Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

Proton Degradation of Light-Emitting Diodes†

A. H. Johnston, B. G. Rax, L. E. Selva and C. E. Barnes
Jet Propulsion Laboratory
California Institute of Technology
Pasadena, California

Abstract 1
Proton degradation was investigated for several types of 850
light-emitting diodes with wavelengths in the near infrared

Light Output (Normalized)

region. All irradiations were done with 50-MeV protons. 660
Several basic light-emitting diode (LED) technologies were
compared, including homojunction and double-heterojunction 825
devices. Homojunction LEDs fabricated with amphoteric 0.1
dopants were far more sensitive to displacement damage than
double-heterojunction LEDs, and were strongly affected by 50 MeV Protons
injection-enhanced annealing. Unit-to-unit variability remains IF = 5 mA
an important issue for all LED technologies. For some
technologies, degradation of the forward voltage
characteristics appears to be more significant than degradation
of light output. 880
0.01
1x109 1x1010 1x1011 1x1012
Proton Fluence (p/cm2)

I. INTRODUCTION Figure 1. Degradation of LEDs of different wavelengths from

50 MeV protons
The severe degradation of some types of optocouplers in
space has been shown to be mainly due to proton displacement II. LED TECHNOLOGIES
damage in the light-emitting diodes that are used within the
optocouplers [1,2]. A variety of LED technologies can be A. Diffused LEDs
used in optocouplers and their sensitivity to proton
displacement damage varies by about two orders of Two very different approaches can be used to fabricate
magnitude, as shown in Figure 1. The data at 880 nm are from light-emitting diodes. Categorizing LEDs by the basic type of
LEDs used by Optek in the 4N49 optocoupler [1]. The other fabrication process makes sense from two standpoints: first,
data are for LED evaluation samples provided by Micropac the physical details of the underlying structure are very
(from unspecified manufacturers) that are intended for use in different; and second, the evidence to date suggests that LEDs
optocouplers. Optocouplers are very simple hybrid devices, made with these processes have widely differing sensitivities
and the type of LED can be readily changed by the to radiation damage [6].
manufacturer with little cost impact. Many optocoupler Many infrared LEDs are fabricated with an older
manufacturers purchase LEDs from outside sources with little processing technique using liquid-phase epitaxy with an
knowledge or control of the manufacturing process used for amphoteric dopant (silicon). This process relies on the fact
the LED, leading to the possibility of very dramatic that the impurity is an n-type dopant during epitaxial growth at
differences in radiation response (JPL has observed such high temperature, but changes to a p-type dopant when the
differences for one type of optocoupler that is used in a hybrid growth is done below a critical temperature. It is possible to
power converter). form a p-n junction by gradually changing the temperature
Increased understanding of LED degradation is needed not during epitaxial growth using the natural transition of the
only because of their use in optocouplers, but also for basic impurity from n- to p-type to create the junction. The
applications of LEDs in optoelectronic systems. This paper resulting structure is shown in Figure 2. It is a simple process
investigates displacement damage in near-IR light-emitting because only a single dopant is required. GaAs and AlGaAs
diodes. Earlier work showed that amphoterically doped LEDs LEDs can be made with this process. The fabrication
were sensitive to proton irradiation [3-5]. Present-day technique produces junctions with a graded impurity profile
devices degrade even more than the devices studied at that over a relatively wide region (50 to 100 µm). GaAs LEDs
time. typically have a wavelength of 930 nm. The wavelength of
AlGaAs devices depends on the aluminum concentration at the
--------- junction and can range from 830 to 900 nm. AlGaAs provides
†The research in this paper was carried out by the Jet Propulsion approximately twice as much light output as GaAs. These
Laboratory, California Institute of Technology, under contract with structures are typically edge emitting. The gradual transition
the National Aeronautics and Space Administration, Code AE, under from n- to p-impurity results in a low doping density at the
the NASA Microelectronics Space Radiation Effects Program junction; this in turn limits the frequency response to
(MSREP).
approximately 1-10 MHz [5].

1
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

Temperature III. EXPERIMENTAL APPROACH

change during Anode
growth creates p-n junction Several different device types were selected for this study,
junction as shown in Table 1. They include simple, low-cost diffused
p -GaAs (Si doped)
Light LEDs (amphoterically doped) as well as double-heterojunction
emission
LEDs. Some of the LEDs are identical to LEDs contained
n (regrowth) within various types of optocouplers. The 700 nm LEDs from
Hewlett-Packard (made with a conventional diffusion process)
are not available commercially other than as part of a complete
optocoupler assembly. The devices that were tested were
n - GaAs (Si doped) removed from 6N140 optocouplers. All of the other LED
types are available commercially as discrete LEDs. They all
have a maximum DC operating current of 100 mA, although
most space applications will derate the operating current
Cathode
substantially because the light output of LEDs degrades with
time under high-current operation [9]. That wearout
Figure 2. Diagram of an Amphoterically Doped LED mechanism has no counterpart in silicon technology.
It is also possible to form LEDs by conventional diffusion Table 1. LED Technologies Investigated in this Study
processes, using dopants of different types in the two
junctions. Diffused LEDs have more abrupt doping profiles λ (nm) Construction Material Manuf.
Response
than those made with amphoteric doping. They have Time (ns)
improved frequency response, but diffused LEDs in the near 930 * Diffused (Si) GaAs Optek 1000
infrared region generally have less light output than 890 * Diffused (Si) AlGaAs Optek 500
comparable LEDs manufactured with amphoteric dopants. 880 * Diffused (Si) AlGaAs Optodiode 500
850 Heterojunction AlGaAs Optek 6
B. Heterojunction LEDs 800 Heterojunction GaAs Optodiode 60
700 Diffused (Zn) GaAsP HP <40
More sophisticated processes can be use to form double
heterojunction structures that confine the direct-gap active
region by locating layers with higher bandgap on either side * Amphoterically Doped
[7,8]. These layers are a different type of semiconductor than
the semiconductor in the active region, and form
heterojunctions on either side. Figure 3 shows a representative Absolute comparison of the light output of different LED
structure for an edge-emitting double-heterojunction LED. types is difficult because the high index of refraction of GaAs
The layers in these structures are very thin, on the order of a (and AlGaAs) causes large Fresnel losses when the device is
few µm compared to the regions present in amphoterically coupled to a medium with low refractive index. The amount
doped LEDs, typically 50-100 µm. Processing for of light that is actually transmitted depends on physical details,
heterojunction LEDs requires well controlled growth including the angle over which the light is accepted and the
techniques, involving many more steps than the processes used properties of coatings and index matching materials. For this
for diffused LEDs. Heterojunction LEDs have more abrupt reason, the data in the paper are normalized to the initial light
doping profiles than amphoterically doped homojunction output measured by a silicon photodetector with an acceptance
LEDs, with much faster response time. angle of approximately ±20 degrees.
Initial radiation tests on a small sample of devices were
n-electrode
done for all of the LED types. More thorough testing with
large sample sizes was done for two different device types
n-GaAs from Optodiode, Inc. which manufactures high-reliability
LEDs that are space qualified as well as for two different types
AlGaAs of LEDs from Optek, a second major manufacturer of LEDs.
guiding layer
The wavelength listed in Table 1 is the nominal value for
Active layer the particular device type. There are substantial variations in
n-AlGaAs
(GaAs) wavelength even within specific lots of LEDs, particularly for
Light
the amphoterically doped devices. The wavelength of
emission individual LEDs was measured with a spectrometer before and
p-AlGaAs
after irradiation. The spectrometer resolution was less than 1
guiding layer nm. Typical spectral width was 50-70 nm, depending on the
p-electrode type of LED.
p-GaAs
Proton testing was done at UC Davis using 50 MeV
Figure 3. Diagram of a Double-Heterojunction LED protons. Irradiations were done in steps of approximately
1-3 x 1010 p/cm2. The beam intensity was varied so that each
irradiation step required about 5 minutes to complete.

2
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

Devices were removed from the irradiation area after each In contrast, the OD800 heterojunction devices exhibited
incremental irradiation step and tested over a range of little or no dependence on bias conditions during irradiation.
operating currents using a phototransistor (connected as a Note also that damage in the OD800s was about the same at
photodiode) to measure the light output. A Keithley low and high measurement currents, whereas significantly less
microammeter was used to measure the detector photocurrent. damage -- approximately 30%, depending on the fluence level
Measurements of the forward-biased diode characteristics -- occurred in the OD880s when they were measured at high
were also made before and after each irradiation step. forward current.
A special fixture was fabricated to allow twelve devices to
be irradiated simultaneously. Test devices were soldered to a Figure 4. Degradation of two Optodiode LED technologies
board in a circular pattern. A matching array of photodiodes measured at a forward current of 1 mA.
was used for optical measurements between irradiations (the
photodiodes were not irradiated). A spacer block -- carefully 1
machined to line up with the photodiode and LED arrays --
was used to provide a consistent way to align the LEDs and OD800 (DH)

Light Output (Normalized)

0.8
respective photodiodes, clamping the two assemblies so that
the spacing between the LED and photodiode was Solid lines: Device off during irradiation
0.6
reproducible. Holes in the spacer block were slightly greater Dashed lines: Device biased near maximum
than the diameter of the LEDs and photodiodes. operating current during
irradiation
0.4
Each measurement sequence could be completed in less
than 5 minutes. Different groups of devices were irradiated OD880 (Diffused)
under different forward bias conditions. Some were unbiased, 0.2

while others were irradiated at a fixed operating current. The

highest current used was near the maximum typical operating 0
0 5x1010 1x1011 1.5x1011 2x1011
current. Measurement currents for the unbiased devices were
restricted to low values because of the possibility of current- Proton Fluence (p/cm2)
enhanced annealing. Figure 5. Degradation of two Optodiode LED technologies
measured at a forward current of 37.5 mA.
IV. EXPERIMENTAL RESULTS
Although it appears from these results that the 800 nm
A. Devices Produced by Optodiode double heterojunction LEDs would be a better choice for
Optodiode Incorprated is a major manufacturer of high- space applications because of the lower radiation degradation,
reliability LEDs for aerospace and other critical applications. the 800 nm LEDs are much less efficient, producing only
Although that company does not produce optocouplers, they about 15% of the optical power of the 880 nm devices prior to
manufacture LEDs with both diffused and heterojunction irradiation. Thus, although the 800 nm LEDs are, on average,
structures for the aerospace market. Initial work was done much less affected by radiation, their reduced initial light
comparing two basic technologies from Optodiode because output must also be considered in device selection
they were being used in JPL systems. Another important issue is the uniformity of the radiation
Comparisons of the radiation degradation of diffused and response. More than 80 of the 880 nm devices were subjected
double-heterojunction devices manufactured by Optodiode are to radiation under various bias conditions, and although
shown in Figures 4 and 5. The solid lines show mean values differences of approximately a factor of two occurred in the
for 12 unbiased devices, while the dashed lines show mean relative degradation of the best and worst device from the total
values for 12 devices that were biased at 37.5 mA during population, none of the devices behaved in an abnormal way.
irradiation (results for intermediate forward bias currents fell Figure 6 shows the variability of the radiation response of a
between these two limiting curves). Degradation of the group of 12 OD880 devices, irradiated without bias and
OD880 (diffused, amphoterically doped technology) depended measured at low forward current. At the lowest level the light
strongly on operating current. Significantly less degradation output ranges from just over 50% to nearly 80% of the initial
was observed for devices that were biased at high current light output. The mean value of devices from a second lot are
during irradiation compared to unbiased devices or devices also shown in the figure. The mean value of the second lot
biased at currents of a few milliamps. was nearly identical to the worst devices in the first lot.

OD800 (DH)
Light Output (Normalized)

0.8

0.6 Solid lines: Device off during irradiation

Dashed lines: Device biased near maximum
operating current during
0.4 irradiation

OD880 (Diffused)
0.2

0 3
0 5x1010 1x1011 1.5x1011 2x1011
Proton Fluence (p/cm2)
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

1 1
Best, worst and mean of 12 devices
Unbiased during irradiation
0.8 M ean of 15 parts
Measured at IF = 1 mA 0.8

LIGHT OUTPUT (NORMALIZED)

0.6 Mean from first lot Worst of 17 devices
(dashed lines show best and worst devices) 0.6

0.4
Mean from second lot 0.4

0.2
Solid lines: 5 x 1010 p/cm 2
0.2
Dashed lines: 8 x 1010 p/cm 2
0
0 2 x 1010 4 x 1010 6 x 1010 1 x 1011
8 x 1010 0
Protons/cm2 0 20 40 60 80 100 120 140 160
LED CURRENT (mA)
Figure 6. Range of damage exhibited by a group of 12 OD880
LEDs, irradiated with no bias and measured at a forward current of 1 Figure 7. Dependence of normalized output power on forward
mA. The mean value of a second lot of six devices is also shown. current for the OD800 LED showing the behavior of an abnormal
unit.
Optodiode LEDs from the double-heterojunction process
exhibited much more variability than devices from the and a double-heterojunction AlGaAs device (850 nm). The
amphoterically doped process. Two of the 800 nm devices latter device is intended for a different class of applications,
degraded quite differently from the majority of the devices in and has a much higher bandwidth than the other LED
the test sample (a total of 17 devices). An example is shown technologies that were evaluated from that manufacturer.
in Figure 7. Initially all devices worked satisfactorily even at Degradation of the amphoterically doped AlGaAs LEDs
very low forward currents (1 mA). However, after the first from Optek are shown in Figure 8 (mean values are shown).
radiation level the minimum current for operation (effectively The magnitude of the degradation is very similar to that of the
a threshold current) increased to about 10 mA for one device. amphoteric AlGaAs LEDs from Optodiode (see Figure 4).
Its light output was far lower than that of typical devices from Both device types show 20-30% less degradation when they
the group until the forward current was increased to about 40 are biased at high operating current during irradiation.
mA. This threshold current continued to increase at higher
radiation levels, as shown in the figure. At low current the
1.0
forward voltage after irradiation was nearly 0.5 V lower than
the forward voltage of devices that behaved normally.
Optek OD233
Forward voltage of typical devices from the group changed by 0.8
Light Output (normalized)

AlGaAs - 890 nm
less than 20 mV. A second unit from the population also measured at 1 mA
behaved abnormally, with similar characteristics. The extreme
damage that occurred for those two parts at low currents was 0.6
unaffected by operating current, and appeared to be a stable
condition that did not recover after irradiation. Six months
0.4
after they were irradiated, the I-V characteristics were
identical to the results obtained just after they were irradiated. biased at 37.5 mA
Results for these two types of LEDs from Optodiode 0.2
illustrate that variability in radiation response can be very unbiased
important for some types of LEDs. This must be taken into 0
account in planning radiation tests and in interpreting radiation 0 2x1010 4x1010 6x1010 8x1010 1x1011
test data. The two part technologies are intended for the same Proton Fluence (p/cm2)
types of basic applications. One technology performs much
better, on average, in a radiation environment, but appears to Figure 8. Degradation of AlGaAs LEDs from Optek (amphoterically
doped).
have far greater unit-to-unit variability. Potential hardness
assurance tools for identifying devices with abnormal response Degradation of the Optek GaAs LEDs, which have a
are discussed in Section VII of the paper. slightly longer wavelength, are shown in Figure 9. The
magnitude of the degradation is very similar to that of the two
other types of amphoterically doped LEDs. However, the
B. Devices Fabricated by Optek and Hewlett- GaAs LEDs show somewhat less dependence on bias
Packard conditions than the AlGaAs LEDs from either Optek or
Optek is a major manufacturer of optoelectronic devices, Optodiode. The optical efficiency of GaAs LEDs is
producing discrete light-emitting diodes as well as approximately a factor of two lower than that of AlGaAs
optocouplers. As shown in Table 1, three types of LEDs from LEDs, largely because of the longer wavelength which affects
Optek were evaluated: diffused, amphoterically doped GaAs internal absorption and coupling efficiency. Thus, AlGaAs
(930 nm); diffused amphoterically doped AlGaAs (890 nm) LEDs are probably a better choice for most applications.

4
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

1.0 manufacturer that are used in most of their optocouplers with

the other types of LEDs. The Hewlett-Packard LEDs use
GaAsP and have a shorter wavelength (700 nm ) than the other
Light Output (normalized)

0.8 Optek OD130

GaAs - 930 nm LED technologies. They are fabricated with a conventional
Measured at I F = mA diffused process using zinc as a dopant.
0.6
Degradation of the 700 nm GaAsP devices from Hewlett-
Packard is shown in Figure 11. The degradation of that device
0.4
type was only slightly affected by bias conditions. The HP
biased at 37.5 mA devices are more than an order of magnitude more resistant to
0.2 proton damage than LEDs made with amphoterically doped
unbiased processes.
0
0 2x1010 4x1010 6x1010 8x1010 1x1011 1.2x1011 1x10-1
Proton Fluence (p/cm2) Zn-Diffused LED from
1x10-2

Diode or Detector Current (A)

HP 6N140 optocoupler
Figure 9. Degradation of GaAs LEDs from Optek. These devices (700 nm) Forward
1x10-3 Voltage
are also amphoterically doped, but have longer wavelengths that
AlGaAs LEDs. 1x10-4
Pretest
1x10-5
Degradation of the heterojunction LEDs from Optek are 1x10-6 1.5x1011
1.5x1011
shown in Figure 10. The figure shows two different 1x10-7
Light
Pretest Output
parameters: output power and forward voltage characteristics.
1x10-8
The output power of those devices degraded much less than
any of the other LED types that were tested in this study. 1x10-9
They are intended for fiber-optic applications and have much 1x10-10
higher bandwidth than the other device types. For those 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9
devices, changes in forward voltage were actually more Forward Voltage (V)
significant than the decrease in output power. Note the large
change in forward voltage characteristics that occurred at the Figure 11. I-V and P-V characteristics of the 700 nm diffused LEDs
highest irradiated level, 3 x 1011 p/cm2. from Hewlett-Packard optocouplers.
The forward voltage characteristics of those devices also
began to degrade at high radiation levels. Although it is
1x10 -1 difficult to see in the reduced size of the figure, there is a
slight change in slope in the preirradiation V-I characteristics
Diode or Detector Current (A)

Double-heterojunction LED
850 nm
1x10 -3 Diode Pretest that occurs at the same forward voltage that the device begins
Current to emit light. The threshold is not significantly affected by
1x10 -5
1.5x1011
irradiation, although the maximum light output is significantly
3x1011
reduced at higher forward voltages, which correspond to
3x1011
1x10 -7 1.5x10 11 actual use conditions. I-V characteristics are further discussed
Current in
Photodetector in Section VI.
Pretest
1x10 -9
V. CURRENT-ENHANCED ANNEALING
1x10-11
0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 Barry, et al. did annealing experiments on unbiased
Forward Voltage (V) amphoterically doped LEDs over a two-week time interval
Figure 10. Change in I-V and P-V characteristics of the Optek [10]. They found that less than 5% of the damage recovered.
OPF320 LED at high proton irradiation levels. Our measurements of amphoterically doped LEDs that
remained unbiased after irradiation are consistent with their
The OPF320 LEDs degrade far less than the DH LEDs results, leading to the conclusion that little or no damage
from Optodiode. Although only 12 units were tested none of recovery occurs in unbiased devices of that type.
the OPF320 LEDs exhibited the abnormal I-V characteristics However, LED damage can be annealed under forward
that were observed in many of the OD800 DH LEDs from injection [2,3]. This was also noted by D’Ordine in studies of
Optodiode. Optek does not make double-heterojunction optocouplers [11]. Only the amphoterically doped devices in
devices for applications that are equivalent to their extensive our study were sensitive to that effect. Figure 12 shows how
line of amphoterically doped LEDs. Consequently, direct the damage in the three different amphoterically doped devices
comparisons of the OPF320 with the other LEDs tested in the recovered when a moderate current, 5 mA, was passed through
study that were produced by Optek do not have the same the devices after they were irradiated. All the devices were
significance as the comparisons made earlier for the two irradiated without bias. They were irradiated to approximately
Optodiode LEDs. 8 x 1010 p/cm2, which reduced the light output to 9-12% of the
LEDs from a Hewlett-Packard optocoupler (the 6N140) initial value prior to irradiation. The ordinate
were also tested in order to directly compare LEDs from that

5
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

2
The amount of damage that recovered during post radiation
Optodiode OD880
annealing is roughly the same as the difference in the
Relative Light Output

1.5
degradation of devices that were irradiated at low and high
currents when the operating current during irradiation is taken
1 Optek OD130
into account (compare the degradation at high and low bias
conditions in Figures 4 and 5), implying that the same basic
effect is involved in reducing the damage for devices that are
0.5 Optek OP233
forward biased during irradiation.
Forward current = 5 mA
Injection-enhanced annealing adds another layer of
0 complexity when interpreting test data for applications.
0 10 100 1000 10,000 100,000 Testing LEDs at high operating currents will cause some of the
Time Bias Is Applied after Irradiation (seconds) damage to anneal, underestimating the amount of damage that
will occur in applications that use lower operating currents, or
Figure 12. Effect of current on post-radiation damage for the three
types of amphoterically doped LEDs. involve devices that are unbiased during most of the time that
they are exposed to radiation. The best way to deal with this
issue is to carefully control the currents and operating time
is normalized to the value after irradiation. Thus, for an LED used for device characterization, and to split the test samples
that degraded to 10% of the initial light level, a factor of 1.5 into groups with low and high bias conditions for
on that scale corresponds to recovery of the damage from measurements. Unlike the double-heterojunction devices,
about 10% to 15% of the light output before the device was amphoterically doped LEDs undergo less damage when they
irradiated. are operated at high currents. The interplay between injection-
We examined the effects of different bias conditions on enhanced annealing and the actual difference in damage at
annealing by passing different amounts of current through high injection makes it difficult to separate these effects unless
irradiated devices. Recovery was much more rapid when high irradiation and measurement conditions closely mimic the
currents were used during the post-irradiation recovery period conditions in the application.
compared to low currents. The maximum current that was
used was 50 mA, one-half the maximum rated current of the
device. Approximately 1/2 of the damage recovered after
several hours of operation at 50 mA, in contrast to the
VI. PARAMETRIC DEGRADATION
unbiased devices which recovered less than 1% during The earlier studies on LED degradation showed that
comparable time periods. The temperature increase during lifetime damage was the mechanism that caused output
steady-state operation at these currents is very slight, so it is degradation [3-5]. Carrier removal, the dominant mechanism
highly unlikely that temperature is a contributing factor. for degradation in GaAs JFETs, is unimportant for 50 MeV
The effect of operating current on annealing could be proton fluences below 1013 cm2 for GaAs devices doped at
analyzed by considering the total charge that flowed through 1016 cm2 or more [13], which is the approximate doping
the device after irradiation. Figure 13 shows how data for concentration of the light-emitting region in the various LED
three different OD880 devices that were annealed under structures in this study. Thus, lifetime damage is expected to
different current conditions compared from the standpoint of be the dominant mechanism, at least within the range of
total charge. The recovery appears to be logarithmic with radiation levels considered here.
time, and begins to saturate for the device that was annealed Although optical output power is the most fundamental
with the largest current. parameter for light-emitting diodes, forward voltage
characteristics provide a way to evaluate the diode
Light Output (Normalized to Initial Value)

0.3
characteristics in a more fundamental way. Zhao, et al. have
OD880 LED used I-V characteristics to study radiation degradation in laser
0.25
Irradiated to 8 x 1010 p/cm2
diodes [12], and Lindquist used I-V characteristics to study
aging effects in diffused LEDs [9].
0.2 50 MeV Protons 10 mA
Basic junction theory predicts that the forward
50 mA characteristics of the LED can be described by the equation
0.15

0.1 5 mA J = k1 e (qV/kT) + k2 e (qV/2kT) (1)

0.05 Results shown for three

different current conditions where J is the total current through the diode, k1 and k2 are
0
constants, q is electronic charge, V is the forward voltage
0.001 0.01 0.1 1 10 100 1000 applied to the diode, k is Bolzmann’s constant, and T is
Total Charge After Irradiation (C) absolute temperature. At low currents, the slope of the diode
characteristics is lower, and the first term --which represents
Figure 13. Normalization of current-dependent annealing to total recombination current -- dominates. Note that recombination
charge.
current does not contribute to light emission. As V increases,

6
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

the second term (corresponding to diffusion current) begins to We also measured the wavelength of these devices before
contribute, and the slope increases. The change in slope and after irradiation. Neither the peak wavelength nor the
corresponds nearly exactly to the threshold for small light spectral width was significantly affected by proton damage.
output of the LED. That relationship appears to hold for all of Thus, the main parameters that are affected are the light output
the diffused LEDs that were examined in the present study. at moderate to high injection and the forward voltage
For example, a typical OD880 device has an initial slope of characteristics.
87 mV/decade with a nearly ideal transition to a slope of 110 A number of factors contribute to the differences in
mV/decade at the threshold. At radiation levels up to about 1 sensitivity of different types of LEDs. Amphoterically doped
x 1011 p/cm2, the slope in both regions is essentially constant. LEDs in the wavelength range of 870-930 nm have very high
The threshold current is also unchanged. The main effect of efficiency, and can be produced at low cost. However, the
the radiation is to decrease the light output, although the processing used to produce those devices results in extended
forward I-V characteristics also shift slightly. transition regions between the n- and p-regions. They require
The heterojunction devices behaved quite differently. long lifetimes because of the extent of the physical structure
Typical heterojunction structures did not exhibit a clear [3-5]. This is also evident from the slow response time in their
transition region between the recombination-dominated and specifications (see Table 1).
diffusion-dominated regions. The slope changed gradually The diffused devices from Hewlett-Packard and the
over several decades of current, and it was generally not double-heterojunction devices from both manufacturers have
possible to identify the threshold region for light output from much shorter response times. Ikeda, et al. have reported an
the forward voltage characteristics alone. Furthermore, there inverse correlation between the doping concentration in the
were significant differences between different devices of the active layer and cutoff frequency in DH LEDs, as well as a
same type. In some cases the current-voltage characteristics dependence on active layer thickness [15]. Thus, high
exhibited a nonlinear region well below the threshold current. operating frequencies require high doping concentrations in
Figure 14 shows an example for the Optodiode OD880 where the active layer along with narrow thicknesses. This reduces
nonlinear I-V characteristics were present before irradiation. the dependence of LED operation on minority carrier lifetime,
For some of the devices this nonlinear region changed although it also reduces efficiency. The OPF320 devices have
markedly after irradiation. The effect on the device was to a response time of only 6 ns, and exhibited very slight
shift the threshold region to very high operating currents (see degradation in light output even at a fluence of 3 x 1011 p/cm2.
Figure 7). That behavior was only observed for a small These more advanced structures are less affected by radiation
number of the double-heterojunction devices, but is potentially than the best devices in earlier radiation studies [3-5]. Our
quite important because it could cause failures in space at experimental results suggest that for this class of devices
relatively low radiation levels. Similar changes in I-V degradation in I-V characteristics due to nonradiative
characteristics have been observed in reliability studies of DH recombination is likely to be a more important failure mode
LEDs by Wittpahl, et al. that appeared to correlate with device than degradation of optical power. Additional work needs to
sensitivity to current stress [14]. This suggests that I-V be done on high-speed LEDs to verify that this conclusion is
measurements should be included in parametric evaluations of valid for a broader range of device types.
LED degradation. However, the underlying mechanism for
the nonlinear I-V behavior is not understood, and warrants
further study.
VII. RECOMMENDATIONS FOR HARDNESS
ASSURANCE
1x100
The variability in the radiation response of LEDs is made
Optodiode OD800
even more important by their extreme sensitivity at very low
820 nm proton fluence levels. As shown in Section IV, some
1x10-2
amphoterically doped LEDs are degraded by more than a
Forward Current (A)

factor of five at 50 MeV proton fluences of 2 x 1010 p/cm2.

Excess recombination
1x10-4 current at low This is equivalent to a total dose level of 2.5 krad(GaAs); 1
forward voltage x 1010 p/cm2 is equivalent to 1.25 krad(GaAs) or 1.59
krad(Si). This make LEDs among the most sensitive
1x10-6 components in environments that are dominated by protons.
(Typical device)
Failures of optocouplers that contain amphoterically doped
LEDs have been observed in Earth-orbiting spacecraft such as
1x10 -8 Topex-Poseidon at approximately 3 x 1010 p/cm2. Screening
the more sensitive devices can be important in successfully
applying them in space.
1x10-10
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 The approximate 50 MeV equivalent proton fluences of
Forward Voltage (V) two earth-orbiting missions are shown in Table 2, assuming a
Figure 14. Preirradiation I-V characteristics for an OD800 LED with spherical shield thickness of 100 mils of aluminum. These
increased recombination losses at low forward current. values take the energy dependence of proton damage into
account, but do not include temporal variations in the trapped
belt intensities.

7
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

Table 2. Annual Proton Fluences for Two Earth-Orbiting Systems relatively small -- approximately a factor of two after the light
output degraded by more than a factor of ten. Although
lifetime degrades, the differences in initial lifetime between
Annual 50 MeV
different samples of the same part are relatively slight. This
Altitude Inclination Equivalent Fluence
makes it difficult to use such measurements for hardness
Mission (km) (deg.) (p/cm2)
assurance at the moderate radiation levels where amphoteric
LEDs are severely damaged. Nevertheless, lifetime damage
Topex-Posiedon 1334 66 1.3 x 1010
remains a useful way to characterize damage at very high
radiation levels.
EOS 705 98 5.5 x 109

Although none of the amphoterically doped devices VIII. CONCLUSIONS

exhibited the abnormal behavior that was seen for a small This paper has examined proton displacement damage in
number of the 800 nm double-heterojunction LEDs, the light-emitting diodes using a variety of bias conditions and a
amount of degradation of the 880 nm devices varied relatively large number of devices for selected technologies.
significantly for different units. There did not appear to be Although double-heterojunction LEDs are less degraded than
any correlation between initial light intensity and radiation amphoterically doped diffused LEDs, the lower output and
sensitivity. However, there was a correlation between the statistical variability of DH LEDs presents a difficult
peak light emission wavelength and radiation sensitivity for challenge for their use in space. Amphoterically doped LEDs
the Optek OD880 devices, as shown in Figure 15. The have higher initial efficiency, and it may be more effective to
spectral width of a typical LED is about 70 nm, so the range of use them under high injection conditions in space, taking
peak emission wavelength is much smaller than the spectral advantage of injection-enhanced annealing, rather than using
width. Note that the worst devices degrade by nearly a factor double-heterojunction devices with much higher unit-to-unit
of two at 8 x 109 p/cm2, while others retain nearly 75% of variability in radiation sensitivity.
their light output at the same radiation level. Note also that
the increased sensitivity of a second lot of LEDs from that Damage in amphoterically doped LEDs depends on
same manufacturer correlated with the different wavelength. operating conditions. It is important to characterize the
Thus, better control and specification of wavelength may be an dependence on bias to make sure that the experimental
effective way of limiting the range of radiation behavior. characterization of damage will actually correspond to circuit
However, this appears to be effective only for diffused LEDs, use conditions. Post-irradiation recovery measurements
and there are likely other factors besides wavelength that are indicate that the amount of damage recovery depends on the
important in determining the radiation sensitivity of individual total charge that passes through the junction after irradiation,
devices. There did not appear to be any correlation between and this appears to be an effective way to characterize the
wavelength and radiation sensitivity of the 800 nm double- dependence of damage on operating conditions.
heterojunction LEDs. Although degradation in light output is important some
types of LEDs exhibit large changes in forward-voltage
880 characteristics at low injection which increase after irradiation.
For one device type, the threshold current of some samples
increased by several orders of magnitude after irradiation
Peak Wavelength (nm)

because of the increase in recombination. Forward voltage

870 characteristics should be included along with measurements of
First Lot optical power for LED technologies.
Many changes have occurred in LED technology during the
last 20 years, and some of the earlier work on radiation
860 degradation has to be modified to account for changes in
Spectral Width ♠ 70 nm efficiency and processing. Even though several different
Second Lot device types were used in this work, the results are not
necessarily applicable to all types of LEDs. LED degradation
850 is a complex topic that deserves further attention, particularly
0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 because of increased interest in using optoelectronic devices in
Fractional Remaining Light Output at 8x109 p/cm2 space.
Figure 15. Correlation between degradation and wavelength for
Optodiode OD880 LEDs (amphoterically doped).

Barry et al. have shown that lifetime measurements can be

used to characterize the behavior of diffused LEDs [10], which
was also noted in the earlier work by Rose and Barnes [3].
We made lifetime measurements on some of the
amphoterically doped LEDs. Although the lifetime changes
somewhat after irradiation, the changes in lifetime were

8
 Published in IEEE Transactions on Nuclear Science,46(6), 1781 (1999)

REFERENCES [9] P. F. Lindquist, “A New Model for Light Output Degradation of

Direct Bandgap Semiconductors,” 1980 International Reliability
[1] B. G. Rax, C. I. Lee, A. H. Johnston and C. E. Barnes, “Total Physics Proceedings, p. 145.
Dose and Proton Damage in Optoisolators,” IEEE Trans. Nucl.
Sci., 43, 3167 (1996). [10] A. L. Barry, A. J. Houdayer, P. F. Hinrischsen, W. G.
Letourneau and J. Vincent, “The Energy Dependence of
[2] H. Lischka, H, Henschel, O. Köhn, W. Lennartz and U. U. Lifetime Damage Constants in GaAs LEDs for 1-500 MeV
Schmidt, “Radiation Effects in Light Emitting Diodes, Protons,” IEEE Trans. Nucl. Sci., 42, 2104 (1995).
Photodiodes and Optocouplers,” RADECS93 Proceedings,
IEEE Document 93TH0616-3, p. 226. [11] M. D’Ordine, “”Proton Displacement Damage in
Optoisolators,” 1997 IEEE Radiation Effects Data Workshop,
[3] B. H. Rose and C. E. Barnes, “Proton Damage Effects on Light IEEE Document 97TH8293, p. 122.
Emitting Diodes,” J. Appl. Phys., 53(3), 1772 (1982).
[12] Y. F. Zhao, A. R. Patwary, R. D. Schrimpf, M. A. Neifeld and
[4] C. E. Barnes, “A Comparison of Gamma-Irradiation-Induced K. F. Galloway, “200 MeV Proton Damage Effects on Multi-
Degradation in Amphoterically Si-Doped GaAs LED’s and Zn- Quantum Well Laser Diodes,” IEEE Trans. Nucl. Sci., 44, 1898
Diffused GaAs LED’s,” IEEE Trans. Elect. Dev., 26, 739 (1997).
(1979).
[13] G. P. Summers, E. A. Burke, M. A. Xapsos, C. J. Dale, P. W.
[5] C. E. Barnes and J. J. Wiczer, “Radiation Effects in Marshall and E. L. Petersen, “Displacement Damage in GaAs
Optoelectronic Devices,” Technical Report AND84-0771, Structures,” IEEE Trans. Nucl. Sci., 35, 1221 (1988).
Sandia National Laboratories, Albuquerque, New Mexico,
[14] V. Wittpahl, Y. Y. Liu, D. S. H. Chan, W. K. Chim, J. C. H.
May, 1984.
Phang, L. J. Balk and K. P. Yan, “A Degradation Monitor for
[6] N. E. Beyer, “Electroluminescence in Amphoteric Silicon- the Light Output of LEDs Based on Cathodoluminescence
Doped GaAs Devices: I. Steady State Response,” J. Appl. Signals and Junction Ideality Factor,” 1996 International
Phys., 41, 1597 (1970). Reliability Physics Proceedings, p. 188.
[7] K. Ideda, A. Horiuchi, T. Tanaka and W. Susaki, “Design [15] K. Ikeda, S. Horiuchi, T. Tanaka and W. Susaki, “Design
Parameters of Frequency Response of GaAs-(Ga,Al)Al Double Parameters of Frequency Response of GaAs-AlGaAs DH LEDs
Heterostructure LED’s for Optical Communications,” IEEE for Optical Communications,” IEEE Trans. Elect. Dev., 24,
Trans. Elect. Dev., 24, 1001 (1977). 1001 (1977).
[8] D. Vanderwater, I.-H. Tan, G. E. Höfler, D. C. Defevere and F.
A. Kish, “High-Brightness AlGaInP Light Emitting Diodes,”
Proc. of the IEEE, 85, No. 11, 1752 (1997).