Rational strategy for characterization of nanoscale particles by asymmetric-flow field flow fractionation: a tutorial.

Analytica Chimica Acta 809 (2014) 9–24

Contents lists available at ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Review

Rational strategy for characterization of nanoscale particles by asymmetric-flow field flow fractionation: A tutorial Julien Gigault ∗ , John M. Pettibone, Charlène Schmitt, Vincent A. Hackley National Institute of Standards of Technology, Materials Measurement Science Division, 100 Bureau Drive, Gaithersburg, MD 20899-8520, USA

h i g h l i g h t s

g r a p h i c a l

a b s t r a c t

• Underlying • • • •

theory and critical parameters are introduced. A rational workflow is proposed to optimize and refine A4F methods. Specific optimization steps and validation parameters are delineated. Pedagogical examples are provided to demonstrate the process. Use and relevance of different detection modalities is addressed.

a r t i c l e

i n f o

Article history: Received 21 August 2013 Received in revised form 31 October 2013 Accepted 8 November 2013 Available online 16 November 2013 Keywords: Field flow fractionation Nanoparticle Nanomaterial Protocol Optimization Validation

a b s t r a c t This tutorial proposes a comprehensive and rational measurement strategy that provides specific guidance for the application of asymmetric-flow field flow fractionation (A4F) to the size-dependent separation and characterization of nanoscale particles (NPs) dispersed in aqueous media. A range of fractionation conditions are considered, and challenging applications, including industrially relevant materials (e.g., metal NPs, asymmetric NPs), are utilized in order to validate and illustrate this approach. We demonstrate that optimization is material dependent and that polystyrene NPs, widely used as a reference standard for retention calibration in A4F, in fact represent a class of materials with unique selectivity, recovery and optimal conditions for fractionation; thus use of these standards to calibrate retention for other materials must be validated a posteriori. We discuss the use and relevance of different detection modalities that can potentially yield multi-dimensional and complementary information on NP systems. We illustrate the fractionation of atomically precise nanoclusters, which are the lower limit of the nanoscale regime. Conversely, we address the upper size limit for normal mode elution in A4F. The protocol for A4F fractionation, including the methods described in the present work is proposed as a standardized strategy to realize interlaboratory comparability and to facilitate the selection and validation of material-specific measurement parameters and conditions. It is intended for both novice and advanced users of this measurement technology. Published by Elsevier B.V.

Contents 1. 2.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A4F theory and principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. General theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Criteria for A4F fractionation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

∗ Corresponding author. Tel.: +1 301 975 5790/+1 301 975 5783; fax: +1 301 975 5334. E-mail addresses: [email protected] (J. Gigault), [email protected] (V.A. Hackley). 0003-2670/$ – see front matter. Published by Elsevier B.V. http://dx.doi.org/10.1016/j.aca.2013.11.021

10 11 11 12

10

3. 4.

5.

6.

J. Gigault et al. / Analytica Chimica Acta 809 (2014) 9–24

Protocol development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Protocol evaluation and application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. General precautions and starting conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1. Calibration: Influence of NP core material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2. Quantity injected . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.3. Mobile phase composition influence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Stepwise application of protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Size distribution and retention measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2. Optimization for component separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. General starting conditions for preliminary A4F analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Protocol performance range: Upper and lower size limits for characterization using A4F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. At the bottom: Nanoclusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Upper size limit: Sub-micrometer populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Charlène Schmitt obtained her M.S. in physics and chemistry for polymers and materials at the University of Pau, France in 2012. Her graduate research centered on the synthesis of polystyrene latex using sustainable saponin surfactants. During 2012, she completed a graduate research sabbatical at the U.S. National Institute of Standards and Technology, where she developed field flow fractionation methods to characterize polymer-based nanoparticles. Currently, she is a Ph.D. student in analytical and polymer chemistry at the University of Pau, where her research focuses on new polymers.

John M. Pettibone, received his Ph.D. in chemical engineering from the University of Iowa. He received a National Research Council Fellowship to conduct research at the U.S. National Institute of Standards and Technology in 2009 and is now a staff scientist. His research focuses on detection, characterization and quantification of nanoscale particles over a wide continuum of states and sizes, and exploration of new measurement approaches to examine complex and dynamic formation and transformational reactions that are critical for predicting the impact and fate of nanoscale materials.

1. Introduction The accurate characterization of nano-object populations (with at least one dimension in the 1 nm to 100 nm range), and, more generally, particles in the submicrometer range (up to 1 ␮m), is an important and relevant measurement challenge spanning a broad range of applications, including industrial, consumer products, environmental remediation, risk assessment, and biomedical research, among others [1–7]. The rapid growth of interest in nanoobject characterization has been driven by the emergence and corresponding commercial penetration of nanotechnology over the past decade. Surface chemistry or functionality in combination with the physical form of nano-objects (i.e., their dimensions, shape and aggregation state) are the primary determinants of

13 16 16 16 16 17 17 19 20 20 20 20 21 22 22 22 22

Julien C. Gigault graduated in analytical and environmental chemistry at the University of Brest, France in 2008, and obtained his Ph.D. from the University of Pau, France in 2011. His Ph.D. focused on field flow fractionation (FFF) for the characterization of singleand multi-walled carbon nanotubes and the study of their environmental behavior. Since 2011 he has been a postdoctoral Associate at the U.S. National Institute of Standards and Technology, where his research seeks to develop and validate novel measurement methods based on FFF coupled to multiple detectors, in order to qualitatively and quantitatively determine the physico-chemical properties of nanoscale particles for environmental risk assessment.

Vince Hackley obtained a Ph.D. in water chemistry (now environmental chemistry and technology) from the University of Wisconsin in Madison, and has held associate and staff scientist positions at the U.S. National Institute of Standards and Technology since 1991. He leads a project team engaged in applied physical and chemical metrology for the analysis of engineered nanoscale materials and their interactions with biological and environmental systems. He is involved in international standards development for nanotechnology and particle characterization through committees in ISO and ASTM, and the development of particle reference materials at NIST. He received the U.S. Department of Commerce Silver Medal for scientific achievement in 2008, and is the author or coauthor of 95+ publications, including two edited books on particle technology.

their behavior in biological and environmental systems, impacting their transformations, transport, fate, biodistribution, metabolism, clearance and toxicity, among other properties [8–13]. Similarly, commercially exploitable optical, magnetic and photochemical properties of nano-objects are directly impacted, if not entirely determined, by their physical form [1,14,15]. Though primarily defined by their dimensions, nano-objects actually represent many different classes of materials characterized by a broad range of physical and chemical properties, including materials of natural, manufactured and anthropogenic sources [16]. Thus, the characterization of these different nano-object classes according to their specific nature, complexity, and source, represents a relatively new and important scientific challenge. International Organization for Standardization (ISO) defines nano-objects as belonging to one of


three principal subclasses based on their dimensional characteristics, to wit: (i) quasi 1-dimensional structures such as nanotubes, nanorods and nanowires, (ii) quasi 2-dimensional structures such as nanoplates, and (iii) materials with all 3 dimensions in the nanoscale range commonly called nanoparticles (NPs) [17,18]. In the present work, a more commonly utilized definition for NPs is applied, in which all discrete nano-objects as defined by ISO are encompasssed. Among the many different particle measurement tools currently available, either commercially or as prototype designs, there is a rapid growing interest in the analytical approach frequently referred to as ‘hyphenation’. In the hyphenated approach, upstream particle fractionation is coupled to one or more downstream on-line detectors [19–21]. Stated succinctly, hyphenation enables size-resolved physico-chemical characterization of complex NP populations under in situ conditions that replicate the native dispersed state with minimal perturbation [21,22]. A variety of upstream fractionation methods have been explored for such applications, including size exclusion chromatography [23–26], electrospray–differential mobility analysis [27–32], various forms of field flow fractionation [21,33–36] and capillary electrophoresis [37–42]. Among commercially available methods, flow field-flow fractionation (flFFF) is perhaps the most broadly adaptable fractionation method for nanoscale-to-microscale particles. Though principally developed for the analysis of polymers [43–46], the dynamic range combined with the advent of commercially available equipment has greatly increased the prominence of flFFF over the past decade, extending its application to NPs and colloidal particulates. The principal benefit of flFFF is its capacity to provide reliable size information and fractionation in complex populations with minimal interaction between the analyte particles and the separation channel [47], by relying on hydrodynamic forces to achieve diffusion-based separation (size and shape). The principal drawback of flFFF is the lack of established protocols and standards for its application, and therefore the persistent burden of repeated sample-specific method development. There are two forms of flFFF, known as symmetric and asymmetric, that differ in the way that the channel is designed and cross-flow applied. For most purposes, the principles and procedures discussed in this work are equally applicable to both forms. However, we have used asymmetric flow field-flow fractionation (A4F) as the test bed, and it will be stressed in the following tutorial. The aim of the present work, and the overall goal of this strategy, is to propose a standardized A4F protocol for optimizing the fractionation of NPs. In order to achieve this goal, different criteria for fractionation efficiency are utilized and procedures are proposed in order to conduct the most relevant fractionation method possible for a particular combination of conditions and analytes. Finally this optimization protocol is pedagogically applied to fractionate representative NPs according to their: (1) Material—the NP core composition (metal, metal oxide, etc.) can have a non-negligible influence on fractionation, (2) Size and shape, (3) Surface charge. To develop and demonstrate these procedures, NPs were chosen based on their relevancy and importance within the nanotechnology research community, and based on the different challenges they present for analysis. It should be emphasized that these particular analytes are a subset of a larger range of materials, including silicon dioxide, silver, and selenium, to which the protocol has been applied and evaluated during our investigations.

11

2. A4F theory and principle 2.1. General theory For FFF techniques, the principle of separation consists of injecting the analytes into a flat channel with a laminar tangential flow (having a parabolic velocity profile), and then to apply a transverse field (that is a cross-flow in the case of flFFF) perpendicular to the laminar flow in order to retain (and thereby fractionate) particles according to different physico-chemical properties (for A4F, separation is based on relative diffusion coefficients of the particles, which is a function of particle size) as illustrated in Fig. 1. Here the term “retain” means that particles are situated in a zone retained by its confinement to flow lines having velocities that are lower than the average velocity present in the FFF channel. There are different elution modes possible in FFF, but for particles in the nanoscale size range the principal elution mode is referred to as normal (or Brownian). In normal mode elution, Brownian dynamics dominate such that particle concentration diffusion balances the flux induced by the perpendicular field (a cross-flow in the case of flFFF) that would otherwise drive the analytes toward the accumulation wall. In this case, translational diffusion is the key parameter that controls the concentration profile of the analyte. Thus, in normal mode elution, the concentration c(x) of particles at distance x from the accumulation wall of the channel will decrease exponentially according to the expression [48]: c(x) = c0 e−x/l

(1)

where c0 is the analyte concentration at the accumulation wall surface (x = 0) and l is the center of gravity distance from the accumulation wall for the analyte cloud at equilibrium. The parameter l can be expressed in a non-dimensional form as = l/ω, where ω is the channel thickness and is the retention parameter. For flFFF, is expressed as [48]: =

V0 D V˙ c ω2

(2)

where V0 is the void volume (mL), V˙ c is the cross-flow rate (mL s−1 ), and D is the analyte diffusion coefficient (m2 s−1 ). A relationship exists between the elution behavior and intrinsic physicochemical parameters contained in , which can be experimentally determined and is expressed by the retention ratio, R. The retention ratio is defined as the ratio of the void time, t0 , and the analyte retention time, tR , and is related to by R = 6 coth(1/2) − 122 . When R is in the range from 0.03 to 0.17, an approximation for in Eq. (2) can be applied to account for the particle shape [49]: D=

V˙ c ω2 R AV0

(3)

where A = 6 for spheres, A = 12 for thin rods and A = 18 for thin disks. Fractionation of analytes in flFFF (both symmetrical and asymmetric channels) is determined by D, and therefore hydrodynamic diameter (dH ) can be calculated by applying the Stokes–Einstein relationship: D=

kB T 3dH

(4)

where is the viscosity of the mobile phase (kg m−1 s−1 ), T is the temperature of the medium (at room temperature, 293 K)and kB is the Boltzmann constant (1.38 × 10−23 kg m2 s−2 K−1 ). More specifically to the channel geometries and flow profiles common for A4F,

12


Fig. 1. Illustration of the A4F channel components and the principal steps for the analysis for mixed populations of particles differing in size. The accumulation wall (membrane) is permeable, which allows flow through the bottom of the channel. Note that this scheme is not to scale. In A4F normal mode elution the separation process occurs in the lower 10% of the channel height.

determination of D based on tR under normal mode elution can be described by: ω2 tR = ln 6D

V˙ c 1+ V˙ p

(5)

where V˙ p is the channel flow (mL min−1 ). 2.2. Criteria for A4F fractionation The retention ratio R is one of two key parameters to evaluate the efficiency of fractionation by A4F in normal mode elution. In this context, efficiency is the balance or compromise between speed of analysis and resolution. When V˙ c is kept constant and tR is sufficiently large (such that R < 0.2), Eq. (3) can be simplified to a linear relationship between D and R by using Eq. (4): D=B×R

(6a)

and thus between dH and tR : dH = C × tR

(6b)

where B and C are constants of proportionality. The optimal range for R is approximately 0.03 ≤ R ≤ 0.2. The linearity of these two equations is one of the initial criteria used to evaluate the quality of fractionation in normal mode. In general, R can be manipulated by changing the ratio of cross-flow to channel flow; increasing this ratio leads to slower elution (higher retention) and higher resolution, but at the cost of increased analysis time and a greater potential for sample loss. In practice, the retention behavior for a specific analyte will also be impacted by other experimental parameters and conditions (see below). In chromatography, selectivity (Sd ) measures the ability of a chromatographic technique to separate two components; Sd is therefore of paramount importance for the analysis of multimodal systems and is an important additional criteria used to establish the

quality of a fractionation method. For flFFF techniques, selectivity can be defined as:

d(ln D) d(ln R)

Sd =

(7)

A higher Sd value translates to higher resolution and reflects a significant change in elution time with a small variation in analyte size. Generally, in flFFF (symmetrical system) the maximum value for Sd is unity. Finally the recovery, R(%), which is defined as the ratio between recovered mass after analysis, m, and initial injected mass, m0 , is expressed as a percentage according to: R(%) =

m × 100 m0

(8)

where R(%) can be used to assess the performance of the fractionation method, i.e., the ability of the method (instrument components + procedures and experimental conditions) to fractionate analytes without a substantial loss of material or a significant loss of information. To determine R(%), quantification of the analyte with an appropriate mass sensitive detection method is necessary (e.g., UV–Vis absorbance, differential refractometer, fluorescence, or ICP–MS). One can then calculate recovery based on two different approaches: (i) Off-line approach, wherein the analyte concentration is determined directly in the sample prior to fractionation and then off-line by the same detection method following fractionation and collection of the eluting peak. (ii) On-line approach, wherein the sample is injected both with and without application of the cross-flow field. In this case, the area under each peak is integrated, and the difference represents the analyte mass loss in the channel. By defining m0 in these two different ways, the performance of the initial conditions can be evaluated, and the percent contributions from both the analytes retained in the channel and the “dissolved phase” can be determined.


13

Fig. 2. Schematic illustration of protocol developed to optimize A4F fractionation for NP characterization. Details regarding each step are discussed in the text. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The final criterion used to evaluate the quality of the employed conditions is the resolution factor, Rs , which determines the degree of separation between two eluting analytes. For flFFF, Rs is defined as: Rs =

tR av

(9)

where av is the average of the two peak widths defined as the half maximum at full width and carry units of time. With Rs = 1.0, two peaks overlap by ≈4%. Rs values less than 1.0 indicate peaks that substantially overlap, while at a resolution of 1.5, the peaks are considered fully separated (i.e., baseline resolved). 3. Protocol development As previously stated, the scientific literature is replete with studies on the application of A4F for size characterization of a wide range of nanomaterials. Although in each of these studies there is typically a summary of optimized conditions determined for fractionation of the specific target analytes, a published systematic protocol for optimization of A4F conditions to fractionate two or more nanoscale components of a given type of material, and the necessary optimization and calibration procedures for particle size characterization of those components, is currently lacking. To address this need, a protocol is presented herein (illustrated schematically in Fig. 2), which demonstrates a single work flow that successfully optimizes parameters for NPs with widely ranging physicochemical properties. Many experimental parameters can influence the A4F fractionation process; the principal contributors are: - Mobile phase composition, - Rate of cross-flow and V˙ c /V˙ p ratio,

-

Channel dimensions, Quantity of analyte injected, Relaxation time, Membrane (material and pore size cut-off).

The initial focus will examine optimization of A4F conditions to adequately resolve the NP size distribution, as this is the most basic objective in the application of A4F, and it represents the most complex issue in terms of the fractionation criteria. 1. Elution. The initial step is to use the conditions presented in Table 1 as a starting point to perform a series of analyses with a test elution program in order to obtain preliminary information on the sample that will help to guide the optimization process. Additionally, material specific properties, such as quantity injected, have to be considered. These issues are discussed in greater detail in Section 5. 2. Channel flow. The channel flow rate (V˙ p ) should be fixed prior to optimization. It will have a direct impact on peak broadening by influencing the residence time in the channel. Typically, V˙ p is set within the range (0.5 to 1.0) mL min−1 . Below 0.5 mL min−1 , extremely slow elution (high retention) is typically observed and immobilization (loss) of a significant quantity of the sample is possible, which can result in low recovery. Above 1.0 mL min−1 , the pressure inside the channel increases substantially and high V˙ c must be applied in order to fractionate very small analytes, thus substantially constraining the available experimental range. 3. Channel components. Channel geometry (spacer thickness, channel width and shape) is a fundamental parameter impacting the elution and fractionation processes. Whereas channel width influences the shape of the eluting peak, it is not a crucial parameter for fractionation. On the other hand, channel thickness, ω, controls the parabolic velocity profile in the channel. The choice of ω is dependent on the diffusion (size) of the NP, and generally ω

14


Table 1 Starting conditions recommended for initiation of A4F optimization for aqueous samples. A4F parameters

Starting conditions

Comments

Mobile phase

NH4 NO3 solution

Spacer

(290 to 370) ␮m

Membrane

Polyethersulfone (PES) 10 kDa MWCO

Channel flow

0.5 mL min−1

Focus flow (relaxation step)

(0.5 to 3.0) mL min−1 for 4 min with injection flow plus 4 min without injection flow

Cross-flow (elution step)

(0.5 to 1.0) mL min−1

Monovalent salt Generally no interaction with NPs Compatibility with all detectors Starting concentration 1 mmol L−1 For metallic NPs: (350 to 370) ␮m For metal oxide NPs: (290 to 320) ␮m Broad compatibility with NPs and polymers MWCO adequate for NP size ranging from (1–100) nm Special condition: For positively charged or neutral NPs, we suggest the use of cationic surfactant in combination with NH4 NO3 in the mobile phase to mitigate attractive interactions with membrane Sufficient to elute NPs without overpressure Wider cross-flow range can be accommodated For monodisperse samples not subject to shear induced modifications: 2.0 mL min−1 For complex mixtures or polydisperse samples: (2.0 to 3.0) mL min−1 For “soft” analytes such as liposomes: (0.5 to 1.0) mL min−1 Special condition: For NPs that have strong interactions with membrane, a small focus flow will be preferred to avoid forced irreversible deposition on the membrane surface and loss of analyte Initially, a cross-flow that is too high can create a broader distribution and interaction with the membrane surface

between 250 ␮m and 490 ␮m is used. To choose an optimal channel height (as defined by the spacer), an understanding of the analyte position in the channel and the position in the velocity profile under the same elution conditions (channel flow and cross-flow rates) should be considered. For example, increasing ω from 250 ␮m to 490 ␮m results in the analyte being positioned at a lower point in the velocity profile and thus will have a correspondingly higher residence time in the channel. Because D (and 1/dH ) is proportional to ω2 (Eq. (5)), a variation of the channel thickness can influence the apparent D (or dH ) determined from tR . Furthermore, due to the compressibility of the underlying membrane and the interaction of the sample with the membrane surface, the final “effective” channel thickness (ωeff ) can be smaller than the nominal value for a given spacer. To determine ωeff it is necessary to analyze a sample with a well-defined D (or dH ) for every test condition used, allowing ωeff to be substituted in to Eq. (5) and calculated by [50]

ωeff =

6DtR

ln 1 + V˙ c /V˙ p

(10)

Typically, a well-defined D reference value can be obtained by one of two approaches: (1) using a DLS detector coupled on-line that allows determination of D directly according to the retention time or (2) using high quality monomodal (spherical) NPs of the same or similar composition to the analyte, with a well-defined size from which D can be calculated. In both scenarios, accuracy (and thus reliability) of the reference sample is optimized by conducting A4F and/or DLS measurements for the reference under conditions identical to that used for the target analyte. 4. Mobile phase. The mobile phase composition is one of most important and flexible parameters in A4F, and should be chosen to both ensure compatibility with the analyte and to improve operational performance. The properties of the mobile phase (composition, ionic strength, pH) directly influence the electrical double-layer (EDL) on the NP and membrane surfaces. The thickness of the EDL is characterized by the Debye length, −1 , which for a monovalent electrolyte at 25 ◦ C in aqueous solution is given by 1 −1 = 0.304 √ I

(11)

where the nDebye length is given in nm and I is the ionic strength (I = 1/2 i=1 ci zi2 , ci is the concentration of each ionic species in

mol L−1 and zi is the charge valence of the ion). The magnitude of the −1 depends only on the properties of the mobile phase and not on properties of the membrane surface, such as its charge or potential [51]. For like-charged membranes and particles, a higher mobile phase ionic strength compresses the EDL thickness (decreases −1 ), allowing the analytes to more closely approach the membrane surface; this can potentially induce irreversible adsorption with attendant loss of sample. Mobile phase ionic strength should thus be set at an appropriate value to strike a balance between providing sufficient electrostatic repulsive force to prevent significant loss of analyte to the membrane surface, and avoiding overly strong repulsive interactions that may result in pre-elution phenomena (entrainment of analyte in the void peak), which is schematically represented in Fig. 3. In general, it is best to choose an indifferent (non-reactive) electrolyte system that enables pH buffering, while also establishing an appropriate ionic strength. The work presented here uses the monovalent salt ammonium nitrate (NH4 NO3 ), because it has minimal adverse impact on recovery for a wide range of NP analytes and it is chemically inert with respect to components (including detectors) typically in contact with the analyte solution. Additional details regarding other beneficial mobile phase compositions are presented in Step 6.B. 5. Fix relaxation and injection. The injected sample volume is determined by the sample loop used on the injector. The quantity (mass) of analyte injected in an elution experiment has a crucial influence on the fractionation quality. Sufficient material must be present to yield a reasonably high detector signal-to-noise ratio (S N−1 > 5). In contrast, if the quantity is too high, this can result in overloading phenomena that alters the retention behavior of the analyte. At the same time, it is necessary to dilute into an aqueous medium that does not significantly modify the analyte form or stability. During injection of the sample into the A4F channel, the analyte is uniformly distributed across the depth of the channel. A relaxation process (referred to as focusing) is then applied in order to achieve analyte equilibration in the mobile phase prior to elution. During relaxation the analyte is concentrated into a thin band located near the channel accumulation wall that is generally positioned directly under the injection port. The position can be adjusted by controlling the balancing flow. Generally, for flFFF techniques, the relaxation period ( ) is


15

Fig. 3. Representation of Debye length −1 and its influence on the electrical double layer as a function of mobile phase ionic strength I. Small changes to I result in large changes to −1 in low I media (highlighted area), but are much less pronounced with increasing I. The interaction between NPs and membrane, shown schematically, can be optimized based on the desired retention behavior in the channel by adjusting the mobile phase I.

chosen based upon the rate of cross-flow generated by the focus flow [48]: = 1.2

V0 V˙ c

(12)

Additional time can be added to to ensure removal of (nearly) all dissolved species in the injected sample solution that could potentially interfere with the measurement. 6. A. membrane optimization. Initially, an elution program should be performed without cross-flow (i.e., no applied field) in order to determine the baseline analyte recovery (vide supra). A low recovery may indicate significant interaction and attachment between the analyte and the membrane. For analysis of nanoparticles, different membrane compositions are commercially available (e.g., regenerated cellulose, polyvinylpyrrolidone or polyethersulfone). Examination of A4F performance with membranes of other compositions has been reported previously [48], and the current work flow inherently contains the appropriate evaluation tools when necessary. For optimizing A4F conditions, polyethersulfone (PES) membranes present several advantages compared to the other commercially available membranes. For instance, PES membranes possess negative charge density for a broad pH range (3 to 11) [52]. Additionally, pore size and porosity of the PES membranes can be controlled by the addition of surfactant to the mobile phase [53,54], which also significantly decreases irreversible fouling and total fouling of the PES membrane. Therefore, we suggest starting with the PES membrane, which is commercially available, provides reproducible results, and is versatile for a broad range of pH values and mobile phase compositions. Based on the observed recovery of the analyte in the channel in the absence of a cross-flow, one of two optimization steps should be followed: for R(%) < 70, the membrane pore size should be addressed first, in order to ensure that there is not a significant loss of material through the membrane pores. If, after decreasing the molecular mass cut-off of the membrane, there is no significant improvement in recovery, then one must conclude that there is a strong interaction with the membrane itself, leading to sample loss. In this case, the mobile phase should be optimized (Step 6.B), after selecting the pore-size that results in the higher recovery value. We found that modifying the mobile phase pH and composition are more effective for controlling membrane-analyte interactions than changing the composition of the membrane. Sample loss may also be evident visually if the analyte exhibits strong optical absorbance (e.g., surface plasmon resonance effects).

6. B. Mobile phase optimization. Previously in Step 4, the mobile phase composition and I were chosen based on the initial information known about the physicochemical properties of the sample (e.g., size range, zeta potential (positive or negative), natural or manufactured source, surface chemistry or nature of capping agent, presence of surfactants, etc.). In order to optimize the mobile phase, the parameters previously outlined controlling the EDL of both the membrane and the analytes should be considered, specifically, ionic strength, pH, and composition. It is generally recommended to use a near-neutral buffer to establish the pH. This may not always be compatible with analyte stability, and thus matching the pH of the mobile phase with that of the sample is a reasonable alternative approach. Moreover, working at a near-neutral pH maintains a near constant membrane surface charge. Operating in extreme pH environments can induce some damage to the integrity of the whole system. Overall, working in a constant near-neutral pH range provides safe operating conditions for the system and provides the user with two parameters, ionic strength and mobile phase composition, to optimize fractionation. As stated previously, the salt concentration, I, is one of the most important parameters due to its direct influence on the electrostatic properties of the system. Increasing I compresses the EDL on all charged surfaces; such charge screening will increase the analyte residence time (i.e., retention time) in the channel if the analyte and membrane are of like charge (i.e., mutually repulsive). In general, a high ionic strength (e.g., I > 0.1 mol L−1 ) will cause a substantial increase in retention, as the electrostatic repulsive interactions are largely screened; however, the “stickiness” of the analyte will also be impacted by the composition and nature of the particles themselves. For instance, the presence of hydrophilic ligands like polyethylene glycol (PEG) can reduce or eliminate attractive interactions between the analyte and membrane. Although NH4 NO3 is employed in the protocol, NaCl, NaNO3 , KCl and other simple electrolytes may also be appropriate if compatible with the analyte and sample medium. Sodium azide (NaN3 ) has also found widespread use as a combination biocide (to maintain sterile conditions and prevent microbial growth at the fluid contact surfaces within an A4F system) and supporting electrolyte [55]; however, we do not recommend using NaN3 because of its intense UV absorption band between (250 to 280) nm and due to hazardous waste storage and disposal issues related to its toxicity. If NaN3 is used, contact with acidic solutions should be avoided, and the salt should be handled and stored appropriately.

16


Finally, another approach to control analyte-membrane interactions is the use of surfactants (cationic, anionic or non-ionic), which can interact with both the analyte and membrane surface [48]. In this case, it is critical to judiciously select compatible surfactants for specific applications, such as SDS or CTAB, and to control the surfactant concentration in the mobile phase. The inclusion of surfactants can impact the ionic strength and pH, and lead to both repulsive and attractive interactions between the analytes and the membrane. Surfactants can also modify the analyte size and state of dispersion, and this should be considered when deciding to employ a surfactant. Poor choice of surfactant and/or inappropriate control of concentration can have deleterious effects on retention behavior, including analyte loss to the membrane surface and pre-elution phenomenon. Additionally, the concentration of a surfactant should be kept well below the critical micelle concentration, as micelle formation can generate an undesired population of eluting species that may interfere with the detection of similar size analytes. For example, in the case of carbon nanotubes, it has been shown that the presence of SDS in the mobile phase at the same concentration (i.e., ionic strength) as an inert salt (NH4 NO3 ) increases interaction between the nanotubes and the membrane [56]. In contrast, the addition of CTAB to the mobile phase provides sufficient repulsion of CTAB coated gold nanorods (AuNRs) to elute without significant loss. Increasing the CTAB mobile phase concentration further eventually results in pre-elution phenomena, demonstrating how the mobile phase composition can significantly affect the retention behavior [57]. Though beyond the scope of this tutorial, off-line measurements of zeta potential and/or size stability over time using, for example, DLS, can help to quickly screen for appropriate mobile phase compositional ranges that are compatible with the analyte. By optimizing mobile phase composition, R(%) > 70 should be attained; if R(%) < 70, this indicates that conditions associated with the sample preparation and/or focus steps are inappropriate for the analyte (including the quantity of analyte injected). To further increase R(%), optimizing the sample may be necessary. 6.C. Sample preparation. If after modifying the membrane and mobile phase composition, without evident improvement, optimization of the sample preparation may be necessary. The most likely cause is an inappropriate analyte concentration. If the concentration is too low, poor signal-to-noise or no apparent signal would be observed. To resolve this issue it is judicious to concentrate the analyte using a stirred ultrafiltration cell with the same membrane (molecular weight cut-off—MWCO and nature) used in the A4F channel. The main advantage of using a stirred cell is the efficiency of the process and the high recovery (typically no significant loss). If NPs are agglomerated during ultrafiltration, it may be possible to redisperse the sample using sonication; this approach has been used to disperse TiO2 , carbon nanotubes and other NPs in aqueous media [58–62]. If the combination of ultrafiltration and sonication are not effective, it may be possible to modify pH and ionic strength in order to optimize stability and preserve the integrity of the sample during the concentration process. Another option for concentrating the analyte is centrifugation, especially if sample quantity is a limiting factor. Both approaches require testing and validation. 7. Optimization of cross-flow. If R(%) is at least 70, the next step is to optimize V˙ c . There are two principal objectives in A4F: (i) to obtain a size distribution from the retention time measurement (red box in Fig. 2) and (ii) resolve two or more components coexisting within the same sample (i.e., chromatographic fractionation, blue box on Fig. 2). The flFFF theory applies most simply when V˙ c is held constant during elution; we therefore recommend determining a V˙ c value that is appropriate for both fractionation (component separation) and size determination (low loss and minimal band broadening with no pre-elution). If the evaluation parameters (Sd ,

R and R(%)) are not in an acceptable range, the next step is to vary and optimize V˙ c /V˙ p by fixing V˙ p and varying V˙ c . If the principal objective is size determination, and no improvement is obtained by changing V˙ c /V˙ p , then the channel thickness is not appropriate and one should proceed to change it based on the criteria in Step 3. If separation (resolving) of multiple components is the principal objective, fixing V˙ c is not a requirement. In this case, an elution program with a gradient V˙ c can enable rapid separation between two populations with an acceptable Rs , i.e., Rs > 1.0. 4. Protocol evaluation and application The principal goal of this tutorial is to demonstrate that the proposed measurement strategy facilitates an efficient method development process for A4F fractionation and characterization of NPs varying in composition, size, shape and dispersion state. The protocol outlined above and shown schematically in Fig. 2 was established by evaluating a broad range of NPs in order to assess the efficacy, robustness and applicability range. The specific NPs selected to develop and validate this strategy represent an important set of materials for industry and for international efforts to address the risks associated with nanotechnology [63]. Discussion on the relevance of detector choice in the context of the specific analytes and desired information is also included. Application of the protocol is demonstrated by first optimizing conditions for polystyrene latex (PSL) and citrate-stabilized gold NPs, referred to as simply AuNPs herein, which will provide a baseline set of optimized conditions for two NPs with distinct physicochemical properties that impact their fractionation and detection. Furthermore, these materials are easy to prepare in the laboratory using well documented procedures or are commercially obtainable. The rational optimization of A4F conditions for other materials with distinct physicochemical properties is then discussed in order to demonstrate the general applicability of the protocol over a broad range of nanomaterials. 4.1. General precautions and starting conditions Before optimizing the conditions for A4F, some starting parameters must first be chosen, such as the quantity injected, the mobile phase composition and the channel thickness. Additionally, it is necessary to choose an appropriate standard material to determine the effective channel height when measuring the diffusion coefficient (or size) based on the retention time. 4.1.1. Calibration: Influence of NP core material As previously stated, calibration of the retention time should be conducted using a known internal reference material comprising the same core material as the target analyte (or as closely matched as possible). If size can be determined using an on-line detector (such as DLS) with sufficient sensitivity, the retention time calibration is not necessary; however, comparison of retention time-based size with on-line DLS results may offer additional insight into the fractionation process for a particular analyte. It was recently demonstrated that for both polymeric and inorganic NPs the core composition has a significant influence on the retention time, an influence not described in the classical FFF equation (see Section S1 and Fig. S1 in the electronic supplementary material, ESM) [64], which demonstrates the importance of selecting the appropriate calibration material and method for diffusion–retention time evaluations. 4.1.2. Quantity injected As previously stated in Step 5, size determination from retention time and the efficacy of the fractionation process are both sensitive


17

was observed for the range of conditions tested (i.e., I and V˙ c ) using these salts (see ESM, Fig. S2). The mobile phase composition can also be modified to manipulate retention behavior by means other than changes to I. In contrast, because commercially available membranes have intrinsic negative charges, positively charged species usually cannot be optimized simply by changing I. An approach to modify the membrane surface and allow the evaluation of positively charged species is the addition of surfactants to the mobile phase. The choice of surfactant must be made carefully. For example, CTAB-stabilized AuNRs require the use of surfactant in the mobile phase (see below) [57,65]. In contrast for PSL NPs, the presence of surfactant (both anionic and cationic) in the mobile phase can lead to substantial interaction between the analyte and the membrane, resulting in lower recovery [48,56,66] and poor reproducibility in retention. Optimization of the mobile phase composition based on specific physicochemical properties of different nanomaterials will be discussed in the constructs of the protocol progression in the next section. 4.2. Stepwise application of protocol

Fig. 4. Influence of the total number of NPs injected into the A4F channel on (a) the measured retention time for the eluting peak (tR = tmax , corresponds to the maximum intensity of the retention peak) and (b) the effective channel thickness determination. Circles represent PSL NPs and triangles are Au NPs. The particle size is the nominal diameter reported by the vendor. The standard deviation of triplicate measurements on the retention time and the effective channel thickness were < 1%.

to the quantity of the analyte injected. Fig. 4 illustrates the influence of the number of NPs injected on the measured retention time and the error associated with retention-based size determination for 30 nm and 60 nm PSL and Au NPs. The data clearly show two distinct regions for both PSL and AuNPs at all sizes. At lower particle numbers (particle number was estimated assuming spherical shape and using the reported mass fraction and nominal size from the provider), tR remains constant with increasing sample addition. However, a threshold is observed at increased particle numbers, resulting in depression of tR caused by sample overloading in the channel. The change in tR directly affects the ωeff determination; therefore, it is imperative to work in a NP concentration range that is insensitive to slight changes in particle number. A particle number range from 107 to 1010 is recommended as for obtaining the most accurate size determination without significant change in the retention time, but validation of the number concentration range should be performed for other analytes. 4.1.3. Mobile phase composition influence Although I strongly influences the EDL on the membrane and analytes, and consequently the position in the channel, further modification of the mobile phase composition may be necessary to optimize Sd , R, and Rs . For PSL and AuNP samples, the effect on retention behavior of using different mobile phases consisting of monovalent salts was examined, including NH4 NO3 , NaN3 and NH4 CH3 CO2 over a range of I values. No significant difference in tR

Application of the protocol is demonstrated by optimizing the conditions for PSL and AuNP samples, summarized in ESM (Table S1), by proceeding through the outlined workflow in Fig. 2. The resulting conditions from the optimization of these two distinct materials provide a baseline to discuss further optimization methods for other nanomaterials with unique physicochemical properties. The choice of initial conditions should reflect the known information of the negatively charged Au and PSL NPs, which in this case were provided by the vendor. General conditions based on known sample information are outlined in Table 1, which should be incorporated into Steps 1–3. The first step in the protocol (Elution) is to inject different quantities of sample to evaluate the detector response relative to the quantity of introduced NPs, so as to achieve sufficient S N−1 and analyte recovery to proceed with analysis (see Fig. 4). An advantage of PSL and Au NPs is their ability to be evaluated with multiple detection modalities (e.g., online DLS and UV–vis). A disadvantage of the Au NPs is the difficulty associated with the radius of gyration, rG , measurement using an online static light scattering detector (e.g., MALS), which is a powerful tool for determining shape in non-optically absorbing NP systems. Proceeding to Step 2, the channel flow, V˙ p , was fixed at 0.5 mL min−1 to optimize the residence time of the NPs while avoiding over-pressurizing the channel. Further optimization of V˙ p can be incorporated into Step 7. To fix the channel components (Step 3), consideration of size (10 nm to 100 nm), shape (spherical) and surface charge (negative) of the NPs is necessary. The choice of the channel thickness affects the NP position in the velocity profile and consequently R. In normal mode elution, R should be less than 0.17, so the fast eluting NPs should approximately follow tR ≥ 6t0 . The determination of t0 is accomplished by injecting species that are not retained in the channel under test conditions (e.g., surfactants, small polymers with the same molar mass as the membrane pore size). Then t0 can be fixed in order to determine a desired spacer height. To approximate the desired experimental spacer thickness, ωdes , substitute ωdes for ω in Eq. (5) to give:

ωdes =

6DtR

ln 1 + V˙ c /V˙ p

(13)

By calculating D using Eq. (4) for the smallest NP in the current study, AuNP10, and a range of R values consistent with normal mode elution and with starting V˙ c values from (0.3 to 1.0) mL min−1 ,

18


Fig. 5. Representation of the recovery obtained for 100 nm PSL NPs as a function of mobile phase ionic strength and cross-flow rate. PSL100 was chosen because it represents the largest NPs used to develop the baseline conditions for the protocol.

a range for ωdes from (450 to 290) ␮m is obtained. Again, the value of V˙ c should be chosen to balance the compression of the sample cloud toward the accumulation wall (decreasing R(%)) and peak broadening that results from lower applied V˙ c . Therefore, based on the physicochemical properties of the smallest analyte and the calculated ωdes , a spacer near 350 ␮m is chosen. Note that the experimental spacer and the effective channel thickness are not equivalent, as the Mylar spacer determines the former and the latter is determined based on the observed retention behavior of the sample. Based on the data from Fig. 5, the mobile phase composition and I for the PSL and AuNP samples are fixed at 0.5 mmol L−1 NH4 NO3 . Again, further optimization of the mobile phase can be addressed in cases of low R(%) and unsuitable R values. Because the NPs being interrogated are reported to be monodisperse, calculating the necessary becomes straightforward using Eq. (13). For the two types of NPs used, a relaxation time of 30 s to 1 min is determined to be sufficient. However, we recommend that additional time be added to the focusing step to ensure the entire volume of the sample from the sample loop is injected into the channel and that the removal of the “dissolved phase” is complete. The volume of the sample injected will be determined by the sample loop employed, but concentration in the channel should be chosen based on the sample stability and behavior in the channel as mentioned previously. The conditions outlined in the initial five steps result in R(%) > 70 for PSL and AuNPs, and in this case the user should proceed to Step 7. However, when the initial five steps result in low recovery or noisy detector baseline signals, the most likely cause is strong interaction between the analyte and membrane (Step 6.A). If the membrane composition is incompatible with the analyte due to strong attractive forces, a range of commercially available membranes can be examined (polyethersulfone, regenerated cellulose, cellulose acetate, polyvinylidene difluoride, etc.) to adjust the affinity between NPs and the membrane. However, most commonly used commercial membranes are negatively charged, restricting the ability to optimize the interactions between the membrane and highly charged analytes. When membrane optimization (Step 6.A) is not available or is inefficient, the mobile phase should be modified (Step 6.B). To reiterate, the mobile phase optimization (Step 6.B) should prevent the loss of material without altering the analyte or changing its normal retention behavior inside the A4F channel. To

Fig. 6. Fractograms for AuNP+ and AuNR with a fixed ionic strength of 0.5 mmol L−1 analyzed in (a) simple salt containing mobile phase and (b) in the presence of an optimized concentration of cationic surfactant (0.35 mmol L−1 CTAB). The conditions for separation are: channel flow rate of 0.5 mL min−1 , cross-flow rate of 0.8 mL min−1 and 350 ␮m spacer thickness.

visualize the influence of I, which controls the thickness of the EDL (−1 ) at both the NP and membrane surfaces, a contour plot showing R(%) for PSL100 NPs as a function of I and V˙ c is presented in Fig. 5. A decrease in recovery is clearly observed as I increases toward 3.0 mmol L−1 for all values of V˙ c , which results in corresponding R(%) values less than the recommended threshold limit for optimization. The highest recovery is observed with I < 2.0 mmol L−1 , indicating the singular importance of employing the appropriate ionic strength. The results from Fig. 5 are summarized in Table 1. Positively charged NPs pose a unique challenge for fractionation, because of the absence of commercially available cationic membranes. Furthermore, optimization of I alone is not sufficient to counter the irreversible adsorption of these NPs onto the commonly used anionic PES membrane. To overcome the limitations of no commercially viable cationic membranes, we use two materials with positive zeta potentials, the dendron-capped, spherical AuNP20 (referred to herein as AuNP+) and CTAB stabilized AuNRs, to demonstrate the role mobile phase composition plays in affecting retention behavior (Fig. 6). Note, in the NH4 NO3 mobile phase (Fig. 6a), no light scattering signal is observed for AuNP+, indicating that nearly all of the analyte is lost to the membrane. For the AuNR sample, two peaks are observed, but signal intensities represent R(%) < 10, which is visually supported by the corresponding noisy baselines and peak splitting. Because the AuNR sample is known to be relatively monodisperse, the second peak can be assigned to characteristic peak splitting associated with strong attractive forces between the analyte and membrane. For these positively charged samples, no significant improvement to R(%) could be achieved


Fig. 7. Representation of selectivity obtained for 100 nm PSL NPs as a function of mobile phase ionic strength and cross-flow rate.

by changing I alone or by modifying monovalent salt composition, demonstrating the need for further modification of the mobile phase composition. In order to improve the elution of both AuNP+ and AuNR, the addition of a cationic surfactant, CTAB, was added to the NH4 NO3 mobile phase to modify the analyte-membrane interaction (Fig. 6b). By surveying different concentrations of CTAB in the mobile phase while maintaining a constant I = 0.5 mmol L−1 , it was possible to achieve R(%) > 90 (based on UV/Vis detection) for both Au samples when a mobile phase composition of 0.15 mmol L−1 CTAB and 0.35 mmol L−1 NH4 NO3 was employed. This level of recovery was not obtainable in the absence of the surfactant. Furthermore, the splitting of the AuNR peak was not observed under optimized conditions, and a sufficient S N−1 was obtained to determine size using online DLS (dH = 16 nm). The decrease in tR also provides evidence for decreasing interaction between the analyte and the membrane (Eq. (2)). As previously mentioned, the addition of too much or too little surfactant in the mobile phase can have undesired effects on native analyte properties (e.g., charge, state, structure), emphasizing the importance of choosing an appropriate surfactant and mobile phase composition that maintains the integrity of the sample. For the CTAB-containing mobile phase, a contour plot similar to that shown in Fig. 5 can be constructed to visualize the optimal concentration for other positively charged samples. Overall, optimization of the mobile phase composition allows commercially available anionic membranes to be used for the fractionation of a broader range of NPs. After sufficient recovery is achieved through the progression outlined in the first six steps, further refinement through crossflow variation provides the final optimization for both principal objectives of A4F: (i) obtaining a NP size distribution; (ii) obtaining size separation (Rs > 1). 4.2.1. Size distribution and retention measurement Optimization of conditions for the determination of the size distribution for PSL and Au NPs is first demonstrated. In Figs. 5 and 7, the R(%) and Sd for PSL100 are shown as a function of V˙ c and I, respectively. Importantly for size determination, Sd for PSL NPs is not significantly affected by I, as clearly illustrated in Fig. 7, indicating that the previous optimization of mobile phase (Steps 4 and 6.B) and R(%) are necessary before determining the size distribution. If the steps were reversed, inherent bias could be incorporated into the size distribution determination if individual analytes exhibited

19

different sensitivities to A4F conditions, i.e., if a lower R(%) was associated with a specific analyte loss in the sample, the size distribution obtained would misrepresent the actual distribution present. Getting back to the role of V˙ c , although Sd is not strongly affected by I, it is dramatically affected by V˙ c (Fig. 7). With increasing V˙ c the value of Sd approaches and surpasses unity, where the optimal Sd is obtained with V˙ c = 1.2 mL min−1 . Higher values of V˙ c may incrementally increase Sd further, but one pays a price in lower R(%), longer elution times, modification in the peak shape, and potentially steric inversion elution phenomena. Theoretically, the maximum Sd achievable for flFFF is unity, but higher values of Sd are achieved due to the asymmetrical geometry of the channel. This optimized Sd is utilized to proceed with protocol validation. In order to evaluate the analytical performance of the proposed scheme to fractionate and characterize PSL and Au NPs, different size populations were analyzed by A4F-UV/vis-MALS-DLS using the optimized fractionation conditions developed in the previous sections. These NPs present the same zeta potential at the pH of the mobile phase. First, PSL NPs of different sizes were analyzed by both DLS and tR using Eq. (5) to obtain size (see Fig. 8a). The optimal fractionation conditions for PSL are as follows: V˙ c from (0.7 to 0.9) mL min−1 , a 2.2 mmol L−1 NH4 NO3 mobile phase composition and an effective spacer of 319 ␮m. To calculate ωeff , tR for the PSL100 sample was used. The results for D obtained by DLS and based on tR were self-consistent for all PSL NPs examined, from (20 to 100) nm; this indicates a linear relationship for all PSL NPs in this size regime. Furthermore, information regarding particle shape can be obtained from the shape factor, = rG /rH , where the radius of gyration is derived from MALS and the hydrodynamic radius, rH , is determined from DLS; 0.775 < < 1 indicates spherical geometry while > 1 indicates prolate shape geometry [67]. From the data in Fig. 8a, all three PSL NP samples are spherical, providing further confidence in the application of Eq. (4) to obtain D using the known size for the PSL NPs. For the AuNPs, optimal fractionation conditions differ from those determined for PSL above. While V˙ c is in the same range at 0.8 mL min−1 , a lower ionic strength is required in the mobile phase (0.5 mmol L−1 NH4 NO3 ) and therefore the effective spacer thickness differs (278 ␮m). By using these optimal fractionation conditions, a linear relationship between tR and D, measured online with DLS, was observed for all samples ranging from 10 nm to 80 nm (Fig. 8b, red trace), which could be represented from Eq. (6b) as tR = 0.49· dH (R2 = 0.9979) and with a corresponding Sd = 1.08 (R2 = 0.9985). Furthermore, when different size AuNPs were mixed (Fig. 8b, black trace), no change in tR was observed, demonstrating the optimized conditions that were determined following the protocol result in high resolution fractionation of the NPs. By comparison, identification of distinct AuNP populations was not achieved for the same mixture when examined with batch DLS (Fig. S3). A limitation of the detection scheme used for the AuNP samples is the determination of rG based on analysis of MALS data due to the strong surface plasmon resonance (SPR) absorbance. The optical properties of AuNPs (imaginary and real refractive index) do not exhibit a typical size-dependent angular dispersion of scattered light, [68] impeding the capacity to collect size information (radius of gyration or geometric radius) with the MALS detector. Overall, the optimization of the fractionation of the PSL and AuNPs described by high Sd and linearity demonstrates the efficacy of the protocol for determining the size distribution present in such samples. From the data in Fig. 8, it is clearly observed that similar but different optimal fractionation conditions exist for these two NPs. It is imperative to reiterate the importance of using an appropriate internal calibrant in A4F. Although similar fractionation conditions were obtained for PSL and Au NPs, the physicochemical properties of these materials are distinct (see Fig. S1), and they elute at

20


Fig. 8. A4F-UV/vis-MALS fractograms obtained using the optimal conditions described in the main text for (a) PSL NPs and (b) mixture of NIST AuNP reference materials (nominally 10–30–60 nm, solid line) and the commercially obtained AuNPs (dashed line) analyzed individually. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

different times even under identical fractionation conditions and despite similar size and charge. The size determination from tR cannot be accurately calibrated for AuNPs using PSL or vice versa, as their elution times under identical conditions vary substantially even though they are nominally the same size. This issue has been demonstrated and discussed in greater detail in a previous publication [64]. It is important to have well characterized internal standards for D (diffusion coefficient or corresponding size) to determine ωeff , especially when size based on light scattering measurements is either not available or not obtainable. 4.2.2. Optimization for component separation The A4F optimization can also be performed with the principal objective to obtain complete separation of NP populations, and as such will require different evaluation criteria as outlined in Fig. 2 (blue box). To demonstrate the application of the protocol to achieve the separation of two or more species, we again optimize V˙ c to separate AuNP+ (Fig. 9). Using the optimized V˙ c = 0.8 mL min−1 found from the previous steps, the separation of the broad AuNP+ peak from t0 was achieved (6b and 9a). However, separation of individual species could not be detected under the current conditions, which is misrepresentative of the sample. By increasing V˙ c from (0.8 to 1.3) mL min−1 , the separation of a previously latent peak is observed at tR = 3.9 min; this peak exhibits the characteristic SPR band for spheroidal AuNPs at 520 nm (not shown). The separation of the peaks results in Rs = 0.98. At V˙ c = 2.0 mL min−1 , further separation of the peaks is achieved, resulting in Rs = 1.24. These data demonstrate the optimization of separating individual populations with distinct D by varying V˙ c . However, the optimization observed in Fig. 9 does not result in the ability to determine size information based on tR for a broad range of sizes. For larger AuNP+ sizes, substantial peak broadening, low recovery and slow elution occurs under these specific fractionation conditions. In the case of larger nanoparticles, further optimization is required to obtain accurate size information. 4.3. General starting conditions for preliminary A4F analysis Together, the data presented thus far demonstrate the application of the proposed protocol for optimizing conditions to achieve the two principal objectives for A4F. We recommend that the previously determined optimal conditions for PSL and Au NPs, summarized in Table 1, be utilized as baseline conditions to begin the process of optimizing A4F fractionation for other negatively charged NP analytes. Note that the tabulated conditions are used simply to provide a starting point in order to obtain sufficient information to enable one to proceed with the protocol presented

schematically in Fig. 2. Furthermore, implementation of different mobile phase compositions to manipulate analyte-membrane interactions for specific systems can be efficiently evaluated following these protocol procedures; this should broaden the range of sample types and matrices that can be evaluated by flFFF with appropriate coupled detection modalities. 5. Protocol performance range: Upper and lower size limits for characterization using A4F Evaluating the size range of an analytical technique or methodology is necessary to validate the efficacy of the protocol used to optimize a given sample. The upper and lower limits for NP fractionation and size characterization have been investigated and are discussed below. 5.1. At the bottom: Nanoclusters To probe the lower size limit capabilities of A4F for both separation and determination of D (or dH ) based on tR , we provide an example borrowed from a recently published study examining the continuum of persistent transformation products in the nanoscale Ag/glutathione (SG) system [69]. Because of the extremely small size of these atomically-precise nanoclusters, the largest available spacer (490 ␮m) was used to optimize the nanoclusters in the parabolic flow profile, resulting in R sufficient for determining size. By using the distribution of SG-protected Ag nanoclusters with both known distinct optical signatures and relative core sizes, [70] the separation capabilities are clearly demonstrated (Fig. 10). In the fractogram two distinct peaks are observed in the 254 nm UV trace (red line) associated with tR at 4.15 min and 5.61 min and denoted Peak 1 and Peak 2, respectively. In this case, R is sufficient for determining size using Eq. (5). Note that the 400 nm trace in the fractogram is included in order to demonstrate the limitations of using a single wavelength to monitor ligated, metallic nanoclusters, which can exhibit strong shifts in optical signatures with small changes in core size (nuclearity), ligand coating, or geometry.[71–74] The calculated dH of Peak 1 and Peak 2 are 1.8 nm and 2.5 nm, respectively, which likely represent an average size (nuclearity) based on previously reported optical spectra (see ESM, Fig. S4) [70]. The optical signature exhibited by the 1.8 nm peak (Fig. 10c) is consistent with Ag32 (SG)19 ,[73] and dH is consistent with the approximate size based on mobility in polyacrylamide gel electrophoresis,[75–77] indicating Eq. (5) is applicable for this size regime. Because the molar absorptivity, ε, of the individual nanoclusters differs due to distinct electronic structures, quantification


Fig. 9. Fractograms for AuNP+ (dH = 16 nm corresponding to peak 2) using a constant mobile phase composition (mixture of 0.15 mmol L−1 CTAB and 0.35 mmol L−1 of NH4 NO3 ), a 350 ␮m spacer and three different cross-flow rates V˙ c : (a) 0.8 mL min−1 , (b) 1.3 mL min−1 , and (c) 2.0 mL min−1 .

with absorbance measurements requires knowledge of ε for each contributing species. Therefore, coupling with ICP-MS and monitoring m/z 107 signal intensity (trace indicated by black dots in Fig. 10) provides quantitative information regarding the Ag mass distribution irrespective of differences in the species present in the peak. Two distinct peaks are observed in the m/z 107 signal consistent with the 254 nm trace. The majority of the Ag mass distribution observed in the fractogram is found in Peak 2, demonstrating the need for appropriate detection schemes based on the analyte system. Changes in the peak intensities of the nanocluster populations over time were also observed (Fig. 10b), clearly demonstrating the ability to quantify the transformations of the nanoclusters over time.[69] Furthermore, although the MWCO for the PES membrane (10 kDa) represents an effective pore size (≈3 nm) that is larger than the size of the Ag nanoclusters, the nanoclusters are still retained in the channel due to the optimized analyte-membrane interactions, indicating that the performance of a membrane for removal

21

Fig. 10. Fractogram of GSH-protected Ag nanocluster solution at (a) t = 0.33 h −1 and (b) t = 24 h after resuspension in DI water examined at V˙ c = 2.0 mL min , Vp = 0.5 mL min−1 , 490 ␮m spacer and 0.5 mmol L−1 NH4 NO3 mobile phase. (c) Online DAD spectra for peak 1 and peak 2 at t = 0.33 h. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

of individual populations should be evaluated as described in Step 6A. Overall, the ability to identify, characterize and quantify metallic species in the sub-5 nm regime with A4F is clearly demonstrated, thus providing a measurement tool capable of probing the smallest size regime in more complex media with minimal sample perturbation. 5.2. Upper size limit: Sub-micrometer populations PSL is widely used to calibrate flFFF systems for the characterization of organic NPs, due to the similarities in density and composition. However, it is also commonly used to calibrate flFFF systems for other classes of particles, a practice we have determined can lead to significant errors in retention-based size determination [64]. A wide size range of PSL is used here to define the upper size limit for “normal mode” elution, with mean diameters ranging from (60 to 600) nm. Two different mixtures of PSL spheres were created, consisting of PSL60-PSL100-PSL200 and

22


possible to identify and determine the transition between the two elution modes, and consequently identify the upper size limit for reliably determining size based on tR . This same experiment could be repeated for different types of NPs, in order to determine, for example, if the transition point is universal or material dependent. 6. Summary

Fig. 11. A4F-UV-MALS fractograms for two mixtures of PSL with diameters from 60 nm to 600 nm. The red trace corresponds to the mixture containing PSL60-100200 and the black trace corresponds to the mixture containing PSL200-400-600, and the measured radius of gyration, rG , for each sample is presented (black ×). Two linear regressions through rG data are plotted: the dashed line corresponds to the fit including all PSL samples; the dotted line corresponds to the fit excluding the PSL600 sample. The deviation from linearity observed at the upper size limit is consistent with a transition from normal mode to steric mode elution. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

PSL200-PSL400-PSL600. PSL200 is common to both mixtures as a point of reference. The optimal conditions determined for this large size range using the protocol are: 250 ␮m spacer thickness, 0.5 mmol L−1 NH4 NO3 mobile phase, and V˙ c = 0.2 mL min−1 . Fig. 11 shows the resulting A4F-UV-MALS fractograms for the two PSL mixtures with regression lines for the correlation between rG and tR . We first highlight that measurement difficulties exist for accurately measuring larger particles by online DLS due to the slow correlation decay and short residence time in the measurement zone. As a result, the DLS data was too noisy in this experiment to be useful, whereas the MALS data was very strong and showed a clear angular dependence. From the data, it is clear that a linear calibration curve does not pass through all points (dashed line), indicating a deviation from linearity at the largest size. In particular, PSL600 nm elutes faster than would be predicted based on measured diameter. Using the optimal conditions for the broad size range reflected in Fig. 11, the smallest samples, PSL60 and PSL100, are not fully resolved (Rs < 1), with insufficient differentiation to confidently assign tR for PSL60. However, tR for PSL200 is the same in both mixtures (tR = 17.6 min), providing evidence that conditions are optimized for a broad range of materials. To determine the upper limits for size determination, the PSL600 data point is removed from the analysis and the remaining data is reexamined (dotted line). With the removal of the PSL600 data, the regression value is improved, suggesting PSL600 is outside the limit for determining size via retention calibration in normal mode elution. The deviation from linearity in this regime also signifies a transition from normal mode to steric elution; the latter is responsible for the increased speed of elution for the PSL600 sample. Extrapolating from the current data set, fractionation of PSL particles with size components exceeding ≈500 nm are likely to deviate from strict normal mode elution. Although this phenomenon has been previously described, accurately determining the upper limits for size determination has not been fully addressed [78], demonstrating the need for more thorough investigations on elution properties near the transition between normal mode and steric mode regimes. By using this protocol, it is

Over the past decade, flFFF techniques have been increasingly utilized for an ever expanding range of applications across many fields, including nanotechnology. The measurement field itself is evolving to address emerging challenges driven by both regulatory and research needs. Previously used by a comparatively small number of specialists, A4F is fast becoming a common tool found in many laboratories alongside (and sometimes in combination with) DLS and other more established techniques. With this growth comes increased need for guidance and standardization of procedures that can be broadly disseminated and that reduce the substantial burden of method development associated with A4F and other flFFF techniques. The lack of published procedures or standards has limited the ability of researchers to effectively apply and validate A4F methods across a wide range of analytes and measurement platforms. Therefore, in the present work, a comprehensive and rational protocol is described that guides the user toward optimization of measurement conditions and parameters based on fundamental criteria such as retention ratio, selectivity, and recovery. The choice and efficacy of detector selection is addressed using examples extracted from published research, past and present. A multi-dimensional approach to detection is encouraged, especially for the analysis of complex mixtures or in complex media where complementary and orthogonal information is most beneficial. Acknowledgments We thank Tae Joon Cho (NIST) for providing dendron-capped cationic AuNPs. We acknowledge Liz Nguyen and Hind El Hadri of NIST for internal review and useful comments. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.aca.2013.11.021. References [1] Y. Arai, S.Y. Jee, S.M. Kim, Y. Kwon, W. Jang, Biomedical applications and safety issues of gold nanoparticles, Toxicology and Environmental Health Sciences 4 (2012) 1–8. [2] M. Elsabahy, K.L. Wooley, Design of polymeric nanoparticles for biomedical delivery applications, Chemical Society Reviews 41 (2012) 2545–2561. [3] S.E. Lohse, C.J. Murphy, Applications of colloidal inorganic nanoparticles: from medicine to energy, Journal of the American Chemical Society 134 (2012) 15607–15620. [4] L.H. Reddy, J.L. Arias, J. Nicolas, P. Couvreur, Magnetic nanoparticles: design and characterization, toxicity and biocompatibility, pharmaceutical and biomedical applications, Chemical Reviews 112 (2012) 5818–5878. [5] D.B. Salunkhe, S.S. Gargote, D.P. Dubal, W.B. Kim, B.R. Sankapal, Sb2 S3 nanoparticles through solution chemistry on mesoporous TiO2 for solar cell application, Chemical Physics Letters 554 (2012) 150–154. [6] A.N. Shipway, E. Katz, I. Willner, Nanoparticle arrays on surfaces for electronic, optical, and sensor applications, ChemPhysChem 1 (2000) 18–52. [7] J. Xie, X. Zhang, H. Wang, H. Zheng, Y. Huang, J. Xie, Analytical and environmental applications of nanoparticles as enzyme mimetics, TrAC—Trends in Analytical Chemistry 39 (2012) 114–129. [8] M. Farré, J. Sanchís, D. Barceló, Analysis and assessment of the occurrence, the fate and the behavior of nanomaterials in the environment, TrAC—Trends in Analytical Chemistry 30 (2011) 517–527.

J. Gigault et al. / Analytica Chimica Acta 809 (2014) 9–24 [9] Y. Ju-Nam, J.R. Lead, Manufactured nanoparticles: an overview of their chemistry, interactions and potential environmental implications, Science of the Total Environment 400 (2008) 396–414. [10] L. Kantiani, M. Llorca, J. Sanchís, M. Farré, D. Barceló, Emerging food contaminants: a review, Analytical and Bioanalytical Chemistry 398 (2010) 2413–2427. [11] E.J. Petersen, L. Zhang, N.T. Mattison, D.M. O’Carroll, A.J. Whelton, N. Uddin, T. Nguyen, Q. Huang, T.B. Henry, R.D. Holbrook, K.L. Chen, Potential release pathways, environmental fate, and ecological risks of carbon nanotubes, Environmental science & technology 45 (2011) 9837–9856. [12] J. Sanchís, M. Farré, D. Barceló, Analysis and fate of organic nanomaterials in environmental samples, Comprehensive Analytical Chemistry (2012) 131–168. [13] C.M. Sayes, D.B. Warheit, Characterization of nanomaterials for toxicity assessment, Wiley interdisciplinary reviews, Nanomedicine and Nanobiotechnology 1 (2009) 660–670. [14] Z. Li, J.C. Barnes, A. Bosoy, J.F. Stoddart, J.I. Zink, Mesoporous silica nanoparticles in biomedical applications, Chemical Society Reviews 41 (2012) 2590–2605. [15] K.W. Powers, M. Palazuelos, B.M. Moudgil, S.M. Roberts, Characterization of the size, shape, and state of dispersion of nanoparticles for toxicological studies, Nanotoxicology 1 (2007) 42–51. [16] M. Hassellöv, R. Kaegi, Analysis and characterization of manufactured nanoparticles in aquatic environments, in: Environmental and Human Health Impacts of Nanotechnology, John Wiley & Sons, Ltd, 2009, pp. 211–266. [17] I.T. 27687.2008(E), in: I.O.f. Standardization (Ed.), Nanotechnologies— Terminology and Definitions for Nano-Objects—Nanoparticle, Nanofibre and Nanoplate, ISO/TS 27687:2008(E), Geneva, Switzerland, 2008. [18] I.T. 80004-1:2010(E), in: I.O.f. Standardization (Ed.), Nanotechnologies— Vocabulary—Part 1: Core terms, ISO/TS 80004-1:2010(E), Geneva, Switzerland, 2010. [19] M.S. Jiménez, M.T. Gómez, E. Bolea, F. Laborda, J. Castillo, An approach to the natural and engineered nanoparticles analysis in the environment by inductively coupled plasma mass spectrometry, International Journal of Mass Spectrometry 307 (2011) 99–104. [20] F.V.D. Kammer, S. Legros, T. Hofmann, E.H. Larsen, K. Loeschner, Separation and characterization of nanoparticles in complex food and environmental samples by field-flow fractionation, TrAC Trends in Analytical Chemistry 30 (2011) 425–436. [21] G. Lespes, J. Gigault, Hyphenated analytical techniques for multidimensional characterisation of submicron particles: a review, Analytica Chimica Acta 692 (2011) 26–41. [22] F.K. Liu, Analysis and applications of nanoparticles in the separation sciences: a case of gold nanoparticles, Journal of Chromatography A 1216 (2009) 9034–9047. [23] T. Arita, T. Yoshimura, T. Adschiri, Size exclusion chromatography of quantum dots by utilizing nanoparticle repelling surface of concentrated polymer brush, Nanoscale 2 (2010) 1467–1473. [24] P. Huve, T. Verrecchia, D. Bazile, C. Vauthier, P. Couvreur, Simultaneous use of size-exclusion chromatography and photon correlation spectroscopy for the characterization of poly(lactic acid) nanoparticles, Journal of Chromatography A 675 (1994) 129–139. [25] F.K. Liu, G.T. Wei, Effect of mobile-phase additives on separation of gold nanoparticles by size-exclusion chromatography, Chromatographia 59 (2004) 115–119. [26] G.T. Wei, Shape separation of nanometer gold particles by size-exclusion chromatography, Analytical Chemistry 71 (1999) 2085–2091. ˜ [27] E. Hontanón, F.E. Kruis, A differential mobility analyzer (DMA) for size selection of nanoparticles at high flow rates, Aerosol Science and Technology 43 (2009) 25–37. [28] L.F. Pease Iii, D.H. Tsai, J.A. Fagan, B.J. Bauer, R.A. Zangmeister, M.J. Tarlov, M.R. Zachariah, Length distribution of single-walled carbon nanotubes in aqueous suspension measured by electrospray differential mobility analysis, Small 5 (2009) 2894–2901. [29] L.F. Pease, D.H. Tsai, K.A. Brorson, S. Guha, M.R. Zachariah, M.J. Tarlov, Physical characterization of icosahedral virus ultra structure, stability, and integrity using electrospray differential mobility analysis, Analytical Chemistry 83 (2011) 1753–1759. [30] H. Saffari, A. Malugin, H. Ghandehari, L.F. Pease, Electrostatic deposition of nanoparticles into live cell culture using an electrospray differential mobility analyzer (ES-DMA), Journal of Aerosol Science 48 (2012) 56–62. [31] K. Takeuchi, J. Yabumoto, Y. Okada, T. Kawai, R.M. Montajir, Y. Goto, A new dual-type DMA for measuring nanoparticles emitted from combustion engines, Journal of Nanoparticle Research 7 (2005) 287–293. [32] D.H. Tsai, T.J. Cho, F.W. Delrio, J. Taurozzi, M.R. Zachariah, V.A. Hackley, Hydrodynamic fractionation of finite size gold nanoparticle clusters, Journal of the American Chemical Society 133 (2011) 8884–8887. [33] S. Anger, K. Caldwell, W. Mehnert, R. Muller, Coating of nanoparticles: analysis of adsorption using sedimentation field-flow fractionation (SdFFF), in: Proceedings of the Controlled Release Society, 1999, pp. 599–600. [34] F. Carpino, L.R. Moore, M. Zborowski, J.J. Chalmers, P.S. Williams, Analysis of magnetic nanoparticles using quadrupole magnetic field-flow fractionation, Journal of Magnetism and Magnetic Materials 293 (2005) 546–552. [35] B. Chen, J.P. Selegue, Separation and characterization of single-walled and multiwalled carbon nanotubes by using flow field-flow fractionation, Analytical Chemistry 74 (2002) 4774–4780. [36] J. Gigault, B.K. Gale, I. Le Hecho, G. Lespes, Nanoparticle characterization by cyclical electrical field-flow fractionation, Analytical Chemistry 83 (2011) 6565–6572.

23

[37] H. Abdel-Haq, E. Bossù, Capillary electrophoresis as a tool for the characterization of pentosan nanoparticles, Journal of Chromatography A 1257 (2012) 125–130. [38] S.K. Doorn, R.E. Fields Iii, H. Hu, M.A. Hamon, R.C. Haddon, J.P. Selegue, V. Majidi, High resolution capillary electrophoresis of carbon nanotubes, Journal of the American Chemical Society 124 (2002) 3169–3174. [39] A. Ibrahim, H. Ohshima, S.A. Allison, H. Cottet, Determination of effective charge of small ions, polyelectrolytes and nanoparticles by capillary electrophoresis, Journal of Chromatography A 1247 (2012) 154–164. [40] U. Schnabel, C.H. Fischer, E. Kenndler, Characterization of colloidal gold nanoparticles according to size by capillary zone electrophoresis, Journal of Microcolumn Separations 9 (1997) 529–534. [41] B. Suárez, B.M. Simonet, S. Cárdenas, M. Valcárcel, Separation of carbon nanotubes in aqueous medium by capillary electrophoresis, Journal of Chromatography A 1128 (2006) 282–289. [42] X. Xu, K.K. Caswell, E. Tucker, S. Kabisatpathy, K.L. Brodhacker, W.A. Scrivens, Size and shape separation of gold nanoparticles with preparative gel electrophoresis, Journal of Chromatography A 1167 (2007) 35–41. [43] J.C. Giddings, M.N. Myers, G.C. Lin, M. Martin, Polymer analysis and characterization by field-flow fractionation (one-phase chromatography), Journal of Chromatography A 142 (1977) 23–38. [44] J.C. Giddings, G.C. Lin, M.N. Myers, Fractionation and size distribution of water soluble polymers by flow field-flow fractionation, Journal of Liquid Chromatography 1 (1978) 1–20. [45] S.L. Brimhall, M.N. Myers, K.D. Caldwell, J.C. Giddings, Polymer Characterization by Flow Field-flow Fractionation, ACS, St Louis, MO, USA, 1984, pp. 48–52. [46] M.A. Benincasa, J.C. Giddings, Separation and molecular weight distribution of anionic and cationic water-soluble polymers by flow field-flow fractionation, Analytical Chemistry 64 (1992) 790–798. [47] J.R. Runyon, S.K.R. Williams, Composition and molecular weight analysis of styrene-acrylic copolymers using thermal field-flow fractionation, Journal of Chromatography A 1218 (2011) 6774–6779. [48] M.E. Schimpf, K. Caldwell, J.C. Giddings, Field-Flow Fractionation Handbook, Wiley-Interscience, New York, 2000, ISBN:9780471184300. [49] R. Beckett, J.C. Giddings, Entropic contribution to the retention of nonspherical particles in field-flow fractionation, Journal of Colloid and Interface Science 186 (1997) 53–59. [50] K.G. Wahlund, J.C. Giddings, Properties of an asymmetrical flow field-flow fractionation channel having one permeable wall, Analytical Chemistry 59 (1987) 1332–1339. [51] J.N. Israelachvili, Electrostatic Forces between Surfaces in Liquids, Intermolecular and Surface Forces, third ed., Academic Press, San Diego, 2011, pp. 291–340, Chapter 14. [52] H. Susanto, M. Ulbricht, Characteristics, performance and stability of polyethersulfone ultrafiltration membranes prepared by phase separation method using different macromolecular additives, Journal of Membrane Science 327 (2009) 125–135. [53] M. Amirilargani, E. Saljoughi, T. Mohammadi, Effects of Tween 80 concentration as a surfactant additive on morphology and permeability of flat sheet polyethersulfone (PES) membranes, Desalination 249 (2009) 837–842. [54] M. Amirilargani, E. Saljoughi, T. Mohammadi, Improvement of permeation performance of polyethersulfone (PES) ultrafiltration membranes via addition of tween-20, Journal of Applied Polymer Science 115 (2010) 504–513. [55] T.J. Cho, V.A. Hackley, Fractionation and characterization of gold nanoparticles in aqueous solution: asymmetric-flow field flow fractionation with MALS, DLS, and UV–Vis detection, Analytical and Bioanalytical Chemistry 398 (2010) 2003–2018. [56] J. Gigault, B. Grassl, G. Lespes, Multi-wall carbon nanotube aqueous dispersion monitoring by using A4F-UV-MALS, Analytical and Bioanalytical Chemistry 401 (2011) 3345–3353. [57] J. Gigault, T.J. Cho, R.I. MacCuspie, V.A. Hackley, Gold nanorod separation and characterization by asymmetric-flow field flow fractionation with UV–Vis detection, Analytical and Bioanalytical Chemistry (2012) 1–12. [58] J. Gigault, B. Grassl, G. Lespes, A new analytical approach based on asymmetrical flow field-flow fractionation coupled to ultraviolet spectrometry and light scattering detection for SWCNT aqueous dispersion studies, Analyst 137 (2012) 917–923. [59] J.S. Taurozzi, V.A. Hackley, M.R. Wiesner, Reporting Guidelines for the Preparation of Aqueous Nanoparticle Dispersions from Dry Materials, 2012. [60] J.S. Taurozzi, V.A. Hackley, M.R. Wiesner, Preparation of Nanoparticle Dispersions from Powdered Material Using Ultrasonic Disruption, NIST, 2012, pp. 1200–1202, NIST Special Publication. [61] J.S. Taurozzi, V.A. Hackley, M.R. Wiesner, Preparation of a Nanoscale TiO2 Aqueous Dispersion for Toxicological or Environmental Testing, NIST, 2012, pp. 1200–1203, NIST Special Publication. [62] J.S. Taurozzi, V.A. Hackley, M.R. Wiesner, Preparation of Nanoscale TiO2 Dispersions in an Environmental Matrix for Eco-Toxicological Assessment, NIST, 2012, pp. 1200–1205, NIST Special Publication. [63] OECD, Working party on manufactured nanomaterials: list of manufactured nanomaterials and list of endpoints for phase one of the OECD testing programme, in: Safety of Manufactured Nanomaterials, Organisation for Economic Co-operation and Development, Paris, 2008. [64] J. Gigault, V.A. Hackley, Observation of size-independent effects in nanoparticle retention behavior during asymmetric-flow field-flow fractionation, Analytical and Bioanalytical Chemistry 405 (2013) 6251–6258.

24


[65] J.R. Runyon, A. Goering, K.T. Yong, S.K.R. Williams, Preparation of narrow dispersity gold nanorods by asymmetrical flow field-flow fractionation and investigation of surface plasmon resonance, Analytical Chemistry 85 (2013) 940–948. [66] S. Dubascoux, F. Von Der Kammer, I. Le Hécho, M.P. Gautier, G. Lespes, Optimisation of asymmetrical flow field flow fractionation for environmental nanoparticles separation, Journal of Chromatography A 1206 (2008) 160–165. [67] W. Burchard, Static and dynamic light scattering from branched polymers and biopolymers, in: Light Scattering from Polymers, Springer, Berlin Heidelberg, 1983, pp. 1–124. [68] H.C. van de Hulst, Light Scattering by Small Particles, Dover Publication, New York, NY, 1981. [69] J.M. Pettibone, J. Gigault, V.A. Hackley, Discriminating the states of matter in metallic nanoparticle transformations: what are we missing? ACS Nano 7 (2013) 2491–2499. [70] S. Kumar, M.D. Bolan, T.P. Bigioni, Glutathione-stabilized magic-number silver cluster compounds, Journal of the American Chemical Society 132 (2010) 13141–13143. [71] J.W. Hudgens, J.M. Pettibone, T.P. Senftle, R.N. Bratton, Reaction mechanism governing formation of 1,3-bis(diphenylphosphino)propane-protected gold anoclusters, Inorganic Chemistry 50 (2011) 10178–10189.

[72] K.P. Hall, D.M.P. Mingos, Homonuclear and heteronuclear cluster compounds of gold, Progress in Inorganic Chemistry 32 (1984) 237–325. [73] A. Desireddy, S. Kumar, J. Guo, M.D. Bolan, W.P. Griffith, T.P. Bigioni, Temporal stability of magic-number metal clusters: beyond the shell closing model, Nanoscale 5 (2013) 2036–2044. [74] I. Diez, R.H.A. Ras, Fluorescent silver nanoclusters, Nanoscale 3 (2011) 1963–1970. [75] Y. Negishi, K. Nobusada, T. Tsukuda, Glutathione-protected gold clusters revisited: bridging the gap between gold(I)-thiolate complexes and thiolateprotected gold nanocrystals, Journal of the American Chemical Society 127 (2005) 5261–5270. [76] M.W. Heaven, A. Dass, P.S. White, K.M. Holt, R.W. Murray, Crystal structure of the gold nanoparticle [N(C8 H17 )4 ][Au25 (SCH2 CH2 Ph)18 ], Journal of the American Chemical Society 130 (2008) 3754–3755. [77] M. Zhu, C.M. Aikens, F.J. Hollander, G.C. Schatz, R. Jin, Correlating the crystal structure of a thiol-protected Au25 cluster and optical properties, Journal of the American Chemical Society 130 (2008), 5883-5585. [78] K.D. Jensen, S.K.R. Williams, J.C. Giddings, High-speed particle separation and steric inversion in thin flow field-flow fractionation channels, Journal of Chromatography A 746 (1996) 137–145.

Separation and characterization of poly(tetrafluoroethylene) latex particles by asymmetric flow field flow fractionation with light-scattering detection.

Asymmetric flow field-flow fractionation in the field of nanomedicine.

Characterization of cationic polymers by asymmetric flow field-flow fractionation and multi-angle light scattering-A comparison with traditional techniques.

Flow-field-flow fractionation: a versatile new separation method.

Feasibility of asymmetric flow field-flow fractionation coupled to ICP-MS for the characterization of wear metal particles and metalloproteins in biofluids from hip replacement patients.

A critical evaluation of an asymmetrical flow field-flow fractionation system for colloidal size characterization of natural organic matter.

Investigation of red blood cell fractionation by gravitational field-flow fractionation.

Analysis of plant ribosomes with asymmetric flow field-flow fractionation.

Field flow fractionation in biomedical analysis.

Feasibility study of red blood cell debulking by magnetic field-flow fractionation with step-programmed flow.

Separation and quantification of monoclonal-antibody aggregates by hollow-fiber-flow field-flow fractionation.

Flow field-flow fractionation: new method for separating, purifying, and characterizing the diffusivity of viruses.

Particle characterization in centrifugal fields. Comparison between ultracentrifugation and sedimentation field-flow fractionation.

Theoretical analysis of the local orientation effect and the lift-hyperlayer mode of rodlike particles in field-flow fractionation.

Elemental ratios for characterization of quantum-dots populations in complex mixtures by asymmetrical flow field-flow fractionation on-line coupled to fluorescence and inductively coupled plasma mass spectrometry.

Simple functionalization strategies for enhancing nanoparticle separation and recovery with asymmetric flow field flow fractionation.

Selection of motile nonchemotactic mutants of Escherichia coli by field-flow fractionation.

Particle size distribution of concentrated perfluorocarbon emulsions by sedimentation field flow fractionation.

Tacticity separation of poly(methyl methacrylate) by multidetector thermal field-flow fractionation.

Analysis of Creutzfeldt-Jakob disease infectious fractions by gel permeation chromatography and sedimentation field flow fractionation.

Characterisation of cationic potato starch by asymmetrical flow field-flow fractionation. Influence of ionic strength and degree of substitution.

Factors affecting measurement of channel thickness in asymmetrical flow field-flow fractionation.

Asymmetric flow field-flow fractionation of manufactured silver nanoparticles spiked into soil solution.

Analyzing the influence of PEG molecular weight on the separation of PEGylated gold nanoparticles by asymmetric-flow field-flow fractionation.