Henini M. Handbook of Self Assembled Semiconductor Nanostructures for Novel devices in Photonics and Electronics

Подождите немного. Документ загружается.

Quantum Dot Charge and Spin Memory Devices 479

remain rather simple [24] . The enhanced Coulomb interactions in QDs due to the strong con-

ﬁ nement give rise to energy shifts between different occupancy states of the order of a few milli-

electronvolts. Each occupancy state of electrons and holes, such as the single exciton (1 X  1 e  ),

bi-exciton (2 X  2 e  ), charged exciton ( X



 2 e  1 h , X



 1 e  ), etc. then have a unique tran-

sition energy. * Furthermore, since the Coulomb energy shifts are much larger than the homo-

geneous linewidth of each transition ( Γ

hom

 1), each occupancy state can be optically selected

via the frequency of either the emitted photon or laser used to address the system. This property

underpins the use of QDs for a diverse range of applications in the ﬁ elds of classical and quantum

information processing.

In order to experimentally observe the single particle non-linearities discussed above, one has

to probe single dots in isolation [25] . Otherwise the minimum energy scale that can be probed is

limited by the inhomogeneous linewidth of the ensemble ( Γ

inhomog.

20  50 nm). This necessity

is nicely illustrated in Fig. 15.2b , which compares PL recorded from an ensemble of InGaAs QDs

with data recorded from the same sample through shadow mask apertures [26] . In this experi-

ment, single dots could be isolated by performing luminescence measurements through submi-

cron diameter apertures in an opaque metal ﬁ lm. This measurement followed many others which,

already at the beginning of the 1990s, provided the ﬁ rst direct evidence for the discrete electronic

structure of QD nanostructures via spatially resolved optical studies of localized excitons in quan-

tum well nanostructures [27, 28] . Shortly after these ﬁ rst studies, various groups reported simi-

lar results on different QD systems using confocal microscopy [28, 29, 30] , near-ﬁ eld optics [31,

32] and cathodoluminescence [33] techniques. In the pioneering work of Brunner et al. [27, 28]

* As a further complication, transitions between quasi 0D orbital states in the dot and delocalized carriers in

the 2D wetting layer continuum can also provide an intrinsic continuum absorption background for energies

above the fundamental s–s transition [71] . This property is manifestly far away from the idea that such nanos-

tructures behave as artiﬁ cial atoms, while other properties of the dots, such as the observation of excitonic

Rabi oscillations, ﬁ rmly underscore the analogy between quantum dots and real atoms.

With Coulomb interactions

s-s

Energy

p-p

Without Coulomb

1200

Single dot

T  10 K

~10 QDs

~10

QDs

1250 1300

Energy (meV)

PL intensity (arb. units)

1350 1400

Figure 15.2 (a) Schematic representation of the form of the interband emission spectrum of a single QD without

(lower) and with (upper panel) Coulomb interactions between conﬁ ned carriers. As discussed in the text, Coulomb

interactions lift the degeneracy of the orbital states allowing each occupancy state to be individually spectrally

addressed. (b) Photoluminescence spectra recorded at low temperatures from an ensemble of InGaAs QDs, revealing

a large inhomogeneous linewidth ( Γ

 70 meV) and contrasting strongly with the homogeneous linewidth of a sin-

gle dot ( Γ

 1 μ eV). The upper single dot spectrum also clearly shows the inﬂ uence of Coulomb interactions on the

emission spectrum.

CH015-I046325.indd 479CH015-I046325.indd 479 6/24/2008 5:06:26 PM6/24/2008 5:06:26 PM

480 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

the discrete and highly non-linear optical properties of single QDs were highlighted, together

with the possibility of controlling the spin of individual charge carriers via the optical polariza-

tion selection rules [27] . Studies of this type revealed that energy splittings between 1X and 2X

transitions typically lie in the range of a few millielectronvolts, with 2X lying either to lower or

higher energy than 1X, depending on the importance of direct Coulomb interactions and few

particle correlation effects [34] .

A further important development in the study and control of few particle states in QDs was

establishing an electrical contact to single dots. This provides the unique possibility to control the

QD charge occupancy using Schottky photodiode structures and has led to a number of advances

such as photocurrent absorption spectroscopy [35] , and the electrical detection of coherent opti-

cal control of excitons [36, 37] . The selective charging of self-assembled QDs was ﬁ rst demon-

strated by Drexler et al. using capacitance spectroscopy of ensembles of InGaAs dots embedded in

the depletion region of a Schottky diode [38] . This was ingeniously extended by Warburton et al.

[39] , followed by other groups [39, 40, 41] , who combined this approach with near-ﬁ eld opti-

cal spectroscopy through metal shadow masks to directly study the interaction of excitons with

additional electrons [39, 40] or holes [41] . These studies of charged excitons in QDs revealed a

wealth of information on Coulomb exchange and correlation effects in single dots populated by a

controlled number of charge carriers. Most recently, single dot absorption spectroscopy has been

shown to be a particularly clean method to directly probe the fundamental quantum properties

of excitonic transitions in QDs [42, 43] and even to facilitate measurement of the spin of a single

electron via spin state selective experiments [44] . An up-to-date overview of the ﬁ eld of single

dot spectroscopy can be found in the proceedings of the biannual International Conferences on

Semiconductor Quantum Dots established in 2000, the most recent of which was held in May

2006 in Chamonix, France [45] .

15.1.3 Optical memory structures based on quantum dots

For the purpose of the present chapter, the single particle non-linearities discussed above play

an important role: they ensure that resonant optical excitation of QDs in the spectral range of

their s interband transition results only in the generation of a single electron–hole pair, the sin-

gle exciton, denoted by the nomenclature 1 X  1 e  . Upon exciting an inhomogeneously broad-

ened ensemble with a narrowband source, one selects a subset of the dots from the ensemble via

their exciton transition frequency. Other excitations, such as 2 X  2 e  X



 1 e  , X



 2 e  ,

etc. are detuned by many thousand linewidths from the exciton transition and are not excited via

any linear optical processes. In 1995 Shunchui Muto suggested that such an inhomogeneously

broadened ensemble of quasi zero-dimensional absorbing centres may be suitable for application

as an ultra-dense, wavelength domain, optical data storage medium [46] . The use of both spa-

tial and wavelength domains for data storage provide the potential to combine both high data

storage capacities (  1 TB cm

 2

) with very low switching energies. In this proposal, each bit of

information is represented by a small number of e  h pairs, potentially even a single charge,

stored within an ensemble of QD nanostructures. High data storage densities are attainable since

data are stored in both spatial and frequency domains. One problem associated with the use of

interband excitations for frequency domain optical data storage is that they are relatively short

lived (typically ~1 ns) and electrically neutral, raising signiﬁ cant problems related to the efﬁ cient

detection of such optically encoded information. This limitation can be overcome by implement-

ing device concepts for separation of the optically generated e–h pair over timescales shorter than

the recombination lifetime, and then sensing these charges to read out the optically encoded

information.

Over the past ten years, these ideas have been explored by a number of groups with ﬁ rst inves-

tigations focused on optical charge generation and non-selective electrical sensing of the stored

charge. In 1995 Yusa and Sakaki demonstrated an optically gated FET that incorporates a layer

of InAs QDs in the vicinity of a two-dimensional electron channel [47] . The observed photo-

switching operation of the channel resistivity in this device was attributed to the preferential

trapping of non-resonantly photo-generated holes in the dot layer. Finley et al. extended this

approach to conditions of resonant optical charge generation in the QD ground states and selective

CH015-I046325.indd 480CH015-I046325.indd 480 6/24/2008 5:06:26 PM6/24/2008 5:06:26 PM

Quantum Dot Charge and Spin Memory Devices 481

exciton ionization [48] . Later on, Shields et al. presented a similar scheme of electric readout in

nanoscale FET structures that eventually led to the observation of discrete photo-charging events

of individual QDs in the transport characteristics [49] . To date, a number of groups have shown

that single charges can be electrically [50, 51,52, 53] or optically [54, 55, 56] injected into QDs.

Alternatively, the quantum dots may be charged optically, an approach that, when combined

with wavelength selective writing or local probe techniques, could approach the situation where

each quantum dot stores a single bit of information as discussed in the introduction.

One of the crucial parameters for the use of InAs–GaAs QDs for optical memory applications

is the thermal activation barrier for the trapped carriers – the larger it is, the longer the storage

time becomes eliminating the need for frequent memory refresh cycles. Capacitance spectroscopy

has been shown to be a powerful method to investigate charge retention times [57] together with

deep-level transient spectroscopy and admittance spectroscopy for both InAs–GaAs [58, 59] and

Ge–Si QDs [60] . GaSb–GaAs dots have similar structural properties to the InAs–GaAs system, but

a signiﬁ cantly larger hole ground state localization energy is expected due to the type II band

alignment attractive for holes [61] . The thermal activation barrier of hole-charged GaSb–GaAs

QDs has been studied by admittance spectroscopy and shown to provide extremely long hole stor-

age times, in excess of milliseconds even at room temperature [62] .

In the following sections we discuss the operation of early prototype optical charge memory

structures with electrical readout of the stored charge. These experiments are then contrasted

with devices that enable all optical charging and readout later in the chapter. These “ all opti-

cal ” memory devices provide a very convenient method to directly orientate the spin of the opti-

cally generated charges via the optical polarization selection rules. We then continue to describe

experiments in which we probe the electron and hole spin relaxation dynamics in self-assembled

QD nanostructures. The spin orientation is shown to be remarkably stable, demonstrating that

motional quantization dramatically suppresses spin relaxation via spin orbit interactions.

15.2 Optically induced charge storage

15.2.1 Electrical detection of stored charge

A number of groups have investigated the effects of a self-assembled InAs QD layer on the lat-

eral transport properties of a two-dimensional electron gas (2DEG) in MODFET-type structures

[51, 56, 63, 64] . These works have shown that the QDs can efﬁ ciently act as controllable charge

traps and scattering centres which can be used to effectively tailor the transport properties of the

2DEG. Furthermore, several authors [51, 56] have demonstrated that such devices are strongly

photosensitive forming the foundation for a basic QD optical memory element or highly sensitive

QD phototransistor with single photon counting capabilities [65] .

15.2.1.1 Charge storage device structure

Optically induced charging of self-assembled InAs QDs has been studied in [48, 66] and using

an optically gated FET structure into which a layer of InAs quantum dots is embedded  50 nm

from a 2D electron or hole conducting channel. The operating principles of the devices are simi-

lar to a MODFET, as illustrated by Fig. 15.3a , with the exception that the metallic gate electrode

is replaced by a layer of self-assembled quantum dots that can be optically charged. By separat-

ing the 2D channel and QD layer using a wider band gap blocking barrier, photo-created excitons

can be selectively ionized leaving electrons (or holes) stored preferentially within the dots.

Figure 15.3b illustrates schematically the fundamental operating principles of these structures.

While the following discussion pertains to an electron storage device, the principles for holes, dis-

cussed below, are entirely analogous. The devices consist of a two-dimensional (2D) electron chan-

nel that was spatially separated by a blocking barrier from a layer of self-organized quantum dots

(upper panel of Fig. 15.3b ). The QD layer is embedded within the intrinsic region of a vertical p - i - n

junction. By fabricating separate ohmic contacts to both the 2D n -channel and buried p -contact,

the electric ﬁ eld in the intrinsic region and conductivity of the n -channel ( σ ) can be tuned by

varying the DC bias applied to the p - i - n junction ( V

). Following resonant optical excitation

CH015-I046325.indd 481CH015-I046325.indd 481 6/24/2008 5:06:27 PM6/24/2008 5:06:27 PM

482 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

of the quantum dots, the photo-created holes escape by thermal activation ( T  100 K) or by tun-

nelling ( T  100 K). By contrast, the electrons remain preferentially stored due to the presence of

the wider band gap-blocking barrier (middle panel of Fig. 15.3b ). The stored electrons within the

QD layer then selectively deplete the 2DEG, resulting in a reduction of its in-plane conductivity

which can be measured electrically. The magnitude of this conductivity change ( Δ σ ) is expected

to reﬂ ect the charge density stored in the QDs. The stored electrons are then removed from the

QDs by forward biasing the vertical p - i - n junction (lower panel of Fig. 15.3b ). This results in the

injection of holes into the QD layer, where they become trapped by the dots, recombining radi-

atively with the stored electrons. The device then switches back to its initial state and another

charge write–store–reset cycle can be performed.

Devices similar to those illustrated in Fig. 15.3b have been grown on a [100] orientated semi-

insulated GaAs substrate and nominally consisted of the following epitaxial layers: 500 nm

p (Be)  2 . 1 0

c m

 3

back contact followed by a 240 nm undoped (u.d.) GaAs spacer. After this,

2.25 ML of InAs was deposited at 530°C/0.04 MLs

 1

which forms the self-assembled QDs. This

QD layer was capped with 5 nm GaAs before growth of the blocking barrier. This consisted of

30 nm undoped Al

0.3

0.7

As and a ﬁ ve-period 2 nm(AlAs)/2 nm (GaAs) short period superlat-

tice to inhibit impurity segregation in the 2D electron channel. The 10 nm thick In

0.1

0.9

electron channel was then grown followed by a 40 nm-wide modulation-doped Al

0.3

0.7

barrier region. The modulation doping was incorporated into two narrow (1.5 nm) δ -doped

(~2  1 0

c m

 1 2

) GaAs quantum wells in an effort to inhibit persistent photoconductivity asso-

ciated with carrier excitation from DX centres in the Al

0.3

0.7

As modulation-doped region [67] .

Finally, the sample was capped with 10 nm of n -doped (4  1 0

c m

 3

) GaAs. A reference sam-

ple was also grown which was nominally identical to the above but without the QD layer. After

growth the wafers were processed into a Hall bar geometry using standard optical lithographic

techniques and separate ohmic contacts were established to both the 2D-electron channel and

the back p -contact to enable switching of the electric ﬁ eld as depicted in Fig. 15.3 .

MODFET

(a) (b)

DotFET

Source

Gate

Substrate

Drain

Drain Drain

Stored

Writing

n-contact

2DEG

InAs-QDs

p-contac

Erase

Drain

EnergyEnergyEnergy

Source

channel

QDs

Gate

 0 V

Gate

 0 V

hω

Figure 15.3 (a) Comparison of a conventional MOSFET and an optically switchable DotFET realized by embed-

ding a single layer of InGaAs quantum dots in the vicinity of a 2D conducting channel. (b) Schematic representation

of the device bandstructure and operating principles as discussed in the text.

CH015-I046325.indd 482CH015-I046325.indd 482 6/24/2008 5:06:27 PM6/24/2008 5:06:27 PM

Quantum Dot Charge and Spin Memory Devices 483

15.2.1.2 Demonstration of selective electron and hole generation

Figure 15.4 compares high power (5 Wcm

 2

) ensemble PL spectra recorded from the sample

with QDs (sample A – full line) and the reference sample without dots (sample B – dashed line).

Over the energy range from 950 to 1200 meV a series of well-resolved peaks labelled E

, n  0 ,

1, 2 and 3 are observed which are separated by 64  3 meV and exhibit a linewidth (FWHM) of

35 meV. The energy at which these features arise together with their characteristic dependence

on the excitation power density identiﬁ es them as arising from radiative recombination of s–s and

p–p transitions in the ensemble of QD ensemble studied. To higher energy, additional PL peaks

are observed which arise from recombination in the In

0.1

0.9

As quantum well (QW) and bulk

GaAs (GaAs) regions of the device. As expected, sample B (dashed line) generates no lumines-

cence over the energy range 800–1200 meV conﬁ rming the absence of optically active centres.

In order to investigate optical charge storage effects a series of spectrally resolved photo-con-

ductivity measurements was performed. This consisted of a 100 s illumination period after which

the temporal dependence of the sample conductivity  σ ( t ) was monitored before a reset electrical

pulse was applied to the p - i - n junction [68] . The excitation energy was then changed and the cycle

repeated. For these measurements the samples were held at T  145 K in a stabilized environment

designed to eliminate parasitic effects arising from ambient illumination and another persistent pho-

toconductive effect that was found to be present for T  100 K. Optical excitation was via a power-

stabilized quasi monochromatic white light from a tungsten halogen lamp ( P

max

 1 mWcm

 2

FWHM  5 meV) and the effect of exciting different quantum dots within the ensemble was studied

by tuning the excitation energy ( E

) through the QD ensemble absorption spectrum.

Figure 15.5 shows σ ( t ) data obtained from sample A as a series of three illumination, recovery

and reset phases were performed with E

 1016 meV, 1024 meV and 1059 meV, respectively. The

onset of illumination is denoted by the label A, with the label B showing the point at which the

shutter is closed. The reset pulse is applied to the sample at C . The channel conductivity prior to

illumination is denoted by σ

, which for this sample was measured to be 1.5  0.2 mS. As can be

seen clearly in Fig. 15.5 , for E

 1016 meV the sample responds only weakly to illumination. By

contrast, after increasing E

the photo-effect becomes much stronger until for E

 1020 meV a

rapid switching of σ is observed following illumination. We focus now on the illumination cycle at

 1059 meV [69] After closing the shutter (B) σ recovers weakly, saturating at a new level σ

Without resetting the sample, σ remained at σ

over timescales longer than eight hours at 145 K

while recovering towards σ

by less than 10%. This directly reﬂ ects the potential for extremely long

charge storage times in the deep trapping potential of the quantum dots, even at elevated tem-

peratures. Sample B, which did not contain quantum dots, exhibited little or no photo-response,

independent of E

, over the energy interval 900 meV  E

 1200 meV [55] .

0.9 1.0 1.1 1.2

Energy (meV)

PL intensity (arb. units)

1.3 1.4

GaAs

1.5

Figure 15.4 Photoluminescence spectra obtained at 293 K from the storage sample containing QDs (sample A – red

full curve) and the reference sample without QDs (sample B – dashed blue curve).

CH015-I046325.indd 483CH015-I046325.indd 483 6/24/2008 5:06:28 PM6/24/2008 5:06:28 PM

484 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

Figure 15.6 shows the spectral dependence of the device photo-response for a number of one

shot measurements over the energy range 950 meV  E

 1260 meV and for different reverse

biases applied to the vertical p - i - n junction (up to V

  2.5 V). The bold line shows the PL

spectrum obtained at 145 K for comparison. The onset of the photo-response discussed above is

very clearly observed for E

 1015 meV, ﬁ rmly establishing the correspondence between the

observed photo-effect and charging of the lowest orbital states of selected quantum dots with a

1.80

1.60

1.40

1.20

0 250

 0.0 V

500 750

Δσ

1000

1.016 eV

145 K – Sample A

1.024 eV 1.059 eV

Time (s)

σ (mS)

Figure 15.5 Temporal dependence of the n -channel conductivity for the QD containing sample (A) as a series of

100 s illumination, recovery and reset cycles are performed as described in the text. The measurements were taken

with excitation energies of E

 1016 meV (leftmost), 1024 meV and 1059 meV (rightmost), respectively. The

arrows labelled A and B denote the points at which the illumination is applied and removed, respectively. The sample

is “ reset ” by forward biasing the p - i - n junction to neutralize the stored charge at point C. The reference sample (B)

showed a much weaker photo-response over the excitation range 900  E

 1200 meV indicating that the meas-

ured response arises from optically stimulated charge storage in the QDs.

160

145 K – Sample A

140

120

100

% Δσ/σ

0.00 V

0.50 V

1.00 V

1.50 V

2.00 V

2.25 V

2.50 V

900 1000 1100 1200

Energy (meV)

1300

PL intensity (arb. units)

Figure 15.6 Spectral dependence of the photoconductive effect for sample A as a function of the reverse bias applied

to the vertical p - i - n junction from V

 0 to  2.5 V. The thick line shows a PL spectrum for comparison.

CH015-I046325.indd 484CH015-I046325.indd 484 6/24/2008 5:06:28 PM6/24/2008 5:06:28 PM

Quantum Dot Charge and Spin Memory Devices 485

single electron. Upon increasing V

, the magnitude of the photo-effect increases strongly (up to

 Δ σ / σ

 100% for V

  2 . 5 V) reﬂ ecting the electrostatic depletion of the n -channel under

reverse bias. The photosensitivity of the devices can be increased by depleting the 2D channel.

Over the same energy range, only a very weak photo-effect ( Δ σ / σ

 5%) is observed from the ref-

erence sample. To higher energies, a series of peaks labelled A

0–2

are observed. Before discussing

these features, we note that the data plotted is the saturation value of the photo-response. Thus,

the magnitude of the charge storage observed includes all possible QD excitation mechanisms

regardless of the relative timescales over which they occur. The energy separation between A

and A

is 75  5 meV, much larger than the splitting between ground and excited quantum dot

peaks ( E

 E

 6 0  5 meV). Furthermore, these resonances are shifted from E

by 33  5 meV,

115  10 meV and 172  20 meV for A

, A

and A

, respectively. This indicates that they do not

simply reﬂ ect direct optical charging of the ground and excited states of the quantum dots. These

observations were attributed to the participation of inelastic processes [55] over long timescales

involved, a superposition conﬁ rmed by optical readout measurements discussed below. Following

direct excitation of the quantum dots ’ excited states, stored electrons are expected to relax rapidly

to the lowest orbital state [70] and multiple charging of the quantum dots may occur via the quasi

continuum absorption band of the QDs associated with crossed 2D–0D transitions [71] .

Similar experiments have also been performed on inverted devices in which holes are optically

stored in the QDs and sensed using a 2D In

0.1

0.9

As p -channel [72] . In all other respects the

device operating principles and fabrication are similar to the electron storage sample discussed

above. Besides being interesting from the fundamental physics viewpoint, hole storage may ena-

ble charge storage up to higher temperatures as a consequence of the larger effective mass, as

mentioned above [62, 73] . As for the electron storage sample discussed above the modulation

doped p -channel was separated by 50 nm from the InAs QD layer which was embedded within

the intrinsic region of a vertical n-i-p junction. The growth conditions for the QD layer were

nominally identical to the electron sample discussed above.

Figure 15.7 (inset) shows the photo-response of the hole storage sample following illumination

at E

 1050 meV, just above the peak of the ground state PL emission ( Fig. 15.7 – bold line, top

panel). As for the electron storage sample discussed above, a strong decrease ( Δ σ ) in the channel

0.4

HOLE STORAGE

T  145 K

 0.5 V

0.3

0.2

0.12

0.10

0.03

0 100 200 300 400

0.1

Δσ/σ

dσ/dt

0.0

0.7

0.6

0.5

0.4

0.3

0.2

800 900 1000

Time (s)

Δσ

Energy (meV)

1100 1200

σ(ms)

Figure 15.7 Inset: Persistent photoconductivity arising from optically stimulated hole storage in the QD layer (top

panel) QD charging spectrum for sample with (open squares) and without (open circles) the QD layer. (lower panel)

Switching rate spectrum ( d σ / dt ) for the hole storage sample at P

 0.02 mWcm

 2

CH015-I046325.indd 485CH015-I046325.indd 485 6/24/2008 5:06:29 PM6/24/2008 5:06:29 PM

486 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

conductivity was observed following illumination ( A ), which remains persistent for many hours

after closing the shutter ( B ). After resetting the device by forward biasing the n-i-p junction ( C )

the conductivity, once again, returns to the pre-illumination level ( σ

). A similar photo-response

is not observed in a separate reference sample, which did not contain quantum dots. Fig. 15.7

(top panel) shows the spectral dependence of the induced photo-effect Δ σ  σ

for both QD (open

squares) and reference (open circles) samples, respectively. Since this data show the saturation

level of the photo-response, care should be exercised while interpreting the detailed form of the

charging spectrum. However, a clear onset is observed close to 1030 meV conﬁ rming that the

photo-effect arises from hole storage within the quantum dots. Furthermore, the peaked struc-

ture observed in the Δ σ  σ

data provides evidence for selective storage of holes within the QDs.

The ﬁ lled squares in Fig. 15.7 (lower panel) show the spectral dependence of the device switch-

ing rate ( P

 0.05 mW/cm

 2

) immediately after illumination for comparison. A series of peaks,

marked by the bold arrows on the ﬁ gure, are clearly observed in d σ / dt arising close to the ground

and excited state PL emission from the QDs. This observation indicates that the observed persist-

ent photoconductivity arises from optically stimulated hole storage in the QDs.

The combination of the AlGaAs blocking barrier adjacent to the QD layer combined with the

switchable electric ﬁ eld in the structure clearly enables charges to be selectively optically gener-

ated within a subset of self-assembled quantum dots. However, while the experiments discussed

above clearly demonstrate the ability to selectively generate charges within the ensemble of dots,

they cannot answer any the following questions:

● How many charges are generated per dot?

● What information does the spectral distribution of stored charge provide about the

response of a QD ensemble to quasi resonant optical excitation?

● What are the mechanisms by which stored charges are lost from the ensemble?

● Can these ideas be extended from optical charge generation to spin generation ?

To address these questions and reveal information about the optical charge storage process, fur-

ther experiments were performed with modiﬁ ed devices that enable optical readout of the stored

charge.

15.2.2 Optically detected charge storage

The ﬁ rst demonstration of full optical readout of optically induced charges was presented by

Lundstrom et al. , employing electric ﬁ eld-induced charge transfer between a QD and a strain-

induced potential minimum in an adjacent quantum well [74] . Optical readout is vital since

spectral information about the stored charge distribution within the ensemble can be preserved

by optical detection schemes; the selective transfer of one of the photo-generated charge carri-

ers out of the QD does not destroy the energetic information encoded during the optical writing

process provided that the second carrier continues to reside in the dot where it was generated.

This is a vital ingredient of the optical memory scheme discussed in [46] . An exciton can then

be re-formed in the ground state of the same QD by subsequent back transfer of charge, leading

to photon emission from an identical quantum mechanical environment to that present during

the initial absorption process. The selective optical generation and readout of a single charge car-

rier in a zero-dimensional electronic environment is therefore a key requirement with regards to

preserving spatial and spectral information. Furthermore, as discussed later in the chapter it can

also provide information about thermally induced charge redistribution [77] and the ideas can

be extended from charge storage to spin storage [75] by utilizing the polarization degree of free-

dom of the excitation source.

We continue to discuss such a wavelength selective QD-based charge storage device with opti-

cally stimulated charge storage (electron or hole) and a demonstration of direct optical readout of

the stored charge distribution in the spectral domain. We will see that similar device concepts are

utilized to those discussed in the previous section, but spectral information is retained in the opti-

cal readout process. This provides further scope to investigate thermally induced charge redistri-

bution and, crucially, electron and hole spin relaxation as discussed in the ﬁ nal sections of the

CH015-I046325.indd 486CH015-I046325.indd 486 6/24/2008 5:06:29 PM6/24/2008 5:06:29 PM

Quantum Dot Charge and Spin Memory Devices 487

chapter. Using these devices, wavelength domain “ data storage ” is demonstrated at T  10 K and

charges are stored over timescales much longer than 25 μ s. Analysis of the spectral distribution

of the stored charge enables the study of resonant and phonon-assisted absorption and, further-

more, temperature dependent measurements allow direct investigations of thermal carrier loss

from the QDs.

The operating principles and band diagram of a fully optical charge memory device are pre-

sented in Fig. 15.8 . A single layer of InGaAs QDs is embedded within the intrinsic region of either

an n - or p -type Schottky photodiode for hole or electron storage, respectively. In the following, we

describe the operation of the hole storage device, the principles for the electron storage sample

being entirely analogous. In the write mode ( Fig. 15.8a ), the device is reverse biased ( V  V

store

Following resonant optical excitation of the QD sub-ensemble ( ω

aser

  ω

write

), photo-generated

excitons are rapidly ionized by the strong vertical electric ﬁ eld ( F ~ 150 kV/cm). While electrons

readily tunnel out of the dot, the holes remain stored by virtue of an AlGaAs barrier immedi-

ately adjacent to the QD layer. Providing that V

store

is chosen such that the exciton ionization

rate far exceeds the radiative lifetime, holes are efﬁ ciently stored with a spectral distribution that

reﬂ ects the response of the QD ensemble to resonant optical excitation. Optical readout of the

stored charge distribution is achieved by application of a forward bias pulse ( V  V

reset

), resulting

in non-resonant injection of majority charge carriers into the QD layer and subsequent neutrali-

zation of the stored charge by radiative recombination. The resulting charge storage spectrum

directly reﬂ ects the spectral distribution of stored charge within the QD ensemble.

p-contact SAQDs

write

Initialization

Readout

Schottky

gate

eV

stor

eV

rese

hω

read

hω

(a)

(b)

Figure 15.8 Schematic structure and operating principles of charge storage devices investigated. (a) In the charge

storage mode ( V

gate

 V

store

) photon to charge conversion is ensured by the negative gate potential and selective exci-

ton ionization. (b) Conversion from charge to photon enabled by forward biasing the Schottky junction, hole reinjec-

tion and detection of the resulting storage EL at  ω

read

  ω

write

as discussed in the text.

As for the electrical memory devices discussed above, the optical memory structures were

grown using molecular beam epitaxy on [100] semi-insulating GaAs substrates. After an undoped

GaAs buffer, a back contact layer was deposited, consisting of 200 nm of heavily ( p - or n -)

doped GaAs ( N

D/A

 10

c m

 3

). Following this, a 50 nm-undoped GaAs buffer was grown followed

CH015-I046325.indd 487CH015-I046325.indd 487 6/24/2008 5:06:29 PM6/24/2008 5:06:29 PM

488 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

by the QD layer that consisted of 6 ML of In

0.5

As, deposited at 520°C and 0.1 ML/s. These

growth conditions resulted in approximately lens-shaped QDs with a mean diameter (height)

of 25 nm (4 nm). Moreover, the optical activity of these QDs is such that the emission from

their fundamental transition is around ~950 nm and can be detected using high-sensitivity

silicon-based detectors. The dot density was ~150 μ m

 2

measured by atomic force microscopy

on a sequentially grown control sample. The QDs were capped by 1 nm GaAs, a 50 nm thick

0.4

0.6

As blocking layer and completed with 10 nm GaAs. The wafers were then processed

into photodiodes, with ohmic contacts established to the buried back contact and completed with

a 10 nm thick, semi-transparent, top Schottky contact.

All optical experiments discussed below were performed using a standard micro-photolumi-

nescence (PL) system and resonant optical excitation pulses were delivered using a mechanically

modulated continuous-wave (CW) Ti–sapphire laser ( ω

write

 1.25–1.38 eV) or the Ti–sapphire

in combination with a laser diode (  ω

write

 1.27 eV) for two colour experiments. An accurately

timed sequence of optical (write) and electrical (readout) pulses were applied to the sample at a

repetition rate of ~20 kHz.In order to ensure efﬁ cient suppression of scattered laser light dur-

ing detection, an Si-based gated avalanche photodiode was used and only switched on during

application of the readout voltage pulse. The charge storage time ( Δ t ) is deﬁ ned as the time delay

between switching off the optical write pulse and the onset of the electrical readout.

Typical charge storage spectra for electron (dashed line) and hole (solid line) devices are pre-

sented in Fig. 15.9a for a storage time of Δ t  1 2 μ s. The overall form of the charge storage

spectra was found to be very similar for both samples, consisting of two principal components:

a sharp peak, labelled R in Fig. 15.9a , very close to the excitation energy and a much broader

band, labelled S , ~20–50 meV to lower energy. For both samples,  ω

write

was tuned to the high-

energy side of the QD ensemble (~1.27 eV – dashed line in Fig. 15.9a ) and the magnitude of the

readout voltage was

||V

reset

 0.8 V corresponding to the onset of majority carrier current ﬂ ow.



storage

1.15

Storage signal (arb. units)

1.20

1.25

1.30 1.35

Energy (eV)

hω



storage

hω

write



hω

write



(a)

(b)

(c)

(d)

Δt  12μs

B  0 T

T  10 K

Figure 15.9 (a) Typical emission spectra after application of the reset bias pulse ( Δ t  1 2 μ s), for electron (lower

panel) and hole (upper panel) storage devices, together with the non-resonantly excited QD ensemble luminescence

spectrum of the hole storage device (dashed lines). The absorption processes resulting in the resonant (R) and low

energy (S) emission features are also schematically depicted. (b) Ground state absorption giving rise to R. (c) Quasi-

resonant excitation via excited state generating the broad emission band S. (d) Phonon-assisted absorption generating

the peaks marked with S and arrows. Redrawn from [105] by permission of the American Physical Society.

CH015-I046325.indd 488CH015-I046325.indd 488 6/24/2008 5:06:30 PM6/24/2008 5:06:30 PM