1. Introduction

Despite the considerable amount of work on laser emission from Er-doped crystals, the laser characteristics of the 1.5 μm ⁴I_13/2 →⁴I_15/2 ground state transition are rather poor. The intrinsic difficulty to obtain laser action on this transition is mainly related to its three-level characteristic, but also other effects have to be taken into account like for example the poor feeding efficiency when pumping into the ⁴I_11/2 level, and the upconversion process that depopulates the ⁴I_13/2 multiplet.

Fluoride hosts are characterized by lower phonon energy than oxides. That increases the probability of radiative infrared transitions. Among fluorides, BaY₂F₈ (BaYF) is a very attractive material thanks to its good thermomechanical properties. In fact, it has a thermal conductivity (~ 0.06 W/cmC [1]) comparable to that of YVO₄ (0.05 W/cmC [2]), moreover its non-cubic structure can compensate the thermal lens under strong pumping as it happens in LiYF₄, rather then in cubic crystals such as YAG [3]. In particular, BaYF appears interesting for the development of a 1.5 μm Er laser because it provides a larger Stark splitting for rare earth energy levels than other fluorides do, allowing good efficiency even for three-level lasers [4].

The Er doping of BaYF has already been investigated as for efficient laser emission at 3 μm [5] or in the green-blue region [6]. To the best of our knowledge, 1.5 μm laser emission in BaYF has been reported only once in [6], where the authors incidentally obtained IR emission while studying the dependence of the blue laser on the pump wavelength. No further efforts have been made up to now to characterize the 1.5 μm laser in this crystal.

Because of the fact that an efficient 1.5 μm Er₃₊ laser oscillation has been demonstrated only in some Erbium-doped glasses so far [7]–[10], to improve the efficiency of population inversion at 1.5 μm, codoping with Cerium has been investigated in some vitreous [11] [12] and crystalline materials [13]–[16] and it has even proved to increase the amplification efficiency in some Er-doped fiber amplifiers[17] [18].

The aim of the present work is to study the effects of Erbium-Cerium codoping in BaYF fluoride host. To this aim, we grew several Er-Ce:BaYF crystals, doped with the same nominal amount of Erbium and different Cerium concentrations and we characterized them as for the Er 1.5 μm emission.

2. Crystal structure and Er-Ce energy transfer in fluorides

BaYF has monoclinic structure and the rare earth dopants enter substitutionally the Y₃₊ sites [19]. It is isostructural to BaEr₂F₈, so that it can be doped up to 100% Er while maintaining the same structure. On the other hand Ce₃₊ ion has an ionic radius larger than Y₃₊, therefore the Ce doping introduces a considerable stress in the crystalline matrix. Moreover, unlike Er, Cerium does not have a compound analogous to BaEr₂F₈. For both reasons, an high Ce doping level in BaYF is not possible.

Due to the complex nature of the Erbium energy level scheme, many processes have to be taken into account when investigating its laser characteristics (see for example [20]). A possible approach to help the population inversion is to deactivate the detrimental level(s) through energy transfer to another rare earth. This approach has successfully been tried in the case of Er 3 μm laser, using Pr as deactivator [20], that effectively depopulates the ⁴I_13/2 multiplet and leaves the ⁴I_11/2 lifetime unchanged.

Codoping with Ce has been investigated in the present work in order to selectively quench the ⁴I_11/2 lifetime instead. Ce has only one electron in the 4f orbital, therefore it has only one excited energy level in the 4f configuration, with an energy about 3000 cm^-1 higher than the ground state. Therefore it is quickly depopulated by strong multiphonon decay at room temperature. The next excited state of Ce is a 5d configuration state. In fluorides, the transition to the ground state is situated in the UV region of the spectrum. Moreover this transition is electric-dipole allowed, leading to lifetimes as short as few nanoseconds.

 

Fig. 1. Schematics of the Er-Ce energy transfer processes in BaYF.

Download Full Size | PDF

Before giving the processes involved in the case of Er-Ce codoped fluoride systems in some detail, it is first important to notice that, since Ce does not absorb the radiation used to pump Er (except UV radiation), it will not adversely affect the pumping efficiency under diode pumped operation. The processes involved in energy transfer in Er-Ce codoped systems, at low (1%) Er concentrations, are shown in Fig. 1. In agreement with the literature, Ce is assumed to primarily increase the ⁴I_11/2 →⁴I_13/2 branching ratio, according to the following path:

As the energy difference between the two Er levels is about 3700 cm^-1, whereas the energy difference between the two lower levels of Ce is about 3000 cm^-1, it is reasonable that the process involves one Ce ion and two lattice phonons.

The ⁴S_3/2 level in fluorides is populated mainly by upconversion from the long-living ⁴I_11/2. Therefore, the depopulation of this latter level also reduces the amount of pump energy that is lost via upconversion. Moreover, by looking at the energy gap between ⁴S_3/2 and ⁴F_9/2 (about 3000 cm^-1) we could also infer a further direct quenching of ⁴S_3/2 by Cerium, according to the scheme:

In addition, the energy gap between the ⁴F_9/2 and ⁴I_9/2 multiplets is also consistent with the Ce upper 4f to ground state energy difference, and therefore it seems reasonable to expect also a ⁴F_9/2 quenching by Cerium as follows:

As a result Ce codoping can give rise to a series of cascade quenching processes which accumulates the energy on the ⁴I_13/2 1.5 μm-emitting level. The cascade quenching stops on the ⁴I_13/2, because the energy gap from this level to the ground state (about 6600 cm^-1) is much higher than the energy separation between the two levels of Ce.

3. Experimental

BaYF single crystals, doped with 1% Er concentration and with different amounts of Ce have been grown by the Czochralski method from oriented seeds. The growth parameters were: temperature around 970 °C, pull rate 0.5 mm/h, rotation rate 10 RPM. We grew a total of five samples, with Ce concentrations of 0%, 0.2%, 0.5%, 2% and 4% in the melt. As the ionic radius of Ce is larger than that of Y, the optical quality of the samples tends to degrade with Ce concentration. For this reason the maximum investigated doping level is 4%Ce,1%Er:BaYF which was still of reasonably good optical quality.

All the crystals were oriented by X-ray backscattering Laue diffractometry and cut with edges parallel to the b and c crystallographic axes in order to investigate polarization-dependent optical properties.

The polarized absorption spectra were performed at room temperature, with a Cary 500 spectrophotometer.

We recorded the emission spectra around 500 nm (⁴S_3/2 → ⁴I_15/2 transition), 1.5 μm (⁴I_13/2 → ⁴I_15/2 transition) and 3 μm (⁴I_11/2 → ⁴I_13/2 transition). The excitation source was a 150-mW diode laser tuned at 968 nm. The luminescence was observed perpendicularly to the exciting beam to avoid pump spurious scattering. It was mechanically chopped and focused by a lens of 75 mm focal length on the input slit of a 25-cm monochromator equipped with a 1200 gr/mm, a 600 gr/mm or a 300 gr/mm grating for the different investigated wavelength regions and filtered by a Si or by a Ge window in the case of 1.5 μm or 3 μm measurements, respectively. The signal was detected by a S20 photomultiplier for the green region or by a cooled InSb detector in the infrared and processed by a lock-in amplifier. The resolution of the fluorescence spectra was 4.5 nm in the visible region, 3 nm at 1.5 μm and 24 nm at 3 μm.

In order to evaluate the 1.5 μm Er emission cross sections we acquired the polarized emission spectra at room temperature on the singly-doped sample. In this case the excitation source was a diode array delivering up to 2 W at 968 nm. The fluorescence was dispersed by a 32-cm monochromator, equipped with a 600 gr/mm grating. Also in this case the detector was a cooled InSb photodiode. The resolution of the polarized 1.5 μm emission spectra was 1 nm.

For the ⁴I_11/2 and ⁴I_13/2 multiplets lifetime measurements we used a pulsed tunable Ti:Al₂O₃ laser with 10 Hz repetition rate and 20 ns pulse width as a source. The laser was tuned to the peak of the ⁴I_15/2 → ⁴I_9/2 absorption band, at 794 nm. To better characterize the Er-Ce energy transfer dynamics, we also measured the green-emitting ⁴S_3/2 level lifetime. In this case the excitation source was the pulsed Ti:Al₂O₃, doubled by a BBO crystal. The excitation wavelength was 407 nm. In all lifetime measurements, the pulse energy was reduced as much as possible in order to minimize power-dependent effects. The fluorescence was collected from a short portion of the sample to observe a uniformly pumped volume. To reduce the spurious decaytime lengthening produced by radiation trapping the laser was focussed at the edge where the observation was made. Detection of the laser scattering was avoided by means of suitable filters. The signal was detected by a cooled InSb detector for the infrared cases and by a S20 photomutiplier for the visible region and was sent, after amplification, to a digital oscilloscope connected to a computer. The response time of the system was about 1 μsec.

4. Spectroscopy

Preliminary unpolarized absorption measurements have been performed for every crystal from 187 nm up to 3 μm, over the whole operating range of the spectrophotometer, in order to check the presence of unwanted contaminants. The recorded spectra allow us to infer that the crystals are free of OH⁻ radicals or other impurities within the sensitivity of our instrument.

In Fig. 2 the Cerium absorption spectrum, of the sample doped with 1% Er and 0.2% Ce, is shown. Because of the very high Ce absorption, the measurements have been performed on samples less than 1 mm thick. The band around 200 nm is relative to the transition ²F_5/2 → ²D_5/2 (the third peak at the shortest wavelength is missing because it lies out of the operating range of the spectrophotometer). The remaining peaks are centered at 246 nm and 301 nm and can be assigned to the ²F_5/2 → ²D_3/2 transition. The first one shows a shoulder at 255 nm that can be ascribed to Er ⁴I_15/2 → ²D_7/2. The spectra of the samples with different Ce doping level have identical shapes, but their intensities are different.

The polarized absorption spectrum of the Er ⁴I_15/2 → ⁴I_11/2 transition at room temperature is shown in Fig. 3. This band has been used for diode pumping. The peak values are α = 1 cm^-1 at 968 nm in E∫∫b polarization, and 0.8 cm⁻¹ at 970 nm for E∫∫c, and are the same for all the samples showing that the Er doping level does not change.

The comparison of the emission spectra intensities at 1.5 μm is shown in Fig. 4 for all the samples except the 0.5%-Ce codoped one, because its emission spectrum was almost undistinguishable from that obtained in case of 0.2%-Ce codoping. The samples were compared in the same geometrical conditions, by using a specially designed mount to reproduce the same experimental conditions in all cases. The intensity increases with the Ce content in the crystals, and in the most heavily Ce-doped one it is five times larger than in the case of solely-Er doped sample.

A confirmation of the quenching effect of Ce is given by the intensity trend of the ⁴I_11/2 → ⁴I_13/2 Er transition. We obtained that the 3 μm fluorescence intensity monotonically decreases with increasing Ce concentration (the opposite of the ⁴I_13/2 →⁴I_15/2 intensity trend). For the sample codoped with 2% Ce, the intensity reduces to about 10% with respect to the solely-Er doped crystal. For the 4% Ce codoped sample, the 3 μm emission is undetectable.

 

Fig. 2. Unpolarized absorption spectrum of the sample with the 0.2% Ce concentration.

Download Full Size | PDF

 

Fig. 3. Polarized absorption spectrum of the ⁴I_15/2 →⁴I_11/2 transition of Er.

Download Full Size | PDF

 

Fig. 4. E ⊥ c emission spectrum of the ⁴I_13/2 →⁴I_15/2 transition of Er.

Download Full Size | PDF

The room temperature lifetimes of the ⁴I_11/2 and ⁴I_13/2 multiplets of Erbium are summarized in Table 1. The ⁴I_13/2 lifetimes are exponential in all the samples, whereas the ⁴I_11/2 ones are exponential only in the samples doped with 0%, 0.2% and 0.5% Ce. In the two higher-Ce concentration crystals these decays are markedly non-exponential giving evidence of a strong quenching of this level by Ce. For a uniform evaluation of the lifetimes both in exponential and non-exponential cases, all the values in Table 1 have been calculated as the integral of the decay curve normalized to the initial intensity. The given lifetime values have been calculated as average of the value of the normalized decay integrals in a series of different subsequent acquisitions, whereas the uncertainties have been evaluated as the standard deviation of the same measurements series.

Our results show that in the sample doped only with Er the decaytime of the upper level is about half of that of the lower one. The Cerium presence selectively shortens the ⁴I_11/2 lifetime, and in the 4% Ce codoped sample the ratio of the lifetimes of the ⁴I_11/2 and ⁴I_13/2 levels reduces to about a fiftieth. The lifetime characteristics and the 3 μm intensity trend are a clear indication of the Ce effect in depopulating the Er ⁴I_11/2 level and in feeding the ⁴I_13/2 multiplet. Moreover the behavior of the 1.5 μm emission intensity confirms this hypothesis even if the lifetime dependence of this level suggests that a slight quenching effect takes place on this level, too. Anyway in this case the intensity of the quenching is much smaller than on the other excited levels, probably thanks to the worse energy matching of the gaps involved. This is clear from three main facts: first the lifetime shortening is less pronounced, secondly the decay behavior is always exponential, and finally the intensity of the emission increases with Ce concentration.

Table 1. Erbium lifetimes as a function of the Cerium codoping.

View Table | View all tables in this article

To fully characterize the Er-Ce:BaYF system, we also analyzed the ⁴S_3/2 →⁴I_13/2 transition, because in case of diode pumping on the 970-nm band, upconversion population of the ⁴S_3/2 green-emitting multiplet is also detrimental to feeding ⁴I_13/2. In fact, one of the most important loss channels for the ⁴I_13/2 population is due to parasitic population of the high-lying levels. We found that the intensity of the green fluorescence dramatically decreases in the Ce-codoped samples (it reduces by about a hundred times when adding 4% Ce). This is related both to the depopulation of the ⁴I_11/2 level and to a direct cascade quenching of the ⁴S_3/2 by Cerium and it increases the efficiency of the 980 nm pumping, allowing the energy of the pump to be mainly deposed on the upper 1.5 μm level. In Fig. 5 the intensity trends of the ⁴I_13/2 →⁴I_15/2 (black squares) and the ⁴S_3/2 →⁴I_15/2 (red triangles) transitions as a function of the Ce concentration are shown together. As we can see, the effect tends to saturate at the highest Ce concentrations we studied.

 

Fig. 5. Comparison of the normalized emission intensities as a function of the Ce concentration: black squares: 1.5 μm transition; red triangles: ⁴S_3/2 transition.

Download Full Size | PDF

To characterize the direct quenching of the Er ⁴S_3/2 by Ce, we measured the lifetime of this level under blue excitation into the ²H_11/2 level, which can be practically considered as a direct ⁴S_3/2 excitation because of the thermal coupling of these two manifolds at room temperature. The decays are not exponential, except for the case of the sample doped only with Er, therefore we calculated the integral of the decay curve normalized to its initial intensity as an estimate of the effective lifetime. In Table 2 the integral lifetimes are listed. Similarly to the ⁴I_11/2 and ⁴I_13/2 cases, the values of the lifetimes and of their respective uncertainties we give in Table 2 arise from the average and the standard deviation we calculated on a series of following measurements. The ⁴S_3/2 lifetime reduces to a fourth in the 4% Ce-codoped sample, compared to the solely-Er doped one. This is a clear indication that a direct quenching of the ⁴S_3/2 multiplet by Ce also takes place (Eq. 2) and, accordingly, a further quenching of the ⁴F_9/2 (Eq. 3) is likely.

As we can see, the Ce codoping in Er-doped BaYF produces two effects: first, a cascade quenching for all the excited energy levels which is larger the smaller their energy gap. Therefore this quenching effect is expected to be maximum for the ⁴S_3/2, ⁴F_9/2 and ⁴I_11/2 multiplets because their energy gaps are close to that of Ce, and minimum for the ⁴I_13/2, whose energy gap is by far the largest. The second effect of Ce is that the depopulation of all the energy levels lying above ⁴I_13/2 (in particular ⁴I_11/2) quenches the upconversion mechanisms towards the high-lying levels. Both these effects improve the population efficiency of the ⁴I_13/2 and make the population inversion for the 1.5 μm laser easier.

Table 2. Erbium ⁴S_3/2 lifetimes as a function of the Cerium codoping.

View Table | View all tables in this article

5. Laser parameters

Erbium emission cross sections at 1.5 μm have been calculated by the integral β-τ method [23], using the polarized room-temperature emission spectra and the measured value of the ⁴I_13/2 lifetime in the case of the sample without Ce. Because ofthe magnetic-dipole contribution to the ⁴I_13/2 → ⁴I_15/2 transition and the monoclinic symmetry of the crystal, we considered the total of the six different polarizations allowed by both the electric and magnetic field directions. In Fig. 6 we show three polarizations only, for clarity. The shape of the obtained cross section curves is strongly dependent on the polarization, the peak values are 6.3 ∙ 10^-21, 10.3 ∙ 10^-21 and 4.5 ∙ 10^-21 cm² at 1538 nm in the (E∫∫c & H∫∫c), (E∫∫c & H∫∫b) and (E⊥b,c & H∫∫c) polarizations, respectively.

 

Fig. 6. Polarized Erbium emission cross sections.

Download Full Size | PDF

An evaluation of the laser gain as a function of the inversion coefficient γ (defined as the ratio of the population of the upper laser level over the total Er₃₊ upper and lower level population densities) is possible by using the formula:

where σ_em labels the emission cross section and σ_abs the absorption one. In Fig. 7 we show the gain cross section for different values of the γ parameter, for the (E∫∫b & H∫∫c) polarization. From the Fig. 7, we are able to estimate the most probable lasing wavelength in this polarization to be around 1615 nm, with a net gain region from 1601.3 to 1625.8 nm for γ = 0.2, and from 1536.7 to about 1640 nm if γ = 0.4.

 

Fig. 7. Gain cross section in the E∫∫b, H∫∫c polarization of Er:BaYF.

Download Full Size | PDF

In order to quantitatively evaluate the Ce effect on the feeding efficiency of the ⁴I_13/2 Er multiplet, we developed a simplified “three-plus-one-level” model (Fig. 8), in which the Ce effect is modelized with a β _Ce parameter which takes into account the ⁴I_11/2 lifetime shortening. The decreasing of the ⁴I_13/2 lifetime is phenomenologically considered by using in the calculations the measured value for every crystal. It is important to underline that our model neglects all upconversion mechanisms. Therefore its validity increases as a function of the Ce concentration in the samples (for fixed pump power) because of the Ce effect that decreases the upconversion processes.

 

Fig. 8. Schematics of the model used for Eqs. 5–8.

Download Full Size | PDF

In case of pure Er doping, the system (neglecting upconversion) is characterized by the following set of coupled rate equations:

where N ₂, N ₁ and N ₀ are the populations of the ⁴I_11/2, ⁴I_13/2 and ⁴I_15/2 levels, respectively, and N_Er is the Er doping density. P is the pump rate (expressed as absorbed photons per unit time). τ ₂ and τ ₁ are the measured lifetimes of the 2 and 1 multiplets, respectively, in the solely-Er-doped crystal.

We can explicitly write the inversion coefficient γ of Eq. 4 as:

In the following, the γ parameter for the crystal without Ce will be indicated as γ_Er , whereas γ_Er-Ce will indicate the parameter in case of Er-Ce codoped systems. When needed for a clear identification of the crystal we will speak about, the concentration of Ce will be explicitly written (for instance γ _Er-2%Ce).

By considering the stationary case, from Eqs. 5(a)–(c) we obtain the following relation between the γ_Er parameter and the pump rate P:

In case of Er-Ce codoping, the rate equations become:

where β_Ce is connected to the measured value τ*₂ of the ⁴I_11/2 lifetime in the sample under investigation by the following equation:

which models the cross-relaxation of Eq. 1 as a linear decay. This model is correct only if the ground state population of Ce is constant, as it seems reasonable to expect given the strong multiphonon relaxation at room temperature of the Ce ³F_7/2. Moreover, from a rate equation analysis we found that the Er ⁴I_11/2 population is always much smaller than the Ce ground state population for every Ce concentration. This ensures that the Ce quenching is always effective.

In Eqs. 8(a)–(c) N ₂, N ₁, N ₀ and N_Er are the populations of the ⁴I_11/2, ⁴I_13/2 and ⁴I_15/2 levels and the Er doping density, respectively and ${\tau }_1^{*}$ is the measured lifetime of the ⁴I_13/2 multiplet in each codoped sample we are considering.

By solving Eqs. 8 in the stationary case, and by substituting in them the expression of the pump rate P, obtained from Eq. 7 as a function of the γ_Er parameter, we obtain the inversion parameter γ_Er-Ce for Er-Ce codoped crystals as a function of γ_Er itself:

This allows to compare, for fixed pump power (i.e., for a given γ_Er value), the inversion efficiency as a function of the Ce codoping. In the calculations, we used the 0.35 value of the branching ratio β ₂₁, given in [20]. In Fig. 9 the γ_Er-Ce /γ_Er ratios in the various samples are compared as a function of γ_Er parameter. For clarity reasons we did not show in the Fig. 9 the case of the sample codoped with 0.5% Ce because its behavior is almost undistinguishable from that of the sample codoped with 0.2% Ce. In the same Fig. 9 we depicted also the γ_Er parameter itself (horizontal black line), to better compare the values obtained in Ce codoped samples. We can observe a monotonic increase of the inversion parameter for fixed pump level as the Ce concentration is increased. The effect is most striking for the sample with the highest Ce concentration, where γ_Er-Ce /γ_Er is about 50% larger than 1 for γ_Er = 0.2, and 33% higher for γ_Er = 0.4.

These results are interesting to give a quantitative evaluation of the Ce effect, even if they have been obtained from a simplified model which neglects upconversion. We have to notice that our model has some approximations that are bigger for crystals with the lowest Ce concentrations and in particular for the crystal doped only with Er, because in this system the upconversion processes are the largest. However, three facts have to be considered: the first one is that the jEr parameter we obtain from our model is an upper limit for the inversion coefficient we would have if we also consider upconversion. In other words: our model overestimates γ_Er for a given pump level. The second fact is that in our crystals the Erbium concentrations are the same (within the resolution of our apparatus), and moreover, all the crystals have been compared at fixed pump rate: therefore it seems reasonable to expect upconversion to play a similar role in all the samples, but for the effect of Cerium. The last consideration, in fact, is that in Ce-codoped crystals the upconversion processes are strongly reduced, and this is mostly true the highest the Ce concentration.

Moreover for the reasons we just spoke about, if we developed a more complete model by taking into account the upconversion processes, probably we would have obtained a difference in the inversion coefficients between the 4%-Ce doped sample and the solely-Er doped one even larger than the 50% increasing we obtained with our simple model. However, our goal was not to perform a detailed analysis of the Er-Ce energy transfer dynamics, but just to demonstrate the Ce usefulness in improving the feeding of the ⁴I_13/2 Er multiplet and therefore in reducing the threshold and increasing the efficiency of a 1.5 μm Er:BaYF laser.

6. Conclusions

In this work the effect of Cerium codoping on Er:BaYF crystals has been investigated. Owing to the intrinsic difficulty to dope BaYF with high concentrations of Cerium, the maximum obtainable Ce concentration was 4% in the melt. However, the results of our investigations show that the Ce presence triggers some cross-relaxation mechanisms from the upper lying manifolds that strongly help the population inversion efficiency for the 1.5 μm emission, without greatly affecting the ⁴I_13/2 level lifetime. In particular, by shortening the ⁴I_11/2 pumping level decaytime, Ce dramatically decreases both the probabilities of parasitic radiative ⁴I_11/2 → ⁴I_13/2 3 μm transition and upconversion population of the green-emitting level ⁴S_3/2. As for the 3 μm transition, it is important to notice that in solely-Er doped crystals, very often this is the preferred lasing channel. Severe suppression of this radiative transition is the only way to obtain laser action at 1.5 μm in such hosts. Regarding the green loss channel, we have to keep in mind that unlike oxides, in fluorides the ⁴I_11/2 lifetime is long enough to produce a meaningful parasitic upconversion population of the high-lying levels in the case of 980 nm pumping, with consequent reduction of the pump efficiency. So, it is reasonable to expect that in fluorides the Ce beneficial effect could be even greater than in already studied oxide hosts, and therefore the Cerium addition to Er:BaYF appears very interesting to obtain a low-threshold, high-efficiency 1.5 μm laser.

 

Fig. 9. Comparison of the ratio between the inversion coefficients γ_Er-Ce /γ_Er in codoped samples and the inversion coefficient γ_Er in the solely-Er doped one for the same pump level.

Download Full Size | PDF