1. INTRODUCTION: Historical Perspectives of Studies of Interstellar Grain Compositions and Sizes

When the existence of solid particles in interstellar space was first convincingly demonstrated by Trumpler (1930) through the discovery of color excess between the photographic (with an effective wavelength _eff 4300 Å) and visual (_eff 5500 Å) magnitudes, the grains were estimated to be 10^-19 g, corresponding to radii of 20 Å (Trumpler 1930). By the end of the 1930s, a ^-1 extinction law (i.e., the interstellar extinction varied approximately inversely with wavelength ) in the wavelength range 1-3 µm^-1 had been well established (Hall 1937; Stebbins, Huffer, & Whitford 1939). At that time, metallic grains with a dominant size of ~ 0.05 µm (Schalén 1936) or a power-law size distribution dn(a) / da ~ a^-3.6 in the size range 80 Å a 1 cm (Greenstein 1938) were proposed to explain the ^-1 extinction law, partly because meteoritic particles (predominantly metallic) and interstellar grains were then thought to have the same origin.

In view of the spatial correlation between gas concentration and dust extinction, Lindblad (1935) suggested that interstellar grains were formed by condensation from the interstellar gas through random accretion of gas atoms. However, it was found later that in typical interstellar conditions, the Lindblad condensation theory would result in a complete disappearance of all condensable gases and the grains would grow to sizes (~ 10 µm) well beyond those which would account for the interstellar extinction (see Oort & van de Hulst 1946 and references therein). By introducing a grain destruction process caused by grain-grain collisions as a consequence of interstellar cloud encounters, Oort & van de Hulst (1946) further developed the interstellar condensation idea and led to the "dirty ice" model consisting of saturated molecules such as H₂O, CH₄, and NH₃ with an equilibrium size distribution which could be roughly approximated by a functional form dn(a) / da ~ exp[ -5 (a / 0.5 µm)³] and an average size of ~ 0.15 µm.

The discovery of interstellar polarization (Hall 1949; Hiltner 1949) cast doubts on the "dirty ice" model since ice grains are very inefficient polarizers. This stimulated Cayrel & Schatzman (1954) to consider graphite grains as an interstellar dust component since aligned graphite particles would be a very efficient interstellar polarizer because of their strong optical anisotropy. The "dirty ice" model was also challenged by nondetection of the 3.1 µm feature of H₂O ice outside of molecular clouds (Danielson, Woolf, & Gaustad 1965; Knacke, Cudaback, & Gaustad 1969). This gave the late J. Mayo Greenberg (1922-2001) the incentive to perform the early experiments on the ultraviolet (UV) photoprocessing of low temperature mixtures of volatile molecules simulating the "original" dirty ice grains (Greenberg et al. 1972) to understand how and why the predicted H₂O was not clearly present. From such experiments was predicted a new component of interstellar dust in the form of complex organics known as "organic refractories", containing a mixture of aliphatic and aromatic carbonaceous molecules (Greenberg et al. 2000).

As an alternative to the interstellar condensation process, Hoyle & Wickramasinghe (1962) proposed that graphite particles with radii a few times 0.01 µm could form in the atmospheres of cool N-type carbon stars, and subsequently be ejected into interstellar space by the stellar radiation pressure, and be responsible for (part of) the interstellar extinction and polarization. The graphite proposal further gained its strength from the prominent 2175 Å interstellar extinction hump detected by the Aerobee rocket observation (Stecher 1965), which was generally attributed to small graphitic grains (Stecher & Donn 1965), although its exact nature still remains unidentified (Draine 1989, 2003a).

Similarly, Kamijo (1963) first proposed that SiO₂ whose size is of the order of ~ 20 Å, condensed in the atmospheres of cool M-type stars and expelled into interstellar space, could serve as condensation nuclei for the formation of "dirty ices". It was later shown by Gilman (1969) that grains around oxygen-rich cool giants are mainly silicates such as Al₂SiO₃ and Mg₂SiO₄. Interstellar silicates were first detected in emission in the Trapezium region of the Orion Nebula (Stein & Gillett 1969); in absorption toward the Galactic Center (Hackwell, Gehrz, & Woolf 1970), and toward the Becklin-Neugebauer object and the Kleinmann-Low Nebula (Gillett & Forrest 1973). Silicates are now known to be an ubiquitous interstellar dust component (see Section 1 in Li & Draine 2001a for a review).

In the 1960s and early 1970s, the extension of the extinction curve toward the middle and far UV ( 3 µm^-1) was made possible by rocket and satellite observations, including the rocket-based photoelectric photometry at = 2600 Å and 2200 Å (Boggess & Borgman 1964); the Aerobee Rocket spectrophotometry at 1200 Å 3000 Å (Stecher 1965); the Orbiting Astronomical Satellite (OAO-2) spectrophotometry at 1100 Å 3600 Å (Bless & Savage 1972); and the Copernicus Satellite spectrophotometry at 1000 Å 1200 Å (York et al. 1973). By 1973, the interstellar extinction curve had been determined over the whole wavelength range from 0.2 µm^-1 to 10 µm^-1. The fact that the extinction continues to increase in the far UV (e.g., see York et al. 1973) implies that no single grain type with either a single size or a continuous size distribution could account for the observed optical to far-UV interstellar extinction (Greenberg 1973). This led to the abandonment of any one-component grain models and stimulated the emergence of various kinds of models consisting of multiple dust constituents, including silicate, silicon carbide, iron, iron oxide, graphite, dirty ice, solid H₂, etc.

In 1956 Platt first suggested that very small grains or large molecules of less than 10 Å in radius grown by random accretion from the interstellar gas could be responsible for the observed interstellar extinction and polarization. Platt (1956) postulated these "Platt" particles as quantum-mechanical particles containing many ions and free radicals with unfilled electronic energy bands. Donn (1968) further proposed that polycyclic aromatic hydrocarbon-like "Platt particles" may be responsible for the UV interstellar extinction.

Greenberg (1968) first pointed out that very small grains with a heat content smaller than or comparable to the energy of a single stellar photon, cannot be characterized by a steady-state temperature but rather are subject to substantial temporal fluctuations in temperature. Under interstellar conditions, grains larger than a few hundred angstroms in radius would attain an equilibrium temperature in the range of 15-25 K (Greenberg 1968, 1971; Mathis, Mezger, & Panagia 1983; Draine & Lee 1984; Li & Draine 2001b). This temperature was long thought to be ~ 3 K before van de Hulst (1949) made the first realistic estimation of 10-20 K for both metallic and dielectric grains.

In a study of the scattering properties of interstellar dust (albedo and phase function) determined from the OAO-2 observations at 1500 Å 4250 Å of the diffuse Galactic light (Witt & Lillie 1973), Witt (1973) first explicitly suggested a bi-modal size distribution for interstellar grains: large grains with radii 2500 Å would provide extinction in the visible region including scattering which is strongly forward directed, and small particles with radii 250 Å would dominate the UV region and contribute nearly isotopic scattering.

The presence of "Platt" particles in a dust cloud near M17 was first suggested by Andriesse (1978) based on an analysis of its IR spectral energy distribution which was shown to be characterized by a combination of widely different color temperatures (e.g., the far-IR and near-IR color temperatures are ~ 38 K and ~ 150 K, respectively) and that the 8-20 µm emission spectrum is similar over a distance of ~ 2' through the source. Andriesse (1978) argued that the color temperature differences cannot be easily ascribed to a spatial variation of the dust temperature in this cloud, because the cloud profile appears to be independent of the wavelength; instead, "Platt" particles with sizes of ~ 10 Å exhibiting temperature fluctuations up to ~ 150 K plus large grains with an equilibrium temperature of ~ 36 K can explain the observed IR spectrum of M17.

In a near-IR (1.25-4.8 µm) photometric and spectrophotometric study of three visual reflection nebulae NGC 7023, 2023, and 2068, Sellgren, Werner, & Dinerstein (1983) discovered that each nebula has extended near-IR emission consisting of emission features at 3.3 and 3.4 µm and a smooth continuum which can be described by a greybody with a color temperature ~ 1000 K. They found that the emission spectrum (i.e., the color temperature for the continuum and the 3.3 µm feature) shows very little variation from source to source and within a given source with distance from the central star. Sellgren (1984) argued that this emission could not be explained by thermal emission from dust in radiative equilibrium with the central star since otherwise the color temperature of this emission should fall off rapidly with distance from the illuminating star; instead, she proposed that this emission is emitted by ultrasmall grains, ~ 10 Å in radius, which undergo large excursions in temperature due to stochastic heating by single stellar photons.

The presence of a population of ultrasmall grains in the diffuse ISM was explicitly indicated by the 12 µm and 25 µm "cirrus" emission detected by the Infrared Astronomical Satellite (IRAS) (Boulanger & Pérault 1988), which is far in excess (by several orders of magnitude) of what would be expected from large grains of 15-25 K in thermal equilibrium with the general interstellar radiation field. Subsequent measurements by the Diffuse Infrared Background Experiment (DIRBE) instrument on the Cosmic Background Explorer (COBE) satellite confirmed this and detected additional broadband emission at 3.5 µm and 4.9 µm (Arendt et al. 1998).

More recently, spectrometers aboard the Infrared Telescope in Space (IRTS) (Onaka et al. 1996; Tanaka et al. 1996) and the Infrared Space Observatory (ISO) (Mattila et al. 1996) have shown that the diffuse ISM radiates strongly in emission features at 3.3, 6.2, 7.7, 8.6, and 11.3 µm. These emission features, first seen in the spectra of the planetary nebulae NGC 7027 and BD+30°3639 (Gillett, Forrest, & Merrill 1973), have been observed in a wide range of astronomical environments including planetary nebulae, protoplanetary nebulae, reflection nebulae, HII regions, circumstellar envelopes, and external galaxies (see Tielens et al. 2000 for a review). Often referred to as "unidentified infrared" (UIR) bands, these emission features are now usually attributed to free-flying PAH molecules which are vibrationally excited upon absorption of a single UV/visible photon (Léger & Puget 1984; Allamandola, Tielens, & Barker 1985; Allamandola, Hudgins, & Sandford 1999; Draine & Li 2001; Li & Draine 2001b) although other carriers have also been proposed such as carbonaceous grains with a partly aromatic character (e.g., HAC [Duley & Williams 1981, Duley, Jones, & Williams 1989], quenched carbonaceous composite [QCC; Sakata et al. 1992], coal [Papoular et al. 1996], fullerenes [C₆₀; Webster 1993]), and surface-graphitized nanodiamonds (Jones & d'Hendecourt 2000).

Since the late 1970s, various modern interstellar dust models have been developed (see Li & Greenberg 2003, Draine 2004, Dwek et al. 2004 for recent reviews). In general, these models fall into three broad categories:

• The silicate-graphite model - this model consists of two (physically) separate dust components each with a power-law size distribution dn(a) / da ~ a^-3.5 in the size range 50 Å a 0.25 µm (Mathis, Rumpl, & Nordsieck 1977 [hereafter MRN]; Draine & Lee 1984). Modifications to this model were later made by Draine & Anderson (1985), Weiland et al. (1986), Sorrell (1990), Siebenmorgen & Krügel (1992), Rowan-Robinson (1992), Kim, Martin, & Hendry (1994), Dwek et al. (1997), Clayton et al. (2003), and Zubko, Dwek, & Arendt (2003) by including new dust components (e.g., amorphous carbon, carbonaceous organic refractory, and PAHs) and adjusting dust sizes (e.g., deriving dust size distributions using the "Maximum Entropy Method" or the "Method of Regularization" rather than presuming a certain functional form).

  Recent developments were made by Draine and his coworkers (Li & Draine 2001b, 2002a, 2002b; Weingartner & Draine 2001a) who have extended the silicate-graphite grain model to explicitly include a PAH component as the small-size end of the carbonaceous grain population. The PAH component, containing 45 × 10^-6 of C relative to H, is represented by a log-normal size distribution dn(a) / d ln a ~ exp{ - [ln(a / a₀)]² / (2²)} with a₀ 3.5 Å and 0.4 for a 3.5 Å (see Li & Draine 2001b). Note that with the PAH component added, the size distribution of the larger grains is changed from the simple MRN power-law (see Weingartner & Draine 2001a; Zubko et al. 2003).

• The silicate core-carbonaceous mantle model - originally proposed by Greenberg (1978) in the context of a complex cyclic evolutionary scenario (see Greenberg & Li 1999), this model consists of larger silicate grains coated by a layer of carbonaceous organic refractory material, produced by UV photolysis of ice mixtures which attempts to simulate the physical and chemical processes occurring in interstellar space. The most recent development of this model was that of Li & Greenberg (1997), who modeled the core-mantle grains as 2:1 (the ratio of the length to the diameter) finite cylinders (to account for the interstellar polarization) with a Gaussian size distribution for the mantle dn(a) / da ~ exp[-5(a - a_c)² / a_i²] and a single silicate core radius a_c 0.07 µm and a "cut-off" size a_i 0.066 µm. In addition, a PAH component and a population of small graphitic grains are added respectively to account for the far-UV extinction rise plus the "UIR" emission bands and the 2175 Å extinction hump.

  Again, modifications to this model were also made by considering different coating materials (e.g., amorphous carbon, HAC), including new dust type (e.g., iron, small bare silicates), and varying dust size distributions (Chlewicki & Laureijs 1988; Duley, Jones, & Williams 1989; Désert, Boulanger, & Puget 1990; Li & Greenberg 1998; Zubko 1999a). The Duley et al. (1989) model exclusively consists of silicate core-HAC mantle grains with a bimodal size distributions: (1) ~ 43% of the total dust mass in grains 100 Å which produce the far-UV extinction rise and the 2175 Å hump through an electronic transition of the OH^- ions in low-coordination sites on or within silicate grains, and (2) the remaining dust mass in grains 0.05 a 0.25 µm with a power-law dn(a) / da ~ a^-3.5 size distribution. The "UIR" bands were attributed to the thermally isolated aromatic "islands" of HAC material - Duley et al. (1989) postulated that an absorbed stellar photon might remain localized in a single aromatic unit of the grain long enough for the unit to cool by IR vibrational relaxation rather than by transferring the photon energy to the phonon spectrum of the grain.

• The composite grain model - realizing that grain shattering due to grain-grain collisions and subsequent reassembly through agglomeration of grain fragments may be important in the ISM, Mathis & Whiffen (1989) modeled the interstellar grains as composite collections of small silicates, vacuum ( 80% in volume), and carbon of various kinds (amorphous carbon, HAC, organic refractories) with a power-law size distribution dn(a) / da ~ a^-3.7 in the size range 0.03 µm a 0.9 µm. In addition, a separate small graphite component containing 59 × 10^-6 of C is needed to account for the 2175 Å extinction hump. Wright (1987) argued that a fractal structure would be expected for interstellar grains formed through the coagulation of small grain fragments created from grain disruption caused by supernova shock waves. In view that the relative abundances of refractory elements in the ISM may be as low as 65% of solar (Snow & Witt 1996), Mathis (1996) revised the composite model in order to satisfy the tighter abundance constraints. To optimize the use of the heavy elements, Mathis (1996) derived a vacuum fraction of ~ 45%. The "UIR" emission, which remained unaccounted for in the models of Mathis & Whiffen (1989) and Mathis (1996), was taken into account in Zubko et al. (2003) by including a population of PAH molecules.

It is fair to say that, as can be seen from the above summarizing description of the key contemporary dust models, a consensus on the interstellar dust compositions and sizes is now approaching among various grain models, although the debate in details is still going on and is not expected to disappear in the near future (except that the IR spectropolarimetric observation at 3.4 µm and 9.7 µm may allow a direct test of the core-mantle model for interstellar dust; see Li & Greenberg 2002) --

• Regarding the grain chemical composition, the most generally accepted view is that interstellar grains consist of amorphous silicates and some form of carbonaceous materials; the former is inferred from the 9.7 µm Si-O stretching mode and 18 µm O-Si-O bending mode absorption features in interstellar regions as well as the fact that the cosmically abundant heavy elements such as Si, Fe, Mg are highly depleted; the latter is mainly inferred from the 2175 Å extinction hump (and the ubiquitous 3.4 µm C-H stretching vibrational band) and the fact that silicates alone are not able to provide enough extinction. For the carbonaceous component, a wide range of dust materials have been suggested including amorphous carbon, coal, C₆₀, diamond, graphite, HAC, PAHs, organic refractory, and QCC (see Pendleton 2004 for a review of the carbonaceous component).

• It is also generally accepted that, through the analysis of the wavelength-dependent interstellar extinction and polarization curves as well as the near and mid IR emission, interstellar grain sizes may be separated into two domains - (1) the "large" grain component (with radii a > 0.025 µm; including the "classical" grains [with a 0.1 µm]) which is primarily responsible for the extinction, polarization and scattering at visible wavelengths and the IR emission at 60 µm; and (2) the "very small grain" component (with a < 0.025 µm) which contributes importantly to the extinction in the vacuum-UV and emit strongly in the near and mid IR at 60 µm (see Figs.8, 16 of Li & Draine 2001b) when transiently heated to high temperatures during quantized absorption events.

  While the size distribution for the "classical grain" component is relatively well constrained by fitting the observed interstellar extinction curve for an assumed dust composition, our knowledge of the size distribution dn/da for the "very small grain" component is better constrained by the interstellar near and mid IR emission due to the fact that, for 0.1 µm, these very small grains are in the Rayleigh limit (x 2 a / << 1) and their extinction cross sections C_ext(a, ) per unit volume V are independent of size, so that the observed UV/far-UV extinction curve A() / N_H (mag cm^-2 H^-1) only constrains the total volume V_tot of this grain component: A() / N_H = 1.086 N_H^-1 C_ext(a, ) (dn / da)da = 1.086 N_H^-1 V_tot (C_ext / V), where N_H is the hydrogen column density. In contrast, the near and mid IR intensity I is sensitive to the grain heat capacity ( a³) which determines the maximum temperature to which the grain can reach when illuminated by a radiation field: I = N_H^-1 C_abs(a, ) (dn / da)da B(T[a]) (dP / dT)dT where C_abs(a, ) is the absorption cross section for a grain of radius a at wavelength (for grains in the Rayleigh limit, C_abs C_ext), B is the Planck function, dP / dT is the dust temperature distribution function. Evidently, dP / dT is a sensitive function of grain size a (see Draine & Li 2001, Li & Draine 2001b; and Section 3, Fig.6 of this paper).

In the following sections of this review, we will focus on the ultrasmall grain component (a 25 nm), with particular emphasis on the photophysical processes including the stochastic heating and the vibrational excitation of PAH molecules, and the excitation of the photoluminescence of silicon nanoparticles. In astrophysical literature, one frequently encounters terms like "ultrasmall grains", "very small grains", "large molecules", "tiny grains", and "nanoparticles"; to avoid confusion, we emphasize here that they are synonymous. In the following we will use the term "nanoparticles" - by "nanoparticles" we mean grains of a few angstroms to a few tens of nanometers in radius.

In Section 2 we summarize the direct and indirect evidence for the existence of nanoparticles in the ISM. The vibrational excitation of nanoparticles is detailed in Section 3. In Section 4 we discuss the physics regarding the excitation of photoluminescence. In Section 5 we present an overview of the nanoparticle species known or proposed to exist in interstellar space. Concluding remarks are given in Section 6.