Viscometer using drag force measurements Nol M.H.,1 Semin B.,1 Hulin J.P.,1 and Auradou H.1 Lab.

FAST, UMR No. 7608, Bt. 502, Universit Paris Sud, F-91405 Orsay Cedex, France. A robust and precise viscometer using the forces exerted by a laminar ow inside a small duct is presented: the force is measured on a long cylindrical sensor dipped into the ow. Two devices of respective volumes 1400 and 31l have been realized, demonstrating that the technique is applicable with small ow volumes. Several Newtonian and non Newtonian uids have been tested at shear rates ranging from 0.3 up to 10 s1 for the rst device and from 85 up to 2550 s1 for the second. For Newtonian uids, of viscosities ranging from 103 to 0.1 Pa.s, the linear response of the device has been veried and a 10% agreement with the values provided by commercial rheometers is obtained. For non Newtonian polymer solutions, the variation of the force with the ow velocity allows one to determine the dependence of the viscosity on the shear rate. Two shear thinning polymer solutions with a power law behaviour at intermediate shear rates have been investigated and their rheological parameters have been determined. Keywords: Suggested keywords I. INTRODUCTION U by the relation: F = ( l0 U ), Measuring the viscosity of biological or chemical uids is a key issue in many elds ranging from medicine and food processing to the chemical and manufacturing industries1 . The viscosity reects the resistance of a uid to ow under an applied shear stress and many dierent types of viscometers have been developed to achieve this measurement. They may classied in four main families: capillary viscometers4,7,11 , rotational or sliding viscometers2,5 , falling spheres3,8 or slender objects9 , and vibrational viscometers10 . In the present study, a viscometer based on the measurement of the friction force exerted by a owing uid is presented. This technique is found to be suitable for measuring rapidly the viscosity of samples of small volume with a precision similar to that of classical rheometers. Moreover, the device is easy to manufacture, it has no moving parts and precise alignments are not required: the technique is thus inexpensive and easy to implement. To our knowledge similar techniques have only be envisioned for high viscosity uids owing in a pipe12 . In these cases, a xed blade transducer coupled to a shaft like probe is immersed in the owing uid and the value of the force on the blade is used to determine the uid viscosity. Here, recent progress in the technology of force sensors has allowed us to extend this approach to uids of low viscosity. In the present system, the probe is a cylindrical object located on the axis of a circular duct used as the ow channel so that the local shear rate on the whole lateral surface of the probe is constant. We show below that, unlike for sliding or rotating plate rheometers2 , a precise coincidence of the probe axis with that of the duct is not required. The force F applied by the uid on the object has been measured at dierent ow rates. For laminar ows of Newtonian uids, F is related to the mean ow velocity 1 (1)

in is the dynamical viscosity, l0 the length of the cylinder inside the ow and a geometrical parameter. Unlike methods using falling objects, these measurements do not require corrections as long as the ow is controlled by viscous forces6 : we demonstrate below that this is achieved practically by using Reynolds numbers Re 50. In Section II, the measurement devices and the viscosity measurement procedure are presented. The technique has been extended to non Newtonian uids: Sec. II B describes the inversion method allowing one to determine the rheological curves in this latter case. In Section III, the results of the viscosity measurements are compared quantitatively to those obtained with standard rotating viscometers. II. EXPERIMENTAL SET-UP AND PROCEDURE a. Description of the apparatus The viscometer is displayed schematically in Fig.1. Two devices have been
TABLE I. Characteristic parameters of the experimental devices. D, L and V : diameter, length and volume of the cylindral duct; d and l0 : diameter and immersed length of the probe; Q: ow rate; : shear rate. D mm 10 1 L mm 90 40 V mm3 1400 31 d mm 1.1 0.14 l0 mm 6080 13-57 Q ml/mn 3-30 0.5-15 s1 0.1-3 85-2550

A B

built and their characteristics are listed in Table I. Device A is large enough so that handling is easy and the control of the experimental conditions is optimal: in this case, a few milliliters of uid are needed to measure the viscosity.

Force sensor

1960 F (N)
g g

Inner cylinder Inlet l0 d

Flow

1920 1880 1840 400 800 Time (s) 1200

Force Outlet

FIG. 2. Variation as a function of time of the force measured in device B during two cycles of stepwise variations of the ow rate.

FIG. 1. Schematic view of the experimental viscometers.

For device B, this volume has been further reduced by a factor of 50. Both devices consist of a cylindrical duct of diameter D and length L inside which a small cylinder is inserted. The ducts are drilled into a PMMA bar with a tolerance of 0.05 mm: they are vertical with the open side at the top. In order to reduce the vortical motions of the uid in the measurement region, the diameter of these ducts is increased to 40 mm at their top. This creates a 10 mm deep bath into which the uid is injected through a small lateral hole: the size of this bath has been chosen so as to to minimize the eect of this transverse ow on the inner cylindrical probe. The latter is attached to a force sensor located above the bath. Its transverse location is adjusted by micrometric screws so that its axis and that of the duct roughly coincide: a small relative displacement of these axis was however found to have only a small inuence on the force measurement in agreement with the results from Refs. 6 and 14. In the duct, uid ows ows either from the top towards the bottom (corresponding to a traction force on the sensor) or from the bottom towards the top (compression force). For device A, the cylindrical probe is a stainless steel rod of diameter 1.1 mm : it is attached by a exible thread to a hook located under a SartoriusTM CP225D scale. The probe hangs freely with its lower part inside the ow duct. The scale allows one to measure forces ranging from 0.1 N to 0.8 N. For device B, the probe is a glass ber of diameter 140m glued to the tip of a MEMs force sensor. The measurement range of this sensor is 1 to 2500 N. b. Fluid preparation and characterization. The Newtonian uids are either pure water or water-glycerol mixtures with a relative mass concentration of glycerol 2

ranging from 0 (pure water) to 85%. Tests were also performed with shear thinning solutions: these are composed of 250 and 1000 ppm of high molecular weight scleroglucan in high purity water (Millipore - Milli-Q grade). Scleroglucan (provided here by Sano BioindustriesT M , France) is a polysaccharide and the solution is protected from bacterial contamination by adding 0.2 g/l of N aN3 . The density of the solutions and their temperature T are measured after each series of experiments by means of an Anton Paar 35N densimeter. c. Experimental procedure: The ow duct and connecting tubes are rst lled with the experimental uid and all trapped air bubbles are removed from the system. The ow rate is then increased by steps from Q = 0 up to the chosen maximum ow rate and in then reduced back to zero in the same way. Fig.2 displays the corresponding variation as a function of time measured by the MEMS sensor of device B. The non zero mean value reects the weight of the probe and the zero-shift of the sensor. Each step lasts from a few seconds (a minimum of 10 s is required so that the measurement stabilizes at a near constant value) up to a few minutes. The cycle is repeated (twice in Fig.2) in order to improve the accuracy of the measurements. The inlet and the outlet of the conduit were connected two a PharmaciaTM double syringe pump: one of the syringes injects the uid at the inlet while the other one sucks its excess at the other end of the duct. The oscillations on each plateau are induced by the pump. Gear or peristaltic pumps have also been used successfully.

A.

Data analysis for Newtonian uids

After each experiment, the mean and the standard deviation of the value of the force F on each plateau are computed. Figures 3a-b display the variation of this mean value as a function of the corresponding mean ow velocity U for two of the experiments performed with pure water. Although, with the present protocol, each

data point may be determined several times during one experiment, only data corresponding to one cycle are shown. Fig. 3a displays data obtained using device A;
6

a geometrical parameter . Here, we use therefore the constant value of this ratio F/U to determine the uid viscosity from the relation: = F l0 U (2)

2
F(N)

-2

-4

a)
50

-4

-2

0 -1 U (m.s )

6x10

-3

Practically, will be determined either by measuring a single couple of values (F, U ) or or by performing a linear regression over a set of dierent measurements of (F, U ) (see Sec.III A below). In both cases, however, the value of the geometrical coecient is required in order to use the Eq.(2). A rst approach is to determine the value of (or rather of l0 ) through a calibration measurement (or a set of measurements) using a uid of known viscosity 0 . Then, the unknown viscosity of the uid of interest is related to o by: (F/U ) = o (F/U )o (3)

40

F (N)

30

20

The parameter may also be estimated analytically by computing numerically the total force induced by the uid ow on the inner cylinder. A rst component of this force is the viscous shear stress force Fs Fs = l0 [ .n d]
C
0.00 0.05 0.10 -1 U (m.s ) 0.15

10

(4)

b)

FIG. 3. Variation of the force on the cylindrical probe as a function of the mean velocity U . (a): device A; solid line: linear regressing on all data of slope 0.735 g.s1 . (b) device B; solid line: linear regression on the rst ve points of slope 0.802 g.s1 .

the ow in the duct is either upward (U > 0) or downward (U < 0). The force F varies linearly with U , at least up to |U | 5 103 ; this velocity corresponds to a Reynolds number Re = |U |D/ 50 ( is the uid density). Moreover, the slope is the same for positive and negative values of U : the direction of the ow does not aect therefore the measurement. Figure 3b displays data obtained using device B: the lower sensitivity of the sensor led us to use a higher range of ow velocities. At rst, the force increases linearly with the velocity up to 0.05m.s1 : the corresponding Reynolds number Re = 50 is of the order of the maximum value reached with device A for which the variation of F with U was also linear. At higher velocities (i. e. for Re > 50), one observes a continuously increasing deviation from the linear variation: this non linearity may be accounted for by the development of inertial eects in the ow6 . In the following only measurements performed in the laminar viscous regime at Reynolds numbers Re 50 will be discussed. In this regime, the ratio l0 of the force by the uid velocity (see Eq. (1) is the product of the immersed length lo by the uid viscosity and by 3

in which is the viscous shear stress tensor which is assumed to be constant along the length lo ; n the unit vector normal to the external surface and C the boundary of a section of the cylinder normal to the axis. A second component is the pressure force Fp created by the dierence between the pressures at the ends of the cylinder: assuming that the pressure gradient p/z induced by the ow is constant, too, along the length lo leads to: d2 p (5) 4 z In order to compute the viscous shear stress and the pressure gradient p/z, we assume (in line with the previous simplications) that the uid velocity V is everywhere parallel to z and that V = v(r)ez due to the rotational and translational symmetries of the system. The governing equation of the ow reduces then to the simple 1D dierential equation13 : Fp = l0 p v(r) + (r )=0 (6) z r r r In which p/z is constant with both z and r. This equation can be solved analytically for zero slip boundary conditions at the surface of the cylinders. After computing the total ow rate in the gap between the coaxial cylinders, the prole v(r) may be related to the mean velocity U by: v(r) =
r ln( D ) U r } {( )2 2 +(12 ) ln() (1 + 2 )ln( 1 ) (1 2 ) d (7)

in which = D/d and U is the mean ow velocity. The gradient p/z can then be related to U from Eq. 6 which provides the value of Fp through Eq. 5. Under the above assumptions, the component along z of the viscous stress .n is equal to the value of for r = d so that: Fs = d lo v(r)/r. Computing the total force F by summing Fs and Fp provides then from Eq. 2 the respective values = 9.9 and 12.2 for devices A and B. Using in Eq.(2) the value of corresponding to the device of interest provides then the viscosity once the ratio F/U and the length lo have been measured experimentally. For instance, the value F/U = 0.735 g.s1 determined by a linear regression on the data of Fig. 3 leads to = 0.99 0.03 mP a.s for lo = 60 mm and = 12.2: this is very close to the value of the viscosity of water found in the litterature at a similar temperature. Systematic measurements of the viscosity of dierent Newtonian uids by this and other techniques are discussed and compared in Sec.III A below. We describe now the procedure developed for determining the rheological characteristics of Non Newtonian uids.

In order to characterize quantitatively the rheological properties of the uid, the variation of F with U must be translated into a rheological characteristic relating the apparent viscosity to . For this purpose we developed rst a procedure for computing the variation of F vs U for any chosen rheological curve = f () characterized by a set of parameters (24 depending on the rheological model selected). Like in the Newtonian case, the force F is determined by means of Eqs. (4) and (5). This requires the determination determination of the pressure and velocity elds between two coaxial cylinders for any specic rheological law f (). While, for Newtonian u ids, p/z and v(r) can be computed analytically, they are determined in the non Newtonian case by a numerical integration of the more general equation of motion replacing Eq. (6): p 1 + (r f ()) = 0 z r r (8)

B.

Processing data for Non Newtonian uids

30 20 F (N) 10 0 -10 -20

-3

with = v(r)/r (p/z is still constant with r and z). Once, v(r) and p/z have been determined for a given velocity U , the forces Fs and Ft on the cylinder can be computed. The force F is then computed for all experimental values of U . Finally, the variation of F with U obtained in this way is compared to the experimental one. The whole process is then repeated while adjusting the parameters of the rheological model by a least mean square method until the computed variation of F (U ) coincides with the experimental one.

III. A.
-4 -2 0 2 -1 U (m.s ) 4 6 8x10

EXPERIMENTAL RESULTS Newtonian uids

FIG. 4. Variation of the force in N for a 250 ppm polymer solution owing in the device A as function of the mean ow velocity U . Triangles and circles show the measurements obtained for two consecutive tests. The solid line represents the force vs velocity that best adjust the experimental data when a power law rheological curve = k is assumed.

Fig.4 shows that, unlike for Newtonian uids, the force varies non linearly with the mean velocity U for the polymer solutions. This deviation reects here the variation of the viscosity with the shear rate: the velocity is indeed too low for inertia eects to appear and they would induce an upward curvature of the curve (see Fig. 3b) and not downward as here. The slower increase of the force with U suggests, from Eq. 2, that the apparent viscosity decreases with U : this is in agreement with the shear thinning properties reported in the litterature for such solutions. 4

The viscosity of dierent water glycerol solutions has been determined by means of the procedure described in section II A. In this case, determining the viscosity requires only a measurement of the force F at a single ow velocity U : then is given by the relation = F/( l0 U ) discussed above. Fig. 5 displays the variation with the shear rate of the values of the viscosity obtained in this way for the less viscous uid studied here: no systematic trend of variation with is visible. Also the width of their distribution is such that these values are compatible both with reference measurements using a Low Shear 30 rheometer and published values from the litterature. More precisely, the average of the experimental values of corresponding to the whole range of shear rates investigated is: 0.99 0.12 mP a.s: this is very close to the value = 0.94 mP a.s from Ref. 7. In the inset of Fig.5 viscosity values measured with an MCR501 Anton Paar rheometer for the same uid are also plotted. These latter values increase sharply at low shear rates ; moreover, in the range of values of used

1.2

1.1 (mPa.s)

1.0

0.9

retical curve and there is a good coincidence of the values for repeat experiments at a same shear rate . In order to remove the eect of temperature variations, the accuracy of the measurements has been characterized quantitatively by computing the relative deviation (mes theo )/theo of all measured viscosities (mes from the theoretical value at the same temperature.

0.8

5 6 7 89

5 6 7 89

(s )

-1

10

(mes- theo)/ theo (%)

40 20 0 -20 -40

FIG. 5. Viscosity of pure water (T = 22.6 o C) measured using dierent devices. (): present apparatus; (): Contraves Low-shear 30 rotating viscometer; (): MCR501 AntonPaar rheometer. Dotted line: value from Ref. 7; Dashed dotted line, experimental value obtained from by a linear regression on the data at dierent ow rates.

in our apparatus, the measured viscosity is more than three times larger than that found in the literature. The tests were then extended to other values of the viscosity by performing measurements on water-glycerol solutions. In this case, and in order to improve the precision, several measurements are performed for each solution at dierent mean velocities U . The ratio a = F/U is determined by performing a linear regression F = a.U on the sets of values of (F, U ) and the experimental viscosity is nally computed by means of Eq. 2. The values obtained in this way are plotted in Fig. 6 together with data from the literature. These latter values (solid line) are obtained by computing analytically the variation of as a function of in the range of shear rates covered by the experiments; we used for that purpose analytical expressionsof as a function of the temperature and of the mass concentration of glycerol in the solution (see Ref. 18). Most experimental values are close to the theo100
6 5 4 3 2

20 40 60 Glycerol concentration (%)

80

FIG. 7. variations of the relative deviation (mes theo )/theo of the measured viscosity mes from the theoretical value as a function of the mass concentration of glycerol. (): device A; (): device B; (+) Contraves Low-shear 30 rotating viscometer; (): MCR501 AntonPaar rheometer

10
6 5 4 3 2

1 0 20 40 60 Glycerol concentration (%) 80

FIG. 6. Variation of the dynamic viscosity as a function of the concentration in mass of glycerol for the water-glycerol solutions used in the experiments. Values obtained using (): device A; (): device B. Solid line: variation of theoretical values theo from Ref. 18 for T =???.

This relative deviation is displayed in Fig.7 as a function of the mass concentration for measurements obtained both with devices A and B (circles and triangles in Fig.7) and with two commercial rheometers (diamonds and crosses). At all glycerol concentrations, the values of the ratio (mes theo )/theo are found to be distributed equally above and below zero with a maximum deviation of the order of 10%. These is no visible trend in the distribution implying that there is no global variation of the deviation with the concentration. A similar deviation is also observed for the two commercial rheometers. Two experimental measurements performed using device B display however a much stronger deviation of the order of 40%: in these cases, the uid was owing upward while no such eect was observed for downward ows. These two results may be accounted for by the buckling of the probe cylinder under the compressive stress induced by the upward ow of the uid. In the geometry considered, (i.e. one end xed and the other free to move laterally), buckling takes place if the hydrodynamic force l0 U becomes larger than the critical value Fc = 2 EI/(2l0 )2 in which E is the value of Youngs modulus for the probe (here a glass ber) and I the moment of inertia of its section (for a cylinder it is I = d4 /32 in which d is the diameter of the probe)19 . Youngs modulus has been determined for the probe by measuring its deection under its own weight: the value E = 39GPa found in this way is close to that usually reported for glass. For a uid of viscosity of 0.01Pa.s and a 5

(mPa.s)

probe of length l0 = 57mm, buckling should therefore occur for a uid velocity of the order of 5.103 m.s1 : this is close to the value corresponding to these experiments.

in the present study as measured using the Low shear30 and Anton Paar rheometers. At shear rates below 100s1, the polymer solutions display a shear thinning behaviour which is well adjusted in the most of the range of values of by the power law variation: = k . (9) The corresponding values of the rheological parameters k and tted are listed in Table II. At higher shear rates, the viscosity of the solutions tends towards a constant value of the same order of magnitude The higher value measured in the double gap conguration the Anton Paar rheometer likely reects an additional dissipation due to the appearance of an hydrodynamic instability: it has therefore been discarded. At very low ow rates, the viscosity should also reach another constant Newtonian plateau value o : the corresponding values of are however below the range of the present experi ments. Measurements using devices A and B In the range of values of U used here, measurements using device A have been found to correspond to the power law variation regime ( was always lower than 100 s1) because of its relatively large size. Data obtained using device B correspond instead to the upper shear rate regime in which the viscosity has the constant value : the values of are indeed higher for this device because of its smaller size. The corresponding values of and are then deter mined from those of F and U respectively by means of the procedures of Sec. II A (device B) and of Sec. II B for non Newtonian ones (device A).
TABLE II. Rheological parameters of scleroglucan solutions obtained with the low shear rheometer (LS), the Anton Paar MCR 501 (M CR) and with our viscometer A. Polymer Conc. ppm 250 Apparatus LS M CR?? A B M CR??? M CR??? A B k 0.018 0.019 0.32 ??? 0.38 0.41 0.49 1.5 0.01 0.80 ???? ???? ???? 0.81 2.8 0.01 mPa.s ???? ????

B.

Non-Newtonian uids

20

(mPa.s)

15

10

10

(s-1)

100

1000

FIG. 8. Variation of the dynamic viscosity of a 250ppm scleroglucan solution at 23.6 o C as a function of . (), (): data obtained respectively using devices A and B. (+) (resp. () data obtained using Low shear 30 (resp. Anton Paar MCR501) commercial rheometers. Solid lines: rheological laws estimated by inverting present data.

100
8

(mPa.s)

6 4

10
8 6 4

1000 ???
1 10 100 -1 (s ) 1000

FIG. 9. Variation of the dynamic viscosity of a 1000ppm scleroglucan solution at 23.6 o C as a function of . The sym bols for the data from devices A and B and the solid lines have the same meaning as in Fig. 8. ( ) (resp. ()) data obtained with an Anton Paar MCR501 commercial rheometer respectively in the cone plate and double gap conguration. Inset: close up view of the data obtained using device A.

Characterization using commercial rheometers. For non Newtonian solutions the uid viscosity varies with the shear rate. Figures 8 and 9 display the rheological characteristic curves of the two polymer solutions used 6

For device B, is determined from the couples of data (F, U ) by means of the procedure of Sec. II A for Newtonian uids. The values of obtained in this way are plotted as lled circles in Fig. 8 and 9 and listed in Table. II. These values are, as expected, independent of and very similar to those given by the commercial rheometers; note that, for the Anton Paar rheometer, only data obtained in the cone plate conguration are used in the comparison as discussed above. For device A, the procedure described in Sec.II B has been used for determining the rheological curve from the

force measurements. Practically a procedure has rst been developed for determining the theoretical variation F (U ) for a uid of rheological characteristic = k . This is achieved by determining rst the velocity and pressure elds between two coaxial cylinders by means of the Matlab software. These elds are then used to estimate the stress on the surface of the inner cylinder. Using these results, the force on the cylinder is easily obtained from Eqs. (4) and (5) as a function of the mean ow velocity. The rheological parameters k and are then determined by ajusting their values so that the experimental force variation F (U ) is well tted by the corresponding theoretical variation. The solid lines in Fig.4 corresponds to optimal ts obtained in this way. These optimal values of k and are listed in Tab.II. The rheological curves obtained in this way for the two polymer concentrations are superimposed in Fig.8 and 9 over the data from the commercial rheometers: a good quantitative agreement is shown. The same procedure might be applied using any usual rheological characteristic (i.e. Carreau or Cross functions, truncated power law ...).

the ratio d/D is kept constant). Hence, in spite of a scale reduction by a factor 10 between devices A and B, the force measured by the sensor are in the same range (i.e. between 1 and 1000N). In device B the measurement volume has been reduced to 30l: this suggests possible possible applications to viscsity measurements in microuidic apparatus. In this respect, further miniaturization of the device and a comparison of its performance to that of other techniques developed in micro or nano uidics2,20? will represent an important challenge. Future developments of the device will also seek to control the temperature of the uid and to reduce the evaporation of the solvent.

ACKNOWLEDGEMENTS This device is patented under the International Application No. PCT/FR2009/001210. We thank L. Auray et R. Pidoux for realizing the experimental setup, and A. Aubertin for the data acquisition program. We also thank CNRS for their nancial support and A. Bouteilly, C. Cozic and H. Corret for their helpful discussions.

IV.

DISCUSSION AND CONCLUSION REFERENCES

In this paper, we have described a new device allowing for the quantitative characterization of the rheological properties of uids from the value of the force exerted by a ow on a cylindrical probe. For Newtonian water-glycerol solutions of viscosities ranging from 103 to 0.1mP a.s, the measured value of is almost constant with the shear rate; even using a single measurement, the accuracy is similar to that of much more complex and costly commercial rheometers. The technique also allows one to determine the rheological characteristic curve of non Newtonian uids: this has been illustrated by measurements of this curve for two water-polymer (scleroglucan) solutions of dierent concentrations. The rheological parameters determined from the variations of the force wit the shear rate a intermediate and high values of correspond well to the values obtained using commercial rheometers. Compared to such rheometers, the present device does not include elements with a tight machining tolerance and, due to the reduced inuence of the positioning of the probe, the set up does not require a very careful alignment or a specic maintenance. Because of its relative simplicity, the technique is adaptable to an industrial environment (and possibly to in-line continuous measurements) and is not restricted to laboratory applications. In addition, the present device is easy to manufacture and (except for the force sensor) inexpensive. Some parts (like the duct or the probe) may therefore be disposed of after each use so that the technique is suitable for medical applications or for the measurement of reactive uids. For a given ow velocity, the force measured on the probe is not inuenced by scale reduction (frovide that 7

D.T.N. Chen, Q. Wen, P.A. Janmey, J.C. Crocker and A.G. Yodh, Annual Review of Condensed Matter Physics 1 (2010). 2 S. Granick and H-W. Hu, Langmuir, 10, 38573866 (1994). 3 J.M. Jose, J.M. Lopez-Pedrosa and M. Bradley, Rev. Sci. Instrum. 79, 094102 (2008). 4 T.S. Rushing and R.D. Hester, Rev. Sci. Instrum. 74, 176181, (2003). 5 D.L. Atkins, J.S. Ervin and L. Shafer, Energy & Fuels, 19, 19351947, (2005). 6 B. Semin, J-P. Hulin and H. Auradou, Phys. of Fluids 21, 103604 (2009). 7 A.F. Collings and V. Bajenov, Metrologia 66 6166 (1983). 8 M. Brizard, M. Meghar, E. Mahe and C. Verdier, Rev. Sci. Instrum. 76 025109 (2005). 9 N.A. Park and T.F. Irvine, Rev. Sci. Instrum. 66, 39823984 (1995). 10 D. C. Ash, M. J. Joyce, C. Barnes, C. J. Booth and A. C. Jeeries, Meas. Sci. Technol. 14 19551962 (2003). 11 G. Meyerho and B. Appelt, Macromolecules 12, 968 971 (1979). 12 U.SU Patents 3.796.088, 4.062.226, 4.148.215, 4.365.519, 4.677.846, 4.757.708 and 5.369.987 and International Patent GO1D 21/00, G01N 11/00 13 J. Happel and H. Brenner, Low Reynolds Number Hydrodynamics (Martinus Nijho, the Hague, Netherlands, 1986). 14 Frei, Ch. and Lscher, P. and Wintermantel, J. Fluid. Mech. 410, 185210 (2000).

R.C. Weast and M.J. Astle, Handbook of Chemistry and Physics (CRC Press, 1982). 16 J. B. Segur and H. E. Oberstar, Ind. Eng. Chem. 43, 2117 (1951). 17 L. W. Bosart and A. O. Snoddy, Ind. Eng. Chem. 20, 1377 (1928).

15

Cheng N-S, Formula for the viscosity of a glycerolwater mixture, Ind. Eng. Chem. 47, 3285-3288 (2008) 19 Timoshenko, S. P., and Gere, J. M., Theory of Elastic Stability, 2 ed., McGraw-Hill, (1961). 20 P. Guillot, P. Panizza, J-B. Salmon, M. Joanicot, A. Colin, C-H. Bruneau and T. Colin, Langmuir, 22, 64386445, (2006).

18