Origin of the 2 E ^ 4 T 2 Fano resonance in Cr 3 +-doped LiCaAlF 6 : Pressure-induced excited-state crossover

M. N. Sanz-Ortiz,1 F. Rodríguez,1,* I. Hernández,1 R. Valiente,2 and S. Kück3 1MALTA Consolider Team, DCITIMAC, Facultad de Ciencias, University of Cantabria, 39005 Santander, Spain 2MALTA Consolider Team, Depto. Física Aplicada, Facultad de Ciencias, University of Cantabria, 39005 Santander, Spain 3Physikalisch-Technische Bundesanstalt, Bundesallee 100, 38116 Braunschweig, Germany Received 6 October 2009; revised manuscript received 13 November 2009; published 12 January 2010

The pressure effect is expected to be particularly important in LiCaAlF 6 :Cr 3+ since the host material exhibits a firstorder phase transition ͑PT͒ from the trigonal P3 ¯1c ͑phase I͒ to monoclinic P2 1 / c ͑phase II͒ between 7 and 9 GPa, which reduces Cr 3+ site symmetry. 13In this work we investigate the effects of pressure-induced PT and the 4 T 2 ↔ 2 E ESCO on the PL properties of LiCaAlF 6 :Cr 3+ .One aim is to induce variations in the Cr 3+ PL by reducing the ͑CrF 6 ͒ 3− complex volume through structural changes along the PT and explore whether the associated pressure hysteresis is able to stabilize phase II at ambient conditions.
Although Cr 3+ ESCO was first observed by Dolan et al., 1 in K 2 NaGaF 6 :Cr 3+ at 6.1 GPa and 154 K, later by Freire et al., 23 in KZnF 3 :Cr 3+ and recently by Hernández et al. 9 in Rb 2 KCrF 6 , the PL change toward a complete rubylike emission ͑ 2 E → 4 A 2 ͒ has never been observed at room temperature ͑RT͒ in fluorides.This is due to the high-pressure con- ditions required to place the emitting state 4 T 2 ͑broadband PL͒ far above the 2 E emitting state ͑narrow-line PL͒ in order to get 2 E mostly populated at RT.As we will see in this paper, an energy separation above 0.2 eV ͑⌬E Ϸ 8 k B T͒ would enable the long-lived 2 E → 4 A 2 transition ͑ = 2.7 ms͒ to be the major contribution to the LiCaAlF 6 :Cr 3+ PL spectrum ͑R1 and R2 lines͒, beyond the short-lived 4 T 2 → 4 A 2 transition, = 170 s in phase I at ambient conditions ͑ =70 s in phase II͒.It is worth noting that this energy is the difference between excited-state configuration minima ⌬E min = E min ͑ 4 T 2 ͒ − E min ͑ 2 E͒ ͓Fig.1͑b͔͒ and not the one obtained for the ground-state equilibrium configuration of ͑CrF 6 ͒ −3 given by the Tanabe-Sugano ͑TS͒ diagram ͓Fig.1͑a͔͒.Therefore, ESCO implies crossing of zero-phonon lines 4 T 2 ↔ 2 E in terms of spectroscopic parameters; i.e., the emitting excited state for the 4 A 2 ground-state configuration corresponds to the high spin 4 T 2 although the excited-state energies associated with the 4 T 2 and 2 E are identical ͓Fig.1͑b͔͒.Actually, the first absorption band for this system exhibits Fano resonance reflecting the energy closeness of the 4 T 2 , 2 E, and 2 T 1 excited states.The knowledge of the linear and quadratic electron-phonon coupling for the PL excited states is essential to draw the configurational energy diagrams for predicting ESCO phenomena.Furthermore the origin of the Fano resonance associated with 4 T 2 ↔ 2 E is investigated through optical spectroscopy under pressure, which unraveled very useful to determine the zerophonon line ͑ZPL͒ energy between 4 A 2 ↔ 2 E even if it cannot be detected at ambient conditions. 9We show that pressureinduced ESCO is attained at 6 GPa at RT. However this pressure corresponds to the crossover of the excited-state configuration-energy minima of the 4 T 2 and 2 E states: , is not able to transform the broadband PL to a narrow-line emission at RT since radiative deexcitation mainly takes place from the short-lived 4 T 2 state.
In this work we investigate whether ESCO can be detected by spectroscopy through the 4 T 2 and 2 E ZPL energies and the minimum pressure required to get narrow-line emission at RT.[31][32]

II. EXPERIMENT
The microcrystals used in pressure experiments ͑70ϫ 50 ϫ 15 m 3 ͒ were obtained by cleavage from a LiCaAlF 6 :Cr 3+ ͑2 mol %͒ single-crystal rod.Time-resolved emission and excitation spectra and PL time-dependent measurements were performed in the 0-35 GPa pressure range using the experimental setup described elsewhere. 33,34These spectra were attained exciting the sample with a Vibrant B355II OPO pulsed laser operating in the 410-2400 nm range.The 407 nm line of a Coherent I-302 Kr + laser was used for obtaining emission spectra under cw excitation.The PL intensity was dispersed by a Chromex 500IS monochromator and detected by photon-counting techniques using a SR400 photon counter measuring photocurrent pulses of 500 s after the laser pulse for 50 averages.Time-resolved emission spectra were corrected for instrumental response using a calibrated tungsten lamp.The PL signal for timeresolved excitation spectroscopy was measured at the band maximum varying the laser wavelength in the 410-730 nm range and was normalized to the pulse intensity averaged for 200 pulses.All the spectra were corrected for wavelength dependence instrumental response.The PL decay curves I͑t͒ for lifetime measurements under pressure were obtained through a Tektronix 2430A scope.All the experimental curves I exp ͑t͒ exhibited a single exponential behavior in the explored pressure range and were fitted to a function I exp ͑t͒ = I back + I 0 exp͑−t / ͒ where the background intensity, I back , was directly measured from I exp ͑t͒ for times before the excitation pulse ͑t Ͻ 0͒.
Absorption spectra were obtained with a Perkin-Elmer Lambda 9 spectrophotometer.The low-temperature absorption and emission spectra were obtained with a closed cycle He Scientific Instruments 202 cryostat with an APD-K temperature controller.High-pressure experiments in the 0-35 GPa range were performed in a diamond anvil cell, High Pressure Diamond Optics, Inc. using silicone oil ͑Dow-Corning 200 fluid 300000 cst͒ as pressure transmitter.Pressure was calibrated through the ruby R-line shift.The pressure and hydrostatic conditions within the cavity at high pressure were explored by three different ruby chips nearby the sample.In addition, we have checked in situ the LiCaAlF 6 :Cr 3+ through its R-line position and broadening, respectively.Up to 35 GPa, hydrostaticity is fairly good with estimated nonhydrostatic pressures smaller than 3%.absorption ͑OA͒ and PL spectra and the associated PL time dependence: 202 and 172 s, respectively.This lifetime decrease with increasing temperature and the increase in the transition-oscillator strength are activated by the odd-parity vibration mechanism characteristic of d-d transitions in centrosymmetric systems such as ͑CrF 6 ͒ 3− in LiCaAlF 6 :Cr 3+ ͑Refs.6 and 35͒.The appearance of the narrow features in the low-temperature absorption and emission spectra states is noteworthy.They are due to electronic transitions between vibronic states and thus provide the vibrational structure of the coupled modes for the excited ͑absorption͒ and the ground ͑emission͒ states. 36An account for the vibronic structure of Cr 3+ electronic spectra is given elsewhere. 9,14,16,19,27he OA spectrum consists of three broadbands, located at 1.99, 2.90, and 4.48 eV that correspond within ͑CrF 6 ͒ 3− to 4 A 2 → 4 T 2 ͑ 2 E and 2 T 1 ͒, 4 A 2 → 4 T 1 ͑a͒, and 4 A 2 → 4 T 1 ͑b͒ crystal-field ͑CF͒ transitions of Cr 3+ , respectively ͓Figs.1͑a͒ and 2͔.The associated Racah and CF parameters obtained by fitting the 300 K OA spectrum ͑Fig.2͒ are; 37 B = 93 meV; C = 390 meV ͑C / B = 4.2͒; and ⌬ = 1.99 eV ͑⌬ / B =21͒ ͓Fig.1͑a͔͒.The corresponding emission spectrum, obtained under excitation in any of the three OA bands, consists of a broadband located at 1.69 eV ͑734 nm͒ at RT and 1.74 eV ͑713 nm͒ at 19 K with the lowest-energy ZPL peaking at 1.833 eV.Within the experimental uncertainty the three absorption bands peak at the same positions at 19 K and RT ͑Table I͒.An account of the thermal dependence of the optical spectra and band shifts is reported elsewhere. 36he triplet structure observed in the first 4 A 2 → 4 T 2 band is noteworthy.A discussion on its origin on the basis of the experimentally determined 4 T 2 and 2 E ZPL energies is given in Sec.0][31][32] The energy of this first absorption band ͑⌬͒ and the Racah parameters indicate that ⌬ / B Ϸ 21, this value corresponding to 4 T 2 ↔ 2 E crossing point for the ground-state ͑ 4 A 2 ͒ equilibrium geometry, according to the TS diagram for Cr 3+ ͓Fig.1͑a͔͒.The emission energy and the ZPL enable us to draw the configurational curves for LiCaAlF 6 :Cr 3+ as depicted in Fig. 1͑b͒.The ESCO would take place once the corresponding energy minima of 4 T 2 and 2 E parabola would cross, i.e., when their ZPL energy equals: Following the TS diagram of Fig. 1͑a͒, the absorption transitions correspond to vertical lines from 4 A 2 to excited states 4 T i ͑i =2,1a,1b͒ at the 4 A 2 equilibrium geometry ͑i.e., ⌬ / B Ϸ 21͒.However, emission takes place from the first excited-state relaxed geometry that, according to the emission energy ͑E emis = 1.69 eV͒, corresponds to ⌬ / B Ϸ 18 what shows that 4 T 2 is the emitting state at ambient conditions ͑Figs. 1 and 2͒.An eventual ESCO transition would require reducing the energy difference between 2 E and 4 T 2 at least by

B. Pressure effects on the excitation and emission spectra, and photoluminescence lifetime. Influence of the pressureinduced trigonal-to-monoclinic structural phase transition
Figure 1͑b͒ depicts the configurational curves of the two emitting excited states as a function of pressure above, below and at the ESCO where both parabola minima ͑and their ZPL͒ coincide.Given that the PL transition rate from 4 T 2 is faster than from 2 E ͑ 2 E → 4 A 2 spin forbidden͒, 2,9 an eventual narrow-line emission is expected at pressure conditions ͑or corresponding CF strength͒ well above the ESCO.
Figures 3 and 4 show the LiCaAlF 6 :Cr 3+ emission spectra and several spectroscopic parameters in the 0-13 GPa pressure range, respectively.From 6 GPa both broadband ͑ 4 T 2 ͒ and narrow-line ͑ 2 E͒ emissions coexist with identical radiative lifetimes due to thermalization processes.Some anomalies in the PL behavior of Cr 3+ are observed at 7.0Ϯ 0.5 GPa caused by the structural PT of LiCaAlF 6 , which was detected by x-ray diffraction around 8 GPa. 13 Pressure-induced PT has important consequences on Cr 3+ PL in LiCaAlF 6 :Cr 3+ .First, the PL band experiences an abrupt redshift ͑−0.17 eV͒ upon pressure ͓Fig.4͑a͔͒ from phase I to phase II producing an increase in the critical pressure needed for the ESCO to take place ͑Figs.3 and 4͒. 13 Second, the narrow features associated with 2 E → 4 A 2 emission, peaking at 1.872 eV at 6 GPa, completely disappear above the PT pressure and reappearing for P Ͼ 10 GPa ͑Fig.3͒.Third, the low-symmetry Cr 3+ site in phase II reduces the PL lifetime from 172 s in phase I to 72 s in phase II ͓Fig.4͑c͔͒ at ambient conditions due to the oscillatorstrength increase by the low-symmetry noncentrosymmetric CF.These results prove that Cr 3+ PL is an efficient probe to detect pressure-induced PT in the host crystal.Both lifetime reduction and PL redshift can render attractive properties to the material as laser medium since PT enhances the absorption coefficient, whereas the emission broadening ͑0.1 eV͒ extends the laser tunability.
An unexpected result is the lifetime reduction from 0 to 7 GPa.In fact, due to the spin-forbidden character of  3+ narrow-line emission in K 2 NaGaF 6 :Cr 3+ from about 6 GPa at low temperature 1,9,20,22 and experimentally demonstrates that a rubylike PL can be obtained in Cr 3+ -doped fluorides at RT but applying much higher pressure.
The initial broadband emission shifts to higher energy with pressure according to the increase in 4 T 2 energy with ⌬ with a pressure rate of 11 meV/GPa ͓Figs.1͑a͒ and 5͑a͔͒.Besides broadband emission, the Cr 3+ R1-R2 lines appear at about 1.88 eV ͑662 nm͒ above 8 GPa for LiCaAlF 6 in phase II as is shown in Figs.5͑b͒ and 5͑c͒.Their intensity increases with pressure at expenses of the 4  These lines are separated by ⌬E = 7 meV ͑⌬ = 2.5 nm͒, which is about twice the difference between ruby R lines: ⌬E = 3.6 meV ͑⌬ = 1.5 nm͒. 38It implies that the lowenergy R1 line at 1.876 eV is four times more intense than the R2 line at 1.883 eV in agreement with their thermal populations at RT.Although the structural PT does not affect significantly the ZPL energy, it however has a strong influence on the ZPL intensity with respect to the broadband 4 T 2 → 4 A 2 intensity.In fact, the 2 E → 4 A 2 splitting into R1 and R2 is as in ruby a direct consequence of the trigonal CF and the spin-orbit interaction.According to x-ray diffraction data in LiCaAlF 6 , 13 the trigonal distortion of the ͑AlF 6 ͒ 3− octahedron in phase I is weaker than in phase II.The small R1-R2 splitting in the emission spectrum at 6 GPa in comparison to the splitting at 13 GPa confirms this feature.
The presence of R lines in the emission spectrum of LiCaAlF 6 :Cr 3+ enables an in situ pressure determination in the host crystal from the pressure shift rate of 0.146 nm/GPa.FIG. 6. ͑Color online͒ ͑a͒ Variation in LiCaAlF 6 :Cr 3+ ͑2%͒ PL spectrum in 0-35 GPa pressure range at 300 K, including the corresponding variations in the emission energy for the broadband 4 T 2 → 4 A 2 and R lines 2 E → 4 A 2 .The solid lines represent least-square linear fits to the experimental points in phase I and II, and for R lines ͑see Eqs.͑2͒͒.͑b͒ Luminescence lifetime of LiCaAlF 6 :Cr 3+ ͑2%͒ in phase II in the 0-35 GPa range at 300 K.The ͑P͒ lines represent the calculated lifetime on the basis of two emitting excited states: 4 T 2 ͑ =72 s͒ and 2 E ͑ = 2.7 ms͒ according to ͑CrF 6 ͒ 3− with O h symmetry given by Eq. ͑1͒ ͑solid line͒, and C 4 symmetry by Eq. ͑3͒ ͑dashed line͒.The time-dependence emission decay at four different pressures is given as an inset.
By contrast the ruby R-shift rate is 0.365 nm/GPa making it more sensitive to pressure changes. 38,39In spite of the smaller sensitivity of LiCaAlF 6 :Cr 3+ R lines, their energy ͑or wavelength͒ dependence E͑P͒ follows a linear behavior in a wider pressure range than ruby thus extending at least up to 35 GPa.This allows us a precise determination of 2 E energy at ambient conditions by extrapolating high-pressure data, what is crucial on discussing the Fano-type resonances in the first absorption band. 40

D. Lifetime measurements
The PL evolution from broadband to narrow-line emission is accompanied by a progressive increase in the lifetime ͑P͒ from 72 s at ambient pressure to 2.7 ms at 35 GPa for LiCaAlF 6 :Cr 3+ in phase II ͓Fig.6͑b͔͒ as the thermal population of 4 T 2 state decreases at expenses of 2 E. This effect is confirmed by the spectroscopically detected 2 E and 4 T 2 ZPLs, the energy separation of which ͑⌬E min = ⌬ ZPL ͒ increases continuously with pressure ͓Fig.6͑a͔͒.The solid line in Fig. 6͑b͒ corresponds to the lifetime fit to the equation where w 1 ͑= 1 −1 ͒ and w 2 ͑= 2 −1 ͒ are, respectively, 4 T 2 → 4 A 2 and 2 E → 4 A 2 transition probabilities, being 12 and 4 their associated total degeneracy in a local O h symmetry ͑CrF 6 ͒ 3− .⌬ ZPL represents the energy difference between both state minima.
Although the fitting quality is appealing, a more realistic description of ͑P͒ must however consider the actual structure of ͑CrF 6 ͒ 3− in the monoclinic P2 1 / c phase II.Particularly, the loss of inversion center and the low-symmetry CF at the Al 3+ site, 13 make the O h model unrealistic.The CF distortion in phase II is mostly responsible for the lifetime decrease and the emission redshift at the PT ͓Fig.4͑c͔͒, and for their respective pressure dependences.The doted line in Fig. 6͑b͒ depicts ͑P͒ variations with model calculations according to the proposed scenario ͑cf.Sec.III E͒.

E. Phase-transition effects in the photoluminescence properties of LiCaAlF 6 :Cr 3+
Figure 3͑b͒ depicts the 4 T 2 state in O h ͑phase I͒ and C 4 ͑phase II͒, and Fig. 6 shows E͑P͒ and ͑P͒ variations according to models.The abrupt emission redshift, ⌬ PT = −0.17eV, at PT is associated with the 4 T 2 ͑phase I͒ splitting into 4 E + 4 A 2 ͑phase II͒.This effect can be hardly detected in excitation ͓Fig.7͑a͔͒ since the energy of the first excitation band in C 4 actually corresponds to the centroid of the two 4 E and 4 A 2 bands.According to Fig. 3͑b͒ scenario, the 4 T 2 ͑O h ͒ splitting in phase II would be ⌬ PT ͑ 4 T 2 ͒ =3⌬ PT / 2 = −0.23 eV.
The PT-induced emission redshift has important consequences in the PL properties of LiCaAlF 6 :Cr 3+ .In the following sections we will discuss the PT effects on: ͑1͒ the ESCO associated pressure, ͑2͒ the lifetime, and ͑3͒ the Stokes shift.

Phase-transition effects on Cr 3+ spectra
The ESCO pressure ͑P ESCO ͒ stabilizing 2 E as the lowest excited state increases due to the 4 T 2 → 4 A 2 emission redshift in phase II.The ESCO energy, defined as   This allows us to observe 2 E in the emission spectrum at 10 GPa though its relative intensity is only 0.2%.

Phase-transition effects on the photoluminescence lifetime
The emission lifetime ͑P͒ changes in phase II as consequence of the 4 T 2 splitting and the corresponding transition rate due to pressure-induced PT.According to energy levels of Fig. 3͑b͒, ͑P͒ in phase II is given by where ⌬ PT = 2 3 ⌬ PT ͑ 4 T 2 ͒ is the energy difference between 4 T 2-I and 4 A 2 ͑ 4 T 2-II ͒ states, ⌬ PT ͑ 4 T 2 ͒ being the 4 T 2 splitting in Phase II ͓Fig.3͑b͔͒, and factors 4 and 8 refer to 4 A 2 and 4 E state degeneracy, respectively.⌬ ZPL = 0.136− ␣P is the energy difference between 2 E and 4 A 2 ͑ 4 T 2-II ͒ ZPLs and the fit parameter ␣ represents the difference between pressure derivatives of E ZPL ͑P͒ for both states.Assuming that the emission band maximum of 4 A 2 ͑ 4 T 2-II ͒ and its ZPL have the same pressure dependence then ␣ is similar to ␣ = 0.011 eV/ GPa.Note that Eq. ͑3͒ reduces to Eq. ͑1͒ in Phase I where ⌬ PT ͑ 4 T 2 ͒ = 0.The dashed line of Fig. 6͑b͒ shows the lifetime data fit to Eq. ͑3͒ with: ⌬ PT ͑ 4 T 2 ͒ = 0.23 eV; w 1 =12 ϫ 10 3 s −1 and w 2 = 0.38ϫ 10 3 s −1 ; ␣ = 0.0095͑10͒ eV/ GPa, ␣ being similar to that derived spectroscopically.
Interestingly, the pressure required to get mostly R-line PL is 28 GPa ͑Figs.5 and 6͒.At this pressure the 4 T 2-II state is 0.17 eV above 2 E and the ESCO around 13 GPa has been clearly surpassed.Such a strong pressure requirement for a major 2 E PL is due to the long-lived 2 E state ͑w 2 −1 = 2 = 2.6 ms͒ compared to the short lived 4 II.Room-temperature excitation and zero-phonon line energies corresponding to the first absorption band of LiCaAlF 6 :Cr 3+ as a function of pressure.The associated Stokes shift, defined as E SS = E exc -E emis , are given below.=80 s͒.Although 4 T 2-II is less populated than 2 E at 15 GPa, it emits faster yielding a rapid depopulation of 2 E into 4 T 2 in order to preserve the Boltzmann population between the two emitting states.The pressure required to get simply R-line emission in the PL spectrum must be higher than 25 GPa according to data of Fig. 6͑a͒; or equivalently the energy separation between their ZPL should be higher than ⌬ ZPL Ϸ 0.15 eV.

Phase-transition effects on the Stokes shift
The PL Stokes shift, E SS , defined as the difference between emission and excitation energies: E SS = E exc -E emis , does not practically change with pressure.Although recent results on Cr 3+ in fluoroelpasolites have shown that E SS undergoes a slight increase with pressure, 9 this work establishes that E SS in LiCaAlF 6 :Cr 3+ remains constant with P as is shown in Fig. 7͑b͒ and Table II.E SS increases from 0.41 to 0.55 eV when LiCaAlF 6 passes from phase I to phase II, pointing out that a LiCaAlF 6 :Cr 3+ four-level laser would be improved if LiCaAlF 6 phase II was used instead of phase I.

F. Origin of the Fano resonance in the
There is some controversy in elucidating the origin of the triplet structure in the 4 A 2 → 4 T 2 + 2 E + 2 T 1 first absorption band as due to Fano resonance or antiresonance. 9It is well known that this triplet structure is due to orbital mixing between the broad 4 T 2 and narrow 2 E + 2 T 1 excited states by spin-orbit interaction.This effect is maximum in those systems showing ESCO ͑⌬ / B Ϸ 21͒ like in LiCaAlF 6 :Cr 3+ .Whether the interaction between 2 E ͑or 2 T 1 ͒ and 4 T 2 through spin-orbit interaction gives rise to a Lorentzian, dispersionlike or inverted Lorentzian in the absorption spectrum, we are dealing with Fano resonance, dispersionlike or Fano antiresonance ͑inset of Fig. 9͒, respectively. 31The pressure results shown through Figs.3-8 are worthy in unraveling the origin of the Fano interaction ͑Fig.9͒.Previous reports, mainly based on band-shape analysis, ascribed such resonant features to Fano antiresonances, [29][30][31][32] as the spectral structure around 2 E + 2 T 1 corresponds to a dip in the emission band ͑Fano antiresonance͒ rather than to a hump ͑Fano resonance͒ or dispersionlike features.According to recent findings, 9,[29][30][31][32]40,41 the resonance origin can be clarified when 2 E energy is exactly known and pressure spectroscopy enables us to obtain the 2 E energy at ambient conditions: E ZPL ͑ 2 E͒ = 1.876 eV ͑R1͒ and 1.883 eV ͑R2͒ at P =0 ͓Eq. ͑2͔͒.This R2͑ 2 E͒ energy ͑marked for comparison purposes in the first absorption band of Fig. 9͒ coincides with the first resonance maximum, clearly revealing that first 2 E ↔ 4 T 2 structure in LiCaAlF 6 :Cr 3+ corresponds to a Fano resonance.Therefore 2 E energy can be directly obtained as the first maximum of the absorption fine structure.However a similar analysis cannot be made for 2 T 1 since the corresponding line is missed in the absorption and emission spectra.However band-shape analysis suggests that 2 T 1 and 4 T 2 interact as a Fano antiresonance as is shown by a dip in Fig. 9 fit, finding E ZPL ͑ 2 T 1 ͒ = 2.030 eV.The present analysis can be considered fairly good if we take into account that only a single Gaussian was used as an effective "density of states" for 4 T 2 and two Fano profiles for 2 E and 2 T 1 , the position of the former was determined from pressure data.The lack of a perfect match between simulated and experimental spectra in Fig. 9 is due to the presence of some vibronic features in the experimental absorption band that are not taken explicitly into account in our simplified model.It must be noted that the energy difference between 2 E and 2 T 1 , ⌬E = 1.883-2.030=−0.147 eV, is higher than the estimated from the TS diagram: ⌬E Ϸ B = −0.0914 The Fano resonance weight in the first absorption band drastically diminishes as 2 E and 4 T 2 become far apart at higher pressure, their excitation energies being 1.872 and 2.12 eV at 10 GPa, respectively ͑Fig.8 and Table II͒.Moreover, 2 E excitation peak is much weaker than 4 T 2 band thus confirming that there is an important decoupling between the two states in the excitation spectrum at 10 GPa.
At 28 GPa Cr 3+ in LiCaAlF 6 :Cr 3+ emits only R lines at 1.864 and 1.872 eV ͑ 2 E͒.This situation is rather similar to ruby at ambient conditions where the 4 A 2 → 2 E spin-flip transition is detected as a tiny feature at 1.786 eV mounted on the low-energy tail of the intense 4 A 2 → 4 T 2 absorption band peaking at about 2.25 eV. 38As a result, ruby PL consists of two R lines 2 E → 4 A 2 separated 3.6 meV in contrasts to LiCaAlF 6 :Cr 3+ at 10 GPa, the PL of which consists of both broadband 4 T 2 → 4 A 2 and R lines 2 E → 4 A 2 simultaneously.Although emission from 4 T 2 is missed at 28 GPa, the 4 A 2 → 4 T 2 excitation band is located at 2.29 eV, providing an energy difference between 2 E and 4 T 2 of 0.43 eV, which is similar to 0.46 eV encountered for ruby at ambient conditions.The CF strength in ruby at ambient pressure is attained in LiCaAlF 6 :Cr 3+ at 30 GPa ͓Fig.7; Eq. ͑2͔͒.These results confirm the role of the CF strength governing PL properties of Cr 3+ irrespective of whether the required CF conditions are produced either by strongly -bonding ligands ͑nephelauxetic series F → O → N͒ or under high-pressure conditions.

G. Crystal-field dependence on the crystal volume and Cr-F distance
According to the TS diagram ͓Fig.1͑a͔͒, the CF splitting ͑⌬͒ is directly obtained from the band maximum of the timeresolved excitation spectra ͑Fig.7͒.Through x-ray diffraction data of LiCaAlF 6 in the 0-12 GPa range 13 and the corresponding equation of state ͑EOS͒ we derived the CF strength dependence on the crystal volume ⌬͑V͒ whose data are represented in Fig. 10.The variation has been fit to a semiempirical CF model through the equation: . We find an exponent n = 2.3 ͑0.1͒.3][44] This behavior is quite general on comparing the variation of ⌬ with the crystal volume V as indicated elsewhere. 45,46The measured exponents, named nK in Refs.45 and 46, are usually smaller than 5 ͑K Ͻ 1͒.However, localized systems such as transition-metal ions and rare earths in fluorides and oxides must be better compared with the local variation in the complex volume rather than with the crystal volume.
The deviation of n ͑or nK͒ from the calculated value is probably due to lattice relaxation around Cr 3+ making the compressibility of the ͑CrF 6 ͒ 3− octahedron different from the LiCaAlF 6 bulk compressibility and thus causes a different ⌬ variation with V 1/3 and R Cr-F .Recent PL studies on Cr 3+ along fluoroelpasolite series have also shown an anomalous variation in ⌬͑R Cr-F ͒ ϰ R Cr-F −n with n = 3.3 ͑Ref.9͒.This unusual dependence was attributed to Fano resonance effects.Nevertheless such effects are minimized in LiCaAlF 6 :Cr 3+ at high pressure since Fano resonance strongly diminishes for P Ͼ 10 GPa ͑Figs.7 and 8͒.In fact, n = 3.8͑0.3͒ is obtained when plotting ⌬ vs R Cr-F instead of ⌬͑V 1/3 ͒ ͑Fig.10͒.R Cr-F was obtained from the ͑CrF 6 ͒ 3− local EOS by fitting R Cr-F ͑P͒ data for K 2 NaGaF 6 :Cr 3+ , 21  35͒ eventually modulates the actual Cr-F bond length around its equilibrium value in LiCaAlF 6 :Cr 3+ at ambient conditions: R Cr-F ͑0͒ = 1.85 Å.This procedure provides a fairly agreement between experimental ͑n = 3.8͒ and calculated ͑n = 4.3͒ exponent, 43,44 thus reinforcing that ⌬ mainly scales with the local Cr-F distance rather than with V 1/3 .

IV. CONCLUSIONS
A relevant conclusion of this study is the high pressure required ͑28 GPa͒ to transform LiCaAlF 6 :Cr 3+ broadband PL to a rubylike R-line emission even though ESCO was already attained at 6 GPa.This behavior is due to the different transition rates of 4 T 2 and 2 E states with ͑ 2 E͒ / ͑ 4 T 2 ͒ = 30.Narrow line emission from 2 E is the dominant feature of the room-temperature PL spectrum from 28 GPa.This pressure is needed to increase the 2 E-4 T 2 energy separation by E ZPL ͑ 4 T 2 ͒ − E ZPL ͑ 2 E͒ Ͼ 0.2 eV in order to decrease the 4 T 2 emitting state population from 2 E and compensate effects due to 4 T 2 faster transition rate.In contrast, this narrow-line emission is observed just above 6 GPa at low temperature.
The high-pressure Phase II of LiCaAlF 6 can be stabilized at ambient conditions after pressure release, improving its laser applications due to enhancement of transition oscillator strength, band broadening, and Stokes shift increase.Phase I is recovered at ambient pressure upon heating LiCaAlF 6 phase II above room temperature.
The evolution of the excitation and emission spectra, and the corresponding lifetime are interpreted in terms of pressure-induced ESCO between 4 T 2 and 2 E. Both lifetime and PL shift with pressure are explained on the basis of the electron-phonon coupling in 4 T 2 and 2 E states.
The origin of the 4 T 2 + 2 E + 2 T 1 resonance was clarified through the knowledge of 2 E energy from optical spectroscopy at high pressure.We conclude that 4 T 2 + 2 E interact as a Fano resonance whereas 4 T 2 and 2 T 1 as a Fano antiresonance.This noteworthy result allows us to derive 2 E and 2 T 1 energies directly from the absorption ͑or excitation͒ spectrum.

Figure 2
Figure2shows ambient pressure 19 and 300 K optical

FIG. 2 .
FIG.2.͑Color online͒ Optical absorption and corresponding emission ͑excitation wavelength, exc = 633 nm͒ spectra of LiCaAlF 6 :Cr 3+ ͑2%͒ single crystal at 19 and 300 K ͑phase I͒.The time-dependence emission decay at each temperature with the corresponding fitted lifetime is given as an inset.Note the presence of Fano resonance in the first 4 A 2 → 4 T 2 absorption band around 1.99 eV.

FIG. 4 .
FIG.4.͑Color online͒ Variation in the ͑a͒ PL emission energy, ͑b͒ bandwidth, ͑c͒ and lifetime with pressure at 300 K around the Phase I → Phase II phase transition at P PT = 7.0Ϯ 0.5 GPa in upstroke for ͑a͒ and ͑b͒ and in upstroke and downstroke for ͑c͒.

FIG. 7 .
FIG.7.͑Color online͒ ͑a͒ Pressure dependence of the time-resolved excitation spectrum of LiCaAlF 6 :Cr 3+ ͑2%͒.Each spectrum was obtained by measuring the PL intensity at the emission maximum as a function of the excitation wavelength using photon-counting techniques.Emitted photons were counted for a three-lifetime interval after 50 ns pulsed excitation for each excitation wavelength.The excitation spectrum was corrected for instrumental response and excitation intensity.͑b͒ Pressure dependence of the excitation and corresponding emission energies associated with the 4 A 2 ↔ 4 T 2 transition.The arrows indicate the magnitude of the Stokes shift.
79 eV͔ lies approximately 0.1 eV below 2 E minimum ͓E ZPL ͑ 2 E͒ = 1.88 eV͔ indicating that the ESCO has not yet taken place and the PL is fully governed by the 4 T 2 first excited state yielding broadband 4 T 2 → 4 A 2 emission.

TABLE I .
LiCaAlF 6 :Cr 3+ emission and absorption energies and Racah and CF parameters at 19 and 297 K. responsible for the initial decrease in ͑P͒ observed in Fig. 4͑c͒.Such effects also influence the energy and bandwidth, yielding PL broadening and lifetime reduction from 172 s ͑phase I͒ to 72 s ͑phase II͒ at ambient conditions ͑hysteresis͒.Interestingly, we observe that the high-pressure phase II remains stable upon pressure release at ambient conditions, thus revealing its first-order character with a hysteresis of at least 6 GPa at RT. Cr 3+ emission spectra in the 0-35 GPa pressure range.The variation illustrates how Cr 3+ broadband emission ͑ 4 T 2 → 4 A 2 ͒ progressively transforms into a two narrow-lines ͑R1 and R2͒ rubylike emission ͑spin-flip 2 E → 4 A 2 transition͒ that becomes the dominant PL feature above 28 GPa.This result contrasts with previous measurements reporting Cr 2 E → 4 A 2 , ͑P͒ should increase with pressure as 4 T 2 approaches 2 E increasing its thermal population.Actually ͑P͒ increases an order of magnitude from 7 GPa ͑ = 100 s͒ to 35 GPa ͑ = 2.7 ms͒ beyond ESCO ͓Fig.6͑b͔͒.Hence pre-transitional effects are 4IG.3.͑Coloronline͒͑a͒ Variation in LiCaAlF 6 :Cr 3+ ͑2%͒ PL spectrum in the 0-13 GPa range at 300 K.Each spectrum was obtained by time-resolved spectroscopy counting emitted photons during a three-lifetime interval after 50 ns pulsed excitation into the4A 2 → 4 T 1 ͑a͒ band using exc = 412 nm.͑b͒ Energy-level diagram of the 4 T 2 , 2 E, and 4 A 2 states of Cr 3+ in LiCaAlF 6 Phase I ͑nearly O h CrF 6 3− ͒ and Phase II ͑nearly C 4 CrF 6 3− ͒.Note that the Phase I 4 T 2 state splits into 4 E and 4 A 2 by ⌬ PT ͑ 4 T 2 ͒ in Phase II.Their energy separation to 4 T 2 state center of gravity is increases from 0.086 eV in phase I to 0.136 eV in phase II ͑Fig.8 and Table II͒.From Figs. 6-8, we obtained that the pressure variations in ZPL energy of 4 T 2 and 2 E for phases I and II are given by 44 + 0.011P͑Phase II͒ ϫ͑E in eV and P in GPa͒.͑2͒Assuming that the pressure dependence of4T 2 ZPL energy is the same as for the emission band maximum, we estimate from Eq. ͑2͒ and Table II that ESCO in LiCaAlF 6 :Cr 3+ takes place at a critical pressure of 7.5 GPa in phase I and 12.4 GPa in phase II.This increase in P ESCO is responsible for the disappearance of the 2 E lines in the emission spectrum at 8 GPa after the PT pressure ͓Fig.3͑a͔͒.This result is confirmed by the excitation and emission spectra at 10 GPa shown in Figs.6͑a͒ and 8.The ZPL for 4 T 2-II and 2 E at 10 GPa and RT can be obtained from Fig. 8 and have values of 1.83 and 1.872 eV, respectively.This means that 4 T 2-II is still the lowlying excited state at 10 GPa and not 2 E. From the 4 T 2 ZPL energy at 10 GPa ͑Fig.8͒, we can now improve our previous estimates on the ESCO pressure thus obtaining P ESCO = 13.8GPa for LiCaAlF 6 :Cr 3+ in phase II.It is worth noting that the relative emission intensity associated with 2 E and 4 T 2 is proportional to their Boltzmann population and respective transition probability.The observation of weak R-line emission below the ESCO pressure, in spite of being 4 T 2 mostly populated, is due to the narrow peak width of 2 E.