Environ. Sci. Technol. 2006, 40, 6738-6743
Size Fractionation and Characterisation of Fresh Water Colloids and Particles: Split-Flow Thin-Cell and Electron Microscopy Analyses ANNA DE MOMI AND JAMIE. R. LEAD* School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, B15 2TT U.K.
Split-flow thin-cell (SPLITT) was employed in conventional mode (CSF), to size-fractionate colloids and particles from a selected freshwater. Imaging and quantification by calculations of particle size distributions (PSDs) and shape factors were performed on sample analyzed by conventional high vacuum scanning electron microscopy (SEM) and environmental SEM (ESEM), to investigate the ability of SPLITT to make accurate and nonperturbing separations. SEM and ESEM images of unperturbed and SPLITT-generated fractions were used in order to obtain qualitative and quantitative information about the properties of colloids and particles. Particle size distributions (PSDs) showed that separations were very good, agreeing with theoretical behavior. ESEM PSDs showed that up to 87-88% of the material in the a fraction (expected to be 1 µm) 87-95% of the material was the expected size. The SEM data indicated a slightly higher contamination of the b fraction with the presence of submicron colloids. Moreover, analysis of conformations indicated significant nonsphericity in unfractionated colloids and particles, but after SPLITT fractionation, shape factors showed that particles were significantly more spherical than before separation.
Introduction Organic and inorganic materials, with dimensions between 1 nm and 1 µm, are defined as colloids. Their behavior is dominated by aggregation and disaggregation phenomena because their small size does not allow them to settle under the influence of gravity. Particles are material >1 µm, and their transport is dominated by settling under gravitation (1-6). Due to their properties and ubiquity in natural waters, colloids and particles are essential to understand problems related to the speciation and bioavailability of trace elements and to regulate processes affecting their transport and biouptake in the environment. As these colloidal and particulate phases are unstable, complex, and present at low concentrations, sensitive and nonperturbing analytical techniques are required to separate and characterize them. Despite the availability of a small number of fractionation techniques such as filtration, fieldflow fractionation (FFF), and centrifugation, none can be * Corresponding author phone: +44 121 414 8147; fax: +44 121 414 5528; e-mail:
[email protected]. 6738
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
used with complete confidence and they are not fully verified for the separation of aquatic colloids and particles (1, 7, 8). Filtration, in frontal or crossflow mode, potentially alters colloid and particle structure through membrane interactions (6, 9), while FFF requires preliminary concentration, potentially resulting in aggregation, and centrifugation can also result in increased aggregation. As an alternative method of fractionation, split-flow thin flow (SPLITT) has been used previously (10), and this method is expected to be minimally perturbing. The SPLITT fractionation (SF) is a separation system that is possibly suitable for rapid processing of environmental and other solid-phase material with the potential to provide a size fractionation in the range of approximately 50 µm down to about 1 µm, dependent on density, flow rate, and other experimental or particle properties. In any single experiment, separation at a single cutoff diameter can be performed. This technique is able to divide the sample in two portions containing particles above and below a specific cutoff diameter in an analogous manner to filtration, but without a membrane. This separation is achieved using well-controlled hydrodynamic flow rates in the thin channel and an external applied (gravitational) field (11, 12). Only a few publications have been reported in the literature using natural aquatic samples. SPLITT has been used to separate diatoms (13), sediments (14, 15), and natural river particles (10, 12), although only partial validations of the method has been performed. More recently, we have fully investigated the ability of SPLITT to perform accurate size fractionations (16), but only in the seldom used ‘full feed depletion mode’ (10). In the present paper, the more commonly used conventional mode of SPLITT operation has been fully investigated and validated for further use.
Materials and Methods Sampling and Size Fractionation. Samples were collected in 1-L polyethylene bottles from Vale Lake, situated on the grounds of the University of Birmingham, West Midlands, UK, between July 29, 2004, and February 4, 2005. Sampling was performed at about 50 cm from the lake bank and a few millimeters below the lake surface level, taking care not to disturb sediments. All the bottles used to stock the natural sample were cleaned with 1% of nitric acid, rinsed with pure water (R ) 18.2 MΩ cm) and with sample prior of sampling. All the washings were discarded and, in order to obtain results that are representative of the natural system, sampling operations were carried out using a standard method (17). SPLITT Cell Apparatus. The SPLITT system is able to perform hydrodynamic separations of colloids and particles in two ways: the so-called conventional mode and full feed depletion mode (Figure 1). Previously, we have validated the full feed depletion mode (16). In the conventional mode (Figure 1) the sample (suspended particles) is continuously introduced into the cell through the top inlet (at a predetermined volumetric flow rate V˙ (a′)). At the same time pure (particle-free) carrier solution is pumped into the cell through the bottom inlet (at a flow rate V˙ (b′)). Where the two inlet streams join to form a single stream a hypothetical “inlet splitting plane” (ISP) is formed. Analogously the “outlet splitting plane” (OSP) separates the two fluid elements at the end of the channel. The position of the ISP and OSP, and consequently the theoretical size of fractionation, is controlled by changing the ratio of flow rates in two inlet V˙ (a′) and V˙ (b′)) and in two outlet (V˙ (a) and V˙ (b)) substreams, respectively. Initially, the particles in the sample feed stream are compressed in a thin lamina between the top wall and the inlet splitting plane, and they migrate toward the bottom 10.1021/es061181t CCC: $33.50
2006 American Chemical Society Published on Web 10/03/2006
FIGURE 1. Side view of a split flow thin cell fractionation (SPLITT) cell operating in conventional mode. wall, along the transverse direction and interacting with the external field, at different velocities depending on their buoyant mass. When the fluid stream reaches the end of the channel, it is mechanically divided into two fractions by an outlet splitter. Slower settling particles emerge from the upper outlet V˙ (a) flow rate), while those that settle faster exit the lower outlet (at V˙ (b) flow rate) (10, 11). Although the theory for standard spheres has been explored, there are uninvestigated difficulties in data interpretation when applying SPLITT to complex and polydisperse natural colloids and particles. In this work, a cutoff diameter of 1 µm was chosen as a borderline between environmental colloids and particles. Knowing that their density can vary from ca. 1 to 5 g mL-1, depending on the component being considered, we used a hypothetical density value of 2.5 mg L-1 as an average of these components and as representative of mineral phases such as silica, which is the major component of many waters. The theory of SPLITT fractionation has been described in the literature (18); therefore, we show only some of the most important equations are shown here. The condition for the best position of the splitting planes is
V˙ (b′) V˙ (b) > V˙ (a′) V˙ (a)
(1)
where V˙ (a) and V˙ (b) are the two volumetric flow rates. The volumetric flow rate passing along the transport region V˙ (t) is related to the substream flows given by
V˙ (t) ) V˙ (a) - V˙ (a′) ) V˙ (b′) - V˙ (b)
(2)
The volumetric flow rate may be also expressed by
∆V˙ ) bLU
(3)
where b is the cell breadth, L its length, and U is the sedimentation velocity. Equation 3 can be rewritten by assuming a spherical model for particles as
∆V˙ )
bLG(Fp - F)d2 18η
(4)
where G is the gravity acceleration, Fp is the density of the particles, F is the carrier density, d is the diameter of the particles, and η is the viscosity of the carrier. The condition for which a particle exits from the outlet b is given by
∆V˙ > ∆V˙ (t)
(5)
TABLE 1. Experimental Parameters Used for SPLITT Cell Fractionation To Separate Lake Water stream flow rate (mL/min-1) dimensional
cutoffa
(µm)
exp 2: conventional SF, 1 µm
a′
b′
a
b
0.4
1.1
1.0
0.5
a
The cutoff is defined as the largest spherical particle that would exit through the outlet a.10
Combining eqs 3 and 5 enables the calculation of the sedimentation coefficient cutoff of particles fractionated by SPLITT cell, given by
Sc )
(V˙ (a′) - V˙ (b)) bLG
(6)
No particles with sedimentation coefficient (S ) U/G) greater than Sc should exit outlet a. However, the fraction a from the outlet b will contain some particles with S < Sc although the SPLITT fractionation process results in the removal of increasing proportions of these particles as (Sc - S) increases. In practice, an approximate size cutoff may be set for the SPLITT fractionations and we have so set it here as 1 µm. The SPLITT model SF1000HC (Postnova) used in this study has dimensions of 20 cm length, 4 cm width, 964 µm thickness. Two peristaltic pumps (Ecoline VC-MS/CA, Ismatec) were used allowing the presence of two independent flow streams at the a′ and b′ inlets for sample and carrier solution introduction, respectively, where the carrier solution was Milli-Q deionized water. The different flow rates used for these experiments are given in Table 1. Electron Microscopy Images. SEM was used in order to acquire information about morphologies and size distribution of colloidal and particulate fractions after air- and vacuumdrying. Droplets of SPLITT-produced samples were placed onto a clean electron microscopy (EM) support stub, allowed to oven dry (40 °C), and coated with a very thin layer of gold or platinum by a sputter coater (Emscope SC 500). The SEM used in this study was a JEOL JSM-6060 LV and the SEM acceleration voltage was 15 kV. Particle size distributions (PSD) were also calculated by measuring the lateral dimension of roughly 400 individual particles. ESEM. A Philips XL30 ESEM-FEG environmental scanning electron microscope was used in wet mode in order to observe morphology and size distribution of colloidal and particulate in their hydrated state, as we have done previously (19, 20). One drop of the SPLITT-generated samples was placed VOL. 40, NO. 21, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6739
FIGURE 2. Selected environmental scanning electron microscopy images of unperturbed and size-fractionated lake water: (a) lake water, (b) 1 µm fraction. Scanning electron microscopy images of unperturbed and size-fractionated lake water: (d) lake water, (e) 1 µm fraction. The cell operated in conventional mode. directly onto a clean glass surface, which was fixed to the ESEM stainless steel support stubs using graphite paint for reducing charge effects around the sample. No staining was required to image the samples. These were in turn positioned on the water-cooled Peltier stage. Four drops of ultrapure water were placed on the cooling stage around the sample to control the sample humidity and to minimize or inhibit any dehydration that may occur during evacuation of the air from the sample chamber (21, 22). Microscopy observations were performed at a constant temperature of 2 °C and a pressure of 5.4 Torr, using a low acceleration voltage of 10 kV (to reduce beam damage) and at 100% relative humidity. These conditions allowed the samples to be fully hydrated but with little bulk water. Particle size distributions were 6740
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
also determined by measuring the lateral dimension of around 400 single particles. Shape factors (SF) were calculated from the following formula SF ) 4π(area/perimeter2); 0 e SF e 1), where a value of 1 was a sphere and values lower than 1 were increasingly less spherical and regular. SF values of imaged particles were calculated using a Gatan Inc. Digital Micrograph program version 3.4.4.
Discussion Electron Microscopic Images of Lake Water and SPLITTGenerated Fractions. The two microscopic techniques chosen in this study have been used in conjunction to exploit
FIGURE 3. Particle size distributions derived from environmental scanning electron microscopy of unperturbed and size-fractionated lake water: (a) lake water, (b) 1 µm fraction. Scanning electron microscopy data using split flow thin cell fractionation in conventional mode: (d) lake water, (e) 1 µm fraction. advantages in their mutual use (22), in line with the commonly agreed necessity of using multiple techniques to ensure data validity when analyzing unstable colloids and particles (16). In particular, the ESEM is a suitable technique for analysis and quantification of organic and inorganic natural colloids in a hydrated state, without the need to coat nonconducting samples or to stain for organic matter. However, ESEM is limited by the practical lateral resolution of the microscope (ca. 30 nm), However, the SEM has a high lateral resolution (ca. 2 nm), giving clear, well-resolved images, but samples must be coated if imaging of nonconductive material is performed. In addition, ultrahigh vacuum (UHV) conditions lead to drying which causes extensive change in the colloid/ particle structure. Images from SEM and ESEM are shown in Figure 2, reflecting these different operating conditions of the electron microscopies. These images were chosen at random and are not fully representative of the PSD or shape factor data presented (which quantified several hundred or thousand particles, respectively) but merely give an indication of the types of particles observed. Further images are shown in the Supporting Information (Figures S1 and S2). Qualitatively, the ESEM images as a whole show more complex morphology than SEM images, indicative of less harsh and perturbing conditions under which analysis was performed. The SEM images show a preponderance of discrete, spherical particles quite different from ESEM images, but with greater
image quality. On the basis of the ESEM images, it is clear, at least qualitatively, that the a fraction (containing material nominally 1 µm) does contain larger aggregates but also a measurable amount of smaller material at about 1 µm. No such distinction can be observed in the SEM images (Figure 2e,f), presumably because of the greater sample perturbation caused by UHV conditions. Accuracy of SPLITT Separations. Particle size distributions (PSDs), on a particle number basis, for the unfractionated lake water and for the a and b fractions (less than and more than 1 µm, respectively) are shown in Figure 3 (conventional SPLITT mode, from ESEM and SEM). Excellent agreement between the observed and expected cutoff values was shown, and 88.3% of particles in the a fraction was less than the 1 µm fraction and 87.5% in the b fraction was greater than the 1 µm fraction observed by ESEM. In addition, 88.1% in the a fraction was less than the 1 µm fraction and 71.8% in the b fraction was greater than the 1 µm fraction observed by SEM, both using conventional-mode SPLITT. These details are summarized in Table 2 and confirm the qualitative observations discussed above. Similar observations were made for different SPLITT modes (16). Interestingly, the percentage of particles in the correct SPLITT fraction as VOL. 40, NO. 21, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6741
TABLE 2. Comparison of Relative Amounts (%) of Colloidal Material SPLITT-Generated, in Full Feed Depletion and Conventional Modes, Determined by SEM and ESEM
lake
fraction a expected dc e 1 µm
fraction b expected dc g 1 µm
experimental cutoff (µm)
SEM (%)
ESEM (%)
SEM (%)
ESEM (%)
SEM (%)
ESEM (%)
exp 2: conventional SF dc e 1 µm dc g 1 µm
62.5 37.5
69.8 38.2
88.1 11.9
88.3 11.8
28.2 71.8
12.5 87.5
quantified by SEM and ESEM agree well for the small colloidal fraction (Table 2). However, ESEM shows that substantially greater numbers of larger particles are present in the b fraction, compared with the SEM. The results suggest that some changes in the particles occur during drying in the UHV SEM, possibly due to dehydration and shrinkage. Previous data from the literature has given only a partial, qualitative validation of SPLITT by selected images from SEM (10) or semiquantitatively by PSDs of the SPLITT derived fractions, taken from SEM images (14, 15). Recently, we showed that the full feed depletion mode performed relatively accurate size fractionations (16), albeit with difficulties associated with contamination of the >1 µm fraction by smaller sized material. Data in this paper has fully and quantitatively shown the ability of conventional-mode SPLITT to make very effective size fractionations between colloids and particles by quantifying the initial unperturbed water samples and by using a nonperturbing analysis method (ESEM). We conclude that, in terms of size fractionation and mass balance data, SPLITT, especially in the conventional mode, is an excellent separative method for distinguishing between colloidal and particulate material. Perturbation of Colloids and Particles during Size Fractionation. Clearly, SPLITT performs a very good size fractionation, confirming our previous work in other modes (16) and previous, less quantitative work by others (10, 14, 15). Nevertheless, when we compared shape factors of particles, which gives an indication of symmetry, especially sphericity, of particles, values were significantly higher in the a and b fractions compared with the original lake water. The shape factor of particles in the lake fraction was 0.37 ( 0.21 (mean ( standard deviation, n ) 1655), of fraction a was 0.40 ( 0.20 (n ) 1840), and of fraction b was 0.40 ( 0.20 (n ) 1711). The differences are small, smaller than those observed in full feed depletion mode (16), but they are significant (p < 0.01). Comparison of particles in the a and b fractions showed that the shape factors were not significantly different from each other. Clearly, we are seeing some small but real changes in the conformation of aquatic colloids and particles due to the SPLITT fractionation. We do not as yet know the mechanism by which change occurs, but it may be on-channel aggregation or mechanical damage due to fractionation. Changes in the structure of individual particles due to solution changes on channel may be less likely, as the eluent was matched to the sample and the full feed depletion mode used previously (16), which does not use an eluent, causes greater change in colloids and particles. SPLITT as a method of size fractionation was chosen as it is likely to be minimally perturbing to aquatic samples compared with filtration (frontal and/or crossflow), FFF, or other methods, since with SPLITT no membrane or preconcentration is used. The time spent on-channel is on the order of 1 min or less. In addition, flow in the channel is laminar, so aggregation exacerbated by turbulent mixing is minimal. Nevertheless, despite these potential advantages over other 6742
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
systems, quantitative evidence of changes in colloid and particle structure and morphology under these minimally perturbing conditions has been provided. Given that the change is in the same direction for both fractions, i.e., toward more spherical material, this cannot simply be an effect of separating the fractions. The changes must be either due to the preferential loss of nonspherical particles or due to changes in particle morphology brought on by the fractionation. While it may be possible, there is little reason to assume that the most irregular particles would be preferentially lost to channel walls. Most likely SPLITT separation alters particle morphology slightly but significantly. This reasoning is qualitatively backed up by changes observed between SEM and ESEM; the more perturbing SEM analysis produces more spherical particles than the ESEM. Nevertheless, we expect SPLITT to be among the least perturbing of size fractionation methods, but changes are still induced. That changes in colloidal and particulate material can be produced, even under minimally perturbing conditions, is supported by a previous atomic force microscopy (AFM) study indicating that a 2 h sedimentation of water under ambient conditions (again a minimally invasive procedure) resulted in the loss of a significant fraction of
