Automated system to acquire fluorescence, polarization and anisotropy maps within liquid flows

Maps of polarization and anisotropy can be helpful for flow analysis systems (FIA, CFA, etc.) with reactions dependent on the intermolecular alignment as well as for dispersion control. Maps can be acquired manually, but when a scan over a sample area is required, the acquisition becomes tiresome and has low precision. The paper describes an automatic flexible system for high-precision sample positioning with closed loop self control, remote data acquisition and storage controlled by a BASIC program. The system was developed to acquire maps up to 850 mm2 of the sample (liquid flows, solids, interfaces, etc.), with up to 100 μm2 precision. To evaluate the equipment, performance is presented as the scan of a thin liquid film of monoethylene glycol (MEG) flowing on borosilicate. Tests were performed with and without surfactantes at submicellar concentrations: two concentrations of sodium dodecyl sulphate (SDS) and one of polyethylene oxide (PEO). For pure MEG, the intermolecular alignment initially increased, then decreased. When SDS was added, both polarization and anisotropy only increased progressively with the flow. This might be explained by the surfactant decrease of interfacial interaction. When PEO was added, both polarization and anisotropy decreased pronouncedly over the entire map, which might be due to macromolecular aggregates within the bulk generating misaligned molecular domains. The system presented as sample positioning repeatability of 0.1% and a high polarization reproducibility (error margin < 6 in 1000).


Introduction
Usually,¯uorescence spectroscopy monitors the sample in a single spot as a function of several factors like wavelength, time, incident light intensity, temperature, etc. This is the case of typical analysis methods like¯ow injection (FI) [1] and enzyme-linked immunosorbent assays (ELISA) [2]. For some applications, it is also necessary to map the sample by acquiring the¯uorescence at several points.
Most of the work in the literature studies laser-induced uorescence (LIF). Nevertheless, some applications require the knowledge of the intermolecular alignment within the sample, as in the case of tensions shear stress during resin extrusion within polymer melts [3] and of dynamic interfacial tension dependence of a liquid owing on surfaces with diå erent chemical constitution [4]. This information can be obtained by¯uorescence depolarization.
Although there are several works in the literature that usē uorescence to monitor liquid¯ows as the¯ow progresses, those that study intermolecular alignment are not many. Kenyon et al. [5,6] and Quintella et al. [7,8] reported for the ® rst time steady-stat e¯uorescence depolarization maps of thin liquid¯ows. They used a nonautomated sample positioning with vertical and horizontal resolution of 0.25 and 0.05 mm, respectively. The operator visually acquired their data. Owing to the high amount of points in each map, the manual acquisition was lengthily and tiresome, and could take up to 12 h.
The present paper presents a fully automated laser spectrometer, which was developed to acquire highresolution maps of intermolecular alignment (polarization and anisotropy) and total¯uorescence, over a predetermined area of liquid¯ows, solids and interfaces.
Here is presented, as an application of the system developed, the¯ow on borosilicate of a thin liquid sheet of monoethylene glycol (MEG) with and without submicellar concentrations of surfactants.
Surfactants are present in several¯ow systems, such as in waste waters [9] where they may become environmental contaminants. They can be used to improve chemical analysis, sampling and storage like in polycyclic aromatic hydrocarbons (PAHs) in water [10] and may interfere in sample recovery. Their hydrophobic± hydrophilic properties make them perfect solubilizers for mixtures of polar and non-polar compounds, avoiding precipitation. Analytical methods may use surfactants with concentration either below or over critical micellar concentration (CMC), like micellar extraction [11], and enhancement of neutral analytes separation [12], which determination may involve¯ow injection analysis (FIA) [9,13]. In uorescence spectroscopy, surfactants increase the quantum yields of¯uorimetric sensors [14,15] as well as enhancing chemiluminescence of low¯uorescent reagents [16]. There are some rheological studies at premicellar and micellar concentrations [17] that have studied thē ow evolution at a single point.
Although surfactants are present in a wide range of chemical analysis processes, their contribution to the intermolecular orientation within the¯ow have only recently been addressed more thoroughly.

Theory
The theory has been presented before [5,6]. Brie¯y, vertically polarized laser light excites the¯uorescent probes within the sample. Their absorption is proportional to the cosine square of the angle between the molecular dipole and the laser electric ® eld. If the dipole is parallel to the longitudinal molecular axis, mainly nearly vertical molecules will absorb laser radiation and a photoselection of the probe occurs.
Fluorescence detection can be done by two laser spectroscopic techniques: Laser induced¯uorescence (LIF) and uorescence depolarization or polarized laser-induced uorescence (PLF) [4,6]. In LIF the¯uorescence emission is collected by a detection system regardless to its polarization and the data give direct information on how the probes align upon excitation. When the sample is a liquid¯owing vertically with shear imposed intermolecular alignment, the LIF signal is proportional to the fraction of molecules aligned along the¯ow direction at the excitation moment.
In PLF, the¯uorescence emission is discriminated according to its polarization and compared with the laser polarization [5,6]. While excited, the photoselected probes can rotate or not as a function of the chemical environment. When¯uorescence takes place, the light emitted is not 100% polarized as the probe transition dipole moment will depend on the molecular orientation of the probe.
The intermolecular alignment data can be interpreted either as a bidimensional phenomenon in terms of polarization (P) [5,6], or as a three-dimensional phenomenon in terms of anisotropy (r) [18]: Anisotropy r will depend on the local rotation diå usion time ¿ [19]. For steady-stat e¯uorescent depolarization it is given by: where r 0 is the initial anisotropy, ½ is the lifetime of the excited state, ² is the viscosity, V is the volume rotating unit, R is the gas constant and T is the temperature.
If the excited state lifetime is of the same order of the probe rotational diå usion, then¯uorescent probes become eae cient intermolecular alignment sensors. Thus, the¯uorescent species behaves as a medium of varying birefringence that permits molecular-level observation of uid¯ow, the birefringence being particularly sensitive to the velocity gradient within the¯uid.

Experimental
The equipment is an ensemble of mechanical hardware, electronic hardware, optic devices, a light detection system and software. The main idea here was to develop a technological tool capable of a high level of¯exibility and precision in sample positioning, closed loop self control, and high-density data acquisition and storage.

Automatic sample positioning system
The spectrometer can scan samples up to 30.9 mm wide and 27.4 mm long, covering a ¹ 850 mm 2 surface, with a ® ne precise path across the stream (X) and down stream (Y). This constraint imposes the need of an especially high precision and high repeatability mechanical hardware, allied to a¯exible positioning control software.
At each point, the intermolecular alignment data is automatically acquired.
The mechanical design uses two perpendicular stainlesssteel endless-screws assembled in a brass frame. The screws are driven by two independent (1.88) step-motors (1 and 2 in ® gure 1A) with 10 mm precision in each direction. The set-up positioning repeatability was determined by successive back and forth cycles. At each cycle end, the distance to the mechanical frame origin was determined using a digital calliper, being the positioning repeatability > 0:1%.
The sample holder was designed to be a very open and exible structure. The simplicity in its mechanical design allows for the improvisation of diå erent holding options, allowing the sample adjustment over 3 degrees of freedom: Y, angle in the XZ plane and the angle in the YZ plane.
The electric signals driving and controlling the stepmotors were generated by the data acquisition and control system (see below). An interlock system prevents any potential damaging caused by an improper manoeuvre.

Electronic hardware
The electronic hardware can be divided in two parts. The ® rst comprehends the two step-motors' interface. Figure 2 shows the command latch, the power drive outputs and the interlock buå er input. A software realtime driver was chosen to drive the motors instead of an onboard hardware driver. This choice was based on the search for a high¯exibility interfacing system, allowing diå erent drive options (full step, half step, low and high torque). This has been possible as the output from the photodiode (the¯uorescence detector) is a slow signal, leaving a long no operation delay between consecutive measures.
The second part of the electronic hardware consists on an analogue section where a high impedance low noise ampli® er prepares the¯uorescence signals to be digitized. The digitalization was performed by a Burr Brown ADS7804, 10 ms, analogue-to-digita l converter (ADC) [20]. A logic chip set allowed communication with a personal computer (PC) by the standard parallel printer port [21]. A Coherent Inova60 argon laser generated the 514.5 nm light beam. The laser intensity was kept at 50 mW, a safe value as saturation phenomena started over 80 mW. The laser light was de¯ected by a mirror (E) and focused into the sample (SPS), on a 0.02 mm 2 diameter spot, by a biconvex borosilicate lens (L 1 ) with a 400 mm focal length. A vertical Glan-Thompson polarizer (P 1 ) guarantees 100% polarization. The¯uorescent probes within the sample absorb the laser radiation. After a given time delay, depending on their excited state lifetime, theȳ uoresce. The¯uorescence is collected by a biconvex borosilicate lens (L 2 ), with 50.8 mm focal length, being focused into the photodiode (PD) at a 0.02 sr solid angle. A 550 nm high-pass ® lter (F) blocked the non-absorbe d laser light, avoiding light mixing and PD over¯ow.

Light detection system
To measure both vertical (I k ) and horizontal (I ? )¯uorescence components with the same optical set-up, the Conoptics 350-50 photo-elastic modulator (PEM), driven by a sinusoidal generator (G) at 600 Hz, periodically rotated the¯uorescence light electric axis around the propagation axis (Z). The PEM-modulated output light beam was selected by the horizontal Glan-Thompson polarizer (P 2 ). The PD detected an alternated signal V AC superposed to a continuous signal V DC .
The OPT301 Burr Brown photodiode [22] is an optoelectronic integrated circuit, containing a photodiode and a transimpedance ampli® er on a single di-electrically isolated chip. The integrated combination of photodiod e and ampli® er eliminated leakage current errors, noise pick-up and gain peaking. The light detection sensibility can be adjusted by a set of dipswitch and external resistors. The spectral responsivity is a broad curve covering from the near IR to the near UV, ensuring that the system can work in a very large light range, allowing the use of several¯uorescence probes.
The PD signal was split up in two by a passive R-RC network (SB). The direct coupled portion of this signal (V DC ) goes both to a digital voltmeter (M) and then to ADC 2 (® gure 2) in the interface (I). The capacitorcoupled portion of the photodiode signal V AC goes to the lock-in (AS) to be detected in phase with the reference sinusoidal PEM modulating signal. The ® ltered V AC is phase locked, ampli® ed and then sent to ADC 1 (® gure 2), in the interface (I), to be digitized.
V AC corresponds to the diå erence between vertical (I k ) and horizontal (I ? )¯uorescence components and V DC to their sum [5,6]. The polarization P is obtained as: where f C is the correction factor due to birefringence of optical components, being 0.20 for both vertical and horizontal polarized light.   Digital images can be acquired with a Hitachi VM-E230A video camera (36£ zoom) connected to the PC.

Software
As mentioned, the concepts behind this equipment design are full automation,¯exibility, precision, repeatability, easy data acquisition and reliable data storage. The PLF spectrometer was projected to be compatible with a small general propose PC and to have high level of portability.
In a hardware sense, the portability was reached using a fully IBM-PC interfacing design, specially in terms of the data communication channel, the standard Centronics parallel port (SPP) [23].
In a software sense, the conception scenario is a little diå erent. Although Microsoft Windows has been the leader for about 15 years, Linux now menaces its position. We have chosen an interpreted programming language, trying to be independent of a speci® c compiler. The code can be interpreted by the BASIC for MSDOS or, eventually, by BASIC for Linux. Figure 4 shows the simpli® ed software block diagram as the program itself has more than 500 code lines. In the start-up routine, the program searches for the SPP location in the PC IO-map and starts the communication protocol. All constants and positioning parameters are loaded up, passing to the auto-zeroing phase for both motors. Then, the program exhibits a dialogue screen asking for one of the operation choices: spectrum acquisition or motor displacement to an arbitrary position or a ® xed point, long-term data acquisition.
The spectrum acquisition routine starts showing a graphic screen (intensity versus position). During data acquisition, there are diå erent options in the operation mode: measuring delay, channel 1 gain, channel 2 gain, numerical oå -set, number of steps, X increment and Y increment. This interactive screen is used as a monitoring tool, where one can visualize in a rough way the data being acquired, i.e. V AC and V DC , and decide to save, or not, a given dataset. The acquisition routine is not a ® nal data-analysin g facility as the saved data should be the object of post-data treatment and analysis.
The motor displacement routine starts loading up the present position and asks for the new location. In that sense, it is almost a manual operation mode, where one can bias the XY frame origin. This routine is particularly useful during the experiment preparation and adds manȳ exibility and test possibilities. Throughout the duration of any motor displacement, the operation can be stopped and the automatic zeroing can be started.
The third option is the ® xed-point long-term data acquisition as a function of time. The subroutine starts asking for a position and, once there, keeps taking data. The operator determines the position and acquisition time intervals.

Data reproducibility and control tests
Polarization and anisotropy were monitored as a function of liquid¯ow temperature, PEM modulation frequency and acquisition time at a ® xed position.
A temperature increase produces an initial signal increase, stabilizes at a maximum value and then decreases (® gure 5A, C). When the temperature is changed, the dye probe rotational diå usion is also varied. For¯uorescence depolarization, the ideal value is when the rotational diå usion of the probe is of the same order of its excitation lifetime (see above). This is achieved between 14 and 17 8C.
The anisotropy and polarization response to modulation frequency was monitored from 10 to 1000 Hz. The frequency 600 Hz was chosen to avoid noise from other electrical systems.
The manual scan had a resolution of (0.35 mm), whilst the automatic had 100 mm 2 . The sample positioning repeatability was better than 0.1%. To validate the system, the ® rst lobe of a liquid thin jet was scanned in the same conditions as those presented by Quintella et al. [8].

Application
The sample system (® gure 1B) consisted of a liquid thin sheet¯owing on a Perfecta borosilicate microscope slide (75 £ 25 £ 0:8 mm, roughness < 5 nm) at 80 8 with the horizontal. Before the experiment, the slide was kept for 10 min in an ultrasound bath with ethanol and dried for 15 min in a stove at 50 8C.
Ethylene glycol (MEG) (Merck 10921, 99.5% purity) passed through a polished sapphire slit nozzle with 100 mm £6 mm, generating a thin free jet and impinging perpendicularly on the solid surface. The liquid¯owed at 220 cm s ¡1 constant average velocity. Its thickness ranged from 150 mm at the middle to 1000 mm at the borders. A home-made thermal bath kept the temperature at 15:0 § 0:5 8C.
Rhodamine 6G (R6G) (Lambdaphysik , 99.99% purity) was used as¯uorescent probe. It has transition dipole moment parallel to the molecule longitudinal direction [24] and was pumped from state S 0 to S 1 [25]. The concentration of 1:9 £ 10 ¡3 mol l ¡1 was chosen to maximize the signal and avoid quenching processes due to excitation migration like Fo« ster transfer [26]. Under this experimental condition, the viscosity was 0.225 g cm ¡1 s ¡1 . The rotational diå usion time and excitation lifetime are in the order of 4.2 and 4.9 ns, respectively [27].
Both SDS and PEO increased the liquid sheet width. It is well known [28] that surfactants decrease the liquid± solid interfacial tension, becoming good wetting agents. This eå ect is more pronounced in dynamic systems, as opposite to static, as surfactants tend to migrate to the surfaces, increasing their concentration locally.
Within pure MEG (® gure 6A), there is an initial decrease in polarization and anisotropy, followed by an increase as the¯ow progresses.
The initially low MEG polarization and anisotropy may be due to strong interfacial interactions with borosilicate [4]. The liquid has hydrogen atoms and hydroxyls, whilst the solid surface is composed by highly electronegative atoms and has been cationized. It is likely that hydrogen bond-type interactions are established between the liquid and surface, slowing the molecules¯owing near the interface and causing local microturbulences, i.e. misaligned molecular domains [29]. As MEG has strong intermolecular bonds, these turbulent domains may have a medium range within the liquid, decreasing the intermolecular alignment at the bulk. Further down the stream, this eå ect is overcome and the intermolecular alignment increases.
When SDS is added (® gure 6B), both polarization and anisotropy increase monotonicall y as the liquid¯ows. The interfacial tension is lowered by the surfactant. Thus, the molecules¯owing near the interface will have weaker interfacial interactions, becoming more aligned with the¯ow. It is even possible that a certain degree of slip velocity occurs at the interface. It is likely that the dispersion [30] will also decrease as the molecular domains become more laminar. When the SDS concentration is increased (® gure 6C), the polarization and anisotropy stay almost identical. This may be due to the interacting sites of the interface being already saturated by surfactant molecules, thus this concentration increase did not cause signi® cant eå ects.
When the solute is PEO (® gure 6D), the interfacial tension is also lowered, which should increase polarization and anisotropy. Similarly to SDS, the long-chain PEO may interact with the borosilicate at several points of the same molecule, generating misaligned molecular domains at the interface and causing the molecules owing near the interface to be more turbulent. Nevertheless, a wide and nearly homogenous region of low values replaces the polarization and anisotropy gradation seen with SDS. This may be explained by the PEO long chains within the bulk being in between the small MEG molecules, disturbing the intermolecular bonds network and misaligning the molecular domains. It is also well known that long-chain polymers like PEO tend to form molecular aggregates [28]. These structures may also decrease signi® cantly the intermolecular alignment of the MEG¯ow.

Conclusion
The automatic system described here for sample XY positioning and remote acquisition of polarization and anisotropy made it possible to acquire higher de® nition maps to reduce acquisition time and increase data reproducibility and resolution.
The intermolecular alignment of a thin sheet of liquid MEG¯owing on borosilicate, evaluated through polarization and anisotropy maps, decreased initially due to strong interfacial tension, thus increasing the misaligned molecular domains and consequent dispersion.
Addition of the surfactant SDS at two submicellar concentrations increased the intermolecular alignment progressively with the¯ow. This may be explained by the decrease interfacial tension, allowing the molecular domains to align themselves with the¯ow.
Addition of the long-chain polymer PEO decreased pronouncedly the intermolecular alignment. This may be due to formation of polymer-aggregate d structures within the bulk-generating misaligned molecular domains.
Flow analysis systems, with intermolecular alignmentdependent reactions, that make use of surfactants may have their values altered by the non-isotropic intermolecular alignment distribution of the molecular domains, and their subsequent dispersion eae ciency.