On-line and in situ optical detection of particles of organic molecules formed by rapid expansion of supercritical solutions (RESS) of CO2

The formation of particles by the rapid expansion of supercritical solutions (RESS) of four organic substances (n-undecane, naphthalene, trans-stilbene and benzoic acid) dissolved in supercritical CO2 has been investigated employing two complementary on-line in situ detection techniques.

For the first time, a laser-based shadowgraphy (LABS) setup was applied, which allows the recording of the diameter, morphology, size distribution and concentration of particles down to sizes of roughly 8 μm.

In addition, laser-based three-wavelength extinction measurements (3-WEM) have been employed to determine the average particle diameter, the width of the particle size distribution and the concentration.

The average particle sizes determined from 3-WEM for n-undecane, naphthalene, trans-stilbene and benzoic acid are 2660, 1260, 550 and 430 nm, respectively, and the standard deviation of the (assumed) logarithmic normal distributions 0.37, 0.42, 0.78 and 0.60.

The particle size distributions at large diameters from LABS compare favourably with the tails of the 3-WEM distributions.

An analysis of the morphology of trans-stilbene particles and n-undecane droplets reveals an, on average, slightly elliptical shape.


Micrometre- to nanometre-sized particles can be formed by expanding a supercritical solution containing a solute of interest through a micrometre-sized orifice.

Petersen et al. coined the term “rapid expansion of supercritical solutions (RESS)” for this process.1

The method has already been applied on a considerable scale,2 ranging from the treatment of pharmaceuticals like ibuprofen3 or cholesterol4 to the production of thin films5 and microfibers.6

Very often, a precise control of the particle size distribution (PSD) is needed, e.g., the efficient inhalation of drugs requires particle sizes in the range between 1 and 5 μm.7

Efficient ways of on-line and in situ monitoring PSDs already during the production process of micrometre- to nanometre-sized RESS particles are therefore desirable to minimize the consumption of raw material, production times and costs.

Two standard off-line techniques widely used for the characterization of RESS particles are scanning electron microscopy (SEM) and optical microscopy.8

In addition, cascade impactors9 and scanning mobility particle sizers (SMPS)10,11 have been employed to analyze aerosols from different sources.

The only on-line and in situ technique applied so far to the investigation of RESS particles is the three-wavelength extinction measurement (3-WEM).4,9

Optical detection is used to measure the particle-induced attenuation of three laser beams at different wavelengths.

If the complex index of refraction of the particles is known, the average diameter of the particles and the width of the particle size distribution (PSD) can be obtained together with the volume concentration (see Section 3).

So far the method has been applied to a handful of RESS-processed organic substances and pharmaceuticals.4,12

In this contribution, a laser-based shadowgraphy (LABS) technique is used for the first time to study the formation of organic RESS particles.

This method is able to determine the size distribution and morphology of particles with sizes ≥8 μm on-line and in situ.

Both solid particles and droplets can be investigated by recording their magnified shadows by means of a farfield microscope/CCD camera combination (see Section 3).

Results for four different organic substances (n-undecane, naphthalene, trans-stilbene and benzoic acid) will be compared with those from our complementing measurements employing the established 3-WEM technique.


An overview of the experiment is shown in Fig. 1.

Particles are produced using a standard RESS flow setup.

Liquified CO2 at room temperature is pressurized (120–180 bar) using two Isco syringe pumps (Model 100 DX) in a continuous flow mode, capable of delivering flow rates up to 60 mL/min, and passes through a preheater (temperature 50–60 °C).

The supercritical CO2 then flows through the heated and stirred extraction cell, where it is saturated with the solute of interest.

This setup can be used at pressures up to 1000 bar and temperatures up to 200 °C.

All substances are commercially available: n-undecane was obtained from Merck, naphthalene and benzoic acid from Sigma Aldrich, and trans-stilbene from Acros.

The saturated mixture is passed through a heated filter [15 μm (Swagelok) or 10 μm (Nova Swiss)], which prevents clogging of the nozzle with undissolved particles.

The supercritical solution is then expanded through a home-built heated nozzle, employing Teflon seals and a laser-drilled sapphire orifice (Bird Precision, diameter 51 μm, length 254 μm, with tapered inlet), into the acrylic-glass detection chamber, which is equipped with Suprasil I quartz windows.

These allow optical access to the expansion at different distances from the nozzle.

A bypass line can be used for flushing the system with supercritical CO2 or producing solute concentrations lower than the saturation concentration.

On-line monitoring was performed by LABS and 3-WEM detection about 20 mm away from the nozzle.

In the LABS setup, pulses from a frequency-doubled Nd:YAG laser (New Wave Research Solo PIV III-15, λ = 532 nm, 50 mJ pulse−1, pulse length 3–5 ns) were expanded by a diffusor, consisting of a lens and a wavelength shifting fluorescence plate (LaVision, diameter 50 mm, fluorescence decay time constant about 10 ns), which was large enough to illuminate the volume of interest.

Optical detection of the shadowgraphs was performed by the combination of a mirror-based farfield microscope (Questar QM1), a suitable magnification lens and a CCD camera (LaVision, Flowmaster 3S, 1280 × 1024 pixel, frame rate up to 8 Hz).

At a working distance of around 60 cm, the field of view below the nozzle was about 1100 μm × 950 μm, with a depth of field around 500 μm.

Typically, up to 1000 images per single measurement were analyzed using the shadowgraphy module of the DaVis 6.2 software package (LaVision) and afterwards combined to larger data sets.

The analysis yields the diameter and morphology of each detected particle as well as the particle size distribution (PSD) and concentration.

The lower detection limit with the current setup is approximately 8 μm.

3-WEM was performed using the Wizard DQL system (Wizard Zahoransky KG) consisting of a triple diode laser head (λ1 = 637 nm, λ2 = 811 nm and λ3 = 1316 nm) and a photodiode detector with an achromatic focusing lens.

Prior to each measurement, reference intensities I0 without particles in the detection chamber were recorded at the three wavelengths.

During the expansion, the intensity I for each of the wavelengths was measured, which allowed two extinction ratios to be determined.

These are compared with a data field calculated via Mie theory,13 from which the average particle diameter, the width of the PSD, and finally the volume concentration of the particles can be determined (see below).

Results and discussion

Three-wavelength extinction measurement (3-WEM)

The relation between the initial intensity I0 and the attenuated intensity I in a 3-WEM experiment is given by the Lambert–Beer type expression:9where N is the particle density, L the path length in the detection chamber, and ε the extinction coefficient.

The latter is dependent on the particle diameter D, the wavelength λ and the complex index of refraction m = n + ik, where n and k are characteristic constants of the material.

In this work, I0 and I are determined at three different laser wavelengths λ1, λ2 and λ3.

This allows the two extinction ratios ER1,2 and ER2,3 to be determined, and the dependence on N and L can be removed:Several studies suggest that the PSDs for most aerosols can be approximated by a logarithmic normal distribution:14,15 where D is the diameter of a particle, Dav is the average diameter of the particles, and σ the standard deviation of the distribution.

The values for the parameters n and k in the complex index of refraction m can be found in standard tables.16,17

The two remaining unknown variables Dav and σ can then be obtained via eqns. (2) and (3).

Finally, the volume concentration N can be determined via eqn. (1), because the path length L of the detection chamber is known.

Fig. 2 shows a typical example of a 3-WEM signal for trans-stilbene RESS nanoparticles.

40 s after initiating the 3-WEM, the nozzle was turned on, which marks the starting point of the particle production.

The different extinctions at the three wavelengths are clearly visible.

At any time, the two extinction ratios can be calculated, and Dav, σ and N determined.

Fig. 3 and 4 show the results of such an analysis for naphthalene and benzoic acid, respectively.

In the case of naphthalene the nozzle was turned on after 140 s.

Average particle sizes of 1260 nm were obtained with a standard deviation 0.42.

Particle volume concentrations of 4.4 × 10−6 m3 m−3 were reached.

The nozzle was then turned off after 280 s.

Due to the sharp drop of the particle concentration, no 3-WEM signal could be recorded.

After 320 s the nozzle was turned on again, and essentially the same values were determined as before.

This nicely demonstrates how the particle production process can be easily followed in real time using the 3-WEM technique.

In the case of benzoic acid (Fig. 4) considerably smaller particles are generated, with Dav = 430 nm and a standard deviation σ ≈ 0.60.

Volume concentrations are in the order of 4.3 × 10−6 m3/m3.

A summary of the 3-WEM results for all organic substances studied can be found in Table 1.

Note that our results for benzoic acid and naphthalene compare very well with the earlier work of Schaber and co-workers,4,18 though it has to be kept in mind that the experimental conditions in both cases are slightly different.

Laser-based shadowgraphy (LABS)

As an example, Fig. 5 shows four selected shadowgraphs taken from a typical run for n-undecane recorded with the LABS system.

In each of these very short (ca. 5 ns) snapshots, particles of different size and morphology are visible.

An automated image analysis was applied to process the shadowgraphs.

Fig. 6 shows the eccentricity of n-undecane droplets [defined as e = (1 − b2/a2)0.5, where b and a are the short and the long axes of an ellipse, respectively], employing a sum total of 4000 images.

The sizes are relatively broadly scattered, and one obtains an average of e ≈ 0.6, equivalent to a b/a axis ratio of roughly 0.8.

This result is somewhat surprising, because for these small droplets one would expect the surface tension to be the key factor controlling the shape.

This should produce spherical droplets.

However, the droplets possibly emerge from somewhat “violent” generation processes, which might induce strong oscillations of the droplet structures.

Finally, note that the eccentricity is relatively insensitive to the diameter of the droplets, although a final judgment is difficult, as the droplet statistics deteriorate towards larger diameters.

Fig. 7 shows the corresponding plot for trans-stilbene.

Note that in this case there are much less particles (about 0.26 per image) in contrast to about 3.9 particles per image in the n-undecane case.

This is consistent with the larger average diameter of the n-undecane droplets (see Fig. 8 and 9), so more particles are above the lower detection threshold of the LABS setup.

For trans-stilbene we find an average eccentricity e ≈ 0.5 (equivalent to an axis ratio b/a ≈ 0.87), very similar to the n-undecane case.

In this case the observed shape could be due to the fact that particle growth inside the tubular nozzle orifice might prefer the formation of slightly elongated particles.

It is also very instructive to compare the PSDs from LABS and 3-WEM.

In Fig. 8 and 9 this is done for trans-stilbene particles and n-undecane droplets, respectively.

Each inset shows an overview of the complete 3-WEM distribution, whereas the main part contains a blow-up of the large diameter wing of the PSD to show the overlap with the PSD from LABS.

From both types of experiments an absolute particle concentration can be extracted, and thus the relative amplitudes of the 3-WEM and LABS PSDs are uniquely determined.

The latter only accurately reflect the particle distribution above ∼ 8 μm.

Due to the detection limit of the current LABS setup, the PSD histograms below 8 μm certainly represent a lower limit to the real value.

Therefore average particle diameters Dav cannot be deduced from the LABS measurements.

Compared to the 3-WEM results, the PSDs from LABS reach out to higher diameters.

However, one has to keep in mind that the PSDs from 3-WEM for trans-stilbene and n-undecane peak at 550 nm and 2660 nm, respectively.

At their far end, this will at best only give a semi-quantitative description of the decay, especially if one keeps in mind that the width of the 3-WEM PSDs were deduced assuming a spherical particle shape in the analysis.

Considering the slight deviations from spherical shape observed with LABS detection the error for the particle size determination using 3-WEM is however expected to be small.

It is therefore safe to say that in general the LABS and 3-WEM results show an overlap.

The 15-fold higher absolute number of particles for n-undecane compared with trans-stilbene, as counted by LABS, is in complete agreement with the larger average particle size in this system.


We have demonstrated that the combination of laser-based shadowgraphy (LABS) and three-wavelength extinction measurement (3-WEM) provides non-invasive on-line and in situ optical detection for studying RESS particle and droplet size distributions as well as morphologies up to large particle diameters.

In addition, accurate particle concentrations can be deduced from such measurements.

Average particle sizes Dav (including standard deviations σ of the size distributions) from 3-WEM for n-undecane, naphthalene, trans-stilbene and benzoic acid were: 2660 nm (0.37), 1260 nm (0.42), 550 nm (0.78) and 430 nm (0.60), respectively.

The detailed knowledge of the size distributions at larger diameters provided by the LABS method might give some useful input for understanding the RESS process, because state-of-the-art theories for describing particle growth in such expansions still predict much smaller diameters than observed in experiment.8,19

An overlap of the particle size distributions at large diameters from LABS and 3-WEM was found.

In addition, information on the morphology was obtained for trans-stilbene particles and n-undecane droplets, which both showed a slightly elliptical shape.

Measurements of the current type are very promising to study the influence of temperature, pressure, nozzle distance and nozzle diameter on the particle formation process in a systematic way, without the need of time-consuming analysis processes.

In addition, extended comparisons with off-line particle sizing techniques like scanning electron microscopy (SEM) or scanning mobility particle sizing (SMPS) would be interesting.

The SEM images will be particularly valuable to confirm the elliptical particle shape observed in this study.

Research along these lines is currently pursued in our laboratories.