MILLIMETRE-WAVE GUNN DIODE TECHNOLOGY AND

APPLICATIONS
Nigel Priestley and Novak Farrington
e2v technologies- Electronic Devices and Systems Engineering Centre
Carholme Rd, Lincoln, LN1 2SF
Contact:
Nigel Priestley: nigel.priestley@e2v.com
Novak Farrington: novak.farrington@e2v.com
Abstract
Gunn diode oscillators have been used in military, commercial and industrial applications for the past
forty years. In the case of millimetre wave operation, the Gunn diode continues to offer at least equal
and often superior levels of performance to MMIC technology. This paper will discuss the hot electron
injection structures that have been developed over the past twenty years and more recent work that
enables the tailoring of device characteristics to match specific applications and operating
environments. The importance of well controlled device packaging and heat-sinking design will be
shown, with some results of the consequential reliability of Gunn diodes in production. The design of
the Gunn diode oscillator itself is crucial to gain optimum performance and long term reliability; some
mechanically tunable oscillator designs will be described before going on to present a patented novel
planar substrate oscillator developed for use at 77 and 94 GHz. Work undertaken to successfully
manufacture commercial 125GHz Gunn oscillators for millimetre-wave radar and imaging applications
will also be presented.
1. Introduction
The Gunn diode has long been the favoured solid-state device for coherent power generation at mmwave frequencies. Its low phase-noise and moderate output power levels make it ideal for many
RADAR and imaging applications. Although it is known that InP devices offer higher RF powers at
millimeter-wave frequencies than GaAs [1], the InP material system does not support the use of
graded-gap hot electron injection. This means the higher output powers and frequencies offered by
InP devices are achieved at the cost of temperature stability: a factor that is essential for many
applications.
State-of-the-art output powers and frequencies obtained using Gunn diode technology are illustrated
in Figure 1, along with the packaging and heatsinking materials used. The majority of these results
were achieved in the research laboratory and rely on the use of diamond heatsinking (which has
demonstrated output power increases of up to 200% in comparison to gold [2]), and quartz ring or
open packages [3]. The use of these techniques in a manufacturing environment would be extremely
difficult to implement while simultaneously maintaining overall yield, quality assurance, and allowing
integration of the devices into practical real-world rugged systems. All of the devices manufactured
commercially at e2v are based on the use of an integral gold heatsink and are ruggedly packaged
using an alumina ring. This proven technology has undergone stringent lifetime and reliability testing,
including that required by the automotive industrys quality standard, ISO/TS 16949:2002. Although,
from an electrical viewpoint, fused quartz rings offer superior performance due to a lower effective
permittivity (and therefore package parasitic capacitance), quartz suffers from its fragility and high
thermal expansion mismatch with the gold plated copper base (essential for heatsinking purposes)
potentially leading to mechanical failure and reducing operational lifespan.
In recent years alternative technologies offering higher levels of functionality have gradually become
available for power generation at millimeter-wave frequencies. Examples of this include SiGe MMICs
with operating frequencies now extending to above 77GHz [4], a wide range of commercially available
GaAs MMICs up to around 100GHz, W-band GaN MMICs demonstrated in the laboratory [5], and
commercially available InP devices up to 325GHz [6]. However, in terms of cost, power and simplicity,
the GaAs Gunn diode remains an attractive alternative. Making the technology even more attractive is

the recent development at e2v of the 125GHz diodes and oscillators shown in Figure 1 which, to the
knowledge of the authors, demonstrate the highest ever reported power at D-band for a GaAs Gunn
device [7]. The output power from this device is currently far in excess of that available using an offthe-shelf MMIC, while also exhibiting superior phase noise characteristics.
1000
1 denotes e2v technologies result

100
Frequency (GHz)

nic

0.1
30

o
rm

GaAs, InP
1, Alumina Ring, IHS
1, Alumina Ring, Diamond
1, Open Quartz, IHS
1, Open Quartz, Diamond
2, Alumina Ring, IHS
2, Quartz Ring, IHS
2, Open Quartz, Diamond
2, None, IHS
2, None, Diamond
3, Package unknown
3, None, IHS

ha

al
ent

n ic
mo
har

m
nda

10

d
2n

I nP

2nd
As
Ga

s fu
GaA

RF Output Power (mW)

100

with hot-electron injection

fu
nd
am
en
tal

Novak E. S. Farrington, e2v technologies, Lincoln, UK, September 2009

InP

400

Figure 1. Compilation of published state-of-the-art results between 30 and 400 GHz for GaAs and InP
Gunn diodes under CW operation. Legend format: mode of operation (1 denotes fundamental, 2
second-harmonic, etc.), package type, heatsink technology. Solid lines outline the highest powers and
frequencies achieved experimentally to-date from each material in fundamental and second-harmonic
mode.

2. Applications and operating environments

Although only power, frequency and phase noise have been considered so far, many applications
place stringent requirements on other performance characteristics such as turn-on voltage,
temperature drift (both RF power and frequency), and oscillator tuning characteristics. Some typical
applications for a Gunn diode oscillator will now be examined with relation to the performance
required of the diode itself. It is noted at this point that the following discussion does not aim to
represent an exhaustive list, but rather a selection taken to represent the extremes placed on device
performance.
Of particular interest at the present time is the use of Gunn diode oscillators as high-power, highfrequency, and low phase noise sources for use in passive millimeter-wave imaging systems. Here a
Gunn diode oscillator is often used to provide local oscillator (LO) power to the mixer when
heterodyne detection is employed. The diodes RF output may be applied directly to the mixer LO
port, or will sometimes be used as the pump for a frequency multiplication stage. Here the
requirements placed on the Gunn diode itself are relatively simple and are in effect similar to those of
the simple Doppler frequency velocity measuring devices used by law enforcement agencies for
measuring a motorists speed. Here, only sufficient RF output power at a fixed frequency, temperature
stability and satisfactory phase noise are required for correct operation of the system.
A conceptually more complicated class of system is that of millimetre-wave FMCW radar. This is
currently one of the main applications for the GaAs Gunn diodes manufactured by e2v which are used
in a wide variety of FMCW radar systems for applications such as automotive radar, missile guidance,
smart munitions, security radar, and runway foreign object and debris (FOD) detection radar.
Depending on the FMCW system architecture, the Gunn diode can be employed in a number of ways.
Placing the simplest performance requirements on a Gunn diode oscillator would be its use for the
generation of a carrier signal onto which a modulated signal is upconverted: an example of such a

system is shown in Figure 2. In this instance, as with fixed-frequency LO generation and Doppler
systems, the Gunn diode is required only to provide a constant power at a fixed frequency.
2-4GHz VCO
Linear sweep

91GHz Gunn oscillator

DSP

Figure 2. 94GHz FMCW radar imaging sensor using upconversion of the modulated signal to a 91GHz
carrier. (after [20]).

More commonly in an FMCM system, the Gunn oscillator will act as a voltage controlled oscillator
(VCO) with its output directly modulated either through the use of bias voltage pushing (direct tuning)
or through the use of varactor tuning. Direct tuning can typically achieve modulation bandwidths of
around 600MHz at V- and W-bands, while varactor tuning can achieve larger bandwidths of over
1000MHz at the expense of output power (typically 50% less than direct tuning). At this point it is
worth noting that although Gunn diodes are considered to be inherently broadband devices [8], the
resonant circuits on which the oscillator modules are generally based, restrict the overall bandwidth.
77.0

Vbias = 5.4V

4.0V
4.2V
4.4V
4.6V
4.8V
5.0V
5.2V
5.4V

76.5

76.0

Vbias = 4.0V

75.5

Frequency (GHz)

Frequency (GHz)

77.0

Vbias = 5.4V

76.5

76.0

Vbias = 4.0V

75.5

-40 -20

20

40

60

80 100

Temperature (DegC)

-40

-20

20

40

60

80

100

Temperature (DegC)

(a)
(b)
Figure 3. Measured frequency tuning characteristics over temperature of (a) an optimised GaAs Gunn
diode with graded AlGaAs hot electron injector, and (b) a similar diode with un-optimised epitaxial
structure.

In the case of a Gunn diode oscillator acting as a directly modulated VCO, not only does the oscillator
need to provide sufficient power across a specified frequency and ambient temperature range but its
voltage tuning characteristics must remain constant with temperature to avoid the introduction of nonlinearity into the FMCW chirp and subsequent degradation of the radar range resolution. An example
of this is shown in Figure 3 where the typical responses of optimised and un-optimised Gunn diode
epitaxial structures with hot electron injectors are shown. In the optimised case, the spacing between
the individual curves (which represent constant bias voltage) remains constant with temperature and
so the device shows minimal change in tuning characteristic (f/V) with temperature. The variation of
oscillation frequency with temperature in GaAs Gunn devices utilising hot electron injection has been
shown to be around 1/10th of a comparable InP device with no hot electron injection [7].
In all of the above applications, performance over temperature, phase-noise and turn-on voltage have
been shown to be key performance characteristics for the Gunn diode oscillator. In order to improve
on the performance characteristics of conventional Gunn diodes, the use of hot electron injection was
researched and developed in the late 1980s at GEC [9]-[13]. The basic operation of a Gunn diode will
be discussed later, along with the theory of operation, and device structure for a Gunn diode with a
graded bandgap hot electron injector.

3. Reliability
Besides meeting the operational requirements of a given system, the Gunn diode must also often
meet stringent automotive or military reliability requirements. In general the distribution of any
semiconductor device lifetime usually follows a lognormal pattern where initially the failure rate rises
rapidly due to infant failures arising from defects in the semiconductor material itself or packaging
issues. Once these infant failures have been accounted for, the failure rate drops and any further
performance degradation or device failure is generally due to intermetallic effects in the ohmic
contacts, diffusion of the dopant in the semiconductor itself, or package failure. All of these failure
mechanisms are to varying extents, related to thermal effects within the semiconductor device itself.
All devices manufactured at e2v go through an extensive screening process to identify infant failures.
This includes a hot-store, extended burn-in at elevated temperatures, automated optical inspection,
and multiple DC and RF electrical testing stages. Once the infant failures have been rooted out,
thermal management of the device is paramount to the long term reliability as the majority of the long
term device failure and degradation mechanisms will be accelerated with elevated temperatures.
Besides good heatsinking, the operational temperature of and self-heating effects in the device can
be controlled by the hot-electron injector structure; therefore providing long term reliability which can
be engineered into a device at the semiconductor level.
4. Device operation, epitaxial structure and material growth
Only a very brief description of Gunn diode operation is included here as it will be assumed that the
reader is familiar with the transferred electron (or Gunn) effect. A more comprehensive review is
beyond the scope of this paper and can be found in a number of texts and published articles [14],
[15]. The Gunn diode can, at its most basic, be thought of as a DC-to-RF converter: when an applied
bias voltage exceeds a certain threshold, oscillation will occur, the frequency of which will depend on
the material properties and the geometry of the device itself. This is due to the transferred electron
effect exhibited by certain binary and ternary compound semiconductors.
A brief, one paragraph explanation of the transferred electron effect could be that as an electron is
accelerated through the material by an electric field, it accumulates energy and so the probability of it
being scattered from the central conduction band valley to the nearest (in terms of energy and
momentum) satellite valley increases. This probability increases dramatically when the electric field
reaches a certain threshold value related to the distance, in energy and momentum, between the
conduction bands central valley and the nearest satellite valley. The scattering mechanism is
columbic in nature and involves a change in both momentum and energy and so the transfer process
is extremely inefficient with large amounts of energy being lost to phonon excitation (and elevation of
the lattice temperature).
A description of how the transferred electron effect results in oscillation (again in one paragraph)
could then be that the effective electron mass (related to the parabolicity of the conduction band
profile) is greater in the satellite valley than in the central valley. This means electron velocity in the
satellite valley due to an applied electric field is lower than that in the central valley. The effect of this
is a bunching of electrons as they travel along the length of the device, with those in the satellite
valley moving at a slower rate than those in the central valley. This leads to a depleted region
followed by a region of accumulation (known collectively as a high-field domain), which sets up an
internal electric field opposing that applied externally. The effect of this is a continuing reduction in the
net electric field across the device which continues until the threshold electric field is crossed
(preventing the nucleation of any further high-field domains). At this point the high-field domain stops
expanding and propagates at a constant velocity through the device until it reaches the cathode
contact layer at which point it collapses and the electric field through the device increases again.
Oscillation is observed in the current and voltage at the device terminals as the domain forms at the
anode, expands, propagates and collapses at the cathode, before another domain nucleates at the
anode.
The fundamental physics behind conventional Gunn device operation are very much centred on
electron energy, lattice temperature, and random scattering processes. These give rise to the
following inherent limitations of conventional Gunn devices:

The onset of oscillation (turn-on voltage) in a Gunn device varies greatly with ambient
temperature.
The oscillation frequency is highly dependant on temperature as it is defined (in part) by the
length of the region in which the domain propagates (transit region). A small portion (referred
to as the dead zone) of this region is required to accelerate electrons to the point at which
they have sufficient energy to transfer to the satellite valley. The length of the dead zone, and
therefore the effective transit region length and oscillating frequency are also dependant on
temperature.
The random nature of the scattering mechanisms which facilitate domain nucleation, lead to
small variations in the point at which domain nucleation occurs. This effectively varies the
length of the transit region from cycle-to-cycle leading to phase noise at the output.

In order to address these issues the use of a hot electron injection structure was proposed and
demonstrated in the 1980s by GEC engineers [9]-[13]. The reasoning behind this was to raise the
electron energy sufficiently to greatly increase the probability of their direct entry to the conduction
band satellite valley upon delivery to the transit region. The injected electrons have greater energy
than those at equilibrium with the transit region lattice, and so are referred to as hot. With the
majority of the injected electrons directly entering the satellite valley, the dead-zone is effectively
eliminated along with the temperature dependency of the oscillation frequency and turn-on voltage. In
addition the overall conversion efficiency is increased (due to the reduced parasitic positive
resistance) and because the random factors of domain nucleation are reduced, the phase noise is
decreased.

0.5 m
10nm

18

10nm
<50nm

GaAs n=510
GaAs undoped
Al0.017Ga0.983As
to Al0.32Ga0.68As
GaAs undoped
18
GaAs n=110

<2 m

GaAs n=210

16

0.5 m
10 m

GaAs n=510
+
n GaAs

18

50nm

Cathode
Contact layer
Spacer
Step-graded
launcher
Spacer
Doping spike

Transit
region

Buffer
Substrate
Anode

Conduction Band

The epitaxial structure of a conventional GaAs Gunn device is generally a lightly doped n- transit
layer, sandwiched between two heavily doped n+ contact layers (upon which the ohmic contacts are
formed). The epitaxial structure of a device with hot electron injection, optimised to have a freerunning fundamental oscillation frequency of around 45GHz is shown in Figure 4. It has the same
n+n-n+ structure as a conventional Gunn device but with the inclusion of a hot electron injector
between the n+ GaAs cathode contact layer and the n- GaAs transit region.

eV

Figure 4. Epitaxial structure and conduction band diagram for a 45GHz fundamental frequency GaAs
Gunn diode with step-graded AlxGa1-xAs launcher. (Note: not to scale).

The injector itself consists of two components: a graded-gap launcher and thin (<50nm) n+ doping
spike. The doping spike situated between the launcher and the transit region is required to control
the electric field at the start of the transit region while retaining the hot electron properties. Its
thickness is therefore restricted to make probable the ballistic transport of the injected electrons
through it. In its absence a depletion region is formed behind the launcher in forward bias that inhibits
high-field domain nucleation and acts as a dead zone [10], [11]. The graded-gap launcher itself can
have a continuous or discreet profile. The device structure illustrated in Figure 4 contains five layers
with a discrete increase in aluminium content in each towards the anode. Originally the stepped
injector structure was introduced to improve repeatability in the growth of the epitaxial structure
compared with the linear variant

A 10nm nominally undoped GaAs spacer layer is included at each side of the launcher to prevent
dopant diffusion into, and thus degradation of, the launcher [16]. The extremely tight control required
over the thickness and composition of each epitaxial layer requires the growth be performed using
molecular beam epitaxy (MBE). For this type of Gunn device, growth is carried out on n+ GaAs
substrates, around 10m of which will remain as part of the device structure after the fabrication
process. This is to prevent damage or strain in the epitaxial layers during the bonding and packaging
process. However, the amount of substrate material included in the device will generally be kept to a
minimum to reduce the parasitic positive resistance it introduces.
5. Fabrication and packaging for device reliability
As mentioned in earlier, the inter-valley scattering process which is the basis of the transferred
electron effect is extremely inefficient, and the majority of the applied energy is dissipated as heat. It
has been calculated that the power density in a GaAs Gunn diode is ~ 120kW.cm-2 [16] and so for
reliable long term operation, good heatsinking is essential. In order to achieve this, a fabrication
process which supports the formation of an electroplated gold integral heatsink is used, the results of
which are shown in Figure 5.

Au plated anode contact (>6m)

+

n substrate (~10m)
+
n buffer (500nm)
n transit region (<2m)
Injector (<100nm)

n contact
layer
(500nm)

Integral gold heatsink

(a)

(b)

Figure 5. (a) SEM of a fabricated GaAs Gunn diode with integral gold heatsink, and (b) schematic
representation of a GaAs device with electroplated HIS (not to scale).

Preformed gold
cross

Ceramic ring
Gold plated
lid

Alumina ring
housing

Die

Gold plated
copper

Package body

(a)

(b)

Figure 6. GaAs Gunn diode package: (a) micrograph of alumina ring package cross-section (without lid)
for a 77GHz ACC device. ) and (b) schematic of whole package.

Once fabricated, the devices are thermosonically bonded to a commercially available package
consisting of a partially metallised alumina ring (>250m thick) brazed onto a threaded gold-plated
copper stud (Figure 6b). The thickness of the alumina ring is kept to a minimum to reduce as far as
possible the standoff distance between the oscillator circuit resonant disk and the metal plane of the
package top surface. Connection between the top of the alumina ring and the devices top contact is

achieved by thermocompression bonding the centre of a pre-fabricated gold cross to the electroplated
gold on the device anode, and then bonding the four arms of the cross to the metallised region on the
top of the alumina ring (Figure 6). The package is then hermetically sealed by soldering a circular
metallised lid onto the alumina ring.
It is noted here that the fabrication and packaging processes are optimised for yield and device
reliability. For example bonding equipment which provides a calibrated, reproducible force in the zaxis is used to minimize risk of damage to the semiconductor itself during the bonding process. The
thickness and profile of the pre-fabricated gold cross are also optimised for longevity, particularly if
pulsed operation (and high repetition of the heating and cooling cycle) will be required by the
customer.

6. Oscillator design and performance

In the discussions so far, the oscillation frequency of a Gunn diode has been assumed to be the free
running value, solely due to the dynamics of domain formation and propagation in the transit region,
and free from the effects of an external circuit. In practice, a device needs to be attached to a circuit
that provides DC power, and couples out RF power: the circuit will have reactive elements and a
resonant frequency associated with it. The circuit can therefore be designed to resonate with the
Gunn diode in order to optimise output power and efficiency at a given frequency. At millimeter-wave
frequencies the most common circuit configuration used to do this is the waveguide-based secondharmonic resonant disk oscillator. A schematic diagram of the cross-section of a second harmonic
resonant disk oscillator is shown in Figure 7.

Choke
Sliding back-short

Bias post

Resonant disk
WG-28 waveguide

Screw-in diode
mount

Diode
Diode
package

Figure 7. Simplified schematic cross-section of a generic waveguide-based second harmonic resonant

disk oscillator circuit showing all mechanically tunable components.

In a resonant disk oscillator the packaged diode will typically be positioned in the waveguide floor and
the DC power for the diode is supplied through the bias post and choke assembly (often a single
piece of machined metal), which is insulated from the waveguide block. The frequency of oscillation is
coarsely determined by a combination of the resonant disk diameter, post width, and the length of the
section of post between the resonant disk and the bottom of the lowest choke section. The
operational principle is such that three distinct electromagnetic modes of propagation exist [17]: a
TE01 rectangular waveguide mode (in the rectangular waveguide section), a radial mode (between the
resonant disk and the waveguide floor), and a quasi-coaxial (TEM) mode (between the resonant disk
and the bottom of the choke). Two separate resonances are therefore supported: a fundamental
frequency resonance in the TEM mode along the quasi-coaxial line (which is inductive) between the
choke and the disk (a capacitor), and a resonance at the second harmonic between the disk and the
waveguide floor. In order to obtain maximum power at the output, the two resonances must be
synchronized (through optimisation of the disk diameter and post/choke geometry), and the coupling
between the radial mode and the magnetic field of the TE01 waveguide mode maximized (through
optimisation of the disk diameter, distance between the disk and waveguide floor, and backshort
distance).
Figure 8 shows an Ansoft High Frequency Structure Simulator (HFSS) model of a second harmonic
resonant disk oscillator configuration using a circular waveguide back short and frequency tuning pin.
Although not showing the quasi-coaxial TEM mode and its resonance (the plot illustrates the electric

field at second harmonic frequencies, not fundamental), Figure 8 shows power being coupled
between the diode and the waveguide output (left), and a degree of interaction between the resonant
disk and the waveguide floor.

(a)
(b)
Figure 8. Ansoft HFSS model of a second harmonic resonant disk millimetre-wave oscillator with tuning
pin and circular waveguide sliding backshort, (b) simulated field electric field plot for the oscillator at
second harmonic frequencies.

Although this type of oscillator is relatively narrowband, it is noted that wideband mechanically tunable
(Carlstrom) oscillators [8], [18] can be achieved through the use of a choke section which contacts
with and slides up the bias post. This can be used to increase the resonant disk-choke distance and
therefore adjust the oscillation frequency. However, because the resonant disk diameter is not
adjustable, the resonances are only truly synchronised for one post length at which a peak in power is
obtained: before and after this the power tails off.
Although the Carlstrom oscillator can provide relatively wideband tuning, the tuning is performed
mechanically which makes it unsuitable for many applications outside of the laboratory. If practical
wideband tuning is required then a varactor-tuned design can be used: this again is typically
implemented in waveguide. Here, instead of the resonant frequency being determined by the
geometry of the circuit elements, the tuning of a varactor mounted in the waveguide along with the
Gunn diode can be used to alter the frequency. This however comes at the price of output power and
overall efficiency.
One of the main perceived limitations of millimeter-wave Gunn diode technology is often thought to be
the requirement for waveguide technology, particularly if high output powers are required. However,
waveguide-based Gunn oscillators can be mechanically designed for mounting on planar microstrip
circuits with either a stepped impedance transformer, or a microstrip probe coupled transition between
the planar circuit and the waveguide (depending on the required connection angle between the
waveguide and the planar line). Such arrangements have shown insertion losses as low as 0.5dB at
94GHz.
To achieve a true planar Gunn diode VCO, a novel, patented concept has been developed and
successfully demonstrated at e2v [19]. Suitable for high-volume production the design features
relatively wideband varactor tuning and offers a solution where the key drivers of size, cost,
performance and versatility may be satisfied simultaneously. This solution was shown to give ~ 30mW
at 77GHz, which is about half the power produced by an equivalent rectangular waveguide cavity
oscillator. Figure 9(a) shows a solid model of a hermetically packaged planar substrate VCO, while
Figure 9(b) gives an idea of the module size, compared to a regular rectangular waveguide cavity
VCO.

(a)
(b)
Figure 9. (a) Solid model illustration of VCO package. (WG output on lower surface), and (b) Relative
sizes, mm/inch, of VCO's. a) Cavity type, b) new VCO with microstrip output, c) new VCO with waveguide
output.

7. Development of D-band GaAs Gunn diode technology

The benefits of GaAs technology over InP have already been discussed and were the driving force for
the development of 125GHz GaAs Gunn devices and oscillators at e2v. These devices were
developed to meet an increasing demand for solid-state sources at higher frequencies. The MBE
growth of several wafers was performed along with device fabrication test and oscillator refinement.
The result of the work, shown in Figure 1, was the demonstration of a record RF output power for a
GaAs device when 40mW was measured at 122GHz [7].
DC Input Power (W)
2.2

2.4

2.6

2.8

3.2

3.4

3.6

3.8

123

50

122

120

40

1.15%

35
0.9%

30
25
20
15
3.5

4.0

4.5

Bias Voltage (V)

5.0

5.5

48

121.56

Frequency (GHz)

121

50

121.58

46

~38MHz

121.54

44

121.52

42

121.50
40

121.48

38

121.46
121.44

36

121.42

34

121.40
-10

Output Power (mW)

RF Output Power (mW)

45

Oscillation Frequency (GHz)

1.4%

121.60

32
0

10

20

30

40

50

60

70

80

Temperature (Degrees Celsius)

(a)
(b)
Figure 10. Performance characteristics of the D-band Gunn diodes reported in [7]: (a) RF output power
and frequency variation with bias voltage (dashed lines represent lines of constant efficiency), and (b)
variation with temperature.

It is seen in Figure 10 that although overall efficiency is low (as was expected from a second
harmonic device), impressive power levels were achieved, especially considering the use of gold as a
heatsinking material, and alumina as the packaging material. The average RF power from a batch of
10 diodes was around 32mW [7] which is sufficient for many applications. As such these devices
represent the first ruggedly packaged, commercially viable GaAs source of D-band power.
8. Conclusions
This paper has reviewed the state-of-the art in Gunn diode technology before going on to review a
variety of applications into which e2v commercially supply devices. The demands such applications
place upon the oscillator and diode itself have been discussed in detail, along with the fundamental
physical mechanisms behind device operation. This framed the argument for the use of devices
employing hot electron injection, the epitaxial semiconductor structure for which was presented.

The operation of, and performance trade-offs, between a number of different implementations of
Gunn VCOs developed at e2v technologies, has been discussed. The paper has also outlined and
presented results from an ongoing programme in the optimisation of packaged, high-performance Dband hot electron injected GaAs Gunn diodes, suited for volume manufacture.
Acknowledgments
The efforts of all the engineers and scientists from the following organisations who contributed to the
development and advancement of graded-gap hot electron injection Gunn diode technology over the
last two decades are acknowledged: the GEC Hurst Research Centre, Marconi Electronic Devices
Ltd., GEC Plessy Semiconductors, EEV and e2v technologies. In particular Mike Carr, Phil Norton,
Ian Dale, John Stevens, John Bird, and Mike Kelly. Professor Mohamed Missous from the University
of Manchester (formally UMIST) is also acknowledged for MBE growth of wafers.
References
[1] B. K. Ridley, Anatomy of the transferred-electron effect in III-V semiconductors, J. Applied Phys., vol. 48,
no. 2, pp. 754-764, February 1995.
[2] H. Eisele and G. I. Haddad, High-performance InP Gunn devices for fundamental-mode operation in Dband, IEEE Microwave and Guided Wave Lett., vol. 5, no. 11, pp. 385-387, November 1995.
[3] H. Eisele and R. Kamoua, Submillimeter-wave InP Gunn devices, IEEE Trans. Microwave Theory Tech.,
vol. 52, no. 10, pp. 2371-2378, October 2004.
[4] Infineon SiGe IC used in Bosch automotive radar system, Semiconductor Today, 2nd Dec. 2008.
[5] M. Micovic, et. Al, W-Band GaN MMIC with 842 mW Output Power at 88 GHz, IEEE Int. Microwave
Symposium Dig., IMS 2010, Anaheim, CA, May 2010.
[6] Fraunhofer IAF
[7] N. Farrington, P. Norton, M. Carr, J. Sly, and M. Missous, A ruggedly packaged D-Band GaAs Gunn diode
with hot electron injection suitable for volume manufacture, IEEE Int. Microwave Symposium Dig., IMS
2008, Atlanta, pp. 281-284, June 2008.
[8] J. E. Carlstrom, R. L. Plambeck, and D. D. Thornton, A continuously tunable 65-115-GHz Gunn oscillator,
IEEE Trans. Microwave Theory Tech., vol. 33, no. 7, pp. 610-619, July 1985.
[9] N. R. Couch, H. Spooner, P. H. Beton, M. J. Kelly, M. E. Lee, P. K. Rees, and T. M. Kerr, Highperformance, graded AlGaAs injector, GaAs Gunn diodes at 94 GHz, IEEE Electron. Device Lett., vol. 10,
no. 7, pp. 288-290, July 1989.
[10] N. R. Couch, M. J. Kelly, H. Spooner, and T. M. Kerr, Hot electron injection in millimeter wave Gunn
diodes, Solid-state Electron., vol. 32, no. 12, pp. 1685-1688, December 1989.
[11] P. H. Betton, A. P. Long, N. R. Couch, and M. J. Kelly, Use of n+ spike doping regions as nonequilibrium
connectors, Electron. Lett., vol. 24, no. 7, pp. 434-435, March 1988.
[12] S. Neylon, I. Dale, H. Spooner, D. Worley, N. Couch, D. Night, and J. Ondria, Sate-of-the-art performance
millimeter wave gallium arsenide Gunn diodes using ballistically hot electron injectors, in IEEE MTT-S Int.
Microwave Sypm. Dig., Long Beach, CA, USA, June 13 - 15, 1989. pp. 519-522.
[13] H. Spooner, and N. R. Couch, Advances in hot electron injector Gunn diodes, GEC J. Research., vol. 7, pp.
34-45, 1989.
[14] R. van Zyl, W. Perold, and R. Botha, The Gunn diode: Fundamentals and fabrication, Proc. Of the 1998
South African Symposium on Communications and Signal Processing, COMSIG98, pp. 407-412, Sept.
1998.
[15] P. J. Bullman, G. S. Hobson, and B. C. Taylor, Transferred electron devices, Academic Press Inc., 1972.
[16] A. Frster, M. I. Lepsa, D. Freundt, J. Stock, and S. Montarani, Hot electron injector Gunn diode for
advanced driver assistance systems, Applied Physics A, vol. 87, no. 3, pp. 545-558, June 2007.
[17] H. Barth, and W. Menzel, GaAs Gunn oscillators reach the 140GHz range, IEEE Int. Microwave Symposium
Dig., IMS 1985, pp. 367-369, June 1985.
[18] S. H. Jones, M. F. Zybura, J. E. Carlstrom, and T. M. OBrian, A 63-170 GHz second-harmonic operation of
an InP transferred electron device, IEEE Trans. Electron Dev., vol. 46, no. 1, pp. 17-23, January 1999.
[19] N. Priestley, K. Newsome, I. Dale, and P. Norton, A Gunn diode based surface mount 77GHz oscillator for
automotive applications, IEEE Int. Microwave Symposium Dig., IMS 2002, pp. 1863-1866, June 2002.
[20] Y-W. Chang, and M. Johnson, Portable concealed weapon detection using millimetre-wave FMCW radar
imaging, Final report for the National institute of justice office of science and technology, 2001.

10