Transcript No Slide Title

Time-Resolved Photoluminescence Spectroscopy
of InGaAs/InP Heterostructures*
Colleen Gillespie and Tim Gfroerer, Davidson College, Davidson, NC
Mark Wanlass, National Renewable Energy Laboratory, Golden, CO
Abstract
Motivation: TPV Cells and
How they Generate Electricity
- -
E-Field
ELECTRON
Bandgap
Frequency-dependent
Lock-in Signal
fmodulation
Lock-in
signal
HOLE
+
+
E-Field
Absorption
+
8fmodulation
+
Current (good!) or
Recombination (bad!)
When a blackbody photon (with sufficient energy) is absorbed, free charge
carriers are created, which move to generate current.
Caveat: if electrons recombine with holes before they are swept away by the
intrinsic electric field, TPV efficiency decreases.
Introduction: Defect Levels and
Expected Behavior
Band - - - - Conduction
Conduction
- - - - - --- - - - - -Band
------
InGaAs
Hypothesis: Spatially-separated
Recombination
Time
1.2
B-B Pex= 150 W/cm
1.0
B-B Pex= 20 W/cm
Low Excitation:
-
Indirect
transition
hole
electron
0.8
τ 1= 47 ns
τ2= 920 ns
0.4
-
Wavefunction position
+ + + + + + + + ++ + + + + + + + +
+ + +
+ + +
Valence Band
Interface Defect
Lattice defects produce new electronic states in the bandgap. Defect states
facilitate nonradiative recombination (which decreases TPV efficiency). But at
high excitation, defect levels should be filled, saturating this loss mechanism.
One possible explanation for this behavior is that sub-bandgap
recombination is taking place between spatially separated
electrons and holes. The electrons and hole wavefunctions are
separated in physical space.
High Excitation:
Previous Work: Spectra and Results
When the modulation frequency of the laser is increased, a signal
with a slower response time will not be able to keep up, resulting
in a smaller measured amplitude.
Frequency Response Results
System response
time = 18 ns
1.1
0.0
0.0
0.5
1.0
1.5
2.0
Time (micro sec)
Energy
Substrate
(InP)
2
Laser
0.6
0.2
2
-0.2
++
-
Transient PL
If the sub-bandgap recombination is spatially-separated, then the rate
would be much slower than spatially-direct band-to-band
recombination. We measured the band-to-band recombination rates
shown above. However, the sub-bandgap signal is ~1000 times
weaker and could not be detected with this system.
Normalized Lock-in SIgnal
Valence Band
Normalized Intensity
PHOTON
(LIGHT)
ENERGY
ENERGY
Conduction Band
Semiconductor-based thermophotovoltaic cells, which convert thermal radiation into electricity, show potential for an efficient and clean
source of energy. InGaAs alloys are ideal for this process because of their small bandgap energies. However, defects in the material
can give rise to new electronic levels within the bandgap. These defect states usually provide non-radiative recombination paths (which
decrease the conversion efficiency of photovoltaic cells), but we have found a deep level in the near-lattice matched samples that allows
for radiative recombination. Previous work shows that sub-bandgap emission grows super-linearly with excitation power up to and
exceeding 1000 W/cm2. This unusual result suggests that the defect-related radiative process is complex. We hypothesized that the
electrons and holes are physically separated at the defect sites, so that increasing excitation leads to an increase in electron-hole
wavefunction overlap and an increased recombination rate. If the electrons and holes are spatially separated, the resulting transition
times should generally be much longer than those of the band-to-band transitions. We have measured the sub-bandgap and band-toband recombination rates and find that our experimental results are inconsistent with the spatial separation model.
1.0
0.9
2
SBG (Pex= 346 W/cm )
2
B-B (Pex= 224 W/cm )
Laser response
0.8
0.7
0.6
Response time ~ 1/(1000
KHz) ~1 microsec
0.5
0.4
1
New Experimental Setup:
Frequency Response
10
100
Frequency (kHz)
1000
While this is not a very precise test, it clearly shows the similarity in
response times between the band-to-band and sub-bandgap
signals, which is inconsistent with the spatial separation hypothesis.
0
Band-toband (B-B)
-1
10
-2
10
Pex= 3442 W/cm
2
Pex= 1539 W/cm
2
Pex= 156 W/cm
Subbandgap
(SBG)
Pex= 31 W/cm
Overlap increases
2
2
-3
10
-
ENERGY
Normalized PL Intensity (a.u.)
10
Nd/Yag Laser
(1064 nm)
Lock-in
Cryostat
@ 77 K
Fast
Pre-amp
More direct
transition
AOM
++
++
Sample
Photodetector
-4
10
Wavefunction position
ND Filters
-5
10
Increased E-field
-6
10
0.4
0.5
0.6
0.7
0.8
0.9
: Laser Light
Energy (eV)
Defect-related emission is observed in the photoluminescence (PL) spectrum.
But the super-linear growth with increasing excitation (for comparison, the
band-to-band increase is approximately linear) is not consistent with the
expected saturation noted above.
At a high excitation, there is an increased density of electrons and
holes, which leads to an increased electric field and an increased
electron-hole wavefunction overlap. Recombination would occur at
a faster rate, explaining the super-linearity of the sub-bandgap PL.
: Luminescence
Band-pass
filter
So we used a fast lock-in amplifier to measure the frequency
response of the band-to-band and sub-bandgap signals. The lock-in
amplifier is more sensitive than the boxcar averager that we used to
measure transient PL.
Conclusions
• Band-to-band (B-B) transient PL reveals radiative (t1 = 47ns) and
defect-related (t2 = 920ns) mechanisms as expected.
• Sub-band gap (SBG) and B-B response times are similar
(approximately 1 microsec).
• Results are inconsistent with spatial separation hypothesis.
• The similiarity in SBG and B-B response indicates that the
mechanisms are more closely related than expected.
* Project supported by the American Chemical
Society – Petroleum Research Fund