Você está na página 1de 4

REGULAR PAPER

Japanese Journal of Applied Physics 54, 07JB02 (2015)


http://dx.doi.org/10.7567/JJAP.54.07JB02

Characterization of the silicon/-FeSi2 nanocrystallites heterostructures


for the NIR photodetection at low temperature
Alexander Shevlyagin1*, Dmitry Goroshko1,2, Evgeniy Chusovitin1, Konstantin Galkin1,2, and Nikolay Galkin1,2
1

Institute of Automation and Control Processes FEB RAS, 690041 Vladivostok, Russia
Far Eastern Federal University, School of Natural Sciences, 690950 Vladivostok, Russia
E-mail: shevliagin@mail.ru
2

Received October 13, 2014; revised October 30, 2014; accepted November 20, 2014; published online April 24, 2015
Using solid phase epitaxy of thin Fe lms and molecular beam epitaxy of Si, p-Si/-FeSi2 nanocrystallites/n-Si(001) diode structure was
fabricated. The diode exhibited a current responsivity of 15 mA/W and external quantum efciency of about 1% at a wavelength of 1300 nm at
120 K without bias and 200 mA/W and 10%, respectively, at %30 V. The device specic detectivity calculated at 120 K in zero bias conditions of
2.1 ' 1011 cm&Hz1/2/W at a wavelength of 1.3 m is the highest ever reported for Si/-FeSi2 systems. The FranzKeldysh effect gives grounds for
applying such systems not only for the development of optrons but also for that of electro-optical modulators.
2015 The Japan Society of Applied Physics

1.

Introduction

Light detection in the near-infrared region (NIR), especially


at wavelengths between 1.3 and 1.55 m, is an important
issue for optical ber communication systems and optical
interconnections because of the low dispersion and low loss
characteristics of the silica bers.1) Despite the successful
application of AIIIBV-based photodetection materials, such
as InAs=GaSb=AlSb superlattices,2,3) there is a strict requirement for the designed devices of compatibility with the
current Si technology for the future integrated optoelectronics. Therefore -FeSi2 has the advantage in terms of
compatibility and abundance of both Fe and Si in the Earths
crust.4) On the other hand, -FeSi2 demonstrates a band gap
of approximately 0.8 eV and a very large optical absorption
coecient of over 105 cm1 at 1 eV.5)
Unfortunately, most research teams emphasize that there
are two main diculties associated with trap centers for
photogenerated carriers that limit the performance of the
Si=-FeSi2 photodiode structure. First, the diusion of iron
atoms into silicon substrates near the interface during -FeSi2
formation results in the occurrence of deep energy levels.6)
Second, dangling bonds occur in the junction interface and
initial -FeSi2 layer, namely, interface states.7) In particular,
the heterojunction interface between the -FeSi2 lm and the
Si substrate should have dangling bonds not only because of
grain boundaries but also because of lattice mist.8)
In this work, we demonstrate an approach of embedding FeSi2 nanocrystallites (NCs) into the p-type layer of the
silicon pn junction for spectral expansion of its photodetecting properties to NIR. To reduce an iron diusion into
the substrate the deposition of thin 0.2 nm Fe layers being
occurred at room temperature. Such a multilayered structure
would contain at most several nm of Fe, whereas for the same
photodetection performance on -FeSi2 lms it is necessary
to use hundreds of nm.9) In addition, heteroepitaxial stress
may result in changes in the optical properties, band gap, and
even electron dispersion law of the embedded -FeSi2.10)
Furthermore, it has been shown that in the case of -FeSi2
lms on Si substrate photovoltaic properties are strongly
limited by the diusion length of the minority carriers in the
-FeSi2 absorbing layer and results in decreasing photoresponse with further increasing lm thickness.11) In our

previous work we have demonstrated that a light-emitting


silicon diode with embedded -FeSi2 NCs is a high-quality
defect-free structure with room temperature electroluminescence,12) so photogenerated carriers from -FeSi2 NCs move
through a high-quality silicon layer. The points mentioned
above give grounds for applying a silicon-silicide nanocrystallite approach for the development of optrons.
2.

Experimental methods

A sample was grown on n-type (7.5 cm) monocrystalline


silicon substrate with (001) orientation. -FeSi2 NCs were
formed by solid phase epitaxy (SPE) at 630 C followed by
molecular beam epitaxy (MBE) of 200 nm of a weakly doped
p-type Si layer at 750 C. To obtain a multilayer structure, the
SPE and MBE steps were repeated four times. More detailed
information about the growth procedure can be found in our
previous works.13,14) Next, the sample was chemically treated
to form a mesa diode for photovoltaic measurements. Finally,
AuSb alloy and Al were deposited to form contacts to n-Si
and p-Si, respectively. For comparison, a similar diode
structure was formed without NCs on the same substrate
and using the same deposition technique (referred to as the
reference diode) and the only dierence was in the thickness
of the p-type epitaxial layer (600 nm).
Currentvoltage (IV) and capacitancevoltage (CV)
characteristics of the mesa diode were measured on an
immittance meter E7-20 in the dark and under illumination.
Photoresponse measurements were carried out with a
calibrated tungsten lump as a light source and a monochromator (Solar Tii MS3504i) at standard lock-in technique
(Stanford Research Systems DSP lock-in amplier SR830)
with an optical chopper ( f = 882 Hz). All the spectra were
registered in a photocurrent mode in the 120300 K temperature range.
3.

Results

Figure 1 shows IV characteristic of the p-Si=-FeSi2 NCs=


n-Si(001) photodiode structure measured at 120 K. The diode
shows the good rectifying behavior expected of a conventional Si diode: the forward current is roughly more than
three orders of magnitude greater than that under reverse bias.
Estimated values of the series and shunt resistances are 1 k
and 20 M respectively for the photodiode structure

07JB02-1

2015 The Japan Society of Applied Physics

Jpn. J. Appl. Phys. 54, 07JB02 (2015)

A. Shevlyagin et al.

10

Current, A

-4

10

-5

10

0.95

Voltage,V

10

-9

Is=1.2*10 A

-8

Rs=1k

-9

Rp=20 M

10
10
10

-8

-6

0.8
1

-1

10

-2

10

-3

10

-4

10

-5

10

-4

-2

-1

10

-2

10

-3

10

-4

-6

10

10

-7

10

Si diode with -FeSi2 NCs at -30 V

-5

10

Si diode with -FeSi2 NCs at -20 V


Si diode with -FeSi2 NCs at -10 V

10

10

-8

-10

0.85

10

-6
-7

0.9

@120 K

Si diode with -FeSi2 NCs without bias

@120K

Ext. quant. eff. ( ), %

-3

Photon Energy, eV
8

10
7
10
6
10
5
10
4
10
3
10 R = 50 M
2
0
10
-2 -1 0 1

Photoresponse (R), A/W

-2

10

Dynamic resistance,

-1

10

-6

10

reference Si diode

Voltage, V

1300

1400

1500

1600

Wavelength, nm
Fig. 1. IV characteristic of p-Si=-FeSi2 NCs=n-Si(001) photodiode
structure measured in the dark at 120 K. The insets demonstrate the dynamic
resistance and diode parameters estimated from the IV characteristic.

Si diode with -FeSi2 NCs

20

8.0x10

@120K
20

6.0x10

20

1.5x10

20

4.0x10

20

1.0x10

19

1/C , F

-2
2

1/C , F

-2

5.0x10

20

2.0x10

build- in

0.0
-4

-3

-2

-1

Voltage, V

-35 -30 -25 -20 -15 -10

-5

Voltage, V
Fig. 2. (Color online) 1=C 2V characteristic of p-Si=-FeSi2 NCs=
n-Si(001) photodiode structure measured in the dark at 120 K. The point of
intersection of the red line with the voltage axis corresponds to the value of
the built-in potential of the Si pn junction at given doping levels and
temperature.

containing -FeSi2 NCs. Furthermore, a structure with


embedded NCs demonstrated saturation current of 1.2 nA.
The dependences of 1=C 2 versus V, presented in Fig. 2,
showed that regardless of the presence of the NCs the value
of the built-in potential of the structure is 0.88 eV, that is in a
good agreement with the value of 0.9 eV for a conventional
Si pn junction at a given temperature and the given doping
levels of the Si substrate (Nd = 5 1014 cm3) and epitaxial
p-Si layers (Na = 1 1014 cm3). Furthermore, on the 1=C 2
curve there is a linear slope in the reverse bias starting from
1 V, suggesting that it is owing to the reduction of the
interface state of the mesa diode.7)
The measured spectra of the current responsivity of the
structures at 120 K in comparison with the reference Si diode
are presented in Fig. 3. It is obvious that the embedding in
the silicon pn junction of only four layers of -FeSi2 NCs
resulted in: (1) enhancing of the spectral range of photosensitivity up to 16001700 nm (0.8 to 0.7 eV); (2) the

Fig. 3. (Color online) Photoresponse spectra of the p-Si=-FeSi2 NCs=


n-Si(001) structure measured at 120 K under dierent reverse biases and
calculated spectra for the external quantum eciency in comparison with the
reference Si pn junction.

photoresponse (R) being enhanced more than one order of


magnitude at telecommunication wavelengths without applying reverse bias; (3) to the possibility of enhancing the
photodetection performance by more than four orders of
magnitude by varying reverse bias in the wavelength wide
range. On the basis of the measured photoresponse, we
estimated the spectral dependence of the external quantum
eciency () of our diode and the conventional pn junction
following Sze.15) The diode without bias demonstrated
external quantum eciency of near 1% at 1300 nm, and the
applying of 30 V resulted in increased by up to 10%. What
is more interesting, however, is that a marked increase in
eciency to almost the same value was clearly observed in
the second transparency window of silica ber at 1.55 m.
The latter persuaded us to study the dependence of the
photoresponse edge from the value of reverse bias in detail.
To this end, we used the original spectra of the photocurrent
and built dependences as follows: the square of the photocurrent versus the photon energy. Thus, we were able to
determine the position of the edge of the photosensitivity
and with it an absorption edge, as it is well known that the
magnitude of the photocurrent is directly proportional to
the absorption coecient. In addition, the direct-gap nature
of iron disilicide meant that the spectral dependence of the
photocurrent was linearized in quadratic coordinates and
eventually allowed us to estimate the position of the
absorption edge. The results of this analysis are presented
in Fig. 4. With increasing reverse bias one can observe the
red shift of the photosensitivity edge, which we will explain
further. Inter alia, there is a nonlinear increase in eciency
as a whole, and at wavelengths of 1.3 and 1.55 m with
increasing bias, this dependence is exponential.
Figure 5 denotes the temperature dependence of the
integrated value of the external quantum eciency of the
structure. The same trend was observed, for example, in
work9) on the temperature dependence of the dynamic
resistance-area product. Since this quantity is directly related
to the characteristics of diode structures, its behavior will also
determine the ultimate eciency of the detection and that is

07JB02-2

2015 The Japan Society of Applied Physics

Photoresponse Edge, eV

0.94
0.92
0.90

A. Shevlyagin et al.

Integrated ext. quant. eff.,


arb. un.

Jpn. J. Appl. Phys. 54, 07JB02 (2015)

4.

-30

-20

-10

Voltage, V

0.88
0.86

@120 K

0.84

Si diode with -FeSi2 NCs

-50

-40

-30

-20

-10

Voltage, V
Fig. 4. Reverse bias dependences of the photoresponse edge and the
integrated external quantum eciency of the p-Si=-FeSi2 NCs=n-Si(001)
structure measured at 120 K.

Temperature, K
280 240

200

160

120

Integrated Ext. quant. eff.,


arb. un.
diff
usi
on

Si diode with -FeSi2 NCs at -30 V

G-R

p -Si

p -Si

p - -FeSi2

lan

e
ch

a
av

n -Si

Evac

Si

extraction
h

Ec

generation

EF
avalanche

Ev
+

Si

1000/T, K

-1

Fig. 5. (Color online) Temperature dependence in semilog coordinates of


the value of the external quantum eciency integrated in the (1250 to
1700 nm) wavelength range with indicated probable mechanisms for the
transportation of the photogenerated carriers.

why we can use the same analysis. In the temperature range


of 200300 K the integrated value of exhibits linear
dependence suggesting the dominance of the carrier diusion
transportation mechanism. In the temperature range of
120200 K the probable mechanism is carrier generation
recombination, which appears to occur at heterointerfaces
and produces leakage current.
Under the assumption of the low background photon ux
at the zero bias voltage, following Keyes we calculated the
Johnson-noise-limited specic detectivity (D+), that limits
the noise performance of the infrared photodiodes.16) Also,
using this formula, adopted for the non-zero voltage case,3)
we also estimated specic detectivity at the dierent biases.
Obtained device specic detectivity at 120 K in zero bias
conditions is 2.1 1011 cmHz1=2=W at a wavelength of
1.3 m and 6.5 109 cmHz1=2=W at 1.55 m, while at a
bias voltage of 30 V these values are 4.8 1010 and 2.4
1010 cmHz1=2=W, respectively.

Discussion

Given the diode parameters calculated from the IV


characteristic it can be concluded that the introduction of
nanocrystallites into the active region of the silicon photodiode has practically no eect on the properties of the diode:
it has a typical silicon reverse saturation current, and in
addition there is no premature switching of the diode, which
could be caused by the iron atoms or by the interface states.17)
The CV characteristics conrm the fact that embedded
NCs do not form a continuous heterojunction, but only form
a set of local ones, distributed in the silicon matrix. This is
why we measured only the built-in potential dierence of
the silicon and failed to determine the magnitude of the
heterojunction between -FeSi2 and Si because of the small
amount of disilicide and its distribution in the silicon. Despite
the fact that on the initial curve piece of 1=C 2 there is a nonlinear dependence, indicating the presence of trap centers,
starting from 1 V one can observe linearized dependence,
indicating their small concentration.7)
As far as photoresponse is concerned, measured photosensitivity of 1516 mA=W without bias at a wavelength of
1.3 m is almost ve times higher than the value obtained for
iron disilicide lm, and our structure contains 100 times less
disilicide.9)
By applying a reverse bias, we have observed an
exponential increase of eciency, not previously observed
for Si=-FeSi2 heterojunctions. We assume that this is owing
to the onset of avalanche multiplication of carriers at a
suciently high voltage value; furthermore, the nanocrystallites can act as local inhomogeneities, leading to even
greater increase of the electric eld which causes the
avalanche; moreover silicon is widely known as a good
semiconductor for avalanche photodiode design.
To explain the gathering process of photogenerated
carriers we proposed a band diagram of the diode structure
in equilibrium state, as presented in the inset of Fig. 5. After
the generation of carriers in the nanocrystallites, they diuse
into silicon p-type layers where the built-in potential of the Si
pn junction separates them. Then, acceleration of the
carriers by the electric eld takes place and they may get
sucient energy to impact ionization of the silicon atoms,
thereby developing an avalanche.
It is worth noting that the p-Si=-FeSi2 NCs=n-Si(001)
photodiode structure analyzed in this work demonstrates
current responsivity of 3 mA=W and external quantum
eciency of near 1% at a wavelength of 1300 nm at room
temperature while biased at 30 V. Unfortunately, we failed
to obtain a measurable photoresponse without bias, apparently because of an increase in the saturation current to more
than 3 orders of magnitude. Furthermore, one should take
into consideration that carrier diusion length as well as
the eciency of avalanche development in Si are strongly
temperature-dependent; as a result, the built-in potential of
the Si pn is not enough to separate and multiply photogenerated carriers.
Greater eciency in comparison with the thin lm
heterojunctions9) further connected with the fact that photogenerated carriers in our case move to the contacts through a
silicon matrix where they have a fairly large diusion length,
in contrast to that of in the disilicide layer.11)

07JB02-3

2015 The Japan Society of Applied Physics

Jpn. J. Appl. Phys. 54, 07JB02 (2015)

A. Shevlyagin et al.

The additional broadening of the spectral sensitivity in the


infrared region of the spectrum with increasing reverse bias is
in our opinion caused by the FranzKeldysh eect, which is a
red shift of the absorption edge and hence the detection edge
of the semiconductor placed in an electric eld. However,
integrated eciency must be maintained with this eect by
Fermis golden rule, but it should be noted that in this case,
among other things it is going avalanche multiplication of
carriers, the gain of the multiplication being increases with
bias. Thus, this means we can formally bypass the Fermi
requirement. The observed eect can be used to develop
electro-optical modulators, especially considering the fact
that silicon itself shows no electro-optical eects.18)
The calculated values of the specic detectivity of our
structure are superior to the results for lms9) and even
comparable to those for materials such as InAs, PbS or
GaSb.19,20) In addition, our structure exhibits a suciently
high value in terms of the photoresponse and external
quantum eciency, and the detecting ability, not only in the
rst transparent window of the silica ber but also at a
wavelength of 1.55 m, whereas all other publications79,11,2123) describe only detection in the rst region.
5.

Summary

We have fabricated an n-Si=-FeSi2 NCs=p-Si photodiode


structure, grown by combination of SPE for the formation of
nanocrystallites with MBE for the epitaxy of the silicon
capping layers. IV and CV characteristics conrmed that
embedding of the -FeSi2 NCs in the depleted region of
the silicon photodiode does not substantially inuence the
diode rectifying properties and saturation current. Furthermore, the presence of the NCs does not cause formation
of the additional signicant concentration of trap centers
resulted from the interface states or diused iron atoms.
We suppose that it is related to non-defects epitaxy and
embedding of the small-sized NCs into silicon matrix
together with implementation of the SPE, which might have
restricted iron diusion to the silicon substrate to some extent
and prevented formation of deep levels associated with it.
It was demonstrated that the proposed diode structure has a
signicant photoresponse at energies below the silicon band
gap, which is associated with the absorption and carrier
generation in -FeSi2 nanocrystallites. Despite the fact that
the structure contains 100 times less iron disilicide, it exhibits
a value several times greater of photosensitivity, external
quantum eciency and specic detectivity at wavelengths
important for optical ber communication systems.
In addition, the observed red shift of the edge of the
photosensitivity with increasing reverse bias associated with
the FranzKeldysh eect suggests the possibility of developing electro-optical modulators based on this system; in
conjunction with the demonstrated ability to create lightemitting diodes, it can be used to create opto-couples.
Such a good result compared with iron disilicide lms
on silicon can be associated, in our opinion, rst, with the
defect-free incorporation of nanocrystallites in the silicon
matrix, second, with a reduction in the diusion of the iron
atoms into substrate and, third, due to the greater diusion
length of carriers in silicon, compared with a continuous
layer of silicide, as well as the eect of avalanche
multiplication of carriers due to inhomogeneity of the electric

eld caused by the presence of nanocrystallites in the


depletion region of the photodiode.
Further improvements to our device may consist of
optimization of pn junction parameters to operate at zero
or small reverse biases. Another way to improve performance,
especially in the eld of solar cells, would be to embed the
NCs of dierent narrow band gap semiconducting silicides,
such as CrSi2 (0.350.37 eV),24,25) Ca3Si4 (0.350.63 eV)26,27)
together with -FeSi2, to covering both the NIR and shortwave infrared region (1.4 to 3 m) of the solar spectrum.
Acknowledgments

This work has been done under nancial support from


Russian Federation Presidents grant (No. MK-6343.2013.8),
Far Eastern Branch of Russian Academy of Science grant
(No. 15-I-4-008o), and Joint program between FEB RAS
and National Research Council of Taiwan grant (No. 15MHT-001).
1) H. Udono, I. Kikuma, T. Okuno, Y. Masumoto, and H. Tajima, Appl. Phys.
Lett. 85, 1937 (2004).
2) Y.-F. Lao, P. K. D. D. P. Pitigala, A. G. U. Perera, L. H. Li, S. P. Khanna,
and E. H. Lineld, Appl. Phys. Lett. 104, 131101 (2014).
3) A. M. Hoang, G. Chen, A. Haddadi, S. Abdollahi Pour, and M. Razeghi,
Appl. Phys. Lett. 100, 211101 (2012).
4) B. Tatar, K. Kutlu, and M. Urgen, Thin Solid Films 516, 13 (2007).
5) M. C. Bost and J. E. Mahan, J. Appl. Phys. 64, 2034 (1988).
6) T. Isobe, H. Nakashima, and K. Hashimoto, Jpn. J. Appl. Phys. 28, 1282
(1989).
7) S. Funasaki, N. Promros, R. Iwasaki, M. Takahara, M. Shaban, and T.
Yoshitake, Phys. Status Solidi C 10, 1785 (2013).
8) M. Han, M. Tanaka, M. Takeguchi, Q. Zhang, and K. Furuya, J. Cryst.
Growth 255, 93 (2003).
9) S. Izumi, M. Shaban, N. Promos, K. Nomoto, and T. Yoshitake, Appl. Phys.
Lett. 102, 032107 (2013).
10) K. Takarabe, R. Teranishi, J. Oinuma, Y. Mori, T. Suemasu, S. Chichibu,
and F. Hasegawa, Phys. Rev. B 65, 165215 (2002).
11) M. Shaban, K. Nakashima, W. Yokoyama, and T. Yoshitake, Jpn. J. Appl.
Phys. 46, L667 (2007).
12) N. G. Galkin, E. A. Chusovitin, D. L. Goroshko, A. V. Shevlyagin, A. A.
Saranin, T. S. Shamirzaev, K. S. Zhuravlev, and A. V. Latyshev, Appl.
Phys. Lett. 101, 163501 (2012).
13) N. G. Galkin, E. A. Chusovitin, T. S. Shamirsaev, A. K. Gutakovski, and
A. V. Latyshev, Thin Solid Films 519, 8480 (2011).
14) D. L. Goroshko, E. A. Chusovitin, A. V. Shevlyagin, M. V. Bozhenko, R. I.
Batalov, R. M. Bayazitov, and N. G. Galkin, Phys. Status Solidi C 10, 1844
(2013).
15) S. M. Sze and K. K. Ng, Physics of Semiconductor Devices (Wiley, New
York, 2007) p. 666.
16) R. J. Keyes, Optical and Infrared Detectors (Springer, Berlin, 1977) Topics
in Applied Physics, Vol. 19, p. 50.
17) E. F. Schubert, Light-Emitting Diodes (Cambridge University Press, New
York, 2003) p. 88.
18) A. Rickman, Nat. Photonics 8, 579 (2014).
19) A. Rogalski and M. Razeghi, Proc. SPIE 3287, 2 (1998).
20) H. Mohseni, M. Razeghi, C. J. Brown, and Y. S. Park, Appl. Phys. Lett. 78,
2107 (2001).
21) M. Shaban, K. Nomoto, S. Izumi, and T. Yoshitake, Appl. Phys. Lett. 94,
222113 (2009).
22) M. Shaban, S. Izumi, K. Nomoto, and T. Yoshitake, Appl. Phys. Lett. 95,
162102 (2009).
23) N. Promros, K. Yamashita, R. Iwasaki, and T. Yoshitake, Jpn. J. Appl.
Phys. 51, 108006 (2012).
24) V. Bellani, G. Guizzetti, F. Marabelli, A. Piaggi, A. Borghesi, F. Nava,
V. N. Antonov, Vl. N. Antonov, O. Jepsen, O. K. Andersen, and V. V.
Nemoshkalenko, Phys. Rev. B 46, 9380 (1992).
25) N. G. Galkin, A. M. Maslov, and A. V. Konchenko, Thin Solid Films 311,
230 (1997).
26) D. B. Migas, V. L. Shaposhnikov, A. B. Filonov, N. N. Dorozhkin, and
V. E. Borisenko, J. Phys.: Condens. Matter 19, 346207 (2007).
27) S. A. Dotsenko, K. N. Galkin, D. A. Bezbabny, D. L. Goroshko, and N. G.
Galkin, Phys. Procedia 23, 41 (2012).

07JB02-4

2015 The Japan Society of Applied Physics

Você também pode gostar