1. Introduction

Optical cooling of solid materials is an emerging technology that has strongly advanced in the last ten years, after it was first demonstrated in 1995 in a sample of Yb-doped ZBLAN glass [1]. To date cryogenic results have been achieved [2–7] and the current state of the art allows to consider in earnest the development of a first generation of all-solid state optical cryocoolers. Temperatures around 150K are achievable in Yb-doped fluoride single crystals, with steady-state cooling power of the order of 100mW. The record result is the laser cooling of a single crystal YLF:10%Yb to 114K, with room temperature cooling power of 750mW [6]. Optical cooling to 93K has also been obtained in YLF:10%Yb, starting from 270K [7]. Although so far the best results have been achieved in Yb doped fluorides, the optical cooling process has been demonstrated also in Er [8,9] and Tm [10–12] doped glassy and crystalline materials.

It is worth to notice that laser cooling in Yb-doped solids has demonstrated access to temperatures well below the cut-off for standard TECs (that is about 180K), and the potentialities of this technology have not by far been exhausted: 80K can be in principle achieved.

Today the implementation of an all-solid state optical cryocooler is highly appealing for space applications. This technology possesses unique features with respect of conventional mechanical techniques: mainly, zero-vibration and no moving parts. Such advantages would permit significant improvements for several high precision space-based technologies, including focal plane IR detectors for imaging, ultra stable laser cavity [13] for atomic clocks and gravitation wave detection, superconductor devices, X-ray and γ-ray sensors. Moreover, the implementation of a device based on this technology would permit to efficiently replace TECs for applications in a temperature range, between 80K and 180K, where no other vibration-free techniques are currently available. The solid-state optical technology also maintains the advantages of compactness, long lifetime (>10⁴ year), low electromagnetic interference and low sensitivity to magnetic fields; all favorable features for space applications.

Such motivations make the investigation of materials as well as the optimization of the laser system, research lines of great interest in order to enhance the refrigeration performances.

The optical cooling process in solids is based on the principle of anti-Stokes luminescence, achievable in an isolated system such as trivalent rare earths ions, embedded in transparent glassy or crystalline hosts. The cooling efficiency ${\eta }_c$ at a given pumping wavelength $\lambda$ and temperature$T$, defined as the ratio between the cooling power ($P_{cool}$) and the absorbed power ($P_{abs}$), can be modeled as [14]:

In this model, two parameters (the EQE and the background absorption coefficient) characterize the cooling performances of the sample: high cooling efficiency is achievable both with high EQE and low background absorption. These coefficients can be estimated by fitting experimental data of cooling efficiency with the model curve in Eq. (1) in order to quantify the cooling performances of a given material.

The EQE defines the radiative efficiency of the cooling transition, addressing the probability of non-radiative decays and the efficiency of photon extraction. Commonly, the combination of low phonon energy hosts (such as fluorides single crystals) and optical transitions in trivalent rare earth ions allows to achieve high values of EQE.

The background absorption coefficient collects all the parasitic loss mechanisms reducing the number of absorbed photons that can initiate the anti-Stokes process. It is currently expected that ${\alpha }_b$ is mainly due to optically active impurities. Contaminants embedded inside the material produce spurious absorption bands and, in case of rare earths, activate energy-transfer process, commonly assisted by phonon release. Also structural defects affect the parasitic absorption, giving heating mechanisms due to scattering of pumping light.

While high structural quality can be achieved more easily thanks to accurate crystal growth procedures, the control of the optical purity to higher and higher level is still quite a challenging process, being strongly dependent not only on the growth process but also on the purification of starting materials.

Today the background absorption is the more critical parameter for the cooling performances of a given material. The control of its value could permit significant enhancement of the refrigeration efficiency.

Several studies identified transition metals as the main source of the background absorption [6, 15]. However its nature and origin still needs to be investigated in order to fully understand the detrimental sources and the related processes. Recently, we observed that also impurities of other rare earth ions, in particular Er and Ho, can reduce the efficiency of the cooling process because of non-radiative energy-transfer and internal processes [16, 17].

It is worth pointing out that so far the efficiency of the cooling process was determined only by the efficiency of the anti-Stokes mechanism, achieved in a given active ion (such as Yb or Tm or Er) embedded in a solid host, and energy-transfer processes, induced by impurities, as well as non-radiative decays, were resulted in losses mechanisms due to occurrence of phonon release.

The novel concept of energy-transfer enhanced anti-Stokes cooling in co-doped system was theoretically suggested by G.Z. Dong et al. [18] in 2013 for Ho-Tm co-doped materials.

In such scheme, the cooling efficiency of the Ho anti-Stokes process around 2 μm is enhanced by phonon assisted Ho-Tm energy-transfer. Via Ho de-excitation ⁵I₇ → ⁵I₈, part of Tm ions get sensitized to the first excite state, and emits photons at energy higher than Ho anti-Stokes ones, i.e. higher than pumping ones, providing an extra release of thermal energy from the material. Such mechanism can be applied only to selected dopants pairs which show first excited states close enough in energy to allow the virtuous energy-transfer followed by anti-Stokes emission. However such scheme strongly suffer from detrimental energy-transfer to upper states and laser sources around 2-μm are not easily available as well. As far as we know, these kind of co-doping systems have been only theoretically investigated.

In this work we propose a novel scheme for anti-Stokes cooling, based on Yb-Tm controlled co-doping. We identified a virtuous energy-transfer that allows for efficiency enhancement, via cooperative sensitization of Yb pairs. Peculiarity of the Yb-Tm system is that the energy of the acceptor state (Tm ¹G₄) is about twice the energy of the donor excited state (Yb ²F_5/2) while phonon depopulation of the acceptor manifold ¹G₄ is prevented by the absence of close enough underlying states. Basically, this work provides first experimental evidence that some virtuous energy-transfer processes can be employed to enhance the efficiency of the anti-Stokes process.

Virtuous contributions of energy-transfer to anti-Stokes cooling will strongly depend on the relative concentrations of dopants. In the present investigation the aim was to study the effect of low Tm concentration combined to dominant Yb doping levels. In order to achieve a significant effect, a controlled Tm codoping of about 10ppm was chosen as starting point for our investigation of Yb-Tm energy-transfer, as concentrations of Tm of the order of 1ppm are normally included in our cooling sample through fluorine raw powders.

In particular, we investigated the cooling performances of a single crystal of YLF:5at.%Yb with a controlled Tm doping level of 0.0016at.% (16 ppm), and its efficiency has been compared with a sample set of YLF single crystals doped with Yb at concentration varying from 5at.% to 10at.%. We observed that Yb-Tm energy-transfer resulted in net increase of the Yb anti-Stokes efficiency, via a significant decrease of the background absorption coefficient.

2. Experimental set-up and cooling measurements

In order to investigate the effect of Yb-Tm codoping on the efficiency of Yb anti-Stokes process, first we grew a sample set of YLF single crystals with Yb doping level 5at.%, 7.5at.% and 10at.%. Subsequently we grew a single crystal of YLF:5at.%Yb where a controlled Tm doping of 0.0016at.% was added.

Growth is a very delicate process, which strongly affect purity and structural quality of materials, hence the efficiency of optical processes. All the samples investigated in this work have been grown by Czochralski technique in our facility. The raw materials were powders of LiF, YF₃, YbF₃ and TmF₃ with guaranteed 5N purity (99.999%), provided by AC Materials (Tampa, Fl., USA). The growth was carried out in controlled atmosphere, made of high purity (5N) argon and CF₄ The addition of CF₄ was necessary to prevent the reduction of Yb³⁺ ions to Yb²⁺ ions. The pulling rate was 0.5 mm/h, the rotation rate 5 rpm and the temperature of the melt was between 860 and 880 °C. The crystals resulted of good quality, free of internal cracks, microbubbles or inclusions. The structural analysis by X-ray diffraction in Laue chamber proved the single crystalline character and returned the orientation of the optical axes. Oriented samples with edges along the optical axes were cut and properly polished for spectroscopy and laser cooling experiments. Samples of 3x3x12mm³ with the a-axis parallel to the long side were employed for the cooling test, all the six facets were polished to high laser quality.

The actual concentration of Tm in the co-doped sample has been checked with an elemental analysis, performed by means of a Laser Ablation-Inductively Coupled Plasma-Mass Spectrometry (LA-ICP-MS) [19].

For each sample, the cooling efficiency was measured collecting the temperature change as a function of the pumping wavelength, in a standard set-up [16].The crystal was mounted on a steel support, inside a vacuum chamber, suspended on two optical fibers (with 120 μm diameter) in order to minimize conductive heat load. The chamber was evacuated at low pressure (of about 10⁻³ Pa) to minimize heat load due to convection in air. Particular care was necessary to manage the heat load on the sample. Four diode lasers, with emission within the absorption spectrum of Yb, were employed to excite the sample, covering the interval between 940 and 1060 nm. All the junctions were mounted so that the crystal was excited along the c-axis that allows the largest absorption coefficient. The temperature was measured via a contactless technique, employing a thermal camera (Raytheon 2500AS) with microbolometer sensor, previously calibrated to determine the proportionality relation between the pixel intensity and the temperature of the emitting crystal. Assuming as dominant heat load the blackbody radiation, the cooling efficiency results proportional to the ratio of the temperature change to the absorbed power. Thermal videos were acquired for each wavelength while the sample was excited with the laser beam in order to measure the temperature change. The video started with the absence of laser light and was extended until the thermal equilibrium was reached; data analysis confirms the exponential change of the sample temperature.

In the same experimental conditions we studied the cooling efficiency of the YLF sample set and that of the YLF:5%Yb with controlled Tm doping. Table 1 shows the EQE and background absorption values estimated for samples of YLF Yb-doped at 5%, 7.5% and 10%. As can be seen, we observed a significant decrease of the background absorption parameter as the Yb doping level was increased, with consequent enhancement of the cooling efficiency. The diminishing of ${\alpha }_b$results either in red-shift and intensity increase of the efficiency curve peak. The EQE parameter does not show a significant variation as a function of the Yb doping level. The slight decrease as the doping level is increased can be ascribed to a slight shift of the mean emission wavelength, due to higher reabsorption effects. The achieved results, reported in Table 1, are in close agreement to those reported in [6], where an analogous analysis has been performed on a sample set of YLF single crystal with Yb doping level between 1at.% and 10at.%.

Table 1. Estimated values of EQE and background absorption parameters for YLF crystals doped Yb at 5%, 7.5% and 10%.

View Table | View all tables in this article

As in [6], we observed a factor of two decrease of the background absorption between the YLF:5%Yb and the YLF:10%Yb, which is so far the best performing material. This agreement with literature data is a strong indication of the high quality of our samples and provides a good validation of our set-up.

Similar investigation has been carried out for the sample of YLF:5%Yb where a controlled Tm doping of 16ppm was added. The estimated values of EQE and background absorption coefficient are: ${\eta }_{ext}$ = 0.988 ± 0.002, ${\alpha }_b$ = (1.9 ± 0.4) 10⁻⁴ cm⁻¹. Experimental data and fitted model curve are reported in Fig. 1. These achievements provide evidence of the virtuous effect of Yb-Tm co-doping on the efficiency of the Yb anti-Stokes process.

 

Fig. 1 Cooling efficiency data points and fit model curve for YLF:5%Yb with 16ppm of Tm doping (circles and red line), YLF:5%Yb (circles and magenta line) and YLF: 10% Yb (circles and blue line).

Download Full Size | PDF

The addition of controlled Tm doping results in a net increase of the overall cooling efficiency, via a significant decrease of the background absorption parameter. Values of ${\alpha }_b$and peak efficiency comparable to that obtained with a YLF doped Yb at 10% were achieved via a controlled Tm doping in a YLF doped Yb at 5%. Moreover, the addition of Tm doping results in lower decrease of EQE with respect of higher Yb doping level, because reabsorption effects are much less intense due to low concentration of Yb ions.

In Fig. 1 are also reported, for comparison, the cooling efficiency curves and data points relative to YLF:5%Yb and YLF:10%Yb in order to point out the virtuous effect of Yb-Tm co-doping on the Yb anti-Stokes efficiency.

A direct comparison between the two YLF:5%Yb samples, one co-doped with Tm and the other barely doped Yb, highlights the efficiency increase achieved in case of addition of Tm codoping in terms of background absorption decrease. A direct comparison between the YLF:5%Tm-0.0016%Tm and the YLF:10%Yb highlights advantage of Yb-Tm codoping as a route to increase the efficiency maintaining low reabsorption effects .

3. Spectroscopic investigation

In order to investigate the processes involved in the Yb-Tm energy transfer and their effect on the cooling performances, a spectroscopic characterization of our samples has been carried out. In addition to analysis of the Yb transition, a simple observation suggested the investigation of visible and IR emissions by exciting the Yb ions: when pumped in the cooling chamber, the samples with only Yb showed a typical green fluorescence, while the sample with Yb and controlled Tm doping showed a blue emission. This fact is indication that different energy-transfer processes occur in case of Yb-Tm codoping in respect to the bare Yb doping.

The polarized absorption and fluorescence spectra of the Yb transition were acquired, in the same experimental conditions, for both the 5% Yb-doped samples. The spectra showed no appreciable difference in size or structure ascribable to the addition of Tm doping. Therefore the discrepancy in cooling performances appears to be unrelated to the Yb spectroscopic properties.

The visible and IR emissions were acquired by exciting the Yb transition, at 940 nm benefiting of Boltzman thermalization within manifolds at room temperature. For simplicity, in the following we limit the comparison between spectroscopic data only relative to the 10%Yb doped and 5%Yb-controlled-Tm doped YLF samples, as the controlled Yb-Tm codoping reflects in efficiency enhancement effects similar to the increase of Yb doping level from 5% up to 10%. However a similar behavior in visible-IR emissions was detected for all the sample doped only Yb.

The emission spectra for the two samples of YLF:5%Yb-0.0016%Tm and YLF:10%Yb are reported for comparison in Fig. 2; in the inset it has been included the IR fluorescence. To increase the visibility of the features of the spectra, the green and red fluorescence signals have been multiplied for a fixed factor, as indicated in the figure. All the signals are in arbitrary units, but the scale has been respected, except for the IR signal, because of a different detector.

 

Fig. 2 Visible and NIR emissions for YLF:5% Yb with 16 ppm Tm doping (red curve) and YLF:10% Yb (blue curve) obtained by pumping the Yb transition at 940 nm. In the inset is shown the emission in 2 μm region.

Download Full Size | PDF

A first survey of the spectra reveals that the Tm-enriched sample has intense emission bands at energies higher than the photons absorbed by Yb ions. The blue emission is dominant, along with intense emission centered at 790 nm and a minor red band at 650 nm. These bands result strongly depressed in the YLF doped only with Yb.

The emission spectrum of the sample with controlled Tm doping suggests the mechanisms of Yb-Tm energy transfer, and their contribution to the enhancement of cooling efficiency. A basic consideration is the starting point for our preliminary model: the addition of a low Tm doping results in increase of the overall cooling efficiency, so the balance of all the processes involved in the Yb-Tm energy transfer must not be exothermic, i.e. must not involve net phonons emission. Figure 3 shows a subset of energy levels of Tm and Yb ions in YLF crystal along with the main transitions detected in the spectroscopic analysis shown above.

 

Fig. 3 Energy levels of Tm and Yb ions in YLF host with the main transition involved in the energy-transfer process.

Download Full Size | PDF

The blue fluorescence (around 480 nm) is identifiable with the ¹G₄→³H₆ transition of Tm ions (indicated as 1 in Fig. 3). The assignment is made considering the energies of Stark sublevels in YLF host [20].

Energy-transfer in Yb-Tm system can result in blue up-conversion of Tm ions via two main processes: three photons sequential non-resonant energy-transfer [21] or Yb-Yb cooperative sensitization [22–24]. In the latter process, a pair of excited Yb ions coupled and simultaneously transfer its energy to one Tm ion, passing from the ground state ³H₆ to the ¹G₄ state. In case of donor (Yb) concentration dominant over the acceptor (Tm), the two photons process (cooperative sensitization) can be more likely than the three photons sequential non-resonant energy-transfer [25].The Yb-Yb interaction is favored and the direct energy-transfer (three photon process) from Yb to Tm is precluded by the energy gap between the Yb - ²F_5/2 and the closer Tm - ³H₅multiplet. As mentioned earlier, cooperative sensitization is peculiar of the Yb-Tm system because the energy of the donor exited state (Yb - ²F_5/2) is approximately one-half the energy of the excited state of the acceptor (Tm - ¹G₄).

The measurement of blue fluorescence intensity at 477 nm as a function of the pump power at 940 nm proved that the excitation of the ¹G₄ manifold was mostly a two photons process: the fit of experimental data returned a power law with an exponent value of 2.3 ± 0.2. Experimental data and the best fit are shown in Fig. 4. Error bars are smaller than the symbols used.

 

Fig. 4 Blue fluorescence vs 940nm pump power. Black squares are experimental data, the continuous line is the best fit of a power law, with the fitted exponent value 2.3 ± 0.2 .

Download Full Size | PDF

While the three-photons process involves strong phonons emission, the cooperative sensitization energy-transfer is virtuous for the cooling mechanism. Considering the energies of Stark sublevels and the mean emission wavelength of the Yb transition, a pair of Yb ions provides the energy to excite the ¹G₄ manifold just above the first Stark sublevel. Annihilation of lattice phonons is thus required for the thermalization of electronic population within the ¹G₄ manifold. Benefits of Yb-Tm controlled codoping are strongly related to such up-conversion process. Firstly, it contributes to the Yb cooling process because the energy-transfer to Tm ions involves annihilation of lattice photons; secondly provides strong emission at about twice the energy of pumping photons, without phonon losses as consequence of the resonance of energy levels. It is worth pointing out that the Stark width of the ¹G₄ manifold at 10 K is about 600 cm⁻¹, comparable with that of Yb - ²F_5/2manifold (about 400 cm⁻¹) and the cut-off for phonon energy in YLF host is 400 cm⁻¹ [26].

The red fluorescence (around 650 nm) is a characteristic line of the Yb-Tm cooperative sensitization and is attributable to the ¹G₄→³F₄ transition of Tm (indicated as 2 in Fig. 3) [22].

The other intense emission peculiar of the Yb-Tm codoping, and strongly depressed in samples doped only Yb, is that centered at 790 nm. Such emission is identifiable with the ³H₄→³H₆ transition of Tm ions (indicated as 3 in Fig. 3). Several processes might contribute to populating the ³H₄ manifold. At first, radiative emission from the ¹G₄ manifold. Such transition however is nearly resonant with the Yb ²F_5/2 → ²F_7/2, therefore it cannot be easily detected, due to large overlap with the dominat Yb fluorescence. Moreover, non-radiative de-excitation ¹G₄ →³H₄ might provide the energy for the Yb ²F_7/2 →²F_5/2 transition to occur, contributing to population of the Yb excited manifold on the red-edge of the absorption spectrum. Amongst all the processes that might populate the ³H₄ manifold, even up-conversion (³F₄, ³F₄)→ (³H₆, ³H₄) between Tm ions can occur. This process is virtuous for the cooling mechanism as it requires phonons annihilation to occur because of the higher energy gap of the ³F₄→³H₄ transition with respect of the ³F₄→³H₆ transition. The 10K energy mismatch of about 1400cm⁻¹ involves absorption of some phonons at room temperature. A long lifetime (of the order of 2ms) was measured for the ³H₄ manifold, proving that the rate of detrimental cross-relaxation process (³H₆, ³H₄)→(³F₄, ³F₄) is negligible.

The green (around 540 nm) and IR (around 2µm) bands have been ascribed to Er and Ho impurities contained in the starting powders [16,17]. In particular the 2µm band is identifiable with the ⁵I₇→⁵I₈ transition of Ho, while the green emission is ascribable to both Er (⁴S_3/2→⁴I_15/2) and Ho (⁵F₄→⁵I₈). The intensity of these bands is near proportional to the Yb doping, and seems to be unrelated to the Tm concentration.

Er (⁴F_9/2 → ⁴I_15/2) and Ho (⁵F₅ → ⁵I₈) impurities can also contribute to red fluorescence. Such emission channels can be responsible of the different peak ratio, between YLF:5%Yb-0.0016%Tm and YLF:10%Yb curves, observed in the red band with respect to the blue and ~790nm emissions. This fact can be related to activation of different energy-transfer in the two samples, because of dissimilar relative concentrations of Tm, Er and Ho acceptor inside. Table 2 reports the concentration of Tm, Er and Ho in YLF:10%Yb and YLF:5%Yb-0.0016%Tm as measured by means of LA-ICP-MS.

Table 2. Concentration of Ho, Er and Tm impurities in the YLF:10%Yb andYLF:5%-0.0016% cooling samples, measured by means of LA-IPC-MS.

View Table | View all tables in this article

In the co-doped sample, due to remarkable larger concentration of Tm acceptor compared to that of Ho and Er, the Yb-Tm energy-transfer, via cooperative sensitization, can be dominant.

Conversely, the YLF doped only Yb includes a Tm concentration similar to that of Er and Ho. A direct energy-transfer from Yb to Er or Ho impurities, similar to that reported in [17], can thus be highly likely. As a consequence, in the YLF:10%Yb higher emission from Er and Ho can occur than for the Tm-enriched YLF, where red fluorescence is essentially related to the cooperative sensitization process. In case of co-doping, the direct-energy transfer from Yb to Er and Ho can be fairly inhibited by the favored Yb-Tm channel.

In YLF samples doped only Yb cooperative sensitization of Yb ions can certainly occur, but the low concentration of Tm acceptor strongly reduces the efficiency of the virtuous Yb-Tm energy-transfer. It is worth pointing out that Er does not present any manifold resonant to the Yb-Yb pair energy. The closest manifold is the underlying ⁴F_7/2 state. However the transfer of excitation from Yb-Yb pair to such manifold involves phonon emission. Ho instead shows an energy level nearly resonant (⁵F₂) to Yb-Yb pair, as Tm does, but the drawback is the presence of a close underlying state ⁵F₃ which induce strong phonon emission from the upper ⁵F₂ manifold.

4. Concluding remarks

In this work we observed an increase of cooling efficiency in YLF:Yb single crystals in case of addition of a controlled Tm doping, via a significant decrease of the background absorption coefficient. This is first experimental evidence that some energy-transfer between Yb and other peculiar trivalent rare earth ions can virtuously contribute to the cooling process, based on the anti-Stokes mechanism.

The Yb-Tm energy transfer can result in enhancement of the cooling efficiency because of the peculiar respective energy levels configurations and doping concentrations. On the contrary of standard energy-transfer that commonly involves release of lattice phonons giving heating mechanisms, in the Yb-Tm system the energy-transfer is virtuous because of the occurrence of a cooperative sensitization between Yb ions, when controlled concentration of Tm ions are employed along with dominant Yb doping level. Yb-Tm cooperative sensitization requires phonons annihilations to restore the equilibrium of electronic population within the acceptor excited manifold. Moreover it determines radiative emissions at energies higher than the photons absorbed by Yb ions, without phonon losses.

A preliminary model of internal transition within Tm acceptor ions is proposed. Further investigation are required in order to completely define which transitions determines the virtuous blue and NIR radiative emissions in the codoped material and the energetic contribution of energy-transfer to the cooling process. A systematic study of the cooling efficiency in Yb-Tm system for different Tm doping level is in progress in order to determine the behavior of Yb-Tm energy-transfer as a function of the Tm concentration and the optimal relative concentration. Such investigation would also permit to verify the significance of all these energy-transfer paths.

Spectroscopic measurements as a function of temperature are underway as well either in order to estimate the minimum achievable temperature (MAT) in case of controlled Yb-Tm codoping either to investigate the energy-transfer cooling contribution as a function of temperature.

As the temperature is lowered, a slight efficiency decrease for the population of the ¹G₄ is expected, because phonon annihilation is required. However decrease of phonon excitation as a function of temperature is the limit of optical cooling. As the temperature decreases, of the efficiency of the anti-Stokes process reduces as well. The Yb-Yb cooperative sensitization should not be appreciably affected by temperature decrease.

From our preliminary results, improvements of MAT can be expected in case Yb-Tm co-doping. A lower MAT value can be expected as consequence of halving of background absorption between a controlled co-doped YLF:5%Yb and a bare Yb doped YLF. However the slight decrease observed for EQE might hinder a significant decrease of MAT.

To conclude, it worth pointing out that Yb-Tm codoping is of great interest as route to enhance the cooling efficiency in Yb doped materials, basically, because of two fundamental reasons:

The Yb-Tm co-doping might results an appealing tool to obtain more efficient optical cooler.

This novel approach of Yb-Tm controlled co-doping might also open the possibility for a self-cooling Yb laser. A self-cooling Yb laser could be in principle obtained increasing the Yb doping level in order to achieve an efficient absorption coefficient for the laser process at wavelengths longer than the mean emission wavelength of the Yb transition, and exploiting the Yb-Tm energy transfer.