2
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. I , NO. I , APRIL 1995
`
`Ytterbium-Doped Silica Fiber Lasers: Versatile
`Sources for the 1-1.2 pm Region
`
`H. M. Pask, Robert J. Carman, David C. Hanna, Anne C. Tropper,
`Colin J. Mackechnie, Paul R. Barber, and Judith M. Dawes
`
`Invited Paper
`
`Abstract- Ytterbium-doped silica fibers exhibit very broad
`absorption and emission bands, from -800 nm to -1064 nm
`for absorption and -970 nm to -1200 nm for emission. The
`simplicity of the level structure provides freedom from un-
`wanted processes such as excited state absorption, multiphonon
`nonradiative decay, and concentration quenching. These fiber
`lasers therefore offer a very efficient and convenient means
`of wavelength conversion from a wide variety of pump lasers,
`including AlGaAs and InCaAs diodes and Nd:YAG lasers. Effi-
`cient operation with narrow linewidth at any wavelength in the
`emission range can be conveniently achieved using fiber gratings.
`A wide range of application for these sources can be anticipated.
`In this paper, the capabilities of this versatile source are reviewed.
`Analytical procedures and numerical data are presented to enable
`design choices to be made for the wide range of operating
`conditions.
`
`I. INTRODUCTION
`FTER the first report in 1962 of laser action in Yb”-
`
`A doped silicate glass [ I ] . Yb3+ has, until recently, at-
`
`tracted relatively little interest as a laser-active ion. It has been
`overshadowed by the Nd3+ ion with its important advantage of
`a four level transition, whereas Yb3& has only three level and
`quasi-three level transitions. In fact, the most important role
`of the Yb3+ ion has so far been as a sensitizer ion, absorbing
`pump photons over a wide spectral range and then transferring
`the excitation to an acceptor ion, such as Er”, which then acts
`as the laser-active ion 121. [31.
`More recently, interest has been shown in Yb3+ as a laser
`ion, in the form of Yb”-doped
`silica and fluoride fiber lasers
`141-[8], and Yb”-doped YAG [9], [IO]. There are several
`reasons for this growth of interest. As shown in Fig. l(a), the
`Yb3+ energy level structure is a simple one, consisting of two
`manifolds; the ground manifold 2F7/2 (with four Stark levels
`labeled (a)-(d) in the figure) and a well-separated excited
`manifold ’Fj/2 (with three Stark levels labeled (e)-(g) in the
`
`nonradiative decay via multiphonon emission from F 5 p ,
`even in a host of high phonon energy such as silica, and also
`precludes concentration quenching. These features contribute
`to the high efficiency of operation that can be achieved in
`Yb3+ lasers, as does the closeness of the pump and laser
`wavelengths. In fact, this energy defect, which leads to heating
`of the host, is a factor of -3 smaller for Yb:YAG compared
`to Nd:YAG (pumped at 800 nm and lasing at 1064 nm). This
`reduced thermal burden is a motivating interest for Yb:YAG.
`Finally, the Yb3+ spectrum is rather broad both in absorption
`and emission, and particularly so in a germanosilicate host, as
`shown in Fig. I(b). The broad absorption spectrum allows a
`wide choice of pump wavelengths. In the form of a fiber, which
`allows even very weak absorption to be exploited, pumping
`can extend from 800 nm out to 1064 nm. Similarly, for a fiber,
`laser operation can be made to extend well into the weak wings
`of the emission, provided sufficient frequency discrimination
`can be introduced to suppress lasing at the peaks of the
`emission profile. The impressive progress in development
`of fiber gratings [ l l ] has made this matter of frequency
`discrimination very straightforward and practical, so one can
`now contemplate the prospect of narrow linewidth operation at
`any discrete wavelength between -975 and -1200 nm, with
`some degree of tunability (-l)% available by stretching or
`temperature tuning the grating. This range covers a number
`of wavelengths needed for specific applications, and these
`can now be generated very conveniently from an Yb3+-doped
`silica fiber equipped with appropriate gratings. Examples in-
`clude 1020 nm for pumping 1300-nm fiber amplifiers [ 121 and
`upconversion lasers based on Pr3+-doped ZBLAN [ 131-1 151,
`1140 nm for pumping Tm”-doped ZBLAN upconversion
`lasers [ 161, [ 171, and 1083 nm for optical pumping of He [ 181.
`Given the wide range of different operating characteristics
`that might be required for various applications, and the variety
`of ways (e.g., different pump wavelengths) that could be
`used to achieve these characteristics, there is a need for a
`comprehensive discussion of the capabilities of Yb3+ fiber
`lasers, giving quantitative design procedures. Such has been
`the aim of this paper. A general discussion of Yb laser char-
`acteristics is presented in Section 11. In Section 111, analytical
`procedures and numerical data are presented to enable design
`choices to be made. Such calculations depend heavily on the
`availability of accurate absorption and emission cross section
`data, and the relevant spectroscopy is described in Section IV.
`IO77-260X/95$04.00 0 199.5 IEEE
`
`figure), - 10000 cm-’ above the ground level. Thus there is no
`
`excited state absorption at either pump or laser wavelengths.
`The large energy gap between ’FjI2 and 2F7/z precludes
`Manuscript received August 5 . 1994; revised October 10. 1994. This work
`was supported by the Science and Engineering Research Council (SERC) and
`by the RACE 11 GAIN program.
`H. M. Pask, D. C. Hanna A. C. Tropper, C. J. Mackechnie, P. R. Barber,
`and J. M. Dawes are with the Optoelectronics Research Centre, University of
`Southampton, HANTS SO17 IBJ, U.K.
`R. J . Carman is with the Centre for Lasers and Applications. Macquarie
`University, N.S.W. 2109, Australia.
`IEEE Log Number 9409720.
`
`ASML 1229
`
`

`
`PASK er U/.: YTTERBIUM-DOPED SILICA FIBER LASERS
`
`3
`
`2
`F
`5/2
`
`2
`F
`7/2
`
`(a)
`
`- 1
`crn
`9 11630
`f 11000
`e 10260
`
`d 1490
`c 1060
`b 600
`a
`0
`
`
`
`,?-)
`
`a r . e e - ; : -
`(bl
`( a ) The Yb'- energy level structure, consisting of two manifolds, the
`Fig. 1.
`ground manifold -?F7;? (with four Stark levels, labeled (at(d)), and a well
`heparated excited manifold ' Fi12 (with three Stark levels labeled (e)-(g)).
`Approximate energies in wavenumbers above ground energy are indicated.
`(b) Absorption and emission cross sections for a germanosilicate host. The
`principal features of the spectra have been labeled A-E, and are discussed
`in the text.
`
`In Sections V-VIII. some specific examples are drawn from
`both experimental and modeling studies of Yb"-doped
`silica
`laser action both to illustrate the agreement between modeling
`and experiment, and to illustrate the versatility and potential
`of this laser system.
`
`11. GENERAL DISCUSSION
`The most obvious features of the absorption and emission
`spectra in Fig. I(b) have been labeled (A)-(E). The narrow
`line at 975 nm (A) in both absorption and emission, cor-
`responds to transitions between the lowest Stark levels in
`each manifold. The absorption peak at shorter wavelengths
`(B) corresponds to transitions from level a to f and g, while
`the long wavelength shoulder in the absorption spectrum (C)
`corresponds to transitions from level b. The weakness of this
`shoulder is a consequence of the much smaller population of
`this level (-6% of level a at room temperature, as calculated
`from the Boltzmann factor). Despite the weakness of this
`shoulder, it plays a significant role. First, it provides the means
`for pumping using a Nd:YAG or Nd:YLF laser (at 1064 or
`1047 nm respectively) [7]. Second, it is a cause of reabsorption
`
`loss, having a significant effect on threshold for lasing at
`wavelengths within the dip in the emission spectrum.
`Laser action on the narrow peak at 975 nm (A), where
`emission is into the lowest Stark level, is truly three-level in
`character. The second peak in the emission spectrum (D), with
`its tail extending out to 1200 nm, corresponds to transitions
`from level e to b, c, and d. Laser action on these transitions
`becomes nearly four level in character at longer wavelengths,
`as the emission is into essentially empty levels c and d.
`Transitions from level f are also evident (E), although very
`weak due to the small thermal population of the level, and
`laser action has not been observed on transitions from that
`level.
`In general, when generation of a particular laser emission
`wavelength is required, there is a wide choice of possible
`to -1064 nm, and
`pump wavelengths ranging from -800
`therefore, a choice of pump laser sources including AlGaAs
`and InGaAs diodes, Titanium sapphire lasers, Nd:YLF lasers
`and Nd:YAG lasers. As will be demonstrated in subsequent
`sections, extremely efficient laser action can be achieved using
`any of these sources. There are, however, several consid-
`erations other than the availability of sources that can go
`into the choice. First, one can only generate gain and hence
`lasing at longer wavelengths than the pump, so for example,
`1020-nm operation cannot be achieved with a 1064-nm pump,
`but can with 975 nm pumping into the sharp absorption line.
`Second, pumping at 975-nm accesses the largest absorption
`cross section, so this is particularly appropriate where the
`shortest fiber length is required as for example in a single
`frequency laser. It is also appropriate for cladding-pumping,
`enabling the fiber length to be kept down to reasonable values,
`since the cladding-pumping geometry involves a scaling up
`of the absorption length, as described in Section VII. When
`pumping at 975 nm, it may be important to consider the
`effects of amplified spontaneous emission (ASE), as discussed
`below. Third, the slope efficiency with respect to absorbed
`pump power of Yb3+ lasers is usually dominated by the ratio
`of laser to pump photon energies. It is therefore possible
`to significantly enhance the conversion efficiency (e.g., by
`up to -20%
`for laser action at 1140nm) by pumping close
`to the laser wavelength (e.g., with a Nd:YLF or Nd:YAG
`laser). Fourth, the pump power requirements are a function
`of the absorption and emission cross sections at the pump
`wavelength.
`The critical pump power PCp required to achieve a gain
`coefficient of zero at a particular point in the fiber (i.e., to
`reach transparency for the signalbaser wavelength), is given
`
`Here we have assumed no background loss in the fiber. This
`expression can be readily derived from (2)-(5)
`in the next
`section, when the gain coefficient, g ( z ) , is set equal to zero
`and the pump quantum efficiency is assumed to equal unity.
`A . h. nil. and r denote the core area, Planck's constant, pump
`laser frequency, and upper level lifetime respectively,
`and
`
`

`
`4
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. I , NO. I , APRIL 1995
`
`i\i]
`
`100
`loo0
`
`. E40
`
`?
`
`900
`
`950
`
`1100
`1050
`1000
`Laaer Karelength (urn)
`
`1150
`
`1200
`
`Fig. 2. Calculated values of the critical pump powers (for a 3.75pm diameter
`fiber) and intensities required to produce net gain (i.e.. to reach transparency
`for the signal wavelength) in Yb3+-doped silica fiber, plotted as a function
`of signal (lasing) wavelength and for various pump wavelengths.
`
`o,~, the emission and absorption cross sections at the laser
`wavelength, and nep and nap, the emission and absorption
`cross sections at the pump wavelength, respectively. The
`critical power is independent of Yb3+ concentration, and
`whatever the fiber length or dopant concentration, for most
`efficient use of pump power, either at or above threshold, the
`power emerging from the fiber end (neglecting the Fresnel
`reflection) would, in fact, be equal to this critical power. Fig. 2
`shows the critical power and also the corresponding intensities
`(for a typical fiber with diameter 3.75-pm) plotted as a function
`of lasing wavelength for various pump wavelengths of interest.
`This figure is useful in making design choices, as it can be seen
`that the critical powers are quite high for some combinations
`of pump and laser wavelengths. For example, to achieve lasing
`at 975 nm, it would be preferable to use a 900-nm pump source
`rather than an 840-nm pump source.
`To further illustrate the dependence of lasing wavelength on
`pump conditions, we show the calculated gain spectrum for the
`specific case of pumping at 840 nm in Fig. 3(a). For low pump
`powers (5 mW) in the illustrated example, net gain first appears
`at long wavelengths -1 100 nm for which there is negligible
`reabsorption from population in the lower level. Then, as pump
`power increases, the gain maximum moves progressively to
`shorter wavelengths, with the pure three level transition even-
`tually having the dominant gain. The oscillation wavelengths
`under free-running conditions, i.e., without any wavelength
`selection applied, will correspond to the wavelength which has
`peak gain at the threshold for oscillation. With progressively
`harder pumping above threshold, assuming pure homogeneous
`broadening, the emission wavelength should remain clamped
`at its threshold value.
`Some degree of control over the free-running wavelength
`can be achieved by changing the Q of the resonator, or by
`changing the fiber length. Changing the Q of the resonator
`changes the required gain for threshold, and therefore, the
`wavelength of peak gain, as can be seen in Fig. 3(a). Fig. 3(b)
`shows the gain under the same conditions as in Fig. 3(a), but
`with a fiber which is 20 times longer. The gain maxima are
`
`Wavelength (urn)
`(a)
`
`t
`
`. . . . . . . . . . .
`
`1150
`
`1100
`1050
`1000
`Karelength (urn)
`(b)
`Fig. 3. Calculated gain spectrum for the specific case of pumping at 840 nm
`for two fiber lengths (a) lm, (b) 20 m. (fiber diameter 3.75 I‘m, NA-0.17,
`[Yb”+]-550 ppm). Curves are labeled with the launched pump power in
`milliwatts.
`
`1200
`
`moved to longer wavelengths as a result of reabsorption in the
`extra length of fiber (e.g., compare the 30 mW curves in both
`figures). Thus oscillation at shorter wavelengths, e.g., at the
`975 nm or 1040 nm gain peaks, can be suppressed by using
`a longer fiber to introduce reabsorption at these wavelengths.
`Alternatively one can use a longer pump wavelength (e.g.,
`1064 nm) than 975 nm or 1040 nm, thus ensuring no gain at
`these emission peaks. In this case a longer fiber is also needed,
`not to suppress the gain at 975 nm, but to ensure adequate
`absorption of the pump. While these two approaches both
`lead to the suppression of 975-nm gain, there is a difference
`which could become important under some circumstances.
`This is where short wavelength pumping is used and a very
`high gain at 975 nm is established at the input end of the
`fiber, although with net gain at 975 nm over the entire fiber
`length being suppressed by reabsorption further down the fiber.
`Under appropriate conditions strong ASE at 975 nm or even at
`N 1040 nm could occur, and while the ASE travelling down the
`fiber would be reabsorbed and simply lead to a redistribution of
`the gain, the ASE that escapes from the input end would result
`in a loss of pumping efficiency and hence higher thresholds.
`
`

`
`PASK er U / . : YTTERBIUM-DOPED SILICA FIBER LASERS
`
`5
`
`where g ( z ) = N2(z)oel - N1(z)oul. The value of N2(z) and
`hence N l ( z ) = N - N2(z), is determined from the pump
`power Pp(z) at that location. Pp(z) obeys the equation
`
`Here we have assumed that background loss can be ignored.
`Our experimental results confirm that for fiber lengths up to
`-100 m the background loss (estimated at < 10 dB/km) played
`a negligible role.
`To solve for P,(,z) we make use of (1) and of the relation
`
`(4)
`where r is the lifetime of the upper level and 4, is the pumping
`quantum efficiency (in this case, for Yb3+, 4, -1).
`From (l), (3), and (4) we obtain the result
`
`where Pp(0) is the input pump power and P, is the pump
`saturation power (i.e., the power that reduces the absorption
`coefficient by a factor of 2).
`P, is given by
`
`P, =
`
`hv, A
`(a?, + o u p ) 7 4,.
`Thus, p,(z) can be calculated froln (5).
`The single-pass gain exponent (2) for a length 1 of fiber is
`then given by
`
`(6)
`
`I
`
`Data such as those presented in Fig. 3 are also useful
`in indicating the degree of frequency discrimination needed
`to enforce oscillation at wavelengths away from the free-
`running wavelength. Thus, from the data in Fig. 3(a) it is
`seen that to achieve oscillation at 1020 nm for a threshold
`pump power of 50 mW it is necessary to introduce -5 dB
`of discrimination against 1040 nm to suppress its oscillation.
`Under higher gain conditions, i.e., for higher pump powers,
`the degree of frequency discrimination required can be seen
`to increase. As a general point, it should be noted that if the
`peak gain becomes high enough to allow strong ASE (-30-
`40 dB single-pass gain), then frequency discrimination via
`end reflectors, however great, will become ineffective. It is
`therefore clear that when oscillation in the low gain region
`of the emission spectrum is required (at long wavelengths
`or in the dip between the emission peaks) it is necessary
`to use a resonator having low loss at the desired oscillation
`wavelength. Fiber gratings are particularly well-suited to this
`need as they can provide both a high frequency discrimination
`and a very high reflectivity and, when written directly into the
`doped fiber, introduce negligible insertion loss. For example,
`with appropriate gratings we have achieved oscillation out
`to 1180 nm, and with further optimization and reduction
`of losses this could, in principle. be extended somewhat
`further.
`This preceding discussion has indicated some of the main
`features that need to be considered when designing an Yb
`fiber laser to operate under specific conditions. particularly
`It is Seen Ihat
`with regard to pump and operating
`it is important to make appropriate choices of fiber length and
`concentration, resonator Q, and the degree of frequency dis-
`crimination. The next section provides an analytical treatment
`of Yb3+ fiber lasers, from which calculated data, such as in
`Fig. 3, are generated.
`
`111. CALCULATION OF GAIN
`For simplicity, we assume pump and laser intensity profiles
`to be uniform over the area A of the core and also assume
`a uniform dopant distribution within the core. Steady-state
`conditions are also assumed. The total populations of the F712
`and 2Fj/2 manifolds are designated N 1 and N 2 , respectively.
`The total population N is then
`
`N = N I + N2.
`
`(1)
`
`Absorption a,(X)
`cross sections are
`and emission a,(X)
`defined such that the absorption coefficient at wavelength X
`is Nla,(X) - N2a,(A). For brevity we write au(Ap) and
`as nap and or, respectively, and a a ( X l ) and or(Al) as
`ac(X,)
`a,( and
`respectively, where A,
`and AI are the pump and
`laser wavelengths, respectively.
`The gain achieved in a length L is given by
`
`- Na,ll
`
`where
`
`which can be reexpressed with the help of (4) as
`&(gel + CJalITPa
`Ahv,
`is the absorbed power
`Pa = P,(O) - P,(l).
`Thus, the procedure for calculation of gain reduces to calcu-
`lating P,(l) from ( 5 ) , hence, Pa (via (S)), and then the gain
`exponent from (7), or expressed in dB
`
`(7)
`
`(8)
`
`This expression simplifies if o,l can be neglected, for ex-
`ample, for long wavelength operation, such as 1140 nm.
`Similarly, the expression for P, is simplified when a,, is zero,
`as in the case where pumping is at short wavelengths, say less
`than 940 nm.
`A generalization of the equations to deal with cladding-
`pumping is straightforward. If the area of the inner cladding
`into which the pump is launched is z times the area of the core,
`then the pump cross sections should be replaced throughout
`by a,, 1.1: and (T,,
`/ . E . Thus, the effective saturation intensity
`
`

`
`6
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 1. NO. I , APRIL 1995
`
`becomes .r times greater (6) and the right-hand side of ( 5 )
`becomes - Naapz /x.
`The above analysis assumes that all the ions in the fiber
`are characterized by identical absorption and emission cross
`sections, i.e., that spectral broadening is homogeneous. The
`amorphous nature of a glass, however, leads to site-to-site
`variations of the local electric field, producing a degree of
`inhomogeneous broadening in the system. There is insufficient
`information available to accurately model the effect of any
`inhomogeneity. Spectral inhomogeneity could manifest itself
`in the form of spectral hole burning, either in pump absorption
`or laser emission. Since our analysis above is a small-signal
`analysis of the laser gain, however, the effects of spectral hole
`burning due to lasing can be neglected. They may, however
`need to be considered for operation above threshold.
`
`Iv. SPECTROSCOPY
`Successful modeling of the Yb3+ laser performance re-
`quires an accurate knowledge of the absorption and emission
`cross sections. The cross sections shown in Fig. l(b) were
`determined experimentally, as described in this section.
`The sidelight fluorescence spectrum was obtained, using
`an integrating sphere, from a length of fiber stripped of its
`protective jacket and pumped at 840 nm. The form of the
`emission cross section was determined by scaling the measured
`fluorescence spectrum by A’,
`and the absolute value of the
`emission cross section was then obtained by relating the
`integrated cross section to the measured upper level lifetime,
`using the relationship
`
`The decay of sidelight fluorescence from ’F;,’ was mea-
`sured by chopping the pump acoustooptically at -60Hz.
`The fluorescence was collected by a bundle of fibers and
`imaged onto a silicon photodiode. The fluorescence decayed
`exponentially over the first three e-folds. with a decay time
`of -840 p’s. The energy spacing between the two manifolds
`(>10000 cm-l) greatly exceeds the maximum phonon energy
`in silica (- 1100 cm-’), and the observed fluorescence lifetime
`is taken to be the radiative lifetime.
`The form of the absorption cross section was measured
`from the transmission of a white light source using a cut back
`technique. The well resolved peak at 975 nm that appears both
`in emission and absorption corresponds to transitions between
`the lowest energy Kramers doublets in the ground and excited
`manifolds. The absorption spectrum was scaled to give equal
`absorption and emission cross sections at the peak wavelength.
`This is an approximation, because the crystal field splitting is
`larger in the excited manifold than in the ground manifold and
`the degeneracy is lower; a McCumber analysis suggests that at
`room temperature the absorption cross section may be 5-10%
`larger than the emission cross section. This discrepancy is
`comparable with other experimental uncertainties. The profiles
`of the room temperature absorption and emission cross section
`spectra shown in Fig. l(b) agree well with the McCumber
`relationship n,, = (T, PXI)(F - h u ) , where the free-energy
`
`parameter F was taken to be the energy difference correspond-
`ing to the sharp peaks (at 975 nm) in the absorption and
`emission spectra. An independent confirmation of this scaling
`was obtained from a measurement of fluorescence sidelight as
`a function of pump power. The pump power (at 1017 nm)
`required to reach half the maximum fluorescence intensity
`(pump saturation power) is given by (6). This measurement
`gave a value for (a, + a,) of 0.43fO. 1 pm2, in fair agreement
`with the value obtained from the absorption and emission cross
`sections (0.39 pm2). This result justifies the assumption that
`the F 5 p lifetime is predominantly radiative.
`To determine the concentration of Yb3+ in the fibers used,
`cutback measurements were performed using low incident
`powers. Assuming a Beer’s law distribution of pump power
`along the fiber, the small-signal absorption coefficient was
`deduced which is the product of the dopant concentration,
`cross section, and overlap integral of the pump with the dopant.
`The overlap integral is not easily calculated; the distribution of
`the dopant across the fiber core tends to follow the germania
`distribution and varies with fiber drawing conditions. There-
`fore, we have chosen to work with an effective concentration,
`i.e., the product of the concentration and overlap integral. The
`effective concentration estimated in this way is 5501t 100 ppm.
`v. OPERATION AS A FREE-RUNNING
`LASER
`In this section, the operating characteristics of single-spatial-
`mode, free-running Yb3+ lasers are presented. The results have
`been chosen to illustrate the versatility and general principles
`of laser action in Yb3+-doped silica fibers and to show the
`agreement between theory and experiment. The Yb3+-doped
`germanosilicate fibers used for these experiments had numeri-
`cal aperture -0.17, and dopant concentration -550 ppm, and
`diameters of either 3 pm or 3.75 pm. The background loss
`was found to be < I O dB/km. Pump light, from either a Ti-
`sapphire laser, Nd:YLF or Nd:YAG laser was launched into
`the fiber using a x10 or x16 microscope objective, and laser
`action was investigated for resonators with feedback provided
`by two bare, cleaved fiber ends, or one cleaved end and a
`butted dielectric-coated mirror at the other end. Typical launch
`efficiencies varied from -50% for the Ti-sapphire laser to
`-70% for the Nd:YLF laser.
`Laser action has been investigated for pumping at wave-
`lengths from 84&1064 nm. In all cases, very efficient laser
`operation can be achieved, with slope efficiencies of up to 90%
`with respect to absorbed power. Depending on the fiber length
`and optical feedback, laser operation has been demonstrated
`from 103.5-1115 nm.
`In Fig. 4, the results of modeling calculations (line) are
`compared with measurements (symbols) of threshold powers
`and lasing wavelength as a function of fiber length. Threshold
`powers (Fig. 4(a)) are in terms of launched pump power, and
`the calculated thresholds are seen to be within 30% of the
`measured thresholds. The trends predicted by the modeling
`are confirmed by the experimental measurements.
`There is clearly an optimum length for reaching threshold
`for laser action at low values of launched pump powers,
`and this represents a compromise between maximizing pump
`absorption and minimizing the reabsorption losses at the laser
`
`

`
`PASK PI al.: YTTERBIUM-DOPED SILICA FIBER LASERS
`
`I
`
`300 t
`
`V
`
`e vv
`
`V 97411,
`1040nm
`
`i
`
`1
`
`I
`10
`Fibre length (m)
`
`1
`100
`
`(a)
`
`I
`
`1
`
`1100
`
`loooj
`-6 w
`
`950 I
`
`' '
`
`1
`
`I
`
`I
`100
`
`I
`10
`Fibre length (m)
`(b)
`Fig. 1. Some results of modeling calculations (lines) are compared with
`measurements (symbols) of (a) threshold powers and (b) laser wavelength
`as a function of fiber length. Threshold powers are in terms of launched
`pump power (fiber diameter 3.75 pm, N A 4 . 1 7 , [Yb'3+]-550 ppm, pump
`wavelength 840 nm)
`
`wavelength. The laser wavelength shifts to longer values as
`the fiber length is increased, as a consequence of the tail in
`the absorption spectrum which extends toward 1100 nm. This
`was discussed in Section 11.
`Despite the increased threshold for the longer fiber lengths,
`it is possible to obtain very efficient laser action in lengths up
`to 90 m. This is illustrated in Fig. 5 , where laser performance
`is shown for fiber lengths of 8, 28, and 90 m, pumped at
`850 nm with feedback provided by Fresnel reflections from
`the cleaved fiber ends. For these lengths the slope efficiencies
`were 70-80% with respect to launched pump power. For fiber
`lengths over 90 m the laser wavelength remains at 1090 nm.
`The laser wavelength can be changed by varying the amount
`of optical feedback, as discussed in Section 11. For the three
`fiber lengths shown in Fig. 5, a highly-reflecting mirror was
`butted to the output end of the fiber, and this was found to shift
`the laser wavelength from 1038 to 1050 nm for the 8 m length,
`1058 to 1060 nm for the 28 m length and 1090 to 1 106 nm
`for the 90 m length. This behavior was also predicted by the
`modeling. Noticeable features of the laser characteristics in
`Fig. 5 are the soft thresholds, a characteristic of quasi-three
`level laser systems, discussed in detail by Fan and Byer [19].
`
`450
`400
`
`- 350
`
`3
`
`300
`
`+-
`
`L
`
`250
`
`3 : 200
`c 2 150
`2 100
`50
`o &
`
`,
`
`I
`
`I
`
`0 8rn
`0 28m
`
`0
`
`0
`O V
`V
`
`0
`
`0 V
`
`0
`V
`
`V
`
`I
`
`I
`
`I
`
`I
`
`launched pump power (mW)
`Fig. 5. Free-running laser performance for three fiber lengths: 8, 28 and
`90 m, pumped at 850 nm. Feedback is provided by Fresnel reflections from
`the cleaved fiber ends (fiber diameter 3.75 Aim, NA-0.17, [Yb3+]-550 ppm).
`
`In all of the examples we cite in this paper, laser perfor-
`mance has been investigated for pump powers up to many
`times above threshold, in which case, to extract maximum
`power it is more important to maximize the slope efficiency
`than to minimize threshold. Consequently, the fiber lengths
`used experimentally tend to exceed the optimum lengths for
`minimizing threshold, as determined from the modeling, so
`as to ensure that virtually all of the launched pump power is
`absorbed.
`Very efficient laser performance has been demonstrated
`when pumping at 1064 and 1047 nm [7]. Due to the very weak
`absorption of the pump at these wavelengths (only -0.1 dB/m
`at 1064 nm for our 550 ppm doping level), it is necessary to
`use long fiber lengths, typically -100 m, to obtain efficient
`pump absorption. Because of the closeness of pump and
`lasing wavelengths, however, extremely high slope efficiencies
`(>90)% can be achieved. For example Fig. 6 shows the perfor-
`mance at 1102 nm demonstrated using a 90 m length of fiber,
`with one end cleaved and a mirror at the launch end which
`transmitted at 1047 nm and was highly-reflecting at 1102 nm.
`The threshold for laser action occurred for 30 mW launched
`pump power, and the slope efficiency was over 90% with
`respect to launched pump power. It is interesting to note that
`the laser wavelength, when such long fiber lengths are used, is
`essentially independent of pump wavelengths from 840 nm to
`1064 nm. The scaling of Yb3+ lasers to higher powers has also
`been addressed, and output powers as high as 2W at 1090 nm
`have been demonstrated by using a CW lamp-pumped laser at
`1064 nm to pump a 90 m length of fiber with two bare ends.
`No sign of rollover in the laser performance and no damage
`to the fiber were observed at these power levels. These results
`illustrate that Yb3+ lasers offer an extremely efficient way of
`converting high power Nd:YAG or ND:YLF lasers to longer
`wavelengths, particularly when using fiber gratings to select
`particular lasing wavelengths, as described in the next section.
`
`VI. OPERATION AT SPECIFIC WAVELENGTHS
`USING FIBER GRATINGS
`To achieve oscillation outside the band of wavelengths
`accessible by free-running cavity configurations, and also to
`
`

`
`8
`
`IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. I . NO. I , APRIL 1995
`
`600
`
`500
`
`A
`
`v
`L
`
`400
`a
`2 300
`a
`4- 3
`O 200
`E
`N
`0 100
`
`7
`
`7
`
`0
`
`950
`
`1000
`
`1100
`
`1150
`
`1050
`Kavelength (nm)
`Fig. 7. Results of computer modeling, showing gain spectra for various
`values of launched pump powers, with the fiber length chosen to maximize
`available gain at 1020 nm. Optimum fiber length as a function of pump power
`is shown in the inset (fiber diameter 3.0 I‘m, NA-0.17,
`ppm,
`[Yb”]-550
`pump wavelength 840 nm).
`
`maintain maximum flexibility in choice of pump wavelengths
`and cavity length, it is often desirable to introduce some form
`of wavelength discrimination into the cavity. The use of bulk
`elements such as prisms and gratings together with the neces-
`sary intracavity lenses introduce losses which cause significant
`decreases in laser efficiency. A much more attractive approach
`to achieve efficient laser operation at any wavelength within
`the emission band is to incorporate fiber gratings [20] into the
`laser cavity [8]. The recent availability of photorefractive fiber
`gratings has been a major element in expanding the potential of
`Yb3+-doped fiber devices. Laser action has been investigated
`at several wavelengths across the emission spectrum, and
`in this section we present the performance characteristics of
`Yb’+-doped silica lasers operating at 1020 nm and at around
`1140 nm, together with modeling results. The modeling is
`particularly useful in predicting the amount of selectivity
`required to suppress oscillation at free-running wavelengths.
`For our experiments, gratings having reflectivities up to 99%
`at any wavelength across the Yb emission spectrum have been
`fabricated in