Document [original]

micromachines

Article

In-Line Analysis of Diffusion Processes in Micro Channels by

Long Distance Raman Photometric Measurement

Technology—A Proof of Concept Study

Julian Deuerling 1,*,† , Shaun Keck 1,†, Inasya Moelyadi 1, Jens-Uwe Repke 2and Matthias Rädle 1





Citation: Deuerling, J.; Keck, S.;

Moelyadi, I.; Repke, J.-U.; Rädle, M.

In-Line Analysis of Diffusion

Processes in Micro Channels by Long

Distance Raman Photometric

Measurement Technology—A Proof

of Concept Study. Micromachines 2021,

12, 116. https://doi.org/10.3390/

mi12020116

Received: 14 December 2020

Accepted: 19 January 2021

Published: 22 January 2021

Publisher’s Note: MDPI stays neutral

with regard to jurisdictional claims in

published maps and institutional affil-

iations.

Licensee MDPI, Basel, Switzerland.

This article is an open access article

distributed under the terms and

conditions of the Creative Commons

Attribution (CC BY) license (https://

creativecommons.org/licenses/by/

4.0/).

1Center for Mass Spectrometry and Optical Spectroscopy, Mannheim University of Applied Sciences,

68163 Mannheim, Germany; [email protected] (S.K.); [email protected] (I.M.);

[email protected] (M.R.)

2Process Dynamics and Operation, Technical University of Berlin, Fakultät 3 Prozesswissenschaften,

10623 Berlin, Germany; [email protected]

*Correspondence: [email protected]

† These authors contributed equally to this paper.

Abstract:

This work presents a novel method for the non-invasive, in-line monitoring of mixing

processes in microchannels using the Raman photometric technique. The measuring set-up distin-

guishes itself from other works in this field by utilizing recent state-of-the-art customized photon

multiplier (CPM) detectors, bypassing the use of a spectrometer. This addresses the limiting factor of

integration times by achieving measuring rates of 10 ms. The method was validated using the ternary

system of toluene–water–acetone. The optical measuring system consists of two functional units: the

coaxial Raman probe optimized for excitation at a laser wavelength of 532 nm and the photometric

detector centered around the CPMs. The spot size of the focused laser is a defining factor of the

spatial resolution of the set-up. The depth of focus is measured at approx. 85

m with a spot size of

approx. 45

m, while still maintaining a relatively high numerical aperture of 0.42, the latter of which

is also critical for coaxial detection of inelastically scattered photons. The working distance in this

set-up is 20 mm. The microchannel is a T-junction mixer with a square cross section of 500 by 500

a hydraulic diameter of 500

m and 70 mm channel length. The extraction of acetone from toluene

into water is tracked at an initial concentration of 25% as a function of flow rate and accordingly

residence time. The investigated flow rates ranged from 0.1 mL/min to 0.006 mL/min. The resi-

dence times from the T-junction to the measuring point varies from 1.5 to 25 s. At 0.006 mL/min a

constant acetone concentration of approx. 12.6% was measured, indicating that the mixing process

reached the equilibrium of the system at approx. 12.5%. For prototype benchmarking, comparative

measurements were carried out with a commercially available Raman spectrometer (RXN1, Kaiser

Optical Systems, Ann Arbor, MI, USA). Count rates of the spectrophotometer surpassed those of

the spectrometer by at least one order of magnitude at identical target concentrations and optical

power output. The experimental data demonstrate the suitability and potential of the new measuring

system to detect locally and time-resolved concentration profiles in moving fluids while avoiding

external influence.

Keywords:

process engineering; fluid-fluid extraction; Raman effect; photometry; microchannels;

droplets; local concentration measurement; optical measurement

1. Introduction

This article presents a new and fast method for the non-invasive, in-line quality control

of local and time-dependent chemical composition and flow regime in microchannels by

means of inelastic light scattering. All measurements were carried out at the Center for

Mass Spectrometry and Optical Spectroscopy, an interfaculty institution of the University

of Applied Sciences in Mannheim, Germany.

Micromachines 2021,12, 116. https://doi.org/10.3390/mi12020116 https://www.mdpi.com/journal/micromachines

Micromachines 2021,12, 116 2 of 13

Temporally and spatially-resolved optical measurement techniques show great poten-

tial for improving insights into flow characteristics, mixing processes and therefore reaction

monitoring, as well as process control in microchannels [

–

]. Established methods in the

field of flow monitoring such as common image analysis provide 2-dimensional images,

acquired by reflection or when using microchannels, transmission by means of specifically

adapted and manufactured set-ups. This of course is predicated on both the bottom and

lid of the channel being transparent [1].

There are various approaches to increasing image contrast, e.g., limiting the depth of

field in conventional image analysis, since a narrowly defined detection plane results in a

sharp image. Measuring a transmitted signal only gives information on mean concentration

along the transmission axis. Although complex tomographic 3D-scanning instruments

suppress this effect, increased acquisition times will be the consequence of capturing the

necessarily high number of images [1].

Methods which combine microscopy and image analysis have made the spatial distri-

bution of fluids and the temporal changes thereof accessible. Introducing contrasting agents

into moving phases improves optical contrast. Although fluorescence markers require only

low concentrations, they lack selectivity and possibly influence the fluids’ flow. Measuring

auto-fluorescence is only rarely possible [

]. Consequently, the aforementioned techniques

are unable to reliably detect local and temporal shifts in concentrations and are therefore

insufficient for consistent tracking of mixtures, detecting possible occurrences of inho-

mogeneity, deviations from desired reactions and the formation of by-products. To fully

address all of these needs it is paramount to have access to quantitative, molecule-specific

information with a high time and local resolution without imposing external restraints on

the flow regime.

Raman scattering is becoming more frequently used in microfluidics as a means of

detecting target concentrations [

–

]. However, measuring techniques predominantly rely

on spectrometers, as well as small focal lengths. Additionally, spatial resolution remains

challenging. Consequently, the integration time for acquiring quantifiable spectra often lies

in the range of seconds due to the low yield of the Raman effect. Fletcher et al. investigated

the synthesis of ethyl acetate from ethanol and acetic acid inside a T-shaped channel

using a Jobin Yvon Raman spectrometer and confocal optics, reporting acquisition times

of 5 s per spectrum [

]. Rinke et al. monitored the hydrolysis of 2,2-dimethoxypropane

to acetone and methanol in a T-shaped micromixer with a Leica microscope coupled to a

Renishaw Raman spectrometer, with integration times ranging from 10 s to 60 s [

]. Gal-Or

et al. presented the acquisition of Raman spectra of cyclohexane and ethyl acetate from

3D printed microfluidic devices by means of time-gated Raman spectrometry through

picosecond-pulsed laser excitation, with measurement times of approximately 6 min per

sample [4].

The presented measuring set-up distinguishes itself by utilizing recent state of the

art customized photon multiplier (CPM) detectors and a long-distance non-contact probe,

thereby addressing the issue of limiting integration times of spectrometers, while achieving

the necessary spatial resolution for application in microchannel geometries. In this work,

we will show that all criteria can be met using application-specific, preselected Stokes-

Raman-Shifted information in a newly designed spectrophotometric Raman set-up.

2. Materials and Methods

The steadily increasing application of microstructured components in process en-

gineering is a clear indicator of the growing importance of these elements in chemical,

pharmaceutical and life sciences. Specific designs include heat exchangers, reactors, static

mixers and others, and thus vary in size and shape [

]. The essential feature in all applica-

tions is the extremely high surface-to-volume ratio, stemming from small linear dimensions,

which is the basis for the majority of advantages over conventional sized chemical process

equipment [

]. This property improves selectivity and increases the yield of the chemical

reaction, while also facilitating a more efficient heat transfer [

]. Furthermore, the diver-

Micromachines 2021,12, 116 3 of 13

gent behavior of fluids in microscale dimensions in comparison to flow in macroscopic

geometries must be considered. Due to dominating viscous effects, smaller dimensions

result in low Reynolds numbers, which in turn impedes turbulent flow instabilities [

Accordingly, Newtonian fluids inside microstructured components such as microchannels

will form laminar flow, along with its limiting influences on mass transfer and mixing [5].

All mixing experiments were carried out in an aluminium T-junction microchan-

nel (CN AW 2007 AlCu4PbMgMn). This alloy is specifically designed for milling while

maintaining high tensile strength and is therefore easy to handle. Moreover, there are no

interactions of this material with the fluid chemicals used in the experiments, making it

ideally suited for the investigated fluid system. To reduce reflections and optimize wall

flow the channel surfaces were homogenized by ball peening with spherical glass beads

(40–70

m; DIN 52 900). The microchannel had a square cross section with dimensions

of 500 by 500

m, a hydraulic diameter of 500

m and 70 mm channel length, as shown

in Figure 1.

Micromachines 2021, 12, x FOR PEER REVIEW 3 of 13

tions is the extremely high surface-to-volume ratio, stemming from small linear dimen-

sions, which is the basis for the majority of advantages over conventional sized chemical

process equipment [8]. This property improves selectivity and increases the yield of the

chemical reaction, while also facilitating a more efficient heat transfer [7]. Furthermore,

the divergent behavior of fluids in microscale dimensions in comparison to flow in mac-

roscopic geometries must be considered. Due to dominating viscous effects, smaller di-

mensions result in low Reynolds numbers, which in turn impedes turbulent flow instabil-

ities [9]. Accordingly, Newtonian fluids inside microstructured components such as mi-

crochannels will form laminar flow, along with its limiting influences on mass transfer

and mixing [5].

All mixing experiments were carried out in an aluminium T-junction microchannel

(CN AW 2007 AlCu4PbMgMn). This alloy is specifically designed for milling while main-

taining high tensile strength and is therefore easy to handle. Moreover, there are no inter-

actions of this material with the fluid chemicals used in the experiments, making it ideally

suited for the investigated fluid system. To reduce reflections and optimize wall flow the

channel surfaces were homogenized by ball peening with spherical glass beads (40–70

µm; DIN 52 900). The microchannel had a square cross section with dimensions of 500 by

500 µm, a hydraulic diameter of 500 µm and 70 mm channel length, as shown in Figure 1.

Figure 1. Top down view of the aluminium T-junction microchannel with two inlet and one out-

put channels; square cross section of 500 × 500 µm, hydraulic diameter of 500 µm and 70 mm chan-

nel length (acquired with Keyence VHX-7000, Neu-Isenburg, Germany).

The microchannel was covered with a transparent quartz lid for both visual and laser

access to the flowing phases, as well as detecting inelastic scattered photons. Quartz glass

causes a comparatively low background signal in regard to the common laboratory and

industrial alternative borosilicate. To ensure leak tightness, the lid was fixed to the base

body with a UV adhesive (Norland Optical Adhesive 61 LOT 415, Cranbury, NJ, USA).

The adhesive was applied to the base body near the microchannel edges and was subse-

quently cured by UV light irradiation (Philips BLB F8T5, Amsterdam, The Netherlands).

The reason for using this specific microchannel over conventional microfluidic devices

was to verify the measurement system itself, before applying the system to more complex

microchannels.

The optical measuring system for monitoring of mixing consists of two functional

units: the coaxial Raman probe with laser coupling and the spectrophotometric detector.

The complete system is schematically depicted in Figure 2.

Figure 1.

Top down view of the aluminium T-junction microchannel with two inlet and one output

channels; square cross section of 500

500

m, hydraulic diameter of 500

m and 70 mm channel

length (acquired with Keyence VHX-7000, Neu-Isenburg, Germany).

The microchannel was covered with a transparent quartz lid for both visual and laser

access to the flowing phases, as well as detecting inelastic scattered photons. Quartz glass

causes a comparatively low background signal in regard to the common laboratory and in-

dustrial alternative borosilicate. To ensure leak tightness, the lid was fixed to the base body

with a UV adhesive (Norland Optical Adhesive 61 LOT 415, Cranbury, NJ, USA). The adhe-

sive was applied to the base body near the microchannel edges and was subsequently cured

by UV light irradiation (Philips BLB F8T5, Amsterdam, The Netherlands). The reason for

using this specific microchannel over conventional microfluidic devices was to verify the

measurement system itself, before applying the system to more complex microchannels.

The optical measuring system for monitoring of mixing consists of two functional

units: the coaxial Raman probe with laser coupling and the spectrophotometric detector.

The complete system is schematically depicted in Figure 2.

Micromachines 2021,12, 116 4 of 13

Micromachines 2021, 12, x FOR PEER REVIEW 4 of 13

Figure 2. Optical set-up: laser stage #1–3, probe stage #4–7 and #9–12, microchannel #8, photomet-

ric detection stage #13–18, software #19.

Both the excitation of Raman scattering and the collection of photons was realized by

means of a specially designed non-invasive probe (#7 in Figure 2). Here, a fiber-coupled

laser emitting at 532 nm (gem532, P

max

= 500 mW, Coherent) in continuous wave mode

was focused into the flowing liquid using a dichroic mirror and a long working distance

(20 mm) microscope objective (Mitutoyo Apochromatic Objective, MY20X-804, Kawasaki,

Japan) (#1–8 in Figure 2). The Raman effect was excited instantaneously and spherical

scattering took place. The objective then collected inelastically scattered photons into a

collimated beam, limited by its numerical aperture (NA = 0.42). The increased wavelength

of Stokes-shifted Raman photons enabled passage through the dichroic mirror and block-

ing filter (#6,9,10 in Figure 2; AHF Analysetechnik). Photons were then coupled into a

glass fiber (#12 in Figure 2). The probe was mounted vertically on a three-dimensional

stage (Steinmeyer, MP130-50, Albstadt, Germany) with three linear axes for positioning

during top down measuring. Detection of Stokes-shifted photons was accomplished using

a novel spectrophotometric hybrid combining two key elements, a diffracting optic and a

parallel alignment of glass fibers (#14, 15, 16 in Figure 2). Lenses (#2, 4, 12 and 13 in Figure

2) positioned before and after the glass fibers were used to focus or collimate the light.

Inelastically scattered photons from inside the microchannel were relayed by glass fiber

from the coaxial probe into a collimated beam and projected onto a holographic grating

(grating constant 1800 mm

-1

, Thorlabs, Newton, USA). This type of grating characteristi-

cally has a low occurrence of periodic errors and low stray light; the latter is especially

beneficial in measurements with a critical signal-to-noise ratio such as Raman spectros-

copy. The emergent angle of incident photons from the grating surface was a function of

the wavelength, enabling wavelength specific selection. The originally circular cross sec-

tion of the beam was then focused in a linear shape onto the glass fibers. Fibers transfer-

ring relevant molecular information were then connected to customized photon multipli-

ers (CM92N, Proxivision, Bensheim, Germany, #17 in Figure 2) and the information was

then processed in additional hardware and software steps (#18, 19 in Figure 2). Optical

power was measured at 70% of set power level at the probe head using a PS 19 thermopile

sensor (Coherent, Santa Clara, CA, USA) in combination with a FieldMax II. The linearity

of the excitation power was confirmed for the set-up with R

= 0.9969, measuring the peak

intensity of acetone at 1710 cm

-1

Raman shift at various power settings.

Figure 2.

Optical set-up: laser stage #1–3, probe stage #4–7 and #9–12, microchannel #8, photometric

detection stage #13–18, software #19.

Both the excitation of Raman scattering and the collection of photons was realized by

means of a specially designed non-invasive probe (#7 in Figure 2). Here, a fiber-coupled

laser emitting at 532 nm (gem532, P

max

= 500 mW, Coherent) in continuous wave mode was

focused into the flowing liquid using a dichroic mirror and a long working distance (20 mm)

microscope objective (Mitutoyo Apochromatic Objective, MY20X-804, Kawasaki, Japan)

(#1–8 in Figure 2). The Raman effect was excited instantaneously and spherical scattering

took place. The objective then collected inelastically scattered photons into a collimated

beam, limited by its numerical aperture (NA = 0.42). The increased wavelength of Stokes-

shifted Raman photons enabled passage through the dichroic mirror and blocking filter

(#6,9,10 in Figure 2; AHF Analysetechnik). Photons were then coupled into a glass fiber

(#12 in Figure 2). The probe was mounted vertically on a three-dimensional stage (Stein-

meyer, MP130-50, Albstadt, Germany) with three linear axes for positioning during top

down measuring. Detection of Stokes-shifted photons was accomplished using a novel

spectrophotometric hybrid combining two key elements, a diffracting optic and a parallel

alignment of glass fibers (#14, 15, 16 in Figure 2). Lenses (#2, 4, 12 and 13

in Figure 2

)

positioned before and after the glass fibers were used to focus or collimate the light. Inelas-

tically scattered photons from inside the microchannel were relayed by glass fiber from the

coaxial probe into a collimated beam and projected onto a holographic grating (grating

constant 1800 mm

−1

, Thorlabs, Newton, USA). This type of grating characteristically has a

low occurrence of periodic errors and low stray light; the latter is especially beneficial in

measurements with a critical signal-to-noise ratio such as Raman spectroscopy. The emer-

gent angle of incident photons from the grating surface was a function of the wavelength,

enabling wavelength specific selection. The originally circular cross section of the beam was

then focused in a linear shape onto the glass fibers. Fibers transferring relevant molecular

information were then connected to customized photon multipliers (CM92N, Proxivision,

Bensheim, Germany, #17 in Figure 2) and the information was then processed in additional

hardware and software steps (#18, 19 in Figure 2). Optical power was measured at 70%

of set power level at the probe head using a PS 19 thermopile sensor (Coherent, Santa

Clara, CA, USA) in combination with a FieldMax II. The linearity of the excitation power

was confirmed for the set-up with R

= 0.9969, measuring the peak intensity of acetone at

1710 cm−1Raman shift at various power settings.

Micromachines 2021,12, 116 5 of 13

The spot size and depth of focus of the laser focus are a defining factor of the spatial

resolution of the set-up and these were measured using a Scanning-Split Optical Beam

Profiler (BP209-VIS/M, Thorlabs). Several spot sizes were recorded at defined intervals

along the optical axis. It should be noted that these measurements were done in free space.

Since the refraction index of water and toluene is higher than air, the actual laser spot size

will be smaller, thereby increasing the spatial resolution. The depth of focus was measured

at approx. 85

m with a spot size of approx. 45

m (Figure 3). The working distance in this

set-up was 20 mm and therefore suitable for application in microfluidics.

Micromachines 2021, 12, x FOR PEER REVIEW 5 of 13

The spot size and depth of focus of the laser focus are a defining factor of the spatial

resolution of the set-up and these were measured using a Scanning-Split Optical Beam

Profiler (BP209-VIS/M, Thorlabs). Several spot sizes were recorded at defined intervals

along the optical axis. It should be noted that these measurements were done in free space.

Since the refraction index of water and toluene is higher than air, the actual laser spot size

will be smaller, thereby increasing the spatial resolution. The depth of focus was meas-

ured at approx. 85 µm with a spot size of approx. 45 µm (Figure 3). The working distance

in this set-up was 20 mm and therefore suitable for application in microfluidics.

Figure 3. Laser-spot geometry (532 nm laser from coherent exiting probe head (#7 in Figure 2) in

water).

Customized photon multipliers were selected specifically for their applicability in

low light detection. Their key features are extremely low background noise, low light level

detection limits and high dynamic range and gain. In CPMs, sensitivity for large spectral

ranges can be achieved by adapting the cathode material. With a signal amplification of

, they are among the most sensitive photodetectors. Furthermore, their low noise of <

10 cps (counts per second) is favorable when detecting low-yield Raman photons. The

weak point of photon multipliers in general is their sensitivity to strong light influences

(e.g., daylight, stray light, reflected light). The resulting high number of primary electrons

leads to the destruction of the dynodes. In CPMs, such overexposure is prevented by an

automatic quenching system that closes the entrance slit with a reaction time of 180 ns.

The substance system selected for the experimental series was toluene, acetone and

water. As shown in Figure 4, the different fluids were pumped with a syringe pump into

the microchannel at different flow rates.

Figure 3.

Laser-spot geometry (532 nm laser from coherent exiting probe head (#7 in Figure 2)

in water).

Customized photon multipliers were selected specifically for their applicability in low

light detection. Their key features are extremely low background noise, low light level

detection limits and high dynamic range and gain. In CPMs, sensitivity for large spectral

ranges can be achieved by adapting the cathode material. With a signal amplification of 10

they are among the most sensitive photodetectors. Furthermore, their low noise of < 10 cps

(counts per second) is favorable when detecting low-yield Raman photons. The weak point

of photon multipliers in general is their sensitivity to strong light influences (e.g., daylight,

stray light, reflected light). The resulting high number of primary electrons leads to the

destruction of the dynodes. In CPMs, such overexposure is prevented by an automatic

quenching system that closes the entrance slit with a reaction time of 180 ns.

The substance system selected for the experimental series was toluene, acetone and

water. As shown in Figure 4, the different fluids were pumped with a syringe pump into

the microchannel at different flow rates.

Micromachines 2021,12, 116 6 of 13

Micromachines 2021, 12, x FOR PEER REVIEW 6 of 13

Figure 4. Fluid set-up: syringe pump #1, syringe with acetone and toluene #2, syringe with water

#3, microchannel #4, waste container with mineral oil as diffusion barrier #5.

The pure substance Raman spectra were measured with a Raman spectrometer

(RXN1, Kaiser Optical Systems, Ann Arbor, MI, USA) (see Figure 5).

Figure 5. Raman spectra of toluene, acetone and water recorded with RXN1 spectrometer (Kaiser

Optical Systems). The acetone peak at 1710 cm-1 used for concentration determination is framed.

Spectra are offset by 500 and 1000 counts.

This specific combination was chosen for several reasons. Firstly, the three-phase liq-

uid–liquid extraction of this system is well documented in the literature [10–13]. Secondly,

the spectral peak of acetone at 1710 cm−1 does not interfere with the peaks of toluene or

water, as shown in Figures 5 and 6. Moreover, the detection glass fibers had a spectral

bandwidth of approx. 10 nm, which needed to be considered when aligning the central

wavelengths of fibers. For these reasons the 1710 cm−1 acetone peak was suited for the

detection despite its lower signal intensity, since in the investigated ternary system there

was no overlap of any other peaks in the spectrum while using the spectrophotometric

prototype. The band assignment (see Figure 5) for the substances used in these experi-

ments was as follows in Table 1:

Figure 4.

Fluid set-up: syringe pump #1, syringe with acetone and toluene #2, syringe with water #3, microchannel #4,

waste container with mineral oil as diffusion barrier #5.

The pure substance Raman spectra were measured with a Raman spectrometer (RXN1,

Kaiser Optical Systems, Ann Arbor, MI, USA) (see Figure 5).

Micromachines 2021, 12, x FOR PEER REVIEW 13 of 13

Funding: German Federal Ministry of Education and Research, grant number 13FH8I03IA, funded

this research.

Data Availability Statement: The data presented in this study are available on request from the

corresponding author.

Acknowledgments: In this article, the authors draw on contributions from many members of the

CeMOS Research Center. In particular, we would like to thank Stefan Schorz, Steffen Manser and

Julia Siber for providing technology and software. All images and plots without source were created

at the CeMOS institute–Centre for Mass Spectrometry and Optical Spectroscopy, 68163 Mannheim,

Germany.

Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the

design of the study, in the collection, analyses or interpretation of data, in the writing of the

manu-script or in the decision to publish the results.

References

Figure 5.

Raman spectra of toluene, acetone and water recorded with RXN1 spectrometer (Kaiser Optical Systems).

The acetone peak at 1710 cm−1used for concentration determination is framed. Spectra are offset by 500 and 1000 counts.

This specific combination was chosen for several reasons. Firstly, the three-phase

liquid–liquid extraction of this system is well documented in the literature [

–

Secondly, the spectral peak of acetone at 1710 cm

−1

does not interfere with the peaks

of toluene or water, as shown in Figures 5and 6. Moreover, the detection glass fibers had a

spectral bandwidth of approx. 10 nm, which needed to be considered when aligning the

central wavelengths of fibers. For these reasons the 1710 cm

−1

acetone peak was suited for

the detection despite its lower signal intensity, since in the investigated ternary system there

was no overlap of any other peaks in the spectrum while using the spectrophotometric

prototype. The band assignment (see Figure 5) for the substances used in these experiments

was as follows in Table 1:

Micromachines 2021,12, 116 7 of 13

Micromachines 2021, 12, x FOR PEER REVIEW 7 of 13

Table 1. Raman band assignment.

Water (cm-1). Toluene (cm-1) Acetone (cm-1) Type of Vibration

622, 787, 1005, 1091, 1211 787 C-H deformation

1710 C-O stretching

2921 C-H/C-H2 stretching

3066 C-H/C-H2 stretching

>3200 O-H stretching

Figure 6. Section of Raman spectra of toluene, acetone and water. Spectral bandwidth of detection

fibers are framed with vertical lines (bandwidth of maximum and minimum signal ~300 cm-1

each). Spectra are offset by 500 and 1000 counts.

The spectral bandwidth of individual fibers translated to approx. 300 cm−1 in the cur-

rent set-up. Peripheral fibers showed a slightly reduced signal intensity, stemming from

a curvature in the focal plane. This was compensated for by using the signal intensity of

the highest investigated concentration as a fixed reference for signal processing.

Figure 6 shows a section of each Raman spectrum of the water–toluene–acetone sys-

tem, with the solid and dotted vertical lines indicating the spectral ranges of the target

and reference fibers. The overlap stems from a magnification of the terminal detection

optics. In the investigated ternary system and the selected target range, this circumstance

can be used to improve background signal correction, since the slope of the background

can be better approximated as a photometric count rate due to the spectral proximity of

the reference channel to the target channel. Using the prototype, binary mixtures were

measured ranging from pure acetone to pure water with a minimum acetone concentra-

tion of 1.56% at an optical power output of 300 mW. Each mixture had a pre-selected,

defined concentration and was continuously introduced into the microchannel. Every

concentration consisted of 12000 individual data points. The signal was offset-corrected

for empty conditions at approximately 5 counts. Calibration confirmed the linear behavior

of the Raman effect as a function of concentration, with a Pearson coefficient of R² = 0.9775.

The signal was then scaled to the maximum value of pure acetone, as depicted in the fol-

lowing Figure 7. Scaling facilitated the comparability of different experimental data sets,

which varied between individual experiments while being self-consistent in each experi-

ment.

Figure 6.

Section of Raman spectra of toluene, acetone and water. Spectral bandwidth of detection fibers are framed with

vertical lines (bandwidth of maximum and minimum signal ~300 cm−1each). Spectra are offset by 500 and 1000 counts.

Table 1. Raman band assignment.

Water (cm−1) Toluene (cm−1) Acetone (cm−1)Type of Vibration

622, 787, 1005, 1091, 1211 787 C-H deformation

1710 C-O stretching

2921 C-H/C-H2stretching

3066 C-H/C-H2stretching

>3200 O-H stretching

The spectral bandwidth of individual fibers translated to approx. 300 cm

−1

in the

current set-up. Peripheral fibers showed a slightly reduced signal intensity, stemming from

a curvature in the focal plane. This was compensated for by using the signal intensity of

the highest investigated concentration as a fixed reference for signal processing.

Figure 6shows a section of each Raman spectrum of the water–toluene–acetone

system, with the solid and dotted vertical lines indicating the spectral ranges of the target

and reference fibers. The overlap stems from a magnification of the terminal detection

optics. In the investigated ternary system and the selected target range, this circumstance

can be used to improve background signal correction, since the slope of the background

can be better approximated as a photometric count rate due to the spectral proximity of

the reference channel to the target channel. Using the prototype, binary mixtures were

measured ranging from pure acetone to pure water with a minimum acetone concentration

of 1.56% at an optical power output of 300 mW. Each mixture had a pre-selected, defined

concentration and was continuously introduced into the microchannel. Every concentration

consisted of 12000 individual data points. The signal was offset-corrected for empty

conditions at approximately 5 counts. Calibration confirmed the linear behavior of the

Raman effect as a function of concentration, with a Pearson coefficient of R

= 0.9775.

The signal was then scaled to the maximum value of pure acetone, as depicted in the

following

Figure 7

. Scaling facilitated the comparability of different experimental data

sets, which varied between individual experiments while being self-consistent in each

experiment.

Micromachines 2021,12, 116 8 of 13

Micromachines 2021, 12, x FOR PEER REVIEW 8 of 13

Figure 7. Calibration data of acetone signal intensity as a function of concentration measured dur-

ing flow inside the microchannel. Data are scaled to maximum acetone intensity for comparability

of different data sets.

For validation, raw data measured at different flow rates were treated accordingly

and compared to this calibration curve.

3. Results

In a first proof-of-principal experiment, acetone mixed with toluene at a ratio of one

to three was introduced into the microchannel opposite pure water as the extracting sol-

vent. A representative section of 24 s extracted from monitoring of these flow conditions

(flow rate 0.1 mL/min) is depicted in Figure 8.

Figure 8. Measured acetone concentration in toluene droplets alternating with water droplets

(flow rate 0.1 mL/min with 1.5 s time delay after contact of toluene and water).

Figure 7.

Calibration data of acetone signal intensity as a function of concentration measured during flow inside the

microchannel. Data are scaled to maximum acetone intensity for comparability of different data sets.

For validation, raw data measured at different flow rates were treated accordingly

and compared to this calibration curve.

3. Results

In a first proof-of-principal experiment, acetone mixed with toluene at a ratio of one to

three was introduced into the microchannel opposite pure water as the extracting solvent.

A representative section of 24 s extracted from monitoring of these flow conditions (flow

rate 0.1 mL/min) is depicted in Figure 8.

Micromachines 2021, 12, x FOR PEER REVIEW 8 of 13

Figure 7. Calibration data of acetone signal intensity as a function of concentration measured dur-

ing flow inside the microchannel. Data are scaled to maximum acetone intensity for comparability

of different data sets.

For validation, raw data measured at different flow rates were treated accordingly

and compared to this calibration curve.

3. Results

In a first proof-of-principal experiment, acetone mixed with toluene at a ratio of one

to three was introduced into the microchannel opposite pure water as the extracting sol-

vent. A representative section of 24 s extracted from monitoring of these flow conditions

(flow rate 0.1 mL/min) is depicted in Figure 8.

Figure 8. Measured acetone concentration in toluene droplets alternating with water droplets

(flow rate 0.1 mL/min with 1.5 s time delay after contact of toluene and water).

Figure 8. Measured acetone concentration in toluene droplets alternating with water droplets (flow rate 0.1 mL/min with

1.5 s time delay after contact of toluene and water).

After mixing, droplets flowed through the microchannel until they passed the focal

point where the concentration was measured. The signal for these experimental conditions

periodically alternated between a maximum signal of ~24% (acetone concentration in

Micromachines 2021,12, 116 9 of 13

toluene) and 0.9% (acetone concentration in water), both measured at a 10 mm channel

length. The fixed measuring position and the varying flow rates resulted in differing time

delays. The delays increased with the reduction of the flow rate and are given in the

captions of Figures 8–13. Although droplets of toluene and water were distinguishable,

this indicates a negligible mass transfer of acetone from toluene to water at the given flow-

rate. Consequently, to further investigate mass transfer, the flow rate was progressively

decreased, thereby increasing the residence time. At 0.06 mL/min, this resulted in a

maximum concentration of ~21.7% acetone in the toluene phase and ~2.15% in the water

phase (see Figure 9).

Micromachines 2021, 12, x FOR PEER REVIEW 9 of 13

After mixing, droplets flowed through the microchannel until they passed the focal

point where the concentration was measured. The signal for these experimental condi-

tions periodically alternated between a maximum signal of ~24% (acetone concentration

in toluene) and 0.9% (acetone concentration in water), both measured at a 10 mm channel

length. The fixed measuring position and the varying flow rates resulted in differing time

delays. The delays increased with the reduction of the flow rate and are given in the cap-

tions of Figures 8 to 13. Although droplets of toluene and water were distinguishable, this

indicates a negligible mass transfer of acetone from toluene to water at the given flow-

rate. Consequently, to further investigate mass transfer, the flow rate was progressively

decreased, thereby increasing the residence time. At 0.06 mL/min, this resulted in a max-

imum concentration of ~21.7% acetone in the toluene phase and ~2.15% in the water phase

(see Figure 9).

Figure 9. Measured acetone concentration in toluene droplets alternating with water droplets (0.06

mL/min with 2.5 s time delay after contact of toluene and water).

At a flow rate of 0.04 mL/min (see Figure 10) there was slight decrease in the maxi-

mum signal to ~19.15% and an equal increase in the minimum signal to ~3.65%. The time

periods of both the minimum and maximum signals were elongated, as was expected.

Measurement artefacts appeared at some of the droplet boundaries.

Figure 10. Measured acetone concentration in toluene droplets alternating with water droplets

(0.04 mL/min with 3.75 s time delay after contact of toluene and water).

Figure 9.

Measured acetone concentration in toluene droplets alternating with water droplets (0.06 mL/min with 2.5 s time

delay after contact of toluene and water).

At a flow rate of 0.04 mL/min (see Figure 10) there was slight decrease in the maximum

signal to ~19.15% and an equal increase in the minimum signal to ~3.65%. The time