Ultralow loss single-mode silica tapersmanufactured by a microheater

Lu Ding, Cherif Belacel, Sara Ducci, Giuseppe Leo, and Ivan Favero*Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot/CNRS,

10 rue Alice Domon et Léonie Duquet, 75013 Paris, France

*Corresponding author: ivan.favero@univ‑paris‑diderot.fr

Received 21 January 2010; revised 24 March 2010; accepted 31 March 2010;posted 5 April 2010 (Doc. ID 123082); published 23 April 2010

Using a ceramic thermoelectric heater, we show highly reproducible fabrication of single-mode sub-wavelength silica tapers with an ultralow loss level. The reproducibility of the process is studied statis-tically, leading to an average taper transmission of 94%. The best tapers have a transmission superior to99%, above the level commonly reached by other fabricationmethods. The taper profile is inspected alongits length and closely follows the exponential profile predicted by the model of Birks and Li. This highdegree of control over the taper shape allows a detailed analysis of the transition to the single-moderegime during tapering. As an application of this fabrication method, we present a microlooped taperprobe for evanescent coupling experiments requiring fine spatial selectivity. © 2010 Optical Societyof America

OCIS codes: 350.3950, 060.2310, 350.4238, 060.4005, 230.7370.

1. Introduction

Optical fiber tapers are now versatile components inmicro- and nano-optics, with applications in wave-length division multiplexing [1], fiber-based sensors[2] or supercontinuum light generation [3]. After thepioneering work of Knight et al. on a silica micro-sphere [4], they have also become a popular tool toinject light into microcavities or nanocavities bymeans of evanescent coupling [4–8]. For all these ap-plications, a low loss level is highly desirable, whichrequires good geometric regularity of the taper andlow surface roughness, both achievable, to some ex-tent, by pulling the optical fiber in a burner flame.However, this technique becomes increasingly diffi-cult and nonreproducible in the case of submicrom-eter taper fabrication [9]. Indeed, flames are proneto turbulence and can easily fluctuate because ofslight bursts in the gas flow or residual air currentsin the room. Impure gas can also lead to fiber con-tamination. By stabilizing the flame and usinghigh-purity gas, a low linear loss level, down to typi-

cally 10−2 dB=mm for 800nm diameter tapers, could,nevertheless, be reached [10–12]. A record value of afew 10−3 dB=mm was reported in [13], correspondingin this case to a taper transmission of 94% when in-cluding transition zones. A similar linear loss valueof 10−3 dB=mm was also reported in [14] without no-tification of the corresponding taper transmission.

The usual flame temperature employed in these si-lica fiber tapering rigs is 1400–1700 °C, close to sili-ca’s melting point. In the context of taper fabrication,it is, however, worth noting that, at 1100–1200 °C, asilica fiber is already soft enough to be stretchedslowly. A simple butane burner is sufficient for thispurpose [11]. This moderate temperature rangecould also be reached using a thermoelectric heater,which has the advantage of providing a more stableheat source. Such a heater was, for example, used in[15] for tapering compound glass nanowires with avery low softening temperature of 500 °C.

Here we use such microheater for taperingsilica fibers, at a temperature of 1160 °C. This allowsus to circumvent the difficulties inherent to theflame fabrication technique. Thanks to heat sourcestability, 800nm diameter ultralow loss tapers withtransmission of 94% are routinely produced, the best

0003-6935/10/132441-05$15.00/0© 2010 Optical Society of America

1 May 2010 / Vol. 49, No. 13 / APPLIED OPTICS 2441

tapers having transmission of over 99%. The fabri-cated tapers are shown to have a well-controlledshape, which allows observing quantitatively theirtransition to single-mode operation during pulling.We anticipate the high degree of control over taperlosses and shape offered by this fabrication methodto be beneficial for a broad range of fiber-taper appli-cations. We particularly stress its reproducibility andflexibility. As an example of application, we present amicrolooped single-mode taper probe for experi-ments of evanescent coupling to microcavities ornanocavities.

2. Taper Fabrication and Profile Analysis

An external cavity laser diode is launched into asingle-mode telecom fiber for λ ¼ 1:55 μm (CorningSMF-28). The fiber coating is striped over 3 cm andthe fiber is subsequently clamped on two linear trans-lation stages (Newport UTS 100CC) [see Fig. 1(a)],whose clamps are separated by a distance of 17 cm.Thestripedpart of the fiber is thoroughlycleanedwithacetone and isopropanol and positioned in the cavityof a ceramic microheater. This microheater is a centi-meter-sized thermoelectric oven originally developedfor fiber-coupler fabrication [16] [see Fig. 1(b)], andwhich we operate here at a temperature of 1160 °C.After positioning the fiber in the cavity, we wait120 s to reach thermal equilibrium. The two stagesare then set in motion at a constant pulling velocityof 40 μm=s, while the fiber transmission is monitoredon a photodetector. After about 40mmof fiber elonga-tion, the stages are stopped and the taper is slowlyremoved from the cavity.We first mount the produced taper on a glass slide

for inspection. Its radius is measured along the fiberaxis z with a high magnification ratio optical micro-scope (500×). The measured radius rðzÞ is shown inFig. 2(b) and is compared to the model of Birksand Li, which, for a constant hot-zone length duringtapering, predicts [17]

rðzÞ ¼

���������

r0 exp�−z−z0−L0=2

L0

�− z0 − L0=2 ≤ z ≤ −L0=2

rw ¼ r0 exp�−z0L0

�− L0=2 ≤ z ≤ L0=2

r0 exp�z−z0−L0=2

L0

�L0=2 ≤ z ≤ z0 þ L0=2

;ð1Þ

where L0 is the hot-zone length [17], 2z0 is the elon-gation length of the fiber, and z ¼ 0 is the middle ofthe taper. Although the temperature profile of theheater hot zone is closer to Gaussian than to a con-stant flat profile [16], Eq. (1) describes well ourresults. The fit of the measured profile is excellentwhen using the measured values z0 ¼ 19:75mmand r0 ¼ 66 μm, and setting the only adjustable para-meter L0 to 3:96mm. On the taper edges, we note asmall discrepancy due to temperature conditionsduring early pulling that deviate from subsequentsteady-state conditions. However, the data show con-vincingly the validity of the model down to the scaleof submicrometer tapered fibers, whereas it wasoriginally tested only on much larger glass rods, ofdiameter 5mm [17]. Here, the waist diameter is0:9� 0:3 μm. In our fabrication method, we can alsovary L0 by changing the operation temperature of theheater and the pulling velocity. This allows fabricat-ing tapers of various profiles, depending on thetargeted mechanical or optical application.

3. Highly Reproducible Ultralow Loss Level

The profile measured in Fig. 2 is expected to behighly adiabatic for light at λ ¼ 1:55 μm. Here, byusing the thermoelectric oven, we indeed easily ob-tain highly adiabatic tapers, which reproduciblyshow an extremely low loss level. In Fig. 3, we pre-sent a statistical study performed over 80 tapers, allhaving about 40mm of elongation length. The pro-duced tapers have an average transmission level of94%, on par with the best subwavelength taperspreviously reported [10–13]. But, here, 94% is anaverage result and we stress that half of the tapersdisplay, in reality, a transmission between 95% and100%. The best tapers rise to transmission valuessuperior to 99%, which we do not further resolve.If excluding taper transition lossy contributions, thiscorresponds for the uniform waist region to a linearloss level in the 10−3 dB=mm range. Similar ultralowlinear loss values were also inferred in [13,14] whenexcluding transition contributions. However, a morefair comparison would require including these tran-sitions, as they contribute to the measured losses. Inthis case, the total taper transmission itself must beexamined. For what concerns the total taper trans-mission, our fabrication method gives reproducibly

Fig. 1. (Color online) Microheater taper fabrication. (a) Schematic view of the set-up for fiber-taper fabrication. μH is the thermoelectricmicroheater providing a stable hot zone. TS 1,2 are motorized linear translation stages. PD is the photodetector. (b) Zoom-in picture of themicroheater with the fiber positioned in the cavity.

2442 APPLIED OPTICS / Vol. 49, No. 13 / 1 May 2010

excellent results, above the level commonly reachedwith other methods.In ambient conditions, the taper transmission de-

cays upon time because of silica reaction with watermolecules in the air [18]. Wemeasured the decay ratefor various relative humidity (RH) levels at a tem-perature of 25 °C (Table 1). In a wet environment(RH ¼ 35%), the taper lifetime can be as short as afew hours. In a plexiglass enclosure purged withdried air, we observe taper transmissions constantover weeks.

4. Transition to a Single-Mode Taper

In addition to achieving a low loss level, it is, in manyapplications, necessary to produce tapers that aresingle-mode guiding. This is the case in optics experi-ments where ideality relies on perfect mode-to-modecoupling, or in filtering applications, where higher-order modes need be suppressed. Following the cut-off criterion for a weakly guiding taper [19], the taperbecomes single mode when its waist radius rw issmaller than (2:405λ=2πNA), where NA is the numer-ical aperture of the fiber NA ¼ ðnco

2− ncl

2Þ1=2 withnco and ncl as the refractive of the core and cladding,

respectively. For ncl ¼ 1 (air) and nco ¼ 1:45 at λ ¼1:55 μm, this occurs for a waist diameter inferior to1:13 μm. The weak guidance approximation (Δ ≪ 1)is sufficient for our purpose here, but we note that afiber taper is not a very weakly guiding structure atits waist because its profile height parameter Δ ¼ðnco

2− ncl

2Þ=2nco2 amounts to 0.26, when standard

telecom fibers have Δ of order 0.001 [20]. Uponstretching, the fiber transmission as a function ofelongation length 2z0 displays oscillations of varyingamplitude and frequencies (see Fig. 4, upper panel).These oscillations arise from interferences betweendifferent modes supported by the fiber taper andtheir vanishing means that the single-mode opera-tion is reached. However, as we will see below in adetailed study, this criterion must be used with somecare in the experiments.

These oscillations of the transmission can be ana-lyzed using themathematical tool of the Gabor trans-form, which is a specific form of short-time Fourieranalysis [21]. Figure 4 shows a taper transmissionas a function of the elongation (upper panel) andits Gabor transform (lower panel). The latter has adominant component vanishing at an elongationlength of 32mm, exactly when a dramatic changein the beating amplitude occurs. A second componentof smaller weight vanishes later at a pulling length ofabout 34mm. To model this behavior, we use analy-tical approximate expressions (accurate far from cut-off) for the propagation constants β of modes guidedin a dielectric cylinder of radius r [22]:

Fig. 3. (Color online) Reproducible ultralow loss level. Statistichistogram of fiber taper transmission out of 80 samples.

Table 1. Transmission Decay of Fiber-Taper Transmission(in Units of Percent/Hour) with Respect to the Relative Humidity

of the Environment at 25 °C

Relative Humidity (%) Transmission Decay (%/hour)

<20 0.225 930 1734 30

Fig. 4. (Color online) Transition to single-mode operation.Upper panel: recorded fiber-taper transmission as a function ofthe elongation length. Lower panel: Gabor transform of therecorded transmission. Dashed lines are given by our model.

Fig. 2. (Color online) Taper profile. (a) Schematic profile of a sym-metric fiber taper formed in a constant hot zone, where the totaltaper length is 2z0 þ L0. zo is the length of the tapering transitionand L0 is the length of the taper waist (also the hot-zone size here).(b) Measured profile of a typical fiber taper fabricated using themicroheater. Here, the taper waist diameter is 0:9 μm.

1 May 2010 / Vol. 49, No. 13 / APPLIED OPTICS 2443

������βðrÞ ¼

h�2πλ

�2n2co −

�UðrÞr

�2i1=2

UðrÞ ¼ Uð∞Þ × exp�

−λr2πNA

� ;

where Uð∞Þ ¼ 2:405, 3.832, 5.135, and 5.520 for thefour first distinct propagation constants βk¼1;2;3;4 ofguided modes that we consider here. For each valueof 2z0, we express the phase accumulated by a modeof propagation constant βk during transmissionthrough the taper [11]:

Φkð2z0Þ ¼ βkðrwÞL0 þ 2Zz0

0

βkðrðzÞÞdz;

where rw and rðzÞ are given by Eq. (1). To model thetaper transmission T, we take the modulus square ofa weighted sum of the expðiϕkÞ, where the weightsare adjusted to reproduce the transmission beatingamplitude (not shown here). The white dashed linesin the lower panel of Fig. 4 are obtained by a Gabortransform of the obtained T, where the only adjusta-ble parameter L0 is set to 3:795mm. The close agree-ment with data allows us to interpret the dominantcomponent in the Gabor transform as originatingfrom a beating of the taper fundamental modeHE11 with the HE12 mode of propagation constantβ4. The second fainter component relates to a beatingwith other modes of lower symmetry, analogous toLP11. Note that complete modeling close to cutoffis also possible if solving numerically the Helmholtzequation [11]. Figure 4 shows that a detailed analy-sis is absolutely needed to ensure that the taper hasindeed reached the single-mode regime. If simplystopping the pulling when the dominant beating dis-appears at 32mm of elongation, the obtained taperstill supports three guided modes HE11, TE01, andHE21

5. Microlooped Fiber Taper

Straight optical fiber tapers are employed for evanes-cent coupling to small optical cavities, such as silicamicrospheres or microtoroids [4,5], but, in the case ofsubmicrometer cavities integrated on-chip, parasiticcoupling to the substrate becomes limiting with astraight taper. To circumvent this difficulty, variousstrategies have been devised to locally curve thetaper on a distance of about 100 μm to improve thespatial selectivity of the coupling [6–8,12]. Here,building on an earlier method [8], we present andcharacterize a simple technique for fabricating amicrolooped fiber taper for spatially selective evanes-cent coupling applications.In our fabrication process, after removing the taper

from the microheater, the translation stages are bothmoved 1mm back in order to release the taper ten-sion. Using a rotating fiber clamp, the taper is firsttwisted four turns on itself, with the effect of forminga multitwisted looped structure at its center, asshown in Fig. 5(a). The taper is put back under ten-

sion bymoving both stages by 0:3mm, and is glued onone end on an acrylate plate with epoxy. Its oppositeend is then pulled gently while observing the loop un-twist until it reaches the single-loop structure shownin Fig. 5(b). It is eventually glued on the plate and theexact degree of tension is adjusted during curing. Thelooped tapers obtained in this manner have a typicaldiameter of 70 μm [Fig. 5(b), inset], and the transmis-sion spectrum is shown in Fig. 5(c). Periodic oscilla-tions in this spectrum correspond to resonances ofthe microring cavity formed by the loop [23], witha free spectral range of 8:2nm. The taper waist of dia-meter 1 μm, having an effective index of 1.25 for itssingle-guided mode at λ ¼ 1:55 μm, corresponds toa loop diameter of 74 μm, in good agreement withthe microscope observation. The free spectral rangemeasurement provides an in situ control of the loopgeometry, a beneficial feature for adjusting its me-chanical properties in evanescent coupling experi-ments. Using this adjustable fiber-taper structure,we have recently probed optical gallery modes ofGaAs disks of a few micrometers in diameter and200nm thickness (to be described elsewhere).

6. Conclusions

Given the recent success of subwavelength optical fi-ber taper in the arena of microphotonics and nano-photonics, it is desirable to increase access to thesetechniques in optics laboratories. The fabricationmethod used in this article is highly reproducible,low cost, and easy to implement. It circumvents dif-ficulties of the standard flame fabrication and offersultralow loss level tapers on demand, together with ahigh level of control over their shape and single-modeoperation. Further progress on the way to large-scaleapplications could include taper encapsulation foruse in ambient conditions.

We thank Hiromasa Ito and Pascale Senellart forfruitful discussion. This work was funded by C-NanoIle de France.

Fig. 5. (Color online) Microlooped taper. (a) Multitwisted loopedtaper structure. (b) Tensioned single-loop taper. (c) Transmissionspectrum of a single-loop taper.

2444 APPLIED OPTICS / Vol. 49, No. 13 / 1 May 2010

References1. F. Gonthier, S. Lacroix, X. Daxhelet, R. J. Black, and J. Bures,

“Broad-band all-fiber filters for wavelength division multi-plexing application,” Appl. Phys. Lett. 54, 1290–1292 (1989).

2. L. C. Bobb and P. M. Shankar, “Tapered optical fiber compo-nents and sensors,” Microwave J. 35, 218 (1992).

3. T. A. Birks, W. J. Wadsworth, and P. St. J. Russell, “Super-continuum generation in tapered fibers,” Opt. Lett. 25, 1415–1417 (2000).

4. J. C. Knight, G. Cheung, F. Jacques, and T. Birks, “Phase-matched excitation of whispering gallery mode resonancesby a fiber taper,” Opt. Lett. 22, 1129–1131 (1997).

5. M. Cai and K. Vahala, “Highly efficient optical power transferto whispering gallery modes by use of a symmetrical dualcoupling configuration,” Opt. Lett. 25, 260–262 (2000).

6. K.Srinivasan,P.E.Barclay,M.Borselli,andO.Painter,“Opticalfiber based measurement of an ultrasmall volume, high-Qphotonic crystal microcavity,” Phys. Rev. B 70, 081306 (2004).

7. I. K. Hwang, S. K. Kim, J. K. Yang, S. H. Kim, S. H. Lee, and Y.H. Lee, “Curved microfiber photon coupling for photonic crys-tal light emitter,” Appl. Phys. Lett. 87, 131107 (2005).

8. C. Grillet, C. Monat, C. L. C. Smith, B. Eggleton, D. J. Moss,S. Frederick, D. Dalacu, P. J. Poole, J. Lapointe, G. Aers, andR.L.Williams, “Nanowirecouplingtophotoniccrystalnanocav-ities for single photon sources,” Opt. Express 15, 1267–1276 (2007).

9. L. Tong, R. R. Gattas, J. B. Ashcom, S. He, J. Lou, M. Shen, I.Maxwell, and E. Mazur, “Sub-wavelength diameter silicawires for low-loss optical wave guiding,” Nature 426, 816–819(2003).

10. G. Brambilla, V. Finazzi, and D. J. Richardson, “Ultra-low-loss optical fiber nanotapers,” Opt. Express 12, 2258–2263(2004).

11. F. Orucevic, V. Lefevre-Seguin, and J. Hare, “Transmit-tance and near field characterization of sub-wavelength

tapered optical fibers,” Opt. Express 15, 13624–13629(2007).

12. C. P. Michael, M. Borselli, T. J. Johnson, C. Chrystal, and O.Painter, “An optical fiber taper probe for wafer scale micropho-tonics device characterization,” Opt. Express 15, 4745–4752(2007).

13. S. G. Leon-Saval, T. A. Birks,W. J.Wadsworth, P. St. J. Russell,and M. W. Mason, “Supercontinuum generation in submicronfibre waveguides,” Opt. Express 12, 2864–2869 (2004).

14. G. Brambilla, F. Xu, and X. Feng, “Fabrication of optical fibrenanowires and their optical andmechanical characterisation,”Electron. Lett. 42, 517–519 (2006).

15. G. Brambilla, F. Koizumi, X. Feng, and D. J. Richardson,“Compound-glass optical nanowires,” Electron. Lett. 41,400–402 (2005).

16. Y. Takeuchi and J. Noda, “Novel fiber coupler tapering processusing a microheater,” IEEE Photonics Technol. Lett. 4, 465–467 (1992).

17. T. A. Birks and Y. W. Li, “The shape of fiber tapers,” IEEE J.Lightwave Technol. 10, 432–438 (1992).

18. J. L. Mrotek, M. J. Matthewson, and C. R. Kurkjian, “Thefatigue of high strength fused silica optical fibers in lowhumidity,” J. Non-Cryst. Solids 297, 91–95 (2002).

19. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics,2nd ed. (Wiley, 2007).

20. A. W. Snyder and J. D. Love, Optical Waveguide Theory(Chapman and Hall, 1983).

21. http://en.wikipedia.org/wiki/Gabor_transform22. A. W. Snyder, “Asymptotic expressions for eigenfunctions and

eigenvalues of a dielectric or optical waveguide,” IEEE Trans.Microwave Theory Tech. 17, 1130–1138 (1969).

23. M. Sumetsky, Y. Dulashko, and A. Hale, “Fabrication andstudy of bent and coiled free silica nanowires: self-couplingmicroloop optical interferometer,” Opt. Express 12, 3521–3531 (2004).

1 May 2010 / Vol. 49, No. 13 / APPLIED OPTICS 2445