Chapter 7 InAs/GaAs Quantum Ring Infrared Photodetectors
7.3 Investigation of the Device Characteristics
After the device fabrication, the photocurrent spectra were measured by FTIR under normal incidence configuration first. Fig. 7.3 shows the response spectra at 30K and 77K.
Although QRIPs are expected to be less sensitive to the normal incident radiation due to the flattened geometry of QRs, the normal incident photocurrent signal is still strong and clear. The
Fig. 7.2 The PL spectra of the sample at 77K with different excitation powers. The excited state peaks are indicated in the high excitation spectrum.
114
polarization dependent response was further examined with the 45o edge-coupling scheme. The TE (E-field polarized in the QR plane) to TM (E-field polarized in the growth direction) response ratio of the sample was measured to be about 11% as shown in Fig. 7.4. This number, although smaller than that in QDIPs, is still stronger than that in QWIPs (see Fig. 5.4).
the response peaks.
Fig. 7.3 The photocurrent spectra of the sample at 30K and 77K. The insert shows the transitions associated with
Fig. 7.4 Polarization-dependent photocurrent spectra of the sample at 22K under 45o facet-coupled configuration.
115
The photocurrent spectrum shows a broadband detection with the center wavelength around 8μm and a bandwidth of 5μm (Δλ/λ~60%). Unlike the single peak usually seen in QWIPs or QDIPs, multiple peaks were seen in the spectrum. From the AFM image and the PL spectrum, the QRs are quite uniform in size, indicating the wide detection band is not from the multi-modal distribution of the QRs. Instead, the multi-peak spectrum is caused by the transitions of electrons from the QR ground state to different excited states including QR bound states, WL states and GaAs continuum states. It is known that the GaAs bandgap at 77K is 1.507eV [7.17], and the WL transition energy determined by the high power PL spectrum is 1.426eV. Assuming a 7:3 ratio for the conduction/valence band discontinuity at the hetero-interface and with a given ground state energy of 1.238eV, any radiation with the photon energy higher than 188 meV will be able to excite electrons in a QR to the GaAs barrier. The broad peak around 6.2μm (200 meV) ransition from the ground state to the GaAs continuum states. The two response peaks at 7.4 μm and 8.36 μm correspond to the transitions from the ground state to the WL states. Finally, the lowest energy peak at 9.33μm is from the transition to the fourth QR excited state. The above four excitation schemes are also schematically plotted in the insert of Fig. 7.3. It might be suspected that some of the response peaks are from the first excited states to the higher states. However, those transitions will have energy lower than the peaks detected. Furthermore, the doping density used in our sample is much lower than the QR density, and the probability for electrons to populate the excited states is quite low. Finally, given that no obvious change of the response spectra at different bias levels w
bandedge when compared with the quantum states in QDs. The ground state electron is thus from the t
as detected, we conclude that the responsivity peaks are unlikely to be originated from the excited states.
On the other hand, for the QDIPs with the same 2.8ML InAs QDs as in the QRIP sample, only the signal associated with the WL transition was observed (see Fig. 5.2). The quantum states in QRs are weakly confined, and the state energies are closer to the GaAs
116
wavefunction extends further outside of the QR region and enhances the bound to continuum transitions in QRIPs. Due to the higher states energy with respect to the GaAs bandedge, the escape probability for the transition to the QR fourth excited states is much higher in QRs. The deep bound states in QDs prohibit the generation of photocurrent from such transitions.
It is also evident in Fig. 7.3 that the responsivity spectra at 30K and 77K are slightly different at longer wavelength. The long wavelength peaks (8.36 μm and 9.33 μm) decrease from 30K to 77K while other peaks remain at the same level. As mentioned, the carriers excited to the lower energy states need more energy to escape from potential trap before being trapped back to the ground states. At higher temperatures, it was found the relaxation rate increases so much that the excited electrons are more likely to be trapped in the QRs than to contribute to the photocurrent [7.18]. Due to the lower excited state energy, this effect is more severe for the 9.33 μm peak, and it decays more than the 8.36μm peak from 30K to 77K.
The absolute responsivity was then calibrated by a 1000 oC blackbody radiation source using the lock-in technique with the chopper frequency set at 1 kHz. Due to the inserted AlGaAs layer, the responsivity is quite weak under positive biases [7.3], so the following discussions are limited to the negative biases only. The peak responsivity at different negative biases and temperatures are plotted in Fig. 7.5. The responsivity is quite stable from 40K to 100K f
plane.
The fas
or a wide bias range (-2.8V ~ -0.6V). This phenomenon is different from that in QDIPs, in which the responsivity usually increases over a decade in this temperature range. As discussed in section 6.2, for QDIPs, the dramatic increase in responsivity with temperature is due to the rapid increase in current gain, which is the result of the repulsive coulomb potential caused by the extra carriers inside the QDs. For QRIPs, however, the repopulation of carriers is faster within individual QR layer because of the higher ground state energy. Thus, the current gain and therefore the responsivity are more stable against temperature variation. This temperature stability is similar to that of QWIPs in which carriers flow freely in the QW
t carrier flow and the large lateral extent of the QRs with higher ground state energy
Fig. 7.5 The voltage dependence of the peak responsivity of the sample at different temperatures.
also increase the probability for the conductive free carriers being captured into the QRs. The current gain of our QRIP sample at -1V and 77K is only 0.003, which is much lower than that measured under the same conditions in QDIPs with a similar device structure. The carrier flow within the QR plane and the high capture probability of free carriers make QRIPs behave like QWIPs in the transport characteristics.
To verify the difference of the carrier flow within a QR layer and a QD layer, two samples, each with a single layer of QDs or QRs, were prepared. The same amount of InAs was used in both samples, but the QR sample had an additional partially-capping and annealing process for the ring formation. The carrier flow information could be extracted by examining the temperature dependent PL line width variation [7.19]. The PL spectra of the two samples and their FWHM as functions of temperature are compared in Fig. 7.6. As expected, the ground state energy for the QRs is higher than that of the QDs for more than 100meV. The widths of the PL spectra of the two samples both decreased first and then broadened with the increase of temperature. At extremely low temperatures, the width of the PL signal is from the non-equilibrium and random distribution of carriers in QRs or QDs, and is usually wider. With the increase of temperature, the carrier distribution will be closer to the quasi-equilibrium
117
Fig. 7.6 Temperature dependent FWHM of the PL spectra of single-layer QRs (solid symbol) and single-layer QDs
or QDs. For the QD sample, Tm was found to be around 230K.
(unfilled symbol). The insert shows their spectra at 20K.
condition with a common Fermi level for the whole ensemble. The temperature with minimum spectral width indicates that, above this temperature (Tm), the thermal energy is high enough for carriers to flow freely between the QRs
But, for the QR sample, the line width shrank much faster and Tm occurred at 130K.
Such a large difference in Tm indicates that electrons within a QR layer can move more freely to repopulate the QRs. The facilitated carrier repopulation in the QR layer makes the current gain and the responsivity in QRIPs insensitive to the temperature change.
The dark current of the device was also measured at different temperatures with a helium dewar. The measured dark current curves and the extracted activation energy (Ea) are shown in Fig. 7.7. The Ea shown here was calculated assuming
118
⎟⎞
⎜⎛ −
∝ E
T
Id exp a [7.1]. Both the dark current and Ea exhibit two distinct regions in the figure. Ea drops dramatically at a lower bias region but then decreases with a lower slope when the bias is over -0.65V.
Accordingly, the dark current curves also show two different increasing rates with biases for the temperatures higher than 60K. By extrapolation, the zero bias Ea could be estimated to be in
⎠
⎝ kT
Fig. 7.7 The voltage dependence of the dark current of the sample with different temperatures and the dark current activation energies (Ea) at different bias voltages.
the range of 250~260 m nd discontinuity of the
Al0.27Ga0.73As/GaAs hetero-interface (256 me [7.17]). It means that the voltage drop is
peration at elevated temper
e result is shown
8 0.5
7 0.5
eV, which agrees well with the conduction ba V
essentially on the Al0.27Ga0.73As layer from 0V to -0.65V and the dark current is limited by the wide bandgap material. When the bias further increases, the applied voltage starts to drop on the active region, and Ea is then dominated by the effective Fermi-level in the QR layer. In this domain, the activation energy at -0.7V was 112 meV, which is close to the reported activation energy for the QRIP without the AlGaAs barrier[7.15]. Such small activation energy in QRIPs prohibits high temperature operation and is the result of the higher ground state energy. Hence, the adoption of a wide bandgap layer in QRIPs is necessary for the o
atures.
The noise current spectra of the device were then measured and the noise current density at 1 kHz was used to calculate the specific detectivity. Th in Fig. 7.8.
The detectivity measured at 77K and -1V is 2.8×10 cmHz /W. At 140K, the highest detectivity measured for the QRIP was 2.5×10 cmHz /W at -0.1V. In this regime, the dark current was highly suppressed by the AlGaAs layer, so the performance of the QRIP could be
119
Fig. 7.8 The voltage dependence of specific detectivity of the sample at different temperatures.
120
easured up to 140K. It should be mentioned that the device performance measured here is
e WL states and the GaAs continuum states.
The higher confinem
ly enhanced with an Al0.27Ga0.73As current blocking layer.
m
based on the normal incident configuration without any grating couplers. The device performance can be enhanced with a properly designed grating coupler.
7.4 Summary
The characteristics of InAs QRIPs were investigated. The thinner InAs thickness and wider extent of the QR structure generated higher but closer electron state energies and weaker wavefunction confinement. This induces the specific features of QRIPs such as wider photocurrent spectra, more stable responsivity with temperature change, and lower dark current activation energy. The broad spectral response comes from the transitions of electrons from the QR ground state to the QR excited bound states, th
ent state energy and the extended wavefunction in QRIPs induce better carrier flow within each QR layer as well as lower current gain. Such behaviors makes QRIPs have similar transport properties with QWIPs instead of QDIPs. For high temperature operation of QRIPs, wide bandgap layer should be inserted into the devices to compensate for the inherently small ionization energy of carriers. The performance of LWIR QRIPs was great
121
Chapter 8
Conclusions and Future Work
8.1 Conclusions of the Present Studies
In this dissertation, comprehensive studies on the self-assembled, zero-dimensional InAs quantum structures and the associated infrared photodetectors are presented. The primary conclusions are summarized as follows:
The growth techniques to manipulate the sheet density and geometry of
8 -2 11 -2
rings (QRs), were the InAs QR formation depends on the thickness of the layer and the annealing condition. A thinner cap layer, because of a larger interface angle, resu
10 -2
ission with the help of the (1)
self-assembled InAs quantum dots (QDs) with fully in situ growth control were investigated. A wide range of dot densities from ~1x10 cm to ~1.2x10 cm , and the control over the geometric change from QDs, through volcano-shaped structures, to quantum
achieved. The evolution level of partial-capping GaAs
lts in a larger driving force for pulling out the InAs materials from the tips of the dots and therefore more matured rings with deeper center craters. The annealing temperature and duration provide the thermal energy and time needed for the migration of InAs.
By simulation, we find the formation of the ring-like potential inhibits the wavefunctions from accumulating to the ring center and therefore raises the state energies and also narrows the energy separations. With optimized growth conditions, uniform QRs with a density of 2.6x10 cm and strong room temperature PL emission are obtained.
(2) The energy spectra of self-assembled InAs QDs were investigated using the selective excitation PL technique. The dependence of the emission spectra on the excitation energy provides important information about the carrier relaxation mechanism. Three distinct regions can be identified as associated with changes in the excitation energy. At high excitation energies, all QDs are excited and participate in the light em
122
continuum states. At low excitation ene ce of the continuum states is such that the carrier relaxation is restricted to a multi-LO phonon process. At medium energy, the QDs are resonantly excited through discrete electronic states, but the relaxation of the carriers does eet the multi-phonon resonance condition with the help of a continuum tail. For ra-dot electron-hole scattering provides an
ed with the simple QD structure, the InAs/
0.15 0.85
efficient channel for the release of energy via the continuous states of either electrons or holes.
(3) Quantum dot infrared photodetectors (QDIPs) with different confinement barrier schemes for InAs QDs were investigated. Compar
In Ga As dots-in-a-well (DWELL) structure pushes the detection band to a longer wavelength to meet the 8-12 μm atmospheric transmission window and enhances the TE absorption due to the reduced QD confinement in the vertical direction. By means of inserting a thin AlGaAs layer on top of the InAs QDs in the DWELL structure, we design a new confinement-enhanced DWELL (CE-DWELL) structure which further enhances the TE absorption by about 2 times due to the enhanced lateral confinement of the QDs. More importantly, due to the enhanced oscillator strength and the increased escape probability of photo-excited electrons, the quantum efficiency increases more than 20 times and the detectivity is an order of magnitude higher at 77K. By tailoring the confinement schemes for QDs, one can effectively change the device characteristics of QDIPs to fit different application requirements.
(4) The device parameters of the CE-DWELL QDIPs were further investigated to enhance the device performance. The thickness and Al content of the AlGaAs insertion layers influence much not only the absorption property but also the transport property of the device.
The device with thinner Al Ga As layers possesses stronger normal-incident absorption and higher responsivity, while the device with thicker Al Ga As layers possesses better suppression of the dark current and is more suitable for the high temperature operations.
Furthermore, compared with the Al Ga As layer, the Al Ga As layer provides stronger
123
ques to in situ growth semiconductor zero-dimensional quantum structures barrier confinement and causes less In-Ga intermixing for the InAs QDs and thereby generates better performance of the device. With appropriate device parameters of CE-DWELL, it is possible to achieve high quantum efficiency, high operating temperature, and 8-12 μm LWIR detection at the same time.
(5) Quantum ring infrared photodetectors (QRIPs) with the broadband, LWIR detection were investigated. The thinner InAs thickness and wider extent of the QR structure generates higher but closer electron state energies and weaker wavefunction confinement. This induces the specific features of QRIPs such as wider photocurrent spectra, more stable responsivity with temperature change, and lower dark current activation energy. The broad spectral response comes from both the bound-to-bound and the bound-to-continuum transitions. The higher confinement state energy and the extended wavefunction in QRIPs induce better carrier capture into and carrier flow within each QR layer and therefore the temperature-insensitive but lower responsivity, similar to the QWIPs. In order to compensate for the inherently smaller ionization energy of carriers in QRs, an Al0.27Ga0.73As high barrier layer is inserted into the QRIPs and the device operating temperature is effectively improved.
8.2 Future Work
In order to successfully accomplish the next generation optoelectronic technologies, epitaxy techni
should be continuously promoted, especially for improving the uniformity and epitaxial quality of the quantum objects. In our work, the MBE based growth techniques of self-assembled InAs QDs and QRs have been carefully studied. However, the control over the desired size and density should be further improved. QDs with very low density (~108 cm-2) and large enough size (dot height 5nm) are important for the single QD studies and also the precondition for ≧ the more advanced single QR studies. As mentioned in section 3.3, the reproducibility of the segmental indium flux growth QDs is unfortunately not good. More efforts to obtain
124
nd magnetic-PL can be performed on a single ring/volcano to study
tection, we have developed a promising E-DWELL structure for high performance QDIPs. It is worthwhile to further optimize the E-DWELL QDIPs. From the application point of view, the way to enha
well-controlled growth conditions for such low-density but large-size QDs are needed. Once such QDs are reproducibly obtainable, ultra-low density quantum rings or quantum volcanos can be readily achieved using the capping/annealing technique described in section 3.4, and then the PL, PL-excitation a
the energy spectra without inhomogeneous broadening. The possible physical changes on electronic states with the geometric change from dots to rings should be very interesting. On the other hand, for high performance QD detectors and QD lasers, high density QDs ( 5x10≧ 10 cm-2) with appropriate dot size and epitaxial quality are definitely required. From our accumulated experience, the growth control for QDs with the density about 2-3 x1010 cm-2 has been well developed. However, when it comes to the range of 5x1010 cm-2, the QDs’ uniformity and optical quality become hardly controlled. Moreover, severe strain energy accumulation for the growth of multi-layer, high density QDs which are needed for optoelectronic devices would be an issue. More detailed studies on growth control for the optimal QD conditions for QD devices are needed.
For infrared detectors aimed for the LWIR de C
device parameters for C
nce the quantum efficiency (QE) is essential no matter how it is achieved. Unfortunately, the weakest point of QDIPs is exactly the poor QE. The state-of-the-art QE reported for the QDIPs with LWIR detection band is only about 5% [8.1], much lower than that commonly reached by QWIPs (20~30%). In QWIPs, the average carrier density in one well is 3x1011 cm-2 with 30~50 QWs in the active region. In contrary, the usual QD density is less than 5x1010 cm-2
nce the quantum efficiency (QE) is essential no matter how it is achieved. Unfortunately, the weakest point of QDIPs is exactly the poor QE. The state-of-the-art QE reported for the QDIPs with LWIR detection band is only about 5% [8.1], much lower than that commonly reached by QWIPs (20~30%). In QWIPs, the average carrier density in one well is 3x1011 cm-2 with 30~50 QWs in the active region. In contrary, the usual QD density is less than 5x1010 cm-2