Invited Paper: The Excitation Mechanism of Rare-Earth Ions in Silicon Nanocrystals
Invited Paper: The Excitation Mechanism of Rare-Earth Ions in Silicon Nanocrystals
Invited Paper: The Excitation Mechanism of Rare-Earth Ions in Silicon Nanocrystals
1007/s003399900087
Applied Physics A
Materials
Science & Processing
Springer-Verlag 1999
Invited paper
Abstract. A detailed investigation on the excitation mechan- process is extremely efficient. Furthermore, the non-radiative
isms of rare-earth (RE) ions introduced in Si nanocrystals (nc) decay processes typically limiting rare-earth luminescence in
is reported. Silicon nanocrystals were produced by high-dose Si (namely back-transfer and Auger) are demonstrated to be
80-keV Si implantation in thermally grown SiO2 followed by absent in Si nc further improving the overall efficiency of
1100 ◦ C annealing for 1 h. Subsequently some of the samples the process. These data are reported and their implications
were implanted by 300-keV Er, Yb, Nd, or Tm at doses in discussed.
the range 2 × 1012 –3 × 1015 /cm2 . The energy was chosen in
such a way to locate the RE ions at the same depth where PACS: 76.30.Kg; 7.55.-m; 61.72.Tt
nanocrystals are. Finally an annealing at 900 ◦ C for 5 min
was performed in order to eliminate the implantation dam-
age. These samples show intense room-temperature lumines- Silicon, the leading semiconductor in the microelectronics in-
cence due to internal 4 f shell transitions within the RE ions. dustry, has for a long time been considered unsuitable for
For instance, luminescence at 1.54 µm and 0.98 µm is ob- optoelectronic applications which remained the domain of
served in Er-doped nc, at 0.98 µm in Yb-doped nc, at 0.92 µm III-V semiconductors and glass fibers. This is mainly due to
in Nd-doped nc and two lines at 0.78 µm and 1.65 µm in the silicon indirect band gap, which makes it a poor emitter,
Tm-doped nc. Furthermore, these signals are much more and to the absence of linear electro-optic effects. The enor-
intense than those observed when RE ions are introduced mous progress in communication technologies in recent years
in pure SiO2 in the absence of nanocrystals, demonstrating resulted in an increased demand for optoelectronic functions
the important role of nanocrystals in efficiently exciting the integrated with electronic circuits. This would allow us to
REs. It is shown that the intense nc-related luminescence at couple the information processing capabilities of microelec-
around 0.85 µm decreases with increasing RE concentration tronics with the efficient interconnection properties of op-
and the energy is preferentially transferred from excitons in toelectronics. In principle, silicon would be the material of
the nc to the RE ions which, subsequently, emit radiatively. choice, due to its mature processing technology and to its
The exact mechanism of energy transfer has been studied unrivaled domain in microelectronics, the main limiting step
in detail by excitation spectroscopy measurements and time- being the absence of efficient Si-based light sources.
resolved photoluminescence. On the basis of the obtained Recently a strong effort has hence been devoted to study
results a plausible phenomenological model for the energy all of those processes able to circumvent the physical inability
transfer mechanism emerges. The pumping laser generates of silicon to emit light [1]. Since the discovery of light emis-
excitons within the Si nanocrystals. Excitons confined in the sion from porous silicon made in 1990 by Canham [2] a lot of
nc can either give their energy to an intrinsic luminescent cen- work has been performed in studying silicon nanostructures.
ter emitting at around 0.85 µm nor pass this energy to the RE These comprise not only porous silicon but also nanocrystals
4 f shell, thus exciting the ion. The shape of the luminescence produced by several techniques [3–10], as well as silicon-
spectra suggests that excited rare-earth ions are not incorpo- insulator multilayers [11, 12]. The initial problems related to
rated within the nanocrystals and the energy is transferred the instability of the luminescence yield have finally been
at a distance while they are embedded within SiO2 . Rare- solved and today reliable, stable structures, compatible with
earth excitation can quantitatively be described by an effect- the silicon technology have been fabricated. In particular, the
ive cross section σeff taking into account all the intermediate group at Rochester University has now produced silicon-rich
steps leading to excitation. We have directly measured σeff for silicon dioxide electroluminescent devices integrated with sil-
Er in Si nc obtaining a value of ≈ 2 × 10−17 cm2 . This value is icon microelectronic circuitry [13]. Alternative approaches
much higher than the cross section for excitation through dir- comprise the doping of silicon with rare earths [14–22]. In
ect photon absorption (8 × 10−21 cm2 ) demonstrating that this this case the luminescence is due to an internal 4 f shell
4
dNexc Nexc
= σφN − , (1)
Fig. 7. Time decay curves of the nanocrystal-related luminescence at dt τ
0.85 µm in the absence of Er and for different Er contents. The system dNc∗ Nexc N ∗
response is also reported. Data were taken by pumping to steady state at = c − cc , (2)
50 mW and mechanically switching off the laser beam at t = 0 dt τtr τd
8
∗ ∗
dNRE NRE
= R∗ − RE . (3) In this formula NRE is the total RE concentration and we
dt τd describe the exciton rare-earth interaction through an interac-
tion volume Λ. Only those RE ions being inside the volume
Equation (1) describes the formation of excitons (Nexc ) within Λ can absorb energy from the exciton with a transfer time
the nanocrystals. Here the first term is the generation rate per ion of τtrRE . Hence NRE Λ represents the average number
proportional to a generation cross section σ, to the pumping of RE ions contained within the volume Λ. Note that all the
laser photon flux φ and to the nanocrystal concentration N. rare-earth ions can be excited and those that are already in
The second term represents the exciton recombination with an excited state can be further excited to a higher state. The
a characteristic recombination time τ whose dependencies exciton recombination time τ will then be given by:
will be discussed later on. Equation 2 describes the excitation
of the luminescent center (Nc∗ ) leading to the nanocrystal- 1 1 1
related luminescence at around 0.85 µm. The first term rep- = eff + c . (8)
τ τtr τtr
resents the excitation of the center (depicted in Fig. 8) pro-
portional to the exciton concentration through a character- The steady-state exciton concentration of (6) can hence be
istic energy transfer time τtrc , whereas the second term de- written as:
scribes its de-excitation with a typical time τdc (taking into
account both radiative and non-radiative de-excitation pro- τtreff τtrc
Nexc = σφN . (9)
cesses). Finally, (3) describes the RE excitation through ex- τtreff + τtrc
citon recombination and the de-excitation processes of the
∗ Clearly the steady-state exciton concentration decreases
excited RE ions. Here NRE represents the excited RE con-
∗
centration, R is the effective increase of the excited rare- as τtreff decreases, i.e. as the RE concentration increases, as
earth population through energy transfer from an exciton to a result of a new possible recombination route. This effect,
the appropriate excited state of the RE 4 f shell (for in- in turn, affects the steady-state population of excited lumines-
stance, in the case of the 1.54-µm line of Er the first ex- cent centers Nc∗ . In fact, the steady-state solution to (2) is:
cited state 4I13/2 ) and τdRE is the decay time of excited RE
τdc τtreff τdc
taking into account both radiative and non-radiative pro- Nc∗ = Nexc c = σφN eff , (10)
cesses. The luminescence intensity from the luminescent cen- τtr τtr + τtrc
ters at 0.85 µm and from the RE will be given repectively
by: where we have used the results of (9). From the data shown
in Fig. 1 we know that as the RE concentration increases the
Nc∗ transfer of energy from the exciton to the RE ions (in a time
Ic ∝ c , (4)
τrad τtreff ) becomes the dominant mechanism with respect to the
∗
NRE transfer from the exciton to the luminescent center at 0.85 µm
IRE ∝ RE , (5) (with a time τtrc ). Therefore, as NRE increases, τtreff becomes
τrad much smaller than τtrc . In this regime the luminescence inten-
sity of the centers at steady steady is described by:
τrad
c
and τrad
RE
being the radiative decay time of the center
and of the excited RE, respectively. The determination of Nc∗ 1 τtreff τdc
the luminescence intensities, therefore, consists in the deter- Ic ∝ = σφN (11)
mination of the concentration of excited centers (Nc∗ ) and τrad
c
τrad
c
τtreff + τtrc
excited RE ions (NRE ∗
) through the resolution of the rate (1)– τ c τ eff τc τ RE
≈ σφN dc tr c = σφN c d c tr ,
(3). τrad τtr τrad τtr NRE Λ
Let’s first find the steady-state solutions for these equa-
tions. As far as (1) is concerned the solution is simply: where we have used (4), (7), and (10). It is interesting to note
that the nanocrystal luminescence intensity decreases with
Nexc = σφNτ. (6) increasing RE concentration in agreement with the experi-
mental data.
Therefore the steady-state exciton population is propor- Let’s now analyze the RE luminescence intensity. In order
tional to the concentration of nanograins, to the generation to do that, according to (5), we should first solve the rate (3)
cross section, to the photon flux, and to the recombination and find a solution for the RE population in the appropriate
time τ. A question arises on what are the processes affecting excited state from which luminescence occurs (NRE ∗
; for in-
the value of τ. Following our description reported in Fig. 7 4
stance in the case of the 1.54-µm line of Er the I13/2 state).
we assume that an exciton can recombine either by giving In order to solve (3) a plausible hypothesis should be made
its energy to the luminescent center (associated with each for R∗ . This term represents the effective number of RE ex-
nanograin) with the characteristic transfer time τtrc or by ex- citations into the appropriate excited state (for example an Er
citing a RE ion. These two processes are in competition one ion on the first excited state) per unit volume and time. It dif-
another. Of course the transfer of energy to a RE ion depends fers from Nexc /τtreff in that this latter rate takes into account all
on the availability of such an ion close to the exciton in the excitations bringing the RE ion to one of the several excited
nanograin. We can describe this transfer through an effective states. Of course, if a RE ion is excited into a highly excited
interaction time τtreff equal to: state it can subsequently decay to the appropriate excited state
(the first excited state in the case of Er). This process has how-
τtrRE
τtreff = . (7) ever a certain probability to occur and not all of the excited
NRE Λ ions will do so. Moreover, we can also have that some of the
9
excitons will excite to a higher state a RE ion that is already in We have now calculated the main features of the present
∗
the appropriate level thus decreasing the population NRE . We description. In the following section we will make a direct
can then assume that the effective excitation rate R∗ has the comparison with the experimental data.
following expression:
∗ ∗
Nexc (NRE − NRE )Λη Nexc NRE Λ 2.4 Comparison with the experiment
R∗ = − (1 − η) (12)
τtr
RE
τtr
RE
∗ We have measured the room-temperature pump power depen-
Nexc (ηNRE − NRE )Λ
= . dence of the 1.54-µm luminescence in Er-doped Si nanocrys-
τtr
RE
tals and the data, normalized to the maximum intensity, are
reported in Fig. 9. These data can be fitted by means of (15)
The first term in this expression represents the filling rate
using, as a unique fitting parameter, the product σeff τdRE . The
of the appropriate excited state. This will clearly be propor-
fit is also reported in the figure as a continuous line yielding
tional to the exciton population Nexc , to the concentration
∗ a value σeff τdRE = 5 × 10−20 cm2 s. An independent measure
of RE ions that are not already in that level (NRE − NRE ),
of the decay time of the Er luminescence in Si nanocrystals
to the volume of interaction Λ, divided by the transfer time
(τdRE ) would allow us to obtain σeff . We have indeed meas-
τtrRE , and multiplied by the probability η that, once excited,
ured this decay time by pumping the system to steady state
the RE will then decay populating the appropriate level. Note
and mechanically shutting the pump laser. The results are
that not all of the RE ions within the volume Λ can be
shown in Fig. 10 for both 17 K and 300 K and show a de-
excited to the appropriate state but only those that are not
∗ cay time of ≈ 3 ms. It is interesting to note that this decay
already in that state (i.e. (NRE − NRE )). Moreover, only a frac-
time of the luminescence at 1.54 µm is almost temperature
tion η (with η ≤ 1) of the excited ions will then decay to
the metastable appropriate excited state hence increasing the
∗
excited population NRE . The second term in (12) describes
∗
a decrease of NRE due to the fact that also a RE ion in
the appropriate level can be excited to a higher state, thus
depopulating that level. Of course a fraction η of these ex-
citations will then return to the initial metastable level thus
producing no net change in the rate, while a fraction (1 − η)
will be lost forever thus decreasing the effective excitation
rate.
The steady-state solution of (3), under the conditions de-
scribed in (12), and using the result of (6), will be:
∗ σeff τdRE φ
NRE = ηNRE , (13)
σeff τdRE φ + 1
periments of Figs. 9–12) we have estimated that there are of the band gap due to confinement. These lines cannot
≈ 20 Er ions per nanocrystal. The effective cross section for be observed in crystalline Si;
Er excitation through nanocrystal absorption should there- (v) the rare-earth-related luminescence temperature depen-
fore be, at maximum, 20 times less than the nanocrystal dence is generally weak as a result of reduced non-
excitation cross section, i.e. 2 × 10−17 cm2 . It is quite inter- radiative de-excitations. The only different case is Tm
esting and instructive to note that this number, in spite of in which, as result of an internal redistribution, the
the rough estimates made, is similar to that experimentally 1.65-µm line is seen to increase in intensity with in-
determined. creasing temperature, whereas the 0.78-µm line has just
A last observation: in the process we are describing a pa- the opposite behavior;
rameter that matters of course is the ratio between Er ions (vi) the temperature dependence of the rare-earth-related lu-
and nanocrystals. At low Er doses there are more nanocrys- minescence lifetime is very weak confirming the ab-
tals than Er ions and, within each interaction volume Λ, sence of non-radiative de-excitations;
there is, at most, only one Er ion. In this regime the lu- (vii) with increasing rare-earth concentration the nanocrystal-
minescence intensity is limited by the Er concentration. related luminescence at 0.85 µm decreases while the
More complicated is the regime in which several rare-earth rare-earth-related luminescence increases demonstrat-
ions need to be excited by the same nanocrystal. In this ing the presence of an energy transfer. The decrease in
regime, once the nanocrystal is excited, it can transfer the the 0.85-µm luminescence, however, is not associated
energy only to one Er ion. However, if the photon flux is with a decrease in its decay time;
high enough, the nanocrystal can be excited again while (viii) the excitation wavelength dependence of rare-earth-
the first Er is still in the excited state and hence a sec- related luminescence peaks follows exactly the de-
ond ion can be excited. In principle, if the photon flux is pendence of the nanocrystals related luminescence at
sufficiently high, all of the ions within Λ can be brought 0.85 µm demonstrating, once more, that the energy is
to the excited state by the same nanocrystal. The time be- transferred from the nanocrystals to the 4 f shell.
tween two different exciton formation in the same nanocrystal
will be 1/σφ. Using, as a first approximation, the esti- On the basis of these results a phenomenological model
mated value for σ we have that 1/σφ equals the Er lifetime on the excitation mechanisms has been proposed. The pump-
(≈ 3 ms) at a flux of ≈ 8 × 1017 photons/cm2s (corresponding ing laser generates excitons within the Si nanocrystals. These
to ≈ 10 mW). Therefore, at pump powers below this pho- excitons can transfer their energy either to an intrinsic lumi-
ton flux one nanocrystal can have only one excited Er ion nescent center emitting at around 0.85 µm or to a rare earth
at a time within its interaction volume. In this regime lumi- close by, the two processes being in competition one another.
nescence is limited by the concentration of the nanocrystals. With increasing rare-earth concentration the 0.85-µm lumi-
However, with increasing photon flux, the situation changes nescence is hence quenched. The process leading to rare-earth
and, eventually, several ions can be excited by the same excitation needs several intermediate steps, it can however
nanocrystal. be described by an effective cross section σeff taking into
These observations support the idea that nanocrystals are account all of the different effects. Through time-resolved
extremely efficient in acting as sensitizers in rare-earth lumi- measurements we have experimentally determined σeff in the
nescence since many rare-earth ions can indeed be excited by case of Er obtaining a value of ≈ 2 × 10−17 cm2 . This value
a single excitation centre. is much higher than the cross section for excitation through
direct photon absorption (8 × 10−21 cm2 ) typically present in
insulating hosts. This demonstrates that indeed nanocrystals
act as very efficient sensitizers for rare-earth ions embed-
3 Summary ded within silicon dioxide. Moreover, since Auger and back-
transfer non-radiative de-excitations are absent, the overall
In conclusion, we have reported a detailed investigation on process is also much more efficient than rare-earth lumines-
the rare-earth luminescence in Si nanocrystals. We have cence within bulk crystalline silicon. These data demonstrate
shown that rare-earth ions are excited very efficiently through that applications such as light amplification in waveguides
an energy transfer process from excitons confined within the simultaneously containing nanocrystals and rare earths hold
nanocrystals. Several general features can be recognized: great potentialities.
(i) the rare-earth luminescence in Si nanocrystals is much
Acknowledgements. This work has been supported in part by the Project
more intense than that observed in pure SiO2 as a result LUNA by INFM, by the Project MADESS II by CNR and by the ESPRIT
of a very efficient excitation process; Project SCOOP financed by the European Community.
(ii) the rare-earth luminescence in Si nanocrystals is much
more intense than that observed in bulk crystalline Si
as a result of the absence of non-radiative de-excitation References
processes;
(iii) the shape of the luminescence spectra suggests that the 1. L.T. Canham: MRS Bulletin 18, 22 (1993); MRS Bulletin 23, 4 (1998)
excited rare-earth ions are not incorporated within the 2. L.T. Canham: Appl. Phys. Lett. 57, 1046 (1990)
nanocrystals and the energy is transferred at a distance 3. H. Takagi, H. Ogawa, Y. Yamazaki, A. Ishizaki, T. Nakagiri: Appl.
while they are embedded within SiO2 ; Phys. Lett. 56, 2379 (1990)
4. T. Shimizu-Iwayama, K. Fujita, S. Nakao, K. Saitoh, T. Fujita, N. Itoh:
(iv) wavelengths below 1 µm (for example 0.98 µm for Er, J. Appl. Phys. 75, 7779 (1994)
0.98 µm for Yb, 0.92 µm for Nd, 0.78 µm for Tm) can 5. J.G. Zhu, C.W. White, J.D. Budai, S.P. Withrow, Y. Chen: J. Appl.
be excited in Si nanocrystals as a result of the widening Phys. 78, 4386 (1995)
12
6. K.S. Min, K.V. Shcheglov, C.M. Yang, H.A. Atwater, M.L. Brongersma, 17. F. Priolo, G. Franzò, S. Coffa, A. Carnera: Phys. Rev. B 57, 4443
A. Polman: Appl. Phys. Lett. 68, 2511 (1996) (1998)
7. K.S. Min, K.V. Shcheglov, C.M. Yang, H.A. Atwater, M.L. Brongersma, 18. G. Franzò, F. Priolo, S. Coffa, A. Polman, A. Carnera: Appl. Phys.
A. Polman: Appl. Phys. Lett. 69, 2033 (1996) Lett. 64, 2235 (1994)
8. M.L. Brongersma, A. Polman, K.S. Min, E. Boer, T. Tambo, H.A. At- 19. B. Zheng, J. Michel, F.Y.G. Ren, L.C. Kimerling, D.C. Jacobson,
water: Appl. Phys. Lett. 72, 2577 (1998) J.M. Poate: Appl. Phys. Lett. 64, 2842 (1994)
9. H.Z. Song, X.M. Bao: Phys. Rev. B 55, 6988 (1997) 20. J. Stimmer, A. Reittinger, J.F. Nützel, G. Abstreiter, H. Holzbrecher,
10. T. Shimizu-Iwayama, N. Kunumado, D.E. Hole, P. Townsend: J. Appl. Ch. Buchal: Appl. Phys. Lett. 68, 3290 (1996)
Phys. 83, 6018 (1998) 21. C-X. Du, W-X. Ni, K.B. Joelsson, G.V. Hansson: Appl. Phys. Lett. 71,
11. D.J. Lockwood, Z.H. Lu, J.M. Baribeau: Phys. Rev. Lett. 96, 539 1023 (1997)
(1996) 22. G. Franzò, S. Coffa, F. Priolo, C. Spinella: J. Appl. Phys. 81, 2784
12. F.A. D’Avitaya, L. Vervoort, F. Bassani, S. Ossicini, A. Fasolino, (1997)
F. Bernardini: Europhys. Lett. 31, 25 (1995) 23. M. Fujii, M. Yoshida, Y. Kanzawa, S. Hayashi, K. Yamamoto: Appl.
13. K.D. Hirschman, L. Tsybeskov, S.P. Duttagupta, P.M. Fauchet: Nature Phys. Lett. 71, 1198 (1997)
384, 338 (1996) 24. M. Fuji, M. Yoshida, S. Hayashi, K. Yamamoto: J. Appl. Phys. 84,
14. H. Ennen, J. Schneider, G. Pomrenke, A. Axmann: Appl. Phys. Lett. 4525 (1998)
43, 943 (1983) 25. M. Fuji, S. Hayashi, K. Yamamoto: Appl. Phys. Lett. 73, 3108 (1998)
15. J. Michel, J.L. Benton, R.F. Ferrante, D.C. Jacobson, D.J. Eaglesham, 26. X. Zhao, S. Komuro, H. Isshiki, Y. Aoyagi, T. Sugano: Appl. Phys.
E. A Fitzgerald, Y.H. Xie, J.M. Poate, L.C. Kimerling: J. Appl. Phys. Lett. 74, 120 (1999)
70, 2672 (1991) 27. A. Polman: J. Appl. Phys. 82, 1 (1997)
16. J. Palm, F. Gan, B. Zheng, J. Mitchel, L.C. Kimerling: Phys. Rev. B 54, 28. M. Brongersma, Ph.D. Thesis (Utrecht 1998)
17603 (1996) 29. K. Takahei, A. Taguchi: Mater. Sci. Forum 83-87, 641 (1992)