Upload
fabienne
View
223
Download
5
Embed Size (px)
Citation preview
1NtaaAcasnptolocsatfa
fgtevwraoel
1930 J. Opt. Soc. Am. B/Vol. 26, No. 10 /October 2009 S. Ivanova and F. Pellé
Strong 1.53 m to NIR–VIS–UV upconversionin Er-doped fluoride glass for high-efficiency
solar cells
Svetlana Ivanova1,2,* and Fabienne Pellé1
1Laboratoire de Chimie de la Matière Condensée de Paris, UMR 7574 CNRS—ENSCP, Paris Cedex 05, France2Center for Information and Optics Technology, University of Information Technology, Mechanics and Optics,
197101 St. Petersburg, Russia*Corresponding author: [email protected]
Received March 18, 2009; revised June 26, 2009; accepted June 29, 2009;posted August 12, 2009 (Doc. ID 108671); published September 18, 2009
Optical spectra of Er-doped modified ZBLAN glasses are studied at room temperature. Radiative quantumyields of the 4I11/2 and 4I13/2 levels are estimated from the experimentally measured lifetimes and from thespontaneous emission probabilities calculated from the Judd–Ofelt theory. The spectra of upconversion (UC)luminescence excited with 1.53 m cw Er fiber laser are investigated in a wide spectral domain [from the near-infrared (NIR) to UV]. Absolute UC efficiency (i.e., the ratio of UC luminescence power to the absorbed pumppower) is experimentally measured; efficiency of up to 12.7% is obtained. A conclusion is made about perspec-tives of use of the studied glasses as upconverter material for solar cells of enhanced efficiency. © 2009 OpticalSociety of America
OCIS codes: 160.5690, 160.2750, 160.4760, 160.4330, 300.6410, 300.6280.
onraifflmprrcIdqwefcss
1mttcFmdcw
. INTRODUCTIONowadays, the search for efficient upconversion (UC) ma-
erials is driven by numerous potential applications suchs biological labeling [1], UC lasers [2–5], imaging [6],nd, recently, enhanced-efficiency solar cells [7–10].mong factors limiting the efficiency of conventional solarells, one can mention fundamental losses provoked by (i)
loss of solar energy contained in the IR part of solarpectra, or sub-bandgap light transmission, and (ii) inter-al relaxation of hot carriers created after absorption ofhotons with energy much higher than the gap of the ma-erial. In the case of crystalline silicon solar cells, 20%f solar energy contained in the AM1.5 solar spectrum isost due to the transmission [8]. A promising approach tovercome the fundamental problem of sub-bandgap lossesonsists in optical conversion of the IR photons to thehorter wavelengths (upconversion) in solid-state materi-ls doped with rare-earth (RE) ions [9,10]. Theoretically,he Shockley–Queisser efficiency limit [11] can be pushedrom close to 30% up to 40.2% for a silicon solar cell withn upconverter illuminated by nonconcentrated light [9].One of the most popular and high-performing materials
or solar cells today is crystalline silicon, with the energyap 1.12 eV at 300 K corresponding to the shortwaveransmission edge of about 1.1 m. Due to the specificnergy-levels scheme, the Er3+ ion can be considered as aery promising candidate among REs for UC of photonsith wavelengths longer than 1.1 m to the shorter
ange. Encouraging results were published by Richardsnd Shalav, who reported an external quantum efficiencyf about 0.6% for a silicon solar cell with a low-phonon en-rgy NaYF4:Er phosphor used as an upconverter at an il-umination intensity of 0.2 W/cm2 at 1550 nm [9].
0740-3224/09/101930-9/$15.00 © 2
Two important issues govern the choice of the host tobtain a material with high UC efficiency. A host with aarrow phonon spectrum permits to prevent multiphononelaxation from the excited Er levels. Multiphonon relax-tion from the 4I11/2 level, emitting at 1 m, dominatesn most of the oxide hosts with phonon spectrum cut-offrequency about 1000 cm−1, and is partly suppressed inuoride, chloride, chalcogenide, and bromide hosts withaximal phonon energy less than 600 cm−1. Another im-
ortant issue is clustering of Er ions within the host. Non-adiative energy transfers with, in particular, cross-elaxations between Er ions are favored in materialsontaining groups of closely spaced Er ions or Er clusters.n the case of Er ions, due to the specific energy-levelsiagram, energy transfer processes leading both to self-uenching of the luminescence and to UC are possible,hether being beneficial or detrimental for the whole UCfficiency. Finally, glassy or disordered hosts are preferredor the specific application as active layers for highly effi-ient solar cells: inhomogeneous broadening in absorptionpectra of RE ions due to disorder of the host allows ab-orbing a larger part of the solar spectrum.
In this paper, we report on the results of a study of.5 m to NIR UC and absolute UC efficiency measure-ents in Er-doped fluorozirconate glasses. The goal of
his study is to quantitatively evaluate the potential ofhe Er-doped modified ZBLAN glasses for use as an up-onverter material in crystalline-silicon–based solar cells.luoro-zirconate glass was chosen as a host due to its lowaximal phonon energy, broad bands in RE ions spectra
oped in the glass, relatively easy synthesis and highhemical stability compared to chlorides or bromides,hich demonstrate lower maximal phonon energy but
009 Optical Society of America
upfpcZwtbseZtluwtharscdmp
1cSr[ocTrTm
2AIitgBcgBcpeiifoZ
apfe
afoo
BMAC
letpnYow
t
ti8gpwpPspass1ihcUbi
ttsstspptlsp
3AREi
S. Ivanova and F. Pellé Vol. 26, No. 10 /October 2009 /J. Opt. Soc. Am. B 1931
sually are highly hygroscopic. We have studied glassesrepared following two distinct synthesis procedures af-ecting the dopant local environment and favoring or sup-ressing formation of groups of closely spaced Er ions orlusters. Although Er3+ spectroscopic properties inBLAN-type glasses have been extensively investigated,e have studied absorption spectra and calculated the in-
ensity parameters, radiative transition probabilities, andranching ratios for the reason of probable change inpectroscopic parameters due to the modified Er3+ localnvironment. The intensity parameters of Er3+-dopedBLAN-type glasses were obtained in [12–15], radiativeransition probabilities were calculated in [16]. The UCuminescence spectra of Er3+-doped ZBLAN-type glassnder IR excitation at 1520 nm are reported in [17] alongith the discussion of mechanisms leading to the popula-
ion of the Er3+ radiative levels. In the present study weave measured the absolute efficiency of UC of radiationt 1.53 m to the NIR, visible (VIS), and UV spectralanges in the Er-doped modified ZBLAN glasses. The re-ults of the tests of photocurrent generation in arystalline-silicon–based solar cell with the studied Er-oped ZBLAN glasses used as an upconverter at an illu-ination at 1.53 m will be published in a forthcoming
aper.The UC processes are intensively studied since the
960s, when Bloembergen’s idea of the IR quantumounter was proposed [18], and demonstrations of anti-tokes emission in RE-doped solid-state inorganic mate-ials were independently interpreted by F. Auzel19,19(b)] and by Ovsyankin and Feofilov [20]. However,nly a few publications reporting on absolute UC effi-iency measurements have appeared up to date [21–26].hese data are of extreme importance for estimation ofeal feasibility of approaches suggested for applications.o the best of our knowledge, this is the first report on theeasurements of absolute efficiency of 1.5 m to NIR UC.
. EXPERIMENTAL DETAILS. Synthesis
n order to obtain glasses with different distribution ofmpurity ions, fluorozirconate glasses of two differentypes of composition were prepared: (i) diluted Er-dopedlass Z17Er2% with composition 55.6% ZrF4,27.3%aF2,2.9 % LaF3,4.9% AlF3,6.8% NaF, 0.5%InF3 (Er con-entration 3.15·1020 cm−3); and (ii) concentrated Er-dopedlass Z18Er5% with composition 54.0% ZrF4,26.6%aF2,2.8% LaF3,4.8% AlF3 ,6.6% NaF, 0.5%InF3 (Er con-entration 8.07·1020 cm−3); according to the procedureroposed by F. Auzel [27]. Applying this procedure, differ-nt precursors were used for introducing the dopant ionnto the host, which affected the spatial distribution of thempurity ions in the resulting glass and led to rather uni-orm distribution for the Z17Er2% sample and formationf groups of closely spaced Er ions or Er clusters for the18Er5% sample.The constituent chemicals of 4N purity were melted inPt crucible at 850 °C in inert argon atmosphere then
oured into a copper plate preheated at 260 °C and leftor slow cooling inside the furnace. As a result, transpar-nt glasses of light rose color were obtained.
For spectroscopic measurements, the samples were cutnd polished in order to obtain thin plates with parallelaces. Concentration of Er ions per cubic centimeter in thebtained glasses was calculated on the basis of the resultsf the density measurements.
. Spectroscopic and Absolute Upconversion Efficiencyeasurementsbsorption spectra were measured using the VarianARY-5 spectrometer. Kinetics of luminescence from the
4I13/2 and 4I11/2 Er levels was measured at selective pulsedaser excitation (exc=980 and 1530 nm, respectively;mission of an optical parametric oscillator pumped byhe 3rd harmonics of Thales Q-switched Nd:YAG laser;ulse duration 10 ns, energy 3 mJ per pulse). Lumi-escence in the IR was analyzed with a HR 250 Jobin–von monochromator and detected with a PbS (at 2.7 m)r InGaAs (at 1.55 m) detector. Luminescence decaysere recorded with a Tektronix TDS 350 oscilloscope.Refractive index was measured with Abbe refractome-
er at 589.3 nm.Upconversion luminescence was excited with a cw pig-
ailed Er fiber laser (ELT-100 IRE Polus Group) deliver-ng power up to 100 mW via a single-mode fiber with
m core diameter. A splitter and a photodiode inte-rated into the device permitted to control the excitationower. The IR luminescence in the range of 900–1700 nmas analyzed by a monochromator (SpectraPro 750, dis-ersion 1.2 mm/nm), detected by a liquid nitrogen cooledbS detector, and corrected for spectral response of theystem. Glass plates under study were brought as close asossible to the pumping laser fiber termination [Fig. 1(a)],nd the ensemble was introduced inside an integratingphere covered with Spectralon™. An UC luminescenceignal was delivered to a spectrometer (AVASpec-024TEC) by an optical fiber. The whole setup (integrat-ng sphere+spectrometer) is calibrated with the use of aalogen tungsten lamp (10 W tungsten halogen fan-ooled Avalight-HAL). The scheme of setup for absoluteC efficiency measurements is shown in Fig. 1(b). Foroth setups (with a monochromator and with an integrat-ng sphere), the same sample holder [Fig. 1(a)] was used.
The Spectralon™ reflectance is still high 98% athe pump wavelength; however, the pump radiationransmitted through the sample and reflected from thephere inner surface does not constitute an additionalource of excitation. When illuminating the sample withhe pump beam diffused from the sphere surface, no UCignal was detected due to low pump-power density. Thisermits us to take into account only a single pass of theump radiation through the sample. Furthermore, sincehe detection system does not respond at wavelengthsonger than 1.1 m, the pump radiation arriving occa-ionally into the detection system does not introducearasitic signals.
. RESULTS AND DISCUSSION. Absorption Spectra and Judd–Ofelt Calculationsoom temperature absorption cross-section spectra ofr:ZBLAN glasses recorded in the IR-VIS-UV are shown
n Fig. 2. At room temperature, the line shapes in the ab-
slas
s→tcbt(ps
ppudTsfSig
lUTrtlltTtl[
stlette
c
Hla
FmmotltIptelti
FZsffi
1932 J. Opt. Soc. Am. B/Vol. 26, No. 10 /October 2009 S. Ivanova and F. Pellé
orption cross-section spectra of both concentrated and di-uted glass are very similar. The absorption cross-sectiont pump wavelength for the Z17Er2% and the Z18Er5%amples were abs1532 nm=0.4210−20 cm2 andabs1532 nm=0.4910−20 cm2, respectively. Weak ab-orption band around 820 nm corresponding to the 4I15/2
4I9/2 transition is not shown in Fig. 2 due to poor signal-o-noise ratio. The broad absorption band near 1.5 m,orresponding to the 4I15/2→ 4I13/2 transition, is locatedelow the absorption edge of crystalline silicon1.1m. The large FWHM 70 nm of this band andhe high integrated absorption cross-section absd1.2910−18 cm for the Z17Er 2% sample and 1.41
10−18 cm for the Z18Er 5% sample) are convenient forumping with a source characterized by a broadbandpectrum such as the AM1.5 solar radiation spectrum.
From the absorption spectra, the Judd–Ofelt intensityarameters [28,29], spontaneous radiative transitionrobabilities, and branching ratios for Er in the glassesnder study were calculated following common procedureescribed, for instance, by A. Kermaoui and F. Pellé [30].he details will be published in a forthcoming paper. Rea-onable values of the intensity parameters were obtainedor both samples, they are represented in the Table 1.light deviations from the intensity parameters obtained
n [12–15] are explained by different composition of thelass host, and, particularly, by a slightly different Er3+
ig. 1. Experimental setup for absolute UC efficiency measure-ents. (a) Sample is brought as close as possible to the fiber ter-ination and attached to a metallic sample holder. (b) Schematic
f experimental setup. Luminescence excitation source: cw pig-ailed Er fiber laser, exc=1.532 m, pump power=75 mW (stabi-ized, real-time control provided by integrated splitter and pho-odiode); a single-mode fiber with a core diameter 8 m.ntegrating sphere: Avasphere; sphere diameter 5 cm, sampleort diameter 1 cm, knife-edge sphere covered with Spec-ralon™. IR-UV-VIS spectrometer: AVASpec-1024TEC thermo-lectric-cooled CCD fiber-optic spectrometer (75 mm focalength). UC luminescence signal arrives on the end-face of an op-ical fiber in such a way that only diffused reflected luminescences collected.
ocal environment due to the special doping procedure.sing the obtained intensity parameters (summarized inable 1 together with corresponding root mean square er-ors on electric–dipole oscillator strengths), the radiativeransition probabilities, branching ratios, and radiativeifetimes for nine excited Er levels were calculated simi-arly to how it was done in [30]. The results obtained forhe samples Z17Er2% and Z18Er5% are represented inable 1, together with central wavelengths of radiativeransitions and the energy gaps separating the excited Erevels in ZBLAN glasses determined by Y. D. Huang et al.31].
The lifetimes of the 4I13/2 and 4I11/2 Er levels were mea-ured experimentally using selective pulsed laser excita-ion; the obtained results are summarized in Table 1. Theuminescence decays were purely exponential for both lev-ls in both samples. For the 4I13/2 level, the measured life-imes are slightly longer than calculated. This may be at-ributed to the influence of reabsorption. The lumin-scence from the 4I11/2 level was detected at 2.7 m (the
4I11/2→ 4I13/2 transition) in order to avoid reabsorption.Radiative quantum yield of the 4I13/2 and 4I11/2 levels
an be estimated as
i =
jAij
j
Aij + WNR
= j
Aij · iexp. 1
ere, Aij are probabilities of radiative transitions to theower energy levels, WNR is the rate of nonradiative relax-tion and exp is experimentally measured lifetime. The
ig. 2. Room-temperature absorption spectra of Er-dopedBLAN glass (sample Z18Er5%) in the (a) IR and (b) UV-VISpectral domains. Absorption bands correspond to transitionrom the ground state 4I15/2 to the excited states listed in thegures.
i
qctaylunrsnadsdt
BTgltrwlIsoawr
osueUw
an→st→((ttaaop
gm
htsic
l
S. Ivanova and F. Pellé Vol. 26, No. 10 /October 2009 /J. Opt. Soc. Am. B 1933
uantum yield of the luminescence from the 4I13/2 level islose to 100% in both samples, and the quantum yield ofhe 4I11/2 level is =66.5% and =76% for the Z17Er2%nd Z18Er5% samples, respectively. Reduced quantumield of the 4I11/2 level, separated from the next lowerevel by the energy gap E 3450 cm−1, may be attrib-ted to the nonradiative multiphonon relaxation (the pho-on cut-off frequency of fluoride hosts typically lies in theange of 500–600 cm−1, and 6 phonons are needed topan the energy gap). Purely exponential decays of lumi-escence from the 4I11/2 level with exprad signify thebsence of energy transfer UC from the 4I11/2 level (theecays were measured at low excitation power), and mayignify the quenching of luminescence from the 4I11/2 levelue to rapid migration to uncontrolled quenching impuri-ies [32].
. Luminescence Spectrahe luminescence spectra of Er-doped fluorozirconatelasses excited by a cw laser-diode (LD) pumped Er fiberaser =1532 nm, are shown in Fig. 3. The spectra inhe UV–VIS–NIR spectral domain 380–1100 nm wereecorded using the setup with an integrating sphere,hich permits absolute measurements of the power of the
uminescence emitted by the sample. The spectra in theR spectral domain 900–1700 nm were corrected for thepectral response of the registration system and scaled inrder to obtain the same luminescence power of the bandt 0.98 m as that in spectra recorded using the setupith the integrating sphere. Both setups allow one to
ecord spectra in energy per wavelength units. The as-
Table 1. Calculated Spectroscopic ParametersParameters „Ωn… of the Er3+ Ion i
Electronictransition
i
[nm]Ei
cm−1
Z1
Aijed+Aij
md
s−1
4I13/2→ 4I15/2 1530 6230b 109.6 1
4I11/2→ 4I13/2 2707 3450b 19.6 0
4I11/2→ 4I15/2 978 86.1 04F9/2→ 4I9/2 3495 2650b 24.15 04F9/2→ 4I11/2 1946 45.61 04F9/2→ 4I13/2 1133 39.8 04F9/2→ 4I15/2 651 797.9 04S3/2→ 4F9/2 3167 2980b 0.4 04S3/2→ 4I9/2 1662 38.3 04S3/2→ 4I11/2 1207 24.2 04S3/2→ 4I13/2 835 303.2 04S3/2→ 4I15/2 545 686.5 0
i 10−20 cm2 2
2.68 1RMS 10−7
aApproximate central wavelengths i, spontaneous emission probabilities forifetimes i, and energy gaps Ei.
bEnergy gaps between Er excited levels determined by Y.D. Huang et al. 31cExperimentally measured lifetimes of Er3+ levels.
btained luminescence power spectra of Er-doped ZBLANamples are shown in Fig. 3 in energy per wavelengthnits [Figs. 3(a) and 3(b)] and in photon flux per constantnergy interval units [Figs. 3(c) and 3(d)]. The obtainedC luminescence spectra are in qualitative agreementith the results reported in [17].For both samples, the most intensive emission bands
re observed near 1.5 m (transition 4I13/2→ 4I15/2, reso-ant with excitation) and near 0.98 m (transition 4I11/2
4I15/2). Several weaker luminescence bands are ob-erved in the NIR (overlapped bands at 820 nm, transi-ion 4I9/2→ 4I15/2 and at 850 nm, transition 4S3/2
4I13/2), red (660 nm, transition 4F9/2→ 4I15/2), green550 nm, transition 4S3/2→ 4I15/2), and blue-UV410 nm, transition 2G9/2→ 4I15/2) spectral domains. Notehat, for population of the initial levels of these transi-ions, absorption of at least two (4I9/2 level), three (4F9/2nd 4S3/2 levels), or four (2G9/2 level) excitation photonsre needed. The energy-levels scheme of Er permits vari-us paths for population of these levels; the most probablerocesses are summarized in Fig. 4(a).A detailed study of energy transfer in fluorozirconate
lasses is needed in order to identify the most efficientechanisms of population of excited levels 4I11/2, 4I9/2,
4F9/2, 4S3/2, and 2G9/2 upon excitation at exc=1532 nm;ere we have only identified possible processes of popula-ion, relying on the consideration of the Er energy-levelscheme. Since in this paper we are especially interestedn the absolute measurements of 1.5 m NIR and VISonversion efficiency, the detailed study of energy transfer
diative Transitions and Judd–Ofelt IntensityZ17Er2% and Z18Er5% Glassesa
Z18Er5%
i
[ms]Aij
ed+Aijmd
s−1 ij
i
[ms]
9.1(10.5c)
98.5 1 10.1(10.8c)
9.5(6.62c)
22.0 0.18 8.4(6.42c)
97.5 0.821.102 26.3 0.023 0.881
52.2 0.04650.9 0.045
1006 0.8860.950 0.4 0.0003 0.817
45.1 0.03727.3 0.022
338.7 0.277813.2 0.664
6 2 4 6
0.99 2.98 1.40 1.041.26
dipole and magnetic-dipole transitions Aijed+Aij
md, branching ratios ij, radiative
of Ran the
7Er2%
ij
.18
.82
.027
.051
.044
.878
.0004
.036
.023
.288
.652
4
.091.33
electric-
ii
tmclcn
CIpSb
c
wsft3t
qpt
stpvcti
Te
a
t
l
Bf(
Fpaws
1934 J. Opt. Soc. Am. B/Vol. 26, No. 10 /October 2009 S. Ivanova and F. Pellé
n fluorozirconate glasses will be published in a forthcom-ng paper.
For clarity, in Fig. 4(a), only the radiative transitionserminating at the ground state 4I15/2 are shown. Experi-entally, it was possible to detect a weak luminescence
orresponding to transitions originating from the excitedevels 4I11/2, 4F9/2, 4S3/2, and 2G9/2 and terminating at ex-ited states; this will be addressed more thoroughly in theext sections of the paper.
. Upconversion Yieldn order to estimate the fraction of absorbed excitationhotons that are subsequently reemitted in the anti-tokes region, we have followed the procedure proposedy J. F. Suyver et al. [33].The fraction of all photons emitted in the band i is cal-
ulated as
fi =i
int
∀j
jint
, 2
here iint=0
1i ·d denotes the area under the emis-ion band i in the luminescence spectrum observed in therequency domain between 0 and 1, displayed as a pho-on flux per constant energy interval [see Figs. 3(c) and(d)], and summation is over all luminescence bands. Fur-her, let p denote the number of excitation photons re-
ig. 3. Luminescence spectra of Er-doped ZBLAN glasses, (a, choton flux of the UC luminescence bands in the interval of 11000nd the photon flux of the luminescence band at 6500 cm−1 (1 pavelength is shown by an arrow. Excitation power is 75 mW. Id
hown in the figures.
i
uired to induce the emission in the band i. Then theroduct pii
int is the (minimum) number of excitation pho-ons required to induce the emission in the band i.
Next, we make an important assumption that each ab-orbed photon contributes to the emission of a photon, i.e.,he depopulation of all the radiative excited levels is onlyossible via radiative transition to the ground state or UCia excited-state absorption (ESA) or energy transfer up-onversion (ETU). Relying on this assumption, the frac-ion Ri of absorbed excitation photons emitted in the bandcan be calculated as follows:
Ri =pii
int
∀j
pjjint
. 3
he assumption made in Eq. (3) implies that, for eachmitting level,
(i) radiative relaxation dominates multiphonon relax-tion;(ii) branching ratios of the transitions terminating at
he ground state are close to unity;(iii) radiative relaxation dominates cross-relaxation
eading to selfquenching of luminescence or UC.
efore the discussion of the results of calculation of theractions Ri using Eq. (3), we will discuss the criteria (i)–iii), paying the most attention to the 4I and 4I ex-
le Z17Er2% and (b, d) sample Z18Er5%. Note that in (c, d) the0 cm−1 (3- and 4-photon processes) is multiplied by a factor of 20,rocess) is multiplied by (c) a factor of 0.2 and (d) 0.5. Excitation:ation of transitions corresponding to the luminescence bands is
) samp–2700
hoton pentific
13/2 11/2
cn
lttlZnoaTatnmel4
w
t4
ospcf(nvci3pElei
cnmrllCstettutr0c
itr(t
an
Hp3tlg
DSafm
Fp(atctanwam
S. Ivanova and F. Pellé Vol. 26, No. 10 /October 2009 /J. Opt. Soc. Am. B 1935
ited levels, since the most intensive bands in the lumi-escence spectra (Fig. 3) arise from these levels.(i) For the 4I13/2 level, the experimentally measured
ifetimes are slightly longer than calculated (possibly dueo reabsorption), and we estimate the quantum yield ofhis level as close to 100%. The quantum yield of the 4I11/2evel is =69.7% and =76.4% for the Z17Er2% and18Er5% samples, respectively. In a case of intense lumi-escence from the 4I11/2 level, the reduced quantum yieldf this level results in underestimation of the fractions ofbsorbed excitation photons used for UC luminescence.he higher energy radiative levels 4F9/2, 4S3/2, and 2G9/2re probably partly quenched by multiphonon nonradia-ive relaxations (MPNR), since they are separated fromext lower levels by the energy gaps E2650 cm−1,2980 cm−1, 1880 cm−1, respectively [24], and approxi-ately five or fewer phonons are needed to span these en-
rgy gaps. Nevertheless, weak luminescence from theseevels compared to the luminescence from the 4I13/2 andI11/2 levels permits us to suggest that no significant errorill result from obeying the requirement of weak MPNR.(ii) The requirement on the branching ratios of radia-
ive transitions is not fully satisfied for the 4I11/2 and theS levels. However, the weak intensity of luminescence
ig. 4. Energy-levels scheme of Er3+ ion. (a) Mechanisms ofopulation of excited Er levels. Absorption of pump radiationexc=1532 nm): ground-state absorption (GSA) and excited-statebsorptions (ESA1–ESA3) indicated by upward arrows; mul-iphonon nonradiative relaxations (MPNR1 and MPNR2) indi-ated by waved arrows; energy transfers (UC1–UC4); and radia-ive transitions indicated by downward arrows. Numbers nearrrows denoting radiative transitions are wavelengths of lumi-escence in nanometers. (b) Losses of absorbed pump energy. Up-ard arrows, GSA and ESA; waved arrows, nonradiative relax-tion providing thermalization of Stark sublevels of eachultiplet and introducing losses; downward arrow, luminescence
LUM.
3/2
f the transitions originating from the 4S3/2 level with re-pect to that originating from the 4I11/2 and 4I13/2 levelsermits us to assume that the large branching ratio4S3/2→ 4I13/20.28 (Table 1) will not introduce signifi-ant error. The relatively high branching ratio obtainedor the 4I11/2→ 4I13/2 transition 4I11/2→ 4I13/2=0.18Table 1) implies that 18% of radiative transitions origi-ating from the 4I11/2 level, populated essentially by UCia ETU or ESA at 1.53 m pumping, provoke lumines-ence in the Stokes region 2.7 m and are not takennto account in the luminescence spectra presented in Fig.. Consequently, the total fraction of absorbed excitationhotons used for UC luminescence calculated using theq. (3) may be slightly underestimated. For other excited
evels the branching ratios for transitions terminating atxcited levels do not exceed 5% (Table 1) and should notntroduce any significant error.
(iii) The energy-levels scheme of Er does not offer anyross-relaxation scheme leading to self-quenching of lumi-escence for the 4I11/2 and 4I13/2 levels, and the require-ent of domination of radiative relaxation over cross-
elaxation leading to self-quenching is satisfied for bothevels. This is also supported by the fact that the decays ofuminescence from these levels are purely exponential.oncerning higher energy excited levels, concentrationelf-quenching is usually observed in fluoride hosts forhe luminescence originating from the 4S3/2 and 2G9/2 lev-ls for Er concentrations above 1%. Again, we supposehat weak luminescence from these levels compared tohe luminescence from the 4I13/2 and 4I11/2 levels permitss to suggest that self-quenching of luminescence fromhe 4S3/2 and 2G9/2 levels does not introduce significant er-or. The role of energy transfer processes in 1.53 m to.98 and VIS upconversion will be discussed in a forth-oming paper.
The fractions Ri of absorbed excitation photons emittedn the band i, calculated using Eq. (3), are presented inhe Tables 2 and 3. The large fraction of absorbed photonseemitted in the NIR and VIS with the Z18Er5% sample44.5% on total, in the 1.1–0.38 m region) proves thathis glass is very promising for use in the UC devices.
For real applications, the fraction of energy emitted inspecific spectral domain i
exp seems to be a more conve-ient parameter to characterize an UC material:
iexp =
0
1
PmeasUC d
Plumint . 4
ere, PmeasUC represents the UC luminescence power dis-
layed in energy per nanometer units [see Fig. 3(a) and(b)], and Plum
int =380 nm1100 nmPmeas
UC d the total power emit-ed in the whole spectral range of emission. The calcu-ated fractions i
exp from the UC luminescence spectra areathered in Tables 2 and 3.
. Discussionlight discrepancy between the values of fractions Ri ofbsorbed excitation photons emitted in the band i and theractions i
exp of power emitted in a specific spectral do-ain is explained by the losses related to the thermaliza-
tclti
a
wbtci
tspeZeZlas
ys
sitllcmdl[t
a1
ytqp
depadtep→ettE
t tral dom
t
1936 J. Opt. Soc. Am. B/Vol. 26, No. 10 /October 2009 S. Ivanova and F. Pellé
ion of the emitting levels: the power reemitted in a spe-ific luminescence band “i” 0
1PmeasUC d is unavoidably
ess than the power of absorbed excitation photons neededo excite this luminescence pii
int ·hpump, as illustratedn Fig. 4(b).
The relevant correction coefficients may be calculateds
i =
0
1
i · h · d
pi · hpump ·0
1
i · d
=
0
1
i · · d
pi · pump ·0
1
i · d
,
5
here 0
1i ·h ·d represents the power emitted in theand i, and the product pi ·hpump·0
1i ·d representshe (minimum) power required to induce this lumines-ence. The calculated correction coefficients i are listedn the Tables 2 and 3.
Total luminescence energy yield can be calculated ashe ratio of luminescence power emitted in the wholepectral range Plum
int =380 nm1100 nmPmeas
UC d to the absorbedump power tot
exp=Plumint /Pabs
pump. The total luminescence en-rgy yields of 38% and 33% were obtained for the samples17Er2% and Z18Er5%, respectively. Then, the absolutenergy yields of 1.5 m→0.98 m UC can be obtained:
2exp·tot
exp=7.4% and 2exp·tot
exp=11.5% for the samples17Er2% and Z18Er5%, respectively. Similarly the abso-
ute energy yields of 1.5 m→NIR–VIS UC were obtaineds i=2
6 iexp·tot
exp=8.1% and i=26 i
exp·totexp=12.7% for the
amples Z17Er2% and Z18Er5%, respectively.The experimentally measured total luminescence
ields totexp are lower than 100% in both samples under
tudy. This may be partly explained as being due to pos-
Table 2. Details for Emission Bands “i” „
iElectronictransition pi
1 4I13/2→ 4I15/2 12 4I11/2→ 4I15/2 23 4S3/2→ 4I13/2 34 4F9/2→ 4I15/2 35 4S3/2→ 4I15/2 36 4G9/2→ 4I15/2 4
aThe numbers of excitation photons required to induce the emission pi, fractionshe band, correction factors i, and the fractions of power emitted in a specific spec
Table 3. Details For Emission Bands “i”
iElectronictransition pi
1 4I13/2→ 4I15/2 1 72 4I11/2→ 4I15/2 2 23 4S3/2→ 4I13/2 34 4F9/2→ 4I15/2 35 4S3/2→ 4I15/2 36 4G9/2→ 4I15/2 4
aThe numbers of excitation photons required to induce the emission pi, fractionshe band, correction factors , and the fractions of power emitted in a specific spec
iible losses in UC signal measured with the help of thentegrating sphere. These losses are due to the fact thathe reflectance of the metallic sample holder is generallyower than that of Spectralon™. However, we do not be-ieve that more than 50% of luminescence could be lost be-ause of low sample holder reflectance. A similar experi-ental setup with the same integrating sphere and
etection system was used for the measurements of abso-ute efficiency of UC of 0.98 m radiation to the VIS34–38]; no significant loss of luminescence signal was de-ected.
Probable channels of relaxation of the excitation energybsorbed by the samples except luminescence in the.7–0.38 m domain are considered below.First, as was discussed above, the reduced quantum
ield of luminescence from the 4I11/2 level (=66.5% and=76% for the Z17Er2% and Z18Er5% samples, respec-
ively) may be attributed to multiphonon relaxation oruenching on uncontrolled impurities. This can partly ex-lain the reduction of the total luminescence yield.As was discussed in the previous section, a part of ra-
iative transitions, originating from the levels populatedssentially by UC via ETU or ESA at 1.53 m pumping,rovokes luminescence in the Stokes region and thus isbsent in the luminescence spectra in the 1.7–0.38 momain. Possible losses may be identified as being due tohe transition 4I11/2→ 4I13/2 at 2.7 m (branching ratio4I11/2→ 4I13/2 / 4I11/2→ 4I15/2=18: 82). However, roughstimations relying on this branching ratio, fraction ofhotons emitted in the band corresponding to 4I11/2
4I15/2 transition (f2 in Tables 2 and 3), and approximatenergy of photons at 2.78 m show that the energy con-ained in this luminescence band may constitute no morehan 2% of absorbed pump energy for both samples Z17r2% and Z18 Er5%.
5… Resulting from the Z17Er2% Samplea
Ri i iexp
74.2% 99% 78.7%23.6% 76% 19.4%0.4% 62% 0.3%1.0% 77% 0.8%0.8% 97% 0.8%— — —
otons f i emitted in the band, fractions of absorbed excitation photons Ri emitted inain i
exp.
5… Resulting from the Z18Er5% Samplea
Ri i iexp
55.5% 99% 61.5%40.1% 76% 34.8%1.1% 62% 0.7%1.5% 77% 1.2%1.7% 97% 1.7%0.1% 91% 0.1%
otons f i emitted in the band, fractions of absorbed excitation photons Ri emitted inain exp.
i=1–
fi
85.5%13.6%0.2%0.4%0.3%—
of all ph
„i=1–
fi
2.1%6.1%0.5%0.6%0.7%0.01%
of all phtral dom
inmcF4lpsd
mltt
lbahmest
4Ivpde
8ZwUsfisl0almtitoN
eefiacadewpt
taglfatmtnt
cZorrettwrttspmapcut+
ATR“PR0
R
S. Ivanova and F. Pellé Vol. 26, No. 10 /October 2009 /J. Opt. Soc. Am. B 1937
Another source of losses may originate from efficientonradiative relaxation of excitation energy due to ther-alization of the emitting levels and MPNR between
losely spaced levels [e.g., 4I9/2 4I11/2, process MPNR1 inig. 4(a), or 4F7/2 2H11/2 4S3/2, process MPNR2 in Fig.(a)] in a case when radiative relaxation from Er excitedevels is preceded by multiple energy transfers. This hy-othesis needs further investigation on the basis of atudy of energy transfer processes in the considered Er-oped glasses.Finally, the quenching of luminescence due to energyigration to eventual traps via highly populated 4I13/2
evel characterized by milliseconds-range lifetime, quan-um yields close to 100% and typically very high migra-ion rate in fluoride hosts, cannot be completely excluded.
The obtained results provide some evidence that the re-iability of the results of estimations of UC efficienciesased on the analysis of luminescence spectra is limited ifbsolute measurements of UC luminescence intensityave not been performed. In addition to the analysis of lu-inescence spectrum, the absolute measurements of UC
fficiency as it has been described [39–41] or using aetup with integrating sphere can be proposed as a meano accomplish the characterization of UC material.
. CONCLUSIONSn this paper we report on the results of absolute upcon-ersion (UC) efficiency (i.e., the ratio of UC luminescenceower to the absorbed pump power) measurements in Er-oped fluorozirconate glasses in the limiting case of high-xcitation pump power.
Absolute UC efficiency of 1.532 m to NIR and VIS of.1% and 12.7% was obtained in Er-doped ZBLAN glasses17Er2% and Z18Er5%, respectively, under excitationith 75 mW cw pig-tailed Er fiber laser. The most part ofC emission energy is contained in the NIR part of the
pectrum (at 1 m), absolute 1.5 m to 0.98 m UC ef-ciency is 7.4% (Z17Er2%) and 11.5% (Z18Er5%). Thetudied glasses are promising for application as an activeayer of UC solar cells with enhanced efficiency, since the.98 m light has an absorption depth of only 100 mnd is thus strongly absorbed in a wafer-based silicon so-ar cell. For crystalline-silicon–based solar cells, approxi-
ately a half of the energy contained in the VIS part ofhe UC spectrum will be released as heat during thermal-zation of hot carriers created after absorption of VIS pho-ons. To the best of our knowledge, this is the first reportn the measurements of absolute efficiency of 1.5 m toIR UC.The results of this study also reveal the problem of dual
ffect of energy transfer processes on efficiency of UC. Thenergy transfer processes represent, at the same time, ef-cient channels of high-excited-levels population via ETUnd sources of losses via luminescence self-quenching ac-ompanied by dissipation of energy due to phonon-ssisted nonradiative relaxation. Choosing the correctopant concentration, which governs the efficiency of en-rgy transfer, is of great importance especially in caseshen ESA is slightly off-resonant and, consequently, hasoor efficiency. In the studied Er-doped ZBLAN glasseshe higher absolute UC efficiency was demonstrated with
he Z18Er5 sample, where dopant concentration is higher,nd the synthesis favors formation of Er clusters orroups of closely spaced dopant ions. The obtained abso-ute UC efficiency is among the highest values publishedor 0.98 m to VIS upconversion. This permits us to make
conclusion about the positive role of energy transfer inhe hosts under study, in spite of relatively low overall lu-inescence yield achieved, which is attributed mostly to
he dissipation of excitation energy via phonon-assistedonradiative relaxation accompanying multiple energyransfers.
In conclusion, the results of our study of the UC pro-esses under 1.53 m excitation in the Er-doped modifiedBLAN glasses permit us to conclude about the potentialf these glasses for application as an upconverter mate-ial in crystalline-silicon–based solar cells, the latter rep-esenting today the most popular and low-cost type of thexisting solar cells. The use of the efficient conversion ofhe IR radiation near 1.53 m to the wavelengths abovehe c-Si absorption edge demonstrated in this materialill definitely improve the solar cells efficiency thanks to
educing the sub-bandgap losses. It is worth mentioninghat the Er absorption band at 1.5 m corresponds tohe high atmospheric transmission band, and thus thetudied materials are perfectly suitable for on-earth ap-lications. To the best of our knowledge, the results of theeasurements of absolute efficiency of 1.5 m to NIR UC
re reported for the first time. The results of the tests ofhotocurrent generation in the crystalline-silicon solarells with the studied Er-doped ZBLAN glasses used as anpconverter at an illumination at 1532 nm along withhe modelling of the complete setup (solar cellupconverter) will be published in a forthcoming paper.
CKNOWLEDGMENTShis work is supported by the French National Agency foresearch (ANR), program “Solar Photovoltaics,” project
Very high efficiency and photovoltaic innovation” (THRI-V). The authors acknowledge partial support from theussian Foundation for Basic Research (RFBR), grant 07-2-01063.
EFERENCES1. Sh. F. Lim, R. Riehn, W. S. Ryu, N. Khanarian, Ch. Tung,
D. Tank, and R. H. Austin, “In vivo and scanning electronmicroscopy imaging of upconverting nanophosphors incaenorhabditis elegans,” Nano Lett. 6, 169–174 (2006).
2. A. Brenier, “Upconversion lasers,” Encyclopedia of ModernOptics, 508–519 (2005).
3. J. F. Massicott, M. C. Brierley, R. Wyatt, S. T. Davey, andD. Szebesta, “Low threshold, diode pumped operation of agreen, Er3+ doped fluoride fibre laser,” Electron. Lett. 29,2119–2120 (1993).
4. J. Y. Allain, M. Monerie, and H. Poignant, “Tunable greenupconversion erbium fibre laser,” Electron. Lett. 28,111–113 (1992).
5. H. Tobben, “Room temperature CW fibre laser at 3.5 m inEr3+ doped ZBLAN glass,” Electron. Lett. 28, 1361–1362(1992).
6. L. Aigouy, G. Tessier, M. Mortier, and B. Charlot,“Scanning thermal imaging of microelectronic circuits witha fluorescent nanoprobe,” Appl. Phys. Lett. 87, 184105(2005).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
1938 J. Opt. Soc. Am. B/Vol. 26, No. 10 /October 2009 S. Ivanova and F. Pellé
7. T. Trupke, M. A. Green, and P. Würfel, “Improving solarcell efficiencies by up-conversion of sub-band-gap light,” J.Appl. Phys. 92, 4117–4122 (2002).
8. A. Shalav, B. S. Richards, and M. A. Green, “Luminescentlayers for enhanced silicon solar cell performance: Up-conversion,” Sol. Energy Mater. Sol. Cells 91, 829–842(2007).
9. T. Trupke, A. Shalav, B. S. Richards, P. Würfel, and M. A.Green, “Efficiency enhancement of solar cells byluminescent up-conversion of sunlight,” Sol. Energy Mater.Sol. Cells 90, 3327–3338 (2006).
0. S. Ivanova, F. Pellé, A. Tkachuk, M.-F. Joubert, Y. Guyot,and V. P. Gapontzev, “Upconversion luminescencedynamics of Er-doped fluoride crystals for opticalconverters,” J. Lumin. 128, 914–917 (2008).
1. W. Shockley and H. J. Queisser, “Detailed balance limit ofefficiency of p-n junction solar cells,” J. Appl. Phys. 32,510–519 (1961).
2. J. P. McDougall, D. B. Hollis, and J. P. Payne,“Spectroscopic properties of Er3+ in fluorozirconate,germanate, tellurite and phosphate glasses,” Phys. Chem.Glass 37, 73–75 (1996).
3. F. Pellé, N. Gardant, and F. Auzel, “Effect of excited-statepopulation density on nonradiative multiphonon relaxationrates of rare-earth ions,” J. Opt. Soc. Am. A 15, 667–679(1998).
4. J. P. McDougall, D. B. Hollis, and J. P. Payne, “Judd–Ofeltparameters of rare earth ions in ZBLALi, ZBLAN andZBLAK fluoride glasses,” Phys. Chem. Glass 35, 258–259(1994).
5. M. D. Shinn, W. A. Sibley, M. G. Drexhage, and R. N.Brown, “Optical transitions of Er3+ ions in fluorozirconateglass,” Phys. Rev. B 27, 6635–6648 (1983).
6. S. R. Bullock, B. R. Reddy, P. Venkateswarlu, and S. K.Nash-Stevenson, and J. C. Fajardo, “Energy upconversionand spectroscopic studies of ZBLAN:Er3+,” Opt. QuantumElectron. 29, 83–92 (1997).
7. Ch. Xiaobo, “Study of up-conversion luminescence ofEr:ZBLAN excited by 1520 nm laser,” Opt. Commun. 242,565–573 (2004).
8. N. Bloembergen, “Solid state infrared quantum counters,”Phys. Rev. Lett. 2, 84–85 (1959).
9. F. Auzel, C. R. Acad. Sci. (Paris) 262, 1016 (1966); C. R.Acad. Sci. (Paris) 263, 819–821 (1966).
0. V. V. Ovsyankin and P. P. Feofilov, “Cooperativesensitization of luminescence in crystals activated withrare earth ions,” Sov. Phys. JETP Lett. 4, 471–474 (1966).[Zh. Eksp. & Teor. Fiz. Pis’ma 4, 471–474 (1966)].
1. R. H. Page, K. I. Schaffers, P. A. Waide, J. B. Tassano, S. A.Payne, W. F. Krupke, and W. K. Bischel, “Upconversion-pumped luminescence efficiency of rare-earth-doped hostssensitized with trivalent ytterbium,” J. Opt. Soc. Am. B 15,996–1008 (1998).
2. F. Auzel, “Upconversion and anti-Stokes processes with fand d ions in solids,” Chem. Rev. 104, 139–174 (2004), andreferences therein.
3. D. Chen, Y. Wang, Y. Yu, P. Huang, and F. Weng, “Novelrare earth ions-doped oxyfluoride nano-composite withefficient upconversion white-light emission,” J. Solid StateChem. 181, 2763–2767 (2008).
4. Zh. Jin, Q. Nie, T. Xu, Sh. Dai, X. Shen, and X. Zhang,“Optical transitions and upconversion luminescence ofEr3+/Yb3+, codoped lead-zinc-tellurite oxide glass,” Mater.Chem. Phys. 104, 62–67 (2007).
5. V. K. Rai, K. Kumar, and S. B. Rai, “Upconversion in Pr3+
doped tellurite glass,” Opt. Mater. 29, 873–878 (2007).
6. Q. Nie, L. Lu, T. Xu, Sh. Dai, X. Shen, X. Liang, X. Zhang,and X. Zhang, “Upconversion luminescence ofEr3+/Yb3+-codoped natrium-germanium-bismuth glasses,”J. Phys. Chem. Solids 67, 2345–2350 (2006).
7. F. Auzel and D. Morin, French Patent B. F. N 96 13327,Procédé de fabrication de matériaux vitreux dopés etdestinés à l’amplification optique ou laser, October 31,1996.
8. B. R. Judd, “Optical absorption intensities of rare-earthions,” Phys. Rev. 127, 750–761 (1962).
9. G. S. Ofelt, “Intensities of crystal spectra of rare-earthions,” J. Chem. Phys. 37, 511–520 (1962).
0. A. Kermaoui and F. Pellé, “Synthesis and infraredspectroscopic properties of Tm3+-doped phosphate glasses,”J. Alloys Compd. 469, 601–608 (2009).
1. Y. D. Huang, M. Mortier, and F. Auzel, “Stark levelanalysis for Er3+ -doped ZBLAN glass,” Opt. Mater. 17,501–511 (2001).
2. D. L. Dexter and J. H. Schulman, “Theory of concentrationquenching in inorganic phosphors,” J. Chem. Phys. 22,1063–1070 (1954).
3. J. F. Suyver, J. Grimm, K. W. Krämer, and H. U. Güdel,“Highly-efficient near-infrared to visible upconversionprocess in NaYF4/Er3+,Yb3+,” J. Lumin. 114, 53–59 (2005).
4. S. E. Ivanova, F. Pellé, R. Esteban, M. Laroche, J. J.Greffet, S. Collin, J. L. Pelouard, and J. F. Guillemoles,“New concept for efficient upconverter materials,”presented at the Journées Annuelles de la SociétéFrançaise de Métallurgie et de Matériaux, Paris, France,3–6 June 2008; Conference program, abstract I-3-27.
5. S. E. Ivanova, A. M. Tkachuk, F. Pellé, M.-F. Joubert, Y.Guyot, and V. P. Gapontzev, “Energy transfer in RE-dopeddouble fluoride crystals for diode pumped upconversionlasers,” presented at the 13th International Conference onLaser Optics, St. Petersburg, Russia, 23–28 June 2008,Technical program, p. ThR1-p64.
6. F. Pellé, S. E. Ivanova, B. Viana, and A. M. Tkachuk,“Characterization of upconversion materials based on Er-Yb doped fluoride crystals,” presented at the 15thInternational Conference on Luminescence and OpticalSpectroscopy of Condensed Matter (ICL 2008), Lyon,France, 7–11 July 2008, Book of Abstracts, We-P-060,p. 494.
7. S. Ivanova, F. Pellé, B. Viana, and A. Tkachuk, “IR to VISupconversion in Er and Yb doped low phonon energycrystals: spectroscopic study and absolute efficiencymeasurements,” presented at the 3rd EPS-QEODEUROPHOTON Conference on Solid-State, Fiber andWaveguided Light Sources, Paris, France, 2–5 September2008, Conference Digest, THp13, p.46, Technical Digest onCD-ROM.
8. S. I. Ivanova, F. Pellé, R. Esteban, M. Laroche, J. J. Greffet,S. Collin, J. L. Pelouard, and J. F. Guillemoles, “Thin filmconcepts for photon addition materials,” presented at the23rd European Photovoltaic Solar Energy Conference andExhibition, Valence, Spain, 1–5 September 2008,Conference program, abstract 1DV.2.56.
9. Z. Pan, S. H. Morgan, K. Dyer, A. Ueda, and H. Liu,“Host-dependent optical transitions of Er3+ ions in lead-germanate and lead-tellurium-germanate glasses,” J. Appl.Phys. 79, 8906–8913 (1996).
0. R. S. Quimby, M. G. Drexhage, and M. J. Suscavage,“Efficient frequency up-conversion via energy transfer influoride glasses,” Electron. Lett. 23, 32–34 (1987).
1. F. Vetrone, J.-C. Boyer, J. A. Capobianco, A. Speghini, andM. Bettinelli, “980 nm excited upconversion in an Er-doped
ZnO–TeO2 glass,” Appl. Phys. Lett. 80, 1752–1754 (2002).