John W. Olesik

	Single-particle inductively coupled plasma–mass spectrometry (spICP-MS) is becoming widely used to measure the number of nanoparticles (or other sub-micrometer–sized particles) per mL with a particular elemental chemical composition and the average particle size (diameter) or particle size distribution (1,2). As the use of engineered nanoparticles has grown explosively, the need for rapid measurement of engineered nanoparticle size and number concentration has become critical. The Nanodatabase (3) lists 3901 products that contain nanomaterials, and more than 460 journal articles using spICP-MS were published between 2017 and 2020 (based on a CAS SciFinder literature search).	 

 are defined as particles with diameters between 1 and 100 nm. However, many engineered particles have diameters up to 1000 nm and can be measured by spICP-MS, thus the use of the phrase 

Conversion of sub-micrometer–sized particles to ICP-MS signals.

 When sub-micrometer–sized particles, suspended in a liquid, are nebulized into a liquid aerosol and introduced into an ICP, first the liquid evaporates, then the remaining particle is vaporized, atomized, and ionized. A spherical cloud of ions is produced in the plasma from each individual particle (Figure 1). If the number of particles entering the plasma per second is small enough, the ion clouds produced from different particles do not overlap in the plasma. A peak in the ICP-MS signal versus time is produced from each particle. The particle suspension must have a small enough particle number concentration so that it is unlikely that ion clouds produced from two different particles will overlap in the ICP. 

	As the plasma gas flows from the ICP through the interface and into the mass spectrometer, each ion cloud is detected as a peak (0.2 to ~1 ms wide) (Figure 1b). Millions of aerosol droplets enter the plasma per second, each containing the elements that are dissolved; ion clouds produced from dissolved elements in different droplets overlap extensively, resulting in a “constant” signal (Figure 1b). Therefore, the sharp peaks produced from individual particles can be distinguished from the more constant signal produced from the same element in solution if the signals from the sharp peaks are larger than the variations in signal generated from the element in solution.

	The total net ICP-MS signal (counts) at a particular mass/charge across the peak is directly proportional to the amount (in fg) of the measured element in the original, individual particle. The number of detected peaks (“particle events”) (for example, 10 in Figure 1c) is directly proportional to the number of particles/mL in the original suspension. The number of detected peaks also depends on the volume of sample aerosol (typically 20 to 100 μL) that enters the plasma during the duration of continuous ICP-MS signal measurement (typically ~2 to 5 min).

	Particles have been detected using peaks in an atomic spectrometry signal versus time as early as 1968 (4). In 2003, Degueldre and coworkers described ICP-MS detection of individual nanoparticles suspended in liquid for the first time (5).

	However, spICP-MS should still be considered an emerging technique (1,6) and we have much to learn about variables that can affect the accuracy of measurements of particle size, elemental composition, and number per mL. These variables include the stability of stored particle suspensions, sample preparation, calibration, chemical matrix effects, high backgrounds resulting from elements in solution or ions that cause spectral overlaps, detection limits (in terms of particle size and the minimum detectable amount of analyte elements), and the limited dynamic range of the number of particles/mL and particle size. 

Amount of analyte element in each particle.

 The technique of spICP-MS directly measures signal from each ion cloud that is proportional to the amount (in fg) of the element measured in each sub-micrometer–sized particle if: (a) the particle is completely vaporized, (b) the fraction of analyte ionized is the same for the sample particles and the calibration solution or calibration particle suspension, and (c) the fraction of ions produced in the ICP that reach the detector is the same for the sample particles and the calibration solution or particles.

	The maximum size particle that is completely vaporized and converted to elemental ions prior to entering the ICP-MS interface depends on the chemical composition of the particle and the time that the particle spends in the hot plasma (typically ~1 to 3 ms, depending mainly on the center gas flow rate, injector diameter, and plasma power). While increasing the time that the particles spend in the hot plasma allows larger particles to be vaporized, it also decreases sensitivity (signal/fg of analyte element) because there is more time for diffusion of ions (broadening the ion cloud) (7).

	Typically, particles with diameters less than about 1000 nm are completely vaporized in the ICP (7,8). Fluctuations in the time particles spend in the hot plasma (and therefore the extent of diffusion of ions in the ion cloud), in part because of fluctuations in the plasma gas velocity, can result in variations in the relationship between signal intensity (counts) and amount (fg) of analyte element in the particle. 

With spICP-MS, we directly measure the number of particles that enter the ICP per second if each particle produces a detectable ion cloud. As the detection limit is approached, the percentage of particles that produces a detectable ion cloud decreases. To quantitatively relate the number of “particle events” measured to the number of particles/mL in the original suspension, two different calibration methods can be used (1,9): (a) a particle suspension with a known number of particles/mL can be measured, or (b) a solution standard with a known concentration of the analyte element and nanoparticles with a known amount (fg) of analyte are separately measured. For (b) the ratio:	 

	 determines the fraction of nanoparticles delivered to the nebulizer that enter the plasma and are detected. In order to know the amount of analyte element in the nanoparticle the elemental composition of the particle, the nanoparticle size (typically diameter), and the density (g/cm

) must be known. Sometimes, the two different calibration techniques lead to different results. There is some evidence that the sensitivity-ratio–based calibration is more stable over time; that suggests that the number of nanoparticles per volume in a suspension may not be as stable over time.

 Single-particle ICP-MS can be used to measure sub-micrometer particle diameters of spherical particles because spICP-MS determines the amount of analyte element (fg) in each particle. To convert amount of analyte to particle size (for example, diameter) the shape of the particle (for example, spherical), the chemical composition of the particle and the density (g/cm

Application of spICP-MS to engineered sub-micrometer–sized particles.

 Many widely used engineered nanoparticles are elemental (for example, Au or Ag) or oxides (ZnO, TiO

). In addition, many engineered nanoparticles contain elements that are not abundant in nature (for example, Au or Ag). The vast majority of publications to date have focused on measurement of these engineered nanoparticles (or sub-micrometer–sized particles), typically in samples that would not be expected to contain these elements in high abundance in natural sub-micrometer–sized particles. 

	ICP-MS with a quadrupole mass spectrometer can monitor just one element continuously. This provides element-selective detection of sub-micrometer–sized particles such as engineered nanoparticles. Element-selective detection allows specific particles (such as Au, Ag, ZnO, and TiO

 engineered nanoparticles) to be measured independently from many other particles that do not contain detectable amounts of the same element as the particle of interest (for example, carbon-based particles like soot). This is one of the advantages of spICP-MS compared to other particle-measurement techniques, such as dynamic light scattering or a Coulter counter, that are not element-specific. 

Application of spICP-MS to mixtures of sub-micrometer–sized particles including natural particles

. Natural particles and anthropogenic particles can have a wide variety of chemical compositions, including combinations of different elements. Some engineered particles have layers or coatings of different materials. Using spICP-MS with a quadrupole MS that measures one element at a time, it would be impossible to distinguish small single element or oxide particles from larger particles that contained a small portion (but same number of fg) of the same element.

	Single-particle ICP-MS with a time-of-flight mass (TOF) spectrometer can provide a complete elemental mass spectrum (for the elements detectable by ICP-MS) for each sub-micrometer–sized particle (for example, Figure 2). By calibrating the signal/fg, the amount of each detectable element in each individual particle can be measured.

	Single-particle ICP-TOF–MS can also measure the number of particles per mL (all particles with detectable elements or for each type of particle, based on its chemical composition). Single-particle ICP-TOF–MS does have some limitations compared to currently sold ICP-quadrupole-MS instruments, including poorer abundance sensitivity, poorer sensitivity, and limitations in use of gases in collision–reaction cells that may also react with some elemental ions.

	Single-particle ICP-MS is uniquely capable of rapidly measuring the number of sub-micrometer–sized particles/mL with either element-selective detection (using quadrupole or sector field mass spectrometers) or elemental chemical composition of individual sub-micrometer–sized particles (using time-of-flight mass spectrometers). At this point, far more care and expertise are needed to obtain accurate results with spICP-MS than for solution analysis with ICP-MS. Detection limits in the best case scenario (single element nanoparticles in suspensions that contain little if any of the element of interest in solution and with no spectral overlaps) are typically around 6 to 10 nm in diameter. The analyte element mass depends on the cube of the particle diameter; to reduce the diameter detection limit by 2x, the detection limit in the amount of analyte element (fg) would need to be reduced by 8x. 

	Thanks to Paolo Gabrielli (Ohio State University Byrd Polar Climate and Research Center) for comments on this manuscript. Thanks to Luke Monroe, Garret Bland, and Ryan Sullivan for the mass spectrum shown in Figure 2, acquired using the ICP-TOF–MS instrument at Carnegie Mellon University. The mass spectrum was acquired from a particle entrapped in an ice core from Mt. Ortles; the sample was prepared for analysis by Paolo Gabrielli and Aja Ellis. The Mt. Ortles project was supported from NSF awards no. 1060115 and no. 1461422 and from the Ripartizione Opere idrauliche e Ripartizione Foreste of the autonomous province of Bolzano and the Stelvio National Park as well as The Ohio State University, the OSU Institute for Materials Research, The Byrd Polar Research Center, and the Ohio Third Frontier Research Scholar program.	 

M. D. Montaño, J. W. Olesik, A. G. Barber, K. Challis, and J. F. Ranville, 

S. Kaushik, S. R. Djiwanti, and E. Skotti, in R. Prasad, Ed.,

 Microbial Nanobionics: Volume 2, Basic Research and Applications

 (Springer International Publishing, Cham, 2019), pp. 13-33.

H. E. Pace, N. J. Rogers, C. Jarolimek, V. A. Coleman, C. P. Higgins, and J. F. Ranville, 

 is a Research Scientist, an Adjunct Associate Professor, and the Director of the Trace Element Research Laboratory in the School of Earth Sciences at The Ohio State University, in Columbus, Ohio. Direct correspondence to 

Articles

Single-particle inductively coupled plasma–mass spectrometry (spICP-MS) is becoming widely used to measure the number of nanoparticles (or other submicrometer– sized particles) per mL with a particular elemental chemical composition and the average particle size (diameter) or particle size distribution.

Single Particle ICP-MS: From Engineered Nanoparticles to Natural Nanoparticles

	Inductively coupled plasma–optical emission spectrometry (ICP-OES) has been one of the most widely used atomic spectroscopy techniques for the last 40 years (1). This paper assesses its current analytical performance, shortcomings, developments over the last 35 years, and its future prospects.	 

	ICP-OES provides robust, rapid, multi-element analysis of solutions (including digested solid samples). Element “lines” (peaks at characteristic wavelengths) result from bound outer shell electron transitions from an excited state to a lower energy level. Detection limits are a part per billion (ng/mL) or below for most elements. Elements that emit brightly, including Be, Mg, Ca, Sr, and Ba, have detection limits of tens of parts per trillion (pg/mL) or below. A desolvation system following a heated spray chamber or ultrasonic nebulizer can further improve detection limits by ~10x. Once samples are prepared and the instrument is calibrated, the analysis time per sample can be less than 1 min. 

	The biggest challenger to ICP-OES currently is ICP–mass spectrometry (ICP-MS). ICP-MS has lower (up to 1000x) detection limits but has disadvantages compared to ICP-OES for analyses where ICP-OES detection limits are sufficient. ICP-MS costs more (instrument purchase, time and cost of maintenance), is limited to samples with 10x lower total dissolved solids, and can suffer from severe matrix effects.	 

	The main limitations of ICP-OES are continuum background emission, higher detection limits than ICP-MS, spectral overlaps, and matrix effects. 

Continuum background emission occurs at all wavelengths because it involves unbound electron transitions (ion-electron recombination or Bremsstralung). The sensitivity (signal/concentration) and background fluctuations determine detection limits. Near the detection limit the average, sample dependent, background emission intensity can be about 30x to 100x the net analyte line intensity.	 

Elements including As, P, Rb, Se, and S have detection limits of a few part per billion (ppb, ng/mL). Detection limits for F, Cl, and Br are typically in the hundreds of ppb (ng/mL) or more unless the spectrometer can measure lines below about 150 nm. The noble gas elements, O, N, and H, are typically not measured because they suffer from poor sensitivities or intense blank signals, or both, because of entrained air.	 

 Elements may emit a few wavelengths (such as K and Na) or thousands of wavelengths (such as rare earth elements and Th and U). Spectral overlaps can occur and, if not accounted for, accuracy can be degraded. However, because emission peaks are narrow (~0.005 nm or less) the emission spectrum consists of more than 128,000 different, resolvable wavelengths over the 160 to 800 nm range. A typical ICP-MS instrument (with a quadrupole mass spectrometer) has a bandpass of about 0.6 amu, resulting in about 386 resolvable mass/charge values from 6 to 238.	 

 Elements in the sample at high concentrations (≥500 ppm [µg/mL]) can affect sensitivity. Ionization and excitation in the ICP are mainly a result of collisions between atoms or ions and fast-moving electrons. Therefore, if components in the sample cause a change in the electron temperature or concentration in the plasma, the sensitivity can change.	 

 ICP-OES instruments with an autosampler allow unattended analyses. However, the instrument does not inform the operator when a matrix effect or spectral overlap has likely occurred for a particular sample, nor automatically correct for them.	 

More and more, ICP-OES buyers want instruments that require less expertise to operate. Expertise is required to identify and deal with the potential problems (spectral overlaps, matrix effects) and to develop methods (including selection of wavelengths, and internal standards if necessary). Lower cost is also desired. Typically ~16 L/min of Ar keeps the plasma torch from melting and sustains the plasma.	 

	There have been several important advances in ICP-OES over the last 35 years (Table I) and development continues (2).	 

 The most significant advance in ICP-OES in the last 35 years was the use of solid-state imaging detectors, typically with echelle spectrometers (3). The use of these detectors provided rapid, simultaneous multi-element analysis, flexibility in choosing analyte wavelengths, and simultaneous detection of analyte peaks and the nearby background emission. 

 Detection limits have improved ~10x since 1985 as a result of a combination of factors. End-on (axial) viewing of emission collects light from a larger region of the plasma; sensitivity and the relative standard deviation of the background improve. Simultaneous detection allows the correlated fluctuations in the background to be removed.	 

 Solid-state generators have replaced generators with a vacuum tube power amplifier, which degraded over time. Free running solid-state generators allow impedance matching without large, tunable capacitor-based impedance matchers.	 

 The size and cost of ICP-OES instruments have decreased. Some current ICP-OES instruments require ~8–9 L/min Ar rather than 15–20 L/min. Recently developed torch designs may allow much further reduction in Ar gas consumption.	 

Sample introduction systems with faster washout times, in combination with simultaneous detection, can provide sample measurement times of less than 1 min.	 

Multiple linear regression improved correction for spectral overlaps. Pure, single element spectra must be acquired for each analyte and potential spectral overlap on the analyte peak. Multiple linear regression determines the constants (a, b, and c in the equation below) that each of the single-element spectra and the background (blank) spectrum must be multiplied by so that their sum best fits the experimentally measured spectrum from the sample (a is proportional to analyte concentration):	 

	a * analyte element spectrum + b * overlap element spectrum + c* blank spectrum = Sample spectrum                      [1]

 In some cases, the precision and accuracy (that is, the concentration uncertainty) provided by ICP-OES may be insufficient. Uncertainties in measured concentrations (4) are typically much larger (5–25%) than indicated by the precision (~1%). By using closely matched standards, internal standardization, enough replicate measurements, and accounting for drift, it is possible to obtain relative concentration uncertainties better than 0.2% (5).	 

	No optical emission sources have been able to unseat ICPs by providing better analytical performance for solution samples. Most of the current challengers have advantages in cost, portability, or the ability to measure solid samples directly, but also suffer from disadvantages. 

	Laser induced breakdown spectroscopy (LIBS) uses a pulsed laser to vaporize, atomize, and excite the sample (6). LIBS can directly measure solid samples. LIBS can be portable. Detection limits in solids range from <1 ppm to >100 ppm (comparable to detection limits of <0.010 ppm to >1 ppm in a 1% wt/wt solution of a digested solid sample). LIBS sometimes suffers from matrix effects, insufficient reproducibility, and nonlinear calibration, although improvements are being made. 

	Nitrogen microwave plasma-optical emission spectroscopy (MP-OES) has the advantage of not requiring Ar gas. The nitrogen plasma gas can be obtained from air using a nitrogen generator. For most elements, particularly those with high excitation energies (such as As, Se, Cd, P, Sb, Se, and Zn) MP-OES detection limits are poorer than ICP-OES can provide.

	Solution cathode glow discharges and liquid sampling glow discharges have the advantage of requiring far less power (<100 W) than an Ar ICP (~1100 W), making portable instruments possible and cost much lower. Detection limits for many elements are not yet competitive with ICP-OES. However, detection limits are being improved, in some cases dramatically (7).

	Is there an atomic emission source that could provide higher sensitivity (and at the same time lower continuum background emission) to provide detection limits that are at least 10x (100x?) better than current ICP-OES detection limits at similar or lower cost? Not yet.	 

	If we will have self-driving cars, why not intelligent (“self-driving”) ICP-OES instruments that can identify and overcome potential errors without an expert instrument operator? 

	There are many potentially useful diagnostics to identify when spectral overlaps and matrix effects are likely occurring for a particular sample. Ion/atom emission intensity ratios and Ar emission intensities indicate plasma temperature. Comparing measured concentrations from three or four lines instead of one could be used to flag spectral overlaps and matrix effects. Semiquantitative analysis with little, if any, additional analysis time, could determine if an element (or elements) in a sample are in high enough concentration to cause matrix effects or spectral overlaps. Our laboratory often uses these diagnostics but with manual assessment. Spatial emission profiles or gradient dilution, or both could also be used (8). While many potential inputs for intelligent ICP-OES instruments have been identified, developing an intelligent instrument and validating its performance is still a huge challenge. 

	2.     G.L. Donati, R.S. Amais, and C.B. Williams, 

	3.     T.W. Barnard, M.I. Crockett, J.C. Ivaldi, and P.L. Lundberg, 

	5.     M.R. Winchester, T.A. Butler, and G.C. Turk, 

	6.     D. Fernandes Andrade, E.R. Pereira-Filho, and D. Amarasiriwardena, 

	7.     C. Yang, G.C.Y. Chan, D. He, Z. Liu, Q. Deng, H. Zheng, S. Hu, and Z. Zhu, 

	8.     Y. Cheung, A.J. Schwartz, G.C.Y. Chan, and G.M. Hieftje, 

In celebration of Spectroscopy’s 35th Anniversary, leading experts discuss important issues and challenges in analytical spectroscopy.

John W. Olesik

Articles

Single Particle ICP-MS: From Engineered Nanoparticles to Natural Nanoparticles

ICP-OES Capabilities, Developments, Limitations, and Any Potential Challengers?