In this chapter, the measurement result of PC antennas and photonic transmitters are shown and discussed. First, in chapter 5-1 and 5-2, we studied the LTG-GaAs and GaAs:O based PC antennas under pulse and CW excitation respectively. An comparable performance of GaAs:O to serve as substituted material of LTG-GaAs is demonstrated. After that, in chapter 5-3, we discussed the performance of STR-PD combined with slot antenna with a center frequency 500GHz. Such results show our high power and high speed STR-PD could be served as an important part of high power sub-THz emitter. Finally, in the end of this chapter, we further to study in detail the characterization of two high-power photonic transmitters based on two different kinds of high-power photodiodes UTC-PD and STR-PD. Such result help us to learn a fabrication of high power and high speed sub-THz emitter.
5-1 LTG-GaAs & GaAs :O PC Antennas under Pulse Excitation
Figure 5-1 shows the bias voltage dependence of the THz absolute intensity generated from different material with the laser beam focused near its anode with the same antenna structure. They were detected by the bolometer. According the Figure 5-1, we can observe that the THz power increases quadratically with the bias at higher bias voltage. GaAs:O has higher photocurrent than LT-GaAs, and GaAs:O
86
has high power when we biased higher than 60V. The THz peak power of GaAs:O is 5.2mW and LT-GaAs is 3.6mW when biasing at 80V and pumping with 30mW.
GaAs:O generate pulse THz power almost 1.4 times LT-GaAs.
10 20 30 40 50 60 70 80
Figure 5- 1(a) Photocurrent-bias (b)THz power-bias curve of GaAs:O and LT-GaAs antenna with the same antenna structure
87
The increasing ratio of THz power-bias voltage curve of GaAs:O are higher than LT-GaAs, although LT-GaAs has higher THz power at lower bias voltage. And there is no saturation effect in both materials under pulse laser pumping. We used TDS to characterize the GaAs:O and LT-GaAs antenna. Figure 5-2 shows that THz wave form of GaAs:O and LT-GaAs. The measurements are under different bias voltage (20V、40V、60V、80V) and pump power is fixed at 30mW.
0 20 40 60
Figure 5- 2 THz time domain waveform of (a) GaAs:O and (b) LT-GaAs at constant pump power 30mW
88
The THz spectra under different biases are shown in figure 5-3. We observed that the THz peak amplitude increases when bias increases, However, we can’t observe any spectral shift, and this also matches with previous literature[24].
.
Figure 5- 3 THz spectrum of (a) GaAs:O and (b) LT-GaAs at constant pump power 30mW
89
Figure 5-4 (a) show that the main peak amplitude of GaAs:O is higher than LT-GaAs under pump 30mW and bias 80V in pulse mode. And we observed both spectra are about 1THz.
Figure 5-5 (a) shows the pump power dependence of the THz intensity measured for our PC antennas using a constant bias voltage of 80V. Figure 5-5 (b) shows the bias voltage dependence of the THz intensity measured for our PC antennas using a constant pump power of 30mW. Due to the low pump power, we
0 10 20 30 40 50 60
Figure 5- 4 (a) Time domain waveform (b) Spectrum for both materials
90
can not observe the strong saturation effect in Figure 5-5 (a) in pulse mode.
Because the pump power doesn’t excite enough carriers to form reverse electric field. And also we can’t observe the saturation effect in Figure 5-5 (b) due to the low electric field. .
Figure 5- 5 (a) The amplitude-pump power curve at bias voltage 80V (b) The THz power-electric field curve at 30mW.
91
5-2 LTGaAs & GaAs :O PC Antennas under CW Excitation
Figure 5-6 shows the bias voltage dependence of the photocurrent and the THz absolute intensity generated from different materials with the two laser beams focused near its anode with the same antenna structure.
They are also detected by the bolometer. According the Figure 5-6, we can obtain that the THz power are almost the same for both materials when bias is smaller than 40V, the highest THz power of GaAs:O is 2.268μW and LT-GaAs is 1.27μW. GaAs:O has almost 1.8 times to LT-GaAs and there is no saturation effect in high bias voltage.
However we observe THz power saturation effect in low temperature grown GaAs [1].
Analysis and Discussion
Figure 5- 6 (a) Photocurrent-bias voltage curve and (b) THz power-bias voltage curve of GaAs:O and LT-GaAs fabricated with the same dipole antenna
92
Figure 5-7 shows the bias voltage dependence of the photocurrent measured for our photoconductive antennas using a constant pump power of 180mW which are 90mW for each other. According to the result, we observe that the photocurrent ratio of GaAs:O divide to LT-GaAs are increasing when bias is more than 20V, The bias voltage below 20V are not increasing due to non-uniform electric field and low signal-noise ratio. And according to the CW theory, the carrier density is proportional the carrier lifetime (n∝ ), electron drift velocity is proportional the mobility at low electric
τ
e field (v∝μ
e) and DC photocurrent is proportional to the carrier lifetime multiplemobility (IDCph =e
τ μ
e e), so the mobility of GaAs:O is between 2.5 and 3.75 times to LT-GaAs.Figrue 5-7 (a) (b) (c) (d) show the DC photocurrent-electric field curve of GaAs:N GaAs:O and LT-GaAs. We observe the increase of onset voltage for
20 40 60 80
Figure 5- 7 (a) Photocurrent-bias voltage curve of both materials and (b) photocurrent ratio of GaAs:O divide to LT-GaAs
93
quadratic behavior in the I-V curve corresponds to a consistent reduction of saturation in the P-I curve [1]. It means that the efficient of generation THz power with the same photocurrent are different when we implant different dosage density, due to different number of impact ion defects. The absence of additional very large defects (precipitates) which reduce the electron mobility may deteriorate the increase of carrier lifetime with voltage.
(a) (b)
Figure 5- 8 (a) Different dosage density of GaAs:N and LT-GaAs (b) Different dosage density of GaAs:N and LT-GaAs[48]
Figure 5- 9 photocurrent-electric field curve and THz power-DC photocurrent curve for GaAs:O and LT-GaAs DC
94
In figure 5-9, we observe that THz power of LT-GaAs is saturated by the pump power. The reason of the LT-GaAs saturation is due to radiation field screening. Note that at low optical intensities the values of THz power is linear with optical pump that is the inverse of photoconductance and is greater than load impedance, when the optical intensities increase, the inverse of photoconductance has the same order to load impedance, THz power would respond nonlinear with optical pump power. But we didn’t observe saturated effect occurred at GaAs:O, we attribute to higher conductivity compared to LT-GaAs
In figure 5-10 (a) we observe that photocurrent are super-linear behavior on both materials. The reason that photocurrent has quadratic increasing is due to carrier life time increasing with voltage and that would increase the photoconductance and, consequently, the photocurrent. And under radiation field screening and carrier life time increasing we observe that GaAs:O has slightly saturated. Other effects such as impact ionization may play a role as well.
Figure 5- 10 CW THz power-optical pump power for both materials.
0 50 100 150
0 1 2
THz power (μW)
pump power (mW) GaAs:O
LT-GaAs
95
In figure 5-11 show that the spectrum of GaAs:O and LT-GaAs measured by CW heterodyne system. We calculate the resonant frequency the theory value is 0.34THz and our measurement value is 0.36THz. The bandwidth of both materials is about 1THz.
They are identical to the value that we measured by THz-TDS system.
10 20 30 40
0.0 0.2 0.4 0.6 0.8
GaAs:O LT-GaAs
photocurrent (mA)
Flectric Field (kV/cm)
Figure 5- 11 (a) Photocurrent-electric field curve and (b) THz power-electric field curve.
10 20 30 40
0 1 2
THz power (
μW)
Electric Field (kV/cm) GaAs:O
LT-GaAs
96
0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.01
0.1
1 GaAs:O
LT-GaAs
THz power ( μ W)
Frequency (THz)
2
2 2
( )
(1 ( ) )(1 ( ) )
DC ph THz
e a
a
P I
w wR C
R R iX
∝ τ
+ +
= −
Figure 5- 12 THz spectrum of both materials measured by CW heterodyne system
97
5-3 STR-PD Combined with Slot Antennas
We directly excited our device by use of a femto-second Ti:sapphire (λ = 800nm) mode-locked laser to measure the radiated impulse responses of our transmitter. The excitation optical plus, which has the 82MHz repetition rate and 100fs pulse width, was focused by an objective on the input facets of our device. We used two parabolic mirrors to collimate and focus the radiated THz pulse on a liquid-helium-cooled Si bolometer for power measurement. The responsivity of the Si bolometer was calibrated with a blackbody radiation source. We also measured and calibrated the radiated THz beam propagation loss in air by measuring the THz intensity as a function of propagation distance. The propagation loss was about 0.082 cm-1. The RF spectra and waveforms of radiated signal were measured by a Martin-Puplett-type Fourier Transform Infrared Spectrometer (FTIR) with the same Bolometer as in power measurement. The FTIR system is consisted of two wire grid polarizers, the first one is used as reflector at the input port of system and the second one act as a beam splitter. By scanning the movable retro-reflector, the interference spectrum (waveform) of the input sub-THz wave in the time domain could be obtained. The frequency responses of radiated signal can thus be obtained through the use of Fourier-transform techniques. The measured interference patterns and corresponding transformed spectra of devices with a 23μm and a 60μm active lengths are shown in Figure 5-12(a) and (b) and their insets, respectively. Both devices exhibit significant resonance at around 100GHz and 200GHz frequency, and a significant resonance at the designed resonant frequency (~500GHz) of slots antenna has only been observed for the device with a short active length, as shown in (a).
98
Figure 5- 13 The Fourier-transformed spectra of our photonic-transmitters with different active lengths, 23μm (a) and 60μm (b).
( The insets show the original interfere spectrum in the time domain measured by using the FTIR system)
For the device with a long active length, it should have a poorer speed performance and lower output power in the high frequency regime than a short device.
A much less apparent resonance at the designed frequency (~500GHz) of long device has thus been observed. The parasitic oscillations in 0.1~0.2 THz frequency regime may be attributed to the influence of integrated photodiode, which has not been included in our antenna design.
Figure 5-13 shows the measured sub-THz power versus injected optical power under different reverse bias voltages of devices with a 23μm active length. Under
~30mW average optical power excitation, as much as 5μW of sub-THz power was measured, which has been calibrated for around 4.3dB propagation loss after 12cm propagation. Besides, the output sub-THz power shows a near parabolic relation between the injected optical powers when it’s value is below 10mW.
99
Figure 5- 14 The measured sub-THz intensity vs. optical pumping power under different reverse bias voltages
On the other hand, when the optical power is over such value, saturation phenomenon of sub-THz power has clearly been observed. The saturation phenomenon of output sub-THz power under high optical power excitation may be attributed to the high photocarrier density resulted defect saturation in the LTG-GaAs active layer, thus that carrier lifetime will increase and lead the bandwidth degradation.
The measured bias-dependent sub-THz power of device with a 23μm active length under different optical pumping power is shown in Figure 5-14. The inset shows the measured photocurrent versus dc reverse bias under different optical pumping power. The maximum reverse bias voltage (-7V) is limited by the device failure. Regarding with the reported LTG-GaAs based photodetectors and photomixers [3,4,5], they usually suffered from serious bandwidth degradation and the saturation of radiated THz power under high reverse bias voltage and high external applied electric field (>50kV/cm) due to the lifetime increasing effect of
100
LTG-GaAs [3,6].
Figure 5- 15 The measured sub-THz intensity vs. reverse bias voltages under different optical pumping power for our photonic-transmitters with a 23μm active length.
(The inset shows the measured output photocurrent vs. reverse bias voltages under different optical pumping power.)
However, for our devices even under a highest reverse bias (-7V), corresponding to an external applied electric field up to 200kV/cm in the active region, saturation phenomenon of radiated power has not been observed. Furthermore, the output sub-THz power has an ideally quadric relation between reverse bias voltages due to that the output photocurrent is linear proportional to reverse bias voltage, as shown in the inset. The superior bias dependent performance of our transmitter to traditional LTG-GaAs based photomixer is because that, for our STR-PD, the external applied electric field is concentrated in the two high quality GaAs transport layers instead of LTG-GaAs based recombination center [7] and the lifetime increasing effect of LTG-GaAs layer under high electric field (>50kV/cm) can thus be neglected [6].
101
5-4 Comparison of UTC-PT and STR-PT
During device testing, an objective lens with a high numerical aperture (NA) value is used to focus the output of the fs Ti:Sappire mode-locked laser (λ=800nm) onto the edge of our devices. The excitation optical pulse has a repetition rate of 82 MHz and a pulse width of 100fs. Figures 3 (a) and (b) show the bias current versus voltage (I-V) of the two devices under different input powers. The active lengths of the measured UTC-PD and STR-PD based PTs (UTC-PT and STR-PT) is the same, around 23μm.
As can be seen in Figure 3, the average photocurrent generated by the UTC-PT is much higher than that of the STR-PT, even under lower bias and input pulse energy.
This can be attributed to the presence of the LT-GaAs based recombination center inside the STR-PD active layers, which effectively traps the photo-generated carriers and degrades the external quantum efficiency of the device. The generated photocurrent of the UTC-PT increases super-linearly with the reverse bias voltage when it is over 5V. This phenomenon can be attributed to the fact that the value of the external applied electrical field (3.6x105V/cm) in the Al0.12Ga0.88As-based collector layer is close to its breakdown field.
A super-linear increase of photocurrent vs. reverse bias voltage has also been observed for the LTG-GaAs based PD [6], which can be attributed to the lifetime increasing effect of LTG-GaAs layer is much less apparent for our STR-PD, perhaps because most of the external electric field is concentrated in the two high quality GaAs transport layers instead of the LTG-GaAs based recombination center, which has a high defect density and field-screening effect under reverse bias voltages
102
Sub-THz Average power (μW)
Reverse Bias (V)
[8].The lifetime increasing effect can thus be neglected. In addition, the avalanche induced photocurrent in the transport layers is not as significant as the case shown in (a), which may be attributed to the fact that most of the avalanche-generated carriers are also trapped by the recombination centers.
The total output power of the two devices was measured by a liquid-Helium cooled Si bolometer, which was carefully calibrated to the black body source. Two parabolic mirrors were used to collimate and focus the generated sub-THz into the bolometer. The propagation loss (0.082 cm-1) is also calibrated by measuring the THz intensity as a function of the propagation distance at a relative humidity of 40%.
Figure 5- 16 The bias dependent measured sub-THz power of (a) UTC-PT, (b) STR-PT, and (c) LTG-GaAs based PC under different optical pulse energy excitation. (d) shows the top-view of
PC dipole antenna
103
Figures 5-15 (a) and (b) show plots of the sub-THz power of UTC-PT and STR-PT emitted under different optical pumping pulse energies as a function of the bias voltage. As much as 8μW and 9μW of maximum sub-THz average power are measured for the STR - and UTC-PT respectively. We can clearly see that although the maximum average photocurrent of the UTC-PT is around four times higher than that of the STR-PT, as shown in Figure 5-15, the maximum average sub-THz power generated by the STR-PT is similar to that of the UTC-PT under a lower excitation optical pulse energy (0.12nJ vs. 0.48nJ). This result indicates that the recombination center in the STR-PD has eliminated most of the DC component of the photocurrent, so the AC component of the optical pulses can be efficiently converted to an electrical AC signal and then radiated into free-space. The STR-PD requires less operation current for the desired radiated power performance which implies that the thermal effect can be minimized during high-power operation.
In addition, as shown in figure 5-15 (a), when the reverse bias of UTC-PD is less than about -5V, the radiated sub-THz power exhibits an ideal quadratic relation to the reverse bias voltage. This is due to the linear dependence of the output photocurrent on the reverse bias voltage (<-5V), as shown in Fig. 5-14(a). When the reverse bias voltage increases further, however, the output sub-THz power exhibits a linear dependence on the bias voltage. This can be attributed to the fact that the value of the external applied electrical field (~3.6x105 V/cm) in the Al0.12Ga0.88As based collector layer is close to its breakdown field. Thus the avalanche-induced bandwidth degradation phenomenon of the PD limits the super-linear increase of output power versus the bias.
104
As can be seen in Figure 5-14 (a), the measured photocurrent clearly increases super-linearly with the bias voltage when it is over -5V. This indicates that the avalanche phenomenon really does occur. The bias dependent output power of STR-PT shows similar behavior to the measured bias dependent photocurrent, as shown in Figure 5-14(b). When the injected pulse energy increases from 0.01nJ to 0.12nJ, the emitted sub-THz power dramatically increases due to the huge increase of photocurrent; see Figure 5-14 (b). However, when the pulse energy exceeds 0.12nJ, the traces under higher injected pulse energy are crowded due to the phenomenon of recombination center saturation, which will be discussed latter in Figure 5-16.
Figure 5-15 (c) shows the measured bias dependent output sub-THz power from a typical photoconductive (PC) dipole antenna, which is fabricated on the LTG-GaAs layer [9] and mounted on the Si-lens. Such PT has a center frequency at 260GHz with around 1THz bandwidth. The measured values of power are close to those reported for an LTG-PC with an antenna of the same design [9]. Figure 5-15 (d) shows the top-view and geometric size of such device. We can clearly see that both of our
0 10 20 30 40
Sub-THz Average Power (μW) -11V -9V
-5V -1V
Input Power (mW) 0.12
(a) (b)
Figure 5- 17 The power dependent measured sub-THz power of (a) UTC-PT and (b) STR-PT under different reverse bias voltage.
105
devices can have around 10 times higher output sub-THz power than such LTG-GaAs based traditional PT by use of a much lower reverse bias voltage (~10V vs. 35V).
Figure 5-16 (a) and (b) shows the measured optical power dependent output sub-THz power under different reverse bias voltage of UTC-PT and STR-PT, respectively. As can be seen, the value of the saturated injected optical pulse energy of the UTC-PT (indicated by arrows in Figure 5-16(a)) increases with the reverse bias voltage. The observed saturation phenomenon of the UTC-PD should be attributed to the electron induced SCS effect in the collector layer [9], which can be minimized by increasing the reverse DC bias voltage [9,10]. On the other hand, the power dependent saturation behavior of the STR-PD is very different from that of the UTC-PD (indicated by the arrows in Figure 5-16(b)).
We can clearly see that the saturation of the optical pulse energy is the same around 0.12nJ regardless of an increase in the reverse bias voltage. This phenomenon may be attributed to that fact that the dominant saturation mechanism in the STR-PD is the defect saturation of the LTG-GaAs based recombination center,
We can clearly see that the saturation of the optical pulse energy is the same around 0.12nJ regardless of an increase in the reverse bias voltage. This phenomenon may be attributed to that fact that the dominant saturation mechanism in the STR-PD is the defect saturation of the LTG-GaAs based recombination center,