Download Few-cycle terahertz generation and spectroscopy of nanostructures

Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
10.1098/rsta.2003.1315
Few-cycle terahertz generation and
spectroscopy of nanostructures
By J u r a j D a r m o, T h o m a s M ü l l e r, W o l f g a n g Parz,
J o s e f K r ö l l, Gottfried Strasser a n d K a r l Unterrainer
Institute of Solid-State Electronics, Vienna University of Technology,
Floragasse 7, 1040 Vienna, Austria ([email protected])
Published online 19 December 2003
We report on new schemes for terahertz (THz) generation. The THz efficiency of
photoconducting antennas can be increased by using a cavity effect for the nearinfrared pump beam. The cavity is formed by a molecular beam epitaxy grown semiconductor Bragg mirror below the photoconducting layer. The optical confinement is
accompanied by an electrical confinement suppressing undesired leakage currents and
providing a constant electric field in the active layers. The performance of this cavityenhanced emitter is further improved by using a mobility optimized low-temperature
GaAs layer. This emitter is successfully used in a femtosecond Ti:sapphire laser cavity for highly efficient intracavity THz generation, where the photoconductive layer
serves also as a saturable absorber. The broadband THz pulses generated are used
for time-resolved spectroscopy of nanostructures. We study the dynamics of intersubband transitions in semiconductor quantum wells. The relaxation of carriers excited
by a near-infrared pump pulse is investigated by measuring the THz absorption
between the different subbands with our THz pulses. For transition energies below
the optical phonon energy we find relatively long relaxation times with a strong
dependence on the excited carrier density.
Keywords: terahertz; spectroscopy; generation; nanostructures
1. Introduction
The generation of few-cycle terahertz (THz) radiation into free space from a biased
photoconductive gap illuminated by an ultrashort laser pulse has been known for
more than a decade (Grischkowsky et al . 1990). Since then, many attempts have
been presented to increase the generated THz power without compromising the bandwidth of the output radiation (Darrow et al . 1992; Cai et al . 1997). In this paper
we report on a photoconductive THz emitter based on low-temperature grown GaAs
(LT GaAs) and integrated with a semiconductor-based Bragg mirror. This design
improves the emitter’s THz output power by about one order of magnitude. The
optical resonance and confinement of the photogenerated carriers in the high electric
field region of the LT GaAs layer are responsible for the observed enhancement of the
One contribution of 16 to a Discussion Meeting ‘The terahertz gap: the generation of far-infrared radiation
and its applications’.
Phil. Trans. R. Soc. Lond. A (2004) 362, 251–262
251
c 2003 The Royal Society
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
J. Darmo and others
(a)
LT GaAs
}
10–30 pairs
AlGaAs/AlAs
semi-insulating GaAs
substrate
absorption in GaAs layer
252
1.0
(b)
0.8
0.6
0.4
with Bragg mirror
without Bragg mirror
0.2
0
1 2 3 4 5 6 7 8 9
relative optical thickness, d / λ 0 (arb. unit)
Figure 1. (a) Structure schematics of the photoconductive THz emitter. (b) Absorption in the
GaAs layer without (broken line) and with (solid line) underlying Bragg mirror calculated for a
wavelength of 800 nm.
THz emission. In addition, we have focused on an optimization of the growth temperature of the LT GaAs layer with respect to a maximum photoresponse of the material
and a maximum breakdown field. It is known that annealed LT GaAs changes its
properties (resistivity, carriers’ lifetime) with the growth temperature (Look 1993).
Therefore, we designed and tested a multilayer LT GaAs structure grown at optimized temperature to increase the performance of THz emitters. Finally, we show
an application of a Bragg-mirror-based photoconductive emitter for THz generation
in the cavity of an fs Ti:sapphire laser.
In the second part of this paper, we will present results of time- and frequencyresolved spectroscopy of semiconductor nanostructures. We applied an interband
pump/intersubband probe technique to monitor the temporal evolution of the electron population in the first and second subbands of an undoped GaAs/AlGaAs asymmetric double quantum well after an optical interband excitation using broadband
THz pulses. The spacing between the two subbands is smaller than the optical phonon
energy. The time dependence of the intersubband absorption can be explained by a
simple rate equation model. We extract an intersubband lifetime of T21 = 14 ps at
an excitation density of 1 × 1010 cm−2 .
2. Terahertz generation
(a) Optical model of THz emitter
A schematic of the photoconductive THz emitter structure is shown in figure 1a. The
structure consists of a multilayered Bragg mirror with a semi-insulating GaAs substrate below and LT GaAs on the top. The Bragg mirror contains layers of AlGaAs
and AlAs to avoid absorption in the mirror. The optical parameters of all semiconductor materials were taken from Palik (1985). The structure forms an optical
cavity with absorbing medium (semiconductor GaAs) in it. We have modelled the
total absorption and reflection of this structure using the transfer matrix method in
the limit of linear absorption. Due to the optical cavity effect, the calculated absorption is not a monotonic function of the GaAs layer thickness but shows resonance
maxima and off-resonance minima (see figure 1b). The position of the maxima is
given by the optical thickness of the GaAs layer equal to 12 ( 12 + n)λo , where λo is
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
Terahertz generation and spectroscopy
253
a wavelength of the incident light. An absolute maximum is achieved for a layer
with an optical thickness of 54 λo . At this thickness the absorption reaches more than
90% for an absorption coefficient of α = 1.335 × 104 cm−1 in GaAs at 800 nm (Palik
1985). This value is larger than that achievable for a GaAs layer without Bragg mirror due to intrinsic 30% reflection of the incoming light at the air–GaAs interface.
Here, the resonance in the optical cavity effectively cancels the reflection of the light
at the interface. For larger thicknesses both types of the emitter structure approach
the same absorption level regardless of the presence of a Bragg mirror. From the
calculation we draw two main conclusions: nearly 100% absorption can be achieved
in a relatively thin layer of GaAs, and an absorption from a few per cent to 50% can
be obtained for a GaAs layer as thin as 65 nm. While the first case is relevant for
THz emitters in standard mode of operation, the second case applies for applications
where the thickness of the absorbing layer plays a negative role due to dispersion in
it, e.g. in the intracavity generation of the radiation (Darmo et al . 2003).
(b) Extra-cavity terahertz emitters
Although photoconductive THz emitters are routinely used for time-domain spectroscopy and THz imaging (Mittleman 2003), there was no attention paid to the
improvement of the emitter performance. We have focused on two strategies to optimize the THz emitter. First, with respect to the results of the optical model of the
emitter, we have implemented a Bragg mirror into the emitter structure. Second, we
have focused on the properties of the photoconductive LT GaAs layer itself.
A low-temperature molecular beam epitaxy (MBE) GaAs layer grown at temperatures in the range 220–350 ◦ C and annealed in situ at 600 ◦ C (10 min) was used as
a photoconductive material. The modified THz emitter is made of a 50–350 nm thick
layer of LT GaAs grown directly on a Bragg mirror (figure 1a). The Bragg mirror
itself consisted of 10–30 pairs of Al0.2 Ga0.8 As/AlAs layers designed for a centre wavelength of 800 nm. As reference emitter a 2 µm thick LT GaAs layer grown directly
on high resistive GaAs (100) substrate was used. Moreover, a bare semi-insulating
GaAs was also used as reference. Electrical contacts to the photoconductive layer
were made of Ti/Au metal and had the shape of coplanar 20 µm wide striplines
separated by a gap of 300 µm.
Emitter chips 2×3 mm2 in size were mounted onto a highly resistive silicon aplanar
hemisphere lens to efficiently couple the generated THz radiation into free space. The
emitters were excited using ultrashort optical pulses (pulse width ca. 10 fs, centre
wavelength ca. 800 nm, spectral width ca. 120 nm, repetition rate ca. 76 MHz and
pulse energy ca. 2.5 nJ) from a solid-state pumped mode-locked Ti:sapphire laser.
The laser beam was focused onto the 300 µm gap between the polarized electrodes
of the emitter. The shape of the THz pulse and the average output power were
measured by electro-optic sampling (Wu & Zhang 1996) and by an autocorrelation
technique with a calibrated Si bolometer (Kersting et al . 1998), respectively.
(i) Role of a Bragg mirror
First we have studied the contribution of the Bragg mirror to the performance of
the THz emitter. We have compared emitter structures with and without a Bragg
mirror. The spectra obtained by fast Fourier transformation of the time domain THz
signal from both emitter types are shown in figure 2. The THz electric field emitted
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
254
J. Darmo and others
electric field
(arb. units)
spectral intensity (arb. units)
1.0
0.8
0.6
0.4
6
4
2
0
−2
−4
−1 0
1
2
time delay (ps)
3
0.2
0
0.5
1.0
1.5
frequency (THz)
2.0
2.5
Figure 2. Normalized power spectrum of emission from a THz emitter with (circles) and without
(squares) Bragg mirror. THz transients corresponding to the calculated spectra are shown in
the inset.
from the emitter with a thin LT GaAs layer but with a Bragg mirror is about twice
as strong as that of the emitter without the Bragg mirror but with a 2 µm thick LT
GaAs layer. In terms of power, the emitter with the Bragg mirror emits four times
more THz power at the peak frequency (figure 2).
Let us now quantify the change in the efficiency of THz emission due to the inclusion of the Bragg mirror in the emitter structure. The calculated absorption of 800 nm
light is 21% for the 326 nm thick GaAs layer and ca. 69% for the 2 µm thick one.
Thus, with respect to the measured THz output power, which is four times larger,
an increase of more than 10 in the THz emission efficiency is achieved for the THz
emitter with a Bragg mirror (Darmo et al . 2003). The bias dependence of the electric
field amplitude of the generated THz pulses shows for both types of emitters a linear
increase up to 240–260 V. At biases above 260 V, a heating-related device failure
was observed for long-time operation. The observed improved efficiency of the THz
generation for the emitter with a Bragg mirror is explained by a stronger electric
field and a higher density of photogenerated charge carriers in the active region of
the modified THz emitter (Darmo et al . 2003).
(ii) Role of LT GaAs growth temperature
Another way of improving the performance of the THz emitter is to optimize the
photoconductor properties. LT GaAs exhibits a wide range of electrical parameters
depending on the growth conditions and thermal history of the material (Look 1993).
To access the photoresponse of LT GaAs grown at various temperatures, we used an
attenuated unfocused laser beam from the Ti:sapphire laser to homogeneously illuminate the area between the electrodes of the emitter structure. In this configuration
the measured DC photocurrent depends on the lifetime and mobility of the photogenerated charge carriers. As expected, the photocurrent of the structures exhibits
a clear monotonic decrease with the decreasing growth temperature of LT GaAs
(figure 3).
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
Terahertz generation and spectroscopy
105
255
300
200
103
DC photocurrent
(arb. units)
dark conductivity
(arb. units)
104
102
101
100
10−1
G220 T220 T270 T300 T320 T350
100
2
1
0
G220 T220 T270 T300 T320 T350
Figure 3. Dark current and photocurrent for different THz emitters measured at a bias of
100 V. (G220, non-annealed GaAs grown at 220 ◦ C; T, annealed GaAs grown at the temperature
shown.)
600
G220
T220
T270
T300
T320
T350
THz signal (arb. units)
500
400
300
200
100
0
50
100
150
200
emitter bias (V)
250
300
Figure 4. THz emission from different emitters as a function of bias. The emitters are compared with a THz emitter without Bragg mirror. (G220, non-annealed GaAs grown at 220 ◦ C;
T, annealed GaAs grown at the temperature shown.)
To compare the emitters with respect to their THz emission efficiency, we have
excited the emitter structures with pulses from an fs laser. Figure 4 shows the bias
dependence of the amplitude of the THz transients for emitters made of 350 nm thick
LT GaAs grown at temperatures between 220 and 350 ◦ C on the semiconductor Bragg
mirror. The steepest rise of the THz emission with an emitter bias was observed for
the structure made of LT GaAs grown at 320 ◦ C. On the other hand, the emitter
structure made of LT GaAs grown at the lowest temperature of 220 ◦ C exhibits the
weakest increase of the THz radiation with bias. The lowest THz generation efficiency
correlates well with the weakest photoresponse observed for this type of THz emitter
(see figure 3) and is due to the short lifetime of photogenerated carriers and their
low mobility. In all cases, the THz amplitude increases linearly with the applied bias
up to ca. 200–260 V depending on the growth temperature of LT GaAs. At biases
above this level the THz emitters exhibit a heating-related device failure.
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
(a)
1.0
0.8
0.6
0.4
0.2
0
750
800
850
wavelength (nm)
900
spectral density (arb. units)
J. Darmo and others
spectral intensity (arb. units)
256
100
(b)
10−1
10−2
10−3
10−4
0
1
2
3
frequency (THz)
4
Figure 5. (a) Fs laser near-infrared pulse spectrum and
(b) THz emission from the intracavity THz emitter.
(c) Intracavity terahertz emitter
The generation of THz radiation within a femtosecond mode-locked laser cavity
has recently been demonstrated (Sarakura et al . 1997). A saturable Bragg reflector
(SBR) with a single quantum well (SQW) as a saturable absorber was employed.
The radiation was generated by means of quantum beats in the quantum well. More
efficient intracavity THz generation was obtained when the SQW structure and later
InAs were subjected to a static magnetic field (Sarakura et al . 1998; Liu et al . 2000),
but the maximum average THz power generated by these emitters was just 0.8 µW.
Alternatively, photoconducting antenna emitters have been demonstrated to be very
powerful THz generators (Cai et al . 1997) and the THz output power modulation
capability is an inherent property of this type of THz emitter.
We present a voltage-controlled intracavity THz emitter that comprises a Bragg
mirror stack and an LT GaAs layer (Look 1993). We employ an LT MBE grown
GaAs as an alternative to the SQW as a saturable absorber (Keller et al . 1996)
and as a THz emitter simultaneously. The THz radiation in the LT GaAs layer is
generated by photocarriers which are accelerated in the electric field that is applied
between two contacts on the LT GaAs layer. The LT GaAs layer thickness and the
strength of the electric field determine the power of generated THz radiation; hence
the output THz power can be voltage controlled and even modulated in the time
domain. The Bragg mirror stack is made of AlGaAs/AlAs layers grown by MBE
on a highly resistive GaAs substrate. The LT GaAs layer was grown subsequently
onto the Bragg mirror stack at 220 ◦ C and, after the growth annealed in situ, in
the MBE chamber at 600 ◦ C for 10 min. The electrical contacts to the LT GaAs
layer were fabricated by photolithographical patterning of Ti/Au layers and had
the shape of parallel metallic stripes of 20 µm width and 50 µm separation. The
processed emitter structure was mounted onto a highly resistive silicon substrate.
To form an emitter unit, a silicon hemispherical lens was attached to its back. The
emitter unit was placed into the cavity of an ultrafast mode-locked Ti:sapphire laser
(Femtolasers Produktions GmbH) as an end mirror and the intracavity laser beam
was focused onto the gap between the electrical contacts on the LT GaAs layer. The
THz emitter element itself ensures self-starting of the mode-locking. In contrast,
the InAs intracavity THz generators require an additional self-starting element (Liu
et al . 2000). A typical spectrum of the near-infrared laser and of the THz pulses is
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
Terahertz generation and spectroscopy
257
shown in the figure 5. The near-infrared pulse spectrum has a spectral width of 26 nm
at a centre wavelength of 823 nm, while the peak frequency of the THz radiation is
ca. 0.35 THz (Darmo et al . 2002).
The average THz output power scaled quadratically with bias up to 110 V, above
which electrical breakdown of the emitter was observed. The maximum average THz
power generated by this new type of intracavity THz emitter was ca. 7 µW for an
average optical intracavity power of 900 mW. Thus, an enhancement of the efficiency
by more than 50 compared with the previous results was achieved (Sarakura et al .
1997, 1998; Liu et al . 2000). In addition, we have demonstrated the capability of
an LT/SBR THz emitter to modulate the generated THz radiation. We were able
to electrically modulate the THz output with frequencies up to 100 kHz without
disturbing the near-infrared lasing and mode-locking (Darmo et al . 2002).
3. Spectroscopy of nanostructures
Intersubband absorption spectroscopy is an important tool for studying electronic
states in n- or p-doped quantum well structures. Photo-induced intersubband absorption (PIA), where electrons and holes are created in an undoped sample using abovebandgap radiation, has also been reported. Here we present a time-resolved PIA
technique that allows the direct monitoring of the temporal evolution of the electron population in different subbands of an undoped quantum well structure after
interband excitation using broadband THz pulses (Müller et al . 2001). In our experiment an interband pump pulse injects electrons into the two lowest subbands of
an undoped Al0.3 Ga0.7 As/GaAs asymmetric double quantum well (ADQW) with a
level spacing smaller than the longitudinal optical (LO) phonon energy. The time
evolution of the electron population in these two subbands is monitored by probing
the THz intersubband transitions to the empty level 3 (figure 6a).
(a) Experiment
The ADQW was grown by MBE on a semi-insulating substrate. It consists of
40 periods of a 75 Å GaAs well and a 65 Å GaAs well, coupled through a 25 Å
Al0.3 Ga0.7 As/GaAs barrier. Two parallel cleavage facets of the sample were polished
at an angle of 58◦ to the growth axis to obtain a waveguide for the THz radiation.
For the energy spacing between the two lowest electron subbands we obtain E21 =
15 meV by solving the Schrödinger equation in the envelope-function formalism.
In our time-resolved PIA set-up we use a mode-locked Ti:sapphire laser that delivers 12 fs pulses (780 nm central wavelength, 110 nm bandwidth) with an average output power of 800 mW. Half of the laser intensity served as the interband pump to
excite the sample and the other part was focused on a 30 µm thick GaSe crystal
to generate a linear polarized THz probe beam by phase-matched difference frequency mixing (Kaindl et al . 1998). Spectroscopic measurements of the THz probe
were performed by using an interferometric autocorrelation technique (Heyman et
al . 1998). The intensity spectrum is shown in figure 6b. The THz probe pulses were
focused into the sample (T = 5 K) and the transmitted light was measured by using
a liquid-nitrogen-cooled HgCdTe detector. The pump beam that excites the sample
was focused to a spot diameter of ca. 350 µm, centred on the probe spot. A pulse
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
258
J. Darmo and others
100 (b)
3
probe
2
1
spectral density (arb. units)
(a)
10−1
10−2
pump
20
30
40
frequency (THz)
50
Figure 6. (a) Al0.3 Ga0.7 As/GaAs asymmetric double quantum well. (b) Intensity spectrum of
THz probe pulse generated by phase-matched difference frequency mixing in a 30 µm GaSe
crystal.
2–3 1–3
subband population (× ns)
(a)
1 ps
pu 2 ps
m 5 ps
p/
pr 10 ps
ob 20 ps
ed
ela 50 ps
y 100 ps
24
27 30 33 36
frequency (THz)
39
1.0
(b)
0.8
subband 1
0.6
0.4
0.2
subband 2
0
0
25
50
75
time (ps)
100
Figure 7. (a) Intersubband absorption spectra at different time delays after the pump pulse.
(b) Measured (symbols) and calculated (lines) population of subbands 1 and 2 as a function of
time delay after the interband excitation.
shaper was used to select the desired pump pulse spectrum (780 nm central wavelength, 10 nm spectral width), and a motorized translation stage controlled the delay
between the pump and the probe pulse.
(b) Results
Figure 7a shows PIA spectra taken with p-polarized THz probe pulses at different
delay times after excitation. The spectra were obtained by chopping the pump beam
and dividing the measured absorption by the reference spectrum obtained by chopping the probe (pump beam blocked). The spectra clearly exhibit two absorption
peaks: one at ca. 27.1 THz, the other one 3.4 THz above this value. The low-energy
peak is due to the (2–3) intersubband absorption, while the second peak is attributed
to the (1–3) absorption. The amplitude of the first peak decreases with the time delay
after excitation due to intersubband relaxation, while the amplitude of the second
peak first rises slightly and subsequently decreases due to carrier recombination.
Since the area under the peak (i − 3) (i = 1, 2) is directly proportional to the
subband population, ni (t), we are able to determine the population dynamics in
the quantum well on the basis of the time-resolved absorption spectra. Figure 7b
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
Terahertz generation and spectroscopy
259
shows the electron population of the first and second subbands as a function of time
delay after optical excitation (symbols). About 40% of the photo-excited carriers
(1 × 1010 cm−2 ) are injected into the second subband, while the remaining 60%
are injected into the first subband at a higher value of k. The population in the
second subband (squares) shows an exponential decay. The carriers relaxing down
from the second subband add to the population in the ground level. Subsequently,
the population in the ground level drops due to carrier recombination. The lines in
figure 7b are the results obtained from a simple rate equation model. By fitting the
calculation to the experimental data, we deduce an intersubband relaxation time of
14 ps.
(c) Discussion
We will now discuss possible intersubband scattering mechanisms and compare
numerical estimates of scattering rates with the experimental results. Since the
intersubband spacing E21 in our sample is smaller than the LO phonon energy,
the electrons in the second subband do not possess sufficient energy to emit LO
phonons and intersubband relaxation can only be due to acoustic phonon emission
and electron–electron (e–e) scattering. Using the model described in Požela et al .
(2000), we calculated the acoustic phonon scattering time to be T21 = 320 ps for our
structure. This time is much too long to explain our experimental findings. Recent
experiments, however, show that the intersubband e–e scattering rates can be very
high, in some circumstances almost approaching the intersubband scattering rate due
to LO phonon emission (Hartig et al . 1999). We calculated the e–e scattering rates
in the Born approximation using static single subband screening within the random
phase approximation. The most prominent intersubband e–e scattering processes are
labelled 2211 and 2221, where ijf g describes an interaction, where an electron in
state i scatters to f under collision with a second electron, which scatters from j to
g. When working out the total population transfer rate between the first and second
subbands, both scattering processes and the number of electrons which are transferred by each process were taken into account. For a population of 5 × 109 cm−2
electrons in the second subband, we calculated an intersubband e–e scattering time of
18.3 ps. This value is in good agreement with the experimental result. Some authors
have also observed a strong reduction of the intersubband relaxation time due to fast
relaxation of electrons in the high-energy tail of the hot carrier distribution in the
second subband by emission of LO phonons (Levenson et al . 1990). Due to the small
intersubband spacing of our sample and the low excitation density, however, the
injected electron population in the second subband is cold enough that LO phonon
emission is suppressed.
The work was done within the framework of the European Community project IST-1999-10154
(Teravision). The authors acknowledge G. Tempea (Photonics Institute, VUT Vienna) for his
contribution to the optical design of the emitter and Femtolasers Produktions GmbH (Vienna,
Austria) for fruitful collaboration on the intracavity THz emitter. The authors also acknowledge
support from the Austrian Fond zur Förderung des Wissenschaftlichen Forschung (SFB-ADLIS).
References
Cai, Y., Brener, I., Lopata, J., Wynn, J., Pfeiffer, L. & Federici, J. 1997 Design and performance
of singular electric field terahertz photoconducting antennas. Appl. Phys. Lett. 71, 2076–2078.
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
260
J. Darmo and others
Darmo, J., Müller, T., Strasser, G., Unterrainer, K., Le, T., Stingl, A. & Tempea, G. 2002
Voltage-controlled intracavity terahertz generator for self-starting Ti:sapphire lasers. Opt.
Lett. 27, 1941–1943.
Darmo, J., Müller, T., Strasser, G., Unterrainer, K. & Tempea, G. 2003 Terahertz emitter with
integrated semiconductor Bragg mirror. Electron. Lett. 39, 460–462.
Darrow, J. T., Zhang, X.-C., Auston, D. H. & Morse, J. D. 1992 Saturation properties of largeaperture photoconducting antennas. IEEE J. Quant. Electron. 28, 1607–1616.
Grischkowsky, D., Keiding, S., van Exter, M. & Fattinger, C. 1990 Far-infrared time-domain
spectroscopy of dielectrics and semiconductors with terahertz beams. J. Opt. Soc. Am. B 7,
2006–2015.
Hartig, M., Ganiere, J. D., Selbmann, P. E., Deveaud, B. & Rota, L. 1999 Density dependence of
carrier-carrier-induced intersubband scattering in GaAs/AlGaAs quantum wells. Phys. Rev.
B 60, 1500–1503.
Heyman, J. N., Kersting, R. & Unterrainer, K. 1998 Time-domain measurement of intersubband
oscillations in a quantum well. Appl. Phys. Lett. 72, 644–646.
Kaindl, R. A., Smith, D. C., Joschko, M., Hasselbeck, M. P., Woerner, M. & Elsaesser, T. 1998
Femtosecond infrared pulses tunable from 9 to 18 µm at an 88-MHz repetition rate. Opt. Lett.
23, 861–863.
Keller, U., Wiengarten, K. J., Käertner, F. X., Kopf, D., Braun, B., Jung, I. D., Fluck, R.,
Hönninger, C., Matuschek, N. & Aus der Au, J. 1996 Semiconductor saturable absorber
mirrors (SESAM’s) for femtosecond to nanosecond pulse generation in solid-state lasers. IEEE
J. Select. Topics Quant. Electron. 2, 435–451.
Kersting, R., Unterrainer, K., Strasser, G., Kaufmann, H. F. & Gornik, E. 1998 Coherent plasmon in n-doped GaAs. Phys. Rev. B 58, 4553–4559.
Levenson, J. A., Dolique, G., Oudar, J. L. & Abram, I. 1990 Intersubband carrier relaxation in
highly excited GaAs/AlGaAs multiple quantum wells. Phys. Rev. B 41, 3688–3694.
Liu, Z., Ono, S., Ohtake, H., Sarukura, N., Liu, T.-A., Huang, K.-F. & Pan, C.-L. 2000 Efficient
terahertz radiation generation from a bulk InAs mirror as an intracavity terahertz radiation
emitter. Jpn. J. Appl. Phys. 39, L366–L367.
Look, D. C. 1993 Molecular beam epitaxial GaAs grown at low temperatures. Thin Solid Films
241, 61–73.
Mittleman, D. 2003 Sensing with terahertz radiation. Springer.
Müller, T., Bratschitsch, R., Strasser, G. & Unterrainer, K. 2001 Intersubband absorption
dynamics in coupled quantum wells. Appl. Phys. Lett. 79, 2755–2757.
Palik, E. D. 1985 Handbook of optical constants of solid. Academic.
Požela, K., Naran, S. B., Leburton, J.-P. & Heyman, J. N. 2000 Structure variation of intersubband electron-acoustic phonon scattering rate in coupled quantum wells. Appl. Phys. Lett.
77, 265–267.
Sarakura, N., Liu, Z., Ohtake, H., Izumida, S., Yamanaka, T., Segawa, Y., Itatani, T., Sugaya,
T., Nakagawa, T. & Sugiyama, Y. 1997 All-solid-state, THz radiation source using a saturable
Bragg reflector in a femtosecond mode-locked laser. Jpn. J. Appl. Phys. 36, L560–L562.
Sarakura, N., Ohtake, H., Liu, Z., Itatani, T., Sugaya, T., Nakagawa, T. & Sugiyama, Y. 1998
THz-radiation generation from an intracavity saturable Bragg reflector in a magnetic field.
Jpn. J. Appl. Phys. 37, L125–L126.
Wu, Q. & Zhang, X.-C. 1996 Ultrafast electro-optic field sensor. Appl. Phys. Lett. 68, 1604–1606.
Discussion
J. Faist (Institute of Physics, University of Neuchâtel, Switzerland ). I thought that
in photoconductive switches, the limitation was basically the thermal breakdown, the
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
Terahertz generation and spectroscopy
261
voltage being limited by the power you can apply. By including a Bragg reflector,
you are certainly increasing the thermal resistance.
K. Unterrainer. We do not observe that the Bragg reflector increases the thermal
resistance. To get more power, one can simply apply a higher bias, and you are
limited by the breakdown field. However, when you generate a lot of electron–hole
pairs in the substrate, it heats faster. The limit is, therefore, engineering the overlap
between the high electric field and the electron–hole pairs which are generated. You
can then go to higher breakdown fields.
G. Smith (Department of Physics and Astronomy, University of St Andrews, UK ).
You made a comment about Fourier transform instruments, compared with THz
generators. What is the difference in sensitivity between the two techniques at the
moment, forgetting about the time-resolved aspect?
K. Unterrainer. Fourier transform spectrometers are limited by background radiation. You overcome this problem in pulsed systems, because even when you have
a low average power, when you gate it, you get rid of the blackbody thermal radiation. But, with the exception of electro-optic sampling, we use the same detector
technology for the pulsed and the Fourier transform spectrometers. The gating technique makes electro-optic sampling very attractive because it works at room temperature; bolometers you have to cool down. And this is an important part of the
THz pulse technology—you have a room temperature detection technique. Even the
phase-resolved aspect can be done with asymmetric Fourier transform techniques.
G. Smith. But can you give me an order of magnitude in the difference in sensitivity,
or is the sensitivity comparable?
K. Unterrainer. I would compare the signal-to-noise of some trace gas, or ammonia. Maybe Paul Planken should comment. With the THz pulsed technique, do you
think you can detect lower trace gas contents than with a conventional FTIR?
P. Planken (University of Technology Delft, The Netherlands). The electro-optic
sampling is much more sensitive, I think.
X.-C. Zhang (Center for Terahertz Research, Rensselaer Polytechnic Institute,
Troy, NY, USA). If I can make a comment about this comparison. We have just
received a top-of-the line FTIR system from Bruker. We found that for many samples
the FTIR system does not work well below 1 THz, even if we use a 1.7 K bolometer.
We even tried some Space Shuttle foam, and the FTIR was unable to measure below
1 THz because the sample was too thick. Based on two weeks of study, we prefer
the FTIR system for a few THz, but for frequencies around 1 THz, or a little lower,
free-space electro-optic sampling is much, much easier.
H. Roskos (Physikalisches Institut, Johann Wolfgang Goethe-Universität, Frankfurt-am-Main, Germany). In this intracavity THz generation, could you gain by
going from photoconductive to electro-optic materials, because at the moment your
system has some limitations owing to the losses that are introduced by the absorption?
K. Unterrainer. No, because we need the absorption anyway for the saturable
Bragg reflector.
Phil. Trans. R. Soc. Lond. A (2004)
Downloaded from http://rsta.royalsocietypublishing.org/ on October 24, 2016
262
J. Darmo and others
J. Allam (Advanced Technology Institute, University of Surrey, Guildford, UK ).
You contrasted interband and intraband transitions, and said that intraband transitions were very useful for studying nanostructures, because you have a method
for directly detecting the polarization intensity as well as the phase. This type of
technique seems to be available at optical wavelengths with frequency-resolved optical gating (FROG) measurements. Is there a distinction between the THz detection
mechanism compared with those in FROG measurements which also give you the
amplitude and phase information?
P. Haring Bolı́var (Institut für Halbleitertechnik, RWTH Aachen, Germany). Perhaps just a comment with regard to this question. In FROG measurements you
extract some phase information, but that is not a direct measurement. It is basically
an indirect deconvolution of the spectral information. You have to assume certain
parameters for your pulse before you can try to derive the phase information from
the measured FROG data.
D. M. Mittleman (Department of Electrical and Computer Engineering, Rice University, Houston, TX, USA). And in FROG, you are also measuring relative phase,
not absolute phase, and you are not measuring with a time resolution that is less
than one optical cycle. At best, you are measuring time resolutions of several cycles.
Phil. Trans. R. Soc. Lond. A (2004)