AlGaN-based multiple-quantum-well (MQW) heterostructures were irradiated with a pulsed electron beam. Excitation with a beam energy of 12keV and a beam current of 4.4mA produced cathodoluminescense at λ=246nm with a measured peak output power of >200mW. The emission is dominated by radiative recombination from the MQW up to the maximum tested excitation power density of 1MW/cm2, as evidenced by unity slope in a double-logarithmic plot of the light output power vs. excitation power density. Monte Carlo simulations of the depth distribution of deposited energy for different beam energies produced good agreement with the measured peak output power vs. beam energy for an assumed carrier diffusion length of 200nm.

Compact deep-UV light sources have numerous applications in medicine, manufacturing, and research, which include water purification and chemical detection. The AlGaN material system is an excellent candidate for fabricating deep-UV emitters because the Al composition can be adjusted to select the emission wavelength over the range from 206nm to 365nm.1 However, realizing current-injection laser diodes and high-performance light emitting diode (LED) at a high Al mole fraction is challenging because p-doping and ohmic contact formation are difficult to achieve and inefficient.2,3 For example, UV LEDs emitting at a wavelength of 270nm showed a wall plug efficiency (WPE) of only 2.5%,4 which is significantly lower than what can be achieved with visible emitters today. An alternative approach that bypasses the need for p-doping and high-quality ohmic contacts is electron-beam pumping.5–9 In particular, cathodoluminescence (CL) from e-beam-pumped AlGaN multiple-quantum-wells (MQWs) has been reported by different groups,7,10,11 with a claimed output power of 100mW at 238nm,7 but lasing in the deep-UV has yet to be demonstrated.

This study presents findings from the electron-beam pumping of AlGaN multiple-quantum-well (MQW) heterostructures and a record for the CL output power in the deep-UV. A dedicated e-beam pumping system was used which is capable of focused beam currents of up to IB=5mA and beam energies up to EB=30keV. The heterostructures consisted of ten 1.5nm thick Al0.56Ga0.44N/Al0.9Ga0.1N quantum wells with a barrier thickness of about 40nm that were grown by metal organic vapor phase epitaxy (MOVPE). The AlN-on-sapphire templates used in this experiment were prepared by the epitaxial-lateral overgrowth (ELO) technique to achieve a low dislocation-density base layer.12–14 

The measurement setup is shown in Fig. 1. The sample was coated with 100nm of Al and directly attached to the quartz window of the vacuum chamber with a UV-transparent adhesive. A calibrated photo-detector was placed in close proximity to the exit window to measure the output power. The Al coating provided a high UV reflectivity for increased light extraction and was electrically grounded for the electrostatic discharge. The photo-detector could be interchanged with an optical fiber and spectrometer to record emission spectra.

FIG. 1.

The schematic diagram of the measurement setup. The AlGaN MQW heterostructure on sapphire is coated with a 100nm Al cap layer and attached to the quartz window with a UV-transparent adhesive, and the calibrated photo-detector is placed close to the window with a narrow air gap.

FIG. 1.

The schematic diagram of the measurement setup. The AlGaN MQW heterostructure on sapphire is coated with a 100nm Al cap layer and attached to the quartz window with a UV-transparent adhesive, and the calibrated photo-detector is placed close to the window with a narrow air gap.

Close modal

Figure 2(a) shows CL spectra recorded over a range of pulsed-beam currents at a beam energy of 17keV, with a pulse width of 750ns and a repetition frequency of 1kHz. The pump spot diameter was 3mm as used for most of the following measurements if not specified otherwise. The MQW emission peaks at λ=246.4nm with a full-width at half maximum (FWHM) of 10.8nm, as determined from Gaussian fits of the spectra, and the intensity increases linearly with the beam current. As shown in more detail in Fig. 2(b), there are only slight variations of the peak wavelength and spectra width with the beam current.

FIG. 2.

The current-dependent variation of CL with pulsed electron-beam current for the beam energy of EB=17keV and a pump spot diameter of 3mm: (a) CL spectra with 750ns pulse width and 1kHz repetition frequency; (b) peak wavelength and FWHM estimated from Gaussian fits of the spectra in (a).

FIG. 2.

The current-dependent variation of CL with pulsed electron-beam current for the beam energy of EB=17keV and a pump spot diameter of 3mm: (a) CL spectra with 750ns pulse width and 1kHz repetition frequency; (b) peak wavelength and FWHM estimated from Gaussian fits of the spectra in (a).

Close modal

The spectra in Fig. 2(a) suggest that there is no defect luminescence in the UV spectral range. This observation is further substantiated with the semi-logarithmic plot of CL intensity for a range of beam energies in Fig. 3, with the highest intensity observed at EB=12keV. A faint blue defect luminescence was detected, which accounts for less than 2% of the overall measured output power.

FIG. 3.

The semi-logarithmic plot of the CL spectrum for a range of pulsed e-beam energies, with 1mA beam current, 1.25μs pulse width, and 1kHz repetition frequency.

FIG. 3.

The semi-logarithmic plot of the CL spectrum for a range of pulsed e-beam energies, with 1mA beam current, 1.25μs pulse width, and 1kHz repetition frequency.

Close modal

The results from Monte Carlo simulations (software CASINO15) of the deposited energy at different incident e-beam energies EB are shown in Fig. 4(a). The two vertical dashed lines mark the depth of the MQW region (450nm) and the location of an additional 150nm of underlying AlN. A good overlap of the energy distribution with the MQW region is observed over the range of 816keV. The simulation of the interaction volume is shown in Fig. 4(b) for the beam energy of EB=12keV. The curves in Fig. 4(a) were integrated over the MQW region, and additional 100nm and 150nm segments of the underlying AlN were included in the integrations as a means to parametrize the contribution to the output power from carrier diffusion from the AlN layer into the MQWs.

FIG. 4.

(a) Monte Carlo simulations of the deposited energy depth profile for e-beam energies between 8keV and 30keV. The first vertical dashed line marks the depth of the MQW region, and the second extends an additional 150nm into AlN. (b) Simulation of the interaction volume for an e-beam energy of EB=12keV. The maximum of excitation is within the yellow contour. The extent of the interaction volume is about 1μm.

FIG. 4.

(a) Monte Carlo simulations of the deposited energy depth profile for e-beam energies between 8keV and 30keV. The first vertical dashed line marks the depth of the MQW region, and the second extends an additional 150nm into AlN. (b) Simulation of the interaction volume for an e-beam energy of EB=12keV. The maximum of excitation is within the yellow contour. The extent of the interaction volume is about 1μm.

Close modal

The measurements of the optical output power were performed for different beam energies at a constant beam current of 0.5mA, a pulse width of 750ns, and a repetition frequency of 1kHz. The measurements were normalized to allow for a qualitative comparison to the simulations presented in Fig. 4(a) and are shown in Fig. 5.

FIG. 5.

The normalized output power vs. the beam energy comparing measurement with the simulations of the electron energy absorption in the MQW region and with additional 100nm and 150nm of underlying AlN, determined by integrating the curves in Fig. 4(a). Measurements were performed at a constant beam current of 0.5mA, a pulse width of 750ns, a repetition frequency of 1kHz, and a spot diameter of 3mm.

FIG. 5.

The normalized output power vs. the beam energy comparing measurement with the simulations of the electron energy absorption in the MQW region and with additional 100nm and 150nm of underlying AlN, determined by integrating the curves in Fig. 4(a). Measurements were performed at a constant beam current of 0.5mA, a pulse width of 750ns, a repetition frequency of 1kHz, and a spot diameter of 3mm.

Close modal

The measured output power peaks at about 12keV, whereas the simulation peaks at about 10keV when only the MQW region is taken into account. A better agreement between the experiment and simulation is obtained by including carrier diffusion from the underlying AlN into the MQWs, with a good match in the peak energy when the depth integration includes 150nm of the underlying AlN. This suggests an estimate for the carrier diffusion length of 200nm and experimentally verifies the models and assumptions made in the Monte Carlo software CASINO.

The dependence of the peak output power on the pulse width is shown in Fig. 6. The output power decreases with an increase in the pulse width with two distinct time constants. A fast drop in the output power is observed for pulse widths between 0.5μs and 2μs and a slow decay for longer pulse widths. This behavior may relate to similar observations that have been made in GaN-based laser diodes and explained by two distinct self-heating mechanisms that involve the MQW region and crystal lattice, respectively.16 

FIG. 6.

Measurement of the peak output power vs. excitation pulse width for a beam energy of EB=10keV, a beam current of IB=1mA, and 1kHz repetition frequency.

FIG. 6.

Measurement of the peak output power vs. excitation pulse width for a beam energy of EB=10keV, a beam current of IB=1mA, and 1kHz repetition frequency.

Close modal

The measurement of peak output power as a function of e-beam current for the UV-C MQW heterostructure is shown in Fig. 7. The beam energy was 12keV. The beam diameter was 3mm, and the pulse width was maintained at 500ns, with 1kHz repetition rate, to minimize the heating effects and to access the maximum output power. A peak output power of 230mW (at λ=246nm) was achieved at the maximum accessed beam current of 4.4mA. This is more than twice the CL output power reported in Ref. 7. In the present study, special care was taken to ensure that the measured output power was not influenced by the inadvertent detection of the X-ray emission generated by the e-beam bombardment of the sample. For example, a 4mm thick quartz window was used to attenuate the flux of X-ray radiation impinging on the detector to negligible levels. This consideration is particularly critical when estimating the wall plug efficiency and other quantitative numbers.

FIG. 7.

L-I curve for e-beam excitation of an AlGaN MQW heterostructure at an e-beam energy of 12keV, a pulse width of 500ns, and a repetition frequency of 1kHz. The maximum output power was 230mW at IB=4.4mA.

FIG. 7.

L-I curve for e-beam excitation of an AlGaN MQW heterostructure at an e-beam energy of 12keV, a pulse width of 500ns, and a repetition frequency of 1kHz. The maximum output power was 230mW at IB=4.4mA.

Close modal

Figure 8 presents the results of Fig. 7 (for a 3mm spot size) in a double-logarithmic plot of the CL output power vs. the e-beam excitation power density. The power density at the maximum beam current is about 750W/cm2, which corresponds to a carrier density per QW of 8×1016cm3 using the following equation:17 

n=GτApnQWdQW=VbIbQ(1γ)EgApnQWdQWτVbIb3EgApnQWdQWτ,
(1)

where G is the generation rate, τ the carrier lifetime, Ap the pump area, nQW the number of QWs, and dQW the QW thickness. Vb and Ib are the beam voltage and current, and Eg is the band gap of the excited material. Q represents the fraction of the beam-energy that generates electron-hole pairs, which is typically about one third. γ as the fractional electron beam energy loss due to backscattering was neglected. We assume the even distribution of the generated carriers into all QWs. Furthermore, we assume that the diffusion length for electrons and holes exceeds the barrier thickness, i.e., the spacing between the QWs. Finally, we used a carrier lifetime of 450ps for the calculations as determined by the time-resolved PL measurements using the same sample.18 

FIG. 8.

Comparison in a double-logarithmic plot of the CL output power vs. e-beam excitation power density for the incident e-beam diameters of 3mm and 50μm.

FIG. 8.

Comparison in a double-logarithmic plot of the CL output power vs. e-beam excitation power density for the incident e-beam diameters of 3mm and 50μm.

Close modal

The slope of the curve is unity over the entire range of power densities, which indicates that the carrier recombination process is dominated by radiative recombination at all the measured current levels. There is no evidence of significant contribution from non-radiative recombination at the lowest power densities or from Auger recombination or carrier overflow transport processes at the highest power densities.

To further assess this potentially interesting observation, the measurement was repeated with the incident e-beam focused to a diameter of 50μm, and the results are included in Fig. 8. The maximum beam current was 1.72mA, which translates to an excitation power density of 1.05MW/cm2 and a calculated carrier density of 1×1020cm3. The slope remains unity up to the maximum tested excitation power density of 1MW/cm2. The lower maximum output power in comparison to the results for the 3mm spot is ascribed to a reduced internal quantum efficiency (IQE) due to heating. To further substantiate this interesting observation, we offer that MQW heterostructures from earlier in the growth optimization effort do exhibit both slope 2 at low excitation power densities, which corresponds to non-radiative recombination, and slope 2/3 at high power densities, which corresponds to processes with cubic dependence in the ABC model.19 Thus, the results in Fig. 8, with slope of unity, indicate the dominance of radiative recombination even at high power densities.

The wall plug efficiency (WPE) is 0.43% for the measurement at 12keV and 3mm spot size in Figs. 7 and 8, which is within the range of that achieved for LEDs emitting in the deep-UV.20 The WPE can be expressed as the product of the conversion efficiency (ηconv), the collection efficiency (ηcoll), the internal quantum efficiency (IQE), the quantum deficit (QD), and the light extraction efficiency (LEE)

WPE=ηconv·ηcoll·IQE·QD·LEE.
(2)

The conversion efficiency represents the efficiency of converting high-energy electrons to electron-hole pairs and ηconv=33%.17,21 The collection efficiency is the percentage of beam energy that is converted into e-h pairs and can be determined from Monte Carlo simulations, which for 12keV electrons incident on our structure yields ηcoll=75.7%. The quantum deficit approximates the energy that is lost due to the higher band gap of the Al0.9Ga0.1N matrix material (Eg=5.67eV) compared to the emission energy of the QWs (Eg=5.04eV) with QD=88.9%. The light extraction efficiency determines how much of the light that is produced by the MQWs reaches the photo-detector. Light ray tracing simulations of the experimental setup shown in Fig. 1 yielded values for the LEE of between 9%and17% dependent on the assumed roughness of the substrate. The above estimates of the several accessible efficiency factors result in an estimate for the IQE between 12%and23%. By way of comparison, the measured results for a 10keV beam yielded WPE=0.62% and ηcoll=85.7% for IQE between 17%and29%. These estimates are consistent with results from temperature-dependent photoluminescence measurements on the same sample that revealed IQE>23%.18 

In summary, the excitation of Al0.56Ga0.44N/Al0.9Ga0.1N MQW heterostructures with a pulsed electron beam (beam energy 12keV; beam current 4.4mA) produced cathodoluminescense centered at λ=246nm with a measured peak output power of 230mW. Radiative recombination from the MQW dominates the emission process up to the maximum tested excitation power density of 1MW/cm2, which indicates a good material quality, a good carrier injection, and no significant contribution from third-order recombination processes (e.g., Auger). The wall plug efficiency for the specified test sample was 0.43% (at 12keV), and the examination of the several factors contributing to the overall efficiency yielded an estimate for the internal quantum efficiency of about 23%.

This research was developed with funding from the Defense Advanced Research Projects Agency (DARPA). The authors thank Markus Weyers and Sylvia Hagedorn, Ferdinand-Braun-Institut, Leibniz-Institut für Höchstfrequenztechnik, Germany, for providing ELO AlN/sapphire templates and P. Maeda (PARC) for contributing light ray tracing simulations.

The views, opinions, and/or findings expressed are those of the authors and should not be interpreted as representing the official views or policies of the Department of Defense or the U.S. Government.

1.
M.
Feneberg
,
M. F.
Romero
,
M.
Röppischer
,
C.
Cobet
,
N.
Esser
,
B.
Neuschl
,
K.
Thonke
,
M.
Bickermann
, and
R.
Goldhahn
,
Phys. Rev. B
87
,
235209
(
2013
).
2.
Y.
Taniyasu
,
M.
Kasu
, and
T.
Makimoto
,
Nat. Lett.
441
,
325
(
2006
).
3.
M. L.
Nakarmi
,
N.
Nepal
,
C.
Ugolini
,
T. M.
Altahtamouni
,
J. Y.
Lin
, and
H. X.
Jiang
,
Appl. Phys. Lett.
89
,
152120
(
2006
).
4.
J. R.
Grandusky
,
J.
Chen
,
S. R.
Gibb
,
M. C.
Mendrick
,
C. G.
Moe
,
L.
Rodak
,
G. A.
Garrett
,
M.
Wraback
, and
L. J.
Schowalter
,
Appl. Phys. Express
6
,
032101
(
2013
).
5.
H.
Klausing
,
F.
Fedler
,
J.
Daenhardt
,
R.
Jaurich
,
A.
Kariazine
,
S.
Guenster
,
D.
Mistele
, and
J.
Graul
,
Phys. Status Solidi A
194
,
428
(
2002
).
6.
M. D.
Tiberi
,
V. I.
Kozlovsky
, and
P. I.
Kuznetsov
,
Phys. Status Solidi B
247
,
1547
1552
(
2010
).
7.
T.
Oto
,
R. G.
Banal
,
K.
Kataoka
,
M.
Funato
, and
Y.
Kawakami
,
Nat. Photonics
4
,
767
(
2010
).
8.
N.
Gamov
,
E.
Zhdanova
,
M.
Zverev
,
D.
Peregoudov
,
V.
Studenov
,
A.
Mazalov
,
V.
Kureshov
,
D.
Sabitov
,
A.
Padalitsa
, and
A.
Marmalyuk
,
Quantum Electron.
45
,
601
(
2015
).
9.
T.
Klein
,
S.
Klembt
,
V. I.
Kozlovsky
,
M. D.
Tiberi
, and
C.
Kruse
,
J. Appl. Phys.
117
,
113106
(
2015
).
10.
T.
Matsumoto
,
S.
Iwayama
,
T.
Saito
,
Y.
Kawakami
,
F.
Kubo
, and
H.
Amano
,
Opt. Express
20
,
24320
24329
(
2012
).
11.
S. V.
Ivanov
,
V. N.
Jmerik
,
D. V.
Nechaev
,
V. I.
Kozlovsky
, and
M. D.
Tiberi
,
Phys. Status Solidi A
212
,
1011
1016
(
2015
).
12.
V.
Kueller
,
A.
Knauer
,
U.
Zeimer
,
H.
Rodriguez
,
A.
Mogilatenko
,
M.
Kneissl
, and
M.
Weyers
,
Phys. Status Solidi C
8
,
2022
(
2011
).
13.
U.
Zeimer
,
V.
Kueller
,
A.
Knauer
,
A.
Mogilatenko
,
M.
Weyers
, and
M.
Kneissl
,
J. Cryst. Growth
377
,
32
(
2013
).
14.
A.
Knauer
,
A.
Mogilatenko
,
S.
Hagedorn
,
J.
Enslin
,
T.
Wernicke
,
M.
Kneissl
, and
M.
Weyers
,
Phys. Status Solidi B
253
,
809
813
(
2016
).
15.
P.
Hovongton
and
D.
Drouin
, See http://www.gel.usherbrooke.ca/casino/ for “Monte Carlo simulation software tool.”
16.
C.
Eichler
,
S.-S.
Schad
,
M.
Seyboth
,
F.
Habel
,
M.
Scherer
,
S.
Miller
,
A.
Weimar
,
A.
Lell
,
V.
Hrle
, and
D.
Hofstetter
,
Phys. Status Solidi C
0
,
2283
2286
(
2003
).
17.
B. G.
Yacobi
and
D. B.
Holt
,
J. Appl. Phys.
59
,
R1
(
1986
).
18.
C. R.
Haughn
, U.S. Army Research Laboratory, private communication (
2016
).
19.
S.
Karpov
,
Opt. Quantum Electron.
47
,
1293
1303
(
2015
).
20.
M.
Kneissl
and
J.
Rass
,
III-Nitride Ultraviolet Emitters
(
Springer
,
Switzerland
,
2016
).
21.
B.
Dierre
,
X.
Yuan
, and
T.
Sekiguchi
,
Sci. Technol. Adv. Mater.
11
,
043001
(
2010
).