Light-emission and excited-state dynamics in Tm2+ doped CsCaCl3, CsCaBr3, and CsCaI3

Article abstract of DOI:10.1021/jp055930p

The light-emission and photophysical properties of CsCaCl ₃:Tm²⁺ (1.04%), CsCaBr₃:Tm²⁺ (0.48%), and CsCaI₃:Tm²⁺ (0.76%) are presented. We find that Tm²⁺ is a multiple emitter under 21 834 cm^-1 laser excitation at low temperatures in all three compounds. Several distinct types of emission are observed and characterized: sharp and long-lived 4f-4f emission in the infrared (IR) and up to four broad and fast decaying emission bands in the near-IR and visible, originating from the 4f-5d states of Tm²⁺. The optical spectroscopic properties of the samples are compared, and we find that the measured differences in the relative intensities and the shifts in the position of the emissions can be related to the chemical influence on the absorption and emission properties of Tm²⁺. Thus, it nicely illustrates the principle of chemical variation on the optical spectroscopic properties. An investigation of the temperature dependence of the luminescence yields important information about the dynamics of the excited states. The interplay and competition between radiative and nonradiative pathways is explained and modeled using a single configurational coordinate approach.

Full text of DOI:10.1021/jp055930p

J. Phys. Chem. B 2006, 110, 2093-2101
2093
Light-Emission and Excited-State Dynamics in Tm2+ Doped CsCaCl3, CsCaBr3, and CsCaI3
Judith Grimm, J. Freek Suyver,† Eva Beurer, Graham Carver, and Hans U. Gu1del*
Department of Chemistry and Biochemistry, UniVersity of Bern, Freiestrasse 3, 3012 Bern, Switzerland
ReceiVed: October 17, 2005; In Final Form: NoVember 23, 2005
The light-emission and photophysical properties of CsCaCl3:Tm2+ (1.04%), CsCaBr3:Tm2+ (0.48%), and
CsCaI3:Tm2+ (0.76%) are presented. We find that Tm2+ is a multiple emitter under 21 834 cm-1 laser excitation
at low temperatures in all three compounds. Several distinct types of emission are observed and
characterized: sharp and long-lived 4f-4f emission in the infrared (IR) and up to four broad and fast decaying
emission bands in the near-IR and visible, originating from the 4f-5d states of Tm2+. The optical spectroscopic
properties of the samples are compared, and we find that the measured differences in the relative intensities
and the shifts in the position of the emissions can be related to the chemical influence on the absorption and
emission properties of Tm2+. Thus, it nicely illustrates the principle of chemical variation on the optical
spectroscopic properties. An investigation of the temperature dependence of the luminescence yields important
information about the dynamics of the excited states. The interplay and competition between radiative and
nonradiative pathways is explained and modeled using a single configurational coordinate approach.
1. Introduction
on the 4f-4f excited states as these are essentially ligand-field
independent. This prompted us to extend our studies of Tm2+
doped compounds. In the present contribution, we give a detailed
account of the light-emission properties of the Tm2+ doped cubic
perovskites CsCaBr3 and CsCaCl3. For a comparative discussion
of the excited-state dynamics, we include the results of the
isostructural CsCaI3:Tm2+. An account of the spectroscopic
properties of CsCaI3:Tm2+ together with RbCaI3:Tm2+ will be
provided separately.13 The chemical variation along the per-
ovskite series CsCaCl3, CsCaBr3, and CsCaI3 leads to a
systematic variation of the light-emission properties. These can
be understood from the competition between radiative and
The most common valence state of lanthanides in solids is
the trivalent one. The basic features of the optical spectra of
trivalent lanthanide ions in solids are well understood, and an
enormous amount of studies describing the light-emission
properties can be found in the literature. Some rare earth ions
may also occur in the divalent state. Eu2+, Yb2+, and Sm2+ have
received most attention among them.1 These are the lanthanides
that are most easily stabilized in their divalent oxidation state.
In some lattices, Tm2+ can also be stabilized in its divalent state,
but most Tm2+ compounds are air sensitive,2 and they have
received considerably less attention in the past.
nonradiative relaxation processes in the photophysics of Tm2+
.
Several aspects of the optical spectroscopic properties of
Tm2+ in alkaline earth fluorides have been described and
interpreted already in the 1960s by McClure, Kiss, and Loh.3-5
An early research highlight was the demonstration of laser
In section 4.4, we provide a simple model based on rate
equations, which nicely reproduces the observed trends.
2. Experimental Section
2
2
oscillation on the intraconfigurational F5/2 f F7/2 transition
in CaF2:Tm2+ at low temperatures at the dawn of the invention
of optical lasers.6 It has also been known for decades that, in
addition to the 2F5/2 f 2F7/2 emission, the lowest 4f-5d state is
spontaneously emitting.7 Due to this, Tm2+ has been suggested
for application in IR counting devices as early as 1966.7
Spontaneous 5d-4f emission from the lowest 4f-5d excited
state was since reported for Tm2+ doped SrB4O7,8 SrCl2,9
BaZnCl4, SrZnCl4,10 and CsCaBr3,11 whereas this emission was
not reported for Tm2+ in alkaline earth fluorides.12 Recently,
we described the light-emission properties of Tm2+ in CsCaBr3,
where we have the most unusual situation of five different types
of spontaneous emission processes.11 In particular, emission
from higher lying 5d-4f states in Tm2+ was observed and
characterized.
2.1. Synthesis and Crystal Growth. Single crystals of
CsCaCl3, CsCaBr3, and CsCaI3 doped with Tm2+ were grown
by the Bridgman technique. For the synthesis, stoichiometric
amounts of CsX (X ) Cl, Br, I) and CaX2 were mixed and
Tm2+ was prepared in situ by synproportionation of TmX3
(prepared via the ammonium halide route14 from 99.999% pure
Tm2O3 from Johnson Matthey) and Tm metal (Alfa Aesar
99.9%). All starting materials are hygroscopic and were handled
in a glovebox under N2 atmosphere. Dark green crystals of good
optical quality with diameters up to 2 mm × 2 mm × 2 mm
were obtained. The crystals were checked for purity by X-ray
powder diffraction. The absolute concentrations of Tm in the
crystals were determined with ICP-OES and are 1.04%, 0.48%,
and 0.76% in CsCaCl3, CsCaBr3, and CsCaI3, respectively. The
use of tantalum rather than the more commonly used silica
ampules was found to be indispensable for obtaining crystals
that contained no Tm3+. Crystals grown from silica ampules
contain trace amounts of Tm3+ because under such crystal
growth conditions Tm2 is easily oxidized by silica. Due to the
hygroscopic nature of the crystals, the handling occurred under
inert atmosphere at all times. For absorption measurements, the
The 4f-5d states are affected by changes in the ligand field,
whereas a change in the host lattice will have only a small effect
* To whom correspondence should be addressed. E-mail: hans-
ulrich.guedel@iac.unibe.ch. Tel: +41 31 631 42 49. Fax: +41 31 631 43
99.
† Present address: Philips Research, High Tech campus 34, 5656 AE
Eindhoven, The Netherlands.
10.1021/jp055930p CCC: $33.50 © 2006 American Chemical Society
Published on Web 01/17/2006

2094 J. Phys. Chem. B, Vol. 110, No. 5, 2006
Grimm et al.
samples were polished in a drybox and enclosed in an airtight
copper cell equipped with a quartz window. Thermal contact
of the crystal with the sample holder was provided through the
application of copper grease (Lake Shore Cryotronics Inc.). For
luminescence measurements, the samples were sealed into quartz
ampules under partial pressure of He, which serves as an inert
atmosphere as well as a heat transmitter.
2.2. Spectroscopic Measurements. Sample cooling was
achieved with a closed-cycle cryostat (Air Products) for absorp-
tion and with the He gas flow technique for emission measure-
ments. Absorption spectra were recorded on a Cary 6000i
spectrometer (Varian). Emission spectra were excited with the
457.9 nm (21 834 cm-1) line of an Ar+ laser (Spectra Physics
2060-10 SA). The sample luminescence was dispersed either
with a 0.85 m double monochromator (Spex 1402) or a 0.75 m
single monochromator (Spex 1702). A PMT (Hamamatsu
P3310-01) and a photon counting system (Stanford Research
400) or a Ge detector (ADC 403L and 403HS) interfaced to a
lock-in amplifier (Stanford Research 830) were used for
detection of the signal in the visible and IR ranges, respectively.
The luminescence transients were measured with either the Ar+
laser and an acousto-optic modulator (Coherent 305, Stanford
Research DS 345 function generator) or the fundamental (1064
nm ) 9398 cm-1) or the third harmonic (355 nm ) 28 169
cm-1) of a pulsed Nd:YAG laser (Quanta Ray DCR 3, 20 Hz).
Transient signals were detected as described above using a
multichannel scaler (Stanford Research 430) or an oscilloscope
(Tektronix TDS 540a). All luminescence spectra were corrected
for the sensitivity of the detection system and are displayed as
photon counts vs energy.
Figure 1. Survey emission (a) and onset of single-crystal absorption
spectrum (c) of CsCaCl3:1.04% Tm2+ recorded at 10 K. In (b) and (d),
the analogous data are presented for CsCaBr3:0.48% Tm2+. Both
emissions were photoexcited at 21 834 cm-1. Note the scale changes
in (b), (c), and (d). The labels of the emission bands are used in the
text. The range of (3H6, t2g) and the onset of (3F4, t2g) absorptions are
3
1
indicated. The first S ) /2 and S ) /2 absorption bands within the
(3H6, t2g) multiplet are marked by arrows.
TABLE 1: Positions in cm-1 of Emission Bands A-E at 10
K and Integrated Photon Counts as a Percentage of the
Total Photon Flux at 10 K in Tm2+ Doped CsCaCl3,
CsCaBr3, and CsCaI3. Laser excitation occurred at 21 834
cm-1
3. Results
Parts a and b of Figure 1 show the 10 K emission spectra of
CsCaCl3:Tm2+ and CsCaBr3:Tm2+ along with the corresponding
absorption spectra 1c and 1d, respectively. Laser excitation for
the emission spectra occurred at 21 834 cm-1 for both samples.
The observed emission bands are labeled A-E, and the
electronic transitions to which they refer will be discussed in
section 4. The same notation is used for both samples. The band
positions and the partition of the emitted photons at 10 K among
the five bands A-E are given in Table 1 including the results
of CsCaI3:Tm2+. The positions of the broad emission bands
B-E all undergo a redshift of about 300-350 cm-1, as the host
lattice is changed from the chloride to the bromide and about
900 cm-1 from the bromide to the iodide. The sharp emission
A, in contrast, is shifted to lower energy by only about 15 cm-1
from chloride to bromide and about 11 cm-1 from bromide to
iodide. The partition of photons among the bands at 10 K is
very different for the three samples. While D is absolutely
dominant in the bromide and iodide, both B and D have sizable
intensity in the chloride. All the transitions A-E are observable
at 10 K in the bromide and iodide, whereas C could not be
detected in the chloride.
Parts a and b of Figure 2 show the high-resolution 10 K
emission spectra of bandsystem A in CsCaCl3:Tm2+ and
CsCaBr3:Tm2+, along with the corresponding 10 K absorption
spectra 2c and 2d, respectively. Laser excitation for the emission
spectra occurred at 21 834 cm-1 for both samples. Disregarding
the lines marked with an asterisk, which we assign to nonregular
crystal sites, the sharp-line spectra of the chloride and bromide
resemble each other and will be analyzed in section 4.3.
In Figure 3, the emission spectra of CsCaCl3:Tm2+ and
CsCaBr3:Tm2+, respectively, are shown at three different
temperatures, 10, 130, and 300 K. The spectra are scaled within
integrated photon
compound
CsCaCl3
emission
position (cm-1
)
counts (%)
D
B
E
A
D
C
B
E
19 450
12 497
10 647
8810
19 152
13 676
12 237
10 370
8796
36.5
63
0.2
0.3
97.6
0.2
2
0.1
0.1
92.7
2.3
4.6
0.1
0.3
CsCaBr3
CsCaI3
A
D
C
B
E
18 344
12 703
11 348
9503
A
8785
3a and 3b, therefore intensities at different temperatures can be
compared. There is a general shift of photons from high-energy
to low-energy bands in both samples as the temperature is raised.
The temperature dependence of the integrated photon counts
of the three dominant emissions A, B (+ C in CsCaI3:Tm2+),
and D under 21 834 cm-1 excitation is shown in Figure 4 for
CsCaCl3:Tm2+, CsCaBr3:Tm2+, and CsCaI3:Tm2+, respectively.
The data will be analyzed in section 4.4. In Figure 5, the
temperature dependences of the integrated photon counts and
the lifetime of emission B in CsCaCl3:Tm2+ (5a) and CsCaBr3:
Tm2+ (5b) are compared. For both compounds, the lifetime is
constant below 100 K, whereas the intensity shows an increase
with increasing temperature. Above 100 K in the chloride and
150 K in the bromide, the lifetime decreases along with the
integrated photon counts. The lifetimes of the emission bands

Dynamics of Tm2+-Doped Perovskite Compounds
J. Phys. Chem. B, Vol. 110, No. 5, 2006 2095
2
2
Figure 2. The 10 K emission spectrum of the F5/2 f F7/2 transition
(band A) of Tm2+ is shown in (a) for CsCaCl3 and (b) for CsCaBr3.
Both emissions were photoexcited at 21 834 cm-1. The 10 K single-
crystal absorption spectrum of the F7/2 f F5/2 transition of Tm2+ is
shown in (c) for CsCaCl3 and (d) for CsCaBr3. The emission lines that
are labeled with an asterisk are attributed to Tm2+ on nonregular sites
(see text). The bands labeled νCl and νBr are vibronic sidebands of the
2
2
2F5/2 f F7/2 transition in CsCaCl3:Tm2+ and CsCaBr3:Tm2+, respec-
tively.
2
Figure 4. Temperature dependence of the integrated photon counts
(normalized to 1 for the most intense band) of the three dominant
emissions: 4, A; 0, B (+C in CsCaI3:Tm2+); and O, D in (a) CsCaCl3:
Tm2+, (b) CsCaBr3:Tm2+, and (c) CsCaI3:Tm2+. The photon counts
were corrected for the decreasing absorption cross section with
increasing temperature at the excitation energy of 21 834 cm-1. The
solid lines represent simulations of the data using the model in section
4.4 (eq 2). The input and fit parameters are listed in Tables 3 and 4,
respectively.
Figure 3. Emission spectra of Tm2+ in (a) CsCaCl3 and (b) CsCaBr3
at three temperatures: 10, 130, and 300 K. The emission was
photoexcited at 21 834 cm-1. The spectra are shifted arbitrarily along
the vertical axis. The spectra within (a) or (b) have the same scale,
therefore intensities at different temperatures can be compared. Only
the prominent emissions A, B, and D are labeled.
Figure 5. Temperature dependence of emission lifetimes (4) and
intensities (0) of band B in (a) CsCaCl3:Tm2+ and (b) in CsCaBr3:
A, B, and D of Tm2+ are given in Table 2 for CsCaCl3, CsCaBr3,
and CsCaI3. They will be used to determine the decay rate
constants in the rate equation modeling in section 4.4.
Tm2+
.
Pm3m.15,16 The Tm2+ ions replace Ca2+ on site 1(b) with Oh
point symmetry. Both samples undergo phase transitions upon
cooling, which is common for cubic perovskites. The cubic to
tetragonal phase change at T ) 95 K in CsCaCl3 is well
documented in the literature.17,18 The site symmetry of Ca2+
4. Analysis
4.1. Structures and Sites. CsCaCl3 and CsCaBr3 are cubic
perovskites at room temperature, crystallizing in the space group

2096 J. Phys. Chem. B, Vol. 110, No. 5, 2006
Grimm et al.
TABLE 2: Measured Lifetimes τ of the Emission Bands D,
In the absorption spectra in parts c and d of Figure 1, all the
bands above 10 000 cm-1 are due to 4f-5d transitions, in which
one electron is promoted from the 4f orbitals to the 5d orbitals.
These absorption transitions are between 1 and 4 orders of
magnitude more intense than the intraconfigurational 4f-4f
transitions. In contrast to the latter, the 4f-5d transitions are
parity allowed, which can result in a huge increase in the
oscillator strength. The states resulting from the (4f)12(5d)1
electron configuration are represented with a displacement along
the configurational coordinate in Figure 6, in agreement with
the observed broad absorption and emission bands. The
displacement results from the different chemical bonding in the
(4f)13 and (4f)12(5d)1 configurations. The (4f)12(5d)1 electron
configuration has a total degeneracy of 910, and energy splittings
occur as a result of various interactions.
B, and A of Tm2+ Doped CsCaCl3, CsCaBr3, and CsCaI3
emission
Da
T
CsCaCl3
CsCaBr3
CsCaI3
10 K
τ
904 ns
1.5 µs
296 µs
501 µs
1.7 µs
1.8µs
323 µs
534 µs
1 µs
2 µs
391 µs
650 µs
τcalc
τ
B
A
10 K
300 K
τ
a The radiative lifetime τcalc of emission D is calculated from eq 7
(see section 4.4).
An important interaction in the (4f)12(5d)1 configuration is
the ligand field interaction of the 5d electron. The octahedral
ligand field splits the 5d orbitals into t2g and eg sets, separated
by 10 Dq. All bands below 25 000 cm-1 that are shown in the
absorption spectra (parts a and b of Figure 1) can be attributed
to excited states with the 5d electron in the t2g set. The (4f)12
part of the configuration is split into the 2S+1LJ terms of Tm3+
by the Coulombic repulsion and the spin-orbit coupling of the
4f electrons. Thus, the energy multiplets of the Tm2+ (4f)12-
(5d)1 configuration can be roughly characterized as (2S+1LJ, t2g)
and (2S+1LJ, eg). For Tm2+, we expect each of these multiplets
to be split into a set of lower energy high-spin S ) 3/2 and a set
1
of higher energy low-spin S ) /2 states due to the isotropic
Figure 6. (a) Schematic single configurational coordinate (SCC)
diagram of Tm2+ assuming harmonic potentials and equal force
constants for all states. The emitting excited states are schematically
displayed in (b) with the radiative and nonradiative processes indicated
as straight and curly arrows, respectively. The labeling of the transitions
corresponds to Figure 1 and Tables 1 and 2.
exchange part of the Coulombic interaction between the 4f and
5d electrons.11 As a consequence, the lowest-energy (4f)12(5d)1
bands are expected to have formally “spin-forbidden” character,
which will manifest itself in weaker oscillator strengths com-
pared to the formally “spin-allowed” bands.3
The emission band B (Figures 1 and 6) originates from the
lowest-energy 4f-5d state. The corresponding weak absorption
band is shown in Figure 1c for CsCaCl3:Tm2+ and 1d for
CsCaBr3:Tm2+. Band B is thus assigned as a “spin-forbidden”
below the phase change is C4h.17 The distortion from Oh
symmetry at the Ca2+ site is small.18 CsCaBr3 undergoes two
phase changes at 237 and 145 K, respectively.16 CsCaI3
crystallizes in the orthorhombic Pnma phase at 300 K.19 The
Ca2+ in the orthorhombic Pnma phase has site symmetry Ci.
Regardless of the phase changes, the Tm2+ ion is always
coordinated by six X- (X- ) Cl, Br, I). From the spectroscopic
data of the three samples, we conclude that only band splitting
due to symmetry reduction is observable in bandsystem A and
is in the order of a few wavenumbers (vide infra). The data
interpretation of the 4f-5d states is therefore done in the
octahedral approximation.
3
2
5d-4f transition from (3H6, t2g)S ) /2 to F7/2. Band C is
correlated with the first intense 4f-5d absorption band and is
therefore assigned as a “spin-allowed” 5d-4f transition from
1
2
(3H6, t2g)S ) /2 to F7/2. Quite to our surprise, we observed
emission from an even higher lying 4f-5d state: band D is
assigned to a transition from the lowest excited state of the (3F4,
2
t2g) multiplet to the F7/2 groundstate. Finally, band E is the
interexcited-state transition from (3F4, t2g) to 2F5/2. The spin of
the emitting (3F4, t2g) state cannot be specified. Apparently, spin
is not as well defined in the (3F4, t2g) multiplet as it is in (3H6,
t2g).11,21
4.2. Emitting States and Spectral Assignments. The as-
signment of the emission bands has already been discussed for
CsCaBr3:Tm2+ in ref 11. Figure 6a shows a schematic single
configurational coordinate (SCC) diagram of the lowest excited
states of Tm2+ assuming harmonic potentials and equal force
constants for all states. The relevant radiative and nonradiative
relaxation processes are indicated in Figure 6b as straight and
curly arrows, respectively.
4.3. 4f-4f Emission A and Crystal Field Splitting of the
2
2F5/2 and F7/2. The high-resolution absorption and emission
data obtained at 10 K in the region of the 4f-4f transitions
(Figure 2) exhibit considerable fine structure. Some lines around
8800 cm-1 in both compounds coincide in absorption and
emission. They are labeled A0 and A1, and they can be assigned
to electronic origins as shown in the diagram of Figure 7. Since
the intensity ratio of A1 and A0 was found to increase with
temperature between 10 and 30 K, the splitting between A0 and
A1 is necessarily an excited-state splitting. Figure 7 is a
schematic representation of the splitting of 2F7/2 and 2F5/2 in an
Oh (first step) and then C4h or Ci (small distortion from Oh)
crystal field. The octahedral splittings are of the order of several
Tm2+ has a (4f)13 ground-state electron configuration, that
is, it is isoelectronic with Yb3+. The two multiplets arising from
the (4f)13 electron configuration are 2F7/2 and 2F5/2. The energy
of the 2F5/2 multiplet is about 1200 cm-1 lower in Tm2+ than in
Yb3+, reflecting the smaller spin-orbit coupling in Tm2+ 20
. As
seen in Figure 2, the sharp 4f-4f bands are observed in
absorption and emission around 8800 cm-1 in both CsCaCl3:
Tm2+ and CsCaBr3:Tm2+ at 10 K. The fine structure will be
analyzed in section 4.3. For the modeling in section 4.4, the
total 4f-4f emission intensity (transition A in Figure 6) will
be considered.
hundred cm-1 22
, and an unambiguous assignment of absorption
or emission features outside the 8800 cm-1 region to electronic
origins is not possible. The situation is very similar to Yb3+ in

Dynamics of Tm2+-Doped Perovskite Compounds
J. Phys. Chem. B, Vol. 110, No. 5, 2006 2097
Figure 7. Schematic representation of the crystal field (CF) splitting
2
2
of the F7/2 and F5/2 states of Tm2+ in CsCaCl3 and CsCaBr3. The
energies of the CF levels are determined from Figure 2.
many host lattices. Unlike 4f-4f transitions in most other
2
2
lanthanides, the F7/2 T F5/2 bands in (4f)13 are characterized
by the presence of strong vibronic sidebands. The bands labeled
Figure 8. Simplified SCC diagram of Tm2+ to model the temperature
dependence of the three major emissions A, B, and D. The four
emissions in the SCC diagram correspond to the following transitions:
A, |1 f |0 ; B, |2 f |0 ; D, |$ f |0 ; E, |3 f |1 . Curly arrows are
used to indicate nonradiative processes (Wij) and straight arrows for
radiative transitions (Ri). The branching ratios η30 and η31 determine
νCl and νBr in the emission spectra of Figure 2 are such bands.
Their identity as vibronic sidebands is confirmed by the reduced
energy difference to the strong A0 and A1 origins in the bromide,
a result of the smaller vibrational energies in the bromide. At
the phase transitions of the host lattice, the site symmetry of
Tm2+ is reduced from Oh to C4h and Ci in the chloride and
bromide, respectively. The degree of the distortion from Oh must
the ratio of photons in the emission bands D and E, with η30 + η31
)
1. P is the laser power and σ the absorption cross section at the
excitation energy.
2
be relatively small, however, since the splitting of F5/2 (Γ8)
into the two Kramer’s doublets 0′ and 1′ is only 10 and 8 cm-1
.
describe the evolution of the excited-state populations Ni (i )
1, ..., 3) upon excitation
respectively. The bands marked with asterisks in the emission
spectra of Figure 2 are assigned to other species due to their
different excitation spectrum. Tm2+ ions on some defect sites
or Tm2+ pair transitions are possibilities.
dN
3 ) σPN0 - W32N3 - R3N3
dt
4.4. Modeling the Excited-State Dynamics. The very
pronounced redistribution of intensity with temperature among
the three major emissions A, B, and D (Figures 3 and 4) provides
clear evidence for thermally activated nonradiative relaxation
processes. A simple model to treat such nonradiative processes
is based on the single configurational coordinate (SCC) model.23
The SCC diagram for Tm2+ doped halides is shown in Figure
6a where the configurational coordinate represents an effective
accepting mode for the relaxation process. To model the
temperature dependence of emissions A, B, and D, we simplified
the SCC diagram. Since emission C is absent in CsCaCl3 and
very weak at all temperatures in the bromide, we left emission
C out of the calculations. In the iodide, emission C is not
negligible and therefore the integrated photon counts of emission
B and C were added together in order to keep the model simple.
Emission E shows the same temperature dependence as emission
D, because the initial state is the same. The simplified SCC
diagram that contains all the relevant radiative (straight arrows)
and nonradiative processes (curly arrows) that are considered
dN
2 ) W32N3 - (W21 + W20 + R2)N2
dt
dN
1 ) W21N2 + η31R3N3 - R1N1
(1)
dt
The steady state solutions are
σPN0
N3 )
W32 + R3
σPN0W32
N2 )
(W21 + W20 + R2)(W32 + R3)
σPN0(η31R3(W21 + W20 + R2) + W32W21)
R1(W21 + W20 + R2)(W32 + R3)
N1 )
(2)
2
is shown in Figure 8. The state |0 corresponds to the F7/2
2
3
ground state of Tm2+, |1 to F5/2, |2 to (3H6, t2g)S ) /2, and
|3 to (3F4, t2g). The radiative and nonradiative rate constants
are labeled Ri and Wij, respectively. Subscript i stands for the
initial state and subscript j for the final state. The ratio of photons
in the emission bands D and E is given by the branching ratios
The temperature dependence of the nonradiative decay rate
constants Wij is calculated within the framework of the SCC
model.24 We assume linear coupling and harmonic potentials.
Wij can be written as25
Wij ) CijFij
(3)
η
30 and η31, with η30 + η31 ) 1. P is the laser power and σ the
absorption cross section at the excitation energy.
For the laser powers used in our experiments, we can assume
that the ground-state population N0 is not affected by the
excitation. We further assume that multiphonon relaxation from
|1 to |0 is negligible. The following rate equations then
where Cij is the electronic factor and Fij the thermal Franck-
Condon factor.
The temperature-dependent Franck-Condon factor Fij can
be written as25

2098 J. Phys. Chem. B, Vol. 110, No. 5, 2006
Fij(T) )
Grimm et al.
TABLE 3: Input Parameters from the Experimental Data
for the Rate Equation Model Discussed in Section 4.3
1 + m p /2
ij
Ip (2Sij(m(m + 1))0.5) (4)
CsCaCl3:Tm2+
CsCaBr3:Tm2+
CsCaI3:Tm2+
exp(-Sij(1 + 2m))
(
)
ij
m
E32
E21
S32
S21
R3
R2
R1
2700 s-1
2700 s-1
3906 s-1
0.1
2700 s-1
2995 s-1
0.1
where Sij is the Huang-Rhys factor for the |i f |j transition,
3929 s-1
0.1
Ip is the modified Bessel function of the first kind, and m is
ij
2
2
2
the thermal population of the effective vibrational mode
1 106 195-W32a s-1
3378 s-1
1996 s-1
0.00816
588 235 s-1
3096 s-1
1873 s-1
0.00125
954 000 s-1
2555 s-1
1538 s-1
0.00123
1
m )
(5)
pωeff
η31
exp
- 1
(
)
a At T ) 10 K.
kT
with k the Boltzman constant and pωeff the energy of the
effective accepting mode in the SCC model. Pij is the number
of effective phonons required to bridge an energy gap
TABLE 4: Parameters that Are Obtained in the Fit: Wij
(calculated from eq 3 using the fitted Cij) and pωeff
CsCaCl3:Tm2+
CsCaBr3:Tm2+
CsCaI3:Tm2+
W32 at 10 K
W21 at 10 K
W20 at 10 K
pωeff
757 217 s-1
0.05 s-1
2532 s-1
43 761 s-1
Eij
3.2 × 10-7 s-1
1.8 × 10-6 s-1
109 cm-1
1.6 s-1
Pij )
(6)
a
0.02 s-1
pωeff
106 cm-1
212 cm-1
Eij is the adiabatic energy difference between the minima of
the potential energy curves of states |i and |j . For a detailed
derivation of these equations, we refer the reader to refs 24 and
25.
We use eqs 2 and 3 to model the temperature dependence of
the relative photon counts of the emissions A, B, and D shown
in Figure 4. All Wij values are defined as in eq 3 except for W20
which is defined as W20 ) C20W21 so as to introduce a
“branching ratio” that describes the competition between the
nonradiative processes |2 f |1 and |2 f |0 . The fitting
parameters are then the three electronic factors C32, C21, and
C20 as well as pωeff.
All the other parameters are deduced from the spectroscopic
data and fixed to the values given in Table 3. We include the
data obtained for CsCaI3:Tm2+ in this comparative study. The
radiative rate constants Ri are assumed temperature independent
and determined from the lifetimes (Table 2). The high radiative
rate constants R3 of emission D are notable. It might be possible
that the (3F4, t2g) multiplet is located inside the conduction band
and that a mixing of states occurs. States located inside the
conduction band can have high radiative decay rate constants.26
The radiative decay rate constant R3 can be calculated from the
oscillator strength f of the first (3F4, t2g) absorption band
assuming equal degeneracies in the ground and excited state25
a The process W20 is not active in the iodide up to 300 K.
compounds. Similarly, the Huang-Rhys factors S21 and S32 are
set to 2 and 0.1, respectively, for all three compounds. This is
a rather crude approximation, since S21 is expected to vary across
the series. But we have no experimental means of determining
this variation, and we want to keep the model as simple as
possible. The pump rate constant σPN0 is normalized to 1 s-1
This means that we are in the linear regime and we are ignoring
ground-state bleaching.
.
The integrated photon counts of all three emissions A, B,
and D are simultaneously fitted (simple minimization of ø2 over
the parameter space) for a given compound, and the solid lines
in Figure 4 represent the best fits. Overall, the model is able to
reproduce the data of CsCaCl3:Tm2+ and CsCaBr3:Tm2+ very
well. The deviations are larger for CsCaI3:Tm2+. The observed
drop of emission A in CsCaCl3:Tm2+ above 200 K is the result
of energy transfer, see section 5, and thus not within the realm
of our photophysical model.
The parameter values obtained in the least-squares fits are
listed in Table 4 for the three compounds. A graphical
representation of the competition between radiative and non-
radiative processes is provided in Figure 9, where the relevant
rate constants are logarithmically plotted as a function of
temperature for all three systems. It provides insight into the
changes that occur along our chemical coordinate.
n[(n2 + 2)/3]2
R ) f
(7)
Rλ2
We are confident that the fits for CsCaCl3:Tm2+ and CsCaBr3:
Tm2+ and the resulting parameter values represent reasonable
approximations of the physical reality in these systems. The
fitted effective phonon energies for CsCaCl3:Tm2+ and CsCaBr3:
Tm2+ are lower than the highest energy vibrational energies
and are comparable with literature data.27-29 The effective
phonon energy for such nonradiative processes is often lower
n is the index of refraction and is estimated to be 1.58, 1.64,
and 1.73 for CsCaCl3, CsCaBr3, and CsCaI3, respectively. R is
a constant (1.5 × 104 s‚m-2) and λ is the emission wavelength.
We find good agreement of the calculated and measured
radiative lifetimes of emission D in all three materials (see Table
2). It is thus unlikely that mixing with the conduction band states
of the host affects the luminescent properties of our samples.
Luminescence D in CsCaCl3:Tm2+ has a nonradiative contribu-
tion already at 10 K. In this case, R3 is given by (at 10 K)
than the highest energy vibrations.27,30 For CsCaI3:Tm2+
,
however, the simple model is a poor approximation of the
complex photophysical processes. The best evidence for this is
the value of pωeff ) 212 cm-1 obtained for the effective
accepting mode. This is higher than any vibrational energy in
CsCaI3 and thus not physical. Evidence for much more complex
photophysics in iodide hosts is also provided by the behavior
of Tm2+ in RbCaI3, in which the 4f-4f emission A already
dominates at 10 K.13 Since our model is extremely simple, there
are many possible reasons why it is no longer fully adequate in
CsCaI3:Tm2+. The necessary caution will therefore be used in
the interpretation of the iodide fit in section 5. But it must be
R3 ) k3 - W32
(8)
k3 is the measured decay rate constant of emission D at 10 K
(Table 2). The branching ratio η31 is the experimental photon
ratio of bands E/D. The energy gaps Eij are estimated from the
absorption and luminescence data. While E21 can be accurately
determined, there is considerable uncertainty in the value of
E32. We use the same value of 2700 cm-1 for all three

Dynamics of Tm2+-Doped Perovskite Compounds
J. Phys. Chem. B, Vol. 110, No. 5, 2006 2099
to our knowledge, emission from higher than the lowest-energy
“spin-allowed” and “spin-forbidden” 4f-5d states has not been
reported in lanthanide systems.12
On the basis of Figure 9, we can understand how this
exceptional behavior comes about. First, we note that emission
D has the highest radiative rate constant (R3) of all the observed
emissions. R3 is the dominant relaxation path in the iodide up
to 220 K, in the bromide up to 100 K. At these temperatures,
the crossing of W32 with R3 occurs in the two samples and thus
nonradiative relaxation to (3H6, ttg) becomes the dominant
relaxation process at elevated temperatures. These crossover
points are highlighted and connected with a dashed line in Figure
9. In the chloride, R3 has a strong competition already at 10 K
from the nonradiative W32, and above 50 K, R3 loses out
completely; see also data in Figure 4. The comparison of the
three samples shows the strong dependence of W32 on the
chemistry. W32 decreases by 2 orders of magnitude at 10 K in
the bromide compared to the chloride, but for CsCaI3:Tm2+
,
the values lie between the two others. A decrease of W32 (10
K) along the halide series is expected on the basis of decreasing
phonon energies. The observed increase in the iodide clearly
shows the limits of the model. The evolution of W32 with
temperature is similar in the chloride and bromide, whereas in
the iodide, the increase with temperature is much less steep.
The chemical variation thus affects not only the 10 K value of
W32 but also its temperature dependence, with a significant
difference in the iodide.
Emission E has the same temperature dependence and
2
originates from the same state as D, but the final state is F5/2
instead of the ground state. It is extremely weak, with D/E
photon ratios of 123:1, 800:1, and 806:1 in the chloride,
bromide, and iodide, respectively. We ascribe this extreme
branching ratio to a selection rule ∆J ) 1.33 The lowest-energy
component of (3F4, ttg) has, according to Hund’s rule, a J ) 9/2
value, if we only consider the spin angular momentum of the
5d electron. Thus, a transition to 2F7/2 is allowed, whereas 2F5/2
is not accessible in this rough approximation.
Figure 9. Semilog plot of the radiative Ri and nonradiative Wij rate
constants as a function of temperature for Tm2+ doped CsCaCl3 (a),
CsCaBr3 (b), and CsCaI3 (c). The radiative rate constants Ri are given
in Table 3. The nonradiative rate constants Wij are calculated from eq
3 using the parameter values in Tables 3 and 4. The crossover points
of W32 with R3 and W21 with R2 are highlighted and connected with
dashed lines for better illustration of the influence of the chemical
variable on the photophysics of Tm2+
.
Emission B is a broad 5d-4f band in the near-IR around
12 000 cm-1. It is assigned to a transition from the lowest-energy
(3H6, ttg)S ) 3/2 to 2F7/2. It is the dominant feature in the emission
spectrum of the chloride up to 120 K, of the bromide between
100 and 250 K, and of the iodide above 220 K; see also Figures
noted that an analysis of the type provided in Figure 4 with the
parameters in Tables 3 and 4 is extremely ambitious and has
not been reported so far. Fits of the temperature dependence of
emission intensities or lifetimes are usually done for just one
transition in one compound.27-31 We extend this by including
three competing transitions in each of three related compounds,
and we do this without adjusting the model. We thus gain
considerably more insight into the photophysics and their
dependence on the chemical variable in these systems. But we
do reach the limit of applicability of the model.
3
3 and 4. (3H6, ttg)S ) /2 is being populated by multiphonon
relaxation (W32) from the higher (3F4, ttg) metastable state. Since
the decrease of emission D with increasing temperature is also
determined by W32, we can immediately understand that the
observed rise with the temperature of B is correlated with the
decline of D, nicely seen in Figure 4 for all three samples. The
lifetime of emission B is constant up to 70 and 160 K in
CsCaCl3:Tm2+ and CsCaBr3:Tm2+, respectively (see Figure 5),
which proves the radiative nature of the decay up to these
temperatures. Above 120 K in the chloride and 250 K in the
bromide, emission B is taken over by emission A. Their
correlated rise and decline is again nicely observed in Figure
4. The nonradiative relaxation process from (3H6, ttg)S ) 3/2 to
2F5/2 (W21) is responsible for this, see the crossing of W21 with
R2 at 120 and 250 K in the chloride and bromide, respectively,
in Figure 9. This onset of nonradiative relaxation is also
confirmed by the simultaneous decrease of the respective
lifetimes in Figure 5. In the iodide, the W21 process is not
competitive up to room temperature, that is, we observe no
crossing point of W21 with R2.
5. Discussion
Each of the five emissions A-E will be discussed separately,
with the main emphasis on its variation with temperature for a
given compound and its variation along the series of Tm2+
doped halides.
Emission D is the highest energy emission in the chloride
and bromide. In the iodide, we observe a weak emission below
200 K, which lies at even higher energy. This is discussed in
ref 13. In all three compounds, emission D appears bright green
to the eye. It is a broad 5d-4f emission band and has its origin
in the lowest states of the (3F4, ttg) multiplet. With the exception
of 4f-4f emissions in lanthanides and actinides, Kasha’s rule
is a guiding principle for light-emitting materials.32 It states that,
if at all, spontaneous emission occurs from the lowest-energy
excited state. Our emission D is an exception to this rule, and
Emission C is a weak emission only observed in the bromide
and iodide that originates in a higher excited state within the

2100 J. Phys. Chem. B, Vol. 110, No. 5, 2006
Grimm et al.
(3H6, ttg) multiplet. We assign it to the (3H6, ttg)S ) 1/2 f 2F7/2
transition. This is a formally “spin-allowed” transition with a
high oscillator strength. The energy difference between the
The choice of Tm2+ as the active ion turns out to be fortunate
for both chemical and physical reasons. It is possible to stabilize
the divalent Tm in all the halides (chloride, bromide, and iodide).
Physically, a big advantage of the (4f)13 electron configuration
of Tm2+ lies in the fact that 2F5/2 is the only 4f-4f excited state,
and above 9000 cm-1, the properties of the states arising from
the (4f)12(5d)1 configuration can be probed without any interfer-
ence from the (4f)13 configuration. The observation of emissions
from higher excited 4f-5d states is facilitated by the fact that
we are dealing with divalent host lattices with heavy anions. A
practical advantage of the systems studied here is the energy
range of the emissions studies, from the near-IR to the green,
which is experimentally more easily accessible than the vacuum
UV, where the 5d-4f emissions of the trivalent lanthanides are
often found.
1
lowest level of (3H6, ttg)S ) /2 and the highest level of (3H6,
3
ttg)S )
/ is difficult to estimate from the absorption and
2
luminescence spectra: A range of 600-1200 cm-1 is compatible
with the data. This is a very small gap, and in all compounds
studied here, relaxation by multiphonon processes is active down
to the lowest temperatures. In the chloride, multiphonon
relaxation is too efficient for C to be observed. In the bromide,
the intensity of C compared to B is small, whereas in the iodide,
C has gained considerable intensity. The observation of emission
C in the bromide and iodide is surprising considering the small
energy gaps involved. In Ce3+ doped compounds, energy gaps
of up to several thousand cm-1 are observed between the states
deriving from the (4f)0(5d)1 electron configuration.34,35 Gener-
ally, only emission from the lowest state is observed. Mul-
tiphonon relaxation among states of the same spin multiplicity
appears to be efficient. In contrast, the multiphonon relaxation
process from (3H6, ttg)S ) 1/2 to S ) 3/2 requires a reorientation
of the spin, which slows it down. This is considered a key to
the observation of emission from the low-spin state.36 Among
the lanthanides having a more than half-filled 4f shell, “spin-
forbidden” and “spin-allowed” emissions have been reported
Let us finally briefly compare the light-emission properties
reported and analyzed here in chloride, bromide, and iodide
lattices with literature studies of Tm
2+ doped fluorides. All these
studies deal with the 4f-4f emission in the near-IR, and we
found no report of 5d-4f emission in a fluoride, even at
cryogenic temperatures. This may simply be a reflection of a
3
2
very efficient (3H6, ttg)S ) /2 to F5/2 nonradiative relaxation
rate constant W
21. However, there are chemical considerations
that have to be taken into account. Tm2+ is harder to stabilize
in a fluoride than in the heavier halides. As a consequence, Tm3+
for Yb2+ 12 Er3+, and Tm3+ doped compounds.36,37 The ratio
,
of “spin-forbidden” and “spin-allowed” emission was found to
depend on the host and the temperature.
is invariably present in Tm2+
doped fluorides, and this might
quench the emission B of Tm2+ by energy transfer. We have
found some preliminary evidence of emission B in a CaF2:Tm2+
2
Emission A (2F5/2 f F7/2) is the strongest emission at 300
2
K in the chloride and bromide. The emitting state F5/2 is
sample, in which Tm2+
was chemically incorporated and not
populated from (3H6, ttg)S ) 3/2 by the multiphonon relaxation
process with rate constant W21. A takes over from B as the most
intense emission at about 120 and 250 K, respectively; in the
two lattices, see Figure 4 and the crossing points of R2 with
created by irradiation as in some of the other investigations.
Acknowledgment. The authors thank K. W. Kra¨mer and
D. Biner for their assistance concerning crystal synthesis. P.
Gerner and A. Sieber are acknowledged for valuable discussions.
The Swiss National Science Foundation is gratefully acknowl-
edged for financial support.
W21 in Figure 9. In the iodide, this crossing is not reached below
300 K, because W21 is not competitive. It is interesting to note
that a very small A emission of the order of 0.1% is observed
even at 10 K in all three samples; see Table 1. In the bromide,
References and Notes
2
the F5/2 population occurs purely radiatively by process E,
(1) Rubio, J. J. Phys. Chem. Solids 1991, 52, 101.
(2) Meyer, G. Chem. ReV. 1988, 88, 93.
whereas in chloride and iodide there is an additional nonradiative
contribution to the 2F5/2 feeding. In CsCaCl3:Tm2+, the intensity
of emission A decreases above 200 K. This is not due to
multiphonon relaxation to the ground state, for which the energy
gap of 8800 cm-1 is too large. We ascribe it nonradiative losses
by energy migration to unidentified killer traps.
(3) McClure, D. S.; Kiss, Z. J. J. Chem. Phys. 1963, 39, 3251.
(4) Kiss, Z. J. Phys. ReV. 1962, 127, 718.
(5) Loh, E. Phys. ReV. 1968, 175, 533.
(6) Duncan, R. C.; Kiss, Z. J. Appl. Phys. Lett. 1963, 3, 23.
(7) Duncan, R. C. IEEE J. Quantum Electron. 1966, 2, R52.
(8) Schipper, W. J.; Meijerink, A.; Blasse, G. J. Lumin. 1994, 62, 55.
(9) Wenger, O. S.; Wickleder, C.; Kra¨mer, K.; Gu¨del, H. U. J. Lumin.
2001, 94-95, 101.
6. Conclusions and Outlook
(10) Wickleder, C. J. Alloys Compd. 2000, 300-301, 193.
(11) Grimm, J.; Gu¨del, H. U. Chem. Phys. Lett. 2005, 404, 40.
(12) Dorenbos, P. J. Phys.: Condens. Matter 2003, 15, 575.
(13) Beurer, E.; Grimm, J.; Gu¨del, H. U. To be submitted for publication.
(14) Meyer, G. AdV. Synth. React. Solids 1994, 2, 1.
(15) Seifert, H.-J.; Langenbach, U. Z. Anorg. Allg. Chem. 1969, 368,
36.
(16) Seifert, H.-J.; Haberhauer, D. Z. Anorg. Allg. Chem. 1982, 491,
301.
(17) Vaills, Y.; Buzare´, J. Y.; Gibaud, A.; Launay, C. Solid State Comm.
1986, 60, 139.
(18) Lucas, M. C. M.; Rodriguez, F.; Prieto, C.; Verdaguer, M.; Gu¨del,
The temperature-dependent study of the emission properties
of Tm2+ doped into a series of isostructural halide lattices
provides insights into the photophysics of this ion. The interplay
and the competition of the various radiative and nonradiative
relaxation processes are elucidated, and the rate constants of
the relevant processes are quantified. The observation and
characterization of up to five different types of light emission
from a given compound, which is without precedent, provide
the basis for such a detailed analysis. It turns out that the
chemical variation along the series of CsCaCl3, CsCaBr3, and
CsCaCI3 host lattices and its significant consequences on the
light-emission behavior are essential for the relatively detailed
picture obtained. The pronounced temperature dependence of
the nonradiative processes is an important factor, which we are
reproducing with an approximate model in the present paper,
whereas the radiative decay constants are essentially temperature
independent. In the iodide, we reach the limit of the applicability
of the simple model. But it does allow a discussion of the
relevant processes and their temperature dependence.
H. U. J. Phys. Chem. Solids 1995, 56, 995.
(19) Schilling, G.; Meyer, G. Z. Anorg. Allg. Chem. 1996, 622, 759.
(20) Hu¨fner, S. Optical Spectra of Transparent Rare Earth Compounds;
Academic Press: New York, 1978.
(21) Dorenbos, P. J. Phys.: Condens. Matter 2003, 15, 6249.
(22) Richardson, F. S.; Reid, M. F.; Dallara, J. J.; Smith, R. D. J. Chem.
Phys. 1985, 83, 3813.
(23) Struck, C. W.; Fonger, W. H. J. Lumin. 1975, 10, 1.
(24) Donnelly, C. J.; Imbusch, G. F. In AdVances in NonradiatiVe
Processes in Solids; di Bartolo, B., Ed.; NATO ASI Series; Plenum Press:
New York, 1991; pp 175-195.
(25) Henderson, B.; Imbusch, G. F. Optical Spectroscopy of Inorganic
Solids; Oxford University Press: New York, 1989.

Dynamics of Tm2+-Doped Perovskite Compounds
J. Phys. Chem. B, Vol. 110, No. 5, 2006 2101
(26) Dorenbos, P. J. Phys.: Condens. Matter 2003, 15, 2645.
(27) Reber, C.; Gu¨del, H. U. J. Lumin. 1988, 42, 1.
(28) May, P. S.; Gu¨del, H. U. J. Lumin. 1990, 46, 277.
(29) Gamelin, D. R.; Gu¨del, H. U. J. Phys. Chem. B 2000, 104,
10222.
(30) Sturge, M. D. Phys. ReV. B 1973, 8, 6.
(31) Donegan, J. F.; Bergin, J. F.; Glynn, T. J.; Imbusch, G. F. J. Lumin.
1986, 35, 57.
(32) Kasha, M. Discuss. Faraday Soc. 1950, 9, 14.
(33) Peijzel, P. S.; Vergeer, P.; Meijerink, A.; Reid, M. F.; Boatner, L.
A.; Burdick, G. W. Phys. ReV. B 2005, 71, 045116.
(34) Dorenbos, P. Phys. ReV. B 2000, 62, 15640.
(35) Dorenbos, P. Phys. ReV. B 2000, 62, 15650.
(36) Wegh, R. T.; Meijerink, A. Phys. ReV. B 1999, 60, 10820.
(37) van Pieterson, L.; Wegh, R. T.; Meijerink, A.; Reid, M. F. J. Chem.
Phys. 2001, 115, 9382.