TIRF Microscopy for the Study of Nanobubble Dynamics

Chon U Chan & Claus-Dieter OhlDivision of Physics and Applied Physics, School of Physical and Mathematical Sciences,

Nanyang Technological University, Singapore 637371, Singapore(Dated: September 8, 2012)

Nanobubbles can be observed with optical microscopy using the total internal re�ection �uo-rescence (TIRF) excitation. We report on TIRF visualization using Rhodamine 6G at 5µM con-centration which results to strongly contrasted pictures. The preferential absorption and the highspatial resolution allow to detect nanobubbles with diameters of 230 nm and above. We resolve thenucleation dynamics during the water-ethanol-water exchange: Within 4 min after the exchangebubble nucleate and form a stable population. Additionally, we demonstrate that tracer particlesnear to the nanobubbles are following Brownian motion: A remaining drift �ow is weaker than afew micrometers per second at a distance of 400 nm from the nanobubble's center.

Introduction Nanobubbles are nanometer high gasbodies attached to a surface being immersed in a liq-uid [1]. In particular the study of surface stabilizednanobubbles in water has attracted a lot of attention be-cause of their potential role for interfacial water technolo-gies [2, 3]. So far nanobubbles are mostly studied withscanning atomic force microscopy (AFM) which o�ers avery high spatial resolution (below 1 nm). The down-side is the long scanning time which prevents a study oftheir short time dynamics. Higher temporal resolutionhas been achieved with single line scans of the AFM tip[4] or with IR spectroscopy, e.g. [5, 6].

In this Letter we report on the visualization of stabi-lized nanobubbles on a hydrophilic surface with a stan-dard optical microscopy technique. A common procedureto create surface nanobubbles on hydrophilic interfaces isthe water-ethanol-water exchange process. There, thewater is replaced �rst by ethanol and then by wateragain [7]. The higher solubility of gas in ethanol andthe exothermic mixing of ethanol with water leads to therelease of dissolved gas from the ethanol. The subsequentreplacement of the ethanol with water is made responsi-ble for the nucleation nanobubbles from the now super-saturated water, see [8] and further studies summarizedin Ref. [1].

The recent review by Seddon and Lohse [1] pointed outthat for success in understanding the physics of nanobub-bles the experimental reproducibility has to be improved,the nucleation process detailed, and indications on a pos-sible dynamic equilibrium collected. All these questionsdemand for an experimental technique which not onlycan resolve the surface nanobubbles but also study theirdynamics with a much better temporal resolution as cur-rently available. In this Letter we present an experi-mental technique which can help to solve some of theremaining puzzles in nanobubble research.

We �rst describe the experimental technique to visu-alize surface nanobubbles and then discuss the dynamicsof bubble formation during the water-ethanol-water ex-change using this optical technique. The physical princi-ple behind optical nanobubble detection is total internal

re�ection �uorescence (TIRF) microscopy. TIRF allowsto illuminate only a very thin volume of liquid in con-tact with an interface, i.e. between glass and water. Forour geometry the total internal re�ection occurs for anangle θtot > sin−1(nw/ng) ≈ 61◦, where nw = 1.33 andng = 1.52 are the indexes of refraction for water and themicroscope cover glass, respectively. In the experimentwe can achieve conveniently this angle of incidence; themaximum possible angel θ has been determined with aprism and is about 74◦. The corresponding penetrationdepth (that is where the intensity drops to 1/e) can beestimated [9] as 70nm by z0 = λ(4π

√(ngsinθ)2 − n2w)−1,

where λ is the laser wavelength.

FIG. 1: (a) Sketch of the optical path for TIRF microscopy.Epi�uorescence and TIRF microscopy excitations are avail-able by rotating mirror M1. (b) Cross session of the PDMSchannel to study the water-ethanol-water exchange. (c)Schematic of the �ow line consisting of two syringes and sy-ringe pumps, a T-connection, and a 5 cm feed to the mi-crochannel. The microchannel is 5 mm long and 1 mm wide.

Experimental Setup Figure 1a depicts the beam form-ing and injection into the microscope. We use a greenDPSS CW laser (40µW, λ = 532nm) expanded to 12mm in diameter and steered into the side port of an in-verted microscope (Olympus IX71). A mirror and a lensset the angle of incidence θ, the mirror (M1) steers andlens (L3) focuses the beam into the back focal aperture of

2

the microscope objective (Olympus ApoN 60x, NA 1.49).By adjusting the rotating mirror M1 located at the backfocal plane of L3, the incident angle θ can be varied whilethe illumination spot remains �xed and within the �eldof view. Due to the limited working distance of the high-NA microscope objective only thin glass plates (here acover slip glass #1 with 140µm thickness) can be used.A cooled slow scan CCD camera (Sensicam QE, PCO,Germany) with a pixel size of 6.45 × 6.45µm2 is usedfor imaging. We have measured a pixel resolution of 108nm/pixel. The di�raction limited resolution for the ob-jective is about 230 nm (at 550 nm). The camera recordsa �eld of view of 55 × 69µm2 with a typical exposuretime of 25-40ms at 18 frames/s. This is su�ciently fastto resolve the nanobubble nucleation dynamics (see be-low).

Figure 1b sketches the side view of the beam re-�ecting from the glass-liquid interface. The liquid istransported in a microchannel with rectangular cross-section (1 mm width, 20µm height). The channel isfabricated with standard soft lithography technique, thepatterned PDMS channel bonded on a glass cover slip(Menzel-Glaser, Germany) pre-cleaned in an ultrasonicbath. 5µM Rhodamine 6G �uorescent dye is dissolvedboth in the DI water (puri�cation with Sartorius Arium611vf, France) and in 99% ethanol (Riverbank Chemi-cals, Singapore). Both liquids are loaded into separatesyringes. They are discharged in sequence using syringepumps. Before the liquids reach the channel they �owthrough a T-junction which is about 5 cm away from thechannel inlet. When the �ow channel is operated somepressure builds up which eventually �exes the cover slipglass mildly. Therefore the distance of the microscopeobjective has to be adjusted during the experiment. Themanual height adjustment is aided by a back re�ectionof the laser beam. This re�ection passes over an aper-ture and is detected with a photodiode (not shown inFig. 1a). During the experiments the photodiode sig-nal is kept constant by adjusting the stage manually andthus keeping the distance between objective and glassconstant.

Nucleation dynamics during water ethanol exchange

Figure 2a demonstrates a typical �gure recorded afterthe water-ethanol-water exchange in TIRF mode. Thesurface lights up with disc shaped objects �lling almostcompletely the surface of the glass cover slip. Beforethe exchange an unstructured dim light is recorded. Yet,after the exchange protocol strong contrasted objects ap-pear. We attribute these objects with surface nanobub-bles. The typical contrast of the surface bubbles as de-�ned by (Imax − Imin)/(Imax + Imin) is approx. 0.5 andthe signal to noise ratio is 3.

Figure 2b-d are enlarged views of the area indicatedwith a white square in Fig. 2a (scale bar length is 500nm). For magni�cation they are re-sampled using linearsplines. Figure 2b is just the enlarged part of Fig. 2a,

FIG. 2: (a) Nanobubbles observed under TIRF microscopy.Scale bar is 5µm. The square area is zoomed into for com-parison of the di�erent technique (b) TIRF microscopy (scalebar is 500nm), (c) bright�eld and (d) epi�uorescence.

while Fig. 2c and Fig. 2d compare the same view now inbright�eld mode and in epi�uorescence (θ = 0◦), respec-tively. From this comparison it becomes clear that theTIRF microscopy provides a strong contrast for nanobub-ble visualization. Nanobubbles are not visible in thebright�eld mode, and only some of the largest structuresappear but weak in the epi�uorescence mode.

What leads to the stark contrast of nanobubbles underTIR illumination? It is well known that Rhodamine isaccumulating at liquid-gas interfaces. Zheng et al. [10]demonstrate that the �uorescence is about 60 timesstronger from the water-air interfaces as compared to thebulk. Thus we explain the strong contrast of nanobub-bles in TIRF microscopy by the combination of adsorp-tion of Rhodamine at the nanobubble interface and theshort penetration depth of the evanescent wave excitingpreferentially the liquid on the scale of the nanobubbleheight, i.e. about 100 nm.

Although the water-ethanol-water exchange is a com-mon method to generate nanobubbles on hydrophilic sur-faces, very few measurements on the nucleation dynamicsare available. The microbalance techniques predicts thatthe nanobubbles are formed within one minute after theliquid exchange [11].

Figure 3 presents the di�erent stages leading to a sur-face decorated with nanobubbles. For this experiment,both syringes holding water and ethanol are dischargedone after another. Initially the feed line and the wholemicrochannel, see Fig. 1c, is �lled with water. Then afast �ow (Ethanol) of 125µ l/min transports the water-ethanol interface from the T-juntion to the inlet of themicrochannel. Thereby we reduce di�usion at the T-junction liquid-liquid interface. Then a slow �ow of1µ l/min pushes the interface through the microchannel.The dynamics is presented in Fig. 3: at time t = 0the channel is �lled with water only. At time t = 18 s

3

FIG. 3: Snapshots of images obtained during water-ethanol-water exchange. The intensity is normalized for each frame.Upper row: Bubbles are formed when water is replaced byethanol. These bubbles dissolved quickly. Lower row: Bub-bles again nucleate when ethanol is replaced by water. Bub-bles continue to nucleate after the exchange is �nished. Scalebar is 10µm.

the water is partly replaced from the left with ethanoland quickly bright spots appear on the surface. Thesenanobubble however quickly dissolve and after 40 s allbubbles have dissolved. Then, we keep ethanol for about15 minutes in the channel before starting the replace-ment. Again, for this exchange the water is pushed ini-tially with a fast �ow through the feed line followed by aslow �ow through the microchannel. At t = 16min 32 sthe water front arrives at the �eld of view and pushesout the ethanol to the right. At time t = 16min 51 sthe microchannel is completely �lled with water and thenanobubbles start to nucleate. 11 s later many morenanonbubbles have nucleated, see last frame of Fig. 3.

Experiments not detailed here revealed that the �owrate a�ects the nucleation speed. At higher �ow rates(about 50µ l/min) nucleation completes almost instantlyonce the liquid-liquid interface has passed. While atsmaller �ow rates, e.g. 5µ l/min and below, nucleationmay take several minutes before a steady nanobubblepopulation is reached. The temporal development ofnanobubble nucleation at a �ow rate of 5µ l/min is shownin Fig. 4. At time t = 2.5min after water-ethanol-waterexchange a few isolated nanobubbles appear. 2 minuteslater the surface shows about 0.1 bubbles /µm2; another2 minutes the surface density has quadrupled. Figure 5aplots the nanobubble density as a function of time []. Therise time of the nanobubble density, i.e. from 10% to 90%of the maximum bubble density, is about 240 s. An anal-ysis of the nanobubble diameter is presented in Fig. 5b.We understand that bubbles below the resolution limitof about 230nm can't be resolved. Thus, the distributiononly shows relatively large diameter nanobubbles. Simi-lar sizes with diameters of up to 1µm have been observedon rough surfaces [12].

Residual �ow close to nanobubbles The optical de-tection of nanobubbles allows to combine the technique

FIG. 4: The channel is �lled with water after water-ethanol-water exchange. Bubbles nucleate gradually over several min-utes. The exchange �ow is from left to right. The Scale baris 10µm.

FIG. 5: The bubble density plotted as a function of time forthe experiment shown in Fig. 4. The triangles state the timeswhere the snapshots in Fig. 4 are taken. Inset �gure providessize distribution of bubbles formed. Bins increase by the pixelsize. The smallest bin is 2 pixels wide.

with other microscope based observation technique. Us-ing small tracer particles we can probe if the Brownianmotion of particles is a�ected near to nanobubbles. It hasbeen proposed that nanobubbles are stabilized againstdissolution through a re-circulating �ow emanating fromthe apex of the bubble towards its three-phase contactline [17]. Their AFM measurements agree with a model

4

utilizing a Knudsen gas leading to a jetting �ow of severalmeters per second and starting at the nanobubble's apex.This jetting �ow should then drive a bulk re-circulating�ow with a lateral dimension of at least the lateral sizeof the nanobubble. We are probing this �ow by trackingthe motion of 200 nm diameter red �uorescent particlesnear and far from the nanobubbles. Areas on the surfacewith a low density of nanobubbles are �rst chosen. Thenthe locations of the nanobubbles are identi�ed. We photobleach the nanobubbles by increasing the laser power to150mW to increase the contrast of the particles near tothem and then slowly replace the water with DI watercontaining the nanoparticles. The particles are recordedat 21 frames/s (dt ≈ 48ms) with an exposure time of1ms and the tracks are obtained from image process-ing [13]. We then analyze the di�erential speed of theparticle ∆r/∆t(d) = ± |(r(t+ dt)− r(t))| /∆t as a func-tion of the distance d = |(r(t+ ∆t) + r(t))| /2, wherer(t) is the distance vector to nanobubble center in eachframe. The recirculating �ow would lead to a prefer-ential particle motion close to the surface towards thethree phase contact line. Our spatial resolution of ap-prox. 100 nm is able to detect drift velocities larger than2µm/s, which is about 6 orders of magnitude smallerthan the predicted jet velocity. We have analyzed 18nanobubbles with altogether 2000 tracks. Figure 6 ana-lyzes the result from the particle tracking: We measureparticle speeds of up to 30µm/s independent of the dis-tance. The averaged velocity is very close to zero; it isless than +2µm/s for all measured distances. A study ofthe mean-square-displacement [13] demonstrates that theparticles follow Brownian Motion. We �nd no indicationof a re-circulating �ow close to the nanobubble.

FIG. 6: Upper: di�erential speed of the particle as a func-tion of distance to nanobubble center. Lower: mean value ofdi�erential speed at each distance interval. The �rst intervalcentered at 0.43µm. Each interval width is 0.11µm.

In summary we have presented the formation ofnanobubbles on a hydrophilic glass surface during the

water-ethanol-water exchange in a microchannel. Thegrowth dynamic has been captured optically with a TIRFmicroscope. We achieved a temporal resolution with acooled CCD camera and a low power green laser of 56ms. The pixel resolution of our setup allows observingnanobubbles with diameters of 230nm and above. Addi-tionally we demonstrate that the motion of nanoparticlesclose to nanobubbles is dominated by Browninan motion.Even at a distance of 430 nm from the nanobubble centerwe �nd no drift velocity larger than the resolution limitof 2µm/s.

The TIRF based technique may help to address manymore open questions on interfacial nanobubbles. A non-exhaustive list is the superstability to strong tensionwaves [14], their acoustic resonance [15], and their abilityto reduce viscous drag [3, 16] in nanochannels.

We thank Lam Research AG in particular Frank Hol-steyns and Alexander Lippert for their continuous sup-port, and Lin Ma for helpful discussions on dyes. Aftersubmission of the �rst version of this manuscript we be-came aware of an independently developed technique ofnanobubble visualization [18]. We thank Detlef Lohsefor sharing their work during the �nalization of bothmanuscripts.

[1] J. R. T. Seddon and D. Lohse, J. Phys. Condens. Matter23, 133001 (2011).

[2] G. Liu and V. S. J. Craig, Appl. Mater. Interfaces 1, 481(2009).

[3] Y. Wang and B. Bhushan, Soft Matter 6, 29 (2010).[4] S. Yang, P. Tsai, E. S. Kooij, A. Prosperetti, H. J. W.

Zandvliet, and D. Lohse, Langmuir 25, 1466 (2009).[5] X. H. Zhang, A. Khan, and W. A. Ducker, Phys. Rev.

Lett. 98, 136101 (2007).[6] X. H. Zhang, A. Quinn, and W. A. Ducker, Langmuir

24, 4756 (2008).[7] S. Lou et al., Mater. Charact. 48, 211 (2002).[8] S. Lou, Z. Ouyang, Y. Zhang, X. Li, J. Hu, M. Li, and

F. Yang, J. Vac. Sci. Technol. B 18, 2573 (2000).[9] D. Axelrod, T. P. Burghardt, N. L. Thompson, Annu.

Rev. Biophys. Bio 13, 247 (1984).[10] X. H. Zhang, A. Harata, and T. Ogawa, Spectrochim.

Acta. A 57, 315 (2001).[11] X. H. Zhang, Phys. Chem. Chem. Phys. 10, 6842 (2008).[12] B. M. Borkent, S. de Beer, F. Mugele, and D. Lohse,

Langmuir 26, 260 (2010).[13] See the supplementary information on details of estima-

tion of nanobubble density, particle tracking and valida-tion of the Brownian motion.

[14] B. M. Borkent, S. M. Dammer, H. Schonherr, G. J.Vancso, and D. Lohse, Phys. Rev. Lett. 98, 204502(2007).

[15] H. Rathgen, K. Sugiyama, C. D. Ohl, D. Lohse, and F.Mugele, Phys. Rev. Lett. 99, 214501 (2007).

[16] A. Steinberger, C. Cottin-Bizonne, P. Kleimann, and E.Charlaix, Nat. Mater. 6, 665 (2007).

[17] J. R. T. Seddon, H.J.W. Zandvliet, and D. Lohse, Phys.

5

Rev. Lett. 107, 116101 (2011).[18] S. Karpitschka, E. Dietrich, J.R.T. Seddon, H.J.W.

Zandvliet, D. Lohse, and H. Riegler, Non-intrusive opti-

cal visualization of surface nanobubbles, Phys. Rev. Lett.in print (2012).