higher values. At the
operating point, M ~ 50 to 100, with the gradient dM/dV very large.
To avoid issues with variations in M, the ratio V/V
B
must be kept
constant. Since the breakdown voltage V
B
follows variations in
temperature, some system of regulation is required. The involvement
of a Zener diode with a breakdown voltage close to that of the APD
has allowed the multiplication to be regulated within the temperature
range from –30°C to +80°C. Other regulation systems make use of the
noise linked with multiplication.
The impulse response of Si N
+
P APDs (see Figure 3.8) is
dominated by the diffusion time t
Diff
of the charge carriers generated
outside the depleted region. It consists of a brief rise to the signal
(<0.5 ns) followed by a relatively long tail, with the duration of the
impulse generally being insufficient to reach the steady-state value.
We then see a similar behavior for the fall of the signal. The diffusion
82 Optoelectronic Sensors
time limits the repetition frequency, and this can be improved by
limiting the thickness of the absorbing region with the use of epitaxial
P/P
+
structures or by localized thinning of the substrate.
Figure 3.8. Typical impulse response of a Si N+P photodiode
3.6.2. Si N
+
PʌP
+
APDs
To compensate for defects in the N
+
P structure, in 1966 Ruegg
[RUE 66] showed the potential of an N
+
PSP
+
structure, which
combines the advantages of an N
+
P structure (internal gain) with those
of a PIN structure (high quantum efficiency and speed). The result of
this is that the depleted region consists of two adjacent regions: one
relatively narrow, with an intense electric field, where the
multiplication of injected charge carriers takes place, and another
relatively extensive region, with a weak electric field, where light is
absorbed and the photocarriers collected. This separation of the roles
of multiplication and absorption gives the structure a high gain-
bandwidth product. Absorption takes place in the very weakly-doped
S region (<10
14
cm
–3
), whose thickness is adjusted as a function of the
operating wavelength (0.85 μm or 1.06 μm) and the desired
performance. The doping and thickness of the P layer are absolutely
critical.
Avalanche Photodiodes 83
Figure 3.9. Operating principles of a Si P+ʌPN+ APD: a) doping profile, b) typical
electric field, c) typical M(V) characteristics. V
rt
is the reach-through voltage
The operation of N
+
PSP
+
APD structures is shown in Figure 3.9,
which displays the doping profile, the electric field distribution for
three significant bias voltages and a typical multiplication curve. At
weak bias voltages, the depleted region is restricted to the P layer. As
the voltage increases, the edge of the depleted region moves and
enters the S layer for a bias voltage V
rt
known as the reach-through
voltage, for which the multiplication is around 10 to 20, just before the
breakdown of the N
+
P junction. Then, with the depleted region
extending across the whole S layer, the increase in the bias voltage
will only cause a weak increase 'E in the electric field. The
multiplication will vary relatively slowly, especially if the S layer is
wide, and the M(V) characteristic involves a more or less pronounced
“plateau”. Under normal operating conditions, the field in the
absorption region is sufficient to ensure carrier transport at their
84 Optoelectronic Sensors
saturation velocity. However, in order to benefit from a high gain, the
operating point must be beyond the “plateau”.
Consistent achievement of the desired M(V) multiplication
characteristics assumes that we can accurately control (± 10%) the
concentration profiles of the N
+
and P layers. Specifically, if the P
layer is too wide, the breakdown of the N
+
P junction occurs before the
depleted region extends in to the S layer, but if it is too narrow, the
extension into the S layer occurs too soon and the multiplication will
be weak (Figure 3.9c).
For applications at 0.85 μm, the thickness of the S layer can be
adjusted between 15 and 50 μm, as a function of the chosen level of
tradeoff between speed and response. In order to avoid the critical
stage of localized thinning, these APDs are best fabricated on an
epitaxial S/P
+
plate, with illumination of the N
+
P junction only
slightly degrading the excess noise factor.
3.6.3. Si N
+
ʌPʌP
+
APDs
An N
+
SPSP+ structure with buried P layer is a refined version of
the N
+
PSP+ structure. Proposed and studied by Lecrosnier [LEC 75],
it offers a significant reduction in the excess noise factor F(M) thanks
to its narrow multiplication region with a constant electric field
limited by the buried P layer: the field is weaker when this is far
below the surface. The technique of implanting high-energy B
+
ions (1
to 2 MeV), chosen for constructing this P layer, is very well-suited
because by modifying the implantation energy and dose parameters it
is possible to change the breakdown voltage and excess noise behavior
of these APDs. For a given Si S/P
+
structure, the multiplication
electric field decreases as the implantation energy is increases, since
the P layer lies further from the surface, at distances of 2.75 and
3.1 μm respectively for implantation energies of 1.5 MeV and
1.8 MeV.
The typical behavior of N
+
PʌP
+
APDs is only reached over a small
range (± 15%) on either side of the optimal implantation dose. If the
Avalanche Photodiodes 85
dose is too large, the behavior is like that of a N
+
P diode, or of a PIN
diode if it is too weak. Figure 3.10 shows a schematic doping profile,
the associated electric field distribution, and the M(V) characteristics
for different doses at a given implantation energy.
The fabrication procedure for APDs involves two diffusion phases
in order to construct stopper ring and guard ring junctions, followed
by two phases of ionic implantation and a phase of electrical
activation annealing: firstly the low-energy implantation of phosphor
ions to build the active N
+
S junction close to the surface (x
J
~ 0.2 μm),
then the high energy implantation of Boron ions (1.5 or 1.8 MeV,
dosage 1.2 u 10
12
ions/cm
2
) for the buried P layer.
a)
b)
c)
Figure 3.10. Operation principles for a Si N+ʌPʌP+ APD: a) doping profile,
b) typical electric field, c) typical M(V) characteristics for
different implantation doses of Boron (in cm
–2
)
86 Optoelectronic Sensors
Control of the conditions for the two implantations ensures good
reproducibility of the procedure, which is particularly easy to do for
epitaxial S/P
+
wafers, and also a good fabrication yield. A model of
this structure, based on complete photoelectric characterization (I(V),
C(V), M(V), noise) for various APDs was developed by Maille [MAI
80].
APDs obtained in this way offer excellent photoelectric
performance, normally with 250 μm diameter diodes:
– weak dark current (I
0
< 1 nA) close to breakdown;
– useful gain M
u
> 150, resulting in a response coefficient greater
than 75 A/W at S = 0.85 μm and a rise time W
r
~ 0.5 ns for a S layer of
thickness 30 μm;
– an excess noise factor F(M) = M
x
typically with F(M = 100) =
3 – 3.5, corresponding to x = 0.25. This value, close to the lower limit
of x = 0.2 is a result of the low ratio k ~ 0.01 of the ionization
coefficients E
i
and D
i
, due to the fact that the multiplication electric
field is such that E
m
< 3 x 10
5
Vcm
–1
;
– a uniformity in the photoresponse of ± 5%.
Diodes with an active zone diameter of 800 μm and even 1,500 μm
have been achieved thanks to the excellent spatial homogenity offered
by the implantation technique. To compensate for the lack of suitable
high-energy implantation equipment, a number of solutions have been
researched. One of them consists of implanting doubly-ionized B
++
Boron ions at relatively high energies (~300 keV), or alternatively
implanting Boron ions at 800 keV in a preferred direction in a S/P+
epitaxial plate with <110> orientation, to take advantage of a
“channeling” effect, resulting in the P layer lying 2.2 μm below the
surface, instead of 1 μm as would be the case under normal conditions.
The excess noise factors at M = 100 are F = 4–5 with “channeling”,
instead of F = 6–7, with x = 0.30–0.35 and x = 0.40 respectively
[KAN 78]. However, although “channeling” implantation gives
promising results, these are not easily reproducible.
Avalanche Photodiodes 87
Another process exists where the buried P layer is fabricated in two
stages: first localized low-energy implantation of Boron ions on the
S/P
+
plate, followed by an epitaxial stage to obtain the upper S layer
for multiplication. The growth of this layer is delicate. APDs made in
this way have an excess noise factor F = 5 at M = 100, with x = 0.35
[YAM 76], which could be improved upon by reducing the doping
and increasing the thickness of the upper S layer. These results
confirm that the excess noise factor F(M) is weaker when the P layer
is buried more deeply, and hence the electric field is only marginally
increased.
3.6.4. SiPt-Si N Schottky APDs
SiPt-Si N Schottky APDs, obtained through the formation of a very
thin, transparent and non-absorbing layer (20 to 40 nm) of platinum
silicate on top of N-type silicon, with resistivity U ~ 0.5–1 :cm, are
well-suited to the detection of ultraviolet (UV) radiation around
O = 0.5 μm, for which the absorption occurs very close to the surface
(absorption length of the order of 0.1 μm). Their performance depends
closely on the quality of the SiPt-Si N interface, and hence on the
standard of the fabrication technology.
After diffusion of the guard ring, the silica protecting the active
central region is chemically removed. Then the residual native oxides
and the first few atomic layers of silicon (~10 nm) are removed by ion
bombardment during the chemical deposition phase. Then, a 10 to 20
nm platinum film is deposited by radio-frequency sputtering. The
platinum silicate is then formed by in situ annealing at 600°C, with the
thickness of the silicate obtained being twice that of the platinum. The
metal present in the passivation layer is eliminated using aqua regia.
The upper contact is formed with a ring of aluminum deposited by
evaporation. The presence of an anti-reflection coating tuned to O =
0.5 μm is needed to obtain a good response coefficient, as the
reflection coefficient of silicon is around 45% in the near ultraviolet.
Under these conditions, a response coefficient above 0.25 A/W at M= 1
can be obtained for O = 0.4 μm.
88 Optoelectronic Sensors
APDs fabricated in this way possess excellent photoelectric
characteristics [ALA 74]. A barrier height I
BN
= 0.85 V and a good
ideality factor n = 1.02 measured from the I-V characteristics confirms
the good quality of the SiPt-Si N interface, despite a dark current
greater than that of traditional N
+
P or P
+
N junctions in their pre-
breakdown regime. At short wavelengths, multiplication is triggered by
electrons, and values of multiplication M ~ 300 have been measured at
O = 0.435 μm with an excess noise factor M
x
such that x ~ 0.7. The
limit on the bandwidth is imposed by the time constant RC and is
several GHz. In contrast, at O = 0.8 μm, the light is mostly absorbed
outside the depleted region and multiplication is thus triggered by
holes leading to performance degradation: a weaker M, higher excess
noise and a reduced bandwidth.
3.7. Avalanche photodiodes based on gallium arsenide
Si APDs are almost perfect photodetectors for the 0.80–0.85 μm
transmission window in fiber-optics, simultaneously possessing a
weak dark current, a high gain and low noise, all thanks to the
maturity of the technology. Its only drawback stems from the response
coefficient/speed tradeoff imposed by the weak absorption coefficient,
which is in fact not a great problem because the applications are at
low or moderate bitrates (<500 Mbit/s).
Nevertheless, investigations have been carried out into gallium
arsenide (GaAs), a direct-bandgap semiconductor with a bandgap
E
G
= 1.40 eV and photosensitive up to 0.88 μm, with respect to
possible use in integrated circuits and on ternary structures based on
GaAs, materials with a high absorption coefficient. GaAs APDs,
either of the Schottky type or diffused with an implanted guard ring
(H
+
protons or Mg
+
ions) offer performances below that of Si APDs
within the range 0.80–0.85 μm, except in terms of speed. On the other
hand, Stilman et al. [STI 74] showed the presence of a very
pronounced Franz-Keldysh effect for APDs made of lightly-doped
material (n ~ 5 x 10
14
cm
–3
), when biased in the multiplication regime.
Avalanche Photodiodes 89
a)
b)
Figure 3.11. Spectral response of electroabsorption GaAs Schottky APDs:
a) weak bias voltages, b) strong bias voltages
More recently, GaAs/Al
x
Ga
1-x
As APDs with low noise, using a
new concept (see section 3.10), have opened a new research avenue.
They offer a multi-quantum well (MQW) or “staircase” multiplication
90 Optoelectronic Sensors
region with a bandgap width varying continuously between GaAs and
Al
x
Ga
1-x
As. The need to cross discontinuities favors the ionization of
one type of charge carrier. Thus, for the x = 0.45 composition of Al,
the discontinuities are respectively 'E
C
= 0.48 eV and 'E
V
= 0.08 eV,
giving a value of k =E
I
/D
I
~ 0.15 and an excess noise factor F = 3 at
M = 10 [KAG 89]. The presence of an adjacent 1.5 μm absorbing
layer of GaAs enables high quantum efficiency devices that are fast
and low-noise.
3.8. Germanium avalanche photodiodes
Germanium, an indirect-bandgap semiconductor with a small
bandgap E
g
= 0.67 eV at ambient temperature and photosensitive in
the spectral range 0.4 μm to 1.65 μm, became established in the
middle of the 1970s as the photodetector for the first generation of
receivers for fiber-optic links at 1.3 and 1.5 μm, a region where silicon
does not work.
Starting in 1966, Melchior and Lynch [MEL 66] showed that Ge
APDs could be used as fast and sensitive detectors, and various
devices were brought to market. However, competition with Si APDs
limited their use to very specific areas of application. The problem
was that Ge APDs suffered a number of drawbacks linked to the
intrinsic properties of germanium, including:
– a high dark current, dominated by the charge carrier diffusion
current, which is proportional to n
i
2
, n
i
being the intrinsic concentration
for germanium;
– an insufficiently high ratio k of the ionization coefficients of
electrons and holes: k § 2, with the holes being the more easily
ionized, in contrast with silicon;
– the absence of stable native oxides, which makes the process of
surface passivation very difficult;
– a difficulty growing high quality epitaxial structures.
Avalanche Photodiodes 91
On the other hand, due to the high absorption coefficient D
O
up to
1.50 μm, depleted region widths w between 2 and 3 μm are sufficient
to absorb the light signal, thus allowing a good response coefficient
and a high speed. In addition, the dark current does not include a
tunneling current contribution, despite the weak value of E
g
.
3.8.1. Ge APDs with N
+
P, N
+
NP and P
+
N structures for 1.3 μm
communication
The three structures N
+
P, N
+
NP and P
+
N were developed in
succession for use in optical links at 1.3 μm. They have a planar
construction with a guard ring to ensure good reliability. Figure 3.12
shows a cross-section view. The presence of a stopper ring decreases
the surface current.
N
+
P structures were the first to be used, since the diffusion of n-
type dopants (As, Sb) is easier to achieve. The conditions for the
surface preparation and the deposition of dielectric layers (SiO
2
,
Si
3
N
4
), used to limit the diffusion, for passivation, and also for the
antireflection coating, are very critical. For an active region of
diameter ‡ = 100 μm, the total dark current measured at 0.9 V
B
is
around 150 nA and the dark current I
d0
around 50 nA at M = 1. The
excess noise factor for M = 10 is such that F(10) § 10 – 11,
corresponding to an effective ratio between the ionization coefficients
of k
eff
§ 1.3 μm. This high value of F(M), due to the injection of
electrons instead of holes as primary charge carriers, increases still
further with the wavelength.
92 Optoelectronic Sensors
Figure 3.12. Cross-section of Ge APDs: N
+
P, N
+
NP and P
+
N structures
The N
+
NP structure was proposed to improve the noise properties
of Ge APDs, with multiplication taking place close to the NP junction
lying about 2.5 μm from the surface. The N layer is obtained by
implantation of arsenic ions followed by annealing, leading to a
gradual junction. Since the light is mostly absorbed in the N layer, the
holes play the role of primary charge carriers triggering multiplication.
In addition, the electric field E at the junction is weaker (E § 2 x 10
5
V/cm). This results in a sensitive increase in the excess noise factors:
F(10) = 7 instead of 10 as previously. On the other hand, the total dark
current at 0.9V
B
is around 1 μm and the response time is limited by
the diffusion time in the N layer. The main drawback of the N
+
NP
layer is the lack of reproducibility caused by the fabrication conditions
for the N layer.
Avalanche Photodiodes 93
The P
+
N structure represents the final success of Ga APDs for 1.3
μm applications. It possesses the respective advantages of the N
+
P and
N
+
NP structures in terms of speed and excess noise factor, with the
light being absorbed in the N-type depleted region and the
multiplication being triggered by holes. The active region and guard
ring junctions, as well as the stopper ring, are made by ionic
implantation followed by annealing at relatively low temperature. This
“fully implanted” technique offers good reproducibility of the
performance. For diodes with an active region diameter of 100 μm
and 30 μm, the excess noise factor F(10) is 8 and 7 respectively, and
the cutoff frequency is at least 1 GHz at M= 10.
3.8.2. Ge APDs with P
+
NN
-
structures for 1.55 μm communication
Above L=1.52 μm the N+P, N+NP and P+N photodiode structures
become “blind” if the temperature drops below -20°C, because the
absorption coefficient of germanium falls dramatically. The
absorption length at ambient temperatures is around 25 μm at O = 1.55
μm, a window of weak absorption for optical fibers (0.2 dB/km).
Structures developed for 1.3 μm communication are not suitable
because of their low depleted region width w § 2 to 3 μm. Only a
structure analogous to N
+
PSP
+
Si APDs, where absorption and
multiplication take place in separate regions, could attain the required
performance of the response coefficient and speed.
Figure 3.13. Cross-section of a P
+
NN

structure Ge APD
for 1.55 μm communication
94 Optoelectronic Sensors
P
+
NN
-
structured Ge APDs on a weak-concentration N substrate
(§ 5 u 10
14
cm
–3
) have been built to meet these requirements. The
active region of these planar-structure devices with guard rings
(Figure 3.13), of diameter Ø = 80 μm or Ø = 30 μm for coupling to
multimode or single-mode fibers, is best obtained by a double
implantation of boron or indium for the surface P
+
junction and of
arsenic for the N-type multiplication region, followed by activation
and diffusion annealing. The guard ring junction is made either by a
long period of zinc diffusion or by implantation of beryllium ions
followed by annealing. The best performances obtained for 80 μm
diameter APDs at O = 1.55 μm are as follows:
– total dark current at 0.9 V
B
: 1.3 μm;
– dark current I
d0
at M = 1: 70 nA;
– quantum efficiency K = 80% at M = 1;
– excess noise factor F = 6.1 at M= 10;
– cutoff frequency above 500 MHz at M = 10;
– breakdown voltage V
B
§ 70–85 V.
The limitation on the response time is imposed by the transit time
in the depleted region. The impulse response of Ge P
+
NN
-
APDs at
1.55 μm does not contain a “lag” due to the collection of charge
carriers created outside the depleted region. This is also the case for
Ge P
+
N photodiodes, whose cutoff frequency is only 10 MHz.
Additionally, Ge APDs often operate with an optimal gain between 5
and 10, with the response time depending little on the multiplication
factor.
Nevertheless, with the available performance gains for Ge APDs
limited by the fundamental properties of germanium, investigations
carried out by various laboratories have yielded APDs based on III-V
compounds with very promising performances as a second generation
of structures for optical communications at 1.3 and 1.55 μm.
Avalanche Photodiodes 95
Figure 3.14. Cross-section of an InGaAs/InP planar APD
with front illumination, using a SAGM structure
3.9. Avalanche photodiodes based on indium phosphate (InP)
In
0.53
Ga
0.47
As/InP heterojunction APDs rapidly came to the
forefront in the 1980s as second generation detectors for optical
communications at 1.3 and 1.55 μm. This was because the ternary
compound In
0.53
Ga
0.47
As, with a direct bandgap of E
G
= 0.75 eV,
offers strong absorption up to 1.65 μm. Formed into a lattice using InP,
it is well-suited to epitaxial growth of high crystal quality layers.
However, the appearance of a high tunneling current for an electric
field E > 1.5 x 10
5
V/cm precludes all avalanche multiplication,
despite a more favorable ratio of ionization coefficients than that of
germanium (D
I
~ 2.E
I
).
On the other hand, P
+
N APDs based on InP, a material with a large
bandgap (E
G
= 1.35 eV) and transparent above 0.95 m, offer a low
dark current along with an ionization coefficient for holes that is
higher than that of electrons, such that E
I
~ 2D
I
.
96 Optoelectronic Sensors
Figure 3.15. Operation principles of an InGaAs/InP APD with a SAGM structure.
The electric field is given: 1) for a weak bias voltage, 2) for the punch-through
voltage, and 3) in the case of multiplication
The use of a structure analogous to the P
+
SPN
+
structure developed
for silicon, consisting of a separate absorption region (In
0.53
Ga
0.47
As)
and multiplication region (InP), led to the SAGM (Separated
Absorption Graded Multiplication) structure. Figure 3.14 shows a
cross-section through a planar InGaAs/InP APD with front-face
illumination which has a SAGM structure. The “graduated”
quaternary InGaAsP layers, with bandgap widths intermediate
between those of InGaAs and InP, fulfill a double function: to ensure
a “gentle” transition between InGaAs and InP, to facilitate the
transport of holes by reducing the valence band discontinuity at the
InGaAs/InP interface, and to avoid the decomposition of the ternary
layer during growth of the InP layers above it. Their presence is vital
for the growth of a vertical heterostructure by liquid phase epitaxy.
This structure, suited for illumination through front or back faces,
offers the advantage of a pure injection of holes, which are the most
Avalanche Photodiodes 97
ionizing carriers in InP, thus allowing a reduction in the excess noise
factor. It is suited to illumination through the front or rear faces.
Proper operation of these APDs, whose principle is summarized in
Figure 3.15, requires strict conditions to be met in terms of doping and
thickness of the epitaxial layers, especially of the multiplication layer,
since during avalanche the electric field E
QT
at the quaternary/ternary
absorption interface must be below 1.5 u 10
5
V/cm in order to retain a
low dark current.
Construction of the guard ring, required to reach the necessary
reliability levels for detectors targeted at submarine communications,
is even more delicate than for Si and Ge APDs. Specifically, the guard
ring junction, which is slightly deeper than the active region, must
remain within the InP, and the depleted region under the guard ring
must reach the ternary layer before that of the active region’s junction.
This requires that the guard ring junction be gradual with a weak level
of doping. For a long time, the implantation of beryllium Be
+
ions,
followed by redistribution annealing [SHI 83], was the only method of
fabricating effective guard rings. More recently, better understanding
of epitaxial growth techniques has allowed good results to be obtained
by combining localized implantation and epitaxial regrowth [WEB 88].
A realistic predictive model for the behavior of planar InGaAs/InP
APDs was put forward by Harari [HAR 91b] after a detailed analysis
of their operating conditions.
3.9.1. InGaAs/InP APDs for optical communications at 2.5 Gbit/s
An APD heterostructure based on InGaAs/InP on a <100>
oriented InP-N
+
substrate consists of a stack of 6 layers: an InP-N
+
buffer layer (n = 5 u 10
17
cm
–3
, 0.5 μm), an InGaAs absorption layer
(n = 5 u 10
15
cm
–3
, 2.5 μm), 2 InGaAsP layers (not intentionally
doped, 0.2 μm), the InP-N multiplication layer (n = 1.5 u 10
16
cm
–3
),
and the InP-N

covering layer (n = 8 u 10
15
cm
–3
, 2 μm).
The guard ring junction is obtained by implantation of beryllium
ions (Be
+
, 150 keV, 5 u 10
13
cm
-2
) followed by annealing at 700°C in
a sealed ampoule. The active junction is formed by diffusion of
98 Optoelectronic Sensors
cadmium at 620°C in a sealed ampoule, for a duration which depends
on the required characteristics of the multiplication layer. A nitride
layer of Si
3
N
4
ensures surface passivation, and another acts as an
antireflection coating.
Typical performances of such 70 μm diameter APDs are:
– a response coefficient (at M = 1) of 0.85 A/W and 1 A/W at 1.3
μm and 1.55 μm respectively;
– a dark current at 0.9 V
B
< 30 nA;
– a maximum gain M
max
= 30;
– an excess noise factor F = 5 at M= 10;
– a bandwidth of 2.5 GHz.
Figure 3.16. M(V) and I
d
(V) characteristics of a fast InGaAs/InP APD
Experiments on accelerated aging under extreme conditions
(175°C, 100 μA) have demonstrated the robustness of these APDs,
whose median lifetime under these conditions is greater than 4,000
hours. In addition, thanks to their high degree of reliability, they are
one of the key devices for fiber-optic communications at 2.5Gbit/s. To
Avalanche Photodiodes 99
begin with, the variation in concentration and thickness of the layers
deposited through liquid phase epitaxy had a beneficial effect, with
the devices present on a treated wafer offering a wide range of
behaviors. This helped develop and understanding of their operating
conditions and to acquire a good understanding of the fabrication
requirements for such APDs. In this way, fabrication yields greater
than 50% were obtained.
More sophisticated growth techniques, such as molecular jet
epitaxy or vapor phase epitaxy, allowing the deposition of layers that
are very uniform in thickness and concentration have led to even
higher fabrication yields (>80%), while tightening the positioning
requirements for the junction layout with respect to the InP/InGaAs
interface.
3.9.2. Fast InGaAs/InP APDs
In order to meet the need for fast detectors for high-bitrate
communications at 5 or 10 Gbit/s, the normal bandwidth of APDs
(typically 2.5 GHz) must be increased. In order to do this, we must
reduce the “accumulation” of holes due to the large valence band
discontinuity at the InP/InGaAs interface, which is the main limitation
of the bandwidth of these APDs. It is also necessary to reduce the
transit time across the InGaAs absorption layer while keeping its
thickness between 1.0 and 1.5 μm in order not to degrade the response
coefficient. Furthermore, the increase in doping of the InP
multiplication layer, along with a reduction in its thickness,
contributes to a reduction in the avalanche trigger time. Finally, an
active region of 30 μm diameter, comparable with single-mode fiber
illumination, minimizes the capacitance of the junction. APDs with
one, two or three thin quaternary layers, or with a quaternary layer
whose composition varies continuously between InGaAs and InP,
have been fabricated with bandwidths between 7 and 9 GHz. With
only one thin 1.18 μm quaternary layer (the wavelength corresponding
to the bandgap energy of the GaInAsP compound) ensuring
equipartition of the discontinuity of the valence band (0.19 eV instead
of 0.38 eV), a theoretical study [RIP 83] shows performances
100 Optoelectronic Sensors
equivalent to two- or three-layer structures which are more
complicated to fabricate. After fabrication, the measured
performances are as follows:
– dark current < 60 nA at M = 10, with the best values being
around 10 nA;
– dark current I
d0
lying between 0.8 and 4 nA;
– maximum gain M
max
> 30 for the majority of diodes;
– response coefficient at M = 1 of 0.9 A/W at 1.3 μm and
0.95 A/W at 1.55 μm;
– bandwidth > 6 GHz at M = 8;
– excess noise factor F = 6 at M = 10, with a fabrication yield
greater than 75%.
However, these InGaAs/InP APDs targeted at 10 Gbit/s
communications are reaching the limits of performance. System
designers often prefer to use InGaAs/InP APDs at 2.5 Gbit/s in a 4 u
2.5 Gbit/s configuration, when an infrastructure for 2.5 Gbit/s optical
communications is already in place.
3.10. III-V low-noise avalanche photodiodes
The performance of APDs in terms of noise and bandwidth
depends strongly on the ratio of the ionization coefficients of the
charge carriers. For a given material, they will be higher when the
ratio between the ionization coefficients of the least and more easily
ionized charge carrier is weaker. This is the case for silicon, where at
weak electric fields (E < 3 u 10
5
Vcm
–1
), k ~ 0.01, which is close to
the behavior of ionization only by electrons. In contrast, in InP where
the holes are the more easily ionized, this ratio is around k ~ 0.5, a
value already corresponding to a significant noise factor.
In order to reduce the excess noise factor and increase the gain-
bandwidth product for III-V APDs based on InP or GaAs, it is
necessary to strongly increase the ionizability of one type of charge
carrier, using a modified structure or a material which has this
Avalanche Photodiodes 101
property under specific conditions. Chin et al. [CHI 80] suggested
reducing the ionization energy of holes or electrons by taking
advantage of the numerous discontinuities in the valence or
conduction band which can be found in multi-quantum-well (MQW)
structures. Alternatively, Hildebrandt et al. [HIL 81] showed that
ternary GaAlSb structures offer a significant increase in the
ionizability of holes close to a specific composition ratio. This is also
the case for the II-VI ternary compound HgCdTe.
3.10.1. III-V super-lattice or MQW APDs
The multiplication region of a MQW APD consists of a stack of
layers with alternately wide bandgaps (“barriers”) and narrow
bandgaps (“wells”) which produce discontinuities at the interface,
between conduction bands 'E
c
and between valence bands 'E
v
,
whose values depend on the nature of the materials used. The
“barrier/well” pairs favor the transport of electrons when 'E
c
> 'E
v
or
of holes when 'E
c
< 'E
v
. For example, in the case of the
InAlAs/InGaAs pair, an electron which undergoes a transition in a
barrier region crosses the interface and “falls” into the adjacent well,
gaining an energy 'E
c
, or alternatively benefits from an artificial
reduction in its ionization energy, thus favoring ionization in the wells.
Conversely, movement from a well to a barrier reduces the ionizing
power of the electron within the barrier. The crossing of numerous
barrier-well interfaces in succession translates into an increase in the
multiplication triggered by the electrons, equivalent to an increase in
D
I
. On the other hand, the ionizing power of the holes traveling in the
opposite direction is less favored, and overall this results in a
reduction of the effective ratio of ionization coefficients k =E
I
/D
I
,
leading to a reduction in the excess noise factor. When 'E
v
> 'E
c
,
such as in InP/InGaAs, it is the ionizing power of holes which is
favored, leading to a reduction in the ratio D
I
/E
I
for this structure.
Pairs of materials with very asymmetric bandgap discontinuities are
favored; thus, the combination of In
0.52
Al
0.48
As/In
0.53
Ga
0.47
As ('E
c
=
0.55 eV, 'E
v
= 0.2 eV) is more favorable than InP/In
0.53
Ga
0.47
As ('E
c
= 0.22 eV, 'E
v
= 0.38 eV). The number of barrier-well layers and the
102 Optoelectronic Sensors
thickness of those layers both have a significant influence on
performance.
Up to now, the best performances have been obtained [WAT 97]
for a planar APD with an implanted Ti guard ring with a MQW
multiplication layer of InAlGaAs/InAlAs (0.27 μm, 13 repeats) and an
absorption region of InGaAsP (1.3 μm, p = 8 u 10
15
cm
–3
) separated
by a thin doped layer of InP-P
+
(0.04 μm, p = 4 u 10
17
cm
–3
) in order
to reduce the electric field between the two regions. These APDs are
characterized by a weak dark current of 0.36 μA at M = 10, a
bandwidth of 15 GHz, a gain-bandwidth product of 110 GHz, a
response coefficient of 0.8 A/W at 1.55 μm, an excess noise factor
F(M) ~ 3 at M = 10 and a high reliability. They are well fitted to the
specifications of high-sensitivity 10 Gbit/s receivers.
3.10.2. Spin-orbit resonance APDs
3.10.2.1. GaAlSb/GaSb APDs
The strong increase in the ionizing power of holes which is seen in
Ga
1–x
Al
x
Sb/GaSb APDs with x ~ 0.05, sensitive at 1.3 and 1.55 μm, is
linked, according to the interpretation of Hildebrand et al. [HIL 81], to
the ionization mechanism for holes which causes the valence band to
be split in two via spin-orbit coupling. The result is that, around this
composition, the structure of the energy bands specific to this material
has the unusual property of having a bandgap energy E
g
close to the
difference in energy ' separating the maximum values for the valence
band and the additional split band.
If 'E
g
o1, through the resonance effect, the ratio of the ionization
coefficients E/Doˆ and the ionization energy of holes is equal to E
g
.
Subsequent construction of Ga
1–x
Al
x
Sb/GaSb APDs with x ~ 0.05 has
made it possible to obtain devices with a weak excess noise factor
F(M) = 3.8 at M = 10, corresponding to a value of k =E
I
/D
I
~ 5, but
much lower than the value of 20 to 30 that has been anticipated, and a
gain-bandwidth product of 90 GHz [KAT 90]. These performances are
higher for these parameters than those obtained with comparable
Avalanche Photodiodes 103
InGaAs/InP APDs; however, the inability to reduce and stabilize their
dark current or to fabricate a planar APD has led to these
investigations being abandoned.
3.10.2.2. HgCdTe avalanche photodiodes
Research into ternary, mercury-rich alloys Hg
1–x
Cd
x
Te, which had
already been explored with the aim of fabricating detectors for
military applications in the infrared windows at 3.5 μm and 8–14 μm,
was extended at the start of the 1980s for the development of
alternatives to Ge APD detectors for fiber-optic telecommunications
at 1.3 μm and 1.55 μm. The alloys with cadmium-rich compositions (x
~ 0.7 – 0.6) corresponding to these two transmission windows offer
the same electronic properties of spin-orbit coupling, which offers the
opportunity to develop APDs with performances similar to Ga
1–x
Al
x
Sb.
A method has been developed for growing Hg
1–x
Cd
x
Te wafers
with a given stochiometric composition, allowing the construction of
p-type substrates with doping levels around 2 u 10
16
cm
-3
, 40 mm in
diameter, with a high homogenity in terms of doping, and a low level
of dislocations. The devices are constructed in a planar structure using
traditional techniques of insulator deposition by cathode bombardment,
and photolithography. The n-type guard ring is obtained by mercury
diffusion. The active region junction is made by ionic implantation
(Al
++
, 65 keV, 10
14
ions cm
-2
) followed by redistribution annealing.
The result is a PIN junction. Compositions with x close to 0.6 are the
most suitable, because they cover both transmission windows and also
give the highest values of the ratio E
I
/D
I
[ORS 87], with the best being
the composition with x = 0.62, E
g
= 0.72 eV and E
I
/D
I
= 30.
Encouraging initial results obtained in terms of dark current, response
coefficient and excess noise factor have not been confirmed at the
predevelopment stage.
104 Optoelectronic Sensors
3.11. Prospects
A large proportion of the work in progress is aimed at the
development of devices for high-bitrate communications (20 to
40 Gbit/s) that are both fast and low-noise. Particularly significant is
the work on Si/InGaAs APDs, “waveguide” MQW APDs and low
noise APDs with a very thin multiplication region.
3.11.1. Si/InGaAs APDs
These APDs, combining the very low excess noise factor of Si
APDs with the strong absorption of InGaAs, should simultaneously
allow high quantum efficiency and a high bandwidth, a high gain-
bandwidth product and a low level of noise. Preliminary results
obtained for Si/InGaAs structures fabricated using the technique of
wafer fusion are very promising: bandwidths of 13 GHz for gains of
135, corresponding to a gain-bandwidth product of 315 GHz have
been measured. However, the dark current is raised very close to
breakdown and is unstable under a bias voltage [HAW 97].
Important work remains to be done to develop enough
understanding of wafer construction and the stages of APD fabrication
in order to overcome the limitations linked to the large discrepancy in
the lattice parameter between Si and InGaAs on the one hand and, on
the other hand, also linked to the difficulties of passivation.
3.11.2. “Waveguide” MQW APDs
In order to reduce the maximum transit time of charge carriers
without reducing the quantum efficiency, the light signal is injected
laterally, perpendicular to the electric field, in a vertical “waveguide”
structure whose InGaAs core is very thin (~0.5 μm).
Avalanche Photodiodes 105
Since the MQW InAlAs/InGaAlAs or InAlAs/InGaAs multiplication
region is also very thin, the transit time of charge carriers is very short
and bandwidths of 25 GHz have been obtained. The gain-bandwidth
product is more than 5. For a diode of length 20 μm, the response
coefficient at M = 1 is 0.9 A/W at 1.55 μm for a fiber-diode coupling
factor of 80% [COH 97].
3.11.3. Low-noise APDs with a very thin multiplication region
The study of APDs with a very thin multiplication region (0.2 to
0.5 μm) has demonstrated the possibility of making devices with a
low excess noise factor out of materials such as InP, GaAs, InAlAs
whose ionization coefficients are similar, even in the case of a primary
injection by the least ionizing charge carriers. Under such conditions,
McIntyre’s analysis is not relevant because the “dead space”, the
distance required for a charge carrier to acquire enough energy to
ionize, is no longer negligible compared to the thickness of the
multiplication region. Consequently, the ionization process is less
random, and as a result there is less noise.
Excess noise factors comparable to those measured for multi quantum
well APDs (F ~ 4 at M= 10) have been published for traditional [DRI 99]
and waveguide [KIN 00] In
0 53
Ga
0.47
As/In
0.52
Al
0.48
As APDs with a high
bandwidth (~20 to 30 GHz), whose multiplication region in the high-
electric-field monolithic InAlAs material is separated from the low-
field InGaAs absorption region by a very thin but heavily doped InP
layer. The primary injection of electrons makes the presence of
quaternary transition layers unnecessary. This makes such structures
very attractive for high-bitrate applications as they are simpler to
fabricate.
Additionally, a record gain-bandwidth value of 290 GHz was
obtained for InAlAs/InGaAs APDs with a vertical structure
incorporating a resonant cavity centered at O = 1.55 μm, a high
response coefficient (above 0.8 A/W) and a weak dark current for a
thickness of 0.2 μm for the InAlAs multiplication region [LEN 99].
However, these APDs require an excellent understanding of epitaxial
106 Optoelectronic Sensors
growth in order to achieve stacking of the “mirror” layers and the
active region, and are much more delicate to fabricate than the
traditional structure in [DRI 99].
3.12. Conclusion
Initially responding to the demands of military applications, the
development of avalanche photodiodes has occurred in parallel to that
of optical fiber telecommunications, with one not being possible
without the other. The result is that for each generation of fibers
corresponding to a spectral transmission window of silica, 0.85 μm,
1.3 μm and 1.55 μm respectively, there is a specific APD associated,
be it P
+
SPN
+
Si APDs, Ge APDs, and then SAGM InGaAs/InP APDs,
tailored to the relevant emitter. The perfection of high-performance
APDs has been one of the important factors in the achievement of
competitive optical communications. An increase in the sensitivity of
the photoreceiver, consisting of the APD combined with an associated
amplifier, allows either a reduction in the power of the signal emitted
in the first place or a reduction in the number of repeaters along the
link.
In addition, the high level of reliability of SAGM InGaAs/InP
APDs has contributed to the expansion of submarine optical fiber
links over the last decade, which has seen the installation of many
intercontinental links at higher and higher bitrates (5 to 10 Gbit/s)
with shorter and shorter latencies. Traditional SAGM InGaAs/InP
APDs have met, and continue to meet, the needs of designers up to
5 Gbit/s and even 10 Gbit/s. However, the advent of optical links at
bitrates of 20 or 40 Gbit/s and beyond leaves no question that there is
no current APD structure which is really applicable, and further
research is required.
Avalanche Photodiodes 107
3.13. Bibliography
[ALA 74] ALABEDRA R., MAILLE C., LECOY G., RIPOCHE G., “Photodiode Schottky
SiPt-Si N à avalanche”, ESSDERC, 1974.
[CHI 80] CHIN R., HOLONYAK N., STILLMAN G., TANG J., HESS K., “Impact ionisation
in multilayered heterojunction structures”, Electronics Letters, vol. 16, no. 12,
p. 467, 1980.
[COH 97] COHEN-JONATHAN C., GIRAUDET L., BONZO A., PRASEUTH J.P.,
“Waveguide AlInAs/GaAlInAs avalanche photodiode with a gain-bandwidth
product over 160 GHz”, Electronics Letters, vol. 33, no. 17, p. 1492, 1997.
[DRI 99] DRIES J., THOMSON K., FORREST S., “In
0,53
Ga
0,47
As/In
0,52
Al
0,48
As separate
absorption, charge, and multiplication layer long wavelength avalanche
photodiode”, Electronics Letters, vol. 35, no. 4, p. 334, 1999.
[EMM 67] EMMONS R., “Avalanche photodiode frequency response”, Journal of
Applied Physics, vol. 38, no. 9, p. 3705, 1967.
[HAR 91a] HARARI J., Etudes théoriques et expérimentales de photodiodes à
avalanche planaires GaInAs-InP, PhD Thesis, Lille Flandre Artois University of
Science and Technology, 1991.
[HAR 91b] HARARI J., DECOSTER D., VILCOT J-P., KRAMER B., OGUEY C., SALSAC P.,
RIPOCHE G., “Numerical simulation of avalanche photodiodes with guard ring”,
IEEE Proceedings, vol. 138, no. 3, p. 211, 1991.
[HAW 97] HAWKINS A., WU W., ABRAHAMP., STREUBEL K., BOWERS J., “High gain-
bandwidth product silicon heterointerface photodetector”, Applied Physics
Letters, vol. 70, no. 3, p. 303, 1997.
[HIL 81] HILDEBRAND O., KUEBART W., BENZ K., PILKUHN M., “Ga1-xAlxSb
avalanche photodiodes: resonant impact ionization with very high ratio of
ionization coefficients”, IEEE Journal of Quantum Electronics, vol. QE-17, no. 2,
p. 284, 1981.
[INT 66] MCINTYRE R., “Multiplication noise in uniform avalanche diodes”, IEEE
Trans. Electron, Devices, vol. ED-13, no. 1, p. 987, 1966.
[JOH 64] JOHNSON K., “High speed photodiode signal enhancement at avalanche
breakdown voltage”, Digest of Technical Papers, International Solid State
Circuits Conference, vol. 7, p. 64, 1964.
[KAG 89] KAGAWA T., IWAMURA H., MIKAMI O., “Dependence of the GaAs/AlGaAs
superlattice ionization rate of Al content”, Applied Physics Letters, vol. 54, no. 1,
p. 33, 1989.
108 Optoelectronic Sensors
[KAN 78] KANEDA T., KAGAWA S., YAMAOKA T., NISHI H., INADA T., “Low noise
avalanche photodiodes by channeling of 800 keV boron into <110> silicon”,
Journal of Applied Physics, vol. 49, no. 12, p. 6199, 1978.
[KAT 90] KATSUWAKA H., MIKAWA T., MIURA S., YASUOKA N., TANAHASHI T.,
WADA O., “An AlxGa1-xSb avalanche photodiode with a gain bandwidth product
of 90 GHz”, IEEE Photonics Technology Letters, vol. 2, no. 1, p. 54, 1990.
[KAY 53] MCKAY K., MCAFEE K., “Electron multiplication in silicon and
germanium”, Physical Review, vol. 91, no. 5, p. 1079, 1953.
[KIN 00] KINSEY G., HANSING C., HOLMES A., STREETMAN B., CAMPBELL J., DENTAI
A., “Waveguide In
0,53
Ga
0,47
As/In
0,48
Al
0,48
As avalanche photodiode”, IEEE
Photonics Technology Letters, vol. 12, no. 4, p. 416, 2000.
[LEC 75] LECROSNIER D., PELOUS G., AMOUROUX C., BRILMAN M., RIPOCHE G.,
“Optimization of avalanche silicon photodiodes: a new structure”, Technical Digest
of International Electron Devices Meeting, Washington D.C., p. 595, 1975.
[LEN 99] LENOX C., NIE H., YUAN P., KINSEY G., HOMLES A., STREETMAN B.,
CAMPBELL J., “Resonant cavity InGaAs- InAlAs avalanche photodiodes with
gain-bandwidth product of 290 GHz”, IEEE Photonics Technology Letters, vol.
11, no. 9, p. 1162, 1999.
[MAI 80] MAILLE C., Modélisation à l’obscurité et sous éclairement des
photodétecteurs à avalanche N
+
SPSP au silicium, PhD Thesis, Languedoc
University of Science and Technology, Montpellier, July 1980.
[MEL 66] MELCHIOR H., LYNCH W., “Signal and noise response of high speed
germanium avalanche photodiodes”, IEEE Transactions on Electron Devices, vol.
ED-13, no. 12, p. 829, 1966.
[MIL 55] MILLER S., “Avalanche breakdown in germanium”, Physical Review, vol.
99, no. 4, p. 1234, 1955.
[ORS 87] ORSAL B., ALABEDRA R., VALENZA M., LECOY G., MESLAGE J.,
BOISROBERT C., “Les photodiodes à avalanche Hg
0,4
Cd
0,6
Te à O = 1,55 Pm. Bruit
près de la résonance liée au couplage spin-orbite”, Revue de Physique Appliquée,
vol. 22, p. 227, 1987.
[RIP 83] RIPOCHE G., PEYRE J.-L., LAMBERT M., MOTTET S., “High speed (Ů 6 GHz)
InGaAs/InP avalanche photodiodes grown by gas source molecular beam epitaxy
with a thin quaternary grading layer for high bit rate (t 5 Gbit/s) systems”,
Journal de Physique, III, vol. 3, p. 1761, September 1983.
[RUE 66] RUEGG H., “A fast high gain silicon photodiode”, Digest of Technical
Papers, International Solid State Circuits Conference, vol. 9, p. 56, 1966.
[SHI 83] SHIRAI T., MIKAWA T., KANEDA T., MIYAUCHI A., “InGaAs avalanche
photodiodes for 1 μm wavelength region”, Electronics Letters, vol. 19, no. 14,
p. 534, 1983.
Avalanche Photodiodes 109
[STI 74] STILLMAN G.,WOLFE C., ROSSI D., DONNELLY J., “Electroabsorption
avalanche photodiode”, Applied Physics Letters, vol. 25, no. 11, p. 671, 1974.
[SZE 66] SZE S., GIBBONS G., “Effect of junction curvature on breakdown voltage in
semiconductors”, Solid State Electronics, vol. 9, p. 831, 1966.
[WAT 97] WATANABE I., NAKATA T., TSUJI M., MAKITA K., TAGUCHI K., “High
reliability and low dark current 10 Gb/s planar superlattice avalanche
photodiodes”, IEEE Photonics Technology Letters, vol. 9, no. 12, p. 1619, 1997.
[WEB 88] WEBB P., MCINTYRE R., SCHEIBLING J., HOLUNGA M., “Planar InGaAs/InP
avalanche photodiode fabrication using vapor phase epitaxy and silicon
implantation techniques”, Technical Digest of Optical Fiber Communications
Conference, OFC 88, 1988.
[YAM 76] YAMAOKA T., MATSUMOTO H., KANEDA T., “Silicon avalanche photodiodes”,
Fujitsu Scientific & Technical Journal, vol. 12, p. 87, September 1976.
This page intentionally left blank
Chapter 4
Phototransistors
4.1. Introduction
The effects of light on transistors have been studied since
transistors were first created, with the main motivation for this
research being to produce a combination of photodetection and signal
amplification in single device. The first work on phototransistors dates
from 1951 when Shockley et al. [SHO 51] proposed the use of a
bipolar n-p-n or p-n-p structure as a phototransistor operating with a
base current generated by optical means. The first demonstration of
this type of photodetector was achieved two years later when Shive
[SHI 53] described a Ge-based n-p-n phototransistor, operating in the
spectral region of 1.2 μm wavelength, with an optical gain of the order
of 100. In the 1960s, it was mostly bipolar Si phototransistors,
operating in the near-infrared, which were developed. However, since
the optical absorption coefficient of silicon is weak (<10
3
cm
–1
), the
foreseeable applications were not those requiring high speeds.
This interest in phototransistors has recently been revived with the
development of optical fiber transmission systems and the evolution
of the III-V materials technology. This is because, on the one hand,
Chapter written by Carmen GONZALEZ and Antoine MARTY.
112 Optoelectronic Sensors
the best attenuation and absorption properties of optical fibers are
found in the near-infrared, with the three minimal absorption windows
at 0.85, 1.3 and 1.55 μm corresponding to the emission and absorption
ranges of semiconductor materials based on GaAs and InP. On the
other hand, III-V materials have absorption coefficients much higher
than that of silicon (>10
4
cm
–1
). These two characteristics have
opened up the possibility of applying phototransistors based on III-V
materials to the field of optical telecommunications. Heterojunction
bipolar transistors and field effect transistors, made with these
materials, began to be studied as phototransistors in the 1970s [ALF
73, BAA 77].
In section 4.2, we present a summary of the different types of
phototransistors based on the materials used for their fabrication and
as a result their sensitivity to optical wavelengths. We will then
classify them according to their structure. In section 4.3 we describe
the mechanisms of operation of a bipolar phototransistor and its main
properties. Section 4.4 is dedicated to some examples of circuits based
around bipolar phototransistors. Finally, in section 4.5, we review the
main fields of application of this device.
4.2. Phototransistors
In terms of photodetectors, phototransistors can be classed according
to their fabrication material and, just like their transistor analogs, they
can also be sorted according to structure into two categories: unipolar
field effect devices and bipolar devices. This section reflects this dual
classification.
4.2.1. Phototransistors according to their fabrication materials
The first phototransistors were based on a bipolar homojunction
silicon transistor. In this indirect bandgap material, the energy of the
bandgap is 1.12 eV. As a result, silicon is photosensitive to
wavelengths in the near-infrared, 0.6–0.8 μm. Later, starting in the
1970s, technological progress achieved with III-V compounds and
Phototransistors 113
their ternary and quaternary alloys enabled the development of
heterojunction devices: field effect transistors and bipolar transistors.
These materials have a direct band structure, and some of their ternary
compounds have bandgap energies which match the spectral windows
of lowest attenuation in optical fibers: 0.85, 1.3 and 1.55 μm. This is
why two variants based on III-V materials have been developed: the
GaAs variant (for O = 0.85 μm) and more recently the InP variant (for
O = 1.3 and 1.55 μm). Heterojunction phototransistors based on
AlGaAs/GaAs are photosensitive to wavelengths between 0.8 and
0.9 μm, with the GaAs bandgap being 1.43 eV. This value of bandgap
is compatible with the spectral window of optical fibers around 0.85
μm, for which there is an attenuation of around 2 dB/km. The optical
absorption coefficient of GaAs for this wavelength is relatively high:
of the order of 10
4
cm
-1
. The first optical systems, operating initially
at 0.85 μm, were later targeted at 1.3 and 1.55 μm where the
attenuation in optical fibers is 0.5 and 0.2 dB/km respectively.
Thus, it is the ternary and quaternary alloys based on InP which are
best suited for photodetection at these wavelengths. In particular, the
ternary alloy In
x
Ga
1–x
As with x = 0.53, which matches the lattice
spacing of InP, has a bandgap energy of 0.75 eV. This value of energy
is compatible with photodetection at 1.3 and 1.55 μm. The optical
absorption coefficient of InGaAs is 1.16 u 10
4
cm
–1
for O = 1.3 μm
and 6.8 u 10
3
cm
-1
for O = 1.55 μm. The bandgap energy (E
g
) and the
applicable spectral range for the main materials used in
phototransistor fabrication are shown in Table 4.1.
Table 4.1. Bandgap energy and applicable spectral range for various
materials used in the construction of phototransistors
114 Optoelectronic Sensors
4.2.2. Phototransistors classified by structure
Homojunction or heterojunction bipolar transistors (photo-HBTs)
and unipolar field effect transistors (photo-FETs) are three-terminal
devices. When illuminated, the optical input acts as an additional
terminal across which the device can be controlled optically. These
devices can be integrated into MMICs (Monolithic Microwave
Integrated Circuits) to achieve optically-controlled amplification or
switching, or even more sophisticated microwave functions such as
optical locking of oscillators and the mixing of optical and electrical
signals [SEE 90].
4.2.2.1. Unipolar field effect transistors
Unipolar field effect transistors (FETs), based on III-V materials,
only create a single type of charge carrier. Out of the family of field
effect transistors, most research has been focused on heterojunction
phototransistors which are Schottky gate transistors made with GaAs
(MESFETs, Metal-Semiconductor Field Effect Transistors) and high
electron mobility transistors (HEMT). This popularity is mostly due to
the possibilities of using the phototransistor effect in MMICs
(Monolithic Microwave Integrated Circuits).
Figure 4.1 shows the cross-section of an illuminated MESFET
device (photo-MESFET). Two distinct phenomena occur in the photo-
MESFET during illumination by light: a photoconductor effect which
is the result of the increase in conductivity due to the creation of
photocarriers, and a photovoltaic effect which occurs close to the
gate/channel and channel/substrate junctions. Several investigations
[GAU 85, MAD 92] have shown not only that the photovoltaic effect
dominates the photoresponse of the device, but also that this effect is
the cause of the mediocre dynamic performance of photo-MOSFETs.
Physically, we can explain this degradation by the poor coupling
between the incident light and the gate junction (poor overlap between
the optical absorption region and the region with an electric field). A
large proportion of the photocarriers end up trapped in low field
regions (e.g. the barrier), thus distorting the static characteristics of the
device.
Phototransistors 115
Compared to a photo-MESFET without illumination, an important
reduction in performance has been observed for the transition
frequency f
T
and the maximum frequency of oscillation f
MAX
when
illuminated [SIM 86]. The best dynamic performance obtained under
illumination is for a bandwidth less than 100 MHz [BAR 97].
Dynamic optical performances are very poor compared to those
obtained electrically.
Figure 4.1. Structure of a GaAs MESFET under illumination. The channel is an
n-type semiconducting region sandwiched between a semi-insulating substrate and
the space charge region (depletion region) of the reverse-biased Schottky junction
MESFET-based phototransistors have shown themselves to be a
poor approach for developing photodetectors in the microwave and
millimeter range. Other devices in the FET family, notably HEMT,
suffer from the same limitation [ROM 96]
4.2.2.2. Bipolar phototransistors
An alternative to the photo-MOSFET is the silicon-based
homojunction bipolar phototransistor or the heterojunction bipolar
phototransistor (based on GaAs or InP). In the bipolar phototransistor,
there is an overlap between the region of light absorption and the
high-electric-field depletion region. This overlap is as good as in PIN
photodiodes. As will be shown in section 4.3, the frequency response
to a modulated optical signal is directly related to its purely electrical
properties. To make the best use of the performance of a transistor,
different illumination approaches have been considered, such as
116 Optoelectronic Sensors
vertical illumination through the front or rear face, lateral illumination,
or even the use of a waveguide.
4.2.2.2.1. Traditional surface illumination
The first phototransistors were surface-illuminated, which is the
traditional situation, particularly for silicon phototransistors. This
vertical illumination allows easy coupling between the optical fiber
and the phototransistor, making the integration of the device easier.
Several solutions have been proposed: in the case of an opening in the
metallic contact of the emitter, the incident light flux crosses the
emitter without being absorbed, and the electron-hole pairs are created
in the active (or intrinsic) base-collector region of the photo-HBT, that
is, the region which is just below the emitter. If a base contact is
removed, as shown in Figure 4.2a, the optical flux directly illuminates
the base and the collector in the “extrinsic” region of the
phototransistor.
4.2.2.2.2. Rear-face illumination
The aim of rear-face illumination of a transistor is the creation of
photocarriers in the active region of the transistor (see Figure 4.2b)
without modifying the emitter contact. The traditional HBT structure
is used with a substrate which does not absorb the incident light. The
absorption of light takes place in the collector and the base.
Furthermore, the response coefficient is improved because the
metallic contact of the emitter acts as a mirror.
Phototransistors 117
Figure 4.2. Cross-section of a HBT phototransistor with
(a) front-face illumination and (b) rear-face illumination
4.2.2.2.3. Lateral illumination
Lateral illumination (parallel to the layers) offers another approach
for the injection of light into a transistor. It is inspired by the
technological processes developed for photodiodes. This type of
illumination requires a face very vertical to the base-collector island,
in order to obtain the best possible injection efficiency. In laterally-
illuminated phototransistors, the photons and charge carries no longer
propagate in the same direction, as is the case of vertically illuminated
photo-HBTs. The aim of lateral illumination is to simultaneously
improve the conversion of optical power into electrical power
(quantum efficiency) and the speed performance (high operating
frequencies) of the device.
4.2.2.2.4. Lateral illumination with an integrated waveguide
However, lateral illumination does pose significant difficulties in
terms of achieving a good optical coupling between the fiber and the
device. This is why one trick involves the fabrication of a waveguide
integrated into the structure of the device, as shown in Figure 4.3.
With this approach, the matching of the mode leaving the optical fiber
to the propagation mode inside the device is made easier [FRE 96]. In
the best cases, an improvement to the injection efficiency is observed
which rises from 50% to 90%, with an improved frequency response.
118 Optoelectronic Sensors
4.3. The bipolar phototransistor: description and principles of
operation
The bipolar phototransistor is a transistor designed such that the
“signal” current which feeds the base terminal is mostly provided by
photoelectric effects. This “signal” is then amplified by the transistor
effect of the device. First we revisit the principle of operation of a
bipolar n-p-n phototransistor, which consists of two distinct p-n
junctions, as shown in Figure 4.4. Then, we shall discuss the parameters
characterizing the phototransistor: the response coefficient, the static
and dynamic gains, the response time, the conversion gain and the noise.
A cross-section of an n-p-n type bipolar phototransistor, using mesa
technology, is shown in Figure 4.4, along with its bias circuit.
Figure 4.3. Cross-section of an integrated waveguide
phototransistor under illumination
Figure 4.4. Schematic diagram of an n-p-n phototransistor, along with its bias circuit
in common-emitter configuration. The shaded regions are free from mobile charges.
The base, collector and sub-collector parts can be thought of as a PIN photodiode
Phototransistors 119
4.3.1. The phototransistor effect
In Figure 4.4, the n-p-n phototransistor is biased in the common-
emitter configuration, with voltages V
BE
> 0 and V
CE
> 0. It differs
from traditional bipolar structures, having a relatively large side area
of the base-collector junction; this is called the optical window and is
what becomes illuminated. This part of the device is effectively a
photodiode, connected between the collector and base contacts of the
active transistor. When illuminated, electron-hole pairs are created by
the photoelectric effect, in the base and in the space charge region
(SCR) associated with the base-collector junction (we assume for the
sake of simplicity that no absorption takes place in the sub-collector
and that the collector is entirely free of mobile charges). A
photocurrent I
ĭ
, known as the “primary” photocurrent, is established
between the base and collector regions; it is transported by minority
electrons from the base which diffuse towards the collector, and by the
carriers generated in the SCR, which are separated and moved by the
electric field. The flow of this photocurrent I
ĭ
generates a voltage
across the base-emitter and base-collector junctions such that the
transistor finds itself in its normal operating system (V
BE
> 0 and
V
CE
> 0).
The holes attracted by the base will therefore find themselves
blocked by the emitter-base junction. This excess of holes will cause a
reduction in the emitter-base potential barrier, which results in an
injection of electrons from the emitter into the base, from where the
majority will diffuse until they are at the level of the collector. Thus,
this is the traditional behavior of a bipolar transistor. The
amplification of the photocurrent is a purely electrical phenomenon
due to the transistor effect.
4.3.1.1. The main currents in the phototransistor
In the normal operating system, the phototransistor can be
characterized in terms of the following currents (see Figure 4.5):
– Firstly, the emitter injects electrons into the base region. These,
minority carriers in this region, diffuse perpendicular to the junction
layout and, if the base is thin enough that recombination can be
120 Optoelectronic Sensors
ignored, they reach the depletion region of the base-collector junction,
where the high electric field present in this region clears them out
towards the collector region. The flow of these charge carriers gives
the contribution I
ne
, which is shown in Figure 4.5.
– Conversely, a current I
pe
of holes, majority carriers in the base, is
injected from the base towards the emitter.
– Generation-recombination phenomena mostly occur at the level
of the emitter-base junction, I
reb
, and in the base, I
rb
. I
reb
comes from
the recombination of electrons in the SCR of the emitter-base
junction. I
rb
is caused by the recombination of electrons with holes,
majority carriers, in the base.
– Illumination produces the primary photocurrent I
ĭ
. Mostly created
in the SCR of the base-collector junction, I
ĭ
originates from a current
of electrons which migrate directly toward the collector contact, and
from a current of holes which accumulate at the level of the base.
Thus, balancing the different contributions to the current crossing
the two junctions, listed above, allows us to calculate the total currents
at the emitter, collector and base; in this way, we can use the continuity
of the electron and hole currents across the transition regions to write
the following equations. For the base:
I
E
= I
ne
÷ I
pe
÷ I
reb
[4.1]
the emitter:
I
B
= I
pe
÷ I
rb
÷ I
reb
÷I
)
[4.2]
the collector:
I
C
= I
ne
÷I
rb
÷ I
)
[4.3]
Phototransistors 121
Figure 4.5. Distribution of the different currents of electrical and optical origin
in the phototransistor. The SCRs around the emitter-base
and base-collector junctions are shown by dotted lines
Equations [4.1], [4.2] and [4.3] will be used in the following
sections to define the optical gain and the electrical gain of the
phototransistor.
4.3.1.2. Injection efficiency from the emitter in a homojunction
and a heterojunction
The standard parameters for a bipolar transistor, the injection
efficiency from the emitter J, and the transport factor in the base D
T
,
can be extended to the phototransistor with the help of equations [4.1],
[4.2] and [4.3] where the photocurrent term I
ĭ
was introduced. The
injection efficiency J relates the ratio of the electron current I
ne
injected into the base to the total emitter current I
E
, by the following
equation:
J =
I
ne
I
ne
÷ I
pe
÷ I
reb
[4.4]
Recombination currents reduce the injection efficiency and they
must be minimized. If in equation [4.4] we ignore the effect of
recombination in the emitter-base SCR, we obtain:
122 Optoelectronic Sensors
J =
1

I
pe
I
ne
[4.5]
This ratio must be as close as possible to 1 (I
pe
<< I
ne
) in order to
achieve the maximum injection efficiency. Starting from the transport
equations established for a bipolar phototransistor, it has been shown
that the injection efficiency closely follows the following expression
[CAM 85, MOR 72]:
J =
A.J
0
1÷ AJ
0
[4.6]
where A is a constant, a function of the thickness of the base and the
diffusion length of electrons in the base, and J
0
is the Kroemer factor
[KRO 57a and b] established for an emitter-base heterojunction. The
Kroemer factor is a function of the physical parameters of the
materials making up the heterojunction, and can be expressed in the
following way:
J
0
=
D
nb
L
pe
n
e
D
pe
L
nb
p
b

m
nb
*
m
pb
*
m
ne
*
m
pe
*
(
(
·
·
·
·
·
·
)
)

3 2
exp
'E
g
kT
(
(
·
·
·
·
·
)
)

[4.7]
with:
– 'E
g
the difference in bandgap between the emitter and the base;
– D
nb
the diffusion coefficient of electrons in the base;
– D
pe
the diffusion coefficient for holes in the emitter;
– L
pe
the diffusion length for holes in the emitter;
– L
nb
the diffusion length for electrons in the base;
– n
e
the electron density in the emitter;
– p
b
the density of holes in the base;
– m
*
nb
and m
*
pb
the effective masses of the electrons and holes in
the base;
Phototransistors 123
– m
*
ne
and m
*
pe
the effective masses of the electrons and holes in
the emitter.
For a homojunction bipolar transistor, there is no variation in the
bandgap between the emitter and the base, 'E
g
is zero, and the
exponential factor in equation [4.7] is equal to 1. It is clear from [4.6]
and [4.7] that to obtain a high injection efficiency, J § 1, it is crucial
that the emitter should be much more heavily doped than the base.
Conversely, for a heterojunction, J depends mostly on 'E
g
, and the
term exp('E
g
/kT) becomes dominant compared to the n
e
/p
b
ratio; thus,
to obtain a J close to 1, it is no longer necessary to under-dope the
base relative to the emitter and/or overdope the emitter relative to the
base. In heterojunctions based on the materials GaAs/AlGaAs and
InGaAs/InP, the size of the emitter bandgap (AlGaAs and InP) is more
important than that of the base (GaAs and InGaAs), thus offering a
reduction in the injection of majority carriers from the base into the
emitter. For the heterojunction phototransistor, the base can be doped
at high levels without compromising the efficiency of the junction,
leading to a reduction in the resistance of the base. The doping of the
emitter can remain within relatively low limits, thus reducing the
capacitance of the emitter. These two effects combine to give an
improvement in the current gain and an increase in the high frequency
performance of the heterojunction bipolar phototransistor.
4.3.1.3. Transport factor in the base
The transport factor in the base B is defined by the ratio between
the electron current gathered by the collector and the electron current
injected from the emitted into the base:
B =
I
nc
I
ne
=
I
ne
÷I
rb
I
ne
[4.8]
Due to the recombination current, B is always less than 1.
Beginning with equation [4.8], B can be expressed as a function of the
transit time in the base t
B
and the lifetime of electrons in the base W
n
using the following equation [CAM 85, CAS 89, POU 94]:
124 Optoelectronic Sensors
B =1÷
t
B
W
n
[4.9]
According to [4.9], B is closer to 1 when the transit time t
B
is small
compared to the electron lifetime W
n
. The transit time t
B
is smaller
when the base thickness is small. As a result, it is necessary for the
base thickness to be smaller than the diffusion length of the electrons.
For a phototransistor, as for a transistor, the base should therefore be
as thin as possible.
4.3.2. The response coefficient of a phototransistor
The response coefficient of the photodiode base-collector part of
the phototransistor is defined as the ration of the primary photocurrent
I
ĭ
to the received optical power P
opt
:
S
0
=
I
)
P
opt
A W
( )
[4.10]
The quantum efficiency, which is the ratio of the number of
electrons collected to the number of incident photons, is also used to
characterize the optical-electrical conversion of the base-collector
photodiode. This efficiency, often called the external quantum
efficiency, is expressed in the following way:
K =
I
)
q
P
opt
hQ
=
hQ
q
.S
0
[4.11]
and the response coefficient of the phototransistor is characterized by
the ratio of the component, due to optical excitation (I
C
)
opt
of the
current leaving the device I
C
to that same incident optical power:
S =
I
C
( )
opt
P
opt
=
I
C
( )
opt
I
)

I
)
P
opt
A W
( )
[4.12]
As we will see in the following section, the ratio (I
C
)
opt
/I
ĭ
defines
the optical gain of the phototransistor.
Phototransistors 125
Taking into account equations [4.11] and [4.12] we can express
(I
C
)
opt
in the following way:
I
C
( )
opt
=
q
hQ
K G
opt
P
opt
=G
opt
S
0
P
opt
[4.13]
This last equation clearly shows that the phototransistor has an
effective response coefficient G
opt
times greater than that associated
with the base-collector photodiode.
4.3.3. Static electrical and optical gains of the phototransistor
4.3.3.1. Static electrical gains E
0
and D
0
In the static system, the current gain of the phototransistor is
obtained without illumination and, as a consequence, it is defined in
the same way as for a bipolar transistor. In the common-emitter
configuration, the static electrical gain E
0
is given by:
E
0
=
I
C
I
B
=
I
ne
÷I
rb
I
pe
÷ I
rb
[4.14]
This gain can also be expressed as a function of the injection
efficiency J and of the transport factor in the base B taking into
account equations [4.5], [4.8] and [4.14]:
E
0
=
BJ
1÷BJ
[4.15]
Because of this dependence on J, the current gain in the common-
emitter configuration, E
0
, can reach very high values for the
heterojunction bipolar transistor. We also define the current gain in
common-base configuration D
0
, which is the ratio of the collector
current to the emitter current. Its relationship with the gain E
0
is:
D
0
=
I
C
I
E
=
E
0
1÷E
0
[4.16]
126 Optoelectronic Sensors
4.3.3.2. Static optical gain G
opt
The static optical gain of a phototransistor is obtained under
continuous illumination. It is defined as the ratio between the
component linked to the collector current (I
C
)
opt
and the primary
photocurrent I
)
, obtained at the level of the base-collector photodiode:
G
opt
=
I
C
( )
opt
I
)
=
hQ
qK

I
C
( )
opt
P
opt
[4.17]
G
opt
is the equivalent of the electrical gain E
0
under illumination. It
links the gain obtained through amplification, due to the transistor
effect, to the primary photocurrent I
ĭ
. Several authors have shown
that, for a bipolar transistor, G
opt
is proportional to the product of the
external quantum efficiency and the electrical current gain [CAM 85,
CHA 85]:
G
opt
=K 1÷E
0
( )
[4.18]
Equation [4.18] shows that to achieve a given value of G
opt
, it is
necessary to establish a compromise between the quantum efficiency
K and the electrical gain E
0
, depending on the intended application.
The result is that the larger the value of K, the smaller the value of E
0
,
and vice versa. In addition, G
opt
is always smaller than E
0
.
4.3.4. Dynamic characteristics of phototransistors
In the dynamic system, just as for a normal transistor, the
phototransistor effect can be put to good use in the quasi-linear
“small-signal” mode, in the non-linear mode to achieve multiplication
and mixing behavior and in the “large signal” switching system. In
these three cases, the illumination varies with time and we are
interested either in (amplitude) modulated light or in an impulse.
These three dynamic modes of operation of a phototransistor will be
analyzed below.
Phototransistors 127
4.3.4.1. Small-signal operation
In the context of small-signal applications, the excitation by light
may correspond directly to the signal to be amplified, or alternatively
it may include a continuous component (modulated signals) which can
be put to the particular use of ensuring a pre-bias in addition to that
obtained by electrical access to the base. Phototransistors whose base
is electrically accessible have the advantage of allowing, by selecting
the bias to the base, the choice of an operating point which ensures
optimal linearity. The global response time of a phototransistor is no
different to the “electrical” response time of the transistor element
triggered by the current I
ĭ
induced in the base-collector photodiode.
Two time constants are therefore associated with the dynamic
operating system of a phototransistor: the first is due to the intrinsic
transport of photocurrent carriers, while the second is due to the
electrical response of the phototransistor. The first time constant may
be more or less important depending on the geometry of the device;
however, to begin with, we can presume that it is the electrical
response time which is most significant. The electrical response time
of a phototransistor is linked to its transition frequency f
T0
, which is
the frequency at which the optical gain is equal to 1. In an equivalent
manner to a transistor, this sets the limit for the use of a
phototransistor as a photocurrent amplifier. This transition frequency
can be calculated in a similar way to the calculation of the equivalent
small-signal method for the ʌ hybrid configuration.
4.3.4.1.1. Transition frequency calculated using the ʌ hybrid model
The behavior of a phototransistor in the small-signal system is
shown in the equivalent circuit diagram of Figure 4.6b. This is based
on an electrical model of a bipolar transistor known as the Giacoletto
circuit, which is established for the common-emitter arrangement
[LET 78]. The incident illumination is modeled by a photocurrent
source i
ĭ
placed between the base B and the collector C. The output
signal is taken across the terminals of a load resistance R
L
connected
to the collector.
128 Optoelectronic Sensors
Figure 4.6. (a) Electrical model and (b) its equivalent electrical circuit in the small-
signal system of a phototransistor, for a common-emitter configuration
The elements of the above equivalent circuit can be described as
follows:
– R
BE
is the dynamic input resistance of the phototransistor;
– R
BE
= E
0
U
T
/I
C
;
– C
TBE
and C
TBC
are the transition capacitances of the emitter-base
and base-collector junctions;
– C
SBE
is the diffusion capacitance, which represents the effects of
accumulation of minority charge carriers in the base, effects linked to
the transit time W
B
in this region, C
SBE
= W
B
.I
C
/U
T
;
– g
m
defines the transconductance, g
m
= I
C
/U
T
.
Under these conditions the input photocurrent i
ĭ
, called the
primary, can be expressed in the following form:
i
)
=v
BE

1
R
BE
÷ jZ C
TBE
÷C
SBE
( )

l
l
l
l
l
÷ v
BE
÷ R
L
i
C

l
l
l
jZ C
TBC
[4.19]
in which the second term (corresponding to the current component i
indicated in Figure 4.6) represents the internal reaction mechanism
induced by the transition capacitance of the base-collector junction
C
TBC
and by the load resistance R
L
, commonly referred to as the
Miller effect. If we now ignore the current contributions i and i
ĭ
Phototransistors 129
compared with the amplified current g
m
v
BE
, the output collector
current is found to be:
i
C
= g
m
v
BE
[4.20]
Given this, the ratio i
C
/i
ĭ
, which defines the dynamic current gain
of the phototransistor g
opt
, known as the optical gain in the small-
signal system, can be written in the form:
i
C
i
)
jZ
( )
= g
opt
jZ
( )
=
E
0
1÷ j Z E
0

C
TBE
÷C
TBC
g
m
÷W
B
÷ R
C
C
TBC
(
(
·
·
·
·
)
)

[4.21]
At high frequencies, the imaginary part of the denominator of
equation [4.21] becomes dominant and we can write:
g
opt
jZ
( )
=
1
jZ
C
TBE
÷C
TBC
g
m
÷W
B
÷ R
L
C
TBC
(
(
·
·
·
·
)
)

[4.22]
If we assume that the product R
C
.C
TBC
is weak compared to the
two first terms, |g
opt
| is obtained when:
1
Z
=
1
Z
TO
=
C
TBE
÷C
TBC
g
m
÷W
B
[4.23]
Equation [4.23] defines the angular frequency of the transition Z
T0
,
with the transition frequency f
T0
being Z
T0
/2S. Physically, the
transition frequency, f
T0
, corresponds to the total transit time, W
T
, of
photocarriers from the emitter to the collector. It is interesting to
analyze this time constant W
T
, associated with f
T0
. It is defined as:
130 Optoelectronic Sensors
W
T
=
1
Z
TO
=
1
2S f
TO
[4.24]
If we substitute equation [4.23] into [4.24] we find:
W
T
=
C
TBE
g
m
÷
C
TBC
g
m
÷W
B
[4.25]
The first two terms of equation [4.24] correspond to the charging
times of the base-emitter and base-collector junctions respectively.
Equation [4.25] shows that W
T
depends on the collector current I
C
(via
g
m
) up to a certain threshold and, if we take into account equation
[4.13], W
T
is inversely proportional to the incident optical power P
opt
.
For high values of P
opt
, W
T
approaches the limiting value W
B
. On the
other hand, at low values of P
opt
, the terms in C
TBE
and C
TBC
become
dominant and the transition frequency f
T0
is proportional to P
opt
.
This behavior of f
T0
as a function of the incident optical power has
been experimentally observed, as shown in Figure 4.7. This
dependence makes the use of relatively high optical powers (>10 μW)
necessary. Alternatively, we can use an electrical current via the
metallic base contact to pre-bias the structure, thus reducing the
“threshold” optical power needed to reach the maximum value of f
T0
.
With the help of equation [4.22] we can also define the cutoff
angular frequency Z
E
as being the angular frequency at which |g
opt
|
falls by 3 dB. This is equal to:
1
Z
B
= E
0
C
TBE
÷C
TBC
g
m
÷W
B
÷ R
L
C
TBC
(
(
·
·
·
·
)
)

[4.26]
Phototransistors 131
Figure 4.7. Variation in transition frequency f
TO
with incident optical power P
opt
.
At weak P
opt
, f
TO
varies rapidly up to a certain limiting value of
P
opt
, after which it remains constant
Figure 4.8. Frequency response of a phototransistor for different load resistances
It can be clearly seen from equation [4.26] that the cutoff angular
frequency Z
E
depends on the operating point (as does Z
TO
) and on the
value of the load resistance R
C
, as shown in Figure 4.8.
4.3.4.2. Nonlinear operation
A bipolar phototransistor behaves similarly to a bipolar transistor,
and as a result possesses nonlinear characteristics which allow it to act
as an optical-electrical mixer. It can mix a modulated optical signal,
carrying information, with an electrical signal from a local oscillator.
132 Optoelectronic Sensors
This mixing behavior has been the subject of recent research [BET 98,
GON 98, SUE 96] and has been developed for telecommunication
applications. Two mixing configurations are possible:
– transposition of a low frequency input signal into a higher
frequency signal. This mode of operation will be referred to below as
up-conversion;
– transposition of a high frequency input signal into a lower
frequency signal. This mode will be referred to as down-conversion;
4.3.4.2.1. Mixing principles
Any device able to transpose an input signal from a frequency f
E
into another higher or lower frequency is known as a mixer. This
change in frequency originates in the nonlinear properties of the mixer
and can be explained in the following manner: consider a circuit
element whose I/V current-voltage characteristics are nonlinear (a
nonlinear resistance, a Schottky diode, etc.). If it is subjected to a
voltage V
OL
= v
OL
sin(Z
OL
t), its I/V nonlinearity can be described by
expanding the current flowing through the element in terms of a
discrete series that is a function of the voltage V:
I = I V
( )
= I
0
÷a
1
V ÷a
2
V
2
÷a
3
V
3
÷... [4.27]
where I
0
is a continuous bias current and a
1
, a
2
and a
3
are real
constant coefficients. [4.27] shows that we see an infinite number of
powers of V appearing at the output of the circuit, these will enrich the
spectrum of the input signal V. If we add a second signal to the input,
V
FI
= v
FI
sin(Z
FT
t), the spectrum of the output signal becomes even
more complicated due to the presence of the products of V
OL
and V
IF
and their respective harmonics.
If we now replace V with V
OL
+ V
FI
in [4.27], the current (V)
becomes:
Phototransistors 133
I =I
0
÷a
1
V
OL
÷V
FI
( )
÷a
2
V
OL
÷V
FI
( )
2
÷...
=I
0
÷a
1
v
OL
sinZ
OL
t ÷v
FI
sinZ
FI
t
( )
÷a
2
v
OL
sinZ
OL
t ÷v
FI
sinZ
FI
t
( )
2
÷...
=I
0
÷a
1
v
OL
sinZ
OL
t ÷v
FI
sinZ
FI
t
( )
÷a
2
1
2
v
OL
2
1÷cos2Z
OL
t
( )
÷v
OL
v
FI
cos Z
OL
÷Z
FI
( )
t ÷cos Z
OL
÷Z
FI
( )
t

l
l
l
'
|
1
1
+
1
1
÷
1
2
v
FI
2
1÷cos2Z
FI
t
( )
'
¦
1
1
+
1
1
÷...
[4.28]
The interesting frequencies are (Z
OL
+ Z
FI
) and (Z
OL
– Z
FI
) and
they can be extracted by a filter tuned to the target frequency. In a
bipolar mixing transistor, the main nonlinearity contributing to the
mixing behavior is transconductance. For a bipolar phototransistor, the
nonlinearity in the current gain is the most important parameter.
4.3.4.2.2. Performance criteria
As a first approximation, we can say that the performance criteria
established for a mixing transistor can be applied to a mixing
phototransistor. We will discuss two of these criteria: conversion gain
and insulation.
Figure 4.9. Optical and electric inputs of the phototransistor
(in place of the mixing phototransistor)
134 Optoelectronic Sensors
Conversion gain
In a mixing phototransistor, the optical input is physically
separated from the electrical input (see Figure 4.9). The optical signal
“enters” through the optical window and the electrical signal is
applied to the metallic contact of the base. In the common-emitter
configuration, the mixed signal is obtained at the output of the
collector. The traditional definition of the conversion gain G
con
,
applied to a phototransistor, can be expressed as the ratio between the
output electrical power at the “mixed” frequency and the effective
available input optical power at the modulation frequency of the light.
Taking the case of up-conversion, the light is modulated at an
intermediate frequency F
IF
, the frequency of the local oscillator F
OL
is
much higher than F
IF
, and the mixed frequencies F
RF
are:
F
RF
= F
OL
± F
FI
The conversion gain can be written as:
G
con
=
P
elec
F
RF
( )
P
opt
F
FI
( )
[4.29]
In equation [4.29], the denominator refers to an optical power. In
order to determine the electrical power generated by the incident
optical power, we must take into account the primary photocurrent I
ĭ
generated by the illumination at the level of the base-collector junction,
and of the input impedance associated with this junction, Z
BC
. The
electrical power can thus be expressed by the equation:
P
elec
=
1
2
Real Z
BC
( )
I
)
2
[4.30]
and the electrical conversion gain can be expressed as:
G
con
=
P
RF
P
elec
( )
FI
=
2P
RF
Reel Z
BC
( )
I
)
2
[4.31]
Phototransistors 135
Another definition of the conversion gain is possible. It is defined
as being the ratio between the power of the mixed signal F
RF
and the
“primary” power of the incident signal F
IF
, in other words, the power
of F
IF
obtained when the base-emitter junction of the phototransistor
is short-circuited (base-emitter voltage V
BE
= 0 V).
G
con
=
P
RF
P
FI
( )
V
BE 0
[4.32]
When describing experimental results this second definition of
G
con
is the most widely used.
Insulation
All the frequencies generated by the mixing are present at each of
the terminals of the mixer, and this has a significantly detrimental
affect on its conversion gain performance. As Figure 4.9 shows, the
phototransistor has three points of access, two electrical, the base
contact (OL) and the collector contact (RF), and one optical, the
optical window (IF). The insulation is measured between two of the
mixer’s points access, IF-OL and IF-RF, by the ratio between the
power at the frequency F
OL
(or F
RF
) present at the access point OL (or
RF) and the power at the frequency F
OL
(or F
RF
) present at the IF
access point. Let us consider the OL and IF access points at which the
F
OL
and F
IF
signals are injected. If we hypothesize that no electrical
signal originating from the access point of the base can be converted
into an optical signal leaving through the optical window, the IF-OL
insulation is infinite. The same reasoning can be applied to the IF-RF
access point. As a result, we obtain a maximum insulation between
these two pairs of terminals without needing to use external circuitry.
However, for insulations in the opposite direction, OL-IF and RF-IF,
there can be a loss of F
IF
signal power in each of these two electrical
access points.
4.3.4.3. Operation in the strong-signal system
As was mentioned previously, the other application of a
phototransistor concerns the detection of radiation in the strong-signal
system. In this case, the presence or not of the illumination induces a
136 Optoelectronic Sensors
switching type of behavior: the device initially blocks a signal, or is in
the weak conduction system if it is pre-biased. It then enters the
normal, or even saturated, system when light is applied; it will then
return to the initial mode of operation when the input signal ceases.
These system changes give rise to transitory phases, as shown in
Figure 4.10, which defines a response time for each: the lag time t
W
is
inherent to the time it takes the transistor to start conducting (in the
case of operation without pre-bias) and is linked to the charging of the
transition capacitances C
TBE
and C
TBC
. By considering the mean
values of these parameters, this time can be calculated using the
expression:
t
r
=
0.7 C
TBE
÷C
TBC
( )
I
)
[4.33]
Figure 4.10. Representation of the different switching
response times of phototransistors
Phototransistors 137
The rise time t
r
in the normal active system (up to saturation, if
applicable) is more difficult to evaluate, bearing in mind the variation
in numerous electrical parameters that apply in this transitory system;
nevertheless, the dynamic behavior of the phototransistor can be
approached in a simple manner, using the small-signal model. If we
assume that the current gain and its cutoff frequency are independent
of the level of injection; the result is that the first-order model, which
relates I
C
to I
ĭ
in the context of this assumption, suggests that the
collector current response to a step change to the primary photocurrent
in the base follows an exponential law with time constant 1/Z
E
. Thus,
the rise time which separates the start of the growth in the collector
current (conventionally, I
C
equal to 10% of its final value) from the
end of this first transient phase (I
C
equal to 90% of its steady-state
value), being either E
0
.I
ĭ
or I
Csat
, can be calculated as:
t
m
=
2.9
Z
E
[4.34]
when saturation is not reached and:
t
m
=
0.8
Z
T

I
C
sat
I
)
[4.35]
in the alternative case.
For opposite switching, the desaturation time corresponds to the
removal of the accumulated surplus charge in the form of the minority
carriers (electrons) in the vicinity of the base, and is therefore defined
by the lifetime of these charge carriers.
Over the course of the decaying phase of the collector current, the
transistor is once more in the normal active system, up to the point it
stops conducting. The fall time t
f
can again be expressed by equations
[4.34] and [4.35].
138 Optoelectronic Sensors
,
Figure 4.11. Response time of a phototransistor as a function of the
collector current and for different values of the load resistance
Here once more, in the switching system, the rise time t
r
and the
fall time t
f
increase with the load resistance R
L
. We also observe that,
while it leads to an increase in response time, the presence of
saturation leads to a reduction in these rise and fall times. Finally, the
addition of a pre-bias also contributes to an improvement in t
r
and t
f
,
as can be seen in Figure 4.11. Overall, depending on the type of
phototransistor and the circuit used, along with its load resistance and
operating point, these response times vary from a few tenths to tens of
microseconds.
4.3.5. Noise in phototransistors
The phototransistor is the first element in a photoreceiver system.
This means that there is a certain threshold power below which the
photoreceiver cannot detect a signal: a power which is determined by
the power of the noise in the phototransistor. The minimum detectable
power in the photoreceiver system is limited by the noise in the
phototransistor and its load circuit. The noise in the phototransistor is
hence an important criterion for judging the performance of the whole
photodetection apparatus. The most commonly-used approach for
characterizing the noise of a photodetector is the noise equivalent
current generator, applied at the input [CAM 82, SMI 80, WAN 86].
This is the approach that we will develop over the rest of this section.
Phototransistors 139
Four main noise sources can be associated with a HPT:
– the shot noise due to the base current I
B
entering the phototransistor,
this current is the sum of the photocurrent I
ĭ
and the electrical bias
current I
Belec
;
– the shot noise due to the collector current I
C
, at the output of the
HPT;
– the thermal noise due to the load resistance R
C
;
– the thermal noise generated in the base region and, in contrast to
the other quasi-neutral regions, the base, although heavily doped,
presents an effective base resistance R
B
which can reach values as
high as hundreds of Ohms.
The spectral densities of the noise power, i
2
'f in A
2
/Hz,
associated with each source of noise are, for the base, the following:
i
base
2
'f
= 2 2q I
)
÷ I
Belec
( ) ( )
÷
4kT
R
BE
[4.36]
The factor of 2 in equation [4.36] is due to the correlation which
exists between the base-emitter and base-collector junctions [MON
71]. Physically, this arises from the fact that each fluctuation in the
base current 'I
B
simultaneously produces another equivalent
fluctuation in the emitter current to compensate for 'I
B
. These two
currents, in the opposite direction, cancel through recombination, but
their shot noises are independent and add up. For the collector:
i
C
2
'f
output
÷
i
th
2
'f
output
=2qI
C
÷
4kT
R
L
[4.37]
These two spectral densities associated with the collector at the
output of the transistor are linked to the optical input of the
phototransistor in the following manner [THU 99]:
i
C
2
'f
input
÷
i
th
2
'f
input
=
1
g
opt
2
2qI
C
÷
4kT
R
L
(
(
·
·
·
·
)
)

[4.38]
140 Optoelectronic Sensors
where g
opt
is the dynamic optical gain of the phototransistor, defined
in equation [4.21].
Some degree of correlation can exist between the sources of shot
noise in the base and collector. Nevertheless, for frequencies above
the noise cutoff frequency 1/f, the correlation between these two
sources can be assumed to be negligible [ESC 95]. As a result, the
total spectral density of noise power of the phototransistor, in terms of
its optical input, can be expressed as the sum of all the spectral
densities, equations [4.36] and [4.38]:
i
Tot
2
'f
=4 q I
)
÷ I
Belec
( )
÷
4kT
R
B
÷
1
g
opt
2
2 q I
C
÷
4kT
R
L
(
(
·
·
·
·
)
)

A
2
Hz
(
(
·
·
·
·
·
)
)

[4.39]
According to equation [4.39], the total spectral noise density of the
phototransistor is dominated at low frequencies by the sources of
noise linked to the photocurrent I
ĭ
, to the bias current of the phase
I
Belec
and to the base resistance R
BE
; at high frequencies, the noise
source due to the collector current becomes dominant.
4.4. Photodetector circuits based on phototransistors
4.4.1. Amplification circuits
Phototransistors can be used either in the linear detection system
(continuously modulated signals or pulsed signals) or in the switching
system, in circuits with (see Figure 4.12a) or without (see Figure
4.12b) pre-bias, with of course the possibility of adding an additional
level of amplification (see Figure 4.12c). It has been shown that in
every application – detection of weak or strong signals – the speed is
limited by the value of the load resistance of the phototransistor. In
order to improve sensitivity, it is common nowadays to use either an
active load with low input impedance (a common-base transistor) (see
Figure 4.12d) or an operational amplifier providing the current-
voltage conversion (see Figure 4.12e). Finally, we should note that the
phototransistor can be integrated with a second transistor to form a
Phototransistors 141
photo-Darlington pair (see Figure 4.12f). This structure clearly gives a
greater response coefficient at the cost of deterioration in the operation
speed.
Figure 4.12. Main circuit diagrams for phototransistors: a) structure with two
electrodes E and C, b) pre-bias using access to the base, c) supplementary
amplification with a second stage, d) improvement in speed using a low impedance
load (common-base transistor) or e) by current-voltage conversion, f) integration on
silicon of a second stage: the photo-Darlington pair. Note: structures c) and d) are
used in fast circuits
Structures c) and d) have been used to make monolithic amplification
circuits integrating the phototransistor and several HBTs, for wideband
[CHA 92, KAM 00, WAN 86a] and narrowband [GON 00a, KAM 95]
applications.
4.4.2. Nonlinear circuits
Phototransistors can also be used in optoelectric mixing and auto-
oscillation configurations. The simplest circuit is shown in Figure
142 Optoelectronic Sensors
4.13a and consists of a phototransistor and two matching cells. The
first cell is placed between the source OL and the input to the base of
the phototransistor. The second cell is placed between the output of
the collector and the load resistance R
L
. The optical signal is
modulated at the intermediate frequency IF, and the mixed signals at
frequencies OL ± IF are recovered at the output of the collector. The
second circuit, described by Sawada et al. [SAW 99] and depicted in
Figure 4.13b, consists of a phototransistor and a circuit associated
with its base which allows it to auto-oscillate at the OL frequency of
9.567 GHz. This frequency is stabilized using a dielectric resonator
DR. The optical signal is modulated at the frequency of 0.2 GHz, and
the mixed signals recovered with up-conversion are at frequencies of
9.367 and 9.767 GHz. The third circuit, described by Lasri et al. [LAS
00], is shown in Figure 4.13c. It consists of two phototransistors with
their emitters joined together. One phototransistor is used as an auto-
oscillator at 30 GHz locked optically to an optical signal O
1
modulated
at the same frequency OL. This signal OL is coupled with the second
phototransistor, which acts as an optoelectrical mixer between the
30 GHz signal and a second optical signal, O
2
, modulated at 300 MHz.
The two optical signals O
1
and O
2
could be at wavelengths
corresponding to a WDM (wavelength division multiplexing) scheme.
These circuits demonstrate the abilities of phototransistors to combine
the functions of photodetection, amplification, mixing and auto-
oscillation.
4.5. Applications
4.5.1. Galvanic isolation
Silicon phototransistors have been used for a long time in optical
isolators. These devices may or may not have a base contact. Figure
4.14 shows an optical isolator based on a phototransistor. For this
application, all that is required is to connect a structure upstream
which takes care of the conversion of an input electrical signal into a
light beam. This is in order to achieve a transfer of information which
avoids any form of electrical connection. In this way, galvanic
isolation between the controller and the receiver is ensured, and the
Phototransistors 143
achievable voltage difference, which in the static system is determined
by the distances which separate the two components or the
connections to the different electrodes, can reach several kilovolts.
Figure 4.13. Various circuit diagrams for phototransistors acting as (a) a simple
optoelectrical mixer, (b) auto-oscillation and mixing with only one phototransistor
and (c) auto-oscillation and mixing with two phototransistors
Figure 4.14. Schematic representation of an optical isolator
based on a phototransistor
144 Optoelectronic Sensors
The device providing the electron-photon conversion at the input is
normally a light-emitting diode (LED) based on GaAs, whose
emission line of around 850 nm is well suited to optical detection by
the phototransistor or the photo-Darlington pair. The current
conversion gain can reach 500%. An example of its variation as a
function of the input current is shown in Figure 4.15.
,
Figure 4.15. Current transfer level (current from the collector of the
phototransistor/current in the input LED) as a function of the
input current to the optocoupler
Another application of silicon phototransistors involves power
electronics. These types of devices must often meet strict requirements
of galvanic isolation, particularly for control applications and energy
conversion from the industrial electricity grid. Control through optical
means allows this specific problem to be solved in an efficient and
elegant manner: with the thyristor and the triac – which are the two
basic bipolar components in this application domain – constructed
around a photodiode, whose response replaces the current from the
control electrode, the gate. A schematic cross-section for a
photothyristor and the equivalent circuit for a phototriac are shown in
Figure 4.16.
Phototransistors 145
Photothyristors present very interesting possibilities for direct current
power transmission at high voltages (1,000-1,500 A and 5-8 kV)
[ARN 92].
The phototriac is constructed by combining two photothyristors in
parallel, in opposing directions. The near-obligation, for monolithic
integration of this device onto a silicon substrate, of using only one
surface of the substrate minimizes the current capacities and reduces
the voltage stability. Because of this, these devices are applied in the
domain of static relays which involve currents and voltages not
exceeding 1 A and 1 kV.
4.5.2. Phototransistors for optical telecommunications
This application, which requires both a good response coefficient
and speed, mostly involves heterojunction bipolar transistors based on
InP and InGaAs.
Figure 4.16. (a) Schematic cross-section of a photothyristor
and (b) the equivalent electric circuit for a phototriac
In the basic process of an optical transmission system, shown in
Figure 4.17, the phototransistor is the device which converts the
optical signal into an electrical signal. It is situated at the input of the
receiver, acting as an optical detector. In the transmitter, the light is
modulated (directly or indirectly via an optical modulator) with the
signal carrying the information. This signal can be of an analog or
146 Optoelectronic Sensors
digital. After transport through the optical fiber, the optical signal is
demodulated by the photodetector in order to recover the electrical
signal. This signal is subject to noise and distortion, and so the
receiving circuit may need to amplify and reconstruct the signal in
order to extract the original information.
Figure 4.17. Schematic diagram of an optical transmission system
with an analog transported signal
An InP/InGaAs phototransistor is an alternative to traditional
photodetection, consisting of a PIN photodiode associated with a
preamplifier. As was previously stated, this device is of interest
because of its combination of photodetection and amplification
behavior and, since its structure is similar to that of a heterojunction
bipolar transistor, it can attain a frequency performance comparable to
that of a HBT based on InP [BLA 00]. In addition, when acting as an
optical-electrical mixer, it can convert the demodulated signal to other
frequency ranges depending on the required output.
4.5.2.1. InP/InGaAs phototransistors as pre-amplifying
photodetectors
The transport of analog and digital data takes place at frequencies
and bitrates that are constantly increasing. In digital transmission, for
example, currently 10 and even 50 Gbit/s can be achieved. As a result,
phototransistors must become faster and faster while maintaining a
good level of sensitivity. The structure of a phototransistor based on
InP/InGaAs is shown in Figure 4.18. This device is sensitive to optical
wavelengths, O = 1.3 and 1.55 μm. The base and collector layers are
made of InGaAs and the emitter layer is InP. The optical window sits
directly above the base. This has been optimized to achieve an optical
gain of 30 dB and a transition frequency of 88 GHz [GON 00]. These
Phototransistors 147
high performances are the result of many refinements, particularly to
reduce the dimensions of the optical window, the base-emitter and base-
collector junctions (in order to reduce their capacitances), as well as the
use of a base layer with a graded composition, in order to increase the
gain and speed.
,
Figure 4.18. Vertical structure of a phototransistor based on InP/InGaAs
on a semi-insulating substrate of InP doped with Fe
Figure 4.19 shows the response coefficient of the phototransistor in
its two modes of operation, transistor (Tr Mode) and photodiode (PIN
mode), as a function of the modulation frequency of light. The
response coefficient (R) is expressed in dB, R (dB) = 20 log S, with S
expressed in A/W. The PIN mode corresponds with the
phototransistor operating as a photodiode (V
BE
= 0V) and the Tr mode
operating with the transistor effect (V
BE
> 0V). As discussed in section
4.3.2, the optical gain G
opt
, expressed in dB, is equal to the difference
between the response coefficient in the Tr mode and that of the PIN
mode. The response coefficient in the PIN mode, S
0
, is 0.25 A/W.
This response coefficient S
0
has been increased to 0.44 A/W in
another phototransistor, while maintaining an equivalent optical gain
(32 dB). However, as predicted by equation [4.18], the transition
frequency drops to 56 GHz [THU 99a].
148 Optoelectronic Sensors
Figure 4.19. Frequency response of an InP/InGaAs photodiode.
The response coefficient (photoresponse) is expressed in dB
Monolithic photoreceiver circuits for wideband amplification
involving photo-HBTs and HBTs have been constructed. The
amplifier circuit proposed and built by [KAM 00] exhibits a
bandwidth of 40 GHz. The circuit diagram is shown in Figure 4.20.
The circuit consists of a phototransistor and two HBT transistors. The
two types of device have a double heterojunction and use the same
epitaxial layers.
Figure 4.20. Monolithic circuit diagram for wideband amplification consisting of a
phototransistor and two double heterojunction transistors, using InP technology
[KAM 00]. The structure of the epitaxial layers is the same for the two devices
Phototransistors 149
4.5.2.2. InP/InGaAs phototransistors as an optoelectrical mixer
The optoelectrical mixing transistor can be used in mixed
radio/fiber network access, or in satellite communications, to achieve
frequency conversion (up- or down-conversion). It can also be used in
high bitrate communication systems as an integral part of timing-
signal extraction circuits. Suematsu et al. [SUE 96] were among the
first to use a photo-HBT based on GaAs, to achieve conversion from
the modulation frequency of the light at 2 GHz to a frequency band at
50 GHz. Later, other investigations were reported [GON 98, GON 00],
using InP/InGaAs technology, to achieve conversion from an
intermediate frequency (”2 GHz) to frequency bands at 30 and 40
GHz. An example of an experimental setup used to achieve this high
frequency conversion is shown in Figure 4.21. The light emitted by a
laser source at 1.55 μm is amplitude modulated at an intermediate
frequency IF (via an external EOM modulator) in the frequency range
of 200 MHz to 2 GHz. The modulated optical signal is injected into
the optical window of the phototransistor via a single-mode optical
fiber. The base contact is connected to a local oscillator LO at a
frequency of 30 GHz. After demodulation of the optical signal, the
phototransistor mixes the OL and IF signals and the result is
visualized with the help of a spectrum analyzer.
Figure 4.22 shows the output electrical powers of the IF signal (in
PIN mode) and the mixed signal OL + IF, as a function of the IF
frequency. The conversion gain, using the definition given in [4.32], is
the ratio of the output power of the LO + IF signal to that of the IF
signal detected in PIN mode. In the frequency range of IF, from 0.2 to
2 GHz, the conversion gain for the LO + IF signal is relatively
constant, of the order of 7 dB. This positive conversion gain is high
because of the direct amplification of the mixed signal by the
phototransistor.
150 Optoelectronic Sensors
Figure 4.21. Experimental setup for up-conversion from the intermediate
frequency IF to a band at 30 GHz, using the HPT as an O/E mixer
Figure 4.22. Conversion gain for the IF signal transposed to a frequency
of 30 + IF GHz, using the InP/InGaAs phototransistor as an O/E mixer
4.6. Conclusion
In this chapter we have covered the main characteristics of the
bipolar phototransistor. We have also presented its applications in the
field of galvanic isolation for silicon phototransistors, and in the field
of optical telecommunications for InP/InGaAs heterojunction
phototransistors. The study of monolithic circuits incorporating
InP/InGaAs phototransistors, and providing more complex functions
such as optical-electrical mixing and auto-oscillation, is the particular
aim of research intended to exploit the synergy between photonics and
Phototransistors 151
microwaves, with the goal of meeting the ever-growing requirements
in sensitivity, speed and functionality of photodetectors.
4.7. Bibliography
[ALF 73] ALFAROV Z.I., AKHMEDOV F., KOROL’KOV V., NIKETIN V., “Phototransistor
utilizing a GaAs-AlAs heterojunction”, Sov. Phys. Semicond., vol 7, no. 6, p. 780,
1973.
[ARN 92] ARNOULD J., MERLE P., Dispositifs de l’électronique de puissance, Hermes,
Paris, 1992.
[BAA 77] BAAK C., ELZE G., WALF G., “GaAs MESFET: a high-speed optical
detector”, Electron. Lett., vol. 13, p. 139, 1977.
[BAR 97] DE BARROS L.E., PAOLELLA A., FRANKEL M., ROMER M., HERCZFELD P.,
MADJAR A., “Photoreponse of microwave transistors to high-frequency modulated
lightwave carrier signal”, IEEE Trans. on Microwave Theory Tech., vol. 45, no. 8,
p. 1368-1374, 1997.
[BET 98] BETSER Y., RITTER D., LIU C.P., SEEDS A. J., MADJAR A., “A single stage
three-terminal heterojunction bipolar transistor optoelectronic mixer”, J. Light.
Tech., vol. 16, no. 4, p. 605-609, 1998.
[BLA 00] BLAYAC S., RIET M., BENCHIMOL J.L., BERDAGUER P., KAUFFMAN N.,
GODIN J., SCAVENNEC A., “Lateral design of InP/InGaAs DHBTs for 40 Gbit/s”,
Proc. of the Indium Phosphide and Related Materials, IPRM’2000, p. 481-484,
2000.
[CAM 82] CAMPBELL J.C., OGAWA K., “Heterojunction phototransistors for long-
wavelength optical receivers”, J. Appl. Phys., vol. 53, no. 2, p. 1203-1208, 1982.
[CAM 85] CAMPBELL J.C., “Phototransistors for lightwave communications”,
Semiconductors and Semimetals, vol. 22D, p. 389-447, 1985.
[CAS 89] CASTAGNE R., DUCHEMIN J.P., GLOANEC M., RUMELHARD C., Circuits
intégrés en arséniure de gallium, Masson, 1989.
[CHA 85] CHAND N., HOUSTON P.A., ROBSON P.N., “Gain of a heterojunction bipolar
phototransistor”, IEEE Transactions on Electron Devices, vol. ED-32, no. 3, p.
622-627, 1985.
[CHA 92] CHANDRASEKHAR S., LUNARDI L.M., GNAUCK A.H., RITTER D., HAMM
R.A., PANISH G.L., QUA G.L., “A 10 Gbit/s OEIC photoreceiver using
InP/InGaAs heterojunction bipolar transistors”, Electronics Letters, vol. 28, no. 5,
p. 466-468, 1992.
152 Optoelectronic Sensors
[ESC 95] ESCOTTE L., ROUX J. P., PLANA R., GRAFFEUIL J., GRUHLE A., “Noise
modelling of microwave heterojunction bipolar transistors”, IEEE Transactions
on Electron Devices, vol. 42, no. 5, p. 883-888, 1995.
[FRE 96] FREEMAN P., ZHANG X., VURGAFTMAN I., SINGH J., BHATTACHARYA P.,
“Optical control of 14 GHz MMIC oscillators based on InAlAs: InGaAs HBTs
with monolithically integrated optical wave guides”, IEEE Transactions
Electrons Devices, vol. 43, no. 3, p. 373-379, 1996.
[GAU 85] GAUTIER J.L., PASQUET D., POUVIL P., “Optical effects on the static and
dynamic characteristics of a GaAs MESFET”, IEEE Trans. on Microwave Theory
Tech., vol. 33, no. 9, p. 819-822, 1985.
[GON 98] GONZALEZ C., THURET J., RIET M., BENCHIMOL J.L., “Optoelectronic up-
converter to millimetre-wave band using an heterojunction bipolar
phototransistor”, Proc. of the European Conference on Optical Communication,
ECOC’98, vol. 1, p. 443-444, 1998.
[GON 00] GONZALEZ C., MULLER M., BENCHIMOL J.L., RIET M., JAFFRÉ P., LEGAUD
P., “HBT phototransistor for remote upconversion in hybrid fibre radio
distribution systems”, Proc. of the European Conference on Optical
Communication, ECOC 2000, vol. 2, p. 103-104, 2000.
[GON 00a] GONZALEZ C., MULLER M., BENCHIMOL J.L., RIET M., JAFFRÉ P., LEGAUD
P., “A 28 GHz HPT/HBT monolithically integrated photoreceiver for hybrid fibre
radio distribution systems”, Proc. of the 8
th
IEEE International Symposium on
Electron Devices for Microwave and Optoelectronic Applications, EDMO 2000,
p. 55-60, 2000.
[KAM 95] KAMITSUNA H., “Ultra-widebande monolithic photoreceivers using HBT
compatible HPTs with novel base circuits and simultaneously integrated with an
HBT amplifier”, J. Lightwave Tech., vol 13, no. 12, p. 2301-2307, 1995.
[KAM 00] KAMITSUNA H., MATSUOKA Y., YAMAHATA S. AND SHIGEKAWA N., “A 82
GHz-Optical-gain-cutoff-frequency InP/InGaAs double-hetero-structure
phototransistor (DHPT) and its application to a 40 GHz band OEMMIC
photoreceiver”, Proc. of the European Microwave week 2000, EuMC 2000,
EuMC36, 2000.
[KRO 57a] KROEMER H., “Quasi-electric and quasi-magnetic fields in non-uniform
semiconductors”, RCA Rev. vol. 18, p. 332, 1957.
[KRO 57b] KROEMER H., “Theory of a wide-gap-emitter for transistors”, Proc. IRE,
vol. 45, p. 1535, 1957.
[LAS 00] LASRI J., BILENCA A., EISENSTEIN G., RITTER D., ORENSTEIN M., SIDEROV
V., GOLDGEIER S., COHEN S., “A two heterojunction bipolar photo-transistor
configuration for millimeter wave generation and modulation”, Proc. of the
International Topical Meeting on Microwave Photonics, MWP’00, p. 62-65,
2000.
Phototransistors 153
[LET 78] LETURCQ P., REY G., Physique des composants actifs à semiconducteurs,
Dunod University, 1978.
[MAD 92] MADJAR A., PAOLELLA A., HERCZFELD P.R. “Analytical model for
optically generated currents in GaAs MESFET”, IEEE Trans. on Microwave
Theory Tech., vol. 40, p. 1681-1691, 1992.
[MON 71] de LA MONEDA F.H., CHENETTE E.R., VAN DER ZIEL A., “Noise in
phototransistors”, IEEE Transactions on Electron Devices, vol. ED-18, no. 6, p.
340-346, 1971.
[MOR 72] MORIIZUMI T., TAKAHASHI K., “Theoretical analysis of heterojunction
phototransistors”, IEEE Transactions on Electron Devices, vol. 19, no. 2, p. 152-
159, 1972.
[POU 94] POUVIL P., Composants semiconducteurs micro-ondes, Masson, 1994.
[ROM 96] ROMERO M., MARTINEZ M., HERCZFELD, “An analytical model for the
photodetection mechanisms in high-electron mobility transistors”, IEEE Trans. on
Microwave Theory Tech., vol. 44, no. 12, p. 2279-2287, 1996.
[SAW 99] SAWADA H., IMAI N., “An experimental study on a self-oscillating
optoelectronic up-converter that uses a heterojunction bipolar transistor”, IEEE
Trans. on Microwave Theory Tech., vol. 47, no. 8, p.1515-1521, 1999.
[SEE 90] SEEDS A. J., DE SALLES A.A.A., “Optical control of microwave devices”,
IEEE Trans. on Microwave Theory Tech., vol. 38, no. 5, p. 577-585, 1990.
[SHI 53] SHIVE J.N., “The properties of germanium phototransistors”, J. Opt. Soc.
Am., vol. 43, p. 239, 1953.
[SHO 51] SHOOCKLEY W., SPARKS M., TEAL G.K., “p-n junctions transistors”, Phys.
Rev., vol. 83, p. 151, 1951.
[SIM 86] SIMONS R., BHASIN K., “Analysis of optically controlled
microwave/millimeter- wave device structures”, IEEE Trans. on Microwave
Theory Tech., vol. MTT-34, no. 12, p. 1349-1355, 1986.
[SMI 80] SMITH R.D., PERSONICK S.D., Semiconductor Devices for Optical
Communications, Springer, Berlin, Heidelberg, p. 89, 1980.
[SUE 96] SUEMATSU E., IMAI N., “A fiber optic: millimeter-wave radio transmission
link using HBT as direct photodetector and an optoelectronic upconverter”, IEEE
Trans. on Microwave Theory Tech.,vol. 44, no. 1, p. 133-142, 1996.
[THU 99] THURET J., Phototransistor bipolaire à hétérojonction InP/InGaAs pour
conversion optique/bande millimétrique dans les réseaux de distribution hybride
radio sur fibre, Thesis, University of Paris VI, 1999.
154 Optoelectronic Sensors
[THU 99a] THURET J., GONZALEZ C., BENCHIMOL J.L., RIET M., BERDAGUER P.,
“High-speed InP/InGaAs heterojunction phototransistor for millimetre-wave fibre
radio communications”, Proc. of the Indium Phosphide and Related Materials,
IPRM’99, p. 389-392.
[WAN 86] WANG H., Photorécepteur monolithique intégrant un phototransistor et des
transistors bipolaires à heterojonction GaAlAs/GaAs pour transmission par fibre
optique, Thesis, University of Paris XI, 1986.
[WAN 86a] WANG H., ANKRI D., “Monolithic integrated photoreceiver implemented
with GaAs/GaAlAs heterojunction bipolar phototransistor and transistors”,
Electronics Letters, vol. 22, no. 7, p. 391-393, 1986.
Chapter 5
Metal-Semiconductor-Metal
Photodiodes
5.1. Introduction
Among the different semiconductor photodetectors, the metal-
semiconductor-metal (MSM) photodiode has a special place. It is a
planar photodiode based on simple technology, it is easy to integrate,
but it has not, up to now, gained the interest of the large
telecommunication firms. Currently, to the best of our knowledge,
only one device for millimeter applications has been commercialized.
However, we will see that this photodetector is not without benefits,
and indeed has very specific properties linked to its interdigitated
structure. In this discussion, we will first consider the operation and
structure of this photodiode before discussing its different
characteristics: its response coefficient, dynamic behavior and noise.
We will conclude with integration possibilities for the device.
Chapter written by Joseph HARARI and Vincent MAGNIN.
156 Optoelectronic Sensors
5.2. Operation and structure
5.2.1. Fundamentals
An MSM photodiode consists of two Schottky contacts on a
semiconductor. Because of this, with each contact having a current-
voltage characteristic I(V) similar to that of a traditional junction, the
MSM photodiode behaves like two diodes in series, facing in opposite
directions. Briefly, under bias, with one of the diodes necessarily
being reverse-biased, the dark current crossing the structure is weak
and, to the extent that the two Schottky contacts are identical, the
symmetry of the device gives rise to a symmetric global I(V)
characteristic, with each half (positive or negative bias) corresponding
to the behavior of a reverse-biased metal-semiconductor junction.
Figure 5.1. MSM photodetector with its typical interdigitated structure.
Inset: current lines in the structure
If the bias voltage is sufficiently high, the region between the
electrodes is completely depleted and this produces an electric field.
The basic principle of this photodetector is that light is absorbed in a
semiconductor layer forming part of the depletion region, such that the
electron-hole pairs generated are separated under the effects of the
electric field, with each carrier then being directly collected by an
electrode. This effect, familiar in photodiodes, takes place in a planar
structure whose geometry is shown in Figure 5.1. In general, the
electrodes have an interdigitated shape, which allows the carriers to be
collected over a large semiconductor area while maintaining a short
Metal-Semiconductor-Metal Photodiodes 157
inter-electrode distance. If, for example, illumination is from above,
this allows carriers to be collected over the whole surface of the light
beam emitted by a fiber in the near field. The inset in Figure 5.1
shows the current lines between several fingers of the electrode, and
shows that, due to the planar structure, the deeper a carrier is
photogenerated in the material, the longer it will take to reach the
electrode. This leads to a very specific distribution of transit times for
the carriers. As with all photodiodes, an MSM photodetector under
illumination can be characterized by its response coefficient, its cutoff
frequency and its photodetection noise. However, before we consider
these different aspects, we will discuss the behavior of the device in
the dark.
Figure 5.2. Behavior of an MSM photodetector in the dark: a) band structure
in thermodynamic equilibrium, b) band structure under bias,
c) corresponding electric field under bias
158 Optoelectronic Sensors
Figure 5.2 shows the evolution of the band structure of an MSM
structure under bias. The two electrodes are assumed to be of different
metals, so that the band structure of each metal-semiconductor
junction differs. On each side, I
n
is the height of the barrier seen by
the electrons in the metal, I
p
is the height for holes and V
b
is the built-
in bias of each junction.
In thermodynamic equilibrium (see Figure 5.2a), the two depletion
regions depend on the doping of the semiconductor and the work
function which separates the Fermi level in the metal from the vacuum
level in the semiconductor. Since the semiconductor is normally
weakly doped, the carriers mostly cross the potential barriers through
thermoelectric effects [SZE 96], in other words only those carriers
with sufficient energy cross the barrier. The sum of all the currents
across the barriers is zero. Under bias, as shown in Figure 5.2b, the
situation is more complicated, with one of the junctions (number 1 in
this case) being reverse-biased (voltage V
1
), and the other being
forward-biased (voltage V
2
). It is the current densities of electrons J
n1
and J
n2
, and of holes J
p1
and J
p2
which determine the behavior [SZE
71]. If we ignore recombination in the region between the electrodes,
we have:
1 2 1 2
and
n n p p
J J J J = = [5.1]
These relationships show that the forward bias voltage of junction
2 is such that the current flowing across it is equal to the current
allowed across junction 1 in reverse. Under these conditions, it is the
height of the barrier I
n1
which determines the value of J
n1
(and hence
J
n2
), and it is I
p2
which determines the value of J
p2
(and J
p1
). The
value of the total dark current thus depends on the heights of the
barriers, which are linked to the nature of the metal used for each
electrode. In parallel to this process, the depletion region of junction 1
expands, while that of junction 2 contracts slightly. Under these
conditions we can define two particular bias voltages which depend on
the structure of the MSM under consideration: the first is the total
depletion voltage V
TD
, which is reached when the depletion region of
the reverse-biased junction reaches that of the forward-biased junction,
that is, when:
Metal-Semiconductor-Metal Photodiodes 159
W
1
÷W
2
= s [5.2]
where s is the inter-electrode distance.
The second is the flat band voltage V
FB
, reached when W
1
= s. The
electric field present in the inter-electrode region is thus only zero at
electrode 2 (the anode), and it is at a maximum at the other side (at the
cathode). Clearly, an MSM photodetector is only useful when the bias
voltage is greater than V
FB
. If we assume a homogenous N-type
semiconductor, with doping level N
d
and permittivity H
s
, we have:
V
FB
=
q N
d
s
2 H
s
÷V
b1
÷V
b2
[5.3]
When the applied bias is increased, the electric field increases at
the cathode, which leads gradually to a breakdown process linked
either to shock ionization in the semiconductor material or to
tunneling through the potential barriers of the metal-semiconductor
barriers from band to band. All this leads to the typical current-voltage
characteristics displayed in Figure 5.3. For V < V
FB
, the dark current
crossing the photodiode is weak and grows slowly with the bias, until
it reaches a saturation value which is maintained until breakdown,
which can be more or less gradual with voltage. If the two electrodes
are made of the same metal, the I(V) characteristics in the dark are, in
principle, symmetric with respect to the V = 0 V axis. Generally, since
a photodetector must operate with a weak dark current, the best bias
conditions for a photodiode are:
– a bias voltage considerably below the breakdown voltage, in
order to limit the dark current (V < V
C
);
– an inter-electrode region that is completely depleted, with a
sufficiently high internal electric field so that the velocities of the
photogenerated charge carriers are in the saturation system (in any
case, V > V
FB
).
Of course, given the planar structure of the device, some elements
must be added to this general description.
160 Optoelectronic Sensors
Figure 5.3. Typical current-voltage characteristics of an
MSM photodetector in the dark
The biasing of the semiconductor, using electrodes placed above it,
leads to a particular electric field distribution, as shown in Figure 5.4.
Initially, the field decreases with depth in the device, and then there
are local maxima in the electric field just under the edge of each
electrode (or each finger for an interdigitated structure). Consequently,
the breakdown of the device always begins at the edge of the electrode,
and a very high electric field may exist close to the electrodes, but the
field is weaker in the middle of the inter-electrode space and deep in
the photodetector. Finally, as the field is weaker deeper in the
structure, the carriers that are photogenerated deep in the structure not
only have a great distance to travel to reach the electrode (as we have
already seen in Figure 5.1), but also have a small drift speed, which
increases the transit delay and decreases the dynamic performance of
the photodetector. All this confirms that the optimal bias voltage of
such a device must be selected with care in order to obtain good
dynamic performances without degradation.
Metal-Semiconductor-Metal Photodiodes 161
Figure 5.4. Distribution of the modulus of the electric field in an MSM photodetector.
The applied bias is 4 V, and the values of the field are in kV/cm.
The semiconductor is weakly doped (Nd = 5 u 10
14
cm
–3
)
5.2.2. Materials used
5.2.2.1. Electrode composition
Since the composition of the electrodes has a direct influence on
the band structure of the photodetector due to the magnitudes of I
n1
,
I
p1
, I
n2
and I
p2
, as shown in Figure 5.2, it will have an effect on the
dark current resulting from the thermoelectric effect. If we bias far
from the breakdown voltage, the phenomena limiting the dark current
are the thermoelectric passage of electrons across junction 1, and the
thermoelectric passage of holes across junction 2. The total dark
current is thus commonly written as:
J = A
n
*
T
2
exp ÷
q I
n1
÷'I
n1
( )
kT

l
l
l
l
l
÷ A
p
*
T
2
exp ÷
q I
p2
÷'I
p2
( )
kT

l
l
l
l
l
l

l
l
l
l
l
l
1÷exp ÷
qV
kT
(
(
·
·
·
·
)
)

(
(
·
·
·
·
)
)

[5.4]
162 Optoelectronic Sensors
where A
n
*
and A
p
*
are the Richardson constants associated with
electrons and holes, T is the absolute temperature, V is the applied
voltage, I
n1
and I
p2
are the heights of the barriers seen by electrons
and holes respectively, corrected with the terms introduced by the
effects of the electric field. This results in [SZE 96]:
'I
n1, p2
=
qE
1 2
4SH
s
[5.5]
where H
s
is the permittivity of the semiconductor and E
1,2
is the value
of the electric field at the metal-semiconductor junction. Greater
barrier heights thus minimize the dark current. If we assume that the
two electrodes are made of the same material, we have (see Figure
5.2):
I
n1
= E
g
÷I
p2
[5.6]
Simultaneously achieving maximum-height barriers for electrons
and holes requires us to choose the contact metal for which I
n
is as
close as possible to E
g
/2. Measurements of the dark current [ITO 86],
made for a gallium arsenide (GaAs) MSM photodetector for different
electrode metals, are presented in Table 5.1.
Dark current
Table 5.1. Dark current measured for GaAs MSM photodetectors (E
g
= 1.42 eV)
for different electrode metals and photodetectors with the same surface
Because of this, for this type of photodetector it is desirable to
decide on the cathode and anode at the fabrication stage, choosing a
metal with a high electron barrier I
n
for the first electrode and a high
I
p
barrier for the second. Under these conditions, the values of the
dark current are of the order of nano-amps, which are compatible with
all possible uses of the device.
Metal-Semiconductor-Metal Photodiodes 163
Clearly, all this assumes that no other effect, associated with
fabrication defects, surpasses the thermoelectric effect which is
normally the main cause of the dark current. This can occur if the
band structure is perturbed close to the metal-semiconductor junction.
In this case, a narrowing of the potential barrier favors the passage of
carriers through the tunnel effect, to the detriment of the
thermoelectric effect. The tunnel effect can be associated with the
presence of a native oxide layer at the interface, which leads to an
accumulation of electrons or holes; alternatively, the effect can be
assisted by trapping states present in the semiconductor material,
which create energy levels inside the bandgap. As a result, attention
must be paid to the epitaxy, the cleanliness of the semiconductor
surface before the metal deposition, and above all to the passivation of
the structure. In the case of a photodetector illuminated from above,
the dielectric layer used for this final operation can operate as an anti-
reflection coating.
5.2.2.2. Epitaxial structure
The epitaxial structure of an MSM is designed starting with the
absorbing layer, which itself depends on the wavelength to be
absorbed according to the basic law:
h c
O
E
g
[5.7]
where h is the Planck constant and c is the speed of light. The
different materials already reported in this application are given in
Table 5.2 along with the substrate they are grown on (an asterisk
indicates growth with a lattice mismatch) and the associated
wavelengths.
Table 5.2. Absorbing materials already used to fabricate MSM photodetectors,
along with their growth substrate and their associated wavelengths
164 Optoelectronic Sensors
As Figure 5.5 shows, around the absorption layer there are
normally two layers with a higher bandgap. The adaptation layer acts
to isolate the absorbing layer from the substrate during growth, and
the barrier layer acts to increase the potential barrier at the level of the
metal-semiconductor junction [BUR 91, WOH 97]. A barrier layer is
required, for example, when the absorbing material is N-type GaInAs,
on which the Schottky contacts are poor quality. Furthermore, to
improve the passage of carriers from the absorbing layer to the
electrodes, we also introduce one or more transition layers which
reduce the discrepancy in the band structure as seen by free carriers,
between the barrier layer and the absorbing layer [WAD 89, ZHA 96].
These very thin layers allow a gentle transition in the band structure
and prevent the accumulation of carriers in the upper part of the
absorbing layer under the electrodes. This accumulation leads to a
drop in the external quantum efficiency of the device, as a non-
negligible number of free carriers recombine while they are blocked at
the interface. It also simultaneously leads to an increase in the transit
time for charge carriers, which has an adverse effect on the frequency
response of the photodetector.
Figure 5.5. Typical epitaxial structure of an MSM photodetector.
The semiconductor is weakly doped (Nd = 5 u 10
14
cm
–3
)
The thickness of the absorbing layer depends mostly on its
absorption coefficient because, as we will see, this directly influences
the response coefficient of the photodetector when it is illuminated
from above or below. The thickness of the adaptation layer is
normally greater than one micron, while that of the barrier layer is as
Metal-Semiconductor-Metal Photodiodes 165
thin as possible (<0.2 μm in general). In some cases, for example, that
of a gallium arsenide MSM or a silicon MSM, the substrate and the
adaptation layer are also absorbing, but this requires the collection of
carriers photogenerated deep in the structure which, as we have
already seen, decreases the dynamic performance of the photodiode.
To avoid this problem, the adaptation layer of the GaAs MSM can be
made of non-absorbing AlGaAs matched to the lattice of the GaAs
substrate. In the case of silicon, SIMOX wafers are used, which have a
layer of silicon underneath.
5.3. Static and dynamic characteristics
5.3.1. Response coefficient
The response coefficient, which provides the photocurrent
generated per Watt of incident light (in units of A/W as a result),
initially depends on the mode of illumination. The photodetector can
be illuminated from above, across the interdigitated metallic structure,
from below or side-on, through an optical waveguide. The case of
lateral illumination must be treated on a case-by-case basis because
the response coefficient depends both on the optical coupling between
the waveguide and the detector, and between the fiber and the
waveguide. This requires a detailed study of the propagation of the
light in the multi-layer structure of the device. Illumination from
below is similar to illumination from above, apart from the absence of
electrodes to partially occlude the light entering the device; in addition,
the absorption associated with the substrate must be taken into account
if it is not thin and not perfectly transparent. The absorption
coefficient of the substrate is generally weak at the wavelengths under
consideration, but the thickness of the material to be crossed can be
sufficiently large to have an effect on the response coefficient.
Illumination from above is common. In this case, the response
coefficient is:
R=
q
hQ
K [5.8]
166 Optoelectronic Sensors
where q is the electronic charge, hQ the energy of a photon of the
incident light and K the quantum efficiency, which is itself given by:
K = 1÷R
( )

s
s ÷d
( )
1÷exp ÷D W
a
( )

l
l
l
[5.9]
where R is the reflection coefficient at the surface of the device, s is
the distance between two fingers, d the width of a finger, D the
absorption coefficient and W
a
the thickness of the absorbing layer.
The ratio s/(s + d) introduces the shadowing effect of the electrodes
which is typical of this photodetector. In brief, the response coefficient
is directly linked to the characteristics of the absorbing layer and the
geometry of the interdigitated electrodes.
As far as the characteristics of the absorbing layer are concerned,
in the case of a semi-infinite layer, the absorption coefficient must be
sufficiently high so that all the carriers are photogenerated in the
depletion region. This is the case for GaInAs at wavelengths of 1.3
and 1.55 μm, and for GaAs around 0.8 μm. In contrast, for silicon, the
absorption coefficient at 0.8 μm is of the order of 10 μm
–1
. This
reduces the quantum efficiency of the photodetector and causes a
reduction in the bandwidth, because at least some of the carriers
photogenerated far from the depletion region reach it through
diffusion, which is much slower here than the transport under the
effects of the electric field. This clearly depends on the inter-electrode
distance, as we will see. Finally, we can see that optimizing the
transport of the carriers across the planar structure implies a
relationship between the thickness W
a
of the absorbing region and the
inter-electrode distance s.
If we now consider the geometry of the electrodes, we find that
attempts to reduce the shadowing effect leads to two approaches. The
first consists of using opaque electrodes while reducing the finger
width relative to the inter-electrode distance. However, this increases
the resistance of the finger, which has a detrimental effect on the
bandwidth. Measurements made on electrodes deposited on silica
show that the effective resistance of a metallic finger is greater than
that which would be calculated based on the volume of the material.
Metal-Semiconductor-Metal Photodiodes 167
This effect is more important the narrower the finger, importance
being obtained at around 0.1 μm. This may be due to metallic
limitations that the structure places on electron transport becoming
important. In any case, for fingers made of titanium (150 Å) and gold
(350 Å), and of width 0.06 μm, the measured resistance is around 80
:/μm instead of 11 :/μm for d = 0.5 μm. This gives an idea of the
sizes to be considered when designing the electrodes. As a result, s
and d normally take similar values, within the range d ” s ” 4d. The
second approach consists of research into electrodes made of
transparent materials. Several possibilities exist, notably thin layers of
gold [MAT 96] and doped oxides: indium tin oxide (ITO) or cadmium
tin oxide (CTO). In each case, there is a tradeoff between the
transparency at the wavelength under consideration and the resistance
of the material. For example, doped ITO is remarkably transparent at
short wavelengths (O < 0.85 μm), but it requires a compromise
between resistance and transparency at longer wavelengths as doping
reduces its transparency [GAO 94, SEO 93]. The same is true for CTO.
In any case, an optical transmission of 100% cannot be achieved at
long wavelengths. Furthermore, attention must be paid to the
conditions of the deposition, which for the oxides is normally through
sputtering, which does not favor particularly good characteristics of
the electrode-semiconductor interface. In short, while the response
coefficient is an absolute priority, the deposition of transparent
electrodes clearly reduces the shadowing effect, but it is not without
consequences in terms of the dynamic behavior of the photodetector.
Firstly, it leads to an increase in the finger resistance, the
consequences of which can be seen in the equivalent circuit of the
MSM in the dynamic system. Also, the electric field diagram (Figure
5.4) shows that this is weak in the medium under each finger. The
drift velocity of the photocarriers will therefore be weak at these
points. As a result, the electrons photogenerated just under the
transparent cathode, and the holes photogenerated under the anode,
will take a long time to reach the opposite electrode. Not only is their
drift velocity low as they travel through areas of weak field, but they
also have a long way to travel (see Figure 5.1). Thus, transparent
electrodes increase the mean transit time and decrease the bandwidth.
A similar effect on the mean transit time is seen when the device is
168 Optoelectronic Sensors
illuminated from below. In this case, the majority of the carriers are
photogenerated in the lower part of the absorbing layer, as much
underneath the fingers as between them. Since this is a region of weak
electric field, this also leads to a longer mean transit time.
Still on the subject of the electrode geometry, a property that
cannot be ignored in the case of illumination from above is the
transparency of the network of fingers when its period p (p = d + s)
becomes smaller than the wavelength of light. Then we observe
typical optical behavior which is more complicated than a simple
shadowing effect. The few studies carried out on this subject [KUT
94] show that the transmission of light through the metallic network
depends on the ratios of O/p and O/d, on the finger thickness, on the
refractive index n
s
of the semiconductor, and on whether the optical
polarization is parallel or perpendicular to the direction of the fingers.
It is higher in the perpendicular case than in the parallel case, where
only a few percent of the optical energy reaches the semiconductor.
Furthermore, it is a minimum for O/p = 1 and O/p = n
s
. The finger
thickness becomes relevant when it is close to its width. All these
complex properties, which must be modeled on a case-by-case basis,
are sharp enough to make it possible to built a wavelength
discriminator using two parallel MSMs with slightly different periods,
illuminated from the same beam and measuring the ratio of the
photocurrents [CHE 97].
Figure 5.6. Example of the epitaxial structure of an MSM photodetector
on top of an optical waveguide
Metal-Semiconductor-Metal Photodiodes 169
In order to understand all the modes of illumination, we will now
consider the case of lateral illumination. The electrode structure of the
MSM is particularly favorable to the injection of light through an
optical waveguide [SOO 88, VIN 89]. Indeed, the direct lateral
illumination, for which absorption occurs in the first few microns after
the cut face, is difficult to achieve, as is the case for a PIN photodiode.
The example in Figure 5.6, of an MSM structure on top of a
waveguide, is based on an N-type InP/GaInAsP/InP optical waveguide,
the standard for this type of technology. The quaternary GaInAsP
layer (with bandgap E
g
= 1.05 eV), formed into a lattice on InP, is
what forms the guiding layer, with InP acting as the cladding.
Coupling with the photodetector occurs through evanescent waves.
Modeling of the optical propagation through such a structure (see
Figure 5.7) shows how the injected beam couples with the absorbing
layer through the intermediary of the thin quaternary adaptation layer
whose characteristics (thickness and composition) determine the
absorbed wavelength. Of course, the photocarriers are still generated
in the lower part of the absorbing layer, but the thickness of this layer
can be selected to be quite thin without significantly reducing the
quantum efficiency. For a fixed adaptation layer, the effect of the
thickness of GaInAs on the internal quantum efficiency (see Figure
5.8) shows that, for TE and TM optical polarizations at a wavelength
of 1.55 μm, a certain number of optimum cases exist, even for small
thicknesses.
At a conceptual level, to achieve the shortest possible absorption
length, it is necessary to carefully choose both the thickness of the
adaptation layer and that of the absorbing layer. The response
coefficient is:
R= 1÷R
( )
K
q
hQ
K
internal
[5.10]
where K
internal
is the internal quantum efficiency determined by the
epitaxial structure, R is the reflection coefficient at the entrance of the
waveguide and K is the coupling coefficient between the beam leaving
the optical fiber and the semiconductor waveguide. The internal
quantum efficiency can only be calculated through detailed modeling
of the optical transport in the material structure of the
170 Optoelectronic Sensors
photodetector/waveguide. The dynamic behavior of such a structure is
comparable to that seen with illumination from below; however, the
thin width of the absorbing layer allows a short transit time to be
retained. Thus, this type of structure allows a high response coefficient
to be combined with a small absorber thickness.
Figure 5.7. Propagation and absorption of light in a MSM
on a waveguide with evanescent coupling
Figure 5.8. Evolution of the quantum efficiency of the detector/waveguide
as a function of the thickness of the absorbing material
Metal-Semiconductor-Metal Photodiodes 171
To conclude, we will discuss all the internal electronic effects in
the device which affect the response coefficient. Typically, this is
reduced under the effect of recombination and increases with internal
gain. Recombination occurs either at the surface of the semiconductor
between to fingers or at the interface between two epitaxial layers,
wherever there is a potential barrier associated with the transition
between different materials, which blocks the transport of carriers and
leads to their accumulation. In the first case, careful passivation of the
surface of the photodetector eliminates the problem. The second case
is relevant when a barrier layer is introduced to improve the quality of
the contact with the electrode. A transition layer between the absorber
and the barrier is thus required, made of an epitaxial layer with
variable bandgap or even a super-lattice [WAD 89, ZHA 96].
There are many effects which lead to gain [KLI 94]. They mostly
consist of two types: those connected with the trapping of carriers,
which operate at low frequencies (below a few tens of megahertz), and
shock ionization close to the edges of the fingers, whose cutoff
frequency is higher, as can be seen in the case of avalanche
photodiodes. In the case of trapping, the gain effect is due to a
photocarrier becoming trapped in the semiconductor for a time W
p
greater than the inter-electrode transit time W
T
which triggers the
injection of additional carriers in order to maintain electrical neutrality.
As long as this carrier remains trapped, a carrier with the opposite
charge will cross the device to provide electrical compensation for it.
The gain is then given by:
G
trapping
=
W
p
W
T
[5.11]
This trapping can occur at the surface of the semiconductor, as in
the case of GaAs photoconductors, or at the interface between two
epitaxial layers. When it occurs at the surface, the holes are mostly
trapped close to the cathode and electrons trapped close to the anode,
this creates an asymmetric charge distribution. Hetero-interface
trapping is associated with epitaxial growth defects, and normally a
barrier layer is used to eliminate surface effects. Generally, attempts
are made to limit both recombination and trapping in order to obtain a
172 Optoelectronic Sensors
photodetector with constant dynamic behavior up to the high cutoff
frequency, while ensuring a good reliability.
5.3.2. Dynamic behavior
As with all photodetectors, the dynamic behavior of MSMs is
determined by their capacitance and their input and output resistances,
in other words their equivalent electric circuit under modulated
electric power, and the transit time of the photocarriers. The
equivalent small-signal circuit (see Figure 5.9) shows:
– the intrinsic capacitance C
PD
and resistance R
PD
of the device;
– the finger resistance R
F
, the capacitance and the resistance of the
contact, C
CONTACT
and R
CONTACT
;
– the connection wires with inductance L
F
and capacitance C
W
.
The finger resistance is given by:
R
F
=2 R
0

L
N
[5.12]
where R
0
is the resistance per unit length of each finger, L the length
and N the number of fingers. In the equivalent circuit, R
L
is the load
resistance and we see how the finger resistance can, in some cases,
increase the time constant RC of the device. One of the important
terms is the intrinsic capacitance of the MSM. A standard analysis of
the planar structure [LIM 68] shows that:
C
PD
=
A
d ÷s
( )
H
0
1÷H
R
( )

K k
( )
K k '
( )
[5.13]
where A is the surface area of the active region, H
0
and H
R
are the
vacuum and relative permittivities, and finally:
K k
( )
=
dM
1÷k
2
sin
2
M
( )
0
S 2
¹
with k = tan
2
S.d
4 d ÷s
( )
(
(
·
·
·
·
·
)
)

and k ' = 1÷k
2
Metal-Semiconductor-Metal Photodiodes 173
The calculation shows two details connected to the interdigitated
planar structure of the MSM. Firstly, the capacitance of the device
falls when the ratio s/d increases; the decrease in the shadow effect is
thus inextricably linked to the capacitance of the photodetector.
Secondly, for a photosensitive surface and an identical mean transit
time, the capacitance of the MSM is 3 to 4 times weaker than that of a
PIN photodiode. This aspect favors the MSM for high modulation
frequencies (millimeter wavelengths), where the size of the devices is
very small.
Typical values for the equivalent circuit can be given as a guide.
The capacitance of the MSM lies between 10 and 1000 fF, its
resistance between 1 and 100 :, and the capacitance of the contact is
a significant parasitic element, generally less than 100 fF (and in all
cases less than C
PD
). Finally, the connection wires have an auto-
inductance of the order of 10 pH and a capacitance of a few fF linking
them to the body of the device. If we ignore all the parasitic elements,
that is to say the connection wire, the capacitance of the contact, and
the finger resistance, we find a standard time constant: R
L
C
PD
. The
cutoff frequency associated with the capacitance effects is thus given
by:
f
CAPA
=
1
2S R
L
C
PD
[5.14]
Now turning to the transit time, a thin, two-dimensional physical
model is required to account for the effects of the geometry of the
electrode, the thickness of the absorbing layer, of the hetero-interfaces
to be crossed, and of the electric field distribution produced by the
bias voltage [ASH 95].
Nevertheless, a number of rules allow us to evaluate the cutoff
linked to the transit time of the carriers. Firstly, the presence of high
electric field areas just at the edges of a finger mean that it is difficult
to appropriately bias a device with a large inter-electrode distance (s >
3 μm). What happens in this case is that a local electric field is
reached which is above the breakdown voltage, while the field in the
middle of the inter-electrode space is still too weak to give the
photocarriers their saturation velocity. This effect, visible in Figure
174 Optoelectronic Sensors
5.4, is more serious for holes, as they move more slowly than
electrons. Even for small inter-electrode distances, the variation of the
field in the device must be taken into account. The bias voltage is thus
an important parameter that must be adjusted to limit the mean transit
time while avoiding breakdown.
Figure 5.9. Equivalent circuit for an MSM photodetector in the case of a small signal
Secondly, the distribution of the current lines in the device favors a
certain ratio between the thickness of the absorbing layer and the
inter-electrode distance. The effect of transit time is smallest when
this value is close to 2 (W
a
/s = 2). If this is achieved, the mean transit
time in the structure is:
W =G
s
2 v
sat
[5.15]
where v
sat
is the saturation velocity of the carriers (we will take the
same velocity for holes as for electrons), s is the inter-electrode
distance and G accounts for the effects of the planar structure (1 ” G ”
2). The saturation velocity used is taken as equal to that of the holes
for the most pessimistic results. The cutoff frequency associated with
the transit time is thus given by:
f
T
=
1
2S W
[5.16]
The cutoff frequency of the MSM photodiode under these
conditions is:
Metal-Semiconductor-Metal Photodiodes 175
f =
1
f
CAPA
2
÷
1
f
T
2
(
(
·
·
·
·
·
)
)

1 2
[5.17]
An increase in the inter-electrode distance will decrease the
capacitance and increase the transit time. For every structure, then,
there is a compromise to be made.
5.3.3. Noise
The different sources of noise in MSMs, as in other photodetectors,
are thermal noise, 1/f
D
noise and shot noise. Thermal noise is linked to
the variations in energy of the carriers in resistive regions. It is
independent of frequency and its spectral density is linked to the
absolute temperature T and the resistance R by:
i
TH
2
=
4kT
R
in A
2
Hz
( )
[5.18]
The 1/f
D
noise is associated with all the random generation and
recombination effects in the semiconductor material. These effects are
themselves linked to the different imperfections, in terms of
irregularities in the structure of the crystal lattice or in terms of foreign
atoms. This generates a specific noise whose spectral power density
varies as 1/f
D
, where D is normally close to 1. This spectral density is
often higher in horizontal transport devices, such as MSMs or field
effect transistors. Finally, the shot noise, familiar from PIN and
avalanche photodiodes, is linked to the photogeneration process and
the charge transport. Its spectral power density is:
i
SHOT
2
= 2 q I
DARK
÷I
PH
( )
in A
2
Hz
( )
[5.19]
In principle, this is white noise, i.e. its spectral density is
independent of frequency; however, in practice it follows the
frequency response curve of the device, making it possible to obtain
the frequency response of a photodetector using a continuous signal,
by measuring the variation of its shot noise as a function of frequency.
176 Optoelectronic Sensors
Having covered these items, we will now describe the variation of
the spectral noise density of the MSM photodetector in two situations:
in normal operating conditions and close to breakdown in the
conditions where a gain in the photocurrent, linked to the shock
ionization effect, is possible.
In general, if the finger and contact resistances are sufficiently low,
the thermal noise is much weaker than the shot noise, so that we can
ignore its effects. Under normal operating conditions [SOO 91, WAD
88], the 1/f
D
noise dominates for frequencies of the order of 1 to
10 MHz, after which it is the shot noise which dominates and which
must be considered when performing calculations for photoreceivers.
This shot noise increases with the dark current, which is why a low
dark current is desirable, normally below about 100 nano-amps.
However, as breakdown is approached [SCH 90], two changes are
observed. Firstly, the 1/f
D
noise increases, so much so that it
dominates for frequencies up to the order of 100 MHz; secondly, the
shot noise also increases, in association with the multiplication
process connected with shock ionization. The spectral density of the
noise becomes:
i
SHOT
2
=2 q I
DARK
÷I
PH0
( )
G
2
F in A
2
Hz
( )
where G is the gain, I
PH0
the primary photocurrent and F the noise
factor, which can be expressed as a function of the gain G, as is done
in the case of avalanche photodiodes: F = G
x
. If, as is often the case,
the part linked to the dark current is negligible compared to the part
linked to the photocurrent, we have:
i
SHOT
2
=2 q I
PH0
G
2 x
in A
2
Hz
( )
However, values of x measured up to now for MSMs are somewhat
higher: between 1.5 and 3, which is not desirable. Due to these noise
characteristics, MSM photodetectors are not used close to breakdown.
Metal-Semiconductor-Metal Photodiodes 177
5.4. Integration possibilities and conclusion
The functionalities offered by MSM photodiodes are considerable.
It is with these photodetectors that the highest current cutoff
frequencies, above terahertz levels, have been achieved [CHO 92],
while maintaining an acceptable response coefficient. Of course, the
inter-electrode distance required for such a performance is small
(25 nm) and requires the use of an electronic mask, but this shows the
interest in a device whose main advantages are:
– a planar configuration of the electrodes allowing, in particular,
the integration of the photodiode into a coplanar microwave
transmission line;
– weakly doped active layers which remain consistent with
technologies based on semi-insulating substrates;
– a capacitance significantly below that of all other photodetectors
with equivalent transit time characteristics;
– electrode technologies allowing the dark current to be
minimized;
– a photodetection noise equal to that of a PIN photodiode.
These properties allow the MSM to be integrated without difficulty
into all integrated circuit technologies based on III-V materials,
whether on GaAs [BUR 91, HAR 88, ITO 84] or on InP [HOR 96],
with MESFET or HEMT transistors. In general, the weakly-doped
epitaxial structure of the detector is grown before the various layers
required to make the transistors and the metallic deposition of the
interdigitated electrode is the same as that of the transistor gates.
In conclusion, the optical properties of thin metallic lattices, which
are still poorly-understood, offer interesting and original characteristics,
even if the power limitations are more restrictive than those for PIN
photodiodes.
178 Optoelectronic Sensors
5.5. Bibliography
[ASH 95] ASHOUR I.S., HARARI J., VILCOT J.P., DECOSTER D., “High optical power
nonlinear dynamic response of AlInAs/GaInAs MSM photodiode”, IEEE Trans.
on Electron Devices, vol. 42, no. 5, p. 828-834, May 1995.
[BUR 91] BURROUGHES J.H., “H-Mesfet compatible GaAs/AlGaAs MSM
photodetector”, IEEE Photonics Technology Letters, vol. 3, no. 7, p. 660-662,
July 1991.
[CHE 97] CHEN E., CHOU S.Y., “A wavelength detector using monolithically integrated
subwavelength metal-semiconductor-metal photodetectors”, Proceedings SPIE, vol.
3006, p. 61-67, 1997.
[CHO 92] CHOU S.Y., LIU Y., FISCHER P.B., “Tera-hertz GaAs metal-semiconductor-
metal photodetectors with 25 nm finger spacing and width”, Applied Physics
Letters, vol. 61, no. 4, p. 477-479, July 1992.
[GAO 94] GAO W., KHAN A., BURGER P.R., HUNSPERGER R.G., ZYDZIK G., O’BRIAN
H.M., SIVCO D., CHO A.Y., “InGaAs metal-semiconductor-metal photodiodes
with transparent cadmium tin oxide Schottky contacts”, Applied Physics Letters,
vol. 65, no.15, p. 1930-1932, October 1994.
[HAR 88] HARDER C.S., VAN ZEGHBROECK B.J., MEIER H., PATRICK W., VETTIGER P.,
“5.2 GHz bandwidth monolithic GaAs optoelectronic receiver”, IEEE Electron
Device Letters, vol. 9, no. 4, p. 171-173, April 1988.
[HOR 96] HORSTMANN M., SCHIMPF K., MARSO M., FOX A., KORDOS P., “16 GHz
bandwidth MSM photodetector and 45/85GHz ft/fmax HEMT prepared on an
identical InGaAs/InP layer structure”, Electronics Letters, vol. 32, no. 8, p. 763-
764, April 1996.
[ITO 84] ITO M., WADA. O., NAKAI K., SAKURAI T., “Monolithic integration of a
metal-semiconductor-metal photodiode and a GaAs preamplifier”, IEEE Electron
Device Letters, vol. 5, no. 12, p. 531-532, December 1984.
[ITO 86] ITO M., WADA O., “Low dark current GaAs metal-semiconductor-metal
photodiodes using Wsix contacts”, IEEE Journal of Quantum Electronics, vol. 22,
no. 7, p. 1073-1077, July 1986.
[KLI 94] KLINGENSTEIN M., KUHL J., ROSENZWEIG J., MOGLESTUE C., HÜLSMANN A.,
SCHNEIDER J., KOHLER K., “Photocurrent gain mechanisms in metal-
semiconductor-metal photodetectors”, Solid State Electronics, vol. 37, no. 2, p.
333-340, 1994.
[KUT 94] KUTA J.J., VAN DRIEL H.M., LANDHEER D., ADAMS J.A., “Polarization and
wavelength dependence of metal-semiconductor-metal photodetector response”,
Applied Physics Letters, vol. 64, no. 2, p. 140-142, January 1994.
Metal-Semiconductor-Metal Photodiodes 179
[LIM 68] LIMY.C., MOORE R.A., “Properties of alternately charged coplanar parallel
strips by conformal mapping”, IEEE Transactions on Electron Devices, vol. 15,
p. 173-180, 1968.
[MAT 96] MATIN M.A., SONG K.C., ROBINSON B.J., SIMMONS J.G., THOMPSON D.A,
GOUIN F., “Very low dark current InGaP/GaAs MSM Photodetector using semi
transparent and opaque contacts”, Electronics Letters, vol. 32, no. 8, p. 766-767,
April 1996.
[SCH 90] SCHUMACHER H., SOOLE J.B.D., LEBLANC H.P., BHAT R., KOZA M.A.,
“Noise behaviour of InAlAs/GaInAs MSM photodetectors”, Electronics Letters,
vol. 26, no. 9, p. 612-613, April 1990.
[SEO 93] SEO J.W., CANEAU C., BHAT R., ADESIDA I., “Application of indium-tin-
oxide with improved transmittance at 1.3μm for MSM photodetectors”, IEEE
Photonics Technology Letters, vol. 5, no. 11, p. 1313-1315, November 1993.
[SOO 88] SOOLE J.B.D., SCHUMACHER H., ESAGUI R., KOZA M.A., BHAT R.,
“Waveguide integrated MSM photodetector on InP”, Electronics Letters, vol. 24,
no. 24, p. 1478-1480, April 1988.
[SOO 91] SOOLE J.B.D., SCHUMACHER H., “InGaAs metal-semiconductor-metal
photodetectors for long wavelength optical communications”, IEEE Journal of
Quantum Electronics, vol. 27, no. 3, p. 737-752, March 1997.
[SZE 71] SZE S.M., COLEMAN D.J., LOYA J.R.A., “Current transport in metal-
semiconductor-metal structure”, Solid State Electronics, Pergamon Press, New
York, vol. 14, p. 1209-1218, 1971.
[SZE 96] SZE S.M., Physics of Semiconductor Devices, John Wiley & Sons Ltd., New
York, 1996.
[VIN 89] VINCHANT J.F., VILCOT J.P., LORIAUX J.L., DECOSTER D., “Monolithic
integration of a thin and short metal-semiconductor-metal photodetector with a
GaAlAs optical inverted rib waveguide on a GaAs semi insulating substrate”,
Applied Physics Letters, vol. 55, no. 19, p. 1966-1968, May 1989.
[WAD 88] WADA O., HAMAGUCHI H., LE BELLER L., BOISROBERT C.Y., “Noise
characteristics of GaAs metal-semiconductor-metal photodiodes”, Electronics
Letters, vol. 24, no. 25, p. 1574-1575, December 1988.
[WAD 89] WADA O., NOBUHARA H., HAMAGUCHI H., MIKAWA T., TACKEUCHI A.,
FUJII T., “Very high speed GaInAs metal-semiconductor-metal photodiode
incorporating an AlInAs/GaInAs graded superlattice”, Applied Physics Letters,
vol. 54, no. 1, p. 16-17, January 1989.
[WOH 97] WOHLMUTH W., ARAFA M., MAHAJAN A., FAY P., ABESIDA I.,
“Engineering the Schottky barrier heights in InGaAs metal-semiconductor-metal
photodetectors”, Proceedings SPIE, vol. 3006, p. 52-60, 1997.
180 Optoelectronic Sensors
[ZHA 96] ZHANG Y.G., LI A.Z., CHEN J.X., “Improved performance of InAlAs-
InGaAs-InP MSM photodetectors with graded superlattice structure grown by gas
source MBE”, IEEE Photonics Technology Letters, vol. 8, no. 6, p. 830-832, June
1996.
Chapter 6
Ultraviolet Photodetectors
6.1. Introduction
The first mention of the existence of ultraviolet radiation was in
1801 when J.W. Ritter observed that certain types of chemical
reactions were accelerated by a radiation which was not visible to the
eye and whose wavelength was shorter than that of violet light. It was
quickly shown that this particular radiation was subject to the laws of
interference just like visible light (T. Young, 1804). Thus this
observation, along with various others, made it possible to establish at
the start of the 19
th
century that visible and ultraviolet radiation
exhibited the same electromagnetic properties, and differed only in
their wavelength.
We know today that ultraviolet radiation (UV) covers a wide
spectral range which extends from the limit of the visible, around
400 nm (3.1 eV) up to the X-ray boundary at 10 nm (124 eV) (see
Figure 6.1), covering a total spectral range which in terms of energy is
about 80 times as large as that of the visible spectrum. It is
traditionally divided into four categories:
Chapter written by Franck OMNES and Eva MONROY.
182 Optoelectronic Sensors
– UVA, covering wavelengths between 400 and 320 nm (from 3.1
to 3.87 eV);
– UVB, whose wavelengths range from 320 to 280 nm (from 3.87
to 4.43 eV);
– UVC, for wavelengths between 280 and 200 nm (from 4.43 to
6.20 eV);
– the far UV, for wavelengths between 200 and 10 nm (form 6.2 to
124 eV).
Due to its own radiation absorption properties, the terrestrial
atmosphere does not allow the free propagation of light with
wavelengths below 200 nm: as a result, the spectral end of the UV
between 200 and 10 nm cannot really propagate except in a vacuum,
and because of this it is often known as “vacuum UV” (we will use the
term VUV for the rest of the chapter when referring to this particular
spectral range).
Figure 6.1. Spectral ranges of the main types of electromagnetic radiation
The fields of application for UV detectors, of which there is a great
variety, mostly consist of:
– biological and chemical sensors: detection of ozone, pollutants
and biological agents;
– fire detection: fire alarms, detection of missile exhausts, flame
control;
– optical inter-satellite spatial communications: secure transmission
of data in space;
Ultraviolet Photodetectors 183
– the calibration of emitters and UV imaging: instrumentation,
measurement of solar UV, astronomical research.
In general, photodetectors suited to these areas of application must
simultaneously possess a high response coefficient, a good linearity of
the photocurrent as a function of the incident optical power, a large
bandgap energy (in the case of semiconductors), a low level of noise
and a high degree of spectral selectivity. The level of importance and
the priority ascribed to these different requirements clearly depends on
the type of application intended. A high response speed can be
particularly useful in applications requiring rapid real-time processing
of the signal, such as UV imaging.
The sun represents the most important natural source of UV
radiation. It is therefore not surprising that most of the applications of
UV detectors are connected with the measurement of solar radiation.
The solar radiation spectrum is shown in Figure 6.2.
,
,
,
,
,
Figure 6.2. Solar radiation spectrum. Inset: UVA and UVB solar irradiance,
outside the Earth’s atmosphere and at ground level
184 Optoelectronic Sensors
For the vast majority of applications, it is helpful if the UV
photodetectors are not sensitive to visible and infrared radiation. They
basically operate as high-pass filters, and they are classed as visible blind
detectors for cutoff wavelengths between 400 and 280 nm, or solar blind,
assuming that the UVC radiation is blocked by atmospheric absorption
for cutoff wavelengths typically less than or equal to 280 nm. The visible
blind detectors are nowadays quite widely used for flame detection
1
, and
have potentially very promising growth possibilities for applications in
space to optical inter-satellite telecommunications, for UV wavelengths
below 280 nm which clearly cannot be detected at ground level. Outside
the Earth’s atmosphere, UV light represents 9% of the total luminous
power emitted by the Sun. The stratospheric ozone layer completely
absorbs the UVC part of the solar spectrum, and only allows UVA and
UVB to be transmitted to the ground after attenuation (see the inset in
Figure 6.2). It is generally considered that solar radiation of wavelengths
below 280 nm does not reach the surface of the Earth. UVA and B
radiation, on the other hand, are mostly responsible for the biological
effects of solar radiation, since they reach the ground (see Figure 6.3).
Damage to
human DNA
Figure 6.3. Spectral dependence of the harmful effects of
UVA and UVB radiation on human health
1. The applications of flame detection mostly cover the following areas: fire detector
heads, forest fire detection systems, missile track detection systems, etc. Gas flames,
for example, have an emission peak in the UV around a wavelength of 320 nm.
Ultraviolet Photodetectors 185
UVB radiation, being particularly energetic, is the direct cause of
sunburn and skin cancers, and it has also been shown more recently
that UVA radiation, although considered less dangerous, is able to
penetrate the skin and has enough ionizing power to cause deep
damage to living cells
2
. It has been found that UVA radiation can in
particular alter the genetic structure of cells and, thus, also produce a
carcinogenic effect comparable to that of UVB, particularly in the
case of prolonged or repeated exposure. This represents a clear threat
to human heath from UV radiation, but also allows it to play an
extremely useful role in a number of different fields. Among others,
UV radiations are involved in the synthesis of Vitamin D in the human
body, which means that a moderate amount of exposure to sunlight is
important for everybody, and the ionizing properties of these
radiations make them useful for the activation of numerous chemical
reactions, both organic and inorganic. Also, their bactericidal action
makes them useful for many types of biological cleaning and
sterilization procedures
3
. Finally, the ozone layer present in the upper
atmosphere is naturally regenerated through the conversion of oxygen
in the air into ozone, a reaction which is photo-assisted by UV
radiation.
On the other hand, the rapid depletion of the stratospheric ozone
layer, caused in particular by human activity
4
, and the resultant
increase in exposure of the Earth’s surface to UV radiation has raised
considerable concern among scientists and professionals interested in
health and the environment. This has recently driven a significant
2. This issue is particularly interesting since it has public health implications because
of the popularity of tanning salons with the general public.
3. The illumination of objects with high intensity UVB and C radiation can be used
for this, particularly in techniques for the preservation of perishable foodstuffs, the
sterilization of surgical instruments, etc. Housewives have also known since time
immemorial that the drying of washing in the sun in fair weather makes it healthier,
etc.
4. The emission of aerosols and gases containing fluoro-, chloro- and bromo-carbons
is largely responsible for this. Large-scale natural phenomena such as volcanic
eruptions also contribute in a manner which can be very significant. One of the most
significant “holes” in the ozone layer is currently situated above the continent of
Antarctica, but the effect is very widespread and nowadays has a global effect.
186 Optoelectronic Sensors
increase in efforts to develop many specialized optoelectronic devices
which in particular allow the detection and measurement of solar UV
illumination. Thus, visible blind detectors have recently found a
growing number of applications in UVA and B monitoring systems
for the environmental study of ground-level solar illumination and UV
dosimetry in fields which are mostly connected with biology, medical
science or research into cosmetic products (the development of sun
creams or other products which modify the photosensitivity of human
skin to UV radiation, etc.). Personal solar dosimeters able to detect
direct or diffuse UV light are also currently available to the general
public for UVB dosimetry. The spectral response curve of these
sensors reproduces as closely as possible (as we will see later) the
erythema (sunburn) action spectrum on human skin.
There are a wide range of different UV photodetectors suitable for
all the different applications listed above. Traditionally,
photomultiplier tubes have always been used for UV detection and
their use is still widespread, particularly in laboratories. However,
they generally involve detection systems which are cumbersome,
expensive and fragile, although they offer a sensitivity which is
difficult to match with the other types of photodetectors applicable to
these ranges of wavelength. The increasing need for miniaturization
and reliability of these systems for the development of portable or
remote applications has naturally led to the development of UV
photodetectors based on semiconductors, whose active components
can be photoconductors, Schottky junction photodiodes, p-i-n
photodiodes, MSM photodiodes, phototransistors or avalanche
photodiodes.
The use of semiconductors for UV photodetection historically
began with the use of narrow bandgap materials such as silicon and
certain III-V materials (GaP, GaAsP, etc.) whose bandgaps are mostly
situated in the near-infrared (Si) or the red (GaAsP). In exceptional
cases they may be in the yellow, as in the specific case of GaP, which
represents their upper limit. The main drawback of silicon, and small
bandgap semiconductors in general, is that it is impossible to use them
directly for UV photodetection. Firstly, the direct exposure of such
Ultraviolet Photodetectors 187
sensors to daylight clearly results in them being flooded with visible
wavelengths and sometimes even infrared, so that it is simply
impossible to extract information pertaining to the UV spectral
component from this white light. Also, the quantum efficiency of a
semiconductor photodetector (defined as the probability of electron-
hole creation by the absorption of an incident photon) is greatest for
energies above the bandgap of the material, but limited to a spectral
range relatively close to that value. For energies much higher than the
bandgap of materials, such as those that we are interested in for UV
light, part of the energy of the incident light is effectively lost through
heating of the semiconductor (excitation of vibration modes, or
phonons in the crystal lattice), so that the quantum efficiency of the
photodetector decreases in the short wavelength part of the optical
spectrum. The use of semiconductor materials with a small bandgap
energy for UV photodetection therefore rapidly drove the
development of filtering devices, which normally use layers of
selectively-absorbing material, and sometimes a phosphor-based high-
pass filter which absorbs the UV light and re-emits towards the
semiconductor photodetector a light whose spectral maximum is not
far from the bandgap of the semiconductor material being used [GOL
99]. It still needs to be joined on to detection systems that are
relatively complex and also expensive, whose behavior is also
increasingly liable to drift in the long term, due to aging of the filters,
resulting in a certain degree of performance loss.
A number of years ago, a new generation of UV photodetectors
appeared, made from semiconductor materials with large bandgaps
such as silicon carbide (SiC), diamond or gallium nitride (GaN) and
AlGaN alloys. Silicon carbide, whose bandgap at ambient temperature
(T = 300 K) is 2.86 eV (6H-SiC), has allowed sensitive detection of
the spectral range in question. In particular, it has recently given rise
to industrial applications such as the construction of flame detectors
that are visible-blind, operating in the UVC range. The only issue is
that even here the use of high-pass optical filters is standard in order
to tune the photodetection system to the desired spectral window,
meaning that this only offers a partial solution to the problem.
188 Optoelectronic Sensors
Thanks to the development of epitaxial crystal growth techniques,
the recent use of gallium nitride (GaN), and still more recently
aluminum gallium nitride (Al
x
Ga
1-x
N, with x being between 0 and 1),
has helped reduce the problems listed earlier, enabling the fabrication
of high performance and robust visible blind photodetectors. The
direct bandgap of Al
x
Ga
1–x
N materials, whose size is an increasing
function of the concentration of aluminum, is found at ambient
temperature to extend from 3.42 eV (O = 362 nm) (x = 0) to 62 eV (O
= 200 nm) (x = 1). Thus, the spectral range that is covered includes in
particular the cutoff values for UVB (3.87 eV or O = 320 nm) and
UVC (4.43 eV or O = 280 nm). It follows that it becomes possible to
fabricate UV photodetectors using these materials which are free from
any need for intermediate spectral filtering, opening up the possibility
of simplifying photodetection systems through miniaturization,
accompanied by an improved reliability of the devices. In addition, the
high chemical stability of GaN and Al
x
Ga
1–x
N compounds,
comparable to that of SiC, means that these materials are also suitable
for use in the fabrication of photodetectors which are naturally
resistant to extreme operating conditions (high temperatures and high
values of inverse bias voltage in particular). Due to their direct
bandgap at all levels of aluminum, the optoelectronic properties of
Al
x
Ga
1–x
N alloys are also clearly superior to those of SiC, whose
indirect bandgap does not so strongly favor the effects associated with
photo-generation of carriers. For all these reasons, group III nitrides
therefore offer a technical solution for the future, as we will see
throughout the rest of this chapter, a solution which is at the same time
flexible, robust and very well-suited for all visible-blind or solar-blind
photodetection applications, because they allow relatively low-cost
fabrication of effective and reliable UV photodetectors suitable for all
the application windows listed above.
In this chapter, we will first introduce the UV-visible contrast and
UV detectors based on silicon and SiC. We will then describe in detail
several types of UV detectors based on nitrides: photoconductors,
Schottky photodiodes, MSM photodiodes, and p-i-n photodiodes. The
most sophisticated technological advances involved with the recent
Ultraviolet Photodetectors 189
construction of avalanche photodiodes and phototransistors based on
nitrides are also briefly described
5
.
6.2. The UV-visible contrast
The absorption of light is the fundamental effect governing the
dependence of the spectral response of a photodetector as a function
of the energy of the incident photons. In direct-bandgap
semiconductors, we therefore observe a dramatic transition in the
value of the spectral response close to the bandgap of the
semiconductor material, dropping all at once to a very low value when
we drop below the bandgap. In the case of UV photodetectors,
assigning an arbitrary limit to the wavelength of visible light of
400 nm, the UV-visible contrast is defined by the ratio
R
imax
(O = O
G
)/R
i
(O = 400 nm), where R
imax
is the maximum value of
the response coefficient at the wavelength O
G
of the bandgap and
R
i
(O = 400 nm) represents the value of the response coefficient at the
limit of visible light.
The UV-visible contrast is mostly limited by the two following
factors:
– the presence in the bandgap of the semiconductor of deep energy
states which are electrically and optically active. These originate in
particular from extended or localized defects in the crystal structure of
the semiconductor, but can also be attributed to the presence of deep
energy levels due to impurities, or deep levels caused by
imperfections in the surface of the semiconductor;
– in the case of Schottky junction photodetectors with semi-
transparent contacts, the effects of photo-emission of carriers by the
metal.
5. The work carried out by the authors on UV photodetectors, described in this
chapter, was carried out with the support of the European Community, under the
Environment and Climate grant number #ENV4-CT97-0539.
190 Optoelectronic Sensors
6.3. Si and SiC photodetectors for UV photodetection
Contemporary semiconductor-based commercial technical
solutions proposed for UV photodetection mostly rely on Si and SiC-
based technologies. Other applications, a little less widespread and
well-known, use other semiconductor compounds such as GaAs, GaP
and GaAsP. These are mainly used for Schottky barrier photodiodes.
This particular family of devices is in particular very well-suited for
the detection of VUV radiation, for which they offer a remarkable
stability of operation. It is also worth mentioning the recent
emergence of UV photodetectors based on diamond, which allows the
construction of photoconductors with a short cutoff wavelength (225
nm) with a UV/visible contrast of 10
6
[WHI 96]. Although diamond-
based devices are potentially very interesting and well-suited for UV
photodetection applications and the detection of high-energy particles,
due in particular to diamond’s very high resistance to the effects of
irradiation, the development of this last type of photodetector has
nevertheless been rather restricted by the current technical challenges
of achieving epitaxial films with adequate homogenity and
monocrystallinity: the best material obtained to date is a crystal
mosaic with grain boundaries which have a fairly strong mismatch
between one another. Clearly, this results in a high density of
electrically and optically active defects which reduces the
performance of the photoconductors, both in terms of response
coefficient and UV/visible contrast and in terms of response time.
Nevertheless, it is important to emphasize that UV photodetectors
based on thin epitaxial films of diamond have recently become
commercially available, which is the sign of a definite and growing
interest in these devices for UV photodetection and its applications
[MAI 00].
Next we will give, by way of illustration, a brief description of the
main families of Si and SiC-based photodetectors.
Ultraviolet Photodetectors 191
6.3.1. UV photodiodes based on silicon
Silicon p-n junction photodiodes are currently by far the most
economical and widespread commercial solution for UV
photodetection. Silicon gives rise to low-cost detectors which are
suitable for the vast majority of UV detection applications across a
very wide spectral range extending from near-UV to VUV, and even
extends to the high-energy system of soft X-rays. However, the
general shape of their spectral response is not uniform, and in
particular it reaches into the infrared, silicon’s bandgap being 1.1 eV
at ambient temperature (300 K). As a result, their use in selective UV
photodetection applications requires, generally speaking, the use of
absorbing filters between the light source and the photodetector, in
order to select only the spectral window of direct interest. We note in
passing that this often significantly increases the cost of the final
detection system, the filters normally requiring complex technologies
which make them very expensive.
UV photodiodes based on silicon can be subdivided into two large
families: p-n junction photodiodes and metal-oxide-semiconductor
photodiodes with a charge inversion layer [GOL 99, RAZ 96]. The
first family of Si UV photodetectors comprises photodiodes whose p-n
junction is situated at a shallow depth below the surface (typically of
the order of 0.2 μm), and which are covered with a thin layer of SiO
2
.
This insulating layer plays the double role of passivation of the
surface of the semiconductor and of antireflection coating. The depth
of the junction is an extremely important parameter, because
absorption of light is more superficial the more energy the photon has.
In the case of p-n junction photodiodes, it is desirable, in order to
optimize its behavior and obtain the maximum photocurrent, that as
many UV-photogenerated carriers as possible reach the junction
region without recombining. For the same reason, control of surface
recombination effects is of fundamental importance for silicon UV
photodetection. Detailed control and optimization of the Si/SiO
2
interface properties makes it possible to reduce the density of surface
traps, while naturally producing an electric field at the surface of the
192 Optoelectronic Sensors
silicon which limits, or even eliminates, recombination effects at the
interface.
The p-n junction UV photodiode with record performance was
initially developed by Korde et al. [CAN 89, KOR 87]. In order to
form the junction, these authors used phosphor (P) diffusion onto an
Si (p) substrate free from dislocations. A SiO
2
layer of thickness 60
nm was built on the surface of the device. Due to its thickness, this
layer absorbs all radiation of wavelengths below 120 nm, which
unfortunately makes this type of photodiode unusable in applications
involving the detection of high energy UV (VUV in particular).
Nevertheless, this type of photodiode is characterized in particular by
an internal quantum efficiency of 100% in the spectral range of 350-
600 nm. The internal quantum efficiency is greater than 1 in the short-
wavelength UV region (<350 nm), due to secondary impact ionization
effects: a high kinetic energy is imparted to the electron-hole pair
photogenerated by high-energy photons, and this is sufficient to create
other secondary electron-hole pairs through impacts with atoms in the
crystal lattice. On the other hand, the efficiency becomes less than 1
beyond a wavelength of 600 nm, because this long-wavelength light is
absorbed in the deeper layers of the material which are far from the p-
n junction, creating electron-hole pairs which are unable to reach the
junction since their diffusion lengths are too short. As a result, they
recombine in the volume of the semiconductor and do not contribute
to the photocurrent.
It is interesting to dwell for a moment on UV photodiodes with a p-
n junction designed for the detection of VUV up to an energy of
124 eV. Although the base structure of the p-n junction remains the
same as that which we have already described, conversely the oxide
layer present on the surface is much thinner than in other types of UV
Si photodiodes. Its characteristic thickness is in this case no larger
than 4-5 nm, making it a factor of 10 or 20 times smaller than that
normally found in more conventional applications. Transparent to
short wavelengths of light, its useful volume is on the other hand
reduced, in such a way that the possibilities for UV light and humidity
to create high densities of recombinant traps in the oxide (which is
Ultraviolet Photodetectors 193
normally one of the main sources of aging in Si UV photodiodes) is
very significantly reduced compared to the normal situation. It follows
that the stability of operation of these devices is very good when they
are exposed to very high energy photons. The process of photo-
generation produces carriers with very high kinetic energies, which
multiply through secondary impact ionization. In the system of high
photon energies, the internal quantum efficiency K of Si VUV
photodiodes typically follows a linear variation as a function of the
energy E of the photon, which can be written: K = E/3.63 [RAZ 96].
To give a better idea of the orders of magnitude, the internal quantum
efficiency of a Si VUV photodiode is typically of the order of 30, for a
photon energy of 124 eV. These photodiodes are commonly equipped
with a bandpass filter consisting of thin metallic layers, picking out
the 10-50 nm band (Al/C, Al/C/Sc, Ti, Sn, Ag, etc.) [RAZ 96].
Photodiodes with a charge inversion layer [GOL 99, RAZ 96] are
similar in their architecture to metal-oxide-semiconductor structures
used to build field effect transistors: they make use of effects
associated with the presence of the two-dimensional charge induced
through inversion at the Si/SiO
2
interface. Through the effects of the
energy band curvature in the semiconductor, a thin deposited layer of
silica on the surface of the Si (p) material leads to the creation of a
charge bilayer with the opposite sign to the doping of the volume,
which consists of electrons in the case we will describe. This very
localized accumulation of charge results in a high electric field being
established in the immediate vicinity of the semiconductor surface
which, in the case of UV photodetection, offers the particular
advantage that this is the exact region where the main absorption of
the high-energy incident light takes place. As a result, this
configuration is ideal for maximizing the quantum efficiency of the
photodiodes, something which is also helped by the effectively
complete absence of a “dead region” of recombination at the surface,
which tended to be found in the first generations of p-n junction UV
Si photodiodes. Si UV photodiodes with an inversion layer offer a
high quantum efficiency in the spectral band from 250-500 nm and a
cutoff wavelength of 120 nm, which is fundamentally linked to the
absorption of the UV light by the oxide layer above this wavelength. It
194 Optoelectronic Sensors
should, however, be noted that the behavior of this type of photodiode
tends to change over time under UV illumination, since UV light has a
tendency to alter the surface oxide layer. Additionally, the range of
linearity of these photodetectors remains reduced compared to the
possibilities offered by p-n junction UV Si photodiodes, an effect
which seems to be mostly explained by the effects of the high
electrical resistance presented by the charge inversion bilayer at the
SiO
2
/Si interface.
6.3.2. SiC-based UV photodetectors
The best UV photodetectors based on 6H-SiC are p-n junction
photodiodes [RAZ 96]. The optimized structure consists of a layer of
SiC (n+) with a thickness between 0.2 and 0.3 μm, sitting on a SiC (p)
substrate. The doping levels are 5–10 u 10
18
cm
–3
for SiC (n+) and
5–8 u 10
17
cm
-3
for SiC (p). The structure is topped with a SiO
2
passivation layer on the surface. An ohmic contact based on nickel is
present on the n+ side. The response coefficient of p-n SiC
photodiodes made in this way lies in the range from 0.15 to 0.175
A/W at a wavelength of 270 nm, which corresponds to an internal
quantum efficiency of 70 to 85%. At O = 200 nm, the response
coefficient of these same structures is typically 0.05 A/W. The
spectral response of these photodiodes contains a peak lying between
O = 268 and O = 299 nm. The dark current of these devices is very
weak. A dark current value of 10
–11
A for a reverse-bias of –1 V has
been measured for SiC photodiodes, at a temperature of 473 K [EDM
94].
Schottky junction 6H-SiC photodiodes are also currently being
developed, operating in the 200–400 nm spectral range [RAZ 96]. The
best performances have traditionally been obtained for Schottky
devices made from SiC (p). The height of the Schottky barrier is about
twice as large as that normally achieved for n-type SiC. In addition,
the diffusion length for electrons is significantly larger than that for
holes. Very low leakage currents can be obtained for Schottky diodes
made using SiC (p). Leakage currents below 1 pA are currently
measured for a reverse-bias of –10 V. However, Anikin et al. [ANI
Ultraviolet Photodetectors 195
92] report the fairly recent construction of a high quality Schottky
photodiode in SiC (n). In this the Schottky junction is made through
deposition of gold, which gives a Schottky barrier height in SiC which
is typically between 1.4 and 1.63 eV. The behavior of these
photodetectors is characterized by a very low leakage current, of the
order of 100 pA for a reverse-bias voltage of –100 to –170 V, and a
high value of the response coefficient, which reaches 150 mA/W at a
wavelength of 215 nm [RAZ 96].
As we have already emphasized, one of the interests in SiC-based
devices is their remarkable resilience under extreme operating
conditions. This is why only a very small deterioration in their
operational characteristics (spectral response and I(V) characteristics)
is seen when these photodetectors are used at high temperatures,
which can reach or even surpass 300°C without seeing a significant
degradation of the devices.
6.4. UV detectors based on III-V nitrides
Among the various families of materials suitable for UV
photodetection, III-V nitrides represent one of the most interesting and
flexible technical solutions at present. The first work in characterizing
the physical properties of polycrystalline gallium nitride (GaN) dates
back more than 30 years. The development of epitaxial techniques for
this material has for its part been held back for several decades,
mainly due to the poor suitability of the techniques used and the
absence of a substrate with a suitably matching lattice parameter,
which for a long time made the development of applications
impossible. Despite a strong lattice mismatch (it is 16% that of GaN)
and despite a strong difference in thermal expansion coefficient
compared to GaN, sapphire substrates are currently very widely
employed for the growth of thin layers of III-V nitrides and the
resultant applications, for which they are the substrate of choice.
Organometallic vapor phase epitaxy (OMVPE) and, more recently
introduced for nitride growth, molecular jet epitaxy with a plasma or
gas source (MJE) currently allow the growth of thin layers of GaN and
related compounds which, despite a high density of dislocations (of
196 Optoelectronic Sensors
the order of 10
9
to 10
10
cm
–2
in GaN), exhibit structural, optical and
electrical properties on sapphire which are good enough to make them
suitable for many applications, mostly involving light-emitting diodes
at short emission wavelengths (from green to blue, and even into the
near-UV), blue lasers, field effect transistors and UV photodetectors.
6.4.1. Photoconductors
6.4.1.1. Spectral response
Photoconductors based on AlGaN typically consist of an epitaxial
layer of AlGaN on sapphire, doped with silicon, with a thickness of
1 μm, onto which two ohmic contacts are attached (see Figure 6.4).
The electrical bias circuit includes a DC voltage source connected
to the photoconductor, and the device is in series with a low-value
load resistance. The current induced in the device can be easily
deduced from measurement of the voltage drop across the load
resistance when the photoconductor is illuminated. Photoconductors
based on GaN and AlGaN have a gain under continuous illumination
which can reach 10
5
–10
6
and which strongly depends on the incident
optical power. Its variation is a function which decreases in proportion
to P
–J
over a range of five decades, with 0.5 < J < 0.95, where P is the
incident optical power [MON 99a]. This behavior is independent of
the wavelength of the light. The value of J is a decreasing function of
the electrical resistance of the layer, and decreases with temperature.
Figure 6.4. Diagram of a photoconductor based on Al
x
Ga
1–x
N(Si)
Ultraviolet Photodetectors 197
The spectral response of photoconductors based on Al
x
Ga
1-x
N
exhibits a cutoff at 298 nm for x = 0.319 nm for x = 0.22 and 368 nm
for x = 0.35. The UV/visible contrast is significantly smaller than the
value obtained from a simple analysis based on the absorption
coefficient of these materials. The mechanism which explains the
variation of the spectral response of the photoconductor must
therefore involve more than just the optical absorption of the
semiconductor. In addition, a sub-linear dependence of the
photocurrent is observed as a function of the power of the incident
light. In the model developed by Garrido et al. [GAR 98], the
photoconductor’s response coefficient R
i
is the combination of two
terms, one due to the photogenerated free carriers, 'n, and the other
due to the modification by the light of the effective conducting cross-
section, 'S. The response coefficient can be written:
R
i
=
'I
P
opt
=
qV
B
μ
e
LP
opt
S'n÷n'S
( )
[6.1]
where q is the elementary charge of the electron, P
e
is the electron
mobility, L is the distance between the contacts, S is the conductive
cross-section, V
B
is the bias voltage and n is the concentration of free
carriers.
The transverse conduction cross-section in the model of Garrido et
al. is not the same as the geometric cross-section of the device, due to
the presence of space charge regions (SCRs) [HAN 98] around lattice
discontinuities (dislocations that reach the surface, grain boundaries
and interface). The absorption of light produces a reduction in these
depletion regions, leading to a modification of the conduction cross-
section. Given that 'n can be expressed in the form [RAZ 96]:
'
n
=KgW
P
opt
hc O
[6.2]
where K is the quantum efficiency, g the photoconduction gain, W the
lifetime of the free carriers and hc/O the photon energy; by
substituting [6.2] into [6.1], it follows that:
198 Optoelectronic Sensors
R
i
=
'I
P
opt
=
qV
B
P
e
L
KgWS
hc O
÷
n
P
opt
'S
(
(
·
·
·
·
·
)
)

[6.3]
The spectral variation of the first term in this equation is given by
O.K(O). Since the quantum efficiency is a direct function of the
absorption, this first term leads to a significant UV/visible contrast.
However, experiments show that the dominant mechanism in these
photodetectors is represented by the second term of equation [6.3], in
other words, it depends strongly on the modification to the effective
transverse conduction cross-section. This mechanism can also explain
the high value of the spectral response for photon energies above the
bandgap energy: the energy levels responsible for the absorption of
light in the visible can originate either from point defects distributed
homogenously within the semiconductor, either dopants or vacancies,
or in extended defects associated with lattice discontinuities. If these
defects are electrically charged, they create a depletion region around
them, which reduces the effective conduction cross-section of the
device. Complete absorption of light through these defects can be
treated as negligible, so that their presence has little effect on the
absorption coefficient, and hence on the first term of equation [6.3].
However, their effect on the photoconduction spectral response is
large, since the electrical charge concentrated at these lattice
discontinuities is modified by the light, which alters the effective
transverse conduction cross-section.
6.4.1.2. Response time
AlGaN-based photoconductors display significant persistent
photoconductivity (PPC), effects. This means they have a very slow
transitory temporal response (several thousandths of a second) and a
non-exponential tail, which is consistent with the frequency
dependence of their spectral response [SHE 99].
The slow drop-off in the photocurrent can be explained by a model
based on the modulation of the transverse conduction cross-section
[MON 99a]. The curvature of the bands around defects and
dislocations leads to a spatial separation of electrons and holes, which
Ultraviolet Photodetectors 199
can be the cause of the PPC. When the light is extinguished, the
electrons are inclined to recombine, but they must first cross the
potential barrier separating them from their recombination sites. The
height of this barrier depends on the charge present at the defect, so
that it will change over time as recombination occurs. Through this
mechanism, the SCR returns slowly to its original extent in the dark,
and as a result the resistance of the device also slowly returns, in a
non-exponential manner, towards its dark value. The dominant defect
type may be that associated with dislocations at the AlGaN-air
interface or the AlGaN-substrate interface [SHE 99]. However, it is
difficult to discriminate between these two regions, and their relative
significance will depend on the crystalline quality of the material.
Quantitatively, the temporal response of photoconductors can be
calculated in the following manner, if 'Q
SS
is the increase in charge at
the defects, connected to the variation in the width of the SCR, 'x
SCR
.
'Q
SS
can be expressed in the form:
'QSS =qN
D
'x
SCR
= 2N
D
H
S
H
0
<
0
÷ <
il
( )
[6.4]
where N
D
is the level of n-type doping, H
S
and H
0
are the relative
dielectric constant of the semiconductor and the vacuum permittivity
respectively, and <
0
and <
il
are the heights of the barrier around the
defect, in the dark and under illumination respectively. It follows that
the response coefficient of the photoconductor is given by:
R
i
=
A
opt
V
B
P
e
P
opt
L
2
'QSS [6.5]
where A
opt
is the optical surface area of the device.
In the dark, 'Q
SS
originates from the current crossing the potential
barriers around the defects. Including the thermionic emission current
J
th
, as well as the tunneling current J
tunnel
, it follows that:
200 Optoelectronic Sensors
dQ
SS
dt
= J
th
÷ J
tunnel
[6.6]
The thermionic and tunneling currents can be expressed respectively
in the following forms:
J
th
= A*T
2
exp
÷q<
0
kT
(
(
·
·
·
·
)
)

exp
qV
ph
kT
(
(
·
·
·
·
·
)
)

÷1

l
l
l
l
l
[6.7]
J
tunnel
= A*T
2
S E
00
<
0
÷V
ph
( )
kT
q
cosh
qE
00
kT
(
(
·
·
·
·
)
)

exp
÷q <
0
÷V
ph
( )
E
0
(
(
·
·
·
·
·
·
)
)

1÷exp
qV
ph
kT
(
(
·
·
·
·
·
)
)

l
l
l
l
l
[6.8]
with E
00
= (h/4S).(N
D
/m*
e
) and E
0
= coth(qE
00
/kT). V
ph
is the voltage
induced by the light. The drop in voltage across the barrier is V
ph
(t),
which can be expressed as:
V
ph
t
()
= <
0
÷
Q
2
SS t
()
2qH
S
H
0
N
D
[6.9]
To illustrate this, Figure 6.5 shows the temporal response of
AlGaN photoconductors, both measured, and calculated by numerical
solutions of equations [6.4] to [6.9].
, , , , ,
Figure 6.5. Decrease in the normalized photocurrent in Al
0.23
Ga
0.77
N
photoconductors, along with numerical simulation using equations [6.4]–[6.10]
Ultraviolet Photodetectors 201
Figure 6.6. Normalized spectral response of a photoconductor, measured at
different modulation frequencies of the incident optical signal.
The spectral response of a Schottky photodiode made from
the same batch is also shown for comparison
6.4.1.3. Effects of frequency modulation of the incident optical signal
The effect of frequency on the spectral response of
photoconductors is shown in Figure 6.6. When the frequency
increases, the SCRs associated with the defects do not have time to
respond. The result is that the cutoff slope becomes sharper and the
spectrum of the response coefficient tends to approach that of a
photovoltaic detector. As a result, the UV/visible contrast is much
better at higher frequencies. The mechanism at the root of the
response of III-V nitride-based photoconductors therefore makes them
of little interest for applications which requires speed or a high
UV/visible contrast.
202 Optoelectronic Sensors
6.4.2. Schottky barrier photodiodes based on AlGaN
6.4.2.1. Electrical properties
Work published on Schottky barrier photodiodes involves planar or
vertical structures (Figure 6.7), which in theory both give extremely
similar operating characteristics. Although the vertical structure is
particularly interesting for the construction of fast devices with a high
response coefficient, the current limitations in III-V nitride
technologies (connected in particular with damage to the material
during the engraving of the mesa) result in a performance degradation,
particularly in terms of the bandwidth and the noise level.
Figure 6.7. Schematic diagram of AlGaN Schottky photodiodes in
(a) vertical and (b) planar configurations
Figure 6.8 shows the typical current-voltage (I-V) characteristics of
planar Schottky photodiodes. In GaN-based devices, the ideality factor
is of the order of 1.2, with a series resistance of 20-50 : and a leakage
resistance greater than 1 G:. The leakage current increases as a
function of the fraction of aluminum, and the ideality factor also
increases and can reach values close to 4. Consequently, the measured
I-V characteristic of AlGaN-based diodes does not make it possible to
deduce precise and reliable information about the height of the
Schottky barrier, due to the high values of the ideality factor. The
variation of the capacitance as a function of the bias voltage leads to
barrier heights of 0.84 eV and 1.02 eV respectively for semi-
transparent Ni and Au contacts on GaN. The height of the barrier as a
function of the molar fraction of aluminum for Schottky diodes with
nickel and gold contacts is shown in Figure 6.9.
Ultraviolet Photodetectors 203
Figure 6.8. Typical current-voltage characteristics for Schottky diodes
made of Al
x
Ga
1–x
N on sapphire
6.4.2.2. Spectral response
In contrast to photoconductors, the photocurrent of Schottky
barrier devices depends linearly on the incident light power (inset in
Figure 6.10). This linearity is independent of the size of the diode, the
nature of the metal used to fabricate the Schottky diode and the
excitation wavelength, for values situated both above and below the
bandgap energy of the semiconductor. The decrease in the response
coefficient observed when the fraction x of aluminum increases results
from two effects. Firstly, at a constant quantum efficiency, the
response coefficient decreases at the shortest wavelengths. Also, the
penetration depth of the light is larger than the SCR, and so the
response coefficient depends on the diffusion length of the minority
charge carriers, which decreases with the concentration of aluminum.
204 Optoelectronic Sensors
Figure 6.9. Variation of the height of the Schottky barrier
as a function of the molar fraction of Al
Figure 6.10 shows that the UV/visible contrast is more than three
orders of magnitude. The cutoff wavelength develops from 362 nm to
293 nm. The steepness of the cutoff slope of Schottky photodetectors
shows that the only limitation in the UV/visible contrast is that caused
by absorption by deep defects, in contrast to the results obtained for
photoconductors. The spectral response is flat for wavelengths smaller
than those of the bandgap, which is an advantage that Schottky
photodiodes have over other types of photodetectors, particularly p-n
junction photodetectors. Since the SCR lies just below the surface of
the semiconductor, there is no decrease in the quantum efficiency at
short wavelengths as is normally seen in p-n and p-i-n photodiodes,
where the photogenerated carriers must diffuse across the layer above
before reaching the SCR at the level of the junction. In junction
devices based on GaN, this problem is accentuated by the high density
of dopants required to achieve p-type conductivity, which results in a
very thin SCR and a material which has a very low mobility. From
this point of view, the choice of Schottky photodiodes based on
Al
x
Ga
1-x
N has more potential than p-n or p-i-n devices for
constructing wideband UV detectors.
Ultraviolet Photodetectors 205
Figure 6.10. Normalized spectral response at ambient temperature of
Schottky photodiodes for different values of the molar fraction of Al.
Inset:: variation of the photocurrent as a
function of the incident light power
The maximum value of the response coefficient obtained in AlGaN
Schottky photodiodes is limited by the optical transmission of the
semi-transparent Schottky contact, and by the diffusion length L
h
of
the minority carriers in this material. For radiation with an energy
close to the bandgap of the semiconductor, a certain proportion of the
photons may cross the SCR without absorption and penetrate deep
into the volume of the material. In order to be collected, the electron-
hole pairs must therefore diffuse towards the SCR at the surface, and a
low value of L
h
results in a decrease in the response coefficient. L
h
is
a function of the mobility and the lifetime of the minority charge
carriers, and is an increasing function of the quality of the material. It
follows that the use of thick AlGaN layers (thickness greater than
1 μm) is preferred because they allow a reduction in the density of
dislocations, and hence an increase in the diffusion length and the
response coefficient; it also results in a reduction of the dark current.
206 Optoelectronic Sensors
6.4.2.3. Response time
The response time of these photodetectors is mainly limited by the
RC time constant of the device, where C is the sum of the internal
capacitance of the diode and the load capacitance, and R is the sum of
the load resistance and the series resistance of the device. Response
times around 50 nanoseconds are typical for diodes of diameters
between 200 and 300 μm. The capacitance of the depletion region is
given by the expression:
C
SCR
= A
qN
D
H
S
H
0
2 <
0
÷V
( )
[6.10]
where A is the area of the Schottky contact, N
D
is the density of
dopants, H
s
is the dielectric constant of the semiconductor, H
0
is the
vacuum permittivity, <
0
is the height of the Schottky barrier seen
from the semiconductor side, and V is the bias voltage. Since the
series resistance does not depend on the bias voltage, the response
time of the device decreases in proportion to (<
0
+V)
–1/2
.
For planar devices, R
S
and C
SCR
are proportional to (PN
D
)
–1
and
N
D
1/2
respectively, where μ is the electronic mobility. The maximum
bandwidth of these devices is therefore proportional to PN
D
1/2
. Since
the mobility increases as the doping level falls, there exists an
optimum doping level, of the order of 10
18
cm
–3
, for which the
response time is minimized. This effect is accentuated for AlGaN
alloys with high concentrations of aluminum, because of the difficulty
in achieving good quality ohmic contacts.
6.4.2.4. Noise and detectivity
The 1/f noise dominates at low frequencies. Its spectral noise
power density, S
n
, satisfies the relationship:
S
n
=s
0
I
dark
D
f
J
[6.11]
where I
dark
is the dark current, f is the frequency, and s
0
, J and D are
dimensionless parameters whose values are close to 1 and 2
respectively. As well as the 1/f noise, the contribution of the shot noise
Ultraviolet Photodetectors 207
must be included. The noise equivalent power (NEP) can therefore be
written in the form:
NEP =
i
shot
2
÷ i
1 f
2
R
i
[6.12]
where i
shot
is the shot noise current, i
1/f
is the 1/f noise current and R
i
is
the response coefficient of the device. The value of <i
shot
2
> is given
by:
i
shot
2
= 2qI
dark
'f [6.13]
where 'f is the bandwidth of the photodiode, and I
dark
represents the
dark current. NEPs of 8 pW/Hz
1/2
and 41 pW/Hz
1/2
have been
obtained for Schottky GaN/Au and Al
0.22
Ga
0.78
N/Au photodiodes
respectively, at a reverse-bias voltage of –2 V. The corresponding
detectivity values are 6.1 u 10
9
W
–1
H
–1/2
cm and 1.2 u 10
9
W
–1
Hz
1/2
cm respectively.
6.4.2.5. Schottky barrier photodiodes made from GaN grown using
epitaxial lateral overgrowth (ELOG)
Despite these promising characteristics, heteroepitaxial growth of
GaN results in a high density of dislocations (~ 10
8
–10
9
cm
–3
), which
limits the UV/visible contrast in GaN-based photodetectors. The
recent development of the technique of epitaxial lateral overgrowth
(ELOG) has reduced this density of dislocations by at least two orders
of magnitude [NAM 97, USU 97].
Schottky barrier photodiodes made with epitaxial GaN layers
grown using the ELOG technique offer an order of magnitude increase
in the UV/visible contrast compared to GaN-based devices grown on
sapphire using standard methods. The dark current, whose value lies
below 1 nA/cm
2
for a reverse-bias voltage of –1 V, is significantly
smaller than that of Schottky photodiodes based on GaN/sapphire. As
the response time of the device is limited by the time constant RC, the
weaker residual doping in these epitaxial layers makes it possible to
obtain bandwidths greater than 30 MHz, 12 MHz and 8 MHz for
208 Optoelectronic Sensors
devices with diameters equal to 200 μm, 400 μm and 600 μm
respectively. A detectivity as high as 5 u 10
11
W
–1
Hz
1/2
cm has been
measured for GaN ELOG photodiodes with a diameter of 400 μm,
with a bias voltage of –3.4 V.
6.4.2.6. Application of AlGaN Schottky barrier photodetectors for the
simulation of the biological effects of UV light
UV radiation produces a wide range of biological effects [DEG 94,
MCK 87]:
– pigmentation: maximum effect at O = 360–400 nm;
– erythema (sunburn): maximum sensitivity at O < 297 nm;
– synthesis of vitamins D2 and D3: O = 249 – 315 nm. Maximum
yield at 290 nm;
– damage to plants: O < 317 nm;
– bactericidal action: O = 210–310 nm. Maximum effects at
254 nm;
– carcinogenic effects: UVB and UVC. Maximum at 310 nm;
– DNA damage: O < 320 nm. Effect grows rapidly as the
wavelength decreases.
Simple, accurate, reliable and low-cost devices are hence required
to evaluate the biological effects of UV radiation. UV detectors with a
wide spectral band have been developed to study the effects of UVA
and UVB on the skin [MCK 87]. Commercial UV dosimeters use
photodiodes with a small bandgap (Si, GaAs, GaP) combined with a
series of filters inserted into the path of the incident light. Muñoz et al.
[MUN 00] showed that the spectral response of AlGaN Schottky
photodiodes can match the UVA and B erythema action spectrum,
with the appropriate selection of an appropriate molar fraction and
carefully selected growth conditions (Figure 6.11). In this way, the
response coefficients of these devices can give direct information on
the biological effects of UV light. For this application, AlGaN
photodetectors therefore offer a flexible solution with a low cost and
minimum footprint, since they require neither filters nor temperature
stabilization.
Ultraviolet Photodetectors 209
Figure 6.11. Normalized spectral response of an AlGaN Schottky photodiode
compared with the sunburn action spectrum (international standard
set by the International Commission on Illumination (CIE))
6.4.3. MSM photodiodes
MSM photodiodes consist of two interdigitated Schottky contacts
deposited onto a planar surface (see Chapter 5). The main advantage
of these structures, based on AlGaN, is their ease of fabrication. The
typical structure of such an MSM diode involves two interdigitated
Ni-Au Schottky contacts deposited on an epitaxial layer of AlGaN,
not intentionally doped, of micron-scale thickness. Their dark current
is very weak and follows the thermionic emission model with a value
of the Richardson constant equal to 26 Acm
-2
K
-2
and a barrier height
of 1.04 eV for GaN.
Their spectral response is relatively flat above the bandgap, with a
steep cutoff slope which shifts towards shorter wavelengths as the
proportion of aluminum increases. The photocurrent varies in a quasi-
linear manner with the incident optical power. The UV-visible
contrast varies from around three orders of magnitude for the weakest
bias voltages (typically less than 5V) to four or five orders for the
highest bias voltages. This accentuation of the contrast with the bias
voltage can be explained by the appearance of a gain. At present, there
210 Optoelectronic Sensors
is no satisfactory explanation available for the origin of this gain.
Despite the existence of this gain, the dynamic behavior of these
photodetectors is for the most part limited by the RC time constant and
the transit time of the photocarriers [MON 99c].
Their NEP lies between a few pW/Hz
1/2
and a few tens of
pW/Hz
1/2
depending on the geometry of the device and the bias
voltage [MON 99b].
6.4.4. p-n and p-i-n photodiodes
The performance limitations of the first GaN p-n junction
photodiodes were mainly caused by a high resistivity of the p-type
layer, connected with difficulties associated with magnesium p doping,
and the high resistivity of the associated ohmic contact. Progress made
in understanding the growth conditions of the materials, and in the
technology of the ohmic contacts, then made it possible to make p-n
junction GaN photodiodes with a fast temporal response (response
time of the order of 100 ns) and a very low noise level. The
performance of these devices was further improved by the
introduction of a region that was not intentionally doped (p-i-n
photodiodes) and through the use of AlGaN/GaN heterostructures.
Arrays of heterostructure Al
0.20
Ga
0.80
N/GaN p-i-n photodiodes with
dimensions 32 u 32 and 128 u 128 pixels [BRO 00] have been made
with good performance in terms of their response coefficient and
detectivity. p-i-n photodiodes in Al
x
Ga
1–x
N, sometimes with high
levels of aluminum (up to 76% [SAN 00]), have also been described.
The optimization of these last devices is however limited by the
intrinsic difficulties in achieving a high level of p doping in
Al
x
Ga
1–x
N materials, which can only be made with difficulty to
exceed 10
17
cm
–3
, and by the associated ohmic contact whose
resistivity remains high. The behavior of the device is thus doubly
penalized: firstly, the response time of a Al
x
Ga
1-x
N p-i-n
photodetector cannot be made small enough, because the series
resistance component due to the resistance of the electrical contacts
remains large, and the device also contributes a significant level of
noise associated with an ohmic contact of insufficient quality, which
Ultraviolet Photodetectors 211
leads to detectivity values which are notably smaller than those that
can be obtained with p-n and p-i-n photodiodes based on GaN.
6.4.4.1. Static characteristics
In p-n and p-i-n photodiodes based on Al
x
Ga
1–x
N, the linearity of
the photocurrent as a function of the incident light power extends over
a scale of more than five orders of magnitude [WAL 99]. The
response coefficient of p-n and p-i-n homojunction GaN photodiodes
typically lies in the range 0.10–0.15 A/W, which corresponds to
external quantum efficiencies of between 30 and 44%. These results
are improved by the use of a Al
x
Ga
1–x
N layer in the illuminated region
(which can be n- or p-doped), in order to avoid the loss of carriers
through diffusion. A response coefficient of 0.2 A/W at O = 365 nm
has been measured for a sapphire-Al
0.28
Ga
0.72
N(n)-GaN(i)-GaN(p)
heterostructure illuminated by the rear face through the sapphire
[YAN 98]. A maximum in the response coefficient of 57 mA/W was
reported at O = 287 nm, which corresponds to a quantum efficiency of
25% [PAR 99]. An increase in this coefficient is observed in all cases
as a function of the bias voltage [WAL 99]. This result confirms that
the sensitivity is limited by the diffusion length of the charge carriers.
As the bias voltage increases, the width of the SCR increases, such
that the carriers created through photo-ionization processes in the
regions further from the junction are collected.
A UV/visible contrast of more than four orders of magnitude is
observed for all these devices. The small diffusion length of the
carriers can affect the spectral response above the bandgap of p-i-n
photodiodes based on Al
x
Ga
1–x
N [XU 97]. The absorption coefficient
increases as the wavelength decreases, meaning that the absorption of
the light occurs closer to the surface. The result is that electron-hole
pairs are generated above the junction in the upper layer and must
diffuse to the junction in order to be collected. A short diffusion
length therefore leads to a decrease in the response coefficient for
shorter wavelengths, which is all the more marked when the surface
layers are thicker (of the order of 0.2 μm). p-i-n detectors based on
group III nitrides are characterized by a sharp cutoff at long
wavelengths, whose energy value is very closely linked to the
212 Optoelectronic Sensors
bandgap of the material forming the active region. For certain
applications, however, an additional cutoff wavelength is also required
at short wavelengths, defining a spectral window for UV detection. In
their theoretical study, Pulfrey and Nener [PUL 98] investigated the
possibility of using p-i-n Al
x
Ga
1–x
N/GaN heterostructures as bandpass
UV detectors, with a high energy cutoff wavelength at the bandgap of
Al
x
Ga
1–x
N. In this case, the Al
x
Ga
1–x
N layer must be thick enough not
only to accentuate the absolute response peak, but also to act as an
integral high-pass filter. Devices equipped with a 1 μm thick
Al
0.10
Ga
0.90
N layer [KRI 98] have demonstrated that it is possible to
achieve rejection on a level of more than two orders of magnitude in
the short wavelength region.
6.4.4.2. Response time
GaN-based p-n and p-i-n photodiodes have a temporal response
which is in general limited by the RC time constant, and show an
exponential fall in the photocurrent. Response times of the order of 25
nm have been measured with zero bias voltage on devices whose
optical surface area is 200 u 200 μm
2
[OSI 97]. This response time
drops to 10 ns for a bias voltage of –6 V, which can be explained by a
reduction in the capacitance of the junction associated with the applied
voltage. Faster response times can be obtained with AlGaN/GaN
heterojunctions [XU 97]. However, the response time of these devices
is often non-exponential, with characteristic fall times that are more
complicated than a simple RC effect [CHE 95, VAN 97, WAL 98].
This behavior can be attributed to the presence of a defect in the
semiconductor, with an activation energy of 99 meV and a high
concentration (of the order of 10
18
cm
–3
) which can probably be
explained by substituted magnesium atoms.
GaN p-i-n photodiodes with response times below 1 ns have been
constructed [CAR 99]. This result is obtained by increasing the
thickness of the “i” layer to 1 μm, which makes it possible to build a
device which has a low capacitance. In this case, the factor limiting
the temporal response is the transit time of the carriers. The only
drawback of this particular type of device is its low response
Ultraviolet Photodetectors 213
coefficient, which is around 0.03 A/W. Such devices are of course
ideal for the detection of rapid signals.
6.4.4.3. Noise
Under illumination, the spectral density of noise under reverse-bias
is well described by the equation S
n
=2.q.I
ph
(shot noise). A very low
normalized noise equivalent power, of the order of 7 fW/Hz
1/2
, has
been obtained in p-n junction photodiodes for a reverse-bias voltage of
–3 V [OSI 97]. Since these devices have an optical area of 200 u 200
μm, their corresponding normalized detectivity is high, of the order of
3u10
12
W
–1
Hz
1/2
cm.
6.4.4.4. p-i-n photodiodes on laterally-grown GaN
Al
0.33
Ga
0.77
N p-i-n photodiodes on GaN have been built using
lateral growth, by means of a SiO
2
mask consisting of 35 μm
widebands with 5 μm openings [PAR 99]. Diode mesas, with
dimensions of 10 u 100 μm
2
, have been made in the regions of the
material corresponding to the “wings” of the lateral growth (outside
the coalescence region, and outside the openings in the SiO
2
mask).
Additional devices with mesa dimensions of 30 u 100 μm
2
have also
been built by way of comparison. These particular devices also cover
the coalescence region situated at the junction of the “wings” of lateral
growth. Finally, devices with a similar form with square 300 u
300 μm
2
mesas have been made from GaN on standard sapphire. Low
dark current densities, of the order of 10 nA/cm
2
, have been measured
for the smallest devices lying on the “wings” of lateral growth. This
leakage current density is an order of magnitude smaller than that
measured for devices which include the coalescence regions, and more
than 6 orders of magnitude smaller than that of devices made from
GaN on standard sapphire. This low value of the leakage current can
be linked to the strong reduction in the density of dislocations, since
these defects lead to a strengthening of the transport mechanism,
assisted by the tunneling effect [YU 98b]. In addition, the spectral
response has a more abrupt cutoff in the case of devices built on the
“wings” of the growth, in a similar way both when the coalescence
region is covered and when it is not. All these responses are consistent
214 Optoelectronic Sensors
with similar observations made for Schottky photodiodes made with
laterally-grown GaN [MON 99d].
6.4.4.5. Avalanche photodiodes based on GaN
The mechanism of avalanche photodiodes is analyzed in detail in
Chapter 3. The unusual requirement for GaN-based materials is that
very high fields are required to achieve multiplication, as calculations
of the ionization coefficients for electrons and holes in GaN show
[KOL 95, OGU 97]. Although a homogenous avalanche multiplication
of carriers has recently been shown for PIN photodiodes based on
GaN [MCI 00], with a response time between 100 and 500 ns, the
high density of defects in these materials makes it difficult to achieve
a homogenous multiplication across the whole active area of the
device. Also, to overcome the limitations in noise due to ionization
coefficients for electrons and holes that are too close together, Ruden
[RUD 99] proposed a hybrid structure consisting of an AlGaN/Si
heterostructure, with multiplication taking place in the silicon.
6.4.5. Phototransistors
The operation of phototransistors is described in Chapter 4. As far
as GaN-based devices are concerned, the literature refers to bipolar
and field effect transistors [KHA 95, YAN 98].
6.4.5.1. Bipolar phototransistors
In a bipolar transistor, the reverse-biased base-collector
junction acts like a p-n junction photodiode, and its photocurrent
is amplified by the transistor effect. In its most up to date
configuration, the base contact is not connected (floating-base
mode of operation). Yang et al. [YAN 98] report the fabrication of a
GaN(n)/GaN(p)/GaN(i)/Al
0.20
Ga
0.80
N(n) heterojunction phototransistor.
Electrical contacts were present on the collector and the emitter, and
the base was left floating. The UV light enters the device through the
sapphire substrate, crosses the Al
0.20
Ga
0.80
N(n) layer, and is absorbed
in the GaN(i) layer. The photogenerated electron-hole pairs are
separated by the electric field in the i region, and the electrons and
Ultraviolet Photodetectors 215
holes travel to the collector and the base respectively. The
accumulation of holes in the floating base increases the injection of
electrons from the emitter, which results in a current gain. A gain of
more than 10
5
has been demonstrated.
GaN-based phototransistors exhibit a sub-linear behavior of the
photocurrent as a function of the incident light power as well as
showing persistent photoconductivity (PPC), analogous to that also
recently observed in photoconductors. In their normal operating
regime, holes ought to recombine in the base with electrons injected
from the emitter. However, the holes are trapped at defects which
reduces their recombination rate and leads to persistent charge effects.
Yang et al. [YAN 98] showed that this recombination is increased by
the application of a bias to the detector in order to force the holes to
enter the emitter. The voltage impulse acts as a sort of electrical “reset
to zero”.
In order to avoid persistent effects, the phototransistor is subjected
to a “reset to zero” voltage impulse before each measurement of the
response coefficient (a voltage between 7 and 10 V is used for this),
and the photocurrent is then measured at a bias voltage of 3 to 4 V.
The gain is strongly dependent on the frequency, so that the gain-
bandwidth product remains constant. An excellent UV/visible contrast
of 8 orders of magnitude was obtained [YAN 98].
6.4.5.2. Field effect phototransistors
AlGaN(n)/GaN(n)/GaN(i) heterostructure field effect transistors
have also been used in photodetection mode [KHA 95], with
illumination through the substrate (sapphire). The photogenerated
holes travel towards the channel, where they are rapidly drawn
towards the drain by the high electric field. These devices display high
response coefficients which can reach 3,000 A/W, with a steep cutoff
and a temporal response of the order of 200 μs.
216 Optoelectronic Sensors
6.5. Conclusion
The development of visible blind ultraviolet photodetectors is at
present largely motivated by the large number of possible applications
in fields as diverse as UV astronomy, studies of the ozone layer, motor
control, missile plume detection, flame detection, secure
communications in space, etc. With these goals in mind,
photoconductors, Schottky photodiodes, MSM photodiodes, p-n and
p-i-n photodetectors and phototransistors based on Al
x
Ga
1–x
N
materials have been developed in recent years, complementing older
widely-available technologies for Si and SiC-based UV photodiodes.
It is important to emphasize that this range of nitride-based sensors is
currently bringing a new degree of improved flexibility to UV detector
fabrication in general, due mainly to the large bandgaps available with
Al
x
Ga
1–x
N materials which make it possible to build detection
systems which are entirely free from intermediate spectral filters, with
performances that are at least equivalent in terms of reliability and
robustness to other currently-available technologies that are
themselves already very high performance.
Al
x
Ga
1–x
N photoconductors offer a high internal gain at ambient
temperatures (~100 for P
opt
= 1 W/m
2
). However, this gain is
associated with a sub-linear behavior as a function of the incident
optical power, a low UV/visible contrast and significant and
undesirable persistent photoconductivity effects. These drawbacks
make them less suitable for most applications. The use of a modulated
optical signal makes it possible to considerably improve the linearity
and UV/visible contrast. However, these photoconductors lose much
of their advantage in this configuration, above all because the
response coefficients of the devices is significantly reduced in such a
configuration, and the measurement system becomes much more
complex and as a result much more expensive.
Schottky photodiodes have the advantage of a uniform and flat
spectral response when they are excited at energies above the bandgap,
independent of the incident optical power and the temperature. They
also offer a steep cutoff slope, with a UV/visible contrast which is of
Ultraviolet Photodetectors 217
the order of 10
3
. Their temporal response is limited by the RC product,
with minimum response times typically of the order of nanoseconds.
This clearly demonstrates that these devices are suited to
environmental applications and the construction of UV photodetector
arrays.
MSM photodiodes with very low dark currents have been built.
These devices have a photocurrent which varies linearly as a function
of the incident light power and makes it possible to achieve a
UV/visible contrast of 10
4
. Given their significant bandwidth and their
low noise level, these devices can be a judicious choice to meet the
needs of detection in the domain of visible blind optical
communications. Furthermore, the possibility could be considered of
integrating Al
x
Ga
1–x
N MSM photodiodes with field effect transistors
based on group III nitrides in the context of the monolithic fabrication
of optical receivers.
p-n and p-i-n junction photodiodes are linear as a function of the
optical power, and can achieve a UV/visible contrast of 10
4
. However,
their response time is normally limited by effects associated with the
presence of trapping levels linked to magnesium, to which a degree of
degradation in the spectral response can also be attributed. The
minimum cutoff wavelength is currently limited by the difficulty of
achieving a high p doping in Al
x
Ga
1–x
N alloys at high concentrations
of aluminum. In spite of very promising published results,
improvements are still currently needed in the p doping of these
materials, in order to increase the performance of these devices, as
well as their reliability.
Phototransistors offer the combination of a very high gain with a
record UV/visible contrast of 10
8
. As a result, these devices are
promising for applications requiring a high spectral resolution, in spite
of the narrowness of their bandwidth which makes them poorly-suited
to high-frequency operation.
To summarize, current results confirm that Al
x
Ga
1–x
N alloys are
the best choice in terms of semiconductors suited to UV
photodetection. However, the performance of these devices is still
218 Optoelectronic Sensors
currently limited by the high density of defects present in the
heteroepitaxial layers. In consequence, very significant improvements
in photodetector devices are anticipated from work currently in
progress in the fields of epitaxial growth and the technology of group
III nitrides, such as the technique of lateral growth.
6.6. Bibliography
[ANI 92] ANIKIN M.M., ANDREEV A.N., PYATKO S.N., SAVKINA N.S., STRELCHUK
A.M., SYRKIN A.L., CHELNOKOV V.E., “UV photodetectors in 6HSiC”, Sensors
and Actuators: A. Physical 33 (1-2), pp. 91-93, 1992.
[ARU 00] ARULKUMARAN S., EGAWA T., ZHAO G.Y., ISHIKAWA H., JIMBO T., UMENO
M., “Electrical characteristics of Schottky contacts on GaN and Al. 0:11. Ga. 0:89.
N”, Japanese Journal of Applied Physics. 39, L351, 2000.
[BRO 00] BROWN J.D., BONEY J., MATTHEWS J., SRINIVASAN P., SCHETZINA J.F.,
NOHAVA T., YANG W., KRISHNANKUTTY S., “UV-specific digital camera based on
a 128x128 focal plane array of GaNAlGaN PIN photodiodes”, MRS Internet
Journal of Nitride Semiconductor Research, 5, article 6, 2000.
[CAN 89] CANFIELD L.R., KERNER J., KORDE R., “Stability and quantum efficiency
performance of silicon photodiode detectors in the far ultraviolet”, Applied Optics,
28, 3940, 1989.
[CAR 99] CARRANO J.C., LI T., EITING C.J., DUPUIS R.D., CAMPBELL J.C., “Very fast
metal-semiconductor-metal ultraviolet photodetectors on GaN with submicron
finger width”, J. of Electronic Materials, 28, 325, 1999.
[CHE 95] CHEN Q., KHAN M.A., SUN C.J., YANG J.W., “Visible-blind ultraviolet
photodetectors based on GaNp-n junctions”, Electronics Lett. 31, 1995.
[DEG 94] DE GRUJI F.R., VAN DER LEUN J.C., “Estimate of wavelength dependency
of ultraviolet carcinogenesis and its relevance to the risk assessment of a
stratospheric ozone depletion”, Health Physics, 67, 314, 1994.
[EDM 94] EDMOND J.A., KONG H.S., CARTER JR C.H.,” Blue LEDs, photodiodes and
high-temperature rectifiers in 6H–SiC”, Physica B 185, 453, 1993.
[GAR 98] GARRIDO J.A., MONROY E., IZPURA I., MUÑOZ E., “Photoconductive gain
modelling of GaN photodetectors”, Semiconductor Science and Technology, 13,
563, 1998.
Ultraviolet Photodetectors 219
[GOL 99] GOLDBERG Y.A., “Semiconductor near-ultraviolet photoelectronics”,
Semiconductor Science and Technology, 14, R41, 1999.
[HAN 98] HANSEN P.J., STRAUSSER Y.E, ERICKSON A.N., TARSA E.J., KOZODOY P.,
BRAZEL E.G., IBBETSON J.P., MISHRA U., NARAYAMURTI V., DEN BAARS S.P.,
SPECK J.S., “Scanning capacitance microscopy imaging of threading dislocations
in GaN films grown on (0001) sapphire by metalorganic chemical vapor
deposition”, Applied Physics Lett., 72, 2247, 1998.
[KHA 95] KHAN M.A., SHUR M.S., CHEN Q., KUZNIA J.N., SUN C.J., “Gate
photodetector based on GaN/AlGaN heterostructure field effect transistor”,
Electronics Lett., 31, 398, 1995.
[KHA 97] KHAN M.R.H., NAKAYAMA H., DETCHPROHM T., HIRAMATSU K., SAWAKI N.,
“A study on barrier height of Au-Al
x
Ga
1-x
N Schottky diodes in the range 0”, Solid
State Electronics, 41, 287, 1997.
[KOL 95] KOLNIK J., H.OGUZMAN I., BRENNAN K.F., WANG R., RUDEN P.P., WANG Y.,
“Electronic transport studies of bulk Zincblende and Wurzite phases of GaN
based on an ensemble Monte Carlo calculation including a full zone band
structure”, J. of Applied Physics, 78, 1033, 1995.
[KOR 87] KORDE R., GEIST J., “Quantum efficiency stability of silicon photodiodes”,
Applied Optics, 26, 5284, 1987.
[KRI 98] KRISHNANKUTTY S., YANG W., NOHAVA T., RUDEN P.P., “Fabrication and
characterization of GaN/AlGaN ultraviolet-band heterojunction photodiodes”,
MRS Internet Journal of Nitride Semiconductor Research, 3, article 7, 1998.
[KUK 98] KUKSENKOV D.W., TEMKIN H., OSINSKY A., GASKA R., KHAN M.A.,
“Origin of conductivity and low-frequency noise in reverse-biased GaN p-n
junction”, Applied Physics Lett., 72, (11), 1365-1367, 1998.
[MAI 00] MAINWOOD A., “Recent developments of diamond detectors for particles
and UV radiation”, Semiconductor Science and Technology, 15, R55, 2000.
[MCI 00] MCINTOSH K.A., MOLNAR R.J., MAHONEY L.J., MOLVAR K.M., EFREMOWN.
JR., VERGHESE S., “Ultraviolet photon counting with GaN avalanche photodiodes”,
Applied Physics Lett., 76, 3938, 2000.
[MCK 87] MCKINLEY A.F., DIFFEY B.L., “A reference action spectrum for ultraviolet
induced erythema in human skin”, CIE Journal 6, 17, 1987.
[MON 99a] MONROY E., CALLE F., MUNOZ E., OMNES F., BEAUMONT B., GIBART P.,
“Visible-blindness in photoconductive and photovoltaic AlGaN ultraviolet
detectors”, J. of Electronic Materials, 28, 238, 1999.
220 Optoelectronic Sensors
[MON 99b] MONROY E., CALLE F., MUNOZ E., OMNES F., GIBART P., “High
UV/visible contrast photodiodes based on epitaxial lateral overgrown GaN layers”,
Electronics Lett., 35, 240, 1999.
[MON 99c] MONROY E., CALLE F., MUNOZ E., OMNES F., “AlGaN metal-
semiconductor-metal photodiodes”, Applied Physics Lett., 74, 3401,1999.
[MON 99d] MONROY E., CALLE F., MUNOZ E., BEAUMONT B., OMNES F., GIBART P.,
“High UV/visible contrast photodiodes based on epitaxial lateral overgrown GaN
layers”, Electronics Lett., 35, 1488, 1999.
[MON 00] MONROY E., CALLE F., PAU J.L., SANCHEZ F.J., MUÑOZ E., OMNES F.,
BEAUMONT B., GIBART P., “Analysis and modeling of Al
x
Ga
1-x
N-based Schottky
barrier photodiodes”, J. of Applied Physics, 88, 2081, 2000.
[MUN 00] MUÑOZ E., MONROY E., CALLE F., OMNES F., GIBART P., “AlGaN
photodiodes for monitoring the solar UV radiation”, Journal of Geophysical
Research, 105, 4865, 2000.
[NAM 97] NAM O.H., BREMSER M.D., ZHELEVA T., DAVIS R.F., “Lateral epitaxy of
low defect density GaN layers via organometallic vapor phase epitaxy”, Applied
Physics Lett., 71, 2638, 1997.
[OGU 97] OGUZMAN I.H., BELLOTTI E., BRENNAN K.F., KOLNIK J., WANG R., RUDEN
P.P., “Theory of hole initiated impact ionization in bulk zincblende and Wurtzite
GaN”, J. of Applied Physics, 81, 7827, 1997.
[OSI 97] OSINSKY A., GANGOPADHYAY S., GASKA R., WILLIAMS B., KHAN M.A.,
KUKSENKOV D., TEMKIN H., “Low noise p-ʌ-n ultraviolet photodetectors”, Appl.
Phys. Lett., 71, 2334, 1997.
[PAR 99] PARISH G., KELLER S., KOZODOY P., IBBETSON J.P., MARCHAND H., FINI P.T.,
FLEISCHER S.B., DEN BAARS S.P., MISHRA U.K., TARSA E.J., “High-performance
(Al,Ga)N-based solar-blind ultraviolet p-i-n detectors on laterally epitaxially
overgrown GaN”, Applied Physics Lett. 75, 247, 1999.
[POL 98] POLYAKOV A.Y., SMIRNOV N.B., USIKOV A.S., GOVORKOV A.V., PUSHNIY B.,
“Studies of the origin of the yellow luminescence band, the nature of nonradiative
recombination and the origin of persistent photoconductivity in n-GaN films”,
Solid State Electronics, 42, 1959, 1998.
[PUL 98] PULFREY D.L., NENER B.D., “Suggestions for the development of GaN-
based photodiodes”, Solid State Electronics, 42, 1731, 1998.
[RAZ 96] RAZEGHI M., ROGALSKI A., “Semiconductor ultraviolet detectors”, J. of
Applied Physics, 79, 7433, 1996.
Ultraviolet Photodetectors 221
[RHO 88] RHODERICK E.H., WILLIAMS R.H., Metal-Semiconductor Contacts,
Clarendon Press, Oxford, 1988.
[RUD 99] RUDEN P.P., “Materials theory based modeling for III-nitride devices”,
SPIE Proceedings, 3629, 174, 1999.
[SAN 00] SANDVIK P., WALKER D., KUNG P., K.MI, SHAHEDIPOUR F., KUMAR V.,
ZHANG X., J.DIAZ, C.JELEN, M.RAZEGHI, “Solar-blind AlxGa1-xN p-i-n
photodetectors grown on LEO and non-LEO GaN”, SPIE Proceedings, 3948, 265,
2000.
[SHE 99] SHEN B., YANG K., ZHAN L., CHEN Z., ZHOU Y., CHEN P., ZHAN R., HUANG
Z., ZHOU H., ZHENG Y., “Study of photocurrent properties of GaN ultraviolet
photoconductor grown on 6H-SiC substrate”, Japanese Journal of Applied
Physics, 38, 767, 1999.
[USU 97] USUI A., SUNAKAWA H., SAKAI A., YAMAGUCHI A., “Thick GaN epitaxial
growth with low dislocation density by hydride vapor phase epitaxy”, Japanese
Journal of Applied Physics, 36, L899, 1997.
[VAN 97] VAN HOVE J.M., HICKMAN R., KLAASSEN J.J., CHOW P.P., RUDEN P.P.,
“Ultraviolet-sensitive, visible blind GaN photodiodes fabricated by molecular
beam epitaxy”, Applied Physics Lett., 70, 2282, 1997.
[WAL 98] WALKER D., SAXLER A., KUNG P., ZHANG X., HAMILTON M., DIAZ J.,
RAZEGHI M., “Visible blind GaN p-i-n photodiodes”, Applied Physics Lett., 72,
3303, 1998.
[WAL 99] WALKER D., KUNG P., SANDVIK P., WU J., HAMILTON M., LEE I.H., DIAZ J.,
RAZEGHI M., “AlxGa1-xN p-i-n photodiodes on sapphire substrates”, Proc. SPIE,
3629, 193, 1999.
[WHI 96] WHITFIELD M.D., MCKEAG R.D., CHAN S.S.M., JACKMAN R.B., “A thin film
diamond photodiode for ultra-violet light detection”, Applied Physics Lett., 68,
290,1996.
[XU 97] XU G.Y., SALVADOR A., KIMW., FAN Z., LU C., TANG H., MORKOC H., SMITH
G., ESTES M., GOLDENBERG B., YANG W., KRISHNANKUTTY S., “High speed, low
noise ultraviolet photodetectors based on GaN p-i-n and AlGaN (p)-GaN(i)-
GaN(n) structures”, Applied Physics Lett., 71, 2154, 1997.
[YAN 98] YANG W., NOHAVA T., KRISHNANKUTTY S., TORREANO R., MCPHERSON S.,
MARSH H., “Back-illuminated GaN/AlGaN heterojunction photodiodes with high
quantum efficiency and low noise”, Applied Physics Lett., 73, 978,1998.
222 Optoelectronic Sensors
[YU 98a] YU L.S., QIAO D.J., XING Q.J., LAU S.S., BOUTROS K.S., REDWING J.M.,
“Analysis and modeling of Al
x
Ga
1-x
N-based Schottky barrier photodiodes”,
Applied Physics Lett., 73, 238, 1998.
[YU 98b] YU L.S., LIU Q.Z., XING Q.J., QIAO D.J., LAU S.S., REDWING J.D., “The role
of the tunneling component in the current-voltage characteristics of metal-GaN
Schottky diodes”, J. of Applied Physics, 84, 2099, 1998.
Chapter 7
Noise in Photodiodes and
Photoreceiver Systems
In systems transmitting information by means of photons, the
recovery of a useful signal must necessarily involve an optical-
electrical (photon-electron) transformation. This transformation,
which makes use of an optoelectronic system, is, as with any system,
fundamentally limited by the background noise. The aim of this
chapter is to discuss the origin of this noise and also to study various
means of minimizing it in order to obtain the best signal-to-noise ratio,
a value which influences the quality of the optical detection. In section
7.1 we will recall the basics of signal analysis, leading to the
frequency analysis of the power of random signals. After recalling the
different sorts of noise associated with the most relevant physical
processes (sections7.2 and 7.3), in section 7.4 we will present the
main results concerned with the analysis of noisy electrical circuits.
Section 7.5 is dedicated to the study of noise in photodetectors, which
we choose to limit to PIN and avalanche photodiodes. Section 7.6
discusses the problems of background noise associated with a
photodiode-amplifier pair forming the photoreceiver, and lays out the
methods of obtaining the best signal-to-noise ratio. Finally, in order to
Chapter written by Robert ALABEDRA and Dominique RIGAUD.
224 Optoelectronic Sensors
illustrate these techniques, we present the complete background noise
calculation for a PIN-FET photoreceiver.
7.1. Mathematical tools for noise
Before discussing the treatment of random signals, we will recall a
summary of the treatment of known signals in an attempt to justify a
frequency approach as compared to a temporal approach for the
treatment of the power or the energy of random signals.
7.1.1. Known signals with finite energy or power
All systems are governed by a differential equation in the time
domain, and we will characterize them by their impulse response h(t).
The output signal s(t) for an input signal i(t) can be calculated in the
time domain either by solving this differential equation or by
calculating the convolution of h(t) with i(t):
s t
()
=i t
()
* h t
()
= e W
( )
h t ÷W
( )
dW =
·
÷·
¹
h W
( )
e t ÷W
( )
dW
·
÷·
¹
[7.1]
The signal s(t) has the same dimensions as i(t) which implies that
h(t) has the dimensions of inverse time. Thus, in the case of an R–C
circuit, the impulse response is:
h t
()
=
1
RC
exp ÷
t
RC
(
(
·
·
·
·
)
)

in s
1
[7.2]
In order to avoid calculating the convolution, we use Fourier
analysis which transforms the convolution in the time domain to a
simple product in the frequency domain, and we can recover the
original s(t) through the inverse transform. On the other hand, if we
Noise in Photodiodes and Photoreceiver Systems 225
want to analyze the same system from the point of view of energy or
power, where phase information no longer has any meaning, we will
then use the autocorrelation function R
s
(W) of the signal. The
relationship between R
i
(W), the autocorrelation of the input signal, and
R
s
(W), that of the output signal, is:
R
s
W
( )
= h* ÷W
( )
* h W
( )
* Re W
( )
[7.3]
where h*(W) is the complex conjugate of h(W).
This expression will only occasionally be used, given its double
convolution product. We will repeat this same analysis in terms of
frequency, using the concept of spectral density and the Wiener-
Khinchin theorem which shows that the spectral density G
x
(f ) of a
signal x(t) is the Fourier transform of the autocorrelation R
x
(W) of x(t)
such that:
G
x
f
( )
=T F R
x
W
( )

l
l
l
= R
x
W
( )
exp ÷j2S f W
( )
dW
·
÷·
¹
[7.4]
and its inverse
R
x
W
( )
=T F
1
G
x
f
( )

l
l
l
= G
x
f
( )
exp j2S f W
( )
df
·
÷·
¹
[7.5]
If we take W = 0, which is the maximum of the autocorrelation
function, R
x
(0) represents the energy or the power of the signal. We
find:
R
x
0
( )
= G
x
f
( )
df
·
÷·
¹
[7.6]
226 Optoelectronic Sensors
G
x
(f ) represents the energy or power distribution of the signal in
the frequency domain. We take the Fourier transform of equations
[7.1] to [7.3]:
F T R
s
W
( )
= h* ÷W
( )
* h W
( )
* Re W
( )

l
l
l
[7.7]
which can be written as:
G
s
f
( )
= H f
( )
H * f
( )
G
i
f
( )
[7.8]
or:
G
s
f
( )
= H f
( )
2
G
i
f
( )
[7.9]
and which is much more easy to use than equation [7.3].
What we need to do now is to work out how to transfer what we
have already set out for deterministic signals to the random signals
which represent the sources of noise.
7.1.2. Random signals and background noise
In the case of random signals, we have no choice but to treat them
in terms of power or energy, and of course without any consideration
of their phase. As we did for the known signals in section 7.1.1, we
will use the autocorrelation function, the spectral power density and
the same Wiener-Khinchin theorem. This will lead us to define a class
of random signals which, fortunately for us, will cover the majority of
practical situations. This class is characterized by ergodicity and time-
stability.
DEFINITION OF ERGODICITY – The average resulting from many
simultaneous experiments is the same as that obtained by the temporal
mean across a single experiment.
Noise in Photodiodes and Photoreceiver Systems 227
DEFINITION OF TIME STABILITY – In its strict sense, this means that
the mean value is independent of time and the autocorrelation function
only depends on W = t
1
– t
2
.
In this case, we obtain for the random signal x(t):
R
x
W
( )
= E x t
()
x t ÷W
( )

l
l
l
[7.10]
where E is the expected value associated with the probability density
of the process under study (Poisson, Gaussian systems, etc.). Under
these conditions the autocorrelation function is practically analogous
to that for known signals which have the same properties. Now, using
this formalism, we can treat the random signals appearing in the
expression in terms of energy or power, and calculate the signal-to-
noise ratio which characterizes the receiving quality.
7.2. Fundamental noise sources
Here we are interested in noise generated by solid devices in
electronic circuits. As we have seen in section 7.1, we will remain in
the frequency domain in order to avoid convolving the input and
output autocorrelation functions. We will use the spectral densities as
our measurement of scale, and in order to recover a power or an
energy we will integrate these spectral densities over the useful
bandwidth. These spectral densities are expressed in A
2
/Hz or V
2
/Hz.
7.2.1. Thermal noise
At the terminals of a metallic resistance of value R at temperature T
in degrees Kelvin, the Brownian motion of electrons produces a
randomly-varying voltage v(t) at its open-circuit terminals, with a
mean value of zero. Bearing in mind the large number of electrons (of
the order of 10
22
cm
–3
), the voltage v(t) obeys a centered Gaussian law.
For frequencies below the inverse of the mean time between two
228 Optoelectronic Sensors
collisions (W
r
§ 10
–12
s), the spectral density associated with this
random signal can be written in the basis of positive frequencies as:
S
v
f
( )
= 4 kTR V
2
Hz
( )
[7.11]
where k is the Boltzmann constant (1.38 u 10
–23
J.K
–1
). We note that
this spectral density is independent of the frequency (white noise).
7.2.2. Shot noise
Shot noise stems from the discrete nature of electrons and their
random emission and collection over time. Thus, it takes the form of a
random variable governed by the Poisson distribution law. Frequency
domain analysis of this process leads to a spectral noise current
density (short circuit current) of:
S
i
f
( )
= 2q I A
2
Hz
( )
[7.12]
where q is the elementary charge (q = 1.6 u 10
–19
C). Here also we
find a white noise as long as we ignore the transit time of the carriers.
Shot noise is always encountered when the current is caused by
carriers emitted from a reservoir or across a potential energy barrier
greater than kT. This is the case for reverse-biased junctions. For PIN
photodiodes, for example, we have:
S
i total
f
( )
= 2 q I
dark
÷I
ph

l
l
l
[7.13]
where I
dark
is the intrinsic dark current of the photodiode and I
ph
the
photocurrent due to absorbed photons and representing the useful
information. For modern devices, I
dark
<< I
ph
and the noise is mainly
due to the photocurrent. In contrast, for detection of low light
intensities, the dark current will make a significant contribution to the
signal-to-noise ratio.
Noise in Photodiodes and Photoreceiver Systems 229
7.2.3. Multiplication noise
In a similar way to vacuum photomultipliers, avalanche
photodiodes with internal gain are based on the process of
multiplication through the impact of carriers in the space charge
region of the photodiode. In the case of vacuum photomultipliers,
there is only one type of carrier: electrons locally multiplied from one
dynode to the next. If we ignore the multiplication inertia and the
transit time of the electrons, the noise generated by this
photomultiplier is amplified shot noise. The variance in the total
multiplication factor is:
M
2
= M
2
[7.14]
and for large enough M we have:
S
i
f
( )
= 2qI
pho
M
2
[7.15]
where M is the mean value of the multiplication factor and I
pho
is the
primary photocurrent. Setting:
M =
I
total
I
pho
[7.16]
we have:
S
i
f
( )
=2qI
pho
M
2
= 2qI
pho
I
total
2
I
pho
2
[7.17]
in other words, we have a noise proportional to the total current
squared. In the case of semiconductor avalanche photodiodes, the
multiplication process involves both types of carrier across the whole
multiplication region, i.e. in the high electric field region in the space
charge region. Numerous theoretical approaches have been proposed
in the literature. To summarize these approaches, we recall that the
230 Optoelectronic Sensors
two types of carrier are characterized by their ionization coefficient (in
cm
–1
) D for electrons and E for holes, and that these depend on the
physical characteristics of the semiconductor materials. The important
parameter is the ratio k =
D
E
. The ideal material would be one in
which only one of the two types of carrier could be ionized: either
electrons, in which case E = 0 and k = f, or holes, in which case D = 0
and k = 0. The noise characteristics would then approach that of a
vacuum photomultiplier.
For avalanche photodiodes, the expression for the spectral noise
density can be written in the general form:
S
i
f
( )
=2qI
pho
M
x
[7.18]
where x lies between 2 and 4.
Taking the noise of a vacuum photomultiplier as our reference, we
define the excess noise factor F as:
F M
( )
=
M
2
M
2
=
M
x
M
2
= M
x 2
[7.19]
in other words as the ratio of the variance of the multiplication process
in the semiconductor to the variance of the same process in the
vacuum tube. In the case of an initial injection of electrons, and for
values of k close to 1, we can approximate F M
( )
by the expression:
F M
( )
= k M ÷ 2÷ M
1 (
(
·
·
·
)
)

1÷k
( )
[7.20]
In particular, for k = 1 (D = E), as in germanium for example, we
have F M
( )
= M and:
Noise in Photodiodes and Photoreceiver Systems 231
S
i
f
( )
=2qI
pho
M
3
[7.21]
For silicon avalanche photodiodes, D is 50 times larger than E. For
multiplication initiated by electrons, we arrive at the expression:
S
i
f
( )
=2qI
pho
M
2 2
[7.22]
which leads to an excess noise factor of:
F M
( )
= M
0 2
[7.23]
In the limiting case where only electrons are responsible for
multiplication (E = 0), the spectral density must approach that of a
vacuum photomultiplier and:
F M
( )
= M
0
=1 [7.24]
However, multiplication initiated by the less ionizing carriers is of
no practical interest. In this case, we find excess noise factors F(M) of
the order of M
2
, which gives:
S
i
f
( )
=2qI
pho
M
2
F M
( )
=2qI
pho
M
4
[7.25]
These results are all depicted in Figure 7.1.
In order to achieve behavior for solid avalanche devices
approaching that of vacuum photomultipliers, several devices have
been conceived based on engineering of the band structure with the
help of heterojunctions. The aim was to obtain impact ionization by
electrons only. Despite the sophistication of the technologies used, the
results did not live up to the performances initially promised, because
the interfaces resulted in an additional 1/f noise.
232 Optoelectronic Sensors
Figure 7.1. Characteristic plots of the variations in the spectral noise density for
multiplication in avalanche photodiodes as a function of the multiplied current
7.3. Excess noise
In addition to the sources of white noise, there are two other
sources of noise which depend on frequency. These are generation-
recombination (GR) noise and scintillation noise, which is referred to
as the 1/f noise. These two sources of noise are referred to together as
excess noise, as compared to the fundamental sources which are
thermal noise and shot noise.
7.3.1. Generation-recombination noise
This noise originates in the fluctuations of the number of carriers,
linked to the presence of different trapping states in the bandgap of
semiconductors. The spectral density associated with these
fluctuations can be written as:
Noise in Photodiodes and Photoreceiver Systems 233
S
N
f
( )
= 'N
2
4W
1÷Z
2
W
2
[7.26]
where 'N
2
is the second order moment of the random variable N
which represents the number of carriers under consideration. W is the
time constant of the generation-recombination process taking place.
This spectrum, of Lorentzian form, displays a plateau at low
frequencies followed by a 1/f
2
dropoff at high frequencies. These
fluctuations lead to current fluctuations proportional to the square of
the mean current:
S
i
f
( )
=
'N
2
N
2
I
2
4W
1÷Z
2
W
2
A
2
Hz
( )
[7.27]
Generally, the observed time constants lie between 10
–3
and 10
6
s.
However, it should be noted that in some specific cases, these values
can be smaller.
7.3.2. 1/f noise
The origin of this noise is poorly understood. Its spectral density
takes the form:
S
i
f
( )
=
A
f
J
I
2
A
2
Hz
( )
[7.28]
where J usually takes a value close to 1, which leads to the term 1/f
noise, and A is an experimentally-determined constant which varies
from one device to another. We note that the spectral density varies as
the square of the bias current. It is not possible to calculate the
autocorrelation function of the process by the use of the inverse
Fourier transform of S
i
(f) because the integral diverges. There are two
interpretations of the origin of this noise. The first was proposed by
234 Optoelectronic Sensors
McWhorter [MCW 57] who attributed the origin of this noise to the
surfaces and interfaces of the material where a continuum of trapping
states for carriers can be found (surface effects), a situation which
produces an overlapping series of generation-recombination events.
The second interpretation is due to Hooge [HOO 76], who attributes
this noise source to a volume effect in materials, due to fluctuations in
the mobility of the carriers. Whichever of these is correct, it is almost
certain that the amplitude of this 1/f noise is affected by the quality of
the device and its fabrication technology. The frequency f
c
is defined
as that for which the spectral density of the 1/f noise is equal to that of
the fundamental noise (shot noise or thermal noise).
For the rest of this chapter, the main sources of noise that we will
take into account when evaluating the noise characteristics of a
photoreceiver are thermal noise, shot noise and its potential
amplification through multiplication processes, and the 1/f noise when
the optical signal requires a treatment that includes low frequencies. In
practical terms, Table 7.1 summarizes the expressions for the spectral
density for the different sources of noise involved with devices and
systems.
Table 7.1. Expressions and characteristics of the main noise sources
Noise in Photodiodes and Photoreceiver Systems 235
7.4. Analysis of noise electrical circuits
As is the case for deterministic signals, the behavior of a system is
described in terms of its equivalent circuit, with which a certain
number of noise sources must be associated. These sources are
represented using the Thevenin or Norton formalisms. In contrast to
generators of deterministic signals, noise generators will be
characterized by the spectral density associated with the quantities that
they represent. In this way we can avoid all the problems of temporal
analysis while retaining the concept of impendence or admittance as
traditionally defined in circuit analysis. The concepts presented below
will be used later in the evaluation of the noise and the signal-to-noise
ratio of photoreceiver systems.
7.4.1. Representation of noise in bipoles
A real bipole (for example a photodetector) is thus represented by a
random voltage generator (Thevenin), which has a spectral density
S
v
(f ) in series with its impedance Z(f ), which is assumed not to be
noisy; or by a random current generator (Norton) with spectral density
S
i
(f ) in parallel with its admittance Y f
( )
=
1
Z f
( )
.
These representations are illustrated in terms of electrical circuits
in Figure 7.2.
Figure 7.2. Representation of a noisy bipole using a frequency
spectrum approach for random processes
236 Optoelectronic Sensors
Henceforth, all noise generators will be marked with hatching in
order to distinguish them from normal signal generators. The
equivalence of these two representations can be expressed in the
following equation:
S
V
f
( )
= Z f
( )
2
S
i
f
( )
[7.29]
which connects the two spectral densities that we have used.
Other parameters can be used to characterize the noise of a bipole.
If we take the example of shot noise, we can define the equivalent
noise current I
eq
as:
I
eq
=
S
i
f
( )
2q
[7.30]
In the case of thermal noise, we can define:
– either the equivalent resistance R
eq
of the noise:
R
eq
=
S
v
f
( )
4kT
[7.31]
– or the equivalent conductance G
eq
of the noise:
G
eq
=
S
i
f
( )
4kT
[7.32]
– or the equivalent temperature T
eq
of the noise:
T
eq
=
S
v
f
( )
4kRe Z f
( ) ¦ ¦
=
S
i
f
( )
4kRe Y f
( ) ¦ ¦
[7.33]
The representations of the noise given in Figure 7.2 make it
possible to directly treat series or parallel combinations of noisy
Noise in Photodiodes and Photoreceiver Systems 237
bipoles: two bipoles in series are equivalent to a single bipole whose
spectral density of voltage noise can be written:
Sv
t
f
( )
= Sv
1
f
( )
÷Sv
2
f
( )
÷2 Re Sv
1
v
2
f
( ) ¦ ¦
[7.34]
where Sv
1
v
2
(f ) represents the cross-spectrum which encapsulates the
statistical dependence of the two random processes associated with the
two noise sources. This dependence can be represented in terms of a
correlation coefficient defined by:
C f
( )
=
Sv
1
v
2
f
( )
Sv
1
f
( )
Sv
2
f
( )
[7.35]
with 0 ” |C| ” 1. Thus, Sv
T
(f ) can then be written:
Sv
T
f
( )
= Sv
1
f
( )
÷Sv
2
f
( )
÷2 Re C f
( )
Sv
1
f
( )
Sv
2
f
( )
¦ ¦
[7.36]
In the normal case where the two sources of noise are uncorrelated
(C = 0), the resultant spectral density is then simply the sum of the
spectral densities associated with each bipole. In a similar manner, in
the case of a parallel combination of two bipoles, we have, using the
parallel representation:
Si
T
f
( )
= Si
1
f
( )
÷Si
2
f
( )
÷2 Re C f
( )
Si
1
f
( )
Si
2
f
( )
¦ ¦
[7.37]
7.4.2. Representation of noise in quadripoles
For quadripoles (the case of amplifiers placed downstream of
photodetectors), the noise is generally represented by a minimum of
two partially-correlated noise sources. These sources, of voltage
and/or current, can be placed at the input and/or output of the
quadripole. This leads to six equivalent representations, each one
associated with one of six matrices describing the properties of the
238 Optoelectronic Sensors
quadripole. In practice two of these representations are the most
commonly used (see Figure 7.3).
Figure 7.3. The two most commonly-used configurations for
representation of a noisy quadripole
In the first case, the noise is represented by a voltage generator and
a current generator both present at the input of the quadripole (which
is itself assumed noise-free), with spectral densities S
en
(f ) and S
in
(f ).
This representation is well suited to describe noise in amplifiers. In the
second case, two current generators are used, one at the input and one
at the output of the quadripole, with spectral densities Si
1
(f ) and Si
2
(f ).
This representation is useful for describing the intrinsic noise of
bipolar and field effect transistors.
An important quantity for characterizing the noise of a system is its
noise factor, which is defined as the quotient of the input and output
signal-to-noise ratios:
F f
( )
=
S B
( )
e
S B
( )
s
[7.38]
This equation shows that the noise factor can also be written as the
ratio of the total output noise power to what the output noise power
would be if the quadripole was noise-free. In this way it can be shown
that the noise factor is always greater than 1, and that the presence of
the quadripole amplifier degrades the signal-to-noise ratio. Bearing in
mind the first representation given in Figure 7.3a, and assuming a
resistive load R
g
at the input to the quadripole which only generates
thermal noise, we have, for a given frequency f,
Noise in Photodiodes and Photoreceiver Systems 239
F f
( )
=1÷
S
en
f
( )
÷ R
g
2
S
in
f
( )
÷2R
g
Re S
enin
f
( ) ¦ ¦
4kTR
g
[7.39]
As R
g
varies, the noise factor passes through a minimum for the
value:
R
gopt
=
S
en
f
( )
S
in
f
( )
Finally, for a series of quadripoles, it can be shown that under
certain conditions the noise factor of the resultant quadripole can be
written as a function of the individual noise factors F
i
and the power
gains G
i
:
F ÷1= F
1
÷1÷
F
2
÷1
G
1
÷
F
3
÷1
G
1
G
2
÷... [7.40]
This equation (the Friis formula) shows that the first quadripole
controls the noise of the whole system as long as it is not very noisy
and its gain is significant. Thus, when designing a photoreceiver, it
will be necessary to match the photodetector to the first amplifying
component in order to minimize its noise factor and give it a sufficient
gain that the noise of the successive stages can be ignored. These
issues will be covered when we study the PIN-FET system.
7.5. Noise in photodetectors
In order to describe the noise of a photodetector in the presence of
a weak signal, we will need to evaluate a signal-to-noise ratio for the
information carried by the photons. In order to do this, we need to
define a number of optoelectronic quantities and characteristic
parameters of photodetectors, which will allow us to express the noise
associated with the photocurrent generated by the conversion of
photons into electron-hole pairs.
240 Optoelectronic Sensors
7.5.1. Characteristic parameters
In all optoelectronic transducers, there is an electric field present
which separates the two types of carriers generated by absorption of
the energy E
ph
= hQ of the incident photons, of wavelength O. For this
generation to take place, we require that:
E
ph
= hv =
hc
O
_ E
g
[7.41]
where h is the Planck constant, v =
c
O
is the optical frequency and E
g
the
size of the bandgap of the semiconductor material. For hQ = E
g
we
define the cutoff wavelength O
C
given by:
O
c
=
1.24
E
g
in Pm with E
g
in eV
( )
[7.42]
The response coefficient R of the detector allows us to relate the
generated photocurrent I
ph
to the incident optical power P
opt
:
R =
I
ph
P
opt
A W
( )
The calculation of R involves the external quantum efficiency K
ext
and the wavelength of the optical signal. We have:
R =K
ext
qO
hc
=0.8K
ext
O A W
( )
[7.43]
if we express O in μm and c in m/s. In calculations of noise in
photodetectors, only knowledge of the photoresponse R is required.
The photodetector exhibits a basic noise S
i
(f) with which the dark
current is associated:
Noise in Photodiodes and Photoreceiver Systems 241
I
obs
=
S
i
f
( )
2q
We will now consider the incident optical power which would
produce a photocurrent I
ph
such that the associated shot noise 2qI
ph
would be equal to the basic noise referred to above (a signal-to-noise
ratio of 1). This optical power characterizes the photodetector and is
designated the NEP (Noise Equivalent Power). The NEP represents
the smallest optical power which can be detected for a given noise
equivalent bandwidth associated with the electronic system. Based on
the definition of the response coefficient, we then have, for linear
photodetectors and a unit bandwidth:
R =
I
ph
P
opt
=
I
obs
NEP
[7.44]
For a given noise equivalent bandwidth, an identical equation can
be written in terms of rms values. Thus, we can directly express the
signal-to-noise ratio in the form:
S
B
(
(
·
·
·
·
)
)

dB
=10log
P
opt
rms
( )
NEP
[7.45]
We can also define another characteristic parameter of
photodetectors, the detectivity, which is the inverse of the NEP:
D =
1
NEP
W
1
( )
In the case of a photoreceiver, we will keep in mind when defining
the NEP that the basic noise used as a reference consists of the basic
noise of the photodetector and the basic noise of the amplification
system, brought down to the level of the photodetector. For a
bandwidth 'f, we therefore have:
242 Optoelectronic Sensors
NEP =
S
i
f
( )
total
'f
R
[7.46]
where S
i
(f )
total
= S
i
(f )
photodetector
+ S
i
(f )
amplifier.
7.5.2. PIN photodiodes
These are diodes, of mesa or planar structure, in which the n and p
regions are separated by an intrinsic semiconductor region where the
photocarriers are generated and separated (see Figure 7.4).
Figure 7.4. Schematic cross-sections of PIN photodiodes using
(a) mesa and (b) planar technology
The most widely used semiconductor materials are shown in Table
7.2 along with their main characteristics.
Noise in Photodiodes and Photoreceiver Systems 243
Table 7.2. Main characteristics of the most widely used semiconductor materials
A theoretical approach to these devices must take into account the
band structure (direct or indirect) and optimize its geometric
dimensions as a function of the bandwidths, operating temperature,
dark currents, carrier lifetimes, leakage currents, series resistances, etc.
Figure 7.5 shows the ideal I = f (V) characteristics of a photodiode
at different incident light powers.
Figure 7.5. Typical characteristics of a PIN photodiode under different illuminations
Under reverse-bias, which is the normal mode of operation of a
PIN photodiode, the conduction consists of two components: the dark
current I
dark
and the photocurrent I
ph
= RP
opt
. Two shot noise
generators are associated with these two currents, and taking into
244 Optoelectronic Sensors
account the series resistance R
s
which is the source of the thermal
noise, the equivalent noise circuit is shown in Figure 7.6.
Figure 7.6. Equivalent noise circuit for a reverse-biased
PIN photodiode under illumination
Starting from this equivalent circuit, we can calculate the intrinsic
NEP of the device from equation [7.46]. For the large bandgap
materials used to detect short wavelengths, the dark current is
negligible, but technical issues are associated with the metal-
semiconductor contacts due to the very high resistivity of these
materials. For small bandgap materials, which are used for infrared
detection, the dark current will be the main limitation on the operation
of the device at ambient temperature.
7.5.3. Avalanche photodiodes
Avalanche photodiodes are PIN photodiodes which make use of
the physics of impact ionization to achieve multiplication through the
effect of the electric field on the charge carriers, with a mean gain M.
This process generates noise as discussed in section 7.2.3. Figure 7.7
shows a cross-section through two avalanche photodiodes.
Noise in Photodiodes and Photoreceiver Systems 245
Figure 7.7. (a) Structure of a n
+
pʌp
+
silicon avalanche photodiode for detection
at O = 0.9 μm. (b) Heterostructure avalanche photodiode with separate absorption
and multiplication regions, for detection at O = 1.6 μm
The equivalent noise circuit is the same as for a PIN photodiode,
but the photocurrent is written as:
I
ph
= MRP
opt
A
( )
For the shot noise, the spectral density of the noise current can at
first approximation be written as:
S
i
f
( )
=2q RP
opt
÷ I
dark
¦ ¦
M
2
F M
( )
A
2
H
2
( )
[7.47]
In general, the I
dark
component is multiplied by a factor M, which
is different from that for the photocurrent because we do not have
control over the injection into the multiplication region of the carriers
originating from the dark current.
7.6. Noise optimization of photodetectors
The main aim when designing a photoreceiver is to minimize the
optical power that needs to be detected while maintaining a specified
signal-to-noise ratio. Achieving this optimal sensitivity depends on the
246 Optoelectronic Sensors
choice of photodetector and the amplification sequence. Specifically,
at the level of the detector, the signal-to-noise ratio can be written:
S B
( )
detector
=
i
signal
2
Si f
( )
detector
df
'f
¹
[7.48]
which in the presence of a preamplifier becomes:
S B
( )
receiver
=
i
signal
2
Si f
( )
detector
÷Si f
( )
preamp
¦ ¦
df
'f
¹
[7.49]
Thus, the signal-to-noise ratios of the receiver and the detector are
connected by the relationship:
S B
( )
receiver
=
1
F
S B
( )
detector
[7.50]
where F is the noise factor of the preamplifier, whose input load is the
photodetector. In order to achieve the minimum degradation in the
signal-to-noise ratio of the photoreceiver, the amplifier must be
matched to the detector in order to minimize F.
7.6.1. Formulation of the problem
Bearing in mind section 7.4, Figure 7.8 shows the equivalent noise
circuit of the detector-preamplifier system.
The photodetector is represented by its admittance Y
s
in parallel
with a noise current generator I
s
. The preamplifier noise is represented
Noise in Photodiodes and Photoreceiver Systems 247
at its input by a voltage generator e
n
and a current generator i
n
. These
two generators are correlated, and we write i
n
in the form of one part
which is correlated with e
n
(which we will call i
c
), and another
uncorrelated part (i
nc
):
i
n
f
( )
= i
c
f
( )
÷i
nc
f
( )
[7.51]
Figure 7.8. Equivalent noise circuit for a photoreceiver,
positioned at the system input
Since the correlated part is proportional to e
n
at a given frequency,
we will write it in terms of a correlation admittance Y
cor
:
i
c
f
( )
=Y
cor
f
( )
e
n
f
( )
[7.52]
The equivalent circuit in Figure 7.8b can then be drawn in the form
shown in Figure 7.9.
248 Optoelectronic Sensors
Figure 7.9. Equivalent circuit used to calculate the spectral density
of the total noise current of the photoreceiver
Bearing in mind our frequency-based approach to random
phenomena, we have in terms of spectral densities:
S
i
6
f
( )
= S
i
s
f
( )
÷ Y
s
÷Y
cor
2
S
en
f
( )
÷S
i
nc
f
( )
[7.53]
The different spectral densities can be expressed as a function of
their equivalent noise conductances (see section 7.4):
S
i
s
f
( )
=4kTG
s
f
( )
S
en
f
( )
=4kTR
n
f
( )
= 4kTG
n
1
f
( )
S
i
nc
f
( )
=4kTG
nc
f
( )
S
i
n
f
( )
=4kTG
i
n
f
( )
'
¦
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
[7.54]
To avoid encumbering the notation, from hereon in we will not
bother indicating that the conductances are a function of frequency.
The noise factor of the photoreceiver now be written:
F =
S
i
6
f
( )
S
i
s
f
( )
=1÷
1
G
s
G
nc
÷ R
n
G
s
÷G
cor
( )
2
÷ B
s
÷ B
cor
( )
2

l
l
l
l
'
|
1
1
+
1
1
'
¦
1
1
+
1
1
[7.55]
Bearing in mind equations [7.51] and [7.52], we have:
G
nc
=G
in
÷R
n
Y
cor
2
[7.56]
Noise in Photodiodes and Photoreceiver Systems 249
and the noise factor can be written as:
F =1÷
1
G
s
G
in
÷R
n
G
s
2
÷ B
s
2
÷2 G
s
G
cor
÷ B
s
B
cor
( )
¦ ¦ ¦ ¦
[7.57]
In order to minimize the noise factor, we set to zero the derivative
of F with respect to the variables G
s
and B
s
. We find:
B
sopt
=÷B
cor
[7.58]
then:
G
sopt
= G
cor
2
÷G
n
G
nc
= G
in
G
n
÷B
cor
2
[7.59]
Under these conditions, the minimum noise factor can be written:
F
min
=1÷2
G
sopt
÷G
cor
G
n
[7.60]
It follows that:
F = F
min
÷
1
G
s
G
n
G
s
÷G
sopt
( )
2
÷ B
s
÷B
sopt
( )
2 '
|
1
1
+
1
1
'
¦
1
1
+
1
1
[7.61]
Hence, if we constrain F to be greater than or equal to F
min
by
modifying the source admittance Y
s
= G
s
+ jB
s
, we find we must solve
the following equation in terms of G
s
and B
s
:
G
s
2
÷ B
s
2
÷2G
s
G
sopt
÷ F ÷F
min
( )
G
n
¦ ¦
÷2B
s
B
sopt
÷G
sopt
2
÷ B
sopt
2
=0 [7.62]
which in the basis G
s
, B
s
represents a circle with center:
250 Optoelectronic Sensors
G
so
=G
sopt
÷ F ÷F
min
( )
G
n
B
so
= B
sopt
'
|
1
1
1
+
1
1
1
1
[7.63]
and radius:
R = F ÷F
min
( )
G
n
F ÷F
min
( )
G
n
÷2G
sopt
¦ ¦
[7.64]
Figure 7.10. (a) Circles of constant noise factor in the basis G
s
, B
s
.
(b) Plot of the surface defined by the noise factor
We note that for F = F
min
the circle shrinks to its center such that:
G
so
=G
sopt
B
so
=G
sopt
The results obtained are shown in Figures 7.10a and 7.10b.
It must be remembered that the series of circles and their resultant
surface is only representative of a single frequency. The matching of
the detector and preamplifier must therefore be as independent as
possible of this parameter.
Noise in Photodiodes and Photoreceiver Systems 251
7.6.2. Concepts for photodetector-transistor matching
A conceptual diagram for the matching is shown in Figure 7.11.
Figure 7.11. Conceptual diagram for matching a photodetector to its
preamplifier in order to minimize the noise factor
Y
s
is the admittance presented by the matching quadripole at the
input of the preamplifier. In order to have F = F
min
, it is necessary that
Y
s
= Y
opt
. Matching is achieved using a quadripole or matching
transformer and/or through a matching admittance Y
A
. Y
d
is the
admittance of the photodetector. For example, a photodiode with a
capacitance C
d
of 0.1 pF and a series resistance R
s
of 20 : presents an
admittance of Y
d
= 0.03 mS + j 1.2 mS at 2 GHz and Y
d
= 0.78 mS + j
6.3 mS at 10 GHz.
Similarly, for a field effect transistor with a gate of width 200 μm
we have, at 2 GHz: Y
sopt
= 2.5 mS – j 2.1 mS, and at 10 GHz: Y
opt
=
12.8 mS – j 10.8 mS. Thus, in the complex plane of admittances, the
matching of the photodetector to its transistor move from the point Y
D
to the point Y
opt
(see Figure 7.12a).
252 Optoelectronic Sensors
Figure 7.12. (a) Representation in the admittance plane of the photodetector’s
Y
d
and of Y
opt
which leads to a minimum of the noise factor, (b) movement due
to the matching admittance Y
a
= G
a
– jB
a
, (c) matching by lossless admittance
and reduction in the width W of the gate, (d) matching by use of a transformer
In order to do this, we can imagine the following points:
a) connect the photodetector directly to the transistor. The noise
factor will not be minimal. However, we will be close enough to the
axis G of the conductances that C
d
will be weak;
b) place an admittance Y
a
= G
a
+ jb
a
in parallel with Y
d
so that Y
a
=
Y
opt
+ Y
d
(see Figure 7.12b). The noise factor will be minimized, but at
the expense of increasing the source noise. It follows that the signal-
to-noise ratio is degraded: increasing the real part of the admittance
seen by the transistor increases the noise and so is not a worthwhile
solution;
c) reduce the width W of the gate of the field effect transistor in
order to have R
e
{Y
opt
} = R
e
{Y
d
} [I think that R
e
should read Re [i.e.
“the real part of”]] (Figure 7.12c). The signal-to-noise ratio of the
preamplifier is optimized. It should however be noted that the
reduction in W leads to a reduction in the transconductance of the gain.
This may then not be large enough for it to be possible to neglect the
noise of the subsequent stage;
Noise in Photodiodes and Photoreceiver Systems 253
d) use an impedance transformer (Figure 7.12d). Y
d
becomes
K
e
{Y
d
+ jB
a
} = Y
opt
. In this case, it is also necessary to match the
imaginary part by using a lossless admittance jB
a
. The real part of Y
s
has been amplified along with the signal, and the signal-to-noise ratio
of the source remains unchanged. In practice, this requires being able
to build a lossless (and hence noiseless) matching circuit with a
significant K factor. It also requires that the lossless load Y
a
has a
negative susceptance (inductance). Finally, the system must operate in
the useful frequency range, bearing in mind that the admittance of an
inductance decreases with frequency when I
m
{Y
opt
} grows. This
makes this option challenging to achieve in practice.
7.7. Calculation of the noise of a photoreceiver
7.7.1. Basic equations
We need to express the equivalent noise circuit in Figure 7.8 in
terms of the basic noise associated with the transistor which forms the
preamplifier. The behavior of the transistor will be described here in
terms of its admittance matrix Y
ija
, and its background noise will be
represented by two correlated current generators i
b1
and i
b2
at the input
and output of the device respectively (section 7.5.2). This leads to the
arrangement depicted in Figure 7.13.
Figure 7.13. Equivalent circuit for the preamplifier,
highlighting the different sources of noise
The calculation of the noise present at the input is made by
considering the short-circuit output, and this does not detract from the
254 Optoelectronic Sensors
generality of the problem since the current gain will also be calculated
under these same conditions.
We set:
Y
11
=Y
s
÷Y
11a
[7.65]
and for V
2
= 0 we have:
i
2
=i
b2
÷Y
21a
V
1
V
1
=
1
Y
11
i
s
÷i
b1
( )
'
¦
1
1
1
1
1
+
1
1
1
1
1
[7.66]
given that the output short-circuit current gain can be written:
G
i
cc
)
v
2 0
=
y
21a
Y
11
[7.67]
we find that the spectral density of the noise current associated with
current i
2
is:
S
i
2
f
( )
= S
i
b2
f
( )
÷ G
icc
2
S
is
f
( )
÷S
ib1
f
( )
÷2 Re G
icc
S
i
b1
i
b2
f
( )
¦ ¦ ¦ ¦
[7.68]
and so the total noise at the input to the transistor can be written:
S
i
e
f
( )
= S
i
s
f
( )
÷S
i
b1
f
( )
÷
1
G
i
cc
2
S
i
b2
f
( )
÷2 Re
1
G
i
cc
*
S
i
b1
i
b2
f
( )
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
[7.69]
Setting Y
s
=G
s
÷ jB
s
and S
i
b1
i
b2
f
( )
Y
21
a
=G
cor
÷ jB
cor
, we have:
Noise in Photodiodes and Photoreceiver Systems 255
S
ie
f
( )
= S
is
f
( )
÷S
i
b1
f
( )
÷
G
11a
÷G
s
( )
2
÷ B
11a
÷ B
s
( )
2
Y
21a
2
S
i
b2
f
( )
[7.70]
÷2 G
cor
G
s
÷G
11a
( )
÷B
cor
B
s
÷ B
11a
( ) ¦ ¦
Insofar as S
is
(f) can be identified as the thermal noise of G
s
, S
ie
(f)
takes the form:
S
ie
f
( )
=C ÷GG
s
÷ BB
s
÷ A G
s
2
÷ B
s
2
( )
[7.71]
which represents the basic equation for the input noise of a
photoreceiver, with:
C = S
i
b1
f
( )
÷S
i
b2
f
( )
Y
11a
Y
21a
2
÷2 G
cor
G
11a
÷B
cor
B
11a
¦ ¦
G =2
G
11a
Y
21a
2
S
i
b2
f
( )
÷G
cor
'
|
1
1
1
1
+
1
1
1
1
'
¦
1
1
1
1
+
1
1
1
1
÷4kT
B =2
B
11a
Y
21a
2
S
i
b2
f
( )
÷B
cor
'
|
1
1
1
1
+
1
1
1
1
'
¦
1
1
1
1
+
1
1
1
1
A= S
i
b2
f
( )
Y
21a
2
'
¦
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
[7.72]
For a given value of S
ie
(f), equation [7.71] generates a circle in the
admittance plane, similar to that which we saw previously.
7.7.2. Models of transistor noise
In this section, we will elucidate the two noise generators i
b1
and
i
b2
as well as their correlation.
256 Optoelectronic Sensors
7.7.2.1. Case of a field effect transistor
i
b1
represents the noise associated with the gate and i
b2
the noise
from the channel (common-source configuration of the transistor). In
general terms, we have:
S
i
b1
f
( )
=
A
g
f
÷2qIg ÷16S
2
kT
C
gs
2
f
2
g
m
R [7.73]
There is a 1/f noise component, a shot noise component associated
with the gate current, and a component produced by noise in the
channel through capacitative coupling. Here R is a dimensionless
parameter which depends on the bias point of the device:
S
i
b2
f
( )
=
A
c
f
÷4kTg
m
P [7.74]
Here we have a 1/f noise contribution and a component associated
with the diffusion noise (thermal noise) of the channel. P is a
parameter which also depends on the bias point. A
g
, along with A
c
, has
units of A
2
and represents the value of the spectral density of the 1/f
noise at 1 Hz. Working from the definition of the correlation
coefficient C
r
between two noise sources (here the channel and gate
noises):
C
r
=C
re
÷ jC
im
=
S
i
b1
i
b2
f
( )
S
i
b1
f
( )
S
i
b2
f
( )
[7.75]
we have from equations [7.73] and [7.74], where it is assumed that
only the thermal noise contribution leads to correlation:
S
i
b1
i
b2
f
( )
= 4kTCgsZ RP C
re
÷ jC
im
¦ ¦
[7.76]
Noise in Photodiodes and Photoreceiver Systems 257
The small-signal equivalent circuit for a field effect transistor,
along with the two noise generators i
b1
and i
b2
, is shown in Figure
7.14. The access points R
g
, R
s
and R
D
may be noisy and contribute
thermal noise in the form of 1/f noise.
Figure 7.14. Equivalent circuit of a field effect transistor,
with noise current generators at its input and output
7.7.2.2. Case of a bipolar transistor
The two noise sources model the shot noise associated with the
base current:
S
i
b1
f
( )
=2qI
B
= 2q
Ic
E
[7.77]
and the collector current:
S
i
b2
f
( )
=2qI
C
[7.78]
On top of the shot noise, it is useful at low frequencies to add the
1/f noise sources (at least at the level of the base) and maybe account
for the thermal noise of the terminal resistances (particularly for r
bb’
).
At high frequencies, bearing in mind that, from the point of view of
the detector, the transistor is equivalent to its input resistance r
ʌ
in
parallel with a capacitance C
T
, the total input noise to the device can
be written:
258 Optoelectronic Sensors
S
i
T
f
( )
=
2qI
C
E
÷
1
r
S
2
÷C
T
2
Z
2
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
2qI
C
g
m
2
[7.79]
Given that r
S
= E
kT
qI
c
and g
m
r
S
= E , we find:
S
i
T
f
( )
=2qI
c
1
E
÷
1
E
2
'
|
1
1
+
1
1
'
¦
1
1
+
1
1
÷2
k
2
T
2
qI
c
C
T
2
Z
2
[7.80]
An increase in the capacitance can be counteracted by an increase
in the current I
C
. The minimum S
iT
(f) is obtained for the value of the
current I
C
which cancels out the derivative of S
iT
(f), so that:
I
c
=
kT
q
C
T
Z
1
E
1
÷E
2
[7.81]
The input spectral noise density becomes:
S
i
opt
f
( )
=4kTC
T
Z
1÷E
1
E
[7.82]
showing that this optimum varies linearly with frequency. The small-
signal equivalent circuit for the bipolar transistor, along with the two
noise generators i
b1
and i
b2
are shown in Figure 7.15.
Noise in Photodiodes and Photoreceiver Systems 259
Figure 7.15. Equivalent circuit for a bipolar transistor, with
noise generators at its input and output
7.7.3. Example calculation: a PIN-FET photoreceiver
Figure 7.16 shows the structure of a PIN-FET photoreceiver for
direct detection of the high impedance type (without R
f
) or the
transimpedance type (with R
f
).
Figure 7.16. Circuit diagram for a PIN-FET photoreceiver
From this circuit diagram, the elements of the admittance matrix of
the preamplifier can be written (ignoring parasitic elements):
260 Optoelectronic Sensors
Y
11a
=
1
R
f
÷ jZ C
gs
÷C
gd
( )
Y
12a

1
R
f
÷ jZC
gd
Y
21a
= g
m
÷
1
R
f
÷ jZC
gd
= g
m
÷
1
R
f
Y
22a
=
1
R
f
÷ jZC
gd
'
¦
1
1
1
1
1
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
1
1
1
1
1
[7.83]
In addition, combining the PIN diode with its capacitance C
d
, the
source admittance takes the form:
Y
s
=
1
R
pol
÷ jC
d
Z [7.84]
Designating the equivalent resistance of R
pol
and R
f
in parallel as
R
T
, and the equivalent capacitance of C
d
, C
gs
and C
gd
in parallel as C
T
(there may also be parasitic capacitances C
p
to take into account), the
equation Y
11
= Y
11a
+ Y
s
can then be written:
Y
11
=
1
R
T
÷ jC
T
Z [7.85]
The photodetector noise is accounted for in the shot noise that
stems from the dark current I
dark
and the signal current I
signal
, with a
further contribution from thermal noise in the series resistance R
series
of the photodiode, which we can correct by taking account of the
access resistance R
g
at the gate of the transistor:
S
i
s
f
( )
=2q I
obs
÷ I
signal
( )
÷4kT R
series
÷ R
g
( )
Cd
2
Z
2
[7.86]
The total noise entering the input of the system can be written (see
equations [7.67] and [7.69]):
Noise in Photodiodes and Photoreceiver Systems 261
S
i
6
f
( )
= S
i
s
f
( )
÷S
i
b1
f
( )
÷
Y
11
Y
21a
2
S
i
b2
f
( )
÷2 Re
Y
11
*
Y
21a
*
S
i
b1
i
b2
f
( )
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
[7.87]
Thus, we have:
S
i
6
f
( )
=2q I
dark
÷ I
signal
÷ I
g
¦ ¦
÷
A
g
f
÷
A
c
f
1
R
T
2
÷C
T
2
Z
2
g
m
2
÷C
gd
2
Z
2
÷4kT
1
R
T
÷ R
series
÷ R
g
( )
C
d
2
Z
2
÷
Cg
s
2
Z
2
g
m
R÷g
m
P
1 R
T
2
÷C
T
2
Z
2
g
m
2
÷C
gd
2
Z
2
'
|
1
1
1
+
1
1
1
÷2C
gs
Z RPRe C
re
÷ jC
im
( )
1 R
T
÷ jC
T
Z
g
m
÷ jC
gd
Z

l
l
l
l
l
'
¦
1
1
1
+
1
1
1
[7.88]
At this point, we can reorganize this equation to make clear the
frequency behavior of the noise at the input to the preamplifier. For
instance, where g
m
>> Cg
d
Z, we have:
S
i
6
f
( )
=
1
f
Ag ÷
Ac
R
T
2
g
m
2
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
÷2q Ig = I
D
¦ ¦
÷4kT
1
R
T
÷
P
R
T
2
g
m
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
÷f 4 f S
2
Ac
C
T
2
g
m
2
÷16S
kT
R
T
RP
C
gs
g
m
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
÷f
2
16S
2
kT
g
m
C
T
2
P÷C
gs
2
R÷2C
im
C
T
RPC
gs
( )
'
|
1
1
+
1
1
'
¦
1
1
+
1
1
[7.89]
262 Optoelectronic Sensors
As an example, Figure 7.17 shows the variation of pA Hz
in S
i
6
f
( )
as a function of frequency, for a system whose
characteristics are listed in the caption.
Figure 7.17. Calculation of S
i
6
f
( )
as a function of frequency, using the
appropriate values for the characteristic parameters of the photoreceiver.
Dotted lines indicate the contributions from 1/f noise, thermal
and shot noise, as well as those of the f and f
2
contributions
C
d
÷C
p
=0.2pF, C
gs
=0.2pF
I
g
=150 nA, I
Diode
=20 nA
R
p
=30 k:, g
m
=12 mS; P =1.9, R =0.4
C
re
=0.2, C
im
=0.4
A
c
=10
11
pA
2
, A
g
=10
3
pA
2
'
|
1
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
1
At high frequencies, in the case where Cg
d
Z remains smaller than
g
m
, equation [7.88] can be written:
Noise in Photodiodes and Photoreceiver Systems 263
S
i
6
f
( )
=4kT
1
R
T
÷ R
series
÷ Rg
( )
Cd
2
Z
2
÷
C
gs
2
Z
2
g
m
R
'
|
1
1
1
+
1
1
1
÷
P
g
m
1
R
T
2
÷C
T
2
Z
2
(
(
·
·
·
·
·
)
)

÷2C
gs
Z RP Cre
1
R
T
g
m
÷C
im
C
T
Z
g
m
(
(
·
·
·
·
)
)

'
¦
1
1
+
1
1
[7.90]
Starting from equation [7.90], it can be seen how the width W of
the transistor gate can be calculated to match it to the photodetector,
setting:
g
m
= g
mo
W
C
gs
=C
gso
W
C
gd
=C
gdo
W and Co =C
gso
÷C
gdo
We then have:
S
i
6
f
( )
=4kT
1
R
T
÷ R
series
÷ Rg
( )
Cd
2
Z
2
÷
Cgs
o
2
g
m
0
WZ
2

P
g
m
0
W
1
R
T
2
'
|
1
1
1
+
1
1
1
÷P
C
o
2
WZ
2
g
m
0
÷ P
Cd
2
Z
2
g
m
0
W
÷2P
CdCo
g
m
0
Z
2
÷
2Cgso
g
m
0
RPC
im
CoWZ
2
÷
2Cgso
g
m
0
RPC
im
CdZ
2
÷2
Cgso
g
m
0
Z RP
Cre
R
T
'
¦
1
1
1
+
1
1
1
[7.91]
In terms of the variable W, this last equation can be rearranged to
obtain:
264 Optoelectronic Sensors
S
i
6
f
( )
=W
4kT
g
m
0
Cgso
2
R÷ PCo
2
÷2Cgso RPC
im
Co
¦ ¦
Z
2
÷4kT
1
R
T
÷2
P
g
m
0
CdCoZ
2
÷2
Cgso
g
m
0
RP C
im
CdZ
2
÷
Cre
R
T
Z
(
(
·
·
·
·
)
)

'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
÷
1
W
4kT
1
g
m
0
P
1
R
T
2
÷Cd
2
Z
2
(
(
·
·
·
·
·
)
)

÷ R
sˇ rie
÷ Rg
( )
Cd
2
Z
2
Wg
m
0
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
[7.92]
Since the gate access resistance is inversely proportional to:
W Rg =
Rgo
W
(
(
·
·
·
·
)
)

it is reasonable to make the following approximation when integrating
the photoreceiver:
R
sˇ rie
÷ Rg
( )
Cd
2
Z
2
Wg
m
0
= R
sˇ rie
0
÷ Rg
o
( )
Cd
2
Z
2
g
m
0
[7.93]
Additionally, the term representing the noise induced by the channel
in the gate, 4kT
Cgs
2
Z
2
g
m
R, can be written as:
4kT
Co
2
Z
2
g
m
R' with R' =
Cgs
2
Co
2
R [7.94]
Equation [7.92] then takes the final form:
Noise in Photodiodes and Photoreceiver Systems 265
S
i
6
f
( )
= 4kT W
Co
2
g
m
0
R'÷ P÷2 R' PC
im
( )
'
|
1
1
1
+
1
1
1
÷
1
R
T
÷2
P
g
m
0
CdCoZ
2
÷2
Co
g
m
0
R' P C
im
Cd Z ÷
CreZ
R
T
(
(
·
·
·
·
)
)

÷
1
W
1
g
m
0
P
1
R
T
2
÷Cd
2
Z
2
(
(
·
·
·
·
·
)
)

÷Cd
2
Z
2
R
series
0
÷ Rg
o
( )
g
m
0

l
l
l
l
l
'
¦
1
1
1
+
1
1
1
[7.95]
The derivative of this final result, with respect to W, leads us to an
optimal value of W which minimizes the noise, such that:
W
opt
=
Cd
Co
P 1÷
1
R
T
2
Cd
2
Z
2
(
(
·
·
·
·
·
)
)

÷g
m
0
R
series
÷ Rg
o
( )
R'÷ P÷2 RPC
im
[7.96]
We then have:
Si
6
f
( )
= 4kT
1
R
T
÷2
CoCdZ
2
g
m
0
P÷ R' P C
im
÷
Cre
R
T
CdZ
(
(
·
·
·
·
)
)

l
l
l
l
l
'
|
1
1
1
+
1
1
1
÷ P 1÷
1
R
T
2
Cd
2
Z
2
(
(
·
·
·
·
·
)
)

÷g
m
0
R
series
÷ Rgo
( )
'
|
1
1
1
+
1
1
1
'
¦
1
1
1
+
1
1
1
R'÷ P÷2 R' PC
im
( )
[7.97]
In the case where the value of R
T
dominates, we have:
Si
6
f
( )
=8kT
Co
g
m
0
CdZ
2
P÷ PR'C
im
÷ P÷g
m
0
R
series
÷ Rg
o
( )
R'÷ P÷2 R' PC
im ¦ ¦
[7.98]
which can be written as:
266 Optoelectronic Sensors
Si
6
f
( )
min
=16kT
Co
g
m
0
CdZ
2
* [7.99]
with:
* =
1
2
P÷ PR'C
im
÷ P÷ R
series
÷ Rg
o
( )
g
m
0
¦ ¦
R'÷ P÷2 R' PC
im
¦ ¦
'
|
1
1
+
1
1
'
¦
1
1
+
1
1
We can consider * as an equivalent noise parameter for the
channel which allows us to treat the system as if the gate noise is zero
(R = C
r
= 0). Equation [7.99] accounts for the high-frequency noise of
the receiver. Now returning to the sources of noise that we have
ignored up to now, the total input noise after the gate width has been
optimized is:
Si
6
f
( )
min
=2q I
dark
÷ I
signal
÷ I
g
( )
÷
A
g
f
÷ A
c
4S
2
C
T
2
g
m
2
f
÷64S
2
kT
Co
g
m
0
Cd*f
2
[7.100]
This equation gives the frequency dependence of the optimized
noise.
7.8. Comments and conclusions
Several comments can be made based on the example that we have
worked though. There are four sources of noise that are a function of
frequency:
– a 1/f contribution which dominates at low frequencies and is
connected with the manufacturing technology of the devices;
– a white noise component which represents the minimum noise
level of the photoreceiver;
Noise in Photodiodes and Photoreceiver Systems 267
– a component proportional to frequency, associated with the high
frequency performance of the transistor and the capacitative coupling
between the channel and the gate, which brings into play the low
frequency noise of the channel;
– a component proportional to the frequency squared, associated
with the * coefficient defined above.
The noise contribution from each element in the photoreceiver
circuit is clear:
– the intrinsic characteristics of the PIN photodiode (I
obs
, C
d
and
R
series
) contribute to the white noise and the high frequency noise as a
result of its matching to the amplifier;
– the bias resistance of the photodiode has an effect on the white
noise and low frequency noise of the system;
– the transistor characteristics (I
g
, g
m
, C
gs
, C
gd
, A
c
, A
g
and *) affect
the noise performance of the photoreceiver across the whole
frequency range of its operation.
Finally, at the level of the whole photoreceiver, and to summarize
all the concepts presented in this chapter, we calculate the signal-to-
noise ratio for an incident optical power P
opt
and a photodiode
response coefficient R through the equation:
S
B
=
RP
opt
( )
2
Si
6
f
( )
df
B
en
¹
[7.101]
where B
en
represents the equivalent noise bandwidth.
This type of calculation can be applied to different photoreceivers
involving bipolar or field effect transistors connected to PIN or
avalanche photodiodes.
268 Optoelectronic Sensors
7.9. Bibliography
[BER 93] BERTHIER P., Transistors à effet de champ AlInAs/(Al)GaInAs(P) pour
photodétection intégrée à 1,3-1,5 μm, PhD thesis, Montpellier University,
December 1993.
[BER 94] BERTHIER P., GIRAUDET L., SCAVENEC A., RIGAUD.D, VALENZA M., DAVIES
J.I., BLAND S.W., “InGaAsP channel HFET’s on InP for OEIC applications”,
Journal of Light Wave Technology, vol. 12, n° 12, December 1994.
[BIQ 92] BIQUARD M.F., Signaux, systèmes linéaires et bruit en électronique,
Ellipses, 1992.
[DAS 95] DAS M.B., CHEN J.W., JOHN E., “Designing optoelectronic integrated
circuit (OEIC) receivers for high sensitivity and maximally flat frequency
response”, Journal of Light Wave Technology, vol. 13, n° 9, September 1995.
[FIS 93] FISH P.J., Electronics noise and low noise design, Macmillan New
Electronics series, 1993.
[HEN 89] HENTSCHEL C., Fiber Optics Handbook, Hewlett-Packard, 1989.
[HOO 76] HOOGE F.N., “1/f noise”, Physica, vol. B83, 1976.
[JOI 96] JOINDOT I.M., Les télécommunications par fibres optiques, Dunod, 1996.
[KIM 97] KIM M.J., KIM D.K., KIM S.J., DAS M.B., “Determination of bit-rate and
sensitivity limits of an optimized p-i-n/HBT OEIC receiver using SPICE
simulations”, IEEE Transactions on Electron Devices, vol. 40, n° 4, April 1997.
[KRE 93] KRESSEL H., Semiconductors Devices for Optical Communication,
Springer-Verlag, Topics in Applied Physics, vol. 39, 1987.
[LEG 93] LEGROS E., Photorécepteur intégré pour transmission cohérente sur fibre
optique, University Thesis, Orsay, December 1993.
[PAR 93] PARK M.S., MINASIAN R.A., “Ultralow noise 10Gb/s p-i-n-HEMT optical
receiver”, IEEE Photonics Technology Letters, vol. 5, n° 2, February 1993.
[MCW 57] MC WHORTER A.L., Semiconductor Surface Physics, University of
Pennsylvania Press, Philadelphia, 1957.
List of Authors
Robert ALABEDRA
CEM2
University of Montpellier II
France
Baudoin DE CREMOUX
Central research laboratory
Thomson-CST
France
Didier DECOSTER
IEMN
Lille University of Science and
Technology
France
Carmen GONZALEZ
Alcatel Thales III-V Lab
Marcoussis
France
Joseph HARARI
IEMN
Lille University of Science and
Technology
France
Vincent MAGNIN
IEMN
Lille University of Science and
Technology
France
Antoine MARTY
LAAS-CNRS
Toulouse
France
Eva MONROY
INAC
Nanophysics and
Semiconductors Laboratory
Grenoble
France
270 Optoelectronic Sensors
Franck OMNES
Neel Institute
Grenoble
France
Dominique RIGAUD
CEM2
University of Montpellier II
France
Gérard RIPOCHE
Alcatel Research Centre
Marcoussis
France
Index
1/f noise, 14, 206, 207, 231, 232,
233, 234, 256, 257, 262
II-VI compound, 2, 101
III-V compound, 2, 5, 51, 59, 94,
100, 101, 111, 112, 114, 177, 186,
195, 201, 202
IV-IV compound, 2
A
absorption, 3, 10, 16, 19, 28, 33, 36,
39, 42, 43, 44, 45, 46, 47, 50,
51, 52, 54, 57, 58, 74, 80, 82,
83, 87, 88, 91, 93, 95, 96, 97,
99, 102, 104, 105, 111, 112,
113, 114, 115, 116, 119, 164,
165, 166, 169, 170, 182, 184,
187, 189, 191, 193, 197, 198,
204, 205, 211, 240, 245
coefficient, 39, 43, 45, 46, 74, 88,
91, 93, 111, 113, 164, 165, 166,
197, 198, 211
acceptors, 2, 5
AlGaN, 187, 196, 198, 199, 200, 202,
205, 206, 208, 209, 210, 212, 214,
215
autocorrelation function, 225, 226,
227, 233
avalanche, 7, 10, 23, 27, 32, 49, 57,
59, 60, 62, 63, 64, 65, 66, 67,
68, 69, 71, 73, 75, 76, 78, 80,
90, 95, 97, 99, 100, 103, 106,
171, 175, 176, 214, 223, 229,
230, 231, 232, 244, 245, 267
region, 62, 63, 64, 69, 75, 76, 78
B
bandgap, 1, 2, 3, 14, 19, 20, 21, 28,
29, 31, 36, 43, 45, 46, 47, 48, 49,
50, 52, 61, 65, 74, 80, 88, 90, 95,
96, 99, 101, 102, 112, 113, 122,
123, 163, 164, 169, 171, 183, 186,
187, 188, 189, 191, 198, 203, 204,
205, 208, 209, 211, 216, 232, 240,
244
bandwidth, 5, 11, 12, 13, 14, 37, 40,
53, 58, 71, 72, 75, 76, 82, 88, 98,
99, 100, 102, 104, 105, 115, 148,
166, 202, 206, 207, 215, 217, 227,
241, 267
bipolar, 111, 112, 114, 115, 118, 119,
121, 122, 123, 125, 126, 127,
131, 133, 144, 145, 146, 150,
152, 214, 238, 257, 258, 259,
267
transistor, 119, 121, 123, 125,
126, 127, 131, 146, 214, 257,
258, 259
272 Optoelectronic Sensors
breakdown, 7, 23, 27, 31, 32, 33, 41,
57, 59, 60, 64, 65, 66, 67, 68, 76,
77, 78, 79, 81, 83, 84, 86, 88, 94,
104, 159, 160, 161, 173, 176
C
cancers, 185
capacitance, 7, 18, 33, 37, 74, 78, 99,
123, 128, 172, 173, 175, 177, 202,
206, 212, 251, 257, 258, 260
circuit, 16, 23, 25, 26, 27, 28, 34, 37,
39, 40, 41, 43, 47, 53, 71, 88, 118,
127, 128, 132, 138, 141, 143, 144,
145, 146, 148, 167, 172, 173, 174,
177, 196, 224, 227, 228, 235, 244,
245, 246, 247, 248, 253, 254, 257,
258, 259, 267
cutoff frequency, 13, 35, 38, 39, 48,
55, 93, 94, 137, 140, 157, 171,
172, 173, 174
D
dark current, 21, 26, 27, 29, 30, 31,
44, 45, 47, 48, 49, 51, 60, 64, 66,
67, 68, 78, 81, 86, 88, 90, 91, 92,
94, 95, 97, 98, 100, 102, 103, 104,
105, 156, 158, 159, 161, 162, 163,
176, 177, 194, 205, 206, 207, 209,
213, 228, 240, 243, 244, 245, 260
depleted region, 28, 52, 54, 63, 97
diffusion
current, 25, 90
length, 6, 25, 28, 122, 124, 194,
203, 205, 211
dislocations, 80, 103, 192, 195, 197,
198, 205, 207, 213
donors, 2, 5, 8
E
electric field, 3, 4, 5, 6, 7, 9, 10, 16,
17, 18, 23, 24, 25, 26, 28, 31, 32,
33, 34, 36, 57, 60, 61, 62, 63, 64,
67, 69, 76, 77, 79, 82, 83, 84, 85,
86, 87, 92, 95, 96, 97, 102, 104,
114, 119, 120, 156, 157, 159, 160,
161, 162, 166, 167, 173, 191, 193,
214, 215, 229, 240, 244
electrode, 156, 158, 159, 160, 162,
166, 168, 169, 171, 173, 174, 175,
177
electrode, 9, 10, 15, 141, 143
epitaxial growth, 59, 77, 95, 97, 106,
171, 218
F, G
fabrication, 45, 48, 49, 50, 54, 80, 81,
85, 86, 87, 92, 99, 100, 104, 112,
113, 117, 162, 163, 188, 209, 214,
216, 217, 234
filter, 51, 133, 184, 187, 191, 208,
216
GaAlAs, 43, 45
GaAs, 2, 43, 45, 46, 59, 70, 88, 89,
100, 105, 112, 113, 114, 115, 123,
144, 149, 162, 165, 166, 171, 177,
190, 208
gain, 5, 7, 10, 11, 12, 23, 27, 32, 40,
49, 53, 55, 57, 58, 59, 66, 67, 68,
71, 72, 73, 76, 82, 84, 86, 88, 94,
98, 100, 102, 104, 105, 111, 118,
121, 123, 124, 125, 126, 127, 129,
133, 134, 135, 137, 140, 144, 146,
147, 149, 150, 171, 176, 196, 197,
209, 215, 216, 217, 229, 239, 244,
252, 254
GaInAs, 38, 39, 40, 46, 47, 48, 49,
53, 99, 164, 166, 169
gain-bandwidth product, 76, 100,
102, 104, 105
GaN, 2, 187, 188, 195, 196, 202, 204,
207, 209, 210, 211, 212, 213, 214,
215
GaSb, 49, 102
generation-recombination noise, 232
germanium, 2, 46, 47, 48, 59, 60, 61,
62, 72, 90, 93, 94, 95
Index 273
guard ring, 75, 77, 78, 81, 85, 87, 88,
91, 93, 94, 97, 102, 103
H, I, J
heteroepitaxial growth, 207
implantation, 78, 81, 84, 85, 86, 87,
92, 93, 94, 97, 103
InP, 38, 39, 40, 46, 47, 48, 49, 53, 59,
61, 62, 70, 95, 96, 97, 98, 99, 100,
101, 102, 103, 105, 106, 112, 113,
115, 123, 145, 146, 147, 148, 149,
150, 169, 177
junction, 4, 5, 6, 7, 8, 9, 10, 15, 17,
18, 30, 31, 32, 33, 34, 37, 48, 58,
60, 62, 64, 65, 66, 68, 69, 73, 74,
76, 77, 78, 79, 80, 83, 84, 85, 92,
94, 97, 99, 103, 108, 114, 115,
118, 119, 120, 121, 123, 128, 130,
134, 135, 139, 147, 156, 158, 161,
162, 163, 164, 186, 189, 191, 192,
193, 194, 204, 210, 211, 212, 213,
214, 217
L, M
large signal, 126
mixer, 131, 132, 135, 142, 143, 146,
149, 150
MSM photodiode, 156, 174, 188,
209, 217
multiplication factor, 57, 62, 63, 64,
65, 66, 94, 229
N
noise, 11, 13, 14, 49, 58, 59, 61, 66,
68, 69, 70, 71, 72, 73, 76, 79, 80,
81, 84, 86, 87, 88, 89, 91, 92, 93,
94, 97, 98, 100, 101, 102, 103,
104, 105, 118, 138, 139, 140, 146,
155, 157, 175, 176, 177, 183, 202,
206, 207, 210, 213, 214, 217, 220,
223, 224, 226, 227, 228, 229, 230,
231, 232, 233, 234, 235, 236, 237,
238, 239, 240, 241, 243, 244, 245,
246, 247, 248, 249, 250, 251, 252,
253, 254, 255, 256, 257, 258, 259,
260, 261, 262, 264, 265, 266, 267
noise equivalent power (NEP), 11,
13, 207, 210, 241, 244
noise factor, 69, 70, 71, 76, 80, 84,
86, 87, 88, 90, 91, 93, 94, 97, 98,
100, 101, 102, 103, 104, 105, 176,
230, 231, 238, 239, 246, 248, 249,
250, 251, 252
O, P
ohmic contact, 4, 194, 210
optical
fiber, 44, 47, 59, 80, 93, 106, 111,
112, 113, 116, 117, 146, 149,
169
links, 91, 106
photoreceiver, 106, 138, 148, 176,
223, 234, 235, 239, 241, 245, 246,
247, 248, 253, 255, 259, 262, 264,
266, 267
phototransistor, 111, 112, 113, 114,
115, 116, 117, 118, 119, 121, 122,
123, 124, 125, 126, 127, 128, 129,
131, 133, 134, 135, 137, 138, 139,
140, 141, 142, 143, 144, 145, 146,
147, 148, 149, 150, 214, 215, 217
PIN photodiode, 15, 16, 18, 25, 28,
29, 32, 34, 35, 36, 38, 39, 40, 42,
48, 49, 50, 66, 72, 74, 75, 115,
118, 146, 169, 173, 177, 214, 228,
242, 243, 244, 245, 267
Q, R
quantum efficiency, 11, 12, 26, 27,
28, 29, 36, 39, 44, 45, 48, 50, 51,
52, 53, 54, 74, 82, 90, 94, 104,
117, 124, 126, 164, 166, 169, 170,
187, 192, 193, 194, 197, 198, 203,
204, 211, 240
reliability, 58, 79, 91, 97, 98, 102,
106, 172, 186, 188, 216, 217
response coefficient, 11, 12, 13, 14
274 Optoelectronic Sensors
S
Schottky contact, 9, 156, 164, 205,
206, 209
semiconductor, 1, 3, 4, 5, 7, 8, 9, 10,
11, 12, 15, 23, 42, 43, 46, 49, 50,
58, 60, 74, 75, 80, 88, 90, 112,
155, 156, 158, 159, 160, 161, 162,
163, 164, 167, 168, 169, 171, 175,
183, 186, 189, 217, 229, 230, 240,
242, 243, 244
shot noise, 13, 68, 69, 71, 139, 175,
176, 213, 229, 232, 234, 236, 243,
245, 257, 260
SiC, 187, 188, 190, 194, 195, 216
silicon, 2, 21, 23, 43, 44, 45, 46, 59,
60, 61, 62, 70, 72, 75, 87, 90, 96,
100, 111, 112, 115, 116, 141, 144,
145, 150, 165, 166, 186, 187, 188,
191, 196, 214, 231, 245
small signal, 174
submarine links, 59
T
telecommunications, 39, 44, 45, 46,
47, 48, 49, 59, 103, 106, 112, 145,
150, 184
temporal response, 12, 13, 198, 199,
200, 210, 212, 215, 217
thermal noise, 13, 71, 73, 139, 175,
232, 234, 238, 244, 256, 257, 260
transit time, 34, 35, 36, 40, 44, 50,
52, 53, 55, 73, 74, 75, 94, 99, 104,
105, 123, 124, 128, 129, 164, 167,
170, 171, 172, 173, 174, 175, 177,
210, 212, 228, 229
transparent contacts, 189
transport, 15, 16, 54, 83, 96, 101,
121, 122, 123, 125, 127, 146, 166,
167, 169, 171, 175, 213
tunnel effect, 26
U, W
ultraviolet (UV), 10, 11, 44, 87, 181,
216
UV-visible contrast, 188, 189, 209
waveguide photodiodes, 52, 53
wavelength, 4, 12, 20, 27, 36, 39, 43,
45, 51, 52, 53, 58, 80, 81, 88, 112,
113, 142, 146, 163, 167, 168, 169,
181, 184, 186, 187, 189, 190, 192,
193, 194, 195, 196, 203, 204, 208,
211, 217, 240

This page intentionally left blank

Optoelectronic Sensors

This page intentionally left blank

Optoelectronic Sensors
Edited by Didier Decoster Joseph Harari

co. Inc.iste. p. Apart from any fair dealing for the purposes of research or private study. in any form or by any means. as permitted under the Copyright. TK8360. . with the prior permission in writing of the publishers. 1961. Didier. 111 River Street Hoboken. stored or transmitted. 2009 www. I.wiley. ISBN 978-1-84821-078-3 1. Optical detectors.II.25--dc22 2009011542 British Library Cataloguing-in-Publication Data A CIP record for this book is available from the British Library ISBN: 978-1-84821-078-3 Printed and bound in Great Britain by CPI Antony Rowe. Designs and Patents Act 1988. 2.III. English Optoelectronic sensors / edited by Didier Decoster. Image sensors. Chippenham and Eastbourne. Joseph. Decoster. this publication may only be reproduced. 2002 First published in Great Britain and the United States in 2009 by ISTE Ltd and John Wiley & Sons. NJ 07030 USA www. or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA.com The rights of Didier Decoster and Joseph Harari to be identified as the authors of this work have been asserted by them in accordance with the Copyright.First published in France in 2002 by Hermes Science/Lavoisier entitled: Détecteurs optoélectroniques © LAVOISIER. Title. Inc. 1948. Includes bibliographical references and index. Joseph Harari.uk © ISTE Ltd. Harari. Designs and Patents Act 1988.O67D4813 2009 681'. or criticism or review. Library of Congress Cataloging-in-Publication Data Détecteurs optoélectroniques. cm. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address: ISTE Ltd 27-37 St George’s Road London SW19 4EU UK John Wiley & Sons.

. . . . . . .Table of Contents Preface . . . . . . . 1. Franck OMNES 1. . . . . Introduction to Semiconductor Photodetectors . . Noise equivalent power . . .8. .2. . . . . . . . . .3. . . . . . . . . . . Temporal response and bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . .8. . . 2. . Semiconductor devices. . . 1. . . . . . . . 1. . . Static characteristics of PIN photodiodes .2. . . .2.3. . . . . . . . . Avalanche effect in p-i-n structures . . . . . . . . 2. . . . Physical processes occurring in photodiodes . . . . . . . . .3. . . . . Electrostatics in PIN diodes: depleted region . . . . .2. . . . 1. . 2. . . . Introduction . . . .2. . . . . 1. . . . . . . . . . . . . . . . . . . . . 1. Detectivity . . . . Operational parameters of photodetectors . . . Baudoin DE CREMOUX 2. . .8. . . . . . . . . . . . 2. . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . Schottky junction . .8. . . .8. . p-n junctions and p-i-n structures . . . . . . . . . . . . . . PIN Photodiodes for the Visible and Near-Infrared . . . . . . . . . 1. . . . . . . . . . 1. . . .6. . . . . . . . . . . . .1. I/V characteristics and definition of static parameters . . . . . . . . . . . Metal-semiconductor-metal (MSM) structures . . . . . gain and quantum efficiency 1.5. . . . . .2. . . . Brief overview of semiconductor materials . . . Photodetection with semiconductors: basic phenomena 1. . . . . . . . . . . . . . 1. . . . 2. . . .3. . . . . . . . . . . . .7. . . . . . . 2. . . . Transport mechanisms. .3. . .1. . . . . Response coefficient. .3. . . . . External quantum efficiency . .4. . . . . . . . . . . . Mechanisms of electron-hole pair generation . .1. . .2. . . Chapter 1. xi 1 1 3 4 5 7 8 10 11 11 12 13 14 15 15 17 17 19 23 25 25 27 Chapter 2. . . . .1. . . . . .2. . . . .1.4. . . . . . . . . . .

. .1. . Crystal defects and microplasmas . Absorption of semiconductors in the range 400-1. Avalanche Photodiodes . . . . . 3. Traveling-wave photodiodes.6. . . .4. . . . . . . . . . . . Gérard RIPOCHE and Joseph HARARI 3. . . . 2. . . . . . . . Dark current .4. . . Si N+P P+ APDs . . .4. . . . Speed. . . . . . 3. . 3. . . . . . . . . . Introduction . . . . . . . . . . . . . Saturation current. . . . . . . . . . 3.5. . .3. . . . . .6. . .5. . .4. .4. . . . . . . History . . . . . . .3. . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . .4. . . . . . .5. . . . . . . . . . . . . . . . . . . . . From 400 to 900 nm: silicon and the GaAlAs/GaAs family 2.800 nm: germanium. 2. . GaInAsP/InP… . . . . Silicon avalanche photodiodes . . . . . . . . . . . . . . . . . 3.6.1. . . . . . .5. . . . . . . . . . . . . . . Breakdown voltage. . . . . . . . . . . .3. . . . . . 2. . . .6. . . . . From 900 to 1. . . . . . . . . 2. . . . . . . Photodiodes with collinear geometry . . . . . . . . . . . . . . .6. .2. . Chapter 3.800 nm . .3. . Power-frequency compromise. . Noise in avalanche photodiodes. . . . . . . . . . . . . . . . . 3. .3. . . . . . . . . . . . . . . . . 3. . . .5. .5. . . . . . . . . 2. 3. . . . . . . 2. . . . Bibliography . . . 3. . . . . . . . . . . . . .3.vi Optoelectronic Sensors 2. 2. Multiplication factors . . . . . . . . . . . . . .1. . .4. . . . . 2. . . .2. Ionization coefficients . . . . . . . 2. . Beyond the limits of conventional PIN . . . . . . . . .3. . . . . . .4. . . . .5. 2. . . . . . . .5. . . . . . . . . . . . . . . . Dynamic characteristics of PIN photodiodes . Beyond PIN structures . . . . . . . . . . .3.1. . . . . . . .4. . . Signal-to-noise ratio in avalanche photodiodes . . . . . . . . 2. . . Current-voltage characteristics and photomultiplication 3. . . . Semiconductor materials used in PIN photodiodes for the visible and near-infrared . . . . . Technological considerations . . . . . . . . . 3. . . . . . 2. . . . . . 3. . . . . . . . 29 31 32 34 34 37 41 42 42 43 46 49 49 50 52 53 54 55 57 57 58 60 61 62 64 66 66 68 71 73 76 77 78 79 80 80 82 84 . New photodiode structures . .2. . . . . . . 3.4. . .6. .1. 3. 3. .6. .1. . . . . . . . . . . . . . . . .6.1. 2.2. .4. . . . . . . .6. . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . Waveguide photodiodes. . . .5. . . . . 2. . . Properties of avalanche photodiodes . . . . . . . .2. . . . . . . . “Mesa” structures.2. . . . . . . . Intrinsic limitations to the speed of response. . . . . . .3. . . . . . . . . . . . . . . . . . . .1. . . . . .5. . 2. . . . . . . . . . Pf2 “law” . . .7.3. 3. .2. .3. . . . . . . . . . . . . . . . . . . Si N+ P P+ APDs . . . . . . . . . . . . response time and frequency response of avalanche photodiodes . .3. . . . . . . . . . . . . . .6. . . . .4. . . . .3. . . . . . . . The avalanche effect . 3. . . . . . .2. . . . 2. . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Si N+P APDs. . . . . . . . . . Limitations due to the circuit . . . . . . . . . . . . . Guard ring junctions . . . . . . . . . . . Breakdown voltage.

Introduction .11.13. . . . . . 4. . . . 4. . . . . . . . . . . . . . . . Amplification circuits . . . . 3. .1. Conclusion. .2. . . InGaAs/InP APDs for optical communications at 2. 4. .9. Phototransistors for optical telecommunications . . . .1. . . . 3. . . . . . . . . . . . . . . . . . . . . . Bibliography . . . . . . . . . . . 3. . . . . . . . . . Prospects . . 3. . .2. . . . Ge APDs with P+NN. . . . 4. . . . . . . . .3. Applications . . .8. . 3. . . . . . . . . . . . . . . . . 3. .11. . . . . . . . Avalanche photodiodes based on gallium arsenide . . . . . 4. . . . . . . . Low-noise APDs with a very thin multiplication region. . . . . . . . . . . . . . . 3. . . . . . .5. . . . . . . . SiPt-Si N Schottky APDs. . . . . . . . .2. . . . . . . . . . 4. . . . . . Carmen GONZALEZ and Antoine MARTY 4. . .6. .3 μm communication . . 4. . 4. . . . . . . . . . . N+NP and P+N structures for 1. . . . . . . . . . . . . 3. . Bibliography . . . 3. . . . . . . . . . . . .2. . . .4. . . . . . . . . . . .1. . .10. . . . . . . . . . . . . . . . . .5 Gbit/s . . Nonlinear circuits. . . . . 4. . . . . . .3. . . . . . . . Si/InGaAs APDs . .1. . . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . Avalanche photodiodes based on indium phosphate (InP). . 3. . . . . 4. . Spin-orbit resonance APDs . . . .9. . . . . . . . . . . . . 87 88 90 91 93 95 97 99 100 101 102 104 104 104 105 106 107 111 111 112 112 114 118 119 124 125 126 138 140 140 141 142 142 145 150 151 Chapter 4. . . . . . . . . . . . . . . . . . 3. . . . .8. . . . Fast InGaAs/InP APDs . . . . .3. The response coefficient of a phototransistor . . . . . .8. . . . . . . . . . . . .11. . . . . . . . .7. . . . . . Phototransistors . . . 3.5. . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . Phototransistors according to their fabrication materials 4. . . . . . . . . . . . .structures for 1. Galvanic isolation . . . . . . . . . . . . . . . . .5. .9. . 4. . . . . . . . . . . .4. . . . . . . . . . . . . . . . . .2. . .12. .4. . . . . . . . . .10. . . . . . . . . Phototransistors . . . .1.3. . . . . . . . . .2. . . Germanium avalanche photodiodes . . .3. . . . III-V super-lattice or MQW APDs . . . . . . . . . . . . . . . The phototransistor effect. .3. . . .1. . . . . .2. . . . . 4. . . Conclusion . . . . . .11. . . . . . 3.6. . . . . . . . . . . . . Photodetector circuits based on phototransistors. . . . . . . 4. .3. . . . . . . . . .5.3. . . . . . . . .2. . .4. . . . .Table of Contents vii 3. . . . . .7. . . . . . III-V low-noise avalanche photodiodes. . . 3. .1. .2. . . . . .55 μm communication . . . . . . . . Static electrical and optical gains of the phototransistor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Noise in phototransistors .4. . . . . . . . . . . . . . 4. . . . . . . . . . . . . The bipolar phototransistor: description and principles of operation . . . . . .2 Phototransistors classified by structure. . 3. . . .10. . . . . . . “Waveguide” MQW APDs .1. . Dynamic characteristics of phototransistors .2. . . Ge APDs with N+P. 4. . . . . .

. . . . . . . . .viii Optoelectronic Sensors Chapter 5. . . . . . . . . . 6. . . . 7. . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Franck OMNES and Eva MONROY 6. . . . Si and SiC photodetectors for UV photodetection . . . .2. . . . . . . . . . . . . . . 6. . . . . 7. 6. . 5. . . . . . . 6. . . . . . . . . . . . . . . . UV detectors based on III-V nitrides . .3. Thermal noise . . . . . . Introduction . . . .4. . . . . . . . . . .4. . . . . . .3. . . . .2. .4. .4. Phototransistors . .3. . Introduction . . . . . . UV photodiodes based on silicon . . . . . . . . . . . . . . . . 6. . 5. p-n and p-i-n photodiodes . . . . .1. . . . . . . . Operation and structure . . . Dynamic behavior . . . . . . . . . . . . . . . . 7. . . . . . . . . 6. . . . . . . . . . . . . . Fundamental noise sources . Response coefficient . .1. . . . . . . . Materials used . . . .3. . 5. . Bibliography . . . .3. .4. . . . . . . . . . . . . . . . . . . . . Schottky barrier photodiodes based on AlGaN 6. . . . . . . Bibliography . . . . . 6. . . . . . 5.2. . . . . . . . . .4. .3.3. . . . . . Excess noise . . . .3. . . . Random signals and background noise . . . . . . . . . . . . . Noise in Photodiodes and Photoreceiver Systems . . . . .3. . . 7. . .4. . . . Metal-Semiconductor-Metal Photodiodes . . . . . . . . . . . . . . . . . . . . . . . Ultraviolet Photodetectors. . The UV-visible contrast . . . .2. . . . . 7. . . . . . . . . . Robert ALABEDRA and Dominique RIGAUD 7. . . . . . . . . . Shot noise . . . . . 155 155 156 156 161 165 165 172 175 177 178 181 181 189 190 191 194 195 196 202 209 210 214 216 218 223 224 224 226 227 227 228 229 232 232 233 Chapter 6. . . . .6. . . . . . . . . . . . . . . .5. . . . . . . . . . . . . . .2. . . . . . . . . . . Integration possibilities and conclusion 5.2. . Chapter 7. . . . . . . . Known signals with finite energy or power 7. . . . . . . . . . . . . . . . 6. 5. . 5. . Multiplication noise . . . . . . . . . .1. . . .2.2. . . Noise . . . . 7. . .2. MSM photodiodes . . . . . . . . . . . . Static and dynamic characteristics. . . . . . . . . . . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . 6. . . . . . . . . . . .2. . . . . . .2. . . . . . . .1. Mathematical tools for noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . .2. . . . . . . . . . . . . . .1.2. . . 6. . . . . . . . . . . . . . . . .3. . . . . Photoconductors . . . . . . . .2. . . . .1. . . . . . . . . Generation-recombination noise . . . . . . . .5. . . . . . . . 1/f noise.1. . . . . . . . . . . .1. . . . . . . . . .3. . . . . . . Conclusion . . . . . .3. . .5. . . . Fundamentals . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . SiC-based UV photodetectors . . . . . . . . .4. . . . . . . . . . . . Joseph HARARI and Vincent MAGNIN 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . 5. . . . . 5. . . . . . . . . . . 7. . . . . . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . .

. . . 7.4. . . . . . . . 235 235 237 239 240 242 244 245 246 251 253 253 255 259 266 268 269 271 List of Authors . . . . . . .7. . .9. . . . . . . .1. 7. . . . . . . . 7. . . . . . Basic equations . . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . .2. . . . . . . . . . . . . . . . . . . . . .4. . .4. 7. . . . . Noise optimization of photodetectors. . . . . . . . . . 7. . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Calculation of the noise of a photoreceiver . . . . . . . . . Characteristic parameters . . . . . . . . . . . . . . . .1. 7. .7. . . . . . .3.6. . . . . . .6. . . . .5.5. Index . . . . . . Models of transistor noise . . . . . . . . . . . . . .5. . Representation of noise in quadripoles . . . . . . . . . .8. . . . . . . .3. . . Comments and conclusions . . . . . 7. . . . Formulation of the problem . Noise in photodetectors . . . . . . . . . . Representation of noise in bipoles . . . . . . . . . . . . . . . . .6. . . . . 7. . . . .7. . . . . .1. PIN photodiodes . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . Avalanche photodiodes . . . . . . . . . 7.2. 7. . . . . 7. . . . . . .2. . . 7. . . .7. . . . . . . Example calculation: a PIN-FET photoreceiver. . . . . . . . . . . . Analysis of noise electrical circuits . . . . . . . . .Table of Contents ix 7. . . . . . . . . . . . . . .5. . . . . . 7. . . . . . . . . . Bibliography . Concepts for photodetector-transistor matching. .

This page intentionally left blank .

potentially at very high speeds. galvanic isolation. This book gathers together the most detailed and significant contemporary thinking on photodetection for wavelengths from the near-infrared to the ultraviolet1. It is often this. These basic considerations imply that a certain number of performance requirements should be met in order to satisfy the demands of the intended application. It is in this context that various photodetecting structures have been conceived: photoconductors. solar cells. phototransistors. Its content not only gives the reader 1. The range of wavelengths relevant to the application also plays an important role. which dictates the nature of the material(s) used. . When information needs to be transmitted quickly. avalanche photodiodes.and far-infrared requires specific treatment and thus is not covered here. the photodetector must react very fast. p-n and p-i-n photodiodes.Preface Photodetection is found in a large number of professional and mass-market systems. with this signal needing to be as large as possible for an optical flux as weak as possible. The very particular nature of photodetection in the mid. Schottky photodiodes. proximity detectors. MSM (metal semiconductormetal) photodetectors. There are numerous applications including: fiber-based and free-space optical telecommunications. through the intermediary of the semiconductor bandgap. All of these applications are based on the same process: the transformation of optical power into an electrical signal. etc.

Chapter 6. In this case. notably for environmental applications. Among these. Chapter 7 is entirely dedicated to noise.xii Optoelectronic Sensors the grounding to design simple photodetectors with specified performance characteristics. but also discusses the state of the art in photodetection. Chapter 1 begins with an introduction to photodetection and the most well-known photodetector structures. is different. – the phototransistor. under a sufficiently high bias voltage. however. some are more commonly used than others. each of Chapters 2 to 5 concentrates in great depth on a specific photodetector type. Given the significant role that this wavelength range is likely to take in the near future. as this concept is absolutely fundamental for the photodetection of low-intensity signals. allows us to achieve a gain (Chapter 3). all types of photodetector are concerned. . the reader will be able to learn about: – the PIN photodiode which is the photodetector on which the majority of systems are based (Chapter 2). Because of this. In sequence. or offer more exciting possibilities. – the MSM photodiode which is well suited to monolithic integrated circuits such as MMICs (Microwave Monolithic Integrated Circuits) (Chapter 5). this subject deserves a separate chapter. which is another means of obtaining gain in photodetection (Chapter 4). It is dedicated to ultraviolet photodetectors. – the avalanche photodiode. and it is the material which is the source of improvement. notably with the appearance of AlGaN materials with a very large bandgap. which is a refinement of the PIN photodiode and which. Finally.

Preface xiii The chapters of this book have all been written by specialists and we take this opportunity to thank them sincerely and warmly for their contributions. Didier DECOSTER and Joseph HARARI .

This page intentionally left blank .

The energy range lying between the top of the valence band and the bottom of the conduction band is known as the forbidden band. or more commonly the bandgap. this applies as long as the crystal dimensions are typically larger than a few dozen nanometers). whereas an electron situated in the conduction band exists in an excited state. to a continuum of states (which.Chapter 1 Introduction to Semiconductor Photodetectors 1. In general terms. An electron situated in the valence band is in a ground state and remains localized to a particular atom in the crystal structure. in the case of an infinite crystal. means that the characteristic dimensions of the crystal are significantly larger than the lattice parameter of the crystal structure.1. What Chapter written by Franck OMNES. . Brief overview of semiconductor materials A semiconductor material is a continuous crystalline medium characterized by an energy band structure corresponding. in a regime where it interacts very weakly with the crystalline structure. in practice. the energy structure of a semiconductor consists of a valence band corresponding to molecular bonding states and a conduction band representing the molecular antibonding states.

InP. Se.). are substituted in place of a silicon atom in the crystal structure: since silicon is tetravalent. and to insulators when the bandgap is more than 6 eV: above this. magnesium) substituted in place of group . These electrons become free to move. the main donor impurities are those which. being from group V of the periodic table (arsenic. subject to a weak activation energy provided by thermal agitation. Semiconductor materials are mostly divided into two large classes: elemental semiconductors (group IV of the periodic table): silicon. sometimes considerably. An impurity is known as a donor when it easily releases a free electron into the conduction band. diamond. and acceptors are group II (zinc. For example. III-V (GaAs. GaN) and II-VI (CdTe. In the case of silicon. in the case of compound semiconductors in group IV of the periodic table such as silicon. and then completes its own outer-shell electronic structure by capturing an electron from its fourth nearestneighbor silicon atom. Impurities can be introduced into the volume of the semiconductor material and can modify its electrical conduction properties. In this case we refer to n-type doping. ZnS. phosphorous. In the case of III-V composites. these atoms naturally form four covalent bonds with the silicon atoms around them. Such an impurity is known as an acceptor. A hole carrying a positive elementary charge and corresponding to a vacant energy state in the valence band is therefore left in the crystal structure of the silicon. slightly below the conduction band. a group III element incorporated into the crystal structure of silicon naturally forms three covalent bonds around it. and doping with acceptors is known as p-type doping. the solar spectrum arriving on the Earth’s surface is unable to produce inter-band transitions of electrons situated in the valence band of the material.2 Optoelectronic Sensors differentiates semiconductors from insulators is essentially the size of the bandgap: we refer to semiconductors where the bandgap of the material is typically less than or equal to 6 eV. InSb. again subject to a weak thermal activation energy.). and compound semiconductors: IV-IV (SiC). ZnSe. Te) substituted in place of group V elements. or group VI elements (S. etc. etc. and also easily give up their surplus electron to the crystal structure. etc. germanium. The characteristic energy level of the impurity is therefore in the bandgap. the donors are mostly atoms from group IV (silicon) substituted in place of group III elements.

phosphorous. When a moderate n-type doping is added.e. The chemical potential. and acceptors belong to either group I (lithium. Photodetection with semiconductors: basic phenomena Photodetection in semiconductors works on the general principle of the creation of electron-hole pairs under the action of light. In this case it is referred to as degenerate. When the level of ntype doping becomes large. the absorbed photons promote electrons from the valence band into excited states in the conduction band. the semiconductor is said to be degenerate when the Fermi level is below the top of the valence band.) or to group V (nitrogen. one free from n and p impurities) is found in the middle of the bandgap of the material. In this last case. the most commonlyencountered donors belong to group VII (chlorine. the positively-charged holes left in the valence band contribute to electrical conduction by moving from one atomic site to another under the effects of the electric field. the Fermi level can cross the bottom of the conduction band and be found inside this band (Mott transition). and whose intensity at a given . In addition. When a semiconductor material is illuminated by photons of an energy greater than or equal to its bandgap. arsenic. of an intrinsic semiconductor (i. the group V element is substituted in place of a group VI element in the semiconductor crystal structure. The semiconductor then behaves like a metal and for this reason is called a semi-metal. or Fermi energy.) substituted in place of group VI elements.Introduction to Semiconductor Photodetectors 3 III elements. etc. by an increasing amount as the level of doping rises.2. whereas group I acceptors are substituted in place of group II elements. etc. which refers by definition to the fraction of the photogenerated free charge-carriers collected at the edges of the material by the electrodes of the photodetecting structure. 1. In the case of II-VI composites. In this way the separation of electron-hole pairs generated by the absorption of light gives rise to a photocurrent. where they behave like free electrons able to travel long distances across the crystal structure under the influence of an intrinsic or externally-applied electric field. the Fermi level rises from the middle of the bandgap towards the conduction band. In the case of p-type doping. etc).

3. which consist of a simple layer of semiconductor simply with two ohmic contacts. and p-i-n photodetectors which include a thin layer of semiconductor material between the p and n region which is not deliberately doped). A fixed voltage of magnitude VB is applied between the two end contacts. We will now briefly introduce the main physical concepts at the root of the operation of the different semiconductor photodetector families. the photogenerated changes produced under the effect of the applied electric field lead to a photocurrent IPH which is added to the bias current. which use the internal electric field of a p-n or Schottky (metalsemiconductor) junction to achieve the charge separation. Here the emphasis is placed on a phenomenological description of the working mechanisms of the devices in question. . the corresponding formalism has been deliberately kept to an absolute minimum in the interests of clarity and concision. in such a way that a bias current IB flows through the semiconductor layer. 1. This last term covers p-n junction photodetectors (photovoltaic structures consisting of a simple p-n junction.4 Optoelectronic Sensors wavelength is an increasing function of the incident light intensity. where the electric field leading to the collection of the charge-carriers is provided by applying a bias voltage between the contacts at either end. Semiconductor devices Photoconductors represent the simplest conceivable type of photodetector: they consist of a finite-length semiconductor layer with an ohmic contact at each end (Figure 1. effectively increasing the conductivity of the device. When it is illuminated. and photovoltaic photodetectors. simply following Ohm’s law. The active optical surface is formed from the region between the two collection electrodes. as well as all Schottky junction photodetectors (Schottky barrier photodiodes and metalsemiconductor-metal (MSM) photodiodes).1). On this level we can distinguish between two large categories of photodetectors based on the nature of the electric field. which causes the charge separation of photogenerated electron-hold pairs: photoconductors.

Introduction to Semiconductor Photodetectors 5 Figure 1. On the other hand. In equilibrium we therefore find a region with no free charge carriers immediately around the junction. where there are. where ionized donors and acceptors provide fixed charges). positively ionized donors and. its other operational parameters (bandwidth. the metallurgical linkage of a region of a p-type doped semiconductor and a region of n-type doping forms a p-n junction. on the n side.4. infrared sensitivity) are generally below that of other types of photodetectors. similar to a charged capacitor. on the p side. negatively ionized acceptors (this zone is known as the space charge region (SCR). Diagram of a photoconducting device The main point of interest in a photoconducting device is its increased gain. 1. in equilibrium. where the joining of the Fermi levels in equilibrium mostly occurs through a flow of charge between the n and p regions. UV/visible contrast. which often greatly limits the scope of its potential applications (this is particularly the case for photoconductors based on III-V nitrides.1. forms an energy barrier between the two regions: the . p-n junctions and p-i-n structures In p-n diodes. the response of photoconductors being typically several orders of magnitude greater than that of photovoltaic detectors for a given material. The presence of charged donors and acceptors produces an electric field in that region which curves the energy bands and. as we will see later on).

which significantly improves the efficiency of the separation of electron-hole pairs by increasing the electric field within the junction. to some extent. and also increase the spatial extent of the SCR. An electron-hole pair produced in this SCR (situation 2 in Figure 1. Figure 1. where the forward direction is defined as the direction of flow of the majority charge carriers (from the n to the p region in the case of electrons.6 Optoelectronic Sensors bottom of the conduction band and the top of the valence band on the n side are below the corresponding levels on the p side (Figure 1. The band structure of the junction implies that the photocurrent will consist of minority charge carriers. Moreover.2). .2). the application of an opposing external electric field (Vp–Vn < 0) allows us to increase the height of the energy barrier in the vicinity of the junction. and so does not recombine. the photocurrent flows in the opposite direction to the bias on the diode. and vice versa for holes). Curvature of the energy bands and mechanisms of photocurrent generation in a p-n junction The width of the SCR is a decreasing function of the level of doping in the material. those generated at a distance from the junction less than or equal to the diffusion length (situations 1 and 3 in Figure 1. These are the charge carriers which contribute to the photocurrent.2) is therefore separated by the effect of the internal electric field of the junction.2. to which we can add. For this reason. while the height of the energy barrier is an increasing function of it.

1. Additionally. where in order to increase the photoresponse it is desirable to ensure that the mechanisms of electron-hole pair generation through incident light take place predominately inside the SCR. . increasing the width of the SCR reduces the capacitance of the structure. The gain is therefore greater than 1 for the generation of charge carriers by light. whose width is then largely determined by the thickness of the “i” layer.Introduction to Semiconductor Photodetectors 7 We note that when the doping level is moderate.and p-type doping are generally adjusted to high values above 1018 cm-3 to maximize the intrinsic electric field of the junction. this is generally the case when the bias voltage at the terminals reaches a few dozen volts). Avalanche effect in p-i-n structures When the reverse-bias voltage established at the terminals of a p-in structure increases sufficiently that the electric field established in the junction reaches values close to the breakdown field (in structures of micron-scale thickness. the width of the SCR is important. Such a structure is interesting because it is possible to maintain high levels of doping in the n and p regions without significantly reducing the extent of the SCR. and this gain can even typically reach 10 or 20 under favorable conditions. which makes p-i-n structures particularly well-suited for high-speed operation. This effect is exploited in what are called avalanche photodiodes where the levels of n. the photogenerated charge carriers in the SCR (which is effectively the region that is not intentionally doped) are accelerated enough to separate other secondary charge carriers from the atoms in the lattice that they impact in the course of their motion: this is the avalanche effect which results in a multiplication of the charge carriers in the SCR. A simple means of increasing the spatial extent of the SCR is to introduce between the n and p regions a thin layer of intrinsic semiconductor material which is not intentionally doped: the structure is therefore referred to as p-i-n.5. This effect is beneficial in the case of p-n junction photodetectors.

when the Fermi levels of the metal and the semiconductor are equalized. Schottky junction A Schottky junction is formed by bringing a metal and a semiconductor into contact.3.3. a transfer of electronic charge occurs from the semiconductor to the metal in the case where the work function q. and a SCR appears at the edge of the semiconductor of width xd next to the junction. A curvature of the energy bands therefore occurs at the junction.6. Figure 1. M of the metal (q being the elementary charge) is greater than the electron affinity X of the semiconductor. Formation of a Schottky junction (in an n-type semiconductor) In thermal equilibrium. where the only charges present are the positively-ionized donors. which leads to the appearance of an energy barrier between the metal and the .8 Optoelectronic Sensors 1. The basic phenomena which lead to the formation of a Schottky junction with an n-type semiconductor are summarized in Figure 1.

along with an increase in the width of the SCR.1] In equilibrium. therefore. Figure 1. It is possible. with the electron-hole pairs being separated by the effect of the electric field in the Schottky junction.4). we find an intrinsic electric field immediately next to the metal-semiconductor junction which is comparable in form to that found in a p-n junction. and only the minority carriers (holes) generated by external excitation (in particular photogeneration) can reach the Schottky contact and hence produce an electric current: as in the case of the p-n . This last effect is of course favorable for photodetection.Introduction to Semiconductor Photodetectors 9 semiconductor. called a Schottky barrier. which leads to an increase in the height of the effective barrier. whose height is given to first approximation by the expression: q Bn q M [1. the application of a negative voltage between the semiconductor and the metal electrode of the Schottky contact has the effect of reverse-biasing the Schottky junction. Consequently. Indeed. to modify the intensity of the internal electric field in the junction by applying a bias voltage V between the semiconductor and the metal of the Schottky contact (Figure 1. as in the case of the p-n junction. Reverse-bias of a Schottky junction (n-type semiconductor material) In the case of an n-type semiconductor. it follows that the majority charge carriers (electrons) cannot flow towards the Schottky contact.4. it is the phenomenon of photogeneration of charge carriers inside and near to the SCR which is responsible for the appearance of a photocurrent.

characterized by a very small thickness of metal (of the order of 100 Å) selected to ensure sufficient optical transmission: while a thin layer of gold of 100 Å thickness transmits up to 95% of the incident light in the infrared. and the other junction is forward-biased. as is the case for example with sapphire). Metal-semiconductor-metal (MSM) structures An MSM structure consists of two Schottky electrodes. A bias voltage can be applied between the two electrodes. 1.7. we resort to a semi-transparent Schottky contact. producing a SCR of increased width. which would be the case if all the photogenerated charge carriers were collected by the electrodes at the ends of the device. thus creating the photocurrent. simply acts as a collection electrode. we therefore find that the current flows in reverse through the Schottky junction. The gain of p-i-n photodiodes (other than the specific case of avalanche photodiodes) and Schottky photodiodes is at most 1. The band diagram of the device under increased bias voltage (VB) is represented schematically in Figure 1. The illumination of Schottky photodiodes can occur through the front or rear face (often this second option is chosen in the case where the substrate material is transparent to the light to be detected. consisting of a forward-biased (and hence transmissive) Schottky junction. the percentage transmitted in the ultraviolet is around 30% in the range 300-370 nm.5. The absorption of light near the reverse-biased junction creates electron-hole pairs which are separated under the effects of the electric field present in the SCR. leaving a free semiconductor surface between the two contacts which forms the active region in which light will be absorbed. in order to break the initial electrical symmetry of the contacts: one of the Schottky junctions is reverse-biased. The other electrode. often interlinked in the form of a comb structure. in which L is the distance between two adjacent contact fingers. 0 is the height of the Schottky barrier and . that is. In the case of illumination through the front face.10 Optoelectronic Sensors junction. from the semiconductor towards the Schottky contact.

1. It follows that photogenerated electronhole pairs are more easily separated and collected by the electrodes at either end. allows the electric field of the junction to extend more easily into semiconductor regions some way from the contact.2] . for a given bias voltage. Ri. gain and quantum efficiency The response coefficient of a photodetector. effect of illumination 1. Operational parameters of photodetectors The main parameters which define the behavior of an ultraviolet photodetector are respectively the response coefficient.5. the quantum efficiency. Response coefficient. 1. the gain. MSM photodetectors normally use semiconductor materials which are not intentionally doped. Energy band diagram for an MSM structure under electrical bias.8. links the photocurrent Iph to the power of the incident light Popt through the relationship: I ph Ri Popt [1. The SCRs associated with Schottky junctions made of these materials are hence of significant width which.8.Introduction to Semiconductor Photodetectors 11 Iph is the photocurrent. are chemically very pure and electrically very resistive. the noise equivalent power (NEP) and the detectivity. Figure 1. the bandwidth.

2.6] 1. in the ideal case. Considering that all the incident light is absorbed in the semiconductor material. the rate G of electron-hole pair generation per unit time is thus given by: G Popt / h c [1. the photocurrent as well as the incident optical power are both. The quantum efficiency is defined as the probability of creating an electron-hole pair from an absorbed photon. Temporal response and bandwidth The speed of response of a photodetector may be limited by capacitative effects. from which we obtain the expression for the response coefficient of the detector: Ri q g / hc [1. is given by: Popt / h c [1. These phenomena all lead to a reduction in the response of the photodetector . then the photocurrent is given by the equation: I ph q G g q Popt / h c g q / hc g Popt [1. proportional to the active optical surface.12 Optoelectronic Sensors It is important to note in passing that the response coefficient is a quantity independent of the active optical surface of the photodetector structure: indeed.4] If we now introduce the gain parameter g which corresponds to the number of charge carriers detected relative to the number of photogenerated electron-hole pairs.3] where h is the Planck constant and c is the speed of light.8.602 x 10-19 C). by the trapping of charge carriers or by the saturation speed of charge carriers in the semiconductor. which is defined as the number of photons reaching the active surface per unit time.5] where q is the elementary charge (1. At a given wavelength . the flux of photons arriving on the semiconductor surface.

The temporal response of a photodetector is characterized by the fall time f (or the rise time r). which is defined as the time needed for the photocurrent to fall from 90% to 10% of its maximum (or to rise from 10% to 90% of it). in permanent vibration due to thermal motion. there are five sources of noise: – shot noise. the noise current Ib increases as the square root of the bandwidth of the photodetector device. In the case of a transient exponential response with a time constant . – thermal noise.7] 1. The cutoff frequency fC of the photodetector is defined as the frequency of optical signal for which the response coefficient is half that for a continuous optical signal. it is the smallest optical power which can be measured.2 / 2 d [1. mainly due to the random nature of the collisions of incident photons.8.3. It follows that the NEP parameter is given by the equation: NEP I b / Ri in W [1.8] In the case of white noise. caused by the separation of the electric current into two parts flowing across separate electrical contacts. In other words. Noise equivalent power The NEP is defined as the incident optical power for which the signal-to-noise ratio is 1.Introduction to Semiconductor Photodetectors 13 in the high-frequency domain. . – partition noise. normalized with respect to the bandwidth BW: 12 in W Hz 1 2 [1. the following relationship links the bandwidth BW and the temporal response of the photodetector: BP 1/ 2 2. It follows that it is preferable and customary to use the following expression for the NEP. due to random collisions of charge carriers with the atoms of the crystal lattice.2 / 2 m 2. and hence the photocurrent Iph is equal to the noise current Ib.9] NEP* NEP BW In semiconductors.

the bandwidth and the noise current of the photodetector device. 1. The measurement of the noise current must be made in darkness. written: D* D Aopt BP 12 Ri I b Aopt BP 12 in W 1 cm. Detectivity This figure of merit is defined by the equation: D NEP 1 Ri / I b in W 1 [1. caused by the random generation and recombination of charge carriers.10] In general terms.4.11] The normalized detectivity is the most important parameter for characterizing a photodetector because it allows direct comparison of the performance of photodetectors using technologies and methods of operation which are at first glance very different. . the photocurrent signal increases in proportion to the active optical area Aopt. – 1/f noise. It is clear from the preceding definitions that the determination of the NEP and the detectivity requires measurement of three parameters: the response coefficient. either band to band or via trapping levels situated in the bandgap.8.14 Optoelectronic Sensors – generation-recombination noise. The device is biased using a very stable voltage source. It follows that the preferred method of comparing between different photodetectors is to use an expression for the detectivity normalized with respect to these parameters.Hz1 2 [1. This last type of noise dominates at low frequencies. and in addition the noise current increases with the square root of the product of the active optical area with the bandwidth BW. and the entire measurement system must itself have an intrinsic noise level considerably lower than the intrinsic noise of the photodetector device. associated with the presence of potential barriers at the level of the electrical contacts.

They are normally reversebiased (biased in the non-conducting direction) and used to convert optical fluxes into electrical currents.Chapter 2 PIN Photodiodes for the Visible and Near-Infrared 2. intrinsic and N respectively. a structure which aims to optimize the characteristics compared to those of simple PN junction photodiodes. . Introduction Photodiodes are optoelectronic devices with two electrodes and asymmetric electrical characteristics. orthogonal on the right Chapter written by Baudoin DE CREMOUX. The term PIN refers to the three doped semiconductor layers in their active part. Figure 2. Schematic layout of PIN photodiodes: collinear optical flux and charge transport on the left.1. of types P.1.

16 Optoelectronic Sensors Figure 2. which have a window where required. section 2. Section 2. determines the materials used to fabricate photodiodes.3 and 2. Finally.5 then describes the properties of photodiodes with collinear geometry. This chapter is organized in order to first provide the reader with the basic tools to understand the concepts of PIN photodiodes for a given application.2 recalls some of the physics of semiconductors. then to the external circuit through the metallic contacts deposited on the surfaces. either using radiative or nonradiative processes. currently used in a numerous professional and mass-market systems.700 nm. We use it as the reference type for the analysis of photodiode characteristics. The spectral domain of interest.6 describes . section 2.1 shows the possible arrangements for the optical flux and the electrical current. Figure 2. After this.4 establish the simple quantitative relations between the structure of devices with typical geometries and their static and then dynamic characteristics. then it describes the devices which are currently available or which should be available soon. the absorption of a photon in the I region leads to the transport of an electron in the external circuit. We will stress the compromises required among these characteristics if we are to keep a particular geometry. Sections 2. – transport under the effects of the electric field present in the I region: of the electron towards the N region and of the hole towards the P region. mostly from 400 to 1. Collinear geometry is the most commonly used. or their central layer is at least weakly doped compared to the P and N layers. we will consider orthogonal geometry. Thus. and as a result almost all photodiodes used in this domain have a PIN structure. This design is the most mature of the semiconductors. most probably in the I region. the pertinent characteristics of the materials and the notations used later in the text. With this aim in mind. emphasizing the processes taking place in photodiodes.1 shows the basic processes taking place in a PIN photodiode: – generation of an electron-hole pair through absorption of a photon.

Physical processes occurring in photodiodes 2.2. Electrostatics in PIN diodes: depleted region Depleted region Figure 2.2. profiles of fixed charge density. By integrating the Poisson equation twice in the z direction: d 2V dz 2 z [2.1] .PIN Photodiodes for the Visible and Near-Infrared 17 advanced devices with novel structures which push the boundaries of the previously-discussed compromises. electric field and potential across a reverse-biased PIN diode In the neighborhood of a PN junction there is a region devoid of free charge carriers (the depleted region) where fixed ionized impurities can be found. These are for the most part laboratory devices which could find applications in real-world systems. negative on the P side and positive of the N side.2. forming a space charge region (SCR). 2. From top to bottom.1.

In the case of an abrupt PN junction. as for a conventional capacitor. In the realistic approximation that the densities of ionized impurities are much greater than the density of residual impurities in the I region (typically by a factor of 100). by: CJ S d [2. its value is E V d [2.3] These approximations will be used throughout the remainder of the chapter unless explicitly stated otherwise. It is given. insofar as V >>VBI. of the electric field and of the potential. showing in sequence the profiles across the structure of the density of fixed charges. The case of a PIN diode structure is considered in Figure 2. we can obtain the profiles of the electric field E and the potential V across the structure. the capacitance of the junction decreases when the reverse-bias voltage increases in absolute value.18 Optoelectronic Sensors Where (z) is the charge density and is the dielectric constant. – the electric field is constant in the depleted region and.2] where S is the area of the junction. the thickness of the depleted region varies as (VBI –V) where VBI is the built-in or diffusion potential present in the absence of external bias applied to the diode.2. – the capacitance of the junction is independent of the reverse-bias voltage. and V is the voltage across the terminals of the diode. . As a result. is a very important parameter in the definition of a PIN photodiode. represented by d in the rest of this chapter. we draw the following conclusions: – the extension of the depleted region into the P and N regions is negligible. and the depleted region is to all intents and purposes coincident with the I region whose thickness.

2.5] . shown schematically in Figure 2. a photon can be absorbed. according to the relationship: dF Fdz [2.PIN Photodiodes for the Visible and Near-Infrared 19 2. Thus.4] Figure 2. The parameter EG is thus a criterion in the choice of materials when designing a photodetector. is the inverse of the process of recombination which gives rise to stimulated emission of photons in laser diodes.1. Mechanisms of electron-hole pair generation 2. This process of radiative generation.2. Radiative generation The material is transparent to photons with lower energies. Radiative generation and optical absorption Electronic transitions between permitted levels in the valence band (VB) and levels in the conduction band (CB) can occur under the action of photons transferring their energy to the electrons. It follows from this that the only photons absorbed efficiently are those whose energy h is greater than the width of the bandgap EG of the material under consideration: h EG EC EV [2.3. Optical absorption is characterized by the coefficient of absorption (cm–1). creating an electron-hole pair.2.2.3.

The shape of these graphs and the magnitude of depend on the direct or indirect nature of the band structure of the material. Figure 2.4. the number of electronhole pairs created per unit volume and time. the variations in energy of the electrons. The level of radiative generation GR (cm–3s–1).4 which represents. is therefore: GR F [2.16 shows graphs for the main semiconductors used in the spectral domain under consideration in this chapter. for the two types of material. with the wavevector k describing their motion in the periodic crystal lattice.6] It follows from these definitions that the level of radiative generation in a homogenous material.20 Optoelectronic Sensors where F (cm–2s–1) is the photon flux propagating in the z direction. This nature is defined in Figure 2. Figure 2. for a photon flux F0 falling on a surface at z=0. the knowledge of variation of with photon energy (or with their wavelength) is essential for the choice of materials for use in a photodetector. Direct bandgap on the left and indirect bandgap on the right . Furthermore. can be written: GR F0 exp z [2.7] We see that 1/ (cm) gives an order of magnitude for the distance required to absorb the radiation.

This is one of the contributions to the dark current of photodetectors. radiative transitions must simultaneously conserve the energy and momentum of the particles involved (the photon and electron). and radiative transitions require the involvement of the third particle.2. and as a result radiative transitions are effectively vertical and connect with a high probability those states at the top of the VB and those at the bottom of the CB.5. The opposite is true for materials with an indirect bandgap (such as silicon). Furthermore. in the SRH (Shockley Read Hall) process shown in Figure 2.PIN Photodiodes for the Visible and Near-Infrared 21 For materials with a direct bandgap (such as gallium arsenide) the main minimum in the conduction band is found at the same vector k as the maximum in the valence band. 2.2. to provide the difference in momentum between the initial and final states.2. the required energy is “borrowed” from the thermal motion of the crystal lattice. The momentum of the electron is much greater than that of the photon. Non-radiative generation As well as the radiative processes there are also non-radiative processes of generation-recombination.5 in generation mode. These transitions are much less probable and grows much less quickly with the energy of the photons above EG. Thus. a detrimental parasitic phenomenon. a phonon. Figure 2. Non-radiative SRH generation .

10] where NC and NV are the equivalent densities of state for the CB and VB. thermal velocity th – NT. etc. electron and hole densities. – kT. intrinsic electron density.). – n and p.22 Optoelectronic Sensors The transition takes place in two parts. 35]: GNR n n p th N T ni2 p np p ni exp ET kT EI n ni exp ET kT EI [2. and define an effective GNR ni th NT 2 ni eff [2. We can draw the following conclusions from this: . density of deep states.9] We also know that: ni2 N C NV exp EG kT [2. p. In the depleted region of a photodiode we have np = ni2 . – vth. p capture cross-sections for electrons and holes by the 3k T m * . If we further assume that lifetime eff: n= p= and ET = EI. The rate GNR (cm–3s–1) of SRH generation-recombination can be written [SZE 81. – ni. via a “deep state” of energy ET. Boltzmann factor. which is close to the middle of the forbidden band and is the result of a crystal defect (impurity.8] with the following notation: – n and deep state. – EI. vacancy. middle of the forbidden band.

– GNR is not an intrinsic characteristic of the material. It is exploited as a gain mechanism in the avalanche photodiodes which are described in Chapter 3. are transported in the device to the two electrical contacts that transfer it to the external circuit – specifically. Another process of non-radiative generation is shock ionization which occurs in a semiconductor material under a strong electric field.PIN Photodiodes for the Visible and Near-Infrared 23 – GNR increases very rapidly as the ratio EG/kT decreases. limiting the reverse-bias voltage that can be applied. for silicon the most effective is the deposition of its natural oxide SiO2. to the load resistance across the terminals. Transport mechanisms The charge carriers. In an electrically neutral region of type p(n). across which the electrical signal is measured.2. 2. and whose structure is optimized with this in mind.11] SRH non-radiative generation is particularly active at the surface of devices. It can be reduced by a passivation treatment. where the electron (hole) deficiency relative to thermodynamic equilibrium is n( p). silica. but depends for its purity on NT. . electrons and holes generated by the aforementioned processes. It grows with EG and the breakdown field is of the order of EB = 105 V/cm for the materials considered.3. if we define n( p) as the lifetime of the minority electrons (holes). This effect is one cause of breakdown in diodes. it follows that: G NR n th N T n n [2. where the density of defects is very high.

24 Optoelectronic Sensors 2.6. as shown in Figure 2. the velocities are proportional to E (linear regime) and we have: n. the electron velocity passes through a maximum (high mobility regime).6. Characteristics n(E) and p(E) in logarithmic coordinates For electric fields such that E < 103 V/cm.12] where n and p (cm2/Vs) are the respective mobilities of electrons and holes. Transport under electric field and resultant current In the depleted region of a photodiode (normally reverse-biased) an elevated electric field is present. Figure 2.2. . p E [2. p E n. Between these two regions. and for certain materials. and the electrons and holes are driven at a speed which depends on the intensity of the field. the velocities are fairly constant and equal to a saturation velocity of the order of vs = 107 cm/s.3.1. For electric fields such that E > 104 V/cm.

From this we can extract the parameters commonly used to characterize photodiode operation.14] 2.2.2. p Dn. p [2. p Dn. Transport along the concentration gradient and diffusion current In the contiguous electrically neutral regions in the depleted region. The various recombination processes are also active in these regions. The photocurrent IP produced by a photodiode is proportional to the illumination across a large range of optical power P. p kT q cm2 s are the diffusion coefficients. .PIN Photodiodes for the Visible and Near-Infrared 25 2. p n.13] where Dn.3. The coefficient of proportionality is the response coefficient R (A/W) (responsivity). Their fluxes Fn and Fp can be written: Fn. such that the electrons and holes are only efficiently transported over a distance known as the diffusion length and given by: I n. which measures the minimum optical power detectable by an optical detector. and which depends on the circuit into which the photodiode is inserted and the specifications of the signal to be received.1. p n. I/V characteristics and definition of static parameters Figure 2. not to be confused with the sensitivity. whose relations to the structure of the device are described in the following sections.3.7 shows I/V curves. p z [2. linking the current crossing a photodiode to the voltage across its terminals for three illumination conditions.3. electrons and holes can be transported in the absence of an electric field under the effect of their concentration gradients. Static characteristics of PIN photodiodes 2. p n.

26 Optoelectronic Sensors Figure 2.4).3.7. we have: R Ip P q h e [2. The saturation current IM is the photocurrent above which several mechanisms limit the linearity of the response: – the intrinsic mechanism. . I/V characteristics of a photodiode. inserted in the circuit shown and under several illumination conditions The external quantum efficiency e links the generated electron flux to the incident photon flux.3. is caused by nonradiative generation and/or the tunnel effect.3.15] where q is the charge of an electron and h is the energy of a photon. The dark current I0. Thus. discussed in section 2. due to the screening of the electric field in the depleted region by the space charge that is produced by the mobile charge carriers created by the illumination (see section 2.

1.PIN Photodiodes for the Visible and Near-Infrared 27 – the circuit-dependent mechanism. which are both dependent on the wavelength. by depositing an antireflection .3. For V=VB the dark current grows very rapidly and the power dissipated in the diode can lead to its destruction. used as the reference situation. Optical efficiency The optical refractive index n of the semiconductors under consideration is in the region of 3. External quantum efficiency The external quantum efficiency defined above can be thought of as the product of an internal quantum efficiency i and an optical efficiency o. caused by the voltage drop across the load resistance (see the circuit in Figure 2. are considered in the following text.7). 2. These two ratios. 2.16] can be obtained under a high reverse-bias voltage VAV in avalanche photodiodes (see Chapter 3). A gain. characterized by the multiplication coefficient M defined by: M I Ip I0 [2. the Frensel transmission coefficient determines the optical efficiency: 4n o n 1 2 [2. the ratio of the photon flux entering the device to the flux of photons incident on its surface.2. and as a result the incident light rays are refracted and reflected at its surface.17] which is of the order of 70% and depends little on the wavelength and the angle of incidence.2. The breakdown voltage VB is the highest reverse-bias voltage the diode can support.3. It is possible to achieve optical efficiencies close to 100% at a given wavelength . At normal incidence.

which is below the energy of the photons to be detected. 2. leading to the passage of one electron for each absorbed photon: thus. The optical efficiency therefore depends on and The electron-hole pairs created by the absorption of photons entering the device may.3. If generation is in the depleted region. Internal quantum efficiency n and 4 n . The curves 1 and 2 represent the profiles of the radiative generation rate GR across two structures.28 Optoelectronic Sensors coating on the surface. In the case of generation in the electrically neutral P and N regions. depending on where they are generated in the structure.1. the electrons and holes are separated by the electric field. and part of the radiation is absorbed there. The quantum efficiency can therefore be written: i exp dp 1 exp d [2. give rise to a current flow in the external circuit. Figure 2. and has a quantum efficiency of 100%. or may not. only the electrons (or holes) generated less than one diffusion length away are transported into the depleted region. as shown in Figure 2.8 shows a model which provides a good evaluation of the internal quantum efficiency of a PIN photodiode. Finally. where the electric field is weak. Another part is absorbed in the i region. as a function of the nature of the material on whose surface the radiation falls.8. the internal quantum efficiency is 100%. The structure for curve 1 is that of a homojunction PIN. It is therefore desirable to design a photodiode in such a way that the radiation should be entirely absorbed in the depleted region. where all of the materials used have the same width of bandgap. the contribution of the residual radiation entering the n region is also negligible.2. and the internal quantum efficiency is in most cases weak. .18] where the definitions of dP and d can be found in Figure 2.2. The radiation is assumed to be incident from the p side. for example of refractive index thickness on the angle of incidence. where the quantum efficiency is negligible.

These two processes are represented in Figure 2. curve 2 shows a heterojunction PIN photodiode Curve 2 corresponds to a PIN photodiode. Optical generation rate across the structure of two PIN photodiodes: curve 1 shows a homojunction PIN photodiodes.19] 2. The same reasoning provides us with the quantum efficiency: i 1 exp d [2.PIN Photodiodes for the Visible and Near-Infrared 29 Depleted region Figure 2. . Dark current The dark current is the sum of several contributions: non-radiative generated currents requiring an input of energy (such as the SRH process) and the tunneling current.3.8. with the capital P indicating that the material in the P region has a bandgap width greater than the energy of the photons to be detected and which is therefore transparent to them.9 for a sharp p+n homojunction. which does not require such an input.3.

9. Figure 2. the dark current contribution of generation in the depleted region is: I0 SdGNR Sdni eff [2.2) – the generation contribution in the neutral p and n regions can be written: I0 Sqni2 Ln N A n Lp N D p [2.21] This is the traditional expression for the reverse current of a diode according to Shockley.9].3. Profile of the VB and the CB of a reverse-biased p+n junction. p ni2 N AD .20] Taking account of equation [2.11] – of the fact that the shortage of electrons (holes) in the p(n) material near the depleted region is n.1. the tunnel process occurs at constant energy .30 Optoelectronic Sensors 2. and the fact that the thickness of the p(n) material is Ln. Generation currents According to equation [2.p (see section 2.3. The generation process (shown here in the depleted region) requires an input of energy.3.2.

3. such that the contribution of the tunneling current to the dark current is most significant for photodiodes made of materials with a small bandgap.2. From equation [2. the two parameters which reduce the width and the height of the potential barrier at the PI junction.2). mostly as a result of the increase in E. The tunneling effect can be the dominant one in the materials used for detection of the longest wavelengths considered in this chapter (see section 2.20] and [2.22] we determine that. This mainly applies to materials with a narrow depleted region and a small bandgap. if the doping level of the i zone is too high [TAK 80].1.3. IT increases when EG decreases. .5.22] where E=V/d is the electric field in the depleted region.4.3.3.3. IT increases very rapidly with the voltage V across the terminals of the diode. 2. Tunneling current Using the tunneling effect electrons can also be transported at a constant energy across the triangular potential barrier present at the junction. we see from equations [2. as long as it is narrow enough.21] that the generation current is primarily a function of ni and varies very rapidly with EG/kT. which is constant for a PIN diode. for a given material. Breakdown voltage 2. A calculation of the current gives the following expression [STI 82]: 2m * q3 EV 1 4 3h2 EG 2 3 2m * EG 2 IT S exp 4qhE [2. 2. Zener breakdown via the tunneling effect The very rapid variation of the tunneling dark current with the reverse-bias voltage is the first process of breakdown.PIN Photodiodes for the Visible and Near-Infrared 31 Thus. Additionally.4.

23] 2.4.7 that for a bias voltage V0. the photodiode must remain reverse-biased (V<0) so that the carriers are separated by the electric field in the depleted region. Avalanche breakdown We can define a critical electric field EC.3.5. the voltage drop in the load resistance RL can cancel out the voltage V across the diode terminals for a limiting current IM such that: IM V0 RL [2.2.3.2.5. most often around the edge of the junction where the point effect increases the electric field.5. Thus.1. In the case of PIN photodiodes. where the multiplication coefficient becomes infinite and leads to breakdown.32 Optoelectronic Sensors 2. . which are not designed to take advantage of avalanche gain. Extrinsic the load resistance limitation by the voltage drop across In order to have a linear response.25] in order to avoid breakdown of the junction. it can be seen in the simple bias diagram of Figure 2. breakdown can arise locally.24] 2.3. of the order of 2 x 105 V/cm for the lightly-doped materials considered in this chapter. Specifically.3. The breakdown voltage is therefore approximated below by: VC EC d [2. E must satisfy the following conditions: 0 E EB [2. Intrinsic limitation by the space charge of mobile charge carriers It is especially important to take into account the contribution of mobile charge carriers to the electric field E in the depleted region for high levels of incident optical power. Saturation current 2.

10 shows the density profiles of the fixed and mobile charges and the electric field profile across the structure.PIN Photodiodes for the Visible and Near-Infrared 33 We will determine the maximum value IM of the photocurrent that simultaneously satisfies these conditions for a heterojunction PIN structure.26] where CJ is the capacitance of the junction. obtained when E E B . Depleted region Depleted region Figure 2. and all the illumination is absorbed close to the Pi junction. Figure 2. with a coefficient of absorption in the i region such that d>>1. the d and the electric field variation in the depleted region is E maximum current IM. for I=0 (dotted line) and for I=IM (solid line). the electric field varies from 0 to the breakdown field EB . We further assume that the mobile electrons in the i region are transported with the saturation velocity vs. independent of the electric field.10. Given the hypotheses set out in this section. where J is the current density. Space charge density and electric field. it follows that the charge density due to electrons is J s . For I=IM. can be written: IM CJ s EB [2.

2. for this value of current. the voltage across the terminals of the junction is V EB d 2 VB 2 [2.4.11 shows the electric charge profile in the depleted region at a time t after a brief impulse signal.34 Optoelectronic Sensors Note that. Intrinsic limitations to the speed of response 2. 2. Dynamic characteristics of PIN photodiodes 2. regardless of the components in the bias circuit. Figure 2. which is a good representation of the properties of real PIN diodes. Its position depends on the values of V0 and RL. Depleted region Figure 2.11.1. Figure 2.1. which was applied to the photodiode at time 0. t) .1. Density of mobile charges produced in the depleted region of a PIN photodiode in response to an optical impulse (z. of energy W0. The packet of generated electrons is transported at speed vs and takes time t =dvs to cross the depleted region. Its analysis is carried out here with the same approximations as in section 2.1.4.10 provides an example where RL = 0 and the voltage across the diode terminals is constant and equal to the bias voltage V0 = VC/2. Transit time in the depleted region The transit time of photocarriers in the depleted region is the first limit to the response time of the photodiode.27] We can easily show that.4. the part of the curve representing the variation of the electric field in the depletion region (marked ) rotates around a fixed point as the photocurrent varies.

Impulse and frequency response of a PIN photodiode limited by the transit time The solution of Maxwell’s equations for the case RL =0 shows that the current generated by the diode is the rectangular impulse shown in Figure 2. In the case where d <<1 and the generation is uniform in the depleted region.4.PIN Photodiodes for the Visible and Near-Infrared 35 Figure 2.2 shows that we obtain the following numerical result: ft 0. of duration t and of intensity: I0 q h W0 t [2.55 t [2.12.31] We see that regardless of the value of d.2.12.44 2 is: [2. This is shown in Figure 2. which we will continue to use: . the more accurate model presented in section 2.30] We see that this varies as 1/d.12 and given by: I f I0 sin I0 t f 1 t f t [2.29] The 3 dB cutoff frequency I f ft 0. we can apply this approximate numerical expression.28] The frequency response can then be obtained by Fourier transforming the impulse response.

n or p. which has a small bandgap. 2.6.32] and [2.3.33] that the product only depends on the characteristics of the material in the i region: i ft s 2 [2.1.19] shows that if d <<1 we have: i d [2. In practice the diffusion time is always greater than the transit time in the depleted region and has the effect of reducing the response speed of photodiodes. Trapping at the Pi heterojunction An additional parasitic mechanism can slow the speed of heterojunction PIN photodiodes. excluding some exceptions (see section 2.32] Also. Diffusion time in the neutral regions When the optical generation occurs in one of the electrically neutral regions. This process can be characterized by a diffusion time d which we can show has an order of magnitude: d d 2 Dn p [2.36 Optoelectronic Sensors ft 12 t [2. it is desirable to design these devices such that the absorption of photons takes place in the depleted region. can be trapped at the interface .4. both to optimize the speed of response and the quantum efficiency. The holes generated in the i region. 2.4. Therefore. we must accept a compromise between the internal quantum efficiency and the speed of response.34] from which we conclude that. equation [2. the minority photocarriers must initially be transported by diffusion to the edges of the depleted region.2.33] i ft Thus we see from equations [2.35] for electrons (holes) generated in the p(n) region. for the geometry considered and for a particular wavelength and material.1.4). in which they can be transported by the electric field.

4. whose effective mass is greater than that of electrons. into the P region.2. with a negligible impedance of the source at high frequencies. although with delay. 2. The most simple circuit is shown in Figure 2.4. Limitations due to the circuit 2.13. . Response of the i region of the basic circuit. a photodiode is inserted into a circuit whose components also contribute to limit the bandwidth of an optical detector. The proposed remedy [FOR 82] is the use of gradual heterojunction or layers with an intermediate-sized bandgap in order to reduce the potential barrier presented to the holes and to make it easier for them to cross through tunneling or thermal effects. whose response is independent of the voltage at its terminals. consisting of the capacitance of the junction CJ in parallel with a load resistance RL. This circuit is driven by the photodiode generated current. This effect is particularly marked for holes.2. They can then be emitted by thermo-ionic effects. We will assume below that the structures of the devices are designed in such a way that these effects are rendered negligible.PIN Photodiodes for the Visible and Near-Infrared 37 with the P region by discontinuity in the valence band.1. optimum thickness In order to produce a signal.

. The frequency response is given by the expression: I f 1 I0 2 fRL C J 2 [2. 3 dB cutoff frequency of a PIN photodiode.13. thus there is an optimal thickness for the depleted region which maximizes the global cutoff frequency f3dB.36] and the 3 dB cutoff frequency is: f RC 1 2 RL C J [2.13 shows the cutoff frequencies variation with d for a typical PIN photodiode. The cutoff frequencies f and fRC vary inversely with d. as a function of the thickness of the depleted region.38 Optoelectronic Sensors Figure 2. The material is assumed to be GaInAs/InP. Figure 2. the width of the active region 10μm and the load resistance 50 The impulse and frequency response of such a resistor-capacitor network are well known.37] We see that fRC varies with d.

with five examples of the parameters d and L listed in Table 2.8 106 cm/s.39] 2. The load resistance was kept constant: RL = 50 . designed for optical telecommunications at a wavelength of 1.3 – the different saturation velocities for electrons and holes n = 6 10 cm/s and p = 4.1 along with the performances obtained.14 shows the response curves of the photodiode inserted in its circuit.PIN Photodiodes for the Visible and Near-Infrared 39 The optimal thickness of d is obtained when f = fRC . and the diameter of the active region is assumed to equal 30 μm. caused by the connecting wires. 4 6 = 1.4. – the inevitable parasitic capacitances of the circuit: CP = 50 fF.2. . – a parasitic inductance L.2. taking into account: – the finite value of the optical absorption coefficient: 10 cm–1. Figure 2.38] The maximum cutoff frequency is then f3dB s 2 2d [2. the cutoff frequency and the internal quantum efficiency. The illumination is assumed to be incident at the N side of the structure. adjustable and kept as a free parameter. giving: t RL C J [2. Response of a real PIN photodiode It is also interesting to consider the results of a more complete analysis [BOW 87] of a photodiode made of GaInAs/InP.3 μm.

Curve 3 corresponds to a depleted region whose width is close to the optimum. The parameters of each curve are given in Table 2.17 Figure 2. For a higher value of inductance (curve 5).1. the bandwidth is reduced. .40 Optoelectronic Sensors Table 2. inserted into the circuit shown. but a gain in signal is obtained.3μm. Frequency response of a PIN photodiode made of GaInAs/InP at 1.1 Curves 1 and 2 show behaviors limited by the transit time and by the circuit respectively. while curve 4 shows that its response can be further improved by parasitic induction.14. Values of the parameters for the curves in Figure 2.

defined mainly by the thickness d of the depleted region and by the area S of the active region.15) between 0 and the maximum value IM given by equation [2. the photocurrent must vary sinusoidally (see Figure 2. the problem is now to design a photodiode. The maximum effective power is therefore: PM 1 VC 2RL 4 2 2 EB d 2 32RL [2. Power-frequency compromise.40] . as well as the voltage varying between the breakdown voltage VB given by equation [2. as in Figure 2.25]. Given the maximum frequency to be transmitted. Excursion from the operating point of the photodiode in a plot of its I/V characteristics: allowing us to determine the maximum RF electrical power.4.3. A’ and A’’ are respectively the mean and extreme points In order to maximize the RF power dissipated in the load. We assume that the incident optical power is modulated at a level of 100%. Points A. P f 2 “law” For certain optical transmission systems. as well as select the bias voltage V0 in a circuit.23] and the voltage VB/2 (equation [2. In particular. Figure 2.15.PIN Photodiodes for the Visible and Near-Infrared 41 2. this allows us to simplify the electronics of the receiver. and is made possible by the current availability of powerful modulated optical sources or optical preamplifiers.27]).7. it is desirable to use a distortion-free increased electrical power in the sinusoidal regime across the resistance RL loading the photodiode.

in addition. thermal effects can severely limit it.1. 2 the product PM f3dB only depends on the properties of the semiconductor material used for the i region. Absorption of semiconductors in the range 400-1. these will be mentioned briefly later on. They can be chosen as a function of the wavelengths to be detected.800 nm The absorption functions of the main semiconductor materials or semiconductor families used to build PIN photodiodes in the spectral range considered in this chapter are shown in Figure 2.41] This expression shows that. 2. Semiconductor materials used in PIN photodiodes for the visible and near-infrared 2.5. We can note that this limitation a photodiode can generate on the RF power is of a purely electronic origin.39] and [2.40].5. Other materials have been studied in the past. Equation [2. in particular for components designed for low frequencies. .16. it follows that: 2 EC 2 s 2 PM f3dB RC 256 [2. for a fixed value of the load resistance. but have not taken off in practice for numerous reasons. We should note that the numerical coefficient 1/256 is a maximal value which may in fact depend on the detailed specifications of the nonlinearity of the transmission system considered.42 Optoelectronic Sensors If we assume that the photodiode is optimized for its frequency response.41] is analogous to the known result for electronic components such as transistors [JOH 64]. and we eliminate d from [2.

.2. . Silicon Silicon has an indirect bandgap and its absorption coefficient slowly increases below the corresponding wavelength ( G = = 800 nm. produces very few electronic states in the bandgap.2. as yet. contributed to the design of heterojunction optoelectronics with competitive performance. and as such they are the natural choice for this spectral domain. On the other hand. SiO2. this semiconductor has not. Figure 2. However.100 nm).5.5. the shape of their absorption functions demonstrates their difference in bandgap nature. . . From 400 to 900 nm: silicon and the GaAlAs/GaAs family These two materials have achieved a good technological maturity under the demands of the integrated circuit market.PIN Photodiodes for the Visible and Near-Infrared 43 . which is indirect for silicon and direct for gallium arsenide.1. reaching the value of 103 cm–1 for interface with its natural oxide. The 1. Absorption spectra for semiconductor materials used in the visible to near-infrared range 2. and ultimately very high density integrated circuits. 2.16. . which is the main factor enabling the development of planar and MOS technologies.

and offered at a low price. thanks to the effectiveness of the surface silicon passivation. in particular for components whose response has been extended into the infrared.17 shows the spectral response of a silicon photodiode designed for optical fiber telecommunications. they are used for short-range telecommunications in numerous professional systems (automation.). in particular for detection of the = 1. photographic equipment. etc.060 nm radiation from solid-state neodymium-doped YAG lasers. potentially with multiple regions of sensitivity. As a result. The dark current is mostly caused by current generation in the depleted region. can also be minimized in the near-ultraviolet by several methods. Lidar. The maturity of silicon technology allows the construction of a variety of photodiode types. which is of the order 10–7 A/cm2.44 Optoelectronic Sensors Figure 2. CCD and MOS imagers also form part of the family of silicon-based photodetectors. It is possible to extend the spectral response slightly towards the infrared by increasing the thickness of the intrinsic region. The structure of the device is chosen (with a depleted region thickness around 30 μm) in order to increase quantum efficiency to between 600 and 900 nm where weakly-attenuating fibers and high performance electroluminescent emitters are available. The drop in efficiency at short wavelengths. due to optical absorption in the surface layer. .) or mass-market systems (optical disk readers. etc. The response speed is often limited to less than 1 GHz by the transit time.

Gallium arsenide and similar materials Gallium arsenide has a direct bandgap. and components showing good performance have been demonstrated [LAW 79]. Spectral response of photodiodes.17.2. it is available as a high-quality substrate for creating defect-free heterojunctions with Ga1–xAlxAs. absorption in the direct-bandgap i material. Highly developed for the construction of microwave integrated circuits and diode lasers. In . A spectral response function is given in Figure 2. Ga1–xInxAs1–yPy and AlxGayIn1–x–yAs alloys.5. The density of the dark current is of the order of 10–8 Acm–2 at VB/2 and a product i ft of the order of 20 GHz can be achieved for conventionally-structured devices. Alongside heterojunctions. Figure 2. its absorption coefficient is approximately 104 cm–1 for = 800 nm and increases very rapidly below the corresponding wavelength G ( G = 870 nm). showing the typical shape for a heterojunction photodiode: rapid cutoff at long wavelengths.17.2.PIN Photodiodes for the Visible and Near-Infrared 45 . the use of Schottky diodes based on GaAs has also been demonstrated. and cutoff at short wavelengths due to absorption from the large-bandgap window layer. made of silicon for optical telecommunications and of experimental GaAlAs/GaAs without antireflection treatment 2. internal quantum efficiency close to 100%. In fact. there has been interest in heterojunctions from the GaAs/GaAlAs system for photodiode and solar cell fabrication since the 1970s [ALF 70].

initially in germanium – which predated the emergence of optical telecommunications – and then in various families of semiconductor compounds. several factors finally brought an end to their application in this domain. 2.3. discussed in Chapter 5. GaAs-based photodiodes have not seen significant development since 2001. then more rapidly for < 1. except in solar cells. can also be thought of as Schottky photodiodes. Germanium Germanium has an indirect bandgap and its absorption coefficient increases slowly below G. This is probably due to the cheap cost of silicon components whose speed of operation is adequate for realworld systems. the deposition of ITO (indium tin oxide). 2. with the GaInAs alloy on InP substrate finally prevailing. a conductive and transparent material.5. These systems have driven the development of photodetectors.1.3. The need for photodetectors for telecommunications was the drive for an important effort in the 1970s in the development of germanium photodetectors [KAN 85].46 Optoelectronic Sensors addition. From 900 to 1. This situation may change if the need arises for optical transmissions above 1 Gbit/s within the spectral range covered in this section. Despite these encouraging and already well-known characteristics. has allowed the development of fast photodiodes based on GaAs [PAR 87].4). along with the increasing use of GaInAs/InP: – non-existence of materials allowing the construction of heterojunctions.5.800 nm: germanium. MSM (metal-semiconductor-metal) photodiodes. GaInAsP/InP… This spectral range is mostly used for long distance transmissions and broadband over silica fibers. .500 nm where its conduction band also has a direct minimum for k = 0 (see Figure 2. However.

43In0.2x) all have an indirect bandgap.650 nm). . . due to the shape of the absorption spectrum.PIN Photodiodes for the Visible and Near-Infrared 47 – limitation of the spectral response around 1.5. The dark current densities are of the order of 10–4 Acm–2 and the cutoff frequencies are limited by the circuit to a few hundred MHz. .57As having the smallest bandgap of the family ( G = 1. 2.3. Indium phosphate and related materials Solid solution crystals from the GaxIn1–xAsyP1–y family formed into a crystal lattice on InP substrates (y 2.16.18 shows the spectral response of a germanium photodiode aimed at scientific instrumentation. depending on the active surface. . .18. The composition of the solid solution can be selected to give a bandgap width anywhere between those two values. Figure 2. . – high dark current due to an insufficient control of the surface preparation. Spectral response of state-of-the-art photodiodes in 2000: in germanium for instrumentation and in GaInAs/InP for optical fiber telecommunications Figure 2. is shown in Figure 2.550 nm for fast photodiodes. as well as that of InP ( G = 920 nm). Its absorption spectrum. with Ga0.2.

but we note in particular that the PIN junction emerges at the surface in the wider-bandgap InP material.000 nm with the use of unmatched materials. It is analogous to that of the devices described above. 2) deposition of the dielectric and etching of openings. 3) zinc diffusion (p-type impurity). as we saw in section 2.2. This can be extended beyond 2.4. required for long distance transmissions. at the expense .700 nm when the absorbing layer is exactly matched to the InP substrate. Fabrication stages of PIN photodiodes in GaInAs/InP by “planar” technology: 1) epitaxial deposition of the absorber and window layers.2. The spectral response is limited to around 1. Figure 2. The transparency of the substrate allows the possibility of the illumination being in this direction (rear illumination). This diode has had an antireflection treatment applied which allows it to approach an external quantum efficiency of 100% across a large spectral band. The main stages in the “planar” fabrication procedure for PIN heterojunction photodiodes are shown in Figure 2.19.19.18 shows the spectral response of a photodiode targeted for optical telecommunications. The dark current density is of the order of 10-5 Acm-2 and. which contributes significantly to the reduction in the dark current. and 4) attachment of metallic contacts Figure 2. electroluminescent diodes and laser diodes. It has been of benefit to photodetectors which have established themselves as rivals to germanium photodetectors. the cutoff frequency of a photodiode with an optimized structure can surpass 20 GHz while retaining a high quantum efficiency.48 Optoelectronic Sensors The technology of these materials was initially driven by the development of electroluminescent emitters.

Nevertheless. also used for electroluminescent emitters. a weaker additional avalanche noise was achieved (see Chapter 3). In this way it has become possible to exceed the limit of the product i ft. On the other hand. 2. it was possible to demonstrate photodiodes with competitive quantum efficiencies. etc. although with a higher dark current (by a factor of ~10) than GaInAs/InP photodiodes. Mercury Cadmium Telluride) integrated into CdTe substrates is the dominant material for detection and imaging in the mid-infrared. the industrial significance of GaInAsP/InP.3.6.720 nm). for certain applications (imaging. towards the end of the 1970s. New photodiode structures 2.PIN Photodiodes for the Visible and Near-Infrared 49 of the dark current.3.6. as in the case of GaInAsSb/GaSb. By combining it with GaAlAsSb. for the fabrication of photodiodes for optical transmissions. The HgCdTe (MCT. Attempts were made to re-use the technological achievements in the infrared for the construction of detectors for telecommunications. Other materials [PEA 85] The integration of the GaAlAsSb family into GaSb substrates was also considered by several groups. Beyond the limits of conventional PIN The development in the last decades of new epitaxial techniques for the integration of very thin layers (only a few atoms thick) of semiconductor compounds has enabled the design of optoelectronic devices. seems to have prevented investment in research which might have taken this technology to new levels. GaSb is a direct semiconductor whose bandgap is close to that of GaInAs ( G = 1. the potential gains were insufficient to justify its industrial development. spectroscopy.) [LIN 92]. Although the performances obtained were comparable to those of GaInAs/InP.1. including photodetectors with far more complex structures than the simple PIN photodiodes with three useful layers. 2.5. which .

after reflection from a mirror at the back (see Figure 2. Photodiodes of collinear geometry with an improvement in the i ft tradeoff .50 Optoelectronic Sensors we examined in section 2.20). 2.1 (for the PIN photodiode geometries discussed). as shown in Figure 2. Photodiodes with collinear geometry A first approach to increase the absorption of radiation by a layer of thickness d << 1 is to cause the light to pass through the absorbing layer again.2. Several avenues were explored with this aim. as shown in the figure. One technique for increasing the number of crossings of the absorbing region is to insert the i layer between two mirrors. which results in a doubling of the absorbing thickness. Such a mirror can be fabricated either by metalizing the rear face of the structure or. whose order of magnitude is 30 GHz for indirect bandgap materials. These mirrors made from a stack of layers with an alternating sequence of bandgaps (and hence refractive indices) of thickness /2n allow reflection coefficients approaching 100% to be achieved.1. possibly Bragg mirrors.20. They were first used in the fabrication of vertical cavity surface emitting lasers (VCSELs) and more generally in microcavity devices. by a Bragg semiconductor mirror. all based on making the radiation pass through the narrow i absorption region multiple times (as shown in Figure 2.6. to form a Fabry Perot cavity (or interferometer). Figure 2.20) in order to increase the internal quantum efficiency while maintaining a short electronic transit time.4.20.

but it turns out to be difficult to show improvement in the i ft product. However. RCE photodiodes can be thought of as the monolithic integration of a photodiode and an optical filter whose selectivity can be exploited beneficially.17 and 2. as can the internal quantum efficiency. which gives rise to the name RCE (Resonant Cavity Enhanced) photodiodes. – for a given wavelength. in particular if it can be made tunable. . Analysis of the response speed [UNL 95] shows that a product i ft approaching 100 GHz should be obtainable by RCE photodiodes. the resonance condition is satisfied only for certain angles of incidence. The experimental results obtained up to now for most of the families of III-V materials have indeed confirmed the potential of RCE photodiodes in terms of quantum efficiency and selectivity. – the possibility of using partially incommensurate i materials without loss of crystal quality if its thickness is below a critical value. this resonant behavior has a marked effect on the characteristics of the device: – the spectral response consists of a series of peaks which become even narrower as the fineness of the cavity is increased. which is in contrast to the normally broad response of conventional photodiodes (see Figures 2. This enhancement is obtained for a series of resonant wavelengths satisfying the above condition.PIN Photodiodes for the Visible and Near-Infrared 51 The analysis of the behavior of such a structure [KIS 91] shows that when the effective distance between the mirrors is an integer number of half-wavelengths of light in the material.20]). the absorption can be considerably enhanced. and the i layer is placed at an antinode of the electromagnetic field. such that the directivity is much higher than that of conventional devices.18). In addition. other advantages of the use of very thin detector layers are anticipated including: – the reduction of the contribution of pair generation to the dark current (equation [2.

21. are transparent and also have a lower refractive index. giving the device the shape of an optical waveguide. a geometric optics representation.3. Numerous variations on this principle have been investigated. where we see an absorbing i layer inserted between several other layers which have a wider bandgap. This is the case in Figure 2. Figure 2. allowing independent adjustment of the transit time (by varying the thickness of the i depleted region) and the waveguide thickness (which determines the degree of coupling with the incident light and consequently the external quantum efficiency [KAT 92]): .52 Optoelectronic Sensors 2. Figure 2. Structure. leading to an internal quantum efficiency of 95% for a waveguide of length 300 μm.21. the light rays propagate in the structure through a series of total internal reflections. but is difficult to reduce below 100 cm–1. In reality. such as the active part of the photodiode being localized on only part of the waveguide.6. since the dimensions of the waveguide are of the order of the wavelength of light in the material.21. a more rigorous analysis of the electromagnetic field shows that only certain configurations (or modes) are supported. this can be adjusted by the design of the waveguide. of a waveguide photodiode with orthogonal geometry and with an improved i ft tradeoff In Figure 2. and cross the i layer several times. Waveguide photodiodes Another approach consists of using an orthogonal geometry. The radiation is progressively absorbed with an effective coefficient of absorption depending on the overlap between the guided mode and the absorbing layer. width of the bandgap and doping of the layers.21 shows a structure with 5 active layers. which is analogous to that of a diode laser.

However. can in practice be partially undone by the difficulty of coupling the radiation into the mode(s) supported by the waveguide. Traveling-wave photodiodes Waveguide photodiodes have the same limitation as conventional structures.300 nm). Such devices have been proposed and analyzed under the term “traveling-wave photodiodes” [GIB 92]. = 1. this can require coupling optics of prohibitive cost. aside from the transit time. 2.PIN Photodiodes for the Visible and Near-Infrared 53 – InP N layer ( G = 920 nm).650 nm).4. This limitation stems from the lumped-element circuit analysis. the circuit bandwidth is limited by the mismatch between the propagation speeds of the two guided waves. = 1. G – GaInAsP P layer ( G = 920 nm). In this way it has been possible to demonstrate a significant gain in the merit factor i ft 30 for GaInAs. it is necessary to use a distributed element model and to think of a photodiode as a waveguide coupled to a transmission line (see Figure 2. it should be noted that the gain in internal quantum efficiency obtained. . however.4. Beyond this. It follows that. in terms of the power/frequency tradeoff discussed in section 2.300 nm). optical and microwave – a limit which is not fundamental in nature. Although it is in theory possible to couple the mode from a single-mode fiber into a photodiode of single mode with an efficiency of 100%. The analysis of [GIB 92] predicts values for the product i ft of greater than 100 GHz. G – GaInAsP N layer ( – GaInAs i layer ( – InP P layer ( G = 1. which is a good approximation as long as the length of the guide remains smaller than the wavelength of the electrical signal (f = 100 GHz.22). = 3 mm). but which is difficult to overcome in practice. terminated at both ends in its characteristic impedance in order to maximize the bandwidth.3.6.

region and the photodetection.6. possibly with the additional use of waveguide photodiode or resonant cavity concepts. three active regions. until now. with the i region fulfilling both the role of radiation absorption and also the role of the depleted region.22. can be doped without reducing the internal quantum efficiency and the speed. An initial example is that of electroluminescent diodes whose optimization leads to a P+p+N–N+ double heterojunction structure. here p+.54 Optoelectronic Sensors fabrication difficulties appear.4. It is possible to imagine photodiodes where these functions are separated so that they can be optimized separately. Beyond PIN structures Whatever their geometry.2). on the condition that it is thin enough to ensure both efficient and rapid transport of photocarriers by diffusion (see sections 2.2 and 2. to have seriously limited the practical performances relative to the potential performance: the challenge still remains to conceive and then construct structures allowing simultaneous improvement in i ft and Pf 2 (the figure of merit). P +p+N–N+ structures The absorbing region. the photodiodes considered up to this point have had. Under reverse-bias the depleted region grows into the N. quantum efficiency and speed characteristics are sufficient for some applications .5. Traveling-wave photodiode 2. where the emitting layer is designated p+.6.1.5.3. Figure 2. by definition. 2.1.2.

p.S. 36..E. Very fast response speeds have e more recently been discovered to be at the expense of low quantum efficiency 0. Sov. LT-5.. [GIB 92] GIBONEY K. IEEE J. p.W.R. in order to equalize the transit time of electrons generated in the i layer towards the N+ layer and of holes towards the P+ layer. KAWANO K. [ITO 00] ITO H. 1363-1365.E.... P+IiIN+ structures The placement of the absorbing i region in a thicker I depleted region. KIM O. 10. 2. “Ultrawide-band long-wavelength p-i-n photodetectors”.. Lett. f RC 300 MHz [POU 80]. ANDREEV V. 1339-1350. HATA S. . allows us in principle to gain a factor of 2 on the cutoff frequency ft [UNL 95]. Appl. [KAN 85] KANEDA T.. 4. [BOW 87] BOWERS J.6. no. “Semiconductors and semimetals”. 2728-2735. vol. 1. 95-98. ZIMOGOROVA N. vol.S. chap. Bibliography [ALF 70] ALFEROV Z. 247-328. 3. p. part D. TRET’YAKOV D.. f3dB 300 GHz in a conventional geometry. 12. 1965.. IEEE Int.. 2.J.PIN Photodiodes for the Visible and Near-Infrared 55 0. [KAT 92] KATO K. “Travelling-wave photodetectors”. 11. 1373-1376. “Optical response time of In0... “Photoelectric properties of AlxGa1-xAs heterojunctions”. no. Phys. KODAMA S. vol. 27-34.. [JOH 64] JOHNSON E. BOWERS J. Phys..6. Electron.. vol.. FURUTA T. RODWELL M. no. p. October 1987. 1809-1810... 21. “A high-efficiency 50 GHz InGaAs multimode waveguide photodetector”. 12 October 2000. [FOR 82] FORREST S. 22. 3. p.477As/InP avalanche photodiodes”.I.. pt. Rec. December 1992. “InP/InGaAs uni-travellingcarrier photodiode with 310 GHz bandwidth”. vol. Conf.. December 1992. July 1982..5. vol. KOZEN A. Lightwave Technology.G.. 28. J.M. May 1970.2. p. no.N. 1985.06. Quantum Electronics. 41. p. e making more use of the high mobility regime of electron transport [ITO 00]. ISHIBASHI T.53Ga0. Lightwave Communication Technology.. no.5. IEEE Photonics Technology Lett.-Semicond. no..O.7. p.K.. vol. YOSHIDA J. Academic Press. SMITH R.. 12. BURRUS C. “Physical limitations on frequency and power parameters of transistors”.A. Lett.

. no. 27. 13551371.P. no. [MAT 87] MATHIEU H.. ED-29. OLSEN G. p. [POU 80] POULAIN P. Phys.E. SAY P.53 for response to longer wavelength (> 1. no.. IEEE J... 8. John Wiley & Sons.56 Optoelectronic Sensors [KIS 91] KISHINO K. DAPKUS P... 1050-1055. Japanese J. HANSOM A. Lightwave Technology. S. Chapter 2. “110 GHz highefficiency photodiodes fabricated from indium tin oxide/GaAs”. April 1980.. vol. vol. UNLU M. no. KOSONOCKI W. Physique des semi-conducteurs et des composants électroniques. vol...S... . vol. March 1995. TABATABAIE N.6 μm) detectors for fiber-optical communications”. no. August 1992.M. G.R. p. “Dark current analysis and characterization of InxGa1-x As/InAsyP1-y graded photodiodes with x > 0. Appl.H. 13. “Long-wavelength (1.M. 527-528. Quantum Electronics.D.. 1981... Electron.... J.F. KAWASHIMA M. [SZE 81] SZE S. Academic Press.7 μm)”. BULMAN G. TOMASETTA L..R. Physics of Semiconductor Devices. 10.. April 1980. Masson. 173-247.. NAKANO K. vol... REED J. G. SIBBETT W. [LIN 92] LINGA K. Phys. Part D. [TAK 80] TAKANASHI Y.... CHYI J-I. MORKOC H. ONAT B. p.M. 2nd edition. no. vol. p.. 2025-2034.. LEBLECI Y. Phys. 180-182.. p. September 1982.. Japanese J. August 1991. 1985. ARSENAULT L.3 to 1. 22. CHIN R. 3. [STI 82] STILLMAN G.. Lightwave Communication Technology.. 1987. Lightwave Technology.. “State-of-the-art performance of GaAlAs/GaAs avalanche photodiodes”. 35. [UNL 95] UNLÛ S. July 1979. on Electron Devices. 19. 4. vol. JOSHI A. 9.. Lett. Lett. COOK L.. no.W. Appl. vol. p. E. 8. POLLACK M. “Semiconductors and semimetal”. “Diffusion limited transient response of heterojunction photodiodes”.D. BAN V.. Appl. p.. [PAR 87] PARKER D. “Resonant cavity enhanced (RCE) photodetectors”. [PEA 85] PEARSALL T. 19. DE CREMOUX B. “Transient simulation of heterojunction photodiodes – Part II: analysis of resonant cavity enhanced photodetectors”. IEEE Trans. p... HORIKOSHI Y. 693-701.A. 406-415. 1987. 4. L189L192. p. [LAW 79] LAW H.M.. vol. 2. “Required donor concentration of epitaxial layers for efficient InGaAsP avalanche photodiodes”. J. 23.

known by the term shock ionization or impact ionization.). . In this chapter we will discuss the following subjects: – theoretical consideration of the avalanche effect (ionization coefficients. and can themselves create other electron-hole pairs: this leads step-by-step to an amplification of the primary photocurrent. – technical conditions necessary for proper operation of APDs. is also at the root of the reversible breakdown of reversebiased p-n junctions. thus acquiring enough energy to ionize the atoms of the crystal lattice. multiplication factor. etc. This mechanism. avalanche photodiodes (APDs) are very attractive devices because they benefit from an internal gain due to the multiplication of the charge carriers generated by absorption of incident light.Chapter 3 Avalanche Photodiodes 3. thereby creating new electron-hole pairs which are immediately separated. Introduction Photodetectors are one of two contrasting families of devices that dominate the field of optoelectronics. Chapter written by Gérard RIPOCHE and Joseph HARARI.1. This occurs when these charge carriers cross a region of high electric field (>105 V/cm). Among these.

although benefiting from a high sensitivity thanks to a noise-free internal gain process. led to the birth of optoelectronics and to the subsequent boom that it has seen. However. InGaAs.. and the creation at the same time of the first electroluminescent diodes. History The discovery of solid-state. suffer from two serious drawbacks for many applications: their size and high operating voltage. etc. and short response times). These make them very attractive to users compared to photomultipliers which. the amplifier stage limits the performance of the receiver in two ways: by reducing the bandwidth of the system and by lowering the global sensitivity due to the additional noise that it introduces. the electrical signal generated by the absorption of incident light by a photodiode is. – state of the art for APDs adapted to different ranges of wavelength. however to take advantage of the considerable potential of these new light sources. – new low-noise APD structures based on superlattices or heterogenous structures (Si/InGaAs). ease of use and reliability.2. reverse-biased p-n junction photodiodes offer numerous advantages: low bias voltages. Among the possible candidates. based on Si. – photoelectric properties of APDs. particularly in the field of optical communications.58 Optoelectronic Sensors – APD structure. small footprint. mostly semiconductor lasers and light emitting diodes (with direct modulation at high modulation frequencies. 3. there is a need for fast and sensitive detectors operating at the emitted wavelengths. In boosting the signal to a sufficient level. Ge. in most cases. gas and semiconductor lasers between 1958 and 1962. . too weak to be directly usable and must be amplified. robustness.

such as superlattice InGaAlAs/InAlAs APDs or heterogenous Si/InGaAs APDs. both in the epitaxial growth of heterostructures and in their design and technology.3 μm.3 μm (~2 dB/km) and at 1. The development of working InGaAs/InP APDs required important progress throughout the 1980s. the first investigations were carried out to evaluate the possibilities offered by detectors based on III-V compounds.2 dB/km). The 1970s saw the optimization of Si APDs for fiber-optic telecommunications at 0.06 μm telemetry and for optical communications around 0. and also the development of germanium avalanche photodiodes (Ge APDs) whose photosensitivity extends to 1. From then on.85 μm. investigations focused on low-noise devices for high bitrates.65 μm. As the performance of these was limited. in 1964 Johnson [JOH 64] highlighted the advantages of primary photocurrent multiplication through the “avalanche effect”. even at 1.55 μm (~0. Since the future of fiber-optic telecommunications is very much based in the nearinfrared due to the weak attenuation of optical fibers around 1. as well as the exploratory study of gallium arsenide (GaAs) APDs and those based on GaAs compounds. which was reported by McKay and McAfee [KAY 53] in 1953 during their study of reversible breakdown in Si p-n junctions. despite the additional noise due to the random nature of the photomultiplication mechanism. This internal gain mechanism gives enough improvement to the signal-to-noise ratio to make the use of avalanche photodiodes (APDs) very attractive for many applications.85 μm through the use of “lownoise” designs. . in particular for optical telecommunications. Studies carried out in the 1960s focused on the development of silicon avalanche photodiodes (Si APDs) for 1. several laboratories worked on improving Ge APDs. which lead to the creation of devices tailored to the needs of fiber-optic submarine links and in particular to those links that can operate at bitrates of over 10 Gbit/s. starting with indium phosphate (InP).Avalanche Photodiodes 59 In order to eliminate these problems and to address the need for amplification of the electric signal generated by the light incident on the photodiode.

which enters the space charge region (SCR) where there is a fairly strong electric field E (>105 V/cm). Figure 3. similar to that observed in gas discharges. Diagram showing the principles of the shock ionization mechanism . as long as the kinetic energy accumulated by each of the charge carriers between successive collisions is greater than the ionization energy Ei.1. electron or hole. observed when the bias voltage is increased towards the breakdown voltage VB. The secondary charge carriers (electron and hole) that are generated.1. A primary charge carrier.60 Optoelectronic Sensors 3. The avalanche effect is shown schematically in Figure 3. is strongly accelerated and can acquire a mean kinetic energy greater than the characteristic ionization energy Ei of the semiconductor. The avalanche effect The study of reversible p-n junction breakdown by McKay and McAfee in silicon. showed that the growth of the dark current. and by Miller [MIL 55] in germanium. is caused by a mechanism commonly known as the “avalanche effect”. can then independently create new electronhole pairs. creating an electron-hole pair through ionization and promoting an electron from the valence band to the conduction band. This process is cumulative and can under certain conditions lead to the breakdown of the junction.3. before colliding with an atom of the crystal lattice. which is a characteristic of a given device. as well as the primary charge carrier.

for silicon and germanium on the left and for InP on the right. If it can be reduced by some means.1] where ai and bi are parameters that depend on the material and on the type of charge carrier. Silicon is unusual in two ways: the electrons are the most easily ionized carriers and the ratio of the ionization coefficients k = i / i is very high for relatively weak electric fields (E ~ 3 x 105 Vcm–1). i = f (1/E) curves. and they quantify the ionizing power of the charge carriers. i ai exp bi E [3. follow empirical relationships of the form: i. The nature of the carriers triggering the multiplication and the ratio k play an important role in the design of devices. which are the base materials for APDs developed for optical communications. .2a and 3. as we will see later. where Eg is the energy corresponding to the width of the bandgap.2b show typical i . Their variations. greater than or equal to 3/2.1. that of a model with two parabolic energy levels.Avalanche Photodiodes 61 3. Ionization coefficients The ionization energy Ei is. this can lead to the preferential ionization of one type of charge carrier. and directly influence the photoelectric performance of APDs. The ionization energy Ei depends on the type of charge carriers for a given material. It is worth noting that the variation seen in the experimental curves published in the literature is a good illustration of the difficulties inherent in determining the coefficients i and i from measurements of photomultiplication or noise. Figures 3. they grow very rapidly with the field. in the simplest case. Depending primarily on the electric field. determined experimentally.3.Eg. The ionization coefficients of electrons i and of holes i are defined as the average number of ionizing collisions produced per unit length (cm) by a charge carrier moving parallel to the constant electric field E.

which itself is only a function of the x coordinate for a given junction. For each type of charge carrier. – mobile charge carriers do not disturb the electric field distribution.2.3. Multiplication factors The impact ionization mechanism leads to a multiplication in the number of charge carriers injected into the SCR. germanium and InP as a function of the inverse of the electric field 1/E 3.62 Optoelectronic Sensors Figure 3.2. Ionization coefficients in silicon. with this taking place in the avalanche region where there is a high electric field E. – recombination of carriers is negligible in the depleted region. the multiplication factor Mn or Mp is defined as the ratio of the number of charge carriers extracted to the number of charge carriers injected. . The following approximations are normally used in the calculation: – the ionization coefficients i and i only depend on the electric field E.

5] 0 with Mn Jn0.Avalanche Photodiodes 63 – the levels of thermal and optical carrier generation in the depleted region are not included.6] 0 . the multiplication factor Mp is given by a similar expression: 1 1 Mp w i w x exp x i x i x dx ' . leads to the expression: 1 1 Mn w i x x exp 0 i x i x dx ' .dx [3.2] [3. The system of equations describing steady-state behavior is as follows: dJ n x dx J Jn x dJ p x dx Jp x i x Jn x i x Jp x [3. Calculation of the multiplication factors Mn and Mp is carried out by solving the continuity equations for the electron current Jn and the hole current Jp under a strong electric field.4] Solving this differential equation for the case of multiplication triggered by a pure injection of electrons from the P+ region.3] C ste which can be rewritten as: dJ n x dx i x i x Jn x i x J [3. where Jn0 is the current of injected electrons at x = 0. In the case of an injection of holes from the N+ region. by integrating Jn(x) over the avalanche region (between x = 0 and x = w).dx [3.

the above equation becomes: 1 1 M i w [3. the two expressions take the form: 1 1 M w i x dx [3. for a given bias voltage V.3. hence: E V w [3.8] where i (E) is a function of the bias voltage V as w is constant. leading to a multiplication factor somewhere between the two which takes into account the “weighting” of each contribution.7] 0 In the case of a PIN junction with a constant electric field in the depleted region. carriers of both types are normally injected into the avalanche region.3. In reality. In the case where the ionization coefficients are equal ( i = i). which occurs for a bias voltage VB – the breakdown voltage. These which conditions occur when the multiplications factors Mn .9] 3. Jpw being the current of injected holes at x = w.64 Optoelectronic Sensors with Mn Jpw . the electric field increases and approaches the critical value for which the dark current grows extremely fast: this is the volume breakdown of the junction. It also follows that the experimental determination of the ionization coefficients i and i from measurements of multiplication factors is very sensitive. Breakdown voltage As the bias voltage increases. it requires the separation of the effects of both types of charge carrier. Mn occurs for: . since in addition to a precise knowledge of the electric field distribution E(x).

1020 2 5 [3. where Eg is the width of the bandgap of the material in question.10] and: w w i 1 0 x exp x i x x dx ' dx [3.14] . we point out the empirical relationship established by Miller [MIL 55] between the multiplication factor M and the breakdown voltage VB: 1 1 M V VB n [3.11] respectively. established a simple semi-empirical relationship between the breakdown voltage VB and the concentration N in the weakly-doped region of an abrupt junction.1 6 5 a 3.12] for an abrupt junction and: VA 60 EG 1. or the concentration gradient a in a gradual junction. In addition.13] for a gradual junction.1 3 2 N 1016 3 4 [3. Finally.Avalanche Photodiodes 65 w x i 1 0 x exp 0 i x x dx ' dx [3. These relations can be applied to different materials. expressed in eV. studying the behavior of P+N and N+P junctions in the avalanche regime. The experimental results are well-matched by the following expressions: VB 60 EG 1. Sze and Gibbons [SZE 66].

This expression satisfactorily describes the variation of the multiplication factor M with the bias voltage V. the photocurrent Iph generated by the light incident on the active region barely varies with the operating voltage.66 Optoelectronic Sensors where n is an exponent which depends on the material. for APDs. I and Id are respectively the photocurrent. we will refer to Iph0 and Id0 as the primary photocurrent and the primary dark current (i. 3. and the dark current. These are denoted as Iph(V0) and Id(V0) in equation [3. and the frequency response. these being at operating voltages in the pre-avalanche regime of the junction. reaching its maximum value when the bias voltage is equal to the breakdown voltage VB of the junction. The gain M is given by the equation: M I ph V I ph V0 I V I V0 Id V I d V0 [3. In the following. the noise linked to the photocurrent.4. In contrast.15]. the non-multiplied currents).e. Properties of avalanche photodiodes The behavior of avalanche photodiodes differs from that of standard PIN photodiodes only in terms of the properties that are involved in the avalanche multiplication mechanism: the gain. This equation allows us to plot the M(V) characteristics of an APD measured under well- . above a certain bias voltage the photocurrent Iph grows steadily.1.15] where Iph. Current-voltage characteristics and photomultiplication For a PIN junction. the total current flowing across the junction.4. 3. the type of junction and the nature of the charge carriers triggering the multiplication. V and V0 are the operating voltage and the reference voltage for which M = 1. and it is often applied to classical P+N and N+P junctions. in contrast to the low voltages at which PIN photodiodes are normally used.

14]. which applies for weak values of the current I. The maximum gain obtained (Mmax) is very sensitive to the conditions of the measurement.Avalanche Photodiodes 67 defined experimental conditions.17] This expression shows the value of devices with a weak dark current (Id 0 << Iph0) in order to promote photomultiplication. for which M depends little on the incident light power. and hence the power of the incident light P0: M max 1 I ph0 or M max 1 P0 [3. These may have a decreasing or increasingly strong influence on the measured values. Taking these considerations into account. close to breakdown. M = Mmax and this can be simplified to: M max VB n Id 0 I ph0 R [3.18] . A reduction in Mmax can be partially explained by a reduction in the avalanche electric field caused by an opposing electric field created by the mobile charge carriers. this contrasts with significant saturation effects that are seen at high gains. the maximum gain is inversely proportional to the square root of the primary photocurrent Iph0.16] for V = VB. which are present in significant numbers in the depleted region. although an APD may behave fairly linearly for relatively low values of gain (M < 10). Also. can for high currents be written: 1 1 M V Id 0 I ph0 VB M R n [3. equation [3. the applied bias voltage is reduced because of the voltage drop across the equivalent resistance R of the diode. In particular. In these conditions.

the currents across the junction undergo a multiplication M. to the range 50–150 V.20] 3. the limitation on the gain is imposed by the dark current and in these conditions.68 Optoelectronic Sensors Taking into account equation [3.4. However. which expresses the breakdown voltage VB as a function of the concentration N. corresponding to relatively high breakdown voltages VB. The random nature of the impact ionization process results in a fluctuation in the gains associated with the different carriers.2. for devices with a strong dark current (Id 0 >> Iph0). and whose spectral density is: i2 2 q Id 0 I ph0 [3. Noise in avalanche photodiodes Noise in detectors is discussed in Chapter 7 and we will limit ourselves here to revisiting the main results for noise associated with the operation of APDs in the avalanche regime. contributing to the appearance of . would be of benefit in obtaining high gains M.21] When the APD operates in the avalanche system. we have: M max VB n Id 0 R [3. Finally. In the absence of multiplication. we find: 3 M max N8 [3. the APD behaves like a traditional photodiode whose noise is essentially shot noise caused by the dark current Id 0 and the primary photocurrent Iph0 across the junction.12]. where possible. which can occur with Ge APDs. the diode breakdown voltage is limited.19] Weak concentrations N in the depleted region. and hence the spectral density of the noise is multiplied by M2. for convenience of use.

22] where F(M). have been derived by McIntyre [INT 66] under the assumption of an avalanche region with a constant electric field. F(M) expressions for the two types of primary injection. which is considered constant. Instances of a pure injection of electrons or of holes are of particular interest.p >> 1.24] Where Mn.25] 1 k Fp M 1 M k p 21 [3.23] and for a primary injection of holes: Fp M Mp 1 1 1 k Mp Mp 1 2 [3. The value of F(M) depends on the characteristics of the p-n junction and on the nature of the primary carriers. These are expressed as a function of M and the ratio of the ionization coefficients k = i/ i. represents the contribution of multiplication to the noise. The spectral density of the shot noise of the multiplied current can be written: i2 2 q Id 0 I ph0 M2 F M [3. these equations can be expressed in the following simplified forms: Fn M kM n 21 k [3.Avalanche Photodiodes 69 extra noise linked to the multiplication.26] and for very high values of multiplication: . electrons and holes. The excess noise factor for a primary injection of electrons can be expressed by the following: Fn M Mn 1 1 k Mn 1 Mn 2 [3. the excess noise factor.

etc. Excess noise factor as a function of different values of k for a pure injection of holes In order to obtain low excess noise values. GaAs. with this parameter playing an important values of the ratio k i i role in the optimization of APDs.3. Furthermore. only silicon satisfies these two requirements. in the literature. Attempts can be made to approach this value through modification of the ionization energy or by careful design of the device. the excess noise factor takes a minimum value of F(M)=2. Ge. electrons or holes. independent of M.3. InP.). consideration of the excess noise factor allows us to approach high . Finally. Figure 3. and a material must be used whose ionization coefficients are very different. The typical variation of the noise factor as a function of M for different values of k is shown in Figure 3. the excess noise factor F(M) is often written in the form: F M Mx [3.28] . Among the normal semiconductors (Si. two conditions must be met: the multiplication must be triggered by the more ionizing carrier.70 Optoelectronic Sensors Fn M kM n and Fp M 1 M k p [3.27] In the specific case of ionization by only one type of carrier.

the limiting noise of traditional photodiodes comes from the thermal noise associated with the load resistance RL or. For k = 1. 3. includes two sources of noise outside the diode.Avalanche Photodiodes 71 which comes directly from the spectral density of the noise. the multiplication is triggered by electrons. For low values of k. consisting of the APD and amplifier. For some application types. independent of the multiplication (the load resistance and the amplifying stage). more commonly. allows us to compare the different devices.29] The exponent x. the injection of holes is favored. from the noise of the first stage of the amplification circuit required to make use of the electrical signal produced by the photodiode. the internal gain of APDs allows us to improve the signal-to-noise ratio (SNR) of the detector. expressed as: i2 2 q I ph0 M 2 x [3. Signal-to-noise ratio in avalanche photodiodes In wideband applications. which depends on the value of the k ratio. This can often be found in device specifications alongside the value of F(M) for a given M. The mean squared value of the photocurrent is given by: .4. given by the equation: 2 iext 4k BTe RL f [3. This detector. the excess noise factor is F(M) = M for the two types of injection. whereas for high values of k.30] where Te is the equivalent temperature of the noise.3. and the exponent x therefore takes the value x = 1. and also includes sources of shot noise connected with the multiplication process: 2 iM 2q I d 0 I ph0 M2 F M f [3.31] where f is the bandwidth of the system.

72 Optoelectronic Sensors i2 ph 1 2 2 m I M2 ph0 2 [3. Figure 3.4. since for low values of F(M) the thermal noise contribution is divided by M2. The SNR for a system with bandwidth f therefore follows the relationship: S B 1 2 2 m I M2 ph0 2 2q I d 0 I ph0 M 2 F M 4k BTe M 2 RL [3. which depends closely on the point at .34] demonstrating the advantage of APDs over PIN photodiodes. Variation of the SNR as a function of the gain M for two semiconductors. the SNR grows and reaches a maximum for an optimal gain value Mopt.33] f which can be rearranged into the following form: S B 1 2 2 m I ph0 2 2q I d 0 I ph0 F M 4k BTe M 2 RL f [3. silicon and germanium As the gain increases.32] where m is the modulation index of the incident light signal.

the time needed to reach a stable multiplication value for a given bias voltage. or light signals modulated at high frequency. in the case where Id 0 << Iph0: M opt 2k BTe qI ph0 RL 1 2 x [3. such that. The variation of the SNR with the gain is shown in Figure 3.4.Avalanche Photodiodes 73 which the multiplied shot noise and the thermal noise are equal. Thus. The response time of an APD operating in the avalanche regime is limited by: – the transit time ttr of carriers in the SCR. Speed. – the avalanche build-up time tav. APDs allow us to improve the sensitivity of photodetectors as long as the shot noise remains lower than the thermal noise entering the amplifier stage. The optimal gain value Mopt depends on the incident radiation power P0 through the equation: I ph0 q h c P0 [3. is a topic of research for numerous applications. response time and frequency response of avalanche photodiodes The ability of avalanche photodiodes to detect and amplify optical impulses with short rise and fall times.4. – the RC time constant of the APD/load system. – the diffusion time tDif.36] As a result.4. .35] For higher multiplication values. Mopt will increase as P0 decreases. the time taken by the electron-hole pairs created outside the depleted region to reach the p-n junction. with the contribution of shot noise becoming the dominant factor. the SNR drops. 3.

The diffusion time tDif of electron-hole pairs towards the junction is dominated by the lifetime of the minority charge carriers in the lightly-doped region of the junction. In order to increase the speed of APDs for high-frequency applications or broadband applications. the thickness w of the depleted region is adjusted so that all the light is absorbed there. In practice. because the light will be absorbed in a thinner width of material. Its value (between a few μs and a few hundred μs depending on the material) makes some devices unusable for many applications. This result is approximately achieved for: w 2 [3.37] where vsat is the saturation velocity for the slowest charge carriers. The transit time ttr can be expressed by the equation: ttr w sat [3. This results in a compromise between the quantum efficiency and the speed which must be treated on a case by case basis. the diffusion time tDif and the RC time constant all depend on the width w of the bandgap. we will seek to reduce the thickness w. as far as possible. The RC time constant is only significant at high frequencies. In this case R~50 .4). in a similar way to the tradeoff in PIN photodiodes (see section 2. The capacitance of the junction Cj is related to the area A of the junction and the width w of the depleted region by: Cj A w [3.74 Optoelectronic Sensors The transit time ttr . This limitation appears for certain Si and Ge APDs.38] where is the absorption coefficient for light of a given wavelength in a semiconductor.39] . This reduction will in turn produce an increase in the capacitance Cj of the junction and a reduction in the response coefficient. the value of the load resistance.

Avalanche Photodiodes 75 where is the permittivity of the semiconductor. The avalanche build-up time tav is therefore expressed by the equation: ta M [3. This is because the high number of carrier-lattice interactions means the carriers “bounce” back and forth. independent of the multiplication M. the impulse response time is controlled by the transit time across the depleted region of the non-ionizing secondary carriers created by the primary carriers. Emmons [EMM 67] analyzed the frequency response of PIN photodiodes operating in the avalanche system and established the following expression: Nb i i 1 [3. This optimization is equivalent to that for PIN photodiodes (see section 2. the bandwidth of the APD. prolonging the time that secondary carriers are present in the avalanche region. This depends very strongly on the value of k = i / i and can become very small if i and i are very different.40] where is the effective transit time of a carrier in the avalanche region between two ionizing collisions. As the multiplication process is not instantaneous. we need to minimize the area of the active region and that of the guard ring. As a result. is not very different from that of a PIN photodiode. the response time of APDs can be greater than that of PIN photodiodes of comparable structure. When the two types of carriers contribute to the multiplication. the steady-state value of M will be reached after a time which will be longer if M is higher and if the ionization coefficients of the carriers i and i are close to each other. and select a value of w compatible with an acceptable transit time ttr .41] where Nb is a number which varies slowly between 1/3 and 2 as the ratio k = i / i varies from 10–3 to 1. and is very sensitive to the ionization conditions.4). in the case of ionization by only one type of carrier. Under these conditions. and 1 is the actual transit time in . which is similar to that of silicon. Hence.

for M > i / i . These inhomogenities can be caused by the structure itself.76 Optoelectronic Sensors the avalanche region.42] where M( ) and M0 are the values of the gain as a function of the frequency f = /(2 ) and for a constant signal. which must be as homogenous as possible throughout the active region. Emmons [EMM 67] showed that. 3. the gainbandwidth product is constant.5. Regardless of their origin. Any inhomogenity present in the SCR that can cause a local increase in the electric field will lead to a premature breakdown which will degrade the APD performance.43] and depends on the material through k and the characteristics of the avalanche region through 1 = la/vsat. . A high value of the gain-bandwidth product is desirable as it allows a higher value of gain for a given frequency. where la is the length of the avalanche region. A low value of k minimizes the effective transit time and is also favorable for the response time and for the excess noise factor. as in a traditional planar junction obtained by localized diffusion in a circular geometry. they always need to be eliminated. As a result there is a very intense electric field (~4-5 x 105 V/cm) present in the avalanche region. or linked to imperfections in the crystal structure. It follows that the product will be much larger if the ratio k of the coefficients of ionization is weak and if the avalanche region is narrow. The gain-bandwidth product is given by: M f Nb k 1 1 [3. Technological considerations APDs work best when they are biased with a voltage close to the volume breakdown voltage in the active region. The gain M varies with frequency according to the following law: M 1 M0 2 M0 2 1 2 2 [3.

In order to suppress peripheral breakdown. the active region is obtained by a shallower . as can be seen by solving Poisson’s equation [HAR 91a]. two kinds of structures are commonly used. Cross-section of a planar photodiode 3. particularly in the fields of crystallogenesis and techniques of epitaxial growth and doping.5. In the first structure (see Figure 3. Breakdown due to edge effects therefore occurs for a voltage sufficiently below the breakdown voltage of the active region so that light falling on the central region is not amplified or barely amplified. The suppression of crystal defects depends closely on the progress of technology. as long as a geometry is used which is free from angular parts liable to cause point effects.1.6a). the increase in the injected global photocurrent due to the high electric field present in a very narrow annular region around the edge can give the illusion of a photomultiplication effect.5. which have annular guard rings modifying the electric field distribution around the outer perimeter. and quite deep in order to increase the radius of curvature and thus reduce the peripheral field. such a photodiode cannot operate as an APD.5). the use of junctions with a guard ring or the use of a mesa structure allows the suppression of peripheral breakdown. if the illuminated region extends beyond the diffused area. On the other hand. Regardless.Avalanche Photodiodes 77 In the first case. the guard ring is formed by an annular diffused junction which is relatively weakly doped. the electrical field inside the depleted region is higher in the curved region around the edge of the diffused area than in the active region. Guard ring junctions For a junction constructed using planar technology (Figure 3. Figure 3.

Figure 3. The breakdown voltage of the gradual junction of the guard ring is much larger than that of the active region.78 Optoelectronic Sensors diffusion at a higher concentration. This structure.2. the extension of the avalanche region at high electric fields outside the edges of the etched region can lead to migration of susceptible ion species. The n+– guard ring junction which surrounds the n+p junction of the active region is much deeper than the latter. the footprint of the active region of the APD is defined by chemical etching or by dry etching followed by surface reconstruction using a suitable chemical bath.6. the guard ring and active region junctions are created by diffusion or ion implantation within a weakly-doped substrate .6b). the extent of the SCR is larger in the former than in the latter. different passivation techniques have . In order to avoid these drawbacks.5. because for a given bias voltage. the addition of a guard ring of the first type is necessary to avoid peripheral breakdown. This is the P PN+ structure. Cross-section of (a) an APD with a traditional guard ring and (b) a P PN+ APD In the second structure (see Figure 3. causing a growth in the dark current which is often accompanied by instability of the diode. This is referred to as a P+ P N+ structure. However. or even superficial. while eliminating the guard ring junction. offers the sometimes deceptive advantage of simplicity. 3. partially overlapping the guard ring.7. “Mesa” structures In mesa structures. It allows us to reduce the capacitance of the device. but the weak concentration in the substrate precludes breakdown. When the n+– junction is not very deep. shown in Figure 3.

metallic precipitation. Observation of the current induced in a scanning electron microscope also allows the detection of inhomogenities in the electric field. Figure 3. In the case of Si APDs. and which appear at well-defined bias voltages which may or may not be close to the volume breakdown voltage. present in a PN junction introduce localized or microplasma breakdowns: isolated regions of a few μm2 where high current densities cross.7.Avalanche Photodiodes 79 been used with moderate success. Unless these microplasmas appear at a voltage very close to the breakdown voltage. Cross-section of a mesa APD 3. etc.. In the absence of microplasmas. Microplasmas can be eliminated using substrates with a low level of . inspection under a microscope allows the microplasmas to be visualized.5.3. stacking faults. At higher voltages. In addition. Crystal defects and microplasmas Crystal defects such as dislocation. the noise linked to electrical discharges triggered by the random injection of charge carriers in the microplasma is so significant that it precludes any interest in the use of such diodes. a diffuse. interspersed with resistive sections in the relationship. the presence of microplasmas leads to vertical jumps of several μA in the inverse characteristics Id(V). pale light uniformly illuminates the whole active surface when the breakdown voltage is reached. their presence decreases APD performance by limiting the photomultiplication factor to that of the breakdown voltage of the first microplasma. which makes such a structure difficult to commercialize on account of its lack of reliability. thus enabling the electrical sorting of APDs through systematic measurement of the Id(V) characteristics via probes. appearing in the form of bright red spots.

1 μm. an indirect-bandgap semiconductor. which is the emission wavelength of the Nd lasers that are very popular in military applications.4 μm. Si N+P APDs Si APDs with an abrupt-junction N+P structure are fabricated on a P substrate with a concentration around 1016 cm–3.6. Also. reliable devices with a high fabrication yield depends closely on the quality of the base material and on the standard of technology used in each stage of the fabrication.63 μm). The absorption length.9 μm. Commercialized in 1966. .06 μm.80 Optoelectronic Sensors dislocations and non-aggressive fabrication techniques tailored to the specific materials used. considerably less than 1 μm at = 0. with electrons having the higher ionization energy ( i >> i). varies between 10 and 10 μm in the wavelength range 0. is sensitive to light in the wavelength range of 0. taking advantage of an atmospheric transmission window allowing targeting at 20–25 km with high precision (of the order of a meter).06 μm. with a bandgap energy Eg = 1. The construction of high-performance.7 μm (ruby lasers emit at 0. Silicon avalanche photodiodes Silicon. they were initially developed for military applications such as laser telemetry at 1. 3. these APDs are well-suited to the detection of light signals with wavelengths below 0. 3.6. and is more than 200 μm at = 1.8–0. These characteristics play an important role in the design of devices targeted at different applications.1.1 eV at ambient temperature. As the width of the depleted region is between 2 and 3 μm. the first transmission window for optical fibers (attenuation ~ 2 dB/km). with the aim of achieving a weak excess noise factor and an acceptable compromise between sensitivity and response time.4–1. the ionization coefficients of electrons and holes are very different.

with the gradient dM/dV very large. Since the breakdown voltage VB follows variations in temperature. the choice of substrates with fairly high dislocation levels (~3. The diffusion . M ~ 50 to 100.7 μm. with the duration of the impulse generally being insufficient to reach the steady-state value. with a breakdown voltage VB ~ 60 – 80 V. Additionally. Other regulation systems make use of the noise linked with multiplication. At short wavelengths absorbed in the N+ layer. These APDs. It consists of a brief rise to the signal (<0. the guard ring and active region junctions are made by a double diffusion of phosphor impurities. The impulse response of Si N+P APDs (see Figure 3. The homogenity and reproducibility of phosphor ion implantation used as a predeposition technology have greatly helped the achievement of microplasma-free APDs. For > 0. whereas they were at best around 20% with the “zero dislocation” substrates advocated in the literature. the multiplication is initiated by holes. We then see a similar behavior for the fall of the signal.000 cm–2). along with a high fabrication yield.000–1. To avoid issues with variations in M. has allowed the attainment of fabrication yields considerably higher than 50%.8) is dominated by the diffusion time tDiff of the charge carriers generated outside the depleted region. but which are stable under various thermal treatments at high temperatures (T 1. At the operating point. masked using a layer of silica obtained by thermal oxidation.Avalanche Photodiodes 81 For the first APDs we will study.5 ns) followed by a relatively long tail. the multiplication is mostly initiated by electrons and takes higher values. the ratio V/VB must be kept constant. with the exponent n in the Miller equation taking a value close to 4. offer M(V) characteristics which vary extremely rapidly as breakdown is approached. The involvement of a Zener diode with a breakdown voltage close to that of the APD has allowed the multiplication to be regulated within the temperature range from –30°C to +80°C. the fabrication conditions for the different stages of the construction procedure have been developed to ensure high performance in terms of the dark current and photomultiplication. some system of regulation is required.200°C).

Figure 3.85 μm or 1. whose thickness is adjusted as a function of the operating wavelength (0. where the multiplication of injected charge carriers takes place. Typical impulse response of a Si N+P photodiode 3. Si N+P P+ APDs To compensate for defects in the N+P structure. and this can be improved by limiting the thickness of the absorbing region with the use of epitaxial P/P+ structures or by localized thinning of the substrate.06 μm) and the desired performance.2. This separation of the roles of multiplication and absorption gives the structure a high gainbandwidth product.82 Optoelectronic Sensors time limits the repetition frequency. and another relatively extensive region. The result of this is that the depleted region consists of two adjacent regions: one relatively narrow. where light is absorbed and the photocarriers collected. in 1966 Ruegg [RUE 66] showed the potential of an N+P P+ structure. . Absorption takes place in the very weakly-doped region (<1014 cm–3). with a weak electric field.8.6. The doping and thickness of the P layer are absolutely critical. which combines the advantages of an N+P structure (internal gain) with those of a PIN structure (high quantum efficiency and speed). with an intense electric field.

Under normal operating conditions. just before the breakdown of the N+P junction.9. As the voltage increases. the depleted region is restricted to the P layer. At weak bias voltages. Then. especially if the layer is wide. c) typical M(V) characteristics. the increase in the bias voltage will only cause a weak increase E in the electric field. and the M(V) characteristic involves a more or less pronounced “plateau”.Avalanche Photodiodes 83 Figure 3. The multiplication will vary relatively slowly. Vrt is the reach-through voltage The operation of N+P P+ APD structures is shown in Figure 3. b) typical electric field. the field in the absorption region is sufficient to ensure carrier transport at their . the electric field distribution for three significant bias voltages and a typical multiplication curve.9. Operating principles of a Si P+ PN+ APD: a) doping profile. the edge of the depleted region moves and enters the layer for a bias voltage Vrt known as the reach-through voltage. for which the multiplication is around 10 to 20. which displays the doping profile. with the depleted region extending across the whole layer.

If the . these APDs are best fabricated on an epitaxial /P+ plate. in order to benefit from a high gain. as a function of the chosen level of tradeoff between speed and response. The typical behavior of N+P P+ APDs is only reached over a small range (± 15%) on either side of the optimal implantation dose. Si N+ P P+ APDs An N+ P P+ structure with buried P layer is a refined version of the N+P P+ structure.75 and 3. For applications at 0.1 μm respectively for implantation energies of 1.3. For a given Si /P+ structure. The technique of implanting high-energy B+ ions (1 to 2 MeV).5 MeV and 1.9c). the thickness of the layer can be adjusted between 15 and 50 μm. the extension into the layer occurs too soon and the multiplication will be weak (Figure 3. the multiplication electric field decreases as the implantation energy is increases. Proposed and studied by Lecrosnier [LEC 75].84 Optoelectronic Sensors saturation velocity. In order to avoid the critical stage of localized thinning. since the P layer lies further from the surface. chosen for constructing this P layer. Specifically. However. is very well-suited because by modifying the implantation energy and dose parameters it is possible to change the breakdown voltage and excess noise behavior of these APDs. the breakdown of the N+P junction occurs before the depleted region extends in to the layer. 3. at distances of 2. if the P layer is too wide.6. Consistent achievement of the desired M(V) multiplication characteristics assumes that we can accurately control (± 10%) the concentration profiles of the N+ and P layers.85 μm. it offers a significant reduction in the excess noise factor F(M) thanks to its narrow multiplication region with a constant electric field limited by the buried P layer: the field is weaker when this is far below the surface. but if it is too narrow. the operating point must be beyond the “plateau”. with illumination of the N+P junction only slightly degrading the excess noise factor.8 MeV.

Avalanche Photodiodes 85 dose is too large. b) typical electric field.8 MeV. dosage 1. Figure 3. the associated electric field distribution.2 μm). c) typical M(V) characteristics for different implantation doses of Boron (in cm–2) .2 1012 ions/cm2) for the buried P layer.5 or 1. the behavior is like that of a N+P diode. The fabrication procedure for APDs involves two diffusion phases in order to construct stopper ring and guard ring junctions.10 shows a schematic doping profile.10. Operation principles for a Si N+ P P+ APD: a) doping profile. a) b) c) Figure 3. then the high energy implantation of Boron ions (1. and the M(V) characteristics for different doses at a given implantation energy. followed by two phases of ionic implantation and a phase of electrical activation annealing: firstly the low-energy implantation of phosphor ions to build the active N+ junction close to the surface (xJ ~ 0. or of a PIN diode if it is too weak.

and also a good fabrication yield. resulting in a response coefficient greater than 75 A/W at = 0.30–0. a number of solutions have been researched. resulting in the P layer lying 2. One of them consists of implanting doubly-ionized B++ Boron ions at relatively high energies (~300 keV). The excess noise factors at M = 100 are F = 4–5 with “channeling”. To compensate for the lack of suitable high-energy implantation equipment. close to the lower limit of x = 0. C(V).86 Optoelectronic Sensors Control of the conditions for the two implantations ensures good reproducibility of the procedure. – a uniformity in the photoresponse of ± 5%. corresponding to x = 0. normally with 250 μm diameter diodes: – weak dark current (I0 < 1 nA) close to breakdown. A model of this structure.5. Diodes with an active zone diameter of 800 μm and even 1. M(V). which is particularly easy to do for epitaxial /P+ wafers.35 and x = 0. – an excess noise factor F(M) = Mx typically with F(M = 100) = 3 – 3. although “channeling” implantation gives promising results. instead of 1 μm as would be the case under normal conditions. or alternatively implanting Boron ions at 800 keV in a preferred direction in a /P+ epitaxial plate with <110> orientation.25. instead of F = 6–7. APDs obtained in this way offer excellent photoelectric performance.500 μm have been achieved thanks to the excellent spatial homogenity offered by the implantation technique. with x = 0. to take advantage of a “channeling” effect.2 is a result of the low ratio k ~ 0.01 of the ionization coefficients i and i.40 respectively [KAN 78]. . This value. these are not easily reproducible. – useful gain Mu > 150. due to the fact that the multiplication electric field is such that Em < 3 x 105 Vcm–1.85 μm and a rise time r ~ 0.5 ns for a layer of thickness 30 μm.2 μm below the surface. noise) for various APDs was developed by Maille [MAI 80]. based on complete photoelectric characterization (I(V). However.

1 μm). The upper contact is formed with a ring of aluminum deposited by evaporation.25 A/W at M = 1 can be obtained for = 0. Then. and hence on the standard of the fabrication technology.Avalanche Photodiodes 87 Another process exists where the buried P layer is fabricated in two stages: first localized low-energy implantation of Boron ions on the /P+ plate. After diffusion of the guard ring. APDs made in this way have an excess noise factor F = 5 at M = 100. The metal present in the passivation layer is eliminated using aqua regia.6. a 10 to 20 nm platinum film is deposited by radio-frequency sputtering. the silica protecting the active central region is chemically removed.5 μm. and hence the electric field is only marginally increased. These results confirm that the excess noise factor F(M) is weaker when the P layer is buried more deeply. with the thickness of the silicate obtained being twice that of the platinum. Their performance depends closely on the quality of the SiPt-Si N interface. The platinum silicate is then formed by in situ annealing at 600°C. followed by an epitaxial stage to obtain the upper layer for multiplication. which could be improved upon by reducing the doping and increasing the thickness of the upper layer. The presence of an anti-reflection coating tuned to = 0. with resistivity ~ 0. 3.4. Under these conditions. SiPt-Si N Schottky APDs SiPt-Si N Schottky APDs. Then the residual native oxides and the first few atomic layers of silicon (~10 nm) are removed by ion bombardment during the chemical deposition phase. as the reflection coefficient of silicon is around 45% in the near ultraviolet. for which the absorption occurs very close to the surface (absorption length of the order of 0. are well-suited to the detection of ultraviolet (UV) radiation around = 0.4 μm. .35 [YAM 76].5–1 cm. transparent and non-absorbing layer (20 to 40 nm) of platinum silicate on top of N-type silicon. The growth of this layer is delicate. a response coefficient above 0. with x = 0.5 μm is needed to obtain a good response coefficient. obtained through the formation of a very thin.

80–0. At short wavelengths. except in terms of speed. Its only drawback stems from the response coefficient/speed tradeoff imposed by the weak absorption coefficient. Stilman et al.02 measured from the I-V characteristics confirms the good quality of the SiPt-Si N interface. In contrast. [STI 74] showed the presence of a very pronounced Franz-Keldysh effect for APDs made of lightly-doped material (n ~ 5 x 1014 cm–3). . Nevertheless. with respect to possible use in integrated circuits and on ternary structures based on GaAs. higher excess noise and a reduced bandwidth. The limit on the bandwidth is imposed by the time constant RC and is several GHz.85 μm.7. A barrier height BN = 0. 3.80–0.435 μm with an excess noise factor Mx such that x ~ 0. and values of multiplication M ~ 300 have been measured at = 0. the light is mostly absorbed outside the depleted region and multiplication is thus triggered by holes leading to performance degradation: a weaker M. simultaneously possessing a weak dark current. a high gain and low noise.88 Optoelectronic Sensors APDs fabricated in this way possess excellent photoelectric characteristics [ALA 74]. GaAs APDs. Avalanche photodiodes based on gallium arsenide Si APDs are almost perfect photodetectors for the 0. On the other hand.7. which is in fact not a great problem because the applications are at low or moderate bitrates (<500 Mbit/s). despite a dark current greater than that of traditional N+P or P+N junctions in their prebreakdown regime. a direct-bandgap semiconductor with a bandgap EG = 1.88 μm. all thanks to the maturity of the technology.85 μm transmission window in fiber-optics. investigations have been carried out into gallium arsenide (GaAs). materials with a high absorption coefficient.8 μm. multiplication is triggered by electrons. when biased in the multiplication regime. either of the Schottky type or diffused with an implanted guard ring (H+ protons or Mg+ ions) offer performances below that of Si APDs within the range 0. at = 0.85 V and a good ideality factor n = 1.40 eV and photosensitive up to 0.

10). using a new concept (see section 3. They offer a multi-quantum well (MQW) or “staircase” multiplication . Spectral response of electroabsorption GaAs Schottky APDs: a) weak bias voltages.11. b) strong bias voltages More recently.Avalanche Photodiodes 89 a) b) Figure 3. GaAs/AlxGa1-xAs APDs with low noise. have opened a new research avenue.

65 μm. ni being the intrinsic concentration for germanium.08 eV. including: – a high dark current. which is proportional to ni2.15 and an excess noise factor F = 3 at M = 10 [KAG 89]. Thus.5 μm. . for the x = 0.48 eV and EV = 0. an indirect-bandgap semiconductor with a small bandgap Eg = 0.3 and 1. Germanium avalanche photodiodes Germanium. Starting in 1966.45 composition of Al. which makes the process of surface passivation very difficult. competition with Si APDs limited their use to very specific areas of application. giving a value of k = I / I ~ 0. The problem was that Ge APDs suffered a number of drawbacks linked to the intrinsic properties of germanium.4 μm to 1.67 eV at ambient temperature and photosensitive in the spectral range 0.90 Optoelectronic Sensors region with a bandgap width varying continuously between GaAs and AlxGa1-xAs. dominated by the charge carrier diffusion current. 3. a region where silicon does not work. and various devices were brought to market.5 μm absorbing layer of GaAs enables high quantum efficiency devices that are fast and low-noise. The presence of an adjacent 1.8. became established in the middle of the 1970s as the photodetector for the first generation of receivers for fiber-optic links at 1. with the holes being the more easily ionized. The need to cross discontinuities favors the ionization of one type of charge carrier. However. in contrast with silicon. – a difficulty growing high quality epitaxial structures. Melchior and Lynch [MEL 66] showed that Ge APDs could be used as fast and sensitive detectors. – the absence of stable native oxides. the discontinuities are respectively EC = 0. – an insufficiently high ratio k of the ionization coefficients of electrons and holes: k 2.

thus allowing a good response coefficient and a high speed.Avalanche Photodiodes 91 On the other hand. For an active region of diameter = 100 μm. Ge APDs with N+P. due to the injection of electrons instead of holes as primary charge carriers. This high value of F(M). the total dark current measured at 0. Sb) is easier to achieve.9 VB is around 150 nA and the dark current Id0 around 50 nA at M = 1. for passivation. due to the high absorption coefficient up to 1. In addition. N+NP and P+N structures for 1. the dark current does not include a tunneling current contribution. corresponding to an effective ratio between the ionization coefficients of keff 1. They have a planar construction with a guard ring to ensure good reliability. are very critical. N+NP and P+N were developed in succession for use in optical links at 1. Si3N4). . The conditions for the surface preparation and the deposition of dielectric layers (SiO2. The excess noise factor for M = 10 is such that F(10) 10 – 11.3 μm. Figure 3.3 μm communication The three structures N+P.1. and also for the antireflection coating.50 μm.12 shows a cross-section view. despite the weak value of Eg. increases still further with the wavelength. The presence of a stopper ring decreases the surface current.3 μm. used to limit the diffusion. depleted region widths w between 2 and 3 μm are sufficient to absorb the light signal. 3. N+P structures were the first to be used.8. since the diffusion of ntype dopants (As.

N+NP and P+N structures The N+NP structure was proposed to improve the noise properties of Ge APDs. leading to a gradual junction. In addition.12. the total dark current at 0. This results in a sensitive increase in the excess noise factors: F(10) = 7 instead of 10 as previously. Since the light is mostly absorbed in the N layer. with multiplication taking place close to the NP junction lying about 2. Cross-section of Ge APDs: N+P. The N layer is obtained by implantation of arsenic ions followed by annealing. . On the other hand. the electric field E at the junction is weaker (E 2 x 105 V/cm).92 Optoelectronic Sensors Figure 3.9VB is around 1 μm and the response time is limited by the diffusion time in the N layer. the holes play the role of primary charge carriers triggering multiplication.5 μm from the surface. The main drawback of the N+NP layer is the lack of reproducibility caused by the fabrication conditions for the N layer.

The absorption length at ambient temperatures is around 25 μm at = 1. where absorption and multiplication take place in separate regions. are made by ionic implantation followed by annealing at relatively low temperature. This “fully implanted” technique offers good reproducibility of the performance.Avalanche Photodiodes 93 The P+N structure represents the final success of Ga APDs for 1.55 μm communication .55 μm.2. because the absorption coefficient of germanium falls dramatically. the excess noise factor F(10) is 8 and 7 respectively. N+NP and P+N photodiode structures become “blind” if the temperature drops below -20°C. Cross-section of a P+NN– structure Ge APD for 1. could attain the required performance of the response coefficient and speed. For diodes with an active region diameter of 100 μm and 30 μm. and the cutoff frequency is at least 1 GHz at M = 10.8.2 dB/km). Ge APDs with P+NN. The active region and guard ring junctions.structures for 1.3 μm applications.13. Structures developed for 1. It possesses the respective advantages of the N+P and N+NP structures in terms of speed and excess noise factor. a window of weak absorption for optical fibers (0. 3.55 μm communication Above L=1. with the light being absorbed in the N-type depleted region and the multiplication being triggered by holes. as well as the stopper ring.3 μm communication are not suitable because of their low depleted region width w 2 to 3 μm.52 μm the N+P. Only a structure analogous to N+P P+ Si APDs. Figure 3.

The guard ring junction is made either by a long period of zinc diffusion or by implantation of beryllium ions followed by annealing.3 and 1.55 μm.9 VB: 1. of diameter Ø = 80 μm or Ø = 30 μm for coupling to multimode or single-mode fibers. whose cutoff frequency is only 10 MHz. The limitation on the response time is imposed by the transit time in the depleted region. – cutoff frequency above 500 MHz at M = 10.55 μm does not contain a “lag” due to the collection of charge carriers created outside the depleted region. This is also the case for Ge P+N photodiodes.13).1 at M = 10. – excess noise factor F = 6. investigations carried out by various laboratories have yielded APDs based on III-V compounds with very promising performances as a second generation of structures for optical communications at 1. followed by activation and diffusion annealing. is best obtained by a double implantation of boron or indium for the surface P+ junction and of arsenic for the N-type multiplication region.55 μm are as follows: – total dark current at 0. Ge APDs often operate with an optimal gain between 5 and 10. – breakdown voltage VB 70–85 V. The active region of these planar-structure devices with guard rings (Figure 3. – quantum efficiency = 80% at M = 1. – dark current Id0 at M = 1: 70 nA.3 μm. The best performances obtained for 80 μm diameter APDs at = 1. Nevertheless. with the response time depending little on the multiplication factor.structured Ge APDs on a weak-concentration N substrate ( 5 1014 cm–3) have been built to meet these requirements.APDs at 1. Additionally. with the available performance gains for Ge APDs limited by the fundamental properties of germanium. The impulse response of Ge P+NN.94 Optoelectronic Sensors P+NN. .

despite a more favorable ratio of ionization coefficients than that of germanium ( I ~ 2. However.35 eV) and transparent above 0.65 μm.53Ga0.5 x 105 V/cm precludes all avalanche multiplication. it is well-suited to epitaxial growth of high crystal quality layers.14. offer a low dark current along with an ionization coefficient for holes that is higher than that of electrons. the appearance of a high tunneling current for an electric field E > 1. . Formed into a lattice using InP.55 μm. Avalanche photodiodes based on indium phosphate (InP) In0. using a SAGM structure 3. offers strong absorption up to 1. such that I ~ 2 I . with a direct bandgap of EG = 0.95 m. a material with a large bandgap (EG = 1.Avalanche Photodiodes 95 Figure 3. I).9. On the other hand.75 eV.47As.47As/InP heterojunction APDs rapidly came to the forefront in the 1980s as second generation detectors for optical communications at 1. Cross-section of an InGaAs/InP planar APD with front illumination. This was because the ternary compound In0.3 and 1. P+N APDs based on InP.53Ga0.

96 Optoelectronic Sensors Figure 3.47As) and multiplication region (InP). with bandgap widths intermediate between those of InGaAs and InP.14 shows a cross-section through a planar InGaAs/InP APD with front-face illumination which has a SAGM structure. This structure. and to avoid the decomposition of the ternary layer during growth of the InP layers above it. Operation principles of an InGaAs/InP APD with a SAGM structure. suited for illumination through front or back faces. The electric field is given: 1) for a weak bias voltage. offers the advantage of a pure injection of holes. led to the SAGM (Separated Absorption Graded Multiplication) structure. and 3) in the case of multiplication The use of a structure analogous to the P+ PN+ structure developed for silicon.15.53Ga0. Figure 3. consisting of a separate absorption region (In0. The “graduated” quaternary InGaAsP layers. to facilitate the transport of holes by reducing the valence band discontinuity at the InGaAs/InP interface. which are the most . 2) for the punch-through voltage. Their presence is vital for the growth of a vertical heterostructure by liquid phase epitaxy. fulfill a double function: to ensure a “gentle” transition between InGaAs and InP.

Proper operation of these APDs. 0.2 μm). the InP-N multiplication layer (n = 1.5 105 V/cm in order to retain a low dark current. an InGaAs absorption layer (n = 5 1015 cm–3. is even more delicate than for Si and Ge APDs. InGaAs/InP APDs for optical communications at 2. since during avalanche the electric field EQT at the quaternary/ternary absorption interface must be below 1.5 Gbit/s An APD heterostructure based on InGaAs/InP on a <100> oriented InP-N+ substrate consists of a stack of 6 layers: an InP-N+ buffer layer (n = 5 1017 cm–3. The guard ring junction is obtained by implantation of beryllium ions (Be+. especially of the multiplication layer. More recently. The active junction is formed by diffusion of . 0. It is suited to illumination through the front or rear faces. the implantation of beryllium Be+ ions. 5 1013 cm-2) followed by annealing at 700°C in a sealed ampoule. 2 μm). 150 keV.1. and the depleted region under the guard ring must reach the ternary layer before that of the active region’s junction. This requires that the guard ring junction be gradual with a weak level of doping.5 μm). followed by redistribution annealing [SHI 83]. For a long time. was the only method of fabricating effective guard rings. better understanding of epitaxial growth techniques has allowed good results to be obtained by combining localized implantation and epitaxial regrowth [WEB 88]. requires strict conditions to be met in terms of doping and thickness of the epitaxial layers.9. the guard ring junction. whose principle is summarized in Figure 3.15. required to reach the necessary reliability levels for detectors targeted at submarine communications. Construction of the guard ring. A realistic predictive model for the behavior of planar InGaAs/InP APDs was put forward by Harari [HAR 91b] after a detailed analysis of their operating conditions.5 1016 cm–3). Specifically. thus allowing a reduction in the excess noise factor. which is slightly deeper than the active region.5 μm). 2 InGaAsP layers (not intentionally doped. 2. and the InP-N– covering layer (n = 8 1015 cm–3. 3.Avalanche Photodiodes 97 ionizing carriers in InP. must remain within the InP.

whose median lifetime under these conditions is greater than 4. – an excess noise factor F = 5 at M = 10.55 μm respectively. – a maximum gain Mmax = 30.16. A nitride layer of Si3N4 ensures surface passivation. 100 μA) have demonstrated the robustness of these APDs.85 A/W and 1 A/W at 1.000 hours. M(V) and Id(V) characteristics of a fast InGaAs/InP APD Experiments on accelerated aging under extreme conditions (175°C. and another acts as an antireflection coating.98 Optoelectronic Sensors cadmium at 620°C in a sealed ampoule. they are one of the key devices for fiber-optic communications at 2. In addition. – a bandwidth of 2. To .5 GHz. Typical performances of such 70 μm diameter APDs are: – a response coefficient (at M = 1) of 0. – a dark current at 0. Figure 3. thanks to their high degree of reliability.5Gbit/s.9 VB < 30 nA. for a duration which depends on the required characteristics of the multiplication layer.3 μm and 1.

along with a reduction in its thickness.2.18 μm quaternary layer (the wavelength corresponding to the bandgap energy of the GaInAsP compound) ensuring equipartition of the discontinuity of the valence band (0.38 eV). With only one thin 1. the variation in concentration and thickness of the layers deposited through liquid phase epitaxy had a beneficial effect.0 and 1. two or three thin quaternary layers.9. APDs with one.5 GHz) must be increased. comparable with single-mode fiber illumination. with the devices present on a treated wafer offering a wide range of behaviors. such as molecular jet epitaxy or vapor phase epitaxy. minimizes the capacitance of the junction. which is the main limitation of the bandwidth of these APDs. an active region of 30 μm diameter.5 μm in order not to degrade the response coefficient. 3. or with a quaternary layer whose composition varies continuously between InGaAs and InP. It is also necessary to reduce the transit time across the InGaAs absorption layer while keeping its thickness between 1. More sophisticated growth techniques. we must reduce the “accumulation” of holes due to the large valence band discontinuity at the InP/InGaAs interface. Finally. a theoretical study [RIP 83] shows performances . This helped develop and understanding of their operating conditions and to acquire a good understanding of the fabrication requirements for such APDs.Avalanche Photodiodes 99 begin with. fabrication yields greater than 50% were obtained. have been fabricated with bandwidths between 7 and 9 GHz. the normal bandwidth of APDs (typically 2. Fast InGaAs/InP APDs In order to meet the need for fast detectors for high-bitrate communications at 5 or 10 Gbit/s. while tightening the positioning requirements for the junction layout with respect to the InP/InGaAs interface. In this way. contributes to a reduction in the avalanche trigger time. the increase in doping of the InP multiplication layer. In order to do this. allowing the deposition of layers that are very uniform in thickness and concentration have led to even higher fabrication yields (>80%).19 eV instead of 0. Furthermore.

a value already corresponding to a significant noise factor.3 μm and 0. – response coefficient at M = 1 of 0.95 A/W at 1. this ratio is around k ~ 0. these InGaAs/InP APDs targeted at 10 Gbit/s communications are reaching the limits of performance. the measured performances are as follows: – dark current < 60 nA at M = 10. In contrast.or three-layer structures which are more complicated to fabricate. using a modified structure or a material which has this . when an infrastructure for 2. However. – bandwidth > 6 GHz at M = 8. where at weak electric fields (E < 3 105 Vcm–1).5 Gbit/s configuration. For a given material. III-V low-noise avalanche photodiodes The performance of APDs in terms of noise and bandwidth depends strongly on the ratio of the ionization coefficients of the charge carriers.9 A/W at 1. – dark current Id0 lying between 0. – maximum gain Mmax > 30 for the majority of diodes.10. In order to reduce the excess noise factor and increase the gainbandwidth product for III-V APDs based on InP or GaAs.100 Optoelectronic Sensors equivalent to two. with the best values being around 10 nA.5 Gbit/s optical communications is already in place. it is necessary to strongly increase the ionizability of one type of charge carrier.5. k ~ 0. This is the case for silicon. which is close to the behavior of ionization only by electrons. 3. they will be higher when the ratio between the ionization coefficients of the least and more easily ionized charge carrier is weaker.5 Gbit/s in a 4 2. After fabrication. – excess noise factor F = 6 at M = 10.8 and 4 nA. System designers often prefer to use InGaAs/InP APDs at 2. with a fabrication yield greater than 75%. in InP where the holes are the more easily ionized.55 μm.01.

or alternatively benefits from an artificial reduction in its ionization energy. an electron which undergoes a transition in a barrier region crosses the interface and “falls” into the adjacent well.55 eV. III-V super-lattice or MQW APDs The multiplication region of a MQW APD consists of a stack of layers with alternately wide bandgaps (“barriers”) and narrow bandgaps (“wells”) which produce discontinuities at the interface.47As ( Ec = 0. thus. movement from a well to a barrier reduces the ionizing power of the electron within the barrier.47As ( Ec = 0. Alternatively.2 eV) is more favorable than InP/In0. whose values depend on the nature of the materials used.Avalanche Photodiodes 101 property under specific conditions. gaining an energy Ec .38 eV). The “barrier/well” pairs favor the transport of electrons when Ec > Ev or of holes when Ec < Ev. leading to a reduction in the ratio I / I for this structure. such as in InP/InGaAs.1. the ionizing power of the holes traveling in the opposite direction is less favored. Pairs of materials with very asymmetric bandgap discontinuities are favored. The crossing of numerous barrier-well interfaces in succession translates into an increase in the multiplication triggered by the electrons.53Ga0.48As/In0. Chin et al. and overall this results in a reduction of the effective ratio of ionization coefficients k = I / I. it is the ionizing power of holes which is favored.52Al0. The number of barrier-well layers and the . Ev = 0. [HIL 81] showed that ternary GaAlSb structures offer a significant increase in the ionizability of holes close to a specific composition ratio. For example. between conduction bands Ec and between valence bands Ev. On the other hand.10. leading to a reduction in the excess noise factor. thus favoring ionization in the wells.53Ga0. Hildebrandt et al. the combination of In0. equivalent to an increase in I. Conversely. in the case of the InAlAs/InGaAs pair.22 eV. When Ev > Ec . 3. This is also the case for the II-VI ternary compound HgCdTe. [CHI 80] suggested reducing the ionization energy of holes or electrons by taking advantage of the numerous discontinuities in the valence or conduction band which can be found in multi-quantum-well (MQW) structures. Ev = 0.

an excess noise factor F(M) ~ 3 at M = 10 and a high reliability. a gain-bandwidth product of 110 GHz. around this composition. 3. GaAlSb/GaSb APDs The strong increase in the ionizing power of holes which is seen in Ga1–xAlxSb/GaSb APDs with x ~ 0. 13 repeats) and an absorption region of InGaAsP (1. Spin-orbit resonance APDs 3. to the ionization mechanism for holes which causes the valence band to be split in two via spin-orbit coupling. through the resonance effect.55 μm.05.2.1.05 has made it possible to obtain devices with a weak excess noise factor F(M) = 3. and a gain-bandwidth product of 90 GHz [KAT 90]. according to the interpretation of Hildebrand et al. p = 8 1015 cm–3) separated by a thin doped layer of InP-P+ (0. a bandwidth of 15 GHz. the ratio of the ionization coefficients / and the ionization energy of holes is equal to Eg. These APDs are characterized by a weak dark current of 0.36 μA at M = 10.10.8 at M = 10. Up to now. They are well fitted to the specifications of high-sensitivity 10 Gbit/s receivers. but much lower than the value of 20 to 30 that has been anticipated. corresponding to a value of k = I / I ~ 5.3 μm.04 μm. is linked. p = 4 1017 cm–3) in order to reduce the electric field between the two regions.2. Subsequent construction of Ga1–xAlxSb/GaSb APDs with x ~ 0. The result is that. If Eg 1. sensitive at 1. These performances are higher for these parameters than those obtained with comparable .8 A/W at 1. the best performances have been obtained [WAT 97] for a planar APD with an implanted Ti guard ring with a MQW multiplication layer of InAlGaAs/InAlAs (0.10. a response coefficient of 0. [HIL 81].27 μm.55 μm. the structure of the energy bands specific to this material has the unusual property of having a bandgap energy Eg close to the difference in energy separating the maximum values for the valence band and the additional split band.3 and 1.102 Optoelectronic Sensors thickness of those layers both have a significant influence on performance.

The n-type guard ring is obtained by mercury diffusion. response coefficient and excess noise factor have not been confirmed at the predevelopment stage.72 eV and I / I = 30.6) corresponding to these two transmission windows offer the same electronic properties of spin-orbit coupling. The result is a PIN junction.Avalanche Photodiodes 103 InGaAs/InP APDs. with a high homogenity in terms of doping. The active region junction is made by ionic implantation (Al++. and photolithography.7 – 0. 3. was extended at the start of the 1980s for the development of alternatives to Ge APD detectors for fiber-optic telecommunications at 1. Encouraging initial results obtained in terms of dark current. . A method has been developed for growing Hg1–xCdxTe wafers with a given stochiometric composition.2.55 μm. 40 mm in diameter. with the best being the composition with x = 0.6 are the most suitable. which offers the opportunity to develop APDs with performances similar to Ga1–xAlxSb.5 μm and 8–14 μm. and a low level of dislocations.3 μm and 1. The alloys with cadmium-rich compositions (x ~ 0. HgCdTe avalanche photodiodes Research into ternary. which had already been explored with the aim of fabricating detectors for military applications in the infrared windows at 3. Eg = 0.10.62. Compositions with x close to 0. however. 1014 ions cm-2) followed by redistribution annealing. because they cover both transmission windows and also give the highest values of the ratio I / I [ORS 87]. allowing the construction of p-type substrates with doping levels around 2 1016 cm-3.2. mercury-rich alloys Hg1–xCdxTe. 65 keV. The devices are constructed in a planar structure using traditional techniques of insulator deposition by cathode bombardment. the inability to reduce and stabilize their dark current or to fabricate a planar APD has led to these investigations being abandoned.

Preliminary results obtained for Si/InGaAs structures fabricated using the technique of wafer fusion are very promising: bandwidths of 13 GHz for gains of 135.104 Optoelectronic Sensors 3. on the other hand. Si/InGaAs APDs These APDs.1. 3. corresponding to a gain-bandwidth product of 315 GHz have been measured. the dark current is raised very close to breakdown and is unstable under a bias voltage [HAW 97]. Important work remains to be done to develop enough understanding of wafer construction and the stages of APD fabrication in order to overcome the limitations linked to the large discrepancy in the lattice parameter between Si and InGaAs on the one hand and. 3. in a vertical “waveguide” structure whose InGaAs core is very thin (~0.11.11. also linked to the difficulties of passivation. a high gainbandwidth product and a low level of noise. Prospects A large proportion of the work in progress is aimed at the development of devices for high-bitrate communications (20 to 40 Gbit/s) that are both fast and low-noise. combining the very low excess noise factor of Si APDs with the strong absorption of InGaAs. However.11.5 μm). “Waveguide” MQW APDs In order to reduce the maximum transit time of charge carriers without reducing the quantum efficiency. . the light signal is injected laterally. perpendicular to the electric field. “waveguide” MQW APDs and low noise APDs with a very thin multiplication region. should simultaneously allow high quantum efficiency and a high bandwidth.2. Particularly significant is the work on Si/InGaAs APDs.

is no longer negligible compared to the thickness of the multiplication region.11.55 μm. Under such conditions.52Al0. and as a result there is less noise. a high incorporating a resonant cavity centered at response coefficient (above 0.3. the ionization process is less random. whose multiplication region in the highelectric-field monolithic InAlAs material is separated from the lowfield InGaAs absorption region by a very thin but heavily doped InP layer. a record gain-bandwidth value of 290 GHz was obtained for InAlAs/InGaAs APDs with a vertical structure = 1. 3.55 μm for a fiber-diode coupling factor of 80% [COH 97]. these APDs require an excellent understanding of epitaxial . InAlAs whose ionization coefficients are similar. For a diode of length 20 μm. Low-noise APDs with a very thin multiplication region The study of APDs with a very thin multiplication region (0.8 A/W) and a weak dark current for a thickness of 0.Avalanche Photodiodes 105 Since the MQW InAlAs/InGaAlAs or InAlAs/InGaAs multiplication region is also very thin.5 μm) has demonstrated the possibility of making devices with a low excess noise factor out of materials such as InP.9 A/W at 1. the transit time of charge carriers is very short and bandwidths of 25 GHz have been obtained. McIntyre’s analysis is not relevant because the “dead space”. This makes such structures very attractive for high-bitrate applications as they are simpler to fabricate. Additionally. even in the case of a primary injection by the least ionizing charge carriers. the response coefficient at M = 1 is 0. However.47As/In0. The primary injection of electrons makes the presence of quaternary transition layers unnecessary. the distance required for a charge carrier to acquire enough energy to ionize. The gain-bandwidth product is more than 5. Excess noise factors comparable to those measured for multi quantum well APDs (F ~ 4 at M = 10) have been published for traditional [DRI 99] and waveguide [KIN 00] In0 53Ga0. Consequently.2 μm for the InAlAs multiplication region [LEN 99]. GaAs.2 to 0.48As APDs with a high bandwidth (~20 to 30 GHz).

there is a specific APD associated. which has seen the installation of many intercontinental links at higher and higher bitrates (5 to 10 Gbit/s) with shorter and shorter latencies. In addition. the advent of optical links at bitrates of 20 or 40 Gbit/s and beyond leaves no question that there is no current APD structure which is really applicable. 0. Ge APDs. and further research is required. the needs of designers up to 5 Gbit/s and even 10 Gbit/s.3 μm and 1. tailored to the relevant emitter. The perfection of high-performance APDs has been one of the important factors in the achievement of competitive optical communications. 3. with one not being possible without the other. and are much more delicate to fabricate than the traditional structure in [DRI 99]. Traditional SAGM InGaAs/InP APDs have met. and continue to meet. Conclusion Initially responding to the demands of military applications.55 μm respectively.85 μm.12. The result is that for each generation of fibers corresponding to a spectral transmission window of silica. 1. An increase in the sensitivity of the photoreceiver. allows either a reduction in the power of the signal emitted in the first place or a reduction in the number of repeaters along the link.106 Optoelectronic Sensors growth in order to achieve stacking of the “mirror” layers and the active region. . and then SAGM InGaAs/InP APDs. the development of avalanche photodiodes has occurred in parallel to that of optical fiber telecommunications. the high level of reliability of SAGM InGaAs/InP APDs has contributed to the expansion of submarine optical fiber links over the last decade. However. be it P+ PN+ Si APDs. consisting of the APD combined with an associated amplifier.

38. vol. OGUEY C.13.. 987. [JOH 64] JOHNSON K.. FORREST S. Lille Flandre Artois University of Science and Technology. no.. [HAR 91a] HARARI J. 303. [HAW 97] HAWKINS A.. p. [HIL 81] HILDEBRAND O. [CHI 80] CHIN R.. 1991. 3. 284. “Waveguide AlInAs/GaAlInAs avalanche photodiode with a gain-bandwidth product over 160 GHz”. vol. 9. [INT 66] MCINTYRE R. vol. no.. KUEBART W. p. KRAMER B. [DRI 99] DRIES J.. 35. Devices. Digest of Technical Papers. 12. RIPOCHE G. p. LECOY G. 211.. MIKAMI O.. p. WU W.. vol. vol.. 1492.. [COH 97] COHEN-JONATHAN C.... International Solid State Circuits Conference. 3. ESSDERC. VILCOT J-P.P. HESS K.. THOMSON K. IEEE Proceedings. [HAR 91b] HARARI J. HOLONYAK N.. 1964. “In0. no. STREUBEL K. 4. vol.. vol. “Ga1-xAlxSb avalanche photodiodes: resonant impact ionization with very high ratio of ionization coefficients”. no. charge. Electron. 16. “Avalanche photodiode frequency response”. 2.. PRASEUTH J. SALSAC P. vol.53Ga0. 33. 334. “Dependence of the GaAs/AlGaAs superlattice ionization rate of Al content”.48As separate absorption.. STILLMAN G. 7. vol. 1974. 1980. 1997. 1966. Applied Physics Letters... GIRAUDET L... “High gainbandwidth product silicon heterointerface photodetector”. 64. 33. 467. TANG J.47As/In0. no.. Applied Physics Letters. QE-17. 1991. no. 70. Etudes théoriques et expérimentales de photodiodes à avalanche planaires GaInAs-InP. no. DECOSTER D. BONZO A. 1997. p. Journal of Applied Physics. IWAMURA H.. p.. “High speed photodiode signal enhancement at avalanche breakdown voltage”.. “Multiplication noise in uniform avalanche diodes”. ABRAHAM P. p. . 1999. Electronics Letters.Avalanche Photodiodes 107 3.. and multiplication layer long wavelength avalanche photodiode”.. “Photodiode Schottky SiPt-Si N à avalanche”. 1981. no.. “Numerical simulation of avalanche photodiodes with guard ring”.. MAILLE C. Electronics Letters. ED-13. vol. PILKUHN M.52Al0... 1. IEEE Journal of Quantum Electronics. [EMM 67] EMMONS R. p... [KAG 89] KAGAWA T. 1989. 138. Electronics Letters. p. 1. 3705. no... IEEE Trans. 17. BENZ K. 54. BOWERS J. Bibliography [ALA 74] ALABEDRA R.. PhD Thesis. p. “Impact ionisation in multilayered heterojunction structures”. RIPOCHE G. 1967.

p. MIKAWA T. 2000. VALENZA M. IEEE Photonics Technology Letters. 1953.. [LEC 75] LECROSNIER D. 4. “InGaAs avalanche photodiodes for 1 μm wavelength region”. p.. 1983. International Solid State Circuits Conference... CAMPBELL J.48Al0. “Low noise avalanche photodiodes by channeling of 800 keV boron into <110> silicon”. no. no. July 1980. [MIL 55] MILLER S. “Electron multiplication in silicon and germanium”. Electronics Letters. p. MESLAGE J. STREETMAN B. 416. p.. vol. p. 6199. 1999. p.. IEEE Photonics Technology Letters. “High speed ( 6 GHz) InGaAs/InP avalanche photodiodes grown by gas source molecular beam epitaxy with a thin quaternary grading layer for high bit rate ( 5 Gbit/s) systems”.. NISHI H.. HOLMES A. ALABEDRA R. HANSING C. 5. 56. KAGAWA S. vol. LAMBERT M. Revue de Physique Appliquée.C.6Te à = 1.. 9. [MEL 66] MELCHIOR H. no. p.. Digest of Technical Papers. NIE H. Montpellier. PEYRE J. AMOUROUX C. “Signal and noise response of high speed germanium avalanche photodiodes”. 1990. 22.. vol. 12.. 829. 1761. 1079. “An AlxGa1-xSb avalanche photodiode with a gain bandwidth product of 90 GHz”. 1955. 1978... MCAFEE K.. no.. [KIN 00] KINSEY G. LYNCH W. STREETMAN B. 2. “A fast high gain silicon photodiode”. [ORS 87] ORSAL B. Journal of Applied Physics. Technical Digest of International Electron Devices Meeting. [RUE 66] RUEGG H. 19. vol.. no.. 11.. [LEN 99] LENOX C. 1.. no. 49. p.108 Optoelectronic Sensors [KAN 78] KANEDA T. “Waveguide In0.. vol. 1975. 1966.. 14. p.. YUAN P. vol. 1162... 12.. INADA T. PELOUS G... 595. 1966.-L. MIURA S. IEEE Transactions on Electron Devices.. 91.. 54.48As avalanche photodiode”. Physical Review. DENTAI A.. 534. 1234. p. BRILMAN M... vol. MOTTET S... 99. KANEDA T. [KAY 53] MCKAY K. vol.. vol.. Journal de Physique. ED-13. “Resonant cavity InGaAs. BOISROBERT C.. no.InAlAs avalanche photodiodes with gain-bandwidth product of 290 GHz”. Bruit près de la résonance liée au couplage spin-orbite”. Modélisation à l’obscurité et sous éclairement des photodétecteurs à avalanche N+ P P au silicium. CAMPBELL J. Washington D. WADA O..53Ga0. IEEE Photonics Technology Letters. “Les photodiodes à avalanche Hg0. MIYAUCHI A. p.. 4. 227. “Avalanche breakdown in germanium”. vol. 12. [MAI 80] MAILLE C. no. KINSEY G. “Optimization of avalanche silicon photodiodes: a new structure”.. MIKAWA T.. 3. [SHI 83] SHIRAI T. 1987. HOMLES A. 9.47As/In0. Physical Review. III. YAMAOKA T. [KAT 90] KATSUWAKA H. TANAHASHI T.55 m..... . Languedoc University of Science and Technology.4Cd0. p. [RIP 83] RIPOCHE G.. LECOY G.. PhD Thesis. vol. September 1983.. YASUOKA N.. RIPOCHE G..

[YAM 76] YAMAOKA T. p. MCINTYRE R. Applied Physics Letters.. KANEDA T. p.. September 1976. 1619.. IEEE Photonics Technology Letters. 11. no. NAKATA T. [WAT 97] WATANABE I. 1966.. HOLUNGA M. “High reliability and low dark current 10 Gb/s planar superlattice avalanche photodiodes”. 1997. “Silicon avalanche photodiodes”. TSUJI M.. SCHEIBLING J.. 831. ROSSI D. Technical Digest of Optical Fiber Communications Conference. p.Avalanche Photodiodes 109 [STI 74] STILLMAN G.. Fujitsu Scientific & Technical Journal.WOLFE C. vol. 1988. 1974... OFC 88... 87. TAGUCHI K. “Effect of junction curvature on breakdown voltage in semiconductors”. 9. MATSUMOTO H... 9. . “Electroabsorption avalanche photodiode”.... MAKITA K. 671. p. vol.. [WEB 88] WEBB P. no. [SZE 66] SZE S. 25. vol. Solid State Electronics. GIBBONS G.. “Planar InGaAs/InP avalanche photodiode fabrication using vapor phase epitaxy and silicon implantation techniques”. vol. 12. DONNELLY J. 12.

This page intentionally left blank .

Chapter 4 Phototransistors 4. This is because. it was mostly bipolar Si phototransistors.2 μm wavelength. with an optical gain of the order of 100. .1. which were developed. operating in the near-infrared. since the optical absorption coefficient of silicon is weak (<103 cm–1). the foreseeable applications were not those requiring high speeds. with the main motivation for this research being to produce a combination of photodetection and signal amplification in single device. The first work on phototransistors dates from 1951 when Shockley et al. However. The first demonstration of this type of photodetector was achieved two years later when Shive [SHI 53] described a Ge-based n-p-n phototransistor. Chapter written by Carmen GONZALEZ and Antoine MARTY. on the one hand. [SHO 51] proposed the use of a bipolar n-p-n or p-n-p structure as a phototransistor operating with a base current generated by optical means. operating in the spectral region of 1. This interest in phototransistors has recently been revived with the development of optical fiber transmission systems and the evolution of the III-V materials technology. Introduction The effects of light on transistors have been studied since transistors were first created. In the 1960s.

On the other hand. just like their transistor analogs. Heterojunction bipolar transistors and field effect transistors.112 Optoelectronic Sensors the best attenuation and absorption properties of optical fibers are found in the near-infrared. 1. silicon is photosensitive to wavelengths in the near-infrared. Section 4.8 μm. BAA 77].85.4 is dedicated to some examples of circuits based around bipolar phototransistors. began to be studied as phototransistors in the 1970s [ALF 73. 4. In this indirect bandgap material. These two characteristics have opened up the possibility of applying phototransistors based on III-V materials to the field of optical telecommunications.3 and 1. the energy of the bandgap is 1. In section 4. with the three minimal absorption windows at 0. they can also be sorted according to structure into two categories: unipolar field effect devices and bipolar devices.5. made with these materials. Phototransistors according to their fabrication materials The first phototransistors were based on a bipolar homojunction silicon transistor.55 μm corresponding to the emission and absorption ranges of semiconductor materials based on GaAs and InP. III-V materials have absorption coefficients much higher than that of silicon (>104 cm–1).2. As a result. This section reflects this dual classification.12 eV. Finally. starting in the 1970s. We will then classify them according to their structure.1. 0. phototransistors can be classed according to their fabrication material and. In section 4.3 we describe the mechanisms of operation of a bipolar phototransistor and its main properties. we review the main fields of application of this device. technological progress achieved with III-V compounds and . in section 4.2. Phototransistors In terms of photodetectors. Later. 4.2. we present a summary of the different types of phototransistors based on the materials used for their fabrication and as a result their sensitivity to optical wavelengths.6–0.

Bandgap energy and applicable spectral range for various materials used in the construction of phototransistors . which matches the lattice spacing of InP. 1. Heterojunction phototransistors based on AlGaAs/GaAs are photosensitive to wavelengths between 0.55 μm).43 eV. operating initially at 0.8 and 0. These materials have a direct band structure.55 μm. The optical absorption coefficient of GaAs for this wavelength is relatively high: of the order of 104 cm-1.9 μm.85 μm. were later targeted at 1.55 μm where the attenuation in optical fibers is 0.5 and 0. for which there is an attenuation of around 2 dB/km. In particular.55 μm.3 and 1. it is the ternary and quaternary alloys based on InP which are best suited for photodetection at these wavelengths. This value of energy is compatible with photodetection at 1.85 μm.55 μm.85 μm) and more recently the InP variant (for = 1.85. has a bandgap energy of 0. This is why two variants based on III-V materials have been developed: the GaAs variant (for = 0.8 103 cm-1 for = 1.1. with the GaAs bandgap being 1.3 and 1. The optical absorption coefficient of InGaAs is 1.3 μm and 6. and some of their ternary compounds have bandgap energies which match the spectral windows of lowest attenuation in optical fibers: 0. Thus.Phototransistors 113 their ternary and quaternary alloys enabled the development of heterojunction devices: field effect transistors and bipolar transistors.3 and 1.3 and 1. Table 4. the ternary alloy InxGa1–xAs with x = 0. The first optical systems.2 dB/km respectively.1. This value of bandgap is compatible with the spectral window of optical fibers around 0.53. The bandgap energy (Eg) and the applicable spectral range for the main materials used in phototransistor fabrication are shown in Table 4.16 104 cm–1 for = 1.75 eV.

1. and a photovoltaic effect which occurs close to the gate/channel and channel/substrate junctions. most research has been focused on heterojunction phototransistors which are Schottky gate transistors made with GaAs (MESFETs. A large proportion of the photocarriers end up trapped in low field regions (e. Unipolar field effect transistors Unipolar field effect transistors (FETs). thus distorting the static characteristics of the device. the optical input acts as an additional terminal across which the device can be controlled optically. Physically. the barrier). or even more sophisticated microwave functions such as optical locking of oscillators and the mixing of optical and electrical signals [SEE 90].2. These devices can be integrated into MMICs (Monolithic Microwave Integrated Circuits) to achieve optically-controlled amplification or switching.1 shows the cross-section of an illuminated MESFET device (photo-MESFET). . we can explain this degradation by the poor coupling between the incident light and the gate junction (poor overlap between the optical absorption region and the region with an electric field). This popularity is mostly due to the possibilities of using the phototransistor effect in MMICs (Monolithic Microwave Integrated Circuits). Out of the family of field effect transistors. MAD 92] have shown not only that the photovoltaic effect dominates the photoresponse of the device. based on III-V materials. but also that this effect is the cause of the mediocre dynamic performance of photo-MOSFETs. Several investigations [GAU 85.2. Figure 4.2. only create a single type of charge carrier. Metal-Semiconductor Field Effect Transistors) and high electron mobility transistors (HEMT).g.2. Phototransistors classified by structure Homojunction or heterojunction bipolar transistors (photo-HBTs) and unipolar field effect transistors (photo-FETs) are three-terminal devices. 4. When illuminated. Two distinct phenomena occur in the photoMESFET during illumination by light: a photoconductor effect which is the result of the increase in conductivity due to the creation of photocarriers.114 Optoelectronic Sensors 4.

Phototransistors

115

Compared to a photo-MESFET without illumination, an important reduction in performance has been observed for the transition frequency fT and the maximum frequency of oscillation fMAX when illuminated [SIM 86]. The best dynamic performance obtained under illumination is for a bandwidth less than 100 MHz [BAR 97]. Dynamic optical performances are very poor compared to those obtained electrically.

Figure 4.1. Structure of a GaAs MESFET under illumination. The channel is an n-type semiconducting region sandwiched between a semi-insulating substrate and the space charge region (depletion region) of the reverse-biased Schottky junction

MESFET-based phototransistors have shown themselves to be a poor approach for developing photodetectors in the microwave and millimeter range. Other devices in the FET family, notably HEMT, suffer from the same limitation [ROM 96] 4.2.2.2. Bipolar phototransistors An alternative to the photo-MOSFET is the silicon-based homojunction bipolar phototransistor or the heterojunction bipolar phototransistor (based on GaAs or InP). In the bipolar phototransistor, there is an overlap between the region of light absorption and the high-electric-field depletion region. This overlap is as good as in PIN photodiodes. As will be shown in section 4.3, the frequency response to a modulated optical signal is directly related to its purely electrical properties. To make the best use of the performance of a transistor, different illumination approaches have been considered, such as

116

Optoelectronic Sensors

vertical illumination through the front or rear face, lateral illumination, or even the use of a waveguide. 4.2.2.2.1. Traditional surface illumination The first phototransistors were surface-illuminated, which is the traditional situation, particularly for silicon phototransistors. This vertical illumination allows easy coupling between the optical fiber and the phototransistor, making the integration of the device easier. Several solutions have been proposed: in the case of an opening in the metallic contact of the emitter, the incident light flux crosses the emitter without being absorbed, and the electron-hole pairs are created in the active (or intrinsic) base-collector region of the photo-HBT, that is, the region which is just below the emitter. If a base contact is removed, as shown in Figure 4.2a, the optical flux directly illuminates the base and the collector in the “extrinsic” region of the phototransistor. 4.2.2.2.2. Rear-face illumination The aim of rear-face illumination of a transistor is the creation of photocarriers in the active region of the transistor (see Figure 4.2b) without modifying the emitter contact. The traditional HBT structure is used with a substrate which does not absorb the incident light. The absorption of light takes place in the collector and the base. Furthermore, the response coefficient is improved because the metallic contact of the emitter acts as a mirror.

Phototransistors

117

Figure 4.2. Cross-section of a HBT phototransistor with (a) front-face illumination and (b) rear-face illumination

4.2.2.2.3. Lateral illumination Lateral illumination (parallel to the layers) offers another approach for the injection of light into a transistor. It is inspired by the technological processes developed for photodiodes. This type of illumination requires a face very vertical to the base-collector island, in order to obtain the best possible injection efficiency. In laterallyilluminated phototransistors, the photons and charge carries no longer propagate in the same direction, as is the case of vertically illuminated photo-HBTs. The aim of lateral illumination is to simultaneously improve the conversion of optical power into electrical power (quantum efficiency) and the speed performance (high operating frequencies) of the device. 4.2.2.2.4. Lateral illumination with an integrated waveguide However, lateral illumination does pose significant difficulties in terms of achieving a good optical coupling between the fiber and the device. This is why one trick involves the fabrication of a waveguide integrated into the structure of the device, as shown in Figure 4.3. With this approach, the matching of the mode leaving the optical fiber to the propagation mode inside the device is made easier [FRE 96]. In the best cases, an improvement to the injection efficiency is observed which rises from 50% to 90%, with an improved frequency response.

118

Optoelectronic Sensors

4.3. The bipolar phototransistor: description and principles of operation The bipolar phototransistor is a transistor designed such that the “signal” current which feeds the base terminal is mostly provided by photoelectric effects. This “signal” is then amplified by the transistor effect of the device. First we revisit the principle of operation of a bipolar n-p-n phototransistor, which consists of two distinct p-n junctions, as shown in Figure 4.4. Then, we shall discuss the parameters characterizing the phototransistor: the response coefficient, the static and dynamic gains, the response time, the conversion gain and the noise. A cross-section of an n-p-n type bipolar phototransistor, using mesa technology, is shown in Figure 4.4, along with its bias circuit.

Figure 4.3. Cross-section of an integrated waveguide phototransistor under illumination

Figure 4.4. Schematic diagram of an n-p-n phototransistor, along with its bias circuit in common-emitter configuration. The shaded regions are free from mobile charges. The base, collector and sub-collector parts can be thought of as a PIN photodiode

Phototransistors

119

4.3.1. The phototransistor effect In Figure 4.4, the n-p-n phototransistor is biased in the commonemitter configuration, with voltages VBE > 0 and VCE > 0. It differs from traditional bipolar structures, having a relatively large side area of the base-collector junction; this is called the optical window and is what becomes illuminated. This part of the device is effectively a photodiode, connected between the collector and base contacts of the active transistor. When illuminated, electron-hole pairs are created by the photoelectric effect, in the base and in the space charge region (SCR) associated with the base-collector junction (we assume for the sake of simplicity that no absorption takes place in the sub-collector and that the collector is entirely free of mobile charges). A photocurrent I , known as the “primary” photocurrent, is established between the base and collector regions; it is transported by minority electrons from the base which diffuse towards the collector, and by the carriers generated in the SCR, which are separated and moved by the electric field. The flow of this photocurrent I generates a voltage across the base-emitter and base-collector junctions such that the transistor finds itself in its normal operating system (VBE > 0 and VCE > 0). The holes attracted by the base will therefore find themselves blocked by the emitter-base junction. This excess of holes will cause a reduction in the emitter-base potential barrier, which results in an injection of electrons from the emitter into the base, from where the majority will diffuse until they are at the level of the collector. Thus, this is the traditional behavior of a bipolar transistor. The amplification of the photocurrent is a purely electrical phenomenon due to the transistor effect. 4.3.1.1. The main currents in the phototransistor In the normal operating system, the phototransistor can be characterized in terms of the following currents (see Figure 4.5): – Firstly, the emitter injects electrons into the base region. These, minority carriers in this region, diffuse perpendicular to the junction layout and, if the base is thin enough that recombination can be

– Conversely. Thus. Irb is caused by the recombination of electrons with holes. in the base. – Illumination produces the primary photocurrent I . we can use the continuity of the electron and hole currents across the transition regions to write the following equations. The flow of these charge carriers gives the contribution Ine. and in the base.5. balancing the different contributions to the current crossing the two junctions. I originates from a current of electrons which migrate directly toward the collector contact. For the base: IE I ne I pe I reb [4. a current Ipe of holes.120 Optoelectronic Sensors ignored. listed above. and from a current of holes which accumulate at the level of the base.3] . Ireb comes from the recombination of electrons in the SCR of the emitter-base junction. Mostly created in the SCR of the base-collector junction. Irb . allows us to calculate the total currents at the emitter.1] the emitter: IB I pe I rb I reb I [4. they reach the depletion region of the base-collector junction. – Generation-recombination phenomena mostly occur at the level of the emitter-base junction. majority carriers. is injected from the base towards the emitter. collector and base. where the high electric field present in this region clears them out towards the collector region. which is shown in Figure 4. Ireb . in this way. majority carriers in the base.2] the collector: IC I ne I rb I [4.

we obtain: .1].3] will be used in the following sections to define the optical gain and the electrical gain of the phototransistor. Injection efficiency from the emitter in a homojunction and a heterojunction The standard parameters for a bipolar transistor. by the following equation: I ne I ne I pe I reb [4.1. The SCRs around the emitter-base and base-collector junctions are shown by dotted lines Equations [4. The relates the ratio of the electron current Ine injection efficiency injected into the base to the total emitter current IE. [4. and the transport factor in the base T. the injection efficiency from the emitter . can be extended to the phototransistor with the help of equations [4.2] and [4.4] Recombination currents reduce the injection efficiency and they must be minimized. Distribution of the different currents of electrical and optical origin in the phototransistor.3. 4.5. If in equation [4. [4.Phototransistors 121 Figure 4.1].2.2] and [4.4] we ignore the effect of recombination in the emitter-base SCR.3] where the photocurrent term I was introduced.

and can be expressed in the following way: Dnb L pe ne 0 * mnb m* pb * mne 32 D pe Lnb pb m* pe exp Eg kT [4. – Dpe the diffusion coefficient for holes in the emitter.122 Optoelectronic Sensors 1 1 I pe [4. – Lnb the diffusion length for electrons in the base. it has been shown that the injection efficiency closely follows the following expression [CAM 85.5] I ne This ratio must be as close as possible to 1 (Ipe << Ine) in order to achieve the maximum injection efficiency.7] with: – Eg the difference in bandgap between the emitter and the base. – m*nb and m*pb the effective masses of the electrons and holes in the base. – pb the density of holes in the base. . – ne the electron density in the emitter.6] 0 where A is a constant. MOR 72]: A. – Lpe the diffusion length for holes in the emitter. The Kroemer factor is a function of the physical parameters of the materials making up the heterojunction. 0 1 A [4. – Dnb the diffusion coefficient of electrons in the base. Starting from the transport equations established for a bipolar phototransistor. a function of the thickness of the base and the diffusion length of electrons in the base. and 0 is the Kroemer factor [KRO 57a and b] established for an emitter-base heterojunction.

1. thus. and the exponential factor in equation [4.8]. 4. it is crucial that the emitter should be much more heavily doped than the base.7] is equal to 1. Conversely. for a heterojunction. POU 94]: .8] Due to the recombination current.6] and [4.3. B can be expressed as a function of the transit time in the base tB and the lifetime of electrons in the base n using the following equation [CAM 85.Phototransistors 123 – m*ne and m*pe the effective masses of the electrons and holes in the emitter. It is clear from [4. Transport factor in the base The transport factor in the base B is defined by the ratio between the electron current gathered by the collector and the electron current injected from the emitted into the base: B I nc I ne I ne I ne I rb [4. Beginning with equation [4. the base can be doped at high levels without compromising the efficiency of the junction.3. depends mostly on Eg.7] that to obtain a high injection efficiency. to obtain a close to 1. and the term exp( Eg /kT) becomes dominant compared to the ne/pb ratio. it is no longer necessary to under-dope the base relative to the emitter and/or overdope the emitter relative to the base. In heterojunctions based on the materials GaAs/AlGaAs and InGaAs/InP. These two effects combine to give an improvement in the current gain and an increase in the high frequency performance of the heterojunction bipolar phototransistor. For a homojunction bipolar transistor. thus reducing the capacitance of the emitter. The doping of the emitter can remain within relatively low limits. there is no variation in the bandgap between the emitter and the base. 1. For the heterojunction phototransistor. leading to a reduction in the resistance of the base. the size of the emitter bandgap (AlGaAs and InP) is more important than that of the base (GaAs and InGaAs). thus offering a reduction in the injection of majority carriers from the base into the emitter. CAS 89. B is always less than 1. Eg is zero.

. The response coefficient of a phototransistor The response coefficient of the photodiode base-collector part of the phototransistor is defined as the ration of the primary photocurrent I to the received optical power Popt: S0 I Popt AW [4. is also used to characterize the optical-electrical conversion of the base-collector photodiode. due to optical excitation (IC)opt of the current leaving the device IC to that same incident optical power: S IC opt IC I opt I Popt Popt AW [4. For a phototransistor. B is closer to 1 when the transit time tB is small compared to the electron lifetime n.S q 0 [4.2.12] As we will see in the following section. as for a transistor.9]. The transit time tB is smaller when the base thickness is small.3. it is necessary for the base thickness to be smaller than the diffusion length of the electrons. the ratio (IC)opt/I defines the optical gain of the phototransistor. often called the external quantum efficiency. As a result.9] According to [4. is expressed in the following way: I q Popt h h . 4. This efficiency.11] and the response coefficient of the phototransistor is characterized by the ratio of the component.124 Optoelectronic Sensors B 1 tB n [4. which is the ratio of the number of electrons collected to the number of incident photons. the base should therefore be as thin as possible.10] The quantum efficiency.

the current gain of the phototransistor is obtained without illumination and. it is defined in the same way as for a bipolar transistor.15] Because of this dependence on .14] This gain can also be expressed as a function of the injection efficiency and of the transport factor in the base B taking into account equations [4. In the common-emitter configuration. Its relationship with the gain 0 is: 0 IC IE 0 1 [4. Static electrical and optical gains of the phototransistor 4. the static electrical gain 0 is given by: 0 IC IB I ne I pe I rb I rb [4.11] and [4.3. as a consequence. Static electrical gains 0 and 0 In the static system.5]. the current gain in the commonemitter configuration.1.3. which is the ratio of the collector current to the emitter current.16] 0 . [4.3. 4. 0.8] and [4.12] we can express (IC)opt in the following way: IC opt q h Gopt Popt Gopt S0 Popt [4.Phototransistors 125 Taking into account equations [4. We also define the current gain in common-base configuration 0.13] This last equation clearly shows that the phototransistor has an effective response coefficient Gopt times greater than that associated with the base-collector photodiode.14]: B 0 1 B [4.3. can reach very high values for the heterojunction bipolar transistor.

CHA 85]: Gopt 1 0 [4. for a bipolar transistor. just as for a normal transistor. Dynamic characteristics of phototransistors In the dynamic system. It links the gain obtained through amplification. obtained at the level of the base-collector photodiode: Gopt IC I opt h q IC opt Popt [4. to the primary photocurrent I . it is necessary to establish a compromise between the quantum efficiency and the electrical gain 0. In these three cases. 4. the smaller the value of 0.18] shows that to achieve a given value of Gopt. These three dynamic modes of operation of a phototransistor will be analyzed below.4. the phototransistor effect can be put to good use in the quasi-linear “small-signal” mode. the illumination varies with time and we are interested either in (amplitude) modulated light or in an impulse.2. In addition.126 Optoelectronic Sensors 4. in the non-linear mode to achieve multiplication and mixing behavior and in the “large signal” switching system. It is defined as the ratio between the component linked to the collector current (IC)opt and the primary photocurrent I .17] Gopt is the equivalent of the electrical gain 0 under illumination.3. due to the transistor effect. Gopt is always smaller than 0. .3. depending on the intended application.18] Equation [4. Several authors have shown that.3. The result is that the larger the value of . Gopt is proportional to the product of the external quantum efficiency and the electrical current gain [CAM 85. Static optical gain Gopt The static optical gain of a phototransistor is obtained under continuous illumination. and vice versa.

or alternatively it may include a continuous component (modulated signals) which can be put to the particular use of ensuring a pre-bias in addition to that obtained by electrical access to the base.4. The global response time of a phototransistor is no different to the “electrical” response time of the transistor element triggered by the current I induced in the base-collector photodiode.3. This is based on an electrical model of a bipolar transistor known as the Giacoletto circuit. Two time constants are therefore associated with the dynamic operating system of a phototransistor: the first is due to the intrinsic transport of photocurrent carriers. Small-signal operation In the context of small-signal applications.6b.1. the excitation by light may correspond directly to the signal to be amplified. this sets the limit for the use of a phototransistor as a photocurrent amplifier. which is established for the common-emitter arrangement [LET 78]. This transition frequency can be calculated in a similar way to the calculation of the equivalent small-signal method for the hybrid configuration. Transition frequency calculated using the hybrid model The behavior of a phototransistor in the small-signal system is shown in the equivalent circuit diagram of Figure 4. which is the frequency at which the optical gain is equal to 1. In an equivalent manner to a transistor. The incident illumination is modeled by a photocurrent source i placed between the base B and the collector C. Phototransistors whose base is electrically accessible have the advantage of allowing. The electrical response time of a phototransistor is linked to its transition frequency fT0.4. The first time constant may be more or less important depending on the geometry of the device.1. 4. to begin with. the choice of an operating point which ensures optimal linearity. however. while the second is due to the electrical response of the phototransistor.1. The output signal is taken across the terminals of a load resistance RL connected to the collector. .3. we can presume that it is the electrical response time which is most significant.Phototransistors 127 4. by selecting the bias to the base.

Under these conditions the input photocurrent i . CSBE = B.6) represents the internal reaction mechanism induced by the transition capacitance of the base-collector junction CTBC and by the load resistance RL.6. – CSBE is the diffusion capacitance.128 Optoelectronic Sensors Figure 4. (a) Electrical model and (b) its equivalent electrical circuit in the smallsignal system of a phototransistor. effects linked to the transit time B in this region. – gm defines the transconductance. – CTBE and CTBC are the transition capacitances of the emitter-base and base-collector junctions. called the primary. gm = IC /UT . for a common-emitter configuration The elements of the above equivalent circuit can be described as follows: – RBE is the dynamic input resistance of the phototransistor. commonly referred to as the Miller effect.19] in which the second term (corresponding to the current component i indicated in Figure 4. which represents the effects of accumulation of minority charge carriers in the base.IC /UT . If we now ignore the current contributions i and i . can be expressed in the following form: i v BE 1 RBE j CTBE CSBE v BE RL iC j CTBC [4. – RBE = 0UT /IC.

21] B RC CTBC At high frequencies.21] becomes dominant and we can write: gopt j j 1 CTBE gm CTBC B [4. can be written in the form: iC j i 0 gopt j 1 j 0 CTBE gm CTBC [4. which defines the dynamic current gain of the phototransistor gopt. corresponds to the total transit time. the imaginary part of the denominator of equation [4.CTBC is weak compared to the two first terms. with the transition frequency fT0 being T0 /2 . known as the optical gain in the smallsignal system. Physically.Phototransistors 129 compared with the amplified current gmvBE.23] defines the angular frequency of the transition T0. fT0. the transition frequency. T. the ratio iC /i . associated with fT0. It is interesting to analyze this time constant T. It is defined as: . of photocarriers from the emitter to the collector.22] RL CTBC If we assume that the product RC.20] Given this. the output collector current is found to be: iC g m v BE [4. |gopt| is obtained when: 1 1 TO CTBE gm CTBC B [4.23] Equation [4.

23] into [4. thus reducing the “threshold” optical power needed to reach the maximum value of fT0.24] If we substitute equation [4. at low values of Popt. With the help of equation [4.24] we find: CTBE gm CTBC gm T B [4.24] correspond to the charging times of the base-emitter and base-collector junctions respectively. T approaches the limiting value B.7.130 Optoelectronic Sensors 1 T TO 1 2 fTO [4. Alternatively.22] we can also define the cutoff as being the angular frequency at which |gopt| angular frequency falls by 3 dB. the terms in CTBE and CTBC become dominant and the transition frequency fT0 is proportional to Popt. Equation [4.13]. On the other hand.26] . T is inversely proportional to the incident optical power Popt. we can use an electrical current via the metallic base contact to pre-bias the structure. This behavior of fT0 as a function of the incident optical power has been experimentally observed. For high values of Popt. This dependence makes the use of relatively high optical powers (>10 μW) necessary. This is equal to: 1 0 B CTBE gm CTBC B RL CTBC [4. if we take into account equation [4. as shown in Figure 4.25] shows that T depends on the collector current IC (via gm) up to a certain threshold and.25] The first two terms of equation [4.

fTO varies rapidly up to a certain limiting value of Popt.Phototransistors 131 Figure 4. with an electrical signal from a local oscillator. Variation in transition frequency fTO with incident optical power Popt. Frequency response of a phototransistor for different load resistances It can be clearly seen from equation [4.7. . carrying information.26] that the cutoff angular depends on the operating point (as does TO) and on the frequency value of the load resistance RC. as shown in Figure 4. 4. after which it remains constant Figure 4.4. At weak Popt.3. and as a result possesses nonlinear characteristics which allow it to act as an optical-electrical mixer. It can mix a modulated optical signal.8.8. Nonlinear operation A bipolar phototransistor behaves similarly to a bipolar transistor.2.

GON 98. This mode of operation will be referred to below as up-conversion..). If we now replace V with VOL + VFI in [4.4. a2 and a3 are real constant coefficients. [4.27] where I0 is a continuous bias current and a1. Mixing principles Any device able to transpose an input signal from a frequency fE into another higher or lower frequency is known as a mixer. SUE 96] and has been developed for telecommunication applications. This change in frequency originates in the nonlinear properties of the mixer and can be explained in the following manner: consider a circuit element whose I/V current-voltage characteristics are nonlinear (a nonlinear resistance. its I/V nonlinearity can be described by expanding the current flowing through the element in terms of a discrete series that is a function of the voltage V: I I V I0 a1V a2V 2 a3V 3 . [4. these will enrich the spectrum of the input signal V.132 Optoelectronic Sensors This mixing behavior has been the subject of recent research [BET 98. – transposition of a high frequency input signal into a lower frequency signal. 4.27] shows that we see an infinite number of powers of V appearing at the output of the circuit. the spectrum of the output signal becomes even more complicated due to the presence of the products of VOL and VIF and their respective harmonics. If we add a second signal to the input. etc. a Schottky diode.. Two mixing configurations are possible: – transposition of a low frequency input signal into a higher frequency signal.27]. VFI = vFI sin( FT t).3. This mode will be referred to as down-conversion.1. the current (V) becomes: .2. If it is subjected to a voltage VOL = vOLsin( OLt).

. We will discuss two of these criteria: conversion gain and insulation.. we can say that the performance criteria established for a mixing transistor can be applied to a mixing phototransistor. a2 vOL sin OL t t vFI sin FI t 2 .2. OLt FI t OL FI [4. 4.2...4. Figure 4. Performance criteria As a first approximation. the main nonlinearity contributing to the mixing behavior is transconductance. Optical and electric inputs of the phototransistor (in place of the mixing phototransistor) . For a bipolar phototransistor..28] t cos OL FI 1 2 v 1 cos 2 2 OL FI t vOLvFI cos .. t 1 2 v 1 cos 2 2 FI The interesting frequencies are ( OL + FI ) and ( OL – FI ) and they can be extracted by a filter tuned to the target frequency.9.Phototransistors 133 I I0 I0 I0 a2 a1 VOL a1 vOL sin a1 vOL sin VFI OL a2 VOL t vFI sin vFI sin OLt VFI FI 2 . In a bipolar mixing transistor.3. the nonlinearity in the current gain is the most important parameter.

Taking the case of up-conversion. In order to determine the electrical power generated by the incident optical power. the frequency of the local oscillator FOL is much higher than FIF. applied to a phototransistor. the light is modulated at an intermediate frequency FIF.31] .9).29]. the denominator refers to an optical power. can be expressed as the ratio between the output electrical power at the “mixed” frequency and the effective available input optical power at the modulation frequency of the light. the mixed signal is obtained at the output of the collector.29] In equation [4. and of the input impedance associated with this junction.30] and the electrical conversion gain can be expressed as: Gcon PRF Pelec FI 2 PRF Reel Z BC I 2 [4. The electrical power can thus be expressed by the equation: Pelec 1 Real Z BC 2 I 2 [4. and the mixed frequencies FRF are: FRF = FOL ± FFI The conversion gain can be written as: Gcon Pelec FRF Popt FFI [4. we must take into account the primary photocurrent I generated by the illumination at the level of the base-collector junction. the optical input is physically separated from the electrical input (see Figure 4. ZBC.134 Optoelectronic Sensors Conversion gain In a mixing phototransistor. The optical signal “enters” through the optical window and the electrical signal is applied to the metallic contact of the base. In the common-emitter configuration. The traditional definition of the conversion gain Gcon.

and one optical. the power of FIF obtained when the base-emitter junction of the phototransistor is short-circuited (base-emitter voltage VBE = 0 V).3. the optical window (IF). Operation in the strong-signal system As was mentioned previously. 4. by the ratio between the power at the frequency FOL (or FRF) present at the access point OL (or RF) and the power at the frequency FOL (or FRF) present at the IF access point. The insulation is measured between two of the mixer’s points access.9 shows. we obtain a maximum insulation between these two pairs of terminals without needing to use external circuitry. the other application of a phototransistor concerns the detection of radiation in the strong-signal system. and this has a significantly detrimental affect on its conversion gain performance. there can be a loss of FIF signal power in each of these two electrical access points. Insulation All the frequencies generated by the mixing are present at each of the terminals of the mixer. the phototransistor has three points of access. It is defined as being the ratio between the power of the mixed signal FRF and the “primary” power of the incident signal FIF. the presence or not of the illumination induces a . OL-IF and RF-IF. As Figure 4.4. However. in other words. the base contact (OL) and the collector contact (RF). for insulations in the opposite direction. If we hypothesize that no electrical signal originating from the access point of the base can be converted into an optical signal leaving through the optical window.32] When describing experimental results this second definition of Gcon is the most widely used. the IF-OL insulation is infinite. Gcon PRF PFI VBE 0 [4. The same reasoning can be applied to the IF-RF access point.Phototransistors 135 Another definition of the conversion gain is possible. As a result.3. two electrical. Let us consider the OL and IF access points at which the FOL and FIF signals are injected. IF-OL and IF-RF. In this case.

It then enters the normal.33] Figure 4.10. as shown in Figure 4. Representation of the different switching response times of phototransistors . or even saturated. These system changes give rise to transitory phases. which defines a response time for each: the lag time t is inherent to the time it takes the transistor to start conducting (in the case of operation without pre-bias) and is linked to the charging of the transition capacitances CTBE and CTBC.10. By considering the mean values of these parameters.136 Optoelectronic Sensors switching type of behavior: the device initially blocks a signal. system when light is applied.7 CTBE I CTBC [4. or is in the weak conduction system if it is pre-biased. this time can be calculated using the expression: tr 0. it will then return to the initial mode of operation when the input signal ceases.

If we assume that the current gain and its cutoff frequency are independent of the level of injection. and is therefore defined by the lifetime of these charge carriers. if applicable) is more difficult to evaluate.35] in the alternative case. suggests that the collector current response to a step change to the primary photocurrent in the base follows an exponential law with time constant 1/ . For opposite switching. which relates IC to I in the context of this assumption.I or ICsat. nevertheless. the desaturation time corresponds to the removal of the accumulated surplus charge in the form of the minority carriers (electrons) in the vicinity of the base. the rise time which separates the start of the growth in the collector current (conventionally.34] and [4. Over the course of the decaying phase of the collector current. the result is that the first-order model. up to the point it stops conducting. IC equal to 10% of its final value) from the end of this first transient phase (IC equal to 90% of its steady-state value). Thus. .34] when saturation is not reached and: 0. The fall time tf can again be expressed by equations [4. being either 0.35].8 ICsat I T tm [4. the dynamic behavior of the phototransistor can be approached in a simple manner. using the small-signal model. can be calculated as: tm 2. bearing in mind the variation in numerous electrical parameters that apply in this transitory system.9 [4.Phototransistors 137 The rise time tr in the normal active system (up to saturation. the transistor is once more in the normal active system.

Figure 4. as can be seen in Figure 4.3. Overall. the rise time tr and the fall time tf increase with the load resistance RL. Response time of a phototransistor as a function of the collector current and for different values of the load resistance Here once more. Finally. This means that there is a certain threshold power below which the photoreceiver cannot detect a signal: a power which is determined by the power of the noise in the phototransistor. along with its load resistance and operating point. while it leads to an increase in response time.5. depending on the type of phototransistor and the circuit used. The most commonly-used approach for characterizing the noise of a photodetector is the noise equivalent current generator. Noise in phototransistors The phototransistor is the first element in a photoreceiver system. The minimum detectable power in the photoreceiver system is limited by the noise in the phototransistor and its load circuit. SMI 80.11.138 Optoelectronic Sensors . 4. This is the approach that we will develop over the rest of this section. the addition of a pre-bias also contributes to an improvement in tr and tf . applied at the input [CAM 82. WAN 86].11. . these response times vary from a few tenths to tens of microseconds. We also observe that. in the switching system. the presence of saturation leads to a reduction in these rise and fall times. The noise in the phototransistor is hence an important criterion for judging the performance of the whole photodetection apparatus.

37] These two spectral densities associated with the collector at the output of the transistor are linked to the optical input of the phototransistor in the following manner [THU 99]: iC f input 2 ith f input 2 1 gopt 2 2qIC 4kT RL [4. – the shot noise due to the collector current IC. in contrast to the other quasi-neutral regions. the base.38] . this current is the sum of the photocurrent I and the electrical bias current IBelec. cancel through recombination. this arises from the fact that each fluctuation in the IB simultaneously produces another equivalent base current fluctuation in the emitter current to compensate for IB. for the base. Physically. – the thermal noise due to the load resistance RC. in the opposite direction.36] is due to the correlation which exists between the base-emitter and base-collector junctions [MON 71]. i 2 f in A2/Hz. although heavily doped. but their shot noises are independent and add up.36] The factor of 2 in equation [4. at the output of the HPT. For the collector: iC f output 2 ith f output 2 2qIC 4kT RL [4. These two currents. associated with each source of noise are. presents an effective base resistance RB which can reach values as high as hundreds of Ohms. – the thermal noise generated in the base region and. The spectral densities of the noise power. the following: i base f 2 2 2q I I Belec 4kT RBE [4.Phototransistors 139 Four main noise sources can be associated with a HPT: – the shot noise due to the base current IB entering the phototransistor.

equations [4.36] and [4.39] According to equation [4.12b) pre-bias.21].39].140 Optoelectronic Sensors where gopt is the dynamic optical gain of the phototransistor. Nevertheless. the total spectral noise density of the phototransistor is dominated at low frequencies by the sources of noise linked to the photocurrent I .12c). the correlation between these two sources can be assumed to be negligible [ESC 95].12e). at high frequencies. we should note that the phototransistor can be integrated with a second transistor to form a . As a result. for frequencies above the noise cutoff frequency 1/f. In order to improve sensitivity. in circuits with (see Figure 4.12d) or an operational amplifier providing the currentvoltage conversion (see Figure 4. the total spectral density of noise power of the phototransistor. defined in equation [4.1. Amplification circuits Phototransistors can be used either in the linear detection system (continuously modulated signals or pulsed signals) or in the switching system. can be expressed as the sum of all the spectral densities. Photodetector circuits based on phototransistors 4. it is common nowadays to use either an active load with low input impedance (a common-base transistor) (see Figure 4.12a) or without (see Figure 4. the noise source due to the collector current becomes dominant.4.4. It has been shown that in every application – detection of weak or strong signals – the speed is limited by the value of the load resistance of the phototransistor.38]: iTot f 2 4 q I I Belec 4kT RB 1 gopt 2 2 q IC 4kT RL A2 Hz [4. 4. Some degree of correlation can exist between the sources of shot noise in the base and collector. with of course the possibility of adding an additional level of amplification (see Figure 4. to the bias current of the phase IBelec and to the base resistance RBE. in terms of its optical input. Finally.

Note: structures c) and d) are used in fast circuits Structures c) and d) have been used to make monolithic amplification circuits integrating the phototransistor and several HBTs. Main circuit diagrams for phototransistors: a) structure with two electrodes E and C.Phototransistors 141 photo-Darlington pair (see Figure 4.12f). Nonlinear circuits Phototransistors can also be used in optoelectric mixing and autooscillation configurations. b) pre-bias using access to the base. d) improvement in speed using a low impedance load (common-base transistor) or e) by current-voltage conversion. for wideband [CHA 92. KAM 95] applications. This structure clearly gives a greater response coefficient at the cost of deterioration in the operation speed. Figure 4.2. f) integration on silicon of a second stage: the photo-Darlington pair. KAM 00. The simplest circuit is shown in Figure .4. 4. c) supplementary amplification with a second stage. WAN 86a] and narrowband [GON 00a.12.

These circuits demonstrate the abilities of phototransistors to combine the functions of photodetection. The optical signal is modulated at the frequency of 0. For this application.13a and consists of a phototransistor and two matching cells. The third circuit. Figure 4.5.5. and the mixed signals recovered with up-conversion are at frequencies of 9. [LAS 00].767 GHz. amplification. The second circuit. and the mixed signals at frequencies OL ± IF are recovered at the output of the collector.2 GHz. The two optical signals 1 and 2 could be at wavelengths corresponding to a WDM (wavelength division multiplexing) scheme.567 GHz. The optical signal is modulated at the intermediate frequency IF. described by Sawada et al. This frequency is stabilized using a dielectric resonator DR.13b. The second cell is placed between the output of the collector and the load resistance RL. Galvanic isolation Silicon phototransistors have been used for a long time in optical isolators. These devices may or may not have a base contact. and the . 4.14 shows an optical isolator based on a phototransistor. The first cell is placed between the source OL and the input to the base of the phototransistor. consists of a phototransistor and a circuit associated with its base which allows it to auto-oscillate at the OL frequency of 9. This signal OL is coupled with the second phototransistor.142 Optoelectronic Sensors 4. 2. In this way. which acts as an optoelectrical mixer between the 30 GHz signal and a second optical signal. It consists of two phototransistors with their emitters joined together. [SAW 99] and depicted in Figure 4. modulated at 300 MHz. described by Lasri et al. Applications 4.367 and 9. This is in order to achieve a transfer of information which avoids any form of electrical connection. mixing and autooscillation. all that is required is to connect a structure upstream which takes care of the conversion of an input electrical signal into a light beam. galvanic isolation between the controller and the receiver is ensured. is shown in Figure 4.1. One phototransistor is used as an autooscillator at 30 GHz locked optically to an optical signal 1 modulated at the same frequency OL.13c.

Phototransistors 143 achievable voltage difference. can reach several kilovolts. Various circuit diagrams for phototransistors acting as (a) a simple optoelectrical mixer. Figure 4. (b) auto-oscillation and mixing with only one phototransistor and (c) auto-oscillation and mixing with two phototransistors Figure 4. Schematic representation of an optical isolator based on a phototransistor . which in the static system is determined by the distances which separate the two components or the connections to the different electrodes.14.13.

the gate.16. Current transfer level (current from the collector of the phototransistor/current in the input LED) as a function of the input current to the optocoupler Another application of silicon phototransistors involves power electronics. Control through optical means allows this specific problem to be solved in an efficient and elegant manner: with the thyristor and the triac – which are the two basic bipolar components in this application domain – constructed around a photodiode. . An example of its variation as a function of the input current is shown in Figure 4. whose response replaces the current from the control electrode.15. The current conversion gain can reach 500%. . whose emission line of around 850 nm is well suited to optical detection by the phototransistor or the photo-Darlington pair.15. A schematic cross-section for a photothyristor and the equivalent circuit for a phototriac are shown in Figure 4.144 Optoelectronic Sensors The device providing the electron-photon conversion at the input is normally a light-emitting diode (LED) based on GaAs. particularly for control applications and energy conversion from the industrial electricity grid. These types of devices must often meet strict requirements of galvanic isolation. Figure 4.

17. Figure 4. The phototriac is constructed by combining two photothyristors in parallel. Phototransistors for optical telecommunications This application.16. In the transmitter. acting as an optical detector.2. the light is modulated (directly or indirectly via an optical modulator) with the signal carrying the information. It is situated at the input of the receiver.Phototransistors 145 Photothyristors present very interesting possibilities for direct current power transmission at high voltages (1. for monolithic integration of this device onto a silicon substrate. The near-obligation. mostly involves heterojunction bipolar transistors based on InP and InGaAs. shown in Figure 4. 4. in opposing directions. (a) Schematic cross-section of a photothyristor and (b) the equivalent electric circuit for a phototriac In the basic process of an optical transmission system. these devices are applied in the domain of static relays which involve currents and voltages not exceeding 1 A and 1 kV. This signal can be of an analog or .500 A and 5-8 kV) [ARN 92]. Because of this. of using only one surface of the substrate minimizes the current capacities and reduces the voltage stability.000-1.5. which requires both a good response coefficient and speed. the phototransistor is the device which converts the optical signal into an electrical signal.

As a result. when acting as an optical-electrical mixer.5. As was previously stated. Schematic diagram of an optical transmission system with an analog transported signal An InP/InGaAs phototransistor is an alternative to traditional photodetection. it can attain a frequency performance comparable to that of a HBT based on InP [BLA 00].1. In digital transmission. In addition. it can convert the demodulated signal to other frequency ranges depending on the required output. These .3 and 1. phototransistors must become faster and faster while maintaining a good level of sensitivity. consisting of a PIN photodiode associated with a preamplifier. The optical window sits directly above the base. InP/InGaAs photodetectors phototransistors as pre-amplifying The transport of analog and digital data takes place at frequencies and bitrates that are constantly increasing. 4. this device is of interest because of its combination of photodetection and amplification behavior and. for example. Figure 4. = 1. currently 10 and even 50 Gbit/s can be achieved. and so the receiving circuit may need to amplify and reconstruct the signal in order to extract the original information. This device is sensitive to optical wavelengths.146 Optoelectronic Sensors digital. since its structure is similar to that of a heterojunction bipolar transistor.18.2.17. The base and collector layers are made of InGaAs and the emitter layer is InP.55 μm. This has been optimized to achieve an optical gain of 30 dB and a transition frequency of 88 GHz [GON 00]. After transport through the optical fiber. The structure of a phototransistor based on InP/InGaAs is shown in Figure 4. This signal is subject to noise and distortion. the optical signal is demodulated by the photodetector in order to recover the electrical signal.

18. The response coefficient in the PIN mode. particularly to reduce the dimensions of the optical window. in order to increase the gain and speed. as predicted by equation [4. transistor (Tr Mode) and photodiode (PIN mode). This response coefficient S0 has been increased to 0. As discussed in section 4.18]. Vertical structure of a phototransistor based on InP/InGaAs on a semi-insulating substrate of InP doped with Fe Figure 4. the base-emitter and basecollector junctions (in order to reduce their capacitances). the optical gain Gopt.19 shows the response coefficient of the phototransistor in its two modes of operation. expressed in dB. The response coefficient (R) is expressed in dB.3. is equal to the difference between the response coefficient in the Tr mode and that of the PIN mode.25 A/W. is 0. as well as the use of a base layer with a graded composition.44 A/W in another phototransistor. However. S0. .2.Phototransistors 147 high performances are the result of many refinements. the transition frequency drops to 56 GHz [THU 99a]. The PIN mode corresponds with the phototransistor operating as a photodiode (VBE = 0V) and the Tr mode operating with the transistor effect (VBE > 0V). R (dB) = 20 log S. Figure 4. while maintaining an equivalent optical gain (32 dB). . with S expressed in A/W. as a function of the modulation frequency of light.

The structure of the epitaxial layers is the same for the two devices .20. Figure 4. The two types of device have a double heterojunction and use the same epitaxial layers.20.19.148 Optoelectronic Sensors Figure 4. Frequency response of an InP/InGaAs photodiode. The circuit diagram is shown in Figure 4. The response coefficient (photoresponse) is expressed in dB Monolithic photoreceiver circuits for wideband amplification involving photo-HBTs and HBTs have been constructed. Monolithic circuit diagram for wideband amplification consisting of a phototransistor and two double heterojunction transistors. using InP technology [KAM 00]. The circuit consists of a phototransistor and two HBT transistors. The amplifier circuit proposed and built by [KAM 00] exhibits a bandwidth of 40 GHz.

from 0. other investigations were reported [GON 98.5. This positive conversion gain is high because of the direct amplification of the mixed signal by the phototransistor. to achieve conversion from an intermediate frequency ( 2 GHz) to frequency bands at 30 and 40 GHz. Later.21.22 shows the output electrical powers of the IF signal (in PIN mode) and the mixed signal OL + IF. In the frequency range of IF. to achieve conversion from the modulation frequency of the light at 2 GHz to a frequency band at 50 GHz. [SUE 96] were among the first to use a photo-HBT based on GaAs.55 μm is amplitude modulated at an intermediate frequency IF (via an external EOM modulator) in the frequency range of 200 MHz to 2 GHz.2. using the definition given in [4. GON 00]. After demodulation of the optical signal. to achieve frequency conversion (up. It can also be used in high bitrate communication systems as an integral part of timingsignal extraction circuits. or in satellite communications. The conversion gain.32].2. as a function of the IF frequency. InP/InGaAs phototransistors as an optoelectrical mixer The optoelectrical mixing transistor can be used in mixed radio/fiber network access. using InP/InGaAs technology. Suematsu et al. The modulated optical signal is injected into the optical window of the phototransistor via a single-mode optical fiber. . An example of an experimental setup used to achieve this high frequency conversion is shown in Figure 4. of the order of 7 dB. is the ratio of the output power of the LO + IF signal to that of the IF signal detected in PIN mode.2 to 2 GHz. the phototransistor mixes the OL and IF signals and the result is visualized with the help of a spectrum analyzer. Figure 4. The base contact is connected to a local oscillator LO at a frequency of 30 GHz.Phototransistors 149 4. The light emitted by a laser source at 1. the conversion gain for the LO + IF signal is relatively constant.or down-conversion).

150 Optoelectronic Sensors Figure 4. Conversion gain for the IF signal transposed to a frequency of 30 + IF GHz. Conclusion In this chapter we have covered the main characteristics of the bipolar phototransistor.21. and in the field of optical telecommunications for InP/InGaAs heterojunction phototransistors. using the HPT as an O/E mixer Figure 4.22. We have also presented its applications in the field of galvanic isolation for silicon phototransistors.6. is the particular aim of research intended to exploit the synergy between photonics and . Experimental setup for up-conversion from the intermediate frequency IF to a band at 30 GHz. using the InP/InGaAs phototransistor as an O/E mixer 4. The study of monolithic circuits incorporating InP/InGaAs phototransistors. and providing more complex functions such as optical-electrical mixing and auto-oscillation.

IPRM’2000. [CAM 85] CAMPBELL J.. p.Phototransistors 151 microwaves. no. QUA G. 1992. J. vol 7. RUMELHARD C. no. no. [ARN 92] ARNOULD J. with the goal of meeting the ever-growing requirements in sensitivity.. 5.. vol. IEEE Trans.. 1989.. J. vol.. p.. ELZE G. [BAR 97] DE BARROS L.. 13. 1985. SEEDS A.. 2. 605-609.. . Light..7. p. Phys.... Proc. Dispositifs de l’électronique de puissance. “Gain of a heterojunction bipolar phototransistor”. OGAWA K. Tech. 389-447. MADJAR A.L. 1982. HOUSTON P. vol.P.. LIU C. 28. 1973. p. Paris.C. Bibliography [ALF 73] ALFAROV Z.H...P... LUNARDI L. NIKETIN V. [BET 98] BETSER Y. Appl... Phys. Electron. on Microwave Theory Tech. KAUFFMAN N. 1977. RITTER D. KOROL’KOV V. 16. Lett. 8.. GNAUCK A.. [BLA 00] BLAYAC S.M.. RITTER D.. 53.N. “A 10 Gbit/s OEIC photoreceiver using InP/InGaAs heterojunction bipolar transistors”. ROMER M. HERCZFELD P. 3.. p. p. HAMM R. 6. 4. Electronics Letters. 1992. RIET M. 2000. 466-468. FRANKEL M. of the Indium Phosphide and Related Materials. GLOANEC M. 1998.. no.. 1203-1208. Sov. no. Semiconductors and Semimetals. “Lateral design of InP/InGaAs DHBTs for 40 Gbit/s”. WALF G.. J. Masson. p.. “Photoreponse of microwave transistors to high-frequency modulated lightwave carrier signal”. 1368-1374. 1997. MADJAR A. BENCHIMOL J... p. Circuits intégrés en arséniure de gallium. GODIN J..A. MERLE P.. Semicond. “Phototransistor utilizing a GaAs-AlAs heterojunction”. [CAM 82] CAMPBELL J.. 4. “GaAs MESFET: a high-speed optical detector”...C. “Heterojunction phototransistors for longwavelength optical receivers”. [BAA 77] BAAK C. vol.. [CHA 85] CHAND N.. IEEE Transactions on Electron Devices...I. Hermes. 780.... no.L.L.. vol. “Phototransistors for lightwave communications”. vol. ED-32..E. AKHMEDOV F. 139. 22D. speed and functionality of photodetectors. PAOLELLA A. [CHA 92] CHANDRASEKHAR S. PANISH G. 481-484. ROBSON P. p. BERDAGUER P.A. 622-627. DUCHEMIN J. SCAVENNEC A... 1985. vol. 45. [CAS 89] CASTAGNE R.. “A single stage three-terminal heterojunction bipolar transistor optoelectronic mixer”.

. vol. BILENCA A. ORENSTEIN M. MULLER M. Proc. “Optoelectronic upconverter to millimetre-wave band using an heterojunction bipolar phototransistor”.L. 1996. ECOC’98. MWP’00.. 18... 55-60. p. 2000.. 1957. 45..L. POUVIL P... “Quasi-electric and quasi-magnetic fields in non-uniform semiconductors”. [KRO 57a] KROEMER H. AND SHIGEKAWA N. GOLDGEIER S... 1998. 1985.. BENCHIMOL J. [GON 00a] GONZALEZ C. EuMC36. . of the International Topical Meeting on Microwave Photonics. BENCHIMOL J. LEGAUD P. Proc. of the European Conference on Optical Communication. [GON 98] GONZALEZ C. p. of the 8th IEEE International Symposium on Electron Devices for Microwave and Optoelectronic Applications. no. BHATTACHARYA P. “Optical control of 14 GHz MMIC oscillators based on InAlAs: InGaAs HBTs with monolithically integrated optical wave guides”. no. [KRO 57b] KROEMER H... 103-104. 819-822.. RIET M. of the European Microwave week 2000. 1995. SIDEROV V. 12. 43. “Theory of a wide-gap-emitter for transistors”. 1535. vol. on Microwave Theory Tech. p. vol. EISENSTEIN G. Proc. p. no. BENCHIMOL J. [GAU 85] GAUTIER J.. EDMO 2000. 2301-2307.. 5... “Ultra-widebande monolithic photoreceivers using HBT compatible HPTs with novel base circuits and simultaneously integrated with an HBT amplifier”. VURGAFTMAN I. p.. RCA Rev. 332.. no. [GON 00] GONZALEZ C..... vol. Lightwave Tech. 373-379. SINGH J. 2000. “A two heterojunction bipolar photo-transistor configuration for millimeter wave generation and modulation”. [LAS 00] LASRI J. [KAM 00] KAMITSUNA H. 33. p. MATSUOKA Y... 1.. RITTER D. IRE. ECOC 2000. LEGAUD P. 2000. ROUX J.152 Optoelectronic Sensors [ESC 95] ESCOTTE L.. vol. 1957.. IEEE Trans.. Proc. 9.. 3. GRAFFEUIL J. ZHANG X. PASQUET D.L.. YAMAHATA S. p.. p.. 2000. p. IEEE Transactions Electrons Devices.. 42. 2. 1995.. 62-65. THURET J. IEEE Transactions on Electron Devices. “Noise modelling of microwave heterojunction bipolar transistors”. JAFFRÉ P. RIET M. [KAM 95] KAMITSUNA H. RIET M. MULLER M.. J..L. COHEN S. Proc. “HBT phototransistor for remote upconversion in hybrid fibre radio distribution systems”. vol. GRUHLE A. vol 13. JAFFRÉ P. “A 82 GHz-Optical-gain-cutoff-frequency InP/InGaAs double-hetero-structure phototransistor (DHPT) and its application to a 40 GHz band OEMMIC photoreceiver”. of the European Conference on Optical Communication.. “A 28 GHz HPT/HBT monolithically integrated photoreceiver for hybrid fibre radio distribution systems”. p.. EuMC 2000.. 443-444. P. “Optical effects on the static and dynamic characteristics of a GaAs MESFET”.. PLANA R... [FRE 96] FREEMAN P. 883-888. vol. Proc..

IEEE Trans. p. “A fiber optic: millimeter-wave radio transmission link using HBT as direct photodetector and an optoelectronic upconverter”.. SPARKS M.. vol.. [MON 71] de LA MONEDA F..H. 133-142.. CHENETTE E. 1953. vol. p. IMAI N.. 1980. “Analysis of optically controlled microwave/millimeter. p. 43..wave device structures”. 12. PAOLELLA A. vol. 340-346. IEEE Trans. [POU 94] POUVIL P.. 151. 44.. Soc. [SIM 86] SIMONS R... Berlin. Thesis. IEEE Transactions on Electron Devices...R. 1996. vol.. no.. vol. p. p. Phys. no. 1951.. TAKAHASHI K. 89. VAN DER ZIEL A. IEEE Trans.. 1681-1691. HERCZFELD. 1978. no. 1996. ED-18. [THU 99] THURET J. PERSONICK S...Phototransistors 153 [LET 78] LETURCQ P. MARTINEZ M. 5. HERCZFELD P. 1. [MOR 72] MORIIZUMI T. on Microwave Theory Tech. 577-585. 1999. on Microwave Theory Tech.. no. [MAD 92] MADJAR A. on Microwave Theory Tech. vol.. p... Springer. 12. 2. “The properties of germanium phototransistors”. Rev. “Analytical model for optically generated currents in GaAs MESFET”. on Microwave Theory Tech. IMAI N.. [SMI 80] SMITH R. 1990.A. 239. Physique des composants actifs à semiconducteurs. “p-n junctions transistors”.. BHASIN K. [SUE 96] SUEMATSU E. “An experimental study on a self-oscillating optoelectronic up-converter that uses a heterojunction bipolar transistor”. [SEE 90] SEEDS A.D... p. Am. 1349-1355. “Theoretical analysis of heterojunction phototransistors”.. IEEE Trans. 38. J. p. 8.. p. no. 1971.D. 1992. 1986. “An analytical model for the photodetection mechanisms in high-electron mobility transistors”. on Microwave Theory Tech.K. Masson. Dunod University. [SAW 99] SAWADA H... . J.vol. Heidelberg.. no. 44. MTT-34. REY G. IEEE Trans. vol. 152159. p. Composants semiconducteurs micro-ondes. Semiconductor Devices for Optical Communications.1515-1521.A. [SHI 53] SHIVE J. Opt. 19. 83.N.. 2279-2287. University of Paris VI. 6. 1994. IEEE Transactions on Electron Devices.R.. on Microwave Theory Tech. no. vol. Phototransistor bipolaire à hétérojonction InP/InGaAs pour conversion optique/bande millimétrique dans les réseaux de distribution hybride radio sur fibre. IEEE Trans. “Noise in phototransistors”. [SHO 51] SHOOCKLEY W. 47. [ROM 96] ROMERO M. TEAL G. 1999. 1972. p. “Optical control of microwave devices”. 40. DE SALLES A. vol..

IPRM’99. . [WAN 86a] WANG H..L. ANKRI D. RIET M. p. 1986.. p. BENCHIMOL J... 7... Thesis. Photorécepteur monolithique intégrant un phototransistor et des transistors bipolaires à heterojonction GaAlAs/GaAs pour transmission par fibre optique. vol. no. of the Indium Phosphide and Related Materials. 22. [WAN 86] WANG H. 1986. BERDAGUER P. University of Paris XI. 391-393. Electronics Letters. “High-speed InP/InGaAs heterojunction phototransistor for millimetre-wave fibre radio communications”.. “Monolithic integrated photoreceiver implemented with GaAs/GaAlAs heterojunction bipolar phototransistor and transistors”. GONZALEZ C. 389-392. Proc..154 Optoelectronic Sensors [THU 99a] THURET J.

However. up to now. In this discussion. Currently. Chapter written by Joseph HARARI and Vincent MAGNIN. we will see that this photodetector is not without benefits. only one device for millimeter applications has been commercialized. the metalsemiconductor-metal (MSM) photodiode has a special place. it is easy to integrate. gained the interest of the large telecommunication firms. Introduction Among the different semiconductor photodetectors. and indeed has very specific properties linked to its interdigitated structure. but it has not. .Chapter 5 Metal-Semiconductor-Metal Photodiodes 5. we will first consider the operation and structure of this photodiode before discussing its different characteristics: its response coefficient. dynamic behavior and noise.1. It is a planar photodiode based on simple technology. to the best of our knowledge. We will conclude with integration possibilities for the device.

156

Optoelectronic Sensors

5.2. Operation and structure 5.2.1. Fundamentals An MSM photodiode consists of two Schottky contacts on a semiconductor. Because of this, with each contact having a currentvoltage characteristic I(V) similar to that of a traditional junction, the MSM photodiode behaves like two diodes in series, facing in opposite directions. Briefly, under bias, with one of the diodes necessarily being reverse-biased, the dark current crossing the structure is weak and, to the extent that the two Schottky contacts are identical, the symmetry of the device gives rise to a symmetric global I(V) characteristic, with each half (positive or negative bias) corresponding to the behavior of a reverse-biased metal-semiconductor junction.

Figure 5.1. MSM photodetector with its typical interdigitated structure. Inset: current lines in the structure

If the bias voltage is sufficiently high, the region between the electrodes is completely depleted and this produces an electric field. The basic principle of this photodetector is that light is absorbed in a semiconductor layer forming part of the depletion region, such that the electron-hole pairs generated are separated under the effects of the electric field, with each carrier then being directly collected by an electrode. This effect, familiar in photodiodes, takes place in a planar structure whose geometry is shown in Figure 5.1. In general, the electrodes have an interdigitated shape, which allows the carriers to be collected over a large semiconductor area while maintaining a short

Metal-Semiconductor-Metal Photodiodes

157

inter-electrode distance. If, for example, illumination is from above, this allows carriers to be collected over the whole surface of the light beam emitted by a fiber in the near field. The inset in Figure 5.1 shows the current lines between several fingers of the electrode, and shows that, due to the planar structure, the deeper a carrier is photogenerated in the material, the longer it will take to reach the electrode. This leads to a very specific distribution of transit times for the carriers. As with all photodiodes, an MSM photodetector under illumination can be characterized by its response coefficient, its cutoff frequency and its photodetection noise. However, before we consider these different aspects, we will discuss the behavior of the device in the dark.

Figure 5.2. Behavior of an MSM photodetector in the dark: a) band structure in thermodynamic equilibrium, b) band structure under bias, c) corresponding electric field under bias

158

Optoelectronic Sensors

Figure 5.2 shows the evolution of the band structure of an MSM structure under bias. The two electrodes are assumed to be of different metals, so that the band structure of each metal-semiconductor junction differs. On each side, n is the height of the barrier seen by the electrons in the metal, p is the height for holes and Vb is the builtin bias of each junction. In thermodynamic equilibrium (see Figure 5.2a), the two depletion regions depend on the doping of the semiconductor and the work function which separates the Fermi level in the metal from the vacuum level in the semiconductor. Since the semiconductor is normally weakly doped, the carriers mostly cross the potential barriers through thermoelectric effects [SZE 96], in other words only those carriers with sufficient energy cross the barrier. The sum of all the currents across the barriers is zero. Under bias, as shown in Figure 5.2b, the situation is more complicated, with one of the junctions (number 1 in this case) being reverse-biased (voltage V1), and the other being forward-biased (voltage V2). It is the current densities of electrons Jn1 and Jn2, and of holes Jp1 and Jp2 which determine the behavior [SZE 71]. If we ignore recombination in the region between the electrodes, we have:

J n1

J n 2 and J p1

J p2

[5.1]

These relationships show that the forward bias voltage of junction 2 is such that the current flowing across it is equal to the current allowed across junction 1 in reverse. Under these conditions, it is the height of the barrier n1 which determines the value of Jn1 (and hence Jn2), and it is p2 which determines the value of Jp2 (and Jp1). The value of the total dark current thus depends on the heights of the barriers, which are linked to the nature of the metal used for each electrode. In parallel to this process, the depletion region of junction 1 expands, while that of junction 2 contracts slightly. Under these conditions we can define two particular bias voltages which depend on the structure of the MSM under consideration: the first is the total depletion voltage VTD, which is reached when the depletion region of the reverse-biased junction reaches that of the forward-biased junction, that is, when:

Metal-Semiconductor-Metal Photodiodes

159

W1

W2

s

[5.2]

where s is the inter-electrode distance. The second is the flat band voltage VFB, reached when W1 = s. The electric field present in the inter-electrode region is thus only zero at electrode 2 (the anode), and it is at a maximum at the other side (at the cathode). Clearly, an MSM photodetector is only useful when the bias voltage is greater than VFB. If we assume a homogenous N-type semiconductor, with doping level Nd and permittivity s, we have:
VFB q Nd s 2 s Vb1 Vb2

[5.3]

When the applied bias is increased, the electric field increases at the cathode, which leads gradually to a breakdown process linked either to shock ionization in the semiconductor material or to tunneling through the potential barriers of the metal-semiconductor barriers from band to band. All this leads to the typical current-voltage characteristics displayed in Figure 5.3. For V < VFB, the dark current crossing the photodiode is weak and grows slowly with the bias, until it reaches a saturation value which is maintained until breakdown, which can be more or less gradual with voltage. If the two electrodes are made of the same metal, the I(V) characteristics in the dark are, in principle, symmetric with respect to the V = 0 V axis. Generally, since a photodetector must operate with a weak dark current, the best bias conditions for a photodiode are: – a bias voltage considerably below the breakdown voltage, in order to limit the dark current (V < VC); – an inter-electrode region that is completely depleted, with a sufficiently high internal electric field so that the velocities of the photogenerated charge carriers are in the saturation system (in any case, V > VFB). Of course, given the planar structure of the device, some elements must be added to this general description.

160

Optoelectronic Sensors

Figure 5.3. Typical current-voltage characteristics of an MSM photodetector in the dark

The biasing of the semiconductor, using electrodes placed above it, leads to a particular electric field distribution, as shown in Figure 5.4. Initially, the field decreases with depth in the device, and then there are local maxima in the electric field just under the edge of each electrode (or each finger for an interdigitated structure). Consequently, the breakdown of the device always begins at the edge of the electrode, and a very high electric field may exist close to the electrodes, but the field is weaker in the middle of the inter-electrode space and deep in the photodetector. Finally, as the field is weaker deeper in the structure, the carriers that are photogenerated deep in the structure not only have a great distance to travel to reach the electrode (as we have already seen in Figure 5.1), but also have a small drift speed, which increases the transit delay and decreases the dynamic performance of the photodetector. All this confirms that the optimal bias voltage of such a device must be selected with care in order to obtain good dynamic performances without degradation.

The applied bias is 4 V.2. If we bias far from the breakdown voltage. n2 and p2. the phenomena limiting the dark current are the thermoelectric passage of electrons across junction 1. it will have an effect on the dark current resulting from the thermoelectric effect.2.1. and the thermoelectric passage of holes across junction 2. p1. The semiconductor is weakly doped (Nd = 5 1014 cm–3) 5. as shown in Figure 5.2. Distribution of the modulus of the electric field in an MSM photodetector. The total dark current is thus commonly written as: J * AnT 2 exp q n1 n1 kT qV kT A* T 2 exp p q p2 p2 kT [5.2.Metal-Semiconductor-Metal Photodiodes 161 Figure 5.4] 1 exp . Materials used 5. Electrode composition Since the composition of the electrodes has a direct influence on the band structure of the photodetector due to the magnitudes of n1.4. and the values of the field are in kV/cm.2.

. choosing a metal with a high electron barrier n for the first electrode and a high p barrier for the second. made for a gallium arsenide (GaAs) MSM photodetector for different electrode metals. This results in [SZE 96]: n1.2 is the value of the electric field at the metal-semiconductor junction. are presented in Table 5. the values of the dark current are of the order of nano-amps. If we assume that the two electrodes are made of the same material. Under these conditions. V is the applied voltage. Dark current Table 5. Measurements of the dark current [ITO 86].42 eV) for different electrode metals and photodetectors with the same surface Because of this.162 Optoelectronic Sensors where An* and Ap* are the Richardson constants associated with electrons and holes. Greater barrier heights thus minimize the dark current. T is the absolute temperature.6] n1 p2 Simultaneously achieving maximum-height barriers for electrons and holes requires us to choose the contact metal for which n is as close as possible to Eg /2. n1 and p2 are the heights of the barriers seen by electrons and holes respectively.1. Dark current measured for GaAs MSM photodetectors (Eg = 1. p2 qE1 2 4 s [5. we have (see Figure 5. which are compatible with all possible uses of the device. for this type of photodetector it is desirable to decide on the cathode and anode at the fabrication stage. corrected with the terms introduced by the effects of the electric field.5] where s is the permittivity of the semiconductor and E1.1.2): Eg [5.

Metal-Semiconductor-Metal Photodiodes

163

Clearly, all this assumes that no other effect, associated with fabrication defects, surpasses the thermoelectric effect which is normally the main cause of the dark current. This can occur if the band structure is perturbed close to the metal-semiconductor junction. In this case, a narrowing of the potential barrier favors the passage of carriers through the tunnel effect, to the detriment of the thermoelectric effect. The tunnel effect can be associated with the presence of a native oxide layer at the interface, which leads to an accumulation of electrons or holes; alternatively, the effect can be assisted by trapping states present in the semiconductor material, which create energy levels inside the bandgap. As a result, attention must be paid to the epitaxy, the cleanliness of the semiconductor surface before the metal deposition, and above all to the passivation of the structure. In the case of a photodetector illuminated from above, the dielectric layer used for this final operation can operate as an antireflection coating. 5.2.2.2. Epitaxial structure The epitaxial structure of an MSM is designed starting with the absorbing layer, which itself depends on the wavelength to be absorbed according to the basic law:
h c Eg

[5.7]

where h is the Planck constant and c is the speed of light. The different materials already reported in this application are given in Table 5.2 along with the substrate they are grown on (an asterisk indicates growth with a lattice mismatch) and the associated wavelengths.

Table 5.2. Absorbing materials already used to fabricate MSM photodetectors, along with their growth substrate and their associated wavelengths

164

Optoelectronic Sensors

As Figure 5.5 shows, around the absorption layer there are normally two layers with a higher bandgap. The adaptation layer acts to isolate the absorbing layer from the substrate during growth, and the barrier layer acts to increase the potential barrier at the level of the metal-semiconductor junction [BUR 91, WOH 97]. A barrier layer is required, for example, when the absorbing material is N-type GaInAs, on which the Schottky contacts are poor quality. Furthermore, to improve the passage of carriers from the absorbing layer to the electrodes, we also introduce one or more transition layers which reduce the discrepancy in the band structure as seen by free carriers, between the barrier layer and the absorbing layer [WAD 89, ZHA 96]. These very thin layers allow a gentle transition in the band structure and prevent the accumulation of carriers in the upper part of the absorbing layer under the electrodes. This accumulation leads to a drop in the external quantum efficiency of the device, as a nonnegligible number of free carriers recombine while they are blocked at the interface. It also simultaneously leads to an increase in the transit time for charge carriers, which has an adverse effect on the frequency response of the photodetector.

Figure 5.5. Typical epitaxial structure of an MSM photodetector. The semiconductor is weakly doped (Nd = 5 1014 cm–3)

The thickness of the absorbing layer depends mostly on its absorption coefficient because, as we will see, this directly influences the response coefficient of the photodetector when it is illuminated from above or below. The thickness of the adaptation layer is normally greater than one micron, while that of the barrier layer is as

Metal-Semiconductor-Metal Photodiodes

165

thin as possible (<0.2 μm in general). In some cases, for example, that of a gallium arsenide MSM or a silicon MSM, the substrate and the adaptation layer are also absorbing, but this requires the collection of carriers photogenerated deep in the structure which, as we have already seen, decreases the dynamic performance of the photodiode. To avoid this problem, the adaptation layer of the GaAs MSM can be made of non-absorbing AlGaAs matched to the lattice of the GaAs substrate. In the case of silicon, SIMOX wafers are used, which have a layer of silicon underneath.
5.3. Static and dynamic characteristics 5.3.1. Response coefficient

The response coefficient, which provides the photocurrent generated per Watt of incident light (in units of A/W as a result), initially depends on the mode of illumination. The photodetector can be illuminated from above, across the interdigitated metallic structure, from below or side-on, through an optical waveguide. The case of lateral illumination must be treated on a case-by-case basis because the response coefficient depends both on the optical coupling between the waveguide and the detector, and between the fiber and the waveguide. This requires a detailed study of the propagation of the light in the multi-layer structure of the device. Illumination from below is similar to illumination from above, apart from the absence of electrodes to partially occlude the light entering the device; in addition, the absorption associated with the substrate must be taken into account if it is not thin and not perfectly transparent. The absorption coefficient of the substrate is generally weak at the wavelengths under consideration, but the thickness of the material to be crossed can be sufficiently large to have an effect on the response coefficient. Illumination from above is common. In this case, the response coefficient is:
q h

R

[5.8]

166

Optoelectronic Sensors

where q is the electronic charge, h the energy of a photon of the incident light and the quantum efficiency, which is itself given by:
1

R

s s d

1 exp

Wa

[5.9]

where R is the reflection coefficient at the surface of the device, s is the the distance between two fingers, d the width of a finger, absorption coefficient and Wa the thickness of the absorbing layer. The ratio s/(s + d) introduces the shadowing effect of the electrodes which is typical of this photodetector. In brief, the response coefficient is directly linked to the characteristics of the absorbing layer and the geometry of the interdigitated electrodes. As far as the characteristics of the absorbing layer are concerned, in the case of a semi-infinite layer, the absorption coefficient must be sufficiently high so that all the carriers are photogenerated in the depletion region. This is the case for GaInAs at wavelengths of 1.3 and 1.55 μm, and for GaAs around 0.8 μm. In contrast, for silicon, the absorption coefficient at 0.8 μm is of the order of 10 μm–1. This reduces the quantum efficiency of the photodetector and causes a reduction in the bandwidth, because at least some of the carriers photogenerated far from the depletion region reach it through diffusion, which is much slower here than the transport under the effects of the electric field. This clearly depends on the inter-electrode distance, as we will see. Finally, we can see that optimizing the transport of the carriers across the planar structure implies a relationship between the thickness Wa of the absorbing region and the inter-electrode distance s. If we now consider the geometry of the electrodes, we find that attempts to reduce the shadowing effect leads to two approaches. The first consists of using opaque electrodes while reducing the finger width relative to the inter-electrode distance. However, this increases the resistance of the finger, which has a detrimental effect on the bandwidth. Measurements made on electrodes deposited on silica show that the effective resistance of a metallic finger is greater than that which would be calculated based on the volume of the material.

Metal-Semiconductor-Metal Photodiodes

167

This effect is more important the narrower the finger, importance being obtained at around 0.1 μm. This may be due to metallic limitations that the structure places on electron transport becoming important. In any case, for fingers made of titanium (150 Å) and gold (350 Å), and of width 0.06 μm, the measured resistance is around 80 /μm instead of 11 /μm for d = 0.5 μm. This gives an idea of the sizes to be considered when designing the electrodes. As a result, s and d normally take similar values, within the range d s 4d. The second approach consists of research into electrodes made of transparent materials. Several possibilities exist, notably thin layers of gold [MAT 96] and doped oxides: indium tin oxide (ITO) or cadmium tin oxide (CTO). In each case, there is a tradeoff between the transparency at the wavelength under consideration and the resistance of the material. For example, doped ITO is remarkably transparent at short wavelengths ( < 0.85 μm), but it requires a compromise between resistance and transparency at longer wavelengths as doping reduces its transparency [GAO 94, SEO 93]. The same is true for CTO. In any case, an optical transmission of 100% cannot be achieved at long wavelengths. Furthermore, attention must be paid to the conditions of the deposition, which for the oxides is normally through sputtering, which does not favor particularly good characteristics of the electrode-semiconductor interface. In short, while the response coefficient is an absolute priority, the deposition of transparent electrodes clearly reduces the shadowing effect, but it is not without consequences in terms of the dynamic behavior of the photodetector. Firstly, it leads to an increase in the finger resistance, the consequences of which can be seen in the equivalent circuit of the MSM in the dynamic system. Also, the electric field diagram (Figure 5.4) shows that this is weak in the medium under each finger. The drift velocity of the photocarriers will therefore be weak at these points. As a result, the electrons photogenerated just under the transparent cathode, and the holes photogenerated under the anode, will take a long time to reach the opposite electrode. Not only is their drift velocity low as they travel through areas of weak field, but they also have a long way to travel (see Figure 5.1). Thus, transparent electrodes increase the mean transit time and decrease the bandwidth. A similar effect on the mean transit time is seen when the device is

Example of the epitaxial structure of an MSM photodetector on top of an optical waveguide . In this case.6. which must be modeled on a case-by-case basis. The few studies carried out on this subject [KUT 94] show that the transmission of light through the metallic network depends on the ratios of /p and /d. the majority of the carriers are photogenerated in the lower part of the absorbing layer. Still on the subject of the electrode geometry. Since this is a region of weak electric field. Then we observe typical optical behavior which is more complicated than a simple shadowing effect. this also leads to a longer mean transit time. are sharp enough to make it possible to built a wavelength discriminator using two parallel MSMs with slightly different periods. as much underneath the fingers as between them. Figure 5. All these complex properties. Furthermore. and on whether the optical polarization is parallel or perpendicular to the direction of the fingers. on the refractive index ns of the semiconductor. It is higher in the perpendicular case than in the parallel case. on the finger thickness. illuminated from the same beam and measuring the ratio of the photocurrents [CHE 97]. it is a minimum for /p = 1 and /p = ns. a property that cannot be ignored in the case of illumination from above is the transparency of the network of fingers when its period p (p = d + s) becomes smaller than the wavelength of light. where only a few percent of the optical energy reaches the semiconductor.168 Optoelectronic Sensors illuminated from below. The finger thickness becomes relevant when it is close to its width.

Coupling with the photodetector occurs through evanescent waves. VIN 89]. Modeling of the optical propagation through such a structure (see Figure 5. of an MSM structure on top of a waveguide. to achieve the shortest possible absorption length.Metal-Semiconductor-Metal Photodiodes 169 In order to understand all the modes of illumination. the direct lateral illumination. The response coefficient is: R 1 R K q h internal [5. For a fixed adaptation layer.10] where internal is the internal quantum efficiency determined by the epitaxial structure.6. the photocarriers are still generated in the lower part of the absorbing layer. The electrode structure of the MSM is particularly favorable to the injection of light through an optical waveguide [SOO 88. as is the case for a PIN photodiode. is difficult to achieve. it is necessary to carefully choose both the thickness of the adaptation layer and that of the absorbing layer. a certain number of optimum cases exist. for TE and TM optical polarizations at a wavelength of 1. formed into a lattice on InP.7) shows how the injected beam couples with the absorbing layer through the intermediary of the thin quaternary adaptation layer whose characteristics (thickness and composition) determine the absorbed wavelength.8) shows that. Indeed. At a conceptual level. The quaternary GaInAsP layer (with bandgap Eg = 1. The internal quantum efficiency can only be calculated through detailed modeling of the optical transport in the material structure of the . with InP acting as the cladding. the effect of the thickness of GaInAs on the internal quantum efficiency (see Figure 5.05 eV). is based on an N-type InP/GaInAsP/InP optical waveguide. R is the reflection coefficient at the entrance of the waveguide and K is the coupling coefficient between the beam leaving the optical fiber and the semiconductor waveguide. The example in Figure 5.55 μm. we will now consider the case of lateral illumination. for which absorption occurs in the first few microns after the cut face. but the thickness of this layer can be selected to be quite thin without significantly reducing the quantum efficiency. even for small thicknesses. is what forms the guiding layer. Of course. the standard for this type of technology.

however.170 Optoelectronic Sensors photodetector/waveguide. Thus. the thin width of the absorbing layer allows a short transit time to be retained. Propagation and absorption of light in a MSM on a waveguide with evanescent coupling Figure 5.7. Evolution of the quantum efficiency of the detector/waveguide as a function of the thickness of the absorbing material . this type of structure allows a high response coefficient to be combined with a small absorber thickness. Figure 5. The dynamic behavior of such a structure is comparable to that seen with illumination from below.8.

In the case of trapping. and shock ionization close to the edges of the fingers. as in the case of GaAs photoconductors. a carrier with the opposite charge will cross the device to provide electrical compensation for it. There are many effects which lead to gain [KLI 94]. careful passivation of the surface of the photodetector eliminates the problem. wherever there is a potential barrier associated with the transition between different materials.11] This trapping can occur at the surface of the semiconductor. this is reduced under the effect of recombination and increases with internal gain. The gain is then given by: Gtrapping p T [5. whose cutoff frequency is higher. A transition layer between the absorber and the barrier is thus required. In the first case. which blocks the transport of carriers and leads to their accumulation. the holes are mostly trapped close to the cathode and electrons trapped close to the anode. Generally. ZHA 96]. When it occurs at the surface. As long as this carrier remains trapped. this creates an asymmetric charge distribution. as can be seen in the case of avalanche photodiodes. attempts are made to limit both recombination and trapping in order to obtain a . which operate at low frequencies (below a few tens of megahertz). made of an epitaxial layer with variable bandgap or even a super-lattice [WAD 89.Metal-Semiconductor-Metal Photodiodes 171 To conclude. They mostly consist of two types: those connected with the trapping of carriers. or at the interface between two epitaxial layers. The second case is relevant when a barrier layer is introduced to improve the quality of the contact with the electrode. Recombination occurs either at the surface of the semiconductor between to fingers or at the interface between two epitaxial layers. and normally a barrier layer is used to eliminate surface effects. the gain effect is due to a photocarrier becoming trapped in the semiconductor for a time p greater than the inter-electrode transit time T which triggers the injection of additional carriers in order to maintain electrical neutrality. we will discuss all the internal electronic effects in the device which affect the response coefficient. Typically. Hetero-interface trapping is associated with epitaxial growth defects.

– the connection wires with inductance LF and capacitance CW. while ensuring a good reliability. The equivalent small-signal circuit (see Figure 5. Dynamic behavior As with all photodetectors. – the finger resistance RF. the dynamic behavior of MSMs is determined by their capacitance and their input and output resistances.2. increase the time constant RC of the device. and finally: 2 are the K k 0 d 1 k sin 2 2 with k tan 2 .3. One of the important terms is the intrinsic capacitance of the MSM. The finger resistance is given by: RF 2 R0 L N [5. In the equivalent circuit.13] and R where A is the surface area of the active region. the capacitance and the resistance of the contact. L the length and N the number of fingers. and the transit time of the photocarriers. in some cases.172 Optoelectronic Sensors photodetector with constant dynamic behavior up to the high cutoff frequency.9) shows: – the intrinsic capacitance CPD and resistance RPD of the device. vacuum and relative permittivities. RL is the load resistance and we see how the finger resistance can. in other words their equivalent electric circuit under modulated electric power. CCONTACT and RCONTACT. A standard analysis of the planar structure [LIM 68] shows that: C PD A d s 0 1 K k R K k' 0 [5.12] where R0 is the resistance per unit length of each finger.d and k ' 4 d s 1 k2 . 5.

for a photosensitive surface and an identical mean transit time. Secondly. a number of rules allow us to evaluate the cutoff linked to the transit time of the carriers. the thickness of the absorbing layer. while the field in the middle of the inter-electrode space is still too weak to give the photocarriers their saturation velocity. and the capacitance of the contact is a significant parasitic element. Firstly. The cutoff frequency associated with the capacitance effects is thus given by: fCAPA 2 1 RL C PD [5. and the finger resistance.Metal-Semiconductor-Metal Photodiodes 173 The calculation shows two details connected to the interdigitated planar structure of the MSM. two-dimensional physical model is required to account for the effects of the geometry of the electrode. the capacitance of the MSM is 3 to 4 times weaker than that of a PIN photodiode. visible in Figure . its resistance between 1 and 100 . the decrease in the shadow effect is thus inextricably linked to the capacitance of the photodetector. If we ignore all the parasitic elements. and of the electric field distribution produced by the bias voltage [ASH 95]. Finally. generally less than 100 fF (and in all cases less than CPD). Typical values for the equivalent circuit can be given as a guide. What happens in this case is that a local electric field is reached which is above the breakdown voltage. a thin. the capacitance of the contact. the connection wires have an autoinductance of the order of 10 pH and a capacitance of a few fF linking them to the body of the device. where the size of the devices is very small. of the hetero-interfaces to be crossed. This aspect favors the MSM for high modulation frequencies (millimeter wavelengths). Nevertheless. the capacitance of the device falls when the ratio s/d increases. The capacitance of the MSM lies between 10 and 1000 fF.14] Now turning to the transit time. that is to say the connection wire. Firstly. we find a standard time constant: RLCPD. This effect. the presence of high electric field areas just at the edges of a finger mean that it is difficult to appropriately bias a device with a large inter-electrode distance (s > 3 μm).

the mean transit time in the structure is: s 2 vsat [5. s is the inter-electrode distance and accounts for the effects of the planar structure (1 2).4. the variation of the field in the device must be taken into account. as they move more slowly than electrons. The cutoff frequency associated with the transit time is thus given by: fT 1 2 [5. If this is achieved.174 Optoelectronic Sensors 5. Figure 5. The effect of transit time is smallest when this value is close to 2 (Wa /s = 2). is more serious for holes. The bias voltage is thus an important parameter that must be adjusted to limit the mean transit time while avoiding breakdown. The saturation velocity used is taken as equal to that of the holes for the most pessimistic results.15] where vsat is the saturation velocity of the carriers (we will take the same velocity for holes as for electrons). the distribution of the current lines in the device favors a certain ratio between the thickness of the absorbing layer and the inter-electrode distance. Equivalent circuit for an MSM photodetector in the case of a small signal Secondly.9. Even for small inter-electrode distances.16] The cutoff frequency of the MSM photodiode under these conditions is: .

by measuring the variation of its shot noise as a function of frequency. Its spectral power density is: 2 iSHOT 2 q I DARK I PH in A2 Hz [5. there is a compromise to be made. its spectral density is independent of frequency. 1/f noise and shot noise. such as MSMs or field effect transistors. the shot noise. this is white noise.e. i.3. is linked to the photogeneration process and the charge transport. It is independent of frequency and its spectral density is linked to the absolute temperature T and the resistance R by: 2 iTH 4 kT in A2 Hz R [5.19] In principle. where is normally close to 1. in practice it follows the frequency response curve of the device. These effects are themselves linked to the different imperfections. . This spectral density is often higher in horizontal transport devices.3. familiar from PIN and avalanche photodiodes.18] The 1/f noise is associated with all the random generation and recombination effects in the semiconductor material. 5. Finally. This generates a specific noise whose spectral power density varies as 1/f . are thermal noise. in terms of irregularities in the structure of the crystal lattice or in terms of foreign atoms. then.Metal-Semiconductor-Metal Photodiodes 175 f 1 2 fCAPA 1 fT2 12 [5. as in other photodetectors. however. For every structure. Noise The different sources of noise in MSMs.17] An increase in the inter-electrode distance will decrease the capacitance and increase the transit time. making it possible to obtain the frequency response of a photodetector using a continuous signal. Thermal noise is linked to the variations in energy of the carriers in resistive regions.

we have: 2 iSHOT 2 q I PH 0 G 2 x in A2 Hz However. the thermal noise is much weaker than the shot noise. the shot noise also increases. However. two changes are observed. normally below about 100 nano-amps. Under normal operating conditions [SOO 91. The spectral density of the noise becomes: 2 iSHOT 2 q I DARK I PH 0 G 2 F in A2 Hz where G is the gain.5 and 3.176 Optoelectronic Sensors Having covered these items. which is why a low dark current is desirable. as is often the case. which is not desirable. we will now describe the variation of the spectral noise density of the MSM photodetector in two situations: in normal operating conditions and close to breakdown in the conditions where a gain in the photocurrent. This shot noise increases with the dark current. in association with the multiplication process connected with shock ionization. In general. WAD 88]. as breakdown is approached [SCH 90]. the part linked to the dark current is negligible compared to the part linked to the photocurrent. If. MSM photodetectors are not used close to breakdown. . linked to the shock ionization effect. the 1/f noise dominates for frequencies of the order of 1 to 10 MHz. which can be expressed as a function of the gain G. if the finger and contact resistances are sufficiently low. Due to these noise characteristics. after which it is the shot noise which dominates and which must be considered when performing calculations for photoreceivers. IPH0 the primary photocurrent and F the noise factor. is possible. so that we can ignore its effects. values of x measured up to now for MSMs are somewhat higher: between 1. Firstly. secondly. so much so that it dominates for frequencies up to the order of 100 MHz. the 1/f noise increases. as is done in the case of avalanche photodiodes: F = Gx.

the integration of the photodiode into a coplanar microwave transmission line. offer interesting and original characteristics. above terahertz levels. – weakly doped active layers which remain consistent with technologies based on semi-insulating substrates. HAR 88. – electrode technologies allowing the dark current to be minimized. but this shows the interest in a device whose main advantages are: – a planar configuration of the electrodes allowing. the weakly-doped epitaxial structure of the detector is grown before the various layers required to make the transistors and the metallic deposition of the interdigitated electrode is the same as that of the transistor gates. These properties allow the MSM to be integrated without difficulty into all integrated circuit technologies based on III-V materials. while maintaining an acceptable response coefficient. in particular. Of course. with MESFET or HEMT transistors. which are still poorly-understood.Metal-Semiconductor-Metal Photodiodes 177 5. – a capacitance significantly below that of all other photodetectors with equivalent transit time characteristics.4. whether on GaAs [BUR 91. . the inter-electrode distance required for such a performance is small (25 nm) and requires the use of an electronic mask. – a photodetection noise equal to that of a PIN photodiode. Integration possibilities and conclusion The functionalities offered by MSM photodiodes are considerable. ITO 84] or on InP [HOR 96]. the optical properties of thin metallic lattices. have been achieved [CHO 92]. even if the power limitations are more restrictive than those for PIN photodiodes. It is with these photodetectors that the highest current cutoff frequencies. In general. In conclusion.

9. O. vol.. p.G.. 61-67.. HÜLSMANN A. no.. [HAR 88] HARDER C. LIU Y.. IEEE Electron Device Letters.R. KORDOS P. IEEE Photonics Technology Letters. 42. no. no. . 140-142.. 171-173. PATRICK W. 4. [HOR 96] HORSTMANN M. Applied Physics Letters. IEEE Electron Device Letters. 2. no. p. 477-479. [KUT 94] KUTA J. January 1994. SAKURAI T. vol. Applied Physics Letters. July 1986.J. p. 828-834.S.B. HUNSPERGER R. p... [GAO 94] GAO W. 1930-1932. no.. BURGER P.. no.. SIVCO D.. 12. 7. 22. DECOSTER D.S. 8.Y. vol. 660-662..A... 1994.. Applied Physics Letters. 4. vol. 3006. LANDHEER D. 64. ADAMS J. 65. vol... 32. July 1991... [CHO 92] CHOU S.. 333-340.. vol. “Tera-hertz GaAs metal-semiconductormetal photodetectors with 25 nm finger spacing and width”. [ITO 84] ITO M... VAN ZEGHBROECK B. p.Y. vol. [KLI 94] KLINGENSTEIN M. CHO A. “InGaAs metal-semiconductor-metal photodiodes with transparent cadmium tin oxide Schottky contacts”. NAKAI K. “Monolithic integration of a metal-semiconductor-metal photodiode and a GaAs preamplifier”... IEEE Trans.J. “A wavelength detector using monolithically integrated subwavelength metal-semiconductor-metal photodetectors”. p. December 1984. 763764.15. MARSO M. KUHL J. July 1992. [BUR 91] BURROUGHES J. Proceedings SPIE. vol. p. vol. HARARI J. CHOU S. VILCOT J. 1073-1077. SCHIMPF K. 1997.M..H. IEEE Journal of Quantum Electronics.... ROSENZWEIG J. KHAN A.Y.M.. “16 GHz bandwidth MSM photodetector and 45/85GHz ft/fmax HEMT prepared on an identical InGaAs/InP layer structure”.. April 1996. “H-Mesfet compatible GaAs/AlGaAs MSM photodetector”. no. 61.. p. vol.. 5. no. “Low dark current GaAs metal-semiconductor-metal photodiodes using Wsix contacts”. FOX A. KOHLER K.. O’BRIAN H. no. on Electron Devices. April 1988.5. 7. 2. VETTIGER P.. 37. 531-532. MOGLESTUE C..2 GHz bandwidth monolithic GaAs optoelectronic receiver”. SCHNEIDER J. 5..178 Optoelectronic Sensors 5. October 1994.. p.P. vol.. p. [CHE 97] CHEN E. VAN DRIEL H. FISCHER P. p.. “High optical power nonlinear dynamic response of AlInAs/GaInAs MSM photodiode”.. [ITO 86] ITO M. WADA O. Bibliography [ASH 95] ASHOUR I.. WADA. “Photocurrent gain mechanisms in metalsemiconductor-metal photodetectors”.. “5. Electronics Letters. “Polarization and wavelength dependence of metal-semiconductor-metal photodetector response”. Solid State Electronics. no. May 1995. MEIER H. 3. ZYDZIK G..

.A. “InGaAs metal-semiconductor-metal photodetectors for long wavelength optical communications”... no. 55. IEEE Photonics Technology Letters. [WAD 89] WADA O. MOORE R. vol. KOZA M. no. CANEAU C. vol. 52-60... THOMPSON D. ABESIDA I. 1313-1315. 24. “Application of indium-tinoxide with improved transmittance at 1. [SEO 93] SEO J. “Very low dark current InGaP/GaAs MSM Photodetector using semi transparent and opaque contacts”. [SOO 91] SOOLE J. Pergamon Press. “Engineering the Schottky barrier heights in InGaAs metal-semiconductor-metal photodetectors”.. 1968. 14. no. p. Applied Physics Letters. “Monolithic integration of a thin and short metal-semiconductor-metal photodetector with a GaAlAs optical inverted rib waveguide on a GaAs semi insulating substrate”. SOOLE J..... vol.. Solid State Electronics. 3006. p.B. vol.. 27. [SZE 71] SZE S.A. BOISROBERT C. no. SCHUMACHER H. ARAFA M.... SIMMONS J. 8.. “Current transport in metalsemiconductor-metal structure”. 1971.... BHAT R.B. LEBLANC H. SONG K. TACKEUCHI A. Proceedings SPIE... 16-17.. 26... 54. HAMAGUCHI H. 1209-1218.. vol.. [SOO 88] SOOLE J. 24. HAMAGUCHI H. vol. DECOSTER D. GOUIN F. BHAT R.D.Y. vol.. VILCOT J. 25. vol. 612-613.... [MAT 96] MATIN M. FAY P.J.M. April 1996. SCHUMACHER H. MAHAJAN A.. Electronics Letters. no.R. no. LE BELLER L. p. 1478-1480. vol.A.. 1574-1575. 1997.. vol. ESAGUI R. MIKAWA T..D. Electronics Letters.M.P. ADESIDA I. 32. p. p. 766-767. March 1997. . Applied Physics Letters..C. 11..B.3μm for MSM photodetectors”. 24.A. “Noise characteristics of GaAs metal-semiconductor-metal photodiodes”. no. 19. November 1993. p. IEEE Journal of Quantum Electronics. “Noise behaviour of InAlAs/GaInAs MSM photodetectors”. IEEE Transactions on Electron Devices. LORIAUX J. 1. [WOH 97] WOHLMUTH W. 173-180.. COLEMAN D. vol. Electronics Letters. KOZA M. 3. December 1988.. p. p. 737-752. NOBUHARA H. LOYA J. [SZE 96] SZE S.. April 1988. p. [WAD 88] WADA O. Physics of Semiconductor Devices. “Waveguide integrated MSM photodetector on InP”. 5.. New York. January 1989..A.F.P. John Wiley & Sons Ltd.. April 1990.L.C. p.Metal-Semiconductor-Metal Photodiodes 179 [LIM 68] LIM Y. 15.. FUJII T. 9.W.D. p.. BHAT R. “Very high speed GaInAs metal-semiconductor-metal photodiode incorporating an AlInAs/GaInAs graded superlattice”. Electronics Letters...J. 1966-1968. New York. [VIN 89] VINCHANT J.G. no. 1996. ROBINSON B. [SCH 90] SCHUMACHER H. “Properties of alternately charged coplanar parallel strips by conformal mapping”.A.. May 1989.

180 Optoelectronic Sensors [ZHA 96] ZHANG Y...Z. 8. .G. “Improved performance of InAlAsInGaAs-InP MSM photodetectors with graded superlattice structure grown by gas source MBE”. vol.X. p. 830-832. June 1996. no. 6. LI A. IEEE Photonics Technology Letters. CHEN J..

Chapter 6 Ultraviolet Photodetectors 6. It is traditionally divided into four categories: Chapter written by Franck OMNES and Eva MONROY. covering a total spectral range which in terms of energy is about 80 times as large as that of the visible spectrum.1). around 400 nm (3. along with various others. 1804).1. Thus this observation. It was quickly shown that this particular radiation was subject to the laws of interference just like visible light (T. made it possible to establish at the start of the 19th century that visible and ultraviolet radiation exhibited the same electromagnetic properties. Ritter observed that certain types of chemical reactions were accelerated by a radiation which was not visible to the eye and whose wavelength was shorter than that of violet light. Young. We know today that ultraviolet radiation (UV) covers a wide spectral range which extends from the limit of the visible.W. Introduction The first mention of the existence of ultraviolet radiation was in 1801 when J. and differed only in their wavelength.1 eV) up to the X-ray boundary at 10 nm (124 eV) (see Figure 6. .

pollutants and biological agents.1.1 to 3. whose wavelengths range from 320 to 280 nm (from 3.87 eV).2 to 124 eV). and because of this it is often known as “vacuum UV” (we will use the term VUV for the rest of the chapter when referring to this particular spectral range). Figure 6. Spectral ranges of the main types of electromagnetic radiation The fields of application for UV detectors.87 to 4. of which there is a great variety. covering wavelengths between 400 and 320 nm (from 3. for wavelengths between 200 and 10 nm (form 6.20 eV). – optical inter-satellite spatial communications: secure transmission of data in space.182 Optoelectronic Sensors – UVA. .43 eV).43 to 6. – UVC. the spectral end of the UV between 200 and 10 nm cannot really propagate except in a vacuum. – fire detection: fire alarms. detection of missile exhausts. – the far UV. the terrestrial atmosphere does not allow the free propagation of light with wavelengths below 200 nm: as a result. Due to its own radiation absorption properties. for wavelengths between 280 and 200 nm (from 4. – UVB. mostly consist of: – biological and chemical sensors: detection of ozone. flame control.

.Ultraviolet Photodetectors 183 – the calibration of emitters and UV imaging: instrumentation. Inset: UVA and UVB solar irradiance. It is therefore not surprising that most of the applications of UV detectors are connected with the measurement of solar radiation. . Figure 6. Solar radiation spectrum. a large bandgap energy (in the case of semiconductors). measurement of solar UV. astronomical research. such as UV imaging. The sun represents the most important natural source of UV radiation. In general.2. A high response speed can be particularly useful in applications requiring rapid real-time processing of the signal. . . The level of importance and the priority ascribed to these different requirements clearly depends on the type of application intended. a good linearity of the photocurrent as a function of the incident optical power. photodetectors suited to these areas of application must simultaneously possess a high response coefficient. a low level of noise and a high degree of spectral selectivity. outside the Earth’s atmosphere and at ground level .2. . The solar radiation spectrum is shown in Figure 6.

3. assuming that the UVC radiation is blocked by atmospheric absorption for cutoff wavelengths typically less than or equal to 280 nm. for example. The stratospheric ozone layer completely absorbs the UVC part of the solar spectrum. and they are classed as visible blind detectors for cutoff wavelengths between 400 and 280 nm. it is helpful if the UV photodetectors are not sensitive to visible and infrared radiation. UV light represents 9% of the total luminous power emitted by the Sun. Damage to human DNA Figure 6. for UV wavelengths below 280 nm which clearly cannot be detected at ground level. and have potentially very promising growth possibilities for applications in space to optical inter-satellite telecommunications. UVA and B radiation. . on the other hand. forest fire detection systems.184 Optoelectronic Sensors For the vast majority of applications. since they reach the ground (see Figure 6. or solar blind. They basically operate as high-pass filters. missile track detection systems. and only allows UVA and UVB to be transmitted to the ground after attenuation (see the inset in Figure 6. are mostly responsible for the biological effects of solar radiation. Spectral dependence of the harmful effects of UVA and UVB radiation on human health 1. Gas flames. The visible blind detectors are nowadays quite widely used for flame detection1.3). Outside the Earth’s atmosphere. etc. have an emission peak in the UV around a wavelength of 320 nm. It is generally considered that solar radiation of wavelengths below 280 nm does not reach the surface of the Earth. The applications of flame detection mostly cover the following areas: fire detector heads.2).

and it has also been shown more recently that UVA radiation. is able to penetrate the skin and has enough ionizing power to cause deep damage to living cells2. Housewives have also known since time immemorial that the drying of washing in the sun in fair weather makes it healthier. Among others. their bactericidal action makes them useful for many types of biological cleaning and sterilization procedures3. the ozone layer present in the upper atmosphere is naturally regenerated through the conversion of oxygen in the air into ozone. also produce a carcinogenic effect comparable to that of UVB. but the effect is very widespread and nowadays has a global effect.and bromo-carbons is largely responsible for this. particularly in techniques for the preservation of perishable foodstuffs. but also allows it to play an extremely useful role in a number of different fields. This has recently driven a significant 2. It has been found that UVA radiation can in particular alter the genetic structure of cells and. This issue is particularly interesting since it has public health implications because of the popularity of tanning salons with the general public. the rapid depletion of the stratospheric ozone layer. being particularly energetic. One of the most significant “holes” in the ozone layer is currently situated above the continent of Antarctica. 4. the sterilization of surgical instruments. although considered less dangerous. is the direct cause of sunburn and skin cancers. and the resultant increase in exposure of the Earth’s surface to UV radiation has raised considerable concern among scientists and professionals interested in health and the environment. etc. etc. On the other hand. The emission of aerosols and gases containing fluoro-. which means that a moderate amount of exposure to sunlight is important for everybody. The illumination of objects with high intensity UVB and C radiation can be used for this. Large-scale natural phenomena such as volcanic eruptions also contribute in a manner which can be very significant. Also. thus. both organic and inorganic. particularly in the case of prolonged or repeated exposure. . 3. a reaction which is photo-assisted by UV radiation. and the ionizing properties of these radiations make them useful for the activation of numerous chemical reactions. UV radiations are involved in the synthesis of Vitamin D in the human body. chloro. This represents a clear threat to human heath from UV radiation.Ultraviolet Photodetectors 185 UVB radiation. caused in particular by human activity 4 . Finally.

phototransistors or avalanche photodiodes. The use of semiconductors for UV photodetection historically began with the use of narrow bandgap materials such as silicon and certain III-V materials (GaP. etc. Personal solar dosimeters able to detect direct or diffuse UV light are also currently available to the general public for UVB dosimetry.). they generally involve detection systems which are cumbersome. expensive and fragile. There are a wide range of different UV photodetectors suitable for all the different applications listed above. Firstly. whose active components can be photoconductors. GaAsP. Traditionally. etc. However. p-i-n photodiodes. particularly in laboratories. The increasing need for miniaturization and reliability of these systems for the development of portable or remote applications has naturally led to the development of UV photodetectors based on semiconductors. the direct exposure of such .) whose bandgaps are mostly situated in the near-infrared (Si) or the red (GaAsP). visible blind detectors have recently found a growing number of applications in UVA and B monitoring systems for the environmental study of ground-level solar illumination and UV dosimetry in fields which are mostly connected with biology. photomultiplier tubes have always been used for UV detection and their use is still widespread. Thus. as in the specific case of GaP. The main drawback of silicon. MSM photodiodes.186 Optoelectronic Sensors increase in efforts to develop many specialized optoelectronic devices which in particular allow the detection and measurement of solar UV illumination. is that it is impossible to use them directly for UV photodetection. and small bandgap semiconductors in general. although they offer a sensitivity which is difficult to match with the other types of photodetectors applicable to these ranges of wavelength. medical science or research into cosmetic products (the development of sun creams or other products which modify the photosensitivity of human skin to UV radiation. In exceptional cases they may be in the yellow. The spectral response curve of these sensors reproduces as closely as possible (as we will see later) the erythema (sunburn) action spectrum on human skin. which represents their upper limit. Schottky junction photodiodes.

Silicon carbide. resulting in a certain degree of performance loss. so that it is simply impossible to extract information pertaining to the UV spectral component from this white light. such as those that we are interested in for UV light. or phonons in the crystal lattice). has allowed sensitive detection of the spectral range in question. a new generation of UV photodetectors appeared. so that the quantum efficiency of the photodetector decreases in the short wavelength part of the optical spectrum. A number of years ago. Also. The only issue is that even here the use of high-pass optical filters is standard in order to tune the photodetection system to the desired spectral window. operating in the UVC range. the quantum efficiency of a semiconductor photodetector (defined as the probability of electronhole creation by the absorption of an incident photon) is greatest for energies above the bandgap of the material. whose behavior is also increasingly liable to drift in the long term. For energies much higher than the bandgap of materials. whose bandgap at ambient temperature (T = 300 K) is 2. . due to aging of the filters. In particular. The use of semiconductor materials with a small bandgap energy for UV photodetection therefore rapidly drove the development of filtering devices.86 eV (6H-SiC). part of the energy of the incident light is effectively lost through heating of the semiconductor (excitation of vibration modes. diamond or gallium nitride (GaN) and AlGaN alloys. meaning that this only offers a partial solution to the problem.Ultraviolet Photodetectors 187 sensors to daylight clearly results in them being flooded with visible wavelengths and sometimes even infrared. It still needs to be joined on to detection systems that are relatively complex and also expensive. but limited to a spectral range relatively close to that value. made from semiconductor materials with large bandgaps such as silicon carbide (SiC). which normally use layers of selectively-absorbing material. it has recently given rise to industrial applications such as the construction of flame detectors that are visible-blind. and sometimes a phosphor-based highpass filter which absorbs the UV light and re-emits towards the semiconductor photodetector a light whose spectral maximum is not far from the bandgap of the semiconductor material being used [GOL 99].

188

Optoelectronic Sensors

Thanks to the development of epitaxial crystal growth techniques, the recent use of gallium nitride (GaN), and still more recently aluminum gallium nitride (AlxGa1-xN, with x being between 0 and 1), has helped reduce the problems listed earlier, enabling the fabrication of high performance and robust visible blind photodetectors. The direct bandgap of AlxGa1–xN materials, whose size is an increasing function of the concentration of aluminum, is found at ambient temperature to extend from 3.42 eV ( = 362 nm) (x = 0) to 62 eV ( = 200 nm) (x = 1). Thus, the spectral range that is covered includes in particular the cutoff values for UVB (3.87 eV or = 320 nm) and UVC (4.43 eV or = 280 nm). It follows that it becomes possible to fabricate UV photodetectors using these materials which are free from any need for intermediate spectral filtering, opening up the possibility of simplifying photodetection systems through miniaturization, accompanied by an improved reliability of the devices. In addition, the high chemical stability of GaN and AlxGa1–xN compounds, comparable to that of SiC, means that these materials are also suitable for use in the fabrication of photodetectors which are naturally resistant to extreme operating conditions (high temperatures and high values of inverse bias voltage in particular). Due to their direct bandgap at all levels of aluminum, the optoelectronic properties of AlxGa1–xN alloys are also clearly superior to those of SiC, whose indirect bandgap does not so strongly favor the effects associated with photo-generation of carriers. For all these reasons, group III nitrides therefore offer a technical solution for the future, as we will see throughout the rest of this chapter, a solution which is at the same time flexible, robust and very well-suited for all visible-blind or solar-blind photodetection applications, because they allow relatively low-cost fabrication of effective and reliable UV photodetectors suitable for all the application windows listed above. In this chapter, we will first introduce the UV-visible contrast and UV detectors based on silicon and SiC. We will then describe in detail several types of UV detectors based on nitrides: photoconductors, Schottky photodiodes, MSM photodiodes, and p-i-n photodiodes. The most sophisticated technological advances involved with the recent

Ultraviolet Photodetectors

189

construction of avalanche photodiodes and phototransistors based on nitrides are also briefly described5. 6.2. The UV-visible contrast The absorption of light is the fundamental effect governing the dependence of the spectral response of a photodetector as a function of the energy of the incident photons. In direct-bandgap semiconductors, we therefore observe a dramatic transition in the value of the spectral response close to the bandgap of the semiconductor material, dropping all at once to a very low value when we drop below the bandgap. In the case of UV photodetectors, assigning an arbitrary limit to the wavelength of visible light of 400 nm, the UV-visible contrast is defined by the ratio Rimax ( = G)/Ri ( = 400 nm), where Rimax is the maximum value of the response coefficient at the wavelength G of the bandgap and Ri ( = 400 nm) represents the value of the response coefficient at the limit of visible light. The UV-visible contrast is mostly limited by the two following factors: – the presence in the bandgap of the semiconductor of deep energy states which are electrically and optically active. These originate in particular from extended or localized defects in the crystal structure of the semiconductor, but can also be attributed to the presence of deep energy levels due to impurities, or deep levels caused by imperfections in the surface of the semiconductor; – in the case of Schottky junction photodetectors with semitransparent contacts, the effects of photo-emission of carriers by the metal.

5. The work carried out by the authors on UV photodetectors, described in this chapter, was carried out with the support of the European Community, under the Environment and Climate grant number #ENV4-CT97-0539.

190

Optoelectronic Sensors

6.3. Si and SiC photodetectors for UV photodetection Contemporary semiconductor-based commercial technical solutions proposed for UV photodetection mostly rely on Si and SiCbased technologies. Other applications, a little less widespread and well-known, use other semiconductor compounds such as GaAs, GaP and GaAsP. These are mainly used for Schottky barrier photodiodes. This particular family of devices is in particular very well-suited for the detection of VUV radiation, for which they offer a remarkable stability of operation. It is also worth mentioning the recent emergence of UV photodetectors based on diamond, which allows the construction of photoconductors with a short cutoff wavelength (225 nm) with a UV/visible contrast of 106 [WHI 96]. Although diamondbased devices are potentially very interesting and well-suited for UV photodetection applications and the detection of high-energy particles, due in particular to diamond’s very high resistance to the effects of irradiation, the development of this last type of photodetector has nevertheless been rather restricted by the current technical challenges of achieving epitaxial films with adequate homogenity and monocrystallinity: the best material obtained to date is a crystal mosaic with grain boundaries which have a fairly strong mismatch between one another. Clearly, this results in a high density of electrically and optically active defects which reduces the performance of the photoconductors, both in terms of response coefficient and UV/visible contrast and in terms of response time. Nevertheless, it is important to emphasize that UV photodetectors based on thin epitaxial films of diamond have recently become commercially available, which is the sign of a definite and growing interest in these devices for UV photodetection and its applications [MAI 00]. Next we will give, by way of illustration, a brief description of the main families of Si and SiC-based photodetectors.

Ultraviolet Photodetectors

191

6.3.1. UV photodiodes based on silicon Silicon p-n junction photodiodes are currently by far the most economical and widespread commercial solution for UV photodetection. Silicon gives rise to low-cost detectors which are suitable for the vast majority of UV detection applications across a very wide spectral range extending from near-UV to VUV, and even extends to the high-energy system of soft X-rays. However, the general shape of their spectral response is not uniform, and in particular it reaches into the infrared, silicon’s bandgap being 1.1 eV at ambient temperature (300 K). As a result, their use in selective UV photodetection applications requires, generally speaking, the use of absorbing filters between the light source and the photodetector, in order to select only the spectral window of direct interest. We note in passing that this often significantly increases the cost of the final detection system, the filters normally requiring complex technologies which make them very expensive. UV photodiodes based on silicon can be subdivided into two large families: p-n junction photodiodes and metal-oxide-semiconductor photodiodes with a charge inversion layer [GOL 99, RAZ 96]. The first family of Si UV photodetectors comprises photodiodes whose p-n junction is situated at a shallow depth below the surface (typically of the order of 0.2 μm), and which are covered with a thin layer of SiO2. This insulating layer plays the double role of passivation of the surface of the semiconductor and of antireflection coating. The depth of the junction is an extremely important parameter, because absorption of light is more superficial the more energy the photon has. In the case of p-n junction photodiodes, it is desirable, in order to optimize its behavior and obtain the maximum photocurrent, that as many UV-photogenerated carriers as possible reach the junction region without recombining. For the same reason, control of surface recombination effects is of fundamental importance for silicon UV photodetection. Detailed control and optimization of the Si/SiO2 interface properties makes it possible to reduce the density of surface traps, while naturally producing an electric field at the surface of the

192

Optoelectronic Sensors

silicon which limits, or even eliminates, recombination effects at the interface. The p-n junction UV photodiode with record performance was initially developed by Korde et al. [CAN 89, KOR 87]. In order to form the junction, these authors used phosphor (P) diffusion onto an Si (p) substrate free from dislocations. A SiO2 layer of thickness 60 nm was built on the surface of the device. Due to its thickness, this layer absorbs all radiation of wavelengths below 120 nm, which unfortunately makes this type of photodiode unusable in applications involving the detection of high energy UV (VUV in particular). Nevertheless, this type of photodiode is characterized in particular by an internal quantum efficiency of 100% in the spectral range of 350600 nm. The internal quantum efficiency is greater than 1 in the shortwavelength UV region (<350 nm), due to secondary impact ionization effects: a high kinetic energy is imparted to the electron-hole pair photogenerated by high-energy photons, and this is sufficient to create other secondary electron-hole pairs through impacts with atoms in the crystal lattice. On the other hand, the efficiency becomes less than 1 beyond a wavelength of 600 nm, because this long-wavelength light is absorbed in the deeper layers of the material which are far from the pn junction, creating electron-hole pairs which are unable to reach the junction since their diffusion lengths are too short. As a result, they recombine in the volume of the semiconductor and do not contribute to the photocurrent. It is interesting to dwell for a moment on UV photodiodes with a pn junction designed for the detection of VUV up to an energy of 124 eV. Although the base structure of the p-n junction remains the same as that which we have already described, conversely the oxide layer present on the surface is much thinner than in other types of UV Si photodiodes. Its characteristic thickness is in this case no larger than 4-5 nm, making it a factor of 10 or 20 times smaller than that normally found in more conventional applications. Transparent to short wavelengths of light, its useful volume is on the other hand reduced, in such a way that the possibilities for UV light and humidity to create high densities of recombinant traps in the oxide (which is

Ultraviolet Photodetectors 193 normally one of the main sources of aging in Si UV photodiodes) is very significantly reduced compared to the normal situation. These photodiodes are commonly equipped with a bandpass filter consisting of thin metallic layers. which tended to be found in the first generations of p-n junction UV Si photodiodes. Ag. in the case of UV photodetection. the internal quantum efficiency of a Si VUV photodiode is typically of the order of 30. As a result. which consists of electrons in the case we will describe. This very localized accumulation of charge results in a high electric field being established in the immediate vicinity of the semiconductor surface which. which is fundamentally linked to the absorption of the UV light by the oxide layer above this wavelength. this configuration is ideal for maximizing the quantum efficiency of the photodiodes. picking out the 10-50 nm band (Al/C.) [RAZ 96]. To give a better idea of the orders of magnitude. It follows that the stability of operation of these devices is very good when they are exposed to very high energy photons. In the system of high of Si VUV photon energies. Photodiodes with a charge inversion layer [GOL 99.63 [RAZ 96]. Ti. which multiply through secondary impact ionization. The process of photogeneration produces carriers with very high kinetic energies. etc. Sn. Si UV photodiodes with an inversion layer offer a high quantum efficiency in the spectral band from 250-500 nm and a cutoff wavelength of 120 nm. the internal quantum efficiency photodiodes typically follows a linear variation as a function of the energy E of the photon. Through the effects of the energy band curvature in the semiconductor. offers the particular advantage that this is the exact region where the main absorption of the high-energy incident light takes place. a thin deposited layer of silica on the surface of the Si (p) material leads to the creation of a charge bilayer with the opposite sign to the doping of the volume. It . for a photon energy of 124 eV. Al/C/Sc. something which is also helped by the effectively complete absence of a “dead region” of recombination at the surface. which can be written: = E/3. RAZ 96] are similar in their architecture to metal-oxide-semiconductor structures used to build field effect transistors: they make use of effects associated with the presence of the two-dimensional charge induced through inversion at the Si/SiO2 interface.

2 and 0. In addition. the range of linearity of these photodetectors remains reduced compared to the possibilities offered by p-n junction UV Si photodiodes. the diffusion length for electrons is significantly larger than that for holes. be noted that the behavior of this type of photodiode tends to change over time under UV illumination.175 A/W at a wavelength of 270 nm. The doping levels are 5–10 1018 cm–3 for SiC (n+) and 5–8 1017 cm-3 for SiC (p). the response coefficient of these same structures is typically 0. an effect which seems to be mostly explained by the effects of the high electrical resistance presented by the charge inversion bilayer at the SiO2/Si interface. operating in the 200–400 nm spectral range [RAZ 96]. The spectral response of these photodiodes contains a peak lying between = 268 and = 299 nm. The optimized structure consists of a layer of SiC (n+) with a thickness between 0. Leakage currents below 1 pA are currently measured for a reverse-bias of –10 V. The structure is topped with a SiO2 passivation layer on the surface.3 μm. The best performances have traditionally been obtained for Schottky devices made from SiC (p). An ohmic contact based on nickel is present on the n+ side. sitting on a SiC (p) substrate. SiC-based UV photodetectors The best UV photodetectors based on 6H-SiC are p-n junction photodiodes [RAZ 96]. Very low leakage currents can be obtained for Schottky diodes made using SiC (p). Anikin et al.194 Optoelectronic Sensors should. since UV light has a tendency to alter the surface oxide layer. The dark current of these devices is very weak. A dark current value of 10–11 A for a reverse-bias of –1 V has been measured for SiC photodiodes. at a temperature of 473 K [EDM 94].05 A/W. [ANI .15 to 0. Additionally. The height of the Schottky barrier is about twice as large as that normally achieved for n-type SiC. At = 200 nm. however. Schottky junction 6H-SiC photodiodes are also currently being developed.3. The response coefficient of p-n SiC photodiodes made in this way lies in the range from 0. which corresponds to an internal quantum efficiency of 70 to 85%. However.2. 6.

which reaches 150 mA/W at a wavelength of 215 nm [RAZ 96]. and a high value of the response coefficient. III-V nitrides represent one of the most interesting and flexible technical solutions at present. which can reach or even surpass 300°C without seeing a significant degradation of the devices. sapphire substrates are currently very widely employed for the growth of thin layers of III-V nitrides and the resultant applications. of the order of 100 pA for a reverse-bias voltage of –100 to –170 V. which for a long time made the development of applications impossible. As we have already emphasized. This is why only a very small deterioration in their operational characteristics (spectral response and I(V) characteristics) is seen when these photodetectors are used at high temperatures. for which they are the substrate of choice.Ultraviolet Photodetectors 195 92] report the fairly recent construction of a high quality Schottky photodiode in SiC (n). mainly due to the poor suitability of the techniques used and the absence of a substrate with a suitably matching lattice parameter.4. The development of epitaxial techniques for this material has for its part been held back for several decades.63 eV. despite a high density of dislocations (of . one of the interests in SiC-based devices is their remarkable resilience under extreme operating conditions. Organometallic vapor phase epitaxy (OMVPE) and. The behavior of these photodetectors is characterized by a very low leakage current. UV detectors based on III-V nitrides Among the various families of materials suitable for UV photodetection. molecular jet epitaxy with a plasma or gas source (MJE) currently allow the growth of thin layers of GaN and related compounds which. more recently introduced for nitride growth.4 and 1. which gives a Schottky barrier height in SiC which is typically between 1. In this the Schottky junction is made through deposition of gold. Despite a strong lattice mismatch (it is 16% that of GaN) and despite a strong difference in thermal expansion coefficient compared to GaN. 6. The first work in characterizing the physical properties of polycrystalline gallium nitride (GaN) dates back more than 30 years.

This behavior is independent of the wavelength of the light. Spectral response Photoconductors based on AlGaN typically consist of an epitaxial layer of AlGaN on sapphire.95. with 0. where P is the incident optical power [MON 99a].4. Diagram of a photoconductor based on AlxGa1–xN(Si) . Photoconductors based on GaN and AlGaN have a gain under continuous illumination which can reach 105–106 and which strongly depends on the incident optical power. optical and electrical properties on sapphire which are good enough to make them suitable for many applications. blue lasers. The current induced in the device can be easily deduced from measurement of the voltage drop across the load resistance when the photoconductor is illuminated. doped with silicon.4. Its variation is a function which decreases in proportion to P– over a range of five decades.4. exhibit structural. The electrical bias circuit includes a DC voltage source connected to the photoconductor.1. field effect transistors and UV photodetectors.196 Optoelectronic Sensors the order of 109 to 1010 cm–2 in GaN). with a thickness of 1 μm.1. The value of is a decreasing function of the electrical resistance of the layer. and decreases with temperature. and even into the near-UV). and the device is in series with a low-value load resistance.1.5 < < 0.4). Figure 6. Photoconductors 6. mostly involving light-emitting diodes at short emission wavelengths (from green to blue. onto which two ohmic contacts are attached (see Figure 6. 6.

the lifetime of the free carriers and hc/ the photon energy. S. In addition. grain boundaries and interface). The mechanism which explains the variation of the spectral response of the photoconductor must therefore involve more than just the optical absorption of the semiconductor.2] where is the quantum efficiency. it follows that: . The UV/visible contrast is significantly smaller than the value obtained from a simple analysis based on the absorption coefficient of these materials.35.319 nm for x = 0. and the other due to the modification by the light of the effective conducting crosssection. is not the same as the geometric cross-section of the device. L is the distance between the contacts. VB is the bias voltage and n is the concentration of free carriers. Given that n can be expressed in the form [RAZ 96]: n g Popt hc [6.22 and 368 nm for x = 0.Ultraviolet Photodetectors 197 The spectral response of photoconductors based on AlxGa1-xN exhibits a cutoff at 298 nm for x = 0. The absorption of light produces a reduction in these depletion regions. g the photoconduction gain. S is the conductive cross-section. a sub-linear dependence of the photocurrent is observed as a function of the power of the incident light. one due to the photogenerated free carriers. [GAR 98]. n.1] where q is the elementary charge of the electron. the photoconductor’s response coefficient Ri is the combination of two terms. e is the electron mobility.1]. leading to a modification of the conduction crosssection.2] into [6. The transverse conduction cross-section in the model of Garrido et al. The response coefficient can be written: Ri I Popt qVB μe S n LPopt n S [6. due to the presence of space charge regions (SCRs) [HAN 98] around lattice discontinuities (dislocations that reach the surface. by substituting [6. In the model developed by Garrido et al.

which reduces the effective conduction cross-section of the device. However.2. they create a depletion region around them. in other words.3] The spectral variation of the first term in this equation is given by .3]. If these defects are electrically charged. so that their presence has little effect on the absorption coefficient. since the electrical charge concentrated at these lattice discontinuities is modified by the light. ( ). Since the quantum efficiency is a direct function of the absorption. However. The slow drop-off in the photocurrent can be explained by a model based on the modulation of the transverse conduction cross-section [MON 99a]. Response time AlGaN-based photoconductors display significant persistent photoconductivity (PPC). and hence on the first term of equation [6. This means they have a very slow transitory temporal response (several thousandths of a second) and a non-exponential tail. or in extended defects associated with lattice discontinuities. This mechanism can also explain the high value of the spectral response for photon energies above the bandgap energy: the energy levels responsible for the absorption of light in the visible can originate either from point defects distributed homogenously within the semiconductor.198 Optoelectronic Sensors Ri I Popt qVB L e g S hc n S Popt [6. experiments show that the dominant mechanism in these photodetectors is represented by the second term of equation [6. The curvature of the bands around defects and dislocations leads to a spatial separation of electrons and holes. which alters the effective transverse conduction cross-section. which is consistent with the frequency dependence of their spectral response [SHE 99].4. which .1. effects. it depends strongly on the modification to the effective transverse conduction cross-section. this first term leads to a significant UV/visible contrast. either dopants or vacancies. Complete absorption of light through these defects can be treated as negligible. 6.3]. their effect on the photoconduction spectral response is large.

the electrons are inclined to recombine.Ultraviolet Photodetectors 199 can be the cause of the PPC. it is difficult to discriminate between these two regions. S and 0 are the relative dielectric constant of the semiconductor and the vacuum permittivity respectively. Including the thermionic emission current Jth. Through this mechanism. When the light is extinguished. QSS originates from the current crossing the potential barriers around the defects. as well as the tunneling current Jtunnel. It follows that the response coefficient of the photoconductor is given by: Ri AoptVB Popt L 2 e QSS [6. In the dark. the SCR returns slowly to its original extent in the dark. xSCR. The height of this barrier depends on the charge present at the defect. However. but they must first cross the potential barrier separating them from their recombination sites. the temporal response of photoconductors can be calculated in the following manner. QSS can be expressed in the form: QSS qN D xSCR 2N D S 0 0 il [6. and 0 and il are the heights of the barrier around the defect.5] where Aopt is the optical surface area of the device. towards its dark value. so that it will change over time as recombination occurs. The dominant defect type may be that associated with dislocations at the AlGaN-air interface or the AlGaN-substrate interface [SHE 99]. connected to the variation in the width of the SCR. it follows that: .4] where ND is the level of n-type doping. Quantitatively. in the dark and under illumination respectively. if QSS is the increase in charge at the defects. in a non-exponential manner. and as a result the resistance of the device also slowly returns. and their relative significance will depend on the crystalline quality of the material.

.4] to [6.9].4]–[6. Figure 6. along with numerical simulation using equations [6.200 Optoelectronic Sensors dQSS dt J th J tunnel [6.5 shows the temporal response of AlGaN photoconductors.(ND /m*e) and E0 = coth(qE00 /kT). and calculated by numerical solutions of equations [6. Decrease in the normalized photocurrent in Al0. . Figure 6.5.23Ga0.7] J tunnel A*T 2 E00 0 V ph qE00 kT cosh q kT exp q 0 V ph E0 1 exp qV ph kT [6.6] The thermionic and tunneling currents can be expressed respectively in the following forms: J th A * T 2 exp q kT 0 exp qV ph kT 1 [6. .8] with E00 = (h/4 ).77N photoconductors. . both measured. . which can be expressed as: V ph t Q 2 SS t 0 2q S 0ND [6.9] To illustrate this.10] . The drop in voltage across the barrier is Vph(t). Vph is the voltage induced by the light.

As a result.4. measured at different modulation frequencies of the incident optical signal. The mechanism at the root of the response of III-V nitride-based photoconductors therefore makes them of little interest for applications which requires speed or a high UV/visible contrast.3. the SCRs associated with the defects do not have time to respond. the UV/visible contrast is much better at higher frequencies. The spectral response of a Schottky photodiode made from the same batch is also shown for comparison 6.1.6. When the frequency increases. Normalized spectral response of a photoconductor. . Effects of frequency modulation of the incident optical signal The effect of frequency on the spectral response of photoconductors is shown in Figure 6.6.Ultraviolet Photodetectors 201 Figure 6. The result is that the cutoff slope becomes sharper and the spectrum of the response coefficient tends to approach that of a photovoltaic detector.

due to the high values of the ideality factor.84 eV and 1. In GaN-based devices.7.4. Electrical properties Work published on Schottky barrier photodiodes involves planar or vertical structures (Figure 6. The height of the barrier as a function of the molar fraction of aluminum for Schottky diodes with nickel and gold contacts is shown in Figure 6.7). and the ideality factor also increases and can reach values close to 4.02 eV respectively for semitransparent Ni and Au contacts on GaN. particularly in terms of the bandwidth and the noise level. Schematic diagram of AlGaN Schottky photodiodes in (a) vertical and (b) planar configurations Figure 6. Schottky barrier photodiodes based on AlGaN 6.9.2. Consequently. which in theory both give extremely similar operating characteristics. The variation of the capacitance as a function of the bias voltage leads to barrier heights of 0.4. the current limitations in III-V nitride technologies (connected in particular with damage to the material during the engraving of the mesa) result in a performance degradation. with a series resistance of 20-50 and a leakage resistance greater than 1 G .2. The leakage current increases as a function of the fraction of aluminum.2. Although the vertical structure is particularly interesting for the construction of fast devices with a high response coefficient.1.8 shows the typical current-voltage (I-V) characteristics of planar Schottky photodiodes. Figure 6. the measured I-V characteristic of AlGaN-based diodes does not make it possible to deduce precise and reliable information about the height of the Schottky barrier. the ideality factor is of the order of 1.202 Optoelectronic Sensors 6. .

Spectral response In contrast to photoconductors. the penetration depth of the light is larger than the SCR.2.8. the response coefficient decreases at the shortest wavelengths.Ultraviolet Photodetectors 203 Figure 6. and so the response coefficient depends on the diffusion length of the minority charge carriers. for values situated both above and below the bandgap energy of the semiconductor. Firstly.2. Typical current-voltage characteristics for Schottky diodes made of AlxGa1–xN on sapphire 6. the photocurrent of Schottky barrier devices depends linearly on the incident light power (inset in Figure 6. The decrease in the response coefficient observed when the fraction x of aluminum increases results from two effects. at a constant quantum efficiency. which decreases with the concentration of aluminum.10). Also. .4. the nature of the metal used to fabricate the Schottky diode and the excitation wavelength. This linearity is independent of the size of the diode.

particularly p-n junction photodetectors. Variation of the height of the Schottky barrier as a function of the molar fraction of Al Figure 6. which results in a very thin SCR and a material which has a very low mobility.9. in contrast to the results obtained for photoconductors.10 shows that the UV/visible contrast is more than three orders of magnitude. In junction devices based on GaN.204 Optoelectronic Sensors Figure 6. From this point of view. . Since the SCR lies just below the surface of the semiconductor. The spectral response is flat for wavelengths smaller than those of the bandgap. there is no decrease in the quantum efficiency at short wavelengths as is normally seen in p-n and p-i-n photodiodes. The steepness of the cutoff slope of Schottky photodetectors shows that the only limitation in the UV/visible contrast is that caused by absorption by deep defects. where the photogenerated carriers must diffuse across the layer above before reaching the SCR at the level of the junction. the choice of Schottky photodiodes based on AlxGa1-xN has more potential than p-n or p-i-n devices for constructing wideband UV detectors. this problem is accentuated by the high density of dopants required to achieve p-type conductivity. which is an advantage that Schottky photodiodes have over other types of photodetectors. The cutoff wavelength develops from 362 nm to 293 nm.

the electronhole pairs must therefore diffuse towards the SCR at the surface. and by the diffusion length Lh of the minority carriers in this material. and hence an increase in the diffusion length and the response coefficient.10. Normalized spectral response at ambient temperature of Schottky photodiodes for different values of the molar fraction of Al. it also results in a reduction of the dark current. . In order to be collected. and is an increasing function of the quality of the material. and a low value of Lh results in a decrease in the response coefficient. a certain proportion of the photons may cross the SCR without absorption and penetrate deep into the volume of the material. For radiation with an energy close to the bandgap of the semiconductor.Ultraviolet Photodetectors 205 Figure 6. Inset:: variation of the photocurrent as a function of the incident light power The maximum value of the response coefficient obtained in AlGaN Schottky photodiodes is limited by the optical transmission of the semi-transparent Schottky contact. It follows that the use of thick AlGaN layers (thickness greater than 1 μm) is preferred because they allow a reduction in the density of dislocations. Lh is a function of the mobility and the lifetime of the minority charge carriers.

The maximum bandwidth of these devices is therefore proportional to ND1/2. 0 is the height of the Schottky barrier seen from the semiconductor side. Noise and detectivity The 1/f noise dominates at low frequencies. Its spectral noise power density. Since the mobility increases as the doping level falls.206 Optoelectronic Sensors 6. satisfies the relationship: Sn s0 I dark f [6. for which the response time is minimized. Sn. and s0. 6.2. and are dimensionless parameters whose values are close to 1 and 2 respectively. Since the series resistance does not depend on the bias voltage.11] where Idark is the dark current. f is the frequency. Response times around 50 nanoseconds are typical for diodes of diameters between 200 and 300 μm. the response time of the device decreases in proportion to ( 0+V)–1/2. The capacitance of the depletion region is given by the expression: CSCR A qN D 2 0 S 0 V [6. As well as the 1/f noise. there exists an optimum doping level. because of the difficulty in achieving good quality ohmic contacts. RS and CSCR are proportional to ( ND)–1 and ND1/2 respectively.4.3.4. and V is the bias voltage. where C is the sum of the internal capacitance of the diode and the load capacitance.2.10] where A is the area of the Schottky contact. 0 is the vacuum permittivity. and R is the sum of the load resistance and the series resistance of the device.4. ND is the density of dopants. Response time The response time of these photodetectors is mainly limited by the RC time constant of the device. of the order of 1018 cm–3. For planar devices. where μ is the electronic mobility. s is the dielectric constant of the semiconductor. This effect is accentuated for AlGaN alloys with high concentrations of aluminum. the contribution of the shot noise .

The noise equivalent power (NEP) can therefore be written in the form: NEP 2 ishot 2 i1 f Ri [6. whose value lies below 1 nA/cm2 for a reverse-bias voltage of –1 V.2 109 W–1Hz1/2 cm respectively. i1/f is the 1/f noise current and Ri is the response coefficient of the device. the weaker residual doping in these epitaxial layers makes it possible to obtain bandwidths greater than 30 MHz. 12 MHz and 8 MHz for . which limits the UV/visible contrast in GaN-based photodetectors.5.22Ga0.78N/Au photodiodes respectively. Schottky barrier photodiodes made with epitaxial GaN layers grown using the ELOG technique offer an order of magnitude increase in the UV/visible contrast compared to GaN-based devices grown on sapphire using standard methods. 6.4. As the response time of the device is limited by the time constant RC. The value of <ishot 2 > is given by: 2 ishot 2qI dark f [6. The dark current. at a reverse-bias voltage of –2 V. The corresponding detectivity values are 6. and Idark represents the dark current. USU 97].Ultraviolet Photodetectors 207 must be included.2. NEPs of 8 pW/Hz1/2 and 41 pW/Hz1/2 have been obtained for Schottky GaN/Au and Al0. is significantly smaller than that of Schottky photodiodes based on GaN/sapphire. The recent development of the technique of epitaxial lateral overgrowth (ELOG) has reduced this density of dislocations by at least two orders of magnitude [NAM 97.12] where ishot is the shot noise current.1 109 W–1H–1/2 cm and 1. Schottky barrier photodiodes made from GaN grown using epitaxial lateral overgrowth (ELOG) Despite these promising characteristics.13] where f is the bandwidth of the photodiode. heteroepitaxial growth of GaN results in a high density of dislocations (~ 108–109 cm–3).

Application of AlGaN Schottky barrier photodetectors for the simulation of the biological effects of UV light UV radiation produces a wide range of biological effects [DEG 94. < 317 nm. = 249 – 315 nm. MCK 87]: – pigmentation: maximum effect at – synthesis of vitamins D2 and D3: yield at 290 nm. A detectivity as high as 5 1011 W–1Hz1/2 cm has been measured for GaN ELOG photodiodes with a diameter of 400 μm. < 297 nm. [MUN 00] showed that the spectral response of AlGaN Schottky photodiodes can match the UVA and B erythema action spectrum. Effect grows rapidly as the Simple. since they require neither filters nor temperature stabilization. with a bias voltage of –3.4 V.208 Optoelectronic Sensors devices with diameters equal to 200 μm. . reliable and low-cost devices are hence required to evaluate the biological effects of UV radiation.4. In this way. – DNA damage: wavelength decreases. 400 μm and 600 μm respectively. AlGaN photodetectors therefore offer a flexible solution with a low cost and minimum footprint. Maximum – erythema (sunburn): maximum sensitivity at – carcinogenic effects: UVB and UVC. = 210–310 nm. 6. – damage to plants: – bactericidal action: 254 nm. the response coefficients of these devices can give direct information on the biological effects of UV light.2.6. accurate. GaP) combined with a series of filters inserted into the path of the incident light. GaAs. UV detectors with a wide spectral band have been developed to study the effects of UVA and UVB on the skin [MCK 87]. Muñoz et al. < 320 nm. Maximum effects at = 360–400 nm. with the appropriate selection of an appropriate molar fraction and carefully selected growth conditions (Figure 6. Commercial UV dosimeters use photodiodes with a small bandgap (Si.11). Maximum at 310 nm. For this application.

with a steep cutoff slope which shifts towards shorter wavelengths as the proportion of aluminum increases. This accentuation of the contrast with the bias voltage can be explained by the appearance of a gain. The main advantage of these structures. there . Their dark current is very weak and follows the thermionic emission model with a value of the Richardson constant equal to 26 Acm-2K-2 and a barrier height of 1. At present.3. of micron-scale thickness. based on AlGaN. The photocurrent varies in a quasilinear manner with the incident optical power. is their ease of fabrication.4. MSM photodiodes MSM photodiodes consist of two interdigitated Schottky contacts deposited onto a planar surface (see Chapter 5). The typical structure of such an MSM diode involves two interdigitated Ni-Au Schottky contacts deposited on an epitaxial layer of AlGaN.04 eV for GaN.11. Normalized spectral response of an AlGaN Schottky photodiode compared with the sunburn action spectrum (international standard set by the International Commission on Illumination (CIE)) 6. The UV-visible contrast varies from around three orders of magnitude for the weakest bias voltages (typically less than 5V) to four or five orders for the highest bias voltages. Their spectral response is relatively flat above the bandgap.Ultraviolet Photodetectors 209 Figure 6. not intentionally doped.

The behavior of the device is thus doubly penalized: firstly. p-i-n photodiodes in AlxGa1–xN.4. Their NEP lies between a few pW/Hz1/2 and a few tens of pW/Hz1/2 depending on the geometry of the device and the bias voltage [MON 99b]. sometimes with high levels of aluminum (up to 76% [SAN 00]). The performance of these devices was further improved by the introduction of a region that was not intentionally doped (p-i-n photodiodes) and through the use of AlGaN/GaN heterostructures.210 Optoelectronic Sensors is no satisfactory explanation available for the origin of this gain. connected with difficulties associated with magnesium p doping.4.20Ga0. and by the associated ohmic contact whose resistivity remains high. the response time of a AlxGa1-xN p-i-n photodetector cannot be made small enough. 6. Arrays of heterostructure Al0. and in the technology of the ohmic contacts. and the device also contributes a significant level of noise associated with an ohmic contact of insufficient quality. p-n and p-i-n photodiodes The performance limitations of the first GaN p-n junction photodiodes were mainly caused by a high resistivity of the p-type layer. which can only be made with difficulty to exceed 1017 cm–3. the dynamic behavior of these photodetectors is for the most part limited by the RC time constant and the transit time of the photocarriers [MON 99c]. because the series resistance component due to the resistance of the electrical contacts remains large. The optimization of these last devices is however limited by the intrinsic difficulties in achieving a high level of p doping in AlxGa1–xN materials. then made it possible to make p-n junction GaN photodiodes with a fast temporal response (response time of the order of 100 ns) and a very low noise level. have also been described. Despite the existence of this gain. Progress made in understanding the growth conditions of the materials. which . and the high resistivity of the associated ohmic contact.80N/GaN p-i-n photodiodes with dimensions 32 32 and 128 128 pixels [BRO 00] have been made with good performance in terms of their response coefficient and detectivity.

which corresponds to a quantum efficiency of 25% [PAR 99].4. As the bias voltage increases. which corresponds to external quantum efficiencies of between 30 and 44%.or p-doped). meaning that the absorption of the light occurs closer to the surface.4. such that the carriers created through photo-ionization processes in the regions further from the junction are collected. A maximum in the response coefficient of 57 mA/W was reported at = 287 nm.Ultraviolet Photodetectors 211 leads to detectivity values which are notably smaller than those that can be obtained with p-n and p-i-n photodiodes based on GaN. This result confirms that the sensitivity is limited by the diffusion length of the charge carriers. A short diffusion length therefore leads to a decrease in the response coefficient for shorter wavelengths. p-i-n detectors based on group III nitrides are characterized by a sharp cutoff at long wavelengths. which is all the more marked when the surface layers are thicker (of the order of 0. An increase in this coefficient is observed in all cases as a function of the bias voltage [WAL 99]. The response coefficient of p-n and p-i-n homojunction GaN photodiodes typically lies in the range 0. A UV/visible contrast of more than four orders of magnitude is observed for all these devices.28Ga0. whose energy value is very closely linked to the . These results are improved by the use of a AlxGa1–xN layer in the illuminated region (which can be n.15 A/W. 6. The absorption coefficient increases as the wavelength decreases.10–0. A response coefficient of 0. The result is that electron-hole pairs are generated above the junction in the upper layer and must diffuse to the junction in order to be collected. Static characteristics In p-n and p-i-n photodiodes based on AlxGa1–xN. The small diffusion length of the carriers can affect the spectral response above the bandgap of p-i-n photodiodes based on AlxGa1–xN [XU 97]. the linearity of the photocurrent as a function of the incident light power extends over a scale of more than five orders of magnitude [WAL 99]. in order to avoid the loss of carriers through diffusion.72N(n)-GaN(i)-GaN(p) heterostructure illuminated by the rear face through the sapphire [YAN 98].2 μm). the width of the SCR increases.2 A/W at = 365 nm has been measured for a sapphire-Al0.1.

Faster response times can be obtained with AlGaN/GaN heterojunctions [XU 97].10Ga0. The only drawback of this particular type of device is its low response . This behavior can be attributed to the presence of a defect in the semiconductor. Response times of the order of 25 nm have been measured with zero bias voltage on devices whose optical surface area is 200 200 μm2 [OSI 97]. which makes it possible to build a device which has a low capacitance. Response time GaN-based p-n and p-i-n photodiodes have a temporal response which is in general limited by the RC time constant. GaN p-i-n photodiodes with response times below 1 ns have been constructed [CAR 99]. an additional cutoff wavelength is also required at short wavelengths. 6. Pulfrey and Nener [PUL 98] investigated the possibility of using p-i-n AlxGa1–xN/GaN heterostructures as bandpass UV detectors. In their theoretical study. However. Devices equipped with a 1 μm thick Al0. This result is obtained by increasing the thickness of the “i” layer to 1 μm. with an activation energy of 99 meV and a high concentration (of the order of 1018 cm–3) which can probably be explained by substituted magnesium atoms. In this case. WAL 98]. the AlxGa1–xN layer must be thick enough not only to accentuate the absolute response peak. the factor limiting the temporal response is the transit time of the carriers. but also to act as an integral high-pass filter. defining a spectral window for UV detection. VAN 97. however. with characteristic fall times that are more complicated than a simple RC effect [CHE 95.4.4.212 Optoelectronic Sensors bandgap of the material forming the active region.2.90N layer [KRI 98] have demonstrated that it is possible to achieve rejection on a level of more than two orders of magnitude in the short wavelength region. which can be explained by a reduction in the capacitance of the junction associated with the applied voltage. the response time of these devices is often non-exponential. For certain applications. and show an exponential fall in the photocurrent. This response time drops to 10 ns for a bias voltage of –6 V. In this case. with a high energy cutoff wavelength at the bandgap of AlxGa1–xN.

4.Iph (shot noise).4. with dimensions of 10 100 μm2. assisted by the tunneling effect [YU 98b]. Low dark current densities. by means of a SiO2 mask consisting of 35 μm widebands with 5 μm openings [PAR 99]. Such devices are of course ideal for the detection of rapid signals. their corresponding normalized detectivity is high. has been obtained in p-n junction photodiodes for a reverse-bias voltage of –3 V [OSI 97].4. the spectral density of noise under reverse-bias is well described by the equation Sn=2. Noise Under illumination. Finally. Diode mesas. In addition. and outside the openings in the SiO2 mask).4. This leakage current density is an order of magnitude smaller than that measured for devices which include the coalescence regions.3. 6.4. Since these devices have an optical area of 200 200 μm.Ultraviolet Photodetectors 213 coefficient.77N p-i-n photodiodes on GaN have been built using lateral growth. of the order of 3 1012 W–1Hz1/2cm. 6. have been made in the regions of the material corresponding to the “wings” of the lateral growth (outside the coalescence region. have been measured for the smallest devices lying on the “wings” of lateral growth. of the order of 7 fW/Hz1/2.q. All these responses are consistent . the spectral response has a more abrupt cutoff in the case of devices built on the “wings” of the growth. since these defects lead to a strengthening of the transport mechanism. which is around 0. This low value of the leakage current can be linked to the strong reduction in the density of dislocations.33Ga0. in a similar way both when the coalescence region is covered and when it is not. These particular devices also cover the coalescence region situated at the junction of the “wings” of lateral growth.03 A/W. of the order of 10 nA/cm2. devices with a similar form with square 300 300 μm2 mesas have been made from GaN on standard sapphire. A very low normalized noise equivalent power. p-i-n photodiodes on laterally-grown GaN Al0. Additional devices with mesa dimensions of 30 100 μm2 have also been built by way of comparison. and more than 6 orders of magnitude smaller than that of devices made from GaN on standard sapphire.

[YAN 98] report the fabrication of a GaN(n)/GaN(p)/GaN(i)/Al0. crosses the Al0. Avalanche photodiodes based on GaN The mechanism of avalanche photodiodes is analyzed in detail in Chapter 3. and the base was left floating.5. In its most up to date configuration. The photogenerated electron-hole pairs are separated by the electric field in the i region. to overcome the limitations in noise due to ionization coefficients for electrons and holes that are too close together. The unusual requirement for GaN-based materials is that very high fields are required to achieve multiplication. As far as GaN-based devices are concerned.5. OGU 97]. Phototransistors The operation of phototransistors is described in Chapter 4.4. with a response time between 100 and 500 ns. the high density of defects in these materials makes it difficult to achieve a homogenous multiplication across the whole active area of the device.4.4. the literature refers to bipolar and field effect transistors [KHA 95. as calculations of the ionization coefficients for electrons and holes in GaN show [KOL 95. the reverse-biased base-collector junction acts like a p-n junction photodiode.1. Yang et al.4. The UV light enters the device through the sapphire substrate.20Ga0. Electrical contacts were present on the collector and the emitter.80N(n) heterojunction phototransistor. and the electrons and . 6. Ruden [RUD 99] proposed a hybrid structure consisting of an AlGaN/Si heterostructure. Also. YAN 98].5. the base contact is not connected (floating-base mode of operation). and its photocurrent is amplified by the transistor effect. 6.20Ga0. and is absorbed in the GaN(i) layer. with multiplication taking place in the silicon. 6.80N(n) layer. Although a homogenous avalanche multiplication of carriers has recently been shown for PIN photodiodes based on GaN [MCI 00]. Bipolar phototransistors In a bipolar transistor.214 Optoelectronic Sensors with similar observations made for Schottky photodiodes made with laterally-grown GaN [MON 99d].

These devices display high response coefficients which can reach 3.4. However. The accumulation of holes in the floating base increases the injection of electrons from the emitter. The voltage impulse acts as a sort of electrical “reset to zero”. where they are rapidly drawn towards the drain by the high electric field. A gain of more than 105 has been demonstrated. In order to avoid persistent effects.5.000 A/W. analogous to that also recently observed in photoconductors. so that the gainbandwidth product remains constant. and the photocurrent is then measured at a bias voltage of 3 to 4 V.2. [YAN 98] showed that this recombination is increased by the application of a bias to the detector in order to force the holes to enter the emitter. the phototransistor is subjected to a “reset to zero” voltage impulse before each measurement of the response coefficient (a voltage between 7 and 10 V is used for this). 6. An excellent UV/visible contrast of 8 orders of magnitude was obtained [YAN 98]. holes ought to recombine in the base with electrons injected from the emitter. GaN-based phototransistors exhibit a sub-linear behavior of the photocurrent as a function of the incident light power as well as showing persistent photoconductivity (PPC).Ultraviolet Photodetectors 215 holes travel to the collector and the base respectively. Yang et al. the holes are trapped at defects which reduces their recombination rate and leads to persistent charge effects. In their normal operating regime. The photogenerated holes travel towards the channel. which results in a current gain. . with a steep cutoff and a temporal response of the order of 200 μs. Field effect phototransistors AlGaN(n)/GaN(n)/GaN(i) heterostructure field effect transistors have also been used in photodetection mode [KHA 95]. The gain is strongly dependent on the frequency. with illumination through the substrate (sapphire).

a low UV/visible contrast and significant and undesirable persistent photoconductivity effects. photoconductors. Schottky photodiodes have the advantage of a uniform and flat spectral response when they are excited at energies above the bandgap. However. this gain is associated with a sub-linear behavior as a function of the incident optical power. above all because the response coefficients of the devices is significantly reduced in such a configuration. studies of the ozone layer. missile plume detection. These drawbacks make them less suitable for most applications. MSM photodiodes. with performances that are at least equivalent in terms of reliability and robustness to other currently-available technologies that are themselves already very high performance. AlxGa1–xN photoconductors offer a high internal gain at ambient temperatures (~100 for Popt = 1 W/m2). With these goals in mind. independent of the incident optical power and the temperature. The use of a modulated optical signal makes it possible to considerably improve the linearity and UV/visible contrast. They also offer a steep cutoff slope.5.216 Optoelectronic Sensors 6. Conclusion The development of visible blind ultraviolet photodetectors is at present largely motivated by the large number of possible applications in fields as diverse as UV astronomy. due mainly to the large bandgaps available with AlxGa1–xN materials which make it possible to build detection systems which are entirely free from intermediate spectral filters. secure communications in space. It is important to emphasize that this range of nitride-based sensors is currently bringing a new degree of improved flexibility to UV detector fabrication in general. Schottky photodiodes. flame detection. etc. p-n and p-i-n photodetectors and phototransistors based on AlxGa1–xN materials have been developed in recent years. motor control. with a UV/visible contrast which is of . complementing older widely-available technologies for Si and SiC-based UV photodiodes. However. these photoconductors lose much of their advantage in this configuration. and the measurement system becomes much more complex and as a result much more expensive.

the performance of these devices is still . Phototransistors offer the combination of a very high gain with a record UV/visible contrast of 108. The minimum cutoff wavelength is currently limited by the difficulty of achieving a high p doping in AlxGa1–xN alloys at high concentrations of aluminum. Furthermore. in order to increase the performance of these devices. current results confirm that AlxGa1–xN alloys are the best choice in terms of semiconductors suited to UV photodetection. the possibility could be considered of integrating AlxGa1–xN MSM photodiodes with field effect transistors based on group III nitrides in the context of the monolithic fabrication of optical receivers. This clearly demonstrates that these devices are suited to environmental applications and the construction of UV photodetector arrays. However. these devices are promising for applications requiring a high spectral resolution. However. improvements are still currently needed in the p doping of these materials. to which a degree of degradation in the spectral response can also be attributed.Ultraviolet Photodetectors 217 the order of 103. their response time is normally limited by effects associated with the presence of trapping levels linked to magnesium. Given their significant bandwidth and their low noise level. in spite of the narrowness of their bandwidth which makes them poorly-suited to high-frequency operation. In spite of very promising published results. as well as their reliability. these devices can be a judicious choice to meet the needs of detection in the domain of visible blind optical communications. Their temporal response is limited by the RC product. with minimum response times typically of the order of nanoseconds. As a result. and can achieve a UV/visible contrast of 104. MSM photodiodes with very low dark currents have been built. To summarize. These devices have a photocurrent which varies linearly as a function of the incident light power and makes it possible to achieve a UV/visible contrast of 104. p-n and p-i-n junction photodiodes are linear as a function of the optical power.

“Visible-blind ultraviolet photodetectors based on GaNp-n junctions”. UMENO M.. BONEY J. Semiconductor Science and Technology.. CARTER JR C. 6. SRINIVASAN P. 453. “Very fast metal-semiconductor-metal ultraviolet photodetectors on GaN with submicron finger width”. 28.E.6.. SUN C.W.. ISHIKAWA H. EGAWA T. ANDREEV A.. Electronics Lett. LI T. YANG J.. CHELNOKOV V. Ga.. ZHAO G.. Bibliography [ANI 92] ANIKIN M. JIMBO T. 13.. Physica B 185. “Photoconductive gain modelling of GaN photodetectors”. 1994. PYATKO S.. “Stability and quantum efficiency performance of silicon photodiode detectors in the far ultraviolet”. 1993.. STRELCHUK A.A.. CAMPBELL J.J. EITING C.M......D..C.. SYRKIN A.. Physical 33 (1-2). [EDM 94] EDMOND J. 314. 67. very significant improvements in photodetector devices are anticipated from work currently in progress in the fields of epitaxial growth and the technology of group III nitrides.. Japanese Journal of Applied Physics.J. MUÑOZ E. [GAR 98] GARRIDO J. NOHAVA T. In consequence. MATTHEWS J.. 3940. 31. 1995. such as the technique of lateral growth.F. SAVKINA N. 91-93. N”.. 1992.. MONROY E. of Electronic Materials..A. pp. 0:11. 28. VAN DER LEUN J. 1999.C.A. article 6. YANG W. [ARU 00] ARULKUMARAN S.M. “UV-specific digital camera based on a 128x128 focal plane array of GaNAlGaN PIN photodiodes”.. KHAN M. SCHETZINA J. 563.. IZPURA I. “UV photodetectors in 6HSiC”.R. 2000.. L351. 325. [DEG 94] DE GRUJI F.. 39. photodiodes and high-temperature rectifiers in 6H–SiC”. 1989. KERNER J. “Estimate of wavelength dependency of ultraviolet carcinogenesis and its relevance to the risk assessment of a stratospheric ozone depletion”.H. MRS Internet Journal of Nitride Semiconductor Research.N. 5. [CHE 95] CHEN Q. DUPUIS R....C. [BRO 00] BROWN J. 1998. KORDE R.S.. KRISHNANKUTTY S. [CAR 99] CARRANO J. “Electrical characteristics of Schottky contacts on GaN and Al.N..” Blue LEDs.R.. Applied Optics. Sensors and Actuators: A.S. 2000..D.Y. [CAN 89] CANFIELD L..218 Optoelectronic Sensors currently limited by the high density of defects present in the heteroepitaxial layers... J. 0:89..L.. . Health Physics. KONG H..

S. [HAN 98] HANSEN P... . 2000.... GIBART P.. VERGHESE S. CALLE F.. 78. 17.E. CHEN Q.. 1998. DETCHPROHM T.J. 238.. OSINSKY A.. MOLVAR K. SAWAKI N. R41...H..R.F. 3938. Applied Physics Lett. KUZNIA J. 1999.. 1987.Ultraviolet Photodetectors 219 [GOL 99] GOLDBERG Y. Electronics Lett... “Visible-blindness in photoconductive and photovoltaic AlGaN ultraviolet detectors”. Applied Optics. MAHONEY L. 1999. WANG R.. [KHA 97] KHAN M. [MON 99a] MONROY E. “Ultraviolet photon counting with GaN avalanche photodiodes”. 28. [KRI 98] KRISHNANKUTTY S. GEIST J. “Origin of conductivity and low-frequency noise in reverse-biased GaN p-n junction”.. MISHRA U... Applied Physics Lett. 15.. 72. 2247...P.G.F.. “Electronic transport studies of bulk Zincblende and Wurzite phases of GaN based on an ensemble Monte Carlo calculation including a full zone band structure”. GASKA R. RUDEN P. 31.. BRAZEL E. NARAYAMURTI V. Applied Physics Lett.J. 3.. Solid State Electronics.J. [KUK 98] KUKSENKOV D.. EFREMOW N. SUN C. 1998. “Gate photodetector based on GaN/AlGaN heterostructure field effect transistor”. of Applied Physics. “Quantum efficiency stability of silicon photodiodes”. [KOL 95] KOLNIK J..A. DIFFEY B.. BEAUMONT B.. NOHAVA T. KHAN M. OMNES F.. 1365-1367. Semiconductor Science and Technology.. 1033.J. 2000. 5284. 1997.. 26. R55.W. 398.. 14. of Electronic Materials.A... article 7. “A reference action spectrum for ultraviolet induced erythema in human skin”.. 41... SHUR M. 76. KOZODOY P. 1995. NAKAYAMA H. TEMKIN H.. H. 1995. MRS Internet Journal of Nitride Semiconductor Research. MUNOZ E. IBBETSON J. [MCI 00] MCINTOSH K. “Recent developments of diamond detectors for particles and UV radiation”.P. STRAUSSER Y.. 72. “Scanning capacitance microscopy imaging of threading dislocations in GaN films grown on (0001) sapphire by metalorganic chemical vapor deposition”. “Fabrication and characterization of GaN/AlGaN ultraviolet-band heterojunction photodiodes”. RUDEN P.. “A study on barrier height of Au-AlxGa1-xN Schottky diodes in the range 0”.N.N.. [MCK 87] MCKINLEY A...S. WANG Y.P.A... TARSA E.. “Semiconductor near-ultraviolet photoelectronics”. 1987... Semiconductor Science and Technology.. [KOR 87] KORDE R. 1998. BRENNAN K.. (11).. [KHA 95] KHAN M.A.OGUZMAN I. JR. J. [MAI 00] MAINWOOD A. SPECK J. CIE Journal 6. DEN BAARS S.M.L.P. HIRAMATSU K. 287.. ERICKSON A.. MOLNAR R... YANG W. J.J.

ZHELEVA T.L.. of Applied Physics... 71.. 105.J. [PAR 99] PARISH G.. OMNES F.P. MUNOZ E. [MON 99d] MONROY E... [POL 98] POLYAKOV A. Appl. 1998. [PUL 98] PULFREY D. “High UV/visible contrast photodiodes based on epitaxial lateral overgrown GaN layers”. GIBART P. RUDEN P. J.V.. TARSA E. SANCHEZ F..D. GANGOPADHYAY S. MONROY E. GASKA R. KOZODOY P. [MON 99c] MONROY E.. 1488. WILLIAMS B.. 75. 42. TEMKIN H. of Applied Physics. 240.F.1999. 3401. 2334. Applied Physics Lett. 7827.... CALLE F... OMNES F. [MUN 00] MUÑOZ E. FLEISCHER S. 35. 88. [OSI 97] OSINSKY A..... 1999. [NAM 97] NAM O. 42....J.. 79. USIKOV A.P.. ROGALSKI A. NENER B.F.. OMNES F.. 1997.A. J.. 2000. 1731. KHAN M.220 Optoelectronic Sensors [MON 99b] MONROY E. GIBART P.B.Ga)N-based solar-blind ultraviolet p-i-n detectors on laterally epitaxially overgrown GaN”... MARCHAND H. OMNES F...K. Applied Physics Lett. PAU J.. CALLE F. BELLOTTI E. 2000. FINI P. 1998. 74. “Studies of the origin of the yellow luminescence band. DEN BAARS S.. OMNES F.. “Suggestions for the development of GaNbased photodiodes”. 1997. BEAUMONT B.. “Analysis and modeling of AlxGa1-xN-based Schottky barrier photodiodes”. 7433.. Electronics Lett.. KOLNIK J. 1999. 1959.. PUSHNIY B... 35. J. GOVORKOV A.D. “AlGaN metalsemiconductor-metal photodiodes”.P.. BREMSER M.... 2081. “High-performance (Al. . Journal of Geophysical Research.. GIBART P.. of Applied Physics. Phys.. 71. SMIRNOV N.. DAVIS R.T. GIBART P.. “High UV/visible contrast photodiodes based on epitaxial lateral overgrown GaN layers”.L..-n ultraviolet photodetectors”. “Low noise p. “Lateral epitaxy of low defect density GaN layers via organometallic vapor phase epitaxy”.H. [MON 00] MONROY E. “Semiconductor ultraviolet detectors”... Solid State Electronics. [OGU 97] OGUZMAN I. Solid State Electronics.Y. KELLER S.. [RAZ 96] RAZEGHI M. “AlGaN photodiodes for monitoring the solar UV radiation”.. MUÑOZ E. WANG R.. 1996.. Electronics Lett. 81.H. 1997. CALLE F.. 247. 1999. BEAUMONT B..... the nature of nonradiative recombination and the origin of persistent photoconductivity in n-GaN films”. 4865. Lett. MUNOZ E.. MUNOZ E. KUKSENKOV D. Applied Physics Lett. IBBETSON J...S. “Theory of hole initiated impact ionization in bulk zincblende and Wurtzite GaN”.. MISHRA U. 2638..B.. BRENNAN K. CALLE F.. CALLE F.

H. 1998..1996. “A thin film diamond photodiode for ultra-violet light detection”.. [WHI 96] WHITFIELD M. RAZEGHI M.H.. SPIE Proceedings. J.. [WAL 98] WALKER D. [SAN 00] SANDVIK P.. 2154. 1999. 265.. SUNAKAWA H.. MCKEAG R. TORREANO R. 38. SPIE Proceedings.J. LU C. 290. Oxford. [RUD 99] RUDEN P. 70.. [USU 97] USUI A..MI. 1988. SALVADOR A. Japanese Journal of Applied Physics. SPIE.. KUNG P.. Proc. Applied Physics Lett.DIAZ. “Ultraviolet-sensitive..... LEE I... ZHAN L. Applied Physics Lett. “Study of photocurrent properties of GaN ultraviolet photoconductor grown on 6H-SiC substrate”. “Back-illuminated GaN/AlGaN heterojunction photodiodes with high quantum efficiency and low noise”.1998..Ultraviolet Photodetectors 221 [RHO 88] RHODERICK E. HUANG Z. Applied Physics Lett. 3948... [WAL 99] WALKER D... 68. CHEN P. visible blind GaN photodiodes fabricated by molecular beam epitaxy”. SANDVIK P. KUNG P. 3629. 1997.. 71. KRISHNANKUTTY S.. KLAASSEN J. YANG K. [VAN 97] VAN HOVE J. ZHOU Y. RUDEN P. WALKER D. YAMAGUCHI A... 1999. RAZEGHI M. Japanese Journal of Applied Physics. 2000.S. CHAN S. 73. [SHE 99] SHEN B. “High speed. “AlxGa1-xN p-i-n photodiodes on sapphire substrates”. KRISHNANKUTTY S. K. KIM W. ZHOU H. CHOW P... ZHAN R. 767. 72. FAN Z.M.... 2282.. KUMAR V.JELEN..P.. C. M. 3629. 193.. DIAZ J. “Materials theory based modeling for III-nitride devices”. 1999. JACKMAN R. ZHENG Y..P..Y..H. CHEN Z. GOLDENBERG B... “Visible blind GaN p-i-n photodiodes”. SMITH G... Clarendon Press. 1997. MORKOC H.. SAKAI A.. HICKMAN R. Applied Physics Lett.D. Metal-Semiconductor Contacts.M. MARSH H. “Solar-blind AlxGa1-xN p-i-n photodetectors grown on LEO and non-LEO GaN”.. WU J. low noise ultraviolet photodetectors based on GaN p-i-n and AlGaN (p)-GaN(i)GaN(n) structures”. ESTES M. Applied Physics Lett..... 1997. 3303.. TANG H. WILLIAMS R. [YAN 98] YANG W.B. “Thick GaN epitaxial growth with low dislocation density by hydride vapor phase epitaxy”.RAZEGHI. ZHANG X. ZHANG X. ... YANG W.. SAXLER A..... NOHAVA T.. L899.... HAMILTON M... 174.... KUNG P.. SHAHEDIPOUR F. DIAZ J.. 978.D.P. HAMILTON M. [XU 97] XU G... 36. MCPHERSON S.

Applied Physics Lett.. LAU S. “The role of the tunneling component in the current-voltage characteristics of metal-GaN Schottky diodes”. 2099. BOUTROS K. LIU Q.J.S.D. “Analysis and modeling of AlxGa1-xN-based Schottky barrier photodiodes”. [YU 98b] YU L.J. 238. 73.S..S.. of Applied Physics. XING Q.. LAU S. REDWING J.S.S. QIAO D.Z. XING Q. 1998. J....222 Optoelectronic Sensors [YU 98a] YU L.. 1998..J.J. QIAO D. REDWING J. . 84.....M.

which we choose to limit to PIN and avalanche photodiodes.4 we will present the main results concerned with the analysis of noisy electrical circuits.1 we will recall the basics of signal analysis. Finally.2 and 7. In section 7. fundamentally limited by the background noise. This transformation. The aim of this chapter is to discuss the origin of this noise and also to study various means of minimizing it in order to obtain the best signal-to-noise ratio.6 discusses the problems of background noise associated with a photodiode-amplifier pair forming the photoreceiver. is. . in section 7.5 is dedicated to the study of noise in photodetectors. Section 7. leading to the frequency analysis of the power of random signals. a value which influences the quality of the optical detection. the recovery of a useful signal must necessarily involve an opticalelectrical (photon-electron) transformation. which makes use of an optoelectronic system. and lays out the methods of obtaining the best signal-to-noise ratio.3). After recalling the different sorts of noise associated with the most relevant physical processes (sections7. as with any system.Chapter 7 Noise in Photodiodes and Photoreceiver Systems In systems transmitting information by means of photons. in order to Chapter written by Robert ALABEDRA and Dominique RIGAUD. Section 7.

1] The signal s(t) has the same dimensions as i(t) which implies that h(t) has the dimensions of inverse time. the impulse response is: h t 1 exp RC t in s RC 1 [7. we use Fourier analysis which transforms the convolution in the time domain to a simple product in the frequency domain.1. we will recall a summary of the treatment of known signals in an attempt to justify a frequency approach as compared to a temporal approach for the treatment of the power or the energy of random signals. and we can recover the original s(t) through the inverse transform. Mathematical tools for noise Before discussing the treatment of random signals. Known signals with finite energy or power All systems are governed by a differential equation in the time domain. On the other hand.1. and we will characterize them by their impulse response h(t). in the case of an R–C circuit. 7.2] In order to avoid calculating the convolution. The output signal s(t) for an input signal i(t) can be calculated in the time domain either by solving this differential equation or by calculating the convolution of h(t) with i(t): s t i t *h t e h t d h e t d [7. Thus. if we .224 Optoelectronic Sensors illustrate these techniques.1. 7. we present the complete background noise calculation for a PIN-FET photoreceiver.

which is the maximum of the autocorrelation function. We find: Rx 0 Gx f df [7. using the concept of spectral density and the WienerKhinchin theorem which shows that the spectral density Gx(f) of a signal x(t) is the Fourier transform of the autocorrelation Rx( ) of x(t) such that: Gx f T F Rx Rx exp j2 f d [7. given its double convolution product. Rx(0) represents the energy or the power of the signal.4] and its inverse Rx T F 1 Gx f Gx f exp j2 f df [7. we will then use the autocorrelation function Rs( ) of the signal. the autocorrelation of the input signal.3] where h*( ) is the complex conjugate of h( ).6] . and Rs( ). that of the output signal. The relationship between Ri ( ).5] If we take = 0. We will repeat this same analysis in terms of frequency.Noise in Photodiodes and Photoreceiver Systems 225 want to analyze the same system from the point of view of energy or power. where phase information no longer has any meaning. This expression will only occasionally be used. is: Rs h* *h * Re [7.

3]. we will use the autocorrelation function. will cover the majority of practical situations.9] and which is much more easy to use than equation [7.7] which can be written as: Gs f H f H * f Gi f [7.1. fortunately for us. we have no choice but to treat them in terms of power or energy.226 Optoelectronic Sensors Gx(f) represents the energy or power distribution of the signal in the frequency domain. Random signals and background noise In the case of random signals. .2. DEFINITION OF ERGODICITY – The average resulting from many simultaneous experiments is the same as that obtained by the temporal mean across a single experiment. This class is characterized by ergodicity and timestability. This will lead us to define a class of random signals which.1.3]: F T Rs h* *h * Re [7. What we need to do now is to work out how to transfer what we have already set out for deterministic signals to the random signals which represent the sources of noise. 7. As we did for the known signals in section 7.1. We take the Fourier transform of equations [7.1] to [7. and of course without any consideration of their phase.8] or: Gs f H f 2 Gi f [7. the spectral power density and the same Wiener-Khinchin theorem.

Now.2. For frequencies below the inverse of the mean time between two . These spectral densities are expressed in A2/Hz or V2/Hz. and in order to recover a power or an energy we will integrate these spectral densities over the useful bandwidth.2. we can treat the random signals appearing in the expression in terms of energy or power. Bearing in mind the large number of electrons (of the order of 1022 cm–3). 7.). the Brownian motion of electrons produces a randomly-varying voltage v(t) at its open-circuit terminals.Noise in Photodiodes and Photoreceiver Systems 227 DEFINITION OF TIME STABILITY – In its strict sense. Thermal noise At the terminals of a metallic resistance of value R at temperature T in degrees Kelvin.10] where E is the expected value associated with the probability density of the process under study (Poisson. As we have seen in section 7. etc.1. the voltage v(t) obeys a centered Gaussian law. this means that the mean value is independent of time and the autocorrelation function only depends on = t1 – t2. and calculate the signal-tonoise ratio which characterizes the receiving quality. Fundamental noise sources Here we are interested in noise generated by solid devices in electronic circuits. In this case. Under these conditions the autocorrelation function is practically analogous to that for known signals which have the same properties. with a mean value of zero. 7.1. we will remain in the frequency domain in order to avoid convolving the input and output autocorrelation functions. using this formalism. We will use the spectral densities as our measurement of scale. Gaussian systems. we obtain for the random signal x(t): Rx E x t x t [7.

the dark current will make a significant contribution to the signal-to-noise ratio.38 10–23 J. Shot noise is always encountered when the current is caused by carriers emitted from a reservoir or across a potential energy barrier greater than kT.13] where Idark is the intrinsic dark current of the photodiode and Iph the photocurrent due to absorbed photons and representing the useful information.2. the spectral density associated with this random signal can be written in the basis of positive frequencies as: Sv f 4 kTR V 2 Hz [7. For PIN photodiodes. Idark<< Iph and the noise is mainly due to the photocurrent.2. 7.12] where q is the elementary charge (q = 1.228 Optoelectronic Sensors collisions ( r 10–12 s).K–1). We note that this spectral density is independent of the frequency (white noise). This is the case for reverse-biased junctions. for detection of low light intensities.6 10–19 C). for example. In contrast. Thus. Shot noise Shot noise stems from the discrete nature of electrons and their random emission and collection over time. . Frequency domain analysis of this process leads to a spectral noise current density (short circuit current) of: Si f 2q I A2 Hz [7.11] where k is the Boltzmann constant (1. Here also we find a white noise as long as we ignore the transit time of the carriers. it takes the form of a random variable governed by the Poisson distribution law. For modern devices. we have: Si total f 2 q I dark I ph [7.

The variance in the total multiplication factor is: M2 M 2 [7.e. we recall that the .Noise in Photodiodes and Photoreceiver Systems 229 7. the multiplication process involves both types of carrier across the whole multiplication region. Multiplication noise In a similar way to vacuum photomultipliers. In the case of vacuum photomultipliers. Numerous theoretical approaches have been proposed in the literature. there is only one type of carrier: electrons locally multiplied from one dynode to the next.16] we have: Si f 2qI pho M 2 2qI pho 2 I total I2 pho [7. If we ignore the multiplication inertia and the transit time of the electrons.15] where M is the mean value of the multiplication factor and Ipho is the primary photocurrent. we have a noise proportional to the total current squared. Setting: M I total I pho [7.17] in other words.14] and for large enough M we have: Si f 2qI pho M 2 [7. the noise generated by this photomultiplier is amplified shot noise. avalanche photodiodes with internal gain are based on the process of multiplication through the impact of carriers in the space charge region of the photodiode. To summarize these approaches.3. in the high electric field region in the space charge region. In the case of semiconductor avalanche photodiodes. i.2.

we have F M M and: . we define the excess noise factor F as: F M M2 M 2 M M x 2 M x 2 [7. The ideal material would be one in which only one of the two types of carrier could be ionized: either electrons.230 Optoelectronic Sensors two types of carrier are characterized by their ionization coefficient (in cm–1) for electrons and for holes. in which case = 0 and k = 0. For avalanche photodiodes. Taking the noise of a vacuum photomultiplier as our reference.20] In particular.19] in other words as the ratio of the variance of the multiplication process in the semiconductor to the variance of the same process in the vacuum tube. and for values of k close to 1. we can approximate F M by the expression: F M kM 2 M 1 1 k [7. as in germanium for example. In the case of an initial injection of electrons. for k = 1 ( = ). The noise characteristics would then approach that of a vacuum photomultiplier. or holes. the expression for the spectral noise density can be written in the general form: x Si f 2qI pho M [7.18] where x lies between 2 and 4. in which case = 0 and k = . The important parameter is the ratio k . and that these depend on the physical characteristics of the semiconductor materials.

22] which leads to an excess noise factor of: F M M 02 [7.Noise in Photodiodes and Photoreceiver Systems 231 Si f 2qI pho M 3 [7. which gives: Si f 2qI pho M F M 2 2 2qI pho M 4 [7. The aim was to obtain impact ionization by electrons only. because the interfaces resulted in an additional 1/f noise. In this case. the results did not live up to the performances initially promised. is 50 times larger than . the spectral density must approach that of a vacuum photomultiplier and: F M M 0 1 [7. For multiplication initiated by electrons. In order to achieve behavior for solid avalanche devices approaching that of vacuum photomultipliers. we arrive at the expression: Si f 2qI pho M 22 [7. we find excess noise factors F(M) of the order of M . Despite the sophistication of the technologies used.24] However. .21] For silicon avalanche photodiodes.25] These results are all depicted in Figure 7. several devices have been conceived based on engineering of the band structure with the help of heterojunctions.1. multiplication initiated by the less ionizing carriers is of no practical interest.23] In the limiting case where only electrons are responsible for multiplication ( = 0).

which is referred to as the 1/f noise. These two sources of noise are referred to together as excess noise. Generation-recombination noise This noise originates in the fluctuations of the number of carriers.1.232 Optoelectronic Sensors Figure 7. 7. The spectral density associated with these fluctuations can be written as: .3. as compared to the fundamental sources which are thermal noise and shot noise. These are generationrecombination (GR) noise and scintillation noise.3.1. Characteristic plots of the variations in the spectral noise density for multiplication in avalanche photodiodes as a function of the multiplied current 7. Excess noise In addition to the sources of white noise. linked to the presence of different trapping states in the bandgap of semiconductors. there are two other sources of noise which depend on frequency.

these values can be smaller. 1/f noise The origin of this noise is poorly understood. These fluctuations lead to current fluctuations proportional to the square of the mean current: Si f N2 N 2 I2 4 1 2 2 A2 Hz [7. The first was proposed by . is the time constant of the generation-recombination process taking place. displays a plateau at low frequencies followed by a 1/f 2 dropoff at high frequencies. We note that the spectral density varies as the square of the bias current.26] where N 2 is the second order moment of the random variable N which represents the number of carriers under consideration. it should be noted that in some specific cases. There are two interpretations of the origin of this noise.3. which leads to the term 1/f noise. 7.2.Noise in Photodiodes and Photoreceiver Systems 233 SN f N2 4 1 2 2 [7. the observed time constants lie between 10–3 and 106 s.28] where usually takes a value close to 1. of Lorentzian form. and A is an experimentally-determined constant which varies from one device to another. This spectrum. Its spectral density takes the form: Si f A f I2 A2 Hz [7. It is not possible to calculate the autocorrelation function of the process by the use of the inverse Fourier transform of Si(f) because the integral diverges. However.27] Generally.

shot noise and its potential amplification through multiplication processes. The frequency fc is defined as that for which the spectral density of the 1/f noise is equal to that of the fundamental noise (shot noise or thermal noise). The second interpretation is due to Hooge [HOO 76]. a situation which produces an overlapping series of generation-recombination events. In practical terms. due to fluctuations in the mobility of the carriers. Whichever of these is correct. the main sources of noise that we will take into account when evaluating the noise characteristics of a photoreceiver are thermal noise. For the rest of this chapter. Table 7. Table 7. it is almost certain that the amplitude of this 1/f noise is affected by the quality of the device and its fabrication technology. who attributes this noise source to a volume effect in materials. Expressions and characteristics of the main noise sources . and the 1/f noise when the optical signal requires a treatment that includes low frequencies.234 Optoelectronic Sensors McWhorter [MCW 57] who attributed the origin of this noise to the surfaces and interfaces of the material where a continuum of trapping states for carriers can be found (surface effects).1.1 summarizes the expressions for the spectral density for the different sources of noise involved with devices and systems.

Analysis of noise electrical circuits As is the case for deterministic signals.4. These representations are illustrated in terms of electrical circuits in Figure 7.1. In this way we can avoid all the problems of temporal analysis while retaining the concept of impendence or admittance as traditionally defined in circuit analysis. Representation of noise in bipoles A real bipole (for example a photodetector) is thus represented by a random voltage generator (Thevenin). Representation of a noisy bipole using a frequency spectrum approach for random processes . 7. These sources are represented using the Thevenin or Norton formalisms. with which a certain number of noise sources must be associated. noise generators will be characterized by the spectral density associated with the quantities that they represent.2. or by a random current generator (Norton) with spectral density Si(f ) in parallel with its admittance Y f 1 Z f . which is assumed not to be noisy. the behavior of a system is described in terms of its equivalent circuit.4.2. In contrast to generators of deterministic signals.Noise in Photodiodes and Photoreceiver Systems 235 7. Figure 7. The concepts presented below will be used later in the evaluation of the noise and the signal-to-noise ratio of photoreceiver systems. which has a spectral density Sv(f ) in series with its impedance Z(f ).

31] – or the equivalent conductance Geq of the noise: Geq Si f 4kT [7. we can define: – either the equivalent resistance Req of the noise: Req Sv f 4kT [7. The equivalence of these two representations can be expressed in the following equation: SV f Z f 2 Si f [7. all noise generators will be marked with hatching in order to distinguish them from normal signal generators.2 make it possible to directly treat series or parallel combinations of noisy .30] In the case of thermal noise. If we take the example of shot noise.33] The representations of the noise given in Figure 7.236 Optoelectronic Sensors Henceforth. we can define the equivalent noise current Ieq as: I eq Si f 2q [7.29] which connects the two spectral densities that we have used.32] – or the equivalent temperature Teq of the noise: Teq Sv f 4kRe Z f Si f 4kRe Y f [7. Other parameters can be used to characterize the noise of a bipole.

SvT(f ) can then be written: SvT f Sv1 f Sv2 f 2 Re C f Sv1 f Sv2 f [7. This dependence can be represented in terms of a correlation coefficient defined by: C f Sv1v2 f Sv1 f Sv2 f [7.37] 7. the noise is generally represented by a minimum of two partially-correlated noise sources.34] where Sv1v2(f ) represents the cross-spectrum which encapsulates the statistical dependence of the two random processes associated with the two noise sources. we have.35] with 0 |C| 1. In a similar manner. Thus. using the parallel representation: SiT f Si1 f Si2 f 2 Re C f Si1 f Si2 f [7.36] In the normal case where the two sources of noise are uncorrelated (C = 0). Representation of noise in quadripoles For quadripoles (the case of amplifiers placed downstream of photodetectors). can be placed at the input and/or output of the quadripole. This leads to six equivalent representations.4.2. in the case of a parallel combination of two bipoles. the resultant spectral density is then simply the sum of the spectral densities associated with each bipole. These sources. of voltage and/or current. each one associated with one of six matrices describing the properties of the .Noise in Photodiodes and Photoreceiver Systems 237 bipoles: two bipoles in series are equivalent to a single bipole whose spectral density of voltage noise can be written: Svt f Sv1 f Sv2 f 2 Re Sv1v2 f [7.

the noise is represented by a voltage generator and a current generator both present at the input of the quadripole (which is itself assumed noise-free). one at the input and one at the output of the quadripole. two current generators are used. In the second case. The two most commonly-used configurations for representation of a noisy quadripole In the first case. and assuming a resistive load Rg at the input to the quadripole which only generates thermal noise. This representation is well suited to describe noise in amplifiers. This representation is useful for describing the intrinsic noise of bipolar and field effect transistors.3).238 Optoelectronic Sensors quadripole. and that the presence of the quadripole amplifier degrades the signal-to-noise ratio. we have. . In this way it can be shown that the noise factor is always greater than 1. with spectral densities Sen(f ) and Sin(f ). with spectral densities Si1(f ) and Si2(f ). Figure 7.3a. Bearing in mind the first representation given in Figure 7.3. for a given frequency f.38] This equation shows that the noise factor can also be written as the ratio of the total output noise power to what the output noise power would be if the quadripole was noise-free. In practice two of these representations are the most commonly used (see Figure 7. An important quantity for characterizing the noise of a system is its noise factor. which is defined as the quotient of the input and output signal-to-noise ratios: F f S B e S B s [7.

it can be shown that under certain conditions the noise factor of the resultant quadripole can be written as a function of the individual noise factors Fi and the power gains Gi: F 1 F1 1 F2 G1 1 F3 1 G1G2 .40] This equation (the Friis formula) shows that the first quadripole controls the noise of the whole system as long as it is not very noisy and its gain is significant. In order to do this. it will be necessary to match the photodetector to the first amplifying component in order to minimize its noise factor and give it a sufficient gain that the noise of the successive stages can be ignored. which will allow us to express the noise associated with the photocurrent generated by the conversion of photons into electron-hole pairs. the noise factor passes through a minimum for the value: Rgopt Sen f Sin f Finally. Noise in photodetectors In order to describe the noise of a photodetector in the presence of a weak signal. we need to define a number of optoelectronic quantities and characteristic parameters of photodetectors.5..Noise in Photodiodes and Photoreceiver Systems 239 F f 1 Sen f 2 Rg Sin f 2 Rg Re Senin f 4 kTRg [7. for a series of quadripoles. we will need to evaluate a signal-to-noise ratio for the information carried by the photons. 7. when designing a photoreceiver. [7.. . Thus.39] As Rg varies. These issues will be covered when we study the PIN-FET system.

8 AW ext ext [7. For h = Eg we define the cutoff wavelength C given by: 1. there is an electric field present which separates the two types of carriers generated by absorption of the energy Eph = h of the incident photons. The photodetector exhibits a basic noise Si(f) with which the dark current is associated: . We have: R q hc 0. only knowledge of the photoresponse R is required. of wavelength . For this generation to take place. we require that: E ph hv hc Eg c [7.43] in μm and c in m/s. In calculations of noise in if we express photodetectors.24 in m with Eg in eV Eg c [7.1.5. Characteristic parameters In all optoelectronic transducers.41] where h is the Planck constant.240 Optoelectronic Sensors 7. v is the optical frequency and Eg the size of the bandgap of the semiconductor material.42] The response coefficient R of the detector allows us to relate the generated photocurrent Iph to the incident optical power Popt: R I ph Popt AW ext The calculation of R involves the external quantum efficiency and the wavelength of the optical signal.

Noise in Photodiodes and Photoreceiver Systems 241 I obs Si f 2q We will now consider the incident optical power which would produce a photocurrent Iph such that the associated shot noise 2qIph would be equal to the basic noise referred to above (a signal-to-noise ratio of 1). brought down to the level of the photodetector. we will keep in mind when defining the NEP that the basic noise used as a reference consists of the basic noise of the photodetector and the basic noise of the amplification system.44] For a given noise equivalent bandwidth. we then have. we therefore have: . we can directly express the signal-to-noise ratio in the form: S B 10 log dB Popt rms NEP [7. The NEP represents the smallest optical power which can be detected for a given noise equivalent bandwidth associated with the electronic system. Thus. Based on the definition of the response coefficient. This optical power characterizes the photodetector and is designated the NEP (Noise Equivalent Power). which is the inverse of the NEP: D 1 W NEP 1 In the case of a photoreceiver. for linear photodetectors and a unit bandwidth: R I ph Popt I obs NEP [7. an identical equation can be written in terms of rms values.45] of We can also define another characteristic parameter photodetectors. the detectivity. For a bandwidth f.

PIN photodiodes These are diodes. of mesa or planar structure. .2. Schematic cross-sections of PIN photodiodes using (a) mesa and (b) planar technology The most widely used semiconductor materials are shown in Table 7.2 along with their main characteristics.5. 7.4.4). in which the n and p regions are separated by an intrinsic semiconductor region where the photocarriers are generated and separated (see Figure 7.46] R where Si (f )total = Si (f )photodetector + Si (f ) amplifier. Figure 7.242 Optoelectronic Sensors NEP Si f total f [7.

Noise in Photodiodes and Photoreceiver Systems 243 Table 7. series resistances. carrier lifetimes. the conduction consists of two components: the dark current Idark and the photocurrent Iph = RPopt. Main characteristics of the most widely used semiconductor materials A theoretical approach to these devices must take into account the band structure (direct or indirect) and optimize its geometric dimensions as a function of the bandwidths.5.5 shows the ideal I = f (V) characteristics of a photodiode at different incident light powers. etc. dark currents. leakage currents. Figure 7. which is the normal mode of operation of a PIN photodiode. Two shot noise generators are associated with these two currents. and taking into . Typical characteristics of a PIN photodiode under different illuminations Under reverse-bias. operating temperature. Figure 7.2.

with a mean gain M.7 shows a cross-section through two avalanche photodiodes. the dark current is negligible. we can calculate the intrinsic NEP of the device from equation [7. For small bandgap materials. which are used for infrared detection.3.46]. Avalanche photodiodes Avalanche photodiodes are PIN photodiodes which make use of the physics of impact ionization to achieve multiplication through the effect of the electric field on the charge carriers. This process generates noise as discussed in section 7. 7. Figure 7. the equivalent noise circuit is shown in Figure 7. but technical issues are associated with the metalsemiconductor contacts due to the very high resistivity of these materials. .6.6.5. For the large bandgap materials used to detect short wavelengths. Figure 7. the dark current will be the main limitation on the operation of the device at ambient temperature.2. Equivalent noise circuit for a reverse-biased PIN photodiode under illumination Starting from this equivalent circuit.244 Optoelectronic Sensors account the series resistance Rs which is the source of the thermal noise.3.

47] In general. the spectral density of the noise current can at first approximation be written as: Si f 2q RPopt I dark M F M 2 A2 H 2 [7. which is different from that for the photocurrent because we do not have control over the injection into the multiplication region of the carriers originating from the dark current.7. for detection at = 1. (a) Structure of a n+p p+ silicon avalanche photodiode for detection at = 0. Achieving this optimal sensitivity depends on the . Noise optimization of photodetectors The main aim when designing a photoreceiver is to minimize the optical power that needs to be detected while maintaining a specified signal-to-noise ratio.6. but the photocurrent is written as: I ph M RPopt A For the shot noise. 7.6 μm The equivalent noise circuit is the same as for a PIN photodiode. the Idark component is multiplied by a factor M .Noise in Photodiodes and Photoreceiver Systems 245 Figure 7.9 μm. (b) Heterostructure avalanche photodiode with separate absorption and multiplication regions.

8 shows the equivalent noise circuit of the detector-preamplifier system. Figure 7. the signal-to-noise ratios of the receiver and the detector are connected by the relationship: S B 1 S B F receiver detector [7. Specifically.246 Optoelectronic Sensors choice of photodetector and the amplification sequence.48] df Si f detector which in the presence of a preamplifier becomes: 2 isignal receiver f S B [7. In order to achieve the minimum degradation in the signal-to-noise ratio of the photoreceiver.49] Si f preamp Si f detector df Thus.4. the amplifier must be matched to the detector in order to minimize F. The photodetector is represented by its admittance Ys in parallel with a noise current generator Is.50] where F is the noise factor of the preamplifier.6. The preamplifier noise is represented . Formulation of the problem Bearing in mind section 7. the signal-to-noise ratio can be written: S B 2 isignal detector f [7. 7.1. at the level of the detector. whose input load is the photodetector.

These two generators are correlated. we will write it in terms of a correlation admittance Ycor: ic f Ycor f en f [7.Noise in Photodiodes and Photoreceiver Systems 247 at its input by a voltage generator en and a current generator in. and another uncorrelated part (inc): in f ic f inc f [7. Equivalent noise circuit for a photoreceiver.8.8b can then be drawn in the form shown in Figure 7.51] Figure 7. positioned at the system input Since the correlated part is proportional to en at a given frequency.9. and we write in in the form of one part which is correlated with en (which we will call ic). .52] The equivalent circuit in Figure 7.

55] Bearing in mind equations [7.51] and [7. we have in terms of spectral densities: Si f Si f Ys Ycor Sen f 2 s Si nc f [7. Equivalent circuit used to calculate the spectral density of the total noise current of the photoreceiver Bearing in mind our frequency-based approach to random phenomena. we have: Gnc Gin Rn Ycor 2 [7.52].54] n To avoid encumbering the notation.248 Optoelectronic Sensors Figure 7.4): Si f 4kTGs f 4kTRn f f f 4kTGnc f 4kTGi n s Sen f Si Si nc 4kTGn 1 f f [7.56] . The noise factor of the photoreceiver now be written: Si Si s F f f 1 1 G Gs nc Rn Gs Gcor 2 Bs Bcor 2 [7.9. from hereon in we will not bother indicating that the conductances are a function of frequency.53] The different spectral densities can be expressed as a function of their equivalent noise conductances (see section 7.

We find: Bsopt Bcor [7.58] then: Gsopt 2 Gcor GnGnc GinGn 2 Bcor [7.62] which in the basis Gs .61] Hence. we set to zero the derivative of F with respect to the variables Gs and Bs.57] In order to minimize the noise factor.59] Under these conditions.60] It follows that: F Fmin 1 Gs Gn Gs Gsopt 2 Bs Bsopt 2 [7. we find we must solve the following equation in terms of Gs and Bs: Gs2 Bs2 2Gs Gsopt F Fmin Gn 2 Bs Bsopt 2 Gsopt 2 Bsopt 0 [7.Noise in Photodiodes and Photoreceiver Systems 249 and the noise factor can be written as: F 1 1 G Gs in Rn Gs2 Bs2 2 Gs Gcor Bs Bcor [7. Bs represents a circle with center: . if we constrain F to be greater than or equal to Fmin by modifying the source admittance Ys = Gs + jBs. the minimum noise factor can be written: Fmin 1 2 Gsopt Gn Gcor [7.

The matching of the detector and preamplifier must therefore be as independent as possible of this parameter.10a and 7.10. (a) Circles of constant noise factor in the basis Gs. Bs. . It must be remembered that the series of circles and their resultant surface is only representative of a single frequency.63] and radius: R F Fmin Gn F Fmin Gn 2Gsopt [7.10b.64] Figure 7. (b) Plot of the surface defined by the noise factor We note that for F = Fmin the circle shrinks to its center such that: Gso Gsopt Bso Gsopt The results obtained are shown in Figures 7.250 Optoelectronic Sensors Gso Bso Gsopt Bsopt F Fmin Gn [7.

Conceptual diagram for matching a photodetector to its preamplifier in order to minimize the noise factor Ys is the admittance presented by the matching quadripole at the input of the preamplifier. and at 10 GHz: Yopt = 12. Yd is the admittance of the photodetector.Noise in Photodiodes and Photoreceiver Systems 251 7.3 mS at 10 GHz. In order to have F = Fmin.5 mS – j 2. Thus.8 mS – j 10.11.2.2 mS at 2 GHz and Yd = 0. Similarly. a photodiode with a capacitance Cd of 0. in the complex plane of admittances.11. the matching of the photodetector to its transistor move from the point YD to the point Yopt (see Figure 7. . at 2 GHz: Ysopt = 2.03 mS + j 1.1 pF and a series resistance Rs of 20 presents an admittance of Yd = 0.78 mS + j 6.12a).6. for a field effect transistor with a gate of width 200 μm we have. Figure 7. For example. Matching is achieved using a quadripole or matching transformer and/or through a matching admittance YA. it is necessary that Ys = Yopt. Concepts for photodetector-transistor matching A conceptual diagram for the matching is shown in Figure 7.1 mS.8 mS.

The noise factor will be minimized. . The noise factor will not be minimal. (a) Representation in the admittance plane of the photodetector’s Yd and of Yopt which leads to a minimum of the noise factor. we will be close enough to the axis G of the conductances that Cd will be weak.12b). we can imagine the following points: a) connect the photodetector directly to the transistor. “the real part of”]] (Figure 7.12. b) place an admittance Ya = Ga + jba in parallel with Yd so that Ya = Yopt + Yd (see Figure 7.e. However. It follows that the signalto-noise ratio is degraded: increasing the real part of the admittance seen by the transistor increases the noise and so is not a worthwhile solution. c) reduce the width W of the gate of the field effect transistor in order to have Re{Yopt} = Re{Yd} [I think that Re should read Re [i. The signal-to-noise ratio of the preamplifier is optimized. but at the expense of increasing the source noise. (d) matching by use of a transformer In order to do this. This may then not be large enough for it to be possible to neglect the noise of the subsequent stage. It should however be noted that the reduction in W leads to a reduction in the transconductance of the gain. (b) movement due to the matching admittance Ya = Ga – jBa.12c).252 Optoelectronic Sensors Figure 7. (c) matching by lossless admittance and reduction in the width W of the gate.

In practice. 7. this requires being able to build a lossless (and hence noiseless) matching circuit with a significant K factor.12d). bearing in mind that the admittance of an inductance decreases with frequency when Im{Yopt} grows. the system must operate in the useful frequency range. and its background noise will be represented by two correlated current generators ib1 and ib2 at the input and output of the device respectively (section 7. Yd becomes Ke{Yd + jBa} = Yopt .13.1. Equivalent circuit for the preamplifier.7.2). Calculation of the noise of a photoreceiver 7. This leads to the arrangement depicted in Figure 7.5. It also requires that the lossless load Ya has a negative susceptance (inductance).Noise in Photodiodes and Photoreceiver Systems 253 d) use an impedance transformer (Figure 7.7. and the signal-to-noise ratio of the source remains unchanged. Figure 7. Finally.8 in terms of the basic noise associated with the transistor which forms the preamplifier. The real part of Ys has been amplified along with the signal. highlighting the different sources of noise The calculation of the noise present at the input is made by considering the short-circuit output. it is also necessary to match the imaginary part by using a lossless admittance jBa. This makes this option challenging to achieve in practice.13. The behavior of the transistor will be described here in terms of its admittance matrix Yija. In this case. Basic equations We need to express the equivalent noise circuit in Figure 7. and this does not detract from the .

67] we find that the spectral density of the noise current associated with current i2 is: Si 2 f Si b2 f Gicc 2 Sis f Sib1 f 2 Re Gicc Si b1ib2 f [7.69] Setting Ys Gs jBs and Si b1ib2 f Y21 a Gcor jBcor .68] and so the total noise at the input to the transistor can be written: 1 Gi 2 cc Si e f Si s f Si b1 f Si b2 f 2 Re 1 Gi* cc Si b1ib2 f [7.254 Optoelectronic Sensors generality of the problem since the current gain will also be calculated under these same conditions.65] and for V2 = 0 we have: i2 V1 ib2 Y21aV1 ib1 1 i Y11 s [7. we have: . We set: Y11 Ys Y11a [7.66] given that the output short-circuit current gain can be written: Gi y21a Y11 cc v 2 0 [7.

71] generates a circle in the admittance plane. 7. equation [7. Sie(f) takes the form: Sie f C GGs BBs A Gs2 Bs2 [7.71] which represents the basic equation for the input noise of a photoreceiver.2. with: Y f 11a Y21a f 2 C Si b1 f G11a Y21a 2 Si b2 2 Gcor G11a Bcor B11a G 2 Si b2 Gcor 4kT [7. we will elucidate the two noise generators ib1 and ib2 as well as their correlation. similar to that which we saw previously. .7.Noise in Photodiodes and Photoreceiver Systems 255 Sie f Sis f Si b1 f G11a Gs 2 B11a 2 Bs 2 Y21a Si b2 f [7.72] f 2 B A 2 Si B11a Y21a b2 2 Si b2 Bcor f Y21a For a given value of Sie(f). Models of transistor noise In this section.70] 2 Gcor Gs G11a Bcor Bs B11a Insofar as Sis(f) can be identified as the thermal noise of Gs.

has units of A2 and represents the value of the spectral density of the 1/f noise at 1 Hz.2. we have: Ag f 2 Cgs f 2 Si b1 f 2qIg 16 2 kT gm R [7.73] and [7.74]. In general terms. P is a parameter which also depends on the bias point.1.256 Optoelectronic Sensors 7. a shot noise component associated with the gate current.73] There is a 1/f noise component. and a component produced by noise in the channel through capacitative coupling.76] . Working from the definition of the correlation coefficient Cr between two noise sources (here the channel and gate noises): Cr Cre jCim Si Si b1 b1ib2 f b2 f Si f [7. Ag. where it is assumed that only the thermal noise contribution leads to correlation: Si f 4kTCgs RP Cre jCim b1ib2 [7. Here R is a dimensionless parameter which depends on the bias point of the device: Si f Ac f 4kTg m P b2 [7.75] we have from equations [7.7. Case of a field effect transistor ib1 represents the noise associated with the gate and ib2 the noise from the channel (common-source configuration of the transistor). along with Ac.74] Here we have a 1/f noise contribution and a component associated with the diffusion noise (thermal noise) of the channel.

it is useful at low frequencies to add the 1/f noise sources (at least at the level of the base) and maybe account for the thermal noise of the terminal resistances (particularly for rbb’).7. the transistor is equivalent to its input resistance r in parallel with a capacitance CT. The access points Rg. Figure 7. At high frequencies.14.2.2. along with the two noise generators ib1 and ib2. Case of a bipolar transistor The two noise sources model the shot noise associated with the base current: Si b1 f 2qI B 2q Ic [7.Noise in Photodiodes and Photoreceiver Systems 257 The small-signal equivalent circuit for a field effect transistor. is shown in Figure 7. the total input noise to the device can be written: .78] On top of the shot noise.14. from the point of view of the detector. Rs and RD may be noisy and contribute thermal noise in the form of 1/f noise. Equivalent circuit of a field effect transistor.77] and the collector current: Si b2 f 2qIC [7. with noise current generators at its input and output 7. bearing in mind that.

The minimum SiT(f) is obtained for the value of the current IC which cancels out the derivative of SiT(f). along with the two noise generators ib1 and ib2 are shown in Figure 7. we find: Si T f 2qI c 1 1 2 2 k 2T 2 2 C qI c T 2 [7. . so that: kT C q T 1 1 2 Ic [7.258 Optoelectronic Sensors Si T f 2qIC 1 r 2 2 CT 2 2qI C 2 gm [7.80] An increase in the capacitance can be counteracted by an increase in the current IC.79] Given that r kT and g m r qI c . The smallsignal equivalent circuit for the bipolar transistor.82] showing that this optimum varies linearly with frequency.15.81] The input spectral noise density becomes: 1 1 Si opt f 4 kTCT [7.

15.7. Circuit diagram for a PIN-FET photoreceiver From this circuit diagram. Equivalent circuit for a bipolar transistor. with noise generators at its input and output 7.16 shows the structure of a PIN-FET photoreceiver for direct detection of the high impedance type (without Rf) or the transimpedance type (with Rf).Noise in Photodiodes and Photoreceiver Systems 259 Figure 7. Figure 7. the elements of the admittance matrix of the preamplifier can be written (ignoring parasitic elements): . Example calculation: a PIN-FET photoreceiver Figure 7.3.16.

69]): .86] The total noise entering the input of the system can be written (see equations [7. the source admittance takes the form: Ys 1 R pol jCd [7.84] Designating the equivalent resistance of Rpol and Rf in parallel as RT.260 Optoelectronic Sensors Y11a Y12 a Y21a Y22 a 1 Rf 1 Rf gm 1 Rf j Cgs j Cgd Cgd 1 Rf j Cgd gm 1 Rf [7.83] j Cgd In addition. combining the PIN diode with its capacitance Cd. with a further contribution from thermal noise in the series resistance Rseries of the photodiode.67] and [7. Cgs and Cgd in parallel as CT (there may also be parasitic capacitances Cp to take into account). and the equivalent capacitance of Cd. the equation Y11 = Y11a + Ys can then be written: Y11 1 RT jCT [7. which we can correct by taking account of the access resistance Rg at the gate of the transistor: Si s f 2q I obs I signal 4kT Rseries Rg Cd 2 2 [7.85] The photodetector noise is accounted for in the shot noise that stems from the dark current Idark and the signal current Isignal.

we have: 1 Si f 2q I dark 1 RT I signal 2 Rg Cd 2 CT 2 Cgd 2 Ig 2 Ag f 2 Cgs 2 2 Ac RT f g2 m 2 2 CT 2 Cgd 2 2 4kT Rseries gm 1 RT gm jCT jCgd R gm P 2 1 RT 2 gm [7.Noise in Photodiodes and Photoreceiver Systems 261 Si f Si s f Si b1 f Y11 Y21a 2 Si b2 f 2 Re * Y11 * Y21a Si b1ib 2 f [7. where gm >> Cgd . we have: Si f 1 Ag f ID 2 Ac 2 2 RT g m 2q Ig 4kT 2 CT 2 gm 1 RT kT RT 2 CgsR P 2 RT g m [7.89] f 4f f 2 16 Ac 16 RP Cgs gm 2 kT 2 C P gm T 2CimCT RPCgs . For instance.87] Thus. we can reorganize this equation to make clear the frequency behavior of the noise at the input to the preamplifier.88] 2Cgs RPRe Cre jCim At this point.

Dotted lines indicate the contributions from 1/f noise.2 pF 150 nA. Calculation of Si f as a function of frequency. R 0.262 Optoelectronic Sensors As an example.4 20 nA 12 mS. Cim 1011 pA2 .2 pF. Figure 7. equation [7. in the case where Cgd remains smaller than gm.2. Figure 7.9. Ag 0.17. thermal and shot noise. I Diode 30 k . P 103 pA2 1. for a system whose characteristics are listed in the caption. Cgs 0.17 shows the variation of pA in Si f Hz as a function of frequency. as well as those of the f and f 2 contributions Cd Ig Rp Cre Ac Cp 0. g m 0. using the appropriate values for the characteristic parameters of the photoreceiver.4 At high frequencies.88] can be written: .

it can be seen how the width W of the transistor gate can be calculated to match it to the photodetector.90] CT gm RP Cre Starting from equation [7. setting: gm Cgs C gd g moW C gsoW C gdoW and Co C gso C gdo We then have: 1 RT 2 Si P f 2 Co W 4kT 2 Rseries 2 Rg Cd 2 CdCo gm 0 2 2 Cgso gm W 2 R 0 P 1 gm W R2 T 0 gm P 0 Cd gm W 0 2P 2 2 [7.91] 2 2Cgso RPCimCoW gm 0 2Cgso RPCimCd gm 0 2 Cgso gm 0 RP Cre RT In terms of the variable W. this last equation can be rearranged to obtain: .Noise in Photodiodes and Photoreceiver Systems 263 Si f P 1 gm R2 T 1 4kT RT 2 CT 2 Rseries 2C gs Rg Cd 2 2 2 Cgs 2 gm 1 RT g m Cim R [7.90].

264 Optoelectronic Sensors Si f W 1 RT 4kT Cgso2 R gm 0 PCo2 2Cgso RPCimCo 2 2 4kT 2 P CdCo gm 0 2 2 Cgso RP CimCd gm 0 Cre RT Wg m [7. 4kT Cgs 2 gm 2 2 R .92] then takes the final form: . can be written as: 4kT Co2 gm R ' with R ' Cgs 2 Co2 R [7.92] 1 1 4kT W gm P 1 2 RT Cd 2 2 Rsˇ rie Rg Cd 2 2 0 0 Since the gate access resistance is inversely proportional to: W Rg Rgo W it is reasonable to make the following approximation when integrating the photoreceiver: Rsˇ rie Rg Cd 2 2 Wg m 0 Rsˇ rie 0 Rgo Cd 2 2 gm 0 [7.93] Additionally. the term representing the noise induced by the channel in the gate.94] Equation [7.

with respect to W.98] g m Rseries 0 P Rgo R' P 2 R ' PCim which can be written as: . we have: Si P f 8kT PR 'Cim Co Cd gm 0 2 [7.Noise in Photodiodes and Photoreceiver Systems 265 Si 1 RT f 4kT W Co2 R' P gm 0 2 R ' PCim R ' P CimCd 2 2 P CdCo gm 0 2 2 Co gm 0 Cre RT Rgo g m [7.96] We then have: Si f 4kT 1 RT 1 2 RT Cd 2 2 2 CoCd gm 2 P R ' P Cim 0 Cre RT Cd 2 R ' PCim [7.95] 1 1 1 P 2 W gm RT 0 Cd 2 2 Cd 2 Rseries 0 0 The derivative of this final result.97] P1 g m Rseries 0 Rgo R' P In the case where the value of RT dominates. such that: P1 Wopt Cd Co 1 2 RT Cd 2 2 g m Rseries 0 Rgo R' P 2 RPCim [7. leads us to an optimal value of W which minimizes the noise.

Comments and conclusions Several comments can be made based on the example that we have worked though. Equation [7.266 Optoelectronic Sensors Si f min 16kT Co Cd gm 0 2 [7. .99] with: 1 P 2 PR 'Cim P Rseries Rgo g m R' P 2 R ' PCim 0 We can consider as an equivalent noise parameter for the channel which allows us to treat the system as if the gate noise is zero (R = Cr = 0). Now returning to the sources of noise that we have ignored up to now.99] accounts for the high-frequency noise of the receiver. There are four sources of noise that are a function of frequency: – a 1/f contribution which dominates at low frequencies and is connected with the manufacturing technology of the devices. – a white noise component which represents the minimum noise level of the photoreceiver.100] This equation gives the frequency dependence of the optimized noise. 7. the total input noise after the gate width has been optimized is: Si 64 f 2 min 2q I dark I signal Ig Ag f Ac 4 2 CT 2 2 gm f Co kT Cd f 2 gm 0 [7.8.

– a component proportional to the frequency squared. and to summarize all the concepts presented in this chapter. This type of calculation can be applied to different photoreceivers involving bipolar or field effect transistors connected to PIN or avalanche photodiodes. – the bias resistance of the photodiode has an effect on the white noise and low frequency noise of the system. we calculate the signal-tonoise ratio for an incident optical power Popt and a photodiode response coefficient R through the equation: S B Ben RPopt Si 2 f df [7. Finally.101] where Ben represents the equivalent noise bandwidth. associated with the high frequency performance of the transistor and the capacitative coupling between the channel and the gate. Cd and Rseries) contribute to the white noise and the high frequency noise as a result of its matching to the amplifier. which brings into play the low frequency noise of the channel. Cgs. associated with the coefficient defined above. Cgd.Noise in Photodiodes and Photoreceiver Systems 267 – a component proportional to frequency. gm. – the transistor characteristics (Ig. Ag and ) affect the noise performance of the photoreceiver across the whole frequency range of its operation. . Ac. at the level of the whole photoreceiver. The noise contribution from each element in the photoreceiver circuit is clear: – the intrinsic characteristics of the PIN photodiode (Iobs.

.. DAS M. Semiconductors Devices for Optical Communication. Hewlett-Packard. vol. SCAVENEC A. Les télécommunications par fibres optiques. December 1993. Bibliography [BER 93] BERTHIER P.. DAVIES J...L. [MCW 57] MC WHORTER A. KIM S. Semiconductor Surface Physics.J. 1989. [JOI 96] JOINDOT I. vol. KIM D. December 1994. Fiber Optics Handbook. April 1997. 1996. Dunod. 5.F.J. . “Designing optoelectronic integrated circuit (OEIC) receivers for high sensitivity and maximally flat frequency response”.9. 1976. December 1993...A.K. Electronics noise and low noise design.. n° 4. BLAND S... Journal of Light Wave Technology. 1987. [KIM 97] KIM M. IEEE Photonics Technology Letters. [HEN 89] HENTSCHEL C.. “InGaAsP channel HFET’s on InP for OEIC applications”. Signaux.M.D. IEEE Transactions on Electron Devices. VALENZA M.. 12. n° 9. JOHN E. 39.B.J.. MINASIAN R. [BER 94] BERTHIER P.W. [PAR 93] PARK M..268 Optoelectronic Sensors 7.. RIGAUD. vol.5 μm... Macmillan New Electronics series. GIRAUDET L. Topics in Applied Physics. Transistors à effet de champ AlInAs/(Al)GaInAs(P) pour photodétection intégrée à 1. B83. Physica. September 1995.. systèmes linéaires et bruit en électronique. 1992. [BIQ 92] BIQUARD M. February 1993. CHEN J. “Determination of bit-rate and sensitivity limits of an optimized p-i-n/HBT OEIC receiver using SPICE simulations”. [KRE 93] KRESSEL H. Springer-Verlag. [FIS 93] FISH P. Photorécepteur intégré pour transmission cohérente sur fibre optique.. n° 2. University of Pennsylvania Press. 13. 40. vol.3-1. 1993.W.. University Thesis. [LEG 93] LEGROS E. 1957. “Ultralow noise 10Gb/s p-i-n-HEMT optical receiver”. Ellipses..B.. Orsay. “1/f noise”.S.. [HOO 76] HOOGE F.. PhD thesis.N. Philadelphia.I. [DAS 95] DAS M. n° 12. Montpellier University. vol. Journal of Light Wave Technology. vol.

List of Authors Robert ALABEDRA CEM2 University of Montpellier II France Baudoin DE CREMOUX Central research laboratory Thomson-CST France Didier DECOSTER IEMN Lille University of Science and Technology France Carmen GONZALEZ Alcatel Thales III-V Lab Marcoussis France Joseph HARARI IEMN Lille University of Science and Technology France Vincent MAGNIN IEMN Lille University of Science and Technology France Antoine MARTY LAAS-CNRS Toulouse France Eva MONROY INAC Nanophysics and Semiconductors Laboratory Grenoble France .

270 Optoelectronic Sensors Franck OMNES Neel Institute Grenoble France Dominique RIGAUD CEM2 University of Montpellier II France Gérard RIPOCHE Alcatel Research Centre Marcoussis France .

197. 164. 183. 52. 225. 93. 80. 211. 114. 226. 244. 125. 232. 205. 76. 259. 196. 202 IV-IV compound. 121. 12. 14. 14. 51. 245 coefficient. 2. 80. 58. 198. 205. 48. 123. 57. 214. 112. 69. 152. 122. 202. 201. 5. 205. 234. 78 B bandgap. 165. 33. 188. 72. 186. 47. 91. 233. 267 region. 66. 101. 187. 11. 97. 259 A absorption. 50. 31. 111. 106. 95. 150. 215. 184. 10. 91. 115. 122. 43. 78. 74. 227. 82. 118. 191. 127. 111. 73. 171. 1. 113. 131. 211. 146. 131. 5. 7. 121. 164. 60. 186. 195. 76. 115. 88. 3. 5 AlGaN. 64. 100. 100. 175. 20. 113.Index 1/f noise. 59. 214. 45. 43. 204. 119. 23. 182. 146. 125. 127. 74. 94. 36. 99. 231. 19. 209. 123. 112. 32. 187. 170. 75. 114. 198. 96. 61. 215 autocorrelation function. 54. 258. 169. 267 transistor. 230. 189. 53. 126. 99. 51. 232. 10. 233 avalanche. 166. 207. 208. 99. 238. 43. 45. 13. 28. 46. 202. 102. 76. 105. 116. 57. 21. 244 bandwidth. 52. 102. 42. 111. 62. 90. 115. 241. 103. 102. 95. 166. 206. 166. 256. 187. 28. 39. 19. 95. 223. 200. 232. 88. 59. 123. 211 acceptors. 45. 216. 208. 75. 46. 104. 62. 164. 210. 217. 257. 214. 148. . 193. 46. 69. 2 68. 88. 65. 93. 262 II-VI compound. 163. 80. 198. 257. 207. 27. 67. 47. 100. 165. 133. 98. 231. 2. 206. 119. 99. 75. 197. 96. 71. 198. 267 bipolar. 44. 82. 258. 144. 145. 212. 240. 39. 71. 245. 104. 191. 63. 199. 112. 90. 203. 111. 50. 14. 37. 204. 40. 114. 36. 119. 97. 171. 176. 49. 209. 2. 64. 101. 16. 227. 74. 83. 88. 87. 58. 29. 63. 112. 177. 105. 214. 229. 101 III-V compound. 206. 126. 240. 169. 189. 2. 3. 113. 257. 49. 65.

258. 156. 81. 163. 145.27