Open Access
Add to BibSonomy
Abstract
Experimental amplification of 10-ns pulses to an energy of 12.2 J at the repetition rate of 1-10 Hz is reported from a diode-pumped room-temperature distributed active mirror amplifier chain (DAMAC) based on Nd:YAG slabs. Efficient power scaling at the optical-optical efficiency of 20.6% was achieved by suppressing the transverse parasitic oscillation with ASE absorbers. To the best of our knowledge, this is the first demonstration of a diode-pumped Nd:YAG active-mirror laser with nanosecond pulse energy beyond 10 joules. The verified DAMAC concept holds the promise of scaling the energy to a 50 J level and higher by adding 10-12 more pieces of active mirror in the chain.
© 2017 Optical Society of America
1. Introduction
High energy diode-pumped nanosecond laser system with high efficiency and high peak power is attracting increasing attention and arousing intense interest for its current and potential industrial and scientific applications such as material processing, particle acceleration, and pump source of X-ray generator. Among the nanosecond pulsed solid-state laser systems in repetitive operation with the output energy beyond 10 joules, the Mercury project firstly demonstrated the amplification to 61 J at 10 Hz in 2007 using Yb:S-FAP slabs [1]. In a similar face-cooled multi-slab configuration, DiPOLE100 system has recently obtained an recorded output energy of 105 J at 10 Hz based on cryogenic gas cooled Yb:YAG disks [2]. For room-temperature operation of Yb:YAG, 14 J per pulse was realized from an active mirror (AM) amplifier by LUCIA in 2013 however with the repetition rate limited at 2-Hz [3], while a second amplifier relying on cryogenically-cooled Yb:YAG is under development targeting at the 30 J level [4].
To overcome the transparent threshold due to quasi-three-level nature, Yb-doped materials in room-temperature operation always require ultrahigh pump intensity and thus the coupling optics with considerable optical losses and complexity, while some of them such as Yb:YAG also suffer from high saturation fluence that is comparable with the damage threshold of optics, severely limiting the available extraction fluence and thus the extraction efficiency. In comparison, Nd-doped alternatives with four-level system have the advantage of operating in the room temperature with moderate pump intensity as well as an extraction fluence several times higher than the saturation fluence. In 2008, HALNA produced an output energy of 21 J at 10 Hz from the side-pumped Nd:YAG zigzag slab amplifiers [5]. Recently, our group has demonstrated the excellent scaling performance of joule-level Nd:YAG and Nd:LuAG AM amplifier [6,7] with the optical-optical efficiency higher than 20%, taking advantage of the round-trip energy extraction of AM geometry. However, further scaling of the Nd-doped AM amplifier much beyond 10 joules necessitates specific measures to deal with the drawback of Nd-doped materials compared with Yb3+ counterparts in terms of lower Stokes efficiency (more heat), shorter upper level lifetime (higher pump peak power required) and lower saturation fluence (higher gain and thus higher depumping loss). In this letter, a novel concept called as distributed active mirror amplifier chain (DAMAC) is proposed, in which the gain and thermal deposition are dispersed over a large number (10~20) of gain modules as well as gain sheets, in contrast to the adoption of quite few number (1~2) of gain modules in previous high-energy laser systems (see Table 1).
Table 1. Parameters comparison of nanosecond laser systems in repetitive operation with output energy beyond 10 joules (RT, room-temperature operation; CR, cryogenically-cooled operation; PA, preamplifier; MA, main amplifier)
The DAMAC configuration has three distinct advantages. First and foremost, the required pump energy (gain) for the amplifier stage is distributed over multiple gain modules, so that the amplified spontaneous emission (ASE) effect within each module can be suppressed at a low level, and simultaneously the pump intensity needed for each module agrees well with the available emitting intensity of diode array (2~4 kW/cm2), having no necessity of coupling optics. Besides, as indicated in Table 1, it can decrease by about 80% the required pump peak power for each array, largely reducing the complexity of achieving pump spectral accordance (both in central wavelength and spectral linewidth) for each bar of the array. Second, the heat accumulation is distributed over multiple gain sheets, thus mitigating the thermal effects and relaxing the requirements on heat removal of the gain module, which is extremely beneficial for high-repetition-rate operation at room temperature. Third, it adds degree-of-freedom and flexibility of optimizing the extraction efficiency, which allows independent tuning of each gain module in spectral (cooling temperature of pump source), temporal (extraction timing relative to the pump pulse) and spatial (pump/laser mode matching) domains. Furthermore, a differentiated design on crystal packaging and cooling structure for each AM in the chain is feasible to compensate the wavefront deformation introduced by every single piece of partially pumped AM [8].
In this paper, the scaling performance of 10-ns pulses by DAMAC is verified by the experimental amplification to 12 J at 1-10 Hz from a diode-pumped room-temperature Nd:YAG DAMAC system. Efficient power scaling at the optical-optical efficiency of 20.6% was achieved by suppressing the transverse parasitic oscillation with ASE absorbers. To the best of our knowledge, this is the first demonstration of diode-pumped nanosecond Nd-doped active-mirror laser with the pulse energy beyond 10 joules. The DAMAC concept verified holds the promise of scaling the energy to 50 J level and higher by adding more pieces of AM in the chain.
2. Experimental setup
Figure 1 shows the schematic diagram of the DAMAC system.
Fig. 1 Experimental layout of diode-pumped DAMAC. FI, Faraday isolator; HR, high-reflection mirror; PBS, polarization beam splitter; QWP, quarter-wave plate; HWP, half-wave plate; TS, telescope; SA, serrated aperture. Inset: detailed layout of the single gain module.
2.1 Front end
The front-end source consists of an electro-optically Q-switched oscillator in the unstable cavity with a variable reflectivity output coupler, generating up to 125 mJ at 1064 nm at a repetition rate of 10 Hz, with the pulse duration of 13 ns (FWHM) and the beam quality factor of M2<1.4. A single Nd:YAG rod, instead of two rods as in [6], with the diameter of 8 mm and the length of 145 mm is used as the gain medium. The oscillating beam output with linear polarization is directed to a Faraday isolator and then a number of AM amplifier stages.
2.2 Booster amplifier
The double-pass booster amplifier has the same setup as introduced in [6], consisting of four 0.6 at. % doped Nd:YAG slabs, having the dimension of 30 × 20 × 8 mm3 and a clear aperture size of 20 × 14 mm2 for the incidence angle of 45°. Each Nd:YAG AM in the booster amplifier is pumped by a 30-bar laser diode stack, with the maximum peak power of 5.6 kW and a maximum peak intensity of 4 kW/cm2 at the pump surface. The pump light from each bar is collimated in fast axis by a microlens, resulting in a divergence angle of 5° and 8° along fast and slow axis respectively. The scaling performance of booster amplifier can be found in [6] and thus is not described in detail here for clarity. The beam output with the pulse energy of 2.2 J from the booster amplifier is expanded to the diameter of 45 mm and then apodized by a rectangular serrated aperture of 32 mm wide and 32 mm high, before entering the main amplification stage of DAMAC.
2.3 Main amplifier
The Nd:YAG slab for each AM in the main amplifier is 0.6 at. % doped, having a thickness of 8 mm and a transverse aperture size of 60 × 40 mm2 for light travelling in the incidence angle of 45°. The extracting beam goes through the crystal twice for every passage in an active-mirror manner, since the front surface of slab is anti-reflection (AR) coated at 1064 nm and at 808 nm relatively to air, while the back surface is high-reflection (HR) coated at 1064 nm and AR coated at 808 nm relatively to water. After the first passage through four pieces of AM, the beam polarization direction is changed from horizontal to vertical, and thus the beam output after the second pass exits the amplifier by the polarized beam splitter. The slab is elastically mounted by a flexible supporter with grooves of 2-mm depth to alleviate the external constraint, as introduced and verified in our previous work [6,7,9]. The passive loss of the double-pass main amplifier is measured as 24%.
For each AM, the pump source is an 808-nm laser diode array having 201 bars arranged in three rows and an emitting area of 33 × 35 mm2, as shown in Fig. 2. The diode array delivers a total of 45.1 kW peak power while the pump width can be controlled by a power supply unit with the maximum pump duration of 300 μs, corresponding to a maximum pump energy of 13.5 J for each gain module. Each diode bar, without any micro-lens shaping, has a divergence angle of 8 degrees and 40 degrees for slow and fast axis respectively. The pump light emitting from the array, without any macro-lens shaping either, illuminates the pump surface of AM after passing through quartz window and the cooling layer. 76% of the pump energy is absorbed by the AM through the single-pass geometry at the coolant temperature of 26.5°C, while a plateau distribution with the pump intensity of 2.2 kW/cm2 is obtained at the pump surface over the pump area of 50 × 36 mm2 at full power.
Each AM is cooled by the flowing water over the slab back surface. A flow homogenizer consisting of multiple guide plates is carefully designed and adopted for uniform cooling of the pump surface of AM. The simulated temperature distribution at full pump power at 10 Hz is obtained by the commercial software Fluent, by combining and solving the Navier-Stokes equations and the thermal conduction equation of AM. Specifically a large-eddy-simulation (LES) model is adopted for turbulent flow cooling [10]. The coolant flow in a 1-mm-thick channel was controlled in the turbulent state, having a flow rate of 5 m/s that corresponds to a heat transfer coefficient of 3.8 × 104 W/(m2K), which is lower than 5 × 104 W/(m2K) as reported in our previous work of 2 J system [6], because the applied pump intensity at the pump surface of AM is lowered from 3.5~4 kW/cm2 in 2 J system to 2.2 kW/cm2 for 12 J system in this paper. Figure 3 indicates that the maximum temperature of 305.0 K occurs at the pump central region while the temperature difference over the beam extracting region is about 1.5 K. Figure 4 shows the actual temperature distribution at the front surface of AM as measured by a DALI thermal imaging camera, with the average temperature of 26.4°C, 28.2°C and 30.2 °C of the imaging area individually under the pump energy of zero, 7.9 J and 13.5 J at 10 Hz respectively. The measured temperature distribution at full pump power and 10 Hz shown in Fig. 4(c) agrees well with the simulated one in Fig. 3. The small temperature rises and good temperature uniformity suggest an outstanding performance of thermal management.
Fig. 4 Measured temperature distribution at the front surface of Nd:YAG AM under different pump condition.
3. Experimental results and discussion
Initially, a bare Nd:YAG slab was used in each gain module without the adoption of ASE absorbers. As shown in Fig. 5(a), the output energy was clamped around 4 J for the single-pass DAMAC at the pump energy of 37.5 J at 10 Hz, with an optical-optical efficiency of 6.9%, while it was clamped around 6 J for the double-pass DAMAC at same pump power, with the optical-optical efficiency of 11.2%. Further increase of pump energy beyond 37.5 J did not boost the output energy, demonstrating the existence of serious parasitic oscillation. To suppress the parasitic oscillation, four ASE absorbers are respectively located along the four side surfaces of slab with the gap of 3 mm, as shown in Fig. 6(a). The absorber material is neutral grey colored glass NS50 provided by Guoguang Optical Glass Co. Ltd, having the refractive index of 1.502 and the absorption coefficient of 6.58 cm−1 at 1064 nm (over 5 cm−1 for the wavelength of 800-1100 nm). Figure 6(b) describes the working principle of the ASE absorber. The 3-mm gap between the slab and the absorber is filled with flowing deionized water. The ASE absorbers are cooled by two largest surfaces through independent channels. The refractive index of the Nd:YAG slab and the deionized water is 1.82 and 1.33, respectively.
Fig. 5 Scaling performance of single-pass and double-pass DAMAC: (a) comparison between the cases with and without the adoption of ASE absorbers; (b) comparison between the predicted and measured results.
Consider the ASE rays that generated within the gain medium and incident at interface A (one of the four side surfaces of slab) between the slab and its adjacent cooling water. The critical angle at the interface A can be calculated as θc = arcsin(1.33/1.82) = 46.95°. On one hand, for ASE rays with the incident angle of θi<θc at interface A, they are all refracted through a water layer, the ASE absorber and the other water layer, with no chance of having total internal reflection at interface B and C. Especially for those ASE rays that reflected by interface D and refracted all the way back into the slab, they experience round-trip absorption within the absorber, leading to the portion of the original ASE energy that can return reaches the maximum value as about 1.2% for the case of θi = 0 and is 0.65% for the case of θi = 30°, assuming the reflectivity of the stainless steel as 60%. On the other hand, for ASE rays with θi>θc at interface A, they are reflected by the slab surface once or twice, and then reach another slab surface perpendicular to interface A, with the incident angle of θi’ = 90°-θi<θc, thus those ASE rays would definitely be refracted out of the slab and guided through the absorber. Furthermore, for ASE rays traveling between the two non-absorbing surfaces of slab, they can also be refracted out of the slab and get absorbed after they propagate close to the surface edge and experience reflection on the absorbing surfaces. Therefore, the absorber efficiently eliminates the parasitic oscillation and the ASE rays trapped in the gain medium by total internal reflection can be neglected.
A great number of literatures have discussed the ASE effect in slab amplifiers. Here we adopt the geometrically dependent gain coefficient (GDGC) model proposed in [11] to describe the ASE behavior occurred in our experiment. According to GDGC model, when the excitation length of the sample is small, the z-dependent gain can be written as
where z is the excitation length. The small signal gain versus z was measured using the ASE absorber structure. A Q-switched seed beam with the fluence of 5 × 10−4 J/cm2 was used as the probe signal. By fitting the experimental data, m and rL was achieved under different pump conditions that we have m, rL and zth as 0.19 cm−1, −0.865 and 2.9 mm respectively for the full pump power. Then the ASE intensity was calculated usingwhere IASE is the ASE intensity, zth is the ASE threshold length where ASE could occur only in the case of z>zth. zth was measured by comparing amplified data in the conditions with and without the ASE absorbers, γ(zth) is the solid angle, Φ is the medium quantum yield, defined by the ratio of the upper state lifetime τf (230 μs for Nd:YAG) and the spontaneous lifetime τs (260 μs for Nd:YAG [12]), and Isvo is the saturated intensity as defined bywhere hν0 is the photon energy, σ is the stimulated emission cross section, τf is the upper state lifetime.The output characteristics can be calculated using a loop iteration method. For each time grid, the saturated gain was achieved using
where Ilaser(t) is the input laser intensity for each grid. Note that Ilaser(0) stands for the input pulse intensity from the oscillator, c is the light velocity and n is the refractive index of the gain medium. The calculated g was used for the Frantz–Nodvik formula. Thus, the output energy could be achieved. The calculated curves of output energy versus pump energy at the full pump power for the cases with and without ASE absorber are illustrated in Fig. 7, which agrees well with the experiment results.Fig. 7 Simulated scaling performance of single gain module with and without ASE absorber, varying with different pump energy.
With the ASE absorbers mounted, the double-pass DAMAC successfully produced an output energy of 12.2 J at the repetition rate of 10 Hz with the seed energy of 1.0 J and the total pump energy of 54.1 J in the main amplification stage, corresponding to an optical-optical efficiency of 20.6%, while it was 9.17 J and 14.7% for single-pass case. It is evident in Fig. 5(a) that the adoption of ASE absorber prevents at high pump intensity the drop of slope efficiency by cutting off the parasitic oscillation, realizing a linear increase of output pulse energy versus the pump energy. In addition, the improved structure dramatically raises the energy extraction long before the threshold of parasitic oscillation is reached as in the non-absorber case (e.g. an increase by 21.0% at 8.4 J of pump energy and by 43.3% at 31.8 J of pump energy), which indicates that the ASE absorber serves to suppress the consumption of inverse population by the ASE effect from the very beginning. All the results reported hereinafter are based on the structure with ASE absorbers. Figure 5(b) describes that the experimental results of both the single-pass and double-pass DAMAC are in good agreement with the calculated curves by our modeling developed on the basis of the Frantz–Nodvik formula, specifically taking into account the influence of overlapped V-shaped light path within AM on energy extraction and mode matching, as well as the temporal overlap between the beams that propagate forwards and backwards in the amplifier stage.
Figure 8 shows the burn pattern of the DAMAC output at 10 J and 10 Hz, which is measured using a laser alignment paper provided by a local manufacturer Leijie technology. Comparison between Fig. 8 and Fig. 2 indicates that the spatial structure in the pump profile along the slow axis (vertical direction) with two valleys at the gap between neighboring rows of bars can still be observed at the near-field profile of amplified beam, although it has been fairly smeared out through the amplification process. Double-pass configuration can improve the beam uniformity than single-pass setup, since the beams in first and second passage may have slight vertical shift with each other and consequently smooth out the spatial structure. For the horizontal direction, it has much better pump uniformity in the first place, while the folding of extracting beam in horizontal plane within the active mirrors further compensate the disparities. In addition, the beam profile has modulated fringes due to Fresnel diffraction from the edge of Nd:YAG rods in the unstable resonator. The beam quality of amplified output can be further enhanced by aberration compensation using a well-designed lens group and the improvement of the seed beam spatial profile.
Figure 9 describes the measured dependence of the output energy on the input seed energy. For the main amplifier with only one piece of AM, there is always a linear increase of output energy at the pump energy of 14 J for single-pass or double-pass amplification, as illustrated in Fig. 9(a), with the extraction energy of 0.5 J and 0.7 J respectively at the seed energy of 2.8 J. However, the gain saturation occurs at full pump power for double-pass one-piece amplifier and becomes so strong with growing joule-level seed injection that the maximum output energy (4.7 J) is almost equal to that of the single-pass case. Figure 9(b) shows the performance of double-pass main amplifier with four pieces of AM, indicating strong gain saturation at the pump energy of 32 J and 54 J. The output dependence on pulse repetition rate is described in Fig. 10 that the output energy at the injected seed energy of 800 mJ and full pump power keeps constant at 10.2 J at the repetition rates from 1 to 10 Hz and slightly decreased to 9.1 J at 15 Hz, which suggests weak thermal effects. Besides, a maximum output energy of 12.2 J at 1-10 Hz and 10.1 J at 15 Hz have been realized at the seed energy of 1.1 J.
Fig. 9 Output energy versus seed energy with different pump power: (a) single- and double-pass main amplifier with one piece of AM; (b) double-pass main amplifier with all four pieces of AM.
For the design of 50 J system, the 10-50 J stage consists of 10-12 AM gain modules, each of which is exactly the same as that for the main amplifier of 12 J system described in this paper. The main difference is that the AM amplifiers following the 10 J stage all adopt single-pass geometry rather than double-pass setup, otherwise the passive loss of double passage may outnumber the saturated gain, since the saturation degree is already high for double-pass amplifier at the 10 J level according to Fig. 9. A critical point is to raise the pump absorption efficiency from current 76% to almost 100%, under which condition our model predicts that the DAMAC with 4 pieces of double-pass AM and 12 pieces of single-pass AM would boost the joule-level seed to 52.7 J at 10 Hz with the total pump energy of 216 J, with the optical-optical efficiency of 23.7%, as shown in Fig. 11. Furthermore, by improving the single-pass passive loss from current level of about 3% per piece to 2.5% per piece, the DAMAC would produce the ultimate output energy of 56.3 J with the optical-optical efficiency of 25.4%, while 10 instead of 12 pieces of single-pass AM can achieve the output energy beyond 50 J. The average output fluence of the 50 J output with the beam size of 32 × 32 mm2 is 4.9 J/cm2, which is fairly below the damage threshold of optical elements. Since the capability of power scaling, thermal management and ASE suppression has been verified for single AM gain module as well as the DAMAC at 10 J level, while the theoretical modeling has also been proved reasonable, the 50 J chain having the same extracting beam size and a series of same gain modules as 12 J system is expected to be realized with the DAMAC concept.
Fig. 11 Predicted output curves of DAMAC at 50 J level, where δs is the single-pass passive loss per piece of AM, and βa is the pump absorption efficiency of each AM gain module.
4. Conclusion
In this letter, a novel concept of DAMAC for room-temperature operation of high energy nanosecond Nd-doped laser has been proposed, in which the gain and thermal deposition are distributed over a number of gain modules and gain sheets. With a total of 8 pieces of Nd:YAG AM in the booster and main amplifier, an output energy of 12.2 J from 1 to 10 Hz as well as 10.1 J at 15 Hz has been obtained with the pulse width of 10 ns, which, to the best of our knowledge, is the first demonstration of diode-pumped nanosecond Nd-doped active-mirror laser beyond 10 joules. In addition, optimized ASE absorbers have been implemented based on developed theoretical modeling, which well proves the capability in suppressing transverse parasitic oscillation and enhances the optical-optical efficiency up to 20.6% at the full power. The experiment results demonstrate the feasibility and validity of the Nd-doped DAMAC configuration in terms of gain optimization, thermal management, and multi-joule stable operation. Power scaling to 50 J level and higher with DAMAC is under development, by adding 10-12 more pieces of AM in the current apparatus, while the optical-optical efficiency around 25% is expected by carefully reducing the passive loss and raising the pump absorption efficiency from 76% to almost 100%.
Funding
The National Key Research and Development Program of China (Grant No. 2017YFB1104500); Beijing Natural Science Foundation (4172030); Tsinghua University Initiative Scientific Research Program, China (Grant No. 2014z21035).
References and links
1. A. Bayramian, P. Armstrong, E. Ault, R. Beach, C. Bibeau, J. Caird, R. Campbell, B. Chai, J. Dawson, C. Ebbers, A. Erlandson, Y. Fei, B. Freitas, R. Kent, Z. Liao, T. Ladran, J. Menapace, B. Molander, S. Payne, N. Peterson, M. Randles, K. Schaffers, S. Sutton, J. Tassano, S. Telford, and E. Utterback, “The Mercury project: a high average power, gas-cooled laser for inertial fusion energy development,” Fus. Sci. Technol. 52(3), 383–387 (2007). [CrossRef]
2. P. Mason, M. Divoky, K. Ertel, J. Pilar, T. Butcher, M. Hanus, S. Banerjee, J. Phillips, J. Smith, M. Vido, A. Lucianetti, C. Hernandez-Gomez, C. Edwards, T. Mocek, and J. Collier, “Kilowatt average power 100 J-level diode pumped solid state laser,” Optica 4(4), 438–439 (2017). [CrossRef]
3. T. Gonçalvès-Novo, D. Albach, B. Vincent, M. Arzakantsyan, and J. C. Chanteloup, “14 J/2 Hz Yb3+:YAG diode pumped solid state laser chain,” Opt. Express 21(1), 855–866 (2013). [CrossRef] [PubMed]
4. T. Gonçalvès-Novo, B. Vincent, and J. Chanteloup, “From 10 to 30 joules with the Lucia laser system: update on current performance and cryogenic amplifier development,” in Advanced Solid-State Lasers Congress, OSA Technical Digest (online) (OSA, 2013), paper ATu3A.19. [CrossRef]
5. R. Yasuhara, T. Kawashima, T. Sekine, T. Kurita, T. Ikegawa, O. Matsumoto, M. Miyamoto, H. Kan, H. Yoshida, J. Kawanaka, M. Nakatsuka, N. Miyanaga, Y. Izawa, and T. Kanabe, “213 W average power of 2.4 GW pulsed thermally controlled Nd:glass zigzag slab laser with a stimulated Brillouin scattering mirror,” Opt. Lett. 33(15), 1711–1713 (2008). [CrossRef] [PubMed]
6. X. Fu, Q. Liu, P. Li, Z. Sui, T. Liu, and M. Gong, “High-efficiency 2 J, 20 Hz diode-pumped Nd:YAG active-mirror MOPA system,” Appl. Phys. Express 8, 092702 (2015). [CrossRef]
7. Q. Liu, M. Gong, T. Liu, Z. Sui, and X. Fu, “Efficient sub-joule energy extraction from a diode-pumped Nd:LuAG amplifier seeded by a Nd:YAG laser,” Opt. Lett. 41(22), 5322–5325 (2016). [CrossRef] [PubMed]
8. D. Albach, G. LeTouzé, and J. C. Chanteloup, “Deformation of partially pumped active mirrors for high average-power diode-pumped solid-state lasers,” Opt. Express 19(9), 8413–8422 (2011). [CrossRef] [PubMed]
9. X. Fu, P. Li, Q. Liu, and M. Gong, “3kW liquid-cooled elastically-supported Nd:YAG multi-slab CW laser resonator,” Opt. Express 22(15), 18421–18432 (2014). [CrossRef] [PubMed]
10. P. Li, X. Fu, Q. Liu, and M. Gong, “Analysis of wavefront aberration induced by turbulent flow filed in liquid-convection-cooled disk laser,” J. Opt. Soc. Am. B 30(8), 2161–2167 (2013). [CrossRef]
11. A. Hariri and S. Sarikhani, “Theoretical application of z-dependent gain coefficient to describe amplified spontaneous emission,” Opt. Lett. 37(6), 1127–1129 (2012). [CrossRef] [PubMed]
12. D. Brown, “Heat, fluorescence, and stimulated-emission power densities and fractions in Nd:YAG,” IEEE J. Quantum Electron. 34(3), 560–572 (1998). [CrossRef]
