Current work is dedicated to theoretical and experimental studies on YAG single crystals performance with non-homogeneous activator ion distribution. Improved lasing characteristics are described and quantified theoretically. The first crystals with such a distribution have been obtained as a result of several growth attempts using Horizontal Direct Crystallization method (Bagdassarov’s method). Extracted from the as-grown boules, samples have been optically treated and studied further. The preparatory phase, growth process and equipment as well as post-growth studies of the extracted samples are equally presented in this paper. Energy-dispersive X-ray analyses were used to quantify the doping distribution values in the grown crystals. Optical transmission experiments were performed to verify the optical characteristics of the as-grown crystals.
©2011 Optical Society of America
With the rapid development of the Diode-Pumped Solid-State Laser (DPSSL) in the recent years, trivalent Ytterbium ion (Yb3+) doped crystals, characterized by lower thermal load and higher capacity for laser energy storage, when compared with Nd3+ doped matrices, have attracted much attention. Yttrium Aluminum Garnet Y3Al5O12 (YAG) is a very interesting host for Yb3+ in high average power DPSSL applications due to its high thermal conductivity, small quantum defect, absence of excited state absorption and up-conversion losses, a relatively long radiative lifetime of the upper laser level (optimizing pump laser diodes capital cost investment). Moreover, YAG is relatively easy to grow over a wide range of doping concentrations and through several distinct techniques.
To our knowledge, all operating laser facilities rely on crystalline gain medium with uniform volume doping distribution. Although, composite structures made of doped and undoped sections are commonly used (usually for thermal and Amplified Spontaneous Emission management purposes), no Yb3+:YAG crystal exhibiting continuously variable doping distribution has ever been used due to the lack of availability for such a gain medium. We illustrate in section 2 the numerous advantages, potentially offered by such structures in terms of thermal and Amplified Spontaneous Emission (ASE) management but also for gain medium volume requirement. These issues are indeed of high relevance for laser programs like the European High Power laser Energy Research facility (HiPER) dedicated to demonstrate the feasibility of laser driven fusion as a future energy source . Section 3 is dedicated to the description of the growth technique used to produce the first 3 boules exhibiting doping gradient profiles. From these boules, several crystals were extracted for further investigations on doping distribution, optical and crystalline quality, as illustrated in section 4 and 5. Fluorescence spectra were recorded and compared with a numerical model (section 6). Finally foreseen laser applications are discussed in the last section.
These studies were performed at the crystal growth department of Lazerayin Tekchnika CSC, Erevan, Armenia and at the Laboratoire d'Utilisation des Laser Intenses (LULI), Ecole Polytechnique, Palaiseau, France where laser applications for gradient doped Yb:YAG are foreseen in the framework of the Lucia program [2,3].
2. Foreseen advantages of gradient doped laser gain medium
Engineering high energy / high average power laser system like the HiPER  laser program requires facing challenges not previously addressed by other multiple 100kJ-class laser systems like the National Ignition Facility (NIF) or Laser Mega Joule (LMJ). Indeed, HiPER’s goal being to demonstrate Inertial Fusion Energy (IFE) production through Fast Ignition in a mode of operation similar to a reactor of a real IFE power plant. Such an operation requires a minimum repetition rate of ~10 Hz . Considering the cost of electricity  analysis and maturity of the available technology, it is very likely that such multiple ~10 kJ/~10 kW beam line facility will rely on diode pumped lasers. Damage threshold limitations and cooling efficiency are among the constraints justifying a high aspect ratio for the gain medium geometry. Figure 1 illustrates two different amplifier architectures currently explored : a dual side pumped slab and single side pumped active mirror concepts.
In addition to common issues with LMJ/NIF facilities (ASE, gain medium large volume management…), a high repetition rate operation brings thermal management on the forefront of key laser issues. Current laser gain media used on such large laser facilities rely on uniformly doped solid state laser materials (glass, crystal or ceramic). We present in this section the advantages of variable doping on gain medium volume requirement, ASE and thermal management. Since this work was performed within the framework of the active mirror based LUCIA DPSSL program at the LULI, the discussion takes place within this specific context. Of course the principle remains true for a dual side pumped slab.
Figure 2 illustrates the case of a 1.9 at% uniformly doped Yb3+:YAG crystal pumped from the left (green arrow) with an intensity of 14 kW/cm2. The pump light is absorbed when travelling through this 0.75 cm thick gain medium, but also on its way back after being reflected on the High Reflectivity (HR) coated (right) face. Only this HR face can practically be actively cooled, which is not optimum since the heat load is located in the vicinity of the pump face, as shown by the curve plotting the stored energy density along the gain medium thickness. Consequently, a noticeable temperature gradient takes place. Resulting internal stresses therefore affect both the amplified laser beam quality and the long term physical integrity of the crystal. Relying on a gradient doped crystal (1.3 at% (pumped face) to 2.3 at% (cooled face)) helps circumventing these issues as illustrated in the right sketch of Fig. 2. The associated stored energy density curve (red) reveals indeed a much more homogeneous heat load distribution.
Considering the required doping level (in the order of 1 at.%) and square meter scaled aperture, the foreseen amplifiers are subject to ASE losses and parasitic lasing, if the product gain [g0 in cm−1] times width [L in cm] becomes too high. A commonly used criterion setting the maximum ASE gain tolerance is g0.L ≤ 4: in that case, exp( + g0.L) = 55 and the loss to such amplified signals must be at least 98% at the edge of the amplifier material in order to avoid any transverse oscillations. Such a loss level can be achieved by the use of i.e. liquid index matching or an absorbing cladding.
In our example, the crystals are 4 cm wide; therefore such criterion is reached when g0 = 1 cm−1. For the gradient doped crystal, the small signal gain is below 1 cm−1 everywhere, as illustrated in Fig. 3 (red curve) - no dramatic losses due to oscillations are expected. In the case of the uniformly doped crystal, the g0.L product will be higher in the layers located between z = 0 and approximately 0.3 cm, reaching values as high as 1.45 cm−1 (dotted curve) nearby the entrance pump face. This location will therefore very likely show transverse oscillations affecting the available local the gain. Three dimensional ASE simulations confirmed the associated gain depletion with a 7 times shorter effective storage lifetime (when compared with the 1 ms nominal value for Yb hosted in YAG).
Let us also mention a specific advantage for LMJ-like large size laser systems, where several thousands of glass slabs weighting more than 100 tons are required . As it can be seen in Fig. 4 , the gradient doping has a positive impact on the extractible energy density for a given gain medium volume. A 0.75 cm thick crystal (central sketch) is sufficient to reach the extraction capacity (~6 J/cm2) of a 1.15 cm thick uniformly doped crystal (left image). In both cases, doping has been adjusted in such a way that the g0 < 1 cm−1 constraint is satisfied for any layer inside the crystal. Such a 35% decrease on the requested volume can have a dramatic impact on HiPER or any large laser requiring high amounts of laser gain media. The same figure also shows (right sketch) the case of an equivalently thin (0.75 cm), but uniformly doped crystal from which a similar extractible energy density would be achievable, when neglecting ASE. Doping ion concentration must then be increased from 1.3 to 1.9 at%, implying a 50% increase for the maximum gain at the entrance face (g0 = 1.5 cm−1). ASE losses will then become important and actually the 6 J/cm2 extractible value will not be achievable.
Finite Element Method (FEM) analysis was performed for both ASE compatible cases (left and centre on Fig. 4, where cooling was achieved through a 295 K air convection on the pumping side (z = 0) characterized by a heat exchange coefficient of h = 10 W/m2/K. The cooling is realized by 288 K cold water circulating on the other side with a heat exchange coefficient of h = 30000 W/m2/K. Repetition rate was set to 10 Hz. At thermal equilibrium, the gradient doped crystal ended up to be roughly 30K colder than the uniformly doped one with internal temperature gradient diminished by a similar amount.
To summarize let us mention that gain media with variable doping distribution offer several advantages in terms of thermal management (reducing thermal gradient and peak temperature), prevention of the onset of parasitic oscillations and gain medium volume requirement. It should finally be pointed out that these characteristics offer engineering/design opportunities relevant not only for very high energy laser facilities but also for much lower energy/power, table top laser systems.
3. YAG Bagdasarov method crystallization with controlled Yb concentration distribution
Due to its congruent melting property, Yb3+:YAG single crystals are mainly grown by RF-heated Czochralski (Cz) method in an inert Nitrogen or an oxygenized atmosphere (typically N2 + 2Vol%O2). However, there are some disadvantages using the Cz method to grow Yb3+:YAG crystals. Defects such as core defects and light-scattering particles easily appear in Cz-grown Yb3+:YAG crystals. Large weight loss of Iridium crucible frequently occurs during Czochralski growth process, especially in oxygenized atmosphere. Another method for growing Yb3+ doped YAG crystals is the so-called Temperature Gradient Technique (TGT). This method is widely used for large YAG growth. However TGT is not capable of growing controlled gradient doped crystals, as all the content of the container is fully molten at the beginning of the process. The gradients obtained with that methods are not controllable, very low. The presence of inhomogeneous doping distribution in Cz and TGT grown crystals is discussed in . A method which helps to partially or fully eliminate all the above mentioned issues was proposed and developed by Bagdassarov and called Horizontal Direct Crystallization (HDC).
Sapfir-2MG furnaces at the crystal growth department of Lazerayin Tekchnika CSC, Erevan, Armenia were used to grow controlled gradient doped Yb3+:YAG single crystals by HDC. These furnaces offer the possibility to control growth parameters, like crucible speed and heater temperature, during the crystallization phase itself. We can therefore control the desired doping distribution throughout the whole growth process, i.e. during the preparatory phase, and later by controlling several key parameters during the crystallization phase itself. Such flexibility gives us the ability to obtain a controlled variably doped boule. Detailed furnace description can be found in . We used boat shape Molybdenum crucibles, where the crystal seed is located at its bow (Fig. 5 ). The width of the crucible is limited by the volume of the heater, but the length can vary within a wide range. The height of the crucibles used for these experiments were limited to 40 mm.
3.1 Preparatory phase
Three Yb3+:YAG boules exhibiting doping gradient distributions were successfully grown, demonstrating the repeatability and robustness of the technique. The YAG seed placed at the crucible bow is oriented along the requested preferential crystallographic axis, in our case the  direction perpendicular to the plane shown on the Fig. 5. The crucible lengths (defining the boule size) used for the three growth sequences were 180, 235 and 225 mm respectively. High purity oxide powders for Y2O3 (X99.999%), Yb2O3 (X99.999%) and Al2O3 (X99.95%) are weighed out in appropriate stoichiometric mole ratios for 0, 20 and 50 at.% concentrations. For each run, the powders are mixed and deposited in the crucible. Then, a first crystallization sequence is performed to obtain homogeneously doped crystals with appropriate percentage of Yb ions. Resulting homogeneously doped crystals are then crackled into small pieces by thermal shock leading to the adequate starting material, which is finally distributed in the crucible according to the three profiles displayed on Fig. 5.
3.2 Crystallization phase
The crystallization phase leading to the three boules being rather similar, we detail only the growth process for the first one. Crucible length and heater width are defining three zones, where the content of the crucible can be described as single crystal, liquid melt and starting material, as illustrated by the top left sketch of Fig. 5. From the same figure, it can be seen that the crucible used for first experiment was 180 mm long in total. The melting zone is first defined by the 140 mm length of the heater. This value keeps evolving during the growth process, because of the increasing size of the grown single crystal section in the crucible. Towards already grown single crystal much more efficient heat transfer indeed occurs from the hot fusion section. This consequently reduces the melting zone length. But this melting zone length can be adjusted by in situ adjustment of growth parameters, such as crucible moving speed and temperature field in the heater (through its driving voltage). Consequently, at the beginning of the growth process, as soon as the first 140 mm starting from the crucible’s bow containing pure YAG are melted, the crucible starts to move forward with a velocity of 2 mm per hour, with crystallization starting from the seed. The melting zone therefore starts moving in the direction of the 20 at.% doped starting material and thereby the melt is being gradually fed with Yb3+ ions. Within the melt, all the components are completely mixed and distributed homogeneously. With continuous motion of the crucible more and more of the 20 at.% starting material is included in the melting zone, therefore each new thin layer of single crystal exhibits higher doping than the previous one. Taking into consideration crucible and heater lengths, as well as the starting material distribution, the doping gradient will therefore appear only in the first 40 mm of this first boule.
The growth process of the third boule differs from this scheme. As illustrated in Fig. 5, 50 at.% doped starting material was used. But the main different is that the grown boule was not extracted from the crucible after the full growth cycle ended. Due to optical inhomogeneities present in the boule (in the form of bubbles) it was actually put again in the furnace to undergo another cycle of growth. From the first grown boule, a 25x10x2.9 mm3 sample was extracted alongside the gradient doping direction (25 mm dimension). From the second boule, a 3 mm thick and 130 mm long crystal was extracted (see Fig. 6 ) for doping gradient evaluation and six 1 to 2 cm3 crystals for fluorescence measurements (top picture of left image). From the third boule (right image), a 3 mm thick and 90 mm long crystal was extracted as a reference for the “body”.
4. Doping distribution evaluation
Spectrally resolved absorption spectra have been recorded with a Cary 500 Spectrometer every 2 mm alongside the extracted sample of boule 1 perpendicular to the expected direction of doping variation. Taking into account the constant thickness of our sample ( ± 4 mm over the full length), the observed absorption variation only depends on the doping distribution. Using these spectral data and Lambert’s law T = exp(-NσZ), where T is the transmission, N the density of absorbing atoms, σ the absorption cross-section for 1 at. % and Z the thickness, we derive the doping distribution shown in Fig. 7 .
Three peak wavelengths were singled out in the absorption spectra: namely at 940 nm, 970 nm and 1030 nm. Doping level was calculated for each of them and Fig. 7 shows the averaged value for each of the measurement positions. Error bars are estimated as standard deviation of those three values. The maximum relative error is less than 2.5%. The resulting doping gradient is 0.45 at.% per cm.
Energy Dispersive X-ray Spectroscopy (EDX) measurements were performed in HTSCLab (High Temperature Superconductivity) IPR NAN, Armenia to verify doping distribution for boules 2 and 3. This method allows qualitative (presence of impurities within the crystals) and quantitative measurement (Yb3+ doping level) of atomic component relative concentration. Before being placed in the Silicon detector (model EDS7378), the samples were coated with a Carbon conductive coating. EDX measurement were performed with reference slices cut from the second and third boules (Fig. 8 ) leading respectively up to 1.8 at.% cm−1 and 3.00 at.% cm−1 doping gradients.
5. Crystal quality
Optical quality of the crystals was evaluated through birefringence measurement with crossed polarizers and with interferometry as illustrated in Fig. 9 . These qualitative measurements reveal satisfying quality although the samples were not completely free of defects.
Birefringence analysis of the first sample (performed perpendicularly to the gradient axis) reveals the absence of constraints, whereas, for the second boule sample, we observe strains especially near the edges (observations are made along the gradient axis). The presence of these stresses might be related to the very special geometry of second boule (longer and narrower crucible). Nevertheless, within this 20x20x10 mm3 sample, a clear 12x12x10 mm3 useful area is available.
Transmission measurements of obtained crystals were performed with a Ti:Sapphire laser at a 800 nm wavelength revealing a value only limited by Fresnel losses. Microscopy observation has shown no bubbles or scattering particles, whereas no scattering of He-Ne laser beam travelling through the sample could be detected.
6. Fluorescence measurements
From the second boule, 10 mm thick crystals samples were extracted for fluorescence measurements (Fig. 10 , top right picture). With help of the 130 mm long reference sample (Fig. 6a, bottom), we selected the extracted crystals labeled “2”, where the doping distribution is linearly varying from one end to the other from 2 ± 0.4 at.% to 3 ± 0.4 at.% as illustrated on left graph of Fig. 8 (Area labeled “2”). A 940 nm CW collimated laser source with 2 W/cm2 of beam intensity is used for pumping along the doping gradient direction while fluorescence in the transverse direction is recorded with the help of a camera (Fig. 10, top left). Then, the same operation is performed with the laser beam propagating along the opposite direction. The resulting fluorescence profiles for both cases are illustrated in Fig. 10, bottom right. If the crystal was homogeneously doped, both curves would be similar. This experiment clearly reveals the effect of the doping gradient previously observed with absorption and EDX measurements.
Expected fluorescence distribution was estimated through a 1D model based on the following system of differential equations:Fig. 11 . This reveals a 0.9 at.% cm−1 doping gradient similar to the 1.0 at.% cm−1 experimentally observed for sample 2. The associated extreme doping levels appear within the error of EDX measurement. It should also be pointed out that absorption of fluorescence light by the ~1 mm thin layer between the observed face of the crystal and laser beam path is not taken into account.
7. Conclusion and outlook
To our knowledge the first gradient doped Yb3+:YAG single crystal boules have been successfully grown with HDC technique, leading to cubic centimeter size laser grade crystals. Spectral absorption investigations of extracted samples demonstrate gradient as high as 3 at. % cm−1. X-ray diagnostics reveal the single crystal nature of the gradient doped samples. Measured fluorescence confirms the impact of gradient doping on energy storing control, offering the possibility to ultimately achieve an almost homogeneous gain distribution along the pumping axis, therefore allowing much more efficient ASE and thermal management for laser applications. Thermal analysis experiments are planned in the short term and will be followed by laser operation in oscillators to evaluate impact on lasing threshold.
We envision also growing boules exhibiting triangular or trapezoidal doping distribution by using crucible with 3 starting material zones instead of 2. The resulting doping profile will therefore be suitable for dual side pumping architectures.
The authors gratefully acknowledge the support of the MŠMT, Ministry of Education, Youth and Sports of the Czech Republic and of the Délégation Générale à l’Armement of the Ministry of Defense of France in supporting this work through the HiPER program.
References and links
1. M. Dunne, “A high-power laser fusion facility for Europe,” Nat. Phys. 2(1), 2–5 (2006). [CrossRef]
2. J.-C. Chanteloup, D. Albach, G. Bourdet, P. Hollander, and B. Vincent, “Impact of variable doped gain medium on HiPER multiple kJ / ~10Hz diode pumped beam lines design,” Advanced Solid State Photonics Topical Meeting and Tabletop Exhibit (ASSP), Denver, Colorado, USA, 1–4 Feb. 2009.
3. D. Albach, M. Arzakantsyan, G. Bourdet, J.-C. Chanteloup, Ph. Hollander, and B. Vincent, “Current status of the Lucia laser system,” Sixth International Conference on Inertial Fusion Sciences and Applications (IFSA 2009), San Francisco, USA, 6–11 September 2009.
4. W. R. Meier, “Systems Modeling for a Laser-Driven IFE Power Plant using Direct Conversion,” The fifth International Conference on Inertial Fusion Sciences and Applications (IFSA2007) IOP Publishing, Journal of Physics: Conference Series (2008), Vol. 112.
5. J.-C. Chanteloup, K. Ertel, J. Hein, and B. J. Le Garrec, “Multi kJ Level Laser Concepts for HiPER facility,” Sixth International Conference on Inertial Fusion Sciences and Applications (IFSA 2009), 6–11 September 2009, San Francisco, USA.
7. X. Xu, Z. Zhao, G. Zhao, P. X. Song, J. Xu, and P. Deng, “Comparison of Yb:YAG crystals grown by CZ and TGT method,” J. Cryst. Growth 257(3-4), 297–300 (2003). [CrossRef]
8. Kh. S. Bagdasarov, Vysokotemperaturnaya kristallizatsiya iz rasplava. M.: Fizmatlit, 2004.
9. Kh. S. Bagdasarov and L. A. Goryainov, Teplo- i massoperenos pri vyrashchivanii monokristallov napravlennoi kristallizatsiei. M.: Fizmatlit, 2007.